Quantification, Qualification, and Context:

A Reply to Stanley and Szabó

 

 Kent Bach

 

 

 

Lawyer: My client didn’t kill his partner. And if he did, he didn’t do it intentionally.

 

 

1. Speaking Loosely

 

We hardly ever mean exactly what we say. I don’t mean that we generally speak figuratively or that we’re generally insincere. Rather, I mean that we generally speak loosely, omitting words that could have made what we meant more explicit and letting our audience fill in the gaps. Language works far more efficiently when we do that. Literalism can have its virtues, as when we’re drawing up a contract, programming a computer, or writing a philosophy paper, but we generally opt for efficiency over explicitness. In most conversation, though, spelling things out is not only unnecessary, it just slows things down, given the articulatory bottleneck in linguistic communication.[1] It is often misleading too, insofar as it guards against something that doesn’t need to be guarded against.

        Here are some simple examples of the kind of loose talk I have in mind. The words in brackets are not part of the sentence (these are not cases of syntactic ellipsis) but the speaker could have uttered them if he deemed that necessary.

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

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

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

            (4)         You’ll succeed if [and only if] you works hard.

            (5)         There are [approximately] 30 students per class.

            (6)         France is hexagonal [roughly speaking].

            (7)         Otto has [exactly] three cars.

            (8)         Felix has always been an honest judge [since he’s been a judge].

            (9)         Adele hasn’t had lunch [today].

            (10)       You’re not going to die [from this cut].

What the speaker means is a qualified version of what he is saying. He would have to utter the words in brackets (or roughly equivalent words—the exact words don’t matter) to make what he meant more explicit.[2] Let’s call the fuller version an ‘expansion’ of the original and what is left out its ‘implicit qualification’. Using a sentence to convey what an expansion of it would express (more) explicitly is the main species of what I call conversational ‘impliciture’ (Bach 1994a), as opposed to implic-a-ture.[3] So, for example, someone who utters (2) is not saying that Jack and Jill are married to each other, any more than he would be saying this with ‘Jack and his sister Jill are married’. Rather, this is implicit in what he is saying or, more precisely, in his saying of it.

        Loose talk of this sort is not a matter of using particular words or phrases nonliterally, or of using vague or ambiguous words. It is not a case of constituent nonliterality, such as metaphor and metonymy, but of what I call sentence nonliterality (Bach 1994b, pp. 71-72). Using sentences nonliterally in this way is so common that we tend neither to be aware of doing it nor to think of it as not literal when others do it. But we do it all the time (as I did just then). Rather than insert extra words into our utterances so as to make fully explicit what we mean, we allow our listeners to read things into them. Being inexplicit is a way of not being literal, though intuitively we may not think of it as such because no words are being used figuratively.[4] Cases like the ones above pass Grice’s test of cancellability. That is, the implicit material is not part of what is said because it may be taken back without contradiction. For example, there is no contradiction in saying, ‘Jack and Jill went up the hill (1) but not together’ or ‘Jill got married and got pregnant (3) but not in that order’.

        What is likely to be communicated in utterances of (1) - (10) is identical to what would be (semantically) expressed by their explicit expansions. This does not mean, however, that what their expansions would express is identical to what the unexpanded sentences do express, even relative to the context. (1) - (10) are not elliptical versions of their expansions, literally containing unvocalized (‘phonologically null’, in linguistic jargon) occurrences of the unuttered material. Nor does it mean that they contain contextual variables whose values are supplied by the context.[5] Making either inference would be to commit what Barwise and Perry call the ‘fallacy of misplaced information’, by mistakenly assuming that ‘all the information in an utterance must come from its interpretation’ (Barwise and Perry 1983, p. 38). Committing this fallacy leads to mislocating certain phenomena, in this case implicit qualification, at some level of linguistic representation, when in fact the phenomenon in question is entirely pragmatic, attributable to sentence use rather than to anything in the sentence.

        Finally, notice that in some cases, even though what is being conveyed is true (let’s assume), the unembellished sentence is likely to be false, at least on a literalistic construal of what is said (as opposed to what is conveyed). Contrary to what (8), (9), and (10) say, Felix hasn’t always been a judge, much less an honest one, Adele has had lunch in the past, and the addressee is going to die. However, the literal falsity of these sentences has nothing to do with how the hearer figures out what the speaker is conveying or with how the speaker intends him to do so. These are not like Gricean quality implicatures, in which an utterance of, e.g., the obviously false ‘I could eat a million of those potato chips’ conveys how irresistible they are. What triggers the hearer’s search for something other than what is said is not its obvious falsity but its lack of relevant specificity. So, for example, if an oncologist says to a cancer patient who demands a frank prognosis, ‘You’re going to die’, the oncologist is conveying that the patient will die from the cancer, not merely that he is mortal. What he is saying, however, is not obviously false but literally true, though not informative—he could have said that to anybody and spoken truly.[6] Similarly, utterances of the negations of (1) - (9) would typically have the same implicit qualifications as utterances of (1) - (9) themselves.

 

2. Quantification and Qualification

Everyday conversation is replete with utterances like (1) - (10). You can do your own sampling to verify that this is a perfectly ordinary phenomenon. The question before us is whether the case of so-called quantifier domain restriction is any different.[7] Offhand, it seems that examples like (11) - (15) illustrate the same phenomenon:

            (11)       Every bottle [I just bought] is empty.

            (12)       Only two students [in this class] know the capital of Latvia.

            (13)       Imelda had no shoes [appropriate to the occasion] to wear.

            (14)       Most people [who vote in the next U.S. election] will vote Democrat.

            (15)       Nobody [famous] goes there any more—it’s too crowded.

Included in the brackets are words the speaker did not use but might have used (again, the exact words don’t matter) to make explicit the qualification he intends on what/who he is talking about. He doesn’t have to do so, of course, since (in normal contexts) he can rely on his audience’s ability to figure out what the intended qualification is. The bracketed qualifications are merely implicit—the relevant adjective, prepositional phrase, or restrictive relative clause is not present at any level of linguistic description. The sentences themselves contain only the explicit restrictions imposed by the nouns.[8] For example, as Stanley and Szabó point out, instead of uttering (11) the speaker (a certain Lisa) could have conveyed what she meant more explicitly by uttering (11+),

            (11+)     Every bottle I just bought is empty.

        Rather than leave it at that, S&S use the example of (11) to set up their problem. As they explain, ‘the difference between the grammatical, the semantic, and the pragmatic solutions lies in the way they spell out the relationship between Lisa’s actual utterance of [(11)] and her hypothetical utterance of [(11+)]’ (p. 25 ms). The grammatical, or syntactic ellipsis, approach says that although what is articulated (the sentences individuated phonologically) in the two utterances are different, what is uttered is the same grammatical sentence (the reduced relative clause ‘I just bought’ is somehow present in (11) but unpronounced). The semantic approach allows that (11) and (11+) are utterances of distinct grammatical sentences but claims that, relative to the context, these sentences semantically express the same proposition (‘what is said’ is the same). And the pragmatic approach maintains that even relative to the context these sentences express distinct propositions, but that what the speaker is communicating is the same.[9]

        S&S reject pragmatic approaches to cases like (11) - (15). On their semantic approach, the quantified noun phrases in these sentences contain a domain variable, whose value is somehow ‘provided’ or ‘assigned’ by ‘context’. The semantic approach allows that there can be literal uses of (11) to convey the (false) proposition that every bottle in the universe is empty, but it construes this as a limiting case of domain restriction. S&S’s  strategy is to show that ‘all versions’ of the pragmatic approach, as well as the syntactic ellipsis approach,[10] are ‘incorrect. It then follows that the role played by context in the providing of a contextually restricted domain to quantifier expressions is semantic’ (p. 33 ms). Needless to say, this assumes that they have considered all versions of the pragmatic approach, that none can be defended from their objections, and that the semantic approach does not have deficiencies of its own. But at least S&S, unlike many others, recognize that the semantic approach has to be argued for.[11]

        The pragmatic approach treats quantifier domain restriction like other cases of implicit qualification. Since the implicit restriction on the quantified noun phrase does not affect what is said but contributes only to its expansion, what is said often turns out to be obviously false, despite intuitions ostensibly to the contrary. S&S find it ‘worrisome’ that on the pragmatic approach ‘one has to abandon ordinary intuitions concerning the truth or falsity of most sentences containing quantifiers’, because ‘accounting for our ordinary judgments about the truth-conditions of various sentences is the central aim of semantics’ (p. 40 ms). Now I should have thought that the central aim of semantics is to account for semantic facts. ‘Ordinary judgments’ or ‘intuitions’ provide data for semantics, but it is an open question to what extent they reveal semantic facts, hence should be explained rather than explained away. [12] Since they are often responsive to non-semantic information, they should not be given too much weight. Besides, they don’t seem to play a role in ordinary communication. People don’t have to be able to make accurate judgments about semantic facts to be sensitive to semantic information. In the course of speaking and listening to one another, people do not consciously reflect on the propositions semantically expressed by the sentences they hear but are focused on what they are communicating and on what is being communicated to them.

        S&S mistakenly suppose that it is essential to the pragmatic approach to claim that in these cases the speaker is exploiting the obvious falsity of the proposition that is semantically expressed. So, for example, when Lisa uses (11) to convey that every bottle she bought is empty, even though the sentence itself expresses ‘the obviously false proposition that every bottle in the universe is empty’,[13] supposedly she relies on the audience to recognize its obvious falsity and on that basis look for ‘a contextual elimination of the pragmatic anomaly’ (p. 31 ms). Using ‘broadly Gricean’ reasoning to explain the ostensible violation of the maxim of quality, the audience ‘looks for some other proposition that the speaker may reasonably be taken to have meant. We call any such approach a pragmatic approach to quantifier domain restriction’ (p. 32 ms).

        This is an implausible rendition of the pragmatic approach. In contrast to typical cases of metaphor, for example, obvious falsity is not the pragmatic anomaly that triggers the audience’s search for an intended domain restriction or other narrowing of content. Consider a case where the pragmatic approach uncontroversially applies. A child is crying because of a tiny cut and his mother calms him by uttering (10), ‘You’re not going to die’.

S&S assume that the pragmatic approach, which they find ‘plausible’ in this case, appeals to the obvious falsity of what she said, since presumably the mother is not assuring her child of his ultimate immortality. But the operative pragmatic anomaly here is not obvious falsity but lack of relevant specificity. Similarly, S&S assume that according to ‘pragmatic approaches,’ someone who hears an utterance of a sentence like (11) - (15) recognizes its literal falsity and ‘uses general pragmatic principles to infer from information available in the context that the speaker intended to communicate a proposition concerning a more restricted domain of quantification’ (p. 39 ms). They dismiss any pragmatic approach that might appeal to ‘some other peculiarity’ because, at least ‘in the particular case of quantifier domain restriction, [it is] unclear to us what this alternative peculiarity could be’. But why couldn’t it be obvious lack of relevant specificity?

        Besides, obvious falsity can’t be the operative anomaly, since obvious truth can have the same effect, as with the negation of (11),

            (~11)     Not every bottle is empty.

It is just as obviously true that not every bottle (in the universe) is empty as it is obviously false that every bottle (in the universe) is empty, but that wouldn’t prevent Lisa, if she uttered (~11), from conveying the proposition that not every bottle she just bought is empty. Whether she utters the obviously false (11) or the obviously true (~11), it should be obvious to her audience that she’s not talking about every bottle in creation (why would she be doing that?). Indeed, that should be obvious before she even completes the sentence.

        Notice here that the implicit quantification in Lisa’s utterance of ‘Not every bottle is empty’, like that in ‘You’re going to die’, passes Grice’s cancellability test for not being part of what is said. Just as the oncologist could say to his terrified patient, ‘You’re going to die—but not from this cancer’, without contradicting himself, so Lisa could say, ‘Not every bottle is empty—but every bottle I bought is empty,’ without contradicting herself. Of course these would be odd things to say, but that’s another matter.

        As it happens, S&S don’t base their rejection of the pragmatic approach on general considerations, such as those discussed above. In fact, they think that for cases like (11) - (15) the contest between the semantic and the pragmatic approach is a toss-up—there is a trade-off between what they regard as the intuitive plausibility of the semantic approach and the syntactic and semantic simplicity of the pragmatic approach (p. 40 ms). What they regard as decisive is the inability of the pragmatic approach to handle the special case of ‘quantified contexts’. This case, they claim, shows that quantifier domain restriction is an inherently semantic phenomenon. They and I agree that if it is, that would override the default position that implicit domain restriction is a purely pragmatic phenomenon. We will take up (in section 5) this special case and how S&S make use of it after we look at (in section 4) what is involved in the semantic approach to the simpler cases. And before we can do that, we need to examine the difference between what S&S call the semantic and the pragmatic role of context.

 

3. Quantification and Context

It is often said 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). It is not so easy to say what context is, what it does, or how it manages to do what it does. These issues are part of what S&S call the ‘foundational problem’ of context dependence, which they defer. They do not explain how context, whatever it is, restricts quantifier domains—they just assume that context somehow manages to do that. This is unfortunate, since it might turn out that their semantic approach is incompatible with the ultimate solution to that problem.

        Also, in their discussion of the ‘descriptive’ problem, S&S tend to blur certain foundational distinctions. Consider how they pose this problem:

In interpreting the utterances of others, we cannot rely exclusively on the permanent features of the words used. For it is often the case that the very same words could be used in a different context to mean something different. The problem of context dependence is the problem of explaining how context contributes to interpretation, that is, process of determining what a speaker meant by making a linguistic utterance on a certain occasion. (p. 2 ms; my italics)

So what is interpreted is an utterance, what is determined is what a speaker meant by it, and what does the determining is the audience (‘we’), not the context. Later, when S&S distinguish different roles of context, they write,

Interpretation, in the broad sense in which we use this word, proceeds from the sentence articulated to the proposition communicated. Context plays a grammatical role in providing the proper lexical and syntactic analysis of the sentence articulated on a given occasion and thereby determining what was uttered; it plays a semantical role in fixing what was said by that occurrence; and finally, it plays a pragmatic role in identifying the proposition communicated by the utterance. To solve the descriptive problem of context dependence for a particular expression, one needs to specify which of these three roles context fundamentally plays in the interpretation of that expression relative to that context. (p. 22 ms; my italics)

Now compare these two passages. The first says that interpretation is of the utterances of others, the process of determining what a speaker meant, whereas the second refers to ‘the interpretation of [an] expression’ and claims that the context, not the audience, does the determining. Indeed, this passage and the subsequent discussion suggest that context is some one thing that plays three distinct roles (grammatical, semantic, and pragmatic) in essentially the same way. The suggestion is that at each of the three stages of interpretation, this unitary thing, context, determines the output of that stage as a function of the previous stage.[14] Also, it seems to be assumed that everything relevant to the interpretation of an utterance other than the permanent features of the words being used automatically counts as part of the context, even including the speaker’s intention.

        In my opinion, this view of context depends for its appeal on blurring a number of distinctions related to the foundational problem. Like the above distinction between interpreting what a speaker means and interpreting the sentence being uttered, the following distinctions contrast something pragmatic with something semantic:

·         It is one thing for content to be determined in context (to vary from context to context) and quite another for it to be determined by context (to be a function of context).

·         It is one thing for something 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 fixed by contextual factors (epistemic vs. constitutive determination).

·         What a sentence is most naturally used to communicate is one thing, what it means may well be something else.

·         It is one thing for context to determine the most plausible interpretation of an utterance and quite another for it to determine the actual interpretation.

Why do these distinctions matter? Because our question is whether quantifier domain restriction operates at the pragmatic or the semantic level, and merely pragmatic phenomena cannot be invoked to support the semantic approach. In my view (see Bach 1999a), the determination of what is communicated is epistemic, and the determination of what is said is constitutive, and they involve context in quite different ways. As I will suggest below, quantifier domain restriction is not constitutively determined by context; it is epistemically determined by the audience in the context, and that is a pragmatic matter.

        Here we need to draw a further distinction, between two kinds of contextual information. The kind that is semantically relevant is limited in scope. Contextual information that combines with linguistic information to determine what is said (in the sense of fixing it) is restricted to a short list of parameters associated with indexicals and tense, such as the identity of the speaker and the hearer and the time and place of an utterance.[15] This determination is independent of speakers’ intentions and beliefs. As Barwise and Perry illustrate the point, ‘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’ (1983, 148).

        When it is said that what a speaker means is ‘a matter of context’ or that ‘context makes it clear’ that he means such and such in uttering something, this is not contextual information in the narrow, semantically relevant sense. Rather, it includes any mutually and readily accessible information that speakers can count on each other to count on in figuring out what they mean when they say what they say. Pragmatic information is made relevant by the act of uttering the sentence. It can even be generated by the act of utterance, since sentences don’t encode the fact that they’re being uttered. The act of uttering a sentence exploits the semantic content of the sentence but by its very performance invokes extralinguistic information and even creates some, namely the fact that the speaker is uttering the sentence. This information provides the basis for the hearer’s identification of the speaker’s communicative intention. It is that intention, not the context, that determines (constitutes) what the speaker is trying to communicate.        Now it might be suggested that the speaker’s intention is part of the context, but this can’t be right. The speaker does not have to recognize his own intention in order to determine what he is trying to communicate.[16] Of course, in implementing his intention he needs to select words whose utterance in the context will enable the hearer to figure out what it is that he is trying to communicate. And the hearer’s so doing is not constitutive but epistemic determination.

        So what are we to make of S&S’s  contention that solving the descriptive problem requires explaining ‘how, together with linguistic rules, context determines what people mean by the utterances they make’ (p. 3 ms; my italics)? This is a a strong claim; it gives a lot of credit to context. It says that what people mean is not merely relative to or determined in the context, but is determined by the context. Moreover, this strong claim is required by their thesis that (implicit) quantifier domain restriction operates at the level of what is said, not merely at the level of what is communicated. For example, Lisa, in uttering (11), ‘Every bottle is empty’, is not merely communicating but actually saying that every bottle she just bought is empty. The claim is that context makes this the case, and only because occurring with ‘every bottle’ is a covert domain variable whose value in the context restricts ‘every bottle’ to those Lisa just bought.[17] Somehow context accomplishes this feat regardless of Lisa’s intentions. In contrast, the pragmatic approach says that context merely provides evidence for what Lisa is communicating in uttering (11), but that this sentence does not, even relative to the context, express the restricted proposition that every bottle Lisa just bought is empty. In assimilating the phenomenon of domain restriction to implicit qualification, which operates at the level not of what is said but of what is communicated, the pragmatic approach denies that context literally ‘assigns’ domains to quantifiers.

        S&S do not clearly distinguish these levels. When they explain how a pragmatic approach is to be distinguished from a semantic one, they write, ‘an approach does not count as pragmatic just because broadly Gricean mechanisms may be used to select a particular domain of quantification’. … [It] counts as pragmatic only if ‘the quantifier domain restriction determined by contextual features does not affect what is said’ (p. 32 ms). Unfortunately, they do not explain what distinguishes what is said from what is (otherwise) communicated. They do describe what is said as the ‘proposition expressed’ by a sentence (relative to a context), but they do not elaborate on this. They do not explain what they think distinguishes the propositional content of the sentence itself from any other proposition the speaker might be communicating in uttering it, and are rather noncommittal about what counts as what is said. However, since they agree with Grice that what is said must correspond to ‘the elements of [the sentence], their order, and their syntactic character’[18] (1989, p. 87), clearly the question to ask is whether among these elements are covert domain variables associated with quantified noun phrases.

 

4. Domain Variables?

S&S claim that sentences like (11) - (15) contain covert domain variables, functions whose contextually determined values map objects onto quantifier domains. These sentences are repeated here with those variables instead of the implicit (bracketed) qualifications included earlier.

            (11)       Every <bottle, f(i)> is empty.

            (12)       Only two <students, f(i)> know the capital of Latvia.

            (13)       Imelda had no <shoes, f(i)> to wear.

            (14)       Most <people, f(i)> will vote Democrat.

            (15)       No <body, f(i)> goes there any more—it’s too crowded.

Somehow the context determines the relevant bottles, students, shoes, people, or individuals, e.g., bottles the speaker just bought, students in the speaker’s class, shoes appropriate to the occasion in question, people who vote in the next Presidential election, and famous individuals.

        The semantic approach has some interesting features and consequences, not all of which are intuitively plausible. First of all, to be consistent it must say that even sentences like (16) - (18), which are naturally understood without any domain restriction, contain domain variables anyway:

            (16)       All <men, f(i)> are mortal.

            (17)       Hardly any <food, f(i)> is blue.

            (18)       There are more <ants, f(i)> than <mosquitoes, f(i)>.

So there must be a contextual domain restriction in these cases, even though it is a restriction without a difference. That is, in a likely context the value of the domain variable does not reduce the set of relevant men, food, ants, or mosquitoes.[19] Although this is the limiting case, the value of the domain variable must still be contextually provided. Otherwise, the sentence would not express a proposition at all. So if sentences (16) - (18), as presented above, strike you as expressing propositions, even though they were displayed only as sample sentences and not actually used, hence not situated in a real context of utterance, that is because you have tacitly imputed a context to them. Context must be playing a semantic role here even when you are unaware of it.[20]

        The semantic approach introduces seemingly needless syntactic complexity into predicative occurrences of indefinite descriptions, for which it is hard to imagine any need for contextual domain restriction. Suppose that Lisa had uttered (19),

            (19)       That is a bottle.

in the same context in which she had uttered (11). Since ‘a bottle’ must have a domain restriction, in this context to bottles Lisa just bought, if she had not just bought the bottle she demonstrated, her utterance of (19) would have been false. S&S might reply that the difference in utterance makes for a difference in context. However, such a difference comes in only at the pragmatic level of interpretation. Whereas in uttering (19), Lisa can plausibly be taken to be describing the thing she is demonstrating, in uttering (11) she cannot plausibly be supposed to be making a statement about bottles in general. There is a pragmatic basis for supposing that she is conveying something more restrictive than that, and presumably it will be obvious to her audience which bottles she is talking about.

        Similarly, there seems to be no need for the domain variable in (20) and (21):

            (20)       Socrates is a <man, f(i)>.

            (21)       Tweety is a <bird, f(i)>.

To be sure, the claim that every quantified noun phrase contains a domain variable does not require that the value of that variable narrow the quantification, since it allows for the limiting case of vacuous restriction. The problem is that in these predicative examples there is no need for the hearer to make an inference that this limiting case obtains. In these cases there seems to be no role for domain restriction at all. The requirement that it be present anyway seems to be a theoretical artifact.[21]

        The contention that quantifier domain restriction is semantically mandated entails that no matter how specific and detailed the explicit restriction on a quantifier, no determinate proposition is expressed unless a value of the ubiquitous domain variable is contextually supplied. Thus, utterances of sentences like (22) and (23), represented as containing domain variables,

            (22)       The first six <presidents of the United States, f(i)> were signers of the

                          Declaration of Independence.

            (23)       Most of the <retired people in Arkansas who voted for Dole in 1996, f(i)>

                          were Republicans.

do not express propositions unless the context makes clear which first six presidents of the United States or which retired people in Arkansas who voted for Dole in 1996 the speaker is talking about. For on the semantic view, the domain variable is always there, waiting to be given a value, no matter how many adjectives, prepositional phrases, and/or relative clauses modify the noun in the quantified noun phrase and no matter how specific the modification.[22]

        Now S&S might reply that syntactic consistency requires that if domain variables are needed in some cases they are needed in all cases. So it is no objection to the semantic approach to point that there there are many sentences (containing quantified noun phrases) whose natural use in normal contexts requires a vacuous restriction at best.[23] However, they have not shown that any of the simple examples considered so far require the semantic approach. The presumption is thus in favor of the pragmatic approach, which can account for the same cases without the hypothesis of covert domain variables. It simply makes the null hypothesis that quantifier domain restriction is just a case of implicit qualification.

        To appreciate the plausibility of this approach, we might recall the strategy behind Grice’s use of his ‘Modified Occam’s Razor’, which proscribes multiplying senses beyond necessity. A strengthened use of a word should not count ‘as a special sense of the word if it should be predictable, independently of any supposition that there is such a sense, that [a speaker] would use the word … with just that meaning’ (Grice 1989, p. 48). Essentially the same strategy may be applied to other linguistic phenomena by considering a variant of the language in question of which it is stipulated that the phenomenon to be accounted for is not linguistically encoded.[24] So imagine a community of speakers who used a version of, say, English of which it is stipulated that there are no domain variables attached to common nouns. It seems that people there would use sentences containing quantified noun phrases, like (11) - (18), to convey the very things we actually use them to convey. They would not expect their audiences to suppose, and their audiences would not suppose, that they are talking about, as the case may be, the maximal set of bottles, students, shoes, etc. Moreover, as language learners, they would have had no reason to suppose that such sentences contained domain variables, certainly not before they encountered and could understand utterances of relatively sophisticated sentences of the sort to be considered below. By then, though, they would have already understood simpler utterances, as of sentences like (11) - (18), without the benefit of recognizing covert domain variables.[25] The same is true of us.

 

5. Quantified Contexts?

One should be leery of any semantic proposal on which semantic structure is not a projection of syntactic (including lexical) structure, including any that posits free variables too freely. Like me, S&S deplore being cavalier about syntax (p. 52 ms) and, if there weren’t a specific reason not to, might well regard quantifier domain restriction as just another instance of the general pragmatic phenomenon of speaking loosely. But they think there is a special reason, provided by the case of ‘quantified contexts’.

        S&S give several examples of sentences in which, they claim, the domain of one quantifier expression varies with the value of the variable introduced by another:

            (24)       In most of John’s classes, he fails exactly three Frenchmen.

            (25)       In every room in John’s house, every bottle is in the corner.

            (26)       Whatever John does, most of the class falls asleep.

S&S argue that there is no way to account for the ‘natural interpretations’ of these sentences without supposing that they contain variables of domain restriction bound by the higher quantified noun phrases.[26] In the case of (24), for example, S&S observe that ‘the domain of the quantifier expression ‘three Frenchmen’ varies with the value of the variable introduced by the quantifier ‘most’. Thus, the quantifier domain variable associated with ‘three Frenchmen’ is bound by the preceding quantifier expression’ (p. 45 ms). Moreover, and this is the key part of their contention, ‘the proposition intuitively expressed by an utterance of (24) is the proposition that, for most x such that x is a class of John’s, John failed three Frenchmen in x’. If (24) has this interpretation, positing a domain variable, bound in this case, is needed to account for it.

        To appreciate what S&S are claiming, compare it with a relatively uncontroversial case, involving relational expressions like ‘stranger’ and ‘local’ (or ‘enemy’, ‘neighbor’, ‘protégé’, ‘familiar’, and ‘distant’). In (27) and (28),

            (27)       Many children talked to some strangers.

            (28)       Everyone in each village was carousing at the local pub.

there seems to be a covert variable present, though not of domain restriction.[27] The truth of these sentences requires that each of many children talked to strangers to that child and that each relevant person was carousing in a pub local to his respective village. Notice, however, that the presence of the covert variable is arguably attributable to the particular lexical item ‘stranger’ or ‘local’, as is clear from examples without multiple quantification:

            (29)       Gabby talked to some strangers.

            (30)       Tipsy was carousing at the local pub.

(29) says that Gabby talked to people who were strangers to Gabby, and (30) says that Tipsy was carousing at a pub that was local to Tipsy.[28] So the case for covert variables in (29) and (30) stands on its own merits and does not depend on the case for them in (27) and (28). Such a variable seems to be imposed by the lexical item itself.

        In contrast, S&S’s  case for covert domain variables associated with quantified noun phrases rests almost entirely on examples like (24) - (26), whose natural interpretations, they claim, can be accounted for only by positing such variables. My main objection is that these natural interpretations are not genuine readings of those sentences but are merely propositions those sentences can be used to convey. In the case of (24),

            (24)       In most of John’s classes, he fails exactly three Frenchmen.

its intuitive reading is claimed to be ‘the proposition that, for most x such that x is a class of John’s, John failed three Frenchmen in x.’ S&S do not explain where the ‘in’ in ‘in x’ comes from. Interestingly, essentially the same, allegedly semantic, interpretation of (24) can be given for (24a),

            (24a)     Usually John fails exactly three Frenchmen.

even though it contains the quantificational adverb ‘usually’ in place of ‘in most of John’s classes’. There is obviously nothing in this sentence to account for the fact that the Frenchmen in each case ( each occasion of John’s failing people) are presumed to be in a class of John’s. That is obviously something intended by the speaker for the hearer to read into the utterance.

        Indeed, even in (24) itself there is nothing to account for the fact that part of the conveyed content is that the three Frenchmen in question belong to John’s various classes. Change the preposition or change the verb phrase (but not the context), and there is no such restriction on ‘exactly three Frenchmen’. So consider this variant of (24) uttered in the same context:

            (24b)     After most of John’s classes, he sips wine with exactly three Frenchmen.

After each class, he may imbibe with different Frenchmen, but they needn’t be in the class in question. This suggests that ‘in’ may be playing a dual role in (24), the relevant one of which is merely pragmatic, since ‘after’ does not have the same effect of restricting ‘exactly three Frenchmen’. So compare (24) with (24c),

            (24c)     In most of John’s classes, he mentions exactly three Frenchmen.

where ‘in’ is present but does not have this restrictive effect. In this case, what is being said to be in most of John’s classes is the discussing, not the Frenchmen. (24c) is analogous to (24a) above, where the adverb ‘usually’ quantifies over events or occasions (of failing). This assumes that, as argued by Terry Parsons (1990, ch. 11), the behavior of such quantificational adverbs shows that there is a covert event or case variable associated with event and action verbs. But it is not the domain variable S&S claim to find in (24).

        S&S imagine that (24) poses a problem for the pragmatic approach, since they think that ‘according to the pragmatic approach, the only reading of (24) is one on which the second part of the sentence is completely unrelated to the first part of the sentence’ (p. 46 ms). However, this objection is answered by the fact that in (24) ‘exactly three Frenchmen’ falls within the scope of the event quantification, so that there can be different Frenchmen failed in each case. (24) expresses a complete proposition, but not the one S&S think it expresses. It is used to convey (24+),

            (24+)     In most of John’s classes, he fails exactly three Frenchmen [in a given class].

The one it semantically expresses parallels the ones expressed by (24b) and (24c), where there is no temptation to suppose that the Frenchmen in question must be in John's classes, much less that this is semantically mandated. The temptation to suppose this about (24) is an illusion derived from illicitly importing pragmatic information into semantic interpretation. Change the situation and you dispel the illusion. Suppose it is understood that John is not a teacher but a student, who is taking various classes in grading. In his various grading classes he grades different exams, including some written by Frenchmen. Obviously an utterance of (24) would not say that he fails exactly three Frenchmen in most of his classes but only that in most of his classes he fails exactly three Frenchmen. The narrow scope of ‘exactly three Frenchmen’ explains why the Frenchmen may be different for each class.

        According to S&S, the proposition ‘intuitively expressed’ by an utterance of (25),

            (25)       In every room in John’s house, every bottle is in the corner.

is ‘that in every room x in John’s house, every bottle in x is in the corner’ (p. 45 ms). ‘In’ also seems to be playing a dual role here (note the syntactically emergent ‘in x’), since it not only quantifies the cases (one for each room in John’s well-stocked house), but restricts the relevant bottles. But if we change the lexical material a little, there is no analogous restriction, again suggesting that when it is present, it is merely pragmatic. In (25a), for example,

            (25a)     In every room in John’s house, the personality of the designer is evident.

The different rooms may have different designers, but the designers needn’t be in the rooms. (25a) is used to convey the impliciture,

            (25a+)   In every room in John’s house, the personality of the designer [of that

                                room] is evident.

Similarly, on the natural interpretation of (25b),

            (25b)           In every house that John rents out, every passing car may be heard.

there are different passing cars for each house but there is no requirement that they be in the house. ‘Every passing car’ is within the scope of ‘every house’, but the restriction on it, to cars passing the house in question, is imposed pragmatically, so that the utterance conveys the impliciture,

            (25b+)   In every house that John rents out, every car passing [that house] may

                                be heard.

        S&S claim that the proposition ‘intuitively expressed’ by an utterance of (26),

            (26)       Whatever John does, most of the class falls asleep.

is ‘that whatever action x John undertakes, most of the class in the situation in which x occurs falls asleep’ (p. 45 ms), and, of course, the class in question may vary. However, it is not clear, given the singular ‘the class’, how S&S’s  account is supposed to explain this ‘natural’ interpretation, unless after ‘class’ there is an implicit ‘in which he does it’. Even then, the semantic interpretation of this sentence involves narrow scope, not semantically based domain restriction. Besides, what someone who utters (26) is likely to mean is much more restrictive than what (26) itself means. Presumably the speaker is not implying that if John watches ‘The Simpsons’ or removes anchovies from his pizza, most of any class falls asleep. The action must take place in a class with the aim of interesting the students in that class. So if on its natural interpretation it is used to convey a truth, there must be an implicit restriction on what John does and an implicit constraint that students’ falling asleep is a response to that. According to the pragmatic approach, the impliciture being conveyed is something like this:

            (26)       Whatever John does [to interest the students in a given class], most of

                          the class [in question] falls asleep [in spite of that].

The multiplicity of classes and the correlation between what John does in each class and the response of the class is part of the impliciture in an utterance of (26).

        Since it might seem that such problems with the semantic approach to (26) are peculiar to the choice of example, Stanley (p.c.) offers some other examples (from Cooper 1995):

            (31)       Whatever John does, most people turn up late for the experiment.

            (32)       Whatever office you go to, the supervisor is unavailable.

But (31) seems to fare no better than (26). Not only that, insofar as (31) can be used to convey John’s woes with various experiments, this use seems to be a case of impliciture, in which what is conveyed is:

            (31+)     Whatever John does [to encourage people to be on time for each of his

                          experiments], most people turn up late for the experiment [in question].

The impliciture for (32) is:

            (32+)     Whatever office you go to, the supervisor [at that office] is unavailable.

So these examples seem to be straightforward cases of relative scope plus implicit qualification.

        The examples we have considered in this section involve multiple quantification, which makes narrow scope possible, but they do not involve quantified domains. They can seem to only if it is assumed that what utterances of them are likely to convey (in likely contexts) are their semantic interpretations. However, the propositions ‘intuitively expressed’ by (24) - (26) are not their semantic interpretations. So these examples do not support the claim that domain variables are semantically mandated.

 

6. Summing Up

The pragmatic approach sees no need to posit a covert domain variable. It treats quantifier domain restriction as a not particularly special case of implicit qualification. This is the most straightforward way to accommodate the fact, which Stanley and Szabó acknowledge (leaving the case of quantified contexts aside), that a speaker who uses a quantified noun phrase could have made any intended restriction on it explicit by modifying with it with adjectives, prepositional phrases, or relative clauses. They themselves grant that this is generally possible, but contend that the explicit restriction adds nothing to what is already said by the unexpanded sentence in the context. Indeed, they think that context literally determines what the restriction is. I have argued that the plausibility of this contention depends on an equivocation on ‘determines’ and on much too expansive a view of what can be done by context (as opposed to in context). Context doesn’t determine restrictions on quantifiers. And if it does, it doesn’t do it semantically.

 

Department of Philosophy

San Francisco State University

San Francisco, CA 94132

USA

kbach@sfsu.edu


 

References

Bach, K. 1975: Performatives are Statements Too. Philosophical Studies, 28, 229-36.

Bach, K. 1994a: Conversational Impliciture. Mind & Language, 9, 124-62.

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

Bach, K. 1995: Standardization and Conventionalization. Linguistics and Philosophy, 18, 677-86.

Bach, K. 1998: Standardization Revisited. In A. Kasher (ed.), Pragmatics: Critical Concepts, vol.IV. London: Routledge, 712-22.

Bach, K. 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. Oxford: Elsevier, 65-84.

Bach, K. 1999b: The Myth of Conventional Implicature. Linguistics and Philosophy, 22, 327-366.

Bach, K. and R. Harnish 1979: Linguistic Communication and Speech Acts.Cambridge, MA : MIT Press.

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

Cooper, R. 1995: The Role of Situations in Generalized Quantifiers. In S. Lappin (ed.),Handbook of Contemporary Semantic Theory. Oxford: Blackwell, 65-86.

Gibbs, R. and J. Moise 1997: Pragmatics in Understanding What is Said, Cognition, 62, 51-74.

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

Kripke, S. 1977: Speaker’s Reference and Semantic Reference. Midwest Studies in Philosophy 2: 255-76.

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

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

Parsons, T. 1990: Events in the Semantics of English. Cambridge, MA: MIT Press.

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

Predelli, S. 1998: I Am Not Here Now. Analysis, 58, 107-115.

Recanati, F. 1989: The Pragmatics of What is Said. Mind & Language, 4, 295-329.

Stanley, J. and Z. Szabó 2000: On Quantifier domain restriction. Mind & Language, 15, 000-000.

Stanley, J. and T. Williamson 1995: Quantifiers and Context-Dependence. Analysis, 55, 291-5.

 


Notes

 

Address for correspondence: Department of Philosophy, San Francisco State University,

1600 Holloway Avenue, San Francisco, CA 94132, USA

E-mail: kbach@sfsu.edu

www: http://userwww.sfsu.edu/~kbach



[1] Stephen Levinson describes this as a ‘design flaw … in an otherwise optimal system. The bottleneck is constituted by the remarkably slow transmission rate of human speech (conceived of as the rate at which phonetic representations can be encoded as discriminable speech signals), with a limit in the range of 7 syllables or 18 segments per second’ (2000, sec. 1.3). In contrast, Levinson adds, ‘the psycholinguistic evidence suggests that all other aspects of speech production and comprehension can run at a much higher rate.’ Levinson interprets the Gricean maxims as heuristics for getting through this bottleneck.

[2] I say ‘more explicit’ rather than ‘fully explicit’, because in some cases the words in brackets do not make fully explicit what is meant. In (1), for example, the relevant hill is not specified. I suspect that most real-life cases require more elaborate qualifications, often in more than one place in the sentence, to make fully explicit what the speaker means.

[3]A more controversial species of impliciture involves ‘completion’ rather than expansion of what is uttered. In these cases the uttered sentence, though syntactically complete, is semantically incomplete—it does not express a complete proposition.

[4] Our seemingly semantic intuitions are especially unreliable when there is a recurrent pattern of nonliterality associated with particular locutions or forms of sentence, as in most of the examples above. These are cases not of conventionalization but of standardization (Bach 1995 and 1998), in particular, of standardized nonliterality (Bach 1994b, pp. 77-85). Like the more commonly recognized phenomenon of standardized indirection (Bach and Harnish 1979, pp. 192-219), including what Grice called generalized conversational implicature (1989, 37-39), in these cases too it seems that the hearer’s inference to what the speaker means is short-circuited, compressed by precedent (though capable of being worked out if necessary), so that the literal content of the utterance is apparently bypassed. For a monumental study of generalized conversational implicature, with a huge range of data, many of which illustrate what I would call ‘impliciture’, see Levinson 2000.

[5] One might suppose, having indexicals in mind as a model, that the original version of each of these sentences expresses, relative to the context in which it is uttered, precisely what its expansion expresses. So, for example, just as (i), relative to a context in which the speaker is Bill Clinton, arguably expresses the same proposition as (ii),

        (i)    I have sinned.

        (ii)   Bill Clinton has sinned.

so it might seem that sentences (1) - (10) express, relative to the context in which they are uttered, precisely what their expansions express. But this is surely the wrong model. Whereas ‘I’, though context-relative, arguably makes the same semantic contribution to (i), relative to the assumed context (its reference is a function of who is using it), as ‘Bill Clinton’ makes to (ii), though without the same context-relativity, the relation between (1) - (10) and their expansions is not comparable to the relation between (i) and (ii). The most obvious difference is that all the lexical material in (1) - (10) occurs in their expansions. The qualificatory material, were it explicitly included, would add content absent from the original sentences. Another obvious difference is that the original, unexpanded sentences express complete propositions, albeit not the ones the speaker is likely to be conveying. Finally, (i) is not cancellable in the way that (1) - (10) are. If Bill Clinton utters (i) but goes on to say, ‘but Bill Clinton hasn’t’, he has denied, not cancelled, what he just said.

[6] Recanati (1989) would call this uninformative proposition the minimal proposition expressed by the sentence. In his view, what is said is not that but the intended expansion of the minimal proposition, hence his title ‘The pragmatics of what is said’. He assumes an ‘Availability Principle’, which prescribes that ‘in deciding whether a pragmatically aspect of utterance meaning is part of what is said, … we should always try to preserve our pre-theoretic intuitions on the matter’ (1989, p. 310). However, he offers no reason to suppose that these intuitions are sensitive to the relevant distinctions. In Bach 1994a (sections 3 and 8) I argue that Recanati’s expansive view of what is said implicitly assumes that the distinction between what is said and what is implicated is exhaustive and that it overlooks the distinction between information available to the hearer and information explicitly exploited by the hearer in inferring what is being conveyed. Some recent tests by Gibbs and Moise (1997) purport to show that people’s intuitions accord with the Availability Principle, but their experimental design imposes a false dichotomy on their subjects by forcing them to choose between what is said and what is implicated. Subjects aren’t offered the in-between the category of implicit qualification.

[7] I say ‘so-called’ because if it isn’t any different, in which case such restrictions are not semantic but merely pragmatic, it is not literally true that what is restricted is a quantifier domain. I’ll use the expression ‘quantifier domain restriction’ anyway.

[8] These may be represented with restricted quantifier notation, which, in contrast to the notation of traditional first-order logic, includes the nominal following the quantifier term as part of the quantificational expression. For example, (11) may be represented as ‘[every x: bottle x](x is empty)’. Not only does this approach accord more closely with grammatical structure, by treating quantificational noun phrases like ‘every bottle’, ‘some people’, and ‘the table’ as syntactic units rather than breaking them up and introducing material that is not present syntactically, it can straightforwardly handle nonstandard quantifiers like ‘most’, ‘many,’ and ‘several’. For further explanation see Neale 1990, pp. 41-43.

[9] In general, from the pragmatic perspective the default position regarding any given linguistic phenomenon that might seem to be due to the presence of some covert syntactic expression is that it should be explained in broadly Gricean terms, i.e., by general considerations on rational communication. This default position may, of course, be overridden by grammatical considerations.

[10] They are too generous with the syntactic ellipsis approach, which strikes me as a nonstarter. I think that what they identify as its main problem, underdetermination, is merely a symptom of how far-fetched it is.

[11] Robin Cooper just takes it to be ‘a well-known fact’, evidently because it is ‘generally assumed’, that ‘the range of quantification for any quantifier has to be limited in some way by the context of use’ (1995, p. 71). Perhaps he is assuming, as formal semanticists often assume, that any element of the content of an utterance belongs to semantics if it can be represented by a free variable. Similarly, Stanley and Williamson write, ‘As is familiar, the truth of an utterance of a sentence containing a quantified expression must be evaluated with respect to a contextually determined domain’ (1995, 291), but at least they offer something of an argument:

An utterance of

         (1)          Every bottle has been put on the shelf.

 would not be true in an ordinary context in which a normal-sized shelf is at issue, unless the domain contained few of all the world’s bottles. Since there are clearly true utterances of (1) in ordinary contexts, it follows that utterances of sentences containing quantified expressions are evaluated with respect to contextually determined domains. (ibid.)

However, they do not explain why this fact about utterances of (1) has semantic consequences for the sentence itself. As Stephen Neale notes, one could maintain that ‘incompleteness [of definite descriptions or other quantificational phrases] is of no [semantic] consequence once one takes into account the distinction between the proposition expressed and the proposition(s) meant’ (1990, pp. 114-15). Even though it clashes with ‘our intuitive ascriptions of truth and falsity’, Neale ‘know[s] of no knock-down argument against this approach’.

[12] This is why Grice should not have 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).

[13] Strictly speaking, (11) does not express the proposition that every bottle in the universe is empty, since it contains nothing corresponding to that prepositional phrase. Rather it expresses the simpler but equivalent—and obviously false—proposition that every bottle is empty, where there is no restriction on ‘every bottle’. Presumably, this is all S&S intend by including ‘in the universe’.

[14] Particularly puzzling is the alleged grammatical role of context, since it is hard to see how the context, rather than the speaker’s intention, could determine which of several like-sounding sentences he is (intends to be) uttering. If after a terrible round a golfer utters ‘I hate my clubs’, the sentence he is uttering could, if he so intended it, contain the word ‘club’ meaning social group. Of course, this won’t be obvious to his audience, who will misidentify the sentence as one containing the word ‘club’ meaning golf stick.

[15] The story must be complicated a bit to handle, e.g., recorded utterances of ‘I am not now here’ on a telephone answering machine. In these cases, as Stefano Predelli (1998) has pointed out, it is the time of hearing, not speaking, that is relevant to fixing semantic content.

[16] If context is 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 is part of the context. But 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. For further discussion of why intention is not part of context, see Bach 1994b, pp. 177-8.

[17] This contrasts with the syntactic approach, which states, quite implausibly, that there is a covert occurrence of the words ‘I just bought’ in sentence (11).

[18] I agree with Grice on this point but, as explained in Bach 1994a and 1999b, I disagree with him on some other fundamental points about what is said.

[19] There are also contexts in which the speaker might be talking about particular subsets of men, food, ants, or mosquitoes, in which case the restriction would not be vacuous.

[20] It would be interesting to see if S&S could suggest a psycholinguistic test for the psychological reality of domain variables in sentences like (16) - (18).

[21] I suppose the semantic view could stipulate that even though values of domain variables are always contextually provided, their default value is the null restriction. But that is starting to sound like the pragmatic approach with a notational embellishment.

[22] There is one other worry worth noting. Because S&S’s semantic approach applies only to cases in which the domain variable is bound by a higher quantified noun phrase, it is not equipped to handle analogous cases of apparent domain restriction lacking such a phrase, such as the following examples of adverbial quantification:

        (i)    Horowitz usually performed on Sundays at 4pm.

        (ii)   Horowitz often cancelled at the last minute.

Likely utterances of these sentences would concern only Horowitz’s scheduled recitals and not pertain to his other activities. Since on most Sundays Horowitz didn’t perform at all, (i) is likely to be used to mean that when Horowitz performed publicly, he usually did so on Sundays at 4pm. (ii) is a little trickier, since it can mean either that Horowitz’s cancellations were often at the last minute or that he often cancelled his performances and did so at the last minute. If S&S were to extend their semantic approach to cases like these, they would have to explain how quantificational adverbs like ‘usually’ and ‘often’ can bind domain variables.

[23] This response is predictable from their observation that ‘syntactic parallelism cuts both ways’ (p. 42). They use this point to object to my argument from syntactic parallelism, whereby I argue (Bach 1994a, pp. 138-9), e.g., that there are no domain variables in (i) because there are no domain variables in (ii),

        (i)    The book is on that table.

        (ii)   A book is on that table.

Here I assume that there is no direct reason to posit a domain variable in (ii), since it could be true regardless of how things are with faraway books. S&S point out that one could just as well argue that (ii) does contain a domain variable because, as they claim to have shown, (i) does.

        In fact, they go further. They argue directly that (ii) contains a domain variable by describing a case where there is a contextually provided domain restriction on ‘a book’ (p. 44 ms). In the situation they describe, where John and Bill are surrounded by copies they’ve printed of Naming and Necessity, Bill uses (ii) to direct John to a book to read during their lunch break. It turns out that the only books on the table in question are copies of Naming and Necessity. So, claim S&S, as used by Bill (ii) is false. They anticipate (correctly) that I would claim that it is literally true, but, since I haven’t given any reason for claiming this, my ‘argument [that quantifier domain restriction is pragmatic] is powerless against its intended audience,’ to whom ‘this occurrence of [(ii)] seems false. Intuitively, that is because [(ii)], relative to this context quantifies over (copies of) books other than Naming and Necessity’ (p. 44 ms). Evidently S&S would think that Bill would have spoken truly had he said,

        (iii)    No book is on the table.

notwithstanding the stacks of Naming and Necessity that are covering it. Evidently they would similarly think that Bill would have contradicted himself if he had uttered not (ii) but (ii+):

        (ii+)   A book is on the table, but not one other than Naming and Necessity.

In my view, he would not be contradicting himself but merely be cancelling the impliciture that a book other than Naming and Necessity is on the table. Members of my intended audience should make sure that their intuitions are responsive to the difference between what is said and what is communicated.

[24] I applied this strategy to argue that performative sentences do not have special performative meanings (Bach 1975), and Kripke (1977) used it to show that the referential-attributive distinction is merely pragmatic.

[25] S&S use the same line of reasoning to argue against the pure indexical version of the syntactic ellipsis approach (p. 37 ms).

[26] S&S call these cases of ‘quantified contexts’, but it’s worth noting, although I won’t press the point, that if there is a quantifier domain variable associated with ‘three Frenchmen’, it is not a variable whose values are contexts of utterance. For since (24) must be uttered in a particular context, the variation in relevant Frenchmen can’t be relative to various utterance contexts. This fact indicates that the phenomenon in question is, if anything, not a case of quantified contexts but quantified domains. So it does not directly support their contention that the phenomenon of quantifier domain restriction involves determination by context.

[27] See Partee 1989 for more examples of this and other sorts.

[28] (29) and (30) also have readings on which the strangers are strangers to the speaker and the pub is local to the speaker, but (27) and (28) also have such readings. These readings are not relevant here.