YOU DON’T SAY

YOU DON’T SAY?

Kent Bach

ABSTRACT. This paper defends a purely semantic notion of what is said against various recent objections. The objections each cite some sort of linguistic, psychological, or epistemological fact that is supposed to show that on any viable notion of what a speaker says in uttering a sentence, there is pragmatic intrusion into what is said. Relying on a modified version of Grice’s notion, on which what is said must be a projection of the syntax of the uttered sentence, I argue that a purely semantic notion is needed to account for the linguistically determined input to the hearer’s inference to what, if anything, the speaker intends to be conveying in uttering the sentence.

In the course of introducing his theory of conversational implicature, Grice adopted a notion of what is said with which to contrast what a speaker implicates in uttering a sentence. He imposed the Syntactic Correlation constraint, as I call it, according to which what is said must correspond to “the elements of [the sentence], their order, and their syntactic character” (1989, p. 87). So if any element of the content of an utterance, i.e., of what the speaker intends to convey, does not correspond to any element of the sentence being uttered, it is not part of what is said.

How could anyone object to the Syntactic Correlation constraint? It might seem that this constraint limits what is said to the context-invariant content of an utterance, but Grice explicitly allowed for indexicality and ambiguity. Even though the referent of an indexical element varies from context to context, it still corresponds (relative to the context) to that element. And even if the uttered sentence is ambiguous, either because it contains ambiguous expressions or is structurally ambiguous, the meaning that is operative in a given context is still a projection of the syntax of the sentence as used in that context. Nevertheless, recent critics have contended that a syntactically constrained notion of what is said is theoretically useless and empirically unmotivated. One common reason for this contention is that many sentences are semantically incomplete or underdeterminate—they fail to express a complete and determinate proposition even after any indexical references are assigned and any ambiguities are resolved—in which case nothing is said in the utterance of such a sentence. Another common reason is that many semantically complete sentences intuitively do not express, and typically are not used to convey, the proposition which, by Syntactic Correlation, comprises what is said. A third reason is that what is said (according to Syntactic Correlation) generally plays no role in the pragmatic processes involved in understanding utterances.

I will argue that a syntactically constrained notion of what is said does have a legitimate theoretical role to play and is empirically significant. In order to show this, I will first (in section 1) suggest several simple refinements to Grice’s original notion. Then (in section 2) I will explain what I take to be right and what I take to be wrong with Grice’s distinction between what is said and what is implicated. This distinction is not exhaustive, for it neglects the phenomenon of conversational impliciture (that’s implic-i-ture, as opposed to implic-a-ture), but taking that into account requires merely refining his distinction.

It has been objected that this distinction, however refined, is ultimately misguided, in that the notion of what is said is inherently pragmatic, just like implicature. This objection has been based on arguments that take roughly the following form:

1. Any sentence can be used to communicate something without implicating it.

2. If an utterance of a given sentence communicates a proposition that it doesn’t implicate, then that proposition is what is said in the utterance of that sentence.

3. Even so, the determination of that proposition is bound up in pragmatics.

4. Therefore, what is said is not a purely semantic matter.

Bad argument. Aside from mistakenly assuming that the distinction between what is said and what is implicated must be exhaustive, this form of argument relies on an overly expansive conception of what is said. The rationale for this conception derives mainly from certain psychological and epistemological considerations involving “pragmatic” processes and “semantic” intuitions. I will argue (in section 3) that such considerations are misdirected and do not undermine a syntactically constrained, purely semantic notion of what is said.

As I will explain in section 4, this overly expansive conception of what is said can seem attractive only insofar as one ignores certain distinctions, each between something semantic and something pragmatic. Once these distinctions are recognized, there is no reason to reject a purely semantic notion of what is said. However, the situation is not as clear-cut as these distinctions might suggest, inasmuch as there are open linguistic questions that remain to be addressed. In section 5, I will raise a few such questions and present a sample of linguistic phenomena that pose legitimate problems for how to draw the line between what is said and what is not.

1. Refining the notion of what is said

In contrasting saying with implicating, Grice took our “intuitive understanding of the meaning of say” (pp. 24-25) to comport with Syntactic Correlation, which requires that what is said correspond to “the elements of [the sentence], their order, and their syntactic character”. His use of ‘order’ and ‘syntactic character’ suggest some sort of compositional principle, such that what is said is determined by the contents of the constituents (‘elements’) of the sentence as a function of their syntactic relationships. And considering that a sentence can include unpronounced elements (Grice did not consider this), such as empty categories, implicit arguments, and syntactic ellipses, Syntactic Correlation should not be construed as requiring that every element of what is said corresponds to an uttered element of the uttered sentence. But it does require that every element of what is said correspond to some element of the uttered sentence. [1]

There was one respect in which Grice’s favored sense of ‘say’ was a bit stipulative. For him saying something entails meaning it. This is why he used the locution ‘making as if to say’ to describe irony, metaphor, etc., since in these cases one does not mean what one appears to be saying. Here he seems to have conflated saying with stating. It is more natural to describe these as cases of saying one thing and meaning something else instead. That’s what it is to speak nonliterally (at least if one does so intentionally). For example, if I utter “You’ll be my friend for life” to someone who has just betrayed me, I am saying that this person will be my friend for life—that’s just not what I mean. Similarly, if I utter “You are pond slime” to this person, I am saying that this person is pond slime, although I mean something rather different (though equally unsavory). Besides nonliterality, there are two other reasons for denying that saying something entails meaning it. A speaker one can mean one thing but unintentionally say something else, owing to a slip of the tongue, a misuse of a word, or otherwise misspeaking. Also, one can say something without meaning anything at all, as in cases of translating, reciting, or rehearsing, where one utters a sentence with full understanding (one isn’t just practicing one’s pronunciation) but is not using it to communicate anything.[2]

So we can replace Grice’s idiosyncratic distinction between saying and merely making as if to say with the distinction (in indicative cases) between explicitly stating and saying (in Austin’s locutionary sense). In so doing, we can distinguish nonliterality from implicature. [3] In this way we have a notion of what is said that applies uniformly to four situations: (1) where the speaker means what he says and something else as well (cases of implicature and of indirect speech acts in general), (2) where the speaker (intentionally) says one thing and means something else instead (nonliteral utterances), (3) where the speaker does not say what he intends to say, as in the misuse of a word or a slip of the tongue, and (4) where the speaker says something but doesn’t mean anything at all. On this conception, what is said, being closely tied to the (or a) meaning of the uttered sentence provides (allowing for indexicality and ambiguity) the hearer with the linguistic (semantic) component of his basis for inferring what, if anything, the speaker is communicating.

Whether a speaker means exactly what he says, something else as well, or something else instead, clearly it is never part of the meaning of a sentence that on a particular occasion of use it is being used to communicate anything at all. That is something the hearer presumes from the fact that the speaker is uttering a sentence at all. This communicative presumption, as Bach and Harnish call it (1979, p. 7), comes into play even if what the speaker means does not extend beyond or depart in any way from the meaning of the sentence he utters. Also, it is never part of what a sentence encodes that it has to be used literally—the hearer must infer (if only by default) that it is being used literally. The utterance does not carry its literalness on its sleeve. It might even contain the word ‘literally’, but even that word can be used nonliterally. So in no case does what a speaker says determine what he is communicating in saying it. To that extent it is true, though trivially so, that every utterance is context-sensitive.

In sum, I am suggesting that we reject Grice’s arbitrary stipulation that saying something entails meaning it but retain his appropriately restrictive Syntactic Correlation constraint on what is said. This has important consequences. First, Syntactic Correlation leaves open the possibility that what is said in uttering a sentence need not be a complete proposition. The mere fact that a sentence is syntactically complete (well-formed) does not entail that it is semantically complete. Second, since saying something does not entail meaning it, what one says does not depend on whether or not one has any communicative intention in saying it. So the speaker’s communicative intention cannot contribute to the determination of what is said.[4] Communicative intentions enter in only at the level of what the speaker means (is communicating). Even if the speaker means exactly what he says (nothing else and nothing more), that he means exactly what he says is a matter of his communicative intention. Third, any proposition that the speaker is communicating that is distinct from what he says is explicitly cancellable (Grice, 1989, p. 44). That is, the speaker can continue his utterance and, without contradicting himself, go on to explicitly indicate that this proposition was not was not part of what he meant. This third feature is really a consequence of the first two. All three of these consequences should be kept in mind in the course of what follows.

2. Refining Grice’s Distinction

Now we can ask whether the distinction between what is said and what is implicated is

exhaustive. Here I assume familiarity with Grice’s account (1989, chs. 2 & 3) and will not go into the details or trot out the stock examples. Suffice it to say that he proposes a general Cooperative Principle and specific “maxims”—of quality, quantity, relation (relevance), and manner—to account for the rationale and success of conversational implicatures. It is arguable, however, that being cooperative (at least in making constative utterances) consists simply in being truthful and relevant, hence that there is no need for four separate maxims. For example, observing the maxim of quantity is just being relevantly informative. Also, the maxims might better be thought of as presumptions (Bach and Harnish, 1979, pp. 62-5), for it is on the presumption that the speaker is being truthful and relevant (and intends to be taken to be) that the hearer figures out what the speaker means in saying what he says.

The question of whether Grice’s distinction is exhaustive is the question of whether a speaker, in saying something, can mean (communicate) something other than by saying it or by implicating it. As I have argued previously (Bach, 1994a), his distinction is not exhaustive, because it leaves out the phenomenon of conversational impliciture (as opposed to implic-a-ture). Implicitures go beyond what is said, but unlike implicatures, which are additional propositions external to what is said, implicitures are built out of what is said. [5] Here are some simple examples of sentences likely to be used to convey implicitures:

(1) Jack and Jill went up the hill.

(2) Jack and Jill are engaged.

(3) Jill got married and became pregnant.

Although these sentences express complete propositions, in uttering them a speaker is likely to have meant something more specific, a qualified version of what he said:

(1e) Jack and Jill went up the hill [together].

(2e) Jack and Jill are engaged [to each other].

(3e) Jill got married and [then] became pregnant.

where the words in brackets, which are not part of the sentence, indicate part of what the speaker meant in saying what he did. He would have to utter those words (or roughly equivalent words—the exact words don’t matter) to make what he meant fully explicit.[6] Let’s call the fuller version an expansion of the original and what is left out its implicit qualification. Utterances of sentences like (1) - (3) illustrate what I call sentence nonliterality (Bach, 1994b, pp. 69-72), as opposed to constituent nonliterality, since no expression in the sentence is being used nonliterally. The proposition the speaker means is not the exact proposition, as compositionally determined, that is expressed by the sentence. So, for example, someone who utters (2) is not saying that Jack and Jill are engaged to each other, any more than he would be saying this with “Jack and his sister Jill are engaged.” Rather, this is implicit in what he is saying or, more precisely, in his saying of it. That it is not part of what is said is clear from the fact that it passes Grice’s test of cancellability. That is, it may be taken back without contradiction. For example, there is no contradiction in saying, “Jack and Jill are engaged, but not to each other” (or “Jack and Jill went up the hill (1) but not together” or “Jill got married and became pregnant (3) but not in that order”).

There is another kind of case, in which the uttered sentence, though syntactically complete, seems to be semantically incomplete, given Syntactic Correlation (and assuming semantic compositionality). Sentences like (4) - (6) do not seem to express complete propositions:

(4) Jack and Jill climbed far enough.

(5) Jack and Jill are ready.

(6) Jill first got married.

On the plausible assumption that what a speaker means must be a complete proposition, it follows that in these cases the speaker cannot mean merely what is expressed by the sentence. What he means must include a completion of what he says, as illustrated in (4c) - (6c) by the contents of the bracketed material:

(4c) Jack and Jill climbed far enough [to get a pail of water].

(5c) Jack and Jill are ready [to get married].

(6c) Jill first got married [and later became pregnant].

A speaker might intend such completions to be understood, but this does not mean that (4) - (6) are syntactically elliptical versions of (4c) - (6c). There is no reason to suppose that they contain suppressed, unpronounced versions of the bracketed material.[7]

Now it might seem if sentences like (4) - (6) do not express complete propositions, nothing is said in utterances of them. It is interesting to observe, however, that utterances of these sentences, and of semantically incomplete sentences in general, pass what I call the IQ Test.[8] That is, there is no problem in indirectly quoting them, even though they do not express complete propositions (let’s imagine they’re uttered by Mother Goose):

(4_IQ) Ms. Goose said that Jack and Jill climbed far enough.

(5_IQ) Ms. Goose said that Jack and Jill are ready.

(6_IQ) Ms. Goose said that Jill first got married first.

So although (4) - (6) do not express complete propositions, Ms. Goose did say something in uttering them. It’s just that in each case what she said, as opposed to what she meant and asserted, was not a complete proposition. The semantic incompleteness of some sentences has no special significance in respect to the distinction between what is said and what is communicated, although it does refute the grammar-school adage that a complete sentence expresses a complete thought. That a sentence, despite being well-formed syntactically, can be semantically deficient is an underappreciated phenomenon of natural language.

Examples like (1) - (3), since they are perfectly ordinary, illustrate the pervasiveness of the gap between linguistic meaning and speaker’s meaning, and those like (4) - (6), though somewhat less common, also illustrate the phenomenon of semantic incompleteness. Such examples have been thought to show that context sensitivity and linguistic meaning are so intermeshed that the notion what is said is a bogus category. As I will argue in the next section, I think this extreme conclusion is unwarranted. It cannot be established by appealing to intuition, introspection, or armchair psychology.

3. Spurious Objections to the Notion of What Is Said

I have endorsed a literalistic notion of what is said, or what François Recanati calls (this issue) “a purely semantic notion of what is said”. Before defending it against various objections, I should indicate what I take to be the relevant difference between semantic and pragmatic. This distinction has been drawn in various ways, but no matter how it is drawn the term ‘semantic’ has something to do with meaning and ‘pragmatic’ has something to do with use.[9] Having previously discussed different formulations of the distinction (Bach, 1999a), here I will simply explain the version of it that I prefer and why I prefer it.

A semantic-pragmatic distinction can be drawn with respect to various things, such as ambiguities, implications, presuppositions, interpretations, knowledge, processes, rules, and principles. For me, it applies fundamentally to types of information. Semantic information is information encoded in what is uttered—these are stable linguistic features of the sentence—together with any extralinguistic information that provides (semantic) values to context-sensitive expressions in what is uttered. Pragmatic information is (extralinguistic) information that arises from an actual act of utterance. Whereas semantic information is encoded in what is uttered, pragmatic information is generated by, or at least made relevant by, the act of uttering it.

This way of characterizing pragmatic information generalizes Grice’s point that what a speaker implicates in saying what he says is carried not by what he says but by his saying it (1989, p. 39). The act of producing the utterance exploits the information encoded but by its very performance creates new or otherwise invokes extralinguistic information. That information, combined with the information encoded, provides the basis for the hearer’s identification of the speaker’s communicative intention. Notice that this pragmatic information is relevant to the hearer’s inference only on the supposition that the speaker is producing the utterance with the intention that the information in question be taken into account. There is no such constraint on contextual information of the semantic kind, which has its effect independently of the speaker’s communicative intention and the hearer’s recognition of that intention. This sort of contextual information is limited to a short list of parameters associated with indexicals and tense, such as the identity of the speaker and the hearer and the time of an utterance.[10] So, for example, “even if I am fully convinced that I am Napoleon, my use of ‘I’ designates me, not him. Similarly, I may be fully convinced that it is 1789, but it does not make my use of ‘now’ about a time in 1789” (Barwise and Perry, 1983, p. 148).

Now we can consider objections to the literalistic or purely semantic notion of what is said. On such a notion, what is said by a speaker in uttering a sentence excludes anything that is determined by his communicative intention (if it included that, then what is said would be partly a pragmatic matter). I am aware of several objections, which I will call the Propositional, the Not-an-Implicature, the Psychological, and the Intuitive objections. Since the first two have in effect been answered in the previous section, here they will be quickly stated and disposed of. The last two deserve more serious attention.

According to the Propositional objection, what is said when somebody utters a sentence must be a proposition. Since, as we have seen, some sentences do not express complete propositions (even after ambiguities and vaguenesses are resolved and indexical references are fixed), nothing is said when such a sentence is uttered. Reply: as illustrated by utterances of sentences (4) - (6), indicative sentences may be indirectly quoted even if they are semantically incomplete—they pass the IQ Test. So there is no basis for the assumption that what is said must be a complete proposition. Besides, why should the syntactic requirements on well-formed sentences preclude the possibility that the semantic interpretation of some well-formed sentences might be an incomplete proposition? It is plausible to suppose that a speaker, when he means something, must mean a complete proposition, but it doesn’t follow that any complete sentence used to communicate a complete proposition must itself express one.

The Not-an-Implicature objection is based on resisting what Robyn Carston has called “the compulsion to treat all pragmatically derived meaning as implicature” (1988, p. 176). Once it is recognized that the contribution of pragmatic processes is not limited to the determination of implicatures, “there is no reason”, according to Carston, “why pragmatics cannot contribute to the explicature, the truth-conditional content of the utterance”, which she equates with what is said. However, insofar as the explicature can be the truth-conditional content of the result of an expansion or completion of the utterance, it cannot be identified with what is said, the proposition, expressed relative to the context, by the uttered sentence. Since, as we have seen, the distinction between what is said and what is implicated is not exhaustive, it is a mistake to suppose that if a conveyed proposition is not an implicature, it must be identified with what is said.

The two main objections to the semantic notion of what is said apply to sentences like (1) - (3), which do express a complete proposition but are typically used to convey not that proposition but an expanded or enriched version of it. Recanati suggests that although the semantic interpretation of such a sentence yields a complete proposition, this “minimal” proposition is an “abstraction with no psychological reality”. He claims that “semantic interpretation by itself is powerless to determinate what is said” in cases like (1) - (3), because “we cannot sever the link between what is said and the speaker’s publicly recognizable intentions”. He reaches these conclusions, as we will see, by adverting to the “pragmatic processes” involved in understanding an utterance and to intuitions about what is said. [11]

The Psychological Objection

This objection assumes that what is said is in the strict, literalistic sense can have no psychological reality unless it is something that a hearer must identify before inferring the speaker’s communicative intention. In developing this objection, Recanati draws several distinctions among what he calls pragmatic processes (the cognitive processes involved in figuring out what a speaker is conveying in uttering a sentence). A primary process yields the first proposition identified by the hearer, whereas any further proposition is arrived at by way of a secondary process. Recanati regards the difference here to be psychologically real, not merely terminological, because only secondary processes are inferential. Primary, or what Ken Taylor (this issue) calls prepropositional, processes cannot be inferential, according to Recanati, because they do not take a proposition as input. He also distinguishes mandatory from optional processes. A process is mandatory if needed to arrive at any proposition at all, and optional otherwise. A process is mandatory if the uttered sentence is semantically incomplete. It is a further question whether or not the sentence actually contains a covert argument slot (needing to be assigned a semantic value for the sentence to express a complete proposition, relative to a given context). So a process is mandatory not because interpreting the sentence requires it but because the semantic interpretation of the (semantically incomplete) sentence does not yield a complete proposition. Finally, notice (this is key to Recanati’s objection) that a process can be primary without being mandatory, because the first proposition that a hearer actually arrives at might not be the one expressed by the sentence in the context.

These are interesting distinctions, although, as Taylor points out, it is not obvious that either of these distinctions marks a psychologically real difference. Perhaps a primary process can be inferential if, when what is said is a complete proposition, it by-passes that proposition, or even if it begins with an incomplete proposition. After all, there is a distinction to be drawn between the possibly non-propositional semantic interpretation of a sentence ‘S’ and the hearer’s thought that the speaker said that S. Since, as we have seen, utterances of semantically incomplete sentences pass the IQ test, the hearer can think that the speaker said that S and proceed to make an inference about why he said it, i.e., about his communicative intention in saying it, even though what is said is not a complete proposition.

The process of utterance comprehension is obviously a very interesting topic for psychology, but it’s hard to see why facts about hearers’ cognitive processes should be relevant to semantics. The relevant question is whether the nature of these pragmatic processes, whatever the details, has any bearing on the legitimacy of a purely semantic notion of what is said. It is a mystery to me why facts about what the hearer does in order to understand what the speaker says should be relevant to what the speaker says in the first place. How could the fact (if it is a fact) that what is said sometimes has no psychological reality for the hearer show that it is a mere abstraction? All this shows is that hearers can infer what a speaker is communicating without first identifying what the speaker is saying. Employing the semantic notion of what is said does not commit one to an account of the temporal order or other details of the process of understanding. This notion pertains to the character of the information available to the hearer in the process of identifying what the speaker is communicating, not to how that information is exploited (Bach and Harnish, 1979, pp. 91-93).

Besides, suppose that it is true that what is said, in the strict, semantic sense, is sometimes not consciously accessed. It is still consciously accessible. This is evident from the fact that people recognize (as we saw with (1) - (3) above) that expansions of it are cancellable. Furthermore, even if in some cases the “minimal” proposition “is not actually computed and plays no role in the interpretation process as it actually occurs” (Recanati, 1993, p. 318), e.g. because of what Recanati calls “local processing” on constituents of the sentence, it can still play a role. Even if a hearer doesn’t explicitly represent what is said by the utterance of a sentence, hence does not explicitly exclude it from being what is meant, still he makes the implicit assumption that it is not what is meant. Implicit assumptions are an essential ingredient in default reasoning in general (Bach, 1984) and in the process of understanding utterances in particular. Communicative reasoning, like default reasoning in general, is a case of jumping to conclusions without explicitly taking into account all alternatives or all relevant considerations. Even so, to be warranted such reasoning must be sensitive to such considerations. This means that such considerations can play a dispositional role even when they do not play an explicit role. They lurk in the background, so to speak, waiting to be taken into account when there is special reason to do so.

The Intuitive Objection

Here the complaint is that the literalistic notion of what is said disregards ordinary intuitions about the truth or falsity, hence about the content, of what is said. For example, it may seem that sentence (7), as uttered on a certain day,

(7) I haven’t taken a bath.

is true if the speaker hasn’t taken a bath that day, although on the literalistic view what he said is false if he has ever taken a bath. Of course what the speaker means is true, but that includes the implicit qualification ‘today’, which is not present in the sentence (it is doubtful that a time frame variable is present either). Since the implicit qualification does not affect what is said but is understood right along with what is said, it can seem counterintuitive that what is said is really false. In fact, this is often given as a reason for adopting a more inclusive conception of what is said, one that is not limited to the “minimal proposition” that the literalistic conception of what is said would insist on.

Recanati even formulates a principle, the Availability Principle, which prescribes that intuitions about what is said be “preserved” in our theorizing. But he needs to explain why the intuitions he takes to concern what is said actually do concern that, as well as why we should be confident in their accuracy. Of course it is true that we have a perfectly good handle on what our words mean, on how they are put together into sentences, hence on what our sentences mean. Who could deny that competent speakers have such knowledge? Grice himself regarded it “as a sort of paradox [that] if we, as speakers, have the requisite knowledge of the conventional meaning of sentences we employ to implicate, when uttering them, something the implication of which depends on the conventional meaning in question, how can we, as theorists, have difficulty with respect to just those cases in deciding where conventional meaning ends and implicature begins?” (1989, p. 49). Grice’s paradox may seem especially troubling if it is supposed, as it often is, that accounting for our ordinary judgments about the truth-conditions of sentences is the central aim of semantics.

This worry is unfounded. It is the central aim of semantics to account for semantic facts, not intuitions. People’s spontaneous judgments or “intuitions” provide data for semantics, but it is an open question to what extent they reveal semantic facts and should therefore be explained rather than explained away. Since they are often responsive to non-semantic information, to what is implicit in what is said, they should not be given too much weight. Besides, these intuitions don’t seem to play a role in ordinary communication. In the course of speaking and listening to one another, we generally do not consciously reflect on the semantic content of the sentences we hear or on what is said in their utterance. We are focused on what we are communicating and on what is being communicated to us, not on what is said. Moreover, we don’t have to be able to make accurate judgments about what information is semantic and what is not in order to be sensitive to semantic information. To “preserve intuitions” in our theorizing about what is said would be like relying on the intuitions of unsophisticated moviegoers about the effects of editing on a film. Although people’s cinematic experience is dramatically affected by cuts and camera angles, there is no reason to suppose that their intuitions are reliable about these effects or about how they are produced. Intuitions about what is said may be similarly insensitive to the difference between the contribution that is made by the semantic content of a sentence and that made by extralinguistic factors to what an utterance communicates. What worried Grice was not a real paradox.[12]

Moreover, it is easy to sensitize people’s intuitions about what is said to Grice’s cancellability test for what is not said. Just present them with sentences like (1) - (3) followed by cancellations of what is not explicit in the utterance, as in (1xx) - (3xx),

(1xx) Jack and Jill went up the hill but not together.

(2xx) Jack and Jill are engaged but not to each other.

(3xx) Jill got married and became pregnant but not in that order.

and see if they sense a contradiction or just a clarification. Or ask them whether what is said by an utterance of explicitly expanded versions of (1), (2), or (3) is the same as what is said by utterances of (1), (2), and (3) themselves:

(1ee) Jack and Jill went up the hill together.

(1) Jack and Jill went up the hill.

(2ee) Jack and Jill are engaged to each other.

(2) Jack and Jill are engaged.

(3ee) Jill got married and then became pregnant.

(3) Jill got married and became pregnant.

They are likely to say that the explicitly expanded utterances say something not said by the original ones. Similarly, ask them whether what a speaker says in uttering (7ee) is the same as in uttering (7),

(7ee) Howard hasn’t taken a bath today.

(7) Howard hasn’t taken a bath.

and they are likely to discern the difference. This contradicts the intuition that in uttering (7) a speaker says that Howard has taken a bath that day. At the very least, the verdict of intuition is less clear when we take into account people’s cancellability judgments and their comparative judgments about what is said by explicitly expanded vs. unexpanded ones.[13]

So it seems that cognitive and intuitive considerations do not undermine a purely semantic notion of what is said. Furthermore, there is a theoretical need for such a notion, namely, to account for the semantic content of an utterance that is independent of the fact that the utterance is actually made. Its semantic content, relative to a context, does not depend on the sentence actually being uttered. The context can be an actual or possible context of utterance. Appealing to pragmatic processes and to intuitions is irrelevant to an account of what is said, or would be said in a hypothetical context, because they do not enter in at the locutionary level. It is only at the illocutionary level that such processes are primary or secondary, mandatory or optional. And it is only at that level that the “intuitive truth-conditions” of an utterance come into play. It is irrelevant that when someone utters (8),

(8) I have had breakfast.

and means that he has already had breakfast that morning that his utterance would be deemed false if he hadn’t had breakfast that day. What is false is what the speaker is claiming, not what he is saying. What he is saying, the content of the sentence relative to the identity of the speaker and time of the utterance, is true even if he hasn’t had breakfast that day, but this is not what people’s intuitive judgments are focused on.

The basic trouble with the various arguments we have considered is their implicit assumption that what is said in uttering a sentence in a context (the semantic content of that sentence relative to that context) can be affected by the speaker’s communicative intention. In fact, what is said is determined linguistically. When a speaker utters a given sentence in a given context, the only intention that is relevant what he is saying is his semantic intention, i.e., his intention concerning the resolution of any ambiguities and the fixing of any indexical references. These may be insufficient to determine a proposition, as when the sentence is semantically incomplete, but that only goes to show that what is said need not be a complete and determinate proposition. This has no special impact on the hearer’s inference to the speaker’s communicative intention (i.e., to what the speaker could plausibly be taken to intend the hearer to take him to mean in saying what he says). Just as the hearer must look for something that is relevantly specific and informative, so, when the sentence is semantically incomplete, he must look for a complete proposition.

4. Diagnosing the Objections: Overlooked Distinctions

How could it even seem that psychological or epistemological considerations about the hearer are relevant to the question of what the speaker says? I believe this is due to overlooking or disregarding certain fundamental distinctions. These distinctions, when heeded, help support a purely semantic notion of what is said. As enumerated below, they all contrast something pragmatic with something semantic. In fact, they derive in one way or another from the basic distinction drawn at the beginning of previous section, between semantic and pragmatic information:

D1. Semantic information is encoded in what is uttered, whereas pragmatic information is generated by, or at least made relevant by, the act of uttering it.

The distinction between semantic and pragmatic information needs to be heeded whenever it is said, as it often is, that the content of an utterance “depends on context” or is “a matter of context”. It is easy to slide from this platitude to the very strong claim that utterance content is “determined by context” (together with linguistic meaning).

That brings in our next distinction:

D2. It is one thing for content to be determined in context and quite another for it to be determined by context.

To say that content is determined in context is just to say that it can vary from context to context, whereas to say that it is determined by context is to say that it is a function of context. The latter occurs only when the sentence contains context-sensitive elements whose semantic values are functions of objective contextual parameters. This is context, i.e., contextual information, in the narrow, semantic sense.

The difference between that and contextual information in the broad, pragmatic sense prompts our next distinction:

D3. It is one thing for content to be determined by context in the sense of being ascertained on the basis of contextual information and quite another for it to be determined by context in the sense of being constituted by contextual factors.

Determination of the first sort is epistemic determination. It does not make it the case that an utterance has a certain content or, in particular, that the speaker has said a certain thing. Rather, it is what is involved in the hearer’s figuring these things out, as well as what the speaker is communicating. Constitutive determination makes it the case that the speaker has said a certain thing and, more generally, that an utterance has a certain content. Only constitutive determination of what is said is relevant to semantics.

What that is can be confusing if we do not heed a further distinction:

D4. What a sentence is most typically used to communicate is one thing; what it means may well be something else.

As illustrated by our earlier examples (1) - (3) and (7) - (8), e.g.,

(7) Howard hasn’t taken a bath.

the most natural use of many sentences is not their literal use. The most natural interpretation of an utterance need not be what is strictly and literally said.

It might be suggested that to the extent that the meaning of the uttered sentence does not determine the content of an utterance, the speaker’s communicative intention does, and that this intention is part of the context. However,

D5. The speaker’s communicative intention is distinct from and not part of the context of utterance.

Now if context were defined so broadly as to include anything other than linguistic meaning that is relevant to determining what a speaker means, then of course the speaker’s intention would part of the context. However, if the context is to play the explanatory role claimed of it, it must be something that is the same for the speaker as it is for his audience, and obviously the role of the speaker’s intention is not the same for both. Context can constrain what the speaker can succeed in communicating given what he says, but it cannot constrain what he intends to communicate in choosing what to say. In implementing his intention in order to achieve communicative success (recognition of his communicative intention), the speaker needs to select words whose utterance in the context will enable the hearer to figure out what he is trying to communicate. But this concerns only the epistemic, not the constitutive, determination of the content of his utterance.

Finally, we need to apply the elementary distinction between sentences and utterances of sentences to draw a distinction between their contents:

D6. The content of a sentence relative to a context is the proposition(s) (possibly incomplete) it expresses relative to that context, whereas the content of its utterance is whatever the speaker is using it to communicate.

In my view, the notion of the content of an utterance of a sentence has no independent theoretical significance. There is just the content of the sentence the speaker is uttering, which, being semantic, is independent of the speaker’s communicative intention, and the content of the speaker’s communicative intention, i.e., what the speaker is using the sentence to communicate. When one hears an utterance, one needs to understand the sentence the speaker is uttering and to recognize the communicative intention with which he is uttering it. Understanding the sentence is independent of context except insofar as there are elements in the sentence whose semantic value are context-relative. Recognizing the speaker’s communicative intention is a matter of figuring out the content of that intention. Utterances themselves (or sentence tokens) have no content apart from the semantic content of the uttered sentence and the content of the speaker’s intention in uttering it.

5. Problematic Issues and Phenomena

There are several unresolved issues to consider. One issue concerns whether semantic incompleteness must be explained by the presence of free variables. It might be argued that, in general, semantically incomplete sentences must contain free variables whose values are somehow supplied by the context, in which case the sentences are incomplete only in the way that sentences containing indexicals are incomplete (see Stanley, 2000). In my view, it is an empirical question whether a sentence that seems to be semantically incomplete really does contain any covert variables, and there is no pat answer to this question—it depends on the type of case. As we will see below, in some cases this is much more plausible than in others. A related issue concerns how to distinguish cases of expansion (of semantically complete sentences) from cases of completion (of semantically incomplete sentences). A third issue arises from the denial that the speaker’s intention is part of the context. This has a fundamental consequence for how we draw the distinction between what is said and what isn’t: if nothing pragmatic affects what is said, then in some cases what is said is not a complete proposition. I will take up take up these issues in reverse order.

Intentions and Semantic Incompleteness

The consequence just mentioned can be appreciated if we consider the rationale for Ken Taylor’s “parametric minimalism”. It begins with the claim that “context-dependent ingredients of sentence meaning more or less tightly constrain the to be contextually determined values of either suppressed or explicit parameters”. The reason for the “more or less” is clear from how pure indexicals like ‘I’ and ‘today’ differ from third-person pronouns like ‘she’ and ‘they’, demonstratives (‘that’, ‘those’), and demonstratives phrases of the form ‘that F’ and ‘those Fs’. As Taylor puts it, with pure “indexicals, there is a precise and well-defined function from objective features of the context, like who is speaking where and when, to semantic values”. He regards this as a special case of something more general: “In the general case, however, the value of the relevant parameter will [merely] be constrained to be of a certain sort”.[14] This is what I have called a referential constraint (Bach, 1994b, pp. 186-92). So, for example, the constraint on ‘she’ is that it be used to refer to some female, and the constraint on ‘that tree’ is that it be used to refer to some tree. The referent is “contextually determined”, i.e., determined in context (not by context) by the speaker’s referential intention. Now here is Taylor’s key point (my italics):

Once determined, the value [of the relevant parameter] will be destined to play a certain role in constituting the completed propositional content of the relevant utterance. But beyond such relatively minimal constraints, the assignment of values will typically be free—that is, entirely dependent on the communicative intentions of the speaker. When parametric values are free in this sense, they typically cannot simply be read off of objective features of the context like who is speaking when, where and to whom. That is why the communication of the contextually determined values of such free but constrained parameters typically presupposes a background of mutual knowledge and the intervention of roughly Gricean considerations like relevance and informativeness.

It follows from my view that these values, when not “read off of objective features of the context”, are not semantic values, precisely because they depend on speakers’ communicative intentions. Even though the values are intended for lexical items in the sentence (or syntactically provided slots, in cases where the sentence contains covert free variables), it does not follow that the proposition resulting from supplying the values (filling the slots) is the semantic content of the utterance (i.e., of the sentence relative to the context). So, for example, on my view it follows that if Sam utters a sentence like (9) to ascribe a property to a certain object he is demonstrating,

(9) That tree is deciduous.

what he is saying, in the strict semantic sense, is not the singular proposition that t (the tree in question) is deciduous (<t, being deciduous>). That counts as the content of his statement but not as what he is saying (besides, it can’t be all of what he is saying, since its truth requires that t be a tree). One reason for denying that what is said by (9) is a singular proposition is that ‘that’ can be used nondemonstratively to make descriptive, rather than objectual, reference (Bach 1994b, p. 66), say to whichever tree dropped leaves on one’s windshield. In that case, no singular proposition is expressed. Nor is what is said the (general) proposition that the tree Sam is demonstrating is deciduous, since that he is demonstrating it is not part of what he is saying. Thus, the most accurate specification of Sam said is given neither by ‘Sam said that t is deciduous’ nor ‘Sam said that the tree he was demonstrating was deciduous’. Rather, I suggest, it is given by something like (9_IQ):

(9_IQ) Sam said that [a certain tree] is deciduous.

Here ‘[a certain tree]’ does not specify what tree (if any) Sam was speaking of but, being enclosed in brackets, merely indicates that Sam referred, objectually or descriptively, to some tree. As a report of what Sam said, (9_IQ) does not specify which tree that is or what (descriptive) condition a tree must meet to be the tree in question. Accordingly, (9_IQ) indicates that what Sam said was an open proposition. It does not specify the proposition that results from determining the identity of the tree. Taylor and I agree that which tree this is is determined by Sam’s referential intention (though this intention could be descriptive rather than objectual) and that this intention is part of his communicative intention (see Bach 1992). However, on my view, as reflected by (9_IQ), this determination goes beyond what is said.

Completion or Expansion?

Another issue concerns how to distinguish cases of completion (of semantically incomplete sentences) from cases of expansion (of semantically complete sentences). It is claimed, for example, that to account for restricted uses of incomplete definite descriptions, and for that matter quantificational phrases in general, we must impute covert domain variables to them. Taylor (this issue) supposes this, when he considers that a sentence like (10),

(10) The cat is on the couch.

isn’t normally used to convey, as Taylor puts it, that the only cat in the universe is on the only couch in the universe. He assumes here that the proposition being conveyed is reached not by an expansion of that absurd proposition but only by assigning values to covert domain variables. The expansionary approach requires no such assumption, but it does entail that typical uses of sentences like (10) are sentence-nonliteral. On this approach, as Stephen Neale recognizes, one can invoke “the distinction between the proposition expressed and the proposition(s) meant” and maintain that the incompleteness of the descriptions is of “no [semantic] consequence”. He “know[s] of knock-down argument against this approach”, but thinks that because “context-dependence is such a ubiquitous feature of the use of natural language”, this approach is likely to conflict with “our intuitive ascriptions of truth and falsity” (1990, pp. 114-5). In my view, its ubiquity explains why such intuitions, if assumed to be semantic, are unreliable. After all (to recall the fourth distinction noted in the previous section), a typical use of a sentence need not be a literal use of it. As I have argue at length elsewhere (Bach, 2000), there is no need for any covertly marked restrictions on incomplete definite descriptions or, for that matter, on incomplete quantificational phrases in general.[15]

The Referential-Attributive Distinction

The false assumption that a typical use of a sentence must be a literal use of it underlies

Sperber and Wilson’s (1986) relevance-theoretic notion of explicature and their thesis of ubiquitous semantic underdetermination. To illustrate the role this assumption can play, consider the thesis, defended in an excellent article by Anne Bezuidenhout (1997), that definite descriptions are semantically underdeterminate with respect to the referential-attributive distinction. Her reason for claiming this is that sentences containing definite descriptions can be used equally to convey either object-dependent or quantified propositions, depending on whether the use is referential or attributive. The argument is essentially this:

1. Description sentences are not semantically ambiguous.

2. They do not univocally express object-dependent propositions.

3. They are not used to convey object-dependent propositions by way of implicature from quantified propositions.

4. The way they are used to convey object-dependent propositions is on a par with the way they are used to convey quantified propositions.

5. Therefore, description sentences express neither object-dependent nor quantified propositions but merely incomplete propositions.

The conclusion says, in other words, that description sentences are semantically underdeterminate (incomplete) with respect to which type of proposition they can be used to convey, hence that descriptions are underdeterminate as between referential and attributive uses.

The trouble with this argument is premise 4. The only evidence presented to show that the way description sentences are used to convey object-dependent propositions is on a par with the way they are used to convey quantified propositions derives from introspection, which reveals no conscious difference between how referential and attributive uses are processed. However, nothing follows from this concerning the semantic content of description sentences. All it shows is that referential uses are typical uses (their typicality streamlines the hearer’s inferential process), not that they are literal uses. The above argument illustrates the fallacy of drawing inferences about semantics from the psychology of the hearer.

Lexical Alternations

There is an interesting complication to the problem of deciding whether or not a sentence whose utterance requires completion contains a covert variable. As the following examples illustrate, there are many lexical items which can occur without complements but which have near synonyms that cannot so occur, as indicated by the asterisk:

(11) a. Fred finished yesterday.

b. *Fred completed yesterday.

(12) a. Sam ate earlier.

b. *Sam devoured earlier.

(13) a. Jack tried three times.

b. *Jack attempted three times.

The b. sentences are ungrammatical because their verbs require an explicit complement. In contrast, the a. sentences, though semantically incomplete, are syntactically complete (well-formed). To understand what the speaker is conveying in uttering these sentences, the hearer must infer some completion, e.g., ‘the job,’ ‘lunch,’ and ‘to call Jill’. The question is whether there is any theoretical reason to impute a covert syntactic node corresponding to the complement argument slot for the verbs ‘finish’, ‘eat’, and ‘try’. It is not clear what the fact that the semantically incomplete a. sentences are grammatical indicates about their covert syntax. Does it suggest that they don’t have covert complements, or merely that they don’t have to have overt complements? So far as I know, this is an unsettled issue in linguistics, but until it is settled, we should not just assume that a sentence can be semantically incomplete only in virtue of containing a lexically generated slot. We need to distinguish sentences which, like sentences containing indexicals, contain constituents in need of semantic values, from sentences which are simply semantically incomplete.

Weather Predicates

The question of covert variables arises in the special case of weather predicates, as in ‘It is raining/snowing/foggy/sunny’. Taylor claims that the verb ‘rain’ has a lexically specified argument place which takes locations as values. He gives no argument for this claim beyond pointing out that rainings are a kind of event that locations undergo and suggesting that this is thematically marked in the argument structure of the verb. The question is whether the verb has that argument structure. If someone utters,

(14) It is raining.

it does seem that the utterance is true only relative to a location. In contrast to ‘It is raining somewhere’, an utterance of (14) is not true merely because it is raining somewhere; it is true only relative to some contextually salient location.[16] Is the fact that (14) is semantically incomplete in this way best explained by the hypothesis that it contains a covert location argument? Taylor is right to say that “the semantic incompleteness [of (14)] is manifest to us as a felt inability to evaluate the truth value of an utterance of [(14)] in the absence of a contextually provided location”, but it does not follow that this felt inability “has its source in our tacit cognition of the syntactically unexpressed argument place of the verb ‘to rain’.” Again, it should not be assumed that a sentence can be semantically incomplete only because it contains a lexically generated slot in need of a semantic value.

Even if (14)’s semantic incompleteness does not show that it contains a covert location argument, perhaps this is shown by the fact that it can be embedded in a quantificational context, such as (15):

(15) Whenever it is raining, most people use umbrellas.

There is no requirement that there be some one contextually salient location where all the relevant rainings are taking place. Moreover, it seems that the truth of an utterance of (15) requires only that most people use umbrellas when it is raining where they are, not when it is raining anywhere. However, it seems to me that it is a semantic illusion that the truth of an utterance of (15) requires only that most people use umbrellas when it is raining where they are. What the truth of an utterance of (15) really requires is that most people use umbrellas when they are out in the rain. However, there is nothing in sentence (15) to suggest that this is semantically required, i.e., that this is what an utterance of (15) says, as opposed to what it communicates. Our real-world knowledge (about when people use umbrellas) intrudes here, but that knowledge is relevant to determining what is communicated, not what is said.

There is another difficulty with the covert argument hypothesis. If ‘rain’ were really marked for a location argument, one would expect noun phrases that designate locations to be able to occur as subjects of ‘rain’. Instead of having to say, for example,

(16) It rains frequently in Seattle.

we would expect to be able to say,

(17) *Seattle rains frequently.

But this is obviously ungrammatical. Indeed, with ‘rain’ and weather predicates generally, an expletive subject is required. And location is provided not by an argument place but by an adjunct, such as the prepositional phrase ‘in Seattle’, which appears to play the same role in (16) as it plays in ‘The Mariners play in Seattle’.

It doesn’t follow from the fact that raining happens to locations and that (14) does not express a complete proposition that the sentence contains a covert location variable. Formal semanticists can use a free variable in representing (14), but that doesn’t mean that the sentence itself contains one. The fact that what is conveyed in an utterance of (14) is a proposition with an unarticulated constituent (Perry 1986) does not show that the sentence used to convey it contains a covert syntactic constituent whose value is that unarticulated propositional constituent.

Short Passives

The last point can be appreciated if we consider the case of short passives, for which there is reason to posit a covert argument slot. With (18), for example,

(18) The land was sold immediately to pay estate taxes.

it is plausible to suppose that an implicit agent argument is associated with ‘was sold’. That there is an implicit argument in (18) is suggested by how it differs from its intransitive counterpart,

(18_int) *The land sold immediately to pay estate taxes.

There is something syntactically anomalous about (18_int), although the same real-world knowledge, that something can be sold only if there is a seller, would seem to account for how the covert subject of the infinitive is understood. Despite their semantic similarity, there is a difference in how quickly people interpret them. Mauner and Koenig (1999) have found empirical evidence for this difference, namely that subjects take longer to interpret sentences like (18_int) than ones like (18).[17] Presumably that is due to a deeper syntactic difference. A plausible explanation for this difference is that the covert subject (the big PRO of GB theory) of the infinitival rationale clause must be controlled by an antecedent: whereas in (18) there is such an antecedent, a covert agent variable associated with the passive ‘was sold’,

(18) The land was sold [by] x immediately PRO_x to pay estate taxes.

there could be no such variable in intransitive (18_int), hence no antecedent there for PRO.

Relational Terms

There is a good case for positing implicit arguments in the case of relational expressions, such as ‘stranger,’ ‘local,’ ‘enemy’, ‘neighbor’, ‘protégé’, ‘familiar’, and ‘distant’. Consider semantically incomplete sentences like (19) and (20),

(19) Ned annoyed the neighbors.

(20) Fred saw a familiar face.

One reason for supposing that they contain covert variables is that they admit of readings that may be plausibly explained on the model of anaphora. (19) can be used to say that Ned annoyed Ned’s neighbors, not just that Ned annoyed the speaker’s neighbors (or, if the speaker is not at home, neighbors of the residence where the speaker is located). Similarly, (20) can be used to say that Fred saw a face familiar to Fred, not just that he saw a face familiar to the speaker. Moreover, as Barbara Partee (1989) has pointed out, relational terms have what seem best explained as bound uses, as in (21) and (22),

(21) People should be friendly to neighbors.

(22) The children were happy to see familiar faces.

(21) seems to say that people should be friendly to their respective neighbors, and (22) seems to say that the children were happy to see faces familiar to them respectively.

Possessives

Possessive phrases seem to examplify what may be called phrasal semantic underdetermination. It has long been recognized that ‘John’s car’, for example, need not be used to mean ‘the car that John owns’. Accordingly, Recanati (this issue) suggests that it means something like ‘the car that bears relation R to John’, where ‘R’ is a free variable whose value is contextually assigned (i.e., is whatever the speaker intends it to be).[18] It is true that ‘John’s car’ might be used to refer to the car John wants to buy or the car John is rooting for, etc. However, it is not clear that these unusual uses are literal uses. Possessive phrases are not so-called for nothing, and ownership is only one kind of possession. John’s car could be the one he has borrowed, the one he has rented, the one he is driving, or even the one he has stolen. There seems to be a certain primacy to the multifarious relation of possession, from which a possessive phrase’s more unusual uses derive. If so, then it is not plausible that there is a free relational variable in possessive phrases.

Polysemy and Adjectival Modification

The phenomenon of polysemy gives rise to plausible cases of phrasal semantic underdetermination. Consider the adjectives ‘sad,’ ‘long,’ ‘dangerous’, ‘generous’, and ‘conscious’, as they occur in the following phrases:

(23) a. sad person / sad face / sad day / sad music

b. dangerous drug / dangerous game / dangerous road

c. fast car / fast track / fast race

d. generous donor / generous gift

e. conscious being / conscious state

Obviously the most plausible import of each of these adjectives varies with the noun it modifies. Indeed, only in the first phrase in each case does the adjective express a property that directly belongs to what the noun denotes. For example, a person can have the property of being sad, but a face, a day, or a piece of music cannot. And a speaker is not likely to ascribe that property other than to sentient beings. However, it is extralinguistic knowledge, not the semantics of these phrases, that keeps the speaker from doing so and the hearer from so taking them. On the other hand, considering that the adjectives in (23) have counterparts in other languages with a similar range of uses, it is not because of ambiguity (linguistic coincidence) that these adjectives have their different uses. Does that mean that their various possible uses can be given a pragmatic explanation? There is an alternative possibility, namely that polysemy involves what James Pustejovsky (1995) calls co-compositionality: what varies from case to case is not a term’s semantic properties but how those properties interact with those of the term it is construction with. Although I am not persuaded by Pustejovsky’s ambitious theory of how this works, I think it is a definite improvement over what he calls “sense enumeration lexicons” (1995, p. 29). Whatever the correct account, polysemy of the sort illustrated in (23) seems too systematic to be relegated to pragmatics.

In and of itself, the semantic incompleteness of a sentence does not show that it contains a covert variable or other context-relative parameter. Whether or not it does is a linguistic question, and appealing to intuition alone does not answer the question in any given case. Intuition may tell us when a sentence of a certain sort does not express a complete proposition, but it does not tell us whether what is missing is the value of a covert variable. Obviously a free variable can be inserted into the semantic representation of the sentence, but this doesn’t show that the alleged covert variable is really there. Linguistic considerations, bolstered by psycholinguistic evidence, are needed to show that. It is rash to assume that a sentence can be semantically incomplete only because it contains a covert free variable. The sentence may simply be semantically underdeterminate.

6. SUMMING UP

In defending a purely semantic conception of what is said, I have stressed the following points:

· The distinction between the semantic interpretation of a sentence and the pragmatic interpretation of an utterance needs to be recognized and respected.

· What counts as what is said in an utterance of a sentence consists of nothing more than its semantic interpretation, that is, the (possibly incomplete) proposition semantically expressed by the sentence (with respect to the context of utterance).

· Free variables should not be posited too freely.

· The nature and role of context that is relevant to semantics is limited to the semantic interpretation of indexical elements.

· The speaker’s communicative intention is not part of the context.

· Intuitions about what is said should be respected only insofar as they are semantic intuitions (not that they play any role in the process of communication).

· Questions about the nature of the cognitive processes involved in understanding an utterance have little if anything to do with the proper formulation of the semantics-pragmatics distinction or with drawing the line between what is said and what isn’t.

References

Austin, J. L.: 1962, How To Do Things With Words, Oxford University Press, Oxford.

Bach, Kent: 1984, ‘Default Reasoning: Jumping to Conclusions and Knowing When To Think Twice’, Pacific Philosophical Quarterly, 65, pp. 37-58.

Bach, Kent: 1992, ‘Intentions and Demonstrations’, Analysis 52, pp. 140-6.

Bach, Kent: 1994a, ‘Conversational Impliciture’, Mind & Language 9, pp. 124-62.

Bach, Kent: 1994b, Thought and Reference, paperback edition, revised with postscript, Oxford University Press, Oxford.

Bach, Kent: 1995, ‘Standardization and Conventionalization’, Linguistics and Philosophy 18, pp. 677-86.

Bach, Kent: 1998, ‘Standardization Revisited’, in A. Kasher (ed.), Pragmatics: Critical Concepts, vol. IV, Routledge, London, pp. 712-22.

Bach, Kent: 1999a, ‘The Semantics-Pragmatics Distinction: What It Is and Why It Matters’, in Ken Turner (ed.), The Semantics-Pragmatics Interface from Different Points of View, Elsevier, Oxford, pp. 65-84.

Bach, Kent: 1999b, ‘The Myth of Conventional Implicature’, Linguistics and Philosophy 22, pp. 327-66.

Bach, Kent: 2000, ‘Quantification, Qualification, and Context’, Mind & Language 15.

Bach, Kent and Robert Harnish: 1979, Linguistic Communication and Speech Acts, MIT Press, Cambridge, MA.

Barker, Chris: 1995, Possessive Descriptions, CSLI Publications, Stanford.

Barwise, Jon and John Perry: 1983, Situations and Attitudes, MIT Press, Cambridge, MA.

Bezuidenhout, Anne: 1997, ‘Semantics, Pragmatics, Semantic Underdetermination, and the Referential/Attributive Distinction’, Mind 106, pp. 375-409.

Carston, Robyn: 1988, ‘Implicature, Explicature, and Truth-theoretic Semantics’, in R. M. Kempson (ed.), Mental Representations: The interface Between Language and Reality, Cambridge University Press, Cambridge, Eng., pp. 155-81.

Gibbs, Raymond and Jessica Moise: 1997, ‘Pragmatics in Understanding What is Said’, Cognition 62, pp. 51-74.

Grice, Paul: 1989, Studies in the Way of Words, Harvard University Press, Cambridge, MA.

King, Jeffrey: 1995, ‘Structured Propositions and Complex Predicates’, Noûs 29: 516-35.

Levinson, Stephen: 2000, Presumptive Meanings: The Theory of Generalized Conversational Implicatures, MIT Press, Cambridge, MA.

Mauner, Gail and Jean-Pierre Koenig: 1999, ‘Lexical Encoding of Event Participant Information’, Brain and Language 68, pp. 178-84.

May, Robert: 1991, ‘Syntax, Semantics, Logical Form’, in A. Kasher (ed.), The Chomskyan Turn, Blackwell, Oxford, pp. 334-59.

Neale, Stephen: 1990, Descriptions, MIT Press, Cambridge, MA.

Partee, Barbara: 1989, ‘Binding Implicit Variables in Quantified Contexts’, Papers from the Chicago Linguistics Society 25, pp. 342-65.

Partee, Barbara: 1995, ‘Lexical Semantics and Compositionality’, in L. Gleitman and M. Liberman (eds.), An Invitation to Cognitive Science, Volume 1: Language, 2nd edition, MIT Press, Cambridge, MA, pp. 311-60.

Perry, John: 1986, ‘Thought Without Representation’, Proceedings of the Aristotelian Society, Supp. Vol. 60, pp. 137-51.

Predelli, Stefano: 1998, ‘I Am Not Here Now’, Analysis 58, pp. 107-15.

Pustejovsky, James: 1995, The Generative Lexicon, MIT Press, Cambridge, MA.

Recanati, François: 1989, ‘The Pragmatics of What is Said’, Mind and Language 4, pp. 294-328.

Recanati, François: 1993, Direct Reference: From Language to Thought, Blackwell, Oxford.

Recanati, François: 1995, ‘The alleged priority of literal interpretation’, Cognitive Science 19, pp. 207-32.

Recanati, François: this issue, ‘What is Said’.

Sperber, Dan and Deirdre Wilson: 1986, Relevance, Cambridge, MA, Harvard University Press.

Stanley, Jason: 2000, ‘Context and Logical Form’, Linguistics & Philosophy 23.

Stanley, Jason and Zoltan Szabó: 2000, ‘On Quantifier Domain Restriction’, Mind & Language 15.

Taylor, Kenneth: this issue, ‘Sex, Breakfast, and Descriptus Interruptus’.

NOTES