Jump to ContentJump to Main Navigation
Language Turned on ItselfThe Semantics and Pragmatics of Metalinguistic Discourse$

Herman Cappelen and Ernest Lepore

Print publication date: 2007

Print ISBN-13: 9780199231195

Published to Oxford Scholarship Online: January 2008

DOI: 10.1093/acprof:oso/9780199231195.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 22 April 2019

The Minimal Theory

The Minimal Theory

Chapter:
(p.123) 11 The Minimal Theory
Source:
Language Turned on Itself
Author(s):

Herman Cappelen (Contributor Webpage)

Ernie Lepore (Contributor Webpage)

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199231195.003.0011

Abstract and Keywords

The following disquotational schema for quotation was noted in Chapter 3: (QS) “e” quotes “e” (where “e” is replaceable by any quotable item). The Minimal Theory of Quotation (MT) is the view that QS is the semantic rule for quotation. MT presumes a principle of Containment — Containment: For any quotable item e, if a quotation expression Q quotes e, then e is contained in Q. This chapter develops the semantic ramifications of MT in some detail; in particular, it aims to explain how MT is compatible with the practices of pure, direct, and mixed quotation.

Keywords:   containment, Minimal Theory of Quotation, semantic theory, pure quotation, direct quotation, mixed quotation

In Chs. 9–10, we argued against semantic theories that treat quotations as proper names, demonstratives, or quantifiers (in particular, as definite descriptions). In Ch. 8, we argued against pragmatic theories according to which quotations lack a semantic function altogether. So, then, what's left? The first moral we want to derive from this prior discussion is that, contrary to what you are led to believe by the entire history of the literature on quotation, theorists ought to take quotation dead seriously as a linguistic phenomenon. To this end, we presume that, unless doing so produces a catastrophe, quotation—like every other legitimate linguistic category—is ineliminable from a language without a loss of expressive power. No other linguistic category in our language can serve the complete function of quotation. This doesn't mean we lack ample resources for metalinguistic discourse in addition to quotation. We can name, describe, and demonstrate expressions; and we can also convey metalinguistic information pragmatically. But these practices, as we have argued over the course of this book, are not the same as quotation. With that consequence in mind, we turn to a semantic theory for quotation that celebrates its unique nature.

11.1. Preliminaries: QS and Containment

Anyone not committed to the truth of (7.7) about the quotation expression ‘ ‘Quine’ ’ has a philosophical axe to grind:

  1. 7.7. ‘ ‘Quine’ ’ quotes ‘Quine’.

As we noted in Ch. 3, the following disquotational schema for quotation is obviously correct:
  1. (QS) ‘ ‘e’ ’ quotes ‘e’

(p.124) (where ‘e’ is replaceable by any quotable item1).

Instances of QS include not only (7.7) but infinitely many other sentences, including (11.1) and (11.2):

  1. 11.1. ‘ ‘Quotation has a certain anomalous feature’ ’ quotes ‘Quotation has a certain anomalous feature’.

  2. 11.2. ‘ ‘has a certain anomalous feature’ ’ quotes ‘has a certain anomalous feature’.

We proffer QS as the semantic rule for quotation. It serves as the full semantic treatment for quotation expressions. In other words, QS is the fundamental axiom schema governing the semantics of quotation expressions.

QS is the simplest, most natural, and most obvious semantic account for quotation expressions. It is indeed often presumed by various authors in passing, as if it were completely obvious. One version may be Richard's (1986: 397) DQR:

  1. DQR: For any expression e, the left quote (lq) followed by e followed by the right quote (rq) denotes e.

We say ‘may’ because DQR uses ‘denotes’ where QS uses ‘quotes’ and it speaks of expressions rather than quotable items (the significance of which will emerge in Ch. 12), but the basic idea behind QS is already present in DQR.

Wallace (1972: 237) may also have QS in mind when he writes

  • . . . the denotation of the result of enclosing any thing in quotes is the thing itself.

QS seems also to be in Ludwig and Ray (1998: 163 n. 43); Mates (1972: 21); Salmon (1986: 6); Smullyan (1957); and Gomez‐Torrente (2001: 145ff., 2005: 129). Each of these authors restricts the quotable items referenced in QS to expressions.

The Minimal Theory of Quotation (MT) is the view that QS is the semantic rule for quotation. It should be obvious that MT presumes a principle of Containment:

  • Containment: For any quotable item e, if a quotation expression Q quotes e, then e is contained in Q.

(p.125) Containment describes a basic feature of quotation expressions. Though it will play no central role in the remainder of this chapter, its significance will become transparent in Ch. 12, where we will sketch a criterion for quotation expression individuation. The remainder of this chapter develops the semantic ramifications of MT in some detail; in particular, it aims to explain how MT is compatible with the practices of pure, direct, and mixed quotation.

11.2. MT and Adequacy Conditions (D1)–(D10)2

If someone asks how quotation functions in natural language, the obvious reply is something along the lines of QS. It is a pleasingly simple schema that requires no complicated assumptions about the surface structure of the sentences in which quotation expressions occur. The devil is always in the details, however. The rest of this chapter is largely devoted to showing how MT cooperates with the other relevant parts of a semantic theory; in particular, we aim to show how the semantic value of a quotation expression composes with the semantic values of other sorts of expressions with which it can grammatically combine productively. (In this regard, it turns out to be somewhat tricky to show how to explain mixed quotation; in particular, how to explain what we described in Ch. 3 as the syntactic chameleon nature of quotation.)

Before we begin, here's a preliminary overview of how MT fares so far against the adequacy conditions of Ch. 3. Some conditions will be deferred to Ch. 12 (since many of these adequacy conditions, we will argue, can be explained only after a viable theory of quotable items is available). But an initial progress report is very much in order.

Recall the adequacy conditions (D1)–(D10) from Ch. 3:

  1. D1. Opacity

  2. D2. Quantifying In

  3. D3. Infinitude

  4. D4. Extant Lexicon

  5. D5. The Proximity Constraint and the Strong Disquotational Schema

  6. (p.126)
  7. D6. Syntactic Chameleonism

  8. D7. Simultaneous Use and Mention in Mixed Quotation

  9. D8. Indexicals Inside Mixed Quotation

  10. D9. Indeterminacy/Context Sensitivity/Ambiguity in Quotation

  11. D10. Iterability in Quotation

A viable theory of quotation should either explain each of (D1)–(D10) or justify why an explanation isn't forthcoming.
  1. (D1). Opacity Can MT explain why we can't derive (1.5) from (1.4) given that ‘bachelor’ and ‘unmarried man’ have the same semantic value?

  2. 1.4. ‘Bachelor’ has eight letters.

  3. 1.5. ‘Unmarried man’ has eight letters.

According to MT, the semantic values of ‘ ‘bachelor’ ’ and ‘ ‘unmarried man’ ’ are the quotable items ‘bachelor’ and ‘unmarried man’, and not the semantic values of these quotable items. The issue about opacity therefore reduces to whether these two quotable items in (1.4) and (1.5) are distinct. Prima facie, they certainly seem distinct, but in order to back up this intuition a theory for how to individuate quotable items is required, one that we will deliver in Ch. 12. (Looking ahead, we'll argue that the semantic values of these two quotation expressions are different simply because the expressions themselves are distinct.)
  1. (D2). Quantifying In Can MT explain why (3.1) can't be derived from (1.4)? According to QS, what sits between the quotation marks in (3.1) is a quotable item, and so ‘ ‘x’ ’ quotes the quotable item ‘x’, thus rendering the quantifier in (3.1) vacuous.

  2. 1.4. ‘Bachelor’ has eight letters

  3. 3.1. (∃x)(‘x’ has eight letters).

  4. (D3). Infinitude QS tells us that any quotable item can be placed inside its instances. Since there's no obvious a priori upper bound on what quotable items are, there's no obvious upper bound on the number of distinct quotation expressions. It is worth pausing here to point out that it follows from the open‐endedness of quotation that we have no guarantee that the set of quotation expressions for a language can be recursively generated. Postal (2004) holds that this fact about quotation creates difficulties for standard views in syntax.

The open‐endedness of quotation rules out the view that there's a finite list (be it an alphabet or some other list of features) from which the (p.127) expressions of natural language can be computed. From this fact, Postal draws dramatic conclusions about the nature of syntax and linguistic theory more generally. We won't discuss those alleged implications here (for further discussion, see Postal 2004; Lepore 1999) for that would take us too far afield from our main topic. What is certainly correct is that quotation requires a revision of the standard view according to which we are in the possession of a computational procedure that operates on items on a finite list.

However, from this it does not follow that a language with quotation contains infinitely many semantic primitives and is accordingly unlearnable. An expression is semantically primitive only if sentences in which it occurs cannot be understood on the basis of understanding sentences in which it does not occur. But on this characterization of a semantic primitive, distinct quotation expressions, according to QS, are not primitive. It is true that quotation marks are themselves semantically primitive—that's what the last four chapters led us to conclude—but from this it does not follow that quotation expressions formed with them are. Once identified, a sentence with a quotation expression can be understood without having to learn any new semantic fact (assuming one understands every other aspect of the sentence). That is, you can understand a quotation expression formed with quotation marks and an enclosed quotable item solely on the basis of understanding sentences in which that quotation expression does not occur.

  1. (D4). Extant Lexicon Quotation is not limited to an extant lexicon because the set of quotable items is unlimited. MT is, as a matter of fact, mute about what that set is. In Ch. 12, we'll argue that the set of quotable items contains items that are not themselves expressions in any language. Hence, the set of quotation expressions cannot be limited to an extant lexicon of any one language.

  2. (D5). The Proximity Constraint and the Strong Disquotational Schema This condition has three interconnected parts. It requires a theory of quotation to explain the proximity constraint (the particularly close relationship between a quotation expression and what it quotes), QS, and the strong disquotational nature of quotation, SDS, e.g. why only the quotation expression ‘ ‘Quine’ ’ can quote the name ‘Quine’. Since QS is the semantic rule governing quotation, it is not up for explanation (other than to say that it constitutes the semantic axiom schema for quotation). Proximity we get for free, so to speak, by virtue of MT's presumption of Containment. It remains to show how SDS follows from MT.

(p.128) In Ch. 6, we argued that quotation expressions are neither context‐sensitive nor indeterminate nor ambiguous. Combining these results with MT buys SDS. Here's an illustration:

  1. From QS it follows that for any quotable item e, ‘ ‘e’ ’ quotes ‘e’. If ‘ ‘e’ ’ quotes some quotable other than ‘e’, say ‘a’, then it follows that the quotation expression ‘ ‘e’ ’ is ambiguous (because it quotes both ‘e’ and ‘a’).

  2. But, as we argued in Ch. 7, quotation expressions are unambiguous (and context‐insensitive and not semantically indeterminate).

  3. So the quotation expression ‘ ‘e’ ’ can only quote ‘e’.

  4. So SDS follows.

  5. (D6). Syntactic Chameleonism With respect to the syntactic chameleon nature of quotation, we will argue later in this chapter that MT is compatible with it. That is to say, we'll provide a grammatical account according to which the same quotation expression takes on different syntactic statuses in distinct linguistic environments (without changing its semantic value). For example, the quotation expression ‘ ‘has a certain anomalous feature’ ’ turns out to be in some contexts a noun phrase, and in other contexts a verb phrase. (That this is so will aid the effort to provide a viable semantics for mixed quotation sentences that respects QS.)

  6. (D7). Simultaneous Use and Mention in Mixed Quotation As we noted in Ch. 3, many authors, including a former time‐slice of ourselves, claim that in (1.10) the quoted words are simultaneously used and mentioned.

  1. 1.10. Quine said that quotation ‘has a certain anomalous feature.’

MT is incompatible with this intuition. However, the intuition itself is not robust, for as we noted in Ch. 6:
  • When quotation marks in many mixed quotation sentences are removed, trouble ensues—in particular, in cases where the mixed quoted expressions contain indexicals, nonsense, or foreign expressions.

The remaining sections of this chapter, then, are for the most part devoted to explaining how we can relinquish this initial first‐blush intuition behind (D7).

  1. (D8). Indexicals Inside Mixed Quotation Since MT denies that in mixed quotation sentences the quote expressions are both used and (p.129) mentioned (see (D7)), we automatically satisfy (D8), i.e. MT avoids the potential problem of claiming that a mixed quotation report implies the indirect quotation report that results from removing the quotes.

  2. (D9). Indeterminacy/Context Sensitivity/Ambiguity in Quotation We rejected (D9): we argued in Ch. 7 that quotations are context‐insensitive. QS does not require the semantic context sensitivity of quotation expressions; hence, MT is consistent with the denial of (D9).

  3. (D10). Iterability in Quotation A number of authors argue that a theory of quotation must explain why quotation is iterable. According to QS, the semantic value of a quotation expression is a quotable item; indeed, it's the quotable item that that expression contains. This quotable item might be a linguistic expression or it might not be (more on this in Ch. 12). Suppose the quotable item is not a linguistic expression. Then, of course, it's not a quotation expression, and so, iterability is inapplicable. Suppose, however, that the quotable item is a linguistic expression. Then either that linguistic expression is itself a quotation expression or it is some other sort of expression. If it is not a quotation expression, then iterability, once again, does not apply to it. And even if it is a quotation expression, it might not be a quotation expression of English (but rather one of another language—we can quote quotation expressions from other languages), and so iterability does not apply to it. But suppose, finally, that the quotation expression that is contained within a quotation expression is itself an English quotation expression. Then, of course, QS applies to it and determines a semantic value for it—namely, its quotable item. In this regard, QS does not prohibit quotation expressions from being iterable. It's not clear what more we can ask of a semantic theory of quotation in this regard.

11.3. Further Developments: Syntactic Chameleonism and Mixed Quotation

In earlier work, we rejected QS because it was incompatible with a feature we then believed held of mixed quotation (see e.g. Cappelen and Lepore 1997b; Lepore 1999). We were wrong: Mixed quotation lacks the feature we thought it had. One goal in this long section is to try to accommodate mixed quotation within the confines of a semantic theory that respects QS. Since quoted expressions are mentioned and not used (p.130) in mixed quotation, we need to devise a semantics that allows a speaker to say something expressed by a clause that combines mentioned and used expressions. We will work our way up to a treatment of mixed quotation through discussions of the truth conditions for pure, direct, and indirect quotation.

11.3.1. The Role of QS in a Semantic Theory

The role of QS in a compositional (interpretive) truth theory for English is to underwrite derivations of (interpretive) T‐sentences like (11.3)–(11.5):

  1. 11.3. ‘ ‘Bachelor’ has eight letters' is true iff ‘bachelor’ has eight letters.

  2. 11.4. ‘Quine said ‘Quotation has a certain anomalous feature’ ’ is true iff Quine said ‘Quotation has a certain anomalous feature.’

  3. 11.5. ‘Quine said that quotation ‘has a certain anomalous feature’ ’ is true iff Quine said that quotation ‘has a certain anomalous feature.’

A semantic theory that aims to assign (interpretive) truth conditions but lacks (11.3)–(11.5) as consequences is woefully inadequate; and it should be equally obvious that a semantics that assigns propositions to pure and direct quotation sentences had better underwrite the propositional assignments (11.3.1)–(11.5.1):
  1. 11.3.1. ‘ ‘Bachelor’ has eight letters' expresses the proposition that ‘bachelor’ has eight letters.

  2. 11.4.1. ‘Quine said ‘Quotation has a certain anomalous feature’ ’ expresses the proposition that Quine said ‘Quotation has a certain anomalous feature.’

  3. 11.5.1. ‘Quine said that quotation ‘has a certain anomalous feature’ ’ expresses the proposition that Quine said that quotation ‘has a certain anomalous feature.’

We'll focus on the assignment of truth conditions as in (11.3)–(11.5) in this chapter (though what we have to say obviously extends to proposition talk); and so our question is how to supplement QS in order to derive these (interpretive) truth condition assignments. In this and the next two subsections we begin with pure and direct quotation sentences like (11.3) and (11.4) respectively, and later combine (p.131) and exploit these two accounts in order to motivate one for mixed quotation sentences like (11.5).

11.3.2. Pure Quotation

How do we supplement QS in order to derive (interpretive) truth conditions for a pure quotation sentence like (1.4)? Assuming that its grammatical form is, roughly (with ‘ ‘bachelor’ ’ a noun phrase (NP) and ‘has eight letters’ a verb phrase (VP)), something along the lines of

  • [NP VP]S

  • 1.4. ‘Bachelor’ has eight letters.

we introduce as relevant semantic axioms:
  • A sentence S of grammatical form [NP VP]s is true iff the semantic value of NP satisfies VP.

  • For any object x, x satisfies ‘has eight letters’ iff x has eight letters.

Using standard first order logic as well as these two semantic axioms (and the relevant instance of QS), we hope it's obvious how to derive (11.3):
  1. 11.3. ‘ ‘Bachelor’ has eight letters' is true iff ‘bachelor’ has eight letters.

11.3.3. Direct Quotation

Accommodating direct quotation sentences poses no further problems, if we make a not unreasonable assumption that their syntax is pretty much the same as that of pure quotation sentences. Assume3 that the syntax of (1.8) is straightforward (with ‘Quine’ as its subject NP1, ‘said’ as its main verb (V), and ‘ ‘Quotation has a certain anomalous feature’ ’ as its grammatical direct object NP2) along the lines of:4

  • [NP1 [V NP2]]VP]S

  • 1.8. Quine said ‘Quotation has a certain anomalous feature.’

(p.132) A derivation (with obvious semantic rules) of the (interpretive) truth conditions for (1.8), then, runs along the lines of:
  • ‘Quine said ‘Quotation has an anomalous feature’ ’ is true iff <SV(‘Quine’), SV(‘ ‘Quotation has an anomalous feature’ ’)> satisfies ‘said’.5

  • <SV(‘Quine’), SV(‘ ‘Quotation has an anomalous feature’ ’)> satisfies ‘said’ iff Quine said ‘Quotation has an anomalous feature’.

This account of direct quotation treats it pretty much like the one for pure quotation; and as with the latter, the semantic values of items inside the directly quoted ‘Quotation has an anomalous feature’ are semantically inert, and so irrelevant to the (interpretive) truth conditions of (1.8). It's as if these items, from a semantic perspective, are as Quine so forcefully pushed, not even constituents of (1.8). (The arguments for this treatment are in Ch. 6.)

It should be obvious how to extend the treatments of pure quotation sentence (1.4) and direct quotation (1.8) to derive (interpretive) truth conditions for endlessly other pure and direct quotation sentences. What's key is that looking over the derivations you can see that QS exhausts the full semantic contribution of quotation expressions to pure and direct quotation sentences.

With these treatments of pure and direct quotation in hand, we are just about positioned to present a semantic treatment of mixed quotation; still missing is a discussion—no matter how brief—of the semantics of indirect quotation.

11.3.4. Indirect Quotation

Anyone even only casually familiar with the literature on the semantics of indirect quotation knows that it is enormous; we cannot hope to do justice to all of it here. However, we need only say enough to motivate a treatment of mixed quotation. For this purpose, we'll go ahead and assume what is a perhaps customary and certainly a widespread view: namely, that the semantic value of a complement clause of an indirect quotation is a proposition or at least something proposition‐like. In the case of (1.9), for example, we will assume for the purposes of this (p.133) discussion that the proposition that quotation has a certain anomalous feature is the semantic value of the complement clause of (1.9).

  1. 1.9. Quine said that quotation has a certain anomalous feature.

(Other candidates include sentences, utterances, notions, and Interpreted Logical Forms (Larson and Ludlow 1993). Nothing relevant to our discussion about mixed quotation hangs on a choice here.)

On the assumption that the complement clauses of indirect quotations have propositions as their semantic values, we will further suppose, again for the purposes of this discussion, that the verb ‘said’ in indirect quotation expresses a relation between a speaker and this proposition. The indirect quotation (1.9), for example, is true, we will suppose, just in case Quine said the proposition that quotation has a certain anomalous feature.

We are finally ready to tackle mixed quotation along with its relations to these other forms of quotation.

11.3.5. Some Adequacy Conditions on a Semantics for Mixed Quotation

Mixed quotation is nothing if not an exceedingly complex phenomenon—one that requires us to account for an extraordinarily wide array of data. But the overarching goal has to be to explain how a quotation expression in the complement clause of a mixed quotation can syntactically combine with other linguistic expressions not quoted in that clause without sacrificing overall grammaticality; and then to show how the semantic value of that quotation expression (as determined by QS) semantically composes with the semantic value(s) of the other expressions in that clause in order to deliver a semantic interpretation for the entire complement, and ultimately, for the entire mixed quotation sentence. To these ends, we begin with a list of semantic desiderata on an adequate semantics for mixed quotation. (We do not mean to suggest that this list exhausts the semantic demands we can make of a theory of mixed quotation; it surely does not.) Semantic desiderata (DMQ1)–(DMQ4) should be familiar; each has been discussed in earlier chapters:

  1. DMQ1. Quotation expressions must respect QS everywhere they occur, including within a mixed quotation, and so, any adequate treatment of mixed quotation must be compatible with a semantically innocent account of quotation. In this regard, mixed quotation is required to be like pure quotation.

  2. (p.134)
  3. DMQ2. Whatever proposition (if any) is expressed by the complement clause of a mixed quotation, quotations must not, once its quotation marks are removed, be assumed to be the proposition that is semantically expressed by the complement of that mixed quotation. That is, a proper semantic treatment of a mixed quotation sentence must ensure that its quoted items in its complement clause are not used. In this regard, mixed quotation is required to be like direct quotation.

  4. DMQ3. Any adequate account of a mixed quotation must ensure that the reported speaker said the quotable item quoted in the report (in order for the report to be true). In this regard, mixed quotation is again required to be like direct quotation.

  5. DMQ4. Any adequate semantic account of a mixed quotation sentence must ensure that the items not quoted in the report are used, and therefore, do make their normal semantic contributions. In this regard, mixed quotation is required to be like indirect quotation.

11.3.6. A Sketch of a Semantics for Mixed Quotation

With these conditions in place, we can proceed to a sketch of a semantic treatment of mixed quotation. This treatment borrows from the above treatments for indirect and direct quotation. Its key ingredients are already in Cresswell and von Stechow (1982: 523), and derive from Richard (2005) (though they use a different formalism). Its development owes to John Hawthorne, Paul Pietroski, and especially Sam Cumming, each of whom has contributed to different stages of the account's development (though we concede that the account as things now stand remains far from complete).

We begin with two familiar intuitions: in direct quotation, the reported speaker is represented as saying a quotable item and in indirect quotation, the speaker is represented as saying a proposition (or something proposition‐like—or at least this is the intuition of generations of semanticists working on the semantics of indirect quotation). The direct quotation sentence (1.8) is true just in case Quine said the quotable item ‘Quotation has a certain anomalous feature’; and the indirect quotation sentence (1.9) is true just in case Quine said the proposition that quotation has a certain anomalous feature.

(p.135) In mixed quotation, we are confronted with a seeming hybrid—a mixture of both direct and indirect quotation. Just as no one can say whatever is expressed by the complement of (1.9) without talking about quotation, so too no one can say whatever is semantically determined by the complement of (1.10) without discussing quotation. In this regard, the semantics of indirect and mixed quotation overlap. Moreover, just as the direct quotation (1.8) cannot be true without someone saying the quotable item ‘Quotation has a certain anomalous feature’, so too no one can say whatever corresponds to the complement of (1.10) without saying its quotable item ‘has a certain anomalous feature.’ In this regard, the semantics of direct and mixed quotation overlap.

Of course, there are ways in which mixed quotation is distinct from these other two forms of report: on the assumption—one strongly defended in this book—that the words quoted in (1.10) (much like those in (1.8)) are not used but only mentioned, the semantics of mixed quotation should not entail that whatever is semantically expressed by the complement of (1.10) is about a certain anomalous feature. In this regard, the semantics of mixed and indirect quotation are distinct. And, whereas the truth of (1.8) requires Quine to have tokened the word ‘quotation’, the truth of (1.10) does not; Quine may have designated quotation in all sorts of ways and still have said whatever (1.10) attributes to him. In this regard, the semantics of direct and mixed quotation are distinct.

Based on what we have said so far it is tempting to conclude that (1.10) is true just in case Quine said the quotable item ‘has a certain anomalous feature’ and spoke of quotation. But the account must go further. The truth of (1.10) requires Quine also to have predicated of quotation whatever corresponds to ‘has a certain anomalous feature’. Or consider another example in (11.6):

  1. 11.6. Quine said that ‘quotation’ has a certain anomalous feature.

It's not enough to say that the truth of (11.6) requires Quine to have said the quotable item ‘quotation’ and spoken of having a certain anomalous feature. We need to guarantee that its truth requires Quine—to coin a phrase—to have subjected ‘quotation’ to having a certain anomalous feature.

In both the case of (1.10) and (11.6) the semantics must ensure that Quine tried6 to apply the quoted expression in some manner or other to whatever non‐linguistic materials are discussed in the mixed quotation. (p.136) We can generalize this idea very roughly. A mixed quotation sentence of the form:

  • A said that F1 Q F2.

(where F1 and/or F2 are either empty or replaced by an appropriately used expression7 and Q is replaced by an appropriate quotation) is true just in case A said <Q; F1, F2>. The saying relation obtains when the speaker A applied Q to whatever F1 and/or F2 discuss.8

Regardless of however complicated this semantic treatment may seem, its key idea is simple. Think of mixed quotation reports as affirming a saying relation between the reported speaker and an ordered pair: in the case of (1.10), the saying relation is between Quine and the ordered pair<‘has a certain anomalous feature’, [quotation ∂]>(where ‘[quotation ∂]’ corresponds to, say, a propositional function which has quotation as its subject matter); in the case of (11.6), the saying relationship is between Quine and the ordered pair<‘quotation’, [∂ has a certain anomalous feature]>(where ‘[∂ has a certain anomalous feature]’ corresponds to a propositional function which has having a certain anomalous feature as its predicative matter).

This account guarantees that no mixed quotation sentence can be true unless the reported speaker tried to apply that quoted item to the materials unquoted in the report.

11.3.7. Two Additional Conditions on an Account of Mixed Quotation

Two further adequacy conditions on a treatment of mixed quotation revolve around the syntax of mixed quotation sentences (a topic about (p.137) which we have said next to nothing so far except for Searle's point in Ch. 6).

  1. DMQ5. Consider these parallel anaphora distributional data for indirect quotation (11.7) and mixed quotation (11.8).

    1. 11.7. Mary said (that) Bill loves himself.

    2. 11.8. Mary said (that) ‘Bill’ loves himself.

    According to our informants, both sentences are grammatically acceptable. Mary's uttering ‘Bill loves himself’ in normal circumstances not only renders (11.7) true, but, according to our informants, it also renders (11.8) true. (In this regard, Mary's uttering ‘William loves himself’ can render only (11.7) true, but not (11.8).)

    It would be surprising if whatever syntactic mechanisms are invoked to account for (11.7) don't also extend to the parallel (11.8).

  2. DMQ6. Consider these data concerning the linguistic phenomenon of VP‐ellipsis. Both the indirect quotation (11.9) and the mixed quotation (11.10) are uniformly taken by our informants to be grammatical:

    1. 11.9. Mary said that Bill loves himself and he does.

    2. 11.10. Mary said that Bill ‘loves himself’ and he does.

  3. A standard syntactic treatment of VP‐ellipsis for indirect quotation requires that the VP ‘loves himself’ be represented in the underlying syntax of the second conjunct and for deletion to occur in its derivation. In other words, at some level of linguistic analysis (11.9) would be represented as something like (11.9.1):

    1. 11.9.1. Mary said that Bill loves himself and he does love himself.

    It would be surprising were this syntactic treatment of VP‐ellipsis, if any good, did not to extend to mixed quotation. For this to be possible, in the relevant level of syntactic analysis for (11.10), the verb phrase ‘loves himself’ (not the quotable expression ‘ ‘loves himself’ ’) must be represented in its second conjunct and deletion must occur on this represented verb phrase.9

(p.138) 11.3.8. (D6) Syntactic Chameleonism

The key to satisfying (DMQ5) and (DMQ6) lies with a solution we want to propose tentatively to explain the syntactic chameleon nature of mixed quotation discussed first in Ch. 3.

Since pure quotations are always NPs, it is tempting to extrapolate to the conclusion that all quotation expressions, wherever they occur are noun phrases (NPs), and so, that quotation expressions in both direct and mixed quotation sentences are NPs as well. This is a mistake. Though QS guarantees the semantic innocence of quotation expressions (the same semantic value being assigned to a quotation expression wherever it occurs), it is completely neutral on whether the syntactic properties of a quotation expression remain constant across linguistic contexts. The quotation expression ‘ ‘has a certain anomalous feature’ ’, by virtue of being its grammatical subject, is an NP in (3.9),

  1. 3.9. ‘has a certain anomalous feature’ is not a complete sentence.

but it can't be one in (1.10), without flouting the obvious syntactic fact that it combines with an NP to form a clause. In this regard, as we noted earlier, quotation is a syntactic chameleon. To explain how this is possible, we suggest a simple rule for converting linguistic expressions of any grammatical category XP into quotation expressions of the same grammatical category. Consider the indirect quotation sentence (1.9):
  1. 1.9. Quine said that quotation has a certain anomalous feature.

In this sentence, the expression ‘has a certain anomalous feature’ is a VP that combines with the NP ‘quotation’ in order to form its complement clause. The rule we want to posit converts, so to speak, (1.9) into (1.10) through a function Q that maps linguistic expressions onto quotations of those expressions; Q maps ‘has a certain anomalous feature’ in (1.9) onto its quotation ‘ ‘has a certain anomalous feature’ ’ in (1.10), while preserving its grammatical status.10

Slightly more formally, Q sits at the end of a node that sister‐adjoins with another node. In the case of (1.10), it sits at the end of a node that sister‐adjoins with a VP node; these two nodes together form a (p.139) VP node (which we'll conveniently label ‘QVP’). The semantic value of QVP (a quotation expression) is determined by QS in its usual way. Since QS is indifferent to the grammatical status of the quotation expressions it interprets, it shouldn't matter what grammatical status a quotation has when QS is applied to it; it assigns the same semantic value (the same quotable item) to a quotation expression wherever it occurs—regardless of whether it is an NP in one context and a VP in another.

In (1.10), QVP corresponds to the quotation expression ‘ ‘has a certain anomalous feature’ ’ and its semantic value is the predicate expression ‘has a certain anomalous feature’. Diagrammatically:

                      The Minimal Theory

In general:

  • QXP → Q XP

This rule enables us to accommodate the syntactic chameleon nature of quotation expressions.

11.3.9. Adequacy Conditions (DMQ1)–(DMQ6)

The introduction of Q permits us in effect to align the accounts of indirect and mixed quotation by extending the shape of the account for indirect quotation to mixed quotation as well.

A semantic theory operates on syntactic/logical forms assigned to various sentences; semantic treatments must respect these assigned forms in deriving, say, interpretive truth conditions for any sentence. For an indirect quotation like (1.9), it is standard to assign a syntactic/logical form something along the lines of:

  • Indirect Quotation (IQ): [NP1 [V [C [NP2 VP2]S]CP]VP1]S

  • 1.9. Quine said that quotation has a certain anomalous feature.

IQ indicates that the main verb ‘said’ of (1.9) takes not an NP as its grammatical object (as we assumed above for the grammar of direct quotation), but rather a complement clause CP. C corresponds to (p.140) a complementizer, in this case ‘that’ (sometimes null, as in ‘Quine said quotation has a certain anomalous feature’). Likewise, for mixed quotation sentences like (1.10), we will posit a syntactic form along the lines of:
  • Mixed Quotation Surface Form (MQSF):11 [NP1 [V [C [NP2 [Q VP2]QVP]S]CP]VP1]S

It should be clear that MQSF is motivated by the grammatical similarity between mixed and indirect quotation. Both IQ and MQSF take complement clauses CP. The key difference is Q. Q in MQSF sister‐adjoins with a node, in this case a VP, to form a new VP (labeled ‘QVP’). In (1.10), QVP corresponds to the quotation expression ‘ ‘has a certain anomalous feature’ ’ and its semantic value is the quotable item ‘has a certain anomalous feature’, which according to MQSF is a VP.

In MQSF, QVP is inside the CP; however, in the logical form of (1.10), we'll assume QVP raises to its head CP, leaving a trace t behind, something along the lines of:

  • Mixed Quotation Logical Form (MQLF): [NP1 [V[[Q VP2]QVP(t)[C[NP2[t]VP2]S]CP1]CP2]S

Diagrammatically, its complement clause looks like this:
                      The Minimal Theory
We hope that you can see, even with only this sketch of the syntax, that this account of mixed quotation provides a compositional semantics, accommodating (DMQ1)–(DMQ4), and it also accommodates syntactic constraints (DMQ5) and (DMQ6).

(p.141) With regards to (DMQ5), (11.8) is grammatical because its QNP ‘Bill’ leaves a trace t that c‐commands ‘himself’ and co‐indexes with it.

  1. 11.8. Mary said (that) ‘Bill’ loves himself.

And we can tell a similar story for (11.10):
  1. 11.10. Mary said that Bill ‘loves himself’ and he does.

Since ‘ ‘loves himself’ ’ is a QVP that has ‘love himself’ as its constituent VP, whatever the story is about the well‐formedness of (11.9) extends to (11.10).

We hope it is also clear why this treatment accommodates (DM‐ Q1)–(DMQ4).

Condition (DMQ1) is met because QS holds for quotation expressions in mixed quotations no less than for those in pure and direct quotation. The semantic value of ‘ ‘a certain anomalous feature’ ’ is the same wherever it occurs; the application of Q together with raising the quotation permits us to respect QS.

Condition (DMQ2) is also accommodated. Quoted materials in a mixed quotation are not used if by that it is meant that a mixed quotation semantically expresses the same proposition as its quotationless indirect quotation counterpart. As a result, the quoted material in a mixed quotation does not contribute standard semantic values to the propositions expressed by that clause; again raising permits us to do this.

Condition (DMQ3) is satisfied because if an expression is mentioned (quoted) in the complement of a mixed quotation sentence it will raise to the first member of the pair said. This, given the semantics, guarantees that a mixed quotation cannot be true unless its quotation expression was said.

Condition (DMQ4) is satisfied since the account guarantees that the structures used in the complement clause of a mixed quotation contribute to the structure and content of the propositional function determined by that complement.

Of course, none of these observations are substitutes for a detailed analysis.

In summary: we believe the account of mixed quotation can explain, in an intuitive way, the data and adequacy conditions placed on it—both syntactic and semantic; both those peculiar to mixed quotation and those requiring a semantic and/or syntactic overlap among direct, indirect, and mixed quotation. We turn now to some residual problems.

(p.142) 11.4. Gibberish, Foreign Expressions, and Indexicals

Given the current treatment of mixed quotation, we can return to the puzzling data from Ch. 6 involving the behavior and legitimacy of nonsense words, foreign expressions, and indexicals in mixed quotations. This treatment will interpret (C2), (6.4), and (6.8) respectively as something along the lines of (C2)Int, (6.4)Int and (6.8)Int:

  • (C2) He says that it will have ‘a more elegant, refined taste than the one I'm making now.’ (Indexical inside the mixed quote.)

  • 6.4. Nicola said that Alice is a ‘philtosopher’. (Gibberish inside the mixed quote.)

  • 6.8. Galileo said that the earth ‘si muove’. (Foreign words inside the mixed quote.)

  • (C2) Int He says <‘a more elegant refined taste than one the one I'm making now’, [it will have t]>

  • (6.4)Int Nicola said <‘philtosopher’, [Alice is a t]>

  • (6.8)Int Galileo said <‘si muove’, [the earth t]>

Key to understanding these three interpretations are:
  • (C2) Int is neutral about whom the reported speaker of (C2) is indexing with his utterance of ‘I’.

  • (6.4)Int is neutral about what Nicola was attributing to Alice with her utterance of ‘philtosopher’.

  • (6.8)Int is neutral about what Galileo was predicating of the Earth with his utterance of ‘si muove’.

This silence is exactly as it should be. Understanding English does not enable its practitioners to know what if anything Nicola said Alice was or what Galileo predicated of the Earth; and furthermore, on the basis of understanding indexicals alone English speakers do not know whom the reported speaker indexed.

Of course, none of this means that speakers can't be reliable judges about what was going on in those original contexts of utterance. Audiences draw all sorts of conclusions on the basis of hearing an utterance, including a mixed quotation one—conclusions often not underwritten by the semantics, but rather on the basis of reasonable assumptions about the reported speaker. For example, since he was (p.143) speaking English and, as a matter of meaning, a use of ‘I’ picks out its user, whoever he is, given (C2) we can reasonably infer that that use of ‘I’ picks out the reported speaker himself.

Based, however, on competence with the English word ‘now’, though we know that when he spoke he was picking out the time of his utterance, we don't know which time that is. We can describe it—based on competence with ‘now’ alone—but we can't refer to it. The same goes for demonstrative expressions.

When someone tells us (11.11),

  1. 11.11. John said that he was never going to like ‘that man’.

we are in no position to demonstrate the man John's reported utterance demonstrated (if any), but we can describe that individual (based on an understanding of the words mentioned in the mixed quotation) as whichever man John demonstrated with his original usage.

With foreign expressions and gibberish, since we ourselves don't speak Italian or recognize ‘philtosopher’, there is nothing to do as a matter of linguistic competence except to describe the speaker as having said something expressed (if anything) with the words in question.

With (1.10) we can of course go further than we can with linguistic expressions we neither understand nor are familiar with since ‘has a certain anomalous feature’ is English and we understand English. Recall that according to QS, we cannot grasp a quotation expression without grasping its quotable item (inasmuch as it is a constituent of the quotation). So competence with English alone enables us to recognize that English expressions are being quoted in (1.10) and also what those words mean. This is what enables us to infer that the reported speaker of (1.10) is telling us that Quine said that quotation has a certain anomalous feature.

11.5. Conclusion

This chapter has been both ambitious and incomplete. We have not done more than sketch an account of mixed quotation and a lot more would need to be said to show that the account is complete.12 However, (p.144) if this account is right (and we believe it or something close to it is), then almost everyone who has ever written on mixed quotation since Davidson (1979), including ourselves, has been confused and mistaken.

What remains to be explained is how to reconcile QS with the strong disquotation principle SDS and the proximity constraint from Ch. 3. Accounting for these features is the most important task confronting a theory of quotation, for these features distinguish quotation as a linguistic phenomenon from every other one in natural language. They are what afford quotation its special place in the linguistic taxonomy. We turn to these important topics in the next and final chapter.

(p.145) Appendix: Direct Quotation—a Second Run

We succeeded in assigning an overlapping syntactic/logical form to indirect and mixed quotation. But we assigned to direct quotation a very distinct syntactic/logical form. This came about from the assumption prevalent throughout the literature that in direct quotation the quotation expression is a noun phrase (see e.g. Simchen 1999: 326, and Saka 1998: 120, 127). But anomalies in this treatment of direct quotation urge a more complicated and subtle treatment.13

Recall from Ch. 3, for example, the occurrence of ‘has a certain anomalous feature’ in (3.10):

  1. 3.10. Quine said ‘has a certain anomalous feature’.

If ‘said’ is a transitive verb, then its direct object ‘ ‘has a certain anomalous feature’ ’ is a noun phrase, but on second glance, the grammar of (3.10) seems more complicated. Ordinary transitive verbs like, say, ‘boils,’ do not permit substitution of their direct object for their subject while preserving meaning. Examples (3.11) and (3.12), though grammatical, are not synonymous.
  1. 3.11. John boiled the water.

  2. 3.12. The water boiled John.

But we can exchange the quotation expression in (3.10) with its subject without a change in meaning. Examples (3.10) and (3.13) are synonymous:
  1. 3.13. ‘has a certain anomalous feature’ said Quine.

This data suggests a more complicated story about the syntactic status of quotations in direct quotations than the one we gave in the chapter.

Furthermore, various data adduced for treating mixed and indirect quotation sentences as syntactically akin seem to extend to direct quotation as well. Recall the data from Ch. 3: namely, that in (11.12)

  1. 11.12. Jones said ‘Smith rules the moon’, and he does.

‘ ‘he’ picks up Smith as its referent instead of the word ‘Smith’ ’ (Seymour 1996: 309). Or consider this other example from Partee (1973: 412):
  1. 11.13. The sign says ‘George Washington slept here’ but I don't believe he really ever did.

(p.146) An account for direct quotation, if possible, should account for the anaphora relationships between ‘he’ and ‘Smith’ in (11.13) and ‘George Washington’ and ‘he’ in (11.13). They seem no less grammatical or interpretable than (11.9) and (11.10).
  1. 11.9. Mary said that Bill loves himself and he does.

  2. 11.10. Mary said that Bill ‘loves himself’ and he does.

The ‘he’ in each of these seems to be picking up on whatever ‘Bill’ or ‘Smith’ or ‘George Washington’ picks out respectively. If this is right, then it would seem to be a presupposition of (11.12) and (11.13) that ‘Smith’ and ‘George Washington’ pick someone out. Furthermore, the ‘does’ (‘did’) would seem to represent a VP deletion of ‘rules the moon’ in (11.12) and ‘slept here’ in (11.13).

Reading more structure into direct quotation sentences than we have so far acknowledged would enable us to account for these various sorts of linguistic data. It may be that we cannot treat direct quotations as NPs. That is, to the extent that these various data are acceptable, there is a greater demand for a uniform syntactic treatment of indirect, mixed, and direct quotation.

Pulling all these different strains together, we want to posit a more complicated syntactic/logical form for direct quotation sentences—something along the lines of:

  • Direct Quotation (DQ): [NP1 [V [[Q [NP2 VP2]S]QS]CP]VP1]S

Q is familiar; in this case it sister adjoins with S to form a QS. DQ suggests the following semantic derivation for (1.8):
  • ‘Quine said ‘quotation has a certain anomalous feature’ ’ is true iff <SV‘Quine’, <SV‘ ‘quotation has a certain anomalous feature’ ’, Ø> satisfies ‘said’ iff Quine said <‘quotation has a certain anomalous feature’, Ø>

(where ‘SV’ stands for ‘semantic value’). Unlike with mixed quotation sentences, the second member of the ordered pair satisfying ‘said’ in direct quotation is null. The speaker with a direct quotation, as a matter of meaning alone, does not commit himself in the direct quotation itself to how the non‐linguistic matters stand. He's said nothing about the non‐linguistic world with his direct quotation. In this regard, with direct quotation, the semantic properties of a directly quoted expression (if any) are rendered semantically inert by Q for the purposes of evaluating a direct quotation sentence. However, inasmuch as the expressions within the direct quotation are still in the underlying form of the sentence, the standard syntactic stories, whatever they may be, can still apply to them—in order to account for, e.g., linguistic data about anaphora and VP‐ellipsis.

Notes:

(1) In this chapter we will not tell you what a quotable item is. This chapter is mostly about the semantics of quotation and a little bit about its syntax. In Ch. 12, we say more about the nature of quotable items.

(2) We argued in Chs. 4 and 5 that (D11) and (D12) are semantically irrelevant, and so they will not be discussed.

(3) We assume that (since it's the subject) of (1.4) the quotation expression is an NP. There are reasons to challenge this assumption about quotation expressions in direct (and mixed) quotation sentences. We'll defend the view that a quotation expression can change its syntactic status without changing its semantic value across diverse linguistic environments below.

(4) In the appendix we'll try to improve upon this initial characterization of the syntax of direct quotation sentences.

(5) ‘SV’ stands here for ‘the semantic value of’.

(6) See sect. 11.4 for why we say only ‘tried.’

(7) No more than one of F1 or F2 can be empty.

(8) To generalize, we need to consider cases of mixed quotation where more than one item is quoted in its report, as in ‘John said that it was ‘raining’ when he ‘left the party’ ’. These are commonplace. To this end, we need something more general roughly along the lines of:

  • A sentence of the form: A said that F1 Q1 F2 . . . Fn Qn

(where at least one Fi is replaced by an appropriately used expression and at least one Qi is replaced by an appropriate quotation and the other Fj's are either empty or replaceable by the appropriately used expressions and the other Qj's are either empty or replaceable by the appropriate quoted items) is true just in case: A said < Q1, . . . ,Qn; F1, . . . , Fn>. So, ‘John said that it was ‘raining’ when he ‘left the party’ ’ is true just in case John said <‘raining’1, ‘left the party’2; [it was ∂1 when he ϕ2]>—where subscripts indicate which place in the propositional function the quoted material corresponds to. More below in sect. 11.3.9.

(9) For a different account of similar data, see Benbaji (2005: 43–5). We reject his account because it assumes quotable items inside mixed quotations are used, something we argued against in Ch. 6.

(10) In this regard, in the grammar for English Q creates quotation expressions, though Q is not responsible for every quotation expression. Each quotation expression in a pure quotation, for example, is not generated by Q but rather, we'll assume, is generated directly in the lexicon.

(11) Of course, this form has its VP mixed quoted; different forms will correspond to sentences with other grammatical components mixed quoted.

(12) For example, on the current account only well‐formed expressions can be mixed quoted. But the data on that is mixed. Is there anything wrong with ‘John said that ‘Bill left’ his mother at home’? If not, then the current account would need to be modified to accommodate such data.

(13) This appendix has benefited from the discussions of the syntactic and semantic complexities of direct quotation in Partee (1973), Cram (1978), and Munro (1982).