Frege maintained that words typically express senses (1892a, 1892b, 1914, 1918). More fully, and stated in today’s taxonomy of linguistic forms rather than Frege’s, he held that senses are expressed by singular terms, by predicates of any order and any acidity, by quantifiers of any order, by various types of function signs, by sentential connectives, and by sentences themselves. They ‘compose’: they combine in a rule-governed manner to generate complex senses out of simpler ones, and they determine, along with the context of utterance, the referents of the expressions, including the truth-values of sentences. The sense of an expression is itself an abstract entity, which can in principle be expressed by more than a single form of words, and can indeed be expressed by forms of words from more than one language. At least in that way, it is independent of language.

Suppose we introduced a term into a language by explicit definition—say a term ‘brote’ into English with ‘df: x is a brote iff x is a dog less than three years old with exactly one white paw’. For the Fregean, the very same sense could be indicated by introducing a term into French, perhaps in place of ‘un chien de moins de trois ans avec exactement une patte blanche’. Translation, according to the picture, is straightforward. A sentence in French translates a sentence in English just in case the two sentences express the same sense—the same ‘thought’, what we now, following Russell, call the same ‘proposition’. The doctrine dovetails with certain more explicitly cognitive ways of speaking. If one understands a sentence (uttered with respect to a context) then one is said to ‘grasp’ the proposition it expresses (in that context). We understand one another on an occasion of utterance just in case we ‘attach’ the same sense to the words uttered on that occasion (skirting over difficulties introduced by indexicals). One might go so far as to say that for Frege, along with matters of the context of utterance, words are mere instruments, pointers, if psychologically necessary, to senses, to propositions or thoughts.

The reasons for thinking or speaking in this way are legion and formidable. I will not stop to rehearse them, leaving it for the body of the argument to emerge insofar as they do emerge (I will however assume that Frege intended his view to cover ordinary language—despite its imperfections and inconstancy—not just to the precise artificial notation of a Begriffsschrift; Weiner, 1990 expresses a contrary view). My thesis is that words in context play the standard roles of Fregean sense. As it might be put rather uncompromisingly but briefly, the sense of a sentence, the proposition it expresses, is the sentence itself. More compromisingly but less briefly, central questions about the sense or proposition expressed by a sentence (at a context of utterance) can be answered in terms of the sentence-at-a-context, by the properties of the sentence together with the context of utterance. Questions about the sense of a sentence independent of context of utterance can be answered in terms of the sentence itself independent of context of utterance.

By ‘sentence’ henceforth I will usually mean not sentence-tokens but sentence-types, including their semantic properties. Same thing with word-types and word-tokens. But I will not stop to explain what sentence-types or word-types are in more exact detail (it is a large and difficult subject, partaking of phonology and morphology as well as grammar or syntax; see Ebbs, 2009 for an explicitly philosophical angle). It is enough to make the relatively innocuous assumption that the existence of sentence-types and of word-types are guaranteed by that of simple word-tokens plus the recursive operation of the principles of grammar. For convenience I will speak of certain properties of tokens as being inherited by their corresponding types; for example the spatial or temporal property of having three syllables, a certain phonic-typographic property.

The thesis will not seem radical to everyone (beside the obvious figures—Wittgenstein, Quine and Davidson—Iacona, 2003 is one recent anti-proposition comrade); but there are great many who maintain the opposite (a recent pro-propositionalist statement is Duží, 2019). Indeed the existence of abstract propositions is often taken for granted in analytic philosophy—in written philosophy but also in the seminar room—sometimes in ways that matter substantively. The case for this thesis will have to be only broad-brush but I hope not simply dismissive of the opposed view. It is not based on the comparatively abstruse ground of underdetermination, radical interpretation or translation, but instead on its broad consistence with certain central and relatively plain facts. (And I assume that, in the name of ontological economy if nothing else, one should not want to accept propositions if they’re not needed, and thus a draw is a win, but I will not argue specifically for this). The case specifically consists of seven considerations: (1) Cognitive Value: Hesperus and Phosphorus; (2) Indirect Sense and Propositional Attitudes; (3) the Paradox of Analysis; (4) the Picture Theory of the Tractatus; (5) Meaning and Syntactical Diagrams; (6) Quantifying-in; (7) Patterns of Use. Frege is merely a convenient stalking horse as well an historic sine qua non: The case carries over to any conception of meaning that agrees with Frege that the abstract proposition and not the material sentence is the essential stuff of meaning and communication—thus in particular to one that denies that proper names express sense over and above their having reference.

1 Cognitive Value: Hesperus and Phosphorus

In fact I will assume a position sometimes attributed to Kripke, that ordinary proper names do not typically express senses (1980; the attribution is erroneous, but this will not matter here). They have only reference—in Mill’s terminology, only denotation and not connotation (or if you like, any connotation they have is irrelevant to their semantic value). The famous historical chains play their role in determining reference but do not normally enter into the understanding of such terms. What, then, do we make of Frege’s argument that (a) there are many cases of co-referential proper names that seem plainly to carry different cognitive value, and his related argument (b) that two propositional attitude ascriptions may have opposite truth-values, the ascriptions’ differing only in that one has a certain proper name in the content clause where the other has a co-referential but distinct one in the same position? In this section I discuss (a), and (b) in the next (I will set aside Frege’s argument from non-referring proper names.)

Nathan Salmon holds that the slack can be taken up by cognitive science rather than semantics (1986, 1989). If one asserts that Hesperus is F but denies that Phosphorus is F, then indeed one is caught in a semantical contradiction, Hesperus and Phosphorus being the same object: there is a certain proposition that one is both asserting and denying. But according to Salmon, one always grasps propositions via certain ‘guises’, which are unique to the individual’s cognitive processing. One grasps the proposition via one guise and believes it; one grasps the same proposition in a different guise and disbelieves it. At the level of thought, at the individualistic cognitive level, there is no contradiction—but at the public, semantic level, there is. Normally for Frege-friendly examples such as ‘Mark Twain’ or ‘Mt. Everest’ much material will in fact be closely similar amongst guises considered interpersonally, but most cases of real proper names will be more idiosyncratic, and there is no general demand from semantics that guises should be shared or similar.

Salmon himself is firmly of the pro-proposition camp; I will not affirm his analysis of singular propositions. I affirm only Salmon’s commitment to the name’s or a representation of the name’s typically figuring in one’s guises, in one’s mental files. For this is enough to show why the names themselves can play the Fregean cognitive role of senses. Trivially, every use of ‘Hesperus’ is a use of that particular word with its phonological and typographical properties. Therefore any publicity requirement on sense or meaning is satisfied. Why can Adam rationally assert ‘Hesperus is F’ whilst rationally denying ‘Phosphorus is F’, when the two singular terms co-refer? Because they are two materially different sentences. (And there is no requirement that competence with proper names ‘a’ and ‘b’ requires that one should know the truth value of ‘a = b’; so Russell’s requirement on logically proper names is not satisfied.)

That is not the whole story, however. As Kripke pointed out, there are cases in which a subject appears to take different attitudes towards two tokens of the same sentence (1979: 265–6). One has some information concerning an object one calls ‘a’, and some other information concerning an object one also calls ‘a’; they are the same object but for whatever reason one mistakenly thinks that they are two objects. One therefore appears to both assent and dissent from ‘a is F’ for some predicate ‘_ is F’ (it might even seem that one at least implicitly takes ‘a = a’ as false, to put it at its starkest). Kripke’s example is Ignacy Jan Paderewski, the Polish prime minister who was also a great pianist: One can imagine a person thinking that these are two different people that both happened to be named ‘Ignacy Jan Paderewski’, thus accepting for example ‘Ignacy Jan Paderewski became PM in 1919’ on one occasion while also rejecting what looks to be the very same sentence on another occasion, without undergoing any relevant cognitive change (I’ve changed the example slightly in order to ward off irrelevant objections). Yet it seems that such a person is no less rational than Adam above, who rationally affirms ‘Hesperus is F’ and rationally disaffirms ‘Phosphorus is F’.

There are different ways of responding philosophically to this phenomenon. For my purposes I will tell only a minimally adequate story, one I take to be consistent with a range of views, and again guided by Salmon’s view. One may know several people named, for example, ‘Claire Smith’. We suppose then that at the individualistic, cognitive level, each use of the name carries with it a tacit further parameter—viz., ‘Claire Smitha’, ‘Claire Smithb’, and so on, which go into one’s mental files. It does not matter for my purposes exactly what the content of these parameters might be, so long as they line up many-one with the bearers. (The tacit parameters might be indexicals—just as utterances of ‘that = that’ can be informative, so can ‘[that] Paderewski = [that] Paderewski’, the material in brackets being tacit; but there is no need to commit to this here; cf. Hanks 2017: 38). To our Paderewski-benighted agent, then, the two utterances of ‘Ignacy Jan Paderewski became PM of Poland in 1919’ will not strictly be utterances of the same sentence—in fact in the agent’s idiolect or mentalese, a sentence of the form ‘Ignacy Jan Paderewskia ≠ Ignacy Jan Paderewskib’ is accepted. The agent’s cognitive estimate and the public, semantical point of view are thus out of kilter. Semantics, in this picture, requires that each person named ‘Claire Smith’ gets at the minimum a distinct parameter which individuates the right Clare Smith at each context of utterance, and so it is with those named ‘Ignacy Jan Paderewski’. The mishap comes to a head in the agent’s not recognising the falsity of his own ‘Ignacy Jan Paderewskia ≠ Ignacy Jan Paderewskib’. The crucial fact for our purposes is that despite initial appearances, the differences in information for the agent between one utterance of ‘Ignacy Jan Paderewski became PM in 1919’ and another correlate with individualistic syntactical features, different word-types in the individual’s mental economy.

2 Indirect Sense/Propositional Attitudes

In Fregean terminology, ‘Hesperus is a star’ expresses a certain direct sense. In the context ‘Adam believes that Hesperus is a star’, ‘that Hesperus is a star’ refers to that direct sense, and expresses a certain indirect sense, which enables the clause to refer as it does. But how, exactly? Staying with Frege’s picture, how is this fact to be assimilated to his function/argument paradigm in semantics? The matter is difficult for Frege, as he holds with ample reason that the referent of ‘Hesperus is a star’ is simply a truth-value, truth or falsity—not the sort of thing to be an argument for a function denoted by ‘that __’ in propositional attitude constructions, or to be fed into such constructions as ‘Adam believes that __’. The information necessary for such fine-grained values as the proposition that Hesperus is a star appears to be unavailable, inaccessible to the semantical machinery.

Carnap (1947), Quine (1956) and then Davidson (1968) proposed sentential analyses of the propositional attitudes. Following Quine,

  1. (1)

    Adam believes that Hesperus is a star.

is rendered as:

  1. (2)

    Believes-true (Adam, ‘Hesperus is a star’).

—understanding by this that the sentence attributes no particular knowledge of English to Adam (indeed, so far as the semantics goes, it is conceivable for example that a mouse could believe-true ‘A cat is there’; the sentence belongs to the ascriber’s language, not the subject’s; Quine, 1956: 186). The idea is merely to maintain the relevant sorting of believers, while making the reference to a sentence explicit. But let us now think of (1) as:

  1. (3)

    Adam believes that ‘Hesperus is a star’.

And then with Davidson we might point out that this is not so monstrous as it seems. Think of ‘that’ in ‘Adam believes that Hesperus is a star’ not as clause-marker or complementizer but as an indexical, referring not to a sentence but to the utterance about to be undertaken (or immediately to the right) of a sentence, as might be represented as:

  1. (4)

    Adam believes that: Hesperus is a star.

—where the indexical with the colon affixed has an effect similar to ‘the following utterance or inscription’, and the fact reported by the whole is a relation of Adam’s cognitive state to the utterance or inscription (again it is the ascriber who supplies the utterance or inscription). But for our purposes it doesn’t matter whether it’s Quine or Davidson we follow, for on either view the embedded sentence itself figures centrally in the semantics of propositional attitude ascriptions. What is important for our purposes is that this apparently solves Frege’s problem as just described.Footnote 1 To put it back to front: One encounters a sentence ‘Hesperus is a star’, or the sentence ‘Hesperus is a star’ in the context of its utterance; one understands the sentence, and then one descries that Adam is being said to believe the sentence, or to believe that, where the demonstrative ‘that’ is being used to pick out the utterance of the sentence. Notice that I have avoided speaking of ‘the content’ of an utterance or inscription, for it is of course meaningful utterances and inscriptions that are at issue; it would be redundant (if harmless) to include those words. Notice also that the idea is not that ‘believes that s’ is to be analysed as “believes that ‘s’ is true” (with respect to which there are ready counterexamples). The form ‘believes-true’ is a neologism designed to make explicit the implicit reference to a sentence in the content clause of the original such as (1).

Now when prefixed by ‘that _', distinct sentences differ in the way they ‘present the thought’ which they express, to invoke a vivid form of words from Frege (the ‘mode of presentation’ differs, to invoke another). Thus it is possible that the following can have opposed truth-values despite the equivalence of the embedded sentences: ‘Adam believes that Hesperus is a star’ and ‘Adam believes that Phosphorus is a star’. To continue with Frege’s idiom, the contribution of the embedded sentences for determining the truth-value of the propositional attitude ascriptions are their ordinary or direct senses, not their truth-values. What then of the senses of the embedded sentences where they appear within ascriptions? It should not be inferred that distinct sentences—e.g. ‘Hesperus is a star’ and ‘Phosphorus is a star’—have different indirect senses in addition to their direct senses. The conclusion should be, rather, that there is no such thing as the indirect sense of a sentence, in addition to its direct sense. What Frege called the indirect sense of a sentence is the way in which its direct sense is presented when demonstratively indicated by a proposition name-forming operator (to get a Quinean reading in keeping with the line I am pushing, erase ‘demonstratively’, replace ‘its direct sense’ with ‘the sentence’, and ‘proposition’ with ‘quotation’). Again, language itself rises to the fore.

A certain response is thus suggested to the type of situation exemplified by the following examples adapted from Mates (1952):

  1. (5)

    Nobody doubts that one who buys a teacup buys a teacup.

  2. (6)

    Nobody doubts that one who buys a teacup purchases a teacup.

Even if we grant that ‘buy’ and ‘purchase’ are synonyms—I will say more about synonymy in §7—we nonetheless can count the first true but the second false, or least as comparatively dubitable, signalling the possibility of a divergence of attitudes towards the two. For ‘One who buys a teacup buys a teacup’ is distinct phonologically and typographically, is made up of slightly different words, from ‘One who buys a teacup purchases a teacup’. One can be competent with the words but still hesitant over the exact equivalence of the two statements.

On this view, we understand such phrases as ‘that p’ or ‘the thought that p’ in virtue of our understanding of the embedded sentences, of those sentences (together with our understanding of the embedding material—‘that __’ and so on). And that language itself must figure in our semantic calculations shows that nothing in principle prevents us from jettisoning the idea of senses as ontologically independent of the words and sentences which we are in the professional habit of saying ‘express’ them. Think of that way of speaking as a somewhat misleading way of describing certain properties of linguistic tools which cannot exist without the tool, any more than the smile can exist without the cat.

A drawback that might be thought to arise, especially for aficionados of Church (1950, 1954; also Thomason, 1977), is that (2) or (3) cannot be satisfactorily translated into for example French, because if the translation preserves the reference to ‘Hesperus is a star’, it will not help a French-only speaker, and if does not, then it cannot have the same sense as the original. I will defer discussion of this point to §5 and §7. I will however acknowledge at this point that Davidson’s scheme might be thought better placed than Quine’s with respect to this situation, because the relevant constraint on the translation of propositional attitude ascriptions is that the utterance used in the content-clause must make the translator a ‘samesayer’ (or rather a ‘samethinker’) of the translatee, a notion which might be thought sufficiently flexible. It might on the other hand be objected that this notion packs in too much into the notion of ‘samesaying’ without explaining it, even that the move risks creating a kind of hermeneutical circle. In any case the position as it emerges in §5 and §7 will render this issue moot.

3 The Paradox of Analysis

Given the foregoing section, a solution to the Paradox of Analysis is straightforward. The Paradox is that if we assume a standard view of definition, and the following is presented as a philosophical definition of its left side in terms of its right—

  1. (7)

    Df: p iff q

—then that overall sentence would express the same proposition or content—in Fregean terms, would express the same sense or thought—as this:

  1. (8)

    Df: q iff q.

But this latter sentence is an uninformative triviality or tautology (‘p’ and ‘q’ are of course mere schematic letters, replaceable by whatever philosophical analysans and analysandum-pair you care to name). Whatever philosophical analysis is, it issues in more substantive statements than that.

The solution is simply that the two sides of (7)—‘p’ and ‘q’—are distinct sentences, even though they imply one another, even necessarily imply one another. If you like, ‘the proposition that p’ differs from ‘the proposition that q’, as in §2.

4 The Picture Theory of Meaning

I’m not positively going to advocate Wittgenstein’s so-called picture theory of meaning as it appears in the Tractatus Logico-Philosophicus (1922), but I will point out that a central and persistently compelling aspect of the theory accords with my main thesis. I will present this aspect dogmatically and on its own, ignoring certain Tractarian complexities. ‘Propositions’, on the Tractarian conception, are nothing but meaningful declarative sentences with respect to a given index—full-blooded (possible) statements, either true or false. Unlike Fregean thoughts, their existence is partly that of (types of) real empirical objects occupying space and time. Take a meaningful use of ‘The cat is on the mat’. The two singular terms—‘the cat’ and ‘the mat’—name, refer to, ‘reach out and touch’ a certain object (a cat) and a certain other object (a mat), respectively (ignore the Tractarian idea that cats and mats are not strictly speaking objects, that they will disappear on analysis, as one says). The conventions of English are that when a referring singular term is connected with another by means of ‘—is on—’, in that order, the result is true if and only if the referent of the first is on top of the referent of the second. (A statement ‘shows how things stand if it is true’, Wittgenstein’s emphasis; 4.022—what Frege called ‘sense’ is displayed by the phonological or typographical structure of complex symbols). If it is true, then one fact (‘the cat’ being joined to ‘the mat’ by ‘—is on—‘) corresponds to another (the cat’s being on the mat). We could equally have a convention that, with respect to the same cat and mat, my coffee cup placed upon my coaster means that the cat is on the mat, so that the relation ‘is on’ is signified by itself; or that a different cat placed upon a different mat means that the original cat is atop the original mat, and so on.

Language has other sorts of declarative sentences that seem not without considerable explanation to fit the cat-on-mat paradigm, and perhaps even on its chosen ground the paradigm will have to be qualified in various ways for certain examples, but as a basic template for atomic sentences pertaining to physical states of affairs it is not so easy to fault. One can see how children might catch on, how it dovetails with such human activities as make-believe, the reading of charts, maps and diagrams, and how it could qualify as a linguistic universal. The main thing from our perspective, once again, is the essential and intricate involvement of language itself in describing meaning and understanding. The flesh-and-blood sentence is directly involved in the semantic explanation of the resulting statement—with objects standing for objects, properties standing for properties, and relations standing for relations. One state of affairs depicts another via logical isomorphism.

5 Syntactical Diagrams and Meaning

Syntactical diagrams elucidate the structure and categories of linguistic expressions, purportedly corresponding in some sense to how they are understood by competent speakers. Take the active and passive forms ‘Linda kissed Paul’ and ‘Paul was kissed by Linda’. Do these mean the same? Do they express a single self-same proposition? What about ‘Linda and Paul stand respectively in the kissed relation’? What about ‘Linda kissed Paul or Linda kissed Paul’? What about ‘Linda kissed Paul or Linda kissed Paul, and Linda kissed Paul or Linda kissed Paul’? What about ‘Linda kissed Paul and Linda is Linda’? Any hesitancy one has over these questions suggests a lack of clarity in the questions. But trivially these will not have the same diagram. They are syntactically distinct. Add in descriptions of (not just the internalist θ-roles) but the full-fledged meaning-properties of the simple items—of ‘Linda’, ‘kissed’ etc.—and you have, arguably, a semantical diagram, a diagram of their meaning that copes easily for example with ambiguity (Larson and Ludlow’s ‘Interpreted Logical Forms’ 1993 contains crucial work for developing this idea). From this point of view, these sentences all have the same truth-condition but their diagrams are not same, and therefore, in a sense that can be recommended just on account of its simplicity and relative clarity, their meanings are not same (see Lewis, 1970; Higginbotham, 1991). So ‘Manhattan is north of Staten Island’ does not have the same meaning as ‘Staten Island is south of Manhattan’. Who really cares? If anything is to blame, it is lack of determinacy in the ordinary concept ‘meaning’. Yet it is still open to insist that no one can understand both without counting both true or both false—that the transition from one to the other is ‘analytic’, if you like.

It is trivial that sentences from different languages will technically have different meanings, even if structurally they have the same diagrams, with identical objects on the part of the items of reference and identical extensions of the predicates that occur in the sentences. But speaking more loosely they will mean the same, precisely because they have structurally the same syntactical diagrams, with identical objects on the part of the items of reference in their respective referring terms, and at least similar explanations of their respective predicates, and so on. One is initially impressed with Church’s (1950) insistence against Carnap that a sentence translates another from a different language just in case the two sentences have the same sense, but that, according to the picture being promoted here, does not hold up. Nevertheless as I say one can make allowance for that alternative way of speaking; I will say more on this in §7.

6 Quantifying In

It is a feature of this picture that we language users are much more adept at darting back and forth between using words and mentioning them, between object-language and meta-language, than some readers of Tarski, Gödel or Carnap might expect. In ordinary speech, we tend to ride roughshod over the distinction between direct quotation and indirect quotation. We often say such things as ‘She said the plumber’s not coming’, simply leaving it ambiguous whether the speech report is direct or indirect, for the most part not caring—sometimes not even being aware of the distinction, especially among illiterate language users. A direct report can be transformed without any further information to an indirect one, and given certain information, we can go the other direction: if we do perchance know the exact words used, the transition from direct to indirect and back, from use to mention and back, is automatic and does not itself involve or require any further information.

This little sermon might help accommodate one to a certain take on the distinction between de dicto and de re belief reports that is suggested by what was said earlier, which focussed on the de dicto style. Again the idea is Quine’s (1956: 185–7). Begin with an ordinary utterance of:

  1. (9)

    Edina believes that Ray smokes.

Paraphrase this as:

  1. (10)

    Believes-true1 (Edina, ‘Ray smokes’).

(The reason for the added subscript will emerge presently). (10) exemplifies the (syntactical) de dicto style, and as per §2 can be held to display what is going on in (9) more explicitly: one who hears and understands (9) recognises the words making up the clause ‘that Ray smokes’. In the typical case, an ordinary hearer implicitly knows that ‘Ray smokes’ is a truth-valued sentence comprising the singular term ‘Ray’ and the predicate ‘__ smokes’ (the name, and what is being said about its bearer). Thus the de re style—where ‘de re’ involves syntactical relations and not simply the character of the referring terms within the content clause—can be rendered:

  1. (11)

    Ray is an x such that believes-true2 (Edina, ‘① smokes’, x)

The step from (10) to (11) illustrates what Quine termed ‘exportation’ (1956: 182). The circled numeral is borrowed from Goldfarb (2003); it indicates what Frege called a gap or hole in the predicate, and is not a variable or pronoun as we understand them nowadays. The reason for the subscript ‘2’ is now evident: it indicates the number of argument places in the belief-predicate (in addition to the subject of the attribution). (11) is of course extensional with respect to the ‘Ray’ position, hence is amenable to quantification, viz.:

  1. (12)

    (∃x)(believes-true2 (Edina, ‘① smokes’, x)).

As you might put it not very ordinarily, “There’s some salient object that Edina believes satisfies the predicate ‘① smokes’”; more ordinarily ‘There’s this guy who Edina thinks smokes’. This is inferred almost trivially from the conjunction of (9) and the assumption that ‘Ray’, on this occasion, successfully refers. Two-place predicates and higher would be handled analogously, e.g. ‘① loves ②’. (A convention would be needed to ensure that the first variable or name goes with the first place indicated by circled numeral, the second with the second, and so on.) Davidson did not officially offer a treatment of the de re attitudes, but his pattern can be developed similarly.

The foregoing is not a rendition of the activities that might take place inside the mind of the man on the Clapham omnibus; it just adds to the case that there no facts that would compel one to recognise language-independent propositions. I go along with Quine in holding that the de dicto implies the de re (1977: 9–10; 1992: 70–71), but that is controversial and at any rate I will leave the issue aside.

7 Patterns of Use

I’ve said that the duties of Fregean sense should be assigned to the words themselves (setting indexicality aside). More exactly, the sense for example of the sentence ‘Hesperus is a planet’ should be thought of as a collection of properties of (a type of) a collection of flesh-and-blood marks or sounds. But what properties? How do they enter in?

This is a large question, but there is an old if programmatic answer. The sense or meaning of a word, or a sentence, is to be sought not in some abstract object, but in its use. It is a pattern of use, a normal way of using a linguistic tool, which might have to be conveyed not by an explicit definition but by means of a few sentences, by similar words, or by examples. Obeying the qualification in §43 of the Philosophical Investigations, this is so ‘[f]or a large class of cases—though not for all’ (1953: 20). The strength of this answer is that it acknowledges that the matter will be very different in different cases—as in Wittgenstein’s tool analogy, Investigations §11—that it is suitably vague, that it is unpretentious, and that it is non-committal as between what would be, from a Fregean or indeed a Gricean point of view, pragmatic matters as opposed to matters properly of sense or meaning. The cases outside the ‘large class’, one might speculate, include examples such as ‘red’ or ‘dog’, which can be defined ostensively. But I do not think it would cross Wittgenstein unduly to think of ostensive definition as a way of teaching the use of a word (or an important aspect of the use of a word), granted that one must always presuppose certain propensities or abilities in the pupil. To say something about those animals—pointing to a sleeping dog—one says for example ‘The dog is sleeping’. The intended use is that to refer to a dog, to say something about one, one can say ‘dog’; that is the main use-property of that particular word, the pattern of use one wants to convey. Talking about things, referring, pointing and so on are skills of one’s burgeoning mastery as a neophyte language speaker, part of grammar in Wittgenstein’s wide sense, part of what one learns (or more likely, they are skills that develop so long as one is exposed to the right environment).

Any serious theory of use must be a least broadly compositional. And—now to lay my cards on the table—I think that any such theory will tend to posit precise facts where there are no precise facts. I thus go along with Quine’s conclusion that a precise theory of meaning, or a precise theory of meaning for a particular language, or even a particular speaker, is not justifiable, and that includes a precise theory of use. This extended quotation from Quine would not go amiss:

Though the word ‘meaning’ is ubiquitous in lexicography, no capital is made of a relation of sameness of meaning. An entry gets broken down into several “meanings” or “senses,” so called, but only ad hoc to explain how to use a word in various dissimilar situations. When a word is partly explained by paraphrasing a sample context, as is so often the way, the paraphrase is meant only for typical circumstances, or for specified ones; there is no thought of sameness of meaning in any theoretical sense. Nor does a practical dictionary observe a distinction between linguistic information about a term and factual information about its denotata. The goal is simply the user’s success in plying the language.

The word ‘meaning’ is indeed bandied as freely in lexicography as in the street, and so be it. But let us be wary when it threatens to figure as a supporting member of a theory. (1970: 83; cf. 1981: 43–54; 1992: 56–9)

Nevertheless this is not to deny the cogency of Chomsky’s posit of an I-language, or rather the possession of an I-language by each mature individual, which is rooted in a universal, genetically constrained internal framework which explains our ability to generate and process novel grammatical strings without upper bound, i.e., the capacity is ‘potentially infinite’ (see Chapter Two of Berwick & Chomsky, 2016 or the essays in Chomsky, 2000 for relatively recent formulations of his view). Chomsky is also sceptical as to the prospects for a rigorous theory of meaning or use—beyond an ‘internalist’ conception—but however imprecisely, we do speak of the meaning of sentences and are understood. My claim is just that the factual undergirding of such talk can be explained as the use of real grammatical strings in context, not to imagined contents or senses associated with the strings.

This brings me to my final subject, an important one. What is synonymy? It is not identity of sense, not sameness of meaning, to link up with the above quotation from Quine. What is it, then? It might be guessed from the quotation together with what I have said so far: As an ordinary word of English, synonymy pertains to the matching of use—where what counts as ‘matching’ depends on the details of the case including the purposes of those involved, and is ultimately a matter of degree, not of kind. It is not all-or-nothing. The aim is successful communication, indeed where what counts as ‘successful’ varies from case to case, and once more is a matter of degree, true as it is that in many cases success is at its maximum. For all uses, or just about all uses, the use of ‘filbert’ matches that of ‘hazelnut’; for most uses, ‘bachelor’ and ‘unmarried man’ match, but in some cases not (fine aspects of their respective connotation differ, even if their denotations match); for some uses ‘dog’ and ‘cur’ match, but not for all. Some questions simply have no clear answer. Does ‘change’ match ‘fluctuate’? Does the verb ‘match’ match the verb ‘copy’? Partly, yes; in some contexts, not without qualification in others. Shifting contexts, interests, social groups, and dialects make generalisation difficult. But these are familiar, fussy and pedestrian kinds of facts. When we translate whole sentences, we often want to match the foreign words—we prefer ‘synonyms’, as we say—but sometimes there aren’t any quite matching words to choose, and sometimes it will feel artificial or forced to insist on exact ‘synonymy’ of word or phrase. One wants a translation that has the ‘same effect’ for the purpose at hand, as much as it makes sense to try to find such a thing.

That is my position concerning synonymy: it is more-or-less matching of use, relativised to context which includes the sometimes obscure purposes of those involved. As an aside, this attitude towards synonymy goes together with my attitude towards belief-reports. I think that even acceptable ones are often irremediably rough. I certainly do not think that there is always a fact of the matter as to the exact equivalence between attitude reports. Does this picture therefore depart from Larson and Ludlow, who deploy their system of Interpreted Logical Forms—a precise version of what in §5 above I called ‘Semantical Diagrams’—to explain the attitudes? They hold that much of the remaining slack can be taken up theoretically at the pragmatic level (pp. 339–342).Footnote 2 I am sceptical but for the purposes here I take no stand. (see Dusche, 1995: 311–14, for something like this response and see 8.2 below for a related point).

8 Conclusion

I offer some brief remarks in closing concerning certain other responses to the bald Fregean demand for language-independent propositions.Footnote 3

(8.1) The above attitude towards synonymy goes together with resisting a definition of ‘the meaning of a word’ or of ‘proposition’ as a set of synonymous words or statements, or as action-types as in Peter Hanks (2011). First, it would be a case where the entities defined into existence would have to take a back seat ideologically to the collection of facts specified in their definition. Although of course there are sometimes practical reasons for speaking as if the entities existed—as when we speak of the way you hold your knife—such a reification would not strictly speaking add anything to the facts on ground, any more than the definitions of three-dimensional geometric shapes adds to the physics of a distribution of matter in the solar system. But second, and more decisively so far as I’m concerned, such would be far from a sharp definition. In many or perhaps most cases a set of synonyms would have at best to be a fuzzy one and not just in the sense in which the set of orange things is fuzzy: it would have to be an open-ended one, one partaking not just of potential Sorities-engendering vagueness but of conceptual vagueness, of (ordinary) indeterminacy.

(8.2) Substantially the same issue arises with Jeff King’s theory expounded in his ‘Structured Propositions and Sentence Structure’ (1996). According to King’s theory, there corresponds to each sentence S a certain syntactical regimentation or representation SI (495–6); then: ‘Given a sentence S, whose SI is constituted by lexical items standing in some complex relation R, the proposition expressed by S consists of the semantic values of those lexical items standing in the very relation R’ (498, author’s emphasis). This supposedly secures propositions as structured abstract entities expressed mutually by sentences of different languages, and is in that sense language-transcendent. But the view trades on interlinguistic sameness of semantic values including those assigned (in effect) to predicates, which is fine for toy philosophers’ examples such as ‘__ is a dog’ but not so much for ordinary language in action. It will be evident from the above remarks on synonymy that I do not believe that there is a profitable scientific task of devising such an account.

(8.3) Another way in which entities might be thought to play the crucial roles supposedly discharged by propositions—including the ascription of propositional attitudes—but without propositions actually being posited, is to accept some version of the ‘Multiple Relation Theory of Judgement’ (originally Moore, 1899, and Russell, 1912: 69–75; a sophisticated recent example is put forward by Lebens, 2017). Thus the fact that b believes that Fc is not analysed as b standing in a relation to the proposition that Fc, but as a certain complex relation holding between b, c, and the universal or property F (the belief is true if Fc, false otherwise; hence it is in the first instance a theory—a correspondence theory—of truth). But again substantially the same issue remains: The account depends on the idea that whenever one has a propositional attitude, the attitude involves properties or relations where the attitude does not depend on any particular language one happens to speak—in other words, it strongly suggests, if it does not strictly imply, that there is some criterion or procedure for determining exactly which properties or relations are at issue irrespective of the details of the subject’s language. It is difficult to see how this would work without a general, well-grounded relation of synonymy, which as above I do not accept. I do accept, as Russell himself did, that if there were no minds, there would be no truth, but this does not prevent us from making statements about times before the evolution of minds or language, or about mind-less or language-less possible worlds (I do not join Russell in his further thought, that there not being any truth would not affect there being facts, i.e., ‘complexes’ of objects and universals).

Speaking of the sense or meaning of a statement is exceedingly useful, and seems forced upon us by different statements, and especially statements from other languages, which seem to mean ‘the same’. Nevertheless I have been arguing that that is only a manner of speaking, that insofar as theoretical matters are at issue, the reification is at best not needed and at worst misleading; the role can perfectly well be filled by the statements themselves.Footnote 4