Quotation and the use-mention distinction

Quotation and the use-mention distinction

Paul Saka

Quote marks, I claim, serve to select from the multiple ostensions that are produced whenever any expression is uttered; they act to constrain pragmatic ambiguity or indeterminacy. My argument proceeds by showing that the proffered account fares better than its rivals–the Name, Description, Demonstrative, and Identity Theories. Along the way I shall need to explain and emphasize that quoting is not simply the same thing as mentioning. Quoting, but not mentioning, relies on the use of conventional devices.

Quotation is one mechanism by which we can mention; as such, it is used for attributing exact words and thoughts to others; for distancing oneself from a given word choice (as in scare quotes); for indicating titles; for expressing irony (Sperber and Wilson 1981, Jorgensen et al. 1984, Groening 1996); and for explaining truth (the disquotation theory), meaning (truth-theoretic semantics), external negation (Horn 1989), and indirect discourse (Cappelen and LePore 1997, Seymour 1994, and the inscriptional theory of Carnap 1937 and others). Finding out how quotation works, therefore, is highly important.

I shall suggest that every expression token (e.g. this particular inscription: cat) ambiguously or indeterminately refers to itself and to various items associated with it (including the inscription-type “cat”, the pronunciation /kaet/, the concept CAT, and the extension of cats). Quote marks–by which I mean double apostrophes as used in the USA, single apostrophes as used in Britain, double angles as used in parts of Europe, italicization, or any like conventional device–help to disambiguate the intended reference, although they are usually neither necessary nor sufficient for doing so.

This approach to understanding quotation is supported by its intrinsic plausibility and by the fact that it fares better than its rivals. In defending it, therefore, I shall summarize the extant theories of quotation and argue that they are inadequate. [sections] 1 and [sections] 2 rehearse the Name and Description Theories and their familiar flaws. [sections] 3 recounts the Demonstrative Theory of Donald Davidson and raises a new objection, that it cannot handle iterated quotation. [sections] 4 untangles the Identity Theory and raises an analogous objection, plus two others: one charging circularity and the other concerning the quotation of abstract types. [sections] 5 expounds my own account of quotation as the formal disambiguation of multiple ostension, and enumerates the ways in which it improves on the preceding theories.

1. The Name Theory

You can refer to anything, including a linguistic expression, by using its name; and you can assign a name to any object present by exhibiting a label along with an act of ostension. For example

(1) green eggs and ham

The label “(1)” now refers to the phrase “green eggs and ham”. In this case the label is well motivated. First–given the scholarly practice of using parentheses in the dubbing of expressions–the parentheses suggest (but do not entail) that an act of dubbing is at hand. Second, the numeral n within parentheses suggests (but does not entail) that the exhibited expression is the nth expression to be named in the present paper. Although these elements–the parentheses and numeral–possess mnemonic value, they are arbitrary from the point of view of compositional semantics. The meaning of “(1)” is not a function of the “meanings” of “(…)” and “I”. “(1)” could have referred, if my dubbing had been more haphazard, to any expression at all. “(1)” is not a description but a name.

In like manner, according to the Name Theory, quoted expressions name their referents. Tarski writes

Quotation-mark names may be treated like single words of a language … the

single constituents of these names … fulfil the same function as the

letters and complexes of successive letters in single words. Hence they can

possess no independent meaning. (Tarski 1933, p. 159)

In the same vein Quine writes

From the standpoint of logical analysis each whole quotation must

be regarded as a single word or sign, whose parts count for no more

than serifs or syllables … The meaning of the whole does not

depend upon the meanings of the constituent words. (Quine 1940,

p. 26]

For Tarski, Quine, and scores of textbooks in logic, quotations are names.

The forward productivity problem

Of course the Name Theory is an utter failure, since the quote mark is a systematically productive device that can be applied to expressions that we have never heard quoted before. This argument is so obvious that one wonders whether Tarski and Quine could have seriously meant that quotations in natural language function just like proper names. Richard (1986) and Bennett (1988) think that they used “name” in the sense of a denoting phrase or singular term, and I would too if it were not for the fact that Tarski and Quine are both so clear about denying that quotations contain meaningful structure. Perhaps Quine meant that quotations function like names only in so far as mathematical logic is concerned, but this position does not appear to be Tarski’s, who claims that the Name Theory “seems to be the most natural one and completely in accordance with the customary way of using quotation marks” (1933, p. 160).

The reverse productivity problem

Just as we can productively go from knowing any expression to knowing its quotation, we can go from knowing the quotation of any expression to knowing the expression itself. Consider Anscombe’s puzzle, which generalizes beyond personal names to all linguistic labels

If I am told “That man’s name is `Smith”‘, his name is mentioned, not

used, and I hear the name of his name but not his name … [Hence] It

is impossible to be told anyone’s name [or the word for anything].

(Anscombe 1957, p. 49)

The patent falsehood of the conclusion establishes the falsity of the premise, that to mention a name is to name the name.

The simultaneity problem

Another argument against the Name Theory is that expressions can simultaneously be used and mentioned. Quine (1943) notes that the mention of “Giorgione” in (2) provides for the interpretation of “so-called” while the use of the very same word provides for the interpretation of “his size”:

(2) Giorgione was so-called because of his size.

(3) Giorgione was called “Giorgione” because of his size.

This would pose a problem for the Name Theory, which imposes a rigid distinction between use and mention, but Quine tries to escape by claiming that (2) is elliptical for (3).

Quine not only fails to describe transformational procedures by which statements of type (2) might derive from statements of type (3), he fails to give any reason for supposing that the necessary kinds of principles even exist. Suffice it to say that other, more problematic, cases of simultaneous use-mention exist

(4) Quine says that quotation “… is weird”.

As Davidson (1979) observes, and Cappelen and LePore (1997) emphasize, the material inside the quote marks (minus the ellipsis, to be strict) is mentioned in so far as I am attributing exact words to Quine. At the same time, the words are being used in so far as they form a predicate rather than a noun phrase, singular term, or name.

The generally acknowledged inadequacy of the Name Theory leads to the Description Theory.

2. The Description Theory

Linguistic expressions can be described in various ways:

EMPIRICALLY: The first sentence of Tale of Two Cities.

LEXICALLY: (with or without an associated phrase structure): The sentence beginning with “It”, then “was”, then “the”, …

ORTHOGRAPHICALLY: The sentence beginning with I, then t, then space, then w, then a, then s, …

PHONOLOGICALLY: The sentence beginning with /I/, then /t/, then /w/, then /e/, then /z/, …

The so-called structural descriptions–those based on the formal (orthographic/phonological) elements or the substantive (lexical) elements–are used in different versions of the Description Theory of Quotation.

According to the orthographic version of the Description Theory, the quotation of an expression describes, letter for letter, what is being quoted. Thus, “cat” = the expression formed by cee followed by ay followed by tee. (Tarski (1933) and Quine (1960) advocate this approach in addition to the Name Theory, and Richard (1986) in effect defends it as well.) This account relies upon naming in so far as cee is the name of “c”, ay is the name of “a”, and tee is the name for “t”, and so on for all characters used in the language. However, the Description Theory is not limited as the Name Theory is when it comes to constructing and interpreting quotations of novel expressions.

Geach (1957, Sc. 18) holds a lexical version of the Description Theory, according to which the quotation of an expression describes, word for word, what is being quoted. This improves on the orthographic version in that it explicitly treats “the cat in the hat” not just as a sequence of meaningless marks but as a string of language. At the same time, it is less plausible than the orthographic version in so far as it postulates a primitive name for every lexeme in the language.

Problems revisited

Clearly, both versions of the Description Theory fall to the same arguments that defeated the Name Theory. (i) The Description theory fails to explain how simultaneous use and mention is possible. (ii) Although the Description Theory, especially the orthographic version, can productively deal with some novel expressions, it still cannot deal with foreign and non-linguistic marks. I can say “[Alpha]” is a Greek letter, and you can understand me even if you had never encountered Greek before in your life. Further, as Searle (1969, p. 76) notices, I can say “The sound of a California jay is …” and you can understand me even though “…” cannot be described by any names for phonic, orthographic, or lexical items in any human language. Moreover, apropos the lexical theory, I can quote nonsense, both comic (5) and tragic (6):

(5) “Twas brillig and the slithy toves/Did gyre and gimble in the wabe…”

(6) Upon expiring, he said “ggrmph”.

Again, you would understand what I have said. Disputing the relevance of cited nonsense and animal sounds, Geach and Bennett respectively write:

Properly speaking, these are not quotations at all; we cannot quote

sheer nonsense, we can only parrot it. (Geach 1957, Sc. 18)

One might occasionally put a scrawl or doodle between quotation

marks, and be understood to have referred to its shape, but only as

a joking extension of conventional quotation. (Bennett 1988, p. 405)

The former response nakedly asserts that there is a difference between quotation and parroting, which is precisely the point in contention between the Description Theory and the picture theories (to be discussed in the remaining sections of this paper). The latter response relies on the intuition that citation of scrawls is fundamentally humorous or non-serious. While I do not share this judgment myself, I can see why you would want to hold it if, like Bennett, you subscribe to a hybrid theory that combines elements of description and demonstration. According to Bennett, every quotation “X” means something like

The type whose every token resembles, in respects [R.sub.1] … [R.sub.n], this: X.

Respects [R.sub.1] … [R.sub.n] are language-relative. In English, the relevant respects for identifying tokens as types involve (for instance) top-bottom orientation (“u” versus “n”), but not ink-color. Reference to [R.sub.1] … [R.sub.n] provides the Bennett account with descriptive content; reference to “this: X” provides it with demonstrative content. The theory avoids some of the problems of the pure Description Theory, but is susceptible to the same criticisms that apply to the “paratactic” Demonstrative Theory.

3. The Demonstrative Theory

We can refer to material objects via deictic demonstration or pointing. When we do this to the ink or soundwaves that make up linguistic tokens, we quote. This is the theory of Davidson (1979); his followers Partee (1973), Goldstein (1984), Garcia-Carpintero (1994), and Cappelen and LePore (1997); and, uncited by them, Christensen (1967). According to Davidson’s version of the theory, quote marks refer to the shape of the quoted material.(1) Thus (7) is equivalent to (8).

(7) “Cats” is a noun.

(8) Cats. That complex of shapes is a noun.

In (8), cats is semantically inert while the demonstrative pronoun does the referring(2) Likewise, in (7), cats is semantically inert while the quote marks do the referring. This approach explains how quotation can be productive, and it holds promise for explaining how simultaneous use and mention is possible (a promise on which Cappelen and LePore (1997) begin to make good).

The punctuation problem

Washington (1992) claims that spoken quotation does not require quote marks, and that (7) can be read without saying “quote-unquote”, without quotative intonation, and without finger-quote gestures. In effect, (9) is a grammatical and true sentence

(9) Cats is a noun.

The point can be made even stronger, I might add. Quote marks are often omitted in writing as well (contra Reimer 1996): it is downright normal, outside of scholarly writing, to exclude quote marks, especially in constructions like “The word cats is a noun”; and even in logic publications, where one might expect the greatest exactitude, it is common for quote marks to be omitted.

If the quoted word is semantically inert and thus can be taken out of the sentence, and if the quote marks are merely optional punctuation with no real semantic role, then (7) ought to be equivalent to (10).

(10) Is a noun.

Yet it is not. Therefore, according to Washington, (7) cannot be analyzed as the Demonstrative Theory claims.

In response, Garcia-Carpintero (1994) grants the legitimacy of (9) and he concurs with Washington and current linguistic theory that (10) contains no implicit subject, demonstrative or otherwise. None the less, he says, some demonstrative element is understood by context and/or conversational implicature.

Garcia-Carpintero’s adversion to conversational implicature is left undeveloped, and I have no clue how it might go. The reference to context, however, is clearer. We often mix language with manifest context in our attempts to communicate. As Christensen (1967) says, for instance

Standing in front of a Studebaker, there is thus no need of saying,

“This car is a Studebaker.” We can simply point at it and say “A

Studebaker!” without use of any name or description [or verbal

deictic]. (Christensen 1967, p. 360)

However, comparing quotation to contextual deixis in this way goes against the evidence. Only a foreign speaker would ever point at a car and say “is a Studebaker”. If you were to omit the subject “this” you would omit the copula “is” as well. The same principle should apply to metalinguistic reference, absent any account that predicts the contrary.

As a matter of terminology, I would suggest that quotation tautologically requires quote marks. This is not to deny the existence of statements like (9), where mentioning occurs without quote marks, but only to insist that mentioning is not the same thing as quoting. An expression without quote marks, even if it is being mentioned, is linguistically distinct from an expression with quote marks, and it will prove convenient below in [sections]5 to reserve a special term for the latter–“quotation” being the most appropriate. Of course, none of this undermines the thrust of Washington’s punctuation argument against the Demonstrative Theory.

The recursion problem

The Demonstrative Theory also fails to account for iterated quotation. Just as we can refer to the word in (11), a verb, by means of forming the metaword in (12), a noun phrase, we can refer to the word in (12) by means of forming the metametaword in (13), ad infinitum. (It is clear that quotations are noun phrases, as they function as grammatical subjects.):

(11) Sit

(12) “Sit”

(13) “`Sit'”

But the Demonstrative Theory translates the perfectly good (14) into (15), which in turn translates either into the uninterpretable (16) or else into (17).

(14) “`Sit'” is a noun phrase.

(15) “Sit.” That is a noun phrase.

(16) Sit. That that is a noun phrase.

(17) Sit. That. That is a noun phrase.

The problem is that (17) fails under every possible construal. (i) The second “that” cannot refer to the first “that”. Although the first “that”–being a pronoun–does in fact constitute a noun phrase, this move gives all iteratively quoted expressions the same reference: “`sit'” and “`eat'” for instance would each denote the interior quote marks; (18) would wrongly come out false while (19, 20) would wrongly come out true.

(18) “`Sit'” contains 5 characters (including two quote marks). (19)

(19) “`Sit'” contains 2 characters Oust interior quote marks).

(20) “`Sit'” is a pair of quote marks.

(ii) Nor can the second that in (17) refer to the entire complex “Sit. That.” For this sequence, being other than a noun phrase, would contradict (14). (iii) Nor can the second “that” in (17) refer to the referent of the first “that”, to “Sit”. For this verb, being other than a noun phrase, would also contradict (14). Besides, there is no linguistic reason for thinking that such a demonstrative in such a position could refer in this transitive way; and even if there were, we would again end up having to deny (18), this time having to say

(21) “`Sit'” contains just 3 characters.

The only way that I see for the Demonstrative Theory to avoid outright falsehood is for the recursive semantics of the language to freeze when it reaches quotations within quotations. But this move is empirically unsatisfactory, for English does not distinguish between exterior quote marks and interior quote marks except as a stylistic device to remind the reader when there is a quote within a quote; the interior quote marks do not, intuitively, possess a sense distinct from the exterior marks. Furthermore, this move is theoretically ad hoc, as there is no independent motivation for treating quotation as the sole exception to the rule that syntactically recursive constructions possess recursive semantics.(3)

4. The Identity Theory

Frege (1892), Quine (1940, pp. 26, 40), Wittgenstein (1953, Sc. 16), Tajtelbaum (1957), Whitely (1957), Searle (1969, p. 75), Washington (1992), and Reimer (1996) can all be taken as advocating the Identity Theory, according to which quotation is “autonymous” (not, as Davidson (1979) repeatedly writes, “autonomous”). Unfortunately the brevity and impreciseness of these works pose problems of interpretation. Washington, who gives a comparatively clear account, says: “in quotation, expressions are used to mention themselves” (1992, p. 583). This is open to at least three construals. (i) It might mean that when an expression is a quotation then it refers to itself. (a) refers to (a).


The problem here is that it makes the Identity Theory circular: it explains quotation (a) in terms of quotation (a), which we cannot understand until we already understand what is being explained. (ii) Alternatively Washington might mean that, inside the context of quotation marks, an expression refers to itself. (b) refers to (b). The problem here is that although it explains the semantic value of (b), it does not explain the semantic value of either the quote marks or the quotation (a) as a whole. The claim that (b) refers to (b) might be true, but unless it is conjoined to some account of the double apostrophe it hardly contributes to the theory of quotation. (iii) Finally, Washington might mean that (a) refers to (b), and that this counts as self-reference because (a) is the same as (b), semantically speaking. This, I believe, is what the Identity Theorist means to say and must say: while quote marks add a pragmatic flourish, they possess no semantic value of their own.

Whereas the Demonstrative Theorist regards quote marks (or context) as referential and the quoted material as an inert adjunct, the Identity Theorist conversely regards the quoted material as (self-)referential and the quote marks as semantically empty. Despite these differences, the Identity and Demonstrative accounts can both be called picture theories, for both claim that quotation resembles its referent, the quoted material. Picture theories are supported by the prevalence of belief in word magic. Word magicians of what I will call Type I obscure the distinction between the use of a word and its referent. In both pre-literate and modem scientific societies, for instance, people commonly believe in the potency of “jinxes”: referring to something bad will make it more likely to happen. Word magicians of Type II obscure the distinction between the mere mention of a word and its use. Thus even the highly educated–teachers and editors–generally refuse to publicly quote obscenities. They omit them altogether or else print ciphers like “f***”. Now what is the difference between “f***” and “fuck”? Whereas the Description Theory is forced to mistakenly hold that there is little difference between the two, picture theories can explain why one is more offensive than the other: the one pictures the actual obscenity whereas the other does not.(4)

Like the Demonstrative Theory, the Identity Theory avoids the productivity problem and (perhaps) the simultaneity problem. But again like the Demonstrative Theory, it faces its own version of the recursion problem–plus problems about speech-acts and multiple ambiguity.

The recursion problem

Since the Identity Theory treats quote marks as semantically empty, “I sunset”‘ must have the same meaning and reference as “sunset”. Duplicating quote marks is like ending an interrogative with two question marks. It may add emphasis, but that is all. Yet “`sunset'” and “sunset” have distinct referents; thus, the Identity Theory must be wrong.

The Identity Theorist may respond that `”sunset'” and “sunset” are not equivalent simpliciter; rather, the use of “`sunset'” is equivalent to the mention of “sunset”. This claim may be right so far as it goes, but it is rather anemic in two different ways. First, it hardly seems characteristic of the Identity Theory; I doubt that any other theory would deny it. Second, it invites an explication of the use-mention distinction, which the Identity Theory never even attempts to give. But this point anticipates the next one.

The speech-act (circularity/vacuity) problem

According to the Identity Theory, “Kim” refers to itself, that is to ‘Kim But we also know that “Kim” refers to Kim. Since “Kim” and Kim are not the same, we are forced into concluding that “Kim” possesses distinct senses or uses. This leaves the Identity Theorist with the task of giving a theory of use. It would be nice to know what “uses” are (even a pretheoretic characterization would be helpful); the criteria for individuating uses; the relation between a word’s having multiple uses and its being polysemous; and so forth.

Considering the link between the Identity Theory and the speech-act tradition, we might wonder whether we can find an answer in terms of illocutionary force. Unfortunately the prospects look dim. To begin with, we cannot simply distinguish use and mention by introducing a “quotative” illocutionary force, as this move is either vacuous or circular. It is vacuous if the quotative force is left undefined, uncharacterized, and unexplained; it is circular if mentioning is explicated by the quotative force, and the quotative force is explicated by mentioning. At the same time, we do not want to contrast quotations and non-quotations in terms of the presence or absence of the familiar illocutionary forces either. We do not want to say, for instance, that whereas (22) makes a genuine assertion, (23) is a mere locution, lacking illocutionary force.

(22) The earth is flat.

(23) “The earth is flat.”

Although it is true that I am not asserting (23), I am quoting someone–perhaps hypothetical, perhaps not–as having made an assertion rather than query or command. Only by acknowledging that the quotation contains assertive force within it can we understand that (23) calls for assent or dissent, that (23) is true or false. Besides which, if quotations lacked illocutionary force, there would be no way to treat iterated quotation. If (23) is a mere locution, distinguished from (22) by lack of assertion, what would distinguish (23) from (24)?

(24) “`The earth is flat.'”

In summary, neither illocutionary forces nor anything in the literature of Identity Theory provides means for distinguishing between quotation and non-quotation.

The multiple ambiguity problem

Tokens refer (or are used to refer); types do not (are not). This is the case for all sorts of words, but is most evident for indexicals: the token of “I” in the abstract of this paper refers to Paul Saka, whereas other tokens in other places and times refer to other individuals. Abstracted from particular speakers, “I” does not refer to anyone. If we accept this view that tokens refer while types do not, and if we agree with the Identity Theorist that quoted expressions refer to themselves, then we must conclude that while expression tokens may be quoted, expression types may not. To put it another way: tokens have the capacity to refer, hence the capacity to refer to themselves, hence the capacity to be used in quotations. Types do not refer, hence do not refer to themselves, hence cannot be used in quotations.

Yet this is not how quotation works (cf. Garver 1965, Christensen 1967, Goldstein 1984, Seymour 1994). Quotation is multiply ambiguous or indeterminate: you can stick quote marks around a token to refer to that very token (25); to refer to some other token of the same type (26); or to refer to its type (27).

(25) “I” refers to me.

(26) “I”, as said by you, refers to you.

(27) “I” does not refer to anyone in particular; only tokens of it do.

Moreover, quotations can refer to words understood as form–content pairings (28); to lexemes understood as words abstracted from their inflectional paradigms (29); to forms, that is spellings or pronunciations (30); and to content both immediate (31) and translated (32).

(28) “Run” is used in the third-person plural but not singular.

(29) “Run” refers to run, runs, ran, running.

(30) “Run” consists of three letters.

(31) The concept “premise” is the same as the concept “premiss”.

(32) Galileo (who spoke no English) said, “The earth moves!”.

It is even possible for quotations to “refer” to non-existents given that types, lexemes, and content are all somewhat controversial theoretical entities. In the event that lexemes, say, do not really exist, “the lexeme ,air”‘ would refer to a certain lexeme, that is to nothing; it would not refer to air. Thus, examples (28-32) show that there really are different kinds of quotation, regardless of whether types, lexemes, and content are real.

For most purposes it is not necessary to distinguish between token-quotation and type-quotation, or between word-quotation and lexeme-quotation, or between form-quotation and content-quotation; the intended interpretation is either immaterial or else obvious. When it matters for scholarly purposes, however, writers will sometimes explicitly stipulate a system of distinct quote marks. Lyons (1977), Atlas (1989); and Horn (1989), for instance, all preface their works by glossing double apostrophes, single apostrophes, italics, and small-capital letters, each with one or more different functions. What this again proves is that quoted tokens can, contra Identity Theorists, refer not only to themselves but to related sorts of linguistic items as well.

Now the Identity Theorist could argue that using an expression is like using a hammer in that it implicates several parts, plus the whole, all at once. Just as in hammering you use a hammer, its handle, your arm, and perhaps a nail, so in speaking you use a word token, a word type, a lexeme token, a lexeme type, etc. By using a word token in mention mode, you quote a word token; by using a lexeme type in mention mode, you quote a lexeme type; and so forth. Any sort of linguistic item at all can thus be referred to by quotation.

But in hammering, you simultaneously use many implicated components. By analogy, the use of an expression in mention mode ought to collectively refer to types, tokens, forms, and content. This is not what happens, of course; statements (25–32) must be read in highly specific ways to be true, or even to be grammatical (each is third-person singular). Therefore the Identity Theory fails regardless of whether or not it invokes the hammer metaphor.(5)

5. Formal disambiguation of multiple ostension

My own understanding of quotation begins with the notion of deferred ostension (Quine 1968, Nunberg 1978, Fauconnier 1985; cp. Goldstein 1984, Reimer 1996). Direct ostension associates a term with its referent via immediate experience, for instance by deictic demonstration or by the simple exhibition of an item. Deferred ostension to an absent object X may be secured by pointing at or describing something present that is saliently related to X. For example, by pointing at a newspaper you can refer to a newspaper office or a newspaper company:

(33) [Pointing at paper] I’ve got to get over there today and place my ad.

(34) [Pointing at paper] They were bought out by Murdoch.

Pointing is not necessary; you can also verbally describe one thing in order to refer to another:

(35) The school [meaning the school building] burned down today.

(36) The ham sandwich [the customer] is waiting for her check.

The “pragmatic functions” that map from description to intended referent are, as you would expect, squishy. None the less, research by Norrick (1981) identifies a number of recurring associations, for example between merchandise and manufacturer (34), institution and physical instantiation. (35), process and product, cause and effect, part and whole, container and content; while research by Nunberg gets at such general principles as “recoverability”.

I claim that every use of language is an act of multiple ostension, partly direct and partly deferred, of at least the following kinds of items.

(a) orthographic form: cat

(b) phonic form: /kaet/

(c) lexical entry:

(d) intension: CAT

(e) extension: (x: x a cat)

First, exposure to the written label cat (a) or the spoken label /kaet/ (b) evokes the corresponding lexeme (c) in every competent speaker of English, where a lexeme is an arbitrary ordered n-tuple including orthographic form, phonic form, syntactic category, meaning, register, etc.(6) This evocation happens automatically and spontaneously as a result of human cognitive architecture trained in English (see Fodor 1983 on the reflex-like nature of language-processing). Second, the lexeme specifies the intension CAT (d) according to the pragmatic function WHOLE-PART. Third, CAT determines the extension (x: x a cat) (e) according to some mysterious but widely assumed function. Thus, the utterance of “cat”–which directly ostends or exhibits the phonic token /kaet/–deferringly ostends the corresponding form type, the lexeme , the concept CAT, the customary referent (x: x a cat), etc. These items form a package deal in which you cannot get the label without getting the rest.

With this understanding of deferred ostension in mind, I propose that use and mention can be understood as follows.(7)

(u) Speaker S uses an expression X iff:

(i) S exhibits a token of X;

(ii) S thereby ostends the multiple items associated with X

(including X’s extension);

(iii) S intends to direct the thoughts of the audience to the

extension of X.

(m) Speaker S mentions an expression X iff:

(i) S exhibits a token of X;

(ii) S thereby ostends the multiple items associated with X;

(iii) S intends to direct the thoughts of the audience to some

item associated with X other than its extension.

For example, let X = the phrase “every cat”. Then (i) S exhibits a token of X by either writing every cat or uttering /Evri kaet/ or even by pointing at someone else’s tokening, as happens in charades.([ii.sub.a]) Such an exhibition directly ostends the token form every cat or /Evri kaet/; and it deferringly ostends the types every cat and /Evri kaet/ and the noun phrase sequence made up of the two lexemes + –that is the noun phrase structure [every.sub.DET] [cat.sub.N]].sub.Np] and its intension EVERY(CAT). ([ii.sub.b]) In addition it deferringly ostends, as reference, every cat. (iii) In the case of use, S intends “every cat” to refer to every cat; in the case of mention, S intends “every cat” to refer to some item that is saliently associated with “every cat” other than its extension, that is to one of the items listed under ([ii.sub.a]).

Although I have characterized the use-mention distinction in terms of extension, referentialist semantics is by no means essential to my account. I myself prefer the anti-realist “internal semantics” of Jackendoff (1983[Beta]), which construes “reference” as a solipsistic mental state; others are invited to adjust (u) and (m) according to their own favored theories of meaning.

The account offered possesses a number of virtues. To begin with, it allows for the existence of mentioning without quote marks. In addition, it reveals how much use and mention have in common. Although the reference clauses (u-iii) and (m-iii) differ from each other, this difference is made possible by the same background conditions: the existence of a conventionalized language (ii) and its exercise (i). Finally, it follows from formulations (u) and (m) that use and mention, though distinct, are perfectly compatible: you can intend to direct the thoughts of your audience to the customary reference of X while at the same time intending to direct the thoughts of your audience to other things as well, for example to X itself. Thus, my account opens a path for treating simultaneous use and mention.

With the foregoing understanding of use and mention in place, we are now prepared to consider the definition of quotation. Syntactically, a pair of quote marks is a discontinuous determiner (a complex symbol which, applied to an argument expression, produces a noun phrase). Semantically, a pair of quote marks is a concept or intension, QUOT, which ambiguously or indeterminately maps its argument expression X into some linguistic item saliently associated with X other than the extension of X. Although quote marks generally do not specify among token, type, form, and concept, they still serve to partially disambiguate, for they rule out customary reference as the intended interpretation. Thus, the speaker who uses quote marks announces “I am not (merely) using expression X but am mentioning it”.

So far we have seen, in the explanation of (u) and (m), how use and mention without quotation work. To see how use and mention with quotation work, suppose X = the singly quoted expression “cat”.(8) Then (i) S exhibits a token of X by writing “cat” or by uttering /kwot /kaet/, [Delta]nkwot/ or by uttering /kaet/ with distinctive intonation or accompanying finger gesture or by just uttering /kaet/, ([ii.sub.a]) Such an exhibition directly ostends the orthographic token “cat” or some spoken counterpart; and it deferringly ostends the corresponding type, and the noun phrase structure [“sub.DET [cat].sub.N]” DET].sub.NP] and its intension QUOT(cat). ([ii.sub.b]) The latter ambiguously specifies the extension: the token cat, the type cat, the token /kaet/, the type /kaet/, the structure [[cat].sub.N], or the concept CAT. (iii) In the case of use, S intends “cat” to refer to the extension of “cat”, namely to one of the items listed under ([ii.sub.b]). In the case of mention, S intends “cat” to refer to some item saliently associated with “cat” other than its extension, namely to one of the items listed under ([ii.sub.a]).

As a consequence of my account, the mention of X and the use of “X” quite properly secure the same reference. Further, whereas the use-mention distinction is characterized in terms of the speaker’s intentions, the distinction between a quotation and a non-quotation is a formal, grammatical affair. Distinguishing between use and mention in a language without quote marks is a purely pragmatic affair, but in a language with quote marks mentioning can be explicitly marked (although such marking is not obligatory).

Now suppose X = the doubly quoted “`cat'”. Then (i) S exhibits a token of X by writing “‘`cat'”. (ii.) Such an exhibition directly ostends the token form “`cat'”, and it deferringly ostends the type “`cat'” and the noun phrase structure [[“sub.DET [“sup.DET [[cat.sub.N] “.sub.DET].sub.NP] “DET]Np and its intension QUOT(“cat”). 00 The latter ambiguously specifies as extension: the token “cat”, the type “cat”, the structure [[“sub.DET] [cat.sub.N] “.sub.DET]].sub.NP], or the concept QUOT(cat). ([ii.sub.b]) In the case of use, S intends “`cat'” to refer to one of the items listed under (iib). In the case of mention, S intends “`cat'” to refer to some item listed under ([ii.sub.a]). In short, the Disambiguation Theory correctly handles iterated quotation.

My account differs from the Demonstrative Theory in that I do not take quote marks as pointing. Rather, I compare quote marks to the subscripts you see on ambiguous words in dictionaries and sometimes in philosophy, except that quote marks help us to steer around a case of systematic ambiguity. For example, “warm” has a number of senses including PRODUCER OF HEAT (“this jacket is warm; it keeps me comfortable”) and PRODUCT OF HEAT (“this jacket is warm; it has been sitting on the stove”). If we agree to mark the instrumental sense with the subscript “inst” then a number of consequences follow. First, “[warm.sub.inst]” means PRODUCER OF HEAT; second, “warm” remains ambiguous (does it lack a subscript because the speaker deliberately intended “not in the instrumental sense”, does it lack a subscript because the speaker was sure that “in the instrumental sense” was so clear as not to require subscripting, or does it lack a subscript because the speaker just did not stop to think about the ambiguity?); third, novel expressions such as “[hot.sub.inst]” PRODUCER OF TOO MUCH HEAT become unambiguously interpretable. Quote marks work analogously, functioning to single out (or rather narrow down) the intended reading. Their presence indicates a metalinguistic use; their absence does not necessarily indicate anything at all; and they generalize to new expressions.

But, you might object, in specifying that the metalinguistic use is operative, don’t the quote marks serve to point at the quoted expression? It is easy to think so, because the quoted word is always present during the act of quotation. None the less, I do not think that the quote marks function deictically (and hence referentially). In the first place, it is possible for a quotation to refer to something that is not physically present, as in the case of word meanings and word types. Second, it is possible to forego the use of quote marks altogether.

My account differs also from the available versions of the Description Theory in that it does not use names for elements. The quote marks, when they are used, do describe which aspect of an ostension is operative; but the panoply of ostension is provided by picture (in so far as the exhibited linguistic label resembles itself) and by other pragmatic relations.(in so far as the form token evokes a type or associated content, etc.).

My account bears some resemblance to the Identity Theory: in both use and mention, the speaker exhibits the same form. However, there are critical differences. To begin with, quotation may refer not only to a given word-form token but also to the corresponding word-form type, to other tokens of that type, to lexemes, to concepts, and so forth. Thus, my account qualifies as a picture theory in its analysis of the quotation of forms (since the inscription “cat” really does picture the inscription cat). But it is not entirely a picture theory, as my analysis of the quotation of content involves no iconic relation at all (since the inscription cat does not resemble the meaning CAT in any visual, acoustic, or other physical manner). Another difference is that my account treats an expression as a quotation or not depending on its linguistic structure, whereas the Identity Theory treats quotation as a matter of function (as the use-mention distinction is for me). Further, on my account the quotation as a whole (quoted material plus quote marks) is referential. For the Identity Theorist, the quoted material alone refers, and for the Demonstrative Theory the quote marks alone refer.

The Disambiguation Theory differs from the Tajtelbaum version of the Identity Theory in particular. For Tajtelbaum (1957), a word refers to itself because “we have the (tacit) convention that a name and its name are denoted by the same word.” According to the Disambiguation Theory, in contrast, mention is not a matter of convention. Whether we use single apostrophes or double–indeed, whether we use any formal device at all–is a matter of convention. But the fact that we can use a word to refer to itself or to associated items is a natural result of the fact that the direct ostension of a form token deferringly ostends type, content, etc. That is, the exhibition of a word form makes the corresponding word type, lexeme, sense, and referent manifest to any human mind that speaks the language and is versed in the notions of word type, lexeme, sense, and referent.

Some may object that the Disambiguation. Theory counts as a version of the Identity Theory, or that the Identity Theory is at least compatible with my proposals, in spirit if not in letter. I do not know how to judge this claim for I do not see any “spirit” in the Identity Theory beyond what it actually says. At any rate, it makes no difference to me whether my account is regarded as a version of the Identity Theory, or distinct from it, so long as my substantive points are acknowledged, for example that mentioning is not the same as quoting.

On this point, in fact, my account distinguishes itself from all others. For the Identity Theory, quote marks are but window dressing, and deploying them or removing them changes nothing. For the Demonstrative Theory, quote marks are so crucial that they are postulated as logically present even when physically absent. My theory, on the other hand, treats an expression with quote marks as distinct from the same expression without quote marks.

Is this really a virtue, though? Some critics have denied that it is, claiming that quoted mention and unquoted mention obviously mean the same. But this is wrong, for it is just false that (37, 38) mean the same.

(37) Chicago has seven characters.

(38) “Chicago” has seven characters.

The fact is that (37) and (38) are ambiguous or indeterminate in different ways. The full explanation of (37) is that it has at least the following two readings, the first one true and the second one false.

(39) The word “Chicago” has seven characters.

(40) The city Chicago has seven characters.

In contrast, (38) has the following two readings, one true and one false.

(41) The word “Chicago” has seven characters.

(42) The expression “`Chicago”‘ has seven characters.

While (39) = (41), the total range of possible readings for (37) differs from the range of possible readings for (38). Thus, (37) and (38) demand different analyses, and on this score my account is indeed attractive.

Another distinctive feature of my account is that it explicitly treats customary reference as a species of ostension on a par with autonymous reference. Use and mention are two sides of one coin; although distinct, they implicate one another. You simply cannot have a language without the potential for both use and mention (although you can have a language without quotation).

Yet problems remain. If you may refer to whatever you ostend, and if the utterance of a word-form deferringly ostends all of the associated linguistic material, then why can’t you quote a word in order to refer to its associated language or dialect, as in (43)?

*(43) “Warshboard” is a stupid dialect.

(44) “Warshboard.” Now there’s a stupid dialect.

The acceptability of (44) proves that the quotation of “warshboard” suffices to ostend or make manifest a certain dialect. So if ostension were equivalent to reference, (43) should be acceptable.

The solution lies in rejecting any simple-minded equivalence of ostension and reference. Sometimes one reference is blocked by a different prior reference.

*(45) The ham sandwich, which was inedible, left without tipping.

Here “ham sandwich” ostends both a comestible and a customer. Once the expression is established in the relative clause as referring to the food, it cannot later be taken as referring to the person. Yet hard-and-fast rules elude us.

?(46) “Ghosttown” has more letters in it than people.

(47) Lee’s dissertation, which weighs five pounds, has already been refuted.

(48) A: What dialect does he speak?

B: “Tomahto”.

Although (46) strikes me as somehow anomalous, as necessarily jocular, it does not seem as bad as (45). In (47) “Dissertation” ostends both a material token and an abstract type; reference to the former does not preclude reference to the latter (Fauconnier 1985). And in (48) the citation of a word does pragmatically refer to a dialect. The general lesson is that developments in the psychology of association and in the pragmatics of reference will better enable us to judge the thesis that quotation formally disambiguates multiple ostension.

6. Conclusion

The Disambiguated Ostension Theory of Quotation consists of two components. First, it assumes that the capacities for both use and mention stem from the same source, namely from the fact that the human mind associates a multiplicity of deferred ostensions with any exhibited token, thus giving rise to pragmatic ambiguity. This thesis is a plausible consequence of the findings of Nunberg (1978) and Norrick (1981), and it is further supported by the fact that surely all languages contain a use-mention distinction. (Even formal languages that officially lack a use-mention distinction still, in practice, possess it. This suggests that the distinction is made not by particular languages per se but rather by what is common across English, Eskimo, the predicate calculus, etc.–namely, human cognitive agency.) While this first component allows for a pragmatic use-mention distinction, the other component of the Disambiguation Theory claims that quote marks formally announce that mentioning is taking place. This view too is plausible if you think that linguistic elements normally signify something, and it is further suggested by the fact that (37) and (38) are ambiguous in distinct ways.

Another consideration in favor of the Disambiguation Theory is that it is more successful than rival theories in dealing with the facts which we have seen any theory of quotation ought to respect. These facts are summarized and extended below.

CONVENTIONALITY: Although the capacity for mention inheres in

the natural connection between a linguistic token and its associated

items, the existence of a device for explicitly marking mention–for

quoting–is conventional. Such a device can be found within most

modern writing systems, and in the grammatical systems of a few

languages (e.g. the indirect discourse mood of German), but it

appears to be absent throughout most of history in most languages.

Furthermore, since convention rests upon a marriage of form and

function, an expression with conventional quote marks possesses a

range of possible significance distinct from that of an expression

without quote marks. This latter fact disproves the Identity Theory.

PRODUCTIVITY: If you are party to the convention, then knowing

any expression will enable you to know its quotation; and knowing

any quotation will enable you to know what is quoted. The fact of

productivity rules out the Name and Description Theories.

ITERATIVITY: Quotations are themselves expressions that may be

quoted in turn, giving yet a different reference. Iterativity rules out

the Demonstrative and Identity Theories.

PICTORIALITY: Belief in word magic is powerful throughout the

world, in both stone-age and industrial cultures–the mere

mention of an obscenity is often taboo or at least somewhat shocking.

Furthermore, constructions like “Pigs are called `pigs'” possess a

sort of a priori truth. Both of these observations suggest that

quotation resembles what is quoted. These facts rule out the Name

and Description Theories.

SIMULTANEITY: In some contexts you may both use and mention

an expression at the same time. In some contexts you may

anaphorically ground an expression for both use and mention. The

fact of simultaneity poses a mortal problem for the Name and

Description Theories. It remains to be seen how it fares under other


MULTIPLICITY: With quotation you can refer to both linguistic

expressions and non-linguistic vocalizations and imprints; to both

form and content; to both types and tokens. On the rare occasions

when we need to make these distinctions explicit, we take advantage

of the diverse devices for marking quotations: double apostrophes,

single apostrophes, italicization, underlining … The fact of

multiplicity disproves the Identity Theory and, at least in

Davidson’s version, the Demonstrative Theory as well.

OMISSIONS: Philosophers, linguists, logicians, and lay writers

commonly omit quotational markings on mentioned expressions

(sometimes with a statement to the effect that “I won’t be fussy”)–for

example “Diabetes comes from Greek.” Instead of treating this as

an error, an outright falsehood, it would be more charitable to take it

as indeterminate/ambiguous between the false “The disease

diabetes itself comes from the Greek language” and the true “The

word `diabetes’ comes from Greek”. The fact that there are omissions

rules out the Demonstrative Theory.

On all of these scores, I have argued, the theory of quotation as formal disambiguation of multiple ostension fares better than the alternatives. A fuller understanding of it now awaits further research in psychopragmatics.(9)


Department of Philosophy

University of Illinois

Urbana, IL 61801


(1) Washington (1992) criticizes Davidson’s theory on the grounds that quotation may refer to size and other features aside from shape, and Garcia-Carpintero (1994) concurs. Granted, I can truthfully say “a” is smaller than “a”. However, “shape” can refer to all sorts of formal properties, including size. After all, size is a geometrical feature, speaking in the strictest, mathematical sense.

(2) For the sake of clarity, I sometimes use double apostrophes and italicization as interchangeable marks of quotation. When I put a word in italics and a set of apostrophes, it is equivalent to being inside of two sets of apostrophes.

(3) Cappelen and LePore (1997, p. 439) note the non-iterativity of quotation within their Demonstrative Theory, though they do not recognize it as a problem. Grandy (1973, p. 106) notes an analogous recursion problem for Davidson’s paratactic theory of indirect discourse.

(4) While it is clear that word magic of Type I is irrational, Type II is a doubtful case. In any event, I apologize if I offend any sensibilities, and hope you see that my point might have been lost had I stuck to ciphers and circumlocutions.

Perhaps I ought to clarify that “word magic” is a standard anthropological term (cf. Ogden and Richards 1923), and I am not using it as an implied ad hominem against anybody. Obviously I am not trying to taint the Name and Description Theorists by this terminology, for they deny the existence, or at least the significance, of word magic. Just as obviously I am not trying to taint picture theorists, for the picture theory, or something like it, is what I am advocating. What I do want to do is to emphasize some robust properties of quotation that get ignored in the philosophical literature.

(5) Boolos (1995) discusses another kind of ambiguity, one due to scope. For instance, (i) denotes either (ii) or (iii):

(i) “a” followed by “b”

(ii) ab

(iii) a” followed by “b

Incidentally, Boolos proposes a way of resolving such ambiguity: in an ideal language you could use subscripts so that each left quote mark gets explicitly and uniquely paired up with some right quote mark.

(6) For simplicity I subsume the phonetic and the phonological under the phonic”.

(7) I use “use” and “mention” in their technical senses (to utilize an expression with customary reference versus to utilize it in reference to itself). It seems that I need to make this clear because some writers insist that, in order to mention an expression, you need to use it (Geach 1950, Ziff 1960, p. 27, Garver 1965, Davidson 1965 and 1979). The sense in which this is obviously true is irrelevant.

(8) I myself am using two layers of quotation–a pair of double apostrophes plus italics–to describe S’s speech. In S’s speech, however, only a single layer is being used.

(9) Most of this paper was written under a University of Illinois University Fellowship in 1995-96, and revisions were made while I held a Beckman Institute Cognitive Science Fellowship in 1997. For reading and commenting on various drafts, I gratefully acknowledge Steve Wagner, Tim McCarthy, Mark Sainsbury, and an anonymous reviewer.


Anscombe, G. E. M. 1957: “Report on Analysis Problem #10”. Analysis, 17, pp. 49-52.

Atlas, Jay 1989: Philosophy without Ambiguity. Oxford: Oxford University Press.

Bennett, Jonathan 1988: “Quotation”. Nous, 22, pp. 399-418.

Boolos, George 1995: “Quotational Ambiguity”, in Paolo Leonardi and Marco Santambrogio (eds.), On Quine. Cambridge: Cambridge University Press, pp. 283-96.

Cappelen, Herman and LePore, Ernie 1997: “Varieties of Quotation”. Mind, 106, pp. 429-50.

Christensen, Niels 1967: “The Alleged Distinction between Use and Mention”. Philosophical Review, 76, pp. 358-67.

Davidson, Donald 1979: “Quotation”, in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press, 1984, pp. 79-92. Originally published in Theory and Decision, 1979.

Fauconnier, Gilles 1985: Mental Spaces. Cambridge MA: MIT Press. Second edition, Cambridge University Press, 1994.

Fodor, Jerry 1983: The Modularity of mind. Cambridge MA: MIT Press.

Frege, Gottlob 1892: “On Sense and Reference”, in Peter Geach and Max Black (eds.), Translations from the Philosophical Writings of Gottlob Frege, 3d edn., Oxford: Basil Blackwell, 1980, pp. 56-78.

Garcia-Carpintero, Manuel 1994: “Ostensive Signs”. Journal of Philosophy, 91, pp. 253-64.

Garver, Newton 1965: “Varieties of Use and Mention”. Philosophy and Phenomenological Research, 26, pp. 230-8.

Geach, P. T. 1950: “On Names of Expressions”. Mind, 59, p. 388.

— 1957: Mental Acts. London: Routledge & Kegan Paul.

Goldstein, Laurence 1984: “Quotation of Types and Types of Quotation”. Analysis, 44, pp. 1-6.

Grandy, Richard 1973: Review of Words and Objections: Essays on Quine. Philosophical Review, 82, pp. 99-110.

Groening, Matt 1996: “Life in Hell”. Village Voice, April 9, p. 3.

Horn, Laurence 1989: A Natural History of Negation. Chicago: University of Chicago Press.

Jackendoff, Ray 1983: Semantics and Cognition. Cambridge MA: MIT Press.

Jorgensen, Julia, Miller, George and Sperber, Dan 1984: “Test of the Mention Theory of Irony”. Journal of Experimental Psychology: General, 113, pp. 112-20.

Lyons, John 1977: Semantics. Cambridge: Cambridge University Press.

Norrick, Neal 1981: Semiotic Principles in Semantic Theory. Amsterdam: John Benjamins.

Nunberg, Geoffrey 1978: “The Non-Uniqueness of Semantic Solutions”. Linguistics and Philosophy, 3, 1979, pp. 143-84. Originally a part of his dissertation in linguistics, The Pragmatics of Reference, City University of New York.

Ogden, C. K. and Richards, I. A. 1923: The Meaning of Meaning. New York: Harcourt, Brace & World.

Partee, Barbara 1973: “The Syntax and Semantics of Quotation”, in Stephen Anderson and Paul Kiparsky (eds.), Festschrift for Morris Halle, New York: Holt, Rinehart & Winston, pp. 410-8.

Quine, Willard 1940: Mathematical Logic, 3d edn. Cambridge MA: Harvard University Press.

— 1943: “Notes on Existence and Necessity”. Journal of Philosophy, 40, pp. 113-27. Most of this material is duplicated in “Reference and Modality”, in his From a Logical Point of View, 2d edn., Cambridge MA: Harvard University Press, 1961.

— 1960: Word and Object. Cambridge MA: MIT Press.

— 1968: “Ontological Relativity”, in his Ontological Relativity, Columbia University Press, 1969, pp. 26-68. Originally published in Journal of Philosophy.

Reimer, Marga 1996: “Quotation Marks: Demonstratives or Demonstrations?”. Analysis, 56, pp. 131-42.

Richard, Mark 1986: “Quotation, Grammar, and Opacity”. Linguistics and Philosophy, 9, pp. 383-403.

Searle, John 1969: Speech Acts. Cambridge: Cambridge University Press.

Seymour, Michel 1994: “Indirect Discourse and Quotation”. Philosophical Studies, 74, pp. 1-39.

Sperber, Dan and Wilson, Deirdre 1981: “Irony and the Use-Mention Distinction”, in Peter Cole (ed.), Radical Pragmatics, New York: Academic Press, pp. 295-318.

Tajtelbaum, Al 1957: “It is Impossible to be Told Anyone’s Name”. Analysis, 17, pp. 52f.

Tarski, Alfred 1933: “The Concept of Truth in Formalized Languages” (tr. J. Woodger), in his Logic, Semantics, Metamathematics. Oxford: Oxford University Press, 1956, pp. 152-78.

Washington, Corey 1992: “Quotation”. Journal of Philosophy, 89, pp. 582-605.

Whitely, C. H. 195 7: “Names and Words”. Analysis, 17, pp. 119-20.

Wittgenstein, Ludwig 1953: Philosophical Investigations. Translated by G. E. M. Anscombe, New York: Macmillan.

Ziff, Paul 1960: Semantic Analysis. Ithaca NY: Cornell University Press.

COPYRIGHT 1998 Oxford University Press

COPYRIGHT 2000 Gale Group