Quotation (Stanford Encyclopedia of Philosophy)

1. How to Characterize Quotation
Problems arise right at the outset since quotation is not an easy
thing to characterize. We start with reflections on how one might try
to do so.
1.1 Quotation Identified Through Examples
There’s an easy and relatively non-controversial way to
identify
quotation: it is the
sort
of linguistic
phenomenon exemplified by the subject in (4) and the direct
object
in (5); these are instances of
pure
and
direct
quotation, respectively.
(5)
Quine said, ‘Quotation has a certain anomalous
feature’.
That leaves open the question of which semantic and syntactic devices
belong to that sort. Any characterization of a more specific nature,
either of a syntactic or a semantic sort, moves into controversial
territory immediately.
1.2 Quotation Identified Syntactically
syntactic characterization
might go something like this:
Take two quotation marks— single apostrophes in Britain, double
in the United States, double angles in parts of Europe — and
put, for example, a letter, a word, or a sentence between the two.
What results is a quotation, as in (4)–(5). But then consider
the following sentences:
(6)
My name is Donald.
(7)
Bachelor
has eight letters.
We can see there are two problems with the syntactic
characterization:
In spoken language, no obvious correlates of quotation marks
exist. Spoken utterances of (6) seem often to be
un
accompanied by lexical items corresponding to
‘quote/unquote’.
Even if attention is restricted to written language, quotation is
not invariably indicated by the use of quotation marks. Sometimes, for
example, italicization is used instead, as in (7).
Other devices employed as substitutes for quotation marks include bold
face, indentation, and line indentation (cf. Quine 1940, pp.
23–24; Geach 1957, p. 82). There’s no clear limit on the
range of distinct written options, other than that they are used as
quotation marks, but this renders the syntactic characterization
incomplete, and thus, unsatisfactory.
1.3 Quotation Identified Semantically
Another tempting strategy is to say that an expression is quoted if it
is
mentioned
. There are two problems with this
characterization.
Several theorists want to distinguish between mention and
quotation (see
Section 3.5
).
This definition would rule their theories out by stipulation.
This characterization is no clearer than the intuitive distinction
between use and mention, and matters become even more complicated as
soon as we do try to characterize ‘mention’ and
‘use’. Isn’t ‘bachelor’ in (4) in some
sense used to refer to itself? If the response is that it is used, but
not with its normal semantic value, then we are left with the
challenge of defining ‘normal’ and ‘abnormal’
semantic values. That, again, leads immediately to controversy.
In order to remain as neutral as possible, we will stick with a simple
identification-through-examples strategy, and emphasize that it is an
open question as to how to identify the sort of linguistic devices to
which the subject in (4) belongs.
2. Basic Quotational Features
Quotation is a subject matter that brings together a rather
spectacular array of linguistic and semantic issues. Here are six
basic quotational features of particular importance (BQ1–BQ6,
for short) that will guide our search for an adequate account:
BQ1.
In quotation you cannot substitute
co-referential or synonymous terms salva veritate
An inference from (4) to (8), for example, fails to preserve
truth-value.
(4)
‘Bachelor’ has eight letters
(8)
‘Unmarried man’ has eight letters
No theory of quotation is adequate unless it explains this feature
(and no theory of opacity is complete before it explains why quotation
has this feature).
BQ2.
It is not possible to quantify into
quotation.
(9), for example, does not follow from (4):
(9)
\((\exists x)\)(‘\(x\)’ has eight letters)
An adequate theory of quotation must explain why not. The product of
quoting ‘\(x\)’ is an expression that refers to the
24
th
letter of the Roman alphabet. The point is that
quotation marks, at least in natural language, cannot be quantified
into because they trap the variable; what results is a quotation that
refers to that very variable. Together, BQ1 and BQ2 place tough
constraints on developing a compositional theory for quotation:
BQ1
would suggest quotational environments are opaque
yet opaque environments often permit quantifying in which, per
BQ2
, quotation doesn’t. For discussion of
compositionality and quotation, see Pagin and Westerståhl 2010,
Bazzoni 2016, and Rabern 2022.
BQ3.
Quotation can be used to introduce novel
words, symbols and alphabets; it is not limited to the extant lexicon
of any one language.
Both (10) and (11) are true English sentences:
(10)
‘\(\Phi\)’ is not a part of any English expression.
(11)
‘❦’ is not an expression in any natural
language.
An adequate theory of quotation must explain what makes this practice
possible.
BQ4.
There’s a particularly close
relationship between quotations and their semantic values.
“lobsters” and its semantic value are more intimately
related than ‘lobster’ and its semantic value, i.e., the
relationship between “lobster” and ‘lobster’
is closer than that which obtains between ‘lobsters’ and
lobsters. Whereas the quotation (i.e., “lobster”), in some
way to be further explained, has its referent (i.e.,
‘lobster’) contained in it, the semantic value of
‘lobsters’, i.e., lobsters, are not contained in
‘lobster’. One way to put it is that an expression e is
in
the quotation of e. No matter how one chooses to spell
this out, any theory of quotation must explain this relationship.
BQ5.
To understand quotation is to have an
infinite capacity, a capacity to understand and generate a potential
infinity of new quotations.
We don’t learn quotations one by one. Never having encountered
the quotation in (10) or (11) does nothing to prohibit comprehending
them (Christensen 1967, p. 362) and identifying their semantic
values.
Similarly, there doesn’t seem to be any upper bound on a
speaker’s ability to generate novel quotations. One natural
explanation for this is that quotation is a
productive
device
in natural language.
BQ6.
Quoted words can be simultaneously used and
mentioned.
This is an important observation due to Davidson,
as exemplified in
(12).
(12)
Quine said that quotation ‘has a certain anomalous
feature’.
(12) is called a ‘mixed quotation’. This is because it
mixes the direct quotation (as in (5)) and indirect (as in (13)).
(5)
Quine said ‘quotation has a certain anomalous
feature’.
(13)
Quine said that quotation has a certain anomalous feature.
In this regard, the quotation in (12) is, in an intuitive sense,
simultaneously used and mentioned. It is used to say what Quine said
viz.
that quotation has a certain anomalous feature), and
also to say that Quine used the words ‘has a certain anomalous
feature’ in saying it.
Mixed quotation had not been much discussed prior to Davidson (1979)
but it has recently taken center stage in discussions of theories of
quotation. For those who believe themselves unfamiliar with the data,
we point out that mixed quotation is one of the most frequently used
forms of quotation. Casually peruse any newspaper and you will find
passages like the following from the
New York Times
NYT
Dec 7, 2004: The court ruled that the sentence was
invalid because the document signed into law by President Bill Clinton
contained a phrase that was illogical. The law said that defendants
like Mr. Pabon, who was convicted two years ago of advertising to
receive or distribute child pornography over the Internet, should be
fined or receive a mandatory minimum sentence of 10 years ‘and
both.’ The appeals court said this language ‘makes no
sense.’
In the 21st century, mixed quotation has become one of the central
topics in the theory of quotation: a range of interesting data has
been brought to light and a flurry of theories have been defended. In
Section 4
below we introduce some of the data.
In what follows we will refer back to these six features and make the
following assumption:
It is a necessary adequacy condition on a theory of quotation that it
either explains how quotations can exhibit features (BQ1)–(BQ6),
or, if it fails to do so, then it must present an argument for why any
unexplained feature doesn’t require explanation.
Before going on, an important note: quotation, perhaps more than any
other area of language, is difficult because not only is there no
consensus about what the correct theory is, but there’s also
basic disagreements about the space of options, and about the data to
be accounted for. We’ll see this below, but it’s important
to bear in mind that when it comes to quotation, the consensus is that
there’s no consensus, and disagreement is the norm.
2.1 Guiding Questions for Theories of Quotation
BQ1–BQ6 play an important role because theories of quotation are
attempts to answer certain questions, and those questions won’t
have satisfactory answers unless BQ1–BQ6 are accounted for.
Three questions can be thought of as the guiding questions for a
theory of quotation:
Q1. In a quotation,
what
does the referring? There are three
options:
The quotation marks
The expression between the quotation marks
A complex of the expression and the quotation marks
Alternatively, one might hold that quotations fail to refer at all,
but rather that speakers refer contingent upon the intentions with
which they use an expression—with or without quotation
marks.
Q2.
How
do quotations refer?
Are they names, descriptions, demonstratives, functors or some
sui
generis
linguistic category?
In addition to Q1 and Q2, theories of quotation often try answer a
third question:
Q3.
What
do quotations refer to?
What kinds of objects are picked out? Is it always the same object or
are quotations
ambiguous?
Our primary focus in what follows will be in Q1 and Q2, but along the
way we will also address Q3.
Section 5
is entirely devoted to Q3.)
2.2 The Use-Mention Distinction
It is standard practice in philosophy to distinguish the
use
of an expression from the
mentioning
of it. Confusing these
two is often taken to be a philosophical mortal sin. Despite its
ubiquitous appeal, it is controversial exactly how to draw the
distinction. The initial thought is easy enough. Consider (D1) and
(D2):
D1. Jim went to Paris
D2. ‘Jim’ has three letters
In (D1) the word ‘Jim’ is used to talk about (or signify,
or denote) a person, i.e. Jim, and the sentence says about that person
that he went to Paris. In (D2) the word is not used in that way.
Instead, ‘Jim’ is used to talk about (or signify or
denote) a word, i.e. ‘Jim’, and the sentence says about
that word that it has three
letters
In (D1), ‘Jim’ is being
used
and in (D2) it is
being
mentioned
The initial thought seems right as far as it goes, but it might be
thought it doesn’t go far enough. One might hope that we could
say
more
to flesh it out in an uncontroversial way, a
theory-neutral way that any party to the debate could agree on. In
fact, though (and this is not recognized enough), this is not
straightforward. To see this, consider a couple of ways to try to
build on the initial characterization:
Expression \(E\) is mentioned in sentence \(S\) just in
case \(E\) is quoted in \(S\).
A problem with this is that according to some theories, an expression
can be mentioned without being quoted (see Objection 3 in section
3.3.2).
Expression \(E\) is mentioned in sentence \(S\) just in
case it is used to refer to itself in \(S\).
This is problematic in a couple of ways. First, notice that according
to (ii), \(E\) must be used in order to mention it; that’s
potentially puzzling. More significantly, it is controversial whether
standard meta-linguistic devices such as quotation are referring
expressions. The theories presented in sections 3.2 and 3.3 treat
quotations as descriptions. If descriptions are quantified
expressions, then quotations are quantifiers, and quantifiers are
typically not treated as referring expressions. Proposals along the
lines of (ii) would also have to ensure that ‘the first seven
words in this sentence’ in (D3) don’t end up referring to
‘the first seven words in this sentence’:
D3. The first seven words in this sentence contain thirty-two
letters.
What this suggests is that any attempt to characterize the distinction
between
use
and
mention
more sophisticated than the
initial characterization in this section will need to address at least
some of the tricky issues that face the various theories of quotation
we describe in what follows.
3. Five Theories of Quotation
There are, roughly, five kinds of theories of quotation that have been
central to the discussion of quotation: the Proper Name Theory, the
Description Theory, the Demonstrative/Paratactic Theory, the
Disquotational Theory, and the Use/Identity Theory. (See here
Maier’s taxonomy (2014b), which just includes the first four
theories, and which evaluates them according in terms of formal
desiderata like compositionality and productivity, and cf. Saka 2013,
which includes ten. This divergence, incidentally, is a good
illustration of the point we made above about lack of consensus among
theorists of quotation.) In the following sections we discuss each of
these and review their strengths and weaknesses.
3.1 The Proper Name Theory
It is now almost a tradition in the literature on quotation to include
a brief dismissive discussion of the Proper Name Theory of Quotation.
This view is described, although not clearly endorsed, in passages in
Quine and Tarski (e.g., Quine 1940, pp. 23–26; 1961, p. 140;
Tarski 1933, p. 159ff), and comments in passing in both Reichenbach
(1947, p. 335) and Carnap (1947, p. 4) strongly suggest they were
adherents. That said, there is some debate about whether Quine and
Tarski ever held the view (see, e.g., Bennett 1988, Richard 1986, Saka
1998, and Gomez-Torrente 2001). And so it is worth mentioning that
although we quote both Quine and Tarski below, we do not thereby
mean to attribute the Proper Name Theory to either of them; we do so
merely because they describe it clearly.
Until recently, it was generally held that the name theory was
‘an utter failure’ (Saka 1998, p. 114), of interest only
as showing us how not to theorize about quotation. In recent years,
however, there have been at least two defenses of something like the
name theory (Gomez-Torrente 2013 and Johnson 2018). We here present
the reasons the name theory was long taken to be a non-starter, as
well as the responses to these problems by the authors mentioned
above.
According to the Proper Name Theory, quotations are unstructured
proper names of the quoted expressions. 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. (Quine 1940, p. 26)
The personal name buried within the first word of the statement
‘‘Cicero’ has six letters’, e.g., is logically
no more germane to the statement than is the verb ‘let’
which is buried within the last word. (Quine 1940, p. 26)
Tarski formulates the view as follows:
Quotation-mark names may be treated like single words of a language,
and thus like syntactically simple expressions. The single
constituents of these names—the quotation marks and the
expressions standing between them—fulfill the same function as
the letters and complexes of successive letters in single words. Hence
they can possess no independent meaning. Every quotation-mark name is
then a constant individual name of a definite expression (the
expression enclosed by the quotation marks) and is in fact a name of
the same nature as the proper name of a man. (Tarski 1933, p. 159)
3.1.1 Strengths of the Proper Name Theory
The Proper Name Theory nicely accommodates (BQ1)–(BQ3); that is,
on this theory we see why co-referential expressions cannot be
substituted for one another. According to the Proper Name Theory, the
name ‘Cicero’ does not occur in
‘‘Cicero’’; from the mere fact that Cicero =
Tully, it does not follow that ‘Tully’ can be substituted
for ‘Cicero’ in ‘‘Cicero’’. As
Quine puts it, ‘[t]o make substitution upon a personal name,
with such a context, would be no more justifiable than to make a
substitution upon the term ‘cat’ within the context
‘cattle’ (Quine 1961, p. 141). The Proper Name Theory
permits the creation of new quotations much as natural languages
permit the introduction of new names. And it prohibits quantifying in,
since each quotation is a single word, and so there is nothing to
quantify into. To see this, consider the following. We obviously can’t
quantify into words. The idea barely makes sense. Imagine we have the
sentence ‘there is a birth dearth in Europe’. We can’t
quantify into, say, the word ‘dearth’. The following
doesn’t make sense, where we’re trying to express an existential
generalization satisfied by the word or string
‘earth’:
(14)
\((\exists x)\) (there is a birth \(dx\) in Europe)
This is nonsense: ‘dx’ just can’t mean ‘d’
combined with the value of the variable ‘x’. Now
consider:
(15)
‘bachelor’ has eight letters.
(16)
\((\exists x)\)(‘\(x\)’ has eight letters)
This is also ill-formed. The proper name theorist can explain it.
Remember, for the proper name theorist, quotations are akin to names
and quotation marks are akin to letters. But then if that’s so, (16)
is on a par with (14): both are trying to quantify into expressions
(the former expressions beginning with ‘d’ and the
latter expressions beginning with ‘ and ending with
‘). The obvious badness of (14) can explain the badness of (16),
for the proper name theorist.
3.1.2 Weaknesses of the Proper Name Theory
Here are three objections to the Proper Name Theory of Quotation. (The
main objections are in Davidson (1979, pp. 81–83), though some
were anticipated by Geach (1957, p.79ff).)
Objection 1:
The Proper Name Theory cannot
explain how we can generate and interpret an indefinite number of
novel quotations (see BQ5).
If quotations were proper names and lacked semantic structure
altogether, then there would be no rule for determining how to
generate or interpret novel quotations. To understand one would be to
learn a new name. (Remember, the quotation marks, according to Quine,
carry no more significance than the serifs you see on these letters.)
But (11), e.g., can be understood by someone who has never encountered
its quoted symbol before.
(11)
‘❦’ is not a letter in any language
Understanding (11) is not like understanding a sentence with a
previously unknown proper name. Upon encountering (11), it would seem
that you know exactly which symbol is being referenced in a way that
you do not with a name you’ve never before encountered.
For a long time, this objection was thought fatal. However, a paper by
Mario Gomez-Torrente (2013) suggests a way to square something like
the name view with
Objection 1
. In brief, he points
out that there are some names that can be indefinitely generated and
interpreted. For example, street names. ‘East 6th Street’,
presumably, is a bona fide name, but one doesn’t either need to
individually bestow it on the particular street it applies to (the
street between East 5th and East 7th), nor does one need to have
already learned it to understand what ‘East 6th Street has some
great bars’ means.
Objection 2
There’s a special relationship
between quotations and their semantic values (see BQ4).
According to the Proper Name Theory, the relationship between
“lobsters” and ‘lobsters’ is no closer than
the relationship between ‘lobsters’ and lobsters. That
seems to miss the fundamental aspect of quotation spelled out in
(BQ4).
Davidson summarizes these first two objections succinctly:
If quotations are structureless singular terms, then there is no more
significance to the
category
of quotation-mark names than to
the category of names that begin and end with the letter
‘a’ (‘Atlanta’, ‘Alabama’,
‘Alta’, ‘Athena’, etc.). On this view, there
is no relation, beyond an accident of spelling, between an expression
and the quotation-mark name of that expression. (Davidson 1979, pp.
81–82; cf., also, Garcia-Carpintero 1994, pp. 254–55)
Again, however, some work has pushed back against these old
objections. Michael Johnson (2018) points to some names which bear
special relationships to their referents. In particular, he points to
onomatopoeia
. For example, ‘buzz’ refers to a
sound, a buzzing sound. But the word ‘buzz’, when uttered,
makes the sound it refers to. Roughly speaking, Johnson holds that we
could view quotations in the same way: as he says, ‘the name of
a person sounds like the name of his name’ (2018). For some
discussion of Johnson’s view, see Miller 2019.
Objection 3
Proper Name Theory leaves no room
for dual use and mention (see BQ6).
If quotations were proper names, and if their interiors lacked
significant structure, there would seem to be no room for dual usage
of the kind found in (12); indeed, on the Proper Name Theory, (12) has
the same interpretive form as (17):
(12)
Quine said that quotation ‘has a certain anomalous
feature’.
(17)
Quine said that quotation Ted.
That is, the Proper Name Theory fails to account for (BQ6).
Other objections have been raised against the Proper Name Theory of
Quotation. For further discussion, see Davidson (1979), Cappelen and
Lepore (1997b) and Saka (1998). Whether the recent defenders of the
name theory will be able to take it off its deathbed remains to be
seen, but it’s fair to say that their work shows that reports of
its demise have been, at least, a bit premature.
3.2 The Description Theory of Quotation
The Description Theory of Quotation was introduced in order to
guarantee that ‘a quoted series of expressions is always a
series of quoted expressions’ (Geach 1957, p. 82) and not
‘a single long word, whose parts have no separate
significance’ (
ibid
., p. 82). According to this theory,
there is a set of basic units in each language: words, according to
Geach (
ibid
., Ch. 18 and 1970); letters, according to Tarski
(1956, p. 160) and Quine (1960, p. 143, p.
212)).
This view retains the Proper Name Theory for basic quotations, e.g.,
according to Quine, “\(a\)” is a name of one letter,
“\(b\)” a name of another, etc. For Geach, each word
has a quotation name. Complex quotations, i.e., quotations with more
than one basic unit, are understood as descriptions of concatenations
of the basic units. Here is an illustration from Geach (where
‘-’ is his sign for concatenation):
…the quotation ‘‘man is mortal’’ is
rightly understood only if we read it as meaning the same as
‘‘man’-‘is’-‘mortal’’,
i.e., read it as
describing
the quoted expression in terms of
the expressions it contains and their order. (Geach 1957, pp.
82–83)
For Quine and Tarski, (4) gets analyzed as (18):
(4)
‘Bachelor’ has eight letters.
(18)
‘B’-‘a’-‘c’-‘h’-‘e’-‘l’-‘o’-‘r’
has eight letters
where ‘-’ is their sign for concatenation and the
individual quotations are names of the letters.
Davidson characterizes the difference between the two versions as
follows:
In primitive notation, which reveals all structure to the eye, Geach
has an easier time writing (for only each word needs quotation marks)
but a harder time learning or describing the language (he has a much
larger primitive vocabulary—twice normal size if we disregard
iteration). (Davidson 1979, p. 84)
That is to say, if Geach’s theory were true, the appropriate
representation of (4) is simply (4), whereas for Tarski and Quine
it’s the much more unwieldy (18). On the other hand, a Geachian
language has (at least) twice as many words as we might expect: for
each word in the dictionary, there would also be the word that stands
for that word (got by enclosing the initial word in quotation
marks).
3.2.1 Strengths and Weaknesses of the Description Theory
In one respect the Description Theory is an immense improvement over
the Proper Name Theory: it deals with no more than a finite set of
basic names, thus, potentially accommodating (BQ5). In other respects,
however, the theory is, by a wide consensus, not much of an
improvement over the Proper Name Theory. At the basic level, the
theory still treats quotations as names. So, at that level, it
inherits many of the problems confronting the simpler Proper Name
Theory. Some of the more obvious objections are these (again we
mention only some of the most obvious ones here since the theory is
not central to contemporary discussions):
At the basic level (i.e., the level of words or letters),
there’s no rule for determining how to interpret and generate
novel quotations (BQ3). This is so because there is no
priori
reason to believe there are finitely many basic
expressions (cf., Lepore 1999).
At the basic level, it doesn’t explain the special
relationship between the expression and the quotation of that
expression (BQ4). It’s obvious to us that “Sam” and
“Alice” do not refer to the same expression, but how can
Geach explain this triviality if both are just proper names; ditto for
Quine with respect to “a” and “b” (cf.,
Davidson 1979, p. 87).
At the basic level, it fails to account for dual use and mention
(BQ6). This is particularly a problem for Geach’s version. Any
account, including the Proper Name Theory, according to which the
semantic function of word-tokens inside quotation marks is just to
refer to word-types (or some other type of linguistic entity) fails to
assign correct truth-conditions to (12). In essence, it predicts that
(12) is syntactically equivalent to something like ‘Quine said
that quotation the phrase’, which is ungrammatical as containing
a definite description where a verb phrase should be.
According to Davidson (1979, pp. 86–87), the Description
Theory can’t explain why we can’t quantify into quotation
(i.e. it fails to account for BQ2). Roughly, the thought seems to be
that we can existentially generalize from (4) to ‘some word has
eight letters’. But by a similar sort of reasoning, we should be
able to move from (18) to ‘some letter is that such that
‘B’-‘a’-‘c’-‘h’-‘e’-‘l’-it-‘r’
has eight letters’. But clearly no such move can be made.
3.3 The Demonstrative Theory of Quotation
The seminal paper on quotation in the twentieth century is, by almost
universal consensus, Davidson’s ‘Quotation’ (1979).
It is without comparison the most discussed and influential paper on
the subject. The view Davidson defends is called the Demonstrative
Theory. (It is also called the Paratactic Theory, though we shall use
the former label in our discussion.) The Demonstrative Theory is
presented in the final pages of ‘Quotation’ and the key
passages are
these:
…quotation marks…help refer to a shape by pointing out
something that has it…The singular term is the quotation marks,
which may be read ‘the expression a token of which is
here’. (Davidson 1979, p. 90)
On my theory which we may call the demonstrative theory of quotation,
the inscription inside does not refer to anything at all, nor is it
part of any expression that does. Rather it is the quotation marks
that do all the referring, and they help to refer to a shape by
pointing out something that has it. (Davidson 1979, p. 90)
Quotation marks could be warped so as to remove the quoted material
from a sentence in which they play no semantic role. Thus instead
of:
‘Alice swooned’ is a sentence.
we could write:
Alice swooned. The expression of which this is a token is a sentence.
(Davidson 1979, p. 90)
The Demonstrative Theory has three central components. First, the
quotation marks are treated as contributing a definite description
containing a demonstrative to sentences in which they occur, i.e., the
quotation marks in (4) become ‘The expression of which this is a
token’, as in (19):
(4)
‘Bachelor’ has eight letters.
(19)
Bachelor. The expression of which that is a token has eight
letters.
Second, in the logical form of a sentence containing a quotation, the
token that occurs between the two quotation marks in the surface
syntax is displaced, so to speak, from the sentence containing the
quotation. What occurs between the quotation marks in the surface
syntax is not part of the sentence in which those quotation marks
occur. Having been displaced from its surface syntactic position, it
serves as an object apt for demonstrative reference (An analogy: if I
utter ‘consider the word ‘apples’.’,
‘apples’ becomes available for demonstrative reference, as
when I continue ‘It is an English word.’)
Third, utterances of quotation marks, by virtue of having a
demonstrative/indexical ingredient, refer to the expression
instantiated by the demonstrated token, i.e., the expression
instantiated by the token that in surface syntax sits between the
quotation marks.
3.3.1 Strengths of the Demonstrative Theory
The Demonstrative Theory is attractive for at least five reasons.
First, to grasp the function of quotation marks is to acquire a
capacity with infinite applications (BQ5). The Demonstrative Theory
explains why: there’s no limit to the kinds of entities we can
demonstrate. Hence, (BQ5) is explained without making quotation a
productive device (for elaboration see Cappelen and Lepore 1997b).
Second, opacity is explained (BQ1): There’s no reason to think
that two sentences demonstrating different objects will have the same
truth-value. (4) and (8) demonstrate different objects, so
there’s no more reason to think the move from (4) to (8) is
truth preserving than there is to think that the move from (20) to
(21) is:
(20)
That’s nice.
(21)
That’s nice.
Third, we have an elegant explanation of mixed quotation, i.e., we can
explain (BQ6). Davidson says:
I said that for the demonstrative theory the quoted material was no
part, semantically, of the quoting sentence. But this was stronger
than necessary or desirable. The device of pointing can be used on
whatever is in range of the pointer, and there is no reason why an
inscription in active use can’t be ostended in the process of
mentioning an expression. (Davidson 1979, p. 91)
This, according to Davidson, is what goes on in (12). A token that is
being used for one purpose is at the same time demonstrated for
another: ‘Any token may serve as target for the arrows of
quotation, so in particular a quoting sentence may after all by chance
contain a token with the shape needed for the purposes of
quotation’ (Davidson 1979, pp. 90–91; cf., also, Cappelen
and Lepore 1997b). On this view, (12) is understood as (22). (Note:
the ‘these’ in (22) is accompanied by a pointing or
indexing to the token of Quine’s words.)
(12)
Quine said that quotation ‘has a certain anomalous
feature’
(22)
Quine said, using words of which these are a token, that quotation
has a certain anomalous
feature.
Fourth, there is no mystery about how to introduce new vocabulary;
since there’s no limit to what can be demonstrated,
there’s no limit to what can be quoted. (BQ3) is explained.
Finally, quantifying-in is obviously ruled out, since the quoted token
is placed outside the quoting sentence, i.e., the Demonstrative Theory
can explain (BQ2).
The Demonstrative Theory is both bold and radical. It triggered an
entire cottage industry devoted to criticizing and defending it. For
proponents, see, for example, Partee 1973; Garcia-Carpintero 1994,
2017, 2018; Cappelen and Lepore 1997b; McCullagh 2007; and Predelli
2008; for critics, see just about anyone else writing on quotation
after 1979.
3.3.2 Weaknesses of the Demonstrative Theory
In what follows we present five criticisms of the Demonstrative
Theory. Needless to say, the list is not exhaustive (e.g., see
Sorensen 2008 and Saka 2011b on the problem of empty quotation), and
indeed, each objection has triggered lively discussion which space
limitations prohibit our taking up here.
Objection 1.
If the Demonstrative Theory were
correct, it should be possible for (4) to demonstrate, e.g., a
penguin.
Here is an argument that mimics a range of objections raised against
Davidson’s account of the semantics for indirect reports (Burge
1986, Stainton
1999).
Recall that according to Davidson the logical form of (4) is (19).
(4)
‘Bachelor’ has eight letters.
(19)
Bachelor. The expression of which that is a token has eight
letters.
(19) contains a demonstrative and demonstratives refer to whatever is
demonstrated with their use. What is demonstrated on a given occasion
depends on the speaker (either the demonstration or the intention or
some combination of the two). It should be possible, then, for a
speaker to utter (4) and
not
demonstrate the exhibited token
of ‘bachelor’. That is to say, if there really is a
demonstrative in (19), that demonstrative should have the same kind of
freedom that other demonstratives have: it should be able to refer to,
for example, a nearby penguin. Of course, no utterance of (4) makes
reference to a penguin. So the Demonstrative Theory is
wrong.
10
(For replies to this objection see Cappelen and Lepore 1999b.)
Objection 2
The Problem of Relevant
Features
According to Davidson, a quotation refers to an expression indirectly,
by referring to a token that instantiates that expression. Davidson
thinks expressions are
shapes
or
patterns
(see
Davidson 1979, p. 90). A problem for this view is that any one token
instantiates indefinitely many distinct shapes or patterns, i.e. many
different expressions. So how, on Davidson’s view, do we get
from a particular token to a unique type, i.e., from a token to an
expression?
Jonathan Bennett formulates the problem as follows:
Any displayed token has countless features, and so it is of countless
different kinds. Therefore, to say the inscription-types instantiated
here: Sheep or what amounts to the same thing, the inscription-type
each token of which is like this: Sheep is to leave things open to an
intolerable degree. How do we narrow it down? That is what I call the
problem of relevant features. It urgently confronts the demonstrative
theory which must be amplified so as to meet it. (Bennett 1988, p.
403, see also Washington 1992, pp. 595–7.)
A related worry is this: Read (4) out loud. It seems obvious that a
spoken utterance says (makes) the same claim as a written utterance of
(4). On the Demonstrative Theory it is unclear why this should be so:
the spoken utterance demonstrates a vocal pattern, and the written
utterance a graphemic pattern. They seem to be attributing properties
to different objects. (Several suggestions are on offer for how to
amend the Demonstrative Theory in this respect: cf., Garcia-Carpintero
1994, Cappelen and Lepore 1997b, 1999c, and Davidson 1999.)
Objection 3
The Problem of Missing Quotation
Marks
According to Davidson, quotation marks are what are used to do the
referring. They are descriptions containing demonstratives whose uses
refer to whatever pattern is instantiated by the demonstrated token.
This makes the presence of quotation marks essential. Much recent work
on quotation argues that we can quote (or do something quote-like)
without
quotation marks and that a theory of quotation should
be capable of explaining how quotation can take place in the absence
of quotation marks. Here is Reimer’s version of this
objection:
Consider the following sentence:
(7)
Bachelor
has eight letters.
Here, we have a case in which an expression is quoted—not by
means of quotation marks—but by means of italicization. But
surely it would be absurd to suppose (consistently with
Davidson’s view) that the italicization of (7)’s subject
term is itself a demonstrative expression! (Reimer 1996, p. 135)
The same idea is expressed by Washington (1992):
In conversation, oral promptings (‘Quote-unquote’) or
finger-dance quotes can often be omitted without impairing the
intelligibility or well-formedness of the utterance. When I introduce
myself, I do not say ‘My name is quote-unquote Corey,’ nor
do I make little finger gestures or even use different intonation in
order to show that it is my name and not myself that is being talked
about. (Washington 1992, p. 588; Saka 1998, pp. 118–19; Recanati
2001; and Benbaji 2004a, 2004b.)
The Demonstrative Theory depends on the presence of quotation marks
(inasmuch as they are what get used to do the referring), so if
quotation can occur without quotation marks (as in the Reimer and
Washington cases), it’s hard to see how the Demonstrative Theory
is adequate.
Proponents have been unimpressed. Consider an utterance of (6):
(6)
My name is Donald.
Several possible replies can be made by proponents of the
Demonstrative Theory. One thing a Demonstrative Theorist might say is
that there are no missing quotation marks in (6): they are in the
logical form of the sentence, not in its surface syntax.
Alternatively, quotation marks for an utterance of (6) could be
generated as conversational or conventional implicatures. (He
can’t be saying that Donald, the person, is his name as he knows
that that is false, so he must be conversationally implicating that
the expression of which he used a token is a name (cf.
Garcia-Carpintero 1994, pp. 262–63)). Or perhaps we could appeal
to the distinction between semantic reference and speaker reference.
(6) is grammatically correct but false; nonetheless, someone can
succeed in communicating something true about Donald’s name if
he succeeds in conveying to his audience his intention to refer to it
(cf., Gomez-Torrente 2001).
Finally, a Demonstrative Theorist can argue that these other
quotation-like phenomena are just that—quotation-like. They
require a separate treatment. There’s no need for a unified
theory (cf., Cappelen and Lepore 2003).
Objection 4
The Problem of Iteration
The Demonstrative Theory seems to have difficulty dealing with
iterated quotation. (24) refers to the quotation in (23):
(23)
‘Smooth’ is an English expression.
(24)
“Smooth” is an English expression.
The Demonstrative Theory’s account for (23) is (25).
(25)
Smooth. The expression of which that is a token is an English
expression
How, then, can the account accommodate (24)? (24), after all, includes
two
sets of quotation marks. It might seem like the
Demonstrative Theory would have to treat it as the ungrammatical (26)
or the unintelligible (27).
(26)
Smooth. That that is an English expression.
(27)
Smooth. That. That is an English expression.
This objection has been raised by Saka (1998, pp. 119–20),
Reimer (1996), and Washington (1992).
In response, Demonstrative Theorists insist that quotations are not
iterative. On this topic, Cappelen and Lepore write:
Quoted expressions are exhibited so that speakers can talk about the
patterns (according to Davidson) they instantiate. The semantic
properties of the tokens are not in active use; they are semantically
inert…So, quotation marks within quotation marks are
semantically inert. (Cappelen and Lepore 1997b, pp. 439–40)
For further discussion of whether quotation is a genuinely iterative
device, see Cappelen and Lepore (1999a) and Saka (2013).
Objection 5.
The Problem of Open Quotation:
Dangling Singular Terms
Recanati (2001) focuses on cases where quoted expressions do not serve
as noun phrases in sentences. He has in mind cases like (29) and
(30):
(29)
Stop that John! ‘Nobody likes me’, ‘I am
miserable’…
Don’t you think you exaggerate a bit?
(30)
The story-teller cleared his throat and started talking.
‘Once upon a time, there was a beautiful princess named
Arabella. She loved snakes and always had a couple of pythons around
her…’
In these cases it looks like the Demonstrative Theory would have to
postulate a dangling singular term, something like (31) or (32):
(31)
Stop that John. That. Nobody likes me. That. I am miserable.
… Don’t you think you exaggerate a bit?
(32)
The story-teller cleared his throat and started talking. That.
Once upon a time, there was a beautiful princess named Arabella. She
loved snakes and always had a couple of pythons around her…
In response to the idea that (29) is elliptical for (33),
(33)
Stop that John!
You say
‘Nobody likes me’,
‘I am miserable’ … Don’t you think you
exaggerate a bit?
Recanati says:
I deny that [29] and [33] are synonymous. Nor are there any grounds
for postulating ellipsis here except the desire to save the theory in
the face of obvious counterexamples (Recanati 2001, p. 654).
3.4 Disquotational Theory of Quotation
Less baroque than the Demonstrative Theory, the Disquotational Theory
is probably the simplest, most natural and obvious account of
quotation. It is endorsed by a wide range of authors, often in
passing, as if completely obvious. A simple version of it can be found
in Richard’s Disquotational Schema (DQR):
DQR: For any expression \(e\), the left quote (lq) followed by e
followed by the right quote (rq) denotes e (Richard 1986, p. 397)
Ludwig and Ray write:
Its semantic function is given by the following reference clause in
the theory: (ref\((\ulcorner\)‘E’\(\urcorner)
=\) E (Ludwig and Ray, 1998, p.163, note 43)
where the ‘\(\ulcorner\)’ and
‘\(\urcorner\)’ are the left and right corner quotes.
(See also Smullyan 1957, Mates 1972, p. 21; Wallace 1972, p. 237;
Salmon 1986, p. 6, Gomez-Torrente 2001, Botteral and Stainton 2005. A
nearby view is the minimalism of Cappelen and Lepore 2007; discussion
of it can be found in Garcia-Carpintero 2011.)
On this account, quotations are not proper names, or descriptions or
demonstratives but rather they are functors that take an expression as
their argument and return it as value.
3.4.1 Strengths of Disquotational Theory
The two most obvious strengths of the Disquotational Theory are its
simplicity and intuitiveness. If asked how quotation functions, the
obvious reply is something along the lines of (DQR). It is also an
axiom (or axiom schema) that’s pleasingly simple and requires no
complicated assumptions about the surface structure of the sentence
(in this respect, it has a clear edge on the Demonstrative
Theory).
In addition to be being exceedingly simple and intuitive, this theory
easily accounts for three of the Basic Facts about Quotation.
First, it explains opacity: ‘bachelor’ and
‘unmarried man’ have different semantic values because
what is between the quotation marks are distinct expressions. An
expression’s semantic value is irrelevant for determining the
semantic value of the quotation of that expression, thus accounting
for (BQ1).
Second, since quotations are functor expressions without internal
structure, (BQ2) is explained: there’s no possibility of
quantifying into a quotation on this view.
Finally, since quotations, as functors, map all expressions onto
themselves, this account can explain the special relationship between
a quotation and the quoted expression—namely,
identity—thus explaining (BQ4).
3.4.2 Weaknesses of the Disquotational Theory
Even with these advantages, at least three serious difficulties
confront the Disquotational Theory.
First, (BQ3) says that we can
use quotation to refer to symbols that are not in the English lexicon,
as in (9) and (10):
(9)
‘\(\Phi\)’ is not a part of any English expression.
(10)
‘❦’ is not an expression in any language.
(DQR) says we can take
any expression
e, put quotation marks
around e, and what results is an expression that refers to e. What
exactly is meant by ‘any expression’ in (DQR)? Richard
offers the following answer:
It is
easy enough
to come up with a
finite
list of
elements (the letters, punctuation symbols, the digits, the space,
etc.) and an operation (concatenation) with which one can generate all
of the concatenates…If we are formalizing a grammar for a
language with quotation names, we would include, as part of the
specification of the lexicon, a proviso to the effect that, for each
concatenate e, the left quote (lq), followed by e, followed by the
right quote (rq) is a singular term (Richard 1986, pp. 386–89,
our emphasis).
If this is how expressions are generated, how then are we to account
for the truth of (9) and (10)? More generally, the worry is this:
(DQR) needs to specify, in some manner or other, the domain of
expressions over which it quantifies. How can it do this without
unreasonably limiting the kinds of symbols that can be
quoted?
11
(See Lepore (1999) for elaboration on this point.)
Second, according to (BQ6), a theory of quotation should leave room
for dual use and mention. It is hard to see how (DQR) leaves such
room. If quotes are referring expressions (as they are according to
DQR), then if we let ‘Ted’ name the expression ‘has
a certain anomalous feature’, (12) should say the same as,
express the same proposition as, (17).
(12)
Quine said that quotation ‘has a certain anomalous
feature’.
(17)
Quine said that quotation Ted.
Not only should (12) and (17) express the same proposition, according
to DQR, they should also have the same logical form. That’s an
obviously incorrect account of (12). (For further discussion of mixed
quotation see
Section 4
below.)
Finally, those who raise Objection 4 against the Demonstrative Theory
would probably raise the same objection here: the Disquotational
Theory fails to account for quotation without quotation marks. All the
work is being done by quotation marks, so there’s no room for
quotation without quotation marks. (DQR) proponents would presumably
be as unimpressed as Demonstrative Theorists and their replies would
be much the same (see above).
3.5 The Identity Theory of Quotation (or, better: the Use-Theory of Quotation)
The label ‘the Identity Theory of Quotation’ was first
used, as far as we know, by Washington (1992), though he attributes
the view to Frege (1892) and Searle (1969). However, the passages in
which this view was allegedly offered prior to Washington hardly count
as presenting a theory; they read more like dogmatic pronouncements
devoid of argumentation. Since Washington’s paper was published
related views have been developed in more detail by Saka (1998, 2004,
2011a), Reimer (1996), Recanati (2000, 2001) and others.
Washington’s presentation of the Identity Theory is somewhat
compressed—the key passage is this:
The quotation as a whole is analyzed into the marks that signify
quotational use of the quoted expression and the quoted expression
itself used to mention an object. All expressions, even those whose
standard uses are not as mentioning expressions, become mentioning
expressions in quotation…a quoted expression is related to its
value by identity: a quoted expression mentions itself. (Washington
1992, p. 557)
There are three important components of this
view:
12
13
The use of quotation marks (as in (4)) is a
derivative
phenomenon. The basic phenomenon is what Washington calls
quotational use
(Washington 1992, p. 557).
The primary function of quotation marks is to indicate that words
are used quotationally (or mentioned) and not (merely) used with their
regular extensions.
Quotation marks do
not
refer according to Washington;
that which is doing the referring in the quotation is the expression
itself, so in (2), for example, it is ‘Aristotle’, not
“Aristotle” that refers to ‘Aristotle’, i.e.,
‘Aristotle’ refers to itself. It refers to itself because
it is being used quotationally (not with its regular extension.)
One way to get a handle on this kind of view is to consider a spoken
utterance of (6)
(6)
My name is Donald.
According to the Use/Identity theory, (6), when spoken, is grammatical
(Washington 1992, pp. 588–90). There are no missing (or
implicit) quotation marks. When uttered by a person whose name is
‘Donald’ it is true (if ‘Donald’ is used
quotationally). The function of quotation marks in written language is
simply to indicate that words are being used in this special,
quotational, way, i.e. not (only) with their regular extensions. In
other words, ‘Donald’ can be used in two different ways:
with its usual semantic value (its regular referent) or quotationally.
In the latter case, its semantic value is an expression.
We wish that the Identity Theory were not called ‘the Identity
Theory’. (In private communication, Washington has expressed the
same sentiment.) The ‘identity’ component is picked up
from the formulation, ‘a quoted expression is related to its
value by identity: a quoted expression mentions itself’. This
formulation, however, is deeply misleading, since according to both
Washington (and later, for example, Saka (1998)), quotations are
ambiguous. They can, according to Washington, refer to types, tokens,
or shapes (see Washington 1992, p. 594) and according to another
proponent of this kind of view, Saka, they are even more flexible
(Saka 1998, see further presentation of Saka’s view below). A
better label would be ‘the Use-Theory of Quotation’, since
this emphasizes the point that a proper understanding of quotation
requires appealing to a special way of using language. In what
follows, we use the ungainly compromise ‘the Use/Identity
Theory’.
It is useful to contrast Washington’s account with the
Disquotational Theory. Recall, that, according to DQR
, quotation
marks
have a semantic function and that function is spelled out
in the disquotational schema (DQR). (DQR) is a semantic axiom. It
treats quotation marks as identity functions. Speakers’
intentions figure not at all in this axiom (other than the intention
to speak English). There is no need, on the Disquotational Theory, to
appeal to a special kind of quotational usage. For Washington, the
quotation marks have no genuine semantic function. They are no more
than a heuristic device for indicating that expressions are used in a
special way, i.e. quotationally.
3.5.1 Other Versions of the Use/Identity Theory: Reimer, Recanati, Saka, and the Picture Theory
Several recent views share important components of Washington’s.
Reimer (2003) combines versions of the Demonstrative Theory and the
Identity Theory. Recanati (2000, 2001, 2010) doesn’t explicitly
discuss any version of the Identity Theory, but his theory
incorporates some of its components; it is distinctive by focusing on
what he calls ‘Open Quotation’ (see Objection 5 to the
Demonstrative Theory above) and the
iconic
aspects of
quotation (see Recanati 2001 for elaboration). Two theories might be
worth a closer look for those interested in exploring Use/Identity
Theories further (the second of these is mostly of historical
interest).
First, Saka (1998, 1999, 2003) has developed a theory that has much in
common with Washington’s (though it also differs in important
respects). He agrees with Washington in emphasizing ‘quotational
use’ (Saka calls it ‘mentioning’), but Saka has more
to say about mentioning than Washington has to say about quotational
use (see Saka 1998 and 2003). Saka also goes further than Washington
in claiming that quotation marks are not required for mentioning even
in written language. (For Washington, it is only in spoken language
that we can quote without quotation marks). According to Saka, (34)
‘is a grammatical and true sentence’ (Saka 1998,
p.118.)
(34)
Cats is a noun.
Even though Saka agrees with Washington that quotation marks
‘announce ‘I am not (merely) using expression X;
I am also mentioning it’’ (Saka 1998, p. 127), he differs
from Washington in that he assigns them a genuine syntactic and
semantic function (Saka, 1998, p. 128). In this respect he
incorporates components of what we above called the Disquotational
Theory.
14
15
Saka’s account also differs from Washington’s in that he
emphasizes that quotations are ambiguous (or indeterminate) and that
what they refer to depends on the speaker’s intentions (see Saka
1998, pp. 123–4). For further discussion of this point, see
Section 5
below.
Second, there is one view we have not discussed and which might
(admittedly with some difficulty) be squeezed into the category of the
Use/Identity Theory. Quine said a number of things about quotation; in
one passage he writes:
…a quotation is not a description but a hieroglyph; it
designates its object not by describing it in terms of other objects,
but by picturing it. (Quine 1940, p. 26)
Davidson (1979), taking his cue from this passage and others, baptized
the view intimated in Quine’s passage as ‘the Picture
Theory of Quotation’ (cf., also, Christensen 1967, p. 362). As
Davidson notes, on this view:
…it is not the entire quotation, that is, expression named plus
quotation marks, that refers to the expression, but rather the
expression itself. The role of the quotation marks is to indicate how
we are to take the expression within: the quotation marks constitute a
linguistic environment within which expressions do something
special… (Davidson 1979, pp. 83–84)
Notice that according to Davidson’s description of this view,
the quotation marks
per se
have no semantic function; rather,
they indicate that the words are being used in a special way. They are
being used ‘autonymously’, that is, to name themselves. So
understood, the Picture Theory has at least this much in common with a
Use/Identity Theory: they agree that quotation marks are inessential;
they only indicate a special use. They indicate that expressions are
being used as a picture (or as a hieroglyph). The Picture
Theory—if it’s even appropriate to call it a
theory—is never elaborated in any great detail and we suspect
that if it were, it would become obvious that it is a version of the
Use/Identity Theory.
3.5.2 Strengths of the Use/Identity Theory
Use/Identity Theorists claim that their theories are explanatorily
more powerful than traditional semantic theories. By seeing quotation
marks as a parasitic phenomenon, they are able to explain the
semantics of both quotation and this more general phenomenon in a
unified way.
There’s a great deal of specific data that these theories claim
to be able to explain. The most important of these is the possibility
of mentioning without quotation marks (as in spoken language and in
written language when no quotation marks are used). If we take
appearances at face value, that means meta-linguistic discourse (call
it mentioning or quotational use) can take place in the absence of
quotation marks. Hence, an account of meta-linguistic discourse must
proceed independently of an account of the semantics (or pragmatics)
for quotation marks.
The following are additional claims made on behalf of the Use/Identity
Theory (we take it to be an open question at this point whether these
points are sustainable):
Whatever can be mentioned can be quoted. If new symbols and signs
can be mentioned, then they can also be quoted; hence, (BQ3) is
satisfied.
There is, on this view, often a particularly close relationship
between the quoted material and the referent; sometimes it is
identity, sometimes it is instantiation, etc, so in various ways we
might say that (BQ4) is satisfied.
If our capacity for mentioning is limitless, then we have an
account of how quotation can be too (i.e., we have at least the
beginning of an account of (BQ1)).
3.5.3 Weaknesses of the Use/Identity Theory
Discussion of the Use/Identity Theories is not yet as extensive as
discussion of the Demonstrative Theory, so there are fewer objections
to report. We discuss four concerns that have surfaced in various
discussions.
Question about the Relevance of Quotational Use/Mention:
Use/Identity Theories put a great deal of weight on the idea that the
semantics for quotation cannot be developed without an account of
quotational usage
mention
in Saka’s
terminology.) There are several reasons for doubting this, two of
which are these:
It is not obvious that the alleged phenomenon of mention without
quotation marks is genuine. It might very well be, as mentioned in
connection with Objection 3 to the Demonstrative Theory, that it is
not possible to mention without using quotation marks. The cases
appealed to might all turn out to be cases in which a conversational
or conventional implicature is generated and where that implicature
contains quotation marks. Alternatively, the quotation marks might
actually be in the logical form of the sentence through some form of
ellipsis. (See Garcia-Carpintero 1994 and Cappelen and Lepore
1999).
Even if we suppose that the phenomenon of mentioning without
quotation marks is genuine, it is not clear why we should consider it
relevant to the semantics for quotation. Suppose that you’re in
the business of trying to develop a semantic theory of sentences with
quotation marks. Suppose it also turns out that it is possible to talk
about language by mentioning without the use of quotation marks. This
might just be a different way of talking about language. An
interesting phenomenon, no doubt, but not one that needs to have
anything to do with the semantics for sentences containing quotation
marks. It does not follow from there being a variety of ways in which
language can be used to talk about language, that all of these ways
are relevant to the semantics of quotation.
Question about the Semantics-Pragmatics Divide
: On the
Use/Identity Theory, a lot of work is done by pragmatic mechanisms.
The appeal to speaker intentions plays a central role on all levels of
analysis. On Saka’s view, for example, what a quoted expression
refers to is largely up to the speaker’s intentions (and, maybe,
what’s salient in the context of utterance). A consequence of
this view is that there is no guarantee, for example, that an
utterance of (35) or (36) will be true:
(35)
‘\(a\)’ = ‘\(a\)’
(36)
‘run’ is a verb in English
The two ‘\(a\)’s in (35) could refer to different
objects; the “run” in (36) might refer to, for example, a
concept (see Cappelen and Lepore 1999b and Saka 1999, 2003, 2011).
Over-generation Problems
: According to Use/Identity Theories,
you can do a lot with quotation. The question is whether this results
in such theories running into problems of over-generation. Take, for
example, Saka’s claim that a quotation refers to an item
associated
with the expression. There’s only one
restriction: This item cannot be the expression’s regular
extension. If this is the
sole
restriction on what quotations
can be used to refer to, we should be able to do things with quotation
that there’s no evidence that we can. It could, for example, be
the case that in a particular context, the (regular) extension of
‘love’, call it
love
, was associated with the
expression ‘money’; maybe, for some reason, that
association was contextually salient. Nonetheless, (37) cannot be used
to say that
love
plays a central role in many peoples’
lives.
(37)
‘Money’ plays a central role in many peoples’
lives.
Use/Identity Theories have to explain what blocks such readings (or
show that they are possible). (See Saka 2003, and Cappelen and Lepore
2003.)
Dual Use-Mention (BQ6):
Washington says that quotation marks
indicate quotational usage and that expressions used quotationally
refer to themselves (or some related entity). If so, the logical form
of (12) should be that of (17), clearly not a correct result.
(12)
Quine said that quotation ‘has a certain anomalous
feature’.
(17)
Quine said that quotation
Ted.
16
It is at least a challenge to Identity/Use theorists to explain how
the theory can accommodate simultaneous use and mention.
In addition to the above points, there’s a lively debate about
the specifics of identity/use theories. For discussions of Saka, see
Cappelen and Lepore (2003), Reimer (2003), and for discussions of
Recanati, see Cappelen and Lepore (2003), Reimer (2003), Benbaji
(2003, 2004a, 2004b), and Cumming (2003).
We now turn from discussions of large-scale theories of quotation to
discussions of how to understand specific aspects of our quotational
practice. Two issues have been particularly important in the recent
discussions: Mixed Quotation and the alleged ambiguity of quotations.
One’s view of these issues has wide reaching implications for
which theory of quotation one favors. We discuss these in turn.
4. Mixed Quotation
From around the millennium, beginning with Cappelen and Lepore’s
1997a, which made the topic mainstream, mixed quotation has been,
arguably, the area in which theorizing about quotation has advanced
the most. The problem has attracted the attention of both philosophers
and linguists, and our understanding of both the data to be accounted
for, and the various possibilities for doing so, has been greatly
improved.
It’s important to reiterate our above note that here as
elsewhere in the theory of quotation what comprises the phenomenon is
up for debate; different theorists emphasise or play down different
bits of data. What follows, accordingly, doesn’t represent a
consensus position on the problem, simply because there isn’t
one (indeed, somewhat unfortunately, there isn’t even consensus
about the terminology, some calling it ‘hybrid’ or
‘impure’, and some subsuming mixed quotation under
‘scare quoting’; De Brabanter 2010 is helpful here, as
well as being a useful survey of the field from a linguist’s
point of view). We will simply present some of the more interesting
and central data points and problems for a theory of mixed quotation
(for the theories themselves, we refer the reader to Maier’s
(2020) overview article which presents some of the main formal
semantic theories out there. For recent theories by philosophers on
mixed quotation, see Gomez-Torrente 2005, Kirk and Ludwig 2017,
McCullagh 2007, 2017. The main linguists to consult are cited in this
section.)
The paradigm example of mixed quotation is the following,
repeated:
(12)
Quine said that quotation ‘has a certain anomalous
feature’.
It is mostly agreed that this sentence puts forward the following
claim: Quine said (expressed the proposition or content) that
quotation has an anomalous feature, and did so by uttering the
expression ‘has an anomalous feature’. This suggests two
initial requirements:
An adequate theory of mixed quotation must account for the fact
that the quoted material is used by the reporter to characterise the
proposition that the reportee expressed.
It must also account for the fact that the quoted material was
uttered by the reportee.
Let’s call these the
use requirement
and the
mention requirement
, respectively.
While these both seem clearly on the right lines (if you doubt this,
open or scroll to a newspaper and read until you find the first mixed
quotation (it shouldn’t take long) and see if the requirements
are operative) there are problems with both of them. It might
initially seem that the mention requirement requires that the reportee
have uttered verbatim the words ascribed to him or her. That would
account for the anomalousness of:
(38)
# Quine said quotation ‘has a certain anomalous
feature’, although he didn’t use the words ‘has a
certain anomalous feature’.
Things aren’t so clear, though. Certain other pieces of data are
in tension with this purported feature of the mention requirement,
most notably that one can mixed quote someone correctly in a language
other than the one they spoke (Cappelen and Lepore 2007 p. 44). Thus
the following, as well as its parenthesised continuation, is fine:
(39)
Kant said that concepts without intuition ‘are empty’,
(although he didn’t use those words: he said ‘sind
leer’).
Moreover, more subtle problems for the use requirement come from
inflected languages (Shan 2010, Maier 2014a). Thus in Italian, nouns
must agree in gender and number with adjectives. Imagine Ken says:
(40)
Gli uomini italiani mi sembrano molto carini.
This means ‘I find Italian men very cute’, and since the
noun ‘uomini’ (men) is m.pl, so is the adjective
‘carini’ (cute). But if one wanted to mixed quote Ken,
using a different word than ‘uomini’, that different word
would have to agree:
(41)
Ken ha detto che le persone italiane ‘mi sembrano molto
carine’
So the mixed quoted material used to report Ken’s utterance is
not what he said verbatim: he said ‘carini’ and he is
reported as saying ‘carine’ (the example is from Shan
2010). Maier presents similar cases for Dutch. Yet a third problem
comes from the fact that one can ‘clean up’ quotations, by
removing grammatical slurs, umms and ahhs, and so on (see Cappelen and
Lepore 2007 p. 45).
What exactly the mention requirement requires is uncertain, then. But
nor, however, is the use requirement without its problems. There are
at least two, which again differ in subtlety. From the beginning,
it’s been realised that indexicals in mixed quotations cause
intriguing problems. Consider:
(42)
Mr. Greenspan said he agreed with Labor Secretary R. B. Reich
‘on quite a lot of things’. Their accord on this issue, he
said, has proved ‘quite a surprise to both of us’.
(Cappelen and Lepore 1997b, p. 429)
If the second mixed quotation were simply being used, then the second
sentence should be true provided accord was a surprise to the people
writing the report, not to Greenspan and Reich. That’s clearly
the wrong result, and shows that the use requirement needs some
modifiction (for discussion about indexicals in mixed quotation, and
whether they call for Kaplanian monstrous operators, see also Recanati
(2001), Cumming (2003), Geurts and Maier (2003), and Maier
(2016)).
The behaviour of indexicals in mixed quotation is important, because
it gives strong evidence against the possibility that mixed quotation
is a pragmatic phenomenon. A typical pragmatic analysis would say that
the semantic contribution of the quoted material in a mixed quotation
is just its unquoted semantic value (so that the literal semantic
content of (12) is simply ‘Quine said that quotation has a
certain anomalous feature’), but that to utter a mixed quotation
is to imply, but not to say, that the quoted material was used by the
speaker to say what they said.
The advantages of a pragmatic analysis of mixed quotation are those
typical of pragmatic analyses in general: no extra semantic machinery
is required to account for the problem cases. As such, versions of
such an analysis have had many adherents at the end of the twentieth
century (see, for example, Recanati (2001), Clark & Gerrig (1990),
Wilson (2000), Sperber & Wilson (1981), Tsohatzidis (1998),
Staintion (1999), Saka (2003) and the discussion in by Reimer
(2003)).
However, the behaviour of indexicals in mixed quotation make the
pragmatic analysis harder to uphold. One needs to say that (42) is
literally false, and while that is a move that can be made, it’s
not particularly attractive. Accordingly, the pragmatic analysis has
gone somewhat out of fashion, and has been argued against explicitly
explicitly by Gomez-Torrente (2017), and implicitly by most of the
linguists working in the field, for whom the behaviour of indexicals
is a key explanandum of a theory of mixed quotation, and not something
to be explained
away
(although see De Brabanter 2019 for a
pragmatic theory of indexicals in mixed quotation).
The second, more subtle, problem is a consequence of the fact that one
can mixed quote an expression belonging to a dialect other than the
one one speaks, or even to no dialect at all (Shan 2010). Thus even in
informal British English, ‘accident’ always takes the
preposition ‘by’ and never ‘on’. A British
speaker can draw attention to this fact, by uttering, about an
American friend of theirs:
(43)
Meagan said she took out the trash ‘on accident’.
But a British speaker can’t felicitiously use ‘on
accident’ because that’s not a part of British English
speakers’ dialect. Similarly, ‘misunderestimated’
isn’t a word in most anybody’s dialects, but we can
say:
(44)
Bush said that his opponents ‘misunderestimated’
him.
A different, more fundamental, source of complication arises from the
fact that it’s unclear exactly how much to include under mixed
quotation. Thus consider:
(45)
Quine wrote a couple of papers on quotation, the ‘anomalous
feature’ of which he thought about a lot.
(46)
A:
The Godfather II
is a total snooze.
B: Well, Pauline Kael said this ‘total snooze’ is a
defining moment in American cinema. (Potts 2007, p 420)
Should we include such phenomena in our theory of mixed quotation? The
first sentence doesn’t contain a speech verb, while the second
does, but the person to whom the mixed quotation is attributed
isn’t the person who uttered the mixed quoted words. It’s
unclear. Things get even more uncertain when we consider scare quotes
(Predelli 2003a,b, Saka 2013, Gomez-Torrente 2017) and what are known
euphemistically as grocers’ quotes:
(47)
Teens are spending more and more time ‘sexting’,
something of concern to parents and even technology companies.
(48)
[written on a desiccate in a new pair of shoes] Do not
‘eat’!
In the former case, there isn’t any one person who used
‘sexting’; rather, the writer is marking that it’s
used in a community which they are not part of or disapprove of. In
the latter case, the quotation marks are serving as emphasis. While we
probably don’t want to spend a lot of time seriously theorizing
about grocers’ quotes (although see Abbott 2003), whether or not
scare quotes should be treated as we treat mixed quotation (indeed,
whether or not they’re one and the same) is an open question
(see McCullagh 2017 for discussion and an affirmative answer to these
questions; Gomez-Torrente 2017 demurs).
This isn’t even an exhaustive characterization. Thus it’s
important to attend to cognate phenomena like dequotation (Shan 2010)
and free indirect discourse (Maier 2015, 2017). And the theorist of
mixed quotation should at least be aware of the problem of mixed
quoting non-constituent clauses (Abbott 2003, De Brabanter 2022), and
the possibility that one can mixed quote without quotation marks (De
Brabanter 2010). The complexity of the data has inspired much
interesting theorizing, and we expect that mixed quotation will
continue to be a focal question in the theory of quotation.
5. What Kinds of Entities do Quotations Refer to?
Running parallel to the debate about
how
quotations refer
(how they manage to hook up with their semantic values) is a debate
about
what
quotations refer to. One view that is widespread
is that quotations are ambiguous or indeterminate. That is, one and
the same quotation, e.g. “lobster” can, on this view,
refer to different objects on different occasions of use, all
depending on the context of utterance. Garcia-Carpintero (1994, p.
261) illustrates this kind of view and the kind of argument typically
given for it. He says that “gone” can refer to any of the
following:
The expression (‘ ‘gone’ is
dissyllabic’);
Different types instantiated by the tokens
(‘ ‘
gone
’ is cursive’);
Different types somehow related to the token (say, the graphic
version of the uttered quoted material, or the spoken version of the
inscribed quoted material, as in ‘ ‘gone’
sounds nice’);
Different tokens somehow related to the quoted token (‘What
was the part of the title of the movie which, by falling down, caused
the killing?—‘gone’ was’);
The quoted token itself (‘At least one of these words is
heavier than ‘gone’ which you should imagine written in
big wooden letters’);
Others think quotation can pick out contents or concepts. Goldstein
says:
For when Elvis says ‘Baby, don’t say
‘don’t’,’ he is not just requiring his baby to
refrain, when confronted with a certain request, from uttering tokens
of the same phonetic shape as ‘don’t’, but from
uttering any tokens that
mean
the same. (Goldstein 1984, p.
4)
Saka (1998, p. 124) concurs and claims that “premise” and
“premiss” in (38) pick out concepts:
(38)
The concept ‘premise’ is the same as the concept
‘premiss’.
Tsohatzidis (1998) claims that since T1 is true, even though Descartes
didn’t speak English, “is a thinking substance” in
T1 can’t refer to an English expression.
(T1) In one of the greatest philosophy books ever written in Latin,
Descartes said that man ‘is a thinking substance’.
These arguments all take the same form: first, they identify a
sentence \(S\) that we are inclined to interpret as true and
suggest that the only way to understand how \(S\) can be true is
to assume that quotations can refer to some kind of object \(O\).
This is then alleged to be evidence that quotations can be used to
refer to objects of kind \(O\).
If quotation has this kind of flexibility, the five theories discussed
above will all have to be evaluated with respect to whether they can
accommodate it. The Proper Name Theory, Description theory, and
Disquotational Theory, all seem to have particular difficulties in
this respect.
Not all, however, are convinced that quotations are flexible in just
this way. Some have expressed skepticism both about the form of
argument (see Cappelen and Lepore 1999a) and about the specific
examples. Cappelen and Lepore (1999a) also argue that the multiple
ambiguity view over-generates, i.e., it predicts that it is possible
to express propositions with quotation sentences that it is not
possible to express with such sentences. For further discussion, see
also Saka (2003).
6. Alternative Quotational Devices
A number of authors over the years have thought that our standard
practices of quotation are not suitable for all purposes and have, in
effect, introduced new technical devices. They cannot all be
summarized here, but here is a brief sketch of some such devices.
6.1 Token Quotes
Reichenbach (1947, p. 284) writes: ‘Whereas the ordinary-quotes
operation leads from a word to the name of that word, the token-quotes
operation leads from a token to a token denoted by that token’.
Reichenbach uses little arrows (\(\tokenquote{\,}\)) for token quotes,
so that the sign (39):
(39)
\(\tokenquote{a}\)
represents not a name for the token of ‘a’, but a token
for it. That is, the token in (39) cannot be repeated. (40), for
example, is not only a token different from (39) but refers to a
different token (Reichenbach 1947, pp. 285–86).
(40)
\(\tokenquote{a}\)
Token quotes then are much like writing a demonstrative expression
like ‘this’ and fastening to it an object to produce a
symbol of that object. (As noted above, some authors (Bennett 1988,
Saka 1998, Washington 1992) opine that this is how ordinary quotation
sometimes functions.)
6.2 Corner (Quasi) Quotation
As Quine (1940, §6) notes, the quotation:
‘\((\mu)\)’
designates only the specific expression therein depicted, containing a
specific Greek letter. In order to effect reference to the unspecified
expression he introduces a new notation of
corners
(namely,
‘\(\ulcorner\)’ and
‘\(\urcorner\)’). So, for example, if we take the
expression ‘Quine’ as \(\mu\), then
\(\ulcorner(\mu)\urcorner\) is ‘(Quine)’.
The quasi-quotation is synonymous with the following verbal
description: The result of writing ‘(’ and then \(\mu\) and
‘)’ (Quine 1940, p. 36). For an interesting suggestion
about quasi-quotation in natural language, see Saka (2017).
6.3 Dot Quotes
Someone who uses the word ‘red’ in speaking or thinking
would generally be held to be employing the same concept as a French
person who uses ‘rouge’. Assuming that ‘rouge’
is a good translation of ‘red’, Sellars (1963) thought it
convenient to have a general term by which to classify words that are
functional counterparts in this way. Such a term is provided by
Sellars’ dot quotes. Dot-quotes form a common noun true of items
in any language that play the role performed in our language by the
tokens exhibited between them, so any expression that is a functional
counterpart to ‘red’ can be described as a
red
. In Sellars’ terminology,
the concept
red
is something that is common and peculiar to
red
s.
6.4 Unambiguous Quotes
Consider (41):
(41)
‘a’ concatenated with ‘b’ is an
expression.
Michael Ernst claims this sentence is ambiguous. Read one way, (41)
means that the concatenation of the first letter of the Roman alphabet
with the second letter is an expression; read another way it means
that the expression between the two outer quote marks is an
expression. Confronted with this ambiguity, Boolos (1995) introduced a
notation for quotation in which every quotation mark ‘knows its
mate’. So understood there are denumerably many distinct
quotation marks, each formed by prefixing a natural number n of
strokes to a small circle, and in order to meaningfully enclose an
expression with Boolos quotation marks we have to choose quotation
marks of a ‘higher order’ than the quotation marks that
occur in the expression to be quoted (if there are any) and each
quotation mark in a grammatical sentence is to be ‘paired’
with the next identical quotation mark (Boolos 1995, p. 291).
Accordingly, reading (41) as containing Boolos quotation marks, we get
the more natural reading according to which it says that the first
letter of the Roman alphabet concatenated with the second is an
expression.
7. Formal-Material Modes
Carnap introduced in
Logical Syntax of Language
(1937) a
distinction between formal and material modes. The material mode is
generally used to describe the non-linguistic world; the formal mode
is generally used to discuss the language that is used to describe the
material world. Thus ‘one is a number’ is a sentence in
material mode; and ‘‘one’ is a number word’ is
its sentential counterpart in formal mode. This distinction is
mentioned in passing here since some authors have thought it
corresponds to the use/mention distinction. Since the use/mention
distinction isn’t particularly about quotation, there
isn’t much that needs to be said about the material/formal mode
here either. Revealing that a statement is basically about the use of
language may succeed in removing some of its metaphysical mystery. And
it may even be that semantic ascent (that is, the device of making a
sentence the topic, instead of what the sentence purports to refer to)
succeeds in demystifying much of what goes on in certain quarters of
philosophy. But as we’ve said, this distinction, and the various
philosophical moves surrounding it, doesn’t have much to do with
quotation per se.