Saka & Johnson - The Semantics and Pragmatics of Quotation - 2017

Perspectives in Pragmatics, Philosophy & Psychology 15
Paul Saka
Michael Johnson Editors
The Semantics
and Pragmatics
of Quotation
Perspectives in Pragmatics,
Philosophy & Psychology
Volume 15
Editor-in-Chief
Alessandro Capone, University of Messina, Italy
Consulting Editors
Keith Allan, Monash University, Melbourne, Australia
Louise Cummings, Nottingham Trent University, UK
Wayne A. Davis, Georgetown University, Washington, USA
Igor Douven, Paris-Sorbonne University, France
Yan Huang, University of Auckland, New Zealand
Istvan Kecskes, State University of New York at Albany, USA
Franco Lo Piparo, University of Palermo, Italy
Antonino Pennisi, University of Messina, Italy
Editorial Board Members

Noel Burton-Roberts, University of Newcastle, UK
Brian Butler, University of North Carolina, Asheville, USA
Felice Cimatti, Università della Calabria, Cosenza, Italy
Eros Corazza, Carleton University, Ottawa, Canada
Marcelo Dascal, Tel Aviv University, Israel
Michael Devitt, Graduate Center, City University of New York, USA
Frans van Eemeren, University of Amsterdam, The Netherlands
Alessandra Falzone, University of Messina, Italy
Neil Feit, State University of New York, Fredonia, USA
Alessandra Giorgi, Ca’ Foscari University of Venice, Italy
Larry Horn, Yale University, New Haven, USA
Klaus von Heusinger, University of Stuttgart, Germany
Katarzyna Jaszczolt, University of Cambridge, UK
Ferenc Kiefer, Hungarian Academy of Sciences, Budapest, Hungary
Kepa Korta, ILCLI, Donostia, Spain
Ernest Lepore, Rutgers University, New Brunswick, USA
Stephen C. Levinson, Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands
Fabrizio Macagno, New University of Lisbon, Portugal
Tullio De Mauro, ‘La Sapienza’ University, Rome, Italy
Jacob L. Mey, University of Southern Denmark, Odense, Denmark
Pietro Perconti, University of Messina, Italy
Francesca Piazza, University of Palermo, Italy
Roland Posner, Berlin Institute of Technology, Germany
Mark Richard, Harvard University, Cambridge, USA
Nathan Salmon, University of California, Santa Barbara, USA
Stephen R. Schiffer, New York University, USA
Michel Seymour, University of Montreal, Canada
Mandy Simons, Carnegie Mellon University, Pittsburgh, USA
Timothy Williamson, University of Oxford, UK
Anna Wierzbicka, Australian National University, Canberra, Australia
Dorota Zielińska, Jesuit University of Philosophy and Education Ignatianum, Kraków, Poland
More information about this series at http://www.springer.com/series/11797

Paul Saka • Michael Johnson
Editors
The Semantics and

Pragmatics of Quotation
123
Editors
Paul Saka Michael Johnson
University of Texas, University of Hong Kong
Rio Grande Pok Fu Lam, Hong Kong
Edinburg, TX, USA
ISSN 2214-3807 ISSN 2214-3815 (electronic)

Perspectives in Pragmatics, Philosophy & Psychology
ISBN 978-3-319-68746-9 ISBN 978-3-319-68747-6 (eBook)
https://doi.org/10.1007/978-3-319-68747-6
Library of Congress Control Number: 2017959052
© Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
In recent years the study of quotation has
seen fervent activity that has broadened and
deepened our understanding of the semantics
and pragmatics of reported language and
reflexive language. To this scholarship we are
pleased to add the articles commissioned
especially for this volume. For their original
contributions to the field we would like to
thank the individual authors, and for their
invaluable assistance in carrying out the
project we would like to thank John Biro,
Alessandro Capone, Julia Jorgensen, Steve
Wagner, and anonymous referees.
This volume is dedicated to the memory of
Dale Jacquette (1953–2016), colleague and
philosopher nonpareil.
Paul Saka
Michael Johnson
Contents
Part I Use & Mention

Scare-Quoting and Incorporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Mark McCullagh
Blah, blah, blah:
Quasi-quotation and Unquotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Paul Saka
Use-Mention Confusions in “Sloppy, Colloquial Speech” . . . . . . . . . . . . . . . . . . . 65
Marga Reimer
A Bridge from the Use-Mention Distinction to
Natural Language Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Shomir Wilson
Part II Quotation Unified

Unity in the Variety of Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Kirk Ludwig and Greg Ray
Semantics vs. Pragmatics in Impure Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Mario Gómez-Torrente
Reference and Reference-Fixing in Pure Quotation. . . . . . . . . . . . . . . . . . . . . . . . . . 169
Manuel García-Carpintero
Quotation in Dialogue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Eleni Gregoromichelaki
Part III New Directions

The Pragmatics of Attraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Emar Maier
vii
viii Contents
Quotation Through History: A Historical Case for the Proper

Treatment of Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Michael Johnson
Ideo- and Auto-reflexive Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Wayne A. Davis
Referential Analysis of Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
Dale Jacquette
Monsters and I: The Case of Mixed Quotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
Kasia M. Jaszczolt and Minyao Huang
Part I
Use & Mention
Scare-Quoting and Incorporation
Mark McCullagh
Abstract I explain a mechanism I call “incorporation,” that I think is at work in a

wide range of cases often put under the heading of “scare-quoting.” Incorporation is
flagging some words in one’s own utterance to indicate that they are to be interpreted
as if uttered by some other speaker in some other context, while supplying evidence
to one’s interpreter enabling them to identify that other speaker and context.
This mechanism gives us a way to use others’ vocabularies and contexts, thereby
extending our expressive capacities on the fly.
Explaining incorporation involves explaining intra-sentential shifts in lexicon
and in context. Shifts of the former sort are familiar to linguists under the heading
of “code-switching.” Shifts of the latter sort have been less explored; accordingly I
explain how to modify Kaplan’s logic of demonstratives to allow for such shifts.
I compare the incorporation account of scare-quoting with accounts offered by
Brandom, Recanati, Geurts and Maier, Benbaji, Predelli, and Shan. Finally I note a
possible implication concerning the speech act of assertion: that you can properly
assert a content you do not believe, let alone know, because part of it is expressed
with words you do not understand.
1 Introduction
My aim in this paper is to explain a mechanism that I think is at work in a wide range
of cases often put under the heading of “scare-quoting.” I call it incorporation.1
Incorporation is flagging some words in one’s own utterance to indicate that they
1
By using this term I do not mean to pick up on what Herbert Clark and Richard Gerrig (1990)
mean by “incorporated quotation.” By “incorporation” they mean a relation that can obtain between
the playing of some bit of music, and the rhythm thereby demonstrated to a listener (789). (That
is the instance with which they introduce the notion.) As will become clear, the relation I have in
mind is very different. (Clark’s and Gerrig’s concerns are largely orthogonal to mine, since their
M. McCullagh ()
University of Guelph, Guelph, Canada
e-mail: mmcculla@uoguelph.ca
© Springer International Publishing AG 2017 3

P. Saka, M. Johnson (eds.), The Semantics and Pragmatics of Quotation,
Perspectives in Pragmatics, Philosophy & Psychology 15,
https://doi.org/10.1007/978-3-319-68747-6_1
4 M. McCullagh
are to be interpreted as if uttered by some other speaker in some other context,

while supplying evidence to one’s interpreter enabling them to identify that other
speaker and context. This mechanism gives us a way to use others’ vocabularies
and contexts. It can be useful not only for purposes of rhetorical vividness but also
in cases where a speaker’s vocabulary lacks some words that express a content he
wishes to express: it enables him to extend his own expressive capacity on the fly.
I’ll start by presenting some examples and remarking on terminology; then I’ll
explain the mechanism and show how it applies to the examples. Finally I’ll contrast
this account with others in the literature. Various writers have proposed accounts of
the sort of data I discuss but they err, I will argue, in trying to assimilate the cases
either to use or to mention. Incorporation is neither of these.
2 Some Data
Discussions of “scare quotation” and of “mixed quotation” are vexed by differences

over the range of cases those terms cover. I’ll discuss the terminological issues after
putting my own range of cases on the table, as my main interest is in explaining the
one mechanism that is at work, I think, in these cases regardless of what headings
different writers would put them under.
I will follow the lead of Stefano Predelli’s (2003) discussion of scare quotes, and
look into what various style and usage manuals say about certain uses of quotation
marks. Predelli cites four usage guides (Perrin and Smith 1968, Kierzek 1977,
Chicago 1982, Ehrlich and Murphy 1991); my excuse for looking at a few additional
ones is that there are some types of use described in them that aren’t described in any
of these four. Another interesting point is that in a couple of cases, the changes over
different editions suggest an increase in the use of the described uses of quotation
marks, or at least a greater acceptance of them.
Here are some uses of quotation marks that the manuals describe and accompany
with examples. With the exception of those from Greenbaum (1996) (which are from
the International Corpus of English), as far as I know the examples that the manuals
give are all unattested. I’ll give under each heading one or more actual examples
from various sources. I should emphasize that I don’t construe these headings as
exclusive; several examples fit under multiple headings. But it is impressive how
much convergence there is: most of the functions listed are described as legitimate
in multiple style manuals.
1. Indicating that the enclosed words are slang (Chicago 1982: 173, Baker 1984:
407, Todd and Hancock 1986: 460, Greenbaum 1996: 548, Birchfield 2004: 646,
Chicago 2010: §7.57 actually deprecates the use of quotation marks to flank slang
expressions, saying they are “rarely needed”).
paper almost entirely concerns the speech-act classification of various quotational or quotation-like
performances rather than a systematic account of their semantics.)
Scare-Quoting and Incorporation 5
Bro Hug In a series of photographs last May, BuzzFeed cruelly broke down
President Obama’s botched farewell hug to his departing press secretary, Jay
Carney, calling it “The Most Awkward Hug in White House History.” Mr.
Carney’s “weird flappy hand” gave way to the president’s “Raptor hand,” which
caused Mr. Carney seemingly to kiss the president’s shoulder, after which the
president “bro yanked” Mr. Carney’s head and gave him a flash of “side eye.”
(New York Times, September 26, 2014)
2. Indicating that the enclosed words are technical jargon (Birchfield 2004: 646,
Stilman 2010: 185 (“enclosing it in quotation marks sends a reassuring signal
to your readers that they aren’t expected to have prior acquaintance with it”);
McCarten 2013: 392, Chicago 1982 says that one use is to flank “technical terms
in special senses.”) (The (unattested) example Chicago gives is of a technical
term used incorrectly, in a sort of metaphorical extension. This is different from
the other manuals that describe technical terms being used with their technical
meanings, as in the following examples.
Aiviq Between 5:34 a.m. and 11:29 a.m., according to a later computer analysis
by Rolls-Royce, the Aiviq’s “wire tensile strength overload” alarm went off 38
times. (New York Times, December 30, 2014, describing a mishap on an offshore
oil-drilling platform)
Tardigrade The tardigrade, which has a natural life span of about a year, is
particularly impressive among the exhibit’s “extremeophiles.” (New York Times,
September 8, 2015)
Chassis At a cabinet meeting led by Vice President Joe Biden, [Energy
Secretary] Chu unveiled thirteen electrofuels grants. A Boston firm planned to
engineer E. coli into “a chassis for iso-octane.” A University of South Carolina
team would study “electroalcoholgenesisbioelectrochemical reduction of CO2 ”
(Grunwald 2012: 5)
Bugs We discovered “bugs in the back end” of live updates. (Twitter post from
a public relations officer, explaining website malfunction. https://twitter.com/
femwriter/status/562586794815156224)
3. Indicating that the writer somehow rejects the use of the enclosed words (Todd
and Hancock 1986: 460 (“Quotation marks may also be used for a word that
a writer disagrees with”), Greenbaum 1996: 548 (“the author does not accept
responsibility for the wording”), Chicago 2010: §7.55 (“Nicknamed scare quotes,
they imply, ‘This is not my term’ or ‘This is not how the term is usually
applied’”), Stilman 2010: 187, McCarten 2013: 392 (“Put quotation marks
around words used ironically”)).
Good Old Days In the “good old days” our great-great-grandmothers walked
several miles to the village. (Greenbaum 1996: 548)
Chomsky The hideous slave labor camps of the new “empire of liberty” were
a primary source for the wealth and privilege of American society, as well as
England and the continent. (Noam Chomsky in New York Times, March 18,
2015)
6 M. McCullagh
Lizard As 12 million Americans “know,” the United States government is run

by lizard people. (The Atlantic, October 31, 2013)
Krugman [R]ight-wing think tanks are still cranking out “studies” purporting
to show that health reform is a failure. (Paul Krugman in New York Times,
August 10, 2015)
Murderer Cross dressing “murderer” moves into Harlem townhouse (New York
Post headline, October 24, 2011)
4. Indicating that the enclosed words are someone else’s choice (McCarten 2013:
392 (“Use quotation marks to set off a pungent or significant word or phrase : : :
justifiable only when the words are controversial, add colour or give the flavour
of an event or the style of a speaker”)).
Denikin After Denikin triumphantly entered Tsaritsyn and attended services in
its Orthodox cathedral, on July 3, he “ordered our armed forces to advance on
Moscow.” (Kotkin 2014: 326)
Bush After a run through of “ideas I strongly reject,” Bush finally got around
to announcing that he was going to “talk about what we’re for.” (New York
Times, March 15, 2008)
Lenin Lenin did speak at the Fourth Comintern Congress, for an hour, in
German, but he was drenched in perspiration and told people that during the
speech he had “forgot what he had already said, and what he still had to say.”
(Kotkin 2014: 418)
5. Indicating that the enclosed words are from some common background, e.g. a
cliché or a canonical text (Chicago 1982 endorses this use but the 2010 edition
deprecates it).
Axe While the initial news of remission is often met with relief, many patients
realize quickly that the good news comes with some grave caveats as they
“wait for the axe to fall.” (New York Times, September 10, 2015)
Militia Another fine day for America’s “well regulated militia”. (John Gruber’s
blog, after a mass shooting in the U.S. http://daringfireball.net/linked/2015/
08/26/virginia-tv-shooting)
6. Indicating that the enclosed words are someone else’s (Gibaldi 1998: 77 (“Place
quotation marks around a word or phrase given in someone else’s sense or in a
special sense or purposefully misused”), Chicago 1982: 173 (indicating that the
expression they enclose is “foreign to the normal vocabulary of the speaker”),
Swan 2005: 240 (“[in headlines,] Quotation marks : : : are used to show that
words were said by somebody else, and that the report does not necessarily claim
that they are true”)).
I don’t give examples under this last functional heading because I believe that
it covers all of the examples given under previous headings. Indeed, my proposal
builds on this core idea. My suggestion will be that in all the examples I gave there
is one semantic mechanism at work. The multiplicity of uses listed in the manuals is
simply due to the fact that there are different reasons that writers can have for using
that one mechanism, and different suggestions they thereby make to their audience
in different cases. The mechanism, of course, is incorporation.
2.1 Terminological Digression
I think that the term “scare quoting” does a decent job covering the phenomenon in
question. I say this with some trepidation since I doubt that that term (or the term
“shudder quotes”) has an agreed meaning even among philosophers and linguists;
probably any use of it to demarcate some field of data, even preliminarily, will
be contestable. (As Paul Saka (2013) notes, the literature on quotation has been
bedeviled by the difficulty of demarcating different subfields of data.) I have found
that some colleagues, perhaps leaning on the word “scare,” take it to be a necessary
condition that there is some criticism, perhaps only implicit, of the words flanked
by quotation marks (or dancing fingers). Others don’t; they would include under the
heading many cases in which nothing like that is being indicated. (When I’ve asked
those people in the former group what label they’d apply in these cases, they’ve
been at a loss.) To some, perhaps, the term is just a catch-all for uses of quotation
marks that are evidently not of one of the kinds long recognized by philosophers
and linguists.
There is also the phenomenon now known as mixed quotation, which I mention
without taking a stand here on whether it really is a distinct phenomenon from
scare-quoting. As far as I know this terminology is due to Herman Cappelen and
Ernie Lepore (1997) who wrote, “The category of mixed quotation is more or less
ignored in the literature” (429n.1), although they do credit Davidson (1979: 81) for
acknowledging it. The working example they use is
Alice said that life “is difficult to understand”
in which we have a reporting verb (“said”) and a “that”-clause as its complement,

part of which is flanked by quote marks.
Davidson did write of a “mix of use and mention” (not “mixed quotation”) in his
1979 paper, but it’s unclear whether by that he means what Cappelen and Lepore
mean by “mixed quotation,” since Davidson included as examples cases in which
no speech act is being reported, e.g.
Dhaulagiri is adjacent to Anapurna, the mountain whose conquest Maurice Herzog
described in his book of the same name. (81)
(I’m not sure that any of the theories I’ll discuss below would apply to examples
like this.) For Cappelen and Lepore, on the other hand, it is essential to “mixed
quotation” that a speech act is being reported:
a basic fact about mixed quotation : : : [is that it] contains a component that serves two
functions concurrently. The quoted part is both employed to report what the speaker said
and it is employed to say, at least partially, what the speaker actually uttered. (436)
8 M. McCullagh
Furthermore they take it that every mixed quotation stands in some entailment
relation with an indirect quotation (1997: 430). What this suggests is that by “mixed
quotation” they, at least, mean a mix of direct and indirect quotation rather than the
sort of mix—of use and mention—that Davidson described.2
It may be the case that the right thing to say about scare quotation is also the
right thing to say about mixed quotation. I won’t take up that question in this paper.
But I will register commitment to a conditional claim: that if mixed quotations are
partly about words—that is, if there is a term in their proper parsing whose semantic
value is the same as that of some pure quotation—then it is a different phenomenon
from scare-quoting. The reason I say this is that on the account I’ll propose, scare-
quotings are not about words (in the sense just described). So for me, the possibility
of an equivalence comes down to the question of whether the mixed quotation about
Alice says or merely suggests that she uttered “is difficult to understand” (in the
course of making the reported statement).3 And evaluating that claim would be too
large a digression from the present topic. (For discussion see Recanati 2001: 659–
61; Cappelen and Lepore 2005: 66; Recanati 2010: 290–302; De Brabanter 2013:
130–32.)
So ends my terminological digression.
It’s likely that scare-quoting is becoming more common. The first edition of
Fowler’s guide (1926) makes no mention of the practice that the 3rd edition (2004)
blandly describes: “Quotation marks may be used to enclose slang and technical
terms” (646). And the 13th edition of the Chicago Manual of Style (1982) doesn’t
even use the term “scare quotes” while the 16th edition (2010) uses it to head an
entire section on their use. Perhaps with the increased access that the World Wide
2
On this terminological point there is variance even among some authors I discuss below. For
example, Geurts and Maier (2005: 109) and Shan (2010: 418) clearly mean the latter while in
addition to Cappelen and Lepore, Recanati (2010: 215) clearly means the former. In line with
this conception, I will take it that a necessary condition for being mixed quotation is that a
reporting verb is being applied to a “that”-clause, part of which is quotation-mark-flanked. (This is
a terminological stipulation I’m making just for clarity’s sake. The following case seems evidently
a reporting use of a quoted phrase even though there is neither a reporting verb nor a “that”-clause:
After [Ian Fleming’s] first heart attack in 1961, he wrote to Christopher Soames, Britain’s
minister of agriculture, hoping to determine “the finest liquor obtainable in England,” now
that he was condemned to a mere three ounces a day. (New York Times, November 24, 2015)
Other such examples are given in De Brabanter (2005: 5).) I think it is clearly not a sufficient
condition, since if there is such a thing as scare-quoting then—as Cappelen and Lepore note (2005:
73n.7)—there should not be any barrier to its being done in part of a “that”-clause complement to
a reporting verb, in which case the resulting sentence does fit the description just given. (One of
the examples I presented, Lenin, illustrates exactly this point.)
3
The parenthetical qualification is important and easy to neglect: as Geurts and Maier (2005)
(discussed below) point out, Potts’s (2007) account of mixed quotation fails on this score inasmuch
as it understands mixed quotations as mere conjunctions of direct and indirect. (Gomez-Torrente
(2017:135) gets the relation right, offering an account on which one of the conventional uses of
quotation marks is to “add an adverbial or prepositional indication” rather than a conjunct.)
Web gives us to different jargons and vocabularies, scare-quoting has more use
nowadays than before.
In addition to showing the variety of uses writers make of scare-quoting there
are two broader points that even a brief survey such as this makes. One is that we
interpret scare quotes very easily, even when they involve indexicals—I doubt that
anyone reading any of the examples was brought up short. (Indeed it’s possible to
misinterpret Lenin as a case of mixed quotation—but I doubt that anyone did that!)
Another is that scare-quoting is a very common device among educated writers.
This distinguishes it from free indirect discourse, which is used primarily in literary
texts (discussed by Johnson 2017), and from the much-mocked uses of quotation
marks to indicate emphasis (Keeley 2010, discussed in Gomez-Torrente 2017, this
volume). Scare-quoting is part of most contemporary writers’ toolkits. Despite what
I’ll argue is some non-trivial machinery under the hood, we take scare-quoting in
stride. We should want our account of it to reflect this. The account I’ll offer does
not involve crediting speakers or interpreters with any special concepts that they
need to bring to bear on utterances involving scare-quotes.
3 What Incorporation Is
I said that incorporation occurs when one speaks while indicating, of some words
in one’s sentence, that they are to be interpreted as if used by another. This affects
how an interpreter is to carry out two tasks.
Any interpreter, of any utterance, must assign meanings to its constituent expres-
sions. She must also select some context in relation to which she interprets any
of the utterances’ context-sensitive expressions. In the typical case the interpreter
works with a lexicon that they believe states the meanings those expressions have
in the speaker’s idiolect. Performing the first task, then, is selecting entries from
this lexicon. Also in the typical case, it is obvious who the speaker is, when they
speak, and where. Performing the second task, then, is choosing these values for
the parameters used in evaluating whatever context-sensitive meanings are selected
from the lexicon.
Things go differently in cases of incorporation from how they go in the typical
case. The two tasks just described must be re-done for the incorporated words.
The interpreter has to shift to a different lexicon: one that states the meanings that
those words have as used by that other speaker. She must also evaluate relative to a
different context: that of the other speaker. And she must shift back, once she has
finished interpreting the incorporated words.
Suppose then that a speaker A utters a sentence S in context cA while indicating,
of some part of S, that it is being incorporated. This means that A is indicating to an
interpreter that that part is to be treated as if spoken by another. Suppose also that
the evidence available to the interpreter suffices for her to identify speaker B as that
other speaker and cB as the context in which items from B’s lexicon are to be taken.
10 M. McCullagh
(Note that cB is not thereby identified as a context in which B utters those words.
Whether that’s the case makes no difference to the interpretative procedure.)
The interpreter must assign meanings to the words in A’s utterance. So in each
case she asks: is this word part of the incorporated part, or not? If it is, then the
interpreter looks up the word in the lexicon she has for B; if it isn’t, then the
interpreter looks it up in the lexicon that she has for A.
The interpreter must also interpret words relative to some context of utterance.
In each case she asks: is this word part of the incorporated part, or not? If it is, then
she interprets the word relative to context cB ; if it isn’t, then she interprets the word
relative to context cA .
These two tasks are, in Kaplan’s terminology, pre-semantic. “[G]iven an utter-
ance, semantics cannot tell us what expression was uttered or what language it
was uttered in. This is a pre-semantic task” (1977: 559). On Kaplan’s conception,
semantics is the business of calculating, on the basis of whatever initial values
(including a parsing of the utterance) are required, a truth value (or condition) for
an utterance; pre-semantics is the business of selecting those initial values.4
Clearly there is nothing quotational, in the traditional sense, about incorporation:
for neither the demarcated words nor the words plus the demarcating marks are
interpreted as denoting linguistic expressions (whether types or tokens), or even
characters.5
Like any mechanism, incorporation can fail. The second supposition that I made
in setting up our schematic case is non-trivial: it is possible for A to fail to ensure
that the interpreter’s evidence does suffice to identify a speaker and context for the
incorporated part. This leaves the interpreter unable to interpret that part. But of
course the possibility of this sort of failure is not unique to incorporation. It is a
possibility that accompanies any use of language. One can hear some words uttered
but not know what language they’re in; or who is uttering them; or where the speaker
is. In such a case one doesn’t know enough about the utterance to be justified in
selecting any particular lexicon (the first case) or giving any particular values to the
contextual parameters (the other cases). Although typically easy, pre-semantics can
be difficult, and it can fail.
4
Setting things up in this way reveals a possibility that I think is not acknowledged in Michael
Johnson’s essay (this volume), where he infers from the claim that quotation marks “don’t
articulate linguistic items” to the claim that they “are irrelevant for assigning semantic contents
to sentences” (289). This inference ignores the role that context plays in determining content:
scare-quotes indicate a shift in context and thereby affect content, even if they aren’t “linguistic
items.” In Kaplan’s terminology the point is that whatever affects pre-semantics consequently
affects semantics, even if it isn’t itself semantic.
5
I say “in the traditional sense” because, as Paul Saka (2013) notes, “there is no consensus on
defining quotation” (940). On his own view, for example, the function of quotation is simply to
“rule out customary reference as the intended interpretation” of the quoted words (1998: 127). In
Saka’s terminology my point would be that there is nothing citational about incorporation.
The second supposition also embodies an idealization, since what is more

common is that the interpreter will be able to identify a type of person or a type
of context. (Several of our examples will be handled in this way.) This is fine if
it makes no difference to the resulting interpretation which particular person, or
context, of that type is selected.
I will discuss this identification task in the course of treating the examples. In
almost all cases it’s completely obvious what the scare-quoted speaker/context is,
which just goes to illustrate how in successful uses of scare-quotes there is more than
enough evidence for the interpreter to succeed in the identification task. But I want
to be explicit that the identification task is a separate one from that of determining a
truth condition. Undertaking the latter task presupposes success in the former. (Thus
on my account it cannot be the case that success in the identification task requires
prior success in the calculation of a truth condition.)
3.1 The Two Shifts That Incorporation Involves
Notably, each of the two sorts of intra-sentential interpretative shifts that incor-
poration demands—in lexicon and in context—is already exemplified separately
in different phenomena; so they already require explanation, no matter what our
view of incorporation or scare-quoting. Intra-sentential lexicon shifts occur in
what linguists have investigated under the heading of “code-switching,” wherein
bilinguals change from one language to another mid-sentence. And intra-sentential
context shifts occur in some very simple cases involving indexicals, e.g. a speaker
using “here” multiple times in a sentence while changing positions during its
utterance. Cases of these two sorts are not cases of incorporation. But the right
stories about how they work combine into a story about how incorporation works.
Code-switching is a phenomenon well-studied by linguists (Poplack 1980;
Poplack and Sankoff 1984 are two important early articles in this literature.)
Bilinguals produce and interpret utterances of this sort with ease. In these cases,
of course, the interpreter is not switching from one speaker’s lexicon to another
speaker’s lexicon, since the one speaker speaks both of the two public languages
involved. But it’s the sort of shift that would be involved in interpreting a non-
bilingual who indicated, of some of his words which are not in the public language
he speaks, that they were to be interpreted as if spoken by another who does speak
a public language that has entries for those words. From the interpreter’s point of
view, the shift is of the same kind in both cases: she consults her lexicon for Spanish
(say) when interpreting part of the speaker’s utterance and that for English when
interpreting the rest of it.
The second sort of shift is one the very idea of which requires some elaboration.
The basic phenomenon of intra-sentential shift in context is acknowledged, espe-
cially in relation to quantifiers (see Stanley and Williamson 1995 and the works
cited therein in n. 4.). But there is no generally accepted formal treatment of it. In
the next section I will explain one treatment—based on a simple modification of
12 M. McCullagh
David Kaplan’s logic of demonstratives—that I think works nicely. Before getting

to that I need to explain what I mean by “indicating,” when I used that notion in
explaining incorporation.
3.2 “Indicating”
When I say that the speaker “indicates” some other speaker and context, it’s
important that by “indicating” I don’t mean describing. The indicating I mean is a
matter of giving evidence to the interpreter to help her with the two tasks of selecting
a lexicon and selecting a context. In the typical case, such evidence is manifest and
ample: the interpreter knows who the speaker is—therefore which lexicon to use—
and where he is, the time of his utterance, and so on—therefore which values to use
for the contextual parameters. But even in cases not involving anything like scare-
quoting, there are ways to indicate that the latter should be changed mid-sentence.
Probably the simplest example concerns the location contextual parameter. There
is a very simple way in which a speaker can often indicate that the interpreter
should change what value she assigns that parameter. It is simply to change his
location. The speaker doesn’t say that he’s changing location; nor does he describe
another location; nor is there a “change parameter value” node in the parsing of
the sentence. He just changes location; doing so gives evidence that—“indicates”
that—his location has changed.
I said “often” because a change in the speaker’s position doesn’t always call
for a change in the value of the parameter with respect to which an interpreter
evaluates position-sensitive expressions. One sometimes moves while one speaks,
and time always passes no matter how quickly one speaks, but in most cases this is
irrelevant to interpretation. In such cases there’s no harm in an interpreter’s leaving
unchanged the values she assigns to the contextual parameters even though there are
changes in the real-utterance features they are meant to reflect. Thus there is nothing
surprising in the fact that the leading theory of context-dependence, Kaplan’s, works
with sentence-context pairs, the idea being that the entire sentence is evaluated
when taken in one context. But in other cases such changes are highly relevant and
matter to truth conditions. Suppose that a speaker changes position—from near the
fireplace to near the window—while uttering the following sentence:
Two Heres It’s warm here, but here it’s quite a bit cooler.
In such a case each occurrence of “here” is to be interpreted as denoting a
different place. (Prima facie, such cases challenge the Kaplan approach; I’ll discuss
this issue below.)
Another type of example is of what Ruth Kempson et al. call “shared utterances,”
in which one speaker completes an utterance begun by another (Kempson et al.
2001; Purver and Kempson 2004). In some such cases we will have two occurrences
of “I” within one (shared) utterance, that evidently denote different people. Thus
the situation evidently demands of an interpreter that she change the value for
the speaker parameter mid-way through her interpretation of the shared utterance.
Neither speaker said that the change is required, but the overall situation evidently
indicates it.
The indicating that incorporation involves is of the same general sort as in
Two Heres or in shared utterances. What incorporation requires is that the speaker
somehow indicate that a mid-utterance re-doing of two crucial pre-semantic tasks
is called for, and indicate also how the re-doing should be done: whose lexicon
should be used, and which context should be used in evaluating the indicated part
of the sentence. Success in incorporation requires that the interpreter have enough
evidence to do as the speaker intends. The evidence available to the interpreter—the
presence of which is part of what I am counting as the speaker’s “indicating”—
should suffice to identify the relevant other speaker and context. The evidence
could be background knowledge, knowledge about what has happened already in
the conversation, or knowledge about the statement itself.
4 Intrasentential Change in Context
I said that the very idea of intra-sentential shift in context requires some defence;
in this section I supply that defence, by explaining how to implement such shifts
in a simple semantic system and by meeting objections to the very idea of such
shifts. Since David Kaplan’s (1977, 1989) work on demonstratives is a touchstone,
I will work within his logic of demonstratives (henceforth, LD). I’ll explain how
intra-sentential changes in context might be implemented by modifications to LD.
4.1 Modifying LD
In the definition of truth in LD that Kaplan gives, truth is assigned to the language’s
well-formed expressions “in” structures, “when taken in” contexts, “under” assign-
ments of values to variables, and “with respect to” times and worlds (1977, 544). So
in a case of a given calculation in that system, the pre-semantic tasks are to decide on
a structure, a context, a variable assignment (irrelevant except for open sentences),
a time and a world. With world and time left unspecified, what we get is a truth
condition rather than a truth value.
What intra-sentential context shifts could we handle in a modification of that
language? One might think that we would have to require that only well-formed
expressions can be context-shifted. Under that restriction, we could have
Fab _ “Gcd ”
as a permissible case (where quotation marks flank the part to be interpreted in

a different context) but not
14 M. McCullagh
Fab _ G “cd ”
since the latter marks only a sequence of term expressions, which isn’t itself a well-
formed expression. But in fact we won’t need any such restriction; context-shifting
of non-constituents (in this sense) is not a problem. We can introduce the approach
by considering one of the simplest clauses in the definition of truth in LD, the one
for conjunction. Here is what Kaplan gives6 :
4.i/ ˆcw . ^ / iff ˆcw and ˆcw
In English: a conjunction is true at a hcontext, worldi iff each of its conjuncts is true
at that hcontext, worldi. Here is what we write instead, in order to accommodate the
possibility of one conjunct being taken in a different context from the other:
4.i/ ˆhc1 ;c2 iw . ^ / iff ˆc1 w and ˆc2 w
In English: a conjunction is true at a hhcontext1 , context2 i, worldi iff its first conjunct
is true at hcontext1 , worldi and its second is true at hcontext2 , worldi.
The change to Kaplan’s scheme that we are introducing is that contexts can be
complex. Let C be the set of contexts in Kaplan’s system LD. Let C* be the closure
of C under ordered n-tupling. C* comprises anything that can be built by ordered
n-tupling out of the raw materials: the original—henceforth simple—LD contexts.
Consider now the clause for atomic sentences. In LD it is:
2: ˆcw ˛1 : : : ˛n iff hj˛1 jcw : : : j˛jcw i 2 I .!/
We modify this to allow each term to be taken in its own context. (We preserve
the idiosyncratic feature of Kaplan’s system, that no predicate letter can be context-
sensitive.)
2 : ˆhc1 ;:::;cn iw ˛1 : : : ˛n iff hj˛1 jc1 w ; : : : ; j˛n jcn wi 2 I .w/
Other clauses are handled similarly; call the resulting system LD*.7 Let’s illustrate
now, using the example I mentioned earlier:
Fab _ G “cd ”
6
I simplify notation somewhat. Where Kaplan has subscripts for context, variable assignment, time
and world, I have one for context and one for world. I also use ’n schematic variables for terms,
glossing over Kaplan’s distinction between “position terms” and “individual terms” (542).
7
Full details are in another paper that I am preparing for publication. This approach to handling
intra-sentential changes in context is significantly different from that proposed by David Braun
(1996) and from the more recent proposal of Geoff Georgi (2015). But this is not the place to get
into such details.
For this sentence we need a context whose structure is hh_,_i, h_,_ii, since (as
mentioned above) there is no context-sensitivity of predicate letters in LD. Since the
arguments to G are marked for context-shifting, and no other shifting is indicated,
we will have something like hhc1 , c1 i, hc2 , c2 ii where c1 ¤ c2 . “Fab” comes out true
(at some world-time pair) if “a” and “b,” when each taken in context c1 , have values
(at that same pair) that are in the extension of the interpretation of “F.” Similarly
for “G‘cd’” except that the two terms are interpreted when each is taken in context
c2 . So even though the part marked for context-shifting—“cd”—isn’t itself a well
formed expression, the evaluation goes through.
4.2 Too un-Kaplanian?
One might wonder whether we have gone to the other extreme from Kaplan. In
his LD each well-formed expression is evaluated relative to one simple context. In
our modified system there is no upper bound to the complexity of contexts. This is
necessary for the system to work, since there is no upper bound to the complexity
of well-formed expressions in LD, and complexity in the latter demands the same in
the former, if a calculation is to be made. Perhaps all this seems unduly complicated.
Although this is not the place for a full investigation, there are two comments
worth making. First, it is easy to model the cases in Kaplan’s system as special
cases of those handled in the modified system. For typically we will want all of
the simple contexts involved in the evaluation of an expression to be the same
simple context. In that case the complexity is otiose: the evaluation according to
the modified rules gives the same result as that according to the rules of LD. That
is, if c is the only simple context out of which the complex context c* is built, then
each simple component of an expression taken at c* will itself be taken at c, which
is the same as what happens by design in Kaplan’s system when c is the one context
in which the whole expression is taken. Second, it must be acknowledged that there
is one new implication of working with complex contexts. In LD, every context is
suitable for the evaluation of every well-formed expression. This is not the case in
the modified system. For example, the complex context hc1 , c2 i is unsuitable for the
evaluation of a simple expression such as “I” or “today.” (This is so regardless of
whether c1 or c2 is itself complex.) Aside from the mismatch in complexity there is
another problem. For “I” requires that the context supply a time. Now, even if each
of c1 and c2 supplies a time, which time would an ordered pairing of them supply?
Although it may be possible to define some notion of the time of a complex context,
it seems better just to require that simple expressions be evaluated when taken in
simple contexts. Here is how we do this in the modified system. Consider the clause
for “I” that Kaplan has:
12: jIjcw D cA
16 M. McCullagh
We reflect the unsuitability of hc1 , c2 i for evaluating “I” by putting the following
clause into our modified system.
12 : If c 2 C; thenW jIjcw D cA
That is, the value of “I” is defined only when it is being taken in a simple context.
So working with complex contexts means giving up on one feature of Kaplan’s
LD: that every context is suitable for the interpretation of every expression. But it
seems a small cost, and I cannot see that there is any great truth about language
or interpretation that entails that every expression can be evaluated when taken in
every context. What we might want to endorse is a much weaker principle, that
every expression can be evaluated when taken in any context in which it is uttered.
(It’s worth noting that LD* does preserve the following principle: that every simple
context is suitable for the evaluation of every simple expression in the language. I
note this just to point out that it’s there for those who like that sort of feature of
Kaplan’s system, not because I think that preservation of this particular principle is
a criterion of adequacy for a semantic treatment of context-sensitivity.)
The application of this modified system to Two Heres and the like is obvious:
each “here” is taken in its own simple context, each of which supplies a position
value that slightly differs from that of the context used in the evaluation of the
next occurrence of “here.” Thus we get the intuitively correct truth condition for
the utterance, which Kaplan’s original system cannot give. Similarly for analogous
cases involving multiple occurrences of “now.”
And at this point we notice an interesting thing, which makes possible incor-
poration in all its aspects. For although we introduced the need for a revision to
Kaplan’s system with a very ordinary example (Two Heres), in fact the revised
system is ready to handle cases that are much less ordinary: cases in which the
contextual feature that requires shifting is the speaker. As far as formal semantics
is concerned, the speaker parameter is simply one more parameter, completely on
a par with the time and location parameters. But while everyone would agree that
shifts in location or time, during the course of an utterance, often occur and are
sometimes semantically relevant, many will be less ready to agree that it’s even
possible, let alone semantically relevant, for a change in speaker to occur during the
course of an utterance.
But in the shared utterances cases mentioned above we already have instances
in which the speaker does change during the course of the utterance. They’re not
cases of incorporation, as I have defined it, since there is no one speaker of the
overall sentence. But they are cases in which the interpreter should, in interpreting
first person indexicals, change part-way through their interpretation of the sentence
the value that they assign to the speaker parameter.
Cases of incorporation are handled in exactly the same way. Where they differ
from the shared utterance cases is that rather than the speaker actually changing,
the speaker indicates that an interpreter is to treat part of the sentence as if it were
produced by a different (indicated) speaker. This indicating is, again, something that
matters to the pre-semantics, not something that changes how the calculation goes
once the initial values are supplied to it.
5 Intrasentential Shift in Lexicon
I’ve dwelt just now on the nature of one of the two interpretative shifts called for
by incorporation. What about the other shift, a shift in lexicon? This is much less
problematic a notion than intra-sentential context shift. In the Kaplanian setting, a
“shift” in lexicon is a shift to a different set of character-specifying lexical axioms
in the formal system.
Suppose that (in our schematic case) A scare-quotes some word, indicating that
it is to be treated as if used by B, and that the word does exist in both A’s and B’s
lexicons but with different meanings. Then the formal system will have two axioms
for that word, one stating the character it has when used by A, the other the one it
has when used by B. “Shifting” to B’s lexicon just means, in such a case, using the
latter axiom rather than the former in a calculation of the utterance’s truth condition.
The word “shift” is slightly misleading as a word for the general case, since we
also include cases in which the scare-quoted expression has no meaning when used
by A: it is not in A’s lexicon. Formally the upshot is the same: the axiom for its
meaning when used by the indicated speaker, B, is used in the calculation. Since
there is no axiom giving its meaning as used by A, there is nothing shifted from.
But this small misleading implication is not problematic in itself—the account still
works in these cases, a calculation can be done—and it is nice to have one term,
“shift,” to cover what is happening both to lexicon and context. So I will continue
to use it.
6 Applying the Account
With accounts of intra-sentential context- and lexicon-shifting in hand, we are ready

to apply the incorporation account to our examples. I’ll select a few that (in my
judgment) bring out challenging or interesting features of the account.
6.1 Chomsky
In this example the phrase “empire of liberty” is scare-quoted. Had the sentence
occurred just on its own, a reader wouldn’t know enough to identify a particular
speaker (or type of speaker) as the other—the B, in our schematic treatment at
the start of §3—whose use is being indicated; the identification task would fail.
But the topic of the larger discourse is the early United States. So an interpreter
18 M. McCullagh
has evidence that the other (whose context and lexicon are being recruited in his
sentence) is someone whose lexical entries are such that the early United States is
in the extension of the phrase “empire of liberty.”8 Interpreting the phrase this way
gives the correct truth condition for the sentence as a whole—again, since it is the
early United States that Chomsky is discussing.
Obviously there is more going on in Chomsky’s scare-quoting than just that it
is being used to get his sentence to have a certain truth condition. (He could, after
all, have achieved that just by using the phrase “the early United States” normally.)
The reader obviously is meant to infer from Chomsky’s highlighting its reliance on
slave labour, that he would not include the early United States in the extension of
the phrase “empire of liberty.” But there are two important points to make about this
fact.
First, it is not a fact that bears on the utterance’s truth condition. For interpreting
the phrase according to Chomsky’s meaning gives the wrong truth condition. On
that interpretation his utterance cannot be about the sources of wealth of the early
United States—which is, again, evidently what he is discussing. For the semantic
task—that of ascertaining the utterance’s truth condition—it is only the other’s
lexical entries (and context) that matter.
Second: in doing the identification task an interpreter doesn’t need any informa-
tion about how “empire of liberty” stands in Chomsky’s lexicon, let alone that he
wouldn’t use it to characterize the early US. All an interpreter needs is the evidence
(in previous sentences) that the topic is the early US. That is enough to carry out
the identification task: it is evidence that the other speaker’s lexicon is such that the
early US is in the extension of “empire of liberty.”9
We can illustrate these points by varying the case. Suppose the sentence had
been:
The lush fields and forests of the new “empire of liberty” were a primary source for the
wealth and privilege of American society : : :
Then there wouldn’t be the same implication of rejection, i.e. of the writer’s not
counting the early US as an “empire of liberty.” For in this variant case it’s possible
that the scare-quoting was being done just to indicate that the writer acknowledges
the provenance of that phrase in some other source. But as in the original case, an
interpreter does have enough information to know that the other whose lexicon and
context are being deployed does include the early US as an “empire of liberty”—for
in this variant case too, that is evidently the topic of the discussion. The identification
task doesn’t rest on that of figuring out how the word stands in the writer’s lexicon.
8
In fact it was a phrase that Thomas Jefferson used in 1780 to stand for the “American union”
he envisioned as the fruit of the Revolution (Boyd et al. 1950–: 4:237–38). Most of Chomsky’s
readers, however, would not know this. I’ll treat the case as a typical reader would interpret it.
9
This is clearly a type identification rather than the identification of a particular. And it must be
admitted that knowing that the early US is in the extension of “empire of liberty” doesn’t suffice
to determine one particular lexical entry for “empire” or “liberty.” Just as with context it often
suffices just to identify a type, so too with lexicon it often suffices to identify a type.
6.2 Krugman
In this example only the word “studies” is scare-quoted. The rest of the material
in the sentence tells the interpreter to treat it as being used by someone in whose
lexicon it applies to products cranked out by right-wing think tanks, purporting to
show that health reform is a failure. Presumably someone affiliated with one of these
think tanks fits the bill. Interpreting “studies” in this way gets us to the correct truth
condition for the utterance.
As with Chomsky, it takes another inference for an interpreter to get to a
conclusion about how “studies” stands in Krugman’s lexicon. Since the word isn’t a
technical term, nor is it a poetically impressive construction by some other speaker,
it is difficult to see any point to the author’s scare-quoting it except to indicate that
unlike the right-wing think-tankers he does not include the documents they produce
in the extension of “studies.”
6.3 Buzzfeed
Here we have many scare-quoted expressions. Let’s focus on the sentence

The president “bro yanked” Mr. Carney’s head and gave him a flash of “side eye.”
The intended reader knows that these are not part of the lexicon that a New York
Times writer would use in an article. The only other speaker relevant in the context in
which this sentence occurs is the Buzzfeed writer. That person is our B. Accordingly
the interpreter looks into the lexical entries for that person in order to interpret the
scare-quoted expressions.
It may be the case that the interpreter doesn’t have such entries at her disposal:
doesn’t know what the Buzzfeed writer means by the scare-quoted phrases. In that
case the sentence is not fully interpretable by her; it is as if she is dealing with a
word used normally, whose lexical entry she doesn’t know.
One might note that a fallback interpretation is available: that the president
did what the Buzzfeed writer calls “bro yanking” to Mr. Carney’s head and gave
him a flash of what the Buzzfeed writer calls “side eye.” Although in many cases
such a fallback quotational interpretation is available, we will see below that there
are reasons not to work with this idea in accounting for scare-quoting. (The main
problems are that (i) such approaches become overly complex when generalized—
as they certainly need to be—beyond syntactic constituents such as noun or verb
phrases, and (ii) the relations and relata these approaches introduce into the analysis
should be available for anaphoric uptake later on, which doesn’t appear to be the
case in general.) For now, it’s worth emphasizing that there’s nothing particularly
illuminating about the availability of a fallback quotational interpretation. Such an
interpretation is also available when unfamiliar terms are used normally. (Imagine
someone saying to you, “I only bro-yank immediate family members.” If you don’t
20 M. McCullagh
know what “bro-yank” means, you still have the fallback interpretation: he said that
he does what he calls “bro-yanking” only to immediate family members.) So the
availability of a quotational fallback interpretation in cases of scare-quoting tells us
nothing about scare-quoting in particular.
What about further inferences, concerning the author’s relation to the scare-
quoted words? Given the conventions of newspaper writing, the most that the
interpreter can infer is that the writer knows that it’s not permissible to use these
terms straightforwardly. She cannot infer that the writer doesn’t have them in his
lexicon; perhaps he does. (Nor can she infer that the writer has them in his lexicon
but with different meanings specified from those of the Buzzfeed writer.) But again:
none of this matters to the truth condition of the statement. With the Buzzfeed
writer’s lexical entries used, we get the correct truth condition for the overall
statement.
6.4 Chassis
This case is similar to Buzzfeed except that rather than being slang, the scare-quoted
expression is a scientific term. Let’s focus on the sentence
A Boston firm planned to engineer E. coli into “a chassis for iso-octane.”
The writer here indicates to us to look in someone else’s lexicon for a meaning of
the mark-flanked phrase. Who then is our B? The reader imagines the author of the
proposal—some employee of the Boston firm—as the person whose lexicon should
be used in the calculation of a truth condition for the overall sentence.
6.5 Denikin
This example differs in several ways from those so far considered. For one thing,
there is no hint of a difference in lexical meanings assigned to some word by the
writer and the other, nor any suggestion that the (perfectly ordinary) words aren’t in
the writer’s lexicon. (Stephen Kotkin is a professor of history at Princeton University
and the selection is from his recent biography of Stalin.) Rather what we seem to
have is a case in which the writer thought that the other’s phrasing was for some
reason worth preserving. Perhaps he thought it somehow evokes the situation. So he
inserted that phrase—“ordered our armed forces to advance on Moscow”—into the
sentence, flanking it with scare-quotes primarily to indicate that the phrasing is not
his.
The second major difference introduces the other component of our account. For
there is an indexical term—“our”—in the scare-quoted part of the sentence. The
“our” refers to Denikin’s people, not to any plurality involving Professor Kotkin.
The account smoothly handles this, as does the reader, who changes the context
in relation to which she evaluates the expressions in the sentence. For the scare-
quoted part, she takes it to be a context in which the speaker is some member of that
plurality. Although we are likely to think that it is Denikin himself whose words
are scare-quoted, it could just as easily be a member of his forces who later wrote a
memoir that Professor Kotkin consulted.
Finally there is the fact that the scare-quoted words are English rather than
Russian. Let’s assume that nobody present on the indicated occasion spoke English.
Then none of those people can be the indicated speaker. Is this a problem for the
account?
No; on the contrary we see its versatility brought out. For (as was noted above)
just as one can incorporate an utterance by B in context cB even if B did not actually
utter anything in context cB , one can also incorporate an utterance by a non-actual
B. All that the mechanism requires—all that the semantic calculation requires—is
that there be some lexicon and some context to work with (for the scare-quoted part).
Whether the lexicon is that of some actual speaker is irrelevant, as is whether the
context is one in which the scare-quoted words are being uttered. Thus in a case
like Denikin the natural thing to say is that the interpreter is being asked to imagine
some English speaker—an English-speaking Denikin, if you like—and to use lexical
entries stating the meanings that the scare-quoted words have for that speaker.
6.6 Lenin
This example introduces yet another twist. By a grammatical criterion it could count
as mixed quotation, since it satisfies our (stipulative) necessary condition: the quote-
mark-flanked part is within the “that”-clause complement of a reporting verb (§2.1).
But if we treat it that way, and our treatment of mixed quotation is one on which
the quote-mark-flanked words are necessarily uttered by the reportee, we will fail
to give the right results for this example, for Lenin did not, of course, use the word
“he” (or the corresponding word in German) to refer to himself. For my part, when
reading this sentence I processed it too smoothly, for that oddity didn’t even register
at first.
Despite the availability of a mixed-quotational construal, however, the reader
seamlessly processes it not as a reporting use of the quotation-marked words but as a
scare-quoting use, from some source—in this case, an account of Lenin’s speech by
someone else. That is the person—not Lenin, not Professor Kotkin—whose context
matters in the interpretation.
6.7 Range of the Account
We have, then, quite a range of cases of scare-quoting whose truth conditions admit
of explanation in terms of incorporation.
22 M. McCullagh
Before moving on to comparing the incorporation account of such cases with

other accounts, I should note that there are some cases superficially similar to the
ones I’ve discussed, which don’t obviously lend themselves to treatment in terms
of incorporation. (These cases were brought to my attention by Jonathan Wright in
personal communication, for which I’m grateful.) Recall Chomsky and consider:
That [1] “empire of liberty” was not an [2] “empire,” and it if had been it wouldn’t have
been one [3] “of liberty.”
(I’ve numbered the scare-quoted fragments for ease of reference.) The worry is
this. No matter whose lexicon we use for all of [1]–[3], we get something self-
contradictory. But the sentence doesn’t read as self-contradictory.
One plausible thing to say about this sort of case rests on the fact that the
incorporation account does not at all require that every scare-quoted part of a
sentence be interpreted using the same speaker-context pair. One could say that for
[1] it is Thomas Jefferson being scare-quoted while [2] and [3] are scare-quotations
of the speaker, which while semantically pointless have the rhetorical effect of
emphasizing that the words are those of the speaker.
Whether that response succeeds is a topic for another occasion; perhaps cases of
this sort fall under the heading of metalinguistic negation (see Horn 1985 and the
ensuing literature). Whether the incorporation account of scare-quoting extends to
an account of metalinguistic negation is also a topic for another occasion.
My claim is that there is a significant range of cases best understood as cases
of incorporation. Congruently with the point noted above about terminological
unclarity, I do not claim that there is a ready label for exactly the range of cases
to which this account so applies. I turn now to a comparison of the incorporation
account of such cases with other accounts that have been offered.
7 Comparison with Other Accounts
Several linguists and philosophers have undertaken to explain how scare-quoting

works. Several of them start from a core idea very much like the one with which I
began: that the quotation-mark-flanked words are to be taken as if spoken by some
other. But there are different ways of elaborating this idea, and none of the proposals
already made relies as mine does on a worked-out account of intra-sentential change
of context. Some other accounts start from the idea that scare-quoting makes no
difference to the content that a sentence expresses; these accounts founder, I think,
on many counterexamples, particularly ones involving indexical expressions.
7.1 “Semanticizing” Accounts
One idea is to do what François Recanati has aptly called “semanticizing the context-
shift” (2010: 283). On approaches of this sort, what makes it the case that the
scare-quoted expression is interpreted as if it were uttered by some other, is that

the sentence somehow says of that expression, that it is to be interpreted in that
way. Recall our schematic scenario: speaker A utters a sentence some part of which,
expression e, is scare-quoted, and indicates B as the “other” whose lexicon and con-
text are to be used in the interpretation of that part. The idea behind semanticizing
approaches is that the presence of the scare-quotes around e effectively replaces
it with a different expression, schematically: “what B meant by ‘e’” (in which the
expression e is purely quoted). We find this idea in several accounts.
• Robert Brandom (1994: 545–47) claims that what is “implicit in” a scare-quoting
sentence such as
Those “freedom fighters” butchered the villagers
is made explicit in a sentence that involves a definite description formed in terms

of a reference relation whose arguments are a term for some speaker and a pure
quotation—in this case,
The ones that he refers to as “freedom fighters” butchered the villagers.
• Recanati pursues an approach of this sort in chapter 17 of his 2000 book, the
idea being that “the contextual shift modifies the character of the sentence” (244)
Recanati (2000). This phrasing is jarring, since context and character are things
that in Kaplan’s scheme can’t modify each other. But I put this to the side. What
matters is how Recanati elaborates this idea. And it is clearly a semanticizing
move, as he himself later categorizes it (2010: 283). Consider the example he
uses:
Your friend “Quine” is not Quine, he is only McPherson.
In this scare-quoting, the effect of what Recanati calls the “context shift” is
to change the character of the word “Quine” into that of the phrase “person
named ‘Quine’ by the addressee” (246). This character is “metalinguistic” (247).
Recanati writes:
In general, a context-shifting function d maps a character ch [in the example, the
character of the name “Quine”] : : : onto a distinct character ch* [that of the phrase
“person named ‘Quine’ by the addressee”] such that, for any context k in its domain,
ch*(k) D ch(k0 ), where k0 is a context obtained by shifting some aspect of the original
context k. Thus d shifts the context from k to k0 . (246)
Although Recanati chooses to say that this character-shift “shifts the context,”
we might prefer to say that by shifting the character, d secures the effect of a shift
in context rather than that it somehow shifts the context, since in this account
there is nothing that actually does that. Surprisingly (given his repeated use of
the phrase “context shift”) Recanati himself is actually explicit about this, for he
24 M. McCullagh
frames his proposal as one that sticks with “the standard notion of the character of
a complete sentence as a function from (single) contexts to propositions” (243).10
We don’t find in Recanati’s discussion any argument in support of his choice
to stick with Kaplan’s original one-sentence/one-context framework. As we’ve
seen, the motivation for the idea of evaluating different parts of a sentence
in different contexts arises in its purest form in cases that have nothing to
do with quotation or “echoic” uses of language: Two Heres and the like. Part
of my reasoning towards the treatment of scare-quoting I’ve proposed is that
incorporation is something whose possibility we see before even approaching
the topic of scare-quoting; there’s no machinery needed specifically for that
phenomenon. So if Recanati’s reluctance to countenance such a treatment of the
scare-quoting cases is based on the idea that it’s too high a price to pay in order
just to handle “echoic” uses of language, then the reluctance is misguided, for the
price already needs to be paid in order to get a theoretically satisfying treatment
of certain cases (e.g. Two Heres) involving the plainest sort of context-sensitivity.
• Geurts and Maier (2005) offer an account that they might apply to the cases I’m
considering in this paper. On their view, “the semantic effect of quotation is to
shift the ordinary meaning of an expression ˛ to a new meaning that involves
mentioning ˛” (119). Applied to Chomsky, for example, this means replacing
the scare-quoted phrase “empire of liberty” with “[x e Q: Ee(x, Q, empire of
liberty)],” meaning: there is an individual x, an event e, and a property Q, such
that “e involves x’s expressing Q by means of” the words “empire of liberty”
(121). That is a semanticizing treatment: we have, in this representation, explicit
reference to a phrase and a speaker.
• Yitzhak Benbaji (2004) gives rules for the interpretation of scare-quoting of terms
or predicates. The rule for terms is that the scare-quoted term “N” refers to
“the thing to which the echoed speaker refers in producing a token that same-
tokens this. N” (542). Here again the idea is that explicit reference to the other
speaker’s meaning, and the other speaker’s context, does the job of what on the
incorporation approach is done by pre-semantic choices.
10
In his 2001 Mind paper Recanati also discusses the same example, offering what seems to be a
completely different account of it. There, the case “can be described in terms of a language shift”
(674). He goes on to explain that lexical selection is a pre-semantic affair, just as I emphasized
above. This proposal is entirely congruent with one component of mine. Whether Recanati sees the
2000 story about context-shift, and the 2001 story about language-shift, as competing accounts is
difficult to tell. What complicates the picture even more is that in the 2001 paper there is a different
account of context-shifting. This is not a semanticizing account, but one on which “a portion of
the utterance is interpreted with respect to a distinct context” (679). Recanati does not elaborate
on the idea, nor note what he noted in the 2000 book, that this conflicts with the standard Kaplan
approach to the evaluation of sentences when taken in contexts. From the point of view of the
incorporation account perhaps Recanati is best viewed as having recognized the need for a theory
of simultaneous context- and lexicon-shift, without having explained how the former would work
(other than as in the 2001 book: a character-shift that does the work of a genuine context-shift).
Approaches of this sort might seem well motivated by the core idea described
above. For reasons I’ll soon explain, I think that on reflection they are not at all well
motivated by it. For now I note some challenges such views face.
First, on this approach the truth of a sentence involving a scare-quotation requires
that the scare-quoted part actually have been uttered by the speaker whose lexicon
is being invoked, with some definite reference on that occasion. But this is requiring
too much.11 Consider Brandom’s example. It works as a scare-quote even if the
invoked speaker never did utter the phrase “freedom fighters.” All that matters is that
they are in the phrase’s extension, in his lexicon, whether he uttered it or not. (Of
course, if “freedom” and “fighters” are in his lexicon then he probably did utter each
of those words, on some occasions, with the meanings being invoked by the scare-
quoting. But it doesn’t matter whether he ever uttered them together.) Similarly for
“empire of liberty” in Chomsky.
One might propose introducing a modality into the analysis in order to accom-
modate this point. For example, rather than what x referred to as “freedom fighters”
we would have what x could have referred to as “freedom fighters.” But this
just generates the opposite problem: now the account requires too little. For any
expression is one that someone could have used to refer to some particular thing
or things—because they could have spoken a language in which that thing is in its
extension.
What this suggests, now, is that we build in reference to the speaker’s lexicon.
So rather than what x referred to as “freedom fighters” we would have what is
in the extension of “freedom fighters” in x’s lexicon. Now we require notions of
lexicon, extension etc. and the more such convolutions we add, the less plausible it
is that this is what ordinary speakers mean. (The incorporation approach becomes
comparatively simpler, too.)
That points to the second problem, at least from one perspective. For on
these accounts one must possess the concepts of quotation and reference (and
perhaps extension, lexicon etc.) in order to engage in scare-quoting. And we
may wish to be cautious in building into our account of one linguistic practice,
the assumption that speakers are proficient with some other, unless there is no
other way to explain the former. Since the incorporation account does explain
the former (in this case, scare-quoting) without assuming proficiency in the latter
(quotation, reference/lexicon specifications, etc.), simplicity considerations favor it
over semanticizing treatments.
Third, semanticizing treatments require generalization beyond noun phrases—
even beyond constituents, as some writers have emphasized (Abbott 2005; Cum-
ming 2005). To this technical problem there is perhaps a technical solution: perhaps
any scare-quoted non-constituent could be seen as the result of scare-quoting
some larger constituent and unquoting as necessary, to get to the scare-quoted
non-constituent. (See Shan 2010; Maier 2015 on unquotation operations.) On this
approach we are presupposing yet another capability in speakers and interpreters:
11
Cf Gomez-Torrente, this volume, p. 135.
26 M. McCullagh
proficiency with the unquotation operation. And again, such postulation is some-
thing that should weigh in the balance in the overall evaluation of the theory.
Finally there is what in my view is the most serious problem these accounts
have. On these accounts there is a hidden (in Brandom’s terminology, “implicit”)
reference to the invoked speaker and reference to the scare-quoted expression. And
I think if these references are there, hidden or not, then it should be possible for
them to be linked to anaphorically later in the discourse. But this often seems not to
be the case with sentences involving scare-quotation, even though it is the case with
their semanticized counterparts.
The point I’m invoking here was made quite nicely by Gabriel Segal (1990).
It is that if there is genuine reference to something in a sentence, even implicitly,
then it is available to be picked up anaphorically: “it is possible to test whether or
not a sentence has implicit semantically significant elements by seeing what kinds
of subsequent linguistic events make sense after the sentence has been uttered”
(105). Davidson’s (1963) account of action sentences passes this test, Segal argues,
for although there isn’t explicit reference to events in “Shem kicked Shaun,” the
hidden reference can be picked up anaphorically: the sentence can meaningfully be
followed by “He did it with gusto,” in which the “it” anaphorically links to a hidden
variable ranging over events.
It is far from clear that the analyses generated by semanticizing accounts pass
this test. In order for them to do so, it would have to be possible to follow any
scare-quoting sentence with another one picking up its hidden reference to some
other speaker or to some (purely) quoted expression. And while in some cases it
seems that this can be done (e.g. Krugman) in others it is far from clear. To see this,
let’s pair an explicit semanticization followed by a felicitous anaphoric uptake of
the terms for expression and/or speaker, with the same—now far from felicitous—
attempted uptake on the original scare-quoting sentence:
• Murderer
Semanticized: Someone that some people called a “murderer” moved into a
Harlem townhouse. They didn’t say that to his face, however.
Original: A cross-dressing “murderer” moved into a Harlem townhouse. *They
didn’t say that to his face, however.
• Militia
Semanticized: Another fine day for what some have called America’s “well
regulated militia.” They called it that because that phrase is in the Constitution.
Original: Another fine day for America’s “well regulated militia.” *They called
it that because that phrase is in the Constitution.
• Tardigrade
Semanticized: The tardigrade : : : is particularly impressive among those items
in the exhibit that some people called “extremeophiles.” They applied that
term to a small proportion of the items.
Original: The tardigrade : : : is particularly impressive among the exhibit’s

“extremeophiles.” *They applied that term to a small proportion of the items.
I should emphasize that I don’t take this test to be decisive against all such
accounts, since it’s not clear whether each of them undertakes the same commit-
ments as the Davidsonian claim about logical form. (Brandom’s claim about what
is “implicit” in a claim, for example, may very well not do so. On the other hand,
the Geurts-Maier DRS approach does, it seems to me.)
Stepping back from their problems, however, on reflection these approaches seem
much less well motivated by the core idea than they might appear at first. From the
point of view of the incorporation approach, there is no more reason to semanticize
the role of lexicon and context in cases of scare-quoting than there is to do so
in the very simplest uses of speech. That is, there is no more to motivate these
approaches to scare-quotation than there is to motivate an approach to all utterances
on which, for example, an utterance by John of “Ann smokes” gets analyzed as
“What John refers to as ‘Ann’ has the property that he expresses with ‘smokes’.”
Presumably that approach to “Ann smokes” has no appeal whatsoever. But it seems
no less applicable than the semanticizing approach to sentences involving scare-
quoting—so why should they seem any better motivated? In short: Why even try to
semanticize the pre-semantic?
7.2 Predelli
I move on now to non-semanticizing accounts.

Stefano Predelli’s 2003 account (which inspired me to do the style manual sur-
vey) applies to scare-quoting an idea proposed in applications to other matters both
by Kent Bach (1999) and Stephen Neale (1999). But this particular application of the
Bach-Neale idea founders, I will argue, on certain data concerning scare-quoting:
most pointedly when there are context-sensitive expressions in the scare-quoted
phrase.
The idea proposed by Bach and by Neale is that a sentence can express more
than one proposition. Bach in particular argues that the cases usually cited as cases
of conventional implicature are not cases of that, but are cases in which a sentence
expresses more than one proposition. Some words or other devices trigger such
multiple propositions. “So, for instance, according to a plausible development of
Bach’s view,” Predelli writes, “the lexical entry for ‘but’ specifies as part of that
expression’s conventional behavior the fact that sentences of the form P but Q
encode further information, besides the content that P and Q” (7). Predelli continues,
elaborating his terminology as follows:
I refer to the content thus generated as the attachment associated with a sentence, as
contrasted with its customary message; I call an expression responsible for the presence
of an attachment, an attachment-trigger. (7)
28 M. McCullagh
While the attachment-trigger makes it the case that some attachment is generated,
contextual factors often determine which attachment is generated. (8)
Predelli’s application of the multiple-propositions idea to scare-quoting is this.

Scare-quotes are attachment triggers, so a sentence with a scare-quoted part
expresses both a message content and an attachment content. The message content is
“compositionally determined on the basis of the customary contributions provided
by the expressions occurring in [the sentence], including those appearing within
quotes” (13–14). That is, it’s the message standardly expressed by the sentence
with its scare-quotes stripped out. The attachment content, while triggered by the
quotation marks, “must be contextually recovered” from the sentence as a whole
and various background factors (14).
This sort of account is understandable as an attempt to capture all of what is going
on in scare-quoting: both what is expressed and what is suggested. (It should be clear
that my concern in this paper has been the former.) It has appealed to other writers
as well: Manuel Garcıa-Carpintero (2005: 102–04) endorses Predelli’s account, and
Cappelen and Lepore (2007: 16–17) find its approach congenial (although they
differ from Predelli on how to categorize the attachment content).
As my focus is on what is semantically expressed by a sentence (in context),
my discussion will concern Predelli’s account of the message content. Predelli’s
account of this seems vulnerable to some pretty clear counterexamples.
Predelli does recognize that there are cases in which his account of how the
message is determined seems to fail. When the scare-quoted aren’t standard English,
there is no “customary contribution” that they make to the meanings of English
sentences. Predelli’s account then entails that if there are such cases of scare-
quoting, then there is no message expressed by them. In his 2003 paper he seems to
embrace this consequence:
the sentence
I am not a ‘philtosopher’
does not in all likelihood express any message regarding my profession, since the non-
word “philtosopher” fails to provide a compositional contribution. (24)
(García-Carpintero concurs, writing, “I think we should deal with [such] cases : : :

by accepting that the messages are the more or less absurd ones obtained according
to the unembellished account” (2005: 104).) Predelli’s line seems to be that scare-
quoting of items outside the speaker’s lexicon can be “acceptable” (24) (despite not
expressing message contents) only when used in what Laurence Horn (1985) has
labelled “metalinguistic negation,” as in
I am not a ‘philtosopher’, I am a philosopher.
But there are many cases of scare-quoting extra-lexical items that are entirely
legitimate and are not metalinguistic negation. As we saw, some style manuals say
that a major use of scare-quotes is precisely to flank expressions that are “foreign
to the normal vocabulary of the speaker” (Chicago 1982). In many cases, then, the
entire point of the scare-quoting is to express a proposition for which the speaker’s
own lexicon is inadequate. We should want truth conditions, not just suggested
contents, for such cases.
In a subsequent paper Predelli modifies the account so that it does not have
this objectionable consequence, or at least doesn’t have it as straightforwardly. His
modification exploits the fact that one of the interpreter’s pre-semantic tasks is that
of selecting lexical entries for the uttered words. And here there is some room
for flexibility concerning mispronunciations and misspellings. So an interpreter
can justifiably select the English lexical entry for “philosopher” when interpreting
a sentence in which “philtosopher” occurs scare-quoted. By emphasizing the
interpreter’s leeway in choice of lexical items, Predelli can now say:
the content semantically derived once the appropriate lexical items have been selected
is straightforwardly obtainable on the basis of the customary regularities governing the
English language. (Predelli 2005: 163, my emphasis)
While this seems correct as far as it goes, I don’t think that it’s enough to make
this a plausible treatment of the message expressed. There are two serious problems
remaining.
One is that the “charitable representational attitude” (163) crucial to this account
can only go so far. It will not go far enough to handle cases of scare-quoting in
which (to work with English as an example) the non-English word is not justifiably
construed as an attempted tokening of an item from the lexicon of English. (Many
cases of scare-quoted slang will be like this, as will scare-quotings of words used
with meanings other than the speaker’s, as in Krugman.)
The deeper and more decisive problem though is that Predelli’s solution
addresses only one component of the interpretative task that scare-quoting imposes,
that of choosing the lexical items for the scare-quoted words. The other dimension,
that of choosing a context for their evaluation, is entirely unaddressed by Predelli’s
line. So his solution for scare-quotings of “philtosopher” and the like, even if it did
work, doesn’t carry over to a solution to scare-quotings of indexical expressions.
In such cases simply stripping off the quotation marks results in a sentence with a
clearly very different truth condition, no matter how much reasonable leeway one
uses in assigning lexical items to the tokened words.12 Taking the Predelli approach
to Denikin, for example, gives a sentence apparently about Professor Kotkin’s
troops:
After Denikin triumphantly entered Tsaritsyn and attended services in its Orthodox
cathedral, on July 3, he ordered our armed forces to advance on Moscow.
12
I believe that this problem also arises for the account offered by Gomez-Torrente, this volume,
who writes that in what he calls “distance uses” of quotation marks—one subset of our data—
“the [quotation] marks work roughly as a parenthetical main clause paratactically added to the
full sentence without quotes, and expressing the mentioned suggestion of the utterer’s” (MS p.
26). Stripping out the scare-quotes does create problems if they flank indexicals, even in cases of
distance use.
30 M. McCullagh
(Similarly with Bush; there the first occurrence of “I” is scare-quoted, and stripping
out the scare-quote marks starkly changes the truth condition.)13
One might wonder whether the problem Predelli’s view has with counterexam-
ples is just a question of detail. Is the message-and-attachment framework wholly
refuted by there being problems with his implementation of it?
In the abstract, perhaps not. But the failure of Predelli’s account of how the
sentence determines a “message” content means that some other account is needed.
And in this task the overarching framework of message and attachment is no help:
we are back at the problem of finding out how the truth conditions of sentences are
determined by the meanings of the component expressions and the manner in which
those meanings combine. Since it is that question to which my proposal is addressed,
there is, at least in the abstract, the possibility of combining that account (of scare-
quoting as incorporation), construed as an account of the “message” content that
a scare-quoting sentence expresses, with an account of how the quotation marks
trigger the generation of “attachment” contents.
Whether there would be any explanatory gain in such a framework is not a
question I will pursue here. My aim here is just to argue that the treatment of
scare-quoting as incorporation does succeed in getting the truth conditions of scare-
quoting sentences right, and that Predelli’s account, while it proposes an interesting
framework, does not succeed in that task.
7.3 Shan
Chung-chieh Shan (2010) gives a categorial grammar for a fragment of English with
a mix of use and mention. (His account seems designed to handle the same range
of cases I’ve put under the heading of “scare-quoting” plus those I’ve classified as
mixed quotation).
His idea is that we conjoin two languages’ syntax and semantics. Take English
and [George W.] Bush-English as distinct languages, both deployed in
Bush is proud of his “eckullectic” reading list.
13
Somewhat remarkably, in light of the forcefulness with which they press data involving mixed
quotations of indexicals as refuting others’ views, Cappelen and Lepore don’t consider the very
same phenomenon in scare quotations as posing a problem for their view that scare-quotes don’t
affect truth conditions:
the quotes in mixed quotations cannot be dropped [without change of semantic content] : : :
In scare-quoting they can be dropped without a loss of semantic content (at least according
to every theory of scare-quoting in print that we know about). (2007, 17)
To me it seems that the indexicals-based case against stripping out quotation marks from scare
quotations is clearly just as strong as that against doing so from mixed quotations.
The quote-marked part is interpreted by taking it (along with the marks) as a

quotation expression, whose syntax is such that when its argument is of grammatical
category A in Bush-English, it itself is of category A in English. (Unenclosed in
quotation marks, the Bush-English expression is of type A0 in English, which has no
role in English except within quotation marks. This is necessary to prohibit “Bush
is proud of his eckullectic reading list” being a sentence of English.) Its semantics is
that the content of the quotation expression is the content of the quoted expression
at a world/context, where the world and the context are somehow determined by
conversational presuppositions or implicatures. (Shan cites Geurts and Maier (2005)
as working with the former idea and Potts (2007) as working with the latter, but puts
the matter to the side.) In this respect quotation is a rigidifier, as is Kaplan’s “dthat”
functor.
This proposal and mine have some ideas in common. One is that the content of
the quotation expression is settled by what the quoted expressions are used to mean
in a context (typically not that of the overall statement) that is somehow indicated to
the interpreter. (Predelli’s proposal, as we saw, fails to take change of context into
account.)
As far as truth conditions go, it seems that Shan’s proposal gives the same results
as does the incorporation treatment. However, it involves some specialized apparatus
introduced for the sole purpose of handling mixed- and scare-quotation. Shan even
generalizes it to handle pure quotation, by introducing a novel conception of a
context of utterance (439–40). Noting that formally we can think of contexts as
functions from expression-world pairs to contents, Shan says that some contexts can
return quotable items as the contents of expression-world pairs. Here again, while
this works formally, it is definitely a huge departure from the Kaplanian conception
of context. There is no such thing in Kaplan’s conception as an “identity context”
that, given an expression and a world, returns that expression.
Shan’s proposal is ingenious and interesting in its own right. What mine
highlights, however, is that such an apparatus isn’t required to get the truth
conditions right. All that is required is a joint deployment of treatments of intra-
sentential change of lexicon and context—and again, these are phenomena we must
acknowledge for reasons having nothing to do with mixed- or scare-quotation. So
while the Shan apparatus works, it’s not needed to handle the phenomena. The
tools for handling code-switching and for handling Two Heres and the like, when
deployed together, suffice—and again, neither of those kinds of phenomena is in
any way quotational.
The incorporation account of scare-quoting fares quite well in comparison with
other accounts. In relation to semanticizing accounts, it has the virtue of relative
simplicity, and fares better on the anaphoric-uptake test; in relation to Predelli’s, it
does a better job getting truth conditions right; and in relation to Shan’s, it sticks
more within the Kaplanian conception of context, and is simpler inasmuch as it
doesn’t require postulating any new syntactic or semantic rules for interpreting the
quotation-mark-flanked expressions.
32 M. McCullagh
8 Incorporation and Assertion
In this paper I’ve explained what incorporation is and argued that scare-quoting is
best explained as a use of incorporation. Along the way we’ve seen some evidence
of the utility of incorporation. As the survey showed, there are a variety of purposes
for which writers (and speakers, using finger-quotes) scare-quote. Sometimes it’s
the poetic or rhetorical qualities that are worth capturing (e.g. Denikin); sometimes
it’s that one wants to remind one’s readers that a phrase was used previously (e.g.
Militia). But what I want to focus on in closing are cases in which one scare-quotes
some phrase because one doesn’t understand it well enough for it to be in one’s
own lexicon (e.g. Bugs, Chassis, Aiviq). For in these cases we see a very interesting
consequence of our use of incorporation.
In these cases the effect of incorporation is that one asserts a content part of which
one doesn’t understand: a term that contributes some content to one’s statement is
not in one’s lexicon. This certainly goes against one longstanding conception of
assertion, on which it is the expression of (part of) the speaker’s mental state—a
judgement (Frege 1919: 294), a belief (Bach and Harnish 1979: 42), or a piece of
the speaker’s knowledge (Williamson 1996, 2000). I take it that for those mental
states, there is no phenomenon analogous to scare-quoting: one cannot somehow
incorporate into a belief or judgement, some content expressed by nothing in one’s
mental lexicon. So in these cases, I would argue, we have instances of the following
schema: one asserts—sincerely, properly, competently—that p while not believing,
judging or knowing that p.
I note this only to highlight one implication of the claim that we engage in
incorporation; a full discussion is for another occasion. Perhaps we have here
some data that favor accounts of assertion on which it is not constitutively tied to
the speaker’s mental states. There are conceptions fitting that general description:
conceptions on which assertion is a matter of updating shared conversational
background (Stalnaker 1978; Lewis 1979), or changing normative status (Brandom
1983, 1994), or simply a conventional act made possible by certain linguistic “force
indicators” (Segal 1990; Kolbel 2010). There are many connections to be explored
between scare-quoting (and other borderline quotational phenomena) and central
philosophical concepts such as that of assertion.14
14
I’m grateful for audiences at presentations of these ideas at the 2015 Pacific Division APA
conference (at a special session organized by Nellie Wieland), the 2015 Canadian Philosophical
Association Congress, and The 1st Context, Cognition and Communication Conference at the
University of Warsaw in June 2016, especially from Jonathan Wright following on the APA
presentation; and to Stefano Predelli and Paul Saka for comments on a previous draft of this paper.
References
Abbott, B. (2005). Some notes on quotation. Belgian Journal of Linguistics, 17, 13–26.
Almog, J., Perry, J., & Wettstein, H. (Eds.). (1989). Themes from Kaplan. Oxford: Oxford
University Press.
Bach, K. (1999). The myth of conventional implicature. Linguistics and Philosophy, 22, 327–366.
Bach, K., & Harnish, R. M. (1979). Linguistic communication and speech acts. Cambridge, MA:
MIT Press.
Baker, S. (1984). The complete stylist and handbook (3rd ed.). New York: Harper & Row.
Benbaji, Y. (2004). A demonstrative analysis of “open quotation”. Mind & Language, 19, 534–547.
Birchfield, R. W. (2004). Fowler’s modern English usage (3rd ed.). Clarendon Press. First edition
edited by Henry Fowler.
Boyd, J. P., Cullen, C. T., Catanzariti, J., Oberg, B. B., et al. (Eds.). (1950). The papers of Thomas
Jefferson. Princeton: Princeton University Press.
Brandom, R. (1983). Asserting. Nous, 17, 637–650.
Brandom, R. (1994). Making it explicit. Cambridge, MA: Harvard University Press.
Braun, D. (1996). Demonstratives and their linguistic meanings. Nous, 30, 145–173.
Cappelen, H., & Lepore, E. (1997). Varieties of quotation. Mind, 106, 429–450.
Cappelen, H., & Lepore, E. (2005). Varieties of quotation revisited. Belgian Journal of Linguistics,
17, 51–75.
Cappelen, H., & Lepore, E. (2007). Language turned on itself: The semantics and pragmatics of
metalinguistic discourse. Oxford: Oxford University Press.
Chicago. (1982). The Chicago manual of style (13th ed.). Chicago: University of Chicago Press.
Chicago. (2010). The Chicago manual of style (16th ed.). Chicago: University of Chicago Press.
Clark, H. H., & Gerrig, R. (1990). Quotations as demonstrations. Language, 66, 764–805.
Cumming, S. (2005). Two accounts of indexicals in mixed quotation. Belgian Journal of
Linguistics, 17, 77–88.
Davidson, D. (1963). Actions, reasons, and causes. Journal of Philosophy, 60, 685–700.
Davidson, D. (1979). Quotation. Theory and decision 11: 27–40. Reprinted in Inquiries Into Truth
and Interpretation, 79–92. Oxford: Oxford University Press.
De Brabanter, P. (2005). Introduction. Belgian Journal of Linguistics, 17, 1–12.
De Brabanter, P. (2013). A pragmaticist feels the tug of semantics: Recanati’s “Open quotation
revisited”. teorema, 32, 129–147.
Ehrlich, E., & Murphy, D. (1991). The HarperCollins concise dictionary of English usage. New
York: HarperCollins.
Fowler, H. W. (1926). A dictionary of modern english usage. Oxford: Oxford University Press.
Frege, G. (1919). The thought: A logical inquiry. BeitraRge zur Philosophie des Deutschen
Idealismus Translated by A. M. and Marcelle Quinton in Mind, 65, 289–311.
Garcıa-Carpintero, M. (2005). Double-duty quotation: The deferred ostension account. Belgian
Journal of Linguistics, 17, 89–108.
Georgi, G. (2015). Logic for languages containing referentially promiscuous expressions. Journal
of Philosophical Logic, 44, 429–451.
Geurts, B., & Maier, E. (2005). Quotation in context. Belgian Journal of Linguistics, 17, 109–128.
Gibaldi, J. (1998). MLA style manual and guide to scholarly publishing (2nd ed.). New York: The
Modern Language Association of America.
Gomez-Torrente, M. (2017). Semantics vs pragmatics in impure quotation. In P. Saka & M.
Johnson (Eds.), The semantics and pragmatics of quotation (pp.135–167). Cham: Springer.
Greenbaum, S. (1996). The Oxford English grammar. Oxford: Oxford University Press.
Grunwald, M. (2012). The new new deal: The hidden story of change in the Obama Era. New
York: Simon & Schuster.
Horn, L. R. (1985). Metalinguistic negation and pragmatic ambiguity. Language, 61, 121–174.
34 M. McCullagh
Johnson, M. (2017). Quotation through history: A historical case for the proper treatment of
quotation. In P. Saka & M. Johnson (Eds.), The semantics and pragmatics of quotation (pp.281–
302). Dordrecht: Springer.
Kaplan, D. (1977). Demonstratives. In Almog et al. (1989), 481–614.
Kaplan, D. (1989). Afterthoughts. In Almog et al. (1989), 567–614.
Keeley, B. (2010). The book of unnecessary quotation marks. San Francisco: Chronicle.
Kempson, R., Meyer-Viol, W., & Gabbay, D. (2001). Dynamic syntax: The flow of language
understanding. Oxford: Blackwell.
Kierzek, J. M. (1977). The Macmillan handbook of English (6th ed.). New York: Macmillan. Sixth
edition revised by Robert F. Willson, Jr.
Kolbel, M. (2010). Literal force: A defence of conventional assertion. In S. Sawyer (Ed.), New
waves in philosophy of language (pp. 108–137). London: Palgrave Macmillan.
Kotkin, S. (2014). Stalin: Volume I: Paradoxes of Power, 1878–1928. Penguin.
Lewis, D. (1979). Attitudes de dicto and de se. Philosophical Review, 88, 513–543. Reprinted in
‘Philosophical Papers’ (1983), (pp.133–159). Oxford: Oxford University Press.
Maier, E. (2015). Quotation and unquotation in free indirect discourse. Mind & Language, 30,
345–373.
McCarten, J. (2013). The Canadian press stylebook (17th ed.). Canadian Press.
Neale, S. (1999). Coloring and composition. In K. Murasugi & R. Stainton (Eds.), Philosophy and
linguistics (pp. 35–82). Boulder: Westview.
Perrin, P. G., & Smith, G. H. (1968). Handbook of current English (3rd ed.). Atlanta: Scott,
Foresman and Company. Third edition revised by Jim W. Corder.
Poplack, S. (1980). Sometimes I’ll start a sentence in Spanish Y TERMINO EN ESPAN˜ OL:
Toward a typology of code-switching. Linguistics, 18, 581–618.
Poplack, S., & Sankoff, D. (1984). Borrowing: The synchrony of integration. Linguistics, 22, 99–
136.
Potts, C. (2007). The dimensions of quotation. In C. Barker & P. Jacobson (Eds.), Direct
compositionality (pp. 405–431). Oxford: Oxford University Press.
Predelli, S. (2003). Scare quotes and their relation to other semantic issues. Linguistics and
Philosophy, 26, 1–28.
Predelli, S. (2005). ‘Subliminable’ messages, scare quotes, and the use hypothesis. Belgian Journal
of Linguistics, 17, 153–166.
Purver, M., & Kempson, R. (2004). Incrementality, alignment and shared utterances. In J. Ginzburg
& E. Vallduvı (Eds.), Catalog ‘04: Proceedings of the eighth workshop on the semantics
and pragmatics of dialogue (pp. 85–92). Barcelona: Department of Translation and Philology,
Universitat Pompeu Fabra, Barcelona.
Recanati, F. (2000). Oratio Obliqua, Oratio Recta: An essay on metarepresentation. Cambridge,
MA: MIT Press.
Recanati, F. (2001). Open quotation. Mind, 110, 637–687.
Recanati, F. (2010). Truth-conditional pragmatics. Oxford: Clarendon Press.
Saka, P. (1998). Quotation and the use-mention distinction. Mind, 107, 113–135.
Saka, P. (2013). Quotation. Philosophy Compass, 8, 935–949.
Segal, G. (1990). In the mood for a semantic theory. Proceedings of the Aristotelian Society, 91,
103–118.
Shan, C.-c. (2010). The character of quotation. Linguistics and Philosophy, 33, 417–443.
Stalnaker, R. (1978). Assertion. Syntax and Semantics, 9, 315–322. Reprinted in ‘Context and
content’ (1999), (pp. 78–95). Oxford: Oxford University Press.
Stanley, J., & Williamson, T. (1995). Quantifiers and context-dependence. Analysis, 55, 291–295.
Stilman, A. (2010). Grammatically correct: The essential guide to spelling, style, usage, grammar,
and punctuation. Cincinnati: Writer’s Digest Books.
Swan, M. (2005). Practical English usage (3rd ed.). Oxford Uinversity Press.
Todd, L., & Hancock, I. (1986). International English usage. London: Routledge.
Williamson, T. (1996). Knowing and asserting. Philosophical Review, 105, 489–523.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press.
Blah, blah, blah:
Quasi-quotation and Unquotation
Paul Saka
Abstract In saying that “A&B” is logically equivalent to “B&A”, I use strict

quotation to assert equivalence between two specific formulas of propositional
logic. In saying that p A&Bq is equivalent to p B&Aq , I use quasi-quotation and
metavariables to assert equivalence between any conjunction and its reversal. In the
case of quasi-quotation, select elements of the quotation are unquoted, unquotations
being those elements, inside of quotation marks, that are read as if they are not
inside. Unquotation and quasi-quotation are found also in natural language, e.g.:
when asked who they are, they say “I am from such-and-such a village.”
Quasi-quotation is ubiquitous in both the technical literature and ordinary speech,
but it is hardly acknowledged; much less is it explained, and the few accounts that
do exist are circular. I therefore describe quasi-quotation, distinguishing among its
varieties, and I sketch a speech-act analysis of the phenomena. Although distinct
varieties of quasi-quotation can be identified, they all depend on the speaker’s ability
to distinguish between using expressions and mentioning them, with interpretation
following preference or default rules that call for discretionary exercise.
1 Strict Quotation vs. Quasi-quotation
The following allows at least two interpretations.
(1) In The Official Rules, Paul Dickson cracks wise: “In an undeveloped
country, don’t drink the water ... In a developed country, don’t
breathe the air.”
On the one hand, in saying (1) I may mention the ellipses so that I may attribute
to my source, Dickson, a dramatic pause, a temporary trailing off, or some other
kind of silence. On the other hand, in saying (1) I may use, in the technical sense,
the ellipses: I may use “...” to signify my silence, as reporter of what Dickson said
(namely my omission of something Dickson expressed between “water” and “In
P. Saka ()
University of Texas, Rio Grande, Edinburg, TX, USA
e-mail: paul-saka@live.com

https://doi.org/10.1007/978-3-319-68747-6_2
36 P. Saka
a developed country”). In the former case I quote everything inside the quotation
marks, giving what may be called strict quotation, and in the latter case we have what
is known as quasi-quotation. (Note that even strict quotation frequently falls short
of being one-hundred percent verbatim (Lehrer 1989), an observation that points us
toward the spectrum between strict and quasi that will emerge later in the paper.)
The English language possesses a variety of quasi-quotational devices: expres-
sions that, when appearing inside of quotation marks, conventionally trigger a
temporary “popping out” of the quotation, alerting the audience that the expression
is not being strictly quoted. The activity of popping out is called unquotation in the
recent linguistics literature (Shan 2010; Maier 2014, 2017; Koev 2017), and this
term extends to that which is popped out (i.e. the portion of the quasi-quotation
that is not strictly quoted). With this terminology in hand, it is now possible to
distinguish among the components of a quotational report such as (2)1 :
(2) Arthur Koestler wrote, “He [Kepler] was persecuted by the religious
authorities.”
(a) reporting speaker/writer of (2) D the author of (2), me
(b) source speaker/writer of (2) D Koestler
(c) source message of (2) D He was persecuted by the religious authorities.
(d) quasi-quotation in (2) D “He [Kepler] was persecuted by the religious
authorities.”
(e) quasi-quotational device in (2), indicator that the quotation is not strict,
D[]
(f) unquotation in (2), the bit of quotation that is not strictly quoted, D
[Kepler]
(g) object of the unquotation in (2) DKepler
To take another example: under the quasi-quotational reading of (1), the quotation
begins with the opening quotation mark and ends with the close, the ellipses
constitute both a quasi-quotational device and an unquotation, and the object of
the unquotation is unrecoverable.
The distinction between strict quotation and quasi-quotation is well known
among formal logicians, but that does not mean it is well understood. In mitigation
I offer this paper. Section 2 addresses the function and forms of quasi-quotation
in formal logic. Quasi-quotation serves as a succinct and convenient means for
expressing generalizations, and it comes in two varieties. (Along the way I
correct use-mention confusions in the logic literature). Section 3 surveys the forms
and functions of quasi-quotation in natural language, touching upon just enough
lexicographical details to demonstrate the need for further research. Section 4 seeks
to identify normative rules that govern the use of quasi-quotation.
1
In this paper I use italics and regular quotation marks as logically interchangeable notational
variants.
Blah, blah, blah: Quasi-quotation and Unquotation 37
2 Quasi-quotation in Logic
Section 2.1 explains the need for quasi-quotation in meta-logic. Quasi-quotation is

dispensable in principle, but in practice it is so convenient that every textbook uses it,
either overtly or covertly. Section 2.2 reviews the only prominent account of quasi-
quotation in the literature, that of Quine, and explains why it is inadequate. Section
2.3 draws morals for our way of thinking about the relation between metalanguage
and object language.
2.1 The Need for Quasi-quotation
According to the standard view, logical relations are relations between sentences
in formal languages, which are sets of syntactically definable strings shorn of the
ambiguity and subjectivity found in natural language. Describing the structure and
functioning of formal languages is the job of logic, and it is the job of logic
books to identify exactly which language is under consideration at any given time.
Acknowledging this point of view, textbooks emphasize the difference between use
and mention, and concomitantly the difference between metalanguage and object
language. The metalanguage is used in order to mention or refer to the sentences of
the object language.
To draw the distinction between metalanguage and object language as sharply as
possible, textbooks typically deploy separate fonts, in the broadest sense of the term
(font type, face, case, and even choice of alphabet). I shall follow suit: my formal
language of logic, £, will consist of upper-case Roman letters in Springer type
(standing for atomic propositions), plus compounds thereof, made from the usual
truth-functional operators and parentheses; my metalanguage will include script ell
plus quotation marks and Greek letters that signify the formulas of £, so that we can
talk about £ without talking in £.
With that understood, textbooks in logic typically contain inference and forma-
tion rules for the likes of £ such as the following example, which is starred because
I will repeatedly refer back to it:
(*) If ® is a (well-formed) formula then ® is a formula.
This is perfectly fine for some audiences, in some contexts, but it is not
technically correct. Construed as a rule that schematizes formulas, so that phi stands
in (substitutes) for arbitrary Roman letters, (*) subsumes (3), which is as ill-formed
as (4):
(3) If R is a formula then R is a formula.
(4) If it is raining is a formula then it is not raining is a formula.

Contrarily, if (*) is taken referentially, so that phi schematically denotes arbitrary
Roman letters, then (*) subsumes:
38 P. Saka
(5) If “R” is a formula then “R” is a formula.

This is ill-formed because we need the tilde inside of our quotation marks.
It would also be a mistake to use strict quotation:
(6) If “®” is a formula then “®” is a formula (in £).

Since phi is part of our metalanguage and not our object language, neither “®” nor
“®” is ever a formula of £. (Also, as a grammatical matter, tildes need to combine
with sentences rather than symbols that refer to sentences.)
Textbook authors usually think they can get around the foregoing problems by
treating the tilde autonymously (not “autonomously”, as some writers persist in
spelling). With autonymy, the metalanguage tilde names the object-language tilde,
and likewise for the other operators, thus rendering (*) equivalent to:
(7) If ® is a formula then “”® is a formula, i.e. if ® is a formula then the

tilde® is a formula.
This is ill-formed, and moreover it is utterly confusing for readers who, having
been taught that formal logic is punctiliously rigorous, expect a technically correct
account.
Another gambit is to refer to form, as in the following inference rule:
(8) From a formula having the form ® ^ § (i.e. the same form as “® ^ §”),
infer a formula having the form ®.
Rule (8) merits attention for two reasons. First, rules that explicitly refer to form
recur in the literature.2 Second, references to form are implied by formulations such
as (*) and “from ® ^ §, infer ®”, which are ubiquitous. Such formulations implicitly
invoke form inasmuch as they generally occur in contexts emphasizing that form
is what determines validity and well-formedness. Form, in turn, is identified with
shape (e.g. Schiller 1912: 3) or other physical pattern in contradistinction to
“content” or meaning (Forbes 1994: 6, Manicas and Kruger 1968: 30, McKay
1989: 106). This point is emphasized by the AI literature, which takes inference
to be a mechanical operation, broadly understood, and hence one defined by the
shape, location, mass, or charge of physical tokens. It is further emphasized by
philosophers of mind who debate whether thought can be reduced to physical
syntax.
Rule (8) is meant to license the likes of “A ^ B ` A”. It does not actually do so,
however, because “A ^ B” does not have the form of “® ^ §”: the former begins
with an angular shape while the latter begins with a curvaceous shape. (We may
miss this because knowing the intended effect of [8] blinds us to what [8] actually
says; indeed, rampant expectation bias distorts readings of philosophical and legal
2
For example: “The following argument forms are used repeatedly in our derivations... A,
A B / B” (Smullyan 1962: 26); “sentences of the form p or not p can’t always be asserted”
(Field 2001: 234).
texts quite generally.) Granted, references to form may intend to invoke partial
form, but this does not, by itself, help: while it is true that form “A ^ B” partially
matches form “® ^ §”, it is also true that “® v §” matches even better, which would
invalidate (8). It would also be a mistake to regard forms as conceived abstractly.
If we abstract just a little bit away from the first two variables in (8), then phi and
psi will be understood to signify two propositional letters having indeterminate but
distinct shapes, and the result will be the failure to license “P ^ P ` P”. Abstracting
a great deal away, however, renders much of our rule system invalid. If we abstract
too much, for instance, then (8) will license “P ^ Q ` R”.
It is perhaps for this reason that some textbooks drop metavariable letters in favor
of blank spaces, blank lines, ellipses, and dashes:
(9) From a proposition having the form “___ ^ ...”, infer a proposition
having the form “___”.
This maneuver, however, is confusing. If it is meant to solve the problem that
arises for variables ® and §, it is an abject failure, as blank lines, ellipses, and
even spaces have their own specific forms that fail to be instanced in object-
language formulas; and if it is meant for some other purpose then the reader is
left wondering what advantage blank spaces and ellipses offer over a more prosaic
use of easy-to-pronounce letters. What’s worse, some textbooks purport to explicate
metavariable letters by means of blanks.3 But this is circular: if you don’t understand
the unquotation of metavariable letters then you can’t be expected to understand the
unquotation of blanks.
Another inadequate specification of our object language appears in the canonical
textbook by Benson Mates (1972: 45, 26), which includes:
(10) If ® and § are formulas then (® v §) is a formula.

This, Mates tells us (some 20 pages away), is an abbreviation of:
(100 ) If ® and § are formulas then the result of writing “(“, followed by ®,
followed by “v”, followed by §, followed by “)” is again a formula.
Though cumbersome, (100 ) gives the correct result, and it illustrates how any rule
or meta-theorem of £ whatsoever could be spelled out. Yet (100 ) does not actually
follow from the rules that Mates sets out for understanding the metalanguage.
His rules are: (i) Greek letters name formulas, (ii) operators and parentheses refer
autonymously, and (iii) “to refer to a compound expression we juxtapose names or
variables referring to its parts”. By rule (ii), (10) gives us juxtapositions:
(11) If ® and § are formulas then the left parenthesis® the disjunction operator
§the right parenthesis is a formula.
3
“These [metavariable letters] are merely placeholders which serve in lieu of blank spaces”
(Manicas and Kruger 1968: 30); also McKay (1989: 26).
40 P. Saka
From rule (iii) we know that these juxtapositions are supposed to refer to some
compound expressions. It does not tell us, however, which compound expressions
are being referred to. On the one hand, the rules construed in English do not tell us
that (10) means the same as (100 ), for the rules give us (11), which is not English.
Aside from the minor fact that “parenthesis®” and “§the” are not words – nor do
they become words when the Greek letters are replaced by names of propositional
letters – a clause of the structure NP NP NP VP is ill-formed and hence formally
uninterpretable. On the other hand, the rules construed in some other language, if
they are to be comprehensible, require our knowing just what this other language is.
To be sure, with charity and common sense we can figure out what Mates intends,
even though he does not say it; but if charity and common sense are to rule the day,
why not count on them in the first place, instead of going through the charade that
clauses (i-iii) constitute a technically correct set of rules for precisely and objectively
understanding the language of logic?
(*), (3), (6), (8), and (9) instantiate different approaches to specifying the forma-
tion and inference rules of a formalized language. Even when supplemented with
autonymous operators and glosses (i–iii), however, they are technically incorrect.
One possible response to this state of affairs is to spell out all of our rules and meta-
theorems in the cumbersome manner found in (100 ) – a manner that not only makes
reading and writing logic awkward but also reduces cognitive tractability. A better
response is to treat (*) and (10) as cases of covert quasi-quotation. The latter move
calls for clearly and accurately explaining quasi-quotation.
2.2 Attempts to Explicate Quasi-quotation
As an efficient way to express generalizations about the object language, Quine

(1961: ch. 1.1) introduces quasi-quotation, which he marks with corner quotes:
(12) If p ®q is a formula then p ® q is a formula.

Quine gives a general definition that treats quasi-quotations as definite descriptions:
p
(13) ______q D the result of putting ® for “®”, § for “§”, ... in “______”.
Unfortunately, Quine’s account of his own device is unsatisfactory.
The chief problem is that if quotation marks always signify strict quotation, as
Quine elsewhere indicates they do (p. 38), then “______” denotes no more and no
less than one blank line, without reference to Greek letters, and so (13) does not
give us the meaning of “p ®q ”. If contrarily the quotation marks in (13) can be
understood as signifying quasi-quotation then – setting aside quibbles concerning
the ellipses – a precise, unambiguous rendering of Quine’s rule should be:
p
(14) ______q D the result of putting ® for “®”, § for “§”, ... in p ______q
(or perhaps in “p ______q ” or perhaps in p “______”q ).
But this is circular, and does no one any good. It would be more honest to stick
with (*).
Quine compounds his lapse by comparing phi and psi to algebraic letters,
adding: “Roughly speaking, the letters ‘x’, ‘y’, etc. may be described as ambiguous
numerals, ambiguous names of numbers”. But if Greek letters are ambiguous names
then (15) amounts to (16):
(15) p ®q is equivalent to p ®q .

(16) “A” is equivalent to “A”, or “B” is equivalent to “B”, or “C” is
equivalent to “C” ...
After all, ambiguity normally carries disjunctive force: “Jan picked up all the files”
means Jan picked up dossiers or tools of abrasion; it does not mean Jan picked up
both (Saka 2007: ch. 6).4
The quasi-quotation “p ® D §q ” “amounts to quoting the constant contextual
background ... ‘ D ’, and imagining the unspecified expressions ® and § written
in the blanks.” It is, says Quine, synonymous with all of the following:
(17) The result of writing ® and then “ D ” and then §;
The result of putting ® and § in the respective blanks of “ D ”;
The result of putting ® for “®” and § for “§” in “® D§”.
Quine’s invocation of imagining, writing, putting, and result all imply intuitionism
or a speech-act theory of logic. (A speech-act theory of logic begins with the usual
thesis that reasoning takes place in the medium of language and adds that language
is to be understood in terms of speech acts.) Since Quine explicitly disavows
intuitionism elsewhere, and makes statements incompatible with speech-act theory,
several interpretive options are open: (i) that I am mistaken in thinking that (17),
taken literally, implies a speech-act theory; (ii) that I am mistaken in construing
Quine elsewhere as distancing himself from speech-act theory; (iii) that Quine’s
views have changed between publications; (iv) that (17) is mere metaphor for
Quine’s true view. One of my critics avers that (iv) is the “natural” view, and that it
is demanded by charity, but I see problems for this option. On the one hand, suppose
(17) is intended to be recognized as mere metaphor because Quine expects the
readers of (17) to be familiar with his larger body of published work and to exercise
the principle of charity. This supposition represents Quine as rather arrogant, and
for that reason is uncharitable. On the other hand, suppose Quine does not intend
for (17) to be recognized as mere metaphor. In that case the “natural” interpretation
is that (17) says what it seems to say, i.e. its constructivist language is to be taken
literally. This same kind of critique also applies against Mates’s (100 ), and a great
deal of the logic literature.
Quine offers a referential view of quasi-quotation: the Greek letters “take the
place of” (are equivalent to) quotations which refer to the formulas of £; they “name”
4
Quine does say that the Greek letters may be compared to ambiguous numerals only roughly
speaking, but my point is that Greek letters are really nothing like ambiguous terms.
42 P. Saka
them. Consequently, “when a Greek letter stands alone in corners, quasi-quotation

is vacuous”: p ®q D ®. (This subsumes “it is raining” D“it is raining”.) There
exists however a different understanding of quasi-quotation that sees Greek letters
as schematizing sentence formulas instead of names or quotations. (In this case,
“p ®q D ®” subsumes a solecism: “it is raining” Dit is raining.) Under the Quinean
referential view, (*) should be revised indifferently as either (*a) or (*b); under the
alternative sentential view, (*) should be revised as (*b).
(*a) If ® is a formula then p ®q is a formula.

(*b) If p ®q is a formula then p ®q is a formula.
These two versions of quasi-quotation are largely indistinguishable in practice. Still,
the fact that two versions populate the literature, with nary any notice, underlines
my point that quasi-quotation is taken too much for granted.
Looking beyond Quine, the only accounts of quasi-quotation that I could find
in the logic literature are those by Grover and Belnap (1973) and Marciszewski
(1981). Both accounts are covertly circular in the same way that Quine’s treatment
is, and both fail to acknowledge the two kinds of quasi-quotation that exist. For this
reason I shall offer my own account of quasi-quotation, one which does not treat
quasi-quotations as definite descriptions.
2.3 Meta-language and Object Language
To sum up my main points so far: quasi-quotation, though dispensable by means of

the likes of (100 ), is highly useful for the purpose of expressing generalizations,
and is used implicitly or explicitly by nearly all logic books. Yet it is almost
never explained, a problem compounded by there being two rival systems of quasi-
quotation, the referential and the sentential, which differ on whether p ®q D ®.
(Neither “quasi-quotation” nor even “quotation” is listed in the indexes of Black-
well’s Companion to Philosophical Logic, the Blackwell Guide to Philosophical
Logic, or the first six volumes of Dov Gabbay’s Handbook of Philosophical Logic.)
Even when it is explained it is explained incorrectly. Quasi-quotation is neither self-
evident nor easy to get right.
The result is not merely technical error that is of concern to the overly
punctilious; it is a pedagogic fog that must surely disorient the uninitiated. The
confusion is exacerbated, not ameliorated, by textbook invocations of an over-
simplistic distinction between metalanguage and object language:
(18) metalanguage (language of textbook) !describes!object language (£)

Because readers have the right to assume that textbooks are written in the language
of the intended audience, (18) claims, in the context of an anglophone community:
(180 ) metalanguage (English) !describes! object language (£)

It would come closer to the truth, however, to acknowledge the interplay of at

least three languages: £; the beginning student’s version of English; and a mongrel
technical language that bridges the two, containing Greek letters and ordinary
English and carets and other special characters and extra rules – an exotic system X
that is powerful enough to efficiently characterize £. At the minimum, therefore, a
better picture looks like (19) or its variant:
(19) metalanguage 1 (English) !describes! object language 1 (X)
metalanguage 2 (X) !describes! object language 2 (£)
(190 ) meta-metalanguage (English) !describes! metalanguage (X)
!describes! object language (£)
Even this picture, however, is incomplete. In his attempt to describe language X,
Quine posits blank spaces and autonymous corner quotes. These expressions are
not part of X, but descriptive of it; and they are not part of ordinary English. (Blank
spaces do exist in English, but there they do not possess the requisite significance,
wherein spaces stand for Greek letters.) Quine thus quietly adds an intermediary
between English and X, call it XX. Consequently, we use the words “blank space”
in language 1 (English) to refer to the blank spaces in language 2 (XX); we use the
blank spaces of language 2 to refer to the Greek letters of language 3 (X); and we
use the Greek letters of language 3 to refer to the formulas of language 4 (£):
(20) meta-meta-metalanguage (English) !describes! meta-metalanguage (XX)
!describes! metalanguage (X) !describes! object language (£)
(Reichenbach 1947 overtly employs a similar four-tier system.)
Against this whole view, it may be said that the boundaries of language are
not regimented: although readers do not have Greek letters and autonymous carets
in their idiolects when they first approach logic, they are capable of absorbing
new lexical items and symbols as they occur in context, all without the benefit of
being told definitions. This would be an argument for retreating from the elaborate
complexity of (20) by regarding a union of English, XX, and X as a single
metalanguage, thus returning us to model (18). But it would equally be an argument
for going further, and denying that £ is a language in its own right, as opposed to a
subset of lexico-syntactic resources that monolingual speakers can make use of on
an opportunistic basis.
Instead of distinguishing between metalanguage and object language, I myself
would distinguish between meta-level speech-acts and object-level speech-acts, and
I would point out that the relation is not anti-symmetric (or even irreflexive). If
utterance U1 is about any utterance U2 then U1 is metalinguistic relative to U2 , even
if the two are in the same language; and this leaves open the possibility that U2 is a
metalinguistic utterance about U1 (as when Socrates says that everything Plato says
is true, and Plato reciprocates). The terms “metalinguistic” and “object-linguistic”
retain legitimacy, but need to be disengaged from the standard Tarskian hierarchy-
of-language model.
One virtue of a speech-act perspective, incidentally, is that it answers a criticism
lodged by Ludwig and Ray (2017: fn 12): “if the use theory [of quotation] is correct,
44 P. Saka
we cannot truly say in English that nurphalisturbia is not a word that has a use in
English, because that would ipso facto illustrate a use of the word in English.” Once
we replace the question of reified language with the question of language users,
we more readily see that Ludwig and Ray equivocate. Does “English” refer to an
abstraction drawn from a community of speakers or to an idiolect? On the one hand,
“nurphalisturbia is not a word used by most speakers of English” is true despites its
use by a few quotation quibblers; on the other hand, “nurphalisturbia is not a word
used in my idiolect, as I speak English” is false as soon as it’s uttered. In neither
case does the assertion generate the contradiction that Ludwig and Ray allege.
Regardless of whether one radically rejects the hierarchy-of-language, or con-
servatively accepts (20), (18) presents a mistaken picture of what’s really going
on. My reflections thus highlight the theoretical point that the relation between
metalanguage and object language stands in need of clarification. They moreover
highlight the pedagogical point that those who labor under misconception (18) will
incline to imagine that since Greek letters are not English letters they must therefore
belong to £.
3 Quasi-quotation in a Natural Language
Any quotation in English, I shall argue, is unspecified, being potentially intended

as either a strict quotation or a quasi-quotation. By default, quotation marks invite a
fairly strict interpretation, but this default can be overridden by context (as in [67]
below); it can be overridden by stipulation, as when a logician announces that a
designated font will be used for metavariables; and it can be overridden by the use
of devices – such as terms and punctuation – that are frequently or customarily
used for unquotation. Such standard or conventional quasi-quotational devices
include “so-and-so”, “you-know-who”, “blah blah”, “something something”, coded
euphemisms, ellipses, slashes, and brackets, each with its own characteristics and
some with local rules promulgated and enforced by usage barons (e.g. publishers,
editors, and organizations such as the MLA). These devices will be discussed as
they appear in direct speech reports (§3.1), indirect speech reports (§3.2), mixed
reports (§3.3), metalinguistic citation (§3.4), and pure use (§3.5). Much of my data
comes from the Corpus of Contemporary American English (COCA), which has
more than one-half billion words drawn in roughly equal parts from newspapers,
magazines, academic journals, fiction, and transcripts of speech from tv/radio (http:/
/corpus.byu.edu/coca/).
In this section my aim is to identify and characterize prominent quasi-quotational
devices. My catalog is not meant to be complete, as I can think of marginal others,
and imagine that significant others are to be found. Nonetheless, Sect. 3 will serve
as a basis for Sect. 4 and for future work.
3.1 Direct Speech Reports (“Direct Quotation”)
In quasi-quotation, one speaker (the reporter) manages to quote another speaker (the
source), while deviating from the source’s actual words. This general function –
deviation – is exercised by the reporter for a variety of reasons: to harmonize the
syntax of the source with that of the matrix report (especially in mixed quotation);
to abbreviate the source’s message for the sake of brevity; to elide distracting
bits of the source message for the sake of foregrounding others; to clarify the
source; to interpolate editorial comments about the source; to save someone from
embarrassment; to keep from getting in trouble; and to communicate a partial
message when in ignorance of the whole (due to obscurity in the original source,
forgetfulness on the reporter’s part, or other failure). These various functions are
exemplified below.
SO -AND -SO AND SUCH -AND -SUCH are primarily found in the position of noun,
noun phrase, or nominal modifier:
(21) A chorus of complaints jackhammers through my skull: “Jon, that deadline is

tomorrow.” “You better call so-and-so, he’s peeved about such-and-such.”
“Daddy, daddy, the VCR ate my Scooby tape!” [COCA]
It is possible that Jon, the reporter, is strictly quoting one vague complaint (“You
better...”), sandwiched between two specific complaints; but it’s just as likely that
Jon is quasi-quoting, with “so-and-so” and “such-and-such” signifying determinate
material in the original source.
In English there is a preference to use “so-and-so” for persons, and “such-and-
such” for non-persons, but it is not a requirement, as attested below in (43) and
(57). The existence of such preferences and defaults and typicality effects is neither
marginal nor incidental in natural language but rather widespread and essential
(cp. Lakoff 1987, Jaszczolt 2005, Garcia-Carpintero 2017, Gomez-Torrente 2017,
Gregoromichelaki 2017, Jaszczolt and Huang 2017). Moreover, in my view it counts
against truth-conditional approaches to understanding quasi-quotation.
According to Sudo (2013), “it is often remarked” that so-and-so and French un tel
“function as place holders for proper names”, and the same claim for whatshisface
is made by Koev (2017). This is a mistake, however. Granted, only a proper name
can substitute for “so-and-so” here:
(22) They’ll say that they might be suspicious of the local bishop or the
hierarchy, so-called, or the institutional church, but they love, you know,
their local pastor, they love Father So-and-So. [COCA]
Furthermore, I acknowledge that names enjoy privilege, at least in our individu-
alistic anglophone culture – who-questions, for instance, most often elicit proper
names as answers. Despite a certain preference for names, however, who-questions
can always be answered instead by kin relation (“my niece”), social role (“the
mayor”), other definite description (“the first customer to come through the door”),
indexical (“him”), or indefinite (“some guy I met”). By the same token, definite
46 P. Saka
descriptions can substitute for “so-and-so” in examples such as (21), and something
other than a name is positively required in “that dirty so-and-so” (where “so-and-so”
euphemistically substitutes for the likes of son of a bitch, not Smith).
SOMETHING AND WHATEVER . These indefinites are used when the reporter
is incapable of providing a complete strict quotation. “Something”, usually in
reduplicated form, typically signals illegibility or other transmission failure between
the original source and the reporter:
(23) As Bakaleynik reads through one file, she mumbles, “59-year-old female who
something something complains of low sodium something something ...”
[COCA]
Reduplication conveys markedness, and inside of quotation marks unquoted mate-
rial is marked; hence it is no accident that a single “something”, occurring inside
a quotation, is more likely meant to be strictly interpreted than a reduplicated
“something”.
“Whatever” typically signifies the reporter’s failure to remember the original
source:
(24) She saw me to her door, never ceasing to chatter. “Such a disgraceful crime ...
profaning a holy place ... like King Henry the Whatever, back in the Christian
era, murdering that archbishop fellow on the altar ...” [COCA]
(Needless to say, a strict quotational reading is permitted here, but the point is that
it is not required.)
YOU -KNOW -WHO . Speakers use “you-know-who” to prevent overhearers from
understanding; to criticize someone without going on record for doing so (which is
a sort of euphemism); to avoid using names of people they detest or are embarrassed
to mention; and for histrionic effect:
(25) I answer him and then I say: “Miguel, hold my wallet because I intend to go
inside to see how you-know-who is doing.” [COCA]
(Again, both strict and quasi-quotational readings are possible.)
ALGEBRAIC VARIABLES . Presumably due to the influence of grade-school
algebra, speakers sometimes use single-letter variables in ordinary speech. Except
when such variables are specially motivated (for instance, “n” for number), the
first variable used in a discourse is usually “X”, the second one “Y”, and the
third one “Z”:
(26) If I catch him about to do something he shouldn’t, I say “Now Jonah, you’re
about to do X. Is that really what you want to do?” [COCA]
This practice (of prioritizing “X” over “Y”, and so on) illustrates another manifes-
tation of preference rules in human communication.
ABBREVIATIONS . It is not uncommon for fiction to use initial letters instead of

full names:
(27) But the other two [officers] called out, “Herr K.! We’re to be beaten
because you made a complaint about us ...” [Franz Kafka, The Trial]
Perhaps Kafka wished to impute universality to K.’s experience, and bleached the
name to imply that it could just as well be anyone. Perhaps, alternatively, Kafka
wished to enhance the realism of his story by pretending that the character’s privacy
needs to be protected, thereby implicating that the character is real – that, indeed,
the character may be Kafka himself. For our purpose what matters is that (27)
admits two possible interpretations: either the two officers are on informal terms
with K., and do use his initial when addressing him; or (as indicated by the larger
text) Kafka’s narrator is expurgating the speech report, in which case we have an
instance of quasi-quotation.
BLAHS AND YADAS are usually used in a succession of three, although they
sometimes occur in a shorter or longer sequence. Indicating omission of quoted
material, they also connote uninterest on the part of the reporter: material is omitted
not just because it is irrelevant but, especially in the case of blah, because it is
tedious, and the original source is portrayed as a blatherer:
(28) All it takes is a feigned interest in your new boots (trust us, a straight man
truly doesn’t care) ... If your response is something like “Thanks! I just
bought ‘em last week at this designer store, they’re Italian-made ... here,
feel the leather! blah blah blah” we sense that you’re materialistic, hard to
please, and a spendaholic. [COCA]
It is also possible to impute blathering to oneself, thus conveying humility:
(29) For a while, I had cards made up that said, “My name’s Hank Azaria. While
you may know me from such films as blah blah blah or such tv shows as
blah blah blah, I’m actually quite shy and your discretion’s appreciated.”
[COCA]
Blah differs from the Greek letters of meta-logic. In (29) the two tokens of “blah
blah blah” each stand in for its own list of titles, whereas in “® v ®” the two tokens
of phi stand in for the same arbitrary formula. This difference – variables coordinate
with themselves in logic but not so much in ordinary language – is to be explained
by pragmatic considerations, I believe. To start with, phi belongs to a family of 24
sisters, which provides for a richness of expressive power lacking in the yada/blah
family (and in the he/she/it family, another domain where variables do not tightly
coordinate). More importantly, precise expression is imperative in logic but not so
much in ordinary language, and for this reason speakers who use blahs have both
less incentive to coordinate them and less practice in doing so.
48 P. Saka
Another superficial difference between logic and ordinary language is that the
metalinguistic variables of logic have something of a universal force whereas so-
and-so and blah tend to have an existential force:
(30) Eric was “blah blah blah” the whole way up. [COCA]
This claims there was something that Eric was (saying) the whole way up; it
does not claim that for any arbitrary fill-in-the-blank, Eric was saying it the whole
way up. Again, however, I believe the difference is but pragmatic. I submit that
metalinguistic variables are quantificationally unspecified, semantically speaking,
and that they are read as universal because of the discourse function that they serve,
namely to express generalizations about well-formedness, validity, consequence,
and so forth. Meanwhile, semantically, the quasi-quotational devices of ordinary
language are likewise quantificationally unspecified. Blah may be read either
existentially (30) or universally (53), such-and-such may be read either existentially
(43) or universally (57), and other examples fit neither mold (22).
Quasi-quotation in English, therefore, is fundamentally the same phenomenon as
quasi-quotation in logic. (This should come as no surprise, since quasi-quotation
as it occurs in logic is not in the artificial language £ but in the meta-logic,
a version of English.) This view is reinforced by the indeterminacy of correct
punctuation practice. As we saw in Sect. 2.2, logicians do not agree on whether
metalinguistic variables in isolation should receive corner marks. By the same token,
some speakers of English use isolated blahs within quotation marks (30) while some
use them without:
(31) Next a Master comes on the screen and starts describing the draft
process, “randomly selected” blah blah blah “everyone will get a card”
blah blah. [COCA]
EUPHEMISMS . Speech is restricted by powerful social norms (in some cases
legal norms, with serious penalties). In today’s anglophone culture, taboos govern
sexually explicit, racist, blasphemous, and libelous language (Anderson and Lepore
2013). To avoid or mitigate taboo, speakers use euphemism. Euphemism comes in
different varieties, which relate to quasi-quotation in different ways.
Euphemism can be achieved by means of grammar, lexicon, phonemic variation,
redaction, and what I call coding. In grammatical cases, the speaker relies on
grammatical resources to navigate around potentially offensive or embarrassing
statements (e.g. “Mistakes were made” instead of “You/I made mistakes”). In
lexical cases, choice words or idioms are used to the same effect (e.g. “husky” for
“fat”). In phonemic cases, vowels and consonants are altered to produce marginally
distinct lexical variants: “God” mutates into gosh, golly, egad; “Christ” becomes
cripes, crikes, criminy; “Jesus” becomes jeez, jeepers, sheesh, gee whiz; “devil”
becomes dickens; “Hell” becomes heck; “damn (it)” becomes darn, dash, daggit;
“shit” becomes shoot (hence the witticism “you’ll go to Heck if you don’t believe
in Gosh”). These are the more conventional euphemisms. (The wealth of such
vocabulary attests to opposing desires to blaspheme and to refrain, the result being
a class of semi-taboo words.) Finally, there is coded language, in which letters

are dropped or replaced by non-letters. (Vowels are more likely to be replaced
than consonants, for writers know that vowels are cognitively easier for readers
to recover.) The symbols *, @, !, # stand in for elements of obscene language, and
many of my students code “God” as “G-d”. When symbols are not coordinated with
each other (e.g. “he is f***ing crazy”), we have a special case of coding, redaction
(in redacted language, letters are removed or blackened out, or the functional
equivalent thereof).
With that typology of euphemism, I would now like to put forth general
theses about quotation. Generalization 1. Direct quotations of grammatical and
lexical euphemism are, by overwhelming default, strict quotations. If you use the
word “torture”, it would be wrong for others to quote you as saying “enhanced
interrogation”, even if doing so made them more comfortable. Generalization 2.
Direct quotations of codes, in contrast, are perfectly ambiguous:
(32) Jan claimed, “Kit is f*cked up”.
If Jan herself used the code, I have strictly quoted; if Jan did not, I have quasi-
quoted. If I have quasi-quoted, my self-censorship may be outer-directed (so as not
to offend my reader) or inner-inspired (motivated by my own sensibilities), or both.
Generalization 3. Direct quotations of phonemic euphemism are a little ambigu-
ous:
(33) Jan warned, “Gosh darn it, don’t ever do that again.”
Aside from its strict quotational interpretation, I believe that (33) does permit a
quasi-quotational interpretation. However, I believe that the strict interpretation is
much preferred. For one thing, the custom of using coded euphemism has now
grown to the point where codes would be expected in quasi-quotation, leaving the
phonemic euphemism in (33) as most likely the result of Jan’s word choice, not
the reporter’s, if (33) is printed rather than spoken. For another thing, society today
appears to be less queasy about blasphemies than obscenities and racist epithets;
many are those who would use the reported euphemisms, but comparatively few
would refuse to quote “God damn it”. That said, I think it is possible to see (33) as
quasi-quotation, and it is especially easy if we imagine that the reporting speaker
is a child who takes seriously the prohibition against using adult language, or if the
reporting speaker lives in a more pious culture.
BRACKETS . Square brackets can be used to indicate commentary made by the
reporter. Commentary can be useful clarification, as when pronouns are amplified
(34) or substituted (35); they can be snooty or small-minded or insecure, as
when one writer wishes to belabor what is obviously a typo; and they can be
extended discussions where the reporter clarifies, elaborates upon, or challenges
some primary source. Common clarifications include “[sic]”, “[emphasis mine]”,
“[italics in original]”, speculative corrections with question marks, and imperatives
instructing the audience to insert material (36, 37):
50 P. Saka
(34) Kepler wrote to Tycho, “Why does he [Ursus] set such value on my
flatteries?” [Koestler in The Sleepwalkers]
(35) Shirley MacLane wrote, “[Jimmy Carter] says his lust is in his heart. I
hope it’s a little lower.”
(36) This is probably scary to a lot of people, particularly the type of people
who tend to get moralistic (or, worse, political) about the sports world,
the ones who say, “I’ll never root for that guy because he (INSERT NAME
OF SOME SORT OF FELONIOUS ACTION ).” [COCA]
(37) According to the university’s Pride Center, the first thing a member of
the UT community should say upon being introduced is: “Oh, nice to
meet you, insert name. What pronouns should I use?” [COCA]
These show that – outside of scholarly registers – brackets may be written as
parentheses, or may even be dropped altogether. In addition, (36) exemplifies my
general point that when speakers change “voice”, shifting from object language
to metalanguage, they may choose to make the medium fit the message, aligning
differences in font, case, indentation, or other physical feature with communicative
intent.
Besides offering interpolated material, brackets can also indicate deleted material
(“[expletive]”) and revised material, if the revision is sufficiently trivial. Capital
letters can be changed to lower case, and vice versa, depending on the needs of
the report’s syntactic context; verb inflections can be changed, so long as the sense
of the original is maintained; the numerical labeling of linguistic examples can be
adjusted to fit new contexts; and measurements can be translated from one unit to
another:
(38) An eyewitness wrote, “Our ship being olde and weake was so tossed, that
she opened at the sterne a [fathom] under water ... and for feare of
sinking we threw and lightened into the sea all of the goods we had.”
[Toby Green, Inquisition]
(The translation of measure terms, incidentally, receives fuller discussion in Saka
2010.)
In scholarship it is common to attribute sources via parenthetical citation. For
brief passages it is normal to put a quotation inside of quotation marks, and then
afterwards cite its source in parentheses. For extended passages, it is normal to
mark a quotation by using block indentation, and then to credit its source at the
end, still within the scope of special indentation. Consistency demands that brackets
be used at this point, for brackets generally signify a shift from the source’s
voice back to the reporter’s voice. Unfortunately, the current custom inexplicably
uses parentheses rather than brackets here (and thus, in this one irregular case,
parentheses carry quasi-quotational significance). A related critique of prevailing
convention, by Pullum (1984) regarding quasi-quotational commas, seems to be
contributing to an overdue shift in punctuation norms.
ELLIPSES . Quotation serves many functions (cp. Johnson 2017: sect. 1). Just in
scholarship alone, when A quotes B it is generally to (i) demonstrate that A knows
the literature; (ii) acknowledge credit or give honor; (iii) ride on or reinforce B’s
authority; (iv) use B’s ideas as expository framing for A’s own ideas; or (v) correct
an error that B has propagated. When the quotation of B goes on for longer than A
wishes it to, there being material irrelevant to purposes (i–v), A may excise part of
the quotation and leave ellipses in its stead.
SLASHES AND BRACES are used quasi-quotationally for the purpose of abbrevi-
ation. For example, (39) may mean (390 ):
(39) “I ran in/into the house” is grammatical.
(390 ) “I ran in the house” and “I ran into the house” are both grammatical.
Unlike brackets and ellipses, the slash does not signal the simple addition or deletion
of material; (39) does not mean (40), for instance.
(40) “I ran in the house and I ran into the house” is grammatical.
Instead, the slash signals that the audience is supposed to follow a procedure to
unpack the intended strict quotational meaning of (390 ).
The use of the slash is quite common among linguists but it is transparent and
sometimes used in ordinary language. More technical is the use of curly brackets,
or braces:
(41) “The ball fis in / fell intog the water” is grammatical. [Leonard Talmy,
Toward a Cognitive Semantics]
The braces clarify that (41) means (410 ) rather than (42):
(410 ) “The ball is in the water” and “The ball fell into the water” are
grammatical.
(42) “The ball is in into the water” and “The ball is fell into the water” are
grammatical.
OTHER EXPRESSIONS ? Todor Koev (2017) claims that whatshisface, whatshis-
name, thingummy, thingy, “etc.” are metalinguistic variables. Of course, if I am
right then any expression can function as a quasi-quotational device in the right
context, but I read Koev as saying that whatshisface etc. are normally or frequently
meant to be unquoted. One problem with this claim is that it is not clear what
“thingy et cetera” refers to. Thingamabob? Thingamajig? Doohickey? Doodad?
These examples come to mind, but they do not strike me as conventional devices
that signal quasi-quotation. A related problem for Koev is that attestations in the
COCA corpus of the items mentioned and entertained in this paragraph, aside from
one allusion, are not metalinguistic. (The corpus reveals that it is not uncommon for
thingy, or even just thing, to be used as a euphemism for genitals.)
52 P. Saka
3.2 Indirect Speech Reports (“Indirect Quotation”)
All of the quasi-quotational devices floated in the preceding section – “so-and-so”,

“whatever”, “blah blah”, coded euphemisms, abbreviations, brackets, ellipses, and
so on – appear in indirect quotations. For example:
(43) I said so-and-so, and she said such-and-such, and for some moments we
kidded back and forth along those lines. –Harold Pickering [PG
Wodehouse, A Few Quick Ones]
Here, Pickering uses “so-and-so” to stand in for some of his own previous discourse,
and “such-and-such” for his fiancee’s discourse. The particular contents of this
exchange are left unspecified and undemonstrated, which is what makes the
subsequent demonstrative “those” so incongruous, and so amusing.
In indirect quotation it is already known that the reporter is taking certain liberties
with what the original source originally said, and hence it may seem that symbols
signaling omission of source material would be otiose. However, because blahs fill
a syntactic role – perhaps any syntactic role whatsoever – they serve a function and
predictably do occur:
(44) He called me and left a long, rambling message on my voicemail,
explaining that he had a limited number of tickets to the party and he had
to uninvite a bunch of his closest friends and he felt so embarrassed and
blah blah blah. [Ron Jeremy, The Hardest (Working) Man in Showbiz]
Ellipses too occasionally occur in indirect quotation. Indeed, they can also be used
without explicit attribution to an original source, i.e. in cases of allusion, including
misquoted allusion:
(45) Judy squinted to try to read the fancy-schmancy handwriting. When in
the bones of human events ... blah blah blah ... we hold these truths ...
more blah blah ... alien rights ... Life, Liberty, and the Purse of
Happiness. [COCA]5
Although the devices discussed in Sect. 3.1 can all appear in indirect quotations,
it’s not clear that they are especially quasi-quotational, for perhaps in indirect
quotations all expressions are unquoted. For example, suppose that I work for Jan,
Jan tells me “The company is hiring an immigrant from Greenland”, and I know that
the only local Greenlander is your godchild, whose name escapes me. Then I may
report to you, my friend:
(46) My boss said that Crapco is hiring your godchild.
5
Note for non-American readers: every US schoolchild is exposed to images of the Declaration of
Independence, with its florid calligraphy, and learns its text: “When in the course of human events,
it becomes necessary for one people to dissolve the political bands ... We hold these truths to be
self-evident, that all men are created equal, ... with certain unalienable rights, that among these are
Life, Liberty, and the Pursuit of Happiness.”
This report would be felicitous even if Jan has no knowledge of you or of the
immigrant’s baptismal history. When I report another’s speech I may characterize
it, an object language, by means of my metalanguage, with my vocabulary and
concepts – Jan needn’t even know what a godchild is, and still my assertion (46)
would be correct.6 Everything in an indirect quotation therefore stands in for what
the source says (even in the vacuous case where the reporter and source use identical
words). Thus, (46) is equivalent to (460 ), where the entire quasi-quotation replaces
what the boss strictly said:
(460 ) My boss said p Crapco is hiring your godchildq .
Going further, it should be added that in indirect quotations indexicals are obliga-
torily unquoted. Even if the source message were “I’m hiring your godchild” – as
said by my boss directly to you while you were visiting me at work – the possessive
pronoun in my report (46) would not be a blind repeat, but independently chosen to
fit its new context of evaluation.
The transparency of speech reports is particularly stark in the case of dys-
phemistic epithets. Suppose Juan says, “I’m going to sue my landlord.” Then I may
report his speech by using direct quotation (47), indirect quotation of a neutral sort
(48), or indirect quotation that specifically reflects my own dysphemistic point of
view (49):
(47) Juan said, “I’m going to sue my landlord.”
(48) Juan said that he’s going to sue his landlord.
(49) Juan said that he’s going to sue the bastard.
In the latter case, the epithet is lifted out of the quotation, so to speak: it is reported
only as belonging to my language. To be sure, the epithet may very well belong to
Juan’s language too, but that’s not how it is reported unless I signal as much by
adding echoic vocal tone or quotation marks:
(50) Juan said that he’s going to sue “the bastard”.
To see that the epithet necessarily expresses the reporter’s point of view, consider
that Juan’s landlord could report any of (47–50) except for (49). In short, the
use of an epithet inside an indirect quotation falls squarely within the reporter’s
language, placing it in the metalanguage rather than the object language. (For more
on metalinguistic epithets, see Dubinsky and Hamilton 1998.)
3.3 Hybrid Speech Reports (“Mixed Quotation”)
Mixed quotation includes elements of both direct and indirect speech:
6
Elsewhere I’ve argued that opaque readings are possible in more syntactic frames than generally
acknowledged (2010, 2018), and here I emphasize that transparent readings are too.
54 P. Saka
(51) Bob Dylan said that he continues his music career because “he made a
vow years ago”. [Maier 2014: 13]
Interpreted as a strict quotation, (51) tells us that Dylan used the word “he” and
thereby indicated someone other than himself. Interpreted as a quasi-quotation, (51)
amounts to:
(510 ) Bob Dylan said that he continues his music career because “[he] made a
vow years ago”.
In this case, “he” is to be evaluated from the point of view of whoever reports (510 ),
referring e.g. to Dylan.
Chung-chieh Shan (2010), committed to the view that only syntactic constituents
can be quoted, interestingly proposes that apparent exceptions are actually cases of
cryptic quasi-quotation. Thus, the underlying structure of (52) is that of (520 ):
(52) David said that he had donated “largish sums to several benign
institutions”.
0
(52 ) David said that he had “[donated] largish sums to several benign
institutions”.
This kind of analysis is extended by Maier (2014) and sympathetically entertained
by Koev (2017).
3.4 Cited Forms (“Pure Quotation”)
Sometimes we can cite expressions, via quasi-quotation, to make intuitively satisfy-

ing claims:
(53) “If blah-blah then blah-blah-blah” is a conditional.

Often, however, we cannot. The following, for instance, will strike many as odd or
unacceptable:
(54) “Some viruses infect other viruses ... They are known as virophages”
contains an unknown number of clauses.
(55) “[t]he octopus has three hearts” begins with a capital letter (as its first
letter).
In each case, I shall argue, the display item is actually equivocal: the immediate
interpretation, stemming from the words perceptually before our very eyes, is false;
and a more distant interpretation is actually true. The immediate interpretation gives
us strict quotation, and when read this way (54) and (55) are not only false but
manifestly false and therefore odd. When read as quasi-quotations, however, they
can be paraphrased as:
(540 ) Some text beginning with “Some viruses infect other viruses” and
proceeding with something unspecified and ending with “They are
known as virophages” contains an unknown number of clauses.
0
(55 ) Some text beginning with a modification of “t” and then proceeding with
“he octopus has three hearts” begins with a capital letter.
A more complicated case is:
(56) It’s not true that “so-and-so” contains vowels.
Reading (56) as a quasi-quotation leaves it still in need of further interpretation.
“So-and-so”, understood as an unquotation, has no particular referent outside of
context any more than pronouns do, and thus lacks both consonants and vowels.
This remains so in instances of direct quotation:
(57) It does not scream “Look at me! I am the work of such-and-such a great
name, an Ozymandias of architecture!” [COCA]
This can roughly be paraphrased with a quantifier, either wide-scope (570 ) or narrow
(5700 ):
(570 ) For any great name x, it does not scream “Look at me! I am the work of
x!”
(5700 ) It is not true for any great name x that it screams “Look at me! I am the
work of x!”
Here “x” functions as a variable. Quasi-quoting it does not secure a determinate
referent that can either have or lack vowels, and I submit that the same point applies
to “so-and-so” in (56).
If I am correct that there exist interpretations by which (54)–(56) are true, how
comes it that the principle of charity does not induce everyone to seize first upon
the true interpretations, instead of fixating on those readings that render (54)–(56)
false? Perhaps some are initially drawn to the quotational readings of (54)–(56)
because it is under these readings that evidence for the truth or falsehood of (54)–
(56) is glaringly available. By mere inspection of the item itself we can tell whether
(54)–(56) are true or false, if we adopt strict quotational readings. When we adopt
quasi-quotational readings, however, our judgments of truth-value must be more
empirical and less certain, which are psychological deterrents. Also: if you really
wanted to say that “The octopus has three hearts” begins with a capital letter, you
can do so more directly and more cooperatively as I have just done it, rather than by
the roundabout means of (55).
3.5 Non-quoted Forms
The quasi-quotational devices we have seen so far can all be used outside of
quotation. For example:
56 P. Saka
(58) I met this lawyer, we went out to dinner, I had the lobster bisque, we
went back to my place, yada yada yada, I never heard from him again.
–Elaine [tv episode of Seinfeld]
Elaine uses “yada yada yada” to avoid divulging certain details, perhaps of her sex
life. It is not that she is uninterested in her own sex life, but she is uninterested in
talking about it.
Blah also occurs outside of all manner of quotation (not direct, indirect, or
mixed), allusion, and sarcasm, as do so-and-so and such-and-such:
(59) They have some kind of special coating that makes them anti-scratch,
anti-fog, anti-UV, anti-blah blah blah. [COCA]
(60) Up until the day my world shattered, 5 years, blah-blah months,
so-and-so days, and whatchamacallit hours ago, I was a deputy with the
Sarasota Sheriff’s Department. [COCA]
(61) An electron “exists” as a mixture of possible states until you make a
measurement, whereupon the wave function “collapses” and, bingo,
there’s one reality – and the electron is in such-and-such a state. [COCA]
In (60), “so-and-so” stands in for a number of days. It is a cliche that if you are
recovering from a trauma, you know to the exact day how long ago it happened,
but in real life harrowing experiences typically render one dazed with regard
to the precise passage of time. Thus repudiating our cliche, the author of (60)
artfully expresses her own mental fuzziness while simultaneously acknowledging
the countability of hours elapsed (hence hinting at their sluggishness and therefore
her unhappiness); meanwhile “blah-blah”, expressing uninterest, adds to the sense
that she is depressed. There is no quotation here, and hence no quasi-quotation.
The preceding examples seem to me to count as evidence against the analysis
proposed by Koev (2017). According to Koev, (62) means (620 ), where italics
indicate presuppositional content:
(62) Maria is dating so-and-so.

(620 ) There is an expression x such that:
Maria is dating the denotation of x and someone has uttered x.
The problem is that I don’t think any of the examples in this subsection presuppose
prior utterances. In saying “yada yada yada”, Elaine avoids constructing a specific
narrative, and so it can hardly be said that it exists for her now to reference. In
saying “so-and-so” (days ago), the former deputy sheriff references a numeral that
is unknown to her, so it too is implausible as a candidate for having been previously
uttered. In saying “such-and-such” (electron state), the physicist refers to a generic
electron, so there’s no way that any prior utterance could single out its state; and
even putting that objection aside, our physicist may refer to a state that is not only
unknown but unknowable, making any assertoric reference to its description highly
unlikely.
4 A Speech-Act Approach
In this section I will try to specify rules of quasi-quotation, building upon my

earlier speech-act account (Saka 1998, 2005, 2011). My earlier work starts by
distinguishing between use and mention:
(U) Speaker S uses expression-string x iff S produces a token of x, intending

to refer to x’s conventional denotation.
(M) S mentions x iff S produces a token of x, intending to refer to x.
Quotation can then be explained as follows:
(Q) S quotes x iff S signals that S is mentioning x.

Along with these old ideas, and my appeals (2005, 2007) to prototypes and
modularity (quick, automatic, reflexive processing), I should perhaps make explicit
the obvious implication that (U) and (M) do not enjoy parity. Normally when
speakers produce tokens, the default interpretation is that the tokens are used, not
mentioned.
In analyzing quotations as references to expressions, I do not mean that quo-
tations refer only to expressions. When I write (63), I both refer to the words in
quotation marks and use them to make an assertion (see also Partee 1984):
(63) As Frank Zappa said, “Without deviation from the norm, progress is not
possible.”
The question of how exactly to integrate the use and mention aspects of a quotation
haunts all theories of quotation; but assuming that any solution that works for strict
quotation will apply to quasi-quotation, for present purposes I put the question aside.
Although my account of strict quotation does not resolve all problems (Saka
1998: 131, elaborated by Gomez-Torrente 2017), I believe it is more successful
than its rivals. I accept it, therefore, as the basis for a refined and extended account
that treats quasi-quotation. As for incidental refinement, I would revise (U) to
accommodate metaphorical speech:
(U*) Speaker S uses expression-string x iff S produces a token of x, intending

to refer to x’s conventional denotation ’ (literal use) or something similar
to ’ (metaphorical use).
I would also revise (U) to accommodate metonymy, hyperbole, and so forth.
Mentioning itself is probably best seen as one species of figurative speech, but
pursuing that idea would take us from the topic at hand.
Extending my speech-act account to quasi-quotation begins with a definition of
quasi-mention:
(MC) S quasi-mentions x iff S produces a token of x, intending to refer to an

admissible substitution of x.
58 P. Saka
Admissibility is the key notion here, and is developed in some detail below. It gives a
handle on quasi-mentioning which in turn feeds into a definition of quasi-quotation:
(QC) S quasi-quotes x iff S signals that S is quasi-mentioning x.

Signaling can be achieved in a variety of ways. First, given that corner quotes are
established notation in logic:
(a) There is a convention whereby if S encloses x in a pair of corner quotes,

S signals that S is quasi-quoting x.
Second, given the practices that prevail in American English:
(b) There are conventions whereby if S encloses x in a pair of

double-apostrophes, italicizes x, specially indents x, or speaks with
finger quotes, S signals that S is either quoting or quasi-quoting x.
Third, even without conventional marks of quotation, quotational intent can be
conveyed by the natural iconic principle “fit the medium to the message”:
(c) If S intentionally sets x off from surrounding text in any way (e.g. by
capitalization or font change or voice change), S signals that S is treating
x in some marked manner (e.g. is emphasizing x or is distancing self
from x or is quoting x).
Fourth, quotations can be signaled by stipulation, for instance if a defective
keyboard leads me to announce that I will mark quotations by means of asterisks.
Stipulations too rely on a natural (universal, convention-independent) principle:
(d) If S announces the intention to use x to signify ’, then when using x S

does signal the intention to signify ’.
Fifth, even when stipulative intentions are not announced they can sometimes
be inferred. When logic textbooks stipulate that Greek letters will serve in the
metalanguage to range over Roman letters in the object language £, then the Greek
presence in (10) will signal, on grounds of charity, (1000 ):
(10) If ® and § are formulas in £ then (® v §) is a formula in £.

(1000 ) If ® and § are formulas in £ then p (® v §)q is a formula in £.
This is a pattern that runs through textbooks, and it can be codified:
(e) If a logician S uses a metavariable in a formula, S signals that S is

quasi-quoting the formula.
Rules (a-e), like linguistic rules more generally, are defeasible.
My account is a bit messy, and will get messier, because English is messy.
Distinct marks of quotation have evolved that occupy niches that are nearly though
not exactly co-terminous, and likewise for the various devices of quasi-quotation.
With minor differences supported by national customs and usage barons (e.g.,
whether commas should appear before a closing quotation mark or after), it is

inevitable that any empirically adequate account of quotation will necessarily have
separate clauses that differ only in boring arbitrary culture-specific detail. That said,
my account possesses a bit more elegance than is evident from the ten principles
offered so far (U*, M, MC, Q, QC, a-e). For example, my account will ultimately
treat quotation as a special case of quasi-quotation and not a separate category,
which simplifies the disjunction at the end of (b).
And now to sketch, all too briefly, some criteria for admissible substitution.
Criterion 1 is a tautology; applying to all contexts and all entities x and y, its scope
goes well beyond just quotation, and just linguistic use:
#1. If x D y then x can substitute for y.

This justifies my comment above that quotation is a special case of quasi-quotation;
(Q) is not an independent principle, but subsumable under (QC).
The remaining criteria apply to quotation contexts in which x and y range over
expressions. Criteria 2 and 3 apply to formal regimented languages:
#2. If x contains metalinguistic terms that schematize sentences in an object

language, the former can be replaced by the latter.
#3. If x contains metalinguistic terms that schematize names of sentences in
an object language, the former can be replaced by the sentences named.
These rules clarify that the difference between sentential quasi-quotation and
referential quasi-quotation has nothing to do with (MC) or (QC), and everything
to do with the choice of metavariable. Sentential metavariable “®” takes different
values from referential “®”, yet sentential p ®q D referential p ®q .
Criterion 4 allows us to use translations in our quotations:
#4. If x is a translation of y then x can substitute for y.

Typically it would be fair to translate Latin to English:
(64) Caesar said, “veni, vidi, vici.”

(65) Caesar said, “I came, I saw, and I conquered.”
And because translation is inexact, it would be fair to take the English above back
to Latin as:
(66) Caesar said, “veni, vidi, et vici.”
But if the source is exactly as (64) says, it would normally be misquotation to report
(66). Hence substitution is not transitive, as further evinced by the bon mot that
sometimes a shower can substitute for sleep, and sometimes food can substitute for
sleep, but food can never substitute for a shower.
The remaining criteria target quasi-quotational devices, i.e. expressions that are
customarily used for quasi-mentioning. For example, criterion 5 allows reporters to
introduce phonemic euphemisms and codes into their quotations:
60 P. Saka
#5. If x D y, except that a taboo expression inside x is phonemically/

orthographically modified to pass the censors, then x can substitute for y.
Criteria 6–8 allow us to use brackets and ellipses in our quotations:
#6. If x D y, except that x contains interpolated material marked by brackets,

then x can substitute for y.
#7. If x D y, except that a trivial change has been made in x and marked by
brackets, then x can substitute for y.
#8. If x D y, except that x omits material in the places marked by ellipses, in
such a way that x preserves the gist of y, then x can substitute for y.
To see how my apparatus works, suppose that we have expressions x and y:
x He [Kepler] was persecuted

y He was persecuted
Then x D y, except that x contains interpolated material marked by brackets. Thus,
by #6 x is an admissible substitution for y; hence, by (MC), intending to refer to y
by means of x counts as quasi-quotation; and finally, by (QC), this intention can be
signaled, or made public, by means of putting quotation marks around x, as done in
(2).
Each quasi-quotational device in English, like any construction in any language,
is attached to its own specific conditions of use. For instance, a complete theory will
indicate that “so-and-so” is preferentially used for persons, and in nominal position;
that “such-and-such” is preferentially used for things, and in nominal or pre-noun-
phrase position; and that “blah blah blah blah” expresses greater dismissal on the
part of the speaker than “blah blah” (reduplicating to express intensity, a cross-
linguistic tendency, is another instance of iconicity). Differences across individual
quasi-quotational devices also point to the spectrum nature of quotation, ranging
from zero difference between source and what is reported (verbatim quotation) to
almost unlimited difference. To put it crudely, as though there were only a single
dimension: strict quotation in legal circles approaches the verbatim end (without
completely reaching it); strict quotation by journalists falls short of that, and
informal strict quotation falls shorter still; the quasi-quotational use of “so-and-so”
is at the middle, signaling source material that is not only absent but not rationally
recoverable; and ellipses are at the opposite extreme, representing deleted source
material of virtually any quantity and quality whatsoever (cp. Gregoromichelaki
2017: sect. 6).
The adduced criteria do not provide necessary or sufficient conditions. To see
that not one of them is necessary for triggering a quasi-quotational reading, consider
the unquotational use of “more” in (45); or consider that (67) appears to include a
clear case of quasi-quotation, even though it is not licensed by any of my proposed
criteria:
(67) She said he was a “pain in the neck”, only she didn’t say “neck”. (Clark
and Gerrig 1990)
Indeed, there seems to be no limit to the number of culture-specific, context-specific

nonce situations where intelligent agents can reasonably guess that proper quasi-
quotation is intended:
(68) The first time you log in to our system, your password will be your date
of birth, “MM/DD/YYYY”.
In this case the password is neither a series of capital letters nor even one’s date
of birth, expressed in any old way, but rather it is date of birth rendered in one
unfortunate unforgivingly specific format.
To see that my criteria are not sufficient for enforcing a quasi-quotational reading,
keep in mind that all of my examples of quasi-quotation permit strict-quotation
readings. Various techniques are used for overcoming ambiguity, but they are not
fool-proof. For example, an author may systematically put brackets around ellipses
when the ellipses are quasi-quoted (representing the reporter’s omission), and use
unbracketed ellipses for strict quotation (representing the source’s omission). This
and other nuanced techniques can reduce interpretive uncertainty in practice, but in
principle they always fail. For one thing, readers who do not realize that a given
reporter is following a certain convention can’t be sure how to interpret a quotation.
For another, whenever an original source includes brackets around ellipses, a strict
quotation of it will contain brackets without quasi-quotational significance. Finally,
there is no way around interpretive indeterminacy because linguistic norms are
largely or entirely defeasible. I have learned, for instance, that some folk have a very
loose sense of exactitude, and would accept even the following as a quasi-quotation.
(69) The exact words that Jan said to me are these: “Today is
you-know-who’s birthday.”
The adduced criteria identify norms, not truth-conditions. They contribute to our
judging whether a given quotation is fair, honest, or assertable, within a specified
context, without fully addressing whether it be true. If Dick says “I deny that I am
a crook”, is it true that he said “... I am a crook”? Is it true that he said “I am a
crook”? Is it true that he said “ich bin kein Krimineller” (I am no crook)? Such
questions miss the mark because linguistic meaning underspecifies truth-conditions
(as emphasized by Lakoff 1987, Atlas 2005, Saka 2007, Recanati 2010, and others).7
What all of this shows is what I have argued for elsewhere: that natural language
is not a set of abstract formal objects possessing intrinsic referential properties.
Formal modeling is desirable, but instead of focusing on truth-conditions, the theory
of meaning needs to attend to speaker intentions, which necessitates the modeling
of context, perceived relevance, and interlocutor interaction. Ultimately, meaning,
interpretability, and even syntactic permissibility are determined by intelligent
7
According to Gomez-Torrente (2005: 144), bracketed material and ellipses “do not make any sort
of truth-conditional contribution” to the quotations they are in. Surely, however, they affect what
thought gets expressed by an assertion, and it is this thought that is relevant when we consider
linguistic meaning.
62 P. Saka
agency and not inert forms. This is not because intelligent agency is a spiritual or
magical force forever beyond the ken of scientific understanding – on the contrary it
is the formalist approach, with its implied platonism, that departs from naturalism –
but because it is implausible to think that the human brain, with its hundred-trillion
neural connections contingently evolved in a social environment, operates according
to the classical computational principles assumed by the formalist paradigm.8
References
Anderson, L., & Lepore, E. (2013). Slurring words. Nous, 47, 25–48.
Atlas, J. (2005). Logic, meaning, and conversation. Oxford: Oxford University Press.
Clark, H., & Gerrig, R. (1990). Quotations as demonstrations. Language, 66, 764–805.
Dubinsky, S., & Hamilton, R. (1998). Epithets as antilogophoric pronouns. Linguistic Inquiry, 29,
685–693.
Field, H. (2001). Truth and the absence of fact. Oxford: Oxford University Press.
Forbes, G. (1994). Modern logic. Oxford: Oxford University Press.
Garcia-Carpintero, M. (2017). Reference and reference-fixing in pure quotation. In P. Saka & M.
Johnson (Eds.), The semantics and pragmatics of quotation (pp. 169–194). Cham: Springer.
Gomez-Torrente, M. (2005). Remarks on impure quotation. In P. De Brabanter (Ed.), Hybrid
quotations (pp. 129–151). Amsterdam: John Benjamins.
Gomez-Torrente, M. (2017). Semantics vs. pragmatics in impure quotation. In P. Saka & M.
Gregoromichelaki, E. (2017). Quotation in dialogue. In P. Saka & M. Johnson (Eds.), The
semantics and pragmatics of quotation (pp. 195–256). Cham: Springer.
Grover, D., & Belnap, N., Jr. (1973). Quantifying in and out of quotes. In H. Lablanc (Ed.), Truth,
syntax and modality. Cham: Reidel. Reprinted in Grover, A Pro-sentential theory of truth,
Princeton University Press, 1992.
Jaszczolt, K. M. (2005). Default semantics, foundations of a compositional theory of acts of
communication. Oxford: Oxford University Press.
Jaszczolt, K. M., & Huang, M. (2017). Monsters and I: The case of mixed quotation. In P. Saka &
M. Johnson (Eds.), The semantics and pragmatics of quotation (pp. 357–382). Cham: Springer.
Johnson, M. (2017). Quotation through history. In P. Saka & M. Johnson (Eds.), The semantics and
pragmatics of quotation (pp. 281–302). Cham: Springer.
Koev, T. (2017). Quotational indefinites. Natural Language & Linguistic Theory, 35, 367–396.
Lakoff, G. (1987). Women, fire, and dangerous things. Chicago: U. of Chicago Press.
Lehrer, A. (1989). Remembering and representing prose. Discourse Processes, 12, 105–125.
Ludwig, K., & Ray, G. (2017). Unity in the variety of quotation. In P. Saka & M. Johnson (Eds.),
The semantics and pragmatics of quotation (pp. 99–134). Cham: Springer.
Maier, E. (2014). Mixed quotation: the grammar of apparently transparent opacity. Semantics and
Pragmatics, 7, 1–67.
Maier, E. (2017). The pragmatics of attraction: Explaining unquotation in direct and free indirect
discourse. In P. Saka & M. Johnson (Eds.), The semantics and pragmatics of quotation
(pp. 259–280). Cham: Springer.
8
A version of this material was presented at the Semantics and Philosophy in Europe conference
(Bochum, 2010). I am grateful to its organizers and audience, and to Stan Dubinsky, Michael
Johnson, Julia Jorgensen, and anonymous reviewers for comments. This work is dedicated to the
memory of Laurence Goldstein.
Mates, B. (1972). Elementary logic, 2d edn. Oxford: Oxford UP.

Manicas, P., & Kruger, A. (1968). Essentials of logic. New York: American Book Company.
Marciszewski, W. (1981). Dictionary of logic as applied in the study of language. Cham: Springer.
McKay, T. (1989). Modern formal logic. New York: Macmillan.
Partee, B. (1984). Compositionality. In F. Landman and F. Veltman (Eds.), Varieties of Formal
Semantics (pp. 281–312). Cham: Foris Publications.
Pullum, G. (1984). Punctuation and human freedom. Natural Language & Linguistic Theory, 2,
419–425.
Quine, W. (1961). Mathematical logic (revised ed.). Cambridge, MA: Harvard University Press.
Recanati, F. (2010). Truth-conditional pragmatics. Oxford: Oxford University Press.
Reichenbach, H. (1947). Elements of symbolic logic. New York: Macmillan.
Saka, P. (2005). Quotational constructions. In P. De Brabanter (Ed.), Hybrid quotations (pp. 187–
212). Amsterdam: John Benjamins.
Saka, P. (2007). How to think about meaning. Cham: Springer.
Saka, P. (2010). Rarely pure and never simple: Tensions in the theory of truth. Topoi, 29, 125–136.
Saka, P. (2011). The act of quotation. In E. Brendel, J. Meibauer, & M. Steinbach (Eds.),
Understanding quotation (pp. 303–322). Berlin: De Gruyter Mouton.
Saka, P. (2018). Superman semantics. In A. Capone, F. Lo Piparo, & P. Perconti (Eds.), Further
advances in pragmatics and philosophy. Cham: Springer.
Schiller, F. (1912). Formal logic. London: Macmillan.
Shan, C.-C. (2010). The character of quotation. Linguistics and Philosophy, 33, 417–443.
Smullyan, R. (1962). Fundamentals of logic. Englewood Cliffs: Prentice-Hall.
Sudo, Y. (2013). Metalinguistic quantification: Evidence from Japanese wh-doublets. http://
web.mit.edu/ysudo/www/pdf/quotation5.pdf. Accessed 1 July 2016.
Use-Mention Confusions in “Sloppy, Colloquial
Speech”
Marga Reimer
Abstract Although commonplace in both ordinary and philosophical discourse,

the use/mention distinction continues to generate controversy among contemporary
analytic philosophers. In what follows, my central aim will be to clarify and
defend Saul Kripke’s (Naming and necessity. Harvard University Press, Cambridge,
MA, 1980) parenthetical remarks regarding the confusion of use and mention in
ordinary (“sloppy, colloquial”) discourse involving proper names. In so doing, I
will be responding to two original and provocative papers that address Kripke’s
stance on this common colloquial confusion: one by Delia Graff Fara (Analysis
72(3):492–501, 2011) and a follow-up paper by John Biro (Analysis 71(2):492–501,
2012). Although Fara and Biro are united in both their interpretation and negative
assessment of Kripke’s remarks, their own positive views are importantly different.
I will suggest below that Kripke’s views, properly understood, are considerably
more plausible than the views of either Fara or Biro. However, I will also raise
some concerns, in the concluding section of the paper, with the blanket suggestion
that “sloppy, colloquial” and otherwise informal discourse is, generally speaking,
irrelevant to the assessment of semantic theories of ordinary language.
1 Saul Kripke
In Naming and Necessity, Kripke writes:

Sloppy, colloquial speech, which often confuses use and mention, may, of course, express
the fact that someone might have been called, or not have been called, ‘Aristotle’ by saying
that he might have been, or not have been, Aristotle. (1980: 62 n. 25)
The foregoing occurs as a parenthetical comment within a lengthy footnote, wherein

Kripke defends the view that names are rigid designators against objections that
appeal to “sloppy, colloquial speech” – speech which, according to Kripke, “often
M. Reimer ()
University of Arizona, Tucson, AZ, USA
e-mail: reimer@email.arizona.edu

https://doi.org/10.1007/978-3-319-68747-6_3
66 M. Reimer
confuses use and mention.” Thus, for example, in colloquial English one may
express (1) by saying (2):
(1) Trump might not have been called ‘Trump’.

(2) Trump might not have been Trump.
While (1) involves one use and one mention of the name ‘Trump’, two uses of
that name are involved in (2). The putative “confusion” lies in the fact that the
speaker utters a sentence (with the appropriate sort of intentions) and thereby
mentions a name that is only used in that sentence. Her speech thus lacks the kind
of transparency that it would have had, had she uttered (1) rather than (2).
This sort of colloquial confusion might occur naturally within the context of
a conversation concerning the Anglicization of Donald Trump’s ancestral name.
Perhaps someone suggests that widespread discrimination against German immi-
grants in late nineteenth century America may have prompted Trump’s paternal
grandfather to change his surname from ‘Drumpf’ to ‘Trump’. Thus, had such
discrimination never occurred, Trump might not have been Trump; he might have
been Drumpf. Less “sloppily,” Trump might not have been called ‘Trump’; he might
have been called ‘Drumpf’.
Importantly, the fact that in colloquial speech one might express (1) by saying (2)
does not, according to Kripke, mean that (2) itself is true and that the rigidity thesis
is thereby undermined. Indeed, if Kripke is right and names are rigid designators,
then (2) is necessarily false.
Kripke provides a helpful mathematical analogy to make the point vivid. Just as
Aristotle might not have been called ‘Aristotle’, 2 2 might not have been called
‘four’. But of course this does not mean that 2 2 might not have been four! Yet in
“sloppy, colloquial speech” one might express (3) by saying (4).
(3) 2 2 might not have been called ‘four’.

(4) 2 2 might not have been four.
The suggestion that colloquial occurrences of sentences like (2) and (4) provide
counterexamples to the rigidity thesis is no mere strawman argument – at least
Kripke doesn’t think that it is. Indeed, in the passage that immediately follows the
one cited above, Kripke writes:
Occasionally, I have heard such loose usages adduced as counterexamples to the applicabil-
ity of the present theory [that names are rigid designators] to ordinary language. (1980: 62,
n.25)
Kripke is not impressed by such “counterexamples”:

Colloquialisms like these seem to create as little problem for my theses as the success of the
‘Impossible Missions Force’ [a fictional espionage agency] creates for the modal law that
the impossible does not happen. (1980: 62, n.25)
Despite the hyperbole, Kripke’s point is a credible one. In particular, the fact
that in colloquial speech one might, for example, express (1) or (3) by saying (2) or
(4), does not appear to undermine the thesis that names, as they occur in ordinary
Use-Mention Confusions in “Sloppy, Colloquial Speech” 67
language, are rigid designators. The Kripkean view that sentences like (2) and (4)
are necessarily false would thus appear untouched by such “sloppy” colloquialisms.
Similarly, while saying (5) in the appropriate sort of context might allow one to
express (6), this is no objection to the claim that (5), an ordinary language sentence,
is a necessary truth.
(5) Trump is Trump.

(6) Trump will do whatever he wants.
To his philosophical adversaries who think otherwise, Kripke might reasonably
reply: You are taking “sloppy, colloquial” language at face value and, in so doing,
are preserving its confusions in your purported counterexamples.
Reflection on Russell’s theory of descriptions is useful in appreciating Kripke’s
point regarding the irrelevance of “sloppy, colloquial speech” to the assessment the
semantic theses of Naming and Necessity. Despite its checkered history, Russell’s
theory, which purports to explicate the meaning (the “logical form”) of natural
language sentences containing definite descriptions, cannot be dispensed with
simply by pointing out that an utterance of (7) might express a truth in a world
containing a multiplicity of cats and mats.
(7) The cat is on the mat.

Much the same point might be made with regard to Kripke’s rigidity thesis
and sentences like (2) or (5). Such sentences might be used, in “sloppy, colloquial
speech,” to express (1) or and (6), yet it would be a mistake to conclude, without
further ado, that Kripke’s thesis is false. We now have the background necessary for
an informed appraisal of the thought-provoking commentaries of Fara (2011) and
Biro (2012). Let’s begin with the former.
2 Delia Graff Fara
Fara (2011) begins her paper by noting that, not only do “most of us know” that
Quine was the author of Word and Object, “most of us also know” that (8) is true:
(8) The author of Word and Object was called Willard.

Referencing the footnote from Naming and Necessity discussed above, Fara
claims that Kripke would take issue with the latter observation. As she explains:
Saul Kripke would complain that although what I meant to convey with [8] is true, I
expressed that truth in “sloppy, colloquial speech”. He would say that I made the horrifying
philosophical blunder of confusing use and mention: in using the phrase ‘was called
Willard’, I used the name ‘Willard’ where I should have mentioned it instead. The proper
statement of that truth should be written, he would claim, as : : : : [9] The author of Word
and Object was called ‘Willard’. (2011: 492)
68 M. Reimer
Fara then continues:

Kripke’s comment makes it seem as though we have to choose between [8] and [9] : : : But
we don’t have to choose. Neither [8] nor [9] involve a use-mention confusion. Not only are
[8] and [9] grammatical, they mean different things.
There are four points to consider here: (i) Fara’s interpretation of Kripke’s
parenthetical remarks, (ii) her claim that neither (8) nor (9) involve a use-mention
confusion, (iii) her claim that both sentences are grammatical, and (iv) her claim
that the two sentences mean different things. Let’s consider these four points in
turn, beginning with (i).
The first thing to note is that, while Fara’s examples involving writing (inscrip-
tions of sentences) Kripke’s remarks concern speech (utterances of sentences).
However, we can easily accommodate this difference by talking more broadly in
terms of “discourse.” Thus, if we are faithful to the spirit of Kripke’s parenthetical
remarks, perhaps we can attribute to him the view that in “sloppy, colloquial”
discourse we may express (9) by writing (inscribing) (8). This would involve a use-
mention confusion insofar as the speaker inscribes a sentence (with the appropriate
sort of intentions) and thereby mentions a name that is only used in that sentence.
I think the extension from speech to discourse more generally is a natural one,
especially considering the ubiquity of texting. Texting is rife with colloquial writing,
full of “loose usages,” and perhaps a bit “sloppy,” as well, given its notoriously
minimalist punctuation and phonetic spelling. However, I do not think (pace Fara)
that Kripke’s parenthetical remarks suggest that he would accuse ordinary speakers
who inscribe sentences like (8) of confusing use and mention – not unless they
insisted that such inscriptions undermined the rigidity thesis or some other semantic
theory of ordinary language. It is, after all, “sloppy, colloquial” speech against which
Kripke inveighs in the footnote to which Fara appeals. Speakers run the risk of
committing “horrifying philosophical blunders,” including use-mention confusions,
when they take such speech at face value and appeal to it in arguing for, or against,
semantic theories of ordinary language. In this way, they internalize and inherit the
confusions that characterize their “sloppy, colloquial speech.”
Let us now shift from Kripke’s view of sentences like (8) and (9) to Fara’s view
of such sentences. This brings us to point (ii) above: Fara’s claim that neither (8)
nor (9) confuse use and mention.

(9) The author of Word and Object was called ‘Willard’.
Although I agree with Fara in denying any use/mention confusion in an inscrip-
tion of (9), I also think (contra Fara) that (8) involves an implicit mention of the
name ‘Willard’ that could easily be made explicit by way of an inscription of (9).
Perhaps the inexplicitness of (8) might be thought (in the spirit of Kripke) to involve
a kind of use/mention confusion insofar as “sloppy, colloquial” English allows one
to express the name-mentioning (9) by inscribing the name-using (8).
What now of (iii): Fara’s claim that both (8) and (9) are grammatical? I think
Fara is right here but would add that (contra Fara) the two sentences mean the
same thing and are used by speakers to communicate the same thing: that which (9)
makes explicit. Thus, (9) makes explicit what is only implicit in (8): that ‘Willard’
is being mentioned rather than used. More generally, (10) and (11) are synonymous,
although the latter is more explicit than the former:
(10) The F is (was) called N.

(11) The F is (was) called ‘N’.
Similarly, sentences like (12) and (13) are both grammatical and mean the same
thing, although the latter is a more explicit than the former insofar as it makes
explicit that the authored work is a text (a book).
(12) The author of word and object was called Willard.

Finally, what of point (iv): Fara’s claim that (8) and (9) mean different things?1
As Fara explains:
To be called Willard is to have ‘Willard’ as a name. To be called ‘Willard’ is for someone
to address you or refer to you using that name : : : You can be called Willard without being
called ‘Willard’, and you can also be called ‘Willard’ without being called Willard : : :
More generally (says Fara):

You can have names that no one ever addresses you by, and you can be addressed by names
other than your own.
Although I agree with the foregoing generalization (as does Kent Bach 2002),
I am less confident about the remarks that precede it, which appear to blur an intu-
itive and theoretically important distinction between naming and calling. I would
instead say (employing conventional punctuation devices to signal mentioning):
To be named ‘Willard’ is to have ‘Willard’ as a name. To be called ‘Willard’ is for someone
to address you or refer to you using that name : : : You can be named ‘Willard’ without being
called ‘Willard’, and you can also be called ‘Willard’ without being named ‘Willard’.
In this way, naming and calling are kept distinct. But what exactly is the distinction
between the two notions?
The crucial distinction between calling and naming (if one is being precise) is
that while calling essentially involves vocalization, naming does not. This sort of
view comports well with the etymology of the verb ‘to call’ which derives from the
Old Norse, ‘kalla’, meaning to summon loudly. More importantly, the verb’s origin
is patent in its various definitions and ordinary usages. Dictionary.com provides and
illustrates 28 definitions of the transitive verb ‘call’! Virtually all of them involve
1
One possible response to Fara’s point, different from that proposed here, would be to claim that
Fara’s mistake is in treating (8) and (9) as sentences. One might instead claim that a string like (8)
is ambiguous between two sentences: one in which Willard is used and one in which Willard is
mentioned. While the first of these leads to an absurd interpretation, the second one leads exactly
to the same interpretation as (9). Thanks to a reviewer for this point.
70 M. Reimer
vocalization. Below is a representative sampling of those definitions, followed by

various illustrations and clarifications provided on the website. They are numbered
exactly as they are in Dictionary.com.
1. To cry out in a loud voice; shout: He called her name to see if she was home.
2. To command or request to come; summon: to call a dog; to call a cab; to
call a witness.
3. To ask to invite to come: Will you call the family to dinner?
4. To communicate or try to communicate with by telephone: Call me when you
arrive.
5. To rouse from sleep, as by a call; waken: Call me at eight o’clock.
6. To read over (a roll or a list) in a loud voice.
15. To attract or lure (birds or animals) by imitating characteristic sounds.
17. To name or address (someone) as: His parents named him James, but the
boys call him Jim.
This last example resonates with me as I have spent half a century explaining my
name(s) by saying things like:
(14) I was named Margaret after my grandmother but everyone calls me Marga
or (15):
(15) My real name is Margaret but everyone calls me Marga.2

I like these sorts of examples because they highlight the important, if subtle, con-
trast between naming and calling. While calling essentially involves vocalization,
naming might be effected by simply inscribing (or otherwise entering) a name on
the appropriate sort of document, such as a birth certificate.
While it is true (as Fara notes) that we occasionally say things like:
(16) Quine was called Willard after an uncle,

such usages are arguably apt only in cases where the named individual is called
(addressed), at least occasionally, by the name she was given. To see this, just
consider (17):
(17) He is called James after his grandfather but everyone calls him Jim.
The natural response to such an utterance would be a puzzled (18):
(18) What do you mean? Does no one call him James?

which would be forestalled by an utterance of something like:
(19) He was named James after his grandfather but everyone calls him Jim.
2
As a reviewer pointed out, a nice example of this kind of remark occurs in “Rocky Raccoon” from
the Beatles’ White Album: Her name was McGill and she called herself Lil/But everyone knew her
as Nancy.
One might, of course, invoke the past tense and say (20) rather than (17):
(20) He was called James after his grandfather but everyone calls him Jim.
Yet this would be unnecessarily confusing, as it would trade on a questionable
ambiguity in the verb ‘call’, easily avoided by using ‘named’ in place of ‘called’
as in (19). Fara might respond that the ambiguity in (20) could be avoided simply
by enclosing the occurrence of ‘Jim’ in (20) in inverted commas. But while such a
device might prove useful in writing, it would be of no use in speech where it would
be silent.
The foregoing reflections are consistent with the idea that ‘called,’ as it occurs in
both (20) and Fara’s (8), refers to naming/dubbing/christening/baptizing, a one-time
event which is over and done with early in life, generally within a period of minutes
if not seconds. My preference, however, would be for ‘named’ (vs. ‘called’), given
its pervasiveness, naturalness, neutrality, and explicit differentiation from ‘called’
which, as I have been suggesting, essentially involves vocalization, consistently
with both its etymology and the myriad definitions and usages that reflect that
etymology.3
On this sort of view, calling (in contrast to naming) is akin to beckoning and
I therefore wonder if Fara might be wrong to dismiss phone calling as irrelevant
to a theory of calling more generally. Phone calling clearly does involve a kind of
vocalization, a kind of beckoning. When the phone rings continuously throughout
the day, I might well say, in exasperation, “I wonder who keeps calling me.” Clearly,
the notion of calling here is akin to that of vocalizing or beckoning – involving an
implicit “Are you there? I want to talk to you.” This sort of view comports well with
the etymology of the verb ‘call’ which, as noted above, derives from the Old Norse
‘kalla’, meaning to summon loudly. It is also supported by Dictionary.com’s fourth
definition of the transitive verb ‘call’:
4. to communicate or try to communicate with by telephone: Call me when you
arrive.
While it is perhaps true that ‘call’ is sometimes used (colloquially) to mean name,
taking this particular use of ‘call’ as theoretically important (as Fara appears to) has
the undesirable effect of blurring the distinction between naming and calling, where
the latter, in contrast to the former, is essentially linked to vocalization. This naming
vs. calling distinction is obvious in sentences like (19), just as it is in the sample
sentence provided in Dicionary.com’s 17th entry for the transitive verb ‘call’:
(21) His parents named him James, but the boys call him Jim.
3
This is certainly not to deny that ‘call’ is sometimes used synonymously with ‘name’ and thus
without any implication of vocalization. I might say that I decided to call my daughter ‘Ingrid’
meaning only that I decided to name her ‘Ingrid’. However, any such use would arguably be “loose”
and so not at odds with the proposed vocalized construal of ‘call’.
72 M. Reimer
Thus, contrary to Fara, I’m not so sure Quine could have been called Willard if no
one ever addressed him as Willard. Indeed, even if his mother had declared upon his
birth:
(22) I will call you Willard

she would not have thereby called Quine Willard; she would rather have named him
Willard. Only upon addressing Quine as Willard would it appropriate to say that
Quine was called Willard. Thus, if Quine’s mother had followed her utterance of
(22) with an utterance of (23):
(23) What do you think of that, Willard?

she would have thereby called Quine Willard. Indeed, had Quine’s mother uttered
(22) and then never addressed Quine as Willard but only as Willy, he might
reasonably say to her:
(24) I thought you said you were going to call me Willard but you never have, not
even once.
For her to respond with something like:
(25) Of course I have; when you were born, I called you Willard after my brother
would be impertinent because patently irrelevant to the remark to which it purports
to respond.
On the view being suggested here, calling essentially involves vocalization and,
for that reason, naming cannot be assimilated, without conflation, to calling. Fara
clearly thinks otherwise, as she claims that the author of Word and Object was called
Willard regardless of whether he was ever so addressed. It is time to ask what leads
Fara to this provocative view.
So far as I can tell, there are at least three factors motivating this view:
(i) Antecedent theoretical commitments

(ii) Apparent intuitiveness
(iii) Adjectival analogies
Because (i) provides psychological rather than logical motivation for Fara’s pro-
posed view, I will ignore it and focus instead on (ii) and (iii), beginning with the
former.
Although Fara does not explicitly characterize her view as “intuitive,” she
apparently believes that it is. For she says (in the opening paragraph of her essay)
that “most of us [reading this paper] know” that (8) is true:

Presumably, if “most of us know” that (8) is true, “most of us” would concede as
much, thereby lending credence to the idea that Fara’s view is an intuitive one.
Perhaps I am mistaken, but I believe that many of us (reading Fara’s paper) would
assent to (8) but regard (26) as a more accurate, more precise, representation of the
historical facts than (8):
(26) The author of Word and Object was named ‘Willard’, even if no one ever
called him that.
This would, of course, be consistent with the view proposed above insofar as it
reflects a distinction between naming and calling. However, instead of speculating
about what “most people” reading Fara’s paper would say about the virtues of (8) vs.
(26), let’s turn to Fara’s intriguing claim regarding the distinction between (8) and
(9). According to Fara, that both sentences are grammatical yet mean different things
is “plain to see” upon consideration of appellative ‘called’ constructions containing
adjectives rather than names.
Fara claims that (27) and (28) – patently grammatical and non-synonymous –
exhibit “the same contrast” as that between (8) and (9).
(27) Maude called me stupid.

(28) Maude called me ‘stupid’.
My concern is with the alleged analogy between (8) and (27). I think that the
two sentences are importantly different, as evidenced by the fact that (27) can be
paraphrased in any number of ways that (8) cannot. Thus, consider any of (29)–
(31):
(29) Maude basically called me stupid.

(30) Maude basically said I was stupid.
(31) Maude more or less said I was stupid.
Thus, while (29) is fine, (32) is not.
(32) The author of Word and Object was basically called Willard.
Moreover, while ‘called’ in (8) might readily be replaced by ‘named’, the same
cannot be said about the occurrence ‘called’ in (27), as evidenced by the oddness of
(33):
(33) Maude named me stupid.

These observations are consistent with Biro’s view that ‘calling’ as it occurs
in (27), but not (8), amounts to characterizing, not naming. I agree with Biro and
believe that his alternative “characterization” interpretation of what is going on in
sentences like (27) undercuts Fara’s analysis. But is Biro’s own positive analysis
any more promising than Fara’s?
74 M. Reimer
3 John Biro
Biro (2012) agrees with Fara that sentence (8) is grammatical but says that the reason
inverted commas around the name are unnecessary to insure its grammaticality is
that the linguistic context, specifically the phrase ‘is called : : : ’, disambiguates the
sentence: it makes it clear that the name is in fact being mentioned rather than used.
In fact, according to Biro, it would be a mistake to put inverted commas around the
name ‘Willard’ in (8), as this would change the expression being mentioned. Instead
of the name’s being mentioned, the name of the name would be mentioned. And that
would, of course, yield a different sentence all together – one with a very different
meaning and truth value.
Biro goes on to provide two additional examples of cases where context
disambiguates in favor of mention over use and where employing inverted commas
would therefore be a mistake, rather than merely unnecessary or redundant.
(34) Someone addressed (referred to) Quine as Willard.

(35) Hitler resented being called Schicklgruber.
I think Biro is mistaken here. While it is true that linguistic context disambiguates
the speaker’s intended meaning in utterances of sentences like (34) and (35), it does
not follow that the inclusion of inverted commas around the names occurring therein
would change the meaning of the sentence. It would merely make the intended (and
disambiguated) meaning explicit.4 In this respect, sentences like (34) and (35) are
like (36) and (37):
(36) Cat has three letters.

(37) I love Ingrid but Ursula is a bit too exotic for me.
where the latter is uttered in a context where a soon-to-be-mother is pondering baby
names. Context, linguistic in the case of (36) and extra-linguistic as well in the case
of (37), arguably makes clear that (36) is about the word cat (about ‘cat’) and that
(37) is about two names, not two famous actresses. However, it does not follow that
making this explicit via the available conventional devices (e.g., inverted commas)
yields sentences with different meanings. But on Biro’s view, if the unnecessary
punctuation alters the sentences’ meanings, then (36) and (37) should not mean
what (38) and (39) mean – but surely they do.
(38) ‘Cat’ has three letters.

(39) I love ‘Ingrid’ but ‘Ursula’ is a bit too exotic for me.
Similarly, while linguistic context disambiguates (34) and (35), making it clear that
the names ‘Willard’ and ‘Schicklgruber’ are mentioned rather than used, that does
mean not that the inclusion of inverted commas about those names would yield
4
This general sort of picture finds expression in the works of Saka (1998) and Recanati (2001).
sentences with different meanings. If it did, (40) and (41) would not mean what
they intuitively (and indeed obviously) do mean – and the sentence that precedes
this one would be irrelevant to the topic of the present paper, as it would concern
not names and named entities but names of names and named names.
(40) Someone addressed (referred to) Quine as ‘Willard’.

(41) Hitler resented being called ‘Schicklgruber’.
Whether we use or mention a name, or any other sort of expression, is presumably
determined by our communicative intentions, not by whether or how we choose to
signify (i.e., publicize) those intentions. If we fail to use inverted commas, contrary
to the conventions of contemporary analytic philosophy, we are perhaps less precise
than we could be but we are not conceptually confused. For we are not using an
expression where we should be mentioning it; we are simply failing to make explicit
that a particular expression is being mentioned.5 Because I take this to be Kripke’s
view, I question Biro when he says:
Kripke is relying on the widely accepted principle that to mention a name, one must enclose
it with inverted commas. He concludes that the absence of these in [8] therefore shows that
the name is being used, not mentioned. (2012: 285)
I would say, and I assume Kripke would concur, that in an inscription of (8)
the name is in fact being mentioned by the speaker, who fails to make this explicit
by using inverted commas, or some other suitable device. Hence the accusation of
“sloppiness,” rather than conceptual confusion – or (as Fara puts it) “philosophical
blundering.” One does not (pace Biro) mention a name by enclosing it in inverted
commas; rather, one encloses an expression in inverted commas in order to indicate,
to make explicit, that it is being mentioned. In much the same way, one does not
refer to a book by italicizing its title; rather, one italicizes a book’s title in order to
indicate, to make explicit, that a book is being referred to. In this way, the relation
between (8) and (9) is akin to that between (42) and (9):

The important point is that conventional devices like inverted commas and italiciza-
tion do not change the meaning of sentences; they simply make the meaning, which
may already be clear, explicit.
5
Thus, while the sentence the speaker uses might involve a use-mention confusion, it does not
follow that the speaker herself is confused on this matter. However, as indicated above, she might
not realize that, taken literally, the sentence she utters involves the use of an expression that she
intends (and succeeds) in mentioning.
76 M. Reimer
4 A Kripkean Alternative
In order to appreciate the plausibility of the Kripkean view sketched above, one has
to appreciate how very modest that view is. Thus, in colloquial discourse we may
express (43) by “saying” or inscribing (44).
(43) : : : ’N’ : : :
(44) :::N:::
In this way, colloquial discourse “confuses use and mention.” For the name
mentioned in what is expressed is not mentioned but used in the sentence uttered or
inscribed. However, the speaker (vs. her speech or writing) is guilty of confusing use
and mention only if she takes that speech or writing at face value and then (perhaps)
invokes it in arguing for or against some semantic theory of ordinary language.
Among such speakers are those who would attempt to discredit Kripke’s rigidity
thesis by pointing out that in colloquial speech we may express (1) by saying (2).
(1) Trump might not have been called ‘Trump’.

(2) Trump might not have been Trump.
What now of the fact that in colloquial writing we may express (9) by inscribing
(8)?
(8) The author of Word and Object was called Quine.

(9) The author of Word and Object was called ‘Quine’.
I have suggested (in the spirit of Kripke) that there is perhaps a use-mention
confusion here insofar as the inscribed sentence expresses something involving the
mention a name (‘Quine’) that is only used in that sentence. I have further suggested
that this sort of phenomenon is usefully understood in terms of the more general
phenomenon of explicitness. Thus, just as (13) is a more explicit version of (12),

(12) The author of word and object was called Willard.
so too (9) is a more explicit version of (8). Finally, I have suggested that
these phenomena are theoretically important in part because of their increasing
pervasiveness, thanks to texting and its minimalist punctuation. All of this is in
the spirit of Kripke insofar as it suggests that we (as philosophers of language) must
exercise caution when drawing on “sloppy, colloquial” discourse in our efforts to
support or undermine semantic theories of ordinary language.
5 Methodology
Both Fara and Biro focus their commentaries on the first sentence of a parenthetical
comment occurring within a lengthy footnote to the first lecture of Naming and
Necessity. Although this makes for some interesting philosophical discussion
regarding the use-mention distinction and its confusion in colloquial language, the
two sentences that follow the first sentence merit equal (if not greater) philosophical
scrutiny. So, in concluding, let us return to that parenthetical comment in its entirely.
Sloppy, colloquial speech, which often confuses use and mention, may, of course, express
the fact that someone might have been called, or not have been called, ‘Aristotle’ by saying
that he might have been, or not have been, Aristotle. Occasionally, I have heard such loose
usages adduced as counterexamples to the applicability of the present theory to ordinary
language. Colloquialisms like these seem to create as little problem for my theses as the
success of ‘Impossible Missions Force’ creates for the modal law that the impossible does
not happen.
Generalizing from Kripke’s understandable concerns about his own semantic

theses, we emerge with a methodological worry about the legitimacy of appeals to
“sloppy, colloquial” discourse in assessing semantic theories of ordinary language:
Philosophers should avoid appealing to colloquial and otherwise informal discourse in
assessing semantic theories of ordinary language as such discourse embodies myriad
confusions that must be eschewed in rigorous semantic theorizing.
Although this methodological principle might seem uncontroversial, even obvious,

it is not. There are several reasons for this. First, ordinary language is rife with
language that is “sloppy,” “colloquial” or otherwise “loose.” This has become
increasingly true with the explosion of texting, tweeting, blogging, and other forms
of social media. The default, if controversial, assumption should therefore be
that such language is relevant to the assessment of semantic theories of ordinary
language. Second, semantic theories like Kripke’s rigidity thesis are not obviously
of a piece with a priori logical laws like “the modal law that the impossible does not
happen.” Thus, the presumptive irrelevance of “loose usages” to laws of logical
does not automatically translate into the irrelevance of such usages to semantic
theories like Kripke’s rigidity thesis. Third, determining whether a given bit of
discourse (whether spoken or written) is “loose” is no simple matter. To see this,
just consider identity statements between names, like Frege’s famous ‘The Morning
Star is the Evening Star’. Such a statement is certainly informative insofar as it
conveys far more than the self-identity of the planet Venus. But the question whether
this information is semantic (and “strict”) or pragmatic (and “loose”) has no easy
answer, and perhaps none that can be arrived at by principled, theory-neutral means.
The potential relevance of colloquial or otherwise “loose” language to semantic
theories of ordinary language is a patently methodological and therefore founda-
tional issue. It is for this reason more important to the philosophy of language than
the interesting question of whether or not utterances of sentences like ‘Trump might
not have been Trump’ or inscriptions of sentences like ‘The author of Word and
78 M. Reimer
Object was called Willard’, involve use-mention confusions. As the pervasiveness of

texting, blogging, and other increasingly informal venues of discourse demonstrate,
attention to, and careful individuation of, such discourse is crucial if contemporary
philosophy of language is to remain (or perhaps become) truly “contemporary.”6
References
Bach, K. (2002). Giorgione was so-called because of his name. Philosophical Perspectives, 16,
73–103.
Biro, J. (2012). Calling names. Analysis, 71(2), 492–501.
Fara, D. (2011). You can call me stupid, : : : just don’t call me stupid. Analysis, 72(3), 492–501.
Kripke, S. (1980). Naming and necessity. Cambridge, MA: Harvard University Press.
6
I would like to thank the editors of this volume for helpful comments on earlier drafts of this
paper.
A Bridge from the Use-Mention Distinction
to Natural Language Processing
Shomir Wilson
Abstract Within computer science, the study of the syntax and semantics of
metalanguage is well developed for formal languages, and this work is applied
prominently in the creation of programming languages and compilers. However,
relatively little work has been done in computer science to address metalanguage
in natural languages. This lack has been to the detriment of language technologies
that could exploit the information expressed in metalanguage to understand users’
utterances. This chapter addresses metalanguage and quotation from the perspective
of mentioned language, a closely related phenomenon, and describes its relevance
to core and applied work in natural language processing (NLP), a field in computer
science concerned with the interaction between computers and natural languages.
Examples are given for how state-of-the-art language technologies fail to cope
with mentioned language. Finally, to promote progress on the computational study
of mentioned language, a rubric is given for identifying the phenomenon in text.
This enables human annotators to work methodically on labeling text to train NLP
systems, a crucial precursor to further computational work.
1 Introduction
To understand the language we use, we sometimes must turn language on itself, and
we do this through an understanding of the use-mention distinction. In particular, we
are able to recognize mentioned language: that is, tokens (words, phrases, sentences,
letters, symbols, sounds, etc.) produced to draw attention to properties that they
possess. The historical lack of attention to the use-mention distinction in natural
language might suggest that it is peripheral to the study of language, but this is far
from the truth. Evidence suggests that human communication frequently employs
the use-mention distinction, and we would be severely handicapped without it
(Perlis et al. 1998). In both written and spoken contexts, the mention of letters,
S. Wilson ()
University of Cincinnati, Cincinnati, OH, USA
e-mail: shomir.wilson@uc.edu

https://doi.org/10.1007/978-3-319-68747-6_4
80 S. Wilson
sounds, words, phrases, or entire sentences is essential for many language activities,
including the introduction of new words, attribution of statements, explanation of
meaning, and assignment of names.
However, algorithmic detection of the distinction is a nontrivial task. While
stylistic cues like italic text or quotation marks are sometimes used to indicate the
mention of language, such cues are not applied (or available) equally in all contexts.
Even when they are applied uniformly, they tend to be “overloaded” with other
uses as well (e.g., emphasis). Cues such as pauses and gestures exist in spoken
conversation, but these also are only approximate indicators and are easily lost in
transcription. Additionally, just as we benefit from mutual recognition of the use-
mention distinction in human communication, the potential exists for us to benefit
from language technologies that recognize it as well. Given a method for computers
to detect mentioned language, the following become possible:
• Dialog systems (such as those spoken with on mobile phones) can be designed
to recognize when a user is attempting to correct misinterpreted statements or
introduce new terms, instead of ignoring or misinterpreting those user activities.
• Systems that use lexical semantics, such as search engines, can take advantage of
information encoded in mentioned language, since it tends to be direct, salient,
and unambiguous.
• Trends in language can be studied with special attention to how new terms are
effectively (or ineffectively) introduced.
• Source attribution tools can be trained to precisely identify which words are being
reproduced in a quotation without the aid of stylistic cues, such as quotation
marks or italics.
• Language learning materials (especially for second language acquisition) can be
prepared with special attention to strategies in metalanguage that have been found
to be most effective.
• Typesetting and copyediting software can be designed to recognize instances of
mentioned language and apply stylistic features to them uniformly.
In natural language processing (NLP, the field of computer science concerned with
the interaction between computers and natural language) it is often necessary to
have a collection of annotated instances of a phenomenon in text before detection
of the phenomenon can be automated. Although automatic annotation is feasible for
some studies, mentioned language has received so little prior attention that human
annotators are appropriate to ensure the results are of a high quality (Wilson 2011).
To ensure the replicability of human-sourced annotations, it becomes necessary to
define the phenomenon and provide a labeling rubric. This chain of requirements
has perhaps stymied work to detect and exploit mentioned language in language
technologies.
This chapter reports on the first effort to break the logjam described above and
enable the computational study of mentioned language. A framework is presented
for identifying and analyzing instances of mentioned language, in an effort to
reconcile elements of previous theoretical works for practical use. Definitions
for mentioned language, metalanguage, and quotation have been formulated with
A Bridge from the Use-Mention Distinction to Natural Language Processing 81
intuitive links between them. Applications are listed for the detection of mentioned
language for NLP; these include fundamental tools that many language technologies
use (e.g., syntactic parsers) and end-user applications such as dialog systems.
Finally, a procedural rubric has been constructed for labeling instances of mentioned
language. This rubric has been applied with success to create a corpus of mentioned
language from a set of English texts (Wilson 2012) and automate the detection of the
phenomenon (Wilson 2013), although those computational results fall beyond the
scope of the present text. Finally, some examples of rubric application are provided
to illustrate its effectiveness and caveats.
2 A Definition for Mentioned Language
Although the reader is likely to be familiar with the use-mention distinction, the
topic merits explanation to establish what precisely is being referred to. Intuitively,
the vast majority of language is produced for use rather than mention, as the roles
of mention are auxiliary (albeit indispensible) to language use. For brevity, this
chapter will adopt the terms language-mention to refer to the act of mentioning
language and mentioned language to refer to linguistic entities produced for the
purpose of mentioning them. The terms language-use and used language will appear
occasionally and will carry the expected complementary meanings.
The use-mention distinction, as one might expect, is the distinction between
using linguistic entities (such as letters, symbols, sounds, words, phrases, or
sentences) and mentioning them. Since this explanation is slightly opaque at best,
some examples and a definition will follow. Consider two example sentences:
(1) The cat is on the mat.

(2) The word “cat” is spelled with three letters.
In (1), the reader’s attention to meaning does not focus on the words themselves, but
instead on the presumed cat on the mat. We say the word “cat” in particular is being
used (to refer to something other than itself, namely to a kind of animal, not even
a word at all) in (1); and so is the word “mat”. This is the usual way any word is
employed in a sentence: to call attention to something beyond the mere word. In (2),
the reader understands that it is the word cat—a string of three letters, as opposed
to any particular cat or cat concept—that is in the focus of the sentence. In such a
case we say the word is being mentioned, not used. Quotation marks around cat in
(2) are a convention to further reinforce that the word is being mentioned, and in
some contexts (such as this sentence) italics may serve the same purpose.1 “Setting
aside” mentioned language via stylistic cues is a common convention, as reflected
1
In a manner of speaking, quotation marks provide a name that refers to the quoted word. Hence
the quotes-plus-word as a unit is being used to refer to the word inside quotes, and the word itself
remains a case of mention. This is not a serious problem, but it foreshadows an issue that will arise
82 S. Wilson
in popular style guides (Chicago Editorial Staff 2010; Strunk and White 1979)
for formal writing. In spoken language, nonverbal cues often delimit mentioned
language, such as prosodic features (e.g., intonation, stress) or gestures.
The other linguistic entities listed above can also be mentioned; for example:
(3) The Classical Latin alphabet did not contain a ‘w’.

(4) Mathematical symbols, such as ‘1’, are available for some fonts.
(5) The rusty hinge emitted a sharp “eeeeek” sound as it closed.
(6) “Behind the eight” is an idiom that originated from the game of billiards.
(7) The sentence “The cat is on the mat” appears in many linguistics papers.
Longer linguistic entities (such as paragraphs) are also subject to language-mention,
though this occurs less frequently and places a greater burden on an audience’s
understanding of the phenomenon. One frequent sentence-length role of language-
mention is quotation, in which language from another source is reproduced as part
of a statement, as in (8) below:
(8) Eric said, “We should meet for lunch.”

In (8), the phrase between quote marks is mentioned as what Eric has said. However,
a reader is likely to react to the quoted text as a string with semantic depth, indicating
that the use aspect of the use-mention distinction is present as well. This mix of use
and mention is common in quotation, as we tend to quote linguistic entities that
possess meaning.
By necessity, discussions of language always invoke metalanguage, which is
language used when language itself is being described. Within the context of formal
languages, metalanguage tends to be distinct and separate from object language—
that is, the language being discussed. However, in natural language this separation
does not hold; we use natural language to talk about natural language. Notably,
sometimes two different natural languages are present when language-mention
occurs, as in
(9) The French word chat refers to a feline animal.

Although English resembles a metalanguage in (9), it is clearly not a metalanguage
in the general sense, as English is used for many “non-meta” functions (and French
speakers can discuss English too).
In spite of the ubiquity of the phrase use-mention distinction, it is difficult to
find an explicit definition for either the distinction itself or its two disjuncts. The
effort here will be to define mentioned language since it is less common and far
less studied than used language. The definition below will specificallycover intra-
later. This chapter will follow convention, however, and speak of a quoted-expression unit as a case
of mentioned language (of the item inside quotation marks).
sentential mentioned language, where the mentioned linguistic entity is referred to

inside of the same sentence that it occurs in. An example of a sentence that fails this
requirement is:
(10) Disregard the last thing I said.

This restriction was necessary to reduce the complexity of the computational tasks
that motivated this framework. Also, although the definition is nominally applicable
as a test to determine whether a token qualifies as mentioned language, it is not
intended for that activity. A rubric for labeling candidate instances will follow later
in this chapter.
Definition For T a token or a set of tokens in a sentence, if T is produced to draw
attention to a property of the token T or the type of T other than denotation or
extension, then T is an instance of mentioned language.
Here, a token is the specific, situated (i.e., as appearing in the sentence)
instantiation of a letter, symbol, sound, word, phrase, or entire sentence (subsumed
in a longer sentence as an independent clause, as in (8) above). A property might
be a token’s spelling, pronunciation, meaning (for a variety of interpretations of
that term), structure, connotation, original source (in cases of quotation), or another
aspect for which language is shown or demonstrated. The type of T is relevant in
some instances of mentioned language (such as in (2)) and the token itself is relevant
in others, as in
(11) “The” appears between quote marks in this sentence.

Constructions like (11) are unusual and are of limited value in practical language
use, but the definition accommodates them for completeness.
3 Relationship with Natural Language Processing
Thus far, the use-mention distinction has received relatively little consideration
from research in natural language processing and artificial intelligence. This section
will illustrate the importance of detecting mentioned language and how some
active areas of research stand to benefit. Section 3.1 will discuss the difficulty that
mentioned language poses to part-of-speech taggers and parsers. However, parsing
is not an end goal in itself, and Sects. 3.2 and 3.3 will discuss how the detection of
mentioned language can impact other automated language tasks.
3.1 Part-of-Speech Tagging and Parsing
Current part-of-speech taggers and parsers are insensitive to the difference between
use and mention, due in part to a lack of conventions on how to tag and parse
84 S. Wilson
mentioned language. An instance of mentioned language ostensibly functions as

a noun (noun phrase), and most instances are labeled as such, simply because they
tend to be nouns when they appear in language use. However, sequences of words
that bear little resemblance to noun phrases (such as independent clauses, which
appear when quoting speech acts) are equally instances of mentioned language.
Existing corpora, to the knowledge of this writer, do not label mentioned language,
though some make concessions toward doing so. The Penn Treebank, for instance,
has tags for foreign words (FW) and symbols (SY) (Marcus et al. 1993), but the
corpus does not differentiate when these entities are used and mentioned. Even when
a word that usually functions as a noun is mentioned, it is unclear what kind of noun
(e.g., proper, common) it should be. In a way, it serves as a name, but it is not a
proper name. Its referent (the word, as it appears) is peculiar when compared to the
usual referents of nouns.
Complicating matters further, mentioned language often involves the production
of linguistic entities in syntactic structures in which they usually do not appear.
Letters, symbols, phonetics, and words that are rarely (if ever) used as nouns can
be mentioned, imparting to them the qualities of nouns. These irregular appearances
tend to lead parsers astray when they are trained on large labeled corpora. Consider,
for example, the two parses in Fig. 1, which were produced by the Stanford Parser
(Klein and Manning 2003). The first parse is structurally correct, but the second
parse does not have the apparent benefit of quotation marks around go. The output
suggests that go is the main verb of the sentence, and this leads to an unusual (and
unusable) syntactic interpretation of the sentence.
Although quotation marks are sometimes used to delimit mentioned language,
they are a tenuous cue at best, as are all other stylistic cues. Quotation marks, bold
text, and italic text are frequently indicators of mentioned language, but each of them
has other common uses, such as distancing and emphasis (Gomez-Torrente 2017;
Jaszczolt and Huang 2017; McCullagh 2017). These cues are often unavailable in
informal texts, and their spoken (approximate) analogues can be complicated to
retrieve from speech recognition systems. Moreover, sometimes stylistic cues make
little difference. Consider for instance the two parse trees in Fig. 2. The sentence in
the first parse contains quotation marks around before, but neither parse assigns a
reasonable label to the word.
Errors in parsing and tagging mentioned language have causes that extend
beyond the parsing system used to generate the above examples. One problem is
the irregularity of mentioned language: parsers are trained upon large text corpora,
and in any given corpus, a common word will appear in use far more often than in
mention. A second problem is the “alienation” imparted upon mentioned language;
the vocabulary and sentence constructions that usually surround a word are often
absent when the word itself is being discussed, leaving statistically trained parsers
with little insight on what to do.
(ROOT
(S
(NP (DT The) (NN word) (`` ``) (NN go) (‘‘ ‘‘))
(VP (VBZ appears)
(PP (IN on)
(NP (DT the) (NN screen)))
(PP (IN after)
(NP (CD five) (NNS minutes))))
(. .)))
(ROOT
(S
(NP (DT The) (NN word))
(VP (VBP go)
(SBAR
(S
(VP (VBZ appears)
(PP (IN on)
(NP (DT the) (NN screen)))
(PP (IN after)
(NP (CD five) (NNS minutes)))))))
(. .)))
Fig. 1 Parses of two stylistic permutations of the sentence “The word go appears on the screen
after five minutes”. The first contains quotation marks around go, while the second does not
3.2 Dialogue Systems
A dialog system is a computer system that converses with a human via natural
language. Dialog systems can process communication between a user and a domain
(e.g., an information source or a controllable device) when spoken or written
conversation is a desirable mode of interaction (Josyula 2005; Lester et al. 2004).
A sufficiently flexible dialog system can ease the user’s “learning curve” when
interacting with a new system (Litman and Pan 2002) or enable human-computer
interaction in situations where speech is the preferred channel of communication
(e.g., while driving a car). Dialog systems are frequently task-oriented (Josyula et
al. 2003): they are designed to cooperate with users to perform specific activities or
achieve certain outcomes.
In contrast with formally written language, conversational language contains a
higher frequency of metalanguage and mentioned language. One study (Anderson
et al. 2004), using a subset of the British National Corpus, found that just over
86 S. Wilson
Fig. 2 Parses of two stylistic (ROOT

permutations of the sentence (S
“The only word on the paper
was ‘before’”. The first (NP
contains quotation marks (NP (DT The) (JJ only) (NN word))
around go, while the second
(PP (IN on)
does not
(NP (DT the) (NN paper))))
(VP (VBD was) (`` ``)
(PP (IN before)))
(‘‘ ‘‘) (. .)))
(ROOT
(S
(NP
(NP (DT The) (JJ only) (NN word))
(PP (IN on)
(NP (DT the) (NN paper))))
(VP (VBD was)
(ADVP (RB before)))
(. .)))
10% of sentences in conversational English contain some form of metalanguage.

Conversation contains a variety of challenges that interlocutors must overcome
to understand each other and advance their respective goals. Channels between
them are often “noisy” (in both the figurative and literal senses), leading to
misunderstood or lost utterances. Conversational language is informal and often
fraught with broken statements, restatements, sub-dialogues, and corrections. An
interlocutor must model their dialogue partner’s knowledge state to understand what
is appropriate to say next and to provide context for what has been said. New terms
are sometimes introduced, and conversation history often must be referenced.
From the imagined perspective of a species of artificially intelligent machines, it
might seem bizarre that humans try to speak with each other at all. However, our
facilities for recovering from perturbations in conversation are well-developed and
require only nominal effort. These facilities are collectively termed conversational
adequacy (Perlis et al. 1998), and the ability to recognize the use-mention distinction
is crucial to them. Figure 3 illustrates this with an excerpt from a conversation with
Let’s Go! (Eskenazi et al. 2008), a dialogue system designed to help users plan
trips on the Pittsburgh bus system. Users speak with it via telephone, and speech
recognition errors occur when audio quality is poor (and even sometimes when it is
not).
In the Let’s Go! fragment, the user says that he wishes to depart from Arlington,
but the dialog system mishears it as “Allegheny West”. The system asks for
confirmation, and the human tries to correct the misunderstanding by mentioning his
Fig. 3 Fragment of a Let’s Go!: Where do you wish to depart from?

human-computer
User: Arlington.
conversation with Let’s Go!
Let’s Go!: Departing from Allegheny West. Is this right?
User: No, I meant “Arlington”.
Let’s Go!: Please say where you are leaving from.
previous utterance. However, Let’s Go! does not recognize the correction and reverts
to a rewording of its original question. More broadly, interlocutors often utilize
mentioned language or metalanguage to track dialogue state, clarify the meaning of
terms, restate lost or misunderstood utterances, report others’ speech acts, and check
dialogue partner comprehension. Without this understanding, a dialogue partner
would likely be severely handicapped in their ability to participate in free-flowing,
robust conversation.
To date, dialogue systems have had to function in spite of this handicap,
often with frustrating results. Current conversational agents are susceptible to
the brittleness problem in artificial intelligence (Anderson and Perlis 2005) when
they are sidetracked by issues such as misunderstandings, misrecognitions, and
user-realized errors (e.g., the user makes a wrong choice, knows it, and wishes
to correct it). Those systems that do have recovery strategies for conversational
perturbations tend to employ methods such as offering available choices again,
repeating a question, or restarting a session. Such strategies come at a cost in
time and user patience. This serves as motivation to develop dialogue systems that,
when appropriate, are capable of responding to problems in dialogue using the same
familiar strategies that humans use when talking with each other.
3.3 Lexical Semantics
Although much of our language learning happens implicitly, explicit statements

about the meanings of words and phrases are essential for us as sources of direct
and (relatively) unambiguous linguistic information. Situations occur when words
and idioms must be introduced with mentioned language, either to highlight their
introduction (alerting the audience of their importance) or to discuss semantics
when doing so otherwise is difficult. After introduction, when the novelty of a
new linguistic entity has diminished, mentioned language is sometimes necessary
to clarify or refine the audience’s understanding of the entity.
Automated detection of mentioned language can make this condensed linguistic
information available to language technologies as well, especially when such
detection is applied to large corpora that contain many instances of the phenomenon.
Such detection could, for example, be used to prioritize (or place greater weight
upon) the processing of sentences that contain explicit information about word
referents. Explanations of idiomatic expressions then would become salient sources
of information, as separated from the instances of their use.
88 S. Wilson
The automated detection of mentioned language affects other topics of research

in natural language processing, including:
• Source Attribution: Presenting quotation explicitly in discourse requires metalan-
guage and mention of language. An understanding of the structure of mentioned
language contributes to efforts to detect where in text sources are cited and what
has been reproduced (e.g., delimiting sentences as mentioned without the aid of
stylistic cues). This is especially true for the quotation of speech acts, when one
person cites the utterances of another.
• Sentiment Analysis: Sentiments expressed in mentioned language are not always
shared by the language user. For example, a person might talk about an issue
in a positive light while quoting others’ arguments in order to refute them.
Instances of mentioned language often provide additional information for a task,
but for some tasks—such as sentiment analysis—those instances should be either
reduced in value or discarded.
• Natural Language Understanding: A general problem remains open on how to
represent statements about language so that intelligent inferences can be drawn
from them. The flexibility with which humans discuss language and the variety
of aspects of it that we refer to both pose a challenge to knowledge representation
schemes.
• Studies in Language Acquisition: In research on second language acquisition,
Ellis has noted the use of informal metalanguage to discuss L2 words and
concepts (e.g., “you need ‘a’ before the noun when a person or thing is mentioned
for the first time”) (Ellis 2009). It is possible that certain metalinguistic strategies
are more effective than others, and discovering this will require the ability to
detect those strategies in language learner corpora and accumulate information
about them.
4 A Rubric for Mentioned Language
A human reader with some knowledge of the use-mention distinction can often
intuit the presence of mentioned language in a sentence, even in the absence of
stylistic cues. However, to operationalize the concept and move toward corpus
construction, it is necessary to create a rubric for labeling mentioned language. The
rubric presented here is based on substitution, and it may be applied, with caveats
described in this section, to determine whether a linguistic entity is mentioned by
the sentence in which it occurs.
Rubric: Suppose X is a linguistic entity in a sentence S. Construct sentence S’ as
follows: replace X in S with a phrase X’ of the form “that [item]”, where [item] is the
appropriate metalinguistic term for X in the context of S (e.g., “letter”, “symbol”,
“word”, “name”, “phrase”, “sentence”, etc.). X is an instance of mentioned

language if, when assuming that X’ refers to X, the meaning of S’ is equivalent
to the meaning of S.2
Several examples will follow to demonstrate the rubric in action. Some minor
adjustments in wording will be necessary for some sentences, and these will
be shown. However, the handling of quotation marks, discussed below, must be
addressed before continuing.
Quotation marks are frequently used as a stylistic cue for mentioned language,
and they pose a slight problem to the rubric. This is because their inclusion or
exclusion in the linguistic entity X can alter the meaning of the transformed
sentence. For instance, consider sentence (6) above when testing “Behind the eight”.
If quotation marks are not included in X, then they persist in S’:
(12) “That phrase” is an idiom that originated from the game of billiards.
The meaning of (12) differs from (6), and “Behind the eight” thus fails the rubric.
(In fact, “That phrase” becomes mentioned language in S’). If the quotation marks
are included in X, then S’ becomes
(13) That phrase is an idiom that originated from the game of billiards.
but “That phrase” refers to “Behind the eight” (i.e., the phrase that includes
the quotation marks), failing the rubric as well. Discarding the quotation marks
altogether when applying the rubric might be sufficient in the general case, but it
poses a problem for sentences like:
(14) The character “ tends to appear at the start of a quotation.

For the purposes of the rubric, quotation marks will be considered informal cues
which aid a reader in detecting mentioned language. Style conventions may call for
them, and in some cases they might be strictly necessary, but a competent language
user possesses sufficient skill to properly discard or retain them as each instance
requires. Similar reasoning can be applied to other stylistic cues, such as bold text
and italic text. To avoid further complications, examples in the subsections that
follow will omit quotation marks or other stylistic cues.
This rubric has some distinct advantages over many prior treatments of the use-
mention distinction:
1. The rubric does not depend on the presence of quotation marks. Many previous
theories of the use-mention distinction require quotation marks to play an integral
role in mentioning language (Davidson 1979; Garcia-Carpintero 2004; Quine
2
For present purposes, truth conditions will be considered equivalent to meaning. However, a
substantial body of work exists on articulating the relationship between these two, as surveyed
by Lynch (2001) and Kirkham (1995).
90 S. Wilson
1940; Tarski 1933).3 In many cases, competent language users are able to
recognize the use-mention distinction without the aid of quotation marks. (The
reader may verify this by reading the example sentences in the next section
without reading their accompanying paragraphs). Punctuation such as quotation
marks is often omitted in informal contexts, and other cues (particularly bold and
italic text) sometimes take their place. The rubric acknowledges the flexibility of
the human recognition of the use-mention distinction, while most previous work
does not.
2. The rubric provides an explicit procedure for identifying mentioned language.
Earlier efforts tend to assume that examples of the use-mention distinction are
obvious, and they work from that assumption to an explanation of the semantics
of mentioned language (Cappelen and Lepore 1997; Maier 2007). The rubric
provides a mechanism for verifying purported examples, which is a desirable
alternative to accepting them at face value.
3. The rubric exploits the relationship between mentioned language and its senten-
tial context. Even when quotation marks are consistently used, it is the sentential
context of mentioned language that selects the relevant aspect of a mentioned
token (e.g., spelling, pronunciation, meaning, etc.). Without this, the meaning of
an instance of mentioned language is uncertain. Notably, Saka’s ostension theory
(1998) recognizes the value of sentential context but does not have the advantage
of item 2 above.
4.1 Examples of Rubric Usage
Some examples will illustrate how the rubric covers the varieties of mentioned
language. For instance, consider
(15) Fancy automobiles are called luxury cars.

where the phrase “luxury cars” is under consideration. Choosing “that phrase” as a
replacement, the sentence becomes
(16) Fancy automobiles are called that phrase.

where “that phrase” is understood to refer to “luxury cars”. While contextual
ramifications may exist (for instance, if a language user specifically wants to utter
the phrase “luxury cars”), the reader can verify that the meaning of the sentence
3
In fact, previous literature surveys (Anderson et al. 2002; Saka 1998) have shown that nearly all
of the proposed theories of the use-mention distinction have required quotation marks to delimit
mentioned language. This practice is so widespread that the literature often uses the term quotation
to refer to mentioned language, causing confusion over the meaning of the term, as it also refers
to the reproduction of language from another source. Reproduction of language is one function of
mentioned language but certainly not the only function.
is essentially unchanged, and “luxury cars” passes the rubric. In contrast, consider
testing the phrase “Fancy automobiles” in (15). The substitution test and a charitable
adjustment to the verb phrase result in
(17) That phrase is called luxury cars.

where “That phrase” is understood to refer to “Fancy automobiles”. It is plausible
(albeit odd) that a speaker might wish to assign the name “luxury cars” to the phrase
“Fancy automobiles”, but it is a very unlikely interpretation of the intent of the
original sentence. The phrase “Fancy automobiles” in (15) fails the rubric.
This rubric requires some adjustment when the sentence already explicitly refers
to X as a word, phrase, or other appropriate entity. In such cases it may be
appropriate to omit the linguistic entity under consideration without substituting,
as in (18) and (19):
(18) The French word chat refers to a feline animal.

(19) That French word refers to a feline animal.
where “That word” is understood to refer to “chat”. Explicit discussion of word
categories requires similar omission without substitution, as in testing “help” in (20)
with the rubric substitution shown in (21) below:
(20) The verb help has several different senses.

(21) That verb has several different senses.
In (21), “that verb” is assumed to refer to “help”; this does not change the meaning
of the sentence, and “help” passes.
The rubric also accounts for explicit discussion of pronunciation, as shown by
testing “canz” in (22) below with its substitution equivalent (23):
(22) Australians say the city name as canz.

(23) Australians say the city name as that pronunciation.
In (23), “that pronunciation” is assumed to refer to the pronunciation represented by
canz. Although the written medium requires an additional level of interpretation—to
distinguish the sequence of letters canz from the pronunciation of canz—the reader
should infer this from the substitution phrase. “canz” thus passes the rubric.
Discussion of characters and spelling is similarly covered, as shown by (24)
(testing “ie”) and its substitution equivalent (25) below:
(24) Garcia spelled his name with i.e. back then.

(25) Garcia spelled his name with that character string back then.
where in (25) “that character string” is assumed to refer to “ie”. This does not
change the meaning of the sentence, and “ie” passes the rubric. However, it might
92 S. Wilson
be tempting to test “Garcia” in (24) as well, and the substitution equivalent of the
sentence is (26) below:
(26) That character string spelled his name with i.e. back then.
Sentence (24) asserts that Garcia is responsible for the said spelling, and sentence
(26) asserts that a character string is responsible for the spelling. Since these are
very different meanings, “Garcia” fails the rubric and is not mentioned language.
Similarly, phrases that simply refer to characters or symbols do not pass the rubric.
Consider the test of “The symbol for infinity” in (27) below, shown with its
substitution equivalent in (28):
(27) The symbol for infinity is a lemniscate.

(28) That phrase is a lemniscate.
In (28), “That phrase” is understood to refer to “The symbol for infinity”. Sentence
(27) asserts that the symbol for infinity is a lemniscate. Sentence (28) asserts that
“The symbol for infinity”—that phrase—is a lemniscate. Since these meanings are
different, “The symbol for infinity” fails the rubric.
The rubric also shows that sentences with almost identical wording can differ on
whether they use or mention the same word or phrase, as in (29) and (30) below
(testing “Spain” in both):
(29) Spain is the name of a European country.

(30) Spain is a European country.
Their substitution equivalents are, respectively:
(31) That name is the name of a European country.

(32) That name is a European country.
where in (31) and (32) “that name” refers to “Spain”. Sentence (31) effectively
asserts that “Spain” is the name of a European country, which does not change the
meaning of the sentence; thus, “Spain” in (29) passes the rubric. However, (32)
asserts that “Spain”—the name itself—is a European country. Since this changes
the meaning of the sentence, “Spain” in (30) fails the rubric.
While many other permutations exist that require minor adjustments in wording,
such untamable variation is inherent in natural language, and the spirit of the rubric
is sufficient for the direction of a human labeling task.
4.2 The Definition-Rubric Relationship
The main intent of the rubric is to validate all and only the linguistic entities that
qualify for the definition of mentioned language when token T in the definition
draws attention primarily to a property of the type of T.4 This partial equivalence is
illustrated with the conjunction of two claims.
Claim #1: If a token T is produced in a sentence to draw attention to a property
of the type of T, then the deictic substitution in the rubric does not alter the truth
conditions of the sentence (thus satisfying the rubric). This is because the properties
of a type are not altered by the substitution in the sentence. In sentence (33) below
(33) “Cat” has three letters.

“Cat” is mentioned to draw attention to the spelling of the type of the word. In (34)
below this is preserved
(34) That word has three letters.

assuming, as the rubric requires, “That word” is understood to refer to “Cat”.
Properties of the type of “Cat” remain intact independent of its presence or absence
in the sentence.
Claim #2 is the converse of the first: if applying the rubric substitution to a
token T in a sentence does not alter the truth conditions of the sentence (thus
satisfying the rubric), then the original sentence draws attention to a property of
the type of T, satisfying the definition. This is because the deictic phrase must have a
relationship with the content of the sentence, and that relationship must match that
of the original token (T) that it displaced. Consider (35) below and (36), with the
rubric transformation on “cheese”:
(35) Cheese is derived from a word in Old English.

(36) That word is derived from a word in Old English.
These two sentences are equivalent in meaning because both of them refer to
a property of “Cheese”, regardless of the need to resolve the referent of “That
word” in (36). By referring to a property of “Cheese”, (35) satisfies the definition.
Moreover, the rubric indicates that “Cheese” in (37) below (transformed in (38)) is
not mentioned language:
(37) Cheese is derived from milk.

(38) That word is derived from milk.
Sentence (37) refers to a property of cheese and (38) refers to a property of an
unknown word indicated by “That word”. Sentence (37) thus does not satisfy the
definition, either.
Together, Claims #1 and #2 above should be sufficient to establish that the rubric
plays the same formal role as the definition in a wide variety of cases.
4
The rubric also covers some (but not all) instances of mentioned language where T is produced in
a sentence to draw attention primarily to a property of the token of T. This will be discussed further
in Sect. 4.3 below.
94 S. Wilson
4.3 SRT Sentences
Sentences that mention T and invoke self-reference to describe some aspect of T’s
role in the sentence will not be handled by the rubric. (For brevity these will be
called “SRT sentences”—Self-Reference to a Token.) For example, consider testing
SRT sentence “Cat” in (39) below, with the substitution result shown as (40):
(39) Cat is the first word in this sentence.

(40) That is the first word in this sentence.
whereas “That” in (40) is understood to refer to “Cat”. Although this sentence
satisfies the definition of mentioned language, the substitution changes its meaning.
Sentence (39) claims that “Cat” is the first word in sentence (39), which is true.
Sentence (40) claims that “Cat” is the first word in (40), which is false. Notably,
token-mention alone does not pose a problem for the rubric; consider (41) and (42)
below, testing the rubric on “cat”:
(41) The token cat has three letters.

(42) That token has three letters.
whereas “That token” is understood to refer to a tokenization of “cat”. These
sentences share the same meaning, even though one contains the token “cat” while
the other makes a deictic reference to a tokenization of “cat”. Sentence (41) thus
passes the rubric. Also, self-reference alone does not trouble the rubric; consider
(43) and (44) below, testing the rubric on the second appearance of “which”:
(43) This sentence, which contains the word which, has two commas.
(44) This sentence, which contains that word, has two commas.
where “that word” is assumed to refer to “which”. The reader may verify that the
meaning of (43) does not differ from (44).
Since SRT sentences can only illuminate properties of a token that involve the
token’s relationship with the sentence, these sentences appear to be of limited
practical value. Excluding them from the rubric was deemed fair because the goal
of creating it was practical: to operationalize the definition and simplify manual
labeling of mentioned language. Still, for their peculiarity, SRT sentences may
deserve further examination in future work.
5 Conclusions
This chapter introduced a framework for enabling the computational study of

mentioned language in natural language. Relationships between metalanguage,
quotation, and mentioned language were described in a way that disambiguates them
while preserving intuitions about them from prior work. A definition for mentioned
language was presented and augmented with a rubric for labeling instances of the
phenomenon in text. The benefits of detecting mentioned language were discussed
for parsing, conversational systems, and other active areas of research. Finally,
justification was given for accepting the rubric as being equivalent to the definition
in a wide variety of cases.
References
Anderson, M. L., & Perlis, D. (2005). Logic, self-awareness and self-improvement: The metacog-
nitive loop and the problem of brittleness. Journal of Logic and Computation, 15, 21–40.
Anderson, M. L., Okamoto, Y. A., Josyula, D., & Perlis, D. (2002). The use-mention distinction and
its importance to HCI. In Proceedings of the Sixth Workshop on the Semantics and Pragmatics
of Dialog (pp. 21–28).
Anderson, M. L., Fister, A., Lee, B., & Wang, D. (2004). On the frequency and types of
meta-language in conversation: A preliminary report. In Proceedings of the 14th Annual
Conference of the Society for Text and Discourse. Retrieved from: http://www.cs.umd.edu/
projects/metalanguage/results/std04_final.pdf
Cappelen, H., & Lepore, E. (1997). Varieties of quotation. Mind, 106(423), 429–450.
https://doi.org/10.1093/mind/106.423.429.
Chicago Editorial Staff. (2010). The Chicago Manual of Style (16th ed.). Chicago: University of
Chicago Press.
Davidson, D. (1979). Quotation. Theory and Decision, 11(1), 27–40.
https://doi.org/10.1007/BF00126690.
Ellis, R. (2009). A typology of written corrective feedback types. English Language Teaching,
63(2), 97–107.
Eskenazi, M., Black, A. W., Raux, A., & Langner, B. (2008). Let’s go lab: A platform for evaluation
of spoken dialog systems with real world users. In Proceedings of Interspeech. Brisbane.
Retrieved from http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.145.2108
Garcia-Carpintero, M. (2004). The deferred ostension theory of quotation. Nous, 38(4), 674–692.
Johnson (Eds.), The Semantics and Pragmatics of Quotation (pp. 135–167). Dordrecht:
Springer.
Jaszczolt, K., & Huang, M. (2017). Monsters and I: The case of mixed quotation. In P. Saka &
M. Johnson (Eds.), The Semantics and Pragmatics of Quotation (pp. 357–382). Dordrecht:
Springer.
Josyula, D. (2005). A Unified Theory of Acting and Agency for a Universal Interfacing Agent. PhD,
thesis College Park: University of Maryland at College Park.
Josyula, D., Anderson, M. L., & Perlis, D. (2003). Towards domain-independent, task-oriented,
conversational adequacy. In Proceedings of the 18th International Joint Conference on
Artificial Intelligence (pp. 1637–1638). San Francisco: Morgan Kaufmann Publishers Inc.
Retrieved from http://portal.acm.org/citation.cfm?id=1630659.1630950
Kirkham, R. (1995). Theories of Truth: A Critical Introduction (1st MIT Press pbk. ed.).
Cambridge, MA: MIT Press.
Klein, D., & Manning, C. D. (2003). Fast exact inference with a factored model for natural
language parsing. In Advances in Neural Information Processing Systems (Vol. 15). The
Stanford Natural Language Processing Group. Retrieved from: http://www-nlp.stanford.edu/
manning/papers/lex-parser.pdf
Lester, J., Branting, K., & Mott, B. (2004). Conversational agents. In M. Singh (Ed.), The Practical
Handbook of Internet Computing. Baton Rouge: Chapman & Hall/CRC Press.
96 S. Wilson
Litman, D. J., & Pan, S. (2002). Designing and evaluating an adaptive spoken dialogue system.
User Modeling and User-Adapted Interaction, 12, 111–137.
Lynch, M. (2001). The Nature of Truth: Classic and Contemporary Perspectives. Cambridge, MA:
MIT Press.
Maier, E. (2007). Mixed quotation: Between use and mention. In Proceedings of the Logic and
Engineering of Natural Language Semantics Workshop. Miyazaki. Retrieved from: http://
ncs.ruhosting.nl/emar/em_lenls_quot.pdf
Marcus, M. P., Marcinkiewicz, M. A., & Santorini, B. (1993). Building a large annotated corpus
of English: The Penn Treebank. Computational Linguistics, 19, 313–330.
McCullagh, M. (2017). Scare-quoting and incorporation. In P. Saka & M. Johnson (Eds.), The
Semantics and Pragmatics of Quotation (pp. 3–34). Dordrecht: Springer.
Perlis, D., Purang, K., & Andersen, C. (1998). Conversational adequacy: Mistakes are the essence.
International Journal of Human-Computer Studies, 48(5), 553–575.
Quine, W. V. O. (1940). Mathematical Logic. Cambridge, MA: Harvard University Press.
Saka, P. (1998). Quotation and the use-mention distinction. Mind, 107(425), 113–135.
Strunk, J., & White, E. B. (1979). The Elements of Style (3rd ed.). New York: Macmillan.
Tarski, A. (1933). The concept of truth in formalized languages. In J. H. Woodger (Ed.), Logic,
Semantics, Mathematics. Oxford: Oxford University Press.
Wilson, S. (2011). In search of the use-mention distinction and its impact on language processing
tasks. International Journal of Computational Linguistics and Applications, 2(1–2), 139–154.
Wilson, S. (2012). The creation of a corpus of English metalanguage. In Proceedings of the 50th
Annual Meeting of the Association for Computational Linguistics ACL (pp. 638–646).
Wilson, S. (2013). Toward automatic processing of English metalanguage. In Proceedings of the
Sixth International Joint Conference on Natural Language Processing (pp. 760–766). Nagoya:
Asian Federation of Natural Language Processing. Retrieved from: http://www.aclweb.org/
anthology/I13-1091
Part II
Quotation Unified
Unity in the Variety of Quotation
Kirk Ludwig and Greg Ray
In quotation not only does language turn on itself, but it does so

word by word and expression by expression, and this reflexive
twist is inseparable from the convenience and universal
applicability of the device.
–Donald Davidson
Abstract This chapter argues that while quotation marks are polysemous, the
thread that runs through all uses of quotation marks that involve reference to
expressions is pure quotation, in which an expression formed by enclosing another
expression in quotation marks refers to that enclosed expression. We defend a
version of the so-called disquotational theory of pure quotation and show how
this device is used in direct discourse and attitude attributions, in exposition in
scholarly contexts, and in so-called mixed quotation in indirect discourse and
attitude attributions. We argue that uses of quotation marks that extend beyond
pure quotation have two features in common. First, the expressions appearing in
quotation marks are intended to be understood, and that they are intended to be
understood is essential to the function that such quotations play in communication,
though this does not always involve the expressions contributing their extensional
properties to fixing truth conditions for the sentences in which they appear. Second,
they appeal to a relation to the expression appearing in quotation marks that plays a
role in determining the truth conditions of the sentences in which they appear.
K. Ludwig ()
Indiana University, Bloomington, IN, USA
e-mail: ludwig@indiana.edu
G. Ray
University of Florida, Gainesville, FL, USA
e-mail: gregray@ufl.edu

https://doi.org/10.1007/978-3-319-68747-6_5
100 K. Ludwig and G. Ray
1 Introduction
Quotation is easy to understand but hard to explain. In this paper, we offer a

semantics for the varieties of quotation listed in (1)–(5).
(1) “Boston” contains six letters.
(2) “psychology” literally means “the study of the soul.”
(3) (a) He said, “Get serious, boy.”
(b) She said, “Gorse is common in Scotland”; she did not say, “Furze is
common in Scotland.”
(c) Caesar literally said, “Veni, Vidi, Vici,” not “I came, I saw, I conquered.”
(d) He said, ‘All mimsy were the borogoves,” but didn’t have anything in
mind by it.
(e) Then Jesus said to them, “Render to Caesar the things that are Caesar’s,
and to God the things that are God’s.”
(f) “Good morning! it was kind of you to push the chair up that hill ... I hope
it wasn’t heavy for you,” said Connie, looking back at the keeper outside the
door.
(g) Davidson said, “Quotation is a device used to refer to typographical : : :
shapes by exhibiting samples.”
(h) “And what is the use of a book,” thought Alice, “without pictures or
conversation?”
(i) “If I had a child!” she thought to herself; “if I had him inside me as a
child!”
(4) (a) In this chapter, Mill attempts to delineate when the authority of society
can rightly limit individuality and the “sovereignty of the individual over
himself.”
(b) Berkeley’s objective in the New Theory of Vision was “to shew the
manner wherein we perceive by sight the distance, magnitude, and situation
of objects. Also to consider the difference there is betwixt the ideas of sight
and touch, and whether there be any idea common to both senses” (NTV §1).
(c) They substituted, as Kant has pointed out, “a physiology of the human
understanding” for the Critical investigation of the claims of reason, and
anthropology for ethics.
(5) (a) Macomber said that he “bolted like a rabbit.”
(b) Alice said that she had “heard nonsense” compared with which that
would be “as sensible as a dictionary.”
(c) She said that it had been many years since “such trifles had broke across
the web of [her] solitude.”
(d) La Rochefoucauld said that jealousy “ends as soon as we pass from
suspicion to certainty.”
(e) Professor Elugardo said that William James said that religious leaders are
“creatures of exalted emotional sensibility.”
(f) She thought that it was curious that that “thin, proud man” should have
had “that little, sharp woman for a mother!”
Unity in the Variety of Quotation 101
(1) is an instance of pure quotation.1 (2) involves pure quotation in its first
appearance and arguably a dual use-mention (to be explained below) in its second
appearance. Examples in (3a–g) illustrate direct discourse. (3a) is an instance of the
use of quotation in dialogue in which understanding the quoted material is essential
to its linguistic function (as is made clear by the fact that in translation dialogue
inside quotation marks is translated as well). However, (3b) illustrates a strict use
of direct discourse that requires the use of the words used by the subject. (3c–d)
illustrate a strict use in which the function of the report does not require its speaker to
understand the words attributed to its subject, or, in the case of (3d), that it have any
meaning at all. (To avoid confusion in discussion of direct and indirect discourse,
we will use ‘speaker’ for the utterer of a sentence of direct or indirect discourse
and ‘subject’ for the person that sentence of direct or indirect discourse is about,
that is, the person the sentence reports as speaking). In contrast, (3e) is an instance
of direct speech in which the quoted words are a translation from the Greek of the
Gospel of Mark, which in turn translate the Aramaic of Jesus, and understanding
them is essential to its function. (3f–g) exhibit the use of ellipsis (in different ways
in each) and exclamation marks in reported speech (in (3f)), and highlight the need
to make sense of punctuation marks inside quotation marks in an account of the
function of quotation in dialogue and direct speech. (3h–i) are examples of the use of
quotation in the analog of direct speech for the attribution of thought. (4a–b) exhibit
uses of quotation in scholarly exposition in which there is a dual use-mention. (4c)
shows a use in which the quoted material is a translation of the original German.
(5a–c) are examples of “mixed quotation” in which quotation marks are used in
the complement clauses of indirect discourse. (5b) raises the problem of how to
handle multiple instances. (5c) raises the question of how to handle interpolations—
authorial brackets—that adjust context sensitive terms to the speaker’s context. (5d)
shows an example in which the quoted material is a translation of the original. (5e)
raises the problem of iterated mixed quotation. (5f) shows these issues extend to
attitude attributions.
1
The term ‘pure quotation’ was introduced into the contemporary semantics literature on quotation
by Cappelen and Lepore (1997), who contrast it with direct (3a–g) and indirect speech, and mixed
cases of direct and indirect speech (5a–e). They are picking up on a distinction drawn in (Davidson
1979), where Davidson contrasts the idea of quotation used to mention an expression that is not
itself used with what he calls mixed cases of use and mention. It is the first half of this contrast that
Cappelen and Lepore have in mind, namely, the use of quotation marks around an expression to
form an expression that refers to the enclosed expression, which is not at the same time used, and
which functions “like a single word” (Davidson 1979, p. 3). More precisely, what we have in mind
is given by (Q) below. This is the predominant use of the term ‘pure quotation’ in the literature on
quotation. For a different suggestion about what contrast should be drawn with ‘pure quotation’ see
(Saka 2013)—though contrast (Maier 2014) writing for the same journal a year later. Saka claims
it is “an analytic truth that all quotations that are not mixed must be pure” and so includes direct
speech under ‘pure quotation’. But it should be noted that Cappelen and Lepore introduce the term
stipulatively, so what counts as an analytic truth with regard to the usage they introduce depends
on what meaning they give it. Our view about the relations between these various uses of quotation
marks will emerge in the following.
We argue that pure quotation is what unifies all of these uses of quotation in the
sense that an account of the semantic function of quotation in each of these examples
involves an instance of the device characteristic of pure quotation. We follow John
Wallace (1970: 135–136) in holding that the semantic rule for pure quotation is
given by the reference clause (Q), where ‘¥’ takes on expressions as values, and we
use square brackets as Quinean corner quotes.2
(Q) For any ¥, [‘¥’] refers to ¥3

This is a precise expression of Tarski’s informal account of the function of quotation
marks.4 Anyone who understands this rule understands all there is to know about the
use of quotation marks in pure quotation. In the following, we sketch an account of
the semantic roles of the varieties of quotation in (2)–(5), which shows that the
device involved in pure quotation plays a central role in each. This will not quite be
to say that quotation marks mean the same thing in each of these uses, but rather
that, as we will suggest, quotation marks are polysemous, and the thread that runs
through the various uses of quotation marks is captured in a generalization of (Q)
we introduce below.
The plan of the paper is as follows. Section 2 takes up pure quotation. We review
desiderata on an adequate account, and five proposals that have been made about
how to understand pure quotation. We argue for the disquotational theory, as it
has come to be called, which invokes a simple rule like that expressed in (Q). We
head off some misunderstandings, respond to some objections, and draw out some
consequences. Section 3 takes up the use of quotation marks in direct discourse.
2
An expression consisting of square brackets (treated as corner quotes) around an expression
containing a metalinguistic variable abbreviates a description of an expression as the concatenation
of the contained expressions and value of the variable, in the order in which they appear inside the
brackets. Thus, [‘¥’] D the expression consisting of “’ followed by ¥ and then “’. [¥ and §] D
the expression consisting of ¥ followed by ‘and’ and then §. And so on. An expression consisting
of square brackets around an expression without a contained metalinguistic variable denotes that
expression. Thus, [Quine] D ‘Quine’. And, [¥] D ¥. For more on Quinean corner quotes, see Saka
(2017).
3
We suppress explicit relativization of semantic predicates to a language except where it is needed.
4
Tarski writes, “We denote by this term [‘quotation-mark names’] every name of a sentence (or
of any other, even meaningless, expression) which consists of quotation marks, left- and right-
hand, and the expression which lies between them, and which (expression) is the object denoted
by the name in question” (1983, p. 156). The rule is expressed here. He later writes, “Quotation-
marks names may be treated like single words of a language, and thus like syntactically simple
expressions” (p. 159). It is this last remark in particular that has led to the ascription of the
proper name theory of quotation to Tarski, though it is noteworthy that he says that they “may
be treated” that way, not that they are syntactically simple. The important point for his purposes
is that quotation terms function like names in the language in the sense of not having semantic
compositional structure. See Gomez-Torrente (2001).
In Sect. 3.1, we first identify a number of contexts in which expressions are referred
to but are also intended to be understood, where being understood is essential for
sentences in which they are contained to fulfill their function in communication,
though their extensional properties do not contribute to fixing the truth conditions
of the sentences in which they are contained. We call this a quasi-use-mention. We
argue that it extends to the use of quotation marks following ‘means’ in (2), and
then, in Sect. 3.2, to certain uses of quotation marks with direct discourse as well,
though this is not a feature that is represented in a compositional account of the truth
conditions of the sentences. We distinguish between strict and non-strict forms of
direct discourse, the former of which requires mentioning the specific words used
by the person to whom the discourse is attributed (the subject) and the latter of
which allows the use of expressions that translate the words the subject used. We
provide truth conditions for both strict and non-strict direct discourse. Non-strict
direct discourse also involves quasi-use-mention though this is not represented in the
truth conditions. In Sect. 3.3, the account is extended to direct attitude attributions
as illustrated by (3h–i). Section 4 takes up quotation in exposition as illustrated in
(4a–c) where what is said is said by the speaker or writer, though some of the words,
in quotation marks, are to be attributed to another. Section 5 takes up mixed indirect
discourse in Sect. 5.1 and mixed attitude attribution in Sect. 5.2. A general account
is provided for any number of distinct uses of quotation marks in complements of
indirect discourse or attitude attributions. Here too we distinguish between a strict
and a non-strict reading. Section 6 takes up the question how to accommodate mixed
indirect discourse in which expressions in quotation marks appear to be intended
to be evaluated in a context other than that of the speaker. This would make ‘x
said that’ what Kaplan called a ‘monstrous operator’, one that operates on character
rather than content, for it would shift the context of evaluation of context sensitive
expressions from the speaker’s context to another context. Without trying to settle
whether the examples that motivate this are well formed, we show that there is
nothing problematic about giving truth conditions for them. A language could make
provision for this. Section 7 takes up a prima facie objection to the account that
rests on the fact that when we translate, e.g., dialogue in a novel, we do not preserve
reference to the expressions in the original language. The answer is that ordinary
translation preserves function over reference when there is a conflict. Section 8 is a
brief summary and conclusion.
2 Pure Quotation
By pure quotation we have in mind a device in written language for referring to

expressions.5 Expressions are strings or configurations of symbols, including the
limiting case of one. We use ‘expression’ and ‘symbol’ in the sense in which they
are used in logic.6 We can speak of complex symbols, so ‘symbol’ and ‘expression’
can be used interchangeably, but we will often use ‘symbol’ when we focus on
a smallest unit in a symbol system. Expressions are types, and they are realized
in tokens, typically spoken, signed, or written. Expressions are not intrinsically
meaningful, and need not appear in any actual language. For example, ‘f&r©n#th§’
is an expression but does not have a meaning or appear in any natural language. As
we are interested in quotation in written language, we focus on written expressions,
that is, inscription types and their tokens. Inscription types are determinants of
expression types. The letters, words, phrases, and sentences in this paper (tokens
of which you are reading) are examples.
A characteristic feature of quotation, as we are interested in it, is that it involves
the construction of a term for an expression by incorporating the expression
5
There is nothing that prevents a similar device from being used in spoken or signed language, or
any other medium of communication. For example, the device in the artificial language Lojban (see
below) makes provision for spoken quotation, and in spoken languages, people will, transferring
the device designed for writing to speech, say “quote”, utter some word or phrase, and then
“unquote,” or use so-called finger quotes. But the device is especially well suited to the written
word and it is with the written word that it originated. Marks for indicating text trace their lineage
back to the second century B.C. in the diple, an arrow-like mark ‘>‘, used as a proofreading device
at the Library of Alexandria by the editor and librarian Aristarchus. But quotation marks in the form
widely used today attained their modern form and function only in the last half of the eighteenth
century, driven by experimentation with methods for setting off dialogue in novels (Houston 2013,
pp. 187–210; Johnson 2017, this volume). There are other uses (or abuses) of quotation marks,
such as scare (or shudder) quotes to indicate that the word in quotation marks is being used in
a non-standard sense or in a sense that the writer herself would not use the word or words to
express (e.g., the Onion headline ‘Jacques Derrida “dies”’), and quotation marks used to identify
a word being defined in a contextual definition, a purpose for which italics are also used (e.g., ‘A
boondoggle/“boondoggle” is a braided cord worn by Boy Scouts as a neckerchief slide, hatband,
or ornament’). Another use is the so-called emphatic use of quotation marks for emphasis in ads
and signs in the way that italics often are, as in, for example: ‘“Fresh” Seafood’, and ‘Lane may be
“slippery” due to oiling’. In these uses, while the quotation marks are used to draw attention to a
word, the word itself in quotation marks is not an object about which the sentences says anything,
and is not used in determining whether it is true or false. The quotation marks function a bit like
meta-remarks about the primary work that the sentence is doing but without being incorporated
into its content or being the content of a distinct meta-level speech act. In this respect, it is like the
conventional implicature of using ‘but’ in the place of ‘and’ to suggest a contrast between what is
expressed before and after ‘but’ (see Predelli 2003 on this in connection with scare quotes). We
therefore set these uses of quotation marks aside.
6
While it is an interesting question what the analysis of the concept of a symbol or expression is
in the sense in which it is used in logic, this is not a task we take up on in this paper, any more
than we take up the analysis of the concept of a language, or a word, or a sentence. Our focus is on
devices for referring to expressions, not the analysis of what they refer to.
itself. A standard form for quotation involves flanking an expression with other
expressions, which we call quotation marks. Using ‘ : : : ’ as a placeholder for an
expression, examples are:
‘ : : : ’ or “ : : : ” English
,:::” Dutch, Romanian, Polish
, : : : ” or ‘ : : : ’ Lithuanian, Macedonian, Icelandic
« : : : » or ‹ : : : › French, German, Russian
› : : : ‹ or » : : : « Hungarian, Polish, Danish
: : : or : : : Chinese, Japanese, Korean
lu : : : li’u Lojban (an artificial language)7
We will call expressions of these forms quotations or quotation terms. We will use
single and double quotation marks for illustration. When we intend pure quotation
henceforth we will use single quotation marks. We will use double quotation marks
for other forms of quotation. The treatment extends straightforwardly to other styles
of quotation marks, like those listed, as well as to using a special font, for example,
italics, or underlining/overlining.
In the following, we will use ‘QUOT(¥)’ to mean the result of performing an
appropriate syntactic operation on ¥ so as to yield a quotation term incorporating
¥ (where as above ‘¥’ is a metalinguistic variable that takes expressions as values).
In pure quotation, quotation marks flank an expression, and the quotation term so
formed refers to the expression enclosed in quotation marks. Call the position of ¥
in an expression QUOT(¥) a quotation context. Thus, we say an expression occurs
in a quotation context iff it occupies the position of ¥ in an expression of the form
QUOT(¥). For the purposes of this section, let this be understood as a pure quotation
context.
There are a number of observations that theories of pure quotation should
accommodate:
1. Quotation contexts are opaque, that is, (a) you cannot in general intersubstitute
coreferring or coextensive terms in quotation contexts salva veritate, (b) you
cannot existentially generalize, and (c) you cannot bind argument positions in ¥
when it appears in a quotation context from outside that context.
2. Quotation can be used to introduce new symbols or expressions and to refer to
expressions that are not in the language one is using, without introducing those
expressions independently of the quotation term.
3. In pure quotation, QUOT(¥) refers (non-accidentally) to the expression in its
quotation context.
7
Lojban is an artificial language based on formal logic. Its typographical quotation marks are by
design pronounceable, and intended to be used in speech in a way analogous to how they are used
in a written text. In distinction from natural languages, grammatical instances are restricted to
containing expressions in a recursively specified syntax for Lojban.
4. In understanding quotation devices, e.g., quotation marks, (a) one understands

how to generate and understand a potential infinity of new expressions, and (b)
one has the capacity to understand novel quotation terms, in the sense of being
able to determine what they refer to, on the basis of recognition of the quotation
term itself.8
Another condition sometimes placed on a theory of quotation is that it make sense
of the possibility that words can be used and mentioned simultaneously. This is not,
however, a feature of pure quotation. We return to the case of dual use and mention
in quotation below.
According to a standard taxonomy, the five main theories of pure quotation
are the name theory, the description theory, the demonstrative theory, the identity
(or use) theory, and the disquotational theory.9 We endorse what has been called
8
This is related to Quine’s remark that “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). While we aim to capture what seems right about this thought, that a quotation term contains
the type to which it refers, we do not treat quotation as literally picturing what it designates. The
device is actually simpler than that suggests.
9
A useful overview of the literature can be found in (Saka 2013). We don’t intend to review every
theory in the following. A referee noted pointedly that we do not discuss the pictorial or iconic
theory of pure quotation. ‘Pictorial theory of quotation’ does not, we think, express a well defined
category. The idea originates in the remark of Quine’s quoted in note 8, though what Quine says
hardly constitutes a theory. Harth (2011, p. 198) argues persuasively that the central idea is not
picturing so much as it is exemplification. Many accounts treat it as a condition of adequacy that
they explain why the appearance of an instance of what is referred in quotation marks is central to
how quotation works. Demonstrative theories (Davidson 1979), identity/use theories (see Harth),
and disquotation theories all seek to explain why the appearance of a token of the type referred
to in utterances of quotations is central to how the device works, though in different ways. In
this sense, one might say that they are all picture theories. But they are very different from one
another. One might give more weight to the idea that the expression in the quotation context
pictures what is being referred to by requiring that the mechanism of reference make some appeal
to a resemblance relation between that expression, or a token of it in speech, and what it refers
to. Perhaps the disquotation theory fares less well on this score, while the identity theory, since
identity is a paradigm of exact resemblance, does better, as does the demonstrative theory, though
in truth neither the identity theory (which invokes the idea of self-reference) nor the demonstrative
theory (which invokes the type-token relation) make much of the idea of resemblance per se. In
any case, we reject the need to invoke resemblance in the explanation of how quotation refers.
Another sort of theory that arguably gives a more central role to the iconic character of quotation is
the demonstration theory (as opposed to a demonstrative theory). On the demonstration theory, one
thinks of the quoted material as being demonstrated in the way in which one might demonstrate
an Australian accent (Clark and Gerrig 1990; Recanati 2000, 2001; De Brabanter 2002). These
theories focus on the production of an instance of what is to be referred to. However, it is still
the case that something, at least a token in use, has to refer, when the quotation occupies a noun-
phrase position in a sentence with truth conditions. So the question is just what the rule of reference
is which competent speakers learn that enables them to figure out what a (pure) quotation refers
to; and we will argue that while a token of the type referred to is indeed displayed, and in some
sense demonstrated (in the sense of the demonstration theory), the rule of reference is given in
(GQ) below, knowledge of which suffices to grasp everything one needs to know about how pure
quotation works.
the disquotational theory, if not everything that has been said about it. The
disquotational theory is easy to state. It asserts that we grasp everything that we
need to know about how pure quotation devices function in grasping the following
rule (GQ) (generalizing (Q)):
(GQ) For any expression ¥, QUOT(¥) refers to ¥

This handles all of the desiderata above. (GQ) obviously explains why QUOT(¥)
refers non-accidentally to the expression in the quotation context, #3. The fact
that intersubstituting in quotation contexts leads to a change of reference explains
why one can’t intersubstitute co-referring terms salva veritate, #1a. The fact that
the contained expression is not functioning semantically explains why you can’t
existentially generalize or bind argument places, #1b–c.10 The fact that the rule
quantifies over all expressions explains why it can be used to introduce new
expressions and refer to expressions in other languages (#2). It clearly explains
how grasp of the rule is sufficient to generate and understand a potential infinity
of new expressions and to determine their referents on the basis of recognition of
the quotation term itself, #4.
Before we explain this further and consider objections, we briefly review the
alternatives.
The proper name theory, which has been attributed to Quine (1940: 26) and
Tarski (1983: 159; though see note 4 above), is nowadays largely rejected, but it
is instructive, and helps bring out what is distinctive about quotation. It holds that
quotations are like ordinary proper names, such as ‘Julius Caesar’ or ‘Mohandas
Gandhi’. The proper name theory explains desideratum #1 because the contained
term is treated as part of a name’s spelling, like ‘bill’ in ‘billabong’. However, it
fails to accommodate desiderata #2–#4: it does not explain why QUOT(¥) refers to
¥ except as an accident of spelling, #3. For this reason, it cannot explain its function
in introducing new expressions (#2), since to figure out that the name refers to a
new expression it must be introduced independently. And for the same reason it
cannot explain how understanding quotation puts one in a position to understand
new quotation terms without having their referents introduced independently, #4.
The description theory (Geach 1957, p. 82; Quine 1960, p. 143, p. 212) retains an
element of the proper name theory, holding that there are primitive quotation names
of either words or letters and that strings of words or letters in quotation marks are
descriptions of their concatenation in the order in which they appear. For example,
‘Caesar crossed the Rubicon’ is interpreted either as the concatenation of ‘Caesar’
with ‘crossed’ : : : or the concatenation of ‘C’ with ‘a’ with ‘e’ : : : . But since this
retains the proper name theory at its core, it inherits its defects.
10
In connection with this, it is worth noting that, contra Saka (2013, p. 941), our view is not that
quotation marks represent a function that takes an argument and yields a value. As just noted, the
position inside quotation marks is not an argument position. It does not take referring terms. It
cannot be bound. And the rule (GQ) does not so represent it.
Davidson’s paratactic or demonstrative theory of quotation (Davidson 1979)

treats quotation marks semantically as a description containing a demonstrative.
Thus the quotation marks in a quotation are taken to be semantically equivalent to
a description, ‘The expression of which this is a token’. This handles desiderata #1
(since the expression is a sample to be demonstrated and not syntactically a part
of the sentence) and #2 (since anything can be the referent of a demonstrative) but
not #3 or #4b, since nothing constrains the referent to be the contained expression.
Another difficulty is that it is hard to see how to extend this to using italics or
underlining for the same purpose, since there is nothing separate from the token
expression itself to serve as the sample to be demonstrated (Reimer 1996, p. 135).
A feature that makes for trouble on this account is the use of the freewheeling
demonstrative ‘this’, and some difficulties could be removed by constraining it to
refer to the expression token in the quotation marks. But how do we do this? Do
we say: the expression type of which a token is contained within these tokens of
quotation marks? But now we have a demonstrative reference to token quotations
marks and similar difficulties can arise. What is wanted is a rule that uses a
description to determine a referent as in (DQ) (demonstrative quotation rule).
(DQ) For any expression ¥, for any utterance act u, if u is an utterance of [‘¥’], u
refers to the expression type of which the use of ¥ in u is a token
(Recall we are using ‘[’ and ‘]’ as the left and right Quinean corner quotes—see note
2.) Once we have got it in this form, however, it seems clear that the same effect is
achieved with a simplification, since by inspection the expression type of which the
use of ¥ is a token is just ¥:
(Q) For any expression ¥, [‘¥’] refers to ¥

This just is the disquotational theory, however, applied to single quotation marks.
From this perspective, the mistake Davidson made was to treat a rule for determining
a referent of an expression as if it gave the meaning of the expression. This is a
mistake equivalent to taking ‘I’ to mean ‘the speaker who is now using “I”’. This
is clearly a mistake since the proposition expressed by ‘I am sitting’ does not entail
that there are any speakers, since it can be true in a possible world in which I am
sitting but neither I nor anyone else is a speaker.
The identity or use theory of quotation (Johnson 2011; Recanati 2000, 2001;
Reimer 1996; Saka 1998; Washington 1992) takes quotation marks (or other similar
devices) to have a function similar to punctuation (see also Johnson 2017, this
volume). On this view, their purpose is to indicate that the expression in the
quotation context (or its token in a token quotation) is being used to refer to itself
(or the (or a) type of which it is an instance).11 Thus, it is not the quotation term
itself, but the contained expression that refers (or a token of it), and the function of
the quotation marks is, as it were, to disambiguate the use of the expression. The
quotation marks do not themselves refer, describe, or demonstrate, and neither does
the quotation term. One motivation for the identity or use theory, as opposed to the
disquotational theory, is that it treats the use of quotation marks as continuous with
verbal reference to expressions made by using them as in (6),
(6) Call me Ishmael

where the speaker is understood to be referring to the name ‘Ishmael’ rather than
(merely) using it. The identity or use theory treats the speaker as using ‘Ishmael’ (or
its token) to refer to itself (or the type of which it is a token). Another motivation
is that it allows quotation to be used to refer to tokens as well as types, and
types of various sorts, depending on the speaker’s intentions. As we will note, the
disquotational view can accommodate this.
The disquotational and identity/use theories are quite similar. There is a sense
in which the disquotational theory might be said to treat quotation marks as
punctuation—they are not given semantic significance independently of use around
an expression. The difference lies in what the rule for quotation marks treats as the
referring term. For the identity/use theory the rule goes as in (QI) (quotation rule for
the identity theory):
11
A referee objected that Recanati is not a use theorist because he denies that either the quotation, or
the quotation marks, or the expression between them, refers. It is true that Recanati (2001) does not
take open quotation, quotation not embedded in a sentence in a NP position, to involve linguistic
reference but only depiction. However, he takes closed quotation, in which the quotation appears in
an NP position in a sentence, which is what is under discussion here, to involve not only depiction,
but also linguistic reference (2001, sec. 2). There are some subtleties in his position. He thinks
of quotation (in use) as involving a demonstration, not in the sense of reference, but in the sense
of a display of a token of an expression type, in something like the way one might demonstrate
how to throw a knuckle ball by doing it. It is the demonstration that refers in closed quotation,
but not constitutively, since demonstration in the relevant sense is present in open quotation as
well and there is no linguistic reference in that case (on Recanati’s view). What is the relation of
the demonstration to the displayed token? Recanati could be clearer on this. At one point (2001,
pp. 650–651) he distinguishes the displayed token from the demonstration, and the demonstration,
in closed quotation, from the demonstration-qua-syntactically-recruited (whatever that is). But he
says both the displayed token and the demonstration-qua-syntactically-recruited have linguistic
meaning, presumably a referent (and the same referent, it seems). But later (2001, p. 655) he says
that it is only the demonstration (presumably qua-syntactically-recruited) that refers and the quoted
material is not part of the sentence (but here it is not clear whether he means type or token). In any
case, the demonstration in closed quotation is supposed to be a singular term and an iconic symbol,
and so to resemble what it refers to. So we can confidently say that on Recanati’s view there is a
token of some type produced in an utterance of a closed quotation that has as its referent that type
of which it is token and which is the type which appears in quotation marks in the quotation. This
seems to us to qualify it as an instance of the use theory as discussed here.
(QI) For any expression ¥, when ¥ appears in an expression of the form [‘¥’], ¥
refers to ¥.12,13
This is, on the face of it, a less straightforward way to understand the rule for
quotation, and seems gratuitous from the standpoint of semantics, since there is
no need to treat a substring of the string the rule applies to as the genuine referring
term. Moreover, this is clearly not something that is transparent to users of quotation
terms, and it is difficult to see what about our use of quotation terms would motivate
taking only the contained string to be the referring term. The motivation to preserve
continuity with examples like ‘Call me Ishmael’ (even granting in this sentence
‘Ishmael’ is used to refer to itself rather than being the deferred ostension of a
tacit demonstrative) is not adequate. First, historically, quotation marks are a device
that arose specifically in the context of written language (see note 5). There is
no reason to take it to be a device continuous with or derivative from any device
in spoken language. Second, in the development of written languages, from the
design standpoint it makes sense to introduce a syntactic device that functions
as characterized by (GQ). In contrast, from the design standpoint, (QI) seems
gratuitous and unmotivated.14 Moreover, we will argue that (GQ) proves to be
particularly fruitful in understanding how quotational devices extend beyond pure
quotation.
In sum, grasp of the rule (GQ).

suffices for anyone who understands it to understand any sentence in which a
quotation name appears (used for pure quotation). Nothing more needs to be added.
But if you do not understand this much, you do not understand quotation. This
suffices to explain the connection between the contained expression and the term’s
referent. It explains how we get from recognition of the term itself to its referent. It
explains how quotation generates an infinite number of expressions and puts one in a
position to understand them (when they are presented—see below). It explains how
12
The use theory also seems to give the wrong result for certain sentences. For example, if the use
theory is correct, we cannot truly say in English that ‘nurphalisturbia’ is not a word that has a use
in English, because that would ipso facto illustrate a use of the word in English. Yet, it seems we
can say that truly. The disquotational theory avoids this result. It is also, perhaps, slightly odd to
think that every expression (of Chinese, or Russian, or Arabic, or yet unimagined languages) is a
name of itself in English, which is a consequence of the theory.
13
For the version that would treat the token or its demonstration (see note 11) as the referring term:
for any expression ¥, any speaker u, any time t, and any token (or demonstration/presentation of)
¥* of ¥ such that ¥* appears in (or occurs in the production of) a token of [‘¥’] produced by u (in
a noun-phrase position in a sentence) at t, ¥* refers to ¥. We will not treat this separately because
it will be obvious from what is said about (QI) how to extend it to this.
14
The point of this is not to say that natural languages were designed or that the person who
introduces an innovation gets to stipulate its future use once taken up, but that there is, ceteris
paribus, no reason to take the device to be more complicated than what is required for it to perform
its function, and thus the burden of proof lies with someone who suggests a semantics that involves
more complications that are needed to accommodate the function of a device in the language.
it can be used to introduce new symbols and refer to expressions in other languages.
Importantly, it captures the sense in which quotation is, in Quine’s terms, like a
hieroglyph, but without the need to treat it as literally picturing itself.
What about the case of using italics (or underlining/overlining) to indicate that
one is referring to the italicized (or underlined/overlined) word? (GQ) applies to this
as well. Underlining and overlining are devices like quotation marks in the sense that
they involve a symbol external to the expression being referred to. But in the case of
italicizing a word, the italicized word is a determinant of the type that the italicized
expression refers to. In this case, the word, though italicized, is used to refer to a
type of which it is a token, though not to the more determinate type the italicized
word represents.
Sometimes the disquotational theory is said to treat quotation marks as a functor
that takes an expression as an argument and returns the expression as a value. This
is a mistake. The rule (GQ) tells you what the quotation term refers to on the basis
of how it is constructed. It carries no more commitments than that. To treat it as
assigning a function to quotation marks is to assimilate it to expressions that have
a different semantic role. The position of ‘¥’ in [‘¥’] is not an argument position
at all because it does not treat the expression that appears there as a referring term.
(GQ) gives a minimalist but sufficient account of the function of pure quotation. No
assimilation to other devices is needed: it is what it is and not another thing. Most of
the mistakes in the theory of quotation derive from trying to understand quotation
in terms of other devices of reference.
A straightforward and interesting consequence of (GQ) is that a language that
contains a quotation device that obeys (GQ) does not have a recursively definable
syntax because it contains a nondenumerably infinite number of semantically
primitive expressions.15 Call the basic vocabulary of a language minus quotation
terms (if any) its basic lexicon. The basic lexicon of any natural language is finite.16
Each item in the basic lexicon must be learned independently of the others. The
rest of the language, excluding again quotation names, can be understood on the
basis of understanding the basic lexicon and rules for their combination. (GQ)
introduces an additional class of expressions that refer to expressions. It does not
generate them from the basic lexicon together with a set of rules for combining
expressions.17 Instead, it quantifies over all expressions, including expressions that
are not in the language, to produce expressions in the language that are about those
15
It is also said to be a counterexample to compositionality because while putting quotation marks
around an expression to form a term referring to it is a syntactic rule, its “semantic value” isn’t
composed from the semantic values of its constituents. Some take this to be a reason to reject
the disquotational account, but since it is perfectly intelligible it is rather this particular form of
compositionality, or its application to quotation, that should be rejected. See Pagin and Westerstahl
(2010) for discussion.
16
What about numerals? Aren’t there an infinite number of them — even non-denumerably many,
since there is no reason in principle not to allow full decimal representations of real numbers?
But the basic numerals are ‘0’ through ‘9’, and the referents of complex numerals and decimal
expressions are given in relation to the primitive numerals.
17
As remarked in note 7, the artificial language Lojban is an exception, for it restricts quotation to
expressions otherwise in the language.
expressions. (This is another way to see why quotation cannot be assimilated to

functional expressions, which can take as argument terms only terms that refer in
the language.) Since the class of expressions is not recursively enumerable, the class
of quotation names is not recursively enumerable either.
In addition, nothing one grasps in grasping the basic lexicon and the rule for
quotation puts one in a position to understand every expression in the language,
specifically, the class of quotation names. It might be thought that this runs afoul
of Davidson’s learnability argument, the upshot of which was supposed to be that
grasp of a finite vocabulary of primitive items had to suffice for understanding
any potential utterance in a natural language on pain of making it unlearnable for
finite beings like us (Davidson 2001, pp. 8–9). Quotation names turn out to show
a limitation in the argument. For, although we are not in advance in a position to
understand every expression in the language, specifically all of the quotation terms,
given the way quotation works, as soon as we are presented with a quotation term,
we can use the rule to determine what it refers to, for in recognizing the quotation
term itself, we are in a position then to see what it refers to.
It has been objected that because the disquotational theory quantifies over
expressions, it may undergenerate quotation terms (Lepore 1999). The thought is
that any symbol can appear in quotation marks, whether or not this symbol is an
expression in any actual spoken language, or even in any possible language. As
noted above, however, we do not distinguish between symbols and expressions. In
addition, any symbol can be used in some language, even if it is not used in any
actual language. Apart from this, the objection is self-defeating because in offering it
the objector must use a term to capture the class of things that he thinks the quotation
device can be applied to which is larger than the class of expressions. This supplies
the proponent of the disquotational theory with a term to use in place of ‘expression’
that will capture the right category of items (supposing the objector is correct).
Finally, one of the themes of the identity/use theory of quotation is that quotation
terms can be used to refer to different sorts of things, not just expression types.
If we think about how the identity/use theorist has to characterize his own rule to
accommodate this, however, we can see that whatever he can say can be adopted by
the disquotational theory. There are two basic ways to do it. First, one can argue that
quotation marks are ambiguous, and the different uses signal different sorts of things
as the referents. Disambiguating would amount to supplying for each different use
a different target. For example, we might introduce a use of quotation marks to refer
to tokens of types that appear in the quotation marks. Suppose that we use asterisks
as quotation marks that are used to refer to the token of the expression that appears
between them. We can give the following clause:
(Q*) For any expression ¥, for any speaker u, time t, [*¥*] as used by u at t refers
to the token of ¥ produced by u at t in producing a token of [*¥*]
For expression types individuated according to different standards, we can replace
the restriction on ¥ in the first quantifier with a term that expresses the appropriate
notion. Second, one could treat quotation marks as context sensitive, with the kind
of thing referred to being determined relative to the speaker’s intention. We still
need to have in mind a class of items over which we can quantify and in relation
to which we can locate other types. So we still need to use ‘expression’ in some
suitable sense in the restriction on our quantifier. But this is fine as long as whatever
else the speaker has in mind can be located in relation to it. So, for example, we can
give the rule (QCS) (quotation as context sensitive):
(QCS) For any expression ¥, for any speaker u, time t, QUOT(¥) as used by u at t
refers to the type or token that bears the relation intended by u at t to the
token of ¥ produced by u at t
Having said this, we add that we think pure quotation is conventionally used to refer
to expression types, and we assume this in the following.18
It is sometime said that the disquotational theory, while suitable for pure
quotation, can’t be the correct theory of quotation because it cannot handle cases
beyond (1) and possibly simple cases of direct quotation as in (2a). In our view, this
desideratum is confused. Quotation, as we said in the introduction, is polysemous. It
has a variety of related uses. The conceptual core of quotation, however, is captured
in the disquotation rule, (GQ), for pure quotation. To try to give one rule for all
the forms of quotation, as opposed to explaining them in relation to the core use,
would be like trying to give a single definition for ‘walk’ that accommodated all
the following uses: ‘I went for a walk’, ‘I walked the dog’, ‘We’ll have to walk
the wardrobe to the bedroom rather than carry it’, ‘She walks the ramparts’, ‘The
workers threatened to walk’. Only confusion can result. In the following, we explain
the uses found in (2)–(4) as exploiting the device of pure quotation for further
purposes.
3 Quotation in Which Contained Expressions Are Intended

to Be Understood
In this section we discuss the varieties of quotation found in examples (2)–(3). We

begin with the idea that in using a quotation term, it can be crucial to its function
that it contains a meaningful expression that is both understood and understood to
be referred to, though the meanings of the quoted expressions do not contribute to
fixing the truth conditions of the sentence containing the quotation term.
18
One might object that we can write on the blackboard: ‘This’ is written in chalk. Well, we get the
idea, of course, but it helps that ‘this’ is the word written! Consider: ‘Boston’ is written in chalk.
Puzzling! We feel like saying: sometimes yes, sometimes no. But then this is about the word being
written in chalk (a chalk token produced) on this or that occasion.
3.1 Quasi-Use-Mention
We begin with (2), repeated here,
(2) “psychology” literally means “the study of the soul.”19

It will be useful in considering the function of quotation in (2) to first consider the
use of the following sentences:
(7) ‘Schnee ist Weiss’ in German means that snow is white
(8) ‘Schnee’ in German means the same as ‘neige’ in French.
(9) ‘Schnee’ in German means snow
(10) ‘Schnee’ in German means snow
In (7) the subject term is simply used to refer to a German sentence, and it is not
presupposed that we know its meaning because the point of (7) is to explain it. But
how does (7) do this? The traditional answer is that the complement, ‘that snow is
white’, refers to the proposition that the German sentence expresses, and by grasping
that proposition and associating it with the German sentence, we come to know
what it means. However, how do we grasp the proposition? Merely referring to
the proposition isn’t sufficient, for if we had named the proposition that snow is
white ‘Betty’ and substituted that for the complement, when we told someone what
‘Schnee ist Weiss’ meant, she would be none the wiser. A Fregean might say that
there is a mode of presentation attached to ‘that snow is white’ that suffices for one
to grasp what it refers to, but this is mysterious—what mode of presentation ipso
facto suffices for entertaining its object if it is a proposition? And how would this
explain the importance of the appearance of the sentence itself in the complement
rather than a name?
There is a simpler answer to the question how (7) informs us of what ‘Schnee ist
Weiss’ means in German, namely, that (i) it is a condition on the truth of (7) that
the sentence in the complement translate ‘Schnee ist Weiss’ and (ii) we understand
that sentence. Once we see this, we can also see that what the complement refers to,
if anything, is irrelevant to the work that the sentence does for us. It might as well
be the sentence itself rather than a proposition, for all the use that the proposition
is to us (Ludwig 2014). So here we have a device for conveying the meaning of a
sentence that involves using a sentence in our language, where it is crucial for the
work that is to be done that our interlocutor understand the sentence, even though the
extensional properties of the sentence do not contribute in any way to determining
the truth conditions of the containing sentence.
Now consider in contrast (8), where we take the quotation terms to be pure
quotation. In this case, it is clear that one could understand the sentence without
19
American printers’ conventions place commas and periods that are part of the containing
sentence inside quotation marks rather than outside. We ignore this in the accounts we develop.
A simple syntactic rewrite of the sentences yields the form that serves as input to the rules given in
the following.
understanding either of the quoted expressions and so remain in the dark about
the meaning of either term. This stands in contrast to (9) and (10) which we
use to explain what a term in German means to someone who speaks English.
The difference is simply that we presuppose that our audience understands the
complement expressions (and so assume that the audience doesn’t fully understand
the sentence without doing so) and that the sentences are true only if the term
appearing to the right of the verb means the same as the term mentioned on the
left. In (8) we relativize the meaning relation to both German and French. In (9) and
(10) we relativize only the term mentioned on the left because in using the terms
on the right in English we fix their interpretations. In (9) we use italics to indicate
that the expression to the right of ‘means’ is being used to give the meaning of the
expression on the left, that is, that it is both to be understood and to be understood
as the same in meaning as the expression on the left. In (10) the same information
is carried simply by context.
With this in mind, when we consider (2), we can see that while pure quotation
is being used on the left, the expression quoted on the right is intended to be
understood, for this is a sentence we use to convey the meaning of one expression by
using another. This use of quotation involves reference to the contained expression,
and that reference is secured by (GQ), but there is another dimension to the use
which, like the function of the complement in (7), does not come out in thinking
about the contribution of the expression to the truth conditions of the sentence. We
call this a quasi-use of the expression, by which we mean that it is essential to its
function in the context that it be understood as used by the speaker (this, e.g., secures
disambiguation and fixes the contributions to meaning by context sensitive terms),
but the extensional properties of the expression contained do not contribute to fixing
the truth conditions of the sentence. We will call quotation of the sort illustrated on
the right side of ‘means’ in (2) quasi-use quotation.
3.2 Direct Discourse
With this in hand we turn to the use of quotation with direct discourse. A common
view about direct discourse is that the function of a sentence such as (3a).
(3) (a) He said, “Get serious, boy.”

is to give a literal rendering of the words spoken by someone on a particular
occasion. However, the practice is more complicated than this. In particular, we will
distinguish between strict and non-strict direct discourse. In strict direct discourse,
the sentence of direct discourse is true iff its subject bears the relation expressed
by the verb to the sentence in quotation marks (excluding punctuation introduced
solely as a result of printers’ conventions).
The strict use is illustrated in (3b–d).
(3) (b) She said, “Gorse is common in Scotland”; she did not say, “Furze is
common in Scotland.”
(c) Caesar literally said, “Veni, Vidi, Vici,” not “I came, I saw, I conquered.”
(d) He said, ‘All mimsy were the borogoves,” but didn’t have anything in
mind by it.
Here what is important are the words themselves rather than any translation of them.
‘Gorse’ and ‘Furze’ mean the same, but switching ‘Gorse’ and ‘Furze’ in (3b) may
change a truth into a falsehood. Similarly, in (3c), ‘I came, I saw, I conquered’ is
a translation of ‘Veni, Vidi, Vici’, but a contrast is being drawn. In (3d) clearly
the meaning of the words can’t be at issue because it is a nonsense sentence. For
strict direct discourse, we can give the truth conditions as in (DQS) (direct quotation
strict), where ‘ref(’, u, t)’ expresses a reference function from a name, speaker and
time to a referent of the name as used by the speaker at that time, that is,
For any x, u, t, ref (’, u, t) D x iff ’ as used by u at t refers to x,
and we relativize the truth predicate to the speaker and time as well.
(DQS) For any speaker u, time t, expression ¥, name ’, [’ said, “¥”] is true(u, t) iff
there is a time t’ < t such that ref (’, u, t) says ¥ at t’.
But not all uses are strict. First, very often understanding the expression quoted
is essential to the function of the sentence, as in dialogue in a novel. This is the
case for (3a). Thus, it must be construed as involving a quasi-use of the contained
expressions. This does not show up in the specification of the truth conditions, but
is an aspect of the conventions associated with the use of direct discourse in certain
uses. Ideally these would be marked with a special sort of quotation mark, but in
practice context disambiguates.
Second, in direct discourse we also often use words in our language to report
what someone said in another language as in (3e).
(3) (e) Then Jesus said to them, “Render to Caesar the things that are Caesar’s,
and to God the things that are God’s.”
In this case, we use a translation of the words of Jesus (as reported in Mark) into
our language, but they are not the words that Jesus used. And despite the contrast
with (3c) above, there is nothing problematic about this use. Thus, in the non-
strict use of direct quotation, we intend a quasi-use, and that the subject have
uttered the sentence referred to or a sentence that translates it. The first of these
requirements, as mentioned, does not show up in the truth conditions. The second
is accommodated in (DQNS) (direct quotation non-strict), in which we introduce
the standard event analysis of action sentences, which treats the action verb as
introducing an existential event quantifier (Ludwig 2010). For this purpose, we use
the following abbreviation (where ‘¢’ is a variable over sentences):
says(e, ref(’, u, t), t’, ¢) D ref(’, u, t) is an agent of e at t’ and e is an utterance

and e is an instance of ¢.
We also make explicit relativization of the semantic predicates to languages. We
introduce this additional complexity because we need to relativize the translation of
¥ to the language the subject used in his utterance of ¢, and there may be no unique
language he speaks in which ¢ is a sentence. The truth conditions are then given in
(DQNS).
(DQNS) For any speaker u, time t, expression ¥, name ’, [’ said, “¥”] is true(u, t,
English) iff there is an event e, a time t0 < t, and a sentence ¢, such that
(i) says(e, ref(’, u, t), t0 , ¢) and
(ii) ¥ in English translates ¢ in the language of ref(’ u, t) at t0 in e.20
(DQNS) accommodates the case in which the subject uses the same sentence as the
speaker, since every sentence is a translation of itself.
There is a further complication presented by the practice of using punctuation
and ellipses in direct quotation (of oral speech) as in (3f).
(3) (f) “Good morning! it was kind of you to push the chair up that hill ... I hope
it wasn’t heavy for you,” said Connie, looking back at the keeper outside the
door.
The exclamation mark indicates something about the mode of delivery. While in
pure quotation [‘¥!’] refers to an expression that has an exclamation mark at the
end, in direct quotation, reporting oral speech, the exclamation mark indicates a non-
symbolic feature of the utterance act type. But this does not need special treatment
because the type picked out by [‘¥!’] has one realization in written language and
another in spoken language. The same extends to question marks, italics, boldface,
caps, and other devices such as repeating a letter to indicate dragging out a syllable,
as in, ‘That was soooo boring’. These indicate aspects of the utterance type that
extend beyond what words were used to include how they were used.
Ellipsis comes in two varieties.21
(i) It can be used to indicate that words in the original are omitted. This is
illustrated in (3g).
(g) Davidson said, “Quotation is a device used to refer to typographical : : :
shapes by exhibiting samples.”
We call this omission ellipsis. Omission ellipsis can be accommodated by condi-
tionalizing on whether or not this sort of ellipsis occurs in a quotation in direct
20
It might be objected that ‘x said : : : ’ doesn’t contain a reference to English in its analysis. Then
replace ‘English’ with ‘the language of u at t in uttering [’ said, “¥”]’.
21
Of course, an ellipsis mark may be a part of an expression referred to in quotation marks. One can
say, for example: when he wrote, “Well : : : I wouldn’t go so far as to say it is entirely worthless!”,
he used the ellipsis ‘ : : : ’ to convey a hesitation of his thought in writing what he did.
discourse and requiring that the subject (Davidson in the case of (1g)) have uttered
a sentence or a translation of a sentence obtained from the sentence with the ellipsis
by inclusion of some additional material in the position of the ellipsis. We will not
write out the clauses.
(II) But in (3f) that is not the function of the ellipsis. The ellipsis indicates
neither that something is omitted nor that the subject uttered or wrote anything that
corresponds to the ellipsis ‘ : : : ’, but just that there was a temporal gap (a hesitation)
between utterance corresponding to the material before the ellipsis in the quotation
and the material after it, though it is intended as a single speech act.
For the case of ellipsis of the sort illustrated in (3f), one might initially think that
we could treat.
x said “—- : : : —-”
as roughly equivalent to
x said “—-” followed by a pause and then “—-”
This could be incorporated straightforwardly into a truth conditional account. But an

ellipsis can occur anywhere, including multiple times in a single sentence such as:
He said, “I think : : : you really don’t want to : : : tease the sasquatch.” To report this
as his saying “I think” followed by a pause, then “you really don’t want to” followed
by a pause, then “tease the sasquatch,” is compatible with his not having said, “I
think you really don’t want to tease the sasquatch.” It is more straightforward to
treat ellipsis in this use—like an exclamation mark, or italics, or a question mark—
as indicating something about the utterance type beyond what words in what order
are included in it, namely, that it includes somewhat greater than usual temporal
gaps between portions of what is uttered.
Importantly, for all these ways of indicating something about the utterance type,
we have to take translation, as it figures in (DQNS), to require preservation of
devices indicating something about the type of utterance beyond its content.
Dialogue in fiction or in reports where the context determines who is speaking
(typically called free direct speech) we treat as translatable into (that is, shorthand
for) direct discourse.
3.3 Direct Attitude Attributions
A striking use of quotation is in direct attributions of thought. These may also

include ellipses and punctuation like exclamation marks and question marks inside
quotation marks. In these cases, we treat the sentences as if they expressed a thought
someone is having or had or will have. We will treat the sentences understood
relative to the thinker and time of thought as expressing the contents of occurrent
thoughts of the thinker (the subject of the sentence). As in (3h), the sentence that is
used to capture the content of a thought may be split between two quotations. First
we join the quotations into one, as in (3h.i).
(3h) (i) Alice thought, “And what is the use of a book without pictures or
conversation?”
This can be thought of as a purely syntactic operation performed before interpre-
tation.22 Then we give the truth conditions for ‘’ thought, “¥”’ in (DQT) (direct
quotation thought).
(DQT) For any speaker u, time t, expression ¥, [’ thought, “¥”] is true(u, t) iff
there is a time t0 < t, and a state s, such that
(i) thinks(s, ref(’), t0 ) and
(ii) Š(s, ¥, ref(’), t0 ))
The variable ‘s’ takes states as values. The relational predicate ‘Š(s, ¥, x, t)’
expresses a sameness-of-content relation between ¥ (in English) interpreted relative
to x at t (that is, with x and t as values for the contextual parameters), on the one
hand, and x’s thought s, on the other. We wish this to be interpreted broadly so that it
subsumes both sameness of propositional content where appropriate but also inner
exclamations, thanks, pauses, and so on, since we wish to accommodate examples
such as (11) and (12).
(11) “Thank goodness!” he thought to himself.

(12) “Well done,” he thought as he smiled.
In this connection, the use of the exclamation point can be taken to reflect an aspect
of the state associated with what we might call the reception of its content—it
marks the thought as emphatic, as we might put it. A question mark indicates that
the psychological mode is “interrogative”—that is, ‘“Which one is it?” he thought’
means the same as ‘“Which one is it?” he wondered’. An ellipsis indicates a pause
(or hesitation) in a thought. One way of thinking about it is to think of the relation
introduced as aiming to capture something about the way that the thinker him- or
22
A referee suggested that (3h) and (3hi) are not equivalent on the grounds that in (3h) ‘Alice
thought’ is syntactically parenthetical and that the verb ‘thought’ therefore cannot take the
quotation as its object. As evidence, the reader suggested that it is odd to say in response to (3h)
That’s not true! but it is not odd in response to (3hi). It seems to us to be an odd thing to say in
response to either in ordinary conversational contexts, as opposed to saying ‘That’s not what she
thought’, but the oddness seems to come from its being unclear whether the response is focused
on the sentence mentioned rather than the sentence it is embedded in, which is not a problem with
‘That’s not what she thought’. Vary the example, though. Consider [*].
[*] “Insanity is doing the same thing over and over again,” Einstein said, “and expecting
different results.”
There is nothing odd about responding, “That’s not correct. While it is a common misattribution,
Einstein never said that.” In any case, there are other tests for whether a shifted clause like this
takes scope over the sentence. It is perfectly acceptable to report what someone said in uttering [*]
by saying: He said that Einstein said, “Insanity is doing the same thing over and over again and
expecting different results.” This would be false, however, if the sentence were not in the scope of
‘Einstein said’. The same thing goes for (3h) and (11) and (12).
herself would express the thought at the time in her language—that if she verbally
expressed the thought at the time of thinking it, in a play-by-play report, as it
were, what she said could be reported in direct discourse using the sentence used to
characterize it. This would mean that such attributions would be limited to linguistic
beings, but on reflection this seems plausible, because it is quite difficult to make
sense of attributing a thought to a nonlinguistic animal using (3h) or (11) or (12),
except under the pretense that it speaks a language or thinks in a language.
An alternative would be to think of these attributions as requiring that the subject
(thinker) explicitly be thinking in words, in the way that we sometimes have a
particular sentence in mind or a particular sentence occurs to us which we do
not utter (‘What a nincompoop!’). However, since we may also want to attribute
something like (12) to someone without supposing that he has in mind, in this way,
the words for expressing the thought, but merely that if he did express it, using those
words to report what he said would be apt, this cannot cover all cases. But there may
be a use that aims precisely to get at having a particular “thought sentence” in mind.
To capture this, we can substitute for ‘thinks’ a relation that expresses being in a
state of verbally thinking a thought at a time, and change the equivalence relation to
require the state to be not just like in content but to be like in respect of involving a
thought sentence that is the same as (on the strict reading) or a translation of (on the
non-strict reading) the sentence in quotation marks.
One difference between direct discourse and direct attribution of thoughts is
worth drawing attention to. In direct discourse, we relate a sentence ¥ used by
the person uttering it to a sentence ¢ uttered by its subject, requiring that the one
be the same as or a translation of the other, where this just requires a translation
manual, not that the sentences be interpreted relative to context (they may contain
nonsense terms even). In direct thought attributions we interpret the complement
sentence relative to the thinker and time of the thought because we are interested in
sameness of propositional content between a sentence and a state. This ensures that
indexical elements like the ‘I’, in the complement of (3i), and demonstratives and
tense are correctly interpreted for the purpose of conveying the content and mode of
the thought.
(3) (i) “If I had a child!” she thought to herself; “if I had him inside me as a
child!”
This can be extended to inner dialogues in fiction and reporting where the context
determines who the thinker is (that is, free direct thought attributions) by treating
them as translatable into (shorthand for) direct attributions.
4 Quotation in Exposition
The examples in (4) involve exposition in which an author is expressing something

in his own voice, but in part using the words of another, as in (4a–b), or a translation
of the words of another, as in (4c).
(4) (a) In this chapter, Mill attempts to delineate when the authority of society
can rightly limit individuality and the “sovereignty of the individual over
himself.”
(b) Berkeley’s objective in the New Theory of Vision was “to shew the
manner wherein we perceive by sight the distance, magnitude, and situation
of objects. Also to consider the difference there is betwixt the ideas of sight
and touch, and whether there be any idea common to both senses” (NTV §1).
(c) They substituted, as Kant has pointed out, “a physiology of the human
understanding” for the Critical investigation of the claims of reason, and
anthropology for ethics.23
We wish first to say that the speaker (of any of (4a–c)) asserts what is expressed by
the sentence stripped of a first layer of quotation marks. Then we wish to say that
the words in quotation marks in the sentence, or a translation of them, were used
by another, whom we will call ‘the subject’, as indicated in the sentence, or by a
citation, or the context. We will not require that the subject have uttered a sentence
that expresses the same thing as what the speaker utters, however. For one may say,
in one’s own voice, e.g.,
I reject, in distinction from Kant, the mere “empirical reality of time” and accept its
“absolute and transcendental reality.”
In addition, one may use words or phrases in a single sentence drawn from different
sentences and even different texts. We will call this variety of quotation ‘scholarly
quotation’ and the quotation marks in this use ‘scholar quotes’.
A question arises about whether we should put into the truth conditions of
sentences like (4a–c) the implication that the subject (of the clause) used the words
in quotation marks, or treat it as a conventional implicature, the truth conditions
being given by the sentence stripped of one layer of quotation marks. In the latter
case, it is a requirement for felicitous utterance, but not for truth, that the subject
mentioned used those words or a translation of them. We will just represent truth
conditions for sentences containing scholarly quotation as in (SQ), putting the
implication into the truth conditions. The alternative would treat (i) in (SQ) as giving
the truth conditions and (ii) as giving the content of the conventional implicature.
For this purpose, we introduce two operators for removing and putting quotations
on expressions, ‘UNQ’ and ‘QUO’. (We will find these useful in the next section on
23
We do not assimilate this to indirect discourse. The parenthetical ‘as Kant pointed out’ is a
comment on what the writer is saying, which is given by the sentence sans phrase and quotation
marks.
mixed quotation in indirect discourse as well.) UNQ(¥) is the result of removing

one “layer” of double-quotation marks from ¥ together with any square brackets
surrounding material inside an outermost layer of quotation marks (cf. Maier
2017). QUO(”) is the result of adding double quotation-marks at both ends of
the expression ” (cf. Gomez-Torrente 2017). Thus UNQ(‘“religious” leaders are
“creatures of exalted emotional sensibility”’) D ‘religious leaders are individuals
of exalted emotional sensibility’. QUO(‘creatures of exalted emotional sensibility’)
D ‘“creatures of exalted emotional sensibility”’. If there are no double quotation
marks in ¥, then UNQ(¥) D ¥. To cut a corner, let SUB(¥, u, t, ”) D the person or
text to which the quoted words ” in ¥ are referred relative to u and t.24
(SQ) For any speaker u, time t, atomic sentence ¥ that contains scholar quotes,
¥ is true(u, t, English) iff
(i) UNQ(¥) is true(u, t, English) and
(ii) For all ” such that QUO(”) occurs in ¥, there is a ” 0 in a language of
SUB(¥, u, t, ”) that is a translation of ” in English and is used by (or in)
SUB(¥, u, t, ”).
We treat ‘is used by’ as tenseless, that is, as equivalent to ‘has been, is, or will be
used at some time or other by’.25
5 Mixed Indirect Discourse and Attitude Attribution
5.1 Mixed Indirect Discourse
We begin with a simple case of mixed indirect discourse and quotation (dubbed
mixed quotation by Cappelen and Lepore (1997).26
(5a) Macomber said that he “bolted like a rabbit.”

Intuitively, (5a) says that (i) Macomber said that he bolted like a rabbit and (ii) he
used “bolted like a rabbit” (or perhaps something that translates that) in saying that
24
There is no simple or uniform syntactic method of identifying the person to whom or text to
which the quoted words are referred. Sometimes it is the subject of the sentence (that is, who the
sentence is about) as in (4a–b). Sometimes an adverbial clause provides the information, as in (4c).
Sometimes a citation to a text is given at the end of a sentence. Sometimes it is given in a footnote.
Sometimes it is left entirely to the context.
25
In Sect. 5, we give a treatment for authorial brackets used in mixed quotation (as in (5c)). The
treatment there can be adapted to scholarly quotation. If “monstrous” operators are to be tolerated,
the treatment in Sect. 6 could be adapted for scholarly quotation. It should be clear how the
extensions go from the treatment below.
26
Of course, scare quotes may be used in the complements of verbs of indirect discourse, but
that has a quite different point relating to the attitude of the speaker to words appearing in the
complement rather than to words that the subject used. See note 5.
of himself (given that the report is of his narration of what he did). We take this to
be a semantic phenomenon. We learn how to interpret the appearance of quotation
marks in complements of indirect discourse sentences in much the same way we
learn how to interpret other constructions, and what we learn can be expressed with
a rule. The goal of this section is to state the rule.
However, we cannot always treat the quoted expression as something the subject
is saying of himself (it may make no reference to the subject at all). So the gloss on
(5a) cannot stand in as a general treatment. In addition, (5b) shows that there may
be multiple instances of quoted material in the complement clause.
(5b) Alice said that she had “heard nonsense” compared with which that would be
“as sensible as a dictionary.”
In giving an account of how the subject of the sentence expressed what is attributed
to her, we must keep track of the grammatical role of the various quoted expressions
in the complement. In fact, given that the same word can be quoted multiple times,
in different roles, it is clear that merely saying that the word was used in the subject’s
saying what she did will not convey what is intended.
(5c) illustrates how an expression may be substituted in the quoted expression
(here in brackets) so that the referent interpreted relative to the speaker corresponds
to the referent of the word used by the subject of the attribution, since in indirect
discourse the complement sentence is, except for the tense, which is governed by the
event time for the main verb, interpreted relative to the speaker’s context (though
see Sect. 6).
(5c) She said that it had been many years since “such trifles had broke across the
web of [her] solitude.”
(5d) illustrates a case in which the subject of the attribution did not use the words in
quotation marks at all.
(5d) La Rochefoucauld said that jealousy “ends as soon as we pass from
suspicion to certainty.”
In this case, the function of the quoted expression is to indicate that these words
are, or are a translation of, the words that the subject used. There are, however,
cases in which it looks like a stricter standard is called for, and so in the case of
indirect discourse, as for direct discourse, we will distinguish between a strict and
a non-strict reading. We will begin with the non-strict reading, and return below to
examples that suggest that a stricter reading is also sometimes intended.
In Ludwig and Ray (1998), we offered a general account of the multiple appear-
ances of quotation in indirect discourse against the background of a sententialist
account of indirect discourse and attitude sentences, but we did not extend it to
interpolations as in (5c) or to using translations of expressions used by the subject
as in (5d). We refine that account here (and fix a few things). We make use of the two
operators introduced in the previous section for removing and putting quotations on
expressions, ‘UNQ’ and ‘QUO’.
A sententialist account contrasts with a propositionalist account of indirect
discourse (and attitude attributions). On the sententialist account, the complement
clause in a report of indirect discourse refers to the contained sentence and
characterizes what the subject said in terms of a content equivalence relation. On the
propositionalist account, it refers to the proposition expressed by the complement,
which is said to be the same as that expressed by something the subject said. While
we prefer the more minimalist sententialist account, after we present the refinement
and extension of our earlier account, we will sketch how to extend the same ideas to
a propositional account, which requires only minor modifications.
On the sententialist account given in Ludwig and Ray (1998), we treated ‘x said
that ¥’ as relating x to the sentence ¥ understood relative to the speaker of the
sentence and time of utterance. In the following, we use the event analysis of action
verbs (as we did for direct discourse) in order to secure that the two clauses are
linked in the right way. For this purpose we use.
saying(e, t0 , x, UNQ(¥), u, t)
to abbreviate e is a saying by x at t0 which samesays UNQ(¥) understood relative
to u at t.
where e is an event, x is the agent of e, t0 is the time at which x is the agent e, t is the
time of utterance, u is the speaker, and ‘samesays’ is a possibly context-sensitive
equivalence relation that relates utterances and sentences.27 For present purposes
it is not necessary to settle the details of how to interpret the samesaying relation,
and, in particular, whether and how it differs from a generalization of the synonymy
relation.
We will develop the account in two stages. In the first stage we ignore pronouns,
and indexicals28 in the complement sentence that need reinterpretation across
contexts, in order to illustrate how to accommodate multiple appearances of quoted
material in the complement. In the second stage, we will fix the problem this leaves
us with pronouns and indexicals. This will provide a template for other verbs of
indirect discourse.
Stage 1. Ignoring pronouns, etc., in the complement, using the notation just
introduced, (MID) (mixed indirect) provides an account of the truth conditions of
indirect discourse using the verb ‘to say’ whether involving mixed quotation or not.
27
Davidson used ‘samesaying’ to relate speakers (1968, pp. 140–141) in an informal rendering of
his account of indirect discourse. Lepore and Loewer appropriated it for use as relating utterances
in a defense of Davidson’s paratactic account (1989, 343), and it has been pressed into service (in
a generalization) as the equivalence relation invoked by ‘says’ between the subject’s utterance and
whatever the complement clause refers to in indirect discourse ever since.
28
Under ‘indexicals’ we include not just what Kaplan called pure indexicals (like ‘I’, ‘now’,
‘today’, etc., whose referents are determined fully relative to a specification of contextual
parameters) but also (what he called true) demonstratives (like ‘this’, ‘that’, ‘then’, ‘there’, and
demonstrative uses of pronouns).
(MID) For all speakers u, times t, sentences F, names ’, [’ said that F] is true(u,
t) iff there is an e, and a time t0 < t, such that
(a) saying(e, t0 , ref (’, u, t), UNQ(F), u, t) and
(b) If UNQ(F) ¤ F, then
(i) There is a sentence ™ such that e is an utterance of ™ and ™ in the
language of ref(a, u, t) at t0 in e, relative to ref(’, u, t) at t0 , is a
translation of UNQ(F) understood relative to u at t,29 and
(ii) For all expressions ”, and formulas §, if F D the result of replacing
‘x’ in § with QUO(”), then
(iii) A translation of ” into the language of ™ as used by ref(’, u, t) has
the same grammatical role in ™ as QUO(”) has in F.
Clause (a) ensures that the subject says what the sentence in the complement says
minus the first layer of quotation marks. Clause (b) ensures that wherever a term
appears in quotation marks in the complement, there is a translation of it that was
used in the same role in a sentence the subject used to express what he did.
Stage 2. Pronouns and indexicals used in the complement clause, as in (13) and
(14),
(13) Betty said that she was “pixelated”

(14) Betty said that I was “pixelated”
make for some complication because in giving the truth conditions for (13) we want
to say that Betty uttered, not a translation of ‘she was pixelated’, but rather of ‘I
am pixelated’, using ‘pixelated’. Similarly, for (14), we want to say that she uttered
not a translation of ‘I was pixelated’ but of a sentence such as ‘You are pixelated’
using ‘pixelated’. In the case of a pronoun cross-indexed with the subject, we want
to shift specifically to a sentence in which the first-person pronoun is used. This can
be incorporated into (MID) by introducing in the place of ‘translation of UNQ(F)’
a defined term, ‘index-shifted translation of F’.
(Def) q0 understood relative to x at t0 is an index-shifted translation of q
understood relative to u at t iff q0 is a mere indexical-variant of a
translation of UNQ(q) such that
(i) for every occurrence of an indexical g in q there is in the same
grammatical role in q0 an occurrence of an indexical (or directly referring
term) d such that d understood relative to u0 at t0 refers to the same thing as
g understood relative to u at t
(ii) and if g is co-indexed with the subject position, then d is the
first-person pronoun agreeing in number with g.
29
We let ‘understood relative to u at t’ do double duty, fixing both interpretation of the sentence
and contextual parameters. See the discussion below of the use of words in complements that
aren’t strictly in English. Ultimately, in a truth-theoretic semantics, we would wish to introduce
the speech act the speaker performs as another contextual parameter, in which case we could add a
further relativization to the use of the sentence in the speech act itself.
To say that q0 is a mere indexical-variant of a translation of UNQ(q) is just to say

that but for possibly these indexical shifts, q0 does not differ from a translation of
UNQ(q). By way of example, suppose when A greets you this morning, A says,
“You were late for a very important date yesterday”. Then ‘I am late for a very
important date today’ understood relative to you and yesterday is an index-shifted
translation relative to A today. Together with the definition of ‘UNQ(x)’ this handles
(5c).
For a propositionalist account of the function of the complement, we can replace
(a) with (a0 ).
(a0 ) saying(e, ’, ref([that UNQ(¥)], u, t)) : : :
We relativize the reference of [that UNQ(¥)] to speaker and time because the
proposition it designates is determined in general relative to the interpretation of
UNQ(¥) relative to the context of utterance.
The approach above extends naturally to cases involving quantification into the
complement clause, as in
There is something such that Galileo said that it moves.
On a sententialist account, we want to say that this is true just in case an assertion
of Galileo’s samesays a completion of the sentence form ‘x moves’, i.e., a sentence
in which ‘x’ is replaced by a singular referring term.30 However, we must provide
a way of generalizing this requirement that allows for the possibility that there is
no completion in English of the sentence form ‘x moves’ that samesays with any
sentence that Galileo uttered. This can be done by providing a satisfaction clause
for open sentences along the lines of (MIDQ) (mixed indirect quantification).
(MIDQ) For all sequences f, all speakers u, times t, formulas Fx, names ’, [’ said
that Fx] is satisfied by f relative to u and t iff ...
The remainder of this condition would be just as in (MID), but in place of ‘F’
we will put ‘F*’ which latter would be shorthand for the result of replacing ‘x’ in F
with a constant b in a language that extends the language of F only by the addition
of b, and in which b refers to f (‘x’).
Clearly, the approach could be generalized to formulas with any number of
free variables. It should be noted that attitude attributions, with the exception of
present-tense attributions, involve quantification into temporal argument positions
in the complement, and so are to be treated in accordance with the account of
quantification given here.
As Cappelen and Lepore have noted (1997), in reporting what people say, we can
put words they use in quotation marks in the complement of indirect discourse even
though they misuse or misspell them or just use expressions that are not words in
the language we are speaking at all, as in (15).
30
We give an account for the sententialist view of the referent of the complement. For the
propositionalist, it is just a matter of quantifying into a complex term that refers to a proposition,
and so relativizing the referent of the complement to the object assigned to the variable (relativized
to speaker and time).
(15) Nicola said that his father is a “philtosopher”31

This is a case in which a stricter standard for what sentence the subject uttered is
called for, for the intention is to indicate that Nicola used ‘philtosopher’, not just any
word that means the same as the speaker’s use of it, to express being a philosopher
(or whatever she expresses by it). For the strict reading, we need only to replace (iii)
in (MID) with (iii*):
(iii*) An occurrence of ” as used by ref (’, u, t) has the same grammatical role in
® as QUO(”) has in F.
The result of applying UNQ to the complement sentence in (18) is ‘his father is a
philtosopher’. That is not a sentence of English, but the account requires only that
it be interpreted relative to the use by the attributor in the utterance context and
the subject at the event time. (What if Nicola didn’t mean anything very definite
by ‘philtosopher’? See the discussion below of nonsense words in complements of
indirect discourse verbs.)
We also sometimes intend to convey that specific words were used rather than
synonyms or translations of them quite apart from cases in which we wish to convey
that someone used a misspelling of a word (or a neologism or a malapropism). For
example, consider this exchange:
A: Brian said that he was allergic to “gorse.”
B: I heard he said he was allergic to “furze.”
A: No, he definitely said that he was allergic to “gorse.”
This makes sense even if A and B both know that ‘gorse’ and ‘furze’ are
synonyms. In this case, we invoke the strict reading to capture the sense of the
exchange.
What about a case in which someone uses a nonsense term (a term that didn’t
mean anything in the mouth of the subject of the sentence), as in (16)?
(16) Nicola said that her father is a “jabberwocky.”

Here we intend the strict reading. These cases too can be accommodated as long as
we allow that ‘her father is a jabberwocky’ can stand in the samesaying relation to
some utterance of Nicola and allow translation to relate expressions that are not fully
meaningful. (iii*) requires that the particular nonsense word in the complement be
used rather than another, so no additional difficulties arise about getting the right
target. Nothing in the account requires sentences that appear in complements to be
31
Is this an instance of scare quotes? Not if the intention is to attribute the word to Nicola. We
can use scare quotes in indirect discourse without there being any suggestion that the subject used
them: John said that he was into “adult films”, though that is not what he called them. However,
there may be a kind of dual use of the quotation marks when one both wants to attribute words to
someone in indirect discourse and to distance oneself from them.
fully meaningful or to have propositional content that determines truth conditions

for them.
A question arises about how to treat iterated indirect discourse as in (5e).32
(5e) Professor Elugardo said that William James said that religious leaders are
“creatures of exalted emotional sensibility.”
The question is whether this should be treated as saying that Professor Elugardo said
something that is the same in content as (17) or (18).
(17) William James said that religious leaders are “creatures of exalted emotional
sensibility.”
(18) William James said that religious leaders are creatures of exalted emotional
sensibility.
On the first reading, we are saying (roughly) that
Professor Elugardo said that: William James said something whose content was that
religious leaders are creatures of exalted emotional sensibility and he [James] used
‘creatures of exalted emotional sensibility’ in the appropriate position.
On the second reading, we are saying (roughly) that Professor Elugardo said
something that is the same in content as (18) while he [Elugardo] used ‘creatures of
exalted emotional sensibility’ in an appropriate position.
Our account says that (5e) attributes to Elugardo an assertion the same in content
as (18) in which he used ‘creatures of exalted emotional sensibility’ (or a translation)
in the grammatical role in which it appears in the complement of (5e)—this is the
second interpretation. To get the first reading, we would use (19).
(19) Professor Elugardo said that William James said that religious leaders are
““creatures of exalted emotional sensibility””.
However, surely it is more natural to give (5e) the first interpretation on which the
words are attributed to James and not Elugardo. One reason it may seem more
natural is that we already know that William James used that phrase. Another is
that the focus is on what James said, so that when quotation marks appear in the
complement of the embedded attribution (in contrast to the NP of the embedded
attribution) we may be inclined to take the speaker to be intending to convey
something about the words with which James said it. But we can also get the other
(second) reading for iterated indirect mixed quotation. Consider (20).
(20) Though she said that he put it more politely, Sue told me that John
complained that he was being “jerked around by his asshole boss”.
In this case, we interpret the speaker of (20) as intending to attribute to Sue the use
of ‘jerked around by his asshole boss’ rather than to John. We react to (5e) and (20)
32
This example was raised by Ray Elugardo as an objection to our earlier account.
by trying to figure out what the speaker could be intending, and we use whatever
knowledge we have to help us arrive at a reasonable interpretation.
We think it is doubtful there is much of a practice attached to iterated mixed
quotation. We have only come across it in examples proposed by philosophers
intended to test our account. Most likely, we interpret these on the fly using whatever
information we can to divine the intentions of the speaker. If this is right, then
we are not faced so much with the question of what interpretation the rules of
the language give to (5e), but how to extend them to cover such cases. For this
purpose, we recommend extending our account since it provides a systematic way
of disambiguating the different readings speakers may intend. One could develop
an alternative which made (5e) literally express the first reading by conditionalizing
on whether F in (MID) was itself an instance of indirect discourse, and giving a
different treatment of embedded complements when it was, but this would make
certain things we might want to say using iterated mixed quotation inexpressible
by any literal interpretation. (Of course, the quotation marks in these cases can be
interpreted strictly or non-strictly.)
5.2 Mixed Attitude Attributions
The case of mixed attitude attributions as illustrated in (5f) is curious.
(5) (f) She thought that it was curious that that “thin, proud man” should have
had “that little, sharp woman for a mother!”
In the case of direct attributions of thought, we treated the quotation term in the
complement as giving the content of the thought interpreted relative to the context
of the thought (the thinker and time of the thought). In indirect attitude attributions
in non-mixed cases, it would seem that the difference is that the sentence in the
complement is interpreted relative to the speaker rather than the subject and time of
the thought. What purpose could interpolation of quotation terms in attributions of
thought on analogy with mixed indirect discourse serve?
It is clear that mixed indirect attitude attributions are intended to provide a more
fine-grained characterization of the thought that is being attributed. In the case
of direct attributions of thought, we suggested that we aim to use a sentence a
translation of which the subject would be disposed to use to express her thought.
With this in mind, we can make sense of the point of mixed indirect quotation,
namely, to indicate that the thinker would be disposed to use translations of the
material quoted in the complement in corresponding positions in announcing the
thought in a sentence. Using (MID) as a template, we can give the following
account in (MIA) (mixed indirect attitude) for attitude attributions, where we replace
the quantifier over events with a quantifier over states. (This is given for the
verb ‘thought’ but as above it provides a template that can be extended to other
propositional attitude verbs.)
(MIA) For all speakers u, times t, sentences F, names ’, [’ thought that F] is

true(u, t, English) iff there is a state s and a time t0 < t, such that
(a) Thinks(s, t0 , ref.(’, u, t), UNQ(F), u, t) and

(b) If UNQ(F) ¤ F, then
(i) Ref(’, u, t) is disposed to express s at t’ using a sentence ™ that relative
to ref(’, u, t) at t0 is an index-shifted translation of UNQ(F) understood
relative to u at t and
(ii) For all expressions ”, and formulas §, if F D the result of replacing
‘x’ in § with QUO(”), then
(iii) A translation of ” into the language of the sentence ™ that ref (’, u, t) is
disposed to use has the same grammatical role in ™ as QUO(”) has in F.
As above, there is a propositionalist version of clause (a) and we can make provision
for a more strict use of quotation marks as well so that what is required is an index-
shifted version of the sentence ™ itself.
6 Monstrous Operators
Kaplan (1989) defined a ‘monstrous operator’ as one that shifts the context of
evaluation away from the context of utterance. Cappelen and Lepore (1997, 2003)
have suggested that mixed quotation can involve this kind of shift. Here is an
example from their 1997 paper:
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”.
In the second sentence, the ‘us’ in quotation marks is clearly intended to be
interpreted relative to the context of Greenspan’s utterance rather than the context of
the report. The practice of replacing pronouns and indexicals in quoted material in
complements so that when interpreted relative to the speaker’s context the referent
comes out correct is illustrated in (5c). In light of this, one might dismiss examples
like these as careless and ungrammatical, an example of changing from indirect to
direct quotation in midsentence and failing to notice.
Without trying to settle this issue about usage, it is interesting to ask from the
theoretical standpoint how such a practice could be accommodated in a semantics
for a language. Take as a sample (21),
(21) She said that “of all of us” she was “the least happy camper”
where we assume that the speaker of (21) is not among those that the subject was
talking about. We want to have the speaker say something to the effect:
(22) She said that of all of them she was the least happy camper and
she used a sentence that translates ‘of all of us I am the least happy camper’
in saying that.
(MID) does not give us (22) because its clause (a) requires the speaker to say that
she said that of all us she was the least happy camper, and this requires the speaker
to include herself among those referred to, contrary to our assumption.
The first thing to do is to transform the complement sentence into a sentence
in which indexicals in quotation marks are replaced by bracketed indexicals that
corefer, taken relative to the speaker’s context, with those they replace taken relative
to the subject’s context. We use the following definition.
(Def) q0 relative to x at t is an index-shifted transform of q relative to y at t0 iff q0 is

a mere indexical variant of q such that
for all g such that QUO(g) appears in q, for every occurrence of any
indexical or demonstrative • in g (unless in authorial brackets), in q0 there is
in the same grammatical role as • in q an occurrence of an indexical (or
directly referring term) •0 in authorial brackets such that •0 understood
relative to x at t refers to the same thing as • understood relative to y at t0
Then we replace (a) in (MID) with (a*):
(a) saying(e, t0 , ref(’, u, t), UNQ(F), u, t)

(a*) there is an F* such that F* relative to u at t is an index-shifted transform of F
relative to ref(’, u, t) at t0 and saying(e, t0 , ref.(’, u, t), UNQ(F*), u, t)
For the rest, we replace ‘F’ with ‘F*’ in clause (b).33
7 Translation Practices
Before concluding, it is worth taking a glance at our practices in translating

quotation, which raise some questions about the account presented here. When
it is pure quotation that is in view, we translate the quotation marks but not the
expression enclosed within it. (This is support for the view that quotation marks
are not semantically inert but a distinct device in the language, and it is evidence
against the proper name theory for which, as remarked, what is interior to the
quotation marks is nothing more than spelling.) Strikingly, however, translations
for other forms of quotation, direct and mixed, often do not preserve the identity
of the material in quotation marks. When we translate dialogue in fiction or in
reporting, we typically translate the sentences that appear in quotation marks as well.
Similarly, when we translate mixed quotation, we typically translate the expressions
33
For arguments that quotation is not actually monstrous, see Jaszczolt and Huang (2017) and
Johnson (2017).
that appear in quotation marks as well. How is this compatible with the view that at
the core of all of these uses of quotation marks is pure quotation?
The answer is that ordinary translation does not aim to preserve faithfully every
aspect of meaning. Sometimes this is because there is no best fit between the
expressions in the target and home language. We can at best paraphrase the German
‘torschlusspanik’ into English as ‘the fear of diminishing opportunities as one ages’,
because we have no one-word equivalent. But in other cases the trouble is that in
preserving certain aspects of meaning we fail to convey something important about
the function of the word or sentence in the original. The most literal translation of
a poem may do a very bad job of conveying the literary qualities of the original.
In other cases, the problem has to do with a function internal to the conventional
rules for using the expressions. Tyler Burge (1978) pointed out that in cases of
self-reference like ‘This sentence is false’ we do not aim, when we translate it
into French or German, to preserve the referent of the subject term to the English
sentence. What we would lose in this case is the fact that the subject term is used
to refer to the sentence in which it appears. In a context in which we are interested
in the phenomenon of self-reference, it is more important to preserve that in the
translation than the reference to the English sentence. Burge noted that the same
thing applies to translation of dialogue in novels. A translation of War and Peace
into English in which all the Russian dialogue was left untranslated would not sell
very many copies. Dialogue is what we called quasi-use quotation. For it to function
as intended, the audience must understand the expressions contained in quotation
marks (so far as possible—dialogue can contain nonsense words too). Preserving
this function is more important in translation that preserving reference to the words
that appear in the quotation marks. Furthermore, given the account of (non-strict)
direct discourse we have given, the original sentence and its translation will still
share the same truth value. These remarks carry over straightforwardly to direct
attitude attributions and to mixed quotation. Thus, the fact that reference to words
in quotation marks, other than in pure quotation, is not preserved in translation is
not a counterexample to the analyses offered above.
8 Conclusion
In this paper we have argued for a unified account of quotation devices in natural
languages. We have not argued that there is one analysis of all uses of quotation
marks, but rather that though quotation is polysemous it is unified in that every form
of quotation (excepting those set aside in note 5) can be seen as making use of a
core function of quotation marks, namely, their use in pure quotation. The general
rule for pure quotation

is a remarkably simple and clear device that exploits the fact that expression types
have tokens by which we recognize them. Uses of quotation that extend beyond pure
quotation have two features in common.34 First, the material appearing in quotation
marks is intended to be understood and that it is understood is essential to the
function that such quotations play in communication (modulo a strict form of direct
and indirect discourse), even though this does not always involve the expressions
contributing their extensional properties to fixing truth conditions for the sentences
in which they appear. Second, they appeal to a relation born to the expression
appearing in quotation marks that plays a role in determining the truth conditions
of the sentences in which they appear. Thus, in all of these uses, a reference to
the expression in quotation marks is made in relation to which other things are
characterized, such as some utterance someone has made or a thought she has had.
References
Burge, T. (1978). Self-reference and translation. In Guenthner & Guenthner-Reutter (Eds.),

Meaning and translation (pp. 137–156). London: Duckworth.
Lepore, E., & Cappelen, H. (2003). Varieties of Quotation Revisited. Belgian Journal of Linguis-
tics, 17, 51–75.
Clark, H. H., & Gerrig, R. (1990). Quotations as demonstrations. Language, 66, 764–805.
Davidson, D. (1968). On saying that. Synthese, 19, 130–146.
Davidson, D. (1979). Quotation. Theory and Decision, 11, 27–40.
Davidson, D. (2001). Theories of meaning and learnable languages. In Inquiries into truth and
interpretation (pp. 3–15). New York: Clarendon Press.
De Brabanter, P. (2002). Making sense of mention, quotation, and autonymy. Dissertation,
University of Brussels. http:tel.archives-ouvertes.fr/docs/00/15/52/94/PDF/BIGdOC.pdf
Geach, P. (1957). Mental acts. London: Routledge and Kegan Paul.
Gomez-Torrente, M. (2001). Quotation revisited. Philosophical Studies, 102, 123–153.
Johnson (Eds.), The semantics and pragmatics of quotation (pp. 135–167). Dordrecht: Springer.
Harth, M. (2011). Quotation and pictoriality. In E. Brendel, J. Meibauer, & M. Steinbach (Eds.),
Understanding quotation (pp. 195–208). Berlin: De Gruyter.
Houston, K. (2013). Shady characters: The secret life of punctuation, symbols & other typograph-
ical marks. London: W. W. Norton.
34
These remarks can be generalized to devices that function like quotation marks in enclosing an
expression to form a term that refers to, or is about, or applies to instances of, a type of entity related
to the enclosed expression in some way. Thus, for example, putting a sentence in angle brackets
to form a name of a proposition [<¥>], or putting a word in all caps to refer to the concept it
expresses ‘RED’, or Wilfrid Sellars’s dot quotes, [•¥•], for types of linguistic entities individuated
by function, all depend on the device that is exhibited in pure quotation. Sellars writes: “I shall use
dot quotes to form the names of expressions : : : which is realized in English by the sign design
illustrated between them” (Sellars, Scharp, and Brandom 2007, pp. 163–4, n. 3). In all of these
cases, the entity referred to is located relative to the sentence or expression flanked by the special
delimiters, and the fact that we understand the contained terms is essential to their functioning in
the way they are intended to.
Jaszczolt, K. M., & Huang, M. (2017). Monsters and I, the case of mixed quotation. In P. Saka
& M. Johnson (Eds.), The semantics and pragmatics of quotation (pp. 357–382). Dordrecht:
Springer.
Johnson, M. (2011). The punctuation theory of quotation. In E. Brendel (Ed.), Understanding
quotation (pp. 209–230). Berlin: De Gruyter Mouton.
pragmatics of quotation (pp. 281–302). Dordrecht: Springer.
Kaplan, D. (1989). Demonstratives. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from
Kaplan (pp. 481–564). New York: Oxford UP.
Lepore, E. (1999). The scope and limits of quotation. In L. E. Hahn (Ed.), The philosophy of
Donald Davidson (pp. 691–714). Chicago: Open Court Publishers.
Lepore, E., & Loewer, B. (1989). You can say that again. Midwest Studies in Philosophy, 14,
338–356.
Ludwig, K. (2010). Adverbs of action and logical form. In T. O’Connor & C. Sandis (Eds.), A
companion to the philosophy of action. New York: Wiley-Blackwell.
Ludwig, K. (2014). Propositions and higher-order attitude attributions. Canadian Journal of
Ludwig, K., & Ray, G. (1998). Semantics for opaque contexts. Philosophical Perspectives, 12,
141–166.
Maier, E. (2014). Pure quotation. Philosophy Compass, 9, 615–630.
Maier, E. (2017). The pragmatics of attraction: Explaining unquotation in direct and free indirect
discourse. In P. Saka & M. Johnson (Eds.), The semantics and pragmatics of quotation
(pp. 259–280). Dordrecht: Springer.
Pagin, P., & Westerstahl, D. (2010). Pure quotation and general compositionality. Linguistics and
Quine, W. V. O. (1940). Mathematical logic. New York: W. W. Norton.
Quine, W. V. (1960). Word and object. Cambridge: MIT Press.
Recanati, F. (2000). Oratio obliqua, oratio recta. Cambridge: MIT Press.
Reimer, M. (1996). Quotation marks: Demonstratives or demonstrations? Analysis, 56, 131–142.
Saka, P. (2017). Blah, blah, blah: quasi-quotation and unquotation. In P. Saka & M. Johnson (Eds.),
The semantics and pragmatics of quotation (pp. 35–63). Dordrecht: Springer.
Sellars, W., Scharp, K., & Brandom, R. (2007). In the space of reasons: Selected essays of Wilfrid
Sellars. Cambridge, MA: Harvard UP.
Tarski, A. (1983). The concept of truth in formalized languages, Logic, Semantics, Metamathemat-
ics (pp. 152–127). Indianapolis: Hackett Publishing Company.
Wallace, J. (1970). On the frame of reference. Synthese, 22, 117–150.
Washington, C. (1992). The identity theory of quotation. The Journal of Philosophy, 89, 582–605.
Semantics vs. Pragmatics in Impure Quotation
Mario Gómez-Torrente
Abstract I defend a semantic theory of quotation marks, according to which these

are ambiguous, as they have several different acceptations involving corresponding
different conventional indications. In particular, in allusion (“mixed”) uses, the
corresponding conventional indication is one with an adverbial or prepositional
content, roughly equivalent to “using the quoted expression or an appropriate
version of it”. And in “scare” uses, the corresponding conventional indication
is that the enclosed expression should be used not plainly but in some broadly
speaking distanced way, or that it is being so used by the utterer. I also defend
this view against some alternative views on which allusion and distance indications
are to be seen as pragmatically conveyed. In particular, I consider several views
that attempt to explain especially allusion and distance indications as pragmatic
suggestions generated from a meager conventional basis, and I argue that they
cannot accommodate a number of linguistic phenomena and reflectively supported
theses about the use of quotation marks. I lay special emphasis on the fact that the
main pragmatic theories fail to pass an extremely plausible test for challenges to
polysemic accounts of an expression.
M. Gómez-Torrente ()
Instituto de Investigaciones Filosoficas, Universidad Nacional Autonoma de Mexico (UNAM),
Cd. México 04510, Mexico
e-mail: mariogt@unam.mx

https://doi.org/10.1007/978-3-319-68747-6_6
136 M. Gómez-Torrente
1 Introduction
Cases of impure quotation are those where the distinctive purpose of using quotation
marks is not the purely referential purpose of referring to the expression enclosed
within the marks.1 Cases of pure or purely referential quotation include the utterance
of the quotation2 in normal utterances of (1):
(1) “Socrates” has eight letters.3

Impure quotation (or at least standard impure quotation; see the discussion of
“emphatic” and other possible non-standard uses of quotation in Sect. 2 below)
includes three kinds of uses of the quotation marks. Paradigmatic cases of the first
kind include the utterances of the quotations in normal uses of (2) and (3); think
of (3) as written in a biography of Henry Ford, in a passage where the author is
explaining Ford’s opinions:
(2) Ford said that thinking is “the hardest work” there is.
(3) Thinking is “the hardest work” there is.
In these uses, the utterer’s distinctive purpose in using the quotation marks is
not (or in any case not merely) to indicate that the quoted expression is being
referred to, but to indicate that this expression was uttered by Ford. We may call
uses of this kind “allusion” uses. Paradigmatic cases of the second kind of impure
quotations include the utterances of the quotations in normal uses of the following
sentences:
(4) The 5000 “smackers” are hidden in the house by the river.
(5) Smith’s “music” records are on the shelf.
In these cases, the utterers’ distinctive purpose in using the quotation marks is to
indicate that “smackers” and “music” are somehow not entirely appropriate in the
relevant contexts. We may call uses of this kind “distance” uses. Finally, the third
1
“Impure” in my usage is thus a purely negative concept, largely free from theoretical presupposi-
tions. Other terms used in the literature to cover the same range of phenomena seem to me to be
inappropriately theory-laden. Thus, for example, “mixed” and “hybrid” as applied to a quotation
mean that the quoted expression is being both “used and mentioned”, but this seems to me already
to load the dice in favor of particular accounts of the relevant uses of quotations.
2
In the technical usage predominant in this paper, “quotation” is a term that applies to the result
of enclosing an expression within quotation marks. In the few cases where “quotation” means
something else, such as the act of forming a quotation, or the act or result of citing someone, I trust
context will make this clear.
3
They also include the utterance of the quotation in cases of so-called “direct quotation”, as
in a normal utterance of Ford said: “thinking is the hardest work there is”. On referential
uses, including a number of pragmatic aspects, see Gomez-Torrente (2001, 2011, 2013, and the
bibliographical references therein).
Semantics vs. Pragmatics in Impure Quotation 137
and least interesting kind of cases of impure quotation include cases such as the
utterances of the quotations in normal utterances of the following:
(6) I find Quine’s “Two Dogmas of Empiricism” difficult.
(7) 1517 lives were lost despite the efforts of the “Titanic” crew.
Here the purpose is to indicate that “Two Dogmas of Empiricism” and “Titanic” as
they are being used name things of certain special kinds, in particular a paper and a
ship, respectively. We may call uses of this kind “special name” uses.
In earlier work (Gomez-Torrente 2005), I defended the view that cases like
normal uses of (2) and (3) exemplify an acceptation of the quotation marks on which
these conventionally indicate that the enclosed expression (or at least a certain part
of it4 ) is a contextually appropriate version of expressions uttered by some agent
or agents who are contextually relevant, while cases like normal uses of (4) and
(5) exemplify an acceptation of the quotation marks on which these conventionally
indicate that the enclosed expression should be used not plainly but in some broadly
speaking distanced way, or that it is being so used by the utterer; cases like normal
uses of (6) and (7) exemplify an acceptation of the quotation marks on which they
indicate that the enclosed expression, as well as the quotation itself, works as a name
of one of certain special sorts—an article or ship or poem or movie or plane, etc.
In this chapter I will again defend a very similar view, modifying only slightly the
postulated content of the conventional indication of the quotation marks in uses like
those of (2) and (3); as explained in Sect. 3 below, I no longer view this indication
as a full propositional indication, and postulate instead that it is an indication with
an adverbial or prepositional content, roughly equivalent in the case of (2) and (3)
to “using ‘the hardest work’ or an appropriate version of it” or to “with a use of
‘the hardest work’ or an appropriate version of it”. But my view was and is that,
being conventional, all three kinds of impure indications—the (impure) “allusion”,
“distance” and “special name” indications, as I will call them—are parts of the
semantics of the quotation marks in these uses. In this chapter I will seek to defend
this view against some alternative views of what is going on in normal uses of (2)–
(5) and related cases of the same kinds, views on which impure allusion and distance
indications are to be seen as pragmatically conveyed. (Cases such as (6) and (7) are
hardly ever discussed (Saka 2005 being an exception), and to my knowledge never
in great depth.)
In particular, I will consider several views that attempt to explain especially
allusion and distance indications as pragmatic suggestions generated from a meager
conventional basis. In Sect. 2, after motivating and introducing these views, I will
seek to argue that they cannot accommodate a number of linguistic phenomena and
reflectively supported theses about the use of quotations; I will lay special emphasis
4
This qualification (already noted in Gomez-Torrente 2005) was meant to accommodate the fact
that ellipsis points (and other devices, like clarifications in brackets), sometimes occur inside
quotation marks but are not implied to have been uttered by the agent that is relevant in the
context. These have been called “unquotation” devices in the recent literature. On this topic see
Shan (2010), Maier (2014), and Saka (2017).
on the fact that the main pragmatic theories fail to pass an extremely plausible test
for challenges to polysemic accounts of an expression. In Sect. 3 I will describe
in more detail my semantic view, introducing the mentioned slight revision of the
content of the conventional indication of allusion uses, and I will explain how the
view accommodates the phenomena and theses mentioned in Sect. 2 as problematic
for pragmatic views.
2 Pragmatic Theories and Their Problems
Pragmatic theories of impure quotation have a number of motivations. The theories

of this kind that I’m aware of (and perhaps all existing theories of this kind)
have been thought to receive very strong support from considerations of theoretical
economy, as indicated e.g. by pragmatic theorists such as Recanati (2010: 300f)
and Saka (2005: 203).5 On the view of mine stated in the Introduction, quotation
marks have, besides their purely referential acceptation, three further conventional
acceptations manifested in allusion, distance and special name uses. My view thus
postulates that there are at least four conventional meanings of the quotation marks.
On the other hand, pragmatic theories seek to explain the variety of distinctive
purposes of uses of the quotation marks without postulating a corresponding variety
of conventional meanings for them. They seek to explain the fact that a variety of
kinds of indications can be conveyed with uses of quotation marks by assuming
just one meaning for them and deriving the possibility of conveying the mentioned
indications via pragmatic mechanisms.
It is indeed an uncontroversial fact that, provided there is such a thing as a literal
meaning m of an expression e, the utterer of an utterance u of e often intends
to convey—and often succeeds in conveying—more than that literal meaning m
with her utterance of u; when successfully conveyed, such non-literal meanings
are said to be pragmatically conveyed. A number of general mechanisms for prag-
matically conveying content have been reasonably well described in the literature
on pragmatics, and we may assume that these mechanisms are there, so to speak,
independently of whatever views we may have about the conventional meaning
or meanings of particular expressions. So if it could be successfully argued that,
5
As an anonymous referee has pointed out to me, this need not mean that the main motivation for a
pragmatic theory is always one of economy. In Recanati’s case, for example, the main motivation
for his theory is arguably what he takes to be the pictographic nature of quotations, which he
believes to virtually necessitate a pragmatic view of most of what is normally conveyed by means
of them. I myself, on the other hand, think that “distance” and “special name” uses of quotations,
at the very least, cannot be said to be pictographic in any natural sense (see Gomez-Torrente 2005,
148, n. 6). In any case, the considerations in this paper will seek to discredit pragmatic views, and
to defend a semantic view, independently of considerations concerning the alleged pictographic
nature of all quotations, and exclusively by focusing on how the different theories handle the
semantics/pragmatics interface in the case of the quotation marks.
by assuming just one conventional meaning for the quotation marks, the other
indications typically communicated by means of quotations could be effectively
conveyed via pragmatic mechanisms, then one would have a strong methodological
consideration of economy in favor of pragmatic theories as against semantic theories
which, like mine, postulate several conventional meanings for the quotation marks:
ceteris paribus, a more economical pragmatic theory should have an advantage over
a less economical semantic theory. It must be emphasized, however, that this kind
of consideration of economy will be decisive only if the ceteris paribus clause holds
good: if some semantic theory can account for phenomena involving quotation that
no pragmatic theory can account for, the consideration of economy will be defeated.
Aside from this general methodological consideration of economy, other more
special facts have been mentioned in defenses of the superiority of a pragmatic
account. Philippe De Brabanter, a champion of such accounts, helpfully collects
some of these in his survey De Brabanter (2010). The first is the possibility
(originally pointed out by Washington 1992) of “quoteless quotation”, i.e. reference
or allusion to expressions without the use of quotation marks, as in
(8) Socrates has eight letters.
(9) Winston Churchill actually read “Mein Kampf”, understood what the plan
was about, wanted to confront it early on, and it was his vision that won the
day and ultimately defeated Hitler, with the help of America and a lot of
blood, toil, sweat and tears. (www.unitedfamilies.org/hannity_article.asp;
cited by De Brabanter 2010: 115)
(De Brabanter notes: “In his maiden speech to the Commons on 13 May 1940,
Churchill declared ‘I have nothing to offer but blood, toil, tears and sweat’. The
context suggests that the writer is mentioning the (slightly modified) phrase as well
as using it” [115].) Typically, (8) would be used with the purpose of referring to
“Socrates” and predicating of it that it has eight letters, and (9) would be used
partly with the purpose of recalling (somewhat inexactly, as it turns out) that “blood,
toil, sweat and tears” was uttered by Churchill. These cases are supposed to be
problematic for semantic accounts because it is sometimes taken to be inherent
to a semantic account to postulate that whenever there is reference or allusion to
expressions in an utterance u, some element corresponding to the expression that is
cited or referred to must be present in the truth conditions or the logical form of u
(cf. De Brabanter 2010: 114). However, such postulation seems ad hoc, as there are
no independent reasons to postulate the existence of such elements in the case of
normal utterances of (8) or (9). By contrast, the possibility of offering a pragmatic
explanation of what is going on in cases such as (8) and (9) suggests itself naturally,
even at first sight. (See below, Sect. 3, precisely for a pragmatic proposal made from
the point of view of the theory defended in the present paper.)
A second fact sometimes mentioned in critiques of semantic theories of impure
quotation is that, in cases where the utterer’s distinctive purpose in using the
quotation marks is an allusion purpose, the quotation need not be under the scope
of a reporting verb. Thus, although in (2) the quotation is under the scope of such a
verb, in (3) it is not, despite the fact that in context it will be clear that the speaker
seeks to indicate that this expression was uttered by Ford. De Brabanter sees this as
a sign that in cases such as (2) and (3) the speech attribution results from a pragmatic
process (De Brabanter 2010: 115–7).
Finally, a third special fact adduced in favor of pragmatic accounts is that in some
cases of allusion uses of the quotation marks, the quoted expression is not a syntactic
constituent:
(10) David said that he had donated “largish sums, to several benign institutions”.
(Abbott 2005: 20; cited by De Brabanter 2010: 117–8)
Some semantic theories of allusion uses of quotation marks explicitly postulate
that a quotation ’ in such uses has a certain compositional meaning, “something
like ‘what the echoed speaker calls ’’” (De Brabanter 2010: 117); De Brabanter
is alluding to Benbaji (2005) and Geurts and Maier (2005); however, expressions
that are not self-standing syntactic constituents presumably have no independent
corresponding meaning or truth-conditional element contributing compositionally
to the meaning or truth conditions of sentences in which they appear. De Brabanter
again sees this as a sign that in cases such as (10) the speech attribution indication
is pragmatically generated (De Brabanter 2010: 117).
As noted in the Introduction, the pragmatic views I will consider here propose to
explain relevant instances of communication with typical impure uses of quotations
as involving pragmatic suggestions generated from a meager conventional basis.
The bases are different in the different theories, and the pragmatic mechanisms
that generate the pragmatically conveyed content from the conventional bases
are different as well. But it seems clear that the crucial aspect of the seman-
tics/pragmatics interface for the quotation marks postulated by the different theories
lies in the nature of the conventional basic indication they postulate, rather than on
the generating mechanism: any particular view about the content of the conventional
indication of the quotation marks could in principle be combined with any particular
view about the specific pragmatic mechanism or mechanisms that generate the
pragmatically conveyed content from that conventional indication. Accordingly, my
main criticisms arise specifically from consideration of the nature of the contents of
the meager conventional indications postulated by pragmatic theories. I will list the
theories roughly in ascending order of strength of the postulated basic conventional
indication.
On the first kind of pragmatic view to be considered, held by Paul Saka, quotation
marks quite generally (not just in impure uses) “signal mentioning, that there is
reference to something other than what is customary” (Saka 2005: 208), i.e. they
carry only a minimal conventional indication that the user is referring to something
other than the conventional reference of the quoted expression (and possibly to this
conventional reference as well). When speaking of the “reference” of the quoted
expression, quite clearly Saka means something looser, such as the meaning of the
quoted expression, as he for example often speaks of propositional indications as
“references”; and in fact he means something looser than what one might read in
the above statement at first sight, for Saka does not mean to presuppose that there
is always a “customary” reference or even a customary meaning for the expression
that is being quoted (this expression might be nonsensical, for example). Perhaps
a more accurate statement of what Saka means would be: quotation marks carry
a minimal conventional indication that the user, in using the quoted expression,
means something different from or additional to what the enclosed expression
conventionally means, if it conventionally means anything at all. (That Saka does
mean “different from or additional to” is evident from many of his examples, as
well as from his formulation (2005: 187) according to which the relevant indication
is that the speaker intends “to pick out something other than the customary referent
(either instead of it or in addition to it).”)6 As for the kind of pragmatic mechanism
via which non-conventional contents are conveyed, Saka seems reluctant to identify
it with any mechanism described elsewhere, but he suggests some points of
similarity with Grice’s “exploitative” conversational implicatures, in particular by
emphasizing that the pragmatically implicated content is in principle and typically
in practice retrievable as a reasonable hypothesis about the speaker’s meaning, when
some conflict with conversational expectations triggers a “higher-level reasoning”
process:
when utterances impinge on audiences, audiences automatically attempt to execute the
conventionally associated lexico-syntactic conceptual structure; quoted matter is defeasibly
used as well as mentioned. The use interpretation is defeated, however, when it is
ungrammatical, as in the case of most citations and titles; and when it is pragmatically
contradicted by the mention interpretation, as in the case of loose-speech scare quotes.
In short, standard quotation marks always direct the audience to the same panoply of
material: to the concepts that are automatically associated with the quoted matter and to
“something else”. It is up to the interpreter to assemble a propositional model using higher-
level reasoning not specifically provided by the particular words at hand. (2005: 208)
On the second pragmatic view we will examine, held by François Recanati,

quotation marks, at least in allusion and distance uses, carry only a minimal
conventional indication that the user is calling attention to the quoted expression,
or “demonstrating” it, in Recanati’s terminology (Recanati 2001, 2010); the view is
considered sympathetically by Abbott (2005) and Reimer (2005). The pragmatic
mechanism by which other indications are generated from this minimal basis,
according to Recanati, is the mechanism of “free enrichment”. A free enrichment
6
Saka (2011: section 5), modifies his 2005 theory somewhat, reducing a bit the field of things
that according to the theory could be pragmatically indicated by quotation marks. On Saka’s most
complete attempt to state the new theory, quotation marks carry a minimal conventional indication
that the user, in using the quoted expression, means something different from or additional to
what the enclosed expression conventionally means and different from or additional to what is
or can be “cognitively generated therefrom” (311, n. 5). (What can be “cognitively generated”
from the conventional meaning of the quoted expression includes, paradigmatically, metaphorical
extensions of the conventional meaning.) I will present my criticisms of Saka’s ideas below
by reference to the old, simpler theory, as they will apply indifferently to it and to the more
difficult to state (and in fact not fully stated by Saka) revised theory. The abstract reason why
the criticisms apply to both theories is that the indications involved in the criticisms will at no
point include indications “cognitively generated” specifically from the particular meanings of the
quoted expressions. See notes 10 and 17 below for exemplifications of this point.
is a process by which the speaker adds a certain non-conventional indication i to

the conventional content of an utterance u in such a way that i “does not remain
external to the intuitive truth-conditions” of u (2001: 672), despite the fact that i can
be canceled, because
it takes some reflection to realize that that fact was not explicitly articulated in the sentence.
From a psychological point of view, the pragmatic suggestion is incorporated into what is
asserted: A single mental representation is constructed using both linguistic and contextual
clues [ : : : ]. In such cases I say that the truth-conditional content of the utterance is
pragmatically enriched. ([footnote 22:] [ : : : ] The extra element contextually provided
[ : : : ] does not correspond to anything in the sentence itself, hence it does not constitute
a component of the compositionally articulated content of the utterance, yet it is part of its
intuitive truth-conditional content.) (2001: 672)
“Free enrichment” is different from Grice’s conversational implicatures in that

conversationally implicated content is supposed to remain “external to the intuitive
truth-conditions of the utterance” and is in some psychological sense more clearly
perceivable as “external” than “enriched content”. (De Brabanter 2013a: 120–1),
on the other hand, in his strongly sympathetic exposition of Recanati, expresses
a preference for a mechanism similar to, but not identical with, that based on the
generation of conversational implicatures.)
A third kind of pragmatic view proposes that quotation marks, at least in
allusion and distance uses, carry a minimal conventional indication that the utterer
is “echoing” someone else’s use of the quoted expression, or at least a previous
use by the utterer (Recanati 2010). Recanati views this as an alternative to the
preferred view explained in the preceding paragraph, an alternative that he is not
able to exclude on present evidence. (But his expositor De Brabanter forcefully
defends Recanati’s original (2001) view that the echoic indication is not universal
and hence not conventional, arguing that the echoic indication in some quotations
is pragmatically conveyed; see De Brabanter (2013b).) Recanati would again
supplement the echoic view (were this to turn out to be preferable) with his theory
of “free enrichment” as providing the mechanism for generating pragmatically
conveyed content.
There are three kinds of criticisms of pragmatic views that I wish to put forward.
The first is that the conventional indications some of them postulate fail to be
intuitive indications in many uses or potential uses of impure quotation; in this
sense, those alleged conventional indications seem simply not to exist as general
conventional indications. A second criticism is based on the observation that, if
the alleged conventional indications postulated by some pragmatic theories really
existed, it should be easy to construct cases in which they are conveyed, so to
speak, “free” from any of their alleged pragmatic add-ons; but this appears to be
difficult or impossible. The third and final kind of criticism, which I take to be
the most significant, is that the assumption that those conventional indications do
exist and are in fact as meager as postulated by some theories implies that many
possible pragmatic indications could be naturally derived from them that are in fact
not actually derivable in a natural way.
I will begin by presenting the first kind of criticism, which affects echoic views.
The view that quotation marks, whether in impure or pure uses, carry always a
conventional indication that the user is “echoing” some previous use of the quoted
expression seems open to clear counterexamples in the case of pure quotation. If I
sincerely utter
(11) Nobody has ever uttered “bigritwesertkil”,
I am certainly not implying (nor will anyone take me to imply) that someone else
or I myself have used “bigritwesertkil” before. Of course, the fact that the echoic
view has clear counterexamples in the case of pure quotation makes it lose much
of any attraction it may have had. For if there is at least one referential non-
echoic acceptation of the quotation marks—as the direct counterexample based on
(11) strongly suggests—then no unified pragmatic explanation of all uses of the
marks can be given, while this would certainly be one important motivation for
sympathizers of the echoic theory.
The weaker view that the echoic indication is a conventional indication of (only)
impure uses (the view not excluded by Recanati 2010) is somewhat trickier, but
seems pretty clearly false as well. In a typical utterance of (4) or (5), there need not
be any intention on the part of a fully competent, careful speaker to communicate
that the quoted expression has been used before. In fact, if an utterer of (5) “cancels”
any possible suggestion that he is implying that the expression “music” has been
used before by uttering I don’t mean to suggest that the expression “music” has
been used before, he will not seem to me to be contradicting himself. But, since the
expression in question has been used before, there will be an air of oddness in such
a “cancelation”, so it is perhaps open to the echoic theorist to claim that some use of
the word is being echoed, perhaps some use by a conceivable or imaginary utterer
or group, such as “the linguistic community”. But there are other counterexamples.
One kind of counterexamples involves impure allusion uses appearing in non-
declarative environments, and where the quoted material is clearly being used for
the first time ever. If I suddenly come up with the worry in (12),
(12) I wonder if anyone ever said that thoughtful armadillos “beautifully
relinquish their numbers” in awe of the icosahedron,
there is no intuitive conventional indication that some utterance of “beautifully
relinquish their numbers” by someone in particular (real or imaginary) is being
echoed; on the contrary, I am precisely asking myself (in part) if there is someone
(of any kind) who has used the expression before.7
A second kind of impure counterexample to the echoic view involves distance
uses. My father enjoyed coming up with new words he had never heard before,
7
This would be a problem also for my earlier (2005) view of the conventional indication of
allusion uses as a straightforward indication that the enclosed expression that the quoted expression
is a contextually appropriate version of expressions uttered by some agent or agents who are
contextually relevant (and for many other views; see note 20 below for an example). See Sect.
3 for the announced modification of my view, which takes care of these counterexamples.
usually by spontaneous combinations of metatheses and epentheses of existing

words, using them playfully when talking to other people. He did not, as far as
I know, use these in written language, but he clearly could have, and he could
perfectly well have put quotation marks around the ensuing cacographies to indicate
he was being playful, as for example in
(13) That kid was so “brutiful” and “aborradle”.

Here we would have the purpose of indicating that the quoted expressions are
somehow inappropriate, but certainly not that of indicating that they have been used
before, let alone that someone else has used them before—in fact, I can perfectly
well imagine my father taking pride in the originality of his inventions. He could
have “canceled” any echoic suggestion of (13) without oddness, by uttering I don’t
mean to suggest that any of the expressions “brutiful” and “aborradle” has been
used before; in fact this would have added to the fun.8
The view that quotation marks carry a conventional indication that the user is
calling attention to the quoted expression and the view that they carry the indication
that the user means something besides what the enclosed expression means (if it
means anything)—the “attention” and “difference” views, as we may call them—
postulate conventional indications which are perhaps too weak for the views to be
susceptible of receiving direct counterexamples. Perhaps in some sufficiently vague
sense all uses of quotation marks are meant to call attention to the quoted expression
and/or to signal that something besides its usual meaning is meant. However, if this
is so, it is natural to think that it should be easy to construct cases in which these
conventional indications are conveyed without the addition of any of the standard
allegedly pragmatic indications; but this appears to be difficult or impossible. This
is my second kind of criticism.
When one uses quotation marks, some clues from the context (broadly under-
stood so as to include information of all kinds provided by the utterance and its
surrounding circumstances) generally make it clear which one of the four standard
indications (the purely referential, allusion, distance and special name indications) is
meant (or if some other non-standard indication is meant; see below the discussion
of the “emphatic” interpretation of quotation marks, the frequent but still non-
standard interpretation of the quotation marks as indicators of emphasis). Reading
(5), the reader is likely to recall that people differ in their musical tastes to the
point that many would refuse to call “music” certain compositions they consider
not sufficiently meritorious; the reader will consequently typically interpret an
utterance of (5) as indicating that the utterer thinks that the compositions in Smith’s
records are not music strictly speaking. Reading (3) in a biography of Henry Ford
where Ford’s opinions are being explained and discussed, the reader is likely not to
think that the author is distancing himself from the quoted expression, but simply
8
Note that the echoic view is also open to counterexamples arising from the “special name” use of
quotation marks, for surely there is no “echoing” when someone simply introduces a title, or the
name of a ship, etc. And for even more against echoic views see Saka (2005, section 3.3).
attributing it to Ford. Reading (1), the reader is likely to take implicitly the quotation
as a name of “Socrates”, for the predicate “has eight letters” makes a lot of sense as
predicated of that expression. If we make an effort to construct examples in which
all contextual clues of this kind have been removed, a pragmatic view presumably
must predict that the standard indications that the view takes as pragmatically
implicated (when they are) are not communicated (which seems correct) and that
the conventionally indicated content is communicated without any expectation on
the part of the audience that one of the standard indications is nevertheless in the
mind of the utterer. But this last prediction seems incorrect. Suppose someone utters,
out of the blue,
(14) “2” C 2 D “4”, “and” “there” are no “polyhedra” “of” seven faces; it’s easy
to go on calling attention to expressions in this way,
or
(15) “2” C 2 D “4”, “and” “there” are no “polyhedra” “of” seven faces; it’s easy
to go on signaling unusual meanings of expressions in this way.
A typical reader, when presented with (14), will not just think The writer is calling
attention to “2”, “4”, “and”, “there”, “polyhedra” and “of” all right, and says
that it’s easy to go on doing this with other words (and perhaps is alluding to
someone’s expressions or expressing “distance” toward them or emphasizing them,
perhaps not). And a typical reader, when presented with (15), will not just think The
writer is definitely signaling that he is using “2”, “4”, “and”, “there”, “polyhedra”
and “of” to mean something besides what these words mean, and says that it’s
easy to go on doing this with other words (and perhaps alluding to someone’s
expressions or expressing “distance” toward them or emphasizing them, perhaps
not). The typical reader, when presented with (14) or (15), will be genuinely puzzled
by these utterances. She will not consider the possibility of accepting them at face
value, but will wonder which one of the standard indications is somehow meant.
Or, perhaps more clearly and significantly, suppose that (14) or (15) are uttered
not out of the blue, but accompanied by suitable cancelations of any allusion
and distance indications (and, if necessary, also of any non-standard “emphasis”
indication)—a special name indication and a purely referential interpretation being
quite obviously excluded.9 A pragmatic view presumably must predict that the
standard indications that the view takes as pragmatically implicated (when they are)
are not communicated (which seems correct) and that the conventionally indicated
content is communicated without any uneasiness on the part of the audience
9
On Recanati’s “attention” theory, an indication that the quotation is being used to refer to the
quoted expression is semantic whenever the quotation is pure, so it cannot be canceled when it’s
the intended indication. But on Saka’s theory, that indication is always pragmatic when it exists.
caused by the fact that all the relevant standard (and if necessary the non-standard)
indications have been canceled or preempted in some way. But this seems clearly
wrong.
(140 ) “2” C 2 D “4”, “and” “there” are no “polyhedra” “of” seven faces; watch
out, I’m not alluding to anyone’s expressions nor using the quotation marks
to distance myself from the quoted expressions; I’m just calling attention to
the quoted expressions.
(150 ) “2” C 2 D “4”, “and” “there” are no “polyhedra” “of” seven faces; watch
out, I’m not alluding to anyone’s expressions nor using the quotation marks
to distance myself from the quoted expressions; I’m just indicating that I’m
using the words with a different meaning.
A reader of (140 ) or (150 ) will then be at least as puzzled as in the other case, and
probably even more, for now the possibility that she might be missing one of the
standard indications will have been explicitly excluded by the utterer, who will
appear to be doing something pretty mad, if not downright contradictory.10 I take all
this to be a sign that the “attention” and “difference” views postulate conventional
indications that are just too weak to recover the actual meaning or meanings of the
quotation marks.11
It might be suggested that this kind of criticism overlooks relevant differences
between pragmatic mechanisms for conveying pragmatic suggestions from a meager
semantic basis. The idea would be that while some pragmatic mechanisms make
10
Note that the same could be said of a more complicated version of (150 ) incorporating
the conventional indication of the quotation marks postulated in Saka’s (2011) revised theory,
mentioned above in note 6. A reader of the following sentence will be at least as puzzled, and
probably more, than a reader of (150 ): “2” C 2 D “4”, “and” “there” are no “polyhedra”
“of” seven faces; watch out, I’m not alluding to anyone’s expressions nor using the quotation
marks to distance myself from the quoted expressions; I’m just indicating that I’m using the words
with unusual meanings not “cognitively generated” from the conventional meanings of the quoted
expressions in this way.
11
A referee suggests that the fact that (140 ) and (150 ) will appear puzzling is predicted at least
by Recanati’s theory, since it’s part of the theory that utterances of quotations (or at least normal
utterances, I suppose) always have a “quotational point”, while (140 ) and (150 ) themselves deny
that they have such a point. To this I would reply, first, that (140 ) and (150 ) do not deny that they
have a “quotational point”; on the contrary, they state that their point is to communicate precisely
the indication postulated as conventional by Recanati’s theory and Saka’s theory, respectively.
And second, that if Recanati’s theory were to include the claim that the “quotational point” of
utterances of impure quotations must always go beyond the indication that attention is being called
to the quoted expression, and thus presumably that it must include either an allusion or a distance
or a special name indication, then the theory would not merely postulate a meager conventional
indication from which other non-conventional indications can be derived pragmatically, but rather a
rich conventional association with the quotation marks, from which specific quotational indications
might be derived in context; it would in fact be a “contextualist” version of a semantic theory of
the quotation marks.
the pragmatically communicated content quite distinct psychologically from the

semantic content of the utterance, others don’t. Thus, while a theory that appeals to
a mechanism similar to that of “exploitative” conversational implicatures (arguably
favored by Saka) may be open to the criticism, a theory based on free enrichment
(such as Recanati’s) may not: in free enrichment, a non-conventional indication
“does not remain external to the intuitive truth conditions” of the utterance, so it
might be argued that in a use of (14), say,12 a pragmatic allusion indication or a
pragmatic distance indication is just bound to be a part of the relevant intuitive
truth conditions. But I should then insist that part of the point is that the theories
we are criticizing presumably predict that in some cases, such as the imagined
utterance of (14), no such indication ought to be communicated, whether as a part
of their intuitive truth conditions or as a part of any other aspect of their conveyed
content. Note that sentences that are sometimes or even frequently used with free
enrichments attached can be used “non-enriched”: to use a modified version of
Recanati’s paradigm case, one can make an utterance of
(16) He took out the keys and opened the door
preceded by a discourse in which it has been emphasized that the male in question
under no circumstance opens the door in question in any other way than by a kick
of his left foot, after taking the keys in question out of the door lock where his poor
wife put them; in such a case, the frequent enrichment that the door was opened
with the taken out keys will be absent. Why should the quotation marks of (14) and
(15) be difficult or impossible to use in environments where the alleged allusion and
distance enrichments are absent (e.g. because they have been preemptively canceled,
as in utterances of [140 ] and [150 ])?
As I said, I have a third and final kind of criticism, which I take to be the most
significant, and which applies again to the “attention” and “difference” theories.
Suppose again, for a reductio, that some one of the conventional indications
postulated by those theories does exist as a conventional indication. As noted
above, these indications are notoriously weak, and for this reason it may well have
to be granted that they are generally meant to be communicated by utterers of
quotation marks, even as a matter of convention. But if so, many stronger non-
standard indications should be conveyable from those weak conventional meanings
via typical pragmatic mechanisms, without producing any sense in the audience that
the utterer has made a mistake. However, such non-standard indications are only
conveyed by producing a sense that something has gone wrong.
A way of illustrating this phenomenon that I find particularly striking involves the
quite frequent but still notoriously non-standard use of “emphatic” quotes—called
12
The same could be said of (15). As noted in the text above, different basic indications can
presumably be combined with different pragmatic mechanisms, so Saka’s basic indication could
be combined with a mechanism of free enrichment; the point to be made will equally hold.
“mystery” quotes by Saka (2013) and “noncitational” quotes by Abbott (2005). Here
are a couple of examples from the amusing Keeley (2010):
(17) “PARENTS”, You are Responsible For your “Childrens” (SAFETY) Do not
let them Run “Free” or “Leave” them “Unattended” “In Shopping Carts”
While Shopping. Management Thank You, (Keeley 2010: 8; seen at a
supermarket)
(18) “NO” Refunds Or Exchanges All Sales “Final” (Keeley 2010: 15; seen at a
store).
The utterers of these examples appear to mean to emphasize the concepts expressed
by the quoted expressions,13 and knowledgeable (or presumably knowledgeable)
speakers can easily figure this out, of course. But knowledgeable speakers find
these uses incorrect, and in fact a source of somewhat perverse fun. Those utterers
are manifestly doing something funny. Why? The natural answer is: because they
are not competent users of the quotation marks, they are not using them as they
should be used, i.e. in order either to name, cite, or express distance toward the
quoted expression or to indicate a special name usage. However, there should be
nothing odd about these utterances if the conventional meaning of the quotation
marks was merely that postulated by the “attention” or “difference” theories. It is
clear, for example, that if the quotation marks had just the “attention” meaning,
then the “emphasis” indication could be communicated straightforwardly via some
pragmatic mechanism, without generating an impression of error. First, the utterers
of (17) and (18) in the examples are intentionally calling attention to the quoted
expressions in their (successful) attempt to emphasize the corresponding concepts,
so they are abiding by the meaning that the marks have on this hypothesis.14 But
second, the “emphasis” indications could then be communicated without oddness
either as “free enrichments” or as “standard” conversational implicatures. Let me
explain this a bit.
“Free enrichments”, recall, are indications so well integrated into the intuitive
truth conditions of the enriched utterances that they remain psychologically hidden;
it takes some reflection to see that they are not semantically expressed. “Standard”
conversational implicatures15 are implicatures communicated without any apparent
violation of the maxims of conversation; they are therefore psychologically hidden
as well, as the audience will not typically actually have to reason their way out
of an apparent violation of the maxims—but according to Recanati they, like all
13
Saka (2013) questions the idea of this sort of quotation as emphatic on the grounds that “reading
aloud examples of mystery quotation does not normally yield the stressed syllables that emphasis
would demand” (938). But I myself, on the contrary, if forced to read aloud (17) and (18), would
think that I would represent best the original utterers’ intentions if I stressed the quoted items.
14
The criticism doesn’t apply to the echoic theory because the echoic theorist could claim that the
problem is that in (17) and (18) there is evidently no echo.
15
The concept is Grice’s (1975: 32), but the terminology “standard” is from Levinson (1983: 104).
implicatures, will remain “external” to the intuitive truth conditions. Grice’s main
example, recall, is one in which a stranded driver tells a passer-by that he is out of
gas, and the passer-by replies with an utterance of
(19) There is a garage round the corner,

the implicated indication being that the garage is open or believed to be open. Now,
the distinction between “free enrichments” and standard conversational implicatures
may not be completely clear-cut. It is somewhat unclear to me, for example, whether
the indication in a typical utterance of (16) that the taken out keys were used to
open the door is “internal” to the intuitive truth conditions of the utterance in a
way in which the indication that the garage is open is “external” to the intuitive
truth conditions of the imagined utterance of (19). But regardless of whether the
distinction is a precise but subtle one or just one of psychological degree, it seems
clear that an audience for (17) and (18) that implicitly assumed an “attention”
meaning for the quotation marks would implicitly grasp that the reason why the
utterers are calling attention to the quoted expressions was in order to emphasize
the corresponding concepts, without this indication being psychologically well-
separated from the alleged basic conventional “attention” indication, without it
being perceived as dissociated from this part of the alleged conventional meaning
(or even the truth conditions) of the utterance.16
A way of seeing that the “emphasis” indications would be pragmatically
communicated without oddness if the “attention” theory was right is to imagine
that we introduce by stipulation a new kind of “attention” quotes. Let’s introduce
by stipulation the attention marks “I ” and “J ”, with the same grammar as the
normal quotes, and with this explicit meaning: when an expression is enclosed
within attention marks, the utterer is conventionally calling attention to the enclosed
16
In any case, the emphasis indication can clearly be derived as a standard implicature, in the
Gricean fashion, from a hypothetical “attention” or “difference” meaning. In the case of (18) and
the “attention” hypothesis:
We presume the writer to be following the plausible maxims regulatory of conversation
identified by Grice when she literally expresses the content that there will be no refunds or
exchanges and all sales will be final, and attention is called to “no” and “final”.
But we must suppose that she is aware of or thinks that the concepts expressed by “no”
and “final” are being emphasized if we are to hold on to this presumption. (For she must
be observing the maxim of Relation directing her to say only things relevant in the context,
and the hypothesis that the concepts expressed by “no” and “final” are being emphasized
is necessary if she is to be supposed to be observing the maxim (while to suppose that she
merely means to call attention to the expressions “no” and “final” amounts to attributing to
her a failure to observe the maxim).)
And surely she thinks that we can see this, so she thinks that we can see that she thinks
that the concepts expressed by “no” and “final” are being emphasized.
So she has conversationally implicated that the concepts expressed by “no” and “final”
are being emphasized.
expression. Now imagine that you can assume normal speakers know this stipula-
tion, and imagine that you encounter utterances of the following:
(170 ) I
PARENTSJ , You are Responsible For your I ChildrensJ (SAFETY) Do
not let them Run I FreeJ or I LeaveJ them I UnattendedJ I In Shopping
CartsJ While Shopping. Management Thank You.
(180 ) I
NOJ Refunds Or Exchanges All Sales I FinalJ .
I think it’s quite clear that you will understand the utterers as not only calling
attention to the “attention-quoted” expressions, but as emphasizing the concepts
corresponding to those expressions. But furthermore, I also think it’s clear that you
will grasp this indication without sensing any oddness in the utterances of (170 ) and
(180 ). Again, there may be alternative explanations of the mechanism via which this
indication would be communicated. Explanations in terms both of “free enrichment”
and standard conversational implicatures seem feasible. As in paradigmatic cases
of “free enrichment”, the “emphasis” indication would not be neatly separated
psychologically from the intuitive truth conditions of the imagined utterances of
(170 ) and (180 ); it would certainly take some reflection even for people to whom
the stipulated meaning of the attention marks was taught to see that the “emphasis”
indication was not semantically expressed. And, as in standard implicatures, the
“emphasis” indication would be communicated without any apparent violation of
the maxims of conversation and without need of any reasoning excluding a real
violation of the maxims. In any case, that the “emphasis” indication would be
communicated without any sense that a mistake was being made is evident.
Mutatis mutandis for the “difference” theory. This says that the marks mean that
the utterer is signaling that something beyond or different from the meaning of
the quoted expression is being meant (and that further indications of the marks are
pragmatic add-ons). And the utterers of (17) and (18) in the examples are signaling
that something beyond or different from the meaning of the quoted expression is
being meant—they are signaling that they want to emphasize the corresponding
concepts. They are again abiding by this alleged meaning of the marks. The further
indication that the utterer is emphasizing the concept expressed by the quoted
expression ought then to be painlessly conveyable as a “free enrichment” or as a
standard conversational implicature of the conventional “difference” indication, and
no oddness or infelicity should be sensed by competent users of quotation marks.
Further, if we repeat the thought experiment of introducing by stipulation a pair
of “difference” marks, we will get results analogous to the ones we got for the
“attention” marks.17
17
And mutatis mutandis for Saka’s (2011) revised theory, mentioned above in note 6. The theory
says that the marks mean that the utterer is signaling that something beyond or different from the
meaning of the quoted expression, though not something “cognitively generated” from it, is being
meant. And the utterers of (17) and (18) in the examples are signaling that something beyond or
different from the meaning of the quoted expression, though not “cognitively generated” from it,
is being meant—they are signaling that they want to emphasize the corresponding concepts. The
The phenomenon we are describing can also be illustrated by thought-

experimenting a bit with the actual quotation marks, rather than thought-
experimenting with marks introduced by an imaginary stipulation. Suppose we
encounter the following piece of written discourse:
(20) There are so many beautiful things in the world : : : There are so many ugly
things in the world : : : Flowers are so beautiful : : : “Slush” is so ugly : : :
Babies are so pretty : : : “Rust” is so hideous : : : I find sunsets, fawns and
snow-capped Himalayan mountains so aesthetically pleasing, and “traffic
jams”, “spiders” and “tin roofs” so unpleasant : : : Recently I spent a week in
gorgeous Vancouver Island, then I went back to “Mexico City”, where I
enjoyed the many enchanting streets of Coyoacán but I suffered the
“circumvallation roads”.
I trust the reader will have conjectured that the utterer means to indicate, whenever
he speaks of something he thinks is ugly, that it is ugly, by putting quotation
marks around the expression that names the thing. But the piece of discourse is
manifestly odd; the utterer is doing something funny. Why? Again the natural
answer is that he is not using the quotation marks as they should be used, because
he is neither intending to name, nor to cite, nor to express distance toward the
use of the quoted expression, nor of course to indicate special name usage. But
there should be nothing odd about this discourse if either the “attention” or the
“difference” theory was right. The utterer of (20) is calling attention to the quoted
expressions, and the further indication that he thinks of the things named by the
quoted expressions as ugly ought then to be straightforwardly conveyable from the
conventional “attention” indication. No oddness or infelicity should appear. Mutatis
mutandis for the “difference” theory: the utterer of (20) is signaling that something
beyond or different from the meaning of the quoted expression is being meant—
he is signaling that he means that the thing named by a quoted expression is ugly.
This further indication ought to be conveyable without difficulty from the alleged
conventional “difference” indication, and the utterance of (20) should not appear
peculiar in any noteworthy way. But it does, and I take this phenomenon to provide
strong evidence that these pragmatic theories are false.
The phenomenon we have just been examining in our third criticism suggests
a more general point. Inspired by his own reactions to challenges to the univocity
of the definite article stemming from Donnellan’s seminal paper on the referential
use of definite descriptions, Kripke (1977) devised a celebrated general test for
challenges to the monosemy of an expression, that went roughly as follows. Suppose
i is an indication of an utterance u of an expression e. In order to determine whether
i is a manifestation of a conventional acceptation of e different from its alleged
unique meaning m, imagine a circumstance in which e has m as its only meaning
further indication that the utterer is emphasizing the concept expressed by the quoted expression
ought then to be painlessly conveyable as a “free enrichment” or as a standard conversational
implicature of that conventional indication. A thought experiment of introducing by stipulation a
pair of marks with the mentioned meaning again reinforces the point.
(perhaps because it has been so stipulated). If i would have been generated by u

even in such a circumstance, then the hypothesis that i exemplifies a conventional
acceptation of e different from m in the present circumstance is not warranted. Now,
our third criticism of pragmatic theories suggests a certain kind of “reverse” general
test for challenges to the ambiguity or polysemy of an expression. Suppose again
that i is an indication generated by an utterance u of an expression e. In order to
determine whether i is a manifestation of a conventional acceptation of e, and is not
a pragmatic indication via u of an alleged more meager meaning m for e, imagine a
circumstance in which e has m as its only meaning (perhaps because it has been so
stipulated). If a certain different indication i’ would have been naturally generated
by another utterance u’ of e in such a circumstance but it would not be generated
by u’ in the present circumstance, then the hypothesis that m is the only meaning of
e in the present circumstance and i is a pragmatic indication of utterance u of e is
not warranted (and the hypothesis that i exemplifies a conventional acceptation of
e in the present circumstance is supported).
Our third criticism can now be seen as involving applications of this general
principle. Suppose i is the indication that “music” in a certain utterance u of the
quotation marks, say one of (5), is being used or should be used in a distanced way.
i is evidently generated by normal utterances of (5). Now imagine a circumstance
in which the quotation marks have as their only meaning the “attention” indication,
say. Then we also ascertain that another utterance of the marks, say one of (18),
would naturally generate in such a circumstance the indication that the concepts
corresponding to the quoted expressions are being stressed. But, at the same time,
we realize that such an indication is not naturally generated by an utterance of (18)
in the present circumstance. We therefore conclude that the hypothesis that the
“attention” indication gives the only conventional acceptation of the marks is not
warranted, and the hypothesis that the distance indication exemplifies a conventional
acceptation of the marks is supported.
Existing tests for ambiguity or polysemy are notoriously problematic, and I don’t
want to claim that the test for a related purpose that I am proposing is necessarily
water-tight. But I think that it is worth considering, and that it gives intuitively
correct results in many cases. Take a traditionally uncontroversial example of
polysemy, “chair”. Suppose i is an indication generated by a certain utterance u
of a sentence containing “chair”, say the indication generated by an utterance of
(21) The chair will give us some money,
that the person referred to by the grammatical subject is the person in charge of some
institution, the kind of person in charge that can give institutional money away. Now
imagine a circumstance in which “chair” has as its only meaning some really meager
indication, such as “thing closely related to an artifact designed for sitting”. Then
we also ascertain that another utterance of “chair”, say the second and third ones in
an appropriate utterance of
(22) I sat on the chair, resting my arms on its two comfortable adjustable chairs,
one on each side; but I realized that the left chair had a nail sticking out from
it, so I ended up resting my left arm on my lap,
would naturally generate in such a circumstance the indication that we are talking
about armrests—after all, an armrest is related to an artifact designed for sitting
in a very close way. But, at the same time, we realize that such an indication is
not naturally generated by an utterance of (22) in the present situation. On the
contrary, an utterance of (22) in the present circumstance would be puzzling. We
therefore conclude that the hypothesis that the indication “thing closely related to
an artifact designed for sitting” gives the (only) conventional acceptation of “chair”
is not warranted, and the hypothesis that the “person in charge” indication naturally
suggested by an utterance of (21) exemplifies a conventional acceptation of “chair”
is supported. I invite the reader to apply the test to other examples.
When, 30-odd years ago, Grice and Kripke among others criticized some
philosophers’ frequent appeals to ambiguity, it was indeed in fashion to postulate
semantic ambiguities or polysemies whenever a newly discovered linguistic intu-
ition involving some uses of an expression seemed not to be covered by a traditional
unitary analysis of its meaning. Grice in fact devised his theory of implicatures to
a large extent in order to confront ambiguity theories of the logical particles and
of the verb “to believe”; Kripke mentions proposals that the propositional “know”
and the identity sign are ambiguous. To this fashion, Grice and Kripke opposed
a healthy cautious attitude: first one should see if a pragmatic explanation of the
newly discovered indication was available without postulating more conventional
meanings than in the traditional analysis. Only after exhausting the pragmatic route
should one postulate ambiguities or polysemies; among other considerations that
favor this way of proceeding, considerations of economy like the ones reviewed at
the beginning of this section played a prominent role.
Some four decades later, the fashion seems to be just the opposite of what it
used to be. Now what is in vogue is to try to explain as many as possible of the
indications that can be conveyed by the use of an expression as pragmatic add-ons
to a meager meaning for it. If Cruse (2000: 197–8) is right, the currently prevalent
view in lexical semantics is monosemy, the view that generally a word has a single
highly abstract meaning rather than the traditional variety of dictionary meanings
for it, an abstract meaning from which the other “meanings” can be derived in
context as pragmatic add-ons (see especially Ruhl 1989). And in philosophical
semantics, some of the significantly represented parties in fashionable disputes
concerning “unarticulated constituents” postulate “minimal”, “underspecific”, and
other kinds of meager meanings for sentences, assigning formidable roles to
pragmatics in the determination or specification of “what is said” with utterances of
those sentences, either via implicatures, or via “enrichment” processes, or whatever.
Grice’s “Modified Ockham’s Razor” is sharper than ever! I myself think that Grice’s
and Kripke’s “program”, as well as some of the post-Gricean appeals to pragmatic
supplementations in lexical semantics and philosophical semantics, have yielded
a large number of correct insights. But if the fashion of 40 years ago went awry
in bloating one’s semantic ontology, the new fashion definitely runs the risk of
seeing fewer things in semantic reality than there actually are. I’m convinced that
the semantics of quotation marks is a case in point, and I hope that the preceding
considerations, including the proposal of the mentioned test for challenges to

polysemic theories of an expression, may contribute a bit to showing this.
Even if pragmatic theories are not correct, it doesn’t mean semantic theories
must be. After all, pragmatic theories had apparently strong motivations behind
them. Perhaps some kind of hybrid theory incorporating insights from both semantic
and pragmatic theories is needed. Or is there a semantic theory that can rise to the
challenges of pragmatic theorists? It will be the task of the next and final section to
argue that at least one semantic theory can.
3 A Semantic Theory and Its Virtues
The theory I wish to propose and defend here, already put forward in a slightly
different form in Gomez-Torrente (2005), can be stated as follows. The quotation
marks have at least (probably exactly) four established acceptations, corresponding
to four different conventions of use. One of these is the purely referential accep-
tation; when used under this acceptation, the quotation marks help create a new
singular term that as a matter of convention semantically refers to the enclosed
expression. In a normal utterance of (1) or (11), the quotation marks will be used
under this acceptation. On a second acceptation, manifested in typical utterances of
(6) and (7) and special name uses in general, the quotation marks again help create
new singular terms that as a matter of convention semantically refer to things such
as articles, poems, movies, ships, planes, etc., that are also named by the enclosed
expressions.
On a third acceptation, manifested in typical utterances of (2) and (3) and
allusion uses in general, the quotation marks conventionally add an adverbial or
prepositional indication, typically to the verb “to say” or a related attitudinal verb
(as in [2]). This will be an indication roughly (but probably not exactly) equivalent
to an adverbial clause of the form of “using ‘—’ or an appropriate version of it”
or to a prepositional phrase of the form of “with a use of ‘—’ or an appropriate
version of it” (where the expression alluded to goes in each case in place of the
dash,18 inside purely referential quotation marks). Such a conventional adverbial
18
The expression alluded to is understood here not to include parts which may be enclosed within
the quotation marks but not for allusive purposes, such as ellipsis points, clarifications in brackets,
etc. (See note 4 above.) The indications of these parts, when they appear, are probably to be
understood, at least typically, as additional content inserted by the utterer, which doesn’t contribute
to truth conditions, which “projects” beyond scopal devices, etc. Thus, for example, if I say Shaw
said that there is no “sincerer [truer] love” than the love of food, what I say is true just in case
Shaw said that there is no sincerer love than the love of food, using the words “sincerer love”; the
clarification that “sincerer” means “truer” is not part of the truth conditions of my utterance. And
if I say Maybe Shaw said that there is no “sincerer [truer] love” than the love of food, what is
said might be the case is that Shaw said that there is no sincerer love than the love of food, using
the words “sincerer love”; the clarification that “sincerer” means “truer” is not part of what is said
might be the case.
or prepositional indication thus contributes compositionally to the truth conditions

of an utterance containing the allusion marks and containing a verb susceptible
of being modified by the corresponding adverbial or prepositional clauses. When
“to say” or a related verb do not appear in the sentence uttered, as in (3), the
adverbial or prepositional indication carried by the allusion marks must modify
some other verb if the result is to be grammatically proper; but the indication
that a relevant proposition is in the scope of “to say” or a related verb will often
be pragmatically conveyed even in the absence of a verb susceptible of being
modified by the adverbial or prepositional clause, and whenever this happens,
the quotation marks will be implicitly understood as adding their conventional
adverbial or prepositional indication to the implicitly understood verb. In such
cases, the conventional adverbial or prepositional indication cannot contribute
compositionally to the truth conditions of the utterance containing the allusion
marks.
On a fourth acceptation, manifested in typical utterances of (4) and (5) and in
distance uses in general, the quotation marks conventionally add to the sentence
without quotes an indication that the enclosed expression should be used not plainly
but in some distanced way, or that it is being so used by the utterer. In these cases the
marks work roughly as a parenthetical main clause paratactically added to the full
sentence without quotes, and expressing the mentioned suggestion of the utterer’s.
Such a conventional indication does not contribute compositionally to the truth
conditions of the utterance containing the distance marks.
Under any of these acceptations, the quotation marks can be put to the use of
communicating non-literal contents, conveyed via any of the well-known pragmatic
mechanisms invoked in pragmatic theories, or via other pragmatic mechanisms. But
according to the theory, as confirmed by the considerations of Sect. 2, there is no
single, meager meaning of the quotation marks from which speakers and readers
generate the indications that the theory postulates as literal and conventional.19
As recalled in Sect. 1, in Gomez-Torrente (2005) I proposed that the quotation
marks in their allusion acceptation had a straightforward full propositional content
as their conventional indication—namely the indication that the enclosed expression
(or a certain part of it) is a contextually appropriate version of expressions uttered
by some agent or agents who are contextually relevant. The present modification
proposes that the marks in allusion uses don’t add a full propositional content, but an
adverbial or prepositional content designed in the most frequent case to modify the
verb “to say” or a related attitudinal verb, a content which thus definitely contributes
19
In Gomez-Torrente (2005) I spoke of the existence of the various acceptations of the quotation
marks as showing that the marks are ambiguous. Some readers have suggested that I should have
said they are polysemous instead. This may be acceptable, provided we become convinced that the
several acceptations of the marks are in fact closely related in some specific sense, as they probably
are (see my own thoughts on this matter below). But my use of “ambiguous” was and is meant to
be for a weak concept that subsumes both polysemy and homonymy, and I saw and see no reason
not to frame the claims of my view in terms of ambiguity, as this highlights the fact that the view
is not essentially committed to a stronger claim of either polysemy or homonymy.
compositionally to the truth-conditional content of the utterance containing the

allusion marks (and containing a verb susceptible of being modified by the
corresponding clauses). This is meant to take care of the fact that impure allusion
quotation marks can appear embedded in contexts where their conventional content
cannot be thought of as being a full propositional indication. In an utterance of
(23) I wonder if Ford said that thinking is “the hardest work” there is,
there will typically be no conveyed indication that Ford did utter “the hardest work”,
while there is surely an indication that the speaker wonders if Ford said that thinking
is the hardest work there is, using the particular expression “the hardest work” (or
a contextually appropriate version of it).20 On the present theory, that utterance is
roughly equivalent to an utterance of
(24) I wonder if Ford said that thinking is the hardest work there is, using “the
hardest work” or an appropriate version of it.
Note that, both intuitively and under the interpretation provided by (24), part of
what I wonder is whether Ford used (an appropriate version of) certain words.
Such a content thus does not intuitively “project” beyond the scope of “to wonder”.
Similarly for an utterance of
(25) Did Ford say that thinking is “the hardest work” there is?
which on the present theory is roughly equivalent to an utterance of
(26) Did Ford say that thinking is the hardest work there is, using “the hardest
work” or an appropriate version of it?
(note that part of what is asked is whether Ford used [an appropriate version of]
certain words) and for an utterance of
(27) Maybe Ford said that thinking is “the hardest work” there is,
which on the present theory is roughly equivalent to an utterance of
(28) Maybe Ford said that thinking is the hardest work there is, using “the hardest
work” or an appropriate version of it.
(Note that part of what is said might be the case is that Ford used [an appropriate
version of] certain words.) In an utterance of
(29) No one can have ever said that thinking is “the hardest work” there is,
20
This fact presents problems for many theories of allusive (or “mixed”, or “hybrid”) quotation.
On Maier’s (2014) important theory, for example, the use of an allusive quotation carries a
presupposition that the quoted phrase was used in an earlier utterance, and I wonder if Ford said
that thinking is “the hardest work” gets analyzed roughly as I wonder if Ford said that thinking is
whatever some salient speaker meant when they uttered “the hardest work”. But this doesn’t have
the intuitively right truth conditions in cases where no one has uttered “the hardest work” before,
and the utterer is just wondering out of the blue if Ford said that thinking is the hardest work by
uttering “the hardest work”.
there is intuitively a conventional indication that every speaker has failed to say that
thinking is the hardest work there is, at least by uttering (an appropriate version of
the expression) “the hardest work” when saying it. This is captured by the present
theory with its postulation that the mentioned utterance is roughly equivalent with
an utterance of
(30) No one can have ever said that thinking is the hardest work there is, using
“the hardest work” or an appropriate version of it.
The examples could be multiplied.
Note that the present theory implies that, when used with an allusion acceptation
in sentences like (2), the quotation marks carry a conventional indication that
turns out to imply a full propositional indication that the enclosed expression is
a contextually appropriate version of expressions uttered by some agent or agents
who are contextually relevant. Thus, (2)’s rough analysis,
(31) Ford said that thinking is the hardest work there is, using “the hardest work”
or an appropriate version of it,
implies
(32) Ford said that thinking is the hardest work there is, and “the hardest work” is
an appropriate version of expressions in Ford’s utterance.
Thus the present theory subsumes the predictions of my earlier (2005) theory for
cases relevantly similar to (2), which are presumably the most frequent cases of use
of the allusion acceptation of the quotation marks.
According to the theory, the successful usability of utterances of sentences like
(3) in order to communicate allusive contents is a sort of by-product of the intended
use of the allusion marks with the verb “to say” or related verbs. Considered out
of the blue, these utterances sound odd, and this oddness may even be traced to
ungrammaticality in some cases if not in all. But when the context makes it clear that
the sayings of some particular agent or agents are at stake, a pragmatic mechanism
will typically convey that the relevant proposition can be understood as in the scope
of “to say” or a related verb, which the marks can be pragmatically understood
as modifying as well. Thus (3) does not in any case literally say the content of
(31), but utterances of it can manage to convey that content. If this theory is right,
one could expect that embeddings of sentences like (3) would be infrequent, for it
would be more difficult to convey and process the multiplicity of (either explicit or
pragmatically conveyed) scopal devices. Thus, an utterance of
(33) I wonder if thinking is “the hardest work” there is
will not typically be used to convey an indication that the speaker wonders if Ford
said that thinking is the hardest work there is, using the particular expression “the
hardest work” (or an appropriate version of it). To the extent that it could be so used,
the present theory would suggest that this is because the utterance in question can
be used to convey pragmatically a certain proposition expressible using “to say” or

a related verb, such as a proposition expressed by an utterance of (24). Similarly for
an utterance of
(34) Is thinking “the hardest work” there is?
which on the present theory could perhaps be used in order to pragmatically convey
the same proposition as an utterance of (26), and for an utterance of
(35) Maybe thinking is “the hardest work” there is,
which on the present theory could perhaps be used in order to pragmatically convey
the same proposition as an utterance of (28). On the present theory also, an utterance
of
(36) Thinking is not “the hardest work” there is,

could perhaps be used to pragmatically convey the same proposition as an utterance
of
(37) Ford did not say that thinking is the hardest work there is, using “the hardest
work” or an appropriate version of it.
The theory’s differential account of (2) and (3) provides an explanation of a
phenomenon I noted in Gomez-Torrente (2005: 139ff.), but left as a puzzling case
there. Compare the following two dialogues (thinking of the second as an e-mail
exchange between Ford biographers, for example):
(2D ) –Ford said that thinking is “the hardest work” there is.
–That’s false, because he didn’t utter the word “hardest”.
(3D ) –Thinking is “the hardest work” there is.
–That’s false, because Ford didn’t utter the word “hardest”.
(2D ) (and relevantly similar exchanges) sounds like an appropriate exchange, while
(3D ) (and relevantly similar exchanges) sounds odd. This led me in Gomez-Torrente
(2005) to express doubts that the conventional indication of allusion marks is part
of the truth-conditional content of the utterances where the marks appear (even
though I inclined toward the view that it is). If the present theory is right, a natural
explanation of the felt difference here would be that while Ford said that thinking
is “the hardest work” there is (D[2]) has the content of (31), and thus its truth
conditions which involve directly in the main clause what Ford said or didn’t say,
Thinking is “the hardest work” there is (D[3]), if it has truth conditions of its own,
has truth conditions involving the question of whether thinking is hard or not (and
not involving the question of whether Ford made a use of “the hardest work” or not),
even if the content of (31) can be pragmatically conveyed via an utterance of (3).
A fact worth noting about the impure allusion acceptation of quotation marks
concerns an ambiguity of certain reports where the marks appear embedded under
the scopes of several occurrences of “to say” or some related attitudinal verb. If
Jones utters (2), I can report Jones’ utterance, apparently accurately and completely,
by means of
(38) Jones said that Ford said that thinking is “the hardest work” there is.
In this case, the content of my report is that of
(38A) Jones said that Ford said, using “the hardest work” or an appropriate
version of it, that thinking is the hardest work there is.
But it is easy to think of cases where I could use (38) to report on a different
utterance of Jones’. Suppose Jones uttered not (2) but
(39) Ford said that thinking is the hardest work there is.
Then I could report Jones’ utterance by means of (38), but clearly the allusion
quotation marks in my utterance would not correspond to any element of the
truth-conditional content of Jones’ utterance of (39)—they would clearly be my
own contribution, so to speak. This would be explained on the present theory by
postulating that in this case (the adverbial or prepositional indications carried by) the
quotation marks modify the first occurrence of “said” in (38), while in the previous
case they modified its second occurrence. The content of my utterance of (38) when
reporting on (39) would thus be the content of
(38B) Jones said, using “the hardest work” or an appropriate version of it, that
Ford said that thinking is the hardest work there is.
By contrast with the quotation marks in their third, or allusion acceptation,
the quotation marks in the fourth, distance acceptation do seem to provide a full
propositional “distance” conventional indication that by default “projects” beyond
scopal elements. Hence the present theory’s postulate that in distance uses the
marks work as a parenthetical added to the full sentence without quotes, and not
as an element susceptible of composing and contributing to truth conditions when
embedded in more complex constructions. Thus, in an utterance of
(40) Jones wonders if the 5000 “smackers” are hidden in the house by the river,
it will typically be difficult if possible at all to convey an indication that Jones won-
ders if the expression “smackers” is somehow inappropriate; the natural indication
(assuming it is contextually clear that the quotation marks are being used to express
distance and not allusion) will be that the utterer thinks that it is inappropriate.
Similarly, and perhaps even more clearly, for utterances of
(41) Are the 5000 “smackers” hidden in the house by the river?
and of
(42) Maybe the 5000 “smackers” are hidden in the house by the river.
In an utterance of
(43) The 5000 “smackers” are not hidden in the house by the river,
it will again be difficult if possible at all to communicate the proposition that the
use of “smackers” is not inappropriate, and the default indication will surely be
the indication proposed as conventional by the present theory, that “smackers” is
thought to be inappropriate by the utterer. The examples could be multiplied.
This is not to say that there cannot be cases where the distance conventional
indication is to be attributed to someone different from the utterer. Consider the
following example, where we are to suppose that the utterer is Smith, the owner of
the records despised by Jones, the utterer of (5):
(44) Jones doesn’t make a secret of his disdain for my artistic tastes : : : Smith’s
fondness of Agatha Christie’s novels is deplorable; these are so naïve
compared to the great masterworks of the detective genre : : : Smith’s
“music” records are a torture; he listens to the Spice Girls! : : : I don’t think I
can get along with Jones.
Here the use of the quotation marks surely conveys an indication that Jones, not
the utterer, thinks that the use of the word “music” as applied to Smith’s records
is somehow inappropriate. But in a case such as this it is pragmatically obvious
that Smith is mimicking an utterance of Jones’. Since an audience will grasp this
fact thanks to the text preceding the quotational sentence, it will consequently be
pragmatically obvious to them that the distance indication is naturally attributable
to the mimicked utterer (Jones) and not to the utterer (Smith). But cases such as
this actually confirm rather than refute the idea that the default attribution of a
quotational distance indication is to the utterer; for under this supposition it is easy
to explain cases like the one we are considering as pragmatic by-products of the
convention of default attribution to the utterer.21
This is one of many examples illustrating the fact that an indication conveyed by
the use of an expression may be conventionally attached to it without necessarily
contributing to truth conditions. In my earlier work, I left unanswered the question
whether the impure allusion and distance indications of the quotation marks
contribute to the truth conditions of the utterances of the quotational sentences
that conventionally express them. In Gomez-Torrente (2005) I gave a detailed
discussion of tests for whether such indications contribute to the truth conditions of
the relevant quotational utterances (including a test based on exchanges relevantly
21
Cases of essentially this kind are presented by Amaral et al. (2007) as counterexamples to the
unqualified claim in Potts (2005) and in many other places that appositives and expressives are
always “speaker-oriented”. In Harris and Potts (2009), Potts accepts these counterexamples to his
earlier general claim, but proposes a pragmatic explanation of the cases.
similar to [2D ] and [3D ] above), and the upshot was that the tests did not yield
a conclusive answer to this question. I also noted, however, that applications of the
tests tended to suggest that the allusion indications do contribute to truth-conditional
content, while distance indications do not. In my present opinion, the projectability
and unprojectability phenomena discussed in the preceding paragraphs make it
overwhelmingly likely that the conventional impure allusion indications do, while
the conventional distance indications do not, contribute to the truth conditions of
the relevant utterances. For it is clear that conventional impure allusion indications
regularly fall by default under the reach of scopal elements applied to sentences
containing the relevant quotations, while conventional distance indications do not.
This is perhaps not yet fully conclusive evidence that allusion indications do
contribute to truth-conditional content and distance indications do not, but it seems
to me that it may well be evidence as conclusive as we can get. The present theory
therefore postulates that impure allusion indications generated by quotation marks
do contribute to truth conditions (at the very least in cases where the quotation marks
are explicitly under the scope of “to say” or a related verb) while distance indications
do not.
Perhaps I ought to stress that, in saying that the distance indications of the
quotation marks do not contribute to the truth conditions of the utterances that
express them, I am of course not denying that those indications are themselves
truth-conditional. Such indications do of course have truth values (at least when
they are not affected by problems that might deprive them of such values, such
as problems having to do with presupposition failures, vagueness and the like).
My only claim is that the truth of those indications appears to be irrelevant to the
truth of the quotational utterances that serve to make them. Thus, for example, the
truth of the indication that an utterer of (4) thinks that “smackers” is inappropriate
appears to be irrelevant to the truth of her utterance of (4), which hangs exclusively
on whether the 5000 dollars in question are hidden in the house by the river. In
my view, distance indications generated by the quotation marks closely resemble
the indications characteristically generated by expressions like “confidentially”, “in
other words” and “to get back to the point”, which, as Bach (1999) has forcefully
argued, can be presumed not to contribute to the truth conditions of the (typical)
utterances of the sentences that contain them, but qualify (some aspect of) the speech
acts performed in uttering them.22
One possible underlying motivation for pragmatic theories not mentioned in
Sect. 2 might be a vague intuition that the different uses of quotation marks would
seem to have something in common; wouldn’t it be natural then to expect that
22
Bach has also argued that “but”, “still”, “even” and other traditional alleged generators of
conventional implicatures do contribute at least typically to the truth conditions of utterances of
sentences that contain them; but I am hesitant to follow him here. In any case, I reject a view like
Predelli’s (2003), which, despite being semantic like the present view, largely assimilates impure
allusion and distance quotation marks to “but”, “still”, etc. and sees no substantive difference
between allusive indications of the quotation marks and distance indications, assigning to both
of them a truth-conditional role (in the case of at least many quotational utterances).
this thing in common was an aspect of the meaning of the marks? And isn’t a
semantic theory that postulates that they are ambiguous ill-equipped to account
for this intuition? The answers are: it is natural to expect that the meanings of the
marks will have something in common, but a semantic theory such as the one just
stated accounts perfectly well for this expectation. As stressed in Gomez-Torrente
(2005) for the essentially similar theory there, all the conventional indications
postulated above for the different acceptations of the quotation marks share an
obvious general feature: they all contain reference to the quoted expression. This
is due to the fact that the general conventions that govern the different uses all
mention the quoted expression: the general convention governing pure quotation
says that “by enclosing any expression within quotation marks one gets a quotation
that stands for the enclosed expression”; the general convention governing the
special name acceptation of quotation marks says that “by enclosing any expression
already naming an article, poem, movie, ship, plane, etc. within quotation marks
one gets a quotation that stands for the thing named by the enclosed expression”;
the general convention governing the allusion acceptation of quotation marks says
that “by enclosing any expression within quotation marks one adds an adverbial or
prepositional indication, roughly equivalent to the meanings of phrases of the form
of ‘using ‘—’ or an appropriate version of it’ or ‘with a use of ‘—’ or an appropriate
version of it’ (where the expression alluded to goes in each case in place of the
dash)”; and finally, the distance acceptation is governed by a general convention
such as “by enclosing any expression within quotation marks one indicates that
the enclosed expression should be used not plainly but in some distanced way,
or that it is being so used by the utterer”. On the present theory, therefore, the
intuition that all uses of the quotation marks have something in common receives a
natural explanation, and in fact one that appeals to a common semantic feature,
a feature shared by all the meanings of the marks. But the explanation in no
way appeals to the idea that quotation marks in all their uses must share a
single meaning.23
In what remains, I will first explain why the present theory is not affected by the
criticisms that I posed to pragmatic theories in Sect. 2. After that, I will argue that
the theory is also not affected by the objections, again from Sect. 2, that pragmatic
theorists have posed to semantic theories.
First, let’s briefly review why the present theory is not vulnerable to the
objections we pressed in Sect. 2 against pragmatic theories, and in fact how it
explains some of the audiences’ reactions in the examples exploited in the criticisms.
A decisive criticism of the strong echoic view was that if I sincerely utter (11) I am
not implying that someone else used “bigritwesertkil” before. The present theory
is not vulnerable to this objection because, in postulating a distinctive referential
meaning of the quotation marks, it can ascribe just this meaning to the utterance of
23
Ludwig and Ray (2017) relatedly propose that quotation marks are polysemous and yet the
semantic rules governing their different acceptations are unified by their common use of pure
quotation.
the marks in the utterance of that sentence. Since this meaning makes no mention of
previous uses of the quoted expression (“bigritwesertkil”, in this case), the intuition
that no such indication is present is preserved, and in fact directly explained.
The weaker echoic view was vulnerable to two objections: (i) one can make
allusive first uses of expressions in environments which by their very nature preclude
any indication that those expressions have been used before, and (ii) one can
spontaneously invent fully new playful words and make first uses of them when
writing to other people, as for example in the imagined utterance of (13). Again
the present theory is not vulnerable to these criticisms, because (i) the conventional
indication of the allusive quotation marks postulated by the theory does not imply
that the quoted expression has been used before, and in fact it’s not even a
full propositional indication; and (ii) the distinctive “distance” acceptation of the
quotation marks postulated by the theory does not imply that there have been
previous uses of “brutiful” and “aborradle”; in both cases the intuition that no such
indication is present is directly explained.
Our next criticism, of the “attention” and “difference” theories, was that, while it
should be easy to construct cases in which their postulated conventional indications
are conveyed without inducing any expectation in the audience that at least one of
the four standard indications (or possibly the non-standard “emphasis” indication)
are being meant, doing this seems difficult or impossible. Thus, if someone utters
(14) or (15) out of the blue, a typical reader will be puzzled by the utterances, will
not interpret them by assigning to them a minimal meaning. And if (14) or (15) are
uttered not out of the blue, but accompanied by cancelations of any allusion and
distance indications (the only intuitively sensible ones), as in (140 ) and (150 ), the
sense of puzzlement will be even greater. But on the present theory, this puzzlement
is perfectly foreseeable. We are trying to imagine a reader who can get content
simply by retrieving the indication that the quoted expressions in (14) and (15) are
being called attention to, or that they are being used in order to convey something
beyond or different from their usual meaning. And on the present theory one cannot
do just one of these things with the quotation marks; in linguistically acceptable
uses, one must utilize them under one of their acceptations, presumably under one
of the four acceptations described in the theory’s statement above. In the uses of the
examples, since it is thoroughly unclear which acceptation is at stake, or it has even
been excluded that the only possible acceptations are at stake, a sense of thorough
puzzlement will arise. The present theory thus explains the puzzlement, and is also
obviously not subject to the criticism, as it directly implies that one cannot use the
quotation marks just to convey a meager “attention” or “difference” indication.
Finally, let’s turn to the third and most significant criticism. Recall that the
utterers of (17) and (18) in the examples are presumably both calling attention to
the quoted expressions and signaling that they intend to convey something beyond
their usual meaning, so they have intentions in line with the meager meanings of the
marks postulated by the “attention” and “difference” theories. And yet the clearly
intended indication that the utterer is emphasizing the concepts expressed by the
quoted expressions is only conveyed at the cost of inducing a sense of oddness

in the reader, a sense that something is not quite right. Nothing like this should
happen if either the “attention” or the “difference” theory was right. By contrast,
the present theory is not subject to this criticism, for the utterers of (17) and (18)
are not intending to use the quotation marks under any of the acceptations that they
have according to the theory: this is evident, for the utterers manifestly do not mean
to convey any of the conventional indications corresponding to those acceptations.
Furthermore, the theory explains the feeling of oddness produced by the utterances
of (17) and (18), also as an effect of the utterers’ not using the quotation marks
under any of the acceptations postulated by the theory.24
Let’s now see how the present theory deals with the criticisms of semantic
theories presented in Sect. 2. First there was the criticism that semantic theories
were uneconomical (while pragmatic theories were economical). That pragmatic
theories are more economical than semantic theories is of course true. But recall
that considerations of economy hold good only ceteris paribus, i.e. provided that
both the pragmatic theory and the semantic theory we are comparing can explain
the same phenomena, or explain them to the same degree of satisfactoriness. And as
detailed in Sect. 2 and in the preceding paragraphs, I don’t think that the semantic
theory presented here and the pragmatic theories of Sect. 2 are tied with regard to
explanatory success. We have seen that there are several phenomena that pragmatic
theories cannot account for or are straighforwardly at odds with, and that the present
theory explains perfectly well. In such a situation, considerations of economy are
simply defeated.
Second, we had the idea that quoteless quotation is problematic for a semantic
account because semantic accounts must postulate unmotivated, ad hoc elements
in logical form corresponding to referential and impure allusion quotations. But
this is simply based on a misunderstanding. Semantic accounts are just accounts of
the semantics of the quotation marks, in the sense of what the marks can be used
conventionally to express. They are fully compatible with any theory of quoteless
quotation, and in particular with plausible theories that do not postulate any hidden
quotational element in the logical form of quoteless utterances and explain the
possibility of quoteless quotation via pragmatic effects. In fact, elsewhere (2001,
2011) I have proposed an explanation of the possibility of indicating reference to
“Socrates” with “Socrates” (instead of “ ‘Socrates’ ”) in (8) in terms of the pragmatic
mechanism of “speaker’s reference”; such an account is evidently fully compatible
with the present theory of the semantics of the quotation marks, which by itself
does not postulate anything at all about what is going on in cases like a normal
24
These remarks leave open the exact nature of the mechanism that allows the communication of
emphasis indications by means of quotations. But the mechanism is in all probability related to
other mechanisms that make possible the communication of indications via the use of misnomers
and catachreses. As noted above in the text, in all their acceptations the quotation marks
conventionally indicate a content involving the quoted expression. It is therefore easy for new
uses to arise that seek to indicate new contents involving the quoted expression, such as the content
that the concept expressed by the quoted expression is being emphasized.
utterance of (8). A related account is clearly perfectly possible for cases where
allusion indications are communicated without the help of quotation marks, as in
(9). In such cases the allusion indication is presumably communicated in virtue of
the fact that the expression alluded to is saliently known to speaker and audience to
have been uttered by some contextually relevant agent or agents. If so, this occurs
as a result of a fully pragmatic mechanism, without need of postulating hidden
elements in logical form, and is again fully compatible with the present theory of
the quotation marks, which are of course fully explicit elements of the sentences in
which they appear.
A third criticism of semantic views was that in some impure allusion cases, such
as in typical utterances of (3), the quotation is not under the scope of an explicit
reporting verb, which would seem to make it likely that the reporting indication
is accomplished pragmatically. The idea here seems to be that since the reporting
verb is absent, there cannot be a conventional indicator of the fact of reporting.
The present theory simply rejects this inference. Both when an allusion quotation
appears under the scope of an explicit reporting verb and when it does not, the
quotation marks conventionally carry the allusion indication. What is presumably
pragmatically accomplished, when the allusion quotation is not under the scope of
an explicit reporting verb, is the suggestion that the reported content is tacitly under
the scope of an implicit reporting verb, with its corresponding implicit subject, but
this fact should not be confused with the pragmatic theorist’s thesis that there is no
conventional indicator of the fact of reporting. In any case, if the present proposal is
to be rejected, it cannot be rejected through the fallacious inference from the absence
of the reporting verb to the absence of a conventional allusion indication carried by
the quotation marks.
The fourth and last criticism of semantic views attempted to exploit cases of
impure allusion uses of the quotation marks in which the quoted expression is
not a syntactic constituent, such as (10). In such cases the quotation presumably
has no self-standing meaning contributing compositionally to utterance meaning
or truth conditions, and yet some semantic theories (such as those in Benbaji
2005; Geurts and Maier 2005) are committed to assigning some such meanings
to quotations. However, the present theory is not such a theory. On the present
theory, the allusive use of the quotation marks conventionally adds an adverbial
or prepositional indication to the verb “to say” or a related attitudinal verb, but the
quotation itself is not (and the quoted expression need not be) a syntactic constituent
of the quotational sentence.
With this I finish the defense of my semantic view of impure quotation.
Semantic views of quotation have lately suffered sustained criticism coming from
the pragmatic theorist’s corner, much of it fueled by the promise of comparatively
economical pragmatic accounts of quotational phenomena. However, I hope to
have shown both that pragmatic theories face inherent limitations that prevent them
from explaining a range of such phenomena, and that the present semantic theory
accounts for these phenomena in a very natural way.25
References
Abbott, B. (2005). Some notes on quotation. In P. De Brabanter (Ed.), Hybrid quotations (pp.
13–26). Amsterdam: John Benjamins.
Amaral, P., Roberts, C., & Smith, E. A. (2007). Review of The logic of conventional implicatures
by Chris Potts. Linguistics and Philosophy, 30, 707–749.
Bach, K. (1999). The myth of conventional implicature. Linguistics and Philosophy, 22, 327–366.
Benbaji, Y. (2005). Who needs semantics of quotation marks? In P. De Brabanter (Ed.), Hybrid
Cruse, D. A. (2000). Meaning in language. An introduction to semantics and pragmatics. Oxford:
Oxford University Press.
De Brabanter, P. (2010). The semantics and pragmatics of hybrid quotations. Lang & Ling
Compass, 4, 107–120.
De Brabanter, P. (2013a). François Recanati’s radical pragmatic theory of quotation. Teorema, 32,
109–128.
De Brabanter, P. (2013b). A pragmaticist feels the tug of semantics: Recanati’s ‘open quotation
revisited’. Teorema, 32, 129–147.
Geurts, B., & Maier, E. (2005). Quotation in context. In P. De Brabanter (Ed.), Hybrid quotations
(pp. 109–128). Amsterdam: John Benjamins.
Gomez-Torrente, M. (2005). Remarks on impure quotation. In P. De Brabanter (Ed.), Hybrid
Gomez-Torrente, M. (2011). What quotations refer to. In E. Brendel, J. Meibauer, & M. Steinbach
(Eds.), Understanding quotation (pp. 139–160). Berlin/New York: Mouton de Gruyter.
Gomez-Torrente, M. (2013). How quotations refer. Journal of Philosophy, 110, 353–390.
Grice, P. (1975). Logic and conversation. In P. Cole & J. Morgan (Eds.), Syntax and semantics,
Vol. 3: Speech acts (pp. 41–58). New York: Academic. Reprinted in Grice, Studies in the Way
of Words (pp. 22–40). Cambridge, MA: Harvard University Press. References to the reprint.
Harris, J. A., & Potts, C. (2009). Perspective-shifting with appositives and expressives. Linguistics
and Philosophy, 32, 523–552.
Keeley, B. (2010). The book of “unnecessary” quotation marks: A celebration of creative
punctuation. San Francisco: Chronicle Books.
Kripke, S. (1977). Speaker’s reference and semantic reference. Midwest Studies in Philosophy, 2,
255–276.
Levinson, S. (1983). Pragmatics. Cambridge: Cambridge University Press.
25
I am grateful to Minyao Huang, Kasia Jaszczolt, Kirk Ludwig and Mark McCullagh for their
reactions to a presentation of part of this material at the 2015 American Philosophical Association
Pacific Division Meeting in Vancouver, and to Paul Saka and an anonymous referee for large
batches of written comments on earlier versions of the paper. Thanks also to audiences at the
UNAM and the University of Texas at Austin for helpful comments. The research was supported
by the Mexican CONACyT (CCB 2011 166502), by the PAPIIT-UNAM project IA 401015, and by
the Spanish MINECO (research project FFI2015-70707-P), and a generous sabbatical leave grant
from the DGAPA-UNAM allowed me to complete work on the paper.
Maier, E. (2014). Mixed quotation: the grammar of apparently transparent opacity. Semantics and
Pragmatics, 7(7), 1–67.
Potts, C. (2005). The logic of conventional implicatures. Oxford: Oxford University Press.
Reimer, M. (2005). Too counter-intuitive to believe? Pragmatic accounts of mixed quotation. In P.
De Brabanter (Ed.), Hybrid quotations (pp. 167–186). Amsterdam: John Benjamins.
Ruhl, C. (1989). On monosemy: A study in linguistic semantics. Albany: State University of New
York Press.
Saka, P. (2005). Quotational constructions. In P. De Brabanter (Ed.), Hybrid quotations (pp.
187–212). Amsterdam: John Benjamins.
Understanding quotation (pp. 303–322). Berlin/New York: Mouton de Gruyter.
Saka, P. (2017). Blah, blah, blah: quasi-quotation and unquotation. In P. Saka & M. Johnson (Eds.),
Shan, C.-C. (2010). The character of quotation. Linguistics and Philosophy, 33, 417–443.
Washington, C. (1992). The identity theory of quotation. Journal of Philosophy, 89, 582–605.
Reference and Reference-Fixing in Pure
Quotation
Manuel García-Carpintero
Abstract Cory Washington has identified three questions that a theory of pure
quotation should address, and on the basis of which such theories could be classified:
(i) what part of a quotation has a referring role, (ii) what the reference of that
referring part is, and (iii) how that reference is fixed. This paper compares the
answers given by the demonstrative, Davidsonian account that I have previously
advocated with those provided by what I regard as the most interesting alternative
to have emerged in recent years: the “disquotational” Tarski-inspired account
advocated by Gomez-Torrente and others. This paper develops three points. First,
the Demonstrative Theory, together with plausible independent principles, does
entail that, in default contexts, the (“semantic”) referent of a quotation is the
linguistic type of the token quoted. The theory explains in this way why instances
of Gomez-Torrente’s “Interiority Principle” are by default true, even if not “as
a matter of meaning alone”. Second, the theory provides the same account for
reference-determination in non-default cases, in which “semantic” referents differ
from linguistic types. Unlike Gomez-Torrente’s principle, the theory does without
a specific general naming convention to account for the determination of the
semantic referents of quotations. Finally, the issues dividing the Demonstrative and
Disquotational theories are not verbal, but substantive, and in principle amenable to
empirical resolution.
1 Introduction: The Deferred Ostension Theory
We use quotations – in written language, typically expressions in which some

quoted material is surrounded by quotation marks, or highlighted by some other
typographical device such as italicization – in theoretically distinguishable ways.
We use them in direct reports, to state what people say, as in ‘Boston is disyllabic’,
Quine said; in mixed reports, as in Quine said that a quotation ‘has a certain
anomalous feature’, which are primarily indirect reports of the contents that
M. García-Carpintero ()
LOGOS-Departament de Filosofia, Universitat de Barcelona, Barcelona, Spain
e-mail: m.Garciacarpintero@ub.edu

https://doi.org/10.1007/978-3-319-68747-6_7
170 M. García-Carpintero
someone said or thought, but in which some of the expressions that were used
for it are indicated; in scare quotes, to somehow dissociate ourselves from the use
we make of an expression, as in the ‘debate’ resulted in three cracked heads; in
emphatic uses, as in we sell ‘fresh’ pastry, to put emphasis on what an expression
contributes.
The central case that traditional theories of quotation have addressed is the pure
use, which might or might not be the core one from which adequate accounts of
these other uses might ensue.1 In such pure uses, quotations occupy the argument
positions that referential expressions such as proper names, descriptions and
indexicals fill up; to that extent they seem to be referential expressions themselves.
Thus, consider (1):
(1) ‘Boston’ is disyllabic.

Moreover, pure quotations appear to pass standard tests of rigidity for referential
expressions (Soames 2002: 251), which contributes to their appearing to be
referential expressions; thus, compare (2), which (to trained ears at least) seems
true, with (3), which seems false:
(2) The object that is ‘Boston’ is such that it could not have existed and failed
to be ‘Boston’ (and no other object could have been ‘Boston’).
(3) The object that is the only city name mentioned in this section is such that it
could not have existed and failed to be the only city name mentioned in this
section (and no other object could have been the only city name mentioned
in this section).
Washington (1992) identified three questions that a theory of pure quotation
should address, and on the basis of which such theories could be classified: (i) what
part of a quotation has a referring role, (ii) what is the reference of that referring
part, and (iii) how that reference is fixed. This paper compares the answers given
by the demonstrative, Davidsonian account that I have previously advocated with
those provided by what I regard as the most interesting alternative to have emerged
in recent years: the “disquotational” Tarski-inspired account advocated by Gomez-
Torrente (2001, 2011, 2013) and others.
In my previous work on quotation (1994, 2004, 2005 and 2012) I have put
together and defended a version of Davidson’s (1979) Demonstrative Theory of pure
quotation, which I call the Deferred Ostension theory (DO henceforth), given its
affinities with Nunberg’s (1993) views; I also showed how it can be used to account
for other uses. On a Reichenbach-inspired account of indexicals (Garcia-Carpintero
1998), there are token-reflexive rules conventionally associated with demonstrative
types, which distinguish semantically, say, ‘he’ from ‘you’. In virtue of them, when
a competent speaker uses a token demonstrative, by default he thereby manifests
1
Gomez-Torrente (2017) argues that quotation marks are semantically ambiguous. I (2011) have
suggested an account on which they are similarly semantically ambiguous, but nonetheless the
pure use remains at its core; cf. also Ludwig and Ray (2017).
Reference and Reference-Fixing in Pure Quotation 171
an intention to refer to an entity of a given conventionally indicated type (male,

in the case of ‘he’), in virtue of the fact that it stands in some contextually
specific existential relation (Peirce’s term) to the token he has produced, made
manifest by the speaker’s communicative referential intentions.2 By default, the
referent (in successful cases) is the demonstrative’s contribution to the proposition
constituting the main speech act it helps to convey; this accounts for its rigidity. The
mutually known descriptive condition also present in successful cases contributes to
characterizing a presupposed proposition (Garcia-Carpintero 2000), by specifying
the content of the speaker’s communicative referential intention.
In the case of demonstratives, sensible manifestation of those intentions will
typically require the speaker to carry out a further case of some standardized
intentional action (a demonstration: a certain pointing gesture, say) to instantiate
the contextually specific existential relation determining the referent. In contrast
to so-called pure indexicals, in typical contexts including the demonstrative there
are several objects satisfying the conventionally indicated type. In some expedient
cases, a single such individual is already manifest, no demonstration being needed;
this is in general the case with pure indexicals. When a successful demonstration
occurs, it picks out an object, the demonstratum, typically coinciding with the refer-
ent. In cases of deferred ostension, the demonstratum is not the referent; it is rather
an index (Nunberg 1993), helping to determine the referent in ways not unlike those
in which the token typically determines it: in virtue of further contextually specific
existential relations, properly indicated by the speaker (cf. Mount 2008: 162).
On DO, quotation works in natural languages along the same lines. Consider
a written utterance of the sentence (1). According to DO, quotation marks are
the linguistic bearers of reference, functioning like a dedicated demonstrative; the
token quoted material plays the role of a demonstrated index, and any expression
appropriately related to the index might be the referent.3 Reference is fixed by
some contextually suggested relation in which it stands to the quoted material; in
default cases the relation will be: : : : instantiates the linguistic expression __, but
there are other possibilities settled by whatever determines the semantic referents
of demonstratives, speaker’s intentions or contextual factors (more on this below).
Quotation marks are thus on this view dedicated demonstratives whose character
can be articulated as the expression in the salient relation to the quoted material.4
2
Some semanticists, following Kaplan (1989), think that there are philosophical problems with
this mention of particulars in a semantic theory, and prefer a more abstract characterization in
which the theory quantifies over types in context, as opposed to tokens. In previous work (1998), I
have argued that the alleged advantages of this are illusory, and the alleged problems non-existent.
Although we should assume a relatively abstract conception of tokens for semantic purposes, it is
crucial that referents are determined through existential relations with particulars. There will be
more on the metasemantics of demonstratives below.
3
I am using ‘quoted material’ to refer to what is inside the outermost quotation marks – Boston in
(2) – and ‘quotation’ for the whole constituted by it and the quotation marks.
4
Predelli (2008) provides a useful formalization of a demonstrative account. In part in order to
deal with issues raised by Cappelen and Lepore (2007) that I will discuss later, he argues for a
Note for later use that there is a significant difference between such dedicated
demonstratives and ordinary demonstratives, even complex demonstratives such as
‘this expression’ that explicitly articulate the sortal covering their referents in any
successful use: unlike the latter, they always include adequate indexes to act as
demonstrata. As a result, a default demonstrative rule for quotations, DDR, can
always safely operate in any context in which they are uttered:
(DDR) A quotation refers to the most salient linguistically significant expression

instantiated by the quoted material.
I’ll provide a justification for DDR in the fourth section, when I discuss the
metasemantics of demonstratives. As I will also explain there, this is just a default;
DO equally accounts for the fact that we do not merely refer with quotations to
expression-types, but (in the appropriate contexts) also to other expressions related
in some way to the token we use: features exhibited by the token distinct from those
constituting its linguistic type, as in (4); features exhibited by other tokens of the
same type but not by the one actually used, as in (5); other related tokens, as in (6);
or poems, songs, books and so on independently named by the instantiated types, as
in (7).5
(4) Use ‘Velázquez’, not ‘Velásquez’.

(5) ‘Hiss’ is a hissing sound.
(6) ‘Batman’ is painted in black.
(7) Almost all English schoolboys used to know by heart ‘The curfew tolls the
knell of parting day’, but not anymore; now they rather know ‘Imagine’.
On a contrasting Frege-Searle Use theory (UT), when an expression is referred to
by means of quotation the quoted material itself is the linguistic referring expression.
Quotation marks are not needed; when they are used, they serve to make clearer
the shift in syntactic and semantic properties effected on the quoted material by its
occupying that linguistic context: whatever its usual syntactic function, the quoted
material functions as a singular term; whatever its usual semantic function, in that
linguistic environment the quoted material refers to itself. In this way, as Washington
(1992) emphasized, UT accounts for those cases (particularly in spoken language)
where no form of quotation marks is used. DO entails that those sentences (say, a
spoken version of ‘cat’ has three letters without any special intonation) without an
distinction between the character and the meaning of quotation marks, taken as demonstrative
expressions, which I do not find necessary. Of course, the distinction is there, in that character is a
function – an abstract mathematical representation of a semantic property. However, to the extent
that the function represents the semantic rule whose default case has been described in the main
text, I do not think we need to invoke it for explanatory purposes – in particular, as I argue below,
it is not needed to deal with Cappelen and Lepore’s arguments.
5
Saka (2005: 315) makes the same point, calling it the “referential multiplicity” of quotations. I
take examples (4)–(6) from Gomez-Torrente (2011: 139–40) and García-Carpintero (1994: 261)
provides analogous cases. Cf. also Cappelen and Lepore (2007, chapter 7). Saka (2005, 206–7),
and Gomez-Torrente (2017, under “special names”) discuss cases like (7).
expression playing the syntactic role of subject should sound syntactically defective
in languages like English which lack the “pro-drop” feature. They are not perceived
to be so, however; the presence of the quoted material is enough for speakers not to
feel any syntactic solecism.
In presenting the two preceding views, I have also indicated what I take to be their
main strengths. These strengths also constitute the main problems for the rival view,
which they typically deal with by invoking the semantics-pragmatics divide. Thus,
for instance, the defender of DO will argue that, while the semantic referent of the
subject in spoken utterances of (1) without quotation marks or any related indication
is a city, and therefore what is said in those utterances is false, the speaker’s referent
will be an expression-type, and the content thereby conversationally implicated will
be true.6 The defender of UT, on the other hand, will argue that, while the “semantic
referents” of quotations are always the very same expression-types they include as
parts,7 their speaker’s referents may well be on many occasions other related entities
like those previously mentioned – tokens, poems, types that are not words, and so
forth.
I would like to emphasize for later use a point I have been careful to make in
presenting the two views: namely, that they differ as regards what they count as
the lexical referring device in quotation. DO does not deny that the quoted material
also counts as a referring expression, according to ordinary intuitive conceptions
of reference and expression. For we intuitively describe indexes in regular cases of
deferred ostension as referring to whatever they help refer to. Moreover, as we just
saw, in those cases in which no quotation marks are used, according to DO they are
the vehicles of speaker reference. The difference between DO and UT lies in that,
according to the former, in the strict sense of referring device in which only tokens
of linguistic lexical types that have a referring function in the system of a natural
language are such, it is tokens of quotation marks that are the referring devices in
quotation; only they convey semantic reference. According to UT, it is rather the
token of the quoted type that is the referring device, in that sense. If, as happens in
(1), the quoted type is itself conventionally a referring device, there is a systematic
ambiguity involved. Quotation marks, or alternative contextual resources, help to
disambiguate.
At first sight, quotation marks look much more like punctuation marks with a
disambiguating role of this kind than like fully-fledged linguistic referring devices.
The impression is even stronger if it is kept in mind that devices like italicization
and, in spoken language, some sorts of intonation are (in my view, at least) among
the different shapes that quotation marks can conventionally adopt.8 I understand
6
See Cappelen and Lepore (1999: 742, 2007: 35–43), and García-Carpintero (2004).
7
The scare quotes on ‘semantic referent’ have to do with the issues about the metasemantics
of demonstratives and their contribution, on the one hand, to semantic content (“compositional
content”) and, on the other, to “what is said” (“assertoric content”), to be discussed in Sect. 4.
8
As explained below, I take linguistic expressions to be abstract entities, which might have different
physical articulations. I am assuming that italicization and certain patterns of intonation are just two
among the different devices for quotation, in addition to the many varieties of graphic marks that
this impression; but no theoretically compelling argument against DO can be based

on it. For the assumption behind such an argument should be that only lexemes –
word-like lexical items – have semantic roles. But, of course, we indicate, say, co-
reference not only with lexemes (anaphoric expressions), but also by using tokens
of the same type, and the relation being of the same type is not a lexeme; we
indicate thematic roles with lexemes (declensional inflection), but also by means
of syntactical relations that are not lexemes at all; we indicate focus by means of
intonation, and so on and so forth. And what we indicate in all these variegated
ways are essential semantic features of natural languages.9
There are serious objections to DO.10 Some by Gomez-Torrente (2001, 2011,
2013) are the strongest I am familiar with; they help to motivate what is, in my
view, the best alternative to have emerged in recent years, a “disquotational” Tarski-
inspired account that Gomez-Torrente has advocated, DT, which he takes to be
a version of a third traditional theory of pure quotation, the Quine-Tarski name
account.11 In this chapter I would like to rehearse responses I have previously
provided to Gomez-Torrente’s criticism, and advance others to more recent ones.
I will not discuss UT any further here; I do not have any new criticisms to add to
those I have made in the past, or others by proponents of DT.12 Without attempting
to be exhaustive (which would of course be out of the question here) I want to outline
a comparison of how DO and DT deal with the evidence that has emerged over the
years as data that adequate theories should explain, focusing on what I take to be
most the sensitive issues to make a theoretically informed decision on the matter.
Gomez-Torrente argues that all theories except for his version of DT end up making
quotation “an eccentric or anomalous phenomenon” (2013: 353). I will aim to turn
the tables on him. I will reply to objections to DO, and I will show that DT fares no
better in the respects that Gomez-Torrente mentions. More decisively, I will show
languages use (cf. Potts 2005, sec. 4). Perhaps this is a good occasion to note that in my view mere
intuitions will not reliably tell us what quotation is; it is for theoretical accounts to characterize the
phenomenon. Of course, we do have intuitions on which we rely to provide paradigm examples
of the phenomenon we want to account for, such as (1), and the pre-theoretical data that must be
explained.
9
I do not assume that semantic devices other than lexemes form a natural class of their own. The
only relevant class is that of lexical items with a semantic role, be they lexemes or not.
10
De Brabanter has pointed out examples in which quotations occur in predicate positions, as
in “They use the editorial ‘we’ : : : ” (an example from Jespersen in De Brabanter 2002: 151; cf.
also De Brabanter 2013: 116, for additional examples. The objection is that demonstratives do
not appear to occur in those positions. However, such cases have been mentioned in the related
discussion of the predicative use of proper names; Rami (2015: 419) offers this example: ‘Leslie is
a she/her, not a he/him’. This could be followed with ‘the alleged she : : : ’.
11
Versions of DT were previously defended by Richard (1986), and afterward by the former
Davidsonians Cappelen and Lepore (2007); its influence is manifest in most recent work, cf. Potts
(2007), Predelli (2008), Pagin and Westerstahl (2010), Maier (2014), Bazzoni (2016) and Ludwig
and Ray (2017).
12
Cf. Gomez-Torrente (2013: 361–2) and Bazzoni (2016: 123–4). Cf. Saka (2005, 2011) for recent
defenses of the use theory and its account of the different varieties of quotation.
that all versions of DT establish an asymmetry in the interpretation of quotations,

dividing them into two groups for which their proponents have not so far given any
justification: those which genuinely, literally and semantically refer to their intuitive
referents, and those which at most “speaker-refer” to them. I’ll thus argue that DO
still affords the “loveliest” explanations (Lipton 2004).
Like most other philosophers, in spite of having written so much about it, I do
not care much about quotation itself. It just happens that it constitutes a sufficiently
isolated environment to test general issues of truly fundamental philosophical
significance. This article illustrates the point. I am much less interested in defending
DO than in signaling certain serious methodological mistakes by defenders of DT
that we all are unfortunately prone to in doing philosophy: appealing to spurious or
irrelevant intuitions, misconstruing the contours of the semantics/pragmatics divide,
and misunderstanding the nature of reference and reference-fixing.
In the next section, I’ll present DT and its main motivation. In Sect. 3 I’ll discuss
the extent to which both DO and DT make quotations context-dependent, at least in
a pre-theoretical sense. Finally, in Sect. 4 I’ll undertake a comparison of the virtues
and vices of both theories.
2 The Disquotational Theory
The main objection that Gomez-Torrente raises against Davidsonian theories includ-
ing DO (2001: 133, 2011: 141, 2013: 359; cf. also Cappelen and Lepore 2007:
69–70) appeals to the intuitive truth of disquotational schemas for quotations such
as:
(8) ‘ ‘Socrates’ ‘ stands for (refers to, denotes) ‘Socrates’.
A problem that (8) is assumed to pose for DO goes as follows. According to
DO, it is just the quotation marks that are linguistic expressions, susceptible of
being in the standing for or referring to relation; the whole quotations are not,
because they include what in fact is a mere extra-linguistic index that helps to
determine the referent. The expression quoted by the grammatical subject of (8)
is a whole quotation: the opening quotation mark followed by ‘Socrates’ followed
by the closing quotation mark. But that, according to DO, is not a term that is in
the standing for relation to anything – only part of it, the quotation marks, are;
moreover, all by themselves, out of context, they do not refer to ‘Socrates’ or to
anything else. A related problem is that instances of (8) need not be true according
to DO, because the subject-term might in some contexts refer to an item that doesn’t
refer to what the object-term refers to.
In earlier work I (2004) provide a rejoinder that Gomez-Torrente (2001: 134)
anticipates, although I (2005: 148) think that it places the Davidsonian in a
comparatively less attractive position vis-à-vis the Tarskian proposal he defends.
I will show below that, however serious it is, the alleged problem also affects
defenders of DT. Leaving the comparison for the final section, the reply was, in
a nutshell, that the intuitions on which the argument relies are not sensitive enough
to the distinction between properly linguistic expressions, part of the expressive
system of a particular natural language, and expressive resources in general, mere
signs. A quotation referred to by the grammatical subject of (8) in the contexts
where these intuitions are prompted is indeed such a sign, “referring” in an extended
sense to ‘Socrates’. This suffices for a theoretical account to adequately honor such
undiscriminating intuitions; the only remaining issue is whether, overall, it provides
a better explanation of all relevant facts.
Let me be explicit here about what I take words to be. I take them to be abstract
types, potentially with instances (tokens), identified by linguistic features, phono-
logical, syntactical and semantic (Wetzel 2009: 114; Bromberger 2011: 499). No
sensible theory can identify linguistic types with physical shapes; in physical terms,
types are highly disjunctive kinds. Just consider morpho-phonological alternation:
the occurrence of ‘nation’ in ‘national’ is one of the same root that also occurs
without the affix, in spite of the altered pronunciation. Words appear to be response-
dependent kinds, constitutively related to speakers’ recognitional capacities. Our
linguistic knowledge is to be described by an appeal to types, and to our ability to
perceptually identify their instances in ordinary utterances (Wetzel 2009, chapter 3;
Bromberger 2011: 490).
In Sect. 4 I will come back to these arguments in a more global setting, comparing
the pros and cons of the two theories. Let us now state the main tenet of DT, which
is supposed to capture the intuitions that prompt the previous objections, and its
variants in different versions of the view – what Gomez-Torrente (2001: 146, 2011:
142, 2013: 370) calls the Interiority Principle (cf. Richard 1986: 398–99; Cappelen
and Lepore 2007: 124; Bazzoni 2016: 124):
(IP) The quotation of an expression refers to the quoted expression.

Gomez-Torrente makes it clear that by ‘expression’ here he doesn’t mean lexical
item, or word, in the sense indicated above, but rather an abstract type (2013: 374)
instantiated by the quoted token, a “linguistically relevant graphical expression type
to which the quoted token belongs” (2011: 149), which might not be a lexical item
in the language of the quotation, or any other. This is why he (2011: 139–40) takes
(4) – in which the quotations refer to two different graphic articulations of what any
sensible account would count as the same lexical item – to be a standard example
of pure quotation, in which the quotations have their semantic referents, in contrast
to (5)–(6), which he accounts for instead as cases in which the intuitive referent is a
speaker referent distinct from the semantic referent (Gomez-Torrente 2011: 152).
It is not evident that the core assumptions of DT allow the quoted material to be
other than a word. I take it that a main goal of Gomez-Torrente’s (2013) account
of reference fixing in quotation is to address this worry, which, as he shows, was
left entirely opaque in Cappelen and and Lepore’s (2007) book-length treatment of
the issue.13 Both Richard (1986) and Gomez-Torrente (2001) take DT to develop the
classical Quine-Tarski “name” view of quotation. Their elaboration (intended to deal
with the Davidsonian “productivity” and “pictoriality” objections to be mentioned
momentarily) consists in taking quotations to be structured, while insisting that this
structure is not “syntactical” (i.e., relevant to a compositional account of how the
meaning of complex expressions depends on the meaning of their constituents and
mode of composition), but rather “lexical” (Richard) or “morphological” (Gomez-
Torrente). According to DT the quoted material is thus a constituent of the quotation,
but not a syntactico-semantic one, and hence it doesn’t need to be a word. As Quine
(1940: 26) famously put it, “each whole quotation must be regarded as a single word
or sign, whose parts count for no more than serifs or syllables.” In a well-developed
version of the view, Bazzoni (2016) articulates this idea by saying that quotables
(entities that can be referred to by means of quotations, in accordance with IP by
being parts of them) are not “syntactic” but – in Bazzoni’s perhaps idiosyncratic
terminology – “semantic” objects, i.e., mere objects to be talked about.14
The fact that quotations are nonetheless structured according to DT deals – as
Gomez-Torrente (2013: 383–390) carefully explains – with Davidsonian objections
to the Quine-Tarski classical name account based on the productivity and “picto-
riality” of the device.15 Gomez-Torrente (2013: 353) runs together both of these
Davidsonian “observations”, providing this characterization: “quotations, unlike
typical names, must somehow get interpreted in a general way that exploits the
salient pre-referential relation between a quotation and the expression between its
quotes, rather than via individual naming stipulations that need not exploit any such
relation”. Actually, I think the pictoriality observation concerns a pre-referential
relation between the meaning-vehicle (the quotation) and its meaning, the quoted
item – not the quoted material. In other places Gomez-Torrente (2013: 358, 386)
13
From the DT camp, Bazzoni (2016: 131–4, 138–9) still raises concerns in this regard for both
Gomez-Torrente and Cappelen and Lepore. I think that Gomez-Torrente’s discussion makes it more
intelligible than other versions of DT how quotations are structured for reference-fixing purposes,
even though their structure is not visible to the syntactico-semantic compositional machinery. I do
have qualms about the empirical adequacy of all DT accounts, discussed below.
14
Bazzoni (2016: 121, 129) denies that quotations are structured. But his reference-fixing rule
is just IP (124). He calls it an “identity function”, but this doesn’t make him a supporter of
“identity” or “use” theories, which he rejects; for him, the quotations include the quotation
marks, and refer to what is inside them. This is the only structure that DT discerns, but it is
a sort of (mereological) structure nonetheless. Bazzoni is right though that calling the structure
“morphological” or “lexical” suggests that what is inside the quotation marks – and thus what
might be semantically referred to by it – must be expressions in the language, thereby creating
compositionality worries; I guess this explains why Cappelen and Lepore (2007: 103) guardedly
say that on their view quotations are “not unstructured”. In fact until I read Gomez-Torrente’s 2011
paper, I took his view to be that quotations can semantically refer only to words.
15
See also Bazzoni (2016: 130.1). I put scare quotes around ‘pictoriality’ because I do not take
quotations to be literally iconic. Given the variety of quotable items, it is not correct to represent
the relation between quotation and quoted item as “picturing”; a quotation can hardly be said to
literally “picture” the embedded linguistic expression (García-Carpintero 1994: 256).
in fact states the observation in this way; and it is understandable that someone
assuming DT should mix them up, because, as we are seeing, the crucial feature of
the view is the identification of the quoted material and the quoted entity. In support
of DT, Gomez-Torrente provides other examples of referring expressions that, while
they are morphologically structured, and this structure is productively invoked
for reference fixing, are nonetheless syntactically and semantically unstructured.
Richard (1986) already mentioned the most obvious example, Arabic numerals;
Gomez-Torrente (2013: 376–383) provides other interesting examples such as some
conventions for street or personal names.
With regard to this issue Cappelen and Lepore turn the tables on demonstrative
theories by correctly pointing out that “[t]here is no requirement built in [them] that
secures any kind of picturing or hieroglyphic relationship between the quotation and
its semantic value. Given that the speaker’s intention (together with other facts about
what is contextually salient) determines the semantic value of a quotation, it also
remains a mystery how we are able to go from a quotation to what it quotes” (2007:
71). DT, as we have seen, explains this datum (which they sensibly call proximity
rather than picturing) in that, by sticking to IP, it contends that the semantic value
of a quotation is already contained in the quoting expression itself. But on DO this
is equally well accounted for by the fact that, as emphasized in the previous section,
quotations are demonstratives that incorporate their index or demonstratum, so that,
in default contexts where DDR applies, sheer linguistic knowledge is enough to
identify the referent.
By not taking the quoted material to be a lexical item, proponents of DT
can accept part of the evidence otherwise suggestive of a demonstrative account
discussed in the previous section, to wit, that quotations can refer to items which
are not expressions of the language to which they belong – or indeed any other
language. Thus, we can quote, in English first-order sentences, Chinese expressions
or objects that are not expressions of any language. On Gomez-Torrente’s view,
all these cases would be covered by the above rule, in which quotations refer to the
“linguistically relevant abstract graphic type” they enclose as quoted material. Other
versions of DT are even more liberal. Unlike Gomez-Torrente, Cappelen and Lepore
(2007: 23) contend that even pictures can be semantically quoted. Unlike both of
them, Richard (2008: 686) also accepts that quotations can semantically refer to
tokens, and suggests dealing with this by invoking a version of IP on which the
quoted material is a token, to which the quotation as a whole refers. Bazzoni (2016:
124) advances a single principle, assuming that the quotable “semantic entities”
that may occur as non-syntactic constituents of quotations might include particulars
such as physical objects (2016, 120, 127), in addition to graphical or acoustic types.
Nonetheless, none of these more liberal versions of DT than Gomez-Torrente’s can
still capture the examples (5)–(7), in which the referred item is not contained in the
quotation.16 Moreover, as I will point out below before we come to the reckoning,
this liberality comes at a price (which, I’ll argue, even Gomez-Torrente’s less liberal
version has to pay), in that the views are thereby exposed to the objections their
proponents raise against DO which I mentioned at the beginning of this section.
As Bazzoni (2016: 127–130) clearly explains, and all other defenders of DT have
insisted, the fact that the quoted item doesn’t occur in a quotation as a syntactic
constituent explains why on this view quotations do not fail to be compositional, in
a rather trivial way. On straightforward assumptions, compositionality is equivalent
to the principle of substitutivity. Quotations appear to falsify this, because we
cannot substitute salva veritate ‘Phosphorus’ for ‘Hesperus’ (assuming they have
the same meaning) in “‘Hesperus’ has eight letters.” But this fails to take notice
of the point that the two terms do not occur in the quotations as such. It is
only the whole quotations that, as two different referring expressions, are subject
to the compositionality and substitutivity principles. In this respect, DT is very
similar to DO. DO doesn’t contradict substitutivity or compositionality either,
just because quoted materials are not on that view expressions subject to such
principles; only the quotation marks (relative to a context) are. Quotations do not
fail compositionality principles, just as ordinary demonstratives don’t. It is not a
failure of compositionality that an utterance of ‘that expression has eight letters’
is true in a context in which the demonstratum is a token of ‘Hesperus’, but false
relative to another in which it is a token of ‘Phosphorus’.
Maier (2014: 625) argues that “a careful analysis” shows that the demonstrative
theory comes with a heavy “pragmasemantics” such that “we thereby effectively
lose compositionality”. His argument, however, is utterly unconvincing. He first
assumes an obviously inadequate character-rule for quotation marks on the demon-
strative account, on which they pick out “the most salient quotation-marked string
tokened in context c” (Maier 2014: 623). Unsurprisingly, he then argues that such
a character rule yields the wrong results in some cases. Indeed, the rule is wrong:
the most salient quotation-marked string tokened in the context of an utterance of
(4) might well be the one referred to by the second quotation, ‘Velásquez’; but
nonetheless the first quotation in the utterance does not refer to it.17 This is not
exactly Maier’s own point against the rule, but his is related. He “fixes” a context,
c, for an utterance of the sentence “‘love’ has four letters”; and then he argues
that, relative to c, the character rule for quotations marks wrongly entails that an
utterance of “‘cat’ has four letters” would express the proposition that ‘love’ has
16
In response to this concern, Ludwig and Ray (2017) suggest (without subscribing to it) a
“liberalization” of DT (cf. their rule (QCS)) which simply abandons Gomez-Torrente’s “interiority
principle” (i.e., their (GQ)) in favor of a version of an indexical view.
17
The problem with Maier’s rule is just a particular case of a well-known one for salience-based
metasemantics for demonstratives, such as Mount’s (2008). As Speaks (2016a) points out, even if
a fiery tiger is the most salient object in our context, this doesn’t make it the referent of ‘that door’
in ‘we can escape through that door!’, said while pointing to a door. Quotation marks are dedicated
demonstratives on DO, which conventionally make the quoted material appear on Kaplan’s (1989:
525) “demonstration platform”.
four letters, and would hence be true, while intuitively it is false. On this basis, he
contends that demonstrative accounts must replace his character rule by another one
he offers, borrowing from ideas in dynamic semantics, which is the one that raises
compositionality issues.
All of this, however, is only an artifact of the inadequate initial characterization of
the character. Consider instead the one given in the previous section, the expression
in the salient relation to the quoted material. Keeping in mind that this is the rule,
when we are asked to evaluate the sentence “‘cat’ has four letters” with respect
to c, assuming a demonstrative account, we can see that there are two different
questions that might be asked: (i) What is the value of the sentence (constituted
only by linguistic expressions), relative to a context in which the quoted material is
‘love’? The answer to this is clear: it is true, of course; for only the quotations marks
are constituents of the sentence, on demonstrative accounts. (ii) What is the value
of the sentence, given the context that comes with it in which the quoted material is
‘cat’, relative to a context in which the quoted material is ‘love’? And the answer
is that there is no way of telling, because the description of the case contains a
contradiction. So we do not need the dynamic “pragmasemantics”, and we have no
reason to fear that we will lose compositionality – as we have seen above we should
not, given any clear-headed demonstrative account.
Both DO and DT thus account for the most significant uncontested pieces of
data. Quotations do not pose a special problem for a compositional account of
the semantics of natural language; they are a productive and systematic device,
and somehow a “pictorial” one, in that there are general rules that allow for the
determination of their meanings on the basis of features of the meaning-vehicle;
entities other than linguistic expressions can be quoted. In Sect. 4 I will address
objections to DO, undertaking a global comparison of its merits and demerits vis-à-
vis DT; but before going into that, I’ll discuss in the next the context-dependence of
quotations.
3 Quotations and Context-Dependence
On all views we have discussed, quotations are ambiguous in a pre-theoretical sense;

but the views provide different theoretical accounts of this pre-theoretical ambiguity,
only on some of which is it properly ambiguous. Thus, compare (4), repeated below,
with (10):
(9) Use ‘Velázquez’, not ‘Velásquez’.
(10) ‘Velázquez’ is the name of a famous Spanish painter.
It is natural to take the quotation appearing as subject in (10) as referring to a
word, a proper name; while in (9) what intuitively looks like the same quotation
refers instead to a particular way of graphically articulating it, different from the
one referred to by the second quotation in (9), which is just a different way of
graphically articulating the same word – one reflecting better the way the name
is acoustically articulated in the Spanish spoken in Latin America, and parts of

Spain.18 For proponents of DO, this ambiguity is just a form of context-dependence.
Not so for proponents of DT; for them, the grammatical subject of (10) and the first
quotation in (9) are in fact different expressions, no matter what intuitively seems to
be the case; for they refer to different items and, in accordance with IP, they do so
by containing as non-syntactic parts those very distinct items: the word, in (10); a
specific graphic articulation thereof, in (9). As Gomez-Torrente (2013: 370) puts it,
“Interiority does not provide a method for assigning a reference to an utterance of a
quotation as a function of an aspect of its context; Interiority assigns a reference to
each quotation type, independently of any sensitivity to contextual factors”.
On DT, the apparent ambiguity is thus “pre-semantic”, as with the intuitive
ambiguity of ‘David’ when correctly used to refer to Hume and to Lewis on the
“multiple homonyms” view of proper names propounded by Kaplan, Kripke and
others, on which they share similarly articulated but different words (Sainsbury
2015). As Kaplan (1989: 562) puts it, context is here “regarded as determining
what word was used” rather than as “fixing the content of a single context-sensitive
word”.19 An alternative indexical view of names (cf. e.g. Pelczar and Rainsbury
(1998)) treats them instead in the way DO accounts for the intuitive ambiguity in
(9)–(10). On this view, ‘David’ as applied to Hume and Lewis is one and the same
indexical word, with a character that determines the referent given a contextually
“dubbing in force”.
At a theoretical level the difference between these two views is significant,
but proponents of DT are sometimes misleading or plainly confused when it
comes to appraising the relevance of such a difference for deciding between DT
and demonstrative accounts on intuitive grounds. Consider Cappelen and Lepore
18
Gomez-Torrente might disagree, given that he takes quotations to refer semantically to “lin-
guistically relevant graphical types”; for reasons I will take up in the next section, this is an
important issue for him. However, he (2011: 151 fn.) grants that they might also refer to words,
and hence end up being “mildly ambiguous, or polysemous”. I do not think he can sensibly deny
such “ambiguity”, given the clear disparities in what counts as linguistically relevant graphical
types instantiated in different contexts by what intuitively are the same expressions. In a given
context the linguistically relevant graphical type for the quotation ‘Velázquez’ might be the quoted
expression written with z, while in another, for what intuitively is the same expression, is a more
abstract one – one such that the relevant phoneme is written either s or z. By the way, I assume
Gomez-Torrente doesn’t mean in the quoted text that quotations are really ambiguous: when a
quotation refers to a word, and when what looks like the same quotation refers instead to a graphic
articulation thereof, according to IP the terms doing the quoting are in fact different, containing the
distinct entities they refer to; so there is no real ambiguity in such cases. There never is on DT, for
the reason he himself gives in the quotation from a later paper immediately below: two quotations
signifying two different semantic referents must themselves be distinct.
19
Cappelen and Lepore (2007: 154–5) deny that the Kaplanian label ‘pre-semantic’ is adequate for
the level at which the disambiguation takes place on DT, because it comes after disambiguation
and language-selection. They are right that this form of disambiguation differs from language- or
homonym-selection. But I do not think this distinction (on which they do not elaborate further)
affects my points below. My uses of the term ‘pre-semantic’ are thus intended to include not just
the processes contemplated by Kaplan, but also whatever is at stake here.
(2007). In sync with their tirades against contextualist views in their work on the
semantics-pragmatics divide, they reject the notion that quotations are context-
dependent. More specifically, they reject the following principle (2007: 68), which
DO endorses:
(QCS) Let S be a sentence with a quotation Q, containing no context-sensitive
expressions other than possibly Q. Two utterances, u and u’, of S can
express different propositions because Q in u and in u’ quotes different
items.
This denial notwithstanding, and at first sight paradoxically, they accept some of
the data emphasized by demonstrative accounts such as DO that suggest the truth of
QCS. For instance, they accept that the subject-term of (10) and the first quotation in
(9) refer to different entities; and also that, while in some contexts (11) below might
be true, in more ordinary contexts it is not. Moreover, they take this to be a semantic
phenomenon – while on Gomez-Torrente’s account it is merely a pragmatic one,
because at least one of the terms (the one on the right-hand side of the identity
sign) does not refer to the linguistic expression ‘Madrid’ but merely to its written
manifestation in a specific font, Helvetica (2007: 77–9):
(11) ‘Madrid’ D ‘Madrid’.
Cappelen and Lepore’s (2007, chapter 12) way of making their rejection of QCS
consistent with their acceptance of these data appeals to the point we have just
mentioned. Quotations might include “quotable items” which, although “signs”,
are not “expressions” (lexemes, or words). Quotations literally have those quotable
items – some of them non-expressions – as parts. Thus, to get the result that (11) is
false, according to IP – or rather according to the equivalent “quotation schema” QS,
‘e’ quotes/refers to e (2007, 123–4) – the quotable items included in the quotations
on the two sides of the identity sign should differ; but this means that the quotations
themselves differ, which is how the rejection of QCS can be upheld: when the quoted
items differ, the quotations doing the quoting differ too.
As a way of theoretically accommodating part of the data highlighted by
context-dependent, demonstrative proposals such as DO, while preserving the
context-independence of an account which relies on IP as the way of capturing
what is essential to quotation, Cappelen and Lepore’s suggestion is acceptable, as
indicated above. As Gomez-Torrente (2013: 364–6), Bazzoni (2016: 132–4) and I
myself (2012: 213–4) point out, however, Cappelen and Lepore are less perspicuous
about their commitments than they might be, to say the least. IP/QS is supposed to
be integrated as an axiom in a semantic theory (2007, 130–1). Given the common
assumption that semantic theories deal with the properties of linguistic expressions,
this might seem to entail that on their view quotations are after all structured in such
a way that the quoted material is itself an expression – which would raise the issues
about substitutivity and compositionality discussed in the previous section. By
remaining unclear on what they take to be the semantically significant structure of
quotation, they make their view either idiosyncratic or plainly obscure. In contrast,
we saw above how Gomez-Torrente (2013) provides a clearer treatment of these
issues.
Be this as it may, what I want to emphasize here is that the means by which DT
manages to avoid making quotations context-sensitive is rather Pickwickian. QCS
has a pre-theoretical interpretation, relative to which it is confirmed by examples
such as (9)–(11); and a theoretical one that, depending on the theory, either embraces
or rejects it. The pre-theoretical interpretation appeals to pre-theoretical notions of
what expressions are and what identity conditions they have. On at least some of
these notions, the second quotation in (11) is the same expression in a context in
which (11) is true (because it refers to the linguistic expression ‘Madrid’), and in
another in which it is false instead (because it refers to its written manifestation
in Helvetica). This pre-theoretical interpretation thus confirms QCS. On DT, the
two quotations are different expressions, and hence the example does not validate
QCS. On DO, QCS is straightforwardly true. Proponents of DT and DO in fact
do not disagree with any of this: in particular, if we understand ‘quotation’ in
a pre-theoretical way, all are fully in agreement with the contextualist diagnosis.
Of course, at a theoretical level at most one view is right, so that at that level
there might be reasons to reject QCS. But we should keep this distinction in mind,
because, as we are about to see, it is unclear whether its rejection at that level spares
proponents of DT the theoretical costs that theories positing context-dependence
like DO thereby incur. To what extent does it yield a more explanatory account to
locate the context-dependence at the pre-semantic, rather than the semantic level?
In my discussion of some of the arguments that Cappelen and Lepore (2007) deploy
against DO, I will argue that the answer to this question is, as rhetorically suggested,
“very little, or nothing at all”.
The first argument relates to the two arguments discussed at the beginning of
the second section in order to motivate IP. Cappelen and Lepore (2007: 69) argue
that demonstrative views accepting the wide-ranging context-dependence assumed
by DO “cannot guarantee the truth of (dis)quotational sentences”, such as (12),
because, on such views, it “should be on a par with” (13):
(12) ‘ ‘Quine’ ‘quotes ‘Quine’.

(13) ‘that’ demonstrates that.
The reply to this on behalf of DO is the one I mentioned above as a rejoinder
to Gomez-Torrente’s (2001: 133, 2013: 359) version. First of all, notice that there
is a significant difference between (12) and (13), as emphasized in the first section:
instances of (12) always include adequate indexes to act as demonstrata, unlike
instances of (13), which is why the default demonstrative rule for quotations (DDR)
can always safely operate in any context in which (12) is uttered. Utterances of (12)
are hence true in such default contexts.20 This explains perfectly well the intuitions
of an asymmetry between (12) and (13). It also explains the intuition that (12) is
true “as a matter of meaning alone” (2007, 70), at least if we leave this claim, as we
should when discussing intuitions, at a merely intuitive level.
20
Cp. the response to Cappelen and Lepore provided by Predelli (2008: 566) on behalf of the
version of the demonstrative account that he formalizes there.
The second aspect of this objection (2007: 69) is that according to DO there are
false instances of QS; (14) might be a case in point, in the suggested context:
(14) ‘ ‘Quine’ ‘ quotes/refers to ‘Quine’.
In reply, I note first that I do not see any non-question-begging reason why this
should be wrong. It would be good to have experimental data on this; for what it
is worth, my bet is that ordinary speakers would simply feel confused if we ask
them whether an utterances of (14) is true, and also that many would find it false if
we set up a context that makes it salient that the Helvetica version of the name is
a candidate referent for the second quotation, like those that Cappelen and Lepore
themselves (2007: 152) provide.
Leaving this aside, what makes Cappelen and Lepore’s argument truly surprising
is that they agree that some utterances of (14) are false. They (2007: 153–4) argue
that this does not contradict the truth of (12) “as a matter of meaning alone”, because
in false utterances of (14) different quotations are at stake: the quotation that is part
of the subject is not the same as the one used as the object. But how can this be
considered an advantage over DO when it comes to accounting for pre-theoretical
intuitions? According to DO, there are false instances of QS (‘e’ quotes/refers to e),
such as an utterance of (14) in the envisaged context; and what determines whether
we have a true or a false instance of QS is whatever determines the interpretation
of demonstratives in context (speakers’ intentions, or coordinating intentions: more
on this below). According to Cappelen and Lepore’s account, all proper instances
of QS are true; an utterance of (14) could either be an instance of QS, and then
true, or not a true instance of QS, and then false; and what determines whether
it is one or the other is whatever grounds the pre-semantic facts, which, as far as
we can tell (they do not say) will be exactly the same sort of thing: intentions,
coordinating intentions or what have you. It is a serious methodological error to
assume that speakers’ intuitions that are acceptable empirical data for semantics
can decide which one of these proposals is right, because the difference between
them lies at a theoretical level that is far beyond the scope of such intuitions. At
the intuitive level, both DO and DT accept violations of QS, because intuitions do
not distinguish the expressions quoted on the left side and used on the right side;
intuitively quotations are ambiguous, whether or not they really are theoretically.
Two additional arguments against demonstrative accounts face exactly the same
problems. Those accounts, Cappelen and Lepore (2007: 71–2) say, “over-generate”;
they predict that instances of (15) can be true, just as instances of (16) can:
(15) ‘a’ ¤ ‘a’.
(16) that ¤ that.
In reply, we can make the same points as before: (i) DDR explains why the schema
feels as if its instances were false “as a matter of meaning”. (ii) Cappelen and Lepore
themselves accept apparent exceptions, illustrated by contexts where (11) is false;
they can only defend the general falsity of (15) by appealing to their theoretical
account of pre-theoretical context-sensitivity, but they agree that schema (15) has
counter-instances when instances are intuitively individuated.
Quotations, they also argue, fail to pass two standard tests for context-sensitivity.
Firstly, the “Inter-context Disquotation Indirect Reporting Test”: If an occurrence of
an expression e in a sentence S tends to block disquotational indirect reports (i.e.,
tends to render them false), then that’s evidence that e, and so S, is (semantically)
context-sensitive (cf. 2007: 73). Secondly the Collectivity Test: If a noun e is
(semantically) context-sensitive (i.e., if it changes semantic value from one context
of utterance to another), then on the basis of merely knowing there are two contexts
of utterance, one in which ‘e is F’ is true and the other in which ‘e is G’ is true,
we cannot automatically infer that there is a third context in which ‘e is F and
G’ (cf. 2007: 74). The points just made apply here as well, mutatis mutandis: (i) the
appearance that quotations fail the two tests is accounted for by cases in which DDR
applies; (ii) there are cases in which quotations pass the two tests, as Cappelen and
Lepore would have to admit if we present them in terms amenable to pre-theoretical
intuitions.
4 Does DT Provide Lovelier Explanations?
As indicated, in his more recent work Gomez-Torrente reproduces the objections

by Cappelen and Lepore to demonstrative accounts such as DO that I have already
discussed. He articulates them from the perspective of his view about which items
quotations semantically refer to, which is less liberal than Cappelen’s and Lepore’s.
As I will show, however, he doesn’t thereby escape the indictment of methodological
misjudgment (unwarranted reliance on intuitions) that I have just made about theirs.
Gomez-Torrente presents the most damaging assumption by DO that he takes issue
with thus (my emphasis): “the demonstrative phrases of Davidsonian analyses can
in principle refer in some contexts to things that quotations (or quotation marks)
as a matter of conventional principle cannot refer to in any context” (Gomez-
Torrente 2013: 359). I understand that he wants what intuitively is one and the same
quotation, such as the first in (9) and the subject-term in (10), to semantically refer
(“as a matter of conventional principle”) to just one entity, a “linguistically relevant
graphical type”, while all other possible referents are accounted for as speaker
references. But we have already seen that this aspiration is doomed to failure, and as
a result the indictment of Cappelen and Lepore’s proposal applies also to his, even
if the number of examples we can provide to show it is smaller.
Thus, consider an instance of the “disquotational principle” QS such as (17), in a
context in which it is made clear (say, by running commentaries while one writes it
on a board in a discussion of this topic) that the quotation mentioned in the subject
is intended to behave as the subject-term of (10), while the one used in the object
works instead like the first one in (9):
(17) ‘ ‘Velázquez’ ’ refers to ‘Velázquez’.
From a pre-theoretical perspective, on which quotations are manifestly (pre-
theoretically) ambiguous and validate QCS, the quotation used as the object in such
a case is the same expression as the one mentioned in the subject. Now, Gomez-
Torrent should admit that, although there are contexts in which the expressions they
both refer to are also the same (say, one and the same word, or graphical type), in
agreement with DT’s disquotational schema, there are also contexts (such as the one
just described) in which both “linguistically relevant graphical types” differ, thereby
disproving the schema at the relevant pre-theoretical level: the quotation mentioned
as the subject refers to a word, while the quotation used as the object refers to a
particular graphical articulation thereof.
Cappelen and Lepore (2007: 69–70) say “to be told that we’re not guaranteed of
the truth of a quotation sentence like [(17)] as a matter of meaning alone, will sur-
prise most competent speakers”. However, if they limit themselves to using notions
deployed in claims that, if competent speakers find them surprising, this is evidence
for semantic theories, they will have to admit that their own view delivers surprising
news of this kind. In a similar vein, Gomez-Torrente (2013: 359) complains that “it
is not guaranteed purely by Davidsonian theory that apparently context-independent
disquotational truisms such as [(17)] are true”. Unfortunately for him, nor is it
guaranteed purely by the version of DT that he himself subscribes to.
Hence, if there is a cost here, it is a cost both for DO and DT, in any version.
The fact that no violation of the schema occurs when “expression” and “ambiguity”
are theoretically disambiguated along the lines of DT is in no way an advantage of
that theory, first and foremost on account of the point that I have been emphasizing:
ordinary speakers’ intuitive judgments empirically relevant for semantic theorizing
do not deploy such theoretically elaborate notions. In addition, DO can claim that
same spurious virtue: no violation of proper instances of the schema occurs either
when the disambiguation is made along its lines. Obviously, ordinary demonstrative
sentences obtained by replacing the outermost quotes in (17) with ‘this expression’
need not be true in contexts in which the quotation mentioned in the subject invokes
for reference-fixing a different index from the quotation used in the object. But in
any case the virtue is spurious: if it is true that ordinary speakers make the judgment
that sentences with the apparent form of (17) are “analytic”, both DO and DT agree
that they are wrong and need to be corrected; and both honor a theoretically nuanced
form of the judgment.
To evaluate another criticism of DO that Gomez-Torrente makes, and a related
reason he provides to uphold his version of DT vis-à-vis Cappelen and Lepore’s,
we need to go into the distinction between semantic and speaker reference, and how
it applies to demonstratives. The complaint against DO is that it “seems to attribute
excessive referential possibilities to quotations” (2011: 144). He illustrates this by
contrasting a case in which one points to a token of ‘Velázquez’, and utters (18),
with one in which one makes a spoken utterance of (19); unlike the former case, he
says, in the latter “it seems one can’t : : : successfully convey to one’s audience the
intended true proposition that Velázquez was a great painter, or at least one can’t do
this without adding complicated peculiarities to the contextual setting” (2011: 145):
(18) This man was a great painter.

(19) ‘Velázquez’ was a great painter.
Another example contrasts “This boldface type is a very dark boldface”,

said while demonstrating a normal, non-bold face Times New Roman token of
‘Velázquez’, with an utterance of “‘Velázquez’ is a very dark boldface”. While
in the former case “I can easily convey and perhaps semantically express the
(let’s suppose, true) proposition that the boldface version of Times New Roman
“Velázquez” is a very dark boldface”, in the latter I cannot “easily manage even
to convey that proposition, as I probably require a fairly complicated contextual
setting to do so” (ibid., 145–6). Note, for later use, that in both cases Gomez-
Torrente’s point is restrained: not that what is possible with ordinary demonstratives
is impossible with quotation marks, but only that the latter is more difficult and
requires special contexts. This is as it should be, because he himself provides good
examples of contexts in which the relevant propositions would be conveyed. This is
ok for him, because his claim is merely that, although quotations might refer in the
ways indicated, these are cases of speaker reference, requiring contrived contexts.
Gomez-Torrente’s criticism of Cappelen and Lepore’s version of DT is the
contrasting one, that it is “insufficient” (2011: 148). I of course agree with the
criticism, which DO upholds, but not with his proposal to deal with it in terms
of speaker reference; as I will argue, the relevant cases involve semantic reference,
in the terms that the notion applies to demonstratives. The criticism is that, as my
previous work emphasized and examples (5)–(7) illustrate, we can use quotations
to refer not just to types distinct from linguistic expressions as in (4)/(9), but also
to entities distinct from those contained in the quotation, no matter how liberal a
notion of part we are using. Consider for example (20), for which Cappelen and
Lepore say that they and their informants find it impossible (or very difficult) to find
a true reading (2007: 72):
(20) ‘I’ tastes like peach.
In spite of the difficulties reported by Cappelen, Lepore, and their informants, it is
not hard at all to find a context in which an utterance of (20) appears to be true;
just imagine that we are speaking about the items in a bag of sweets in the shape of
letters. Not much imagination is needed to contemplate true utterances of (6) above
either. In fact, at other places Cappelen and Lepore are more prudent and do not
deny the intuitive data (2007: 71).
However, the semantic account that IP/QS affords for (9) is unavailable for these
cases, because the quoted item is not part of the quotation. Cappelen and Lepore
(2007: 76) consider explaining them in terms of familiar pragmatic strategies, but
reject them, so they are left without any account for these cases. Gomez-Torrente
(2011: 152–6) provides a detailed elaboration of the first pragmatic proposal
(“conversational implicature”) that they consider, in terms of Kripke’s (1977) notion
of speaker reference: while the quotation keeps its semantic referent (the relevant
“linguistically relevant graphical type”), and hence what is said is false, the speaker
manages to convey a truth about the intuitive referent by relying on familiar Gricean
mechanisms. I’ll come back to Gomez-Torrente’s account presently. But whatever
the result of this debate among proponents of DT, the main problem for both of
them lies in justifying the asymmetries in their accounts of these cases, (9) on the
one hand, (5)–(7) on the other. Firstly, as we have seen, if the arguments they provide
against demonstrative accounts worked in the case of quotations referring to items
non-contained in them, they would also apply to some referring to contained items.
Secondly, as I’ll now argue, once one accepts as a genuine semantic phenomenon
the (more or less restricted) pre-theoretical semantic context-dependence they are
prepared to concede, it is unprincipled not to grant the one they reject.
I have been placing scare quotes above when talking about semantic reference in
the case of demonstratives. Heck (2014) denies that the distinction between seman-
tic and speaker reference can get a grip in that case; reference with demonstratives
(and in general with context-dependent expressions) is just speaker reference. I think
his claims are rhetorically overblown; the qualifications he provides make me think
that he wouldn’t disagree with the limited sense in which, I suggest below, we can
nonetheless make sense of the distinction.
First of all, following Lewis (1980), I have argued elsewhere (Garcia-Carpintero
2006) that we should distinguish semantic content (“compositional value”, as
writers such as Ninan (2012), Rabern (2012) or Yalcin (2014) call it in making
essentially the same point) from assertoric content – or “what is said” in Grice’s
understanding of ‘saying’, which, as Bach (1994: 143) puts it, is a “generic
illocutionary verb that describes any constative act whose content is made explicit”.
Semantic content is ascribed to sentences in order to fulfill core explanatory tasks
for theories of natural languages: accounting for facts about systematicity and
productivity in understanding, communication and acquisition, judgments about
entailments or truth-value or appropriateness relative to given situations. Ultimately,
the intuitive data in all those cases concern assertoric contents, and hence semantic
contents should contribute to them; but, as Lewis (1980) points out, it doesn’t follow
that they need to be identified, and (as argued in the papers just mentioned) there are
very good reasons for not doing so. In particular, when it comes to demonstratives
and context-dependent expressions in general, semantics need not bother identifying
the values they might have in particular contexts. Semantic content is essentially
character-content, and hence not truth-evaluable by itself; as shown in the work
referred above, this is perfectly compatible with the point that compositional rules
typically compose the values of context-dependent expressions in context, not their
characters.21
This doesn’t mean that we should accept “radical contextualism” (or Cappelen
and Lepore’s full-blown “speech act pluralism”) concerning assertoric content, or
“what is said” in the Gricean sense, for (unlike pragmatically implied contents) it
is semantically constrained. I think it is a social notion, or, as Perry (2009: 191)
puts it, a “forensic” one. This can be appreciated by critically taking up an argument
21
As I have also pointed out (García-Carpintero 2013: 518), considerations by minimalists such
as Cappelen and Lepore and Borg in favor of the truth-evaluability of semantic contents invoke
intuitions about assertoric content; however, the truth-evaluable semantic contents that minimalists
posit are as theoretical and remote from the intuitions of ordinary speakers about what is said as
the index-relative constructs that Lewisians assume. Hence, features of the former, such as truth-
evaluability, cannot without further ado be ascribed to the latter.
that some writers (Fricker 2012; Green 2015; Lepore and Stone 2015) make, that
we cannot indirectly conversationally implicate assertions. A first consideration they
deploy invokes the intuitive distinction between lying and misleading. We intuitively
distinguish communicating contents we believe false with the intention to deceive
on the one hand by means of sentences that literally convey them (this would be a
straightforward lie), and on the other by indirectly conveying them as implicatures,
when the literal content is true (this would be merely misleading). This allegedly
shows that we cannot make indirect assertions, because asserting what one believes
to be false with the intention to deceive suffices for lying. A second consideration
to the same effect has to do with the epistemological role of assertion. Assertions
are intended to provide testimonial knowledge; but the indeterminacy created by
indirection, they argue, makes that impossible.
In a recent piece (Garcia-Carpintero forthcoming), I argue that these consider-
ations only show that we intuitively distinguish a subclass of assertions, roughly
corresponding to Gricean sayings: those whose content is made sufficiently explicit,
by being constrained by the semantic content of the uttered sentence. What suffices
for lying is not asserting what one believes to be false with the intention to deceive,
but saying it; indirectly asserting it would be merely misleading. Similarly, sayings
provide for clear-cut cases of acquisition of testimonial knowledge. However, they
themselves are not free from indeterminacy – in fact I argue in the mentioned
paper that assertoric contents themselves are not propositions strictly understood,
but propositional concepts, something with which Heck (2014: 359–61) agrees; and
in the relevant contexts merely implied assertions might be much more determinate.
The resulting notion of what is said is thus inextricably social, as Camp (2006)
insists. It gives us some criteria that might help us to identify the contribution of
specific expressions to assertoric content, or what is said, and thus indirectly to
semantic content, to the extent that the latter constrains the former. Some authors
(cf. Michaelson 2016: 481) have also argued for appealing to the lying-misleading
distinction to trace the semantics/pragmatics divide, applying it to cases such as
debates about referential uses of definite descriptions; and we can similarly appeal
to considerations about the testimonial provision of knowledge.
When it comes to the metasemantics of demonstrative reference, a view along
these lines regarding their contribution to what is said – i.e., their semantic referents
in the suggested understanding – speaks in favor of accounts that rely on the
intentions of the speaker, in the qualified way articulated by Bach (1992) and
elaborated by others including Perry (2009) and Speaks (2016b). On these views,
for o to be the semantic referent of demonstrative d in context c two conditions must
be met: (i) the speaker intends o to be the value of d in c; and (ii) the speaker intends
(and hence, if rational, justifiably believes that what he does makes it more probable
than otherwise) that his audience take o to be the value of d in c.22
22
In previous work (García-Carpintero 2007: 212) I have argued for this view, endorsing Bach’s
(1992) defense of a qualified form of intentionalism against “conventionalist” or “contextualist”
accounts of the metasemantics of demonstratives. I take the specific articulation provided here
Applied to quotation marks understood as dedicated demonstratives, this pro-

posal validates the DDR mentioned above, and it provides adequate responses to
Gomez-Torrente’s objections that it entails excessive referential possibilities. Note
first that quotation marks are dedicated demonstratives; we can think of them as
coming with the sortal “expression”, understood in the intuitive fully general sense
in which it applies to tokens and types that are not words. Note also that the view
just outlined about demonstratives allows for the distinction between speaker and
semantic reference; for instance, cases that Speaks (2016b: 305–6) describes as of
“insufficient intentions”, in which speakers intend to refer to something but fail to
provide sufficient grounds to their audiences to pick it out, are cases of speaker
reference without semantic reference.23
Let us now explain the contrast between “This boldface type is a very dark
boldface”, said while demonstrating a normal, non-bold face Times New Roman
token of ‘Velázquez’, and “‘Velázquez’ is a very dark boldface”. In the first case,
the sortal in the complex demonstrative guides by itself a competent, reasonable
and attentive audience to the intended referent, so both conditions are met. Not so
in the latter case; in any normal context the second condition cannot be taken to be
satisfied, when the intended referent is the boldface type Times New Roman. But of
course, there are “complicated contextual settings” in which it is satisfied, and that
type is the semantic referent. Gomez-Torrente (2011: 146) mentions one: a number
of fonts are being discussed, and every font is represented by the written name of
a different painter. Similar points can be made about the contrast between (18) and
(19): whether it is the sortal presupposed for quotation marks or the one explicit in
‘that man’ makes a difference regarding what it takes to satisfy the two conditions
on semantic reference for the demonstrative. But again, it is not that difficult to think
of “complicated” settings in which the second is satisfied. Imagine for instance that
we are giving examples of three-syllables words that we like, and the reasons why:
‘platypus’ sounds nice; ‘resentment’ is my favorite emotion; ‘Velázquez’ is a great
painter.
from Speaks (2016b: 229–30), whose arguments against King’s (2014) “coordination” account I
find compelling. I have run together Speaks’ two proposals for the second condition, because I do
not share his rejection of the conceptual connection between rational intention and belief on the
basis of which he distinguishes them. In support of this rejection he (2016b: 331) mentions cases
such as that of a basketball player who shoots from behind halfcourt just before time expires; but
the way of stating the connection that I have given in the main text, which I take from Sinhababu
(2013: 681), adequately captures them. Heck (2014: 339–343) endorses a similarly nuanced form
of intentionalism and hence a similar notion of “semantic reference” for demonstratives, even if he
dismisses the label.
23
More difficult are cases of “conflicting intentions” (Speaks 2016b: 306–8), such as the famous
“picture of Carnap/Agnew” example from Kaplan (1978: 30), in which both candidates appear to
satisfy both conditions. I have suggested in the previously cited work that the nesting-intentions-
based proposals of Bach (1992) and Perry (2009) point in the appropriate direction in order to
select which one provides the semantic referent, if any (it is the “means” intention, as opposed to
the “ultimate goal” one), but I cannot go into this here; cf. Speaks 2016b for the thorny problems
these issues raise.
What does the lying/misleading test, and the considerations about the epistemic
role of assertions, tell us about the semantic character or otherwise of reference by
means of quotations to expressions not contained in them, as in (5)–(7)? It seems
clear to me that we intuitively lie, and not merely mislead, when we utter (6) or (20)
knowing very well that what we are saying about the relevant tokens is false, with
the intention to deceive; but, correspondingly, also that we can straightforwardly
put our audience in a position to acquire testimonial knowledge by means of them,
when we instead say what we know.
Thus, there is an important difference between the accounts of quotation that DT
and DO provide. DO can account on equal terms for the contribution of quotations
to what is said in (4), in (5)–(7), and in both true and false utterances of (11). The
version of DT that Cappelen and Lepore defend provides an account of (4) and
(11), but tells us nothing about (5)–(7). The version that Gomez-Torrente defends
does explain (5)–(7), and the intuitively false utterances of (11) – which he also
takes to involve speaker reference, because he takes the semantic referents of both
quotations to be the same “linguistically relevant graphical type” (2011: 158–9).
To the extent that these writers commit themselves to the mechanisms they posit
as having some psychological reality,24 these differences between the two views
have empirical consequences. What do the empirical facts tell us about them? At
least at a phenomenological level, I am not aware of any difference in my conscious
experiences when I interpret quotations in each of those categories, and, as I have
been insisting, I find it methodologically unwarranted to ascribe this awareness to
the intuitive judgments of ordinary speakers. There could certainly be processing
differences at a subpersonal level, and it would be very nice to explore the issue;
but, of course, the authors I am discussing do not provide any evidence of that kind,
and in fact, for all I can tell, they might lack (as I do) even the capacity to surmise
whether it is at all feasible to obtain data of this kind in the present state of our
knowledge on this matter. What, then, should our provisional conclusion be?
Gomez-Torrente’s (2013: 353) main argument against all other theories of
quotation is methodological: he argues that they all end up making quotation “an
eccentric or anomalous phenomenon”. My main claim against DT in this chapter
turns the tables on him on this point. I have provided what I take to be a decisive
reply to his objections to DO, showing that DT fares no better in the respects
he mentions. And I have shown in addition that all versions of DT establish an
asymmetry in the interpretation of quotations, dividing them into two groups, for
which their proponents have not so far given an empirical justification (although
in principle there may be one). On methodological grounds, therefore, to my taste
at least, as far as we can now tell, demonstrative accounts in general, and DO in
particular, provide the loveliest account.25
24
Gomez-Torrente (2011, 153) at least appears to do so.
25
Financial support for my work was provided by the DGI, Spanish Government, research project
FFI2013-47948-P and the award ICREA Academia for excellence in research, 2013, funded
by the Generalitat de Catalunya, and from the European Union’s Horizon 2020 Research and
References
Bach, K. (1992). Paving the road to reference. Philosophical Studies, 67, 295–300.
Bazzoni, A. (2016). Pure quotation, metalanguage and metasemantics. Linguistics and Philosophy,
39, 119–149.
Bromberger, S. (2011). What are words? Journal of Philosophy, 108(15), 486–503.
Camp, E. (2006). Contextualism, metaphor, and what is said. Mind and Language, 129, 280–309.
Cappelen, H., & Lepore, E. (1999). Using, mentioning and quoting: a reply to Saka. Mind, 108,
741–750.
Cappelen, H., & Lepore, E. (2007). Language turned on itself. Oxford: Oxford University Press.
De Brabanter, P. (2002). Making sense of mention, quotation, and autonymy. A semantic and
pragmatic survey of metalinguistic discourse. Ph.D. dissertation, Linguistics: Universite Libre
de Bruxelles.
De Brabanter, P. (2013). Francois Recanati’s radical pragmatic theory of quotation. Teorema, 32(2),
109–128.
Fricker, E. (2012). Stating and insinuating. Proceedings of the Aristotelian Society
Suppl, LXXXVI , 61–94.
Garcia-Carpintero, M. (1994). Ostensive signs: Against the identity theory of quotation. Journal
of Philosophy,XCI , 253–264.
Garcia-Carpintero, M. (1998). Indexicals as token-reflexives. Mind, 107, 529–563.
Garcia-Carpintero, M. (2000). A presuppositional account of reference-fixing. Journal of Philoso-
phy, XCVII (3), 109–147.
Garcia-Carpintero, M. (2004). The deferred ostension theory of quotation. Nous, 38(4), 674–692.
Garcia-Carpintero, M. (2005). Double-duty quotation: The deferred ostension account. Belgian
Garcia-Carpintero, M. (2006). Recanati on the semantics/pragmatics distinction. Critica, 38, 35–
68.
Garcia-Carpintero, M. (2007). Fiction-making as an illocutionary act. Journal of Aesthetics and
Art Criticism, 65, 203–216.
Garcia-Carpintero, M. (2012). Minimalism about quotation. Philosophical Studies, 161, 207–225.
Garcia-Carpintero, M. (2013). Emma Borg’s pursuing meaning. Mind, 122, 515–521.
Garcia-Carpintero, M. (forthcoming). Indirect assertion. Polish Journal of Philosophy.
Gomez-Torrente, M. (2011). Double-duty quotation, conventional implicatures and what is said.
In E. Brendel, J. Meibauer, & M. Steinbach (Eds.), Understanding quotation, Mouton Series in
Pragmatics (Vol. 7, pp. 139–160). Berlin/New York: Mouton de Gruyter.
Gomez-Torrente, M. (2013). How quotations refer. Journal of Philosophy, 110, 353–390.
Green, M. (2015). Speech acts. In Edward N. Zalta (Ed.), The Stanford encyclopedia of philosophy
(Summer 2015 ed.). URL: http://plato.stanford.edu/archives/sum2015/entries/speech-acts/
Heck, R. G., Jr. (2014). Semantics and context-dependence: Towards a Strawsonian account. In A.
Burgess & B. Sherman (Eds.), Metasemantics: New essays on the foundations of meaning (pp.
327–364). Oxford: Oxford University Press.
Innovation programme under Grant Agreement no. 675415, Diaphora. A version of this material
was presented at the Semantics and Philosophy in Europe 4 conference, Bochum, and at a LOGOS
seminar. I am thankful to the audiences for helpful remarks. I am also very grateful to André
Bazzoni, Philippe De Brabanter and Mario Gomez-Torrente for comments on a previous version.
Thanks to Michael Maudsley for the grammatical revision.
Kaplan, D. (1978). Dthat. In P. Cole (Ed.), Syntax and semantics (pp. 221–243). New York:
Academic.
Kaplan (pp. 481–563). Oxford: Oxford University Press.
King, J. (2014). Speaker intentions in context. Nous, 48(2), 219–237.
Kripke, S. (1977). Speaker’s reference and semantic reference. In P. French, T. Uehling, & H.
Wettstein (Eds.), Contemporary perspectives in the philosophy of language (pp. 255–276).
Minneapolis: University of Minnesota Press.
Lepore, E., & Stone, M. (2015). Imagination and convention. Oxford: Oxford University Press.
Lewis, D. (1980). Index, context, and content. In S. Kanger & S. Öhman (Eds.), Philosophy and
grammar. Dordrecht: Reidel.
Lipton, P. (2004). Inference to the best explanation. London: Routledge.
The semantics and pragmatics of quotation (pp. 99–134). Dordrecht: Springer.
Maier, E. (2014). Pure quotation. Philosophy Compass, 9(15), 615–630.
Michaelson, E. (2016). The lying test. Mind and Language, 31(4), 470–499.
Mount, A. (2008). Intentions, gestures, and salience in ordinary and deferred demonstrative
reference. Mind and Language, 23, 145–164.
Ninan, D. (2012). Propositions, semantic values, and rigidity. Philosophical Studies, 158(3),
401–413.
Nunberg, G. (1993). Indexicals and deixis. Linguistics and Philosophy, 16, 1–43.
Pagin, P., & Westerstahl, D. (2010). Pure quotation and general compositionality. Linguistics and
Pelczar, M., & Rainsbury, J. (1998). The indexical character of names. Synthese, 114, 293–317.
Perry, J. (2009). Directing intentions. In J. Almog & P. Leonardi (Eds.), The philosophy of David
Potts, C. (2005). The logic of conventional implicatures. Oxford: Oxford University Press.
Predelli, S. (2008). The demonstrative theory of quotation. Linguistics and Philosophy, 31,
555–572.
Quine, W. V. O. (1940). Mathematical logic. Cambridge, MA: Harvard University Press.
Rabern, B. (2012). Against the identification of assertoric content with compositional value.
Synthese, 189(1), 75–96.
Rami, D. (2015). The multiple uses of proper names. Erkenntnis, 80, 405–432.
Richard, M. (1986). Quotation, grammar, and opacity. Linguistics and Philosophy, 9, 383–403.
Richard, M. (2008). Opacity. In E. Lepore & B. C. Smith (Eds.), The Oxford handbook of
philosophy of language (pp. 667–688). Oxford: Oxford University Press.
Sainsbury, M. (2015). The same name. Erkenntnis, 80, 195–214.
Saka, P. (2005). Quotational constructions. In P. De Brabanter (Ed.), Hybrid quotations. Belgian
Understanding quotation (pp. 303–322). Berlin: Mouton de Gruyter.
Sinhababu, N. (2013). The desire-belief account of intention explains everything. Nous, 47,
680–696.
Soames, S. (2002). Beyond rigidity. Oxford: Oxford University Press.
Speaks, J. (2016a). A puzzle about demonstratives and semantic competence. Philosophical
Studies. https://doi.org/10.1007/s11098-016-0704-5.
Speaks, J. (2016b). The role of speaker and hearer in the character of demonstratives. Mind, 125,
301–339.
Washington, C. (1992). The identity theory of quotation. Journal of Philosophy, LXXXIX ,

582–605.
Wetzel, L. (2009). Types and tokens. Cambridge, MA: MIT Press.
Yalcin, S. (2014). Semantics and metasemantics in the context of generative grammar. In A.
Burgess & B. Sherman (Eds.), Metasemantics: New essays on the foundations of meaning (pp.
17–54). Oxford: Oxford University Press.
Quotation in Dialogue
Eleni Gregoromichelaki
Abstract Quotation is ubiquitous in natural language (NL). Recent grammars that

take a dialogical view on the formal and semantic properties of NLs (Ginzburg,
The interactive stance: meaning for conversation. Oxford University Press, Oxford,
2012; Gregoromichelaki et al. Dialog Discourse 2(1):199–233, 2011; Eshghi et al.
Feedback in conversation as incremental semantic update. In Proceedings of the
11th International Conference on Computational Semantics (IWCS 2015), Queen
Mary University of London, UK April 2015, 261–271, 2015) indicate that quotation
mechanisms need to be integrated within the purview of standard grammatical
frameworks since such mechanisms are crucially involved in metacommunicative
conversational interaction. Accordingly, the account presented in Ginzburg and
Cooper (J Logic Lang Inf 23(3):287–311, 2014, G&C) provides syntactic analyses,
denotations, and pragmatic constraints for quotational constructions that make use
of grammatical entities independently needed for the analysis of conversation.
However, despite the great advances achieved by G&C, the construction-based
grammar employed lacks essential integration of the psycholinguistically grounded
observation that NL use relies crucially on incremental/predictive processing with
context integration at each word-by-word processing stage. For this reason, certain
data showing the grammatical continuum underpinning various quotational con-
structions as well as interactions between quotation mechanisms and conversational
phenomena (split-utterances, Gregoromichelaki et al. Dialog Discourse 2(1):199–
233, 2011) are not amenable to G&C’s discrete constructional approach. Based on
this inadequacy of even such a state-of-the-art, comprehensive model, this chapter
argues that a satisfactory account of the function of quotational devices cannot be
given within standard NL theories involving the division of labour between syntax
and semantics/pragmatics. Instead, it adopts a dynamic, incremental perspective
that takes joint action as the basis for the definition of the grammar as advocated
E. Gregoromichelaki ()
Heinrich Heine Universität Düsseldorf, Düsseldorf, Germany
King’s College London, London, UK
e-mail: elenigregor@gmail.com

https://doi.org/10.1007/978-3-319-68747-6_8
196 E. Gregoromichelaki
within Dynamic Syntax (DS, Kempson et al. Dynamic syntax: the flow of language
understanding. Blackwell, Oxford, 2001) updated with the integration of some of
G&C’s proposed formal constructs (DS-TTR, Purver et al. Splitting the I’s and
crossing the you’s: Context, speech acts and grammar. In Proceedings of SemDial
2010 (PozDial), Poznan, Poland, 2010; Eshghi et al. Feedback in conversation as
incremental semantic update. In Proceedings of the 11th International Conference
on Computational Semantics (IWCS 2015), Queen Mary University of London, UK,
April 2015, 261–271, 2015).
1 Introduction
It has long been noted that quotation is ubiquitous in natural language (NL), either
obliquely in the form of dialogism or heteroglossia (Bakhtin 1981; Morris 1997),
or directly with more or less explicit indications in conversation and written texts.
It is puzzling then that both NL formal grammars and philosophical accounts (e.g.
Davidson 1979) have assumed that quotation constitutes some sort of extraordinary
or abnormal use, or that elements between (possibly implicit) quotation marks need
not be generated by standard syntax or addressed by the semantics. Since Partee
(1973), various counterexamples to this claim have been presented, for example
cases where the formal or semantic properties of the quotation are needed to
construct antecedents for anaphora/ellipsis:
(1) We had the “crap sticks”, according to the translation on the menu, but they were
actually delicious.
(2) She said “I’ve hiked in Grete” and I confirmed she did. Of course it’s called Crete.
(3) George says Tony is his “bestest friend” and indeed he is. (adapted from Geurts and
Maier 2005)
(4) “I talk better English than the both of youse!” shouted Charles, thereby convincing me
that he didn’t (Partee 1973)
Other counterevidence too (examined later) favours the conclusion that quotation
has to be included in the syntactic analysis and grammaticality definitions of any
grammar (see also Postal 2004).
Recent grammars that take a dialogical view of the formal and semantic
properties of NLs (Ginzburg 2012, Gregoromichelaki et al. 2011, Eshghi et al.
2015) provide additional arguments for integrating quotation within the purview of
standard grammatical mechanisms. Quotation mechanisms are crucially involved
in cases of metacommunicative exchanges in dialogue. One such case is the
phenomenon of echo questions, for example clarification requests, whose content
includes mentioning some previously uttered token (Noh 1998, Ginzburg 2012,
Eshghi et al. 2015):
(5) A: Did Bo leave?
B: Bo? (‘Who are you referring to as ‘Bo’?; Did you utter the word “Bo”’)
Quotation in Dialogue 197
Currently a number of accounts have been proposed regarding the grammatical

integration of quotational expressions. However, such accounts introduce ad hoc
entities and rules in order to shift the usual contents attributed to NL elements
and otherwise ignore the interaction of the whole grammatical apparatus, syntax,
phonology, semantics, and, sometimes, even pragmatics in explaining the totality
of the phenomenon. In contrast, the account of quotation presented in Ginzburg
and Cooper (2014)—henceforth G&C—provides syntactic analyses, denotations,
and pragmatic constraints for quotational constructions that utilise independently
needed and antecedently established grammatical entities. In this chapter, the G&C
insights are adopted to a large extent but within a distinct grammatical framework
that combines Dynamic Syntax (e.g. Kempson et al. 2001; Cann et al. 2005) with
Type Theory with Records (e.g. Cooper 2012; Ginzburg 2012). This combination,
DS-TTR (Purver et al. 2010; Gregoromichelaki 2013b; Eshghi et al. 2015; Hough
2015), takes into account the fact that NL use crucially displays incrementality
and predictivity in its processing with contextual integration at each word-by-word
processing stage.1 In DS-TTR, the incrementality and predictivity of linguistic
processing are considered essential enough to be included as part of the design of the
grammar formalism. As a consequence, certain quotation data that are not amenable
to appropriate treatment within the G&C model become naturally accounted for
under the DS-TTR modelling. For example, quotation can appear subsententially,
and discontinuously at any point, which means that contextual parameters regarding
the discourse situation and semantic evaluation variables need to be able to shift
incrementally2 :
(6) “Cities,” he said, “are a very high priority .” [FrameNet]
(7) Wright won’t disclose how much the Nike deal is worth, saying only that “they treat me
well”. (De Brabanter 2010a, from The Face, September 93: 55)
(8) A doctor tells him [Gustave Flaubert] he is like a “vieille femme hystérique ”; he agrees.
(De Brabanter 2010a, from TLS Online, 18 December 1998)
(9) Alice said that life is “difficult to understand ”. (Cappelen and Lepore 1997)
(10) Mary felt relieved. If Peter came tomorrow , she would be saved. (free indirect speech;
Recanati 2000)
Additionally, quotation is one of the environments where the phenomenon of split-
utterances (Gregoromichelaki et al. 2011) is frequently observed due to the assumed
projectibility of the upcoming continuation (Lerner 1991) and, in my view, the
1
Incrementality here refers to the psycholinguistic notion of a stimulus’ multi-level analysis
piece-by-piece as it is temporally encountered (see e.g. Marslen-Wilson 1973; Steedman 1992;
Tanenhaus et al. 1995; Chater et al. 1995).
2
Elements in the data that constitute the main focus of a claim appear shaded in the examples. Bold
font indicates additional highlighting.
opportunity for co-constructing a vivid unified perspective on some other (actual

or imaginary) speech/thought event:
(11) Clinician: So I watch : : : this person being killed and then I go to bed and I’m you
know lying there going, “well.”
Patient: “Did I hear something?” (Duff et al. 2007)
(12) Patient: And it’s like well : : : you know “I just-”
Clinician: “I don’t want to .” (Duff et al. 2007)
(13) Joan: she wz wise she’d pick up the phone en say, : : :
Linda: I’m comin over , (Lerner 1996)
(14) Adam: Well. I can tell you what her view on that is. and that
Sherm: what.
Adam: is, : : : h : : : I’m older, and therefore I’m in a worse competitive
position, and I and I’ve really got to produce .
Sherm: but I’m smarter [LAUGHS] yeah.
Adam: and I’m going to .
Sherm: yeah. (Grimshaw 1987)
Such shared utterances are not amenable to G&C’s standard sentential grammar.
Unlike what is licensed in the G&C model, the contextual parameters relevant to
such cases need to shift mid-sentence, before whole propositional contents have
been derived. Moreover, such role-switches include cases where the same structure
can be used both as expressing a speaker’s own voice and as a subsequent quotation,
as the following show:
(15) Jem: Mary, whatever it is you think you know you mustn’t speak of it.
Not if you want to stay safe .
Mary: says the horse-thief [BBC Transcripts, Jamaica Inn, Episode 1]
(16) Lucy: Me sister’s alive! That girl they found in the field, it wasn’t Catherine. She’s
written to me. She’s coming to visit.
Patience: Oh, God, now I’M pissing myself! What? You don’t think it was written by a
ghost, do you? Or someone pretending to be your sister?
Miriam: That is the nastiest, dirtiest thing anyone has ever done
Patience: says Black Peter’s strumpet! What are you crying for? [BBC Transcripts,
Jamaica Inn, Episode 1]
(17) A: SOMEONE is keen
B: says the man who slept here all night [BBC Transcripts, The A-Word]
In all such cases, issues of “footing” (Goffman 1979), namely changes in perspec-
tives and roles assumed by interlocutors, intersect with syntactic/semantic issues of
direct/indirect speech forms and speech-act responsibility (Gregoromichelaki and
Kempson 2016; Kempson et al. 2017; Goodwin 2007) in ways challenging for
orthodox grammatical frameworks like those of G&C, Maier (2014a, b), and Potts
(2007).
For these reasons, this paper argues that an adequate account of the function
of quotational devices cannot be given within standard NL theories involving the
encapsulation and division of labour between semantics/pragmatics and syntax.
Instead, it adopts a dynamic, incremental perspective that takes a fine-grained
analysis of joint action among conversational participants as the basis of the

definition of the grammar as advocated within Dynamic Syntax (DS, Kempson
et al. 2001; Cann et al. 2005; Gregoromichelaki 2006, 2013a) augmented with
type-theoretic representations (DS-TRR, Purver et al. 2010; Purver et al. 2011;
Eshghi et al. 2015; Hough 2015; Gregoromichelaki and Kempson 2015). As argued
within the context of DS-TTR, the grammar itself needs to incorporate incremental
interpretation and interaction with the context in order to deal, not only with what
have traditionally been analysed as indexicals, e.g. elements like I and you, but also
speech-act information, roles assigned to participants etc., and, most relevantly here,
subsentential shifts of the context of evaluation.
The structure of this chapter is as follows: First it is argued that the echoing and
metarepresentational abilities that underpin quotation are not peculiar to citation and
reported discourse but also occur in conversational interaction, in particular, in cases
of repair and, in general, during the process of “grounding”, i.e., the signalling of
comprehension, correction, (dis)agreement, or clarification of a previous utterance
(Sect. 2). We will then examine two recent proposals implementing an appropriate
contextual integration of syntax and semantics that will provide ingredients for
the present analysis of quotation: (a) The PTT3 model (Poesio and Traum 1997,
1998; Poesio and Rieser 2010), which implements incremental semantic evaluation
by means of subsentential updates of ‘information states’ (Sect. 3); and (b)
Ginzburg and Cooper’s model (2014), which reformulates both syntactic analyses
and semantics via Type Theory with Records (TTR, Cooper 2005, 2012) (Sect.
4). Exploiting the potential of TTR, G&C attempt to integrate a standard view of
quotation within a dialogue model without ad hoc devices that are not independently
needed in the analysis of conversation (Sect. 5). It will then be argued that G&C’s
constructional approach is not sustainable because the lack of incrementality in the
grammar deprives this account of the resources needed to deal with data of various
intermediate phenomena like free, hybrid, and mixed quotation (Sect. 6). In view
of this, an alternative grammar formalism (DS-TRR) is presented that integrates
some of the ideas of the PTT and G&C accounts but within a distinct incremental
processing architecture (DS) that accounts naturally for the properties of these
intermediate phenomena (Sect. 7).
2 Metacommunicative Interaction as the Basis for Quotation
The human ability to employ quotational mechanisms has been proven pernicious
for the formal languages of logic which isolate the descriptive function of NLs. For
this reason, quotational devices are presented as exceptional by philosophers and
logicians, even though they constitute a basic aspect of the ubiquitous potential for
reflexivity in human behaviour and thought. Since NLs, and other communicative
3
PTT is not an acronym but somehow composed out of the names of its inventors.
means, reflect this cognitive potential directly, they cannot be adequately analysed
in abstraction from such “devices of quotation”. This becomes apparent when
one examines the mundane metacommunicative mechanisms required to underpin
coordination during conversational interaction. First, there are various frequently
used reflexively interpreted elements occurring in all NLs, e.g. indexical pronouns
and tenses; such elements require reference to the parameters of the utterance
event itself (e.g. interlocutor roles, event time) for their interpretation (see e.g.
Recanati 2010). Second, achieving coordination in conversation crucially relies
on the ability to express and to perceive certain (aspects of) linguistic signals
as pertaining reflexively to the communication process under way. This is not a
sophisticated human ability. Even prelinguistic children and non-human animals can
engage in activities which involve some kind of play, e.g. pretending to fight instead
of fighting (see e.g. Bateson 1987), demonstrating a dance, play-acting etc. Within
such framings, the actions that occur carry the implicit meta-message that their usual
significance (e.g. hostility, aggression) is suspended and, instead, some other kind
of significance needs to be sought. Similarly, during the mundane linguistic activity
that occurs in everyday conversation, many speech act(ion)s employ linguistic
elements that can function in various ways, instead of being exclusively confined
in their supposedly basic referential function. For example, repetition of a phrase
can be construed as echoing some just used utterance token, for example in cases of
clarification questions (see e.g. Ginzburg 2012):
(18) A: Who came?
B: Who came ? How dare you? (‘Are you asking “who came”? How dare you?’)
(19) A: Did Bo leave?
B: Bo? (‘Who are you referring to by your use of the name ‘Bo’?’; ‘Did you utter the
word Bo?’)
In addition, further phenomena, like correction and related speech amendments
(other- or self -repair), also require similar “echoic” mechanisms. On such occa-
sions, a new token is produced that has to be recognised as similar to an antecedent
one. Subsequently, another token might be produced which has to be recognised as
offering a replacement for that antecedent token:
(20) A: Bo, (not Bo,)(I mean) Joe , left.

(21) A: Bo, Bo Jones, left.
(22) A: Bo left. (Not Bo,)(I mean) Joe .
(23) A: Bo
B: (Not Bo)(You mean) Joe .
A: Yes. He left
(24) A: Bo left.
B: (Not Bo) (You mean) Joe .
A: Yes.
Besides these repair mechanisms affecting the standard interpretation of linguis-
tic tokens, conversation naturally offers an environment where the construction,
interpretation and authorship of utterances is spread across interlocutors (split
utterances, Gregoromichelaki et al. 2011). These are illustrated in turns 3, 4, 5,

12, 14, 21 below (square brackets enclose overlapping speech).
(25)
1. A: Instead of having <name hidden> <unclear> they had to come through the
Dock Commission all of the men, they wanted so and so men for that boat, they
used to come through to me.
2. B: Before that though, <name hidden> and <name hidden> [<unclear> had their
own men ]
3. A: [Had their own men
4. B: unload the boats ?
5. A: unload the boats , yes. They <unclear>
6. B: They were employed directly by
7. A: That’s right but they all came
8. B: <name hidden>?
9. A: They used to work say one week and have about a month off or go on the dole
for a month.
10. B: So then what happened was, did the Dock Commission say you can’t have your
own men anymore?
11. A: That’s right they had to go on a rota.
12. B: Run by the Dock Commission ?
13. A: Run by the Dock Commission. See the dockers then all got together and they
said right so many men for that job, so many for that job and that didn’t matter who
they were, they had to <unclear> their job, all the way round the dock.
14. B: Whether they wanted to go on that job or not ?
15. A: Whether they want to go or not, they take their turn and the employer had to pay
a percentage into the pool what those men earned, so when those men hadn’t work
at all they drew their money from the National Dock Labour Board.
16. B: Is this where the National Dock Labour Board came into existence?
17. A: That’s how how they come into existence, yes <name hidden> he was a man
what introduced that.
18. B: When was this?
19. A: Oh that’s er, I would say about nineteen forty roughly [CLEARS THROAT] I’d
say about nineteen forty that came in, might have been before that.
20. B: Before that then if they were ill
21. A: They get nothing .
22. B: Could they not get any welfare benefit?
23. A: No [British National Corpus H5H: 89–113]
In many cases of split utterances, as in quotation, the current speaker can be
seen as the animator (utterer) but not necessarily the author or principal, i.e.,
the person bearing the responsibility for the speech act(s) performed (Goffman
1979; McCawley 1999; Antaki et al. 1996). For example, in (25)-4 and (25)-12
above, the continuations are offered by interlocutor B accompanied by a request for
confirmation towards A as to whether they reflect A’s view of the situation, i.e.,
whether they provide contents that the actual principal, A, deems as appropriate or
whether they are an appropriate “echoing” of A’s authorship, i.e., what A was going
to say.
Reported speech is one of the environments where the phenomenon of split-

utterances is frequently observed due to the assumed projectibility (high predictabil-
ity) of the upcoming continuation (Lerner 1991). Additionally, I would argue, such
environments also offer the potential for construction of a jointly derived evaluation
of somebody’s behaviour/thought, a jointly derived plan of action, or, in general, an
affiliative opportunity—factors which, in my view, override the normativity of the
usual turn-taking strictures (see also (11)–(14) earlier):
(26) Ken: she’ll say wouldja-

Louise: wanna glassa milk ? hehhh
Ken: No. wouldju like a little bitta he’ing?
Louise: heh ha ha
Ken: wouldja like some crekles ?
Louise: ehh ha ha ha ha
Ken: wouldja like a peanut butter an’ jelly sandwich ? (Lerner 1991)
(27) Roger: they rationalized it. they say heh heh heh
Louise: it wasn’t there it was all in his imagination . (Lerner 1991)
(28) A: mid April. we had reached the point of thinking that we weren’t going to be able to
reach a policy decision
B: that’s right
A: and so we must. Tell these guys [that we’ll carry on .. ]
B: [we’re going to carry on. yep] (Antaki et al. 1996)
(29) Anne: I wish that he’d say- he said, “I have to be back around four because our family is
having something,” and I wish he’d say
Kay: “why don’t you come over honey ”
Anne: Yeah. (Lerner 1991)
As can be seen from these examples, it is not the case that even clear cases
of “reported speech” ((28)–(29)) involve the reproduction (“echoing”) of actual
utterances/thoughts rather than imaginary or future ones. Instead they function as
coordinating devices, e.g. in planning future action.4
These types of joint action pattern syntactically either with direct or indirect
reports depending on the attribution of authorship to the utterance and correlating
with the function it performs. It has been established that indexicals in split utter-
ances are assigned reference according to parameters of context (e.g. interlocutor
roles) that shift incrementally during the unfolding of the action and utterance
(Gregoromichelaki et al. 2011); see e.g. (30) which is a simple continuation of
A’s utterance with no verbal echoing intent, even though confirmation might be
simultaneously requested as to whether this is an appropriate representation of the
semantic content of A’s query:
(30) A: Oh, I am so sorry, did you burn

B: myself ? No, it’s OK / #yourself? No, it’s OK
4
For quotation of thoughts, see Maier (2017).
On the other hand, indexical shifts can reflect subtle distinctions in the function
of the current utterance, for example confirmation of various suggestions about
aspects of the speech act (in (30), which parallels indirect reports), or clarification
of reference (paralleling indirect reports in (31) and direct reports in (32)):
(31) A: Did you leave?

B: Me ? [‘Are you asking about ‘B, the current speaker’?’]
(32) A: Did you leave?
B: You ? [‘Who do you mean “you”?’]
Given that all these phenomena—continuations, clarification, and, in general,
repair exchanges—are universal and present from very early on in language
acquisition, the means and skills involved in the production and comprehension of
reported discourse do not appear idiosyncratic or ad hoc. Taking this assumption
seriously, two recent holistic models of NL use, HPSG-TTR (Ginzburg and Cooper
2014) and Dynamic Syntax-TTR, seek to model reported discourse via the same
mechanisms as those used to analyse such everyday conversational phenomena as
those just cited. We turn to these two models next.
3 The Formalisation of Metapragmatic Awareness:

Information States and Utterance Events
Recent efforts in formal semantics, inspired by work in Situation Semantics and

Discourse Representation Theory (DRT), have shifted attention away from a strict
formulation of a truth theory for sentences in order to develop theories of semantic
interpretation for utterances in context. For this purpose, representational systems
allowing the specification and seamless integration of multiple types of information
have been sought. One strand of this development, based on recent advances in
developing compositional forms of DRT, is the PTT model (Poesio and Traum
1997, 1998; Poesio and Rieser 2010), which expands dynamic semantics to take into
account NL use in interaction. One distinctive feature of Poesio and Rieser (2010)
is the assumption—derived from ideas developed in Situation Semantics (Barwise
and Perry 1983) and psychology (Clark 1996)—that semantic interpretation of utter-
ances relies on the participant’s information state, i.e., an evolving representation
of the context for each participant in a conversation. Such representations also
include the reification and explicit representation of the utterance event/situation,
i.e., the situation providing for the instantiation of the contextual parameters of
the conversation itself in order to account for the reflexive reference of indexical
elements (see also Maier 2017, this volume).
Even more innovatively compared to previous versions of DRT, in this account,
the occurrence of utterances of subsentential constituents is recorded (as micro-
conversational events) in a certain temporal order so that the gradual accumulation
of utterance micro-events becomes part of the structure of the information state. The
occurrence of each such micro-conversational event leads to immediate updates of
the participants’ information states with the initiation of semantic and pragmatic
interpretation processes, thus implementing incrementality (see also Larsson and

Traum 2000; Stone 2004). As regards the characterisation of particular dialogue
actions, in this model speech acts are conceptualised as events too, termed as conver-
sational events. This is based on the fact that interlocutors can make metapragmatic
statements employing such events as the antecedents of anaphoric expressions:
(33) A: You’re an idiot.
B: That was uncalled for. [that D A’s insulting B]
4 Utterance Events in Type Theory with Records
Another recent articulation of the effort to develop theories of semantic interpreta-

tion for utterances in context (albeit excluding the incrementality dimension),5 but
with richer semantic structures, has been via the development of Type Theory with
Records (TTR). TTR provides a general semantic representation format that can
integrate both low-level (sub-symbolic) perceptual information (see e.g. Larsson
2011) and high-level conceptual inference enriching the underspecified, flexible
meanings of NL expressions (see e.g. Cooper 2005, 2012). Such integration allows
the modelling of how NL forms and meanings adapt to the discourse situation via
the formalisation of an evolving, structured notion of the (multi-modal) context.
Consequently, instead of adopting the assumption that the role of semantic theories
is to assign truth conditions to decontextualised sentences, in these approaches
attention has shifted to the modelling of situated utterances and speech acts. This
has led to a significant expansion of the data deemed appropriate for inclusion in
a formal theory of interpretation, namely the modelling of NL use in interaction
and the demands this places on appropriate semantic models (see e.g. Ginzburg and
Cooper 2004; Ginzburg 2012).
TTR as a representation format provides recursive data structures reminiscent
both of the feature structures in Head-Driven Phrase Structure Grammar (HPSG,
Sag et al. 2003) and, semantically, of discourse representation structures (DRSs).
Records, like r below, are structured collections of ‘fields’ consisting of assignments
of entities as values to ‘labels’, the equivalent of discourse referents/variables in
DRT. Each line inside of r below is a field, and x, time, place, and sit are labels.
(34) A record representation:
x = John
time = 12AM_13_Oct_2012
labels r =
place = London
a field
sit = s1
Such records can then be taken as the representation of events/situations in the

world. Importantly, contexts and speech events can be represented as such records
5
At least not until recently; for some preliminary moves in this direction see Ginzburg et al. (2014).
in order to provide for the instantiation of various contextual parameters as imposed

by linguistic forms.
Records (and therefore effectively events/situations) are classified by types which
are called record types. Unlike the basic Montagovian types, record types (like
records) are structured and recursive (i.e. [record] types can be embedded as the
value of a label within a record type). Additionally, dependencies can occur among
the values assigned to the labels: in many cases the vertical ordering of the fields
reflects such dependencies in that if we first introduce a typed entity it can then
yield a context for the subsequent introduction of a new type dependent on it. A
record r belongs to a type T iff each field in r satisfies the constraints specified by T.
For example, as a simplified illustration, the record r in (34) is of the type T in (35)
below (it is a witness for T) because r assigns entities to the labels that satisfy the
type requirements specified by T. This means that the label x is assigned an entity of
type IND(ividual), namely John; the labels place and time are assigned entities that
are places and times respectively; and the event s1 is such that it contains evidence
that John runs—perhaps s1 is some actual event of John’s running6 :
2 3
x W IND
6 time W TIME 7
(35) A record type representation: T D 6 7
4 place W PLACE 5
sit W RUN .JOHN/
Types, which are abstract modelling constructs, can thus be conceived as categorisa-
tions of entities and events/situations that provide the interface between the external
world and cognition; for example record types, namely categorisations of situations,
can be used as structured representations of perceptual judgements, meaning
relations, grammatical information, speech act assignments, etc. In addition, in TTR,
types are first-class citizens of the semantic ontology, not reducible to sets or their
members. So types are intensional and inference can be performed at the level of
types, irreducibly about the types themselves, solving puzzles that traditionally
have been encountered in intensional constructions such as the complements of
propositional-attitude (Cooper 2005) and reporting verbs. Moreover, types are
modifiable by manipulating their structure, e.g. adding or deleting fields/values.
For this reason, the underspecification, enrichment, and general innovation that
permeates type judgements are naturally handled, as during language acquisition,
knowledge adjustment, conversational coordination and, more pertinently, as we
will see now, quoting (some aspects of) another’s speech or predicating properties
of (aspects of) a grammatical sign.
6
This characterisation is related to Martin-Loef’s “propositions as types” implementation. Hence
run(john) is a type of events and it correctly classifies events that constitute “proofs” of John’s
running (see Ranta 1994: chapter 2).
5 Ginzburg and Cooper (2014): TTR-Modelling of Quotation

Constructions
The account presented in Ginzburg and Cooper (2014) provides syntactic analy-
ses, denotations and pragmatic constraints for reporting constructions that utilise
independently needed syntactic/semantic entities. G&C aim to demonstrate that a
dialogical perspective on NL structure and use directly provides the tools to deal
with reported discourse via structures and denotations that are already independently
motivated for the modelling of dialogue phenomena.
5.1 Grounding and Clarification
Following the model most comprehensively detailed in Ginzburg (2012), the

analysis of dialogue involves defining the appropriate updates of richly structured
representations of context (information states) formalised through TTR. Adoption
of the TTR formalism as a uniform representational format allows Ginzburg to
model the interactions of the distinct components of the model: the semantic
ontology, the grammar, and a system of context updates underpinning the modelling
of the interlocutors’ common ground.
To be able to include in the model of contextual updates the metacommunicative
function of certain utterances, for example clarification requests (5) and (18),
dialogue processing is assumed to rely on a process of grounding (Clark 1996).
Grounding is a necessary dialogue phase during which each participant either
confirms that they have understood and agreed with the utterance addressed to them,
thus incorporating it in their information state (i.e. their own version of the “common
ground”), or they seek clarification of aspects that have not been grounded. Ginzburg
formally elaborates the grounding requirement along two dimensions. First, ground-
ing is not monolithic and immediate; instead, it allows partially comprehended
utterances to contribute to the context while ungrounded (parts of) utterances can
remain as “pending” and lead to metacommunicative interaction (clarification)
for their resolution. Second, it is not only semantic content that is recorded and
manipulated in the participants’ information state (context), but also a range of
properties of the utterance that has occurred, e.g. syntactic/phonological information
that would enable the disambiguation and resolution of elliptical utterances that
function metacommunicatively (see (19)–(24) earlier).
Ellipsis resolution for such fragments requires that the grammar be able to
express reference to utterance tokens, conceptualised here as “utterance events”,
that can be grammatically characterised along multiple dimensions. Representation
of utterance events as records (as in (34)) is employed to serve this role. (Partial)
grounding is then formalised through the pairing of an utterance event (a record, a
token) with an utterance type, i.e., a (partially defined) grammatical type (a sign7 )
that classifies it. Such signs, i.e. grammatical types recording multiple grammatical
properties as defined in HPSG, are expressed in the grammar as record types
(as in (35)).
So here a major advantage of the use of TTR becomes evident: the grammar
and the conversational update mechanisms are provided with access to both types
(record types) and tokens (records) of utterances at the object level so they can both
be employed for defining the syntactic and semantic updates that need to be effected.
This forms the basis for modelling the metacommunicative or metalinguistic func-
tions of NL elements. For example, it is argued that the clarification request in (36)
below, which on the surface simply echoes A’s use of Bo, can be enriched in various
(rule-governed) ways, i.e. it requires disambiguation that can be defined through the
formalism. So, to enable the eventual grounding of A’s query, the grammar needs to
specify a reading for the fragment Bo? which queries which individual named Bo A
was referring to in the previous utterance (‘intended content reading’). This needs
to be distinguished from a separate reading regarding what Bo refers to in general—
which is also a possible reading, as can be seen more clearly in the ‘intended content
reading’ of the predicate finagle in (36). These readings also need to be distinguished
from other readings such as the confirmation readings in (c):
(36) a. A: Did Bo finagle a raise?

B: (i) Bo? / (ii) finagle ?
b. Intended content readings:
(i) ‘Who is (the) “Bo” (you’re referring to)?’ / (ii) ‘What does
it mean “to finagle”?’
c. Clausal confirmation readings:
(i) ‘Are you asking if BO (of all people) finagled a raise?’
(ii) ‘Bo FINAGLED a raise (of all actions)?’
If the grammar and the model of the participants’ information states allow for
reference to actual token utterance events, it becomes possible to explicitly model
readings such as that in (36)b(i) by assigning interpretations to the fragment
Bo that match the intuitive paraphrase given involving reference to the specific
utterance event that has occurred, namely A’s uttering Bo. Additionally, reference
to grammatical types allows for the metalinguistic flavour of clarification queries
regarding the meaning of particular utterance tokens as in (36)b(ii).
In order to formally license such constructions and model their disambiguation,
HPSG-TTR incorporates a constructional version of HPSG, also expressed in the
uniform representational framework of TTR. The rich type theory thus included
then allows for the definition of entities modelling both utterance tokens (events as
records) and their characterisation via utterance types (signs, grammatical types)
7
Note that this use of the term ‘sign’ does not coincide with that of Cappelen and Lepore (2007:
chapter 12).
that the grammar and the model of the context can manipulate and reason over. We
are going to see how these are employed in the analysis of quotation and reported
speech next.
5.2 Locutionary Propositions and Abstract Semantic Objects
Two components in the modelling of the process of grounding are relevant for
the analysis of quotation and direct/indirect discourse that concerns us here: (a)
locutionary propositions, and (b) abstract quasi-propositional semantic objects
assigned as contents to sentential units in order to serve as the arguments of speech
act predicates.
5.2.1 Locutionary Propositions
For an utterance in dialogue to be grounded, first it has to be parsed and understood

correctly. In HPSG-TTR, the successful outcome of this process of parsing is
modelled via requiring the truth of a so-called locutionary proposition. Simplifying
somewhat, a locutionary proposition is the pairing of the current utterance-event
token with a fully specified grammatical type (an HPSG-defined sign). Such signs
are structured TTR-types, i.e., representations that include labels characterising
phonology, syntax, semantics and contextual specifications with constraints gov-
erning their correspondence. If the truth of such a locutionary proposition cannot be
established after parsing, i.e., if a complete grammatical type cannot be assigned to
an utterance, various clarifications are licensed to occur that can make reference to
the particular utterance token that causes the trouble.
5.2.2 Abstract Quasi-propositional Objects
In HPSG-TTR, the grammar is assumed to assign to every utterance of a root

sentence a speech-act characterisation. During the grounding process, this enables
reference to be made to the particular speech act performed by the previous
interlocutor, e.g. modelling interpretations like ‘Are you asking q’ and ‘Are you
asserting p?’. The (quasi-)propositional arguments p or q in such speech-act
specifications are abstract semantic objects like propositions, questions, outcomes,
facts, etc. These objects are defined in the semantic ontology and are assigned by the
grammar as the contents derived through the realisation of the speech acts performed
with utterances. For example, root clauses are required by the grammar to include
a speech-act specification, selected from a small number of such specifications, like
Assert, Ask, Order, Exclaim. Which of these speech-act specifications is selected
depends on the semantic object that is compatible with each such specification and
which is attributed to the agent of the speech act. So, a proposition will be what an
agent Asserts, a question will be the complement of the Ask relation, an outcome
the complement of the Order relation, and a fact is the object associated with the
Exclaim relation.
These defined abstract entities, locutionary propositions (i.e. pairings of utterance
events with utterance types), and abstract (quasi-)propositional semantic objects,
having been independently postulated for the explanation of NL use in conversation,
especially metacommunicative interactions, are taken by G&C to naturally extend
to pure, direct, and indirect quotation as we will see now.
5.3 Pure Quotation8
Having assumed a constructional version of HPSG, in extending these conversa-

tional mechanisms to reporting discourse, G&C define constructions for various
quotation phenomena that specify the interacting syntactic, semantic and pragmatic
constraints for their licensing. According to various theories of quotation (identity
theory Washington 1992, proper name theory Tarski 1993; Quine 1940, description
theory Geach 1957), the quotation marks somehow modify the reference of the
expression that is enclosed within them, so that the expression’s usual reference
is suppressed and the expression itself somehow is referred to. However, as G&C
argue, the analysis of pure quotation requires formal explication of the notion
‘expression’ (see also Saka 1998; Cappelen and Lepore 2007). The HPSG-TTR
model offers a specification of this notion via the invocation of a particular
grammar that supplies the TTR-expressed grammatical type characterising such
token expressions (see also Maier 2014a,b; Potts 2007).
So, in analysing pure quotation, G&C introduce the assumption that, in general,
the contextual parameters included in the information state have to include a param-
eter that refers to the particular grammar licensing the type of the expression
used. roughly corresponds to a formalisation of what in Recanati (2010) is
characterised as a shiftable “language”. It is a welcome innovation of the view
that G&C implement that NL grammars are considered as collections of resources
for building context-appropriate ad hoc languages (syntax-semantics specifications)
(Cooper 2012; Cooper and Ranta 2008). therefore will be instantiated in each
particular context as the particular subcollection of grammatical resources relevant
to the licensing of an utterance event under a grammatical type (sign) in that context
(such an instantiation is argued to be a “pre-semantic” process in Recanati 2010;
Maier 2014b).
Under this view of the flexibility of grammatical resources, pure quotation can
be characterised as invoking a particular set of such resources which provide
for its properties as a sign. Pure quotation constructions are usually assumed to
8
The term pure quotation here follows G&C’s usage, namely it refers to citation. For objections to
this usage, see Saka (2013: 939).
occupy an NP/DP position and introduce a referential (singular) term that refers to
the linguistic material enclosed in the quotation marks. Accordingly, G&C formally
define a special grammatical construction for pure quotation. This construction
licenses an NL string, e.g. You love John, to syntactically project a phrase (NP)
whose semantic content is its grammatical type (sign) relative to a particular
grammar . Aspects of this grammatical type, e.g. the phonology, can be picked
up in a case like:
(37) ‘You love John’ starts with a consonant.
The contextual parameters usually assigned by to standard uses of the sign (e.g.
speaker-hearer, time, location etc.) are discarded in such a quotational construction.
This and the fact that the semantic content of the “construction” is now its
grammatical type, instead of its usual compositionally derived reference, explains
the opacity of such uses (e.g. you in (37) does not refer to the current hearer).
According to G&C, use of quotation marks in written discourse indicates this shift
of content for such uses of NL elements.
5.4 Direct and Indirect Quotation
Turning to direct and indirect discourse, G&C analyse these as involving two
components: (a) lexical entries for the reporting verbs (quotative predicates) and
(b) constructions that specify the presumed idiosyncratic properties of each such
structure. Quotative predicates can select for syntactic constituents denoting (a)
locutionary propositions (see Sect. 5.2.1) in direct reports, or (b) quasi-propositional
abstract entities (see Sect. 5.2.2) for indirect reports. Both direct and indirect reports
are analysed as constructions that involve the combination of a reporting verb, like
say or ask, with a clause whose denotation involves such an abstract semantic object.
A direct-quotation construction involves, first, the projection of a direct-
quotational phrase from the quoted material. This phrase can then serve as the
complement of a reporting verb specified to require the type of such a complement.
The derived semantic content of a direct-quotational phrase is a locutionary
proposition, i.e., an utterance event to which a grammatical type, a sign, is assigned
by a particular grammar (see earlier Sect. 5.2.1). The utterance event component
of such a proposition represents the reported utterance event. This reported event is
now associated with the grammatical type assigned to it by the reporter relative to a
grammar thus accounting for the fact that, for example, the quote might be in a
language different from that of the original reported event or other modifications the
reporter might effect (and still be counted as direct quotation).9 Since the grammar
manipulates TTR types as well as tokens, it now becomes possible to express how
the original reported utterance event and the reporting event are deemed to be
9
Recanati speaks of the “language” of the context (2010: 190).
“similar” in some relevant respects (for these, see e.g. Clark and Gerrig 1990).
The grammatical type assigned to the utterance event included in the locutionary
proposition by the reporter’s assumed grammar is constrained to “resemble” the
type of the original event, i.e., there has to exist a contextually defined exceeded
threshold value on a similarity measure between the grammatical types of the
original and the reporting events. Further, via this construction, the contextual
parameters of the standard use of the sign are discarded, as we also saw in pure
quotation cases earlier. However, for direct quotation, at the phrasal level, a new
set of contextual parameters is introduced now relative to the reported event. In
this way, the full content of the reported sign becomes available, unlike cases of
pure quotation (where there is no reported event). This allows for the explanation
of cases of anaphoric reference to the content of the quotation subsequently, as in
the Partee examples (such as (4) earlier) where anaphora/ellipsis resolution relies
on semantic/contextual processing of the quoted material.
A further innovative advantage offered by this analysis is that by analysing direct
quotation complements as denoting locutionary propositions, which include as one
of their components a sign (a grammatical type), we can explain the fact that a single
sentence can contain predications that address both type and token aspects of the
quotation, e.g. use the same quotation as both the complement of a direct-quotation
construction and as a pure quotation, as in (38)–(39):
(38) “Was I snoring” was asked by Bill and is a frequently used interrogative clause.
(39) Bill asked, “Am I snoring?”, a sentence frequently used by men who don’t think they
snore. It is usually answered by “You were before you woke up”. (adapted from G&C)
The direct-quotational phrases whose properties we have just described appear
as the complements of verbs that combine with direct quotations, for example
independent clauses (non-clausal complements for such verbs are also possible).
Many such verbs also take embedded clauses as their complements, resulting in
indirect discourse constructions. This is implemented in this model by defining
such verbs as combining with complements that can have either of two distinct
semantic objects as contents. For example, the lexical entry for ask has two versions.
In the case of direct quotation, the lexical entry for the verb ask specifies that the
complement must have as its content a locutionary proposition, i.e., the combination
of an utterance event with a grammatical type (i.e. a direct-quotational phrase,
as discussed above). As we’ve just seen, due to the direct-quotational phrase
specifications, the utterance event will be the reported event. Additionally, the
lexical entry for the verb ask specifies that the SPEAKER x of the utterance event
included in this locutionary proposition (the original utterer) is identified with the
subject of the main clause. Since the grammar, according to Ginzburg (2012),
conventionally associates speech-act specifications with utterances (see earlier Sect.
5.2), the speech-act characterisation of the original reported event can be available
through the abstract semantic object associated with it. Due to this fact, the content
of the main clause is now identified with the speech-act content introduced in the
grammatical type of the complement of ask so that it comes out as the speech-act
specification Ask(x, q), where q is an abstract semantic object of type question.10
On the other hand, the lexical entry for ask in an indirect-report context specifies
that it combines with a subject x and a sentential complement. Unlike the case
of direct quotation, this sentential complement is not of the type ‘locutionary
proposition’, i.e. the original reported event is not included in the representation;
hence, unlike direct quotation, in indirect report cases, the reported event parameters
cannot affect the current contextual parameters. The only restriction is that the
complement has as its content an abstract semantic object q of type question (e.g.
that derived from whether Bill left). The content of the whole sentence built on the
basis of the lexical entry for ask is then a proposition Ask(x,q) where x is identified
as both the subject x of the main clause and as the agent of the speech act reported
through the use of ask.11
This account is designed to capture the commonalities of direct and indirect
discourse via the lexical entries of verbs that combine with both. As we just saw, the
contents derived for both such structures are identical,12 even though the structure
with the direct-quotational phrase includes reference to the original demonstrated
event. Another commonality this setup is designed to capture is the common
entailments between direct and indirect reports, illustrated by the fact that they both
support common inferences about the characterisation of the semantic object they
combine with. So both (40) and (41) below entail (42), which is explained because,
as we just saw, the contents eventually assigned to the clauses built on the basis of
the two versions of ask are identical:
(40) Zohar asked whether she snored.
(41) Zohar asked ‘naxarti?’.
(42) Zohar asked a question, a question about herself.
So the G&C account successfully captures various properties of reporting
constructions via the attempted identification of the mechanisms of quotation with
mechanisms of repair in conversation. In addition, the TTR modelling proposed is
able to allow for the explanation of new data like the cases of “mixed predications”
in (38)–(39) where a single predication can simultaneously address metalinguistic
and reporting aspects of the same utterance. It also claims to capture the commonal-
ities between indirect and direct reporting and the common and mutual entailments
holding between such structures as seen in (40)–(42) earlier.
From the present point of view, there are some problems with this latter claim,
stemming from the fact that the grammar necessarily associates conventionalised
speech-act specifications with each main clause. For the same reason, in combina-
10
Note that to this a new speech-act specification will eventually be added to the effect that the
final (schematic) content will come out as Assert(Speaker, (Ask(x,q)) for a (schematic) sentence
like ‘Johnx asked q’.
11
The eventual content derived will again be Assert(Speaker, (Ask(x,q)).
12
As a reviewer notes, there are various other subcategorisation possibilities for such verbs which
are not discussed here.
tion with the fact that the grammar is defined in terms of constructions, rather than
general structural constraints, the account does not seem to be able to generalise
to cover all quotational possibilities that have been reported in the literature. The
main technical and conceptual reasons for this are, first, the fact that the intrinsic
incrementality of NL-processing is not part of the grammar, and, secondly, the fact
that syntax is taken as an independent level of analysis with its own categories and
constraints (as is standard for most grammar formalisms).
In order to remedy these shortcomings, in the next sections we will examine these
problems to motivate the claim that we need to introduce an alternative account that
builds on some of G&C’s insights but is formulated within an incremental, dynamic
framework, namely DS-TTR.
6 Free (In)direct Discourse, Mixed Quotation, Hybrid Uses
Recanati (2001) makes a distinction between closed and open quotation. Closed
quotation are instances where the quotation is syntactically employed as an NP (or
DP in some syntactic frameworks) and semantically plays the role of a singular
term. The G&C account is explicitly addressed to such closed quotation cases only.
However, I believe that G&C have provided some of the resources that make a more
inclusive account available, i.e., covering also the phenomenon of open quotation,
where the quoted material is not integrated in a clause. In my view, the only factor
that prevents the integration of such phenomena in the G&C account is the standard
assumption of an independent syntactic level of analysis in the grammar and the
lack of incremental syntactic licensing and interpretation. The same assumptions,
standard in all formal grammatical frameworks, prevent other grammatical accounts
of quotation (e.g. Potts 2007; Maier 2014a) from dealing with the whole range of
data as we will now see.
The G&C constructional account inevitably adheres to the standard strict divi-
sion between direct and indirect quotation. However, this strict distinction can
be disputed as there is a host of phenomena that lie in a continuum between
these two supposed extremes. First, there are languages where there is no such
strict distinction; instead various syntactic features of the utterance can indicate
either more or less syntactic/semantic integration, for example combinations of
complementiser, mood change, or verb-second:
(43) (a) Peter sagte, dass er das nicht machen könne. [subjunctive C complementiser]
Peter said that he that not make can.SUBJUNCTIVE
Peter said that he couldn’t do this
(b) Peter sagte, dass er das nicht machen kann. [indicative C complementiser]
(c) Peter sagte, er könne das nicht machen. [subjunctive C verb-second]
(d) Peter sagte, er kann das nicht machen. [indicative C verb-second]
(German, from Brendel et al. 2011).
In my view, such phenomena show that the grammar needs to provide mechanisms
for processing rather than rigid constructional analyses.
Next, there is always the possibility of introducing quotational elements, for

example elements extraneous to the reporter’s dialect, within a report otherwise
characterisable as indirect (and without the use of [potentially invisible] quotation
marks, contra Maier 2014a):
(44) To which Mr. Bailey modestly replied that he hoped he knowed wot o’clock it wos in gineral
(Clark
. and Gerrig 1990: 791, from Charles Dickens, Martin Chuzzlewit)
Then there is the phenomenon of free direct discourse. In these cases there is
no reporting verb or clause to indicate reporting but indexicals and other devices
conform to the reported context indicating direct quotation:
(45) Hilaryi crept into the back room. Shei saw the curtains, dragged together roughly,
as if – as if – There’s someone behind them. Ii’m sure there’s someone behind
them. Ii must stay calm – Shei reached for the light. (Crystal 2013)
Free indirect discourse Banfield (1973) is similar to indirect reporting in that there
is shift of tenses and indexicals. However, usually, there are no overt reporting
indications, temporal adverbials are evaluated with respect to reported event, and
some features of direct discourse (such as direct questions and vocatives) are
maintained so that there is only a partial shift of perspective towards the reporting
context (see e.g. Eckardt 2014):
(46) Mary felt relieved. If Peter came tomorrow, she would be saved . (Recanati 2000)
(47) Marie was wondering. Did her brother arrive ? (Bonami and Godard 2008)
And there are further “hybrid” cases, for example in English, the interrogative word
order can sometimes be maintained in indirectly reported questions:
(48) The baritone was asked what did he think of Mrs Kearney’s conduct . (Koder and Maier
2014: fn. 1 citing McCloskey 2006, from James Joyce, Dubliners)
These phenomena cannot be handled by the G&C account because their model
requires the grammar to deal with phrasal constructions that specify either direct
or indirect features. In all these cases, however, as in pure quotation, there is
no necessity for a reporting verb to determine the appearance of a quotational
interpretation. Another phenomenon that is excluded for the same reasons is that
of mixed quotation, a combination of direct and indirect discourse, characterised
in written language by the use of quotation marks in the sentential complement of
an indirect-report construction (see e.g. Cappelen and Lepore 1997; De Brabanter
2010a):
(49) Alice said that life is difficult to understand.
In these cases, in common with indirect reporting, the complement of the verb is
a that-clause which is presented as having the same content as what the reported
speaker said. But, as in direct reporting, there seems to be indication that the
reporting speaker, or some other speaker (Recanati 2010), used similar words as
those appearing in quotation marks. Since both these aspects of such reports affect
the truth-conditions of the sentence, they need to be accounted for by an adequate
model of NL use (Recanati 2000; Potts 2007; Geurts and Maier 2005). However,
because such structures in G&C’s account would have to be analysed through
the constructions defined for the reporting verbs, which unify the subject of the
report with the utterer of the quotation, the option of such quotations echoing other
speakers/thoughts is excluded.
Along with the G&C account, all standard syntactic/semantic and pragmatic
models face problems accounting for certain alleged peculiarities of mixed quo-
tation. First of all, like direct quotation and, as we saw earlier in (44), even with
indirect quotation, there is the possibility to shift not only the interpretation of
indexicals but even language in the midst of reporting such utterances:
(50) Wright won’t disclose how much the Nike deal is worth, saying only that ‘they treat
me well’. (De Brabanter 2010a, from The Face, September 93: 55)
(51) A doctor tells him [Gustave Flaubert] he is like a ‘vieille femme hystérique’; he
agrees. (De Brabanter 2010a, from TLS Online, 18 December 1998)
Another issue that arises for formalisms that do not embrace the incrementality
of processing in the grammar but, instead, attempt to characterise and interpret
well-formed sentences, is the fact that the quotation-like interpretation might span
multiple sentences or even within-sentence non-constituents:
(52) She replied, ‘I live alone. My son lives alone too. We both prefer it that way ’. (De
Brabanter 2010a, from Huddleston and Pullum 2002: 1026)
(53) Writing that book, Doyle felt himself “ a slave to reality. I was just dying to write
a big book, and to have a bit of fun ”. (De Brabanter 2010a, from Independent Arts, 17
September 2004)
(54) David said that he had donated “largish sums, to several benign institutions ”. (Abbott
2005)
(55) Mary allowed as how her dog ate “odd things, when left to his own devices ”. (Abbott
2005)
(56) Tim Marlow of London’s White Cube gallery suggested that such self-censorship was
now common, though ‘very few people have explicitly admitted ’ it. (De Brabanter
2010a, from www.guardian.co.uk/commentisfree/2008/oct/01/religion.islam)
(57) [The doctors’] actions defied the instructions of members of Congress, who issued
subpoenas to attempt to block ‘the barbaric ’ removal of her feeding tube on Friday
[ : : : ]. (De Brabanter 2010a, from The Guardian, 20 March 2005)
This cannot be handled by a grammar that requires phrases to be built out of
conventional constituents that just shift interpretation as they project according to
distinct pre-defined constructions. As De Brabanter (2010a) argues, the whole set of
these effects cannot even be handled by the ad-hoc constituency imposed by Maier’s
(2007) account since the continuity and unity of the quoted fragments gets lost. And,
as Recanati (2010) among others points out, such phenomena have truth-conditional
effects, as can be seen from the distinct interpretations obtained when the quotation
marks are removed13 :
(58) Paul says he’s due to present his work in the “paper session”. [Paul calls “paper
session” the ‘poster session’]
(59) Paul says he’s due to present his work in the paper session. (Recanati 2010)
(60) James says that “Quine” wants to speak to us. [James thinks that McPherson is Quine]
(61) James says that Quine wants to speak to us. (Recanati 2010)
(62) Nicola believes that her father is a “philtosopher”.
(63) Nicola believes that her father is a philtosopher. (Cappelen and Lepore 1997)
This clearly indicates that a grammar formalism needs to integrate interaction with
context at a subsentential level, before the semantic contents derived from words
are composed. The cases above have been analysed by Recanati (2010) in terms
of a language-shift. We can implement this, similarly to the G&C account, by
assuming that the contextual parameters that need to be included in a grammatical
analysis must represent various sources for the grammatical resources employed,
for example entities like potential idiolects, dialects, and languages. It seems to
be the case that we need a rather liberal characterisation of such entities since the
processing devices involved (the ‘grammar’ from our point of view) in such uses
are open-ended and are not dependent on any actual folk-linguistic characterisation
as the examples in (58)–(63) show. Cases involving use of quotation marks as
indicating the speaker’s dissociation from some usage of words (‘scare quoting’)
can also be accounted for through such grammar-shifts (see Sect. 7.3.3). The
pragmatic process leading to the (local) shift and the instantiation of such variables
can be conceptualised as described in Saka (2003/2005: Sect. 3.1).14 Other such
phenomena can be treated in addition as potentially ‘echoic’ in the sense that
the contextual parameters will also include a reported utterance- or thought-event
(whether actual or generic, habitual etc.; see also Predelli 2003, Maier 2017). In
my view, all such analyses are available and compatible (contra Saka 2003/2005) if
the grammar provides fine-grained mechanisms rather than static characterisations
of “expression types”. However, for such analyses to be available, it is crucial
that such shifts of contextual parameters be available subsententially during the
interpretation of a fragment of the utterance being processed. Such parameters,
additionally, should neither necessarily project syntactically nor be defined only at
the root level (as in many standard grammatical/semantic frameworks).
In this connection, another related issue that arises for the G&C account is the
fact that the speech-act specification associated with each main clause is taken
13
This does not imply that quotation marks are necessary for such uses. It is just that they constitute
signals that facilitate the alternative interpretations indicated (see also Saka 2003/2005). In a
particular context of use, their presence is not required for such interpretations to be possible.
14
Such a process does not have to be conceived in a Gricean manner as inference driven by the need
to derive the speaker’s intention. Instead it can be implemented mechanistically in the grammar in
the sense, explained later, that the hearer understands the speaker’s actions through mirroring these
actions as specified by her/his own grammar (see Gregoromichelaki et al. 2011).
to be conventionalised, i.e. there is a selection from among a predefined set of

such illocutionary forces (see earlier Sect. 5.2.2). However, what precise speech-
act specification is potentially assigned to each utterance is an open-ended issue and
subject to contextual determination so that there can’t be any default specifications
determined by the grammar (Gregoromichelaki and Kempson 2015); the grammar
just needs to include mechanisms for such optional pragmatic determination to
potentially affect truth-conditional content on the way to deriving contents for the
full utterance. Support for this claim is provided by the fact that indirect report
complements can appear with a multitude of speech-act denoting framing verbs
(and this class of verbs is open-ended):
(64) Replying to another question by the shareholders he characterised as “imaginary

scenario” the possibility of Greece leaving the eurozone, however, he clarified that
“there is no practice or methodology for a country to exit the eurozone.” [Cyprus Mail,
31/05/11]
And the alleged common inferences with direct discourse are equally possible for
such characterisations:
(65) In a reply to publications in the German newspapers, Mario Draghi stated yesterday:
“There is no practice or methodology for a country to exit the eurozone.”
(66) Mario Draghi clarified: “There is no practice or methodology for a
country to exit the eurozone.”
(67) Mario Draghi offered a clarification of his previous statements.
Such alleged “entailments” are not qualitatively different from the ones offered by
G&C in (40)–(42). However, they cannot be explained as arising from a range of
fixed speech-act specifications and special semantic objects defined in the grammar,
which is what provides the explanation of (40)–(42) in the G&C model. If there is a
mechanism for deriving the inferential pattern in (64)–(67) pragmatically, it can also
be used to derive the inferences in (40)–(42) as long as such pragmatically inferred
contents can interact with grammatical specifications at an appropriate level.
On the other hand, the alleged inviolable restrictions implemented for indirect
reporting in the G&C account and others do not hold for mixed quotation, a
construction structurally similar to indirect quotation. So, for example in a mixed
quotation, a first-person indexical need not refer to the speaker performing the
utterance act but, instead, to the subject of the reporting verb (Geurts and Maier
2005; Cumming 2005; Anand and Nevins 2003) as in direct quotation structures:
(68) Bill Watterson said that reality “continues to ruin my life”. (Maier 2014a)
However, wh-extraction is possible out of mixed quotation environments, which

places mixed-quotation on a par with indirect discourse proper and indicates that
quotation marks are not in any way “syntactic opacity” indicators (cf. Schlenker
2011), so that any actual such constraints have to be implemented elsewhere:
(69) Quine remarks that quotations have a “certain anomalous feature” that
“calls for special caution”; (Davidson 1984: 9)
(70) Who did Mary say that she would “never misunderestimate ever again”? (Maier 2014a)
Regarding the lack of syntactic opacity in mixed quotation, Maier (2014a) claims
that certain features of the quoted original in mixed quotation obligatorily have
to be adjusted to fit the new quoting environment. For example, he claims (citing
Shan 2011) that the grammatical gender agreement displayed by a quoted phrase
in gender-determining languages has to be adjusted to fit its new environment.
However, this is not an absolute constraint either but a choice concerning whether
the quotation echoes faithfully the form of an utterance or not. For example, there are
cases like (72) where this alleged restriction does not hold because the incompatibly
gendered characterisation (as shown in (71)) happens to convey exactly Maria’s
words:
(71) *Ta koritsia tis Lenas ine poli psagmenes [Greek]

The girls of Lena are very sophisticatedFEM
NEUT
Lena’s girls are very sophisticated
(72) I Maria ipe oti ta koritsia tis Lenas ine poli “psagmenes ” [Greek]
Mary said that her girls of Lena are very “sophisticatedFEM ”
NEUT
Mary said that Lena’s girls are very sophisticated
In conclusion, these intermediate phenomena—free, hybrid, and mixed
quotation—show that there is no strict distinction between direct and indirect
reporting so that there is no need for distinct phrasal constructions to be defined
for each to account for their alleged distinct properties. Any such formalisation will
prevent the whole range of phenomena from being captured. Instead, as in mundane
conversational interaction, it also has to be assumed for quotation that the grammar
provides fine-grained mechanisms according to which speakers/writers can freely
shift the mode of presentation and perspective of their utterance, indicate who takes
responsibility for its content and form, or draw attention to some of its properties
at any sub- or supra-sentential level. This argues against a model of NL-grammar
that ignores the psycholinguistically established incrementality of processing and
the dynamic nature of context updates. On the other hand, it provides support to the
claim that grammatical semantic/syntactic constraints are not qualitatively different
from pragmatic processing, and, therefore, cannot be segregated in a distinct abstract
static model that provides analyses only for linguistic strings. This is shown most
clearly by the fact that contents provided by NL-utterances can compose with a
variety of demonstrating events, like gestures, noises, or pictorial signs in written
language:
(73) The car engine went [BRMBRM], and we were off. (Clark and Gerrig 1990)
(74) The boy who had scratched her Rolls Royce went [RUDE GESTURE WITH HAND] and
ran away. (Recanati 2010)
(75) Every person who went [DEMONSTRATION OF RUDE GESTURE/BRMBRM] was
arrested. (adapted from Postal 2004)
To capture such phenomena and desiderata as an intrinsic consequence of the

framework, we now turn to a grammar formalism that takes into account the fact
that NL is primarily a form of action, produced and interpreted in context in a
time-linear manner (for similar intuitions in the quotation literature see Saka 1998,
2003/2005). The next sections aim to show that the data mentioned above, which
are highly problematic for other formalisms, find natural explanations from such a
perspective.
7 Dynamic Syntax
In distinguishing between open and closed quotation (see earlier Sect. 6), Recanati
(2010) makes an alleged important distinction: open quotations are primarily
“demonstrations”, involving
the meaning of the speaker’s act of ostensive display. That meaning is pragmatic: it is
the meaning of an act performed by the speaker, rather than the semantic content of an
expression uttered by the speaker. (Recanati 2010: 271, emphasis mine)
Closed quotations in contrast, according to Recanati, carry additional referential

meaning due to their integration in the linguistic system. From that point of
view, this distinction reflects the standard conception of NL-analysis as requiring
a specifically linguistic grammar on the one hand and a separate component of
pragmatic inference, concerned with human action, on the other. In contrast, the
framework adopted here, Dynamic Syntax (DS, Kempson et al. 2001; Cann et
al. 2005; Gregoromichelaki and Kempson 2013, 2015, 2017), presents a more
radical alternative concerning the status of the syntax/semantics components of the
grammar and their integration with pragmatics. Under this conception, syntax is not
a level of representation at all but a set of more-or-less domain-general routinised
mechanisms (packages of actions) for integrating or producing communicable sig-
nals, with the grammar standing in continuity with other categorisation processes of
intentional/non-intentional stimuli. From this point of view, for DS, ‘demonstration’
(whether echoic or not), not ‘reference’, is all there is in linguistic processing
in general so the opposition ‘closed’ vs ‘open’ quotation cannot be adopted
(for a similar but less radical conception regarding syntax see also Saka 1998,
2003, 2003/2005). Nevertheless, Recanati’s insight that ‘closed quotation’, like
other non-linguistic demonstrations (see (73)–(75)), can be recruited as linguistic
constituents can receive natural expression in DS as we are going to see in
Sect. 7.3.1.
DS, instead of conceiving of NL as a code licensing form-meaning cor-
respondences, models the mechanisms of processing, conceived as (epistemic)
act(ion)s interlocutors engage in during the production and comprehension of both
meaning and forms. So all levels of traditional NL analysis are reconceptualised
as actions performed and assigned meaning within a context. In this respect,
DS can be seen as a psycholinguistically inspired formalism that specifies the
‘know-how’ that is employed in linguistic processing, in contrast to standard

formalisms which codify (specifically linguistic) propositional knowledge of rules
and representations. Regarding levels of analysis, DS eschews a string-syntactic
level of constituency as a level of explanation. Instead it implements the assumption
that grammatical constraints are all defined procedurally. Such constraints guide
the progressive development of conceptual representations along with contextual
information (‘information states’), with partial interpretations and strings emerging
step-by-step during social interaction on a more or less word-by-word basis. In the
model adopted here (DS-TTR, Purver et al. 2010), Dynamic Syntax is enriched with
conceptual representations formulated in the Type Theory with Records framework
(TTR, Cooper 2005, 2012; see earlier Sect. 4). TTR is able to integrate information
from perceptual and subsymbolic sources (Larsson 2011, 2015), which captures
directly the fine-grained dynamics of dialogue, its potential for integrating input
from various modalities under a single processing mode, and the potential for
underspecification and enrichment (Purver et al. 2011; Eshghi et al. 2015). Thus
DS-TTR is formulated as a system which crucially involves:
– An action-based architecture (DS) that dynamically models the development of
unitary TTR representations (information states) integrating multiple sources of
contextual information
– Word-by-word incrementality and predictivity within the grammar formalism
– Parser/generator (i.e. speaker/hearer) mirroring and complementarity of process-
ing actions as part of the grammar.
This perspective, when applied to dialogue modelling and quotation devices, sheds
new light on the phenomenon of split utterances seen earlier in (25), taken up
below in Sect. 7.2; how the mechanisms apply there, in combination with some
of the tools provided by the G&C account, allow for modelling the continuity
of mechanisms underpinning pure quotation, direct and indirect discourse, and
mixed/scare quotation (as we will see in Sect. 7.3). Since both dialogue phenomena
and reporting/citation devices are using the same grammatical resources they are
predicted to interact. This is shown with quotation data which receive analysis with
the same means as dialogue phenomena.
7.1 Incrementality/Predictivity and Radical

Context-Dependency in the Grammar
Instead of deriving sentence structures paired with propositional meanings, as in

models of competence, the DS formalism models directly the interlocutors’ perfor-
mance in processing word-by-word NL strings and meanings in interaction with
the non-linguistic context. For NL use in conversation this is a crucial explanatory
factor since many of its metacommunicative features rely on such incremental
production and comprehension. For example, the frequent occurrence of constituent
clarifications (see earlier (5), (19)) in conversation shows that utterances can be
processed and understood partially without having to map a sentential structure

to a full proposition. Moreover, the process of grounding, invoked and modelled
at the propositional level by Ginzburg (2012) (see Sect. 5.1 earlier) relies on the
appropriate positioning of items like inserts, repairs, and hesitation markers, a
positioning which is not arbitrary but systematically interacts with grammatical
categories and derivations at a subsentential level (see e.g. Clark and Fox Tree
2002). During grounding, addressees display their comprehension and assessments
of the speaker’s contribution subsententially, as the utterance unfolds, through
‘back-channel’ contributions like yeah, mhm, etc. (Allen et al. 2001). And speakers
shape and modify their utterance according to such verbal and non-verbal feedback
received from hearers as their turn unfolds (Goodwin 1981). Hence the grammar
must be equipped to deal with such metacommunicative signals in a timely and
integrated manner, namely by incrementally providing online syntactic licensing,
semantic interpretation, and pragmatic integration. In addition, the turn-taking
system (see e.g. Sacks et al. 1974) seems to rely on the grammar, as it is based
on the predictability of (potential) turn endings in order for the next speaker to time
appropriately their (potential) response; in this respect, experimental evidence has
shown that this predictability is grounded mostly on syntactic recognition rather
than prosodic cues, intonation, etc. (De Ruiter et al. 2006).
For all these reasons, the DS-TTR model assumes a tight interlinking of NL per-
ception and action by imposing top-down predictive and goal-directed processing at
all comprehension and production stages so that input and feedback are constantly
anticipated by relying on contextual linguistic and non-linguistic information in
order to implement efficient performance. Concomitantly, coordination among
interlocutors can then be seen not as inferential activity but as the outcome of the fact
that the grammar consists of a set of licensed actions that both speakers and hearers
have to perform in synchrony (Gregoromichelaki and Kempson 2013). These
actions perform step-by-step a mapping from perceivable stimuli (phonological
strings) to conceptual representations or vice-versa. Production uses the testing of
parsing states in order to license the generation of strings while comprehension
invokes prediction of upcoming input in order to constrain efficiently the usual
overwhelming ambiguity of linguistic stimuli.
7.2 Conversational Phenomena in DS-TTR
In DS-TTR, the conceptual contents derived by processing linguistic strings are

represented as trees inhabited by record types (see earlier Sect. 4 and (76) below).
The nodes of these trees are annotated with terms in a typed lambda calculus,
with mother-daughter node relations corresponding to predicate-argument structure
(by convention arguments appear on the left whereas predicates appear as the
right daughters). Abstracting away from details for now, for example the content
associated eventually with the string John arrives will be the functional application
of the lambda term x.Arrive0 x inhabiting the function daughter, to the singular term
derived by processing the name John (in TTR terms, notated as the sole witness x
of type john, x : john, in the display below). x.Arrive0 x has the semantic type of
a one-place predicate, which in the logic and diagrams is shown as: Ty(e ! t)).
The result of functional application will be a propositional type (Ty(t)) (the witness
exemplifying such a propositional type will be an event/situation, notated as p in
the display below). For simplicity we assume that John will trigger the search of
the context for an individual (of semantic type e; Ty(e)) named ‘John’.15 In terms
of representations, such contents are accumulated in fields, recursive label-value
pairs, of TTR record types (see earlier (35) in Sect. 4). Labels (like p or x below)
stand for the witnesses of the types expressing derived conceptual content. The
semantic content is accumulated as the value of a designated content field (indicated
as CONTENT in the simplified diagram below):
(76) (simplified) DS-TTR representation of the conceptual structure derived by processing

John arrives:
A pointer, ˙, which moves around the tree nodes as the result of defined language-
specific processing actions (thus accounting for word order), indicates the current
node of processing, the current locus of attention.
Words and syntactic rules are conceptualised in DS-TTR as lexical and com-
putational actions respectively, i.e., as triggers for inducing packages (macros) of
atomic actions if certain specified conditions are satisfied at the current locus of
attention (the IF specification in (77) below). Such actions include the triggering of
contextual searches for conceptual content (find, (fresh)put, substitute),
building conceptual tree-structure (make), copying values, introducing predictions
of upcoming input, or, finally, aborting (abort) in case the conditions of use of
the word/rule are not satisfied in the current linguistic and non-linguistic context.
In this sense, words and rules can be seen as ‘affordances’, i.e., possibilities for
(inter)action that agents attuned to these possibilities can recognise, predict, and
manipulate (Gibson 1977). For example, the simplified lexical entry for arrives
shown below first checks whether the pointer is at a node predicted to be of predicate
type (indicated as: ?Ty(e ! t)) and, if this condition is satisfied, it introduces, via
the execution of the atomic action put, the conceptual content represented by the
function x.Arrive0 x (the full specification of the macro includes the execution of
15
Two analyses for names currently co-exist in DS: (a) as constants resulting from the contextual
enrichment of metavariables introduced by names, and (b) as iota-terms, namely terms carrying
uniqueness implications and descriptive content. Here no stance is taken on this issue as it does not
affect current concerns.
further actions relating to tense, mood, agreement etc. through the employment of a
set of actions like make[node], go[to node], abort[processing], etc.):
(77) Lexical entry for an intransitive verb:
Following Gregoromichelaki (2006), it is also assumed that a propositional

representation (of type t) always includes an indication that events/situations
belong to some world/time of evaluation (Recanati 2004)16 as a contextually
derived value of specified parameters. Additionally, each predicate type derived as
the CONTENT field at each subnode of the tree includes independently shiftable
world/time/situation parameters to account for well-known cases of differentiation
among the parameters of evaluation for various predicates in a sentence:
(78) The fugitives are now in jail. (Enç 1986)
All such context-dependent values are derived through the fact that various linguistic
elements are defined as initially introducing metavariables in the conceptual repre-
sentation. Metavariables in DS (indicated in capitals and bold font) are temporary
place-holders introduced to enforce their later substitution with values (variables or
constants) from the current context. For example, pronouns, anaphors, ellipsis sites
(auxiliaries in English), tenses, modal verbs, etc., lexically introduce metavariables
of various types (of type e or predicative types) and restrictions constraining their
subsequent replacement by values derived from the linguistic or non-linguistic
context.
Consequently, in order to model the interpretation process of reflexively inter-
preted elements like indexicals I, you, and now, contextual parameters regarding
the utterance event are recorded in a structured CONTEXT field17 on which the
16
For eliminating worlds from the semantics, replacing them with more psycholinguistically
plausible TTR contents in terms of (types of) situations, see Cooper (2005). Here we maintain
the more conservative view for brevity of exposition.
17
The differentiation CONTEXT vs. CONTENT fields is for exposition purposes only, just for the
convenience of shortening reference to fields in the displays; it does not signify any substantial
claim regarding any qualitative differentiation among the parameters handled. In TTR there is
always an intuitive inclusion of the context in that, via the notion of dependent types, subsequent
fields can depend on elements introduced previously (up along the vertical dimension in record
types) but not the other way round. In terms of expressivity, reference to a value in some record
(type) can be indicated via the definition of paths leading to specific values; we show such paths
with dots separating the sequence of steps, for example r.l1 .l2 , refers to the value of label l2
CONTENT field depends. The CONTEXT field records the occurrence of each
word-utterance event (utterance action, notated as e.g. u1 ,u2 , : : : ,un ), including the
‘words’18 that have been uttered, the agent (i.e., the utterer [spkr on the diagrams],
which can be distinct from the agent taking responsibility for the illocutionary
act), the addressee (addr), time/location of the event (following the specification
of micro conversational events in Poesio and Rieser 2010; see earlier Sect. 3), the
world parameter of the context, and various constraints in the relations among these
terms19 :
(79) (simplified) DS-TTR representation for John arrived with contextual parameters:
Processing of a contextually dependent element, e.g. an indexical pronoun like

I, first checks whether the pointer appears on a node predicted to be of type e
which provides the value of l1 in record (type) r. In the displays here, the various fields are freely
simplified and condensed in various ways for uncluttered illustration of the relevant points.
18
Note that ‘words’ in DS-TTR are conceptualised as phonological/graphemic/signed shapes, i.e.
stimuli that serve as the triggers for DS-TTR actions; not, as usual, ‘signs’, or ‘expressions’
(Cappelen and Lepore 2007), or phonology/syntax/semantics feature bundles (cf. Saka (2011) for
discussion about the nature of linguistic elements, leading to distinct conclusions).
19
The initial arrow carrying a word string illustrates the process of scanning, the process of
recognising stimuli as triggers of lexical macros. Subevents are sequentially numbered through
subscripts and further subscripts can be used for mnemonic purposes (the subscripts s, a here stand
for speaker, addressee but will not be maintained further to avoid confusion with occurrences
of subscripts s on types where they indicate the subtype of type e (entities) that are situations
(type es ).
(anticipated but not yet realised ‘goals’ are indicated with a ? in front of the expected
specification). Then, if this condition is satisfied, an appropriate parameter in the
CONTEXT field will be located (the entity that is the speaker) and its value will be
copied as the value on the current node, namely in this case, the current speaker
value and the indication that it is of Ty(e)20 :
(80) Lexical entry for indexical pronoun I:
Interspersed with lexical entries, general computational rules can apply without
‘scanning’/generation of linguistic input. For example, computational rules induce
the concatenation of word-utterance subevents (indicated as u1 ˚u2 ˚ : : : un ) pro-
ducing cumulative utterance events at mother nodes of the tree structure. Such
concatenation is effected in parallel with the computational actions performing
functional application on content-complete nodes (see Purver et al. 2010; Purver
et al. 2011 for details):
(81) Concatenation of subevents in the CONTEXT and parallel function application in the
CONTENT fields:
For our purposes, we note that there can be additional world and event parameters
in the CONTENT field, introduced via the actions of linguistic operators like tense,
modality etc., with accessibility relations represented as TTR-dependencies among
CONTENT and CONTEXT fields (to deal with such phenomena where shift of
20
Bold lower case variables in the lexical macros indicate rule-level variables that unify with
specified values on the current tree descriptions (parse states). These values are then used in the
further execution of the macro (for formal explication see Kempson et al. 2001: 90–91, 311).
evaluation occurs, e.g. conditionals, see Gregoromichelaki 2006; for an alternative

TTR formulation that does away with worlds in favour of types of situation, not
adopted here for simplicity of exposition, see Cooper 2005). Such potential is
needed, independently of quotation, in dialogue since shifts and interactions of
contextual and evaluation parameters can freely occur subsententially:
(82) Nun: I’ll telephone the Wing Governor. Surely she will appreciate the hiatus in care that
has occurred.
Nurse Franklin: Of course she must! In terms of women’s healthcare, we’re in charge,
so I wouldn’t mince your words. [BBC Transcripts, Call the Midwife, Series 3,
Episode 3]21
(83) Stanley: Louis, I just didn’t think
Louis: you’d ever hear from me? [BBC Transcripts, Dancing to the Edge,
Episode 5]
In these CONTEXT parameters, following Ginzburg and Cooper (2014), we now

add an NL-use parameter for each sub-event, indicated initially as a metavariable G
of type linguistic usage (l-use) to represent the reification of the processing of the
utterance as an event/action conforming to some set of computational and lexical
actions, specified for a particular “language”, according to which metalinguistic
judgments can be assessed.22,23 Here the potential to introduce such language-
use metavariables makes explicit the freely available potential for switching the
language, idiolect, or any other variety of usage, and evaluation of metalinguistic
judgments according to such switches, all of which can occur at any subsentential
stage of production/interpretation.
The DS-TTR grammar operates by means of licensing in context word-utterance
events according to their temporal order. As we said, words (and the operation of
“syntax” in general) are modelled in DS-TTR as offering ‘affordances’, opportuni-
21
In the illustration of such phenomena, in my view, scripted dialogue provides valuable evidence
because such occurrences cannot easily be attributed to speech errors.
22
Note that this also shows that the above mentioned CONTEXT vs. CONTENT distinction is indeed
artificial and hence present here only for simplicity of display purposes. The truth values of
“metalinguistic” statements rely on conceptualisations of the instantiation of implicit contextual
variables.
23
Metalinguistic judgements (involving concepts like ‘sentence’, ‘word’, etc.) involve G-
dependent types that range over conceptualisations of NL-use that reflect folk-linguistic concep-
tions but do not necessarily correspond with the analysts’ grammar of a particular language (unless
of course the discourse involves discussion of exactly such a grammar). The actual processing
model (the grammar) used (unconsciously) for processing an utterance will be captured by the
rule-level variable indicated as g in the quotation-related processing actions later. Unlike G&C, this
is an essential reservation for the DS-TTR formalism which does not license form-meaning pairs
(“expressions”) but, instead, interlocutors’ performance, i.e., the production and interpretation of
actions. Any reification of (part of) the products of such actions is then necessarily the outcome of
some coercion and reification of the actual language use.
ties for action, exploited by the interlocutors to facilitate interaction, so that words
and linguistic constructions are not conceptualised as abstract objects, ‘expression
types’, that are associated with referential/semantic values (cf. Cappelen and Lepore
2007: chapter 12). As in DRT (Kamp 1981; Kamp and Reyle 1993) and related
frameworks (see also Jaszczolt 2005), semantic, truth-conditional evaluation applies
solely to contextually enriched conceptual representations. However, unlike all
these other models, truth-conditional evaluation applies incrementally, as each word
is processed (see e.g. Hough 2015 for details). The other distinguishing feature
of DS-TTR, as compared to DRT, is that the process of progressive building of
conceptual structures is the only notion of “syntax” admitted, in that there is no
intermediate level of syntactic structuring where a string of words is assigned
hierarchically organised constituency as syntactic categories, phrases or clauses.
Such constituency is considered in DS-TTR as epiphenomenal on the function-
argument semantic relations as typified in the lambda-calculus analyses of NL
meanings. In consequence, in DS-TTR, all standardly assumed syntactic dependen-
cies have been reformulated in procedural terms, i.e., in terms of how time-linear
processing is affected by semantic dependencies. Such procedural explanations
include, in particular, the classical data used to deny the direct correspondence
between NL-structure and semantic content that led to accounts via transformations
(long-distance dependencies, binding, quantification, etc.; see e.g. Kempson et al.
2001; Cann et al. 2005; Gregoromichelaki 2006, 2011, 2013a). With no privileged
semantic entities corresponding to (types of) expressions, only mechanisms for
processing stimuli, quotation thus offers a crucial test for the legitimacy of these
DS-TTR claims regarding natural languages: When processing a quoted/cited string,
what happens within the quotation marks (or any other indications) following these
assumptions?
To answer the question of NL quotational/citational uses (Sects. 7.3.1, 7.3.2, and
7.3.3) we first need to remember that the application of these DS-TTR grammatical
assumptions to the analysis of quotation is parallel to their application in the analysis
of conversational mechanisms. This is because, as we saw in Sect. 2, and following
the insights of Ginzburg and Cooper (2014), quotational phenomena appear to be
subsumed under the constructs needed to underpin interactional mechanisms and the
modelling of metacommunicative coordination. From this point of view, first, DS-
TTR’s lack of a syntactic level of representation and its sub-propositional semantic
evaluation is an advantage in conversational modelling since it directly provides
the mechanisms for accounting for split utterances and fragmentary discourse in
dialogue (see (25) and the illustration in (85) below). Various cases of subsentential
actions in dialogue are employed to indicate that the words uttered by the current
speaker do not necessarily reflect his/her perspective (as in e.g. (11)–(17), (25)), or
are not being used with the sole purpose of inducing their conceptual content (see
e.g. (18)–(24)).
In DS-TTR, modelling the potential of partially assuming another interlocutor’s

role, being perceived as “demonstrating” what the other interlocutor was going
to say, is achieved unproblematically because the potential for sharing syntac-
tic/semantic dependencies is guaranteed at each step and there is no requirement
to derive a global propositional speech act: Both speaker and addressee perform
processing steps incrementally, guided not solely by the NL string, but also driven
by predictions (introducing ‘goals’) generated by the DS-TTR grammar (in the
displays these anticipated goals are shown with a ? accompanying each predicted,
but not yet realised, specification). These goals are imposed by either the procedures
associated with NL elements (lexical actions) or are system-generated as general
top-down computational goals to be achieved in the next steps. Simplifying for
presentation purposes, for example in English, with its characteristic SVO structure,
a general computational goal will ensure that production and parsing start with
the expectation of the appearance of a subject first (of semantic type e, ?Ty(e)),
followed by a predicate afterwards (of semantic type e ! t). The pointer then shifts
to the ?Ty(e) node, which processing of the first word in the sentence, e.g. John,
annotates with a value of type entity (e.g. the logical representation of the individual
John which is indeed of Ty(e)). Subsequently, if an intransitive verb, like arrive,
follows, it will trigger actions that annotate this predicate node with a function
to be applied to the subject. It will also introduce the event/situation (shown as
the variable s below of type es ) that is taken as the witness of the type derived
by processing the clause (see earlier Sect. 4). Finally, computational actions that
complete the process will follow next (CONTEXT values are omitted for clarity, the
label tn indicates the treenode address which serves as a handle for accessing the
relevant node content, PREDICTION and COMPLETION are examples of the general
non-lexical computational actions employed in DS):
(84) Incremental steps in processing a clause with an intransitive verb24,25 :
24
The representations here employ so-called manifest fields. The notation employing the equality
sign is abbreviatory for a singleton type constructor (see e.g. Cooper 2012; Ginzburg 2012),
indicating subtypes of some type restricted to a single member, that is, only the relevant value
mentioned. So, for example, xDjohn : e means that the value of label x is of the subtype of type e
whose unique witness is the individual John.
25
The notation employing a dot indicates a path to a value, e.g. r.tn indicates that the value needed
is to be found as the value of label tn in record (type) r (see also fn. 24). (Note that this use of the
dot notation is different from its use in separating the œ-bound variable [plus restrictions in TTR]
from the function expression, e.g. x:[xDjohn : e]. Arrive0 x)
If a transitive verb follows instead, its lexical entry will introduce not only the
conceptual content associated with the verb but also the prediction that an argument,
the complement, will follow immediately afterwards. Such complements can be
either of individual entity type (type e) or of propositional type (semantic type t),
the latter for propositional attitude or reporting verbs. The embedding of proposi-
tional types as complements defines one aspect of linguistic recursivity. Another
aspect, related to adjunction (e.g. relative clauses, adverbials, parentheticals), is
implemented by relating trees via a so-called LINK relation, a relation that does not
involve mother-daughter tree-relations. The construction of a LINK relation among
two independent trees offers opportunities for interrupting the construction of one
tree at a specific node in order to elaborate on some of its terms by shifting the
pointer to an auxiliary tree, processing some linguistic input there, and, eventually,
enforcing sharing of this information among the paired trees.
Thus, parsing in DS-TTR incorporates elements of generation (production)
through the constant formulation of predictions for what will ensue next. On the
other hand, production exploits the parsing mechanism in that licensing of the
generation of each word relies on checking that the string thus produced can deliver
a conceptual representation that accords with the (partial) conceptual structure the
speaker attempts to verbalise (called the goal tree). As a result, speaker and hearer
roles involve in part mirroring each other’s actions (Gregoromichelaki et al. 2013a,
b; Pickering and Garrod 2013). From this perspective, it is then unproblematic to
model the sharing of utterances and the joint construction of conceptual structure
in dialogue. As the schematic illustration in (85) shows, the only difference that
registers the change of utterers during simple split utterances is the change of values
in the contextual parameters:
(85) Processing John arrives: final content derived through two micro-conversational events
by different speakers
The sharing of syntactic/semantic dependencies is possible because, as speakers

and hearers simulate the actions of each other, the fulfilment of syntactic/semantic
predictions is attempted at each incremental step, subsententially, for both parser
(hearer) and generator (speaker). Such fulfilment can be achieved by either speaker
or hearer, whether on the basis of the other interlocutor’s input, the context, or by
recourse to the processor’s own resources. As no structure is ever assumed to be
derived for the sentence string, no whole-string “grammaticality” considerations
ever arise. Similarly, no context-independent compositionality restriction applies to
NL strings; only contextually derived conceptual structures are interpreted compo-
sitionally. Hence, fragments that can be processed by fitting into a structure that is
already in the context are licensed directly, that is, they are NOT characterised as
elliptical and there is no requirement that they need to be enriched to a propositional
type to be interpreted:
(86) A: Who left?
B: John?
C: with Mary, yesterday.
Such split utterances are unproblematically processable and are in fact a natural
consequence of such a fine-grained bidirectional incremental system: As predictive
goals are constantly generated by the grammar, to be achieved symmetrically by
both the parser and the producer, the hearer/parser can wait for input from the
speaker in order to fulfil these goals. However, according to the grammar, such goals
are also what activates the search of the lexicon (‘lexical access’) in generation in
order to recover a suitable NL word for the concept to be conveyed. As a result, a
current hearer/parser who achieves a successful lexical retrieval before processing
the anticipated NL input provided by the speaker can spontaneously become the
producer and take over verbalising the continuation of the utterance instead (for
detailed analyses see Eshghi et al. 2010, 2011, 2012, 2015; Gargett et al. 2008,
2009; Gregoromichelaki et al. 2009, 2011, 2012, 2013a; Kempson et al. 2009a, b,
2011a, 2012; Purver et al. 2009, 2010, 2011).
We will now see how these mechanisms which licence split- and non-sentential
utterances in conversation license and interact with reporting and metalinguistic
phenomena.
7.3 Metalinguistic Devices in DS-TTR

7.3.1 Pure Quotation (Citation)
As we’ve already seen, the utterance-situation parameters (speaker, hearer, time of

utterance, etc.) in the information state, the value of CONTEXT in DS-TTR, include
storage of the word forms that have triggered processing. As Ginzburg (2012) has
shown, this is essential for various parallelism effects observed in the processing
of dialogue phenomena (e.g. interpretation of clarifications as echoic; see e.g. (5)).
In addition, CONTEXT also stores the DS-TTR processing actions that have already
been used in deriving conceptual CONTENT structures. This is necessary for the
resolution of anaphora and ellipsis (Kempson et al. 2001, 2011b, 2016; Cann et al
2005, 2007; Gregoromichelaki et al. 2011; Kempson et al. to appear; Purver et al.
2006). Under this view, the processing actions utilised in parsing and production
are first-class citizens in the model in that the grammar includes means for referring
to sequences of actions already stored in the CONTEXT, reasoning over them, and
reemploying them again in subsequent steps (Cann et al. 2007). This is necessary
for the explanation of phenomena like ‘paycheck anaphora’ and ‘sloppy readings’
of ellipsis ((87)–(89) below) where the interpretation changes due to the new local
environment where the anaphoric elements acquire their interpretation (Kempson et
al. 2011a). They also need to be available both subsententially and for anaphoric
and cataphoric employment, the latter shown in (90)–(91) below:
(87) The mani who gave his paycheck to his wife was wiser than the manj who gave it to
his mistress.
[‘manj ’s paycheck’]
(88) John upset his mother. Harry too. [‘Harry upset Harry’s mother’]
(89) The man who arrested John failed to read him his rights. The man who arrested Tom
did too. [‘failed to read Tom’s rights’]
(90) The representations here employ so-called manifest fields. [this document, footnote 23]
(91) It appears that John left.
In cases like (87)–(89), in order to model the rebinding of the anaphoric elements
(indicated in bold) to the newly introduced subjects in the next clause, DS-TTR
retrieves a sequence of actions <ai , ...., aiCn > already performed in processing the
previous clause and therefore stored in the context representation. It then executes
them again in the new sentential environment with the result that the new subject
now provides the local binder of the metavariables introduced by the anaphoric
elements (see e.g. Purver et al. 2006; Cann et al. 2007).
The same action retrieval mechanisms are used in cases of self-repair where
one element that replaces another (surfing to replace swimming below) needs to
re-compose with elements that have already been processed (with Susan below):
(92) Peter went swimming with Susan, um : : : , or rather, surfing, yesterday. [‘Peter went
surfing with Susan yesterday’]
Such cases of repair (whether self- or other-repair, including clarification), in
many cases, require re-execution of already processed material. This is modelled
as the re-running of a sequence <ai , ...., aiCn > of actions stored in CONTEXT
in order for material to be reprocessed (Hough 2015; Eshghi et al. 2015). For
such repair and other purposes, DS-TTR also records the various potential but not
pursued processing options licensed by the grammar at each step: As we saw, the
DS-TTR formalism operates by generating predictions regarding future steps of
processing. This results in the generation of multiple potential processing paths that
the information state can develop into. These paths, whether pursued or not, are
taken as part of the context representation and can be illustrated in the form of a
DAG, a Directed Acyclic Graph (see Sato 2011; Hough 2015 for formal details):
(93) Context DAG showing various potential processing paths
In such a graph, edges correspond to potential DS-TTR computational/lexical

actions and nodes to resulting information states (Purver et al. 2011). This contex-
tual representation is employed for the modelling of various dialogue phenomena,
where the parser needs to backtrack to a previous path, other than the one actually
pursued, and proceed to another interpretation of the input or reformulation of the
utterance (see Hough 2015; Eshghi et al. 2015 for details). The claim that such
abandoned paths need to remain in the context is additionally justified by cases
where “repaired” elements need to be accessible for e.g. anaphoric purposes:
(94) Jill left, no, (I mean) Bill left, she’s in Paris already.
Now let’s examine the cases of Recanati’s “closed quotation” where an NL-
string appears in a regular ‘NP’ position, i.e., where the grammar, under DS-TTR
assumptions, has already generated a prediction for the processing of a singular
term (or any other semantic type).26 Given this prediction, there will always be an
attempt for whatever is processed in such a position to be construed as ‘subject’,
‘object’, modifier etc. Exactly because any DS-TTR grammar for a particular use
of language consists of routinised sequences of actions, this will also be possible
for any set of actions, for example non-linguistic actions as in (73)–(75). Moreover,
given incrementality and the absence of sentential grammaticality licensing, any
DS-TTR model can license the processing of input provided through some language
use distinct from the one providing the tree position that the content of this input will
annotate.
Regarding interpretation, as word forms in DS-TTR are assumed to constitute
triggers for macros of actions, which include importing conceptual content con-
tributions, inevitably any conceptual contribution associated with a cited word
or string of words will become available to the interpreter if it belongs to a
known type of language use; and the same goes for any other conventionalised
non-linguistic signals. However, where the context requires a “metalinguistic”
interpretation for the uttered string, the conceptual value, like other properties of the
stimulus associated with the word-form, even though accessed and built up, ends
26
For the potential of such quoted strings to function as Ns or other categories rather than NPs (e.g.
The whys raised by this issue. These are not ‘I really should’ radishes : : : . (Clark and Gerrig 1990,
from Jon Carroll, San Francisco Chronicle)), see De Brabanter (2005b, 2013).
up embedded as the value of the particular predicted type on the treenode of the
eventual conceptual representation. In such cases, given that the DS-TTR grammar
does not provide form-meaning correspondences but only provides for the parsing
and generation of utterance events, the process of ‘coercing’ some linguistic element
to fill the role of an already predicted conceptual type on the tree can be taken as a
reification of the grammatical process itself (like the demonstration of a car sound
in the same position can be taken as the conceptualisation of some sound experience
that is being demonstrated).
In DS-TTR terms, these assumptions can be expressed as the ad hoc categorisa-
tion of the running of a sequence of actions <ai , ...., aiCn > at a node. Such a sequence
will belong to some particular linguistic use (grammar) indicated as the rule-level
variable g (see earlier Sect. 7.2 and fn. 19, 21, 22) which becomes instantiated by the
intended grammar being invoked. The idea is that embedding the actual execution
of a sequence of actions as the conceptual value of a node on the tree results in their
conceptualisation as an element of the type already predicted in the particular tree
position where the pointer finds itself. The intuition behind this implementation is
that an utterance event (notated as uq below) is performed (demonstrated) under the
assumption of a particular DS-TTR grammar g (captured by binding of g) in order
to provide the content value for the current treenode27 :
(95) Computational action for processing quotation:
The IF condition in the computational macro first checks whether the pointer is at
a node predicted to be of a particular type, e.g. type e, cn (common noun), or any
other type of content that cited strings can be associated with (x2fe, cn, : : : g ). If it can
be shown, as seems to be the case, that content derived from citation can belong to
any semantic type (De Brabanter 2005a, 2013), this restriction (2fe, cn, : : : g) can be
dropped. Suppose that the pointer is at a node predicted to be of type e. The string
to be processed is the following:
(96) ‘John arrives’ is grammatical.
In such a case the action in (95) can apply to provide a value on this node by
processing the upcoming utterance event uq which is immediately provided (note
that events in DS-TTR belong to subtypes of type e, the type es (Gregoromichelaki
2006; for the notation employing the D sign of manifest fields employed here, see
fn. 23). Unlike the anaphoric cases of action re-running we saw earlier in (87)–(89),
here the process is cataphoric: the demonstrating event has been predicted to occur
27
Like x, ai ,..., aiCn are also rule-level variables that become bound to whatever individual actions
the current state provides; see fn. 19.
and its subsequent occurrence and processing duly satisfies this prediction. In this
way, the content that appears at the relevant node, the specific singleton type of
the event uq , its grammatical characterisation in the DS-TTR sense, is constructed
on the fly: this type of event has as its sole witness an upcoming employment of
some grammar (rung ) to execute a sequence of actions <ai ,..., aiCn >. Notice also
that the instantiation of the metavariable indicated by the parameter g will invoke
a language use that can be distinct from the linguistic use applicable to the rest of
the string. The folk-linguistic (or scientific) characterisations that predicate of such
reified linguistic uses can then target aspects of the grammatical processing that has
just been executed. This, in essence, is just a process of explicit categorisation of
various aspects of the stimulus (for the potential properties accessed in such cases,
see e.g. Saka 2011, 2013; for inferences narrowing down such targets, see e.g. Saka
2003/2005).
In (the unusual) cases where the metalinguistic/metacommunicative interpreta-
tion becomes available only after a metalinguistic predicate has been processed, the
parser will need to backtrack along the DAG-recorded path (see (93)) to a previous
parse state in order to pursue this new option as in cases of repair in conversation. We
can then assume that the function of quotation marks or other quotational indications
in spoken or written language is exactly to indicate to the parser that a non-default
processing strategy (i.e. a DAG path of low-probability rating) is to be pursued.
Such cases are similar to those where there is invocation of a previous utterance
event (‘echoing’). For example, in a situation where A rehearses the string John
arrives for a radio play, B, the sound engineer, can say:
(97) “John” was a bit loud.
In such cases, we assume that, in addition to the choice of grammar g, there is

also a contextually available token utterance event (in the display below indicated
by the rule-level variable u belonging to the subtype of entities that constitute
events/situations es ). As in the previous case (95), there is again a quotation event
(uq ) mentioned below, conforming to some linguistic use g (instantiated by g), but
this time (part of) the contextual parameters are set by the contextually available
event (instantiating u):
(98) Computational action for processing sententially embedded echoic quotation:

These non-sentential echoing quotation cases are similar to direct quotation to which
we now turn.
7.3.2 Direct and Indirect Reports
Under DS-TTR, the lexical action for a framing verb (e.g. a verbum dicendi like say)
can be assumed to uniformly provide conceptual content that is able to combine
with related semantic objects (e.g. propositions, of type t, and utterance events
of type es ), provided through distinct combinations of processing actions. Such
combinations of processing strategies can result in cumulative results modelling
either direct and indirect reports, or intermediate phenomena, without postulating
specific types of static syntactic constructions as in G&C and other models.
In DS-TTR, the only factor that accounts for the alleged syntactic differences
between direct and indirect quotation (e.g. parenthetical/reversed word order or wh-
extraction; see e.g. Schlenker 2011; Bonami and Godard 2008) is that the actions
induced by such verbs, like other verbs in English (e.g. eat) can license ‘object-
drop’, a license that is modelled in the DS-TTR account by allowing object-drop
verbs to take as their complement a metavariable. As in the DS-TTR modelling of
ordinary cases of pronominal or elliptical anaphora resolution, such a metavariable
needs to be provided with a value from CONTEXT (in the form of existing conceptual
content or via the rerunning of actions). In ‘direct quotation’ cases, the value for
such a metavariable can be provided by the independent clause processed as an
antecedent either anaphorically or cataphorically, e.g.:
(99) “I talk better English than the both of youse!” John shouted/announced/said.
Framing verbs can also compose directly with non-linguistic actions, e.g. gestures
or sounds, which is straightforwardly modelled in the DS-TTR formalism, as there
is no distinction between linguistic and other actions: both invoke categorisation
processes which, in the case of linguistic performance, are standardly characterised
as the ‘grammar’. However, as has been pointed out previously (Slama-Cazacu
1976; Clark 1996; Postal 2004; De Brabanter 2010b), the grammar needs to be
conceived in a much wider sense to account for cases like the following:
(100) The car engine went [brmbrm], and we were off. (Clark and Gerrig 1990)
(101) The boy who had scratched her Rolls Royce went [RUDE GESTURE WITH HAND] and
ran away. (Recanati 2010)
(102) I didn’t see the [IMITATION OF FRIGHTENING GRUMPINESS] woman today; will she
be back this week? (De Brabanter 2010b)
(103) Piano teacher to student: It’s not [plays passage in manner ]—it’s [plays same
passage in manner ’]. (Horn 1989)
As the DS grammar operates predictively, positions in the conceptual representation

are constantly generated in anticipation of the next input.28 Being processed in such
a position can coerce any perceptual stimulus to induce a processing action that
will compose its derived categorisation (i.e. content) with the rest of the conceptual
representation (Gregoromichelaki 2013b). Since such conceptual representations
are expressed through the TTR formalism in DS-TTR, as in G&C, any perceptual
stimulus can be accommodated by the type system via the subtyping relation (see
Cooper 2012 for formal details), hence allowing for the construction of ad hoc
types. This is commonplace in actual conversational interactions. For example,
Gregoromichelaki (2012) and Gregoromichelaki and Kempson (2013) argue that
non-linguistic actions are regularly the antecedents of ellipsis, clarification requests,
etc. In such cases, constraints on the conceptualisations of such actions are
imposed linguistically via the form of antecedent-requiring elements (e.g. case
requirements29 ; see (104) below) that retrospectively restrict the structure of the
construal underpinning the conceptual representation:
(104) [Context: A is contemplating the space under the mirror while

re-arranging the furniture and B brings her a chair]
A to B:tin karekla tis mamas? / *i karekla tis mamas? Ise treli? [clarification] [Greek]
‘theACC chairACC of mum’s / *theNOM chairNOM of mum’s.
Are you crazy?’
‘Mum’s chair? Are you crazy?’
(105) [Context: A asks who C has invited and D points to B]
C: (Actually,) not him, his sister.
(106) [Context: A comes in the room and punches B]
B to A: Why?
Supporting DS-TTR’s assumed uniformity of lexical, computational and context-
shaping non-linguistic actions, notice that even the sequential process of pars-
ing/production can become the object of anaphoric mention:
(107) The rules of Clouting and Dragoff apply, in that order. (Ross 1970)
The same idea covers cases of direct and indirect quotation: the fact that contents
supplied by framing verbs can acquire propositional complements either directly
via embedding a description of the content of an utterance event (indirect reports)
or indirectly via the echoing of a previous utterance event (propositional or not,
in direct reports) allows us to capture the continuity of direct/indirect discourse.
It also explains the intermediate cases, e.g. mixed quotation and free (in)direct
28
As we saw, the position currently under development is indicated by a ‘pointer’, ˙, which is
what accounts for variable word-orders.
29
According to DS-TTR, case affixes in morphologically rich languages impose the prediction/goal
of an appropriate tree-structural position to accommodate the conceptual content contributed by the
linguistic element carrying it.
discourse structures. Such structures show that fine-grained processing mechanisms

can be combined in various ways, both synchronically, at the discretion of a
current speaker for novelty effects, and diachronically, becoming routinised and
therefore commonplace (i.e. assigned high probability as processing paths in the
context DAG; see earlier (93)), to deliver various conceptually articulated construals
and non-conceptual impressions. Thus modelling these intermediate phenomena
via fine-grained mechanisms that can combine with each other argues against
postulating monolithic, fixed form-meaning correspondences (‘constructions’) since
the available mechanisms can, and will, be freely exploited by human processors
to deliver various novel effects in context. Fixing ab initio the outcome of such
combinations is bound to fail to account for the various potential outcomes of
situated processing (see also Gregoromichelaki and Kempson 2017).
Turning to direct quotation first, we can describe the DS-TTR grammatical
mechanisms allowing for its processing and effects as follows: As the DS-TTR
grammar is articulated in terms of actions, we can postulate that the properties
that characterise “direct reports” are the result of potentially choosing to focus the
hearer’s attention30 on the triggers of the lexical actions (words, as stored in the
context DAG) presented as having been used by another speaker (i.e. demonstrating
(echoing) a contextually available utterance event u retrieved from and stored in
the CONTEXT representation and via instantiation of the g parameter to another
speaker’s grammar as we saw earlier for pure quotation). We can assume that
this can sometimes be indicated by the quotation marks. As discussed earlier
for pure quotation (Sects. 6 and 7.3.1), following G&C, Recanati (2010), and
Predelli (2003), for direct discourse the DS-TTR conceptual representation derived
will involve an embedded utterance event uq , corresponding to the demonstration
the speaker performs. The verb say in English and other languages regularly
combines with utterance events, whether echoic or not or assertional or not (contra
Brandom 1994: 531 whose presentation implies that the propositional-complement
use should be primary in that it makes explicit implicit ‘assertional ascription’
practices):
(108) He said “constraints in agriculture” when he meant “excluded products”.

(109) At 36 months, he had begun developing functional language but could not grasp
concepts like first and second person; he said “I” when he meant “you,” and “you”
when he meant “I.” [https://findingmykid.com/tag/hockey/ retrieved 13/6/16]
(110) Martin approached, listened for a second, looked at the UPS guy, and said, “Oh, hi!”
[https://findingmykid.com/tag/hockey/ retrieved 13/6/16]
(111) “Ouch!”/“Wow”, she said.
(112) Say “Kognitionswissenschaft”.
30
When “focussing attention” is explicitly conceptualised, it is represented in DS-TTR by
introducing additional inferentially derived propositional or sub-propositional contents as LINKed
structures, the representational device used for the processing of adjuncts like relative clauses,
conditionals etc. (see Kempson et al. 2001; Cann et al. 2005; Gregoromichelaki 2006).
In cases of direct reports, this demonstrating event uq , “echoes”, i.e., bears a

contextually determined similarity value, ‘resemblance’ to, another, anaphorically
retrieved utterance event (u). This anaphorically retrieved event supplies (part of) the
contextual parameters, thus accounting for the corresponding change in the values
of indexicals across speakers and turns:
(113) A: Will you say to Nick. . .
B: “I hate you”? Yes, why? [‘B hates Nick’]
(114) A: Did you say to Nick “You . . .
B: “hate yourself”? Yes, why? [‘B said Nick hates Nick’]
Notice that, as in the earlier (104)–(107) cases, the presumed contextually available
element, in this case a speech event, need not be part of the context already; instead,
the introduced requirement that it should be part of the context eliminates DAG
paths where it is not possible for such an event to be conceptualised (e.g. contexts
where the hearer believes that the reportee was unable to communicate) or leads
to the generation of further metacommunicative interaction, e.g. clarification, in
order to be accommodated. This is standard for many cases of direct reports where
what is “reported” has never actually been uttered (see e.g. Tannen 1986; Norrick
2015) and cases intermediate between direct reports and free direct speech where the
contextual parameters again need to recruited from such an imaginary, reconstructed
event:
(115) Adam: Well. I can tell you what her view on that is. and that
Sherm: what.
h I’m older, and therefore I’m in a worse competitive position, and
I and I’ve really got to produce .
Sherm: but I’m smarter [LAUGHS] yeah. [SAID VERY SOFTLY]
Adam: and I’m going to .
Sherm: yeah. [SAID VERY SOFTLY] (Grimshaw 1987)
For cases standardly regarded as clear-cut cases of direct report constructions, in
DS-TTR terms, the only difference with the previous echoing case in (97)–(98) is
that such an otherwise freely available computational action has been “lexicalised”:
it has become part of the routinised macro stored as an option in the lexical entry
of the verbum dicendi,31 so that, in terms of the DAG representation, its execution
constitutes a highly probable option. So, for example to process a string like the
following:
(116) John said “I was loud”.
the following (schematic as regards irrelevant details) lexical entry for say can be
invoked:
31
Alternative options in a lexical entry are listed as embedded in ELSE statements, before abort
is encountered.
(117) Lexical entry for say C LINK sequence32 :
The condition IF here expects the presence of a salient utterance event in the context
(to bind u) whose speaker will provide the value for x. The lexical macro then
ensures that the subject of the proposition will be that speaker x (put(?[x : e])).
Next it constructs the predicate and its object node (abbreviated presentation here,
see fn. 31) and inserts a metavariable U of type utterance event (es ) as a temporary
place-holder. A LINKed node is then introduced (shaded in the display), which is
the device used in DS-TTR for the processing of adjunction (see e.g. Kempson
et al. 2001; Cann et al. 2005). The conceptual value on this LINKed node will be
provided by the execution of the actions needed to process the following string with
contextual parameters provided by the contextually instantiated value of u which
ensures that the indexicals receive appropriate values, e.g. as instantiated by the
utterer of u for a pronoun like I. The DS-TTR constraints governing LINK transitions
will then ensure that the value of the metavariable U will be unified with the content
of the LINKed node.33
Under this analysis, the syntactic/semantic opacity observed in such structures is
explained by (a) the presence of the LINKed node, as is usual in DS-TTR regarding
the modelling of the banning of extraction from adjuncts (Kempson et al. 2001),
and (b) the embedded nature of the propositional content derived on the LINKed
node, in that it is just one of the TTR types characterising the utterance event.
32
As mentioned earlier, make, go, put, run, etc. are elementary DS actions processing strings
and building conceptual structure. They are modelled via accessibility relations among information
states in the Dynamic Logic underpinning DS (see Kempson et al. 2001: chapter 9; Cann et
al. 2007) The specifications object/subject/predicate-node are just schematic name
abbreviations to avoid the clutter of presenting actual DS-TTR step-by-step actions and modalities.
33
Some collections of sequences of actions are indicated as freely ordered or optional through
bracketing to account for variable word-orders.
However, given that even in such cases the conceptual representation contributed by
the quoted string is inevitably derived, the fact that the demonstrating event offers
anaphoric possibilities that can be exploited subsequently both supra-sententially,
subsententially, and across turns is a natural prediction:
(118) “I talk better English than the both of youse!” shouted Charles, thereby convincing me
that he didn’t. (Partee 1973)
(119) “Don’t worry, my boss likes me! He’ll give me a raise” said Mary, but given the
economic climate I doubt that he can. (Maier 2017, to appear)
(120) A: I talk better English than the both of youse!
B: You obviously don’t. (Partee 1973)
Instead of assuming that the availability of such anaphoric resolutions is the result
of presuppositional elements or implicatures (as in Maier 2014a), here the grammar
itself provides the resources for explaining the phenomena. As stated earlier, the
resolution of both ellipsis and pronominal anaphora in DS-TRR is assumed to
involve reuse of terms annotating CONTENT fields on treenodes, non-linguistically
provided content, or the rerunning of processing actions stored in the CONTEXT
(Eshghi et al. 2012; Kempson et al. 2015). Since the demonstrating event is
constituted by a set of such processing actions, and both the ensuing content and its
processing actions are not segregated from the rest of the conceptual representation,
they are stored in the context DAG and are available to be invoked for the resolution
of anaphoric and elliptical occurrences as in (118)–(120). For the same reason, as in
the G&C analysis, we can account for cases of “mixed predication” where both
token and type aspects are addressed simultaneously; recall (38)–(39), repeated
here:
(121) “Was I snoring” was asked by Bill and is a frequently used interrogative clause.
(122) Bill asked, “Am I snoring?”, a sentence frequently used by men who don’t think they
snore. It is usually answered by “You were before you woke up”.
But further than any other account, the present analysis extends to cases where the
continuation of an utterance started by an initial speaker without any quotational
intent can become quotational, i.e., treated as a demonstrating event (123)–(124);
and conversely, structures initiated without an already present reported event which
can be provided a quotational, echoing complement by the actions of another
speaker (125):
(123) Jem: Mary, whatever it is you think you know you mustn’t speak of it. Not if you want
to stay safe.
Mary: says the horse-thief [BBC Transcripts, Jamaica Inn, Episode 1]
(124) Miriam: That is the nastiest, dirtiest thing anyone has ever done
Patience: says Black Peter’s strumpet ! What are you crying for? [Jamaica Inn,
Episode 1]
(125) Noel: What I’m saying is
Stacey: you are IT!
Noel (ironically): Well, yeah...
(adapted from BBC Transcripts, Never Mind the Buzzcocks, 16/10/11)
In cases like (123)–(124), there is anaphoric use of the reported utterance event (that
is, the demonstrating event has already been performed instead of being executed
after it has been announced as the quotational cataphoric uses we have analysed
so far). In accordance with (117) earlier, the actions induced by say are executed
but the value for the metavariable U, the quoted event, is provided by appropriation
of the other speaker’s utterance that has just occurred, instead of being provided
via an independent demonstration on a LINKed node. This utterance also constrains
the value of the upcoming subject via the predicted unification with the speaker
(spkr) value of the reported event available in the CONTEXT part of the IF condition
(see also fn 32; this LINK-unenriched option also accounts for further variable [and
parenthetical] word-order patterns in direct reporting structures)34 :
(126) Lexical entry for say C direct report:
On the other hand, (125), repeated below, is an intermediate case of indirect report in
English where the complementiser that is missing, as shown by the intended values
of the indexicals35 :
(127) Noel: What I’m saying is
Stacey: you are IT! [‘Noel is IT’]
Noel (ironically): Well, yeah...
In such cases, the object node of the verb saying will be provided a value of type t
(Ty(t)), i.e. the type ‘propositional’, which in DS-TTR, is a complex record type
whose tn value, the final type derived, is t and does not carry any assertional
34
Note that due to the implementation of incremental licensing, parsing/generation in DS-TTR
can be initiated from any subpropositional stage, e.g. here starting with the requirement to build a
predicate (?Ty(e ! t)).
35
Complementisers in DS-TTR do not themselves contribute content that appears on treenodes,
they just execute procedural functions of introducing constraints on what can occupy nodes or
predictions of upcoming input.
implications (Gregoromichelaki 2006). Accordingly, this is what explains the

syntactic transparency of such structures (e.g. extraction possibilities; see earlier
(69)–(70)), as is usual in DS (see e.g. Kempson et al. 2001; Cann et al. 2005). This
will be ensured by means of the lexical entry for say that combines its content with
a propositional complement as shown below.
(128) Lexical entry for say C indirect report:
Simplifying the semantics for illustration purposes (see also Maier 2017 for a
similar formal implementation), here the situation derived as part of the content
of the embedded report (the witness of the proposition, the value to replace the
metavariable P) is constrained to exemplify the same type as the propositional
content (u.[CONTENT]) of some uttering event (u) by the contextually invoked
speaker (x, also the subject of the sentence) and to be part of all the worlds in the set
of worlds compatible with what this speaker said (thus instantiating the value of the
metavariable Wx ).36 However, unlike direct reports, this uttering event (u) does not
provide the contextual parameters for the report since the reporter’s utterance just
provides an interpretation of that uttering event. Notice though that, just like direct
reports, such structures can also felicitously embed (descriptions of) conversational
phenomena, e.g. repetition and abandoned sentential strings, which renders essential
their incremental licensing:
(129) I kept up, and anxious not to lose him, I said hurriedly that I couldn’t think of leaving
him under a false impression of my-of my-I stammered. The stupidity of the phrase
appalled me while I was trying to finish it, ... (Clark and Gerrig 1990, from Joseph
Conrad, Lord Jim)
Now returning to (125)/(127), processing the continuation accompanied with
context shift is unproblematic because each word micro-conversational event will
introduce its own contextual parameters, hence accounting for the resolution of
both the contents of I and you to same individual. The result will be a proposition
(‘Noel is “it”’) that matches the hypothesised utterance produced by Noel even
36
Further similarity requirements could be introduced following G&C’s definition of similarity
relations; the complications mentioned by Cappelen and Lepore (1997), regarding similarity of
content rather than replication of contents, could be implemented by loosening the same-type
restriction through appealing to the subtyping relation.
though this utterance has been produced by Stacey describing what Noel would
have said (the pragmatic effect being that Stacey only “pretends” that this is the
utterance that Noel would have produced, so, unlike genuine continuations, she
carries the responsibility for its content so Noel has to confirm it). Since, on the
surface, only content is relevant here, Stacey has to switch indexical when assuming
Noel’s speakership (see (83); cf. (31) and (113)). In modelling the processing of
this structure, we assume that what, which is taken as an anaphoric element in
DS-TTR, has introduced a metavariable for an event to be resolved cataphorically
(for other such grammaticalised cataphoric structures, see (91) earlier and Cann et
al. 2005; Gregoromichelaki 2013a). This metavariable will provide the temporary
place-holder for the binding of the rule-level variable u in the lexical action in (128).
This metavariable can eventually be provided with a proper value only after the
second speaker, Stacey, has uttered her part with subsequent appropriate resolution
of all remaining variables.
What allows the flexibility of such an account is the difference between this
approach and G&C’s, namely the fact that a monolithic utterance event is not
necessarily derived at once for the whole complement of the framing verb. Instead,
as the contextual parameters are reset at each micro-conversational event, there is
the possibility at each subsentential stage for the current speaker/hearer to switch.
For the same reason, the incrementality of DS-TTR also provides for the modelling
of the potential a speaker has, even during a non-shared utterance, to be able to shift
the default context and perform a demonstration. This is what accounts for both
cases of free (in)direct reports and mixed quotation as we are going to see in the
next section. In line with Recanati (2010), we can assume that standard uniform,
non-shared indirect reports are cases where the CONTEXT field values remain
constant throughout the utterance of both the reporting section of the sentence
and the reported-content part. As a consequence, indexical elements receive their
interpretations from the context established by the current utterance event Un .
However, as a consequence of the lexical action introduced by the framing verb, a
new possible world/time (or set of world/times) metavariable W is introduced for the
report to express the fact that it reflects the reportee’s view (see also Recanati 2000).
Such contextual and world/time parameters can be shifted independently of each
other, and the possibility of shifting world and context parameters (including time
of utterance) independently and incrementally, for each word-utterance (each micro-
conversational event; see earlier Sects. 3 and 7.22) as the utterance develops, models
the otherwise puzzling cooccurrences of transposed and untransposed indexicals
considered by Recanati (2000: chapters 15–16) and pronouns and tenses in various
intermediate cases of reporting (Eckardt 2014). Confirming the desirability of such
flexibility, notice the independently established fact (Gregoromichelaki et al. 2011)
that in cases of split reporting utterances, indexicals will acquire values according
to who currently assumes the relevant interlocutor roles (see also (31)):
(130) A: So you say you will live

B: by my pen, yes
(131) A: Did you say to Nick that : : :

B: you injured me? Yes, why? (‘A injured B’)
(132) A: Did you say to Nick that you : : :

B: injured myself? Yes, my doctor says so. (‘B injured B’)
As we said earlier in Sect. 7.1, the eventual representation derived, following
standard DS-TTR procedures, composes the contents derived at the various subsen-
tential stages, as well as recording the various concatenated u1 ,u2 : : : un subevents
that resulted in a (perhaps joint) utterance-event U D u1 ˚u2 ˚ : : : un . Hence the
interpretation derived eventually has the values of the indexicals as intended by the
participants at each previous processing stage in that their lexical actions have been
executed subsententially in line with the then-current context so that the eventual
composition deals with contents only. The fact that there is no level of syntactic
representation for the string of words makes utterances like (132) fully licensed as
joint utterances and provided with appropriate interpretations. Any other grammar
that insists on an independent syntactic analysis of such strings (e.g. Potts 2007;
Maier 2014a) will have trouble with such utterances, as the string of words Did you
say to Nick that you injured myself will have to be characterised as ungrammatical
(and for (130)–(131) it will derive the wrong interpretation).
7.3.3 Free (In)direct Discourse, Mixed Quotation and Scare Quoting
Essentially, along with Maier (2014a), the continuity of pure quotation, direct
reports and mixed quotation is also assumed here; however, in line with G&C, the
grammar does not need to implement this insight by employing special devices.
Unlike G&C, since DS-TTR does not impose a separate level of syntactic analysis
for the string of words, only the conceptual representation derived by processing
the string, there is no issue arising here in terms of characterising distinct syntactic
categories for indirect, direct, free, and mixed quotation structures in contrast to any
other grammatical analysis of quotation (also, in fact, contra Recanati 2000, 2010).
The only mechanism that is needed is the general mechanism in (95) that deals with
pure quotation cases potentially accompanied with the assumption that there is an
echoed event (utterance or thought):
(133) Echoing version of computational action for processing pure quotation with
derived content:
The difference between the macro in (133) and the one in (95) is that in (133)
the content derived by processing the demonstrating event (uq ), under a grammar
instantiating g potentially distinct from the current speaker’s grammar, and the
type of content derived by the echoed event (the instantiation of u) need to match
(as shown by shading). That derived content will occupy the current node under
processing, which can be of any type. Additionally, as an option, the contextual
parameters can be provided by the echoed event as in the intermediate echoic case
in (98) and the direct report cases in (117).
In consequence, to extend the coverage of the insights of the G&C account,
DS-TTR does not need to employ specific constructions to deal with separate
quotational phenomena, only mechanisms that can apply freely, combine with each
other, and interact with the context, while at the same time eschewing a syntactic
level of representation and definitions of abstract ‘expressions’ and ‘expression
types’. We now turn to the various remaining phenomena to exemplify briefly
these mechanisms in various combinations. In the case of free indirect discourse,
in addition to the free non-lexicalised introduction of an echoing demonstrating
event, with or without shift of grammar, there is also a (non-lexicalised) shift in the
CONTENT world parameter (as in the lexicalised option in (128); see Recanati 2000),
for example the event is taking place in a world/time index according to somebody’s
thoughts/beliefs (hence this view reconciles the Maier 2014b and Eckardt 2014
analyses):
(134) Mary felt relieved. If Peter came tomorrow, she would be saved. (Recanati 2000)
Since in DS-TTR these parameters are independent, there is the possibility for inde-
pendent shifting of world/time and CONTEXT parameters as required by particular
linguistic elements and the discourse context (for systematising the grammatical
constraints in this area, see e.g. Eckardt 2014). In DS-TTR, the eventual interpre-
tation emerges via the concatenation of utterance subevents which can define their
contexts independently of each other, corresponding to the sequential shifting in and
out of echoing demonstrations that the speaker performs. Due to this fine-grained
incrementality, there is no problem with having to coordinate the world/time and
context shifts. This account gives results similar to those of Maier (2017, to appear)
but without using ad hoc devices like the “unquotation” mechanism. The results just
follow from the incremental contextual licensing of structures and interpretation
that constitute independently the basis of the DS-TTR model. And, unlike other
grammatical analyses, e.g. G&C and Sharvit (2008), since there is no independent
level of syntactic analysis for the sentence, we do not have to license a complete
sentential string that has to be internally consistent as to indirect/direct report
features and contextual parameters (since, at the final stage, DS-TTR composes
contents and not Kaplanian “characters”; cf. Eckardt 2014). Accordingly, free direct
discourse (see (45) in Sect. 6) is simply a case where the CONTEXT parameters are
also shifted uniformly along with the world parameter.
In the cases of mixed quotation (seen earlier in (49) in Sect. 6) and hybrid cases,
there is no assumption here of any “verbatim requirement” (cf. Maier 2014a), so
no such difference with indirect discourse ensues. Additionally, as Recanati (2010)
has pointed out, the context might make it evident that the words of somebody
else rather than the subject of the framing verb are being echoed. It might also
be the case that nobody has in fact uttered those words (hence scare quoting is
not a separate phenomenon). Such cases can be adequately dealt with through the
processing macros either in (95) or in (133):
(135) Alice said that Clinton is “smooth”, as you would put it. Of course that’s not the word
SHE used. (Recanati 2010)
(136) These are not “I really should” radishes : : : . (Clark and Gerrig 1990)
(137) Dutch is a “that I him have helped” language. (Abbott 2005, from Philippe De
Brabanter)
We can also account for any “syntactic” binding effects in mixed quotation since
even in structures licensed through the lexical entry for verbs with an indirect report
complement, as shown in (128), the speaker, by employing in addition the actions in
(95) or (133) for part of the utterance, can freely shift in and out of a demonstration:
(138) John said that “the queen of each man’s heart” loves only herself. (Johnson 2011)
(139) Which houses did the FBI say they could “search without warrant”? (Johnson 2011)
Non-constituent mixed quotation does not present a fundamental problem for
this account either, since, by definition, the grammar incrementally licenses and
interprets word strings, without relying on what other grammars characterise as
“syntactic constituents” either subsententially or supra-sententially:
(140) She allowed as how her dog ate “strange things, when left to its own devices”. (Abbott
2005)
(141) Pascal suspected that the mercury was really supported by the “weight and pressure of
the air, because I consider them only as a particular case of a universal principle
concerning the equilibriums of fluids.” (Maier 2015)
(142) Also, he categorically stated that “there is no legal way of temporal extension of the
Greek debt without this being regarded as a credit event. Therefore there is no way
that it will be allowed to happen such a credit event in Greece because it would create
negative impact on the whole system.” [Cyprus Mail, 30/5/11]
But we can go even further than that to account for data that are completely
out of reach for other grammars. As we saw earlier (Sect. 7.3.1), given its psy-
cholinguistically inspired nature, the DS-TTR model records the various alternative
options arising during processing including those arising from the processing of
ambiguous strings. Even options less probabilistically favoured, and hence not
currently pursued, are stored temporarily in the context model (context DAG) in
order to be employed for, e.g., the functioning of repair processes, like corrections,
in dialogue (see e.g. Hough 2015; Eshghi et al. 2015). This independently needed
modelling allows us to capture the variable semantic-“constituency” ambiguity of
some mixed quotation strings and the ways they can be exploited by interlocutors,
for example in puns and jokes (as pointed out by Maier 2014a):
(143) The menu says that this restaurant serves “[breakfast] [at any time]” . . . [ so I ordered
[ French toast during the Renaissance] ]. (Maier 2014, from Steven Wright)
Due to the fine-grainedness of the individual DS-TTR mechanisms and the non-
differentiation of grammatical and pragmatic modes of processing, all the “pecu-
liarities” of mixed quotation presented in Maier (2014a) and others (e.g. see earlier
(138)–(139)) are eliminated here. This is because, in DS-TTR, there is no need to
license a level of syntactic constituency or any independent syntactic categories for
strings (see Gregoromichelaki in prep. for full formal implementation of particular
instances).37
8 Conclusion
The view of NLs as codes mediating a mapping between “expressions” and the
world has been abandoned here to give way to a view where utterances are seen
as goal-directed actions aimed at locally and incrementally altering the affordances
of the context for both one’s self and one’s interlocutors.38 As conceived in the
model presented here, such actions employ perceptual stimuli composed not only
of words and syntax but also of elements like visual marks and styles, prosody,
intonation and timing, gestures, facial expressions and gaze. All these aspects of the
stimuli serve as triggers for the invocation not only of conceptual contents but also
37
As an anonymous reviewer points out, there is potential for overgeneration in this overall
approach. However, in my view, this should be handled on a case-by-case basis, given observed par-
ticularities of languages and constructions, not as systemic architectural grammatical constraints
(unless there is solid evidence for the latter), which is what I am concerned with here.
38
Goal-directedness should not be construed as consciously or even subconsciously “intentional” in
the Gricean sense. All (subpersonal) DS-TTR grammatical operations are goal-directed in the sense
that predictions of the next perceptual input are system-generated and, accordingly, constrain which
input will be sought and how such input will be accommodated. For arguments against the Gricean
construal see Gregoromichelaki et al. (2011), Gregoromichelaki (2013b), Gregoromichelaki et al.
(2013b), and Pickering and Garrod (2004); see also Saka (2003/2005) for similar views regarding
the processing of quotation.
time/space/psychological perspectives, remembered experiences, feelings, attitudes,

beliefs, imagistic impressions etc. all of which constitute part of their “meaning”.
Thus, as part of this set, linguistic elements are not conceived as symbols and
operations arbitrarily related to their referents and semantics. Instead they are seen
as intrinsically linked to their phonetic or graphical realisations and the “meanings”
they activate through human categorisation processes. From this perspective, any
aspect of such stimuli can participate in the processes that constitute the “grammar”,
whose function is nothing else but the dynamic categorisation of various perceptual
inputs and their integration with memory and action schemata in the process of
generating the next action steps. This perspective does not allow for any process,
like the alleged operation of “quotation”, that segregates meaning from form,
“demonstration” from reference, or syntax/semantics from pragmatics. During
human interaction, due to the interlocutors’ (partially) shared experiences and goals,
perceptual inputs are able to trigger common action schemata, event invocations,
and associations thus becoming the basis of joint performance coordination via
the intersubjective affordances that they make available. From this point of view,
linguistic knowledge is part of the abilities to coordinate effective interaction with
the environment, one’s own self, or one’s interlocutors. In particular contexts,
some of the various affordances that linguistic stimuli give access to will be more
relevant than others in order to locally coordinate effective responses. Reporting,
echoing, citing or metacommenting on aspects of the process itself are means
through which some of the various aspects of meaningfulness can be foregrounded
in the service of facilitating joint performance. It is not curious then that quotation
bears common features with conversational phenomena: under the present view
this is because it employs the same mechanisms as conversation, and consequently
quotation is expected to interact with such conversational phenomena, e.g. repair
and shared utterances, which also facilitate coordination. DS-TTR, in taking a
psycholinguistically realistic action-grounded view of grammar, aims to model
these interactions by subsuming quotation phenomena in a unified framework under
general conversational coordinative mechanisms.39
References
Abbott, B. (2005). Some notes on quotation. In De Brabanter (2005b, 13–26).

Allen, J., Ferguson, G., & Stent, A. (2001, January). An architecture for more realistic conver-
sational systems. In Proceedings of the 2001 international conference on Intelligent User
Interfaces (IUI).
39
I wish to thank all my collaborators to the DS-TTR project: Ruth Kempson, Ronnie Cann,
Stelios Chatzikyriakidis, Arash Eshghi, Pat Healey, Julian Hough, Chris Howes, Greg Mills,
Matt Purver, and Graham White. I am especially grateful to Paul Saka for various suggestions,
comments, invaluable editorial assistance and tremendous support. In addition, I am very thankful
to an anonymous reviewer for insightful comments, data, and further sources of information. I
acknowledge support from the ESRC (Grant ESRC-RES-062-23-0962).
Anand, P., & Nevins, A. (2003). Shifty operators in changing contexts. In Proceedings of semantics
and linguistic theory (pp. 20–37). Ithaca: CLC Publications.
Antaki, C., Diaz, F., & Collins, A. F. (1996). Keeping your footing: Conversational completion in
three-part sequences. Journal of Pragmatics, 25(2), 151–171.
Bateson, G. (1972). Steps to an ecology of mind: Collected essays in anthropology, psychiatry,
evolution, and epistemology. Chicago: University of Chicago Press. 1987 \mboxfISBN 0-
87668-950-0.g
Bakhtin, M. M. (1981). The dialogic imagination: Four essays. In M. Holquist (Ed.). (C. Emerson
& M. Holquist, Trans.). Austin: University of Texas Press.
Banfield, A. (1973). Narrative style and the grammar of direct and indirect speech. Foundations of
Language, 10(1), 1–39.
Barwise, J., & Perry, J. (1983). Situations and attitudes. Cambridge, MA: MIT Press/Bradford.
Bonami, O., & Godard, D. (2008). On the syntax of direct quotation in French. In S. Mueller (Ed.),
Proceedings of the HPSG08 conference (pp. 358–377). Stanford: CSLI Publications.
Brendel, E., Meibauer, J., & Steinbach, M. (2011). Introduction. In E. Brendel, J. Meibauer, & M.
Steinbach (Eds.), Understanding quotation. Berlin: Mouton de Gruyter.
Brandom, R. (1994). Making it explicit: Reasoning, representing, and discursive commitment.
Cambridge, MA: Harvard University Press.
Cann, R., Kempson, R., & Marten, L. (2005). The dynamics of language. Oxford: Elsevier.
Cann, R., Kempson, R., & Purver, M. (2007). Context and well-formedness: The dynamics of
ellipsis. Research on Language and Computation, 5(3), 333–358.
Cappelen, H., & Lepore, E. (2007). Language turned on itself. The semantics and pragmatics of
Chater, N., Pickering, M., & Milward, D. (1995). What is incremental interpretation? Incremental
interpretation, Edinburgh Working Papers in Cognitive Science, 11.
Clark, H. H. (1996). Using language. Cambridge: Cambridge University Press.
Clark, H. H., & Fox Tree, J. E. (2002). Using uh and um in spontaneous speech. Cognition, 84,
73–111.
Clark, H. H., & Gerrig, R. J. (1990). Quotations as demonstrations. Language, 66, 764–805.
Cooper, R. (2005). Records and record types in semantic theory. Journal of Logic and Computa-
tion, 15(2), 99–112.
Cooper, R. (2012). Type theory and semantics in flux. In R. Kempson, N. Asher, & T. Fernando
(Eds.), Handbook of the philosophy of science, vol 14: Philosophy of linguistics (pp. 271–323).
Amsterdam: Elsevier.
Cooper, R., & Ranta, A. (2008). Natural languages as collections of resources. In R. Cooper & R.
Kempson (Eds.), Language in flux (pp. 109–120). London: College Publications.
Crystal, D. (2013). http://www.davidcrystal.community.librios.com/?idD2914. Accessed 13/1/15.
Cumming, S. (2005). Two accounts of indexicals in mixed quotation. In De Brabanter (2005b,
77–88).
Davidson, D. (1984). Inquiries into truth and interpretation. Oxford: Clarendon Press.
De Brabanter, P. (2005a). Introduction. In De Brabanter (2005b, 1–2).
De Brabanter, P. (Ed.). (2005b). Hybrid quotations. Amsterdam: John Benjamins.
De Brabanter, P. (2010a). The semantics and pragmatics of hybrid quotations. Language and
Linguistics Compass, 4(2). https://doi.org/10.1111/j.1749-818X.2009.00185.x.
De Brabanter, P. (2010b). Uttering sentences made up of words and gestures. In B. Soria & E.
Romero (Eds.), Explicit communication (pp. 199–216). London: Palgrave Macmillan.
De Brabanter, P. (2013). Francois Recanati’s radical pragmatic theory of quotation. Teorema, 32,
109–128.
De Ruiter, J., Mitterer, H., & Enfield, N. J. (2006). Projecting the end of a speakers turn: A cognitive
cornerstone of conversation. Language, 82(3), 515–535.
Duff, M. C., Hengst, J. A., Tranel, D., & Cohen, N. J. (2007). Talking across time: using reported
speech as a communicative resource in amnesia. Aphasiology, 21(6–8), 702–716.
Eckardt, R. (2014). The semantics of free indirect discourse: How texts allow us to mind-read and
eavesdrop. Leiden: Brill.
Enç, M. (1986). Towards a referential analysis of temporal expressions. Linguistics and Philoso-
phy, 9, 405–426.
Eshghi, A., Healey, P. G. T., Purver, M., Howes, C., Gregoromichelaki, E., & Kempson, R..
(2010, September). Incremental turn processing in dialogue. In Proceedings of the 16th annual
conference on Architectures and Mechanisms for Language Processing (AmLAP), York, UK.
Eshghi, A., Purver, M., & Hough, J. (2011). Dylan: Parser for dynamic syntax. Technical report,
Queen Mary, University of London.
Eshghi, A., Hough, J., Purver, M., Kempson, R., & Gregoromichelaki, E. (2012). Conversational
interactions: Capturing dialogue dynamics. In L. Borin & S. Larsson (Eds.), From quantifica-
tion to conversation. London: College Publications.
Eshghi, A., Howes, C., Gregoromichelaki, E., Hough, J., & Purver, M. (2015, April). Feedback
in conversation as incremental semantic update. In: Proceedings of the 11th international
conference on computational semantics (IWCS 2015) (pp. 261–271). Queen Mary University
of London, UK
Gargett, A., Gregoromichelaki, E., Howes, C., & Sato, Y. (2008). Dialogue-grammar correspon-
dence in dynamic syntax. In Proceedings of the 12th SemDial (LonDial), London.
Gargett, A., Gregoromichelaki, E., Kempson, R., Purver, M., & Sato, Y. (2009). Grammar
resources for modelling dialogue dynamically. Journal of Cognitive Neurodynamics, 3(4), 347–
363.
Geach, P. (1957). Mental acts. London: Routledge Kegan Paul.
Geurts, B, & Maier, E. (2005). Quotation in context. In De Brabanter (2005b, 109–28).
Gibson, J. J. (1977). The theory of affordances. In R. E. Shaw & J. Bransford (Eds.), Perceiving,
acting, and knowing. Hillsdale: Lawrence Erlbaum Associates.
Ginzburg, J. (2012). The interactive stance: Meaning for conversation. Oxford: Oxford University
Press.
Ginzburg, J., & Cooper, R. (2004). Clarification, ellipsis, and the nature of contextual updates in
dialogue. Linguistics and Philosophy, 27(3), 297–365.
Ginzburg, J., & Cooper, R. (2014). Quotation via dialogical interaction. Journal of Logic,
Language and Information, 23(3), 287–311.
Ginzburg, J., Fernández, R., & Schlangen, D. (2014). Disfluencies as intra-utterance dialogue
moves. Semantics and Pragmatics, 7(9), 1–64.
Goffman, E. (1979). Footing. Semiotica, 251–2, 1–30.
Goodwin, C. (1981). Conversational organization: interaction between speakers and hearers. New
York: Academic.
Goodwin, C. (2007). Interactive footing. In E. Holt & R. Clift (Eds.), Reporting talk: Reported
speech in interaction (pp. 16–46). Cambridge: Cambridge University Press.
Gregoromichelaki, E. (2006). Conditionals in dynamic syntax. PhD thesis, University of London.
Gregoromichelaki, E. (2011). Conditionals in dynamic syntax. In R. Kempson, E. Gre-
goromichelaki, & C. Howes (Eds.), The dynamics of lexical interfaces. Stanford: CSLI
Publications.
Gregoromichelaki, E. (2012). Review of J. Ginzburg (2012) The interactive stance. Folia Linguis-
tica, 47(1), 93–316.
Gregoromichelaki, E. (2013a). Clitic left dislocation and clitic doubling: A dynamic perspective
on left-right asymmetries in Greek. In G. Webelhuth, M. Sailer, & H. Walker (Eds.), Rightward
movement in a comparative perspective. Amsterdam: John Benjamins.
Gregoromichelaki, E. (2013b). Grammar as action in language and music. In M. Orwin, C. Howes,
& R. Kempson (Eds.), Language, music and interaction. London: College Publications.
Gregoromichelaki, E., & Kempson, R. (2013). The role of intentions in dialogue processing. In A.
Capone, F. Lo Piparo, & M. Carapezza (Eds.), Perspectives on linguistic pragmatics. Cham:
Springer.
Gregoromichelaki, E., & Kempson, R. (2015). Joint utterances and the (split-) turn-taking puzzle.
In Interdisciplinary studies in pragmatics, culture and society (pp. 703–743). Springer.
Gregoromichelaki, E., & Kempson, R. (2016). Reporting, dialogue, and the role of grammar. In
Indirect reports and pragmatics (pp. 115–150). Springer.
Gregoromichelaki, E., & Kempson, R. (2017). Procedural syntax. In R. Carston, B. Clark, & K.
Scott (Eds.), tba Cambridge University Press.
Gregoromichelaki, E., Sato, Y., Kempson, R. Gargett, A., & Howes, C. (2009). Dialogue modelling
and the remit of core grammar. In Proceedings of IWCS, Eighth international conference on
computational semantics.
Gregoromichelaki, E., Kempson, R., Purver, M., Mills, G. J., Cann, R., Meyer-Viol, W., & Healey,
P. G. T. (2011). Incrementality and intention-recognition in utterance processing. Dialogue and
Discourse, 2(1), 199–233. special issue on Incremental Processing in Dialogue.
Gregoromichelaki, E., Kempson, R., & Cann, R. (2012). Language as tools for interaction:
grammar and the dynamics of ellipsis resolution. The Linguistic Review, 29(4), 563–584.
Gregoromichelaki, E., Cann, R., & Kempson, R. (2013a). Coordination in dialogue: subsentential
speech and its implications. In L. Goldstein (Ed.), Brevity. Oxford: Oxford University Press.
Gregoromichelaki, E., Kempson, R., Howes, C., & Eshghi, A. (2013b). On making syntax
dynamic: The challenge of compound utterances and the architecture of the grammar. In I.
Wachsmuth, J. P. de Ruiter, P. Jaecks, & S. Kopp (Eds.), Alignment in communication: Towards
a new theory of communication. Amsterdam: John Benjamins.
Grimshaw, A. D. (1987). Finishing other’s talk: Some structural and pragmatic features of
completion offers. In R. Steele & T. Threadgold (Eds.), Language topics: Essays in honor
of Michael Halliday (pp. 213–235). Amsterdam: John Benjamins.
Horn, L. (1989). A natural history of negation. Chicago: University of Chicago Press.
Hough, J. (2015). Modelling incremental self-repair processing in dialogue. PhD thesis, Queen
Mary, University of London.
Huddleston, R., & Pullum, G. (2002). The Cambridge grammar of the English language.
Cambridge: Cambridge University Press.
Jaszczolt, K. M. (2005). Default semantics: Foundations of a compositional theory of acts of
Johnson, M. (2011). The punctuation theory of quotation. In E. Brendel, J. Meibauer, & M.
Steinbach (Eds.), Understanding quotation. Berlin: Mouton de Gruyter.
Kamp, H. (1981). A theory of truth and semantic interpretation. In J. Groenendijk, T. Janssen,
& M. Stokhof (Eds.), Formal methods in the study of language (pp. 277–322). Amsterdam:
Mathematical Centre Tracts.
Kamp, H., & Reyle, U. (1993). From discourse to logic. Dordrecht: Kluwer.
Kempson, R., Meyer-Viol, W., & Gabbay, D. (2001). Dynamic syntax: The flow of language
understanding. Oxford: Blackwell.
Kempson, R., Gregoromichelaki, E., Purver, M., Mills, G. J., Gargett, A., & Howes, C. (2009a).
How mechanistic can accounts of interaction be? In Proceedings of Diaholmia, the 13th
workshop on the semantics and pragmatics of dialogue.
Kempson, R., Gregoromichelaki, E., & Sato, Y. (2009b). Incrementality, speaker/hearer switching
and the disambiguation challenge. In Proceedings of European Association of Computational
Linguistics (EACL). Athens.
Kempson, R., Gregoromichelaki, E., & Howes, C. (2011a). The dynamics of lexical interfaces.
Chicago: CSLI Publications.
Kempson, R., Gregoromichelaki, E., Meyer-Viol, W., Purver, M., White, G., & Cann, R. (2011b).
Natural-language syntax as procedures for interpretation: The dynamics of ellipsis construal. In
Proceedings of the PRELUDE workshop on games, dialogues and interaction. Lecture Notes
in Computer Science. Berlin: Springer.
Kempson, R., Gregoromichelaki, E., & Chatzikyriakidis, S. (2012, April 26–27). Joint utterances
in Greek: Their implications for linguistic modelling. In Proceedings of 33rd annual linguistics
meeting “Syntactic Theories and the Syntax of Greek”. Thessaloniki
Kempson, R., Cann, R., Eshghi, A., Gregoromichelaki, E., & Purver, M.. (2015). Ellipsis. In S.
Lappin & C. Fox (Eds.), Handbook of contemporary semantics (2nd ed., pp. 114–140).
Kempson, R., Cann, R., Gregoromichelaki, E., & Chatzikyriakidis, S. (2016). Language as mech-
anisms for interaction. Theoretical Linguistics, 42(3–4), 203–276. https://doi.org/10.1515/tl-
2016-0011.
Kempson, R., Cann, R., Gregoromichelaki, E., & Chatzikyriakidis, S. (2017). Action-based
grammar. Theoretical Linguistics, 43(1–2), 141–167.
Koder, F., & E. Maier (2014). Children mix direct and indirect speech: Evidence from pronoun
comprehension. Ms. University of Groningen.
Larsson, S. (2011, September). The TTR perceptron: Dynamic perceptual meanings and semantic
coordination. In Proceedings of the 15th workshop on the semantics and pragmatics of dialogue
(pp. 140–148). SemDial 2011 – Los Angelogue.
Larsson, S. (2015). Formal semantics for perceptual classification. Journal of Logic and Computa-
tion, 25(2), 335–369.
Larsson, S., & Traum, D. R. (2000). Information state and dialogue management in the TRINDI
dialogue move engine toolkit. Natural Language Engineering, 6(3&4), 323–340.
Lerner, G. H. (1991). On the syntax of sentences-in-progress. Language in Society, 20, 441–458.
Lerner, G. H. (1996). On the “semi-permeable” character of grammatical units in conversation:
conditional entry into the turn space of another speaker. In E. Ochs, E. A. Schegloff, & S.
Thompson (Eds.), Interaction and grammar (pp. 238–276). Cambridge: Cambridge University
Press.
Maier, E. (2007). Mixed quotation: Between use and mention. In Proceedings of LENLS 2007,
Miyazaki, Japan.
Maier, E. (2014a). Mixed quotation: The grammar of apparently transparent opacity. Semantics &
Pragmatics, 7(7), 1–67.
Maier, E. (2014b). Language shifts in free indirect discourse. Journal of Literary Semantics, 43(2),
143–167.
Maier, E. (2015). Quotation and unquotation in free indirect discourse. Mind & Language, 30(3),
345–373. https://doi.org/10.1111/mila.12083.
Maier, E. (2017). Mixed quotation. Ms. University of Groningen (survey article for The Companion
to Semantics edited by Lisa Matthewson, Cécile Meier, Hotze Rullmann, Thomas Ede
Zimmermann, Wiley/Blackwell, New York.)
Marslen-Wilson, W. (1973). Linguistic structure and speech shadowing at very short latencies.
Nature (London), 244, 522–523.
McCawley, J. D. (1999). Participant roles, frames, and speech acts. Linguistics and Philosophy,
22(6), 595–619.
McCloskey, J. (2006). Questions and questioning in a local English. In R. Zanuttini, H. Campos, E.
Herburger, & P. H. Portner (Eds.), Cross-linguistic research in syntax and semantics: Negation,
tense and clausal architecture (pp. 87–126). Washington, DC: Georgetown University Press.
Morris, P. (1997). The Bakhtin reader: Selected writings of Bakhtin, Medvedev, Voloshinov.
Oxford: Oxford University Press.
Noh, E.-J. (1998). Echo questions: Metarepresentation and pragmatic enrichment. Linguistics and
Norrick, N. (2015). Indirect reports, quotation, and narrative. In A. Capone, F. Kiefer, & F. Lo
Piparo (Eds.), Indirect reports and pragmatics. Dordrecht: Springer.
Partee, B. (1973). The syntax and semantics of quotation. In S. R. Anderson & P. Kiparsky (Eds.),
A Festschrift for Morris Halle (pp. 410–418). New York: Holt, Rinehart, and Winston.
Pickering, M., & Garrod, S. (2004). Toward a mechanistic psychology of dialogue. Behavioral and
Brain Sciences, 27(02), 169–190.
Pickering, M., & Garrod, S. (2013). An integrated theory of language production and comprehen-
sion. Behavioral and Brain Sciences, 36(4), 329–347.
Poesio, M., & Rieser, H. (2010). Completions, coordination, and alignment in dialogue. Dialogue
and Discourse, 1(1), 1–89.
Poesio, M., & Traum, D. R. (1997). Conversational actions and discourse situations. Computa-
tional Intelligence, 13(3), 309–347.
Poesio, M., & Traum, D. (1998). Towards an axiomatization of dialogue acts. In Proceedings of the
twente workshop on the formal semantics and pragmatics of dialogues (13th twente workshop
on language technology).
Postal, P. M. (2004). Sceptical linguistic essays. Oxford University Press.
Potts, C. (2007). The dimensions of quotation. In P. Jacobson & C. Barker (Eds.), Direct
Purver, M., Cann, R., & Kempson, R. (2006). Grammars as parsers: Meeting the dialogue chal-
lenge. Research Language Computation, 4, 289. https://doi.org/10.1007/s11168-006-9007-x
Purver, M., Howes, C., Gregoromichelaki, E., & Healey, P. G. T. (2009). Split utterances in
dialogue: A corpus study. In Proceedings of SigDial. London.
Purver, M., Gregoromichelaki, E., Meyer-Viol, W., & Cann, R.. (2010). Splitting the I’s and
crossing the you’s: Context, speech acts and grammar. In Proceedings of SemDial 2010
(PozDial), Poznan, Poland.
Purver, M., Eshghi, A., & Hough, J. (2011). Incremental semantic construction in a dialogue
system. In Proceedings of the 9th international conference on computational semantics
(IWCS). Oxford.
Quine, W. V. O. (1940). Mathematical logic. Cambridge, MA: Harvard University Press.
Ranta, A. (1994). Type-theoretical grammar. Oxford: Clarendon Press.
Recanati, F. (2000). Oratio Recta, Oratio Obliqua. Cambridge, MA: MIT Press.
Recanati, F. (2004). Literal meaning. Cambridge: Cambridge University Press.
Recanati, F. (2010). Truth-conditional pragmatics. Oxford: Oxford University Press.
Ross, J. R. (1970). Metalinguistic anaphora. Linguistic Inquiry, 1, 273.
Kempson, R., Gregoromichelaki, E., Eshghi, A., & Hough, J.. (to appear) Ellipsis in Dynamic
Syntax. In J. van Craenenbroeck & T. Temmerman, (Eds.), The Oxford handbook of ellipsis.
ISBN: 9780198712398
Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the organization of
turn-taking for conversation. Language, 50, 696–735.
Sag, I. A., Wasow, T., & Bender, E. M. (2003). Syntactic theory. Stanford: CSLI Publications.
Saka, P. (1998). Quotation and the use-mention distinction. Mind, 107(425), 113–135.
Saka, P. (2003/2005). Quotational constructions. Belgian Journal of Linguistics, 17(1), 187–212.
Saka, P. (2011). Quotation and conceptions of language. Dialectica, 65(2), 205–220.
Saka, P. (2013). Quotation. Philosophy Compass, 8(10), 935–949.
Sato, Y. (2011). Local ambiguity, search strategies and parsing in dynamic syntax. In Kempson et
al. (2011b) (Eds.).
Schlenker, P. (2011). Indexicality and de se reports. In C. Maienborn, K. von Heusinger, & P.
Portner (Eds.), Semantics: An international handbook of natural language meaning (Vol. 2, pp.
1561–1604). Berlin: Mouton de Gruyter.
Shan, C.-c. (2011). The character of quotation. Linguistics and Philosophy, 33(5), 417–443.
Sharvit, Y. (2008). The puzzle of free indirect discourse. Linguistics and Philosophy, 31(3), 353–
395.
Slama-Cazacu, T. (1976). Nonverbal components in message sequence: ‘mixed syntax’. In W. C.
McCormack & S. A. Wurm (Eds.), Language and man: Anthropological issues (pp. 217–227).
The Hague: Mouton.
Steedman, M. (1992). Grammars and processors (Tech. Rep. MS-CIS-92-52). University of
Pennsylvania, Department of Computer and Information Science.
Stone, M. (2004). Intention, interpretation and the computational structure of language. Cognitive
Science, 28(5), 781–809.
Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K. M., & Sedivy, J. C. (1995). Integration of
visual and linguistic information in spoken language comprehension. Science, 268, 1632–1634.
Tannen, D. (1986). Introducing constructed dialogue in Greek and American conversational and
literary narratives. In F. Coulmas (Ed.), Direct and indirect speech (pp. 311–322). Berlin:
Mouton.
Tarski, A. (1993). The concept of truth in formalized languages. In A. Tarski (Ed.), Logic,
semantics, metamathematics (2nd ed., pp. 152–278). Indianapolis: Hackett.
Washington, C. (1992). Quotation. Journal of Philosophy, 89, 582–605.
Part III
New Directions
The Pragmatics of Attraction
Explaining Unquotation in Direct and Free Indirect
Discourse
Emar Maier
Abstract The quotational theory of free indirect discourse postulates that pronouns
and tenses are systematically unquoted. But where does this unquotation come
from? Based on cases of apparent unquotation in direct discourse constructions
(including data from Kwaza speakers, Catalan signers, and Dutch children), I
suggest a general pragmatic answer: unquotation is essentially a way to resolve a
conflict that arises between two opposing constraints. On the one hand, the reporter
wants to use indexicals that refer directly to the most salient speech act participants
and their surroundings (ATTRACTION). On the other hand, the semantics of direct
discourse (formalized here in terms of event modification) entails the reproduction
of referring expressions from the original utterance being reported (VERBATIM).
Unquotation (formalized here also in terms of event modification), allows the
reporter to avoid potential conflicts between these constraints. Unquotation in free
indirect discourse then comes out as a special case, where the salient source of
attraction is the story protagonist and her actions, rather than the reporting narrator
and his here and now.
1 Introduction: Reporting Speech and Thought
1.1 The Direct–Indirect Distinction
Many languages offer two basic options for reporting what someone said or thought.
If we want to report just the content of what someone said or thought, we choose an
indirect discourse construction.
(1) a. Mary said that she would never forgive me.
b. He thought that his heart was broken.
E. Maier ()
University of Groningen, Oude Boteringestraat 52, 9712GL Groningen, The Netherlands
e-mail: emar.maier@gmail.com

https://doi.org/10.1007/978-3-319-68747-6_9
260 E. Maier
On the other hand, if we want to report the actual words uttered (either out loud, or
sub voce, in the case of a thought), we go for a direct discourse construction.
(2) a. Mary said, “Oh no, I’ll never forgive you”
b. “My heart’s broken,” he thought. “If I feel this way my heart must be
broken.”
The two basic reporting modes can be distinguished semantically, and, in many
languages, also syntactically. In the grammar of English, for instance, indirect
discourse involves a complement clause embedded under a verb of saying or
thinking, while direct discourse offers the report as an independent main clause
with a non-subordinating, often parenthetical, reporting frame ( she said/thought).
Semantically, pronouns, tenses, and other indexicals in an English indirect discourse
complement are interpreted with respect to the actual reporting utterance context.
The first person pronoun me in (1a) refers to me, the reporter. Indexicals in direct
discourse, by contrast, are ‘shifted’, i.e., interpreted with respect to the context of
the reported utterance. The I in (2a) does not refer to me but to Mary, the reported
speaker.
Interestingly, these are not the only two modes of reporting. Various construc-
tions in different languages do not fall neatly into either category. In this paper I
will discuss various apparent exceptions to the general rule that indexicals in direct
discourse are shifted. My main focus is on free indirect discourse, which I introduce
in the next subsection.
1.2 The Puzzle of Unquotation in Free Indirect Discourse
Free indirect discourse is a form of reporting speech or thought, characteristic of

narrative contexts (Banfield 1973; Fludernik 1995). Consider the following report
of Mary’s thoughts as she is packing her bags.
(3) Mary was packing her bags. Tomorrow was her last day. Oh how happy she
would be to finally walk out of here. To leave this godforsaken place once
and for all.
Free indirect discourse shares characteristics with both direct and indirect dis-
course.1 Syntactically, free indirect discourse patterns with direct discourse in that
the report clauses are independent main clauses rather than embedded clauses.
Semantically, free indirect discourse displays one of the most salient characteristics
of indirect discourse: all pronouns and tenses are interpreted from the reporting
perspective. However, other context dependent expressions (e.g., tomorrow, here,
1
The term ‘free indirect discourse’ is misleading. Free indirect discourse is not really a species
of indirect discourse (Banfield 1973; Maier 2015). In fact, as we will see, it is closer to direct
discourse, both semantically and syntactically. According to Banfield, the term ‘free’ refers to the
fact that the report clause is not syntactically subordinated.
The Pragmatics of Attraction 261
and this in (3)) are interpreted as in direct discourse, i.e. evaluated relative to the
protagonist’s perspective.
Free indirect discourse is thus truly a hybrid of direct and indirect speech. Given
that we have a pretty good understanding of the semantics of both direct and indirect
discourse, this raises the question if we can reduce free indirect discourse to one of
these basic modes. Since Schlenker’s (2003) introduction of monsters (i.e. context
shifting operators) in indirect discourse, the go-to approach in semantics has been
to try and analyze free indirect discourse as a special kind of indirect discourse
(Schlenker 2004; Sharvit 2008; Eckardt 2014). I argue for the opposite reduction:
free indirect discourse is a species of direct discourse, involving genuine quotation.
In earlier work I have shown that free indirect discourse exhibits some key
characteristics of quotation that cannot be accounted for by existing context shift
analyses (Maier 2015, 2014a). In particular, free indirect discourse, like direct
discourse, allows the reporter (the narrator) to slip into the language, dialect, or
idiolect of the reported speaker/thinker (the protagonist):
(4) Ah well, her fathaire would shoorly help her out, she told John in her thick
French accent.
To account for such data I offered an alternative semantic analysis in which free
indirect discourse is like direct discourse but with holes for the pronouns and tenses.
More precisely, the truth conditions of (3) can be schematically represented as in (5),
in which the quotation marks indicate regular direct discourse and square brackets
indicate “unquotation”.
(5) Mary was packing her bags. “Tomorrow [was] [her] last day. Oh how happy
[she] [would] be to finally walk out of here. To leave this godforsaken place
once and for all.”
Given any reasonable semantics of quotation and unquotation (such as the one
developed in Sect. 2 below), logical forms like these get the truth conditions right.
But this approach does raise a few concerns. Let’s start by addressing the least
worrying ones, viz. that it appeals to invisible quotation and unquotation operators.
First, an appeal to covert quotation is not particularly problematic, as regular direct
discourse is also often left unmarked, for instance in colloquial spoken English,
but even in writing, particularly with direct thought reports. In addition, some
early occurrences of free indirect discourse are in fact overtly marked by quotation
marks2 :
(6) choose an indirect discourse construction.igid requisitions might have been
taken, is of little consequence. Lady Russell’s had no success at all – could
not be put up with – were not to be borne. “What! Every comfort of life
knocked off! Journeys, London, servants, horses, table, – contractions and
2
Helen de Hoop p.c. has collected similar examples from the early Dutch epistolary novel Sara
Burgerhart by Aagje Deken and Betje Wolff, 1782.
262 E. Maier
restrictions every where. To live no longer with the decencies even of a

private gentleman! No, he would sooner quit Kellynch-hall at once, than
remain in it on such disgraceful terms.” ŒJane Austen Persuasion.1817/
Likewise, covert unquotation needn’t worry us either. Shan (2011) and Maier
(2014b) describe various cases of overt mixed quotation where, arguably, some
expressions must be interpreted as covertly unquoted. In Sect. 3 we’ll see many
concrete examples of covert unquotation in direct discourse.
The remaining, most serious problem for the quotation approach is that it
overgenerates. Since, as far as syntax and semantics are concerned, any constituent
in an overt quotation can be unquoted, it remains a mystery why in free indirect
discourse all and only the pronouns and tenses get unquoted. I call this the puzzle
of unquotation in free indirect discourse.
In the current paper I try to solve this puzzle. I will show that the apparent
unquotation restriction to pronouns and tenses in free indirect discourse is not
an isolated phenomenon to be stipulated to get the truth conditions right, but an
instance of a much more general pattern. Following the terminology of Evans (2012)
I propose to explain the observed unquotation patterns in direct and free indirect
discourse in terms of attraction, a pragmatic mechanism that can be used to describe
a wide variety of seemingly unrelated data concerning direct speech in Kwaza, role
shift and pointing in sign languages, and children’s interpretation of pronouns in
direct speech.
Before discussing the pragmatics of unquotation in free indirect discourse and
beyond, I will first present a concrete semantic account of quotation and unquotation
in which to frame the pragmatic discussion more effectively.
2 The Semantics of Quotation and Unquotation
2.1 Direct and Indirect Discourse as Event Modification
The standard account of indirect discourse in formal semantics is that developed

by Kaplan (1989). Kaplan sketches a way of analyzing indirect discourse as an
intensional operator within his two-dimensional Logic of Demonstratives. The idea
is that my utterance of (1a) (Mary said that she’d never forgive me) is true iff
there was an earlier speech act in which Mary said something that expressed that
she’d never forgive me. In other words, the proposition expressed by the report’s
complement clause, as uttered by me in the current context, gives us the content
(but not the form) of the original utterance event.
One way to make this precise is by assuming that a speech act is an event, which
can have an agent and a duration (as usual in neo-Davidsonian event semantics),
but also a linguistic form and a propositional content. We can then treat indirect
discourse reports as asserting that there was an utterance event e whose agent
(agent.e/) is given by the subject term and whose content (content.e/) is given by
the intension of the complement clause.
(7) jjMary said that she’d never forgive mejj = 9eŒsay.e/ ^ agent.e/ D jjMaryjj
^ time.e/ < now ^ content.e/ D ^ jjshe’d never forgive mejj
First, some general notes on the semantic framework and notation I’ll be using.
I’m assuming a traditional Montagovian framework, with a recursive translation
mechanism, jj jj, mapping syntactically well-formed expressions to expressions
in an interpreted formal metalanguage (a higher-order, intensional, typed lambda
calculus with indexicals). For instance, jjMaryjj D mary (a constant of type
e), and jjsayjj D pst xeŒsay.e/ ^ agent.e/ D x ^ content.e/ D p.3 Well-formed
expressions of this formal language can be interpreted in a model (relative to an
assignment f , a Kaplanian context c, and a world w), notated as wf ;c . For instance,
^ ˛wf ;c D w: ˛wf ;c and iwf ;c D the agent/speaker of c. In this paper I’m interested
primarily in the semantics–pragmatics interface, so I will content myself with just
presenting the logical forms of entire sentences, without spelling out the underlying
assumptions concerning syntax, translation, and model-theoretic interpretation.4
On a conceptual level, the idea behind content.e/ is that certain eventualities,
like say-events and belief-states, have a propositional content (Hacquard 2010). For-
mally, content denotes a function from such contentful eventualities to propositions
– in this case mapping an utterance event e to what was said in e.5
This event modification approach to indirect discourse has a number of advan-
tages over the traditional intensional operator approach. First and foremost is
that it extends straightforwardly to direct discourse. Where indirect discourse
complements specify the content of the original speech act, direct discourse specifies
its linguistic surface form.6
3
I also assume existential closure of event lambdas left over at sentence level (as usual in event
semantics), and (the translational analogues of) rules like predicate modification and intensional
function application (Heim and Kratzer 1998).
4
In most cases, filling in the gaps is a straightforward exercise, but in others, some additional
research is needed. I leave this for another occasion.
5
For those skeptical of content as a theoretical primitive, note that an utterance event e occurring
in a world w at time t quite naturally determines a unique Kaplanian context of utterance
(context.e/ WD hagent.e/; t; wi), and also a Kaplanian character (char.e/ WD the character of the
sentence uttered in e. Hence, the skeptic may understand content.e/ as char.e/.context.e// (for any
utterance event e at a given time and world).
6
The equality sign in (8) may be an oversimplification. Direct discourse rarely provides a literal
copy of the original (cf. e.g., Wade and Clark 1993; Saka 2005). We could model this flexibility of
direct discourse by substituting a context-dependent relation of sufficient similarity for the equality
of forms here (cf. Bonami and Godard 2008; Maier 2014b). In addition, the current analysis does
not account for the apparent transparency of direct speech with respect to anaphora and ellipsis.
I refer the interested reader to Michael Johnson’s 2017 contribution to this volume for a criticism
along these lines, and to my 2014b:14–15 for a sketch of a pragmatic solution.
264 E. Maier
(8) jjMary said “No; I’ll never forgive you”jj = 9eŒsay.e/ ^ agent.e/ D mary
^ time.e/ < now ^ form.e/ D pNo; I’ll never forgive youq
Here, form is a function mapping an utterance event to the “linguistic form” of the
utterance.7 There are different kinds of entities that we might call the linguistic
form of an utterance. For written language, I use the simple typographical form, as
represented by a sequence of letters in the Roman alphabet; for spoken language,
I use these same letter sequences as crude approximations of phoneme sequences;
and for signed examples I represent sign sequences using the standard linguistic
notation where signs are represented as English words in capital letters.8 I use so-
called Quine corners around such letter/phoneme/sign sequences to refer to them in
the formal metalanguage: pabcdeqwf ;c Dabcde.9
2.2 Thinking Vs. Saying
So far we have looked only at reported speech, but in Sect. 1.1 we also saw examples
of both direct and indirect thought reports. Indirect thought reports, under the header
of propositional attitude reports, have traditionally received a lot of attention in the
philosophical and linguistic literature. Direct thought reports, like (2b), repeated in
(9), by contrast, are seriously understudied.
(9) “My heart’s broken,” he thought. “If I feel this way my heart must be
broken.” ŒErnest Hemingway; ‘Ten Indians’; in Men Without Women; 1927:
Direct thought reports are particularly relevant for the current investigation, as my
goal is to defend the analysis of free indirect discourse as direct discourse, and free
indirect reports are typically used to convey a character’s inner thoughts.10
I propose to treat direct thought reports exactly like direct speech, i.e., as
asserting the existence of a thinking event with a linguistic form.
(10) 9eŒthink.e/ ^ agent.e/ D x ^ time.e/ < now
^ form.e/ D pMy heart is broken: If I feel this way my heart must be brokenq
A thought event is like an utterance event – think of it as the sub voce utterance
of a sentence in the subject’s language of thought. In other words, in direct thought
reports, thinking is conceptualized as silently speaking to yourself.
7
Cf. Ludwig and Ray (2017) for a similar use of a form-function on speech events.
8
In principle, IPA would be a more appropriate alphabet for spoken utterances, and it might even
be possible to come up with an alphabet of ‘signemes’ for signed utterances.
9
More specifically: pabcdeq is a well-formed expression of type u (cf. Potts 2007), and the five
letter string itself is an entity in the corresponding domain Du .
10
In fact, as Banfield (1973) suggests, free indirect speech reports purporting to represent X’s words
are often better thought of as reporting what the addressee Y is hearing than as what X is saying.
If so, perhaps all so-called free indirect speech reports are really free indirect thought reports.
This conception of thoughts as concrete spatio-temporal entities with linguistic

form is rather different from the conception of beliefs as propositional attitudes
that we find in the traditional semantic analysis of attitude reports. I consider this
a virtue of the present account, as in fact direct thought reports have very different
semantic properties from typical attitude reports. For instance, indirect belief and
other attitude reports are closed under logical consequence, or at least they typically
allow inferences as in (11a), while direct thought and speech reports do not allow
analogous inferences, as shown in (11b).
(11) a. I thought/believed/knew/realized/feared/imagined that John and Mary
were both idiots
) I thought/believed/knew/realized/feared/imagined that John was
an idiot
b. “John and Mary are both idiots”, I thought
6) “John is an idiot”, I thought
The pattern in (11b) is predicted by the event modification approach sketched
in (10). The existence of an event with a given linguistic form does not entail the
existence of another event with a different form.
We might analyze indirect attitude reports as reports of more abstract, con-
tentful states instead of events. Like speech events, these attitudinal states have a
propositional content, i.e. content maps attitudinal states to sets of possible worlds.
To capture the inference in (11a) we further assume that a state s is a state of
believing that John and Mary are idiots if its content entails that they are, i.e., if
the propositional content of s is a subset of the proposition that John and Mary are
idiots.
(12) 9sŒbelieve.s/ ^ agent.s/ D i ^ time.s/ < now
^ content.s/ ^ Œidiot.j/ ^ idiot.m/
The semantics in (12) is arguably appropriate for believes that, but the inference
patterns are less clear for said that or thought that. Hence it is not clear whether we
should switch from the strict event specification analysis in (1a) to the more flexible
state analysis in (12) for these types of reports as well. I will leave this for another
occasion. What’s important for this paper is that direct discourse involves an event
of saying or thinking, which is a spatio-temporally realized concrete particular, with
a certain linguistically specifiable form.
2.3 Speaking and Signing Vs. Writing
Apart from the unification of direct and indirect speech and thought reports, another
major selling point of the event modification approach is that we can incorporate
central insights from the demonstration theory of quotation (Clark and Gerrig 1990;
Davidson 2015) to account for action and speech reports in spoken or signed
266 E. Maier
language. In these modalities reporters may specify not (just) the form or content of
the original event, but demonstrate some relevant features of it.
Let me illustrate with the English be like construction. The hhii notation in (13)
is meant to indicate my saying She was like followed by a shrugging gesture.
(13) She was like hhshrugsii

The shrugging gesture is an event that is ‘recruited’ (in Recanati’s 2001 terms) into
the syntax of the spoken sentence. Note that this kind of recruitment is possible in a
live medium, like speaking or signing, but much harder in printed writing, since you
can’t print an event.11 Davidson (2015) proposes an analysis of demonstrations like
(13) in terms of event modification. In our notation, the logical form of (13) would
look like (14), where d denotes the reporter’s shrugging event:
(14) 9eŒagent.e/ D sue ^ time.e/ < now ^ demonstration.d; e/
The logical form in (14) features Davidson’s demonstration, which relates two
events if they are sufficiently similar in certain contextually salient respects. In this
way, (14) captures the idea that the reporter’s shrugging serves as a demonstration
of what Sue did.
Focusing on signed language, Davidson takes this demonstration analysis of
spoken be like as her starting point to capture canonical direct discourse construc-
tions in spoken and signed languages. Spoken speech reports with demonstrational
(aka iconic or simulative) elements, like (15), would be analyzed as asserting
the existence of a past saying event sufficiently similar to the reporter’s verbal
demonstration, i.e. the event of her uttering Ehh well I don’t know in a creaky voice.
(15) The old man said, hhin a creaky voice Wii “Ehh. . . Well, I don’t know”
I propose – contra Davidson – that examples like (15) are really mixtures,
consisting of a simultaneous direct speech and action report, as shown in (16) (where
o denotes the old man and d the event of uttering the quoted fragment in a creaky
voice).
(16) 9eŒsay.e/ ^ agent.e/ D o ^ time.e/ < now
^ form.e/ D pEhh : : : Well; I don0 t knowq ^ demonstration.d; e/
In this way we maintain a straightforward, uniform analysis of direct discourse
across modalities, while leaving room for demonstrative strengthening in ‘live’
modalities. Printed writing does not really allow demonstrative action reporting,
so cases like (13) and (15) do not occur there.
11
As Michael Johnson (p.c.) points out, smileys and emoticons may be considered the print
counterparts of event demonstrations, as in She was like, .
2.4 The Semantics of Unquotation
In certain genres of factual writing, square brackets are used to indicate editorial
adjustments to a quotation. In previous work I’ve introduced unquotation as the
dual of mixed quotation (Maier 2014b). But we’re interested now in unquotation in
full, clausal, direct discourse. The new event semantics of direct discourse allows a
simpler semantics, bypassing mixed quotation entirely.12
The goal of this subsection then is to capture the truth conditions of direct
discourse constructions containing unquotation brackets, as in (17).
(17) “Find a way to get rid of [me] as soon as possible,” they said.
The intended interpretation is one where the speaker/writer of (17) overheard some
people talking about her behind her back, saying how they wanted to get rid of
her. Now she’s reporting what she heard, using unquoted me (rather than her own
name or some more elaborate description) to refer to herself. Semantically, the
basic idea is that unquotation brackets indicate that the material inside is to be
interpreted “from the reporter’s perspective,” i.e. outside the scope of the quotation.
In the current theoretical framework then, unquoted expressions are to be modeled
as specifications of content, while the surrounding quotation is a specification of
form. I use the symbol \ to denote concatenation of strings (of letters, phonemes, or
signs).
(18) 9eŒsay.e/ ^ agent.e/ D x ^ time.e/ < now ^ 9e0 e Œ
form.e/ D pFind a way to get rid ofq\ form.e0 /\ pas soon as possibleq
^_ content.e0 / D i
In words, (17) asserts the existence of a speech event e with linguistic form find
a way to get rid of . . . as soon as possible, where the ellipsis indicates some
unspecified word (or complex expression), the reference of which is fixed by the
unquoted material. Translating this into our formal notions of events with contents
and forms, (18) says that in addition to the speech event e there should be a
sub-event e0 of uttering some linguistic expression (form.e0 /), which refers to me
(_ content.e0 / D i).13
12
I restrict attention to what Shan (2011) calls ‘semantic unquotation’. My semantics does not
handle ‘syntactic unquotation’ like “Get out, you [expletive]!”, she yelled. Cf. Saka (2017) for
more on the various kinds of unquotation.
13
This is a case where spelling out the exact recursive translation rules is non-trivial. For instance,
we’ll likely need to assume some kind of syntactic movement to get the unquoted expressions
outside the quote. Cf. Maier (2014b) and Koev (2017) for more detailed proposals incorporating
such a movement, or Shan (2011) for an account without movement.
268 E. Maier
2.5 The Semantics of Free Indirect Discourse
Over the previous subsections I’ve developed a concrete semantic framework for
representing direct and indirect speech and thought reports, as well as unquotation.
With this we can now return to the free indirect discourse examples from 1.2
to make the informal paraphrases in terms of quotation and unquotation precise.
Consider, (3) repeated in (19a) below. In (19b) I repeat (part of) the informal
paraphrase from Sect. 1.2, and in (19c) I now add the actual logical form, according
to the semantic theory laid out above.
(19) a. Mary was packing her bags. Tomorrow was her last day. Oh how
happy she would be to finally walk out of here. To leave this
godforsaken place once and for all.
b. . . . “Tomorrow [was] [her] last day” . . .
c. 9eŒthink.e/ ^ agent.e/ D x ^ time.e/ < now ^ 9e0 e; e00 e Œ
form.e/ D pTomorrow q\ form .e0 /\ form .e00 /\ plast dayq ^
_
content.e0 / D jjwasjj ^ _ content.e00 / D jjherjj
The logical form in (19b-c) is still highly simplified. For instance, we’re unquoting
the entire words was and her, but, as I’ve argued elsewhere, what is really unquoted
is just the underlying third person feature and the past tense feature (which,
moreover, is not typically interpreted in situ, but first moved higher up) (Maier
2015).14 Nonetheless, (19) suffices to illustrate how our formalization of direct
discourse and unquotation can be used to explicate the informal free indirect
discourse semantics suggested in Sect. 1.2.
3 The Pragmatics of Unquotation in Direct Discourse
In the previous section we developed a semantics to back up the intuitive quotational

paraphrases of free indirect discourse reports. This leaves us with some pragmatic
issues, like where do the covert unquotations come from? In this section I survey
cases of apparent unquotation in other varieties of direct discourse. In Sect. 4 I then
apply the insights gained here to tackle the central puzzle of unquotation in free
indirect discourse.
14
Sharvit (2008) has argued that these unquoted pronominal features are actually interpreted de se
i.e., the result of so-called ‘sequence of tense’ and ‘sequence of person’ rules in English, triggered
by a silent attitude verb in the syntax. As I have shown elsewhere, this is entirely compatible with a
quotation + unquotation approach, like this one, but I won’t spell out the details here (Maier 2015).
3.1 Second Person Magnetism
Typologists occasionally observe puzzling direct discourse reports in which a

second person pronoun behaves as if it were in indirect discourse, i.e., as referring
to the reporter’s current addressee. Thus, van der Voort (2004) writes about direct
discourse in Kwaza, an isolated language of Brazil15 :
As most examples in this section indicate, the quotative construction contains direct speech
embedded in an extra layer of inflexions. As is seen here [in (20)], this is not the case when
the subject of the quotated utterance is a second person. In that case the interpretation is
one of indirect speech and the identity of the second person equals the hearer in the actual
speech context instead of the (logically expected) speaker. So the quoted second person
represents an exception to the direct speech analysis of the quotation construction presented
in the previous sections. [van der Voort 2004 : 411]
He uses the following example to illustrate the point:
(20) maga’riDa kukui’hQy-xa-’ki-tse

Margarida ill-2sg-DECL-DECL
literal translation: ‘Margarida (says) “you are ill”’
intended interpretation: ‘Margarida says that you are ill’
With (20), the reporter uses a second person inflection to refer to her current
addressee, without thereby suggesting that Margarida’s original utterance was also
second person. In other words, (20) would be true if, for instance, Margarida had
said something like “That annoying linguist is ill”. Given that the Kwaza quotative
construction (glossed as DECL-DECL) otherwise consistently behaves as direct
speech, I suggest that the logical form of (20) involves unquotation:
(21) a. Margarida says “[you] are ill”
b. 9eŒsay.e/ ^ agent.e/Dmarga ^ 9e0 e Œform.e/Dform.e0 /\ pbe illq
^ _ content.e0 /Dyou
Evans (2012) cites similar observations about quotation constructions in Slave
(Canada) and Nez Perce (U.S.).16 He even adds an example from colloquial spoken
English:
X had become confused about which house his daughter-in-law Y lives in, knocked on the
neighbour’s door, and had been directed to the daughter-in-law Y’s house. Later he tells
his daughter-in-law what had happened. ‘They told me, “Oh, youX ’ve got the wrong house,
youY live next door.”’ ŒEvans 2012 W 87
15
This passage is also cited and discussed by Evans (2012).
16
The data Evans mentions for Slave and Nez Perce are ultimately inconclusive, as there is some
additional evidence that the reports in question are indirect rather than direct discourse (Anand and
Nevins 2004; Deal 2014). For these languages, more research is required to compare the relative
merits of a monstrous indirect discourse analysis (à la Schlenker 2003) and a direct discourse with
unquotation analysis (as proposed here).
270 E. Maier
The first you behaves as usual. It refers to the original addressee of the reported
speech act (i.e., the neighbor giving directions to X). The second you, by contrast,
refers to the addressee of the reporting speech act (i.e., X telling his daughter-in-
law about getting lost earlier). Formally, I propose to capture the apparent truth
conditions of (21) in terms of direct speech and unquotation, as in the Kwaza
example.
(22) They told me, “Oh, you’ve got the wrong house, [you] live next door.”
Evans coins the term “second person magnetism”, or, more generally, “speech
act participant attraction”, to describe a pragmatic mechanism that gives rise to
quotation constructions as in Kwaza, Slave, and Nez Perce, as well as occasional
readings17 like (22) in English.
what is basically direct speech can undergo ‘speech act participant attraction’ which leads
any arguments in the quoted passage referring to the addressee (Kwaza, Slave, Nez Perce)
or the speaker (Nez Perce) to trump the person value that they would have as calculated
from the perspective of the reported speech event. This leads the construction to depart
from canonical direct-speech status those values of its person features that are susceptible to
second person attraction, even though in all other respects the construction is direct speech.
ŒEvans 2012
Evans offers no more detailed theory of the semantics or pragmatics of attraction.

Below I will integrate the underlying intuition into our formal semantics of direct
discourse and unquotation. I suggest that the pragmatic attraction by the salient
speech act participants, i.e., the actual speaker and/or hearer in the reporting context,
as described by Evans in the quote above, is a general pragmatic principle. We can
reformulate it as a constraint as follows:
(23) ATTRACTION: when talking about the most salient speech act participants,
use indexicals to refer to them directly.
Concretely, this rule entails that a speaker should always use I to refer to herself
(rather than, say, the speaker or Emar), and you to refer to her current addressee.
By ‘referring directly’ I mean using an indexical that picks out its referent by
being evaluated relative to the current utterance context. Crucially, I’m using
‘referring’ here in its pragmatic sense, i.e., a speaker using an expression to refer
to something.18 In that sense, indexicals in direct discourse reports may be used to
refer indirectly. For instance, in John said, “I am here”, the indexicals I and here
are used to refer indirectly to John and the place where he made his utterance.
17
Evans suggests that this example may be just a “performance error”. Nonetheless, if examples
like this occur regularly, that could be taken as evidence for attraction as a pragmatic principle.
18
In Kaplan’s (1989) Logic of Demonstratives, it is a logical truth that all indexicals refer directly.
This theorem however depends on a semantic notion of direct reference. On Kaplan’s semantic
construal, indexicals in direct discourse have no reference at all, while on the current pragmatic
construal they may still be used to refer to individuals (Maier 2016).
The possibility of indirect reference in direct speech points to a potential conflict

between the semantics of direct discourse and the pragmatics of attraction. I will
make this concrete by formulating a second constraint:
(24) VERBATIM: in direct discourse, faithfully reproduce the linguistic form of
the reported utterance.
This constraint is really just an informal paraphrase of our semantics of direct
discourse in terms of form specification, and hence a semantic rather than a
pragmatic constraint.
The two constraints conflict in cases where we want to quote a speech act about
someone who participates in the reporting speech act, but who played a different role
in the original speech act. The cases considered by Evans and cited above are of this
type. Take the Kwaza example in (20). Margarida was talking about X, let’s say by
describing him in a third person way as that annoying linguist. Later, the reporter
is talking to this linguist X about what Margarida said. According to ATTRACTION
the reporter must use a second person pronoun if he wants to refer to X, his current
addressee. But, since he’s using the direct discourse mode of reporting, VERBATIM
entails he must faithfully reproduce Margarida’s way of referring to X, i.e., by using
that annoying linguist.
Our semantic inventory offers a way to resolve this conflict: use unquotation to
locally suspend VERBATIM. The logical form with unquoted you in (21) (Margarida
says “[you] are ill”) allows the reporter to satisfy ATTRACTION, referring to the
current addressee with a second person pronoun, without violating VERBATIM,
because unquoted material need not be faithful to the original form.
The unquoted you in the logical form of the colloquial English report in (22)
can also be explained as the resolution of a conflict between ATTRACTION and
VERBATIM. The neighbor literally said something like Oh, you’ve got the wrong
house, she lives next door, so VERBATIM would force the direct discourse reporter
to refer to the daughter-in-law indirectly, as she. But since the reporter is telling this
story to his daughter-in-law, ATTRACTION forces him to refer to her as you instead.
We can satisfy both by invoking unquotation, which explains the logical form we
postulated for the example in (22).
But what about the other indexical in this English example? In the first
clause, the reporter refers to himself indirectly with a shifted you, thus violating
ATTRACTION.19 The quote from Evans above suggests that there may be some
cross-linguistic variation in what participant role can give rise to unquotation by
ATTRACTION. Thus, in Kwaza and Slave (and colloquial English?), only the current
speech act’s addressee exerts a force of attraction that is great enough to trump
VERBATIM and thus give rise to unquotation. In other languages, or perhaps just
19
A report that avoids this violation would look like They told me, “Oh I’ve got the wrong house,
you live next door”, with both I and you understood as unquoted. More empirical research, for
instance a corpus study on colloquial, oral storytelling, would have to be done to determine if such
reports indeed occur, or if ATTRACTION in English is restricted to the addressee.
272 E. Maier
other genres, it may be both the speaker and the addressee.20 And on the other end
of the spectrum, in English academic prose for instance, VERBATIM may be so
strong that the role of ATTRACTION is negligible.21
In the remainder of this section I’ll use the pragmatic unquotation-by-attraction
story to explain some findings from sign language and child language research.
3.2 Attraction in Sign Language Role Shift
The sign language analogue of direct discourse is role shift (Padden 1986). In
role shift, the reporter shifts her body and/or breaks eye contact with her current
addressee to indicate that she’s quoting someone else’s words. Consider the
following basic example.
RS
(25) JOHN IX-1 HAPPY
‘John said/signed/thought, “I’m happy”’
First, some notes about notation. As usual in sign language linguistics, lexical
signs and names are represented in capital letters. The sign language equivalents of
pronouns are pointing signs, represented as IX (for “index”). IX-1, the sign language
equivalent of a first person pronoun, is typically realized as a pointing toward the
signer’s own chest; IX-2, the second person pronoun, is realized as a point toward
the addressee. Third person pronouns are realized as pointings toward any other
concrete entities in real space, or to more abstract “discourse referents” associated
with specific locations in signing space. The scope of the non-manual role shift
marking is indicated by an overline.
So, (25) represents the signer signing the name sign for some John, and then, in
role shift, signing I’m happy, which includes a point to the signer’s chest to mimic
John’s utterance of I, i.e. we have a shifted first person referring to the reported
speaker/signer rather than the reporting signer. The verb of saying is often left
implicit. I will use the term utterance as a modality non-specific way of referring to
the event of intentionally producing something with linguistic form and content.22
20
Saka (p.c.) and Evans both raise the possibility of a natural, cross-linguistic hierarchy constrain-
ing this process of unquotation by ATTRACTION, e.g. if a given language (or a given genre,
modality, or construction) allows first person unquotation to satisfy ATTRACTION it must also
allow it for second person.
21
On the other hand, Maier (2015) cites style guides for formal writing that suggest that a reporter
should adjust pronouns and tenses to fit the reporting environment. This may be viewed as an
instance of unquotation by attraction, quite similar in fact to what we’ll see in Sect. 4. The only
difference then is that the unquotation must be marked overtly in such genres.
22
It is often noticed in the sign language literature that a reporter may include gestures and
other iconic elements (especially in so-called classifier constructions, cf. Davidson 2015) into
their report, as demonstrations of certain paralinguistic or extralinguistic events surrounding the
reported utterance. As outlined in Sect. 2.3 for spoken reports, I propose to model this as a
(26) 9eŒutt.e/ ^ agent.e/ D john ^ form.e/ D pIX-1 HAPPYq
As observed by Engberg-Pedersen (1995), Quer (2005), Hübl (2014) and others,

not all indexicals occurring under role shift are semantically shifted. In (27), the
indexical IX-1 is shifted, referring to the agent of the reported thought, while HERE
is not shifted, referring to the location where the report is taking place, in Barcelona.
t RS
(27) IX-a MADRID MOMENT JOAN THINK IX-1 STUDY FINISH HERE
intended interpretation: ‘When he was in Madrid, John thought he would
finish his study in Barcelona.’ ŒLSC.Catalan Sign Language/; Quer 2005
For authors like Quer (2005) and Schlenker (2017), this is evidence that role shift
is not like direct discourse after all, but rather like indirect discourse, with a so-
called monstrous operator to account for the indexicals that do shift, like the IX-1’s
in both (25) and (27). Davidson (2015) argues against the monstrous approach23
and offers a monster-free alternative, based on a purely demonstrational account
of quotation. Though much closer to my own proposal, Davidson’s approach still
makes the wrong predictions for mixed indexicality cases like (27). Concerning this
very example, she points out that her account “makes the prediction that more iconic
indexicals will shift before less iconic indexicals under role shift”. But in (27), both
IX-1 and HERE are pointing signs, and as such they are simply not iconic at all –
after all, I do not resemble a pointing at my own chest (the sign IX-1 in LSC), nor
does Barcelona resemble someone pointing to the ground (the sign HERE in LSC).
In sum, I fully agree with Davidson that role shift is essentially a form of direct
discourse, as opposed to a monstrous operator. However, to deal with unshifted
pronouns, I suggest we use the mechanism of unquotation (for the semantics) and
attraction (for the pragmatics).
Let’s start with the semantics. We can describe the correct truth conditions of (27)
as follows:
(28) a. . . . JOAN THINK “IX-1 STUDY FINISH [HERE]”

b. 9eŒ: : : think.e/ ^ agent.e/ D john ^
9e0 e Œform.e/ D pIX-1 STUDY FINISHq\ form.e0 /
^_ content.e0 / D here
Here is another example, taken from a pilot study about role shift in NGT (Sign
Language of The Netherlands).24 Some signers were presented with a video of
Martine signing (29a) and then Mascha reporting that utterance as in (29b).
Crucially, in (29a) Martine refers to Johan Cruijff by finger spelling his last name,
demonstrational action report modifying the reported utterance event with the additional constraint
demonstration.d; e/ (with d referring to the reporter’s partly verbal demonstration).
23
I’ve added some arguments against the monstrous approach to role shift myself (Maier 2016).
24
This pilot study was performed by myself and Martine Zwets in Nijmegen, 2012. Unfortunately,
the study never moved beyond the exploratory, informal data collection stage.
274 E. Maier
letter by letter, while in (29b) Mascha in her report refers to Cruijff by pointing at a
nearby picture of Cruijff.
(29) a. IX-1 THINK C-R-U-I-J-F-F BEST SOCCER PLAYER

‘I think Cruijff is the best soccer player.’
RS
b. MARTINE IX-1 THINK IX-a.pointing-at-Cruijff/ BEST SOCCER PLAYER
intended: ‘Martine said, “I think Cruijff is the best soccer player”’
When asked to judge whether (29b) is a felicitous way to report (29a), most subjects
judged (29b) equally felicitous as a variant with a fully verbatim reproduction of
(29a) in the role shift (both report variants averaging 3.9 on a 5-point Likert scale). I
propose to capture the apparent truth conditions of (29b), like those of (27), in terms
of quotation and unquotation:
(30) MARTINE “IX-1 THINK [IX-a.pointing-at-Cruijff/ ] BEST SOCCER
PLAYER”
Now the pragmatics. Why is HERE unquoted in (27), or the pointing at Cruijff
in (29b)? My answer follows the attraction explanation for unquotation in Kwaza
and colloquial English. The cases of unquoted you in Sect. 3.1 involved attraction
by a salient speech act participant, viz. the addressee. In the sign language cases
above, we have a signer talking about other salient entities present in their immediate
surroundings: the place where signer and addressee are currently located: Barcelona
in (27), and Cruijff in (29b). Hence, we should modify ATTRACTION to apply to
any salient entity physically present in the current utterance context, not just the
immediate speech act participants (speaker and hearer, as Evans suggested). In this
way, ATTRACTION will pull the reporter toward using indexicals, i.e. pointing signs,
to refer to these salient entities, Barcelona and Cruijff, directly. If role shift is a form
of quotation, this leads to a conflict with VERBATIM, which would force the reporter
to reproduce the signs originally used by the reported speaker to pick out Barcelona
(probably the city’s name sign) and Cruijff (the finger spelled name C-R-U-I-J-F-F).
The unquotations in the logical forms proposed above are then explained as a way
to resolve the conflict.
3.3 Attraction in Child Language
Recent empirical research on the interpretation of pronouns in reported speech

reveals that children may be especially sensitive to attraction (Köder and Maier
2016). In one experiment children and adults were presented with reports like (31)
and asked who gets the object.
(31) a. Hond zei: “Ik/jij/hij krijgt de bal”
‘Dog said, “I/you/he gets the ball”’
b. Hond zei dat ik/jij/hij de bal krijgt

‘Dog said that I/you/he gets the ball’
The results suggest that children from ages 4 to 12 have difficulty interpreting direct
discourse, but not indirect discourse. More precisely, children seem to interpret the
pronouns in direct discourse items as if they were indirect, i.e. they fail to perform
the required context shift.
In the current theoretical framework, we can describe the children’s behavior in
terms of unquotation and attraction. To exclude first the possibility that children
just confuse direct and indirect discourse, note that in Dutch, unlike in English,
indirect speech requires a different word order than indirect speech, and an overt
complementizer. In addition, to mark the direct speech items as such we added
a distinct pre-quotation pause and mimicked the different voices of the quoted
characters. Assuming that children pick up on (some of) these syntactic and prosodic
features of the direct–indirect distinction, I suggest that they interpret a second
person instance of (31a) as in (32).
(32) a. Hond zei, “Jij krijgt de bal”
b. Dog said, “[You] get the ball”
c. 9eŒsay.e/ ^ agent.e/ D dog ^ 9e0 e Œform.e/ D form.e0 /\
pget the ballq^ _ content.e0 / D you
On this semantic analysis, children treat the indexical pronouns in direct discourse
as unquoted and thus referring directly to coordinates of the actual reporting context.
In other words, they treat the Dutch (32) just like a Kwaza speaker would treat its
Kwaza counterpart, and a Dutch signer its NGT counterpart.
Pragmatically, the reason why children should choose to interpret the second
person in (32) as unquoted is by now familiar. The most salient entities physically
present in the context of the reporting utterance include Monkey (the current
speaker), Elephant (the current addressee), and Dog (who is standing just a few
feet away). These salient individuals thus attract the use of indexicals that refer to
them directly, i.e. I for Monkey, you for Elephant, and he for Dog. Since the speech
act roles were divided differently in the original utterance (I referred to Dog, etc.),
this conflicts with the verbatim reproduction requirement imposed by the semantics
of direct discourse. Unquoting the pronoun as shown in (32) resolves the conflict.
One last remaining question is, why are adults not affected? Looking closely
at the adult data, Köder et al (2015) conclude that adults are affected. They make
significantly more “mistakes” in direct than in indirect discourse, and the mistakes
that they make are compatible with the attraction effect, i.e. direct speech pronouns
are interpreted as unquoted. Nonetheless, compared to the children, it looks like
the adult grammar gives more weight to VERBATIM, in many cases ranking
it higher than ATTRACTION. The very late acquisition of this adult constraint
ranking suggests a possible influence of writing instruction in school. Perhaps the
strict verbatim requirement and the necessity of consistent, overt marking of both
quotation and unquotation in formal writing, interferes with adults’ performance in
this experiment.
276 E. Maier
4 Attraction in Free Indirect Discourse
We set out to explain why pronouns and tenses are unquoted in free indirect
discourse. I will argue that unquotation in free indirect discourse derives from the
pragmatics of attraction, exactly as in the above examples of unquotation in spoken
and signed languages. Note however that in the cases discussed in Sect. 3, unquo-
tation was always the result of attraction to salient entities physically present in the
reporter’s immediate surroundings. Thus, we’ve seen unquoted ‘local indexicals’
like I, you, and here. In free indirect discourse the unquotation extends to remote
third person pronouns and past tenses.
To bring out the continuity between pragmatic attraction in direct discourse and
unquotation in free indirect discourse, let’s start with a case of free indirect discourse
in a spoken, first person, present tense narrative.
(33) This woman left me a voice mail, asking all kinds of questions about
you. How well do I know you? Where have we met? Have I ever noticed
anything strange about you?
Here, a speaker/narrator25 is telling his addressee about the content of a voice mail
message from some third person. The questions in (33) are intended as reports of
what this person asked the speaker. They exhibit the key characteristics of free
indirect discourse: the form of the reporting faithfully reproduces the form of the
original questions, except for the pronouns, which are adjusted to fit the reporter’s
perspective. On our quotational analysis, the logical form must be as in (34).
(34) . . . “How well do [I] know [you]?” “Where have [we] met?” “Have [I] ever
noticed anything strange about [you]?”
This is precisely the pattern of quotation and unquotation that we encountered in
Kwaza, sign language, and child language (e.g., Dog said, “[you] get the ball”).
The attraction explanation extends to this case without further modification. On
the one hand, direct quotation of questions entails the use of verbatim, i.e. shifted,
indexicals, as in How well do you know him?. On the other hand, the story is about
two very salient, physically present speech act participants, viz. the speaker and his
addressee, which leads to attraction, i.e., a preference for choosing indexicals to
pick out these salient participants directly: How well do I know you?. Unquotation
of the pronouns referring to the salient participants resolves the conflict.
The exact same explanation holds for literary cases of free indirect discourse in
so-called second person narratives, a somewhat experimental form of storytelling in
which the main protagonist is the narrator’s addressee.
25
Narratologists stress the fundamental differences between the narrator and the author, and
between the narratee and the reader. In this paper I do not make these fine distinctions. I use
the terms narrator and author (or speaker, or writer, for that matter) interchangeably.
(35) Sunlight. A morning. Where the hell are your sunglasses? You hate
mornings – anger rises in you, bubbling like something sour in your throat
– but you grin into the morning because somebody is approaching you,
shouting a magic word. Your name.
Œfrom Joyce Carol Oates; ‘You0 .1970/; cited by Fludernik 1995 W 82
It is not entirely clear where the free indirect reporting in this passage begins or
ends, but the direct question with the expletive is definitely intended as a report of
the second person protagonist’s thoughts – it’s not the narrator asking where your
sunglasses are. So, again, the form is that of a (typographically unmarked) direct
discourse, but the indexical your is adjusted to the narrator’s frame of reference –
the original thought would be in the first person (Where the hell are my sunglasses?).
The logical form then is:
(36) . . . “Where the hell are [your] sunglasses?” . . .
In this case too the unquotation may be explained pragmatically, in terms of

attraction: the most salient agent is the protagonist (i.e., you) so ATTRACTION
requires that the narrator use the appropriate indexical expression to refer to that
agent directly: your. But this is only possible if the VERBATIM constraint imposed
by the semantics of direct discourse is locally suspended, through unquotation.
Now it is but a small step from free indirect speech in first and second person
present tense narratives, to the standard, literary examples, illustrated by (37),
repeated from Sect. 126 :
(37) a. Mary was packing her bags. Tomorrow was her last day. Oh how
happy she would be to finally walk out of here. To leave this
godforsaken place once and for all!
b. . . . “Tomorrow [was] [her] last day.” . . .
The difference between examples like (37) and the examples discussed before (in
this section and the previous), is that the most salient entities under discussion – the
story’s protagonists – are both physically and temporally removed from the actual,
reporting speaker (i.e., the narrator). The kind of stories where we find free indirect
discourse reports like (37) are told by a so-called omniscient, third person narrator.
Such a story does not concern the immediate here and now of the narrator, but is
presented as taking place in the past, dealing primarily with protagonists observed
from a distance by the narrator and her addressee.
So let’s take a closer look at (37). The story is about Mary – a third person,
referred to by her name or by third person pronouns, and whose actions are
described in the past tense. Mary is more salient than the narrator or his addressee.
Likewise, the story time is more salient than the time of the narration.27 Hence,
26
The logical form in (ib) is a simplification of my proposal in (Maier 2015). For instance, I argue
there that in an example like this it’s not the whole pronoun and verb that get unquoted but only
the third person and tense features.
27
The time of narration is a rather abstract concept, distinct from both the time of writing and the
time of reading, but, as pointed out above, such subtleties don’t concern us here. What’s important
278 E. Maier
while in earlier cases the source of attraction was always someone or something
in the reporter’s here and now, in these types of literary third person narratives,
the source of attraction is the protagonist and her here and now. As a result, the
narrator should use the appropriate referential terms to refer directly to those distant
but salient story agents. In such a case, the appropriate directly referential terms
for the narrator to refer to his protagonist and her actions directly are third person
pronouns and past tenses. The pull of ATTRACTION thus explains the unquotation
of pronouns and tenses in cases like (37).
5 Conclusion
The goal of this paper was to solve the puzzle of unquotation in free indirect
discourse: if free indirect discourse is really best described in terms of quotation
and unquotation (as argued elsewhere), why is it that it’s precisely the pronouns and
tenses that get unquoted?
In order to solve the puzzle, I first presented a novel semantic analysis of reported
speech, designed to describe also some less well-studied cases, such as signed and
spoken reports, direct thought reports, and unquotation.
Next, armed with the formal apparatus of quotation and unquotation, I surveyed
cases of apparent unquotation in various languages, drawing on descriptions in
the typological literature, examples from sign languages, and data from child
language experiments. It turned out that in all these diverse cases, unquotation
can be described as the pragmatic resolution of a conflict between two opposing
constraints: (i) ATTRACTION, use referring expressions to directly pick out the most
salient entities relative to your own, i.e. reporting, context; and (ii) VERBATIM,
when reporting directly, be faithful to the form of original utterance. Given our
semantic framework, unquoting the expressions used to pick out the salient speech
act participants (and salient entities surrounding them) allows the reporter to satisfy
both constraints.
Seen in this light, the case of free indirect discourse is not fundamentally different
from the way Kwaza speakers or Catalan signers use and interpret direct speech. The
only difference is that in a typical literary narrative (with a third person omniscient
narrator), the greatest source of attraction is not the narrator and his here and now,
but the story’s protagonist, and her here and now, which are arguably much more
salient.
Reducing the unquotation pattern found in free indirect discourse in this way
to the much more general pragmatic phenomenon of attraction partially solves the
puzzle of unquotation, viz. the question of why pronouns and tenses get unquoted. I
thereby defended the quotational approach to free indirect discourse from a serious
charge of overgeneration. I have shown that unquotation is not an ad hoc stipulation
is that, in the case of a story, the time of narration is less salient (or backgrounded, not-at-issue, if
you will) than the time of the story, whenever these two notions come apart.
to get the truth conditions for free indirect discourse right, but the result of a general
pragmatic constraint resolution strategy found in a wide variety of direct discourse
reporting constructions.
In closing, let me point out that the attraction explanation falls short of a
complete answer to the puzzle of unquotation in free indirect discourse. Attraction
explains why (most) pronouns and tenses are unquoted in free indirect discourse,
but it fails to explain why certain non-pronominal indexicals are not unquoted.
Why does the narrator in (37) reproduce his protagonist’s tomorrow, here or this
verbatim, while unquoting pronouns and tenses to satisfy ATTRACTION? In the
end, it looks like we still have to assume some additional, grammatical constraints
on what can in principle be unquoted in a given direct discourse construction.
Thus, from the very limited data we have encountered in this paper we might
hypothesize that English and Kwaza direct speech in principle allows unquotation
of at least second person pronouns, while the grammar of role shift in some sign
languages may be more flexible, allowing unquotation of any pointing sign. Free
indirect discourse then comes out similar to role shift in allowing unquotation of all
pronominal elements.28
References
Anand, P., & Nevins, A. (2004). Shifty operators in changing contexts. Semantics and Linguistic
Theory (SALT), 14, 20–37.
Banfield, A. (1973). Narrative style and the grammar of direct and indirect speech. Foundations of
Language, 10(1), 1–39.
Bonami, O., & Godard, D. (2008). On the syntax of direct quotation in French. In 15th
International Conference on HPSG (pp. 358–377). CSLI Publications.
Clark, H., & Gerrig, R. (1990). Quotations as demonstrations. Language, 66(4), 764–805.
Davidson, K. (2015). Quotation, demonstration, and iconicity. Linguistics and Philosophy, 38(6),
477–520. doi:10.1007/s10988-015-9180-1.
Deal, A. R. (2014). Nez Perce embedded indexicals. In H. Greene (Ed.), Proceedings of SULA 7:
Semantics of Under-Represented Languages in the Americas (pp. 23–40). Amherst: GLSA.
Eckardt, R. (2014). The semantics of free indirect speech. How texts let you read minds and
eavesdrop. Leiden: Brill.
Engberg-Pedersen, E. (1995). Point of view expressed through shifters. Language, gesture, and
space (pp. 133–154). Hillsdale : Lawrence Erlbaum Associates.
Evans, N. (2012). Some problems in the typology of quotation: a canonical approach. In D. Brown,
M. Chumakina, & G. G. Corbett (Eds.), Canonical Morphology and Syntax. Oxford: Oxford
University Press.
28
I thank Kathryn Davidson, Amy Rose Deal, Nicholas Evans, Petra Hendriks, Franziska Köder,
Josep Quer, Per Erik Solberg, and Ede Zimmermann for helpful discussion about data and theory.
Special thanks to Paul Saka and Michael Johnson detailed editorial reviews. For help gathering and
interpreting the sign language data discussed in Sect. 3.2 I thank Martine Zwets, Mascha van den
Barselaar, Johan Ros, Yassine Nauta, Merel Naomi van Zuilen, and the participants in our 2012
survey. This research is supported by the EU under FP7, ERC Starting Grant 263890-BLENDS
AND by NWO, Vioi GRANT 267-80-004.
280 E. Maier
Fludernik, M. (1995). The Fictions of Language and the Languages of Fiction: The Linguistic
Representation of Speech and Consciousness. London: Routledge.
Hacquard, V. (2010). On the event relativity of modal auxiliaries. Natural Language Semantics,
18(1), 79–114. doi:10.1007/s11050-010-9056-4.
Heim, I., & Kratzer, A. (1998). Semantics in Generative Grammar. Oxford: Blackwell.
Hübl, A. (2014). Context shift (im)possible: Indexicals in German Sign Language. Conference of
the Student Organization of Linguistics of Europe (ConSOLE), 21, 171–183. http://www.hum.
leiden.edu/lucl/research/sole/proceedings
Johnson, M. (2017). Quotation through history: A historical case for the proper treatment of
quotation. In P. Saka & M. Johnson (Eds.), The semantics and pragmatics of quotation (pp.
281–302). Cham: Springer.
Kaplan (pp. 481–614). New York: Oxford University Press.
Köder, F., & Maier, E. (2016). Children mix direct and indirect speech: evidence
from pronoun comprehension. Journal of Child Language, 43(3), 843–866.
doi:10.1017/S0305000915000318.
Köder, F., Maier, E., & Hendriks, P. (2015). Perspective shift increases processing effort of
pronouns: A comparison between direct and indirect speech. Language, Cognition and
Neuroscience, 30(8), 940–946. doi:10.1080/23273798.2015.1047460.
Koev, T. (2017). Quotational indefinites. Natural Language and Linguistic Theory, 35(2), 367–396.
doi:10.1007/s11049-016-9344-x
Maier, E. (2014a). Language shifts in free indirect discourse. Journal of Literary Semantics, 43(2),
143–167. doi:10.1515/jls-2014-0010.
Maier, E. (2014b). Mixed quotation: The grammar of apparently transparent opacity. Semantics
and Pragmatics, 7(7), 1–67. doi:10.3765/sp.7.7.
Maier, E. (2015). Quotation and unquotation in free indirect discourse. Mind and Language, 30(3),
235–273. doi:10.1111/mila.12083
Maier, E. (2016). A plea against monsters. Grazer Philosophische Studien, 93, 363–395.
Padden, C. (1986). Verbs and role-shifting in ASL. In Proceedings of the Fourth National
Symposium on Sign Language Research and Teaching (pp. 44–57).
Compositionality (pp. 405–431). New York: Oxford University Press.
Quer, J. (2005). Context shift and indexical variables in sign languages. Semantics and Linguistic
Theory (SALT), 15, 152–168.
Recanati, F. (2001). Open quotation. Mind, 110(439), 637–687. doi:10.1093/mind/. 110.439.637
Saka, P. (2005). Quotational constructions. In P. de Brabanter (Ed.), Hybrid quotations (Vol. 17,
pp. 187–212). Amsterdam: John Benjamins.
Saka, P. (2017). Blah, blah, blah: Quasi-quotation and unquotation. In P. Saka & M. Johnson (Eds.),
Schlenker, P. (2003). A plea for monsters. Linguistics and Philosophy, 26(1), 29–120.
doi:10.1023/A:1022225203544.
Schlenker, P. (2004). Context of thought and context of utterance: a note on free
indirect discourse and the historical present. Mind and Language, 19(3), 279–304.
doi:10.1111/j.1468-0017.2004.00259.x.
Schlenker, P. (2017). Super Monsters I: attitude and action role shift in sign language. Semantics
and Pragmatics, 10(9), doi:10.3765/sp.10.9
Shan, C.-c. (2011). The character of quotation. Linguistics and Philosophy, 33(5), 417–443.
doi:10.1007/s10988-011-9085-6.
Sharvit, Y. (2008). The puzzle of free indirect discourse. Linguistics and Philosophy, 31(3), 353–
395. doi:10.1007/s10988-008-9039-9.
van der Voort, H. (2004). A Grammar of Kwaza. Berlin: Mouton De Gruyter.
Wade, E., & Clark, H. (1993). Reproduction and demonstration in quotations. Journal of Memory
and Language, 6(32), 805–819.
Quotation Through History: A Historical Case
for the Proper Treatment of Quotation
Michael Johnson
Abstract I present a theory of quotation in speech reports and the resolution of

indexicals in those reports based partially on the historical development of the
practice of punctuating with quotation marks. For the first thousand or so years
of English writing, quotation marks weren’t used. Their use in direct quotation
developed in the eighteenth century through a need created by the rise of the novel,
a need to clearly and frequently mark out when direct speech begins and ends.
The eighteenth century practice differs noticeably from current practice. While
eighteenth century novels always place direct speech forms within quotes, they
sometimes also place indirect forms there as well. I present a diachronically con-
sistent theory of quotation in speech reports. For a variety of reasons, contemporary
direct speech reports may diverge from a verbatim reproduction of the reported
speakers’ words: translation, removal of taboos, and clean-up of infelicities and non-
standard dialect features. Speech that has been minimally altered in these respects
is still appropriately placed within quotation marks. The same was true in the
eighteenth century novel: but those authors allowed that context-sensitive direct
forms could be altered within a quotation to indirect forms, in certain circumstances.
The historical data also supports the claim that quotation marks are punctuation
marks with no semantic effects. Monstrous accounts cannot handle cases where
some but not all context sensitive expressions are “shifted” within quotes. I present
a non-monstrous account of indexicals in reported speech.
1 Introduction
In common parlance, quotation is repeating the words of someone. That someone

need not be distinct from the person quoting: you can quote yourself. That someone
need not exist, and need not be in the past: you can quote Gandalf, quote what
someone should have said, or quote what they will say. Indeed, you can even quote
a source when you aren’t aware that that person was the source of the quote:
M. Johnson ()
Department of Philosophy, Hong Kong University, Pokfulam, Hong Kong
e-mail: micjohns@hku.hk

https://doi.org/10.1007/978-3-319-68747-6_10
282 M. Johnson
Even if you’ve never read a play by Shakespeare, you’ve probably quoted him countless
times without even knowing it : : : Bet the last time you said that you hadn’t “slept one
wink,” you didn’t realize you were quoting Cymbeline.1
There are myriad reasons to quote. Quoting others reveals our social alignments,
our knowledge, preferences, and interests. Quoting is a powerful social signifier:
I get a lot of e-mails from foreigners hoping to secure a visa to work in the United States : : :
a few weeks ago, one of the guys who e-mailed me had a signature that, at the very bottom,
said, “Everything dies, baby that’s a fact, but maybe everything that dies someday comes
back.”
I wished so bad that I could get that guy into the United States at that moment. Because
anyone going around quoting Bruce Springsteen deserves to be allowed into the United
States.2
Quoting song lyrics, religious texts, and literary figures not only reveals our knowl-
edge of social norms and mores, but also appeals to and at the same time enforces
those norms. Quoting can be a powerful form of adding gravitas, credibility, and
authority to our speech, and it can also be a potent means of undermining authority.
For example, when Job appears to be questioning his ability to prove his innocence
to God in 9:2–24, “Job is really exposing his friends, by ironically quoting some
of their absurd maxims” (Clarke 1837: 197). Quotation can even be a means of
creating authority, as when we quote past predictions that came out correct.
In our contemporary, formal, written practice, quotation marks are used to
attribute spoken or written words to others. But it was not always so. The diple,
the forerunner of the quotation mark, began simply as a means for ancient Greek
editors to highlight important or interesting aspects of the text (Finnegan 2011:
86). In the Christian era, the marks were adopted to signal Old Testament passages
that are quoted or paraphrased in the New Testament, and to signal the words of
Jesus (but not others). Quotation marks weren’t used for ordinary dialog: “quotation
quintessentially pertained to the scriptures : : : it was not for the speech of ordinary
mortals” (Finnegan: 97–98). It wasn’t until the eighteenth century and the rise of
the novel that quotation marks began to take on a role closer to their present one
(Finnegan: 98). In novels, multiple speakers exchange remarks, and it is necessary
to indicate when one person’s speech ends and another’s begins. Quote marks soon
became standard in this role.
However, the eighteenth century practice of using quote marks does differ in a
number of significant ways from our current practice, though I will argue that the
differences aren’t very deep. I’ll further argue that there’s no particular reason to
think our current practice is either more or less worthy of a theoretical understanding
than its historical forerunners are. Insofar as we have reasons to have a semantic or
1
Joelle Herr, William Shakespeare: The Complete Plays in One Sitting, Introduction. This is
somewhat of a borderline case because I’m assuming that Herr is assuming that those who are
unknowingly quoting Shakespeare see themselves as quoting someone, they are just unaware of
who they are quoting.
2
Catching Wise (blog), http://www.catchingwise.com/tag/bruce-springsteen-is-america/
Quotation Through History: A Historical Case for the Proper Treatment of Quotation 283
pragmatic (or whatever) theory of quotation as it works now, we have reasons to

have a semantic or pragmatic (or whatever) theory of quotation in the eighteenth
century. In this essay I’ll first introduce and analyze eighteenth century practice;
then I’ll attempt to provide a theory that covers both it and contemporary practice.
2 Q-Marks and the Eighteenth Century Novel
For the majority of time that English has been written, direct speech has gone
unpunctuated. For example at MS Cotton Vitelius A XV (the manuscript containing
Beowulf ) f 137r, there is a transition from the Beowulf poet’s voice to the voice of a
character, the cliff warden, after a verb of saying ‘frægn’ (‘asked’). This transition is
marked only with a dot, the same symbol that punctuates elsewhere ends of various
sentences and clauses. The warden is reported with a direct question in the present
tense ‘hæt syndon ge : : : ’ (‘what are you : : : ?’), and the indexicals ‘ge’ (2nd person
plural, ‘you’) and ‘Ic’ (1st singular, ‘I’) refer to the warden’s interlocutors and the
warden, respectively.
Even into the eighteenth century, quotation marks were not always used by
printers.
Tho’ my Mother refused to move it to my Father, yet as I have heard afterwards, she reported
all the Discourse to him, and that my Father, after shewing a great Concern at it, said to her
with a Sigh, That Boy might be happy if he would stay at home, but if he goes abroad he
will be the miserablest Wretch that was ever born: I can give no Consent to it. [Robinson
Crusoe, W. Taylor, 1719 1st Edition: 6]
The final ‘I’ here refers to the reported speaker, Crusoe’s father, and ‘he’ refers to
the (fictional) author of the passage, Crusoe.
These examples show at least this: current conventions regarding quotation marks
are recent. For the first thousand years English has been written, there were no such
marks (or they served purposes other than our contemporary ones). Nevertheless,
readers had little trouble working out the referents of indexicals including ones
“shifted” to the reported speaker’s context. The supposed inability to get “shifted”
readings from written reports without quotation marks (Cappelen and Lepore, 2007:
55–56) is a malady that uniquely affects present-day readers.
One early use of so-called “quotation marks” in printing was commonplacing.
Between the seventeenth and nineteenth centuries, the educated classes in Britain
and America were taught to keep commonplace books, something like a book of
quotations (“commonplaces”) compiled individually by their owner, though these
books could also contain things like statistics, measurement conversions, and other
information. Commonplace books were a status symbol, signifying the refinement
and erudition of their owners. They could be used as a source of meditation on
profound thoughts, or a means for finding a choice quote in the age before the
internet. Francis Bacon reportedly used his commonplace book to prepare for his
speeches.
284 M. Johnson
The demand for wise words to fill up one’s commonplace book was high enough
that printers used commonplace marks—inverted double commas—to indicate
passages worthy of copying. Editions of Shakespeare were common targets. In
this example from Shakespeare’s Measure for Measure, the printer has marked
Isabella’s statement of the play’s theme, as she chooses to refuse Angelo’s request
to exchange her virginity for her brother’s life:
Then Isabell live chaste, and brother die;
“More then our Brother, is our Chastitie.
Ile tell him yet of Angelo’s request,
And fit his minde to death, for his soules rest.3
Commonplace marks thus had the function of “reverse quotation marks” in that
they did not indicate which parts of a text had been copied from elsewhere (had
been quoted), but rather indicated which parts of a text you were supposed to copy
(to quote).
Still today there are texts that are intended to be copied and re-purposed—chain
emails and so-called “copypasta”—but the practice of explicitly marking such texts
has fallen out of favor. Commonplacing is an evolutionary dead end on the radiation
of functions of “quotation marks” during the early days of printing.
In order to distinguish between the marks and their function, which is only
sometimes quotation, I will call the inverted double commas “Q-marks.” The
earliest practice of Q-marking printed quotations differed from the present-day
practice in some notable ways, ways that I will argue matter for the theory of
quotation.4
The first relevant feature of direct and mixed quotation in the eighteenth century
novel is that it subsumes most of our contemporary practice. So, for example,
there are standard indirect quotes (1) and indirect questions (2) that are punctuated
roughly as we would punctuate them today (the exception being that the indirect
questions end with a question mark). Then there are standard direct quotations
(3), accounting for the majority of quotations. Finally, there are plenty of mixed
quotations, such as (4) and (5).5
(1) : : : For sometimes he said, he looked

on himself as still married, and considered his
Wife as only gone a little before him : : : [vol. I, p. 6]
3
Example from Estill (2014), who cites Mr. Shakespeares Comedies, Histories, & Tragedies
(London: Isaac Iaggard and Ed. Blount, 1623). Note that here, there are no closing Q-marks; only
the second line in the example is commonplaced. If more than one line were commonplaced, they
too would contain opening Q-marks at the beginning (and no closing Q-marks).
4
In what follows in this section, all examples are taken from Henry Fielding’s The History of Tom
Jones, a Foundling (1749), unless otherwise noted. For Volumes I and III, I have used the 1750 A.
Millar edition and for Volume II the 1769 A. Millar edition, all available on Google Books.
5
In the works I examined, mentioning an expression is not effected by putting quote marks around
it. So-called pure quotation has a distinct origin from direct and mixed quotation, or so it seems.
(2) Mrs. Miller then asked what was to be done with Blifil?
(3) ‘I wish,’ answered the surgeon, ‘I could

‘assure your being well in a month or two.
‘ : : : ’ [vol. II, p. 186]
(4) Upon this, Jones related to him the whole

Truth, and earnestly begged him not to punish
the poor Soldier, ‘who, I am confident,’ says
he, ‘is as innocent of the Ensign’s Escape, as he
‘is of forging any Lie, or of endeavouring to
‘impose on you.’ [vol. II, p. 169]
(5) nay, Cicero reports them to have been ‘incredi-

‘bly childish.’ [vol. II, p. 78]
But there is no shortage of examples that are at first instance tough to parse for the
contemporary reader6,7 :
(6) Jones1 answered with a Sigh, ‘He1 feared it was

‘ already too late for Caution;’ [vol. I, p. 264]
In (6), Tom is alone with Sophia, who has just warned him to be cautious. He
presumably says, “I fear it is already too late for caution,” but these are not the
words Fielding uses to report him. Instead, the tense progression is altered to follow
the main verb ‘answer’ and the pronoun is changed to the third-person form.
We might suspect that Fielding is just putting quote marks around indirect
discourse, but again there is no shortage of forms like (1) with indirect discourse
but no quote marks. We might suspect that Fielding is just not being meticulous in
his use of the marks, but three facts belie this claim: first, never once does Fielding
fail to put quotation marks around a direct quotation. It would be strange then for
him to accidentally forget to do so for a large number of indirect quotations. Second,
consider this fragment of quote from Mr. Allworthy:
(7) ‘ : : : I cannot avoid saying,

‘there is no Part of your Conduct which I re-
‘sent more than your ill Treatment of that good
‘young Man (meaning Blifil) who hath behaved
‘with so much Tenderness and Honour towards
‘you.’ [vol. II, p. 58]
6
Henceforth, I will index co-referring terms when necessary for clarity.
7
As a curiosity for the reader, Lynne Truss in her popular book on punctuation, Eats, Shoots, and
Leaves, presents one of these unusual uses of quote marks by Fielding as an example of correct
punctuation, seemingly oblivious to its (current) unacceptability. See p. 155.
286 M. Johnson
Fielding isn’t being sloppy in his use of quote marks. When he adds words that the
original speaker did not use, he puts them in parentheses. This is someone who cares
about distinguishing what he, as narrator, says, from what his characters say. Third,
the existence of mixed quotations highly suggests that a careful distinction is being
made by the use of quote marks.
Some useful context may be provided by (8). Here, Master Blifil is protesting
that he didn’t provoke Tom into assaulting him, so the “his mouth” is the mouth
of Blifil. Fielding reports the content of what Blifil said (that he had made use of
no such appellation)—without quoting it, because again, it’s just the content. Then
he goes on to report what Blifil “added.” The addition here is not more content,
because it’s roughly Blifil just saying that he had made use of no such appellation.
Instead, Fielding is reporting the words that Blifil used in his protests. That’s Blifil’s
“heaven forbid” and Blifil’s “should ever” and Blifil’s exclamation point. Fielding
is reporting Blifil’s speech verbatim—except for the personal pronoun (and perhaps
the tense of ‘should’).
(8) he1 positively insisted, that he1 had made Use

of no such Appellation; adding, ‘Heaven forbid
‘such naughty Words should ever come out of
‘his1 Mouth.’ [vol I, pp. 123–124]
Other examples seem to attest to this general view, that quotation in these cases
suggests verbatim reporting, with grammatical alterations. For example, in (9), it
seems unlikely that Fielding is calling Tom “that fine spark” – this is the petty-
fogger speaking:
(9) Jones1 had no sooner quitted the room, than

the Pettyfogger, in a whispering tone, asked
Mrs. Whitefield2 ‘If she2 knew who that fine
‘Spark1 was?’ [vol. IIp. 214]
Example (9), like (10) and (11), also shows that quote marks sometimes appear
around what otherwise seem like indirect questions, though as (2) demonstrates, not
for every indirect question.
(10) : : : He1 then con-

cluded by asking, ‘who that Partridge was, whom
‘ he2 had called a worthless Fellow?’ [vol. I, p. 82]
(11) : : : she1 asked her2 , ‘How

‘she2 had the Assurance to mention her1 Name
‘with Disrespect?’ [vol. II, p. 116]
In many instances, the grammar seems to force the indirect forms within the
quotation. So, for example, in (12), pronouns referring to the speaker are in the third
person and ‘beg’ takes an infinitival complement, which isnot the language with
which one begs. But since Fielding is reporting what Sophia said (amended only
for grammatical purposes), he puts quote marks around it and continues in the next
sentence to standard direct quoting.
(12) : : : she1 begged him2 , ‘not to

‘make her1 the most miserable Creature on Earth
‘by forcing her1 to marry a Man whom she1 de-
‘ tested. This I1 entreat of you2 , dear Sir,’ said
she1 , ‘for your2 Sake, as well as my1 own, since
‘you2 are so very kind to tell me1 your2 Happiness
‘depends on mine1 .’ [vol. II, p. 38]
We see something similar in (13) with ‘entreat’; here Fielding follows with more
indirect speech imposed by the complementizer ‘that’ (which disprefers direct
speech). It’s also worth noting that when Fielding is reporting what someone is
“adding” or “urging” or “saying” (the gerund participle) we often get the indirect
forms, with quotation marks. This may be grammatical in nature as well.
(13) She1 entreated her Aunt, ‘to have Mercy

‘upon her1 , and not to resent so cruelly her1 Unwilling-
‘ness to make herself1 miserable; often urging
‘that she1 alone was concerned, and that her1
‘ Happiness only was at Stake.’ [vol. II, p. 86]
The data then are these. Quote marks always appear around direct forms and
sometimes appear around indirect forms in eighteenth century novels.8 Quote marks
around indirect forms seem to indicate that the speech contained therein is reported
verbatim, with the exception of necessary grammatical changes in pronouns, in the
tense, in word order (for indirect questions) and in forms of address.9 The reasons
for choosing indirect forms in these cases are sometimes but clearly not always
grammatical. It is also possible, as I’ll suggest, that indirect reporting is not seen by
these authors as incompatible with verbatim reporting.
8
Let me be clear that I mean “some 18th Century novels.” Q-mark usage was highly non-
standardized at this point in time. For example, in 1766 both The Count of Otranto and The Vicar
of Wakefield were published; the former uses no Q-marks and the latter uses Q-marks more or less
as we do (but with a new left quote at the beginning of each new line). So there is no one usage to
be analyzed, just different practices amongst different writers and editors.
9
“Sir,” which appears frequently in direct quotes in Tom Jones, never once appears in a quoted
indirect form. I should also note that none of Fielding’s eye-dialect from French or German
characters shows up quoted in indirect forms either.
288 M. Johnson
3 Meaning as Use (for Punctuation)
Suppose we live in a society where each individual x obeys rule (1). Furthermore,
it’s common knowledge that the rule is obeyed. I won’t assume that it’s a
convention, because it’s not clear what coordination problem obeying the rule solves
(Lewis 1969). Today we observe that Jeb is wearing a pink and white polka-dotted
hat. We reason:
1. For all x, x wears a pink and white polka-dotted hat when and only when it’s
x’s birthday. [Common knowledge]
2. Therefore, Jeb wears a pink and white polka-dotted hat when and only when
it’s Jeb’s birthday. [1, universal elimination]
3. Jeb is wearing a pink and white polka-dotted hat. [Observation]
4. Therefore, it’s Jeb’s birthday. [2, 3, biconditional elimination]
Now, we don’t live in a society where everyone obeys this particular rule, but
the case illustrates how we can acquire information from individuals by observing
their non-linguistic behavior, when this behavior is governed by rules everyone is
presumed to obey.
We do live in a society where most individuals, on most occasions, obey the
following rule: x places Q-marks around all and only quotations of sufficient length.
Indeed, there are institutional forces in our society that police conformity to this
rule: students are failed, or even expelled, people have their degrees revoked, books
are pulled from the shelves, and writers lose careers for not properly punctuating
quotations with Q-marks (or other devices of indicating quotation). So we can and
do reason as follows: Suppose Davidson writes:
Quine said that quotation “ : : : has a certain anomalous feature.”
1. For all x, x places Q-marks around all and only quotations of sufficient length.
[Common knowledge]
2. Therefore, Davidson places Q-marks around quotations of sufficient length.
[1, universal elimination]
3. Davidson placed Q-marks around the words “has a certain anomalous
feature.” [Observation]
4. Therefore, Davidson’s use of “has a certain anomalous feature” was a
quotation. [2, 3, biconditional elimination]
Let me make a few remarks here. First, in order to explain how quotation marks
signal content to readers, it is unnecessary to assume they are linguistic in nature
or that they have semantic content, or affect semantic content. The fact that the
analogue of quotation marks doesn’t occur in speech; that the first 1,000 years of
English writing did not have any such marks; and that all writing systems outside
the European writing systems lack or have borrowed these or similar devices from
Europe strongly suggests to me that Q-marks do not articulate some unpronounced
linguistic element, but instead are recently invented decorations of written material
that signal much the way the polka-dotted hat signals in our earlier example. Q-
marks are punctuation marks.10
I’ll make a somewhat more explicit argument. I assume that there is some
level of linguistic representation in the mind at which things like the syntactic
structure of a sentence are represented, as well as the identities of the lexical
items that occupy that structure. Call these things that are represented at this level
“linguistic items.” Furthermore, I assume that only linguistic items contribute to the
meanings of sentences. Meanings are assigned by compositional mental processes
that take the meanings of the lexical items and compose them in a certain way that
respects the grammatical structure they inhabit. Spoken and written items that don’t
articulate linguistic items are thus not relevant to sentence meanings. Q-marks don’t
articulate linguistic items. They don’t articulate syntactic structures, at least not
on any syntactic theory I’m aware of. Furthermore, they don’t articulate lexical
items. The evidence for this claim is that no lexical item freely appears in all
syntactic positions, but as the data from mixed quotation demonstrates, Q-marks
can appear anywhere, even surrounding non-constituents. Therefore, on the model
I’ve elaborated, Q-marks are irrelevant for assigning semantic contents to sentences.
This is fine: as we’ve seen, an explanation of how Q-marks nevertheless succeed in
conveying information to readers does not require that they have a semantic effect,
or articulate a linguistic item that has a semantic effect.
Depending on one’s view of pragmatics, it may follow that the content conveyed
by Q-marked quotations is pragmatically conveyed (for instance, if one’s view
is that pragmatics is whatever is not semantics). But it is important to see that
this content isn’t conveyed by standard types of pragmatic mechanisms, such as
presupposition or implicature. I’ve argued for this before (see Johnson 2011), so
here I’ll just remark that with regard to presupposition, quotation entailments don’t
project out of the antecedents of conditionals or over negation, like presuppositions
do; and with regard to conversational implicature, quotation entailments aren’t
cancelable,11 as implicatures are. The mechanism of content conveyance is the
one elaborated above: There are rules of punctuation that, in most contexts,
writers are presumed to obey. Institutional pressures and sometimes paid agents
(i.e. copyeditors) police compliance with these rules. This widespread compliance
10
Several of the theses here are defended by others in the literature. Washington (1992) holds that
quotation marks are punctuation marks; Wertheimer (1999) denies that they have semantic values;
and Saka (2011) entertains the thesis that they are non-linguistic marks.
11
The example from Johnson (2011) is this:
Alice told me, “snow is white.” #But she didn’t utter ‘snow is white’ or any of its translations
into foreign languages.
I of course made the example up; negative evidence, by its nature, does not occur in the wild. In
my opinion the burden is on the theorist who thinks quotation entailments are conversationally
implicated to produce compelling examples of cancelation, which abound for standard cases of
conversational implicatures.
290 M. Johnson
licenses inferences about, for example, which turns of phrase a writer takes to be
quotations. To give the meaning of a punctuation mark is to say how it is to be used.
But the rules are never so simple. I have stated the rule for Q-marking quotations
as follows: place Q-marks around all and only quotations of sufficient length. And
I have defined quotation as follows: repeating someone’s words. But we frequently
Q-mark expressions that are not completely faithful reproductions of others’ words.
Quotation is a balancing act. The original speaker of some phrase spoke at a
particular time, in a particular language, to a particular audience. When we quote
that speaker, we occupy a different context. We must therefore balance our duty of
faithfulness to the reported speaker with our (potentially) conflicting duties to our
audience, such as to speak in language that is intelligible, grammatical, and polite.
For example, we Q-mark translations of the words of non-English speakers into
English, as in (14):
(14) Nietzsche said, “God is dead!”
Q-marks can thus appear around words and phrases that the attributed speaker never
said in the strictest sense. In the strictest sense, Nietzsche said, “Gott ist todt!” but no
one bats an eye, in most contexts, if we report Nietzsche with (14). (14) is frequently
preferable, as our interlocutors might not know what Nietzsche meant when he said,
“Gott ist todt!” if that’s how we report him (Cappelen and Lepore 2007: ch. 5).
There are other reasons we choose to alter texts and still accept those texts as
verbatim reproductions (in the strictest sense that meets our needs). Profanity taboos
are quite powerful in our society—even the paper of record sacrifices accuracy in its
direct quotations for the sake of avoiding words like ‘fuck,’ as exhibited in (15).
(15) “I’m like, ‘What are you doing?’” he recalled shouting, using his own
expletive. (New York Times, June 27, 2014)
I assume the target of the quote is something like, “What the fuck are you doing?”
Such an inference requires the background knowledge that the Grey Lady would
print ‘What the hell are you doing?’ but wouldn’t print ‘What the fuck are you
doing?’ unless the latter were spoken by the president, as well as knowledge of the
fact that the reported speaker is not the president. As I said, the rules are never so
simple.
It’s also a regular journalistic practice to report statements like (16) (taken
from the comments section of a sports website)—especially when spoken—in more
Standard English like (17).
(16) Bron is good. Probably top 20 all time. But he no MJ neither man.
(17) The commenter said that LeBron James is “good. Probably top twenty all
time. But he is no MJ [Michael Jordan] either.”
Preserving the non-standard dialect features of (16) (negative concordance, copula
omission) in the report (17) can be viewed as racist and/or classist. First, non-
standard dialect features, though present among whites and non-whites, are dispro-
portionately quoted verbatim for non-whites, and can present them as uneducated.
Second, the quoted speakers may be perfectly competent with standard English;
preserving the nonstandard dialect features may be a form of inappropriately taking
the speaker out of context: presenting his or her claims to an audience in a way that
person would not themselves present those claims to that audience. However that
may be, newspaper policy often requires “clean up” of quotes like (16).
We correct slips of grammar that are obvious slips and that would be needlessly
embarrassing. We remove verbal mannerisms such as ‘ahs’, routine vulgarities and
meaningless repetitions. Otherwise we do not revise quotations. [The Canadian Press’
Stylebook]
The rule for quote marks is something like: if a passage contains the exact words
some speaker used, altered only (if at all) for sufficient reasons that are generally
recognized as sufficient reasons to alter a verbatim report (such as translation to
the reporter’s language, avoiding taboo language, and cleaning up infelicities and
non-standard dialect features), then put quote marks around the speech report. This,
incidentally, explains the unsystematic nature of the types of alterations allowed
within quote marks. Cappelen and Lepore (2007), for instance, chalk the lack of
systematicity up to a complicated, uncharacterizable “saying relation.” Locating
the special exceptions to verbatim reporting in the rules for using quotation marks
feels more natural, and as we shall see, helps to better understand variation through
history.
I think we can understand the eighteenth century novel data as conforming to the
same rule that we conform to today. The difference for Fielding is that, for example,
the dictates of English grammar are sufficient reason (for him) to put words within
quote marks that the speaker didn’t use. When Fielding decides to report Sophia’s
utterance (18), he balances several considerations.
(18) Don’t make me the most miserable creature on earth by forcing me to marry
a man whom I detest.
A maximally faithful report (19) is unfortunately marked: ‘beg’ doesn’t take finite
complements.
(19) She begged him “don’t make me the most miserable creature on earth by
forcing me to marry a man whom I detest.”
Thus he chooses an indirect form, grammatically appropriate to the circumstances
(20):
(20) [S]he begged him “not to make her the most miserable creature on earth by
forcing her to marry a man whom she detested.”
292 M. Johnson
The quotation marks are retained because the quoted material is still what Sophia
said, altered only as much as compelling and sufficient reason demands12 (just as
we today translate, omit profanities, or alter dialect forms).
Even after quote marks stopped appearing around indirect forms like these
that we’d now leave bare authors like Jane Austen retained them for free indirect
discourse, as this passage from Sense and Sensibility (1811) illustrates. Every third-
person singular pronoun in (21) refers to Mr. John Dashwood:
(21) The prospect of four thousand a-year, in addition to his present income,
besides the remaining half of his own mother’s fortune, warmed his heart,
and made him feel capable of generosity.— “Yes, he would give them three
thousand pounds: it would be liberal and handsome! It would be enough to
make them completely easy. Three thousand pounds! he could spare so
considerable a sum with little inconvenience.”13
Our current quotation practice has no way of indicating how close a correspondence
there is between indirect forms in speech reports and the original direct forms.
Cappelen and Lepore (2007) adopt the view that “what these cases [like (14)–
(17)] teach us is no more than something about the conditions under which a speaker
can stand in a saying relation to a certain quoted item” (Cappelen and Lepore 2007:
47). But I think this is perhaps not the best way to understand the data historically.
On such a view, either the saying relation itself has changed since the eighteenth
century (!) or the meaning of ‘say’ and every other verb that can be followed with
direct quotes has changed since the eighteenth century; neither of which possibilities
strike me as particularly compelling. Alternatively, we can instead see the data as
telling us about the rules that govern punctuating reported speech with quote marks.
Indeed, we can understand why Fielding and Austen faced a different calculus
of markedness and faithfulness than we of contemporary practice face. Q-marks
were introduced to indicate quotations in novels. Though we still use them for this
purpose, our practice is more centered around journalism and scholarship. There
is a strong moral obligation to report faithfully the subjects of news items and
the scholarly work of other researchers. No such moral obligation attends to accurate
reporting of fictional characters by their creators. Thus it is unsurprising that early
use of Q-marks tended toward less faithful, but less marked forms.
In conclusion, Q-marks are a type of punctuation mark. To give their meaning
is to state how they are to be used. One of their purposes is to indicate quotations
(hence their common name: quotation marks). But the practice of quotation itself
12
Why is the pronoun ‘me’ shifted to ‘her’? Not for grammatical reasons, as an anonymous
reviewer points out. Though I talk of ‘altering’ forms for sufficient reasons, that perhaps is not
the best model. Rather, Q-marking is more of an ex post facto endeavor. What gets written is
standard speech with a mix of direct and indirect forms: “she begged him not to make her the
most miserable creature : : : This I entreat of you : : : ” Then, afterwards, the question is where
the Q-marks go. Here, they go around the indirect forms, because other than their indirectness,
everything is as the original speaker said it.
13
I haven’t found an earlier edition yet, but the quote marks are present in an 1833 edition published
by Richard Bentley, p. 3.
can involve conflicting obligations: an obligation to faithfully report the quoted

speaker, and an obligation to speak intelligibly, grammatically, and politely to one’s
audience. Over time, as the practice of Q-marking quotations has moved from its
limited role in the Bible, to a broader role in the novel, to writing in all formal
contexts, our views of our obligations have changed, particularly of our duty to
fidelity, and this is reflected in the change in our Q-marking practice over time.
4 Quotation Is Using Someone’s Words
There are lots of “theories of quotation,” and a number of them, in my view, conflate
two, historically distinct uses of Q-marks. As a result, these theories miss what I
believe to be a fundamental feature of quotation, namely, that in quoting, we do not
merely repeat the words of someone, but we re-use them.
In a separate historical development, Q-marks came to be used to indicate when
words are being used to refer to a linguistic object associated with them, such as
their written form, their meaning, or perhaps themselves. Consider two examples:
(22) Das Leben ist kein Ponyhof. This is a very common expression in German
that means “life isn’t a place for riding ponies.”14
(23) As X notes in the comments, it’s not fake as in “created by photoshop”, but it
IS fake in the sense of being added as an ironic joke by a company known for
such things.15
In example (22), the speaker wishes to refer to the meaning of the German
expression ‘Das Leben ist kein Ponyhof,’ and does so by using a Q-marked English
expression that has that meaning. In example (23), the writer wants to distinguish
two senses of the word ‘fake’ when applied to a photo. One of the senses is named
by a definite description: ‘the sense of being added as an ironic joke’; the other sense
is named by a Q-marked expression that has that sense.
Indeed, there is a broad practice of using Q-marked expressions to refer to words
or similar objects (like word types, or spellings).
(24) More so than most words, “as” can be translated to Spanish in many
ways—probably dozens, and you often can’t substitute one of them for
another.16
(25) Some do spell it “Emmenthal” as well, but that’s not the way it is spelt in
Switzerland where the original comes from.17
14
From “23 German Slang Words Your Textbook Isn’t Teaching You,” www.fluentu.com/german/
blog/, 2015.
15
From “Helpful Label,” http://languagelog.ldc.upenn.edu/nll/, 7 July 2014.
16
From “Translating ‘As’ to Spanish,” http://spanish.about.com/od/translationsfromenglish/a/as.
htm
17
Comment on “Kennedy Town Hamburger” on Hong Kong Geo Expat Forums by user PDLM,
https://hongkong.geoexpat.com/forum/321/thread29393.html, 26 July 2008.
294 M. Johnson
In example (25), the pronoun ‘it’ refers to the name ‘Emmental’ (prior context
makes this clear) and ‘Emmenthal’ enclosed in Q-marks refers to one spelling of
that name.
The typological limitations of early printing practices have unfortunate echoes
into the present day. Q-marks have been used for many, diverse purposes: quotation,
commonplacing, distancing, reference to linguistic objects, and so on. But it’s
important not to conflate these different uses,18 which are governed by distinct
rules. In examples (22)–(25) above, no quotations occur. People have indeed said
“created by photoshop” and “as” and “Emmenthal” and “life isn’t a place for riding
ponies,” but the writers here are not quoting any of those people. Instead, they are
using linguistic expressions to refer to linguistic objects that are related to those
expressions. A separate punctuation rule directs us to Q-mark these expressions as
well. But this doesn’t make them quotations and more importantly, it doesn’t make
quotations references to linguistic expressions.
Indeed, in my opinion, this is the fundamental feature of quotation. When we
quote someone, we don’t just produce words they previously produced, and consider
those words as some uninterpreted formal object. We use those words. We use them
to characterize the contents of assertions and beliefs; we use them to assert, to
express propositions, to refer to objects. We use them in our context, and we are
responsible for providing them with sufficient context to be interpretable.
The evidence that this is so has been with us since the beginning of analytic
investigation into the phenomenon. Davidson (1979) taught us that in mixed
quotation, the quoted expressions couldn’t be interpreted as reference to a linguistic
object.19 For example, consider (26):
(26) He described the 2-iron as being “very important and instrumental in my

success here.”20
Mickelson, the reported speaker in (26), did not describe the 2-iron as being an
expression, he described it as being important and instrumental to his success in the
British Open links course. But the report is not false, as the quoted expression is
being used and not merely mentioned. Still, there are holdouts. Recanati (2001), for
18
To fully substantiate the claim that these uses are historically separate would require us to show
that they appeared at different points in history. Unfortunately (a) reference to linguistic objects
is not exceedingly common and (b) historians of punctuation are very interested in quotation, but
not in so-called pure quotation. So I can tell you that the first attestation of quote marks used to
indicate direct speech is in 1574, in a book of poems called The Mirour for Magistrates (Houston
2016), but I don’t have a single instance of pure quotation before the nineteenth Century. In what
context it first appeared and whether it was used alongside other types of Q-marks is, as far as I’m
aware, an open question.
19
I am not here endorsing Davidson’s positive account, which I think also misses the fundamental
feature of quotation, merely giving him credit for the (obviously correct, in retrospect) argument
that mixed quotation cannot be treated as solely referential.
20
“On British Open links courses, irons take on added importance,” Associated Press, 16 July 2014.
example, maintains that direct quotations21 are “closed” and do involve reference
to a linguistic object. As I’ve emphasized, I think such views miss the fundamental
feature of quotation.
Here I will present three arguments that quotation, mixed or direct, involves
re-using the words of someone, not referring to those words. The first involves
propositional-attitude reports. Recall the rule for punctuating quotations: if you
repeat someone’s words, Q-mark them. Since one excellent way of characterizing
someone’s beliefs is using the words with which they expressed those beliefs, we
find Q-marked sentences as the object of propositional-attitude verbs, as in (27):
(27) Thomas Jefferson believed “the Earth is given as a common stock for men to
labour and live on.”22
Here, the sentence, “the Earth is given as a common stock for men to labour and live
on,” cannot coherently be thought of as referring to a linguistic expression. People
don’t believe sentences. However, the contents of their beliefs can be reported by
sentences. In (27), the sentence “the Earth is given as a common stock for men to
labour and live on” characterizes the content of Jefferson’s belief. It’s put in quote
marks because it’s a quote from Jefferson.
Second, consider the case of parenthetical attribution. In standard speech reports,
the attributed content is not asserted. However, suppose we wish to assert something
using someone else’s words. We do this in English with an aside like: as/like so-and-
so said, as in (28):
(28) The way Palestinians are treated by Israel can therefore be characterized as
injustice. And as Martin Luther King said ‘Injustice anywhere is a threat to
justice everywhere’.23
This quote is taken from a pro-Palestine, anti-Israel website. (I am not endorsing any
political positions in this paper.) It is clear that the author is arguing that injustice in
Palestine is a threat to justice everywhere. He asserts as much by using Dr. King’s
words, and parenthetically attributing them. Parenthetical attribution is used in any
cases where we want to use someone else’s words for some purpose other than
characterizing what they said or thought, not just cases of assertion. For example,
consider (29):
(29) If, as Einstein said, “Play is the highest form of research” I am going to do A
LOT of research this year! [tweet from Dr. Cyndi Burnett]
21
By “direct quotation,” I mean of course sentences of the form SPEAKER C VERBA DICENDI
C Q-MARKED SENTENCE. Recanati’s “open quotations” are not closed, but here I’m taking
issue with his view that the other direct quotations are closed.
22
http://www.wealthandwant.com/docs/Foldvary_UTR.htm The Ultimate Tax Reform: Public
Revenue from Land Rent by Fred E. Foldvary, January 2006
23
Adri Neuwhof, “The legacy of Martin Luther King: Injustice anywhere is a threat to justice
everywhere,” Electronic Intifada, https://electronicintifada.net/content/legacy-martin-luther-king-
injustice-anywhere-threat-justice-everywhere/6829
296 M. Johnson
Here again, the quote is used and not for the purpose of reporting Einstein’s speech.
The truth-conditions of the quote are not the same as “if Einstein said, ‘Play is the
highest form of research,’ then Dr. Burnett is going to do a lot of research this year.”
What Dr. Burnett instead said is that if play is the highest form of research then
she is going to do a lot of research (in virtue of her playing a lot). The quote marks
indicate that she is quoting and the prepositional phrase “as Einstein says” attributes
that quote to Einstein, but the function of Dr. Burnett’s utterance is not primarily a
speech report of Einstein.
Sports reporting is a good place to study quotation. A sports reporter is tasked
with telling a story about something with no inherent narrative structure: a sports
match. To establish a narrative, the writer brings in the emotions, thoughts,
intentions, and reactions of players and coaches, typically from interviews before
and after the match. These quotes are then interwoven into the story the writer
tells. Importantly, the quotes don’t just hang there uninterpreted: quotes that contain
context-sensitive expressions are provided context by the reporter so they can be
interpreted as part of the story. Additionally, the quotes introduce new individuals
and events into the narrative, and those objects are then subsequently available for
anaphoric reference.
(30) The Spurs knew their defense had to be better : : : They held Miami to 35%
in the first half. “They played great and I can honestly say I don’t think any
of us were expecting this type of performance,” Heat coach Erik Spoelstra
said.24
Here, we know that Spoelstra’s ‘they’ refers to the Spurs, the current topic in the
article. We know that his ‘this type of performance’ refers to the Spurs holding
Miami to 35 percent field goal percentage in the first half. We’re not just being told
some words that Spoelstra uttered, we’re given the context to interpret it, and we do.
Those interpretations become part of the discourse, available for subsequent use:
(31) The Stanley Cup Final hasn’t ended in a sweep since 1998, when the Detroit
Red Wings did it against the Washington Capitals. “Nothing is done, nothing
is finished,” Quick said. “We still have a lot of work to do.” Not as much as
they would have had if Quick didn’t come through with a vintage
performance.25
Here we have a direct quote from Quick, “We still have a lot of work to do.”
This licenses an anaphor in Rosen’s next sentence. Quick’s ‘we’, referring to his
team the Kings, licenses Rosen’s ‘they’ anaphorically referring also to the Kings.
If the direct quote didn’t contain a reference to the Kings, then the only objects
anaphorically available for Rosen would have been the previously mentioned Red
Wings and Capitals. Additionally, two ellipses in Rosen’s sentence “Not as much
24
“Spurs take 3–1 lead with 107–86 victory over Heat,” Brian Mahoney, nba.com, 13 June 2014.
25
“Kings blank Rangers, move within one win of Cup,” Dan Rosen, nhl.com, 10 June 2014.
(work) as they would have had (to do)” are licensed by Quick’s ‘work’ and ‘to do’,
respectively.
An anonymous reviewer helpfully suggests that it might be unclear how the
data in (30) and (31) are objections to alternative accounts. For example, on
demonstrative theories, while the Q-marked material is not taken to be a syntactic
or semantic part of the sentence, it still exists, with its syntax and semantics. But
noticeably, merely demonstrated linguistic material does not, for example, introduce
discourse referents. Imagine a poster on the wall saying “We still have a lot of work
to do.” Now consider (310 ):
(310 ) ???Quick said that. Not as much as they would have had.
Discourse referents can’t be introduced or retrieved by linguistic material simply in
the vicinity of a discourse—they must be introduced or retrieved by material used
in that discourse.
So let me revise the definition of quotation I’ve been working with: quotation is
re-using the words of someone,26 It’s taking those words from their original context,
and using them in a new context, with the attendant obligations of providing context
for interpretability, and the attendant effects of introducing objects and events into
the discourse. This can be obscured by the separate use of Q-marks to punctuate
reference by linguistic items to related linguistic items, and it can be obscured by
the fact that quotations are most frequently used in speech reports. But we can assert
quoted material and we can use it to help tell our stories—to introduce objects and
events into them.
5 Shifty Indexicals
In direct and mixed quotations, context-sensitive expressions behave unusually.

Consider the quotation in (32):
(32) After beginning with a serious assessment of Wainwright’s pitches, Jeter

pivoted suddenly. “I don’t know, man,” he said, to laughter.27
The standard story regarding the indexical ‘I’ is that its use is governed by a
strict rule, that it always refers to the speaker who uses it. (32) is taken from an
article written by Tyler Kepner, but ‘I’ as it occurs in (32) does not refer to Kepner,
26
This is actually too strong, I believe. For example, the numbered sentences and passages that
appear in this paper as examples are not used by me, but they are quotations. My reason for
asserting the stronger thesis, however, is that I think “normal” cases of quotation in literature
and journalism and scholarship are uses of the quoted material. Only in special cases—like in
linguistics and philosophy of language—do we repeat the words of others in order to examine
them as formal objects.
27
“Jeter Takes Advantage, So Does the American League,” Tyler Kepner, New York Times, 16 July
2014.
298 M. Johnson
but rather to the reported speaker, Jeter. Additionally, Kepner’s report of Jeter is in
the non-past tense, which ordinarily requires that it describe a state or event that is
not in the past. But the reported content is true as long as Jeter didn’t know at the
time of the interview, before the time at which Kepner wrote the article.
On a mere mention account of direct quotation, there’s no conflict between the
characters and the behavior of the indexicals in (32), ‘I’ and the non-past tense
marker:
There is a way to control an indexical, to keep it from taking primary scope, and even to
refer it to another context (this amounts to changing its character). Use quotation marks. If
we mention the indexical rather than use it we can, of course, operate directly on it. (Kaplan
1989: 510–11, emphasis original)
According to a mere mention account, (32) merely mentions an expression of

Jeter’s: the words ‘I’ and ‘don’t’ are not used in (32), and ipso facto they are not
used unusually.
But, as I’ve argued, quotations are re-uses of someone’s words. They aren’t
referred to another context, we use them in the one we’re in. Thus Kaplan’s
proposed mechanism by which the character of a quoted indexical can change is
unavailable to us.
There’s another proposal in the literature for how the character of a quoted
indexical can change: this is the “monster account” of Cumming (2003) and Bittner
(2007). Cumming and Bittner reject Kaplan’s view that there are no “monsters,” that
is, character-shifting operators, in natural language. Indeed, on their view, Q-marks
in quotations articulate these operators. There are two reasons I don’t endorse the
monster account, one theoretical and the other empirical. On the theoretical front,
I have argued that Q-marks can’t articulate any linguistic item, and thus can’t have
semantic effects. The empirical argument once again has to do with the neglected
historical data. Recall (12), repeated here:
(12) : : : she1 begged him2 , ‘not to

‘make her1 the most miserable Creature on Earth
‘by forcing her1 to marry a Man whom she1 de-
‘ tested. This I1 entreat of you2 , dear Sir,’ said
she1 , ‘for your2 Sake, as well as my1 own, since
‘you2 are so very kind to tell me1 your2 Happiness
‘depends on mine1 .’ [vol. II, p. 38]
In a single quotation, both third-person pronouns (her, she) and a first-person
pronoun (I) refer to one and the same person. Both past (detested) and present
(entreat) forms are used to refer to one and the same time. It’s true that no account
of present-day quotation needs to make room for eighteenth century data. But as
I’ve emphasized, there is a strong continuity between the practice of Q-marking
quotations in the eighteenth century and our present practice. It would be strange
indeed if radically different mechanisms were needed to explain the two practices.
So, what remains of our options? Here are three theses that stand in contradiction:
(a) ‘I’ always refers to the speaker who uses it; (b) there are no character-shifting
operators in natural language; and (c) in some quotations, ‘I’ refers to someone
other than the speaker who uses it. Kaplan and other pure-mention theorists reject
(c). The monster account rejects (b). I therefore reject (a). ‘I’ doesn’t always refer
to the speaker who uses it. Sometimes it refers to someone else, and some of those
times are in quotations.28
I suppose I can’t reject such a well-established principle as (a) without giving
some justification beyond, “it’s what my theory needs to be true.” So I want to
engage in a brief discussion of two common types of shiftiness within natural
language.
One well-recognized phenomenon is called the historical present. This is a
narrative technique in which events that occurred in the past are referred to with
the nonpast tense. For example, here is Ophelia explaining to her father Polonius
some odd, past behavior of Hamlet’s:
(33) He took me by the wrist and held me hard;

Then goes he to the length of all his arm;
And, with his other hand thus o’er his brow,
He falls to such perusal of my face
As he would draw it.
I’ve underlined the two non-past tense verbs Ophelia uses to describe the past
events. Importantly, Ophelia could have used the past tense with no change in
referential meaning. She’s not pretending to be in the past. Her context is the present
and from within it, she is referring to past events with the non-past tense. Wolfson
(1989) analyzes performed narratives (people telling stories) and finds that they
seamlessly alternate between past and non-past in describing the past events the
stories are about. For example, in one story she analyzes, “there are 85 verbs that
could, according to their time orientation, have been in the CHP [conversational
historical present], and of these there are 57 that are realized in that tense” 141).29
In Wolfson (1989)’s view, which I share, the historical present (she uses “CHP”)
is used to partition the events of the story from one another. For example, the events
in one story she looks at take place on a boat, then on land, then back on the boat. The
action on the boat is characterized with the past tense, and the action on land with the
non-past tense. She compellingly demonstrates that this partitioning function, and
not enhanced drama by making events seem more “present,” is the true function of
using the historical present. For example, in the boat story, the events characterized
by the historical present are the least dramatic in the story (p. 143).
One thing I want to note is the observation we began with regarding why Q-marks
were co-opted for a quote-indicating function in the first place. In novels, multiple
speakers exchange remarks, and it is necessary to indicate when one person’s speech
28
See Jaszczolt and Huang (2017) for a position along these lines that invokes Default Semantics.
29
Another revealing piece of evidence is that one and the same sentence can have both past tense
and historical present forms: “When we drove up, I see all these kids : : : ” Example from Wolfson,
p. 146.
300 M. Johnson
ends and another’s begins. When people started to write novels, such a device was
needed. But without them, authors could at least distinguish the narrator’s voice
from that of the characters. Most eighteenth century novels are narrated in the past
tense (but not, for example, epistolary novels). The present-tense forms within direct
quotations are themselves a means of partitioning what the characters say from what
the narrator says. This partitioning of contents may be one of the functions of so-
called direct quotation.
There’s another kind of shiftiness that more closely relates to indexicality and
quotation. It might be what Recanati (2001) calls “open quotation,” but I’m not
sure. I’ll start with an example of the phenomenon I’m thinking about. (34) is taken
from a transcript of the show Politically Incorrect, which is a late-night infotainment
program hosted by Bill Maher. Maher is asked why Judd Gregg, a Republican,
accepted a position in Obama’s cabinet and then decided subsequently to turn it
down. He responds:
(34) Well, because it’s a big Washington job. I think it probably was a pay
increase and a status increase and I’m in the Cabinet. And then I guess his
Republican friends got to him—you know, like the mob.
On my view, this is perhaps a case of quotation. As I said before, you can quote
what someone should have said or would have said. But I kind of doubt that Maher
thinks Gregg would have said, “I’m in the Cabinet,” had he been appointed to the
Cabinet. Instead, I think this is a natural sort of shiftiness we engage in when we do
things like characterize others’ reasons. Maher is asked to give Gregg’s reasons for
his actions, and it’s quite natural in this circumstance to take Gregg’s perspective,
and use one’s pronouns as if one were Gregg.
While I said that this may be a case of Recanati’s “open quotation,” much of
what Recanati writes makes it difficult to tell. For example, many of Recanati’s
definitions presume his views. He says, “A quotation which is not closed is (as
one might expect) open,” and “Whenever a linguistic demonstration (a quotation) is
recruited in this way and serves as a singular term, filling a slot in the sentence, I
say that the quotation is closed,” (649). There are a number of issues here. First, as I
indicated, I’m not sure (34) contains a quotation at all, so in that sense, it would
not be in the purview of Recanati’s “open quotation.” Second, when I consider
Recanati’s own examples, the cases he diagnoses as closed strike me as containing
either no quotation or a quotation not used as a singular term. So if I were to apply
the definitions given my own views, most quotations would be open, but that doesn’t
seem to be Recanati’s view. Normally, we might turn to the examples an author gives
of the phenomenon to help nail it down, but here again, things are difficult. Recanati
only positively characterizes one non-mixed-quotation case as an open quotation:
(35) Stop that John! ‘Nobody likes me’, ‘I am miserable’... Don’t you think you
exaggerate a bit?
The problem here is that this particular case seems like mockery, and that is
definitely not the phenomenon I’m interested in pointing to. When we mock others,
we do repeat their words, but we pretend to be them, and pretense doesn’t involve
shifting terms’ referents. When I pretend I’m Chuck Norris, my uses of ‘I’ all
refer to me. It’s just that in the pretense, I’m Chuck Norris. I’m rather interested
in cases, like (34), that seem to involve a shift in reference, engendered by the
explanatory position taken on by the speaker. I thus leave it to others to determine
whether the shiftiness I’m interested in is open quotation or not. Once again, it’s
of note that quotations are frequently used to characterize others’ speech acts and
beliefs, and hence their perspective. I doubt that in the standard case, direct speech
reports involve the sort of perspective-taking that results in Maher’s spontaneous
shiftiness. My point in bringing up the historical present and the spontaneous
taking of other individuals’ perspectives is twofold: first, they illustrate that the
character of indexical expressions is more loosely constrained than might first
appear; and second, they show that speech and attitude reports involve the sorts
of features (partitioning parts of a narrative, characterizing people’s perspectives)
that elsewhere help engender shiftiness. This makes it somewhat more reasonable
perhaps to suppose that speech and attitude reports are yet a third case where such
shiftiness can happen.
6 Conclusion
To quote is to re-use someone’s words (perhaps one’s own). This involves removing
them from their original context and using them in a new context, where they might
be understood differently. We thus must balance a pair of potentially conflicting
duties: the duty to faithfully report our sources and the duty to intelligibly report
them to our audience.
For the first thousand years of English writing, there was no means of indicating
when some piece of writing was a quote. The Q-mark, before the eighteenth century,
was used with a wide array of functions by printers, such as commonplacing. With
the rise of the novel in the eighteenth century, the Q-mark was co-opted into the
function of signaling quoted speech; this helped distinguish the varying voices of
the characters, and the narrator. Early adopters of Q-marks balanced their duties of
fidelity and intelligibility in a different fashion than writers of today, but the essential
function of indicating quoted speech has remained unchanged for the past 275 or so
years.
References
Bittner, M. (2007). Online update: Temporal, modal, and de se anaphora in polysynthetic discourse.
In C. Barker & P. Jacobson (Eds.), Direct compositionality (pp. 363–404). Oxford: Oxford
University Press.
302 M. Johnson
Clarke, A. (1837). The holy bible: Containing the old and new testaments, the text carefully printed
from the most correct copies of the present authorized translation, including the marginal
readings and parallel texts. With a commentary and critical notes designed as a help to a
better understanding of the sacred writings. New York: T. Mason & G. Lane.
Cumming, S. (2003). Two accounts of indexicals in mixed quotation. The Belgian Journal of
Davidson, D. (1979). Quotation. Theory and Decision, 11(1), 27–40.
Estill, L. (2014). Commonplace markers and quotation marks. ArchBook. http://drc.usask.ca/
projects/archbook/commonplace.php. Accessed 20 Jan 2015.
Finnegan, R. (2011). Why do we quote: The culture and history of quotation. Cambridge, UK:
Open Book Publishers.
Houston, K. (2016). Shady characters: The secret life of punctuation, symbols & other typograph-
ical marks. New York: W. W. Norton & Company.
Jaszczolt, K., & Huang, M. (2017). Monsters and I: the case of mixed quotation. In P. Saka & M.
Johnson, M. (2011). The punctuation theory of quotation. In E. Brendel, J. Meibauer,
& M. Steinbach (Eds.), Understanding quotation: Linguistic and philosophical analyses
(pp. 209–230). Berlin: Mouton de Gruyter.
Lewis, D. (1969). Convention. Cambridge, MA: Harvard University Press.
Saka, P. (2011). The act of quotation. In E. Brendel, J. Meibauer, & M. Steinbach (Eds.), Under-
standing quotation: Linguistic and philosophical analyses (pp. 303–332). Berlin: Mouton de
Gruyter.
Wertheimer, R. (1999). Quotation apposition. The Philosophical Quarterly, 49, 514–519.
Wolfson, N. (1989). The conversational historical present. Linx, 20(1), 135–151.
Ideo- and Auto-reflexive Quotation
Wayne A. Davis
Abstract I focus on the two uses of quotation marks illustrated by The word ‘car’
expresses the idea “automobile,” in which the single quotation marks are used
“auto-reflexively” to mark reference to the word quoted and the double quotation
marks are used “ideo-reflexively” to mark reference to the idea expressed by
the word quoted. I account for the difference between auto- and ideo-reflexive
reference within my theory of meaning and indexical reference. Meaning consists
in the expression of mental states, principally thoughts and their parts. Indexicals
express a type of thought part that links to another mental state that determines its
referent. Determinants can be other concepts (as in the anaphoric use of indexicals)
or presentations (as in their deictic use). On my account, the idea “automobile”
expresses the indexical concept expressed by the idea (a definite noun phrase used
restrictively) linked to an introspective presentation of the idea expressed by the
word in quotes. The word ‘car’ expresses the indexical concept expressed by the
word linked to a perceptual presentation of the word in quotes. The quotation marks
are optional punctuation marks, with no referents of their own. I observe that a word
‘w’ can also be used elliptically to mean “the word ‘w.’” I show that my deictic
indexical theory avoids the defects of other accounts, including the demonstrative
theory and Davidson’s theory.
1 Introduction
Quotation marks have a variety of uses. Two are illustrated in the following
sentence: The word ‘car’ expresses the idea “automobile.” The single quotation
marks are used “auto-reflexively” to mark reference to the word quoted. The double
quotation marks are used “ideo-reflexively” to mark reference to the idea expressed
by the word quoted. Whereas I will use single and double quotes to mark this
semantic difference, in standard English both can be used for either purpose, as can
italics. Unmarked words can also be used to refer to themselves or the ideas they
W.A. Davis ()

Georgetown University, Washington, DC, USA
e-mail: davisw@georgetown.edu

https://doi.org/10.1007/978-3-319-68747-6_11
304 W.A. Davis
express. I will account for the difference between auto- and ideo-reflexive reference
within my theory of meaning and indexical reference.
Section 2 provides a brief summary of my account of what it is for words
and other expressions to have meanings, on which words are conventional signs
of mental states, principally thoughts and their parts, and meaning consists of
their expression. Section 3 examines indexical expressions like pronouns, which
have different referents in different contexts. I distinguish three different uses
of indexicals—anaphoric, demonstrative, and deictic—differentiated by the ways
in which their referents are determined. On my account, indexicals express a
distinctive type of concept (thought part), which is capable of being linked to a
determinant, whose referent becomes the referent of the concept. Section 4 looks at
the deictic use, in which the determinant is a presentation focused on the referent.
Section 5 analyzes indexical concepts as composed of a general sortal concept plus
a nondescriptive determiner that determines the specific range of determinants the
concept can link to. Section 6 sets out my view that noun phrases of the form ‘the
G’ used restrictively, along with those of the form ‘this G’ and ‘that G,’ express
indexical concepts whose sortal component is the concept expressed by ‘G.’
After detailing the differences between auto-reflexive and ideo-reflexive refer-
ence in Sect. 7, I focus on ideo-reflexive designators like the idea “automobile” in
Sect. 8. On my account, this designator expresses the indexical concept expressed
by the idea linked to an introspective presentation of the idea expressed by the word
in quotes. Section 9 focuses on auto-reflexive designators like the word ‘car.’ On my
account, this auto-reflexive designator expresses the indexical concept expressed by
the word linked to a perceptual presentation of the word in quotes. The quotation
marks are optional punctuation marks, with no referents of their own. Section 10
looks at the elliptical use of a word ‘w’ to mean “the word ‘w’.” Finally, I show that
my deictic indexical theory avoids the defects of a demonstrative theory in Sect. 11
and of Donald Davidson’s theory in Sect. 12.
2 The Expression Theory of Meaning
I adopt the view that words are conventional signs of mental states, principally
thoughts and ideas, and that meaning consists in their expression. In Davis (2003),
I explain what it is for words to have meaning and express ideas in terms of speaker
meaning and expression, and what it is for a speaker to mean or express something
in terms of intention.
I focus on thinking as a propositional attitude distinct from believing. One can
think the thought that the moon is made of green cheese without believing it, and
one can believe that bats fly without thinking that thought at the moment. Thinking
in this sense differs from believing in being an event in the narrow sense of an
occurrence or activity rather than a dispositional state. We retain our beliefs when
we are asleep or unconscious, but thoughts are something actively going on. The
Ideo- and Auto-reflexive Quotation 305
nouns ‘thought’ and ‘belief’ have the familiar act-object ambiguity, meaning either
thinking and believing, respectively, or their objects. On my view, thoughts (what
we think) are structured events, and a particular kind of mental representation. They
are similar in many ways to sentences, such as having a constituent structure and
in some cases truth values. But their constituents are very different. Sentences
express thoughts, but not vice versa. We think when thoughts occur to us. For S
to think a thought T is for T to occur to S. I define propositions as thoughts with a
declarative structure or, equivalently, as objects of belief and desire. For S to believe
a proposition (declarative thought) P is to stand in a different relation to it than
thinking, one that does not require its occurrence to S.1
I define ideas (or concepts) as thoughts or parts of thoughts, and distinguish
them carefully from conceptions (belief systems) and sensory images (structures of
sensations). Conceptions and images are important forms of mental representation,
but linguistic meaning cannot be defined in terms of them. In addition to occurring
and being parts of thoughts, concepts can be acquired and possessed, and may
become associated with each other.
Unlike Grice (1957), who assumed that speaker meaning entails attempting to
communicate and produce a belief in an audience (false when talking to our pets,
writing individual words, or telling stories), I take expressing a thought, belief, or
other mental state to involve performing an observable action as an indication that
it is occurrent. Thus Sue expressed the idea and meant “automobile” by ‘car’ only
if Sue uttered ‘car’ as an indication that the idea “automobile” is occurring to her.
Indication is a weaker relative of what Grice called natural meaning.
On my account, the meaning of individual words, as well as the non-
compositional meaning of fixed idioms, is given in terms of what ideas they
are conventionally used to directly express. The meaning of phrases and clauses
is provided by a recursion clause, based on conventions to use particular
expression structures to express certain idea structures. ‘Convention’ here means
“customary practice.” Conventions are common actions that are socially useful, self-
perpetuating, and arbitrary. The common goal served by language is preeminently
communication. Conventions are self-perpetuating in a number of ways: precedent
plays a role, as does habit, transmission of tradition, social acceptance and criticism,
and their success in achieving common goals. Conventions are arbitrary in that other
regularities could have served the same purposes and perpetuated themselves in the
same ways.2
My theory is Fregean in important respects, but not all. In particular, I reject the
Fregean assumption that senses and concepts must be descriptive (see especially
Davis 2005). We clearly use ‘Washington’ to express the common component
of the thought that Washington was from Virginia, the thought that Washington
was taller than Adams, and so on. Kripkean (1972) arguments show that a name
1
So in the act sense, thoughts and beliefs are different. But in the object sense, beliefs are thoughts
(though not all thoughts are beliefs).
2
For more on conventions in this sense, see Davis (2003: chapter 9).
306 W.A. Davis
cannot be defined exclusively in descriptive terms. ‘Washington’ does not have a

meaning like that of ‘the first president of the United States.’ Indeed, standard names
appear to be among the primitive or undefinable terms of a language. ‘Washington’
has a nondescriptive sense because it expresses a nondescriptive concept, as do
syncategorematic terms (prepositions, logical constants, etc.)—and indexicals.
3 Indexical Concepts and Determinants
The paradigm indexicals include the personal pronouns I, you, he, she, and it; the
demonstrative pronouns this and that, plus noun phrases with them as determiners;
and the locative pronouns now, here, there, and then. These expressions contrast
markedly with proper names like Washington and definite descriptions like the
fourth planet from the sun in the way their reference is determined by an element
of the context of use. Indexicals have different referents in different contexts even
when used in the same sense and evaluated with respect to the same circumstances.
Indexicals not only have contextually variable referents, but different ways of
being used. Imagine that Thomas Jefferson utters sentence (1) while in a room with
George Washington, Alexander Hamilton, and Benjamin Franklin. John Adams is
elsewhere.
(1) Adams knows that Washington became president after he led the Continental
Army to victory.
(a) Anaphoric: ‘He’ refers to Washington because ‘Washington’ is
the antecedent of ‘he.’
(b) Demonstrative: ‘He’ refers to Hamilton because the speaker is
pointing at Hamilton.
(c) Deictic: ‘He’ refers to Franklin because the speaker is visually
focusing on Franklin.
On the most typical interpretation of (1), the name ‘Washington’ is the antecedent
of ‘he,’ and as a result the pronoun refers to Washington. This use is anaphoric.
But it is also possible that the speaker is using ‘he’ while pointing at Hamilton,
with the result that the pronoun refers to Hamilton rather than Washington. This
use is demonstrative. Finally, it is possible that the speaker is using ‘he’ without
using it anaphorically, and without pointing at anything. The speaker might simply
be visually focusing on Benjamin Franklin, with the result that the pronoun refers to
Franklin rather than Hamilton or Washington. I call this use deictic.3 The indexical
‘he’ is used in the same sense (linguistic meaning) in all three uses. But the referent
is determined in different ways.
3
The terms ‘anaphoric,’ ‘demonstrative,’ and ‘deictic’ are common in linguistics, but there is little
consensus on their usage.
The same three uses can be observed with phrases that are most typically used
demonstratively, as can be seen by replacing ‘he’ in (1) with ‘this patriot.’ The same
three uses are also possible with indexicals that are most typically used deictically,
as (2) illustrates.
(2) (a) Anaphoric: September 1, 1939: World War II began today.
(b) Demonstrative: World War II began today (said pointing at a
particular square on a calendar).
(c) Deictic: World War II began today.
‘Today’ refers to September 1, 1939 in (2a), to the day represented by the square the
speaker is pointing at in (b), and to the day of utterance in (c).4
The propositions and propositional attitudes expressed using pronouns differ
markedly from those expressed using non-indexical terms. Even though the subject
terms in (3) all have the same referent, the sentences do not have the same meaning.
The propositions they express have the same truth conditions, but are not the same.
Someone can believe one without believing the others.
(3) (a) Washington led the Continental Army to victory.
(b) The first U.S. president led the Continental Army to victory.
(c) He led the Continental Army to victory (visually focusing on
Washington).
(d) This patriot led the Continental Army to victory (pointing at a picture
of Washington).
The problem for linguistics is to account for the differences in meaning among
sentences like these. The problem for psychology is to explain how believing what
is expressed by (c) and believing what is expressed by (d) differ from each other
and from believing what is expressed by (a) or (b). This “problem of the essential
indexical” (Perry 1979), a special case of Frege’s (1892) problem for the theory that
the meaning of an expression is its reference, arises with all other indexicals and
attitudes. A busy vice-president might believe “I have a meeting at noon” without
believing “I have a meeting now,” with the result that she misses an important
meeting. A pilot may know that flying in Iranian airspace is prohibited without
realizing that flying here is prohibited, with fatal consequences.
The linguistic and psychological differences both can be explained by the
hypothesis that the subject terms of (3) express different thought parts. The thesis
that the name ‘Washington’ expresses the idea of Washington seems tautological, as
does the hypothesis that ‘the first U.S. president’ expresses the idea of the first U.S.
president. But what ideas do ‘he’ and ‘this patriot’ express? We cannot say that ‘he’
expresses the idea of him, for in the idea of him, ‘him’ has an object-position sense
and requires a particular referent. The idea expressed by ‘he’ in (1) is a subject-
concept and has different referents in different contexts. We can, however, use the
idea “he” to refer to the thought part expressed by ‘he’ in (1a), (b), and (c). I will use
4
I focus on ‘I’ elsewhere (2013, 2017a, b).
308 W.A. Davis
i(he) for the same purpose. ‘This patriot’ expresses the same thought part whether it
is the subject or object of a verb. We still cannot say that ‘this patriot’ expresses the
idea of this patriot because the thought part it expresses does not have a particular
referent. But we can refer to the thought part it expresses as the idea “this patriot.”
The name ‘Washington’ has different senses (it is the name of a president, a
city, and a state, among other things), so it expresses different thought parts or
ideas in English. But each of the ideas it expresses has a fixed referent. Whenever
‘Washington’ is used with the sense intended in (3a), it refers to the same person.
The same is true of ‘the first U.S. president.’ The ideas (thought parts) expressed by
‘he’ and ‘this patriot,’ in contrast, differ from both ‘Washington’ and ‘the first U.S.
president’ in having different referents in different contexts of use. They are what
I call indexical concepts. What determines the referent of an indexical concept?
We will focus here on primary indexical concepts, those that do not contain other
indexical concepts.
I hypothesize that (primary) indexical concepts differ from those expressed by
proper names and definite descriptions in connecting in a particular way with other
representational mental events, including sensory experiences and other concepts,
whose objects become their referents. Indexical concepts are capable of being linked
to a determinant. Consider again sentence (1), which has two different anaphoric
interpretations depending on whether ‘Washington’ or ‘Adams’ is the antecedent of
‘he.’ What makes one name the antecedent rather than another on a given occasion
of use? On my view, that depends on whether (1) is used to express a thought
in which i(he) is linked to i(Washington) or one with i(he) linked to i(Adams). A
speaker S cannot think the first thought unless i(he) and i(Washington) both occur
to S and do so in the right relationship, which I call indexical linkage.
Let ‘’ be a variable for any primary indexical concept and ‘•’ for any
determinant. (Think of the ‘x’ in the symbol as representing the concept’s
potential linkage to a determinant.) On my view, concepts are event types that can
occur at different times as parts of different thoughts. Let ‘• ’ stand for the subtype
of that consists of its occurrence linked to •. Hence:
(4) • occurs to S iff and • occur to S and their occurrences are linked.
i(he)i(GW) and i(he)i(JA) are different subtypes of i(he). Metaphysically, • is to

as driving with friends is to driving, and striking in anger to striking. For a neural
model, suppose is a circuit whose activation is the occurrence of . Let be
capable of being activated by other neural circuits, including •. Then • might
represent the activation of by • over a specific type of neural pathway.
From a semantic standpoint, the most important rule governing • is the derived
reference rule (5). Let ‘extfxg’ denote the extension of x.
(5) extf• g D extf•g, provided the referent of • is in the extension of the sortal
component of .
As explained in Sect. 5, the sortal component of i(he) is i(male). Hence the extension
of i(he)i(GW) is Washington, and the extension of i(he)i(JA) is Adams. So when ‘he’
has ‘Washington’ as its antecedent, it refers to Washington. For ‘Washington’ is the

antecedent of ‘he’ when the indexical concept expressed by ‘he’ has the concept
expressed by ‘Washington’ as its determinant. The speaker uses ‘he’ to express the
generic concept i(he), thus meaning “he” by ‘he.’ The speaker also expresses the
more specific concept i(he)i(GW) , thereby referring to Washington. Indexical terms
are characterized by a “double triangle of signification”: the reference of a term is
in general that of the idea it expresses, and the reference of an indexical idea is that
of the determinant it is linked to.
4 The Deictic Use
Returning to example (1), the pronoun ‘he’ is used anaphorically to refer to George
Washington, on my theory, when ‘he’ is used to express an occurrence of i(he)
linked to the concept i(Washington) expressed by the proper name ‘Washington’ on
that occasion. When the pronoun is used demonstratively, the determinant is related
in a particular way to the speaker’s pointing gesture. But in the deictic usage, there is
no antecedent or demonstration. What then is the determinant? I believe that what is
distinctive about deictic occurrences of indexical concepts is that they are linked to
perceptual, introspective, memory, or even hallucinatory experiences—what I call
collectively presentations. When Jefferson used ‘he’ deictically, he was referring
to Franklin because he was visually attending to Franklin, and i(he) was linked to
that perception. Had Jefferson instead been using ‘he’ to express an occurrence of
i(he) linked to his perception of Hamilton, ‘he’ would have referred deictically to
Hamilton. Had Jefferson been using ‘he’ to express an occurrence of i(he) linked
to an occurrent memory of James Madison, Jefferson would have been referring
to Madison. Note that Jefferson could also have been using ‘he’ to express an
occurrence of i(he) linked to his visual or auditory perception of Washington. In
that case, Jefferson would have referred deictically rather than anaphorically to
Washington, and conceivably, might not have realized that he is Washington.
By ‘presentation,’ I mean a non-epistemic awareness or memory of something,
or a similar non-veridical state of consciousness. Seeing a lemur differs markedly
from seeing that it is a lemur. The latter entails knowing and therefore believing that
the object is a lemur, which in turn entails having the concept of a lemur. A subject
can see a lemur without knowing that it is a lemur: the perceiver might misidentify
it as some kind of raccoon, or without being capable of knowing that anything is
a lemur. The subject might not even know that what is seen exists: a man might
mistakenly think he is hallucinating; a lower animal might not have the concept of
existence.5
5
Since seeing-that entails knowing-that, Dretske (1969) called it “epistemic” perception. Seeing a
lemur is “non-epistemic” perception.
310 W.A. Davis
Attention or focus is an essential element of the process whereby sense-

perceptions or introspections become linked with indexical concepts. Suppose we
are looking at the grid of letters below:
c f c
f u z
We see all six letters in this grid. There is a sensory presentation of the upper left ‘c,’
a sensory presentation of the upper middle ‘f,’ and so on. While remaining aware
of all the letters, we can attend to one and think This is a ‘c.’ Whether the thought
is true or false depends on which of the six letters we are attending to. The shift in
attention does not change the sensory presentation: the array looks exactly the same.
Visual attention is partly a non-sensory process.
We perceive an object by perceiving some of its parts. If Mary is looking at the
Enterprise, she may be seeing its bow or its stern. Suppose now that Mary sees
the aircraft carrier in virtue of seeing its stern (she is looking at it directly from
the rear). Even though they may coincide for a while, the process of perceiving the
stern is different from the process of perceiving the ship. Moreover, Mary may be
attending to the ship rather than the stern even though she does not see more of the
ship than the stern. The purely sensory component of the process may be the same
as if she were attending to the stern. But attending to something has an additional
non-sensory component that differs in the two cases. In virtue of the different non-
sensory components, the subject has different dispositions to respond to changes in
the stimulus. If Mary is attending to the stern, and the ship turns so that more of
the ship is in view, she is liable to keep the stern in her focal point. But if Mary
is attending to the ship, she is liable to shift her focal point away from the stern to
a more central part of the ship. Because the processes of perceiving and attending
to the ship differ from those of perceiving and attending to the stern, the indexical
concept i(this) can be linked to a presentation that is focused on the ship rather than
on the stern, even if her sense-impression of the ship is a sense-impression of the
stern. Consequently, she can truly say “This is thousands of feet long,” referring
to the whole ship rather than just the stern. Sense-impressions (the complexes of
sensations involved in seeing, hearing, feeling, tasting, or smelling) are only part of
the process of perceiving or attending to an object.
Attention can also be directed upon objects of introspective awareness. I can
work with a pain in my foot without attending to it. But I must be attending to it
when I think “It is still there.” A similar process is involved when indexical concepts
are linked to memory or hallucinatory presentations, but is not called attention.
Suppose we are remembering a wedding, and have a vivid image of the scene in
which the groom kissed the bride. We think He kissed her. We may be remembering
the priest too. But our memory presentations of the bride and the groom are mentally
highlighted, just as the sense-impressions of the ‘z’ is when we are attending to it
in the grid above. We cannot say that we are literally attending to the bride and the
groom, but we can say that we are focusing on them. Attention is the specific case
of focusing on an object of current perceptual or introspective observation.
We noted that perceiving and attending differ when it comes to parts and wholes.
We perceive a ship by perceiving one of its parts. But if we attend to a part, we are
not attending to the whole. In the case of tokens and types, we can both perceive and
attend to a type by perceiving and attending to its tokens. When we attend to the ‘c’
in the upper left of the grid, we are not also attending to the ‘c’ in the upper right. So
we are attending to a particular token. But we are also attending to a type of letter
we are not attending to when we attend to the ‘z.’ Focusing differs from attending
in this case. Even though we are attending to both the token ‘c’ and the type ‘c,’ we
may be focused on one or the other. As a result, the thought “This occurs more than
once” may be true or false.
An indexical concept is linked to a presentation only if the subject is focusing
on an object through it. I say in that case that the presentation is focused on the
object. I use the Greek letter ‘¡’ (rho) to denote presentations. If ¡(BF) is a veridical
presentation focused on Franklin, the extension of i(he)(BF) is Franklin.
5 Sortals and Determiners
We have looked at what differentiates occurrences of the generic indexical concept

i(he): they are linked to different determinants. What can we say about the different
generic indexical concepts? How does i(this) differ from i(he) and other primary
indexical concepts? Are they atomic or do they have components? Since purely
descriptive concepts are not indexical, a primary indexical concept cannot be
completely analyzed into descriptive concepts. But that allows to have some
descriptive components as long as it has others that make it non-descriptive. To
account for all the similarities and differences in meaning among indexicals, I
hypothesize that each primary indexical concept is composed of a very general
sortal concept plus a nondescriptive determiner. I present evidence that the sortal
component of i(he) is i(male). The sortal component of i(this), expressed by ‘this’
as a pronoun, is i(thing); that of a phrase of the form ‘this N,’ in which ‘this’ is
a determiner, is i(N). The sortal component generates the condition on the derived
reference rule (5): extf• g D extf•g provided extfıg is in the extension of the sortal
component of . Suppose Terry points at someone he mistakenly believes to be male,
and says ‘He is tall.’ Even though Terry is referring to a woman, ‘he’ cannot refer
to a woman because of its sortal component. His statement suffers from the same
presupposition failure as ‘This/The male is tall.’6
The determiner component of a primary indexical concept does a number of
things. It combines with the sortal component to form a singular subject-concept
that is indexical and therefore capable of linking to determinants. It furthermore
determines the specific range of determinants the concept can link to. No two
primary indexical concepts can occur with the same range of determinants. Each
6
See Davis (2017b) for more on sortal components.
312 W.A. Davis
has a unique set of determinant constraints, which specify the determinants that
cannot link to. Some pronouns, such as ‘this,’ ‘that,’ and ‘it,’ differ only in their
determinant constraints; there is no difference in their sortal components. ‘He’ and
‘himself’ differ in the same way, as illustrated by (6).
(6) (a) Hemingway killed himself.

(b) Hemingway killed him.
In (6a), the reflexive pronoun can only have ‘Hemingway’ as its antecedent.
In (b) the pronoun cannot have the same antecedent. ‘Him’ in (b) could refer
to Hemingway, but only if the speaker is using it demonstratively (pointing at
Hemingway) or deictically (focusing on him).7
6 Indexical Designators
Definite descriptions can be used with either a restricted or an unrestricted

interpretation.8 The unrestricted interpretation was the main focus of the Russell-
Strawson debate. On this interpretation, ‘the G’ means something like “the only
thing in the entire universe that is F,” so that ‘The G is H’ is true only if there is
one and only one G and it is H. ‘The queen of England’ would most naturally have
the unrestricted interpretation. A definite description like ‘the cat,’ in contrast, is
seldom if ever used with the unrestricted interpretation. More commonly, ‘the cat’
means something like “the contextually indicated cat.” If I see a cat and say “The cat
is on the mat,” then ‘the cat’ refers to the cat I am attending to. The fact that there
are millions of other cats and mats around the world will not make my statement
false or truth-valueless. What I said in no way implies or presupposes that there is
just one cat in the whole world. The presence of more than one cat on the mat might
raise questions. But even then there will be no problem if my audience knows which
one I meant.
Let ‘ther G’ represent the general form of definite descriptions used with the sort
of restricted sense ‘the cat’ has in the above example. In using this subscript, I do not
mean to suggest that the definite article itself is ambiguous; it is the whole phrase
that has the restricted interpretation. The rough gloss which says that ‘ther G’ means
“the contextually indicated G” is accurate in its implication that the restricted sense
of a definite description is indexical. ‘The cat’ refers to different cats in different
contexts even though it is used with the same sense and evaluated at the same
world. Restricted definite descriptions can be used anaphorically or deictically, as
illustrated by (7).
7
I discuss these well-known Binding Rules (Chomsky 1995: section 1.4.2) in Davis (2017b).
8
Cf. Strawson (1950: 14–5), Donnellan (1968, 1977), Evans (1977, 1980), Wettstein (1981),
Salmon (1982), Soames (1986), Bach (1987: Sect. 6.4), Recanati (1989a: 232–3, b: 314), Neale
(1990: Sect. 3.7), Bezuidenhout (1997a, b), Reimer (1998a, b), and Powell (2001: 80ff).
(7) Adams knows that after Washington led the Continental Army to victory,
the Virginian became president.
(a) Anaphoric: ‘Washington’ is the antecedent of ‘the Virginian.’
(b) Deictic: The speaker was visually focusing on Thomas Jefferson when
he said ‘the Virginian.’9
Replacing ‘the Virginian’ by ‘he’ in (7) produces little change in its meaning,
except for replacing the presupposition that the referent is a Virginian with the
presupposition that the referent is a man.
‘Ther G’ is very similar in meaning to ‘this G’ and ‘that G,’ which are definite
but not descriptions. ‘This/that Virginian’ could replace ‘the Virginian’ in (7) with
little difference in meaning. One way ‘ther G’ differs markedly from ‘this/that G’ is
that it cannot be used demonstratively. Compare:
(8) (a) This Virginian became president.

(b) The Virginian became president.
If there are a number of Virginians in the room, I can randomly point at one with my
eyes closed and utter (8a). ‘This Virginian’ will then refer to the Virginian I pointed
to. ‘The Virginian’ cannot be used in the same way. I can point when I use it, but
the pointing never determines the referent. The pointing must be interpreted as an
emphatic gesture, not as a determinant of the referent. If I accidentally point at the
wrong Virginian, what I said will not diverge from what I meant.
On my view, ‘ther G,’ ‘this G,’ and ‘that G’ all express primary indexical
concepts. So they are indexical designators. Their referent on any occasion of use
is determined by the determinant linked to the indexical concept it expresses by
rule (5). Thus on the anaphoric interpretation of (7), the determinant is the concept
expressed by ‘Washington.’ Hence ‘the Virginian’ refers to Washington. On the
deictic interpretation of (7), the determinant is the speaker’s visual presentation of
Jefferson. Hence ‘the Virginian’ refers to Jefferson. ‘Ther G’ does not presuppose
there is just one G because its referent is determined not by ‘G’ but by an indexical
determinant.
Since the indexical concepts expressed by ‘ther G,’ ‘this G’ and ‘that G’ all
have the same sortal component—the concept expressed by ‘G’—they must be
differentiated by their determiners. The determiner for i(ther G) differs from that
for i(this G) and i(that G) in excluding demonstrative determinants. ‘That G’ and
‘this G’ differ from each other in their determinant constraints. For example, ‘this
speaker’ can be used demonstratively when the speaker is focusing perceptually
or introspectively on himself; ‘that speaker’ cannot be. Let , K, and ƒ represent
the three different determinant types, and let ♂ represent the generic indexical
determiner (think of the arrow as representing the linkage to the determinant). Then
9
Background facts: Washington and Jefferson were both Virginians; Adams was from Mas-
sachusetts.
314 W.A. Davis
the indexical concepts expressed by ‘ther G,’ ‘this G,’ and ‘that G’ can be analyzed
as follows:
(9) Indexical Concept Definitions:
i(ther G) D ♂ i(G)
i(this G) D ♂K i(G)
i(that G) D ♂ i(G)
♂ , ♂K , and ♂ƒ are the specific determiners that differentiate these three indexical
concepts. These definitions can equally well be reformulated as semantic rules,
saying that expressions of the form ‘ther G,’ ‘this G,’ and ‘that G’ express the
indexical concepts ♂ i(G), ♂K i(G), and ♂ƒ i(G) in English.
7 Two Uses of Quotation Marks
Quotation marks are used for markedly different purposes in sentences (10)–(13):
(10) The word ‘car’ expresses the idea “automobile.”

(11) Sue used ‘car’ to express the idea “railcar.”
(12) Sue uttered ‘A car is a vehicle.’
(13) Sue is thinking “A car is a vehicle.”
In (10) and (11), the single quotes are used to mark reference to a word, which is
a linguistic entity. The words ‘automobile’ and ‘railcar’ are also used in (10) and
(11), but they are not mentioned. The double quotes are used to mark reference
to ideas, which are mental entities. The word ‘car’ is an expression consisting
of three letters or speech sounds. It expresses several ideas in English, including
the ideas “automobile” and “railcar.” The ideas “automobile” and “railcar” are not
expressions. They are expressed by words, but are not themselves words. Words can
be uttered, produced by speaking or writing. Ideas cannot be uttered. They occur to
us when we are thinking. The word ‘car’ is a part of many sentences that people can
utter. The idea “automobile” is a part of many thoughts that people can think. When
quotation marks are used as the single quotes are in (10)–(13) to mark reference
to the word quoted, I say they are used auto-reflexively. When they are used as the
double quotes are to mark reference to the ideas expressed by the words quoted, I
say they are used ideo-reflexively. The quotation marks themselves, pace Davidson
(Sect. 12), have no referents. They punctuate expressions that do have referents.
I have used the difference between single and double quotes to mark syntactically
the difference between auto- and ideo-reflexive usage. It is more conventional,
however, to use either single or double quotes for both purposes. Some editorial
house styles insist on it. Thus (14) has the same meaning as (10).
(14) The word ‘car’ expresses the idea ‘automobile.’

It is also conventional to use italics (or emphasis) for the same two purposes, as in
(15):
(15) The word car expresses the idea automobile.

And especially in speech and non-formal writing, words are often used auto- and
ideo-reflexively without being marked by either quotation or emphasis, as in (16):
(16) The word car expresses the idea automobile.

We will refer to the use of ‘car’ to denote cars rather than words or ideas as a first-
order use. In first-order usage, ‘car’ expresses the idea “car,” but does not refer to
that idea. Without a first-order use, an expression is not a word, phrase, or sentence
in a language, and it has no ideo-reflexive use. Expressions without a first-order use
can be used auto-reflexively, but not to refer to words, phrases or sentences of the
language.
Auto- and ideo-reflexive uses have many different linguistic properties. An
expression used ideo-reflexively, for one thing, has to be grammatical. An expres-
sion used auto-reflexively need not be. Thus (17a) is grammatical and might be true.
(17b) is ungrammatical.
(17) (a) Tom uttered ‘All cars is vehicles.’

(b) Tom thought “All cars is vehicles.”10
Similarly, an expression used ideo-reflexively must be meaningful as well as
grammatical; an expression used auto-reflexively need not be. (18a) is meaningful
and may be true. (18b) is meaningless because the sentence quoted is meaningless.
(18) (a) Tom uttered ‘All mimsy were the borogoves.’

(b) Tom thought “All mimsy were the borogoves.”
Moreover, speakers can use words auto-reflexively without knowing what they
mean, but they cannot properly use words ideo-reflexively without knowing what
they mean.
(19) (a) Tom uttered ‘Polybasite is not sphalerite.’

(b) Tom thought “Polybasite is not sphalerite.”
A speaker who has no idea what ‘polybasite’ or ‘sphalerite’ mean, or even whether
they are meaningful, can properly use (19a) and may know that what it expresses
is true. The same speaker cannot properly use (19b), even though it is meaningful
in English and may be true. A speaker who uses words ideo-reflexively must know
what they mean, and mean something by them. The speaker need not mean anything
by the sentence quoted in (19a).
Ambiguity affects auto- and ideo-reflexive use differently.
10
Note that thought is not the same as inner speech. Tom said to himself “All cars is vehicles” is
grammatical and may be true.
316 W.A. Davis
(20) (a) Tom uttered the word ‘car.’

(b) Tom expressed the idea “car.”
(i) Tom expressed the idea “automobile.”
(ii) Tom expressed the idea “railcar.”
(20b) is ambiguous in a way (20a) is not: (b) could mean either (i) or (ii). Whether
(20b) means (i) or (ii) on any given occasion is determined by whether the speaker
means “automobile” or “railcar” by ‘car.’ Uttering the word ‘car’ and meaning
“automobile,” and uttering the word ‘car’ and meaning “railcar,” are two ways
of uttering the word ‘car.’ In contrast, (20a) is general in meaning rather than
ambiguous: it is true no matter what, if anything, Tom meant by ‘car’ when he
uttered it.
When translating from one language to another, words used ideo-reflexively may
be translated. Words used auto-reflexively cannot be translated without producing a
sentence having a very different meaning (unless the translation of the word happens
to be the same word). Compare (12) and (13) with the French sentences (21) and
(22):
(21) Sue a prononcé ‘Une voiture est un véhicule.’

(22) Sue pensait “Une voiture est un véhicule.”
While (22) has the same meaning as (13), (21) and (12) have very different truth
conditions. For (21) to be true, Sue has to utter a French sentence; for (12) to be
true, Sue has to utter an English sentence. (22) and (13) say that Sue thought the
same thought.11
English auto- and ideo-reflexive designators also have different equivalents
in English. For example, (13) is equivalent to (23a) when (23) has its opaque
interpretation. (23b) makes no sense; sentences cannot be thought.
(23) (a) Sue is thinking that a car is a vehicle.

(b) Sue is thinking the sentence “A car is a vehicle.”
The opposite is true with (24):
(24) (a) Sue uttered that a car is a vehicle.

(b) Sue uttered the sentence ‘A car is a vehicle.’
(24b) is equivalent to (12), but (24a) makes no sense, in this case because it is
ungrammatical. (24a) would be grammatical if said replaced uttered, but Sue said
that a car is a vehicle can be true when (12) is false, and vice versa. Similarly, (25b)
has the same meaning as (a).
11
‘Say’ differs markedly from ‘utter’ in having an illocutionary as well as locutionary sense. In
the former, Sue said “A car is a vehicle” is equivalent to Sue said that a car is a vehicle, and
does not entail that Sue uttered an English sentence. If my grasp of French is correct, ‘prononcé’
corresponds to ‘uttered,’ ‘dit’ and ‘declaré’ to ‘said.’
(25) (a) The idea “automobile.”

(b) The idea of an automobile.12
(26a) is not equivalent to anything like (b), which is not well formed.
(26) (a) The word ‘car.’

(b) The word of a car.
(c) The word for a car.
(26c) is grammatical, and a speaker might use it to refer to the word ‘car.’ But (c)
differs from (a) in presupposing that there is just one word for a car in English, a
presupposition that is false.
The assumption that the ideo-reflexive use of words is at least implicitly auto-
reflexive is natural and quite persistent. For (27) can be used with the same referent
as (25):
(27) The idea expressed by the word ‘automobile.’

However, (25) and (27) do not have the same meaning. First, as noted above, (27)
mentions the word ‘automobile’; (25) uses the word but does not mention it. Second,
while (25) and (27) may have the same referent, they need not. English could
change so that the word ‘automobile’ no longer expresses the idea “automobile.”
Consequently, (28) is at best a contingent a posteriori truth.
(28) The idea “automobile” is the idea expressed by the word ‘automobile.’
(28) is a posteriori, because to know that it is true, we have to know something about
a particular English letter- or sound-sequence. (28) is contingent because it could be
or become false. The difference between (25a) and (27) is in one respect like the
difference between Barack Obama and the person whose name is ‘Barack Obama.’
(25a) is a rigid designator, as is (25b): as used on any given occasion, they have
the same referent in every world. (27) is non-rigid because ‘automobile’ expresses
different ideas in different worlds. Consequently, a counterfactual like (29) is true.
(29) If ‘automobile’ came to mean “bicycle,” it would no longer express the idea
“automobile.”
Note, though, that if ‘automobile’ does mean “bicycle” in some possible world w,
then when used in w, (25a) would rigidly designate the idea of a bicycle.
Second, (27) is a definite description, and as such presupposes that ‘automobile’
expresses just one idea. (25) is not a definite description and does not have such a
presupposition. These differences are more apparent with (30a) and (b).
12
As noted in Sect. 3, the idea “®” and the idea of ® may not be synonymous when ‘®’ is an
indexical like him. Note too that (b) can be used transparently, as in The idea of a sedan is the idea
of an automobile. When it is transparent, it is not synonymous with (a), which has no transparent
interpretation. For further complexities, see Davis (2003: section 7.6).
318 W.A. Davis
(30) (a) The idea “car”

(i) The idea “automobile”
(ii) The idea “railcar”
(b) The idea expressed by ‘car’
The fact that ‘car’ has more than one meaning has different effects on (30a) and (b).
(30a) is ambiguous, with one meaning for each meaning of ‘car.’ But in none of its
meanings does it presuppose that ‘car’ has just one meaning. (30b) is unambiguous,
but presupposes that ‘car’ has exactly one meaning, at least in the context of
utterance. These considerations make it evident that the meaning of (30a) depends
compositionally on the meaning of ‘car,’ whereas the meaning of (30b) is unrelated
to the meaning of ‘car.’
Something more complex than (28) will be true on any occasion in which its
ideo-reflexive subject is used, namely (31):
(31) The idea “automobile” is the idea expressed by the word ‘automobile’ on
this occasion.
Because it contains the indexical phrase this occasion, (31) does not express a
particular proposition except in a given context of use, in which this occasion has a
particular referent. When it refers to an occasion in which the idea “automobile” is
used, the referent of the idea expressed by the word ‘automobile’ on this occasion
can only be the idea expressed by the word ‘automobile’ used ideo-reflexively
on that occasion. Note that the two singular terms flanking is in (31) are not
synonymous either. The one on the left is ambiguous13 but not indexical. The one
on the right is indexical but not ambiguous.
While (31) always expresses a contingent truth, (31) will be true in every context
of use if ‘this occasion’ refers to that context and ‘automobile’ happens to be
unambiguous. But it is not difficult to imagine a context in which (32) is false.
(32) The idea “car” is the idea expressed by the word ‘car’ on this occasion.
Let (32) be used in context c, in which this occasion refers to some other context
c0 . If the speaker is using the idea “car” to refer to the idea of an automobile, but
is mistaken in his assumption that ‘car’ expressed the same idea in c0 , then (32) will
be false in c.
Quotation marks and italics have many other uses. Quotation marks are used
in irony (President Bush’s immigration policy is another “mission accomplished”),
word introductions (This is “dimorphism”), titles (I saw “Lawrence of Arabia”),
representations of form (sentences of the form ‘NP VP,’ propositions of the form
‘p or not-p’), and various forms of speaker quotation (Quine said “To be is to
13
‘Automobile’ means “of or pertaining to cars” (as an adjective) as well as “car” (as a noun).
be the value of a variable”).14 Italics (or intonational stress) is used to signal a

different range of features, including highlighting, contrast, emotional intensity,
foreign words, and technical symbols.
8 Ideo-reflexive Designators
We have seen that ideo-reflexive quotation is one of several productive devices in

English and other natural languages that enable us to refer to an idea by using a
word, phrase, or sentence that expresses that idea (Sect. 7). The device thus exploits
the word-idea pairings established by the first-order conventions of the language for
a second purpose. The word ‘car’ has the meanings it has in English because it is
conventionally used to express certain ideas, including the idea of an automobile
and the idea of a railcar. We can thus talk about automobiles (or railcars) by using
the word ‘car.’ The ideo-reflexive conventions for referring to ideas enables us to
talk about the idea of an automobile (or railcar) by using any of the phrases in (33).
(33) the idea “car,” the idea car, the idea car, the idea of a car
I call such phrases ideo-reflexive designators.
While there are many similarities, ideo-reflexive designators differ from index-
ical designators of the form ‘ther G’ discussed in Sect. 6 in two ways. First,
ideo-reflexive designators are not themselves indexical. When used with the same
sense, they refer to the same idea in every context of use. The idea “car” does have
different referents in different contexts, but only because it is used with different
meanings in different contexts. When ‘car’ means “automobile,” the idea “car”
refers to the idea of an automobile wherever it is used. When ‘car’ means “railcar,”
the idea “car” always refers to the idea of a railcar. Second, the idea “car” does
not have the form ‘the G,’ where ‘G’ is a general term. Hence it does not mean
anything like the only idea “car” or the contextually indicated idea “car.” These
glosses do not even make sense because idea “car” is not a noun or noun phrase.
It cannot be preceded by an indefinite article (‘an’), quantifiers (‘all,’ ‘some,’ ‘no,’
‘most,’ etc.), or numerical determiners (‘one,’ ‘two,’ etc.), and cannot be modified
by an adjective.15
While the idea “car” is not an indexical designator, the idea can be when it
occurs independently as a restricted definite description.
(34) John uses ‘car’ to express the idea “automobile.” Pierre uses ‘voiture’ to
express the idea.
14
I develop a theory of saying reports based on my account of auto- and ideo-reflexive designators
in Davis (2016a).
15
‘Idea of a car’ is a general term, however, when ‘idea’ means “conception” rather than “thought
part.” There are many conceptions of a car.
320 W.A. Davis
The second occurrence of ‘the idea’ in (34) has the idea “automobile” as its
anaphoric antecedent. ‘This idea’ or ‘that idea’ could be used instead of the second
occurrence of ‘the idea’ with little change in meaning; but using them in place of
the first occurrence of ‘the idea’ would be very unnatural. Or suppose a scientist
struggling to answer a theoretical question finally arrives at the answer and says the
following:
(35) This is amazing! The idea is going to revolutionize science.

The scientist used this deictically to refer to an idea she had, and then used the idea
anaphorically with this as its antecedent. As with other expressions of the form ‘ther
G,’ ‘the idea’ cannot be used demonstratively. By (9) above, ‘ther idea’ expresses
the indexical concept ♂ i(idea).
My hypothesis is that the use of the idea “car” as a unit expresses a deictic
subtype of the indexical concept expressed by ‘the idea’ when it is used indepen-
dently as a restricted definite description. The function of ‘car’ is to express an idea
for which our introspective awareness is the indexical determinant of i(the idea)
D ♂ i(idea). It might be either the idea of an automobile or the idea of a railcar.
If the former, then the idea “car” expresses the specific indexical concept i(the
idea)[i(automobile)] , where ¡[i(automobile)] is the subject’s introspective presentation
of the idea i(automobile). It follows from the derived reference rule (5) that when the
idea “car” expresses i(the idea)[i(automobile)] , it refers to i(automobile). Because the
determinant is a presentation, the designator is rigid. If communication is successful,
and the hearer understands what the speaker meant by the idea “car,” then the idea
of a car will occur to the hearer too, and her introspective awareness of that idea will
be linked to her occurrence of ♂ i(idea).
In general, let ‘š ’ be a place-holder for any term expressing idea š. On any
occasion in which the idea “ ” is used, the idea š expressed by the displayed word
‘š ’ on that occasion is occurring to the speaker, who is introspectively aware of it.
His introspective awareness ¡(š) of that idea is linked to ♂ i(idea). Consequently,
the speaker uses the idea “ ” to express ♂ i(idea)( ) . It follows from the extension
rule (5) that the idea “ ” refers to š. Thus whereas ‘ther idea’ expresses the
generic indexical concept ♂ i(idea), which may be linked to either conceptual or
presentational determinants, the idea “ ” expresses the specific indexical concept
♂ i(idea)( ) , whose determinant is the subject’s introspective presentation of a
particular idea, the one expressed by ‘’ on that occasion. Because expressions of
the form the idea “ ” express specific indexical concepts, they are not themselves
indexical. We can thus add (36) to (9).

i(the idea “ ”) D ♂ i(idea)( )
This says that the concept expressed by a phrase of the form the idea ‘ ’ is the
specific indexical concept whose sortal component is i(idea), whose determiner is
♂ , and whose determinant is a presentation of the idea expressed by the word
quoted in the phrase on the occasion of its use. Whereas primary indexical concept
definitions like (9) define generic indexical concepts, specifying determinant types
but no particular determinants, (36) specifies a particular determinant as well as a
determinant type. Hence (36) defines a specific indexical concept as a determinate
of the generic indexical concept ♂ i(idea).
These definition schemas can be generalized by replacing ‘idea’ with a variable
‘C’ holding a place for more specific terms like subject concept, predicate concept,
atomic concept, complex concept, syncategorematic concept, thought, and propo-
sition. Thus i(the thought “A car is a vehicle ”) D ♂ i(thought)( ) , where ¡(š)
is a subject’s introspective presentation of the thought expressed by ‘A car is a
vehicle’ in the current context (i.e., in the thought “A car is a vehicle” as I used
it). Since the proposition that a car is a vehicle is an object of belief, desire, and
other propositional attitudes, we have further equations such as i(the belief “A car
is a vehicle”) D ♂ i(belief)( ) . Further ideo-reflexive constructions can be defined
in terms of ideo-reflexive designators. Thus the verb phrase thinking “A car is a
vehicle” means thinking the thought “A car is a vehicle.”
The thesis that the ideo-reflexive use of a sentence involves using it to refer to the
thought it expresses does not entail the Fregean thesis that the words so used “name
what would customarily be their sense” (Frege 1892: 59). We observed in Sect. 7
that (13) and (23a) are equivalent when the latter is opaque. (37a) and (b) illustrate
the same point.
(37) (a) Sue is thinking “Birds fly.”

(b) Sue is thinking the thought that birds fly.
(c) Sue is thinking the idea of birds the idea of flying.
(37a) and (b) say that Sue is thinking a certain thought, the one expressed by the
sentence ‘Birds fly.’ Frege maintained that the thought is the sense of the sentence.
I do not. (37a) and (b) do not say that Sue is thinking the senses of linguistic
expressions. It is particularly clear that the words ‘birds’ and ‘fly’ in (37a) and (b)
do not name or refer to the ideas they express. If they did, (37a) and (b) should be
equivalent to (37c), which makes no sense at all. ‘Birds’ and ‘fly’ in (37a) and (b) are
used to express the ideas they customarily express, as does the sentence ‘Birds fly.’
As a result, (37a) and (b) refer to the thought expressed by ‘Birds fly.’ Ideo-reflexive
reference enables us to do this without using the names of any ideas.
9 Auto-reflexive Designators
Auto-reflexive reference is a productive device that enables us to refer to a word by

using the word to display what we are referring to, as in (38).
(38) (a) the word ‘car’

(b) the word car
(c) the word car
322 W.A. Davis
All three of these designators refer to the word that is used auto-reflexively. Note that
in the standard use we are focusing on, the three phrases in (38) are not themselves
used auto-reflexively. And whereas it is tautological that the phrases in (38) refer to
the word ‘car,’ the quoted word therein does not refer to the word ‘car’ (more on
this in Sect. 10). So ‘auto-reflexive’ does not mean self-referential. What the third
word in (38) does is display (exhibit, present) the referent of the phrase. To display
something is to show it in the sense of making it perceptible.16 The pentagon in (39)
is functioning similarly.
(39) The shape

In ideo-reflexive reference, in contrast, the quoted word does not display the
referent, but expresses and calls up the referent, which is introspectible. The quoted
word is displayed, but not merely displayed: it must be used to express a particular
idea.
Like ideo-reflexive designators, auto-reflexive designators are not indexical
designators themselves, but they contain a phrase that can be used as an indexical
designator, such as the word in (38). It is my hypothesis that car is functioning as a
“deictic antecedent” in (38). That is, the determinant of i(ther word) is a presentation
of the word displayed, namely car. This accounts for the fact that auto-reflexive
designators refer rigidly. Rule (36) thus has another companion in (40). ¡(‘w’) is a
presentation of the word ‘w.’

i(i(the word ‘w’) D ♂ i(word)(‘w’) .
This says that the concept expressed by a phrase of the form the word ‘w’ is the
specific indexical concept whose sortal component is i(word), whose determiner
is ♂ , and whose determinant is a presentation of the word quoted in the phrase.
Parallel rules also result from replacing ‘word’ with cognate terms like ‘phrase’
or ‘sentence,’ more general terms like ‘expression,’17 the plural ‘words,’ or more
specific terms like ‘noun’ and ‘verb.’
It is well known that word and related expressions have a type-token ambiguity.
Thus there is a sense in which line (41) on the piece of paper you are reading
contains eight words (the token sense) and a sense in which it contains just the
six words ‘the,’ ‘dog,’ ‘chased,’ ‘cat,’ and ‘room’ (the type sense).18
16
Cf. Clark and Gerrig (1990), Wertheimer (1999: 515), Recanati (2001: 639), and Green (2007).
17
The parallel for ‘expression’ implies that the expression ‘car’ is ambiguous in English, most
commonly referring to the word for automobiles, but also possibly referring to the expression
consisting of that word surrounded by single quotation marks. Whereas ‘car’ (a word in quotation
marks) is not a word, it is an expression.
18
Each token of the sentence type contains eight word tokens but no word types. The sentence
type contains eight occurrences of word types, but no word tokens. Occurrences are subtypes, not
tokens. Occurrences of types in types are often mistaken for tokens. See Wetzel (2009: chapter 7)
and Davis (2014).
(41) The dog chased the cat around the room.

The sortal concept in the indexical concepts expressed by auto-reflexive designators
can be either type or token expression concepts. Ideo-reflexive designators have the
same ambiguity.
Mental states are objects of introspection but not sense-perception. Since the
only presentations we have of thoughts and ideas are introspective, the determinants
of the concepts expressed by ideo-reflexive designators are always introspective
presentations. Whenever auto-reflexive designators are used in speech or writing,
the determinants of the concepts they expressed are perceptual presentations. We
hear the words in speech and see them in writing. Auto-reflexive designators do
have introspective determinants when they are used in inner speech, however. When
I say (12) or (13) to myself, I am focusing introspectively on an auditory image of
the sentence ‘A car is a vehicle.’
The rules for auto-reflexive reference differ significantly from those for ideo-
reflexive reference in not requiring that the word ‘w’ express an idea in English,
or even be an English expression.19 The word ‘belle’ is French is a sentence of
English that refers to a French word. The word it refers to has a meaning in French,
but is not being used with its French meaning in that sentence. Its grammatical
category is irrelevant. The speaker is consequently not using French (Davis 2003:
Sect. 7.8). There is not even a restriction that ‘w’ be a word. The word ‘f’ is short
is a grammatical English sentence. Its defect is semantic: it falsely presupposes that
‘f’ is a word, which is a consequence of the meaning assigned to the word ‘f’ by
(40).
10 Elliptical Forms
Speakers often use a word ‘w’ elliptically to mean “the word ‘w’,” as in (42).
(42) (a) Car has three letters.

(b) Chicago is a name.
A reflexive interpretation of (42a) is forced, because a first-order interpretation
would be ungrammatical. To get the auto-reflexive interpretation of (42b), readers
might have to reason “It would not make sense to say that a city is a name, so
the speaker must be referring to the word ‘Chicago.’” This would be especially
true if the conversation had previously been about the city, with no signal of a
metalinguistic shift. But in most contexts, readers will automatically figure out
what is meant. The specific form of ellipsis involved in (42) is sometimes called
“autonymy.”
19
Cf. Cappelen and Lepore (2012: 7, 29). See also Washington (1992: 592–4).
324 W.A. Davis
The fact that car can be used to mean “the word ‘car’” does not mean that it
has a third meaning in English in addition to “automobile” and “railcar.” The word
‘car’ has no meaning in English on which it is a singular term whose referent is a
word. Like figures of speech, ellipsis is one of the ways in which speaker meaning
and reference diverge from word meaning and reference. The use of car to mean
“the word ‘car’” is not itself a convention: people do not use car with that meaning
because people have specifically used car with that meaning before. For all I know,
it may never have been so used. Moreover, we can use expressions autonymically
that have never been used before or are not words of English, as in (43). In this
respect, autonymy differs from figures of speech.
(43) Ogacihc has seven letters.

In saying “Ogacihc has seven letters,” I do not mean anything by the expression
‘ogacihc.’ Nor does ‘ogacihc’ mean anything in that sentence or anywhere else. It
has no referent. The meaningfulness of (43) does not give ‘ogacihc’ a meaning.
Compare also a standard case of noun ellipsis:
(44) John’s brother is anti war, and Bill’s is anti globalization.

In (44), the speaker uses Bill’s to express the idea “Bill’s brother.” But the speaker
does not mean “Bill’s brother” by the possessive Bill’s. The speaker expresses
the idea “Bill’s brother” indirectly, by using Bill’s to directly express the idea
“Bill’s” after expressing the concept “brother” and before expressing the idea “is
an anti-globalization activist.” (44) differs from (42) and (43) in that the term used
elliptically is used with its conventional meaning, “Bill’s.” The speaker expresses
the more complex idea “Bill’s brother” indirectly, by expressing the idea “Bill’s.”
In (42a), the speaker does not use car to express the idea of a car or an automobile,
and in (43) the subject term does not express any idea in English.
While there is no convention to use car to directly express the idea “the word
‘car,’” the use of (42a) does follow a more general convention. Someone who utters
(42a) would express a thought about the word ‘car’ because speakers conventionally
use words elliptically to refer to themselves.20 The use is a convention in the
“customary practice” sense: it is a common action that perpetuates itself through
precedent, habit and association, traditional transmission from one generation to
the next, social acceptance and criticism, and its success in achieving the common
goal of communication. The elliptical auto-reflexive use of words (autonymy) is
not the sort of convention that gives individual words meanings. It is instead
like conventions that give phrases and clauses partially compositional idiomatic
meanings. Even though ‘shoes’ does not mean “situation” and ‘put’ does not
mean “imagine” in English, ‘put oneself in his shoes’ does have the idiomatic
meaning “imagine oneself in his situation” in English. This meaning is partially
compositional (replacing ‘his’ with ‘hers’ or ‘Obama’s’ changes the meaning
20
Cf. Christensen (1967: 362), Wertheimer (1999: 515–6, 518), Saka (2011: 312, 314), and
Cappelen and Lepore (2012: 7). Contrast Saka (2011: 309).
predictably) but not completely compositional (replacing ‘shoes’ with ‘footwear’

or ‘loafers’ eliminates the idiomatic meaning). The ambiguity of the expression is
neither lexical nor syntactic. I do not have space to make the case here, but I believe
the elliptical auto-reflexive interpretation of sentences like (42a) is one of a large and
varied set of “free-form” idioms—idioms that are almost completely compositional
(Davis 2016b).
Logicians, philosophers, and linguists have a well-established convention of
using quoted expressions to refer to the expressions quoted, as in (45).
(45) (a) ‘Car’ has three letters.

(b) ‘Chicago’ is a name.
(c) ‘Ogacihc’ has seven letters.
This usage too is elliptical. (45a) means “The expression ‘car’ has three letters.” The
sentences in (45) also have an autonymous interpretation, on which (45a) says that
the expression consisting of the word ‘car’ plus the quotation marks—the expression
0
‘car’0 —has three letters.
The elliptical auto-reflexive use of italics is especially common in linguistics.
The sentences in (46) would be used with the same meaning as those in (45).
(46) (a) Car has three letters.

(b) Chicago is a name.
(c) Ogacihc has seven letters.
Neither device works best in all contexts. The quotation convention is especially
cumbersome and sometimes unclear when referring to quoted words, particularly if
only one style of quotation marks is permitted. The italics convention can be unclear
when italics is also used for emphasis or ideo-reflexive reference. Unlike quotation,
moreover, italics cannot be used iteratively (cf. (51) below).
Given that quotation marks can be used elliptically as in (45), a sentence like (47)
has a subtle amphiboly.
(47) The word ‘Chicago’ is a noun.

(a) The word, ‘Chicago,’ is a noun.
(b) ‘Chicago,’ the word, is a noun.
(c) ‘Chicago’ (the word) is a noun.
On the simplest and most natural interpretation, the one we have been analyzing, the
grammatical subject of (47) is the phrase The word ‘Chicago,’ which expresses the
specific indexical concept i(the word ‘Chicago’) D ♂ i(word)¡(‘Chicago’) , as given
by (40). But as in (45b), ‘Chicago’ can be elliptical for the word ‘Chicago,’ and
the word in (47b) can be used as a restricted definite description with ‘Chicago’ as
its anaphoric antecedent, in which case the word expresses the indexical concept
♂ i(word)i(the word ‘Chicago’) . On this anaphoric interpretation, (47) has the appositive
syntactic structure and meaning of (47a), and is interchangeable with (b) and (c).
When the grammatical subject of (47) is the auto-reflexive designator, in contrast, it
326 W.A. Davis
cannot be interrupted by a comma and is not interchangeable with (b) or (c).21 The
anaphoric interpretation of (47) is like the only possible interpretation of (48):
(48) The movie Superman is popular.

(a) The movie, Superman, is popular.
(b) Superman, the movie, is popular.
(c) Superman (the movie) is popular.
In (48), ‘The movie Superman’ is not a single noun phrase. The restricted descrip-
tion ‘the movie’ is anaphoric on the name Superman. Hence (48) is more clearly
written with the punctuation of (48a), which is equivalent to (b) and (c).
We have focused so far on the auto-reflexive use of (42a) and (b). These sentences
could also be used ideo-reflexively, but that would not be very natural and the
results would be absurd. Ideas cannot have letters or be names. But with predicates
applicable to ideas, ideo-reflexive interpretations are very natural, as in (49).
(49) (a) Car is a general artifact concept.

(b) Chicago is a singular name concept.
(c) A car is a vehicle is a true proposition.
When (49a) is true, ‘car’ is used elliptically to mean the idea “car.” Quotation marks
or italics can be used elliptically in the same way, as in (50).
(50) (a) Car is a general artifact concept.

(b) ‘Chicago’ is a singular name concept.
(c) “A car is a vehicle” is a true proposition.22
Cognitive scientists, exemplified by Fodor (1998), increasingly use small caps for
the same purpose, as in ‘CAR is a general artifact concept.’
Wertheimer (1999: 519) claims that “the quotational referent can be any manner
of thing indentifiable with the enquoted, anything from a particular phoneme to a
trans-linguistic truth-evaluable proposition.” He was justly rebelling against the idea
that quotation could be used to refer only to the expression quoted, rather than the
concept or proposition expressed by the enquoted. It is not clear what Wertheimer
could have meant by ‘identifiable,’ but it is very clear that the range of entities
associated with the enquoted expression that can be referred to by quoting it is very
limited. We cannot use the word car in quotes to refer to the more general idea of
a vehicle, or the closely associated ideas of driving or wheels. Nor can we use it to
refer to its first letter or phoneme, its transliteration in Greek, the act of uttering it,
the synonymous word auto, or the phonetically and orthographically similar word
cur.
21
Wertheimer (1999: 518–9) observes the appositive interpretation of (47), but erroneously thinks
that it is the only interpretation.
22
Cf. Wertheimer (1999: 515).
11 A Demonstrative Theory
I have argued that an auto-reflexive designator is not indexical, but expresses a

specific indexical concept whose determinant is a presentation focused on the word
displayed in the designator. The word ‘car’ is thus equivalent semantically to a
deictic use of the indexical description the word in which the determinant is a
presentation of the word car. An alternative is suggested by Christensen’s (1967:
361) proposal that the quotation marks in the word ‘car’ are playing the role of
pointers.23 The phrase would be equivalent to a use of this word whose referent
is determined by pointing at an occurrence of car. As Reimer (1996: 135) has
observed, italicizing a word rather than enclosing it in quotation marks would
be another way of pointing at it. So on this theory, auto-reflexive designators
are demonstratively used indexicals. The theory can be extended to ideo-reflexive
designators by claiming, plausibly, that we can refer to an idea demonstratively
by pointing at a word that expresses it, much as we can refer demonstratively to
Washington by pointing at a painting of Washington. We can refer to the idea of an
automobile by pointing at the word ‘automobile’ and saying “This idea occurred to
Henry Ford many times.”
One problem with this theory is that unlike this word, the word cannot be used
demonstratively (Sect. 6). Note too that this word ‘car’ is syntactically like ‘this
movie Superman’ and unlike the word ‘car’ in not forming a single noun phrase
(recall [47] and [48]).
A more serious problem for this demonstrative theory is that if it were true, the
quotation marks would be critical, for they would determine the referent of the
phrase. But as we have seen, they are not. Example (38), among others, shows
this clearly. Quotation marks are more like the gestures that accompany deictic
reference for emphasis or clarity rather than those that determine the reference
of a demonstrative indexical. Since there is no semantic difference among the
three phrases in (38), the quotation marks in (38a) cannot be functioning as
determinants of its referent the way the speaker’s pointing at Hamilton in example
(1b) determines the referent of ‘he.’24 The phrases in (38) are “demonstrative” in
the sense of showing their referent, but not in the sense in which indexicals can
be used demonstratively as opposed to deictically or anaphorically. Ideo-reflexive
designators are not even demonstrative in the showing sense, given that they do not
display their referent.
In response to the objection that auto-reflexive reference is possible without
quotation marks or italics, Reimer responds by observing that “ostensive devices
are not the only features of a context that can assist in securing the reference of a
demonstrative expression” (1996: 139). But this is true only when ‘demonstrative’
23
See also Prior (1971: 60ff), Partee (1973), Davidson (1979: 90), Goldstein (1984: 4), Garcia-
Carpintero (1994, 2017), Reimer (1996: 135), and Cappelen and Lepore (1997, 1999, 2012).
24
See Clark and Gerrig (1990: 765, 801), Washington (1992: 588), Reimer (1996: 133–5), Saka
(1998: 118–19, 2006: 463–67), and Recanati (2001: 653–6).
328 W.A. Davis
is used to denote indexicals generally (as in Kaplan 1977: 489), not indexicals used
demonstratively rather than deictically or anaphorically.25 Cappelen and Lepore
(2012: 25) suggest that the missing quotation marks in cases like (38b) and (c) might
be in the logical form of the sentence rather than its surface structure. I think this
implies incorrectly that quotes are more than punctuation marks. But the important
point here is that there is no sense in which a pointing device can be either implicit
in or a component of logical form. Things like pointing gestures and the affixing
of arrows are not concepts; they are external objects of sense perception when
indexicals are used demonstratively. If they are omitted, the use may be indexical
but not demonstrative.
Unlike deictic reference, demonstrative reference is possible when the speaker is
not observing the referent. In that case, what the speaker is referring to can diverge
from what the demonstrative refers to. As a result, a pronoun used demonstratively
may refer to the wrong object if the speaker does not point in the right direction. In
example (1b), ‘he’ refers to Hamilton even if the speaker thought he was pointing at
Washington. Kaplan (1978: 30) imagines a case in which he pointed without looking
at a place that had long been occupied by a picture of Carnap, and said “That is a
picture of one of the greatest philosophers of the twentieth century.” Unbeknownst
to Kaplan, however, Carnap’s picture had been replaced by one of Spiro Agnew. As
a result, Kaplan inadvertently said that a picture of Agnew is a picture of one of the
greatest philosophers of the twentieth century.
To find a similar case involving auto-reflexive designators, we need to somehow
imagine a writer using quotation marks without being aware of the word she has
quoted. Given normal means of writing, this is not possible. Let us imagine, though,
that the author has a computer with buttons that produce words on a screen or
sheet of paper the author cannot see, and that occasionally produce the wrong word.
Suppose that by hitting the quote key, the ‘car’ key, the quote key, and so on, the
author produces ‘Chicago’ is a four-letter word. The sentence she produced clearly
says (in one sense) that the expression ‘Chicago’ is a four-letter word, and is false.
But did the author say that the expression ‘Chicago’ is a four-letter word the way
Kaplan said that a picture of Agnew was a picture of a famous philosopher? Did the
author express any concept that in fact refers to the word ‘Chicago’? It seems to me
that the answer to both questions is ‘no,’ supporting the deictic hypothesis. In the
case I imagined, the author did not inadvertently say that the expression ‘Chicago’ is
a four-letter word. The computer malfunctioned, not the author. A common mistake
in technical writing is to forget to write the quotation marks. Copy-editing may
catch the omission, and supply the marks. If quotation were a form of demonstrative
reference, this would be as impossible as using ‘this’ demonstratively but forgetting
to point.
25
See also Washington (1992: 605), whose assumption that “naming, describing, and demonstrat-
ing . . . exhaust the standard ways of mentioning non-linguistic objects” (1992: 582) is true only if
‘demonstrating’ covers all forms of indexical reference.
Reimer (1996: 135–6) uses the theory that quotation marks are pointers to argue
that quoted expressions do not strictly speaking refer to themselves on the grounds
that what we point to are tokens, whereas what we refer to are types. This overlooks
the fact that we point at a type by pointing at a token of the type. I can point at the
word ‘car’ by pointing at the token in this line on the page I am viewing. The same
goes for producing and perceiving words. To produce a word is to produce a token
of the word. To see a word is to see one of its tokens.
A genuine problem for the demonstrative theory is a worry for mine as well.
Whenever anyone writes the word ‘Chicago,’ what occurs between the quotation
marks in any context of actual writing is a token of several different word types,
including: a written word type (“The word ‘Chicago’ is in italics”); a more abstract
word-type that has spoken as well as written tokens (“The word ‘Chicago’ has
the accent on the second syllable”); and an even more abstract word-type that has
different inflections (“run and ran are the same word”). On any given occasion,
does the word ‘Chicago’ refer to a token or a type, and if the latter, which type?26
Whether the referent is a token or type is determined by whether the speaker is using
word with its token or type sense. If word has its type sense, then the particular
type is selected by the subject’s focus. We observed in Sect. 4 that when a speaker
is looking at the stern of a ship, he is also looking at the ship. A presentation of
the former is a presentation of the latter. But the speaker may be focusing on the
stern rather than focusing on the ship. His focus determines whether ‘That is over
a thousand feet long’ is true or false. In the same way, someone looking at a word
token may be focused on the token or one of its types. If the latter, he may be
focusing on a written word type, or the more abstract type that has spoken as well as
written tokens. The demonstrative theory cannot appeal to focus, because we focus
on objects through presentations, not demonstrations.
12 Davidson’s Theory
Donald Davidson’s (1979: 90) remark that “quotation marks . . . help refer to a shape
by pointing out something that has it” suggests the demonstrative theory discussed
above. His more detailed formulations are quite different, however. For example:
The singular term is the quotation marks, which may be read ‘the expression a token of
which is here.’ (Davidson 1979: 90)
It is implausible semantically and syntactically to think of the quotation marks in

(45a) all by themselves as singular terms.27 As Reimer (1996: 134) observed, they
are more like “demonstrations” (ostensive devices like pointing) than “demonstra-
26
Christensen (1967: 363–4), Goldstein (1984: 4), Bennett (1988: 403), Washington (1992: 595–7,
599–601), Garcia-Carpintero (1994), Cappelen and Lepore (1997: 441, 1999, 2012: 22–3), Saka
(1998: 124ff, 2011: 315), and Wertheimer (1999: 519).
27
Wertheimer (1999: 516), Saka (2006: 459–63), and Cappelen and Lepore (2012: 22).
330 W.A. Davis
tives” (indexicals whose reference is determined by an ostensive device). But if

‘here’ is demonstrative in Davidson’s gloss, the speaker will point at a place, not an
expression. The most serious problem is that Davidson’s gloss would not refer to
the word ‘car’ unless ‘here’ refers to the right location. And if it did, the resulting
reference would not be rigid, since different tokens are at that location in different
possible worlds.
Another Davidsonian suggestion is even more defective:
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:
[A] ‘Alice swooned’ is a sentence
we could write:
[B] Alice swooned. The expression of which this is a token is a sentence. (Davidson
1979: 91)
First, moving ‘Alice swooned’ out of the quotation marks in this way results in
[B] asserting something [A] does not, namely, that Alice swooned. [B] would be
equivalent to [A] only if there were some indication that the first sentence in [B]
was merely being displayed. [B] has no such interpretation in English. Second, [B]
implies [A] only if ‘this’ refers to a token of ‘Alice swooned.’ Nothing in Davidson’s
gloss guarantees that.28 Davidson consequently had to add “imagine the token of
‘this’ supplemented with fingers pointing to the token of ‘Alice swooned’” (1979:
91). Davidson’s analysis fails to capture the primary function of auto-reflexive
quotation, which is to determine what expression is referred to. Third, Davidson’s
view has some of the defects of the token-reflexive theory (Reichenbach 1947: 248).
On Davidson’s gloss, we cannot say that sentence [A] is true, since [A] is a sentence
type tokened in numerous copies of books. There is a token on the physical page you
are reading, and another on the one I am looking at now. The grammatical subject
of [A] refers to the type of which they are tokens; [A] itself contains no tokens.29
At best we could say that some of [A]’s tokens are true. Finally, Davidson’s gloss
makes iterated quotation impossible. When it is applied to (51), yielding (51a), the
result has only the second and third defects. But when Davidson’s gloss is then
applied to the quoted material in (51a), the result makes no sense (Saka 1998: 119).
0
(51) ‘Chicago’0 is an expression consisting of a name in single quotes.
(a) ‘Chicago.’ The expression of which this is a token consists of a name
in single quotes.
(b) Chicago. The expression of which this is a token. The expression of
which this is a token consists of a name in single quotes.
If the second ‘this’ in (51b) refers to either of the previous “sentences,” then (51b)
is obviously false. (51b) cannot possibly be equivalent to (51a) because the referent
of ‘this’ in (51a) does not appear in (51b).
28
Compare Burge (1986), Stainton (1999), and Cappelen and Lepore (1999, 2012: 14).
29
Contrast Goldstein (1984: 4) and Saka (1998: 123ff). What sentence type [A] contains is an
occurrence of the sentence ‘Alice swooned.’ Recall fn. 14.
13 Summary
Noun phrases of the form the word ‘w’ are used to refer to the word quoted.
They express a specific indexical concept, consisting of the primary indexical
concept expressed by the word used as a restricted definite description linked to
a presentation focused on the word quoted. Primary indexical concepts are a type of
thought part consisting of a sortal concept and a determiner. The sortal component of
the concept expressed by the word is the concept expressed by word on the occasion
of use. The determiner of an indexical concept enables it to link to a specific range of
determinants. The referent of an indexical concept is the referent of its determinant.
Since the referent of a presentation is the object it is focused on, noun phrases of the
form the word ‘w’ refer to the word in quotes.
Noun phrases of the form the idea “w” are used to refer to the idea expressed
by the word quoted. They express a specific indexical concept, consisting of
the primary indexical concept expressed by the idea used as a restricted definite
description linked to a presentation focused on the idea expressed by the word
quoted. The sortal component of the concept expressed by the idea is the concept
expressed by idea on the occasion of use. While the term has other meanings, I use
it to mean thought or thought part. Since the referent of a presentation is the object
it is focused on, noun phrases of the form the idea “w” refer to the idea expressed
by the word in quotes.
The quotation marks in these phrases are optional punctuation marks. Italics
or emphasis can be used for the same purposes: to help indicate the object of
the presentation that determines the referent, either an expression or the idea it
expresses. Unmarked expressions of the form the word w and the idea w are
commonly used in the same ways, especially in speech, and w or ‘w’ alone can
be used elliptically to mean “the word w” or “the idea w.” Quotation marks in
these uses have no referents of their own. They are like the act of pointing that
accompanies demonstrative uses of indexicals, but are not determinants of reference.
Auto- and ideo-reflexive descriptions are deictic rather than demonstrative.
References
Bach, K. (1987). Thought and reference. Oxford: Oxford University Press.

Bennett, J. (1988). Quotation. Noûs, 22, 399–418.
Bezuidenhout, A. (1997a). The communication of de re thoughts. Noûs, 31, 197–225.
Bezuidenhout, A. (1997b). Pragmatics, semantic underdetermination and the referential/attributive
distinction. Mind, 106, 375–407.
Burge, T. (1986). On Davidson’s ‘Saying that’. In E. Lepore (Ed.), Truth and interpretation (pp.
190–208). Oxford: Blackwell Publishing.
Cappelen, H., & Lepore, E. (1999). Reply to Saka. Mind, 108, 741–750.
Cappelen, H., & Lepore, E.. (2012). Quotation. In E. N. Zalta (Ed.), Stanford encyclopedia of
philosophy (pp. 1–56). Stanford. http://plato.stanford.edu/archives/spr2012/entries/quotation/
332 W.A. Davis
Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press.

Christensen, N. (1967). The alleged distinction between use and mention. Philosophical Review,
76, 358–367.
Clark, H. H., & Gerrig, R. J. (1990). Quotations as demonstrations. Language, 66, 764–805.
Davidson, D. (1979). Quotation. Theory and Decision, 11, 27–40. Reprinted in Inquiries into Truth
& Interpretation (pp. 79–92). Oxford: Clarendon Press, 1984.
Davis, W. A. (2003). Meaning, expression, and thought. New York: Cambridge University Press.
Davis, W. A. (2005). Nondescriptive meaning and reference. Oxford: Oxford University Press.
Davis, W. A. (2013). Indexicals and de se attitudes. In N. Feit & A. Capone (Eds.), Attitudes ‘De
Se’: Linguistics, epistemology, metaphysics (pp. 29–58). Palo Alto: CSLI Publications.
Davis, W. A. (2014). On occurrences of types in types. Australasian Journal of Philosophy, 92,
349–363.
Davis, W. A. (2016a). A theory of saying reports. In A. Capone, F. Kiefer, & F. Lo Piparo (Eds.),
Indirect reports and pragmatics: Interdisciplinary studies (pp. 291–332). Cham: Springer.
Davis, W. A. (2016b). Irregular negatives, implicatures, and idioms. Cham: Springer.
Davis, W. A. (2017a). The property theory and de se attitudes. In K. Korta, & M. de Ponte (Eds.),
Reference and representation in thought and language. Oxford: Oxford University Press.
Davis, W. A. (2017b). Pronouns and implicature. In P. Cap, & M. Dynel (Eds.), Implicitness: From
lexis to discourse. Amsterdam: Benjamins.
Donnellan, K. (1968). Putting humpty dumpty together again. Philosophical Review, 77, 203–215.
Donnellan, K. (1977). The contingent a priori and rigid designators. Midwest Studies in Philoso-
phy, 2, 6–27.
Dretske, F. (1969). Seeing and knowing. Chicago: University of Chicago Press.
Evans, G. (1977). Pronouns, quantifiers, and relative clauses (I). Canadian Journal of Philosophy,
7, 467–536.
Evans, G. (1980). Pronouns. Linguistic Inquiry, 11, 337–362.
Fodor, J. A. (1998). Concepts. Oxford: Oxford University Press.
Frege, G. (1892). On sense and reference. Zeitschrift fur Philosophie und Philosophische Kritik,
100, 25–50. In P. Geach and M. Black (Eds.), Translations from the Philosophical Writings of
Gottlob Frege (pp. 56–78). Oxford: Basil Blackwell, 1952.
Garcia-Carpintero, M. (1994). Ostensive signs: Against the identity theory of quotation. Journal
of Philosophy, 91, 253–264.
Garcia-Carpintero, M. (2017). Reference and reference-fixing in pure quotation. In P. Saka & M.
Johnson (Eds.), The semantics and pragmatics of quotation. Dordrecht: Springer.
Goldstein, L. (1984). Quotation of types and other types of quotation. Analysis, 44, 1–6.
Green, M. S. (2007). Self-expression. Oxford: Oxford University Press.
Grice, H. P. (1957). Meaning. Philosophical Review, 66, 377–388.
Kaplan (pp. 481–563). Oxford: Oxford University Press. 1989.
Kaplan, D. (1978). Dthat. In P. Cole (Ed.), Syntax and semantics, 9, pragmatics (pp. 221–253). New
York: Academic. In P. Yourgrau (Ed.), Demonstratives. (pp. 11–49). Oxford: Oxford University
Press, 1990.
Kripke, S. (1972). Naming and necessity. In D. Davidson & G. Harman (Eds.), Semantics of natural
language (pp. 253–355., 763–69). Dordrecht: D. Reidel.
Neale, S. (1990). Descriptions. Cambridge, MA: MIT Press.
Partee, B. (1973). Some structural analogies between tenses and pronouns in English. Journal of
Perry, J. (1979). The problem of the essential indexical. Noûs, 13, 3–21.
Powell, G. (2001). The referential-attributive distinction – a cognitive account. Pragmatics and
Cognition, 9, 69–98.
Prior, A. N. (1971). Objects of thought. Oxford: Clarendon Press.
Recanati, F. (1989a). Referential/attributive: A contextualist proposal. Philosophical Studies, 56,
217–249.
Recanati, F. (1989b). The pragmatics of what is said. Mind & Language, 4, 293–329.

Reichenbach, H. (1947). Elements of symbolic logic. New York: Macmillan.
Reimer, M. (1996). Quotation marks: Demonstratives or demonstrations? Analysis, 56, 131–141.
Reimer, M. (1998a). The Wettstein/Salmon debate: Critique and resolution. Pacific Philosophical
Quarterly, 79, 130–151.
Reimer, M. (1998b). Donnellan’s distinction/Kripke’s test. Analysis, 58, 89–100.
Saka, P. (2006). The demonstrative and identity theories of quotation. Journal of Philosophy, 103,
452–471.
Saka, P. (2011). The act of quotation. In E. Brendel (Ed.), Understanding quotation (pp. 303–322).
Berlin: De Gruyter Mouton.
Salmon, N. (1982). Assertion and incomplete definite descriptions. Philosophical Studies, 42,
37–45.
Soames, S. (1986). Incomplete definite descriptions. Notre Dame Journal of Formal Logic, 27,
349–375.
Stainton, R. J. (1999). Remarks on the syntax and semantics of mixed quotation. In K. Murasugi
& R. Stainton (Eds.), Philosophy and linguistics (pp. 259–278). Boulder: Westview.
Strawson, P. F. (1950). On referring. Mind, 59, 320–344.
Wertheimer, R. (1999). Quotation apposition. The Philosophical Quarterly, 49, 514–519.
Wettstein, H. K. (1981). Demonstrative reference and definite descriptions. Philosophical Studies,
40, 241–257.
Wetzel, L. (2009). Types and tokens: On abstract objects. Cambridge, MA: MIT Press.
Referential Analysis of Quotation
Dale Jacquette
Abstract A non-Fregean solution is offered to the problem of understanding the

meaning of quotation statements. Quotations are analyzed by Frege in terms of
a distinction that is judged unnecessary and counterintuitive, between customary
and quoted indirect reference interpreted as customary sense. There are intuitive
objections to Frege’s maneuver that are avoided in a proposed referential analysis in
which quoting, like being taller than, is explained as a relation between two existent
entities, a quoted subject and the exact linguistic type entity that the quoted subject is
correctly quoted as articulating. The referential analysis is further extended to model
four difficulties for standard neo-Fregean semantics, including (a) the use-mention
distinction, (b) the intensionality-extensionality distinction, (c) the intensionality of
syntax numbering contexts, and (d) criticism of Quine’s nine planets argument for
the intensionality of alethic modal contexts.
1 Quotation
Quotation statements are generally thought to be semantically peculiar. Whereas

proper names and definite descriptions are understood extensionally, the content
within quotations often does not support intersubstitution of coreferential terms or
materially equivalent sentences salva veritate. Quotation statements are categorized
in contemporary theory of meaning as intensional – off-limits to intersubstitutions of
coreferential terms or sentences. Unquoted contexts, until proven otherwise, can be
considered fair-game purely extensional, admitting intersubstitutions without truth-
value sacrifice.1
1
The classic secondary source critical of Frege (1970) is Linsky (1967: especially 22–38). A recent
useful collection of essays on indirect discourse reportage in quotation is Capone et al. (2016).
D. Jacquette (deceased) ()
University of Bern, Bern, Switzerland

https://doi.org/10.1007/978-3-319-68747-6_12
336 D. Jacquette
2 Truth and Substitution in Quotations
Contrast ‘A is taller than Z’, where ‘A’ and ‘Z’ refer to objects that are said to stand
in a certain relation of physical height to one another, with ‘S is quoted as saying,
“q”.’ If A is taller than Z, and A D B, then B is taller than Z. There is no room for
dispute where purely extensional relational terms like ‘taller than’ are concerned in
the inference, provided only that the assumptions are true.
Quotation, whether direct or indirect, is supposed to be different because it
purports to be exact reportage of a speaker’s utterance on a particular occasion. If S
says ‘q’, and ‘q’ is true if and only if ‘p’ is true, it does not follow that S said ‘p’;
and if a police spokesperson says that the alleged criminals have been apprehended,
it might be true that the persons identified as alleged criminals are actually innocent
of any offense, a fact to appear only later in the course of their trial at law. Of
course, the police spokesperson does not say that individuals actually innocent of
any offense have been apprehended, but would be misquoted as saying anything of
the kind.
Understood in this way, quotation sentences are always either contingently true
or contingently false. They propose as a contingently true fact or falsehood that
a quoted, hence intelligent and linguistically articulate, subject S, at some time
uttered (or inscribed, etc.) the exact quoted sentence ‘q’. The intensionality of
the conventional quotation context ‘ : : : ’ implies that as truth conditions for any
quotation sentence of the form, ‘S said (etc.) “q”,’ no Fregean proper name or
definite description or concept-functor appearing in ‘q’ is permitted substitution
by any other coreferential proper name or definite description, nor can another
sentence-type be substituted for ‘q’ in such a way as to satisfy the same truth grounds
or truth-makers as does ‘q’, in any exact quotation statement of the form, ‘S said
“q”.’ S said or did not say precisely sentence ‘q’ and none other on the occasion
intended, consisting in each instance of a particular string of particular words. If so,
then in saying ‘q’, S did not, for example, say ‘p’, even if p $ q, since ‘q’ ¤ ‘p’.2
3 Analysis Proposal Overview
A quasi-Fregean solution to the problem of understanding the semantics of quo-

tation statements is forthwith explained and defended on intuitive grounds. The
guiding idea is that when someone is quoted as saying ‘q’, the entire term ‘q’ refers
to the sentence constructed as an exactly grammatically isomorphic construction
of the very same words that the quoted subject is supposed to have uttered or
inscribed on the intended occasion when the quoted subject’s words were quoted.
By implication, ‘q’ refers in the same stroke to the past utterance or inscription
2
For a discussion of the non-identity and material equivalence substitution failure in connection
with slingshot arguments, see Jacquette (2010: 136–140).
Referential Analysis of Quotation 337
event in which S is quoted. ‘S’ refers to the quoted subject, and ‘said “q”‘ refers to
the putative fact of sentence type ‘q’ and none other being uttered or otherwise
linguistically expressed on the intended occasion by quoted subject S. Implicit
reference to the intended occasion in which the quoted remarks are quoted thereby
becomes an undetachable part of the total meaning of the statement of the quoted
subject’s quoted assertion.
As in ‘A is taller than Z’, the proposed analysis of quotation contexts makes both
predicational relata purely referential. Treating ‘q’ purely referentially means in part
that we need not care what specific proposition q is intended to represent, or whether
or not q is true. Nor need we correctly understand what ‘q’ means in any language,
its import, propositional content, modality, or appropriate usages, in order for it to
be true and for us to be able in principle to know whether or not it is true that S
said “q” in quoting S as saying ‘q’. This essay develops, within a generally Fregean
semantic framework, a non-Fregean referential analysis of quotation contexts that
tests its mettle by being further extended to model four difficulties for standard
neo-Fregean semantics. The applications include: (a) the use-mention distinction,
(b) the intensionality-extensionality distinction, (c) the intensionality of syntax
numbering contexts, and (d) criticism, via the referential analysis of quotation, of
Quine’s famous number-of-planets argument for the intensionality of modalities
involving intersubstitutions of coreferential numerical identities across true non-
iterative cognitively informative identity statements as well as within the scopes of
alethic modal operators.
4 Reference on Fregean Semantic Principles
Gottlob Frege’s two-dimensional semantics of propositional meaning distinguishes

between the sense and reference of every singular referring expression and of
every genuine true or false Fregean thought (Gedanke), which is to say, abstract
proposition. Frege takes a special interest in identity relation predications, as they
appear in the theorems of elementary arithmetic in formulas like 1 C 1 D 2. These
are among the most basic mathematical truths that he wants to be able ultimately
in Grundgesetze der Arithmetik to reconstruct on purely logical terms. Identity
statements in arithmetical equations are an indispensable part of mathematical
expression. There are many specialized truth-preserving algorithms that are estab-
lished precisely for manipulating syntax constructions on both sides of an identity
relation sign in ways that are formally theoretically interesting or practically useful.
We see the importance of identity relations for Frege’s project, and hence the
motivations for his 1892 landmark essay, ‘Über Sinn und Bedeutung’ (‘On Sense
and Reference’).3
3
Frege (1970: 56–78). More detailed analysis of Frege’s theory of names and oratio obliqua
contexts is offered by Jacquette (1998: 63–79, 2016: chapter 8).
338 D. Jacquette
There are two apparently exhaustive alternate logical forms of identity statements
that Frege’s general semantic theory takes notice of. They are iterative a D a
and non-iterative a D b. The two forms occasion Frege’s general two-dimensional
theory of meaning applied to both iterative and non-iterative logical forms of
identity statements. Iterative identity statements are cognitively uninformative,
but when true they refer to the True just as much as true non-iterative identity
statements. The referential meaning (Bedeutung) of true iterative and non-iterative
identity statements is always identical. The only other available referential meaning
of an identity statement is the False when the identity relation expressed does not
obtain.
Frege argues on the basis of the distinction in their respective predicational forms
that iterative versus non-iterative identity statements can nevertheless be different in
sense meaning (Sinn). Frege predicates the potential cognitive informativeness of
non-iterative identity statements on the difference in sense meaning despite identity
of referential meaning when it happens that a non-iterative identity predication a
D b, in total or considered as a whole, has both sense and referential meaning. An
identity predication having referential meaning depends compositionally for Frege
entirely on whether or not the proper name ‘a’ has referential meaning in iterative
identity a D a, or on whether both proper names ‘a’ and ‘b’ have referential meaning
in non-iterative identity statement a D b. The putative identity relation, Sherlock
Holmes D the private detective who solved the Hound of the Baskervilles case, is
not true on the authority of the Arthur Conan Doyle story, for Frege, nor is it false
for subtle reasons likely to elude most readers of Georgian-Victorian-Edwardian
British crime fiction. The sentence instead, despite possessing Fregean sense, is
referentially meaningless, referring to neither the True nor the False.4
Frege holds that proper names including definite descriptions must have Fregean
sense meaning, but they need not refer. If they do refer, then they can refer only
to existent entities in a purely extensionally comprehended referential semantic
domain. Fregean Gedanken (thoughts) have a sense that by semantic composition-
ality is a function of the senses of the thought’s predicational components’ senses.
There is supposed to be a difference in the sense of the Fregean proper name ‘The
Morning Star’ as contrasted with the sense of ‘The Evening Star’, although the two
proper names have the same reference in the existent planet Venus. The ostensible
difference in Fregean sense of ‘The Morning Star’ and ‘The Evening Star’ despite
their identical reference allows Frege to explain the cognitive informativeness of
the non-iterative type identity statement, ‘The Morning Star D The Evening Star’.
There is accordingly in Frege’s semantics a referential dimension and an epistemic
dimension to the meaning of identity statements. Frege’s two-dimensional theory
4
Frege allows that proper names including definite descriptions and sentences expressing thoughts
or propositions can have sense while lacking reference. ‘Sherlock Holmes’ is an example of the
former; ‘Sherlock Holmes is a detective’ of the latter. Frege’s semantic compositionality thesis
implies that the sentence has sense but no truth-value when its subject term proper name has sense
but no reference to any existent entity in the Fregean referential semantic domain consisting of all
and only existent entities.
of identity statement meaning is predicated on his more general two-dimensional

semantics of the sense and reference of any fully meaningful predication, non-
iterative identity statements included.
That there are serious difficulties in Frege’s sense-reference distinction and his
effort to apply the distinction in explaining the cognitive informativeness of non-
iterative identity statements is no longer a dark secret of the trade. I consider a
number of these distributed about the discussion to follow as rationale for replacing
Frege’s semantics of quotation contexts with a more broadly conceived Fregean-in-
spirit-and-inspiration non-Fregean referential analysis. It is difficult for one thing to
see how the senses of coreferential Fregean proper names can possibly be distinct,
if Fregean sense is supposed to determine reference, especially with respect to fine
differences among referents scattered across logically possible worlds.
We thinkers and language users can have different opinions about the objective
sense of a Fregean proper name, as Frege remarks in his footnote on the third
page of ‘On Sense and Reference’ (1970: 58), concerning the sense meaning
of the proper name ‘Aristotle’. Frege does not seem to want the explanation to
take that direction, perhaps because it seems too psychologistic. Thus, among his
architectonic resources he must distinguish between the senses of ‘The Evening
Star’ and ‘The Morning Star’ themselves in the identity statement, ‘The Evening
Star D The Morning Star’. Venus is not a star, in the first place, although to the
naked eye it looks like one, so there is trouble from the outset in transcribing Fregean
senses directly from these proper names. We cannot in good conscience argue that
‘The Evening Star’ means exclusively the first bright heavenly body seen in the
evening, because that must also be part of the sense of the Fregean proper name,
‘The Morning Star’, if the senses of these names are to converge precisely on the
same reference in all possible worlds.
If we write referentially equivalently, Hesperus D Phosphorus, then most of us
will be hard-pressed to explain their presumed difference in Fregean sense, cashing
out the supposedly distinct Fregean sense meaning of ‘Hesperus’ versus that of
‘Phosphorus’ in any generalization of Frege’s solution to the problem of cognitive
informativeness in non-iterative identity statements. Algebraically, as Frege should
be more sensitive, there is scant prospect for distinguishing the Fregean senses
of ‘a’ and ‘b’, when all we know in a D b is that a single arithmetical value is
designated by the two proper names. What is the sense of ‘a’? Being identical to b?
And b, is it identical to a? Frege wisely avoids this trap, on the assumption that
‘Being identical to ____’ is a purely extensional context that gratefully accepts
intersubstitutions of coreferential terms, exactly like ‘a’ and ‘b’, salva veritate.
The imagined difference immediately comes to nothing. Since, however, cognitive
informativeness seems anyway to be a matter of individual subjective psychology,
distinctive background knowledge and experience, and so on, as to whether it is
news to a certain individual to be told that The Evening Star D The Morning
Star, it would seem more appropriate for Frege to back away from differences in
sense meanings of the two names, but take refuge instead in the different opinions
that different thinking subjects can have concerning the sense meanings of the
proper names and definite descriptions contained especially in non-iterative identity
340 D. Jacquette
statements. My brother believes that ‘The Evening Star’ has as part of its sense
meaning the property of being the last bright heavenly body to be seen in the
morning, whereas I believe that the sense meaning of ‘The Evening Star’ is only to
be the first bright heavenly body to be seen in the evening. It comes as a cognitively
informative surprise to me but not to my brother on this occasion to learn that The
Evening Star D The Morning Star. Indeed, I had been praying and making nightly
burnt offerings to The Evening Star to destroy The Morning Star!
Remarkably, for all his reliance on reference as a basic semantic relation, Frege
says next to nothing about how a proper name refers to an existent object or a
proposition refers to the True or to the False. It just does. Names and referents go
together, like knives and plates around the dinner table in the Humean image Frege
exploits to explain his concept of beiderseits eindeutigen Zuordnung (a one-to-one
correspondence), in his 1884 Grundlagen der Arithmetik (2007: 72), and we need
not inquire how the correlation is instantiated. Another line of explanation would
have it that a proper name refers because there is an explanatorily further irreducible
conceptually primitive intentionality relation connecting the use of the name with
its intended object. For Frege the intentionality of reference option must suffer from
two objections. First, that to speak of there existing an intentional relation between
a name’s use and its intended object is to say nothing more enlightening than that
the name refers to the object. Second, that insofar as intentionality adds anything
to reference it contributes only an objectionable psychologism of thinking subjects
intending something when a name is used. Frege need not doubt that naming and
sentential reference involve intentionality, while insisting that referential meaning
in the objective scientific semantic sense transcends mundane facts of individual
psychology that depend on rather than determine extant mind-independent abstract
or third-realm referential relations (Frege 1977: especially 26–30).
Fregean referential semantics need not provide any deeper analysis of reference
than Frege ventures in his theory of meaning. Proper names refer because there
is a many-one relation between alternative coreferential proper names and their
referents (a, b; ‘The Morning Star’, ‘The Evening Star’; ‘Hesperus’, ‘Phosphorus’),
when indeed a proper name has both Fregean sense meaning and reference. Fregean
reference instantiates a special many-one naming relation, by which a linguistic
symbol represents an existent entity. Explanation has to end somewhere, even if the
concept of what it means for a symbol to represent, as opposed to intend or refer to,
an existent entity does not seem to improve understanding.
Whatever reference is, it seems difficult to define without spooning up synonyms
that collectively do not amount to an analysis of the concept of reference, of
reference relations or the referential meaning of proper names and referentially
meaningful sentences. If reference is not something psychological then, as Frege
seems to prefer, it can only be an abstract correlation of linguistic items and
corresponding existent objects. We invoke reference on Fregean semantic principles
as the abstract many-one relation of referring expressions and the referents they
symbolize, without expecting to explain how a given proper name or sentence refers
to or symbolizes an existent entity or Fregean reified truth-value, the True or the
False.
5 Fregean Model of Quotational Meaning as Indirect

Reference
The treatment of quotation in Frege’s theory of meaning depends on yet another

semantic distinction. Frege establishes different categories for customary as opposed
to indirect reference, and speaks of customary and indirect sense, despite there being
no place, further discussion, or obvious use in his semantic system for a category of
indirect sense.
Frege deploys the distinction in order to forestall embarrassing intersubstitution
failures in what are supposed to be purely extensional referential contexts. The
proposal applies to quotation as a special case, for which sake it is made the subject
of comparison below with another preferred solution. It provides the basic template
for further discussion of a referential analysis of quotation and its applications. Frege
explains the semantics of oratio obliqua indirect discourse or quotation in these
terms in his 1892 essay, ‘On Sense and Reference’:
If words are used in the ordinary way, what one intends to speak of is their reference.
It can also happen, however, that one wishes to talk about the words themselves or their
senses. This happens, for instance, when the words of another are quoted. One’s own words
then first designate the words of the other speaker, and only the latter have their usual
reference. We then have signs of signs. In writing, the words are in this case enclosed in
quotation marks. Accordingly, a word standing between quotation marks must not be taken
as having its ordinary reference : : : It is quite clear that in this way of speaking words do not
have their customary reference but designate what is usually their sense. In order to have
a short expression, we will say: In reported speech, words are used indirectly or have their
indirect reference. We distinguish accordingly the customary from the indirect reference of
a word; and its customary sense from its indirect sense. The indirect reference of a word
is accordingly its customary sense. Such exceptions must always be borne in mind if the
mode of connexion between sign, sense, and reference in particular cases is to be correctly
understood. (Frege 1970: 58–59)
When it is true that ‘S says “q”‘, then S is being quoted as saying ‘q’. The
intensionality marked by quotation marks by which ‘q’ is surrounded protect
it conventionally from intersubstitution of coreferential terms within ‘q’ or any
entire expression materially equivalent to q. Frege’s ingenious interpretation of the
situation is that in ‘S says “q”‘, ‘S’ makes customary reference to S, an obliging
frequently quoted and otherwise unknown thinking and linguistically competent
subject, while ‘q’ is involved in what Frege calls indirect reference. A component of
a referentially meaningful expression must itself have referential meaning, if Frege’s
semantic compositionality thesis is correct, and if proposition ‘q’ cannot be assumed
to make normal sentential reference to the True or the False merely from the fact that
someone has uttered the sentence. Frege, looking around at the available distinctions
already in play within his scheme of category distinctions, in turn identifies indirect
reference with customary sense. If S is quoted as saying, ‘Aristotle is a philosopher’,
then the quoted expression according to Frege does not refer to either the True or the
False, but to the customary Fregean sense of the sentence, viz. that something named
Aristotle has the property of being a philosopher. The proposition’s customary sense
342 D. Jacquette
identified with its indirect reference is further analyzable as the set of Fregean sense
meaning properties belonging to the sense of the proper name ‘Aristotle’, including
prominently the property of being a philosopher.
Frege’s proposal is neat and comprehensive. The distinction between customary
and indirect reference as customary sense unfortunately also has disadvantages
that caution against unreserved enthusiasm. It is unclear what Frege’s cross-
categorical solution is supposed to imply. When a sentence is quoted, the sentence
no longer has its customary reference to the True or the False, we have seen, but
according to Frege to its customary sense. The customary sense of the quoted
sentence is the Fregean sense meaning of the sentence, which is a compositional
function, never structurally identified, of the individual customary sense meanings
of the sentence’s component meaningful terms. The customary Fregean sense of a
sentence is a function on a presumably ordered set of properties, although Frege,
like Wittgenstein in Tractatus Logico-Philosophicus, does not offer much guidance
for others to follow in projecting the analysis.5 There is evidently supposed to be
an identity between the sense meanings of terms in an analyzed Fregean Gedanke
and the properties that cash out each term’s Fregean sense meaning. The latter, no
matter how extensive, can be chunked as property n . I consider the customary
Fregean sense of a quoted sentence to have the following simplified schematic
logical form. Where quoted sentence ‘q’ has an articulated grammatical structure
of elements ‘x1 ’ and ‘x2 ’ through ‘xn ’, Frege’s theory of quotational meaning
prescribes that, when denotes the Fregean sense of any element to which it applies,
f < (‘ x1 ’ ), (‘ x2 ’ ), : : : , (‘ xn ’ )> D f <f 1 g, f 2 g, f n g>.6
When Frege’s analysis is expanded, the implausibility of supposing that the
referential meaning of a quoted sentence is its customary sense, a function on
an ordered set of properties, is immediately apparent. What is wanted is rather
something still directly or customarily referential, a state of affairs to which
the sentence might alternatively be thought to refer or, for that matter, even in
quotation, the quoted sentence’s truth-value. Why suppose that when a politician
S is quoted as saying that ‘q’, the meaning of what S says no longer makes
truth-value reference to the True or the False in Frege’s semantics? Quotation
5
Wittgenstein (1922) in Tractatus Logico-Philosophicus is more easily let off the hook than
Frege. Wittgenstein in 2.172, 4.12, 6.13 and elsewhere considers the logical analysis of colloquial
propositions to include their form of representation and correlation that, like their logical form in its
symbolic aspect where analysis is supposed to hold in precise logical isomorphisms, transcends the
perceivable sign. If logical analysis for the early Wittgenstein transcends experience, then he cannot
be blamed for not presenting an instructive analysis for inspection. Neither Frege nor Wittgenstein
seems to have an applicable method of analysis that can be exhibited in moments of process or end
result. It is the possibility of analysis or of the structures that an analysis would reveal, if only it
could, that matters to both Frege and Wittgenstein. Satisfied of that point, the structure of analysis
except in the most general terms is not undertaken, because it is not supposed or expected to be
capable of being carried out in practice.
6
Of the voluminous literature on Frege’s semantic compositionality thesis especially recommended
are Partee (1984: 281–312), Pelletier (1994: 11–24), Hodges (2001: 7–28), and Janssen (2001:
115–136).
might make truth-value explanatorily irrelevant to anyone quoting S, but it is

presumably not irrelevant to S, or S is unlikely to have said what S is quoted as
saying. Nor is the truth-value of what is quoted obviously explanatorily irrelevant in
understanding the meaning of the quotation. Generally, when someone is quoted, the
truth-value of the quoted statement, even if the individual responsible for offering
the quotation does not independently know how things stand, is not supposed to
disappear by being changed from the otherwise expected customary reference to
indirect reference, construed as Fregean sense, when a compositional function is
applied to the component terms in the quoted sentence. What justifies Frege’s
demotion of the referential truth-value of the quoted sentence, replacing it with
a sense meaning function on ordered senses of terms and ultimately on ordered
sets of properties? A news reporter quoting a politician saying that ‘Aristotle is a
philosopher’ would likely be astonished to discover that the meaning of the sentence
quoted in attribution to the politician no longer means referentially what the quoted
politician meant by it, but rather the original sentence’s Fregean sense, dished out as
functions on ordered sets of the senses of the sentence’s individual terms or on the
sets of properties by which the sense meanings of the sentence’s individual terms
are ultimately explained.
Let the problem serve to represent a host of kindred theoretical infelicities engen-
dered by Frege’s semantics of quotation contexts. Whether or not we are driven
from Frege’s solution to a preferable alternative, it is worthwhile to compare Frege’s
original attempt at analyzing the meaning of quotation sentences in a referentially
existence-presuppositional, purely extensionalist semantics with another approach
that is different from but sympathetic to the main principles of Frege’s semantic
philosophy. There is good enough reason already in place to make it valuable to see
whether Frege’s analysis of the meaning of quotation statements can be improved
in ways Frege might imaginably have approved, even if the proposal is essentially
different from the solution Frege historically presents.
6 Non-Fregean Fregean Referential Analysis
A non-Fregean referential analysis of quotation sentences that in other ways is

Frege-inspired is now sketched. The analogy previewed is that by which to say
that ‘A is taller than Z’ involves (customary) reference to existent entities A and
Z, that are said in a certain order to stand in the relation of one’s being taller than
the other. Quotation in the proposed non-Fregean Fregean referential analysis is
modeled isomorphically on this elementary relational predication of a relation said
to be instantiated between an existent thinking subject entity and a quoted existent
linguistic entity. Although the Fregean distinction between customary and indirect
reference (D customary sense) is not further invoked, the instructive contrast is that
the referential analysis of quotation implies that the quoted sentence always has
customary and never indirect reference. Compositionality is satisfied canonically
by the construction of a quotation as a relation between the quoting subject’s proper
name and the quoted speech act in its entirety with quotation marks attached as the
proper name of a particular linguistic object.
344 D. Jacquette
Quotation in pure form is understood as a relation structurally similar to the

relation of something’s being taller than something else. Quotation is a relation
ideally between two things: someone who is quoted and what the quoted individual
is supposed to have written or said. Quotation must be exact to be correct (pace
Gregorimichelaki 2017; Saka 2017; see also Jaszczolt and Huang 2017; Johnson
2017). It goes afoul when someone is quoting another speaker and, not knowing
how to spell a key word, substitutes another less phonetically challenging term
that is hoped to mean about the same thing. That is not quoting someone’s speech
act. Quotation marks are used to set off in text the exact thing an individual is
quoted as saying, as though a photograph of words used on a certain occasion
were made as a document and inset into the quotation statement attributing exactly
those words to the person quoted. Anything less logic should not recognize as
quotation, journalists beware. It is an exact replication of an expression that
is supposed to be accomplished by literal quotation. These assumptions about
quotation are exploited in crafting a more generally Fregean referential theory
of the meaning of quotation statements that does not follow Frege historically
in distinguishing between customary sense and customary and indirect reference.
Frege’s customary and indirect reference as customary sense can appear contrived
and opportunistic, diverting indirect reference to customary sense just as it threatens
to raise interesting difficulties for Frege’s sense-reference semantics. What reason
is given for supposing that a quoted sentence is bleached of truth-value by virtue
of being quoted? Are we no longer to entertain the question whether a quoted
sentence is true or false? That would be only too convenient for politicians speaking
to journalists.
Frege’s semantics of quotation statements obliterates the truth-value of whatever
is quoted, depriving it of sentential referential meaning as identical to the True or the
False, its only poor cousin indirect reference now being its customary sense. Why
should this happen to the Fregean meaning of a sentence just because it has been
quoted? There is strictly no need for anything as drastic as Frege proposes in order
to provide a plausible semantics of quotation. Semantic theory can proceed in more
generally Fregean terms, relationally speaking. ‘A is taller than Z’ relates existent
entities A and Z. ‘S said “q”’ on the proposed referential analysis of quotation
context meaning similarly relates existent entities S and ‘q’. There is q to begin
the hierarchy, an existent or nonexistent state of affairs. If q exists, then sentence ‘q’
refers to the True in Frege’s reification of truth-values. If q does not exist, then ‘q’
refers to the False. If S utters or inscribes ‘q’, then S can be quoted as doing so by
writing ‘S said “q”.’
Quotation marks indicate that what they enclose is supposed to be the exact
content of whatever S said. Quotation in the sense of exact correct replication of a
speech act accepts no coreferential or materially equivalent substitutions, including
translations from other languages, for which special provision must be made
to avoid confusions of meaning. Often conversational implicature does duty for
explicit qualification that what appears as quotation is the translation of a quotation.
When a speaker is quoted, the quoting subject predicates a specific linguistic object
standing in expressive relation to the quoted speech act subject. The subject of the
quoted speech act expresses exactly the quoted expression attributed to the subject
by the highest intensionally impenetrable ‘q’ convention. We are permitted to infer
from ‘q’ that therefore either q or not-p, but importantly not from S said ‘q’ to q or
not-p, and certainly not ‘q or not-p’. Quoting S as saying that ‘God exists’ does not
make it true that God exists. If what S says on record is ‘The Deity exists’, then S
is misquoted if a different sentence ‘God exists’ is substituted for S’s exact words
when S is being quoted, even if ‘The Deity’ is a coreferential Fregean proper name
for ‘God’ (which it probably is not). Lexical differences among coreferentials are
often important, even if they are not considered as such by those who do the quoting
and their information consumers. It is not what S is supposed to have said, even if it
sometimes amounts to more or less the same thing and we cannot always afford to
be too fussy.
The quotation statement ‘S said “q”’ relates two existent objects, a subject
of speech acts S and a specific linguistic object ‘q’. There is nothing more to
the referential analysis of quotation except to explain that S is supposed to have
performed a speech act by means of the exact linguistic object ‘q’. That object ‘q’
has referential conditions and implications of its own. It is lifted directly out of
the quotation context wherein ‘q’ is attributed to S. If ‘q’ is treated referentially
as that exact thought expressed in exactly the terms of intensional context-shielded
sentence ‘q’, it will be immune from intersubstitutions of coreferential terms or
materially equivalent sentences salva veritate or non.
The quotation relation, when it obtains, holds strictly between whoever is quoted
as saying something and what the quoted speech act maker said, expressed as an
exact linguistic type. This is the ideal to which quotation reportage should aspire,
not only in catching the words of current events figures, but historical exposition and
criticism. If the quotation relation is between a quoted speech act maker and an exact
quoted expression, then the quoted expression becomes an object of reference like
any other existent entity in a broadly Fregean referential existence-presuppositional,
purely extensional semantic domain. S exists and stands in a speech act relation
to the exact sentence type q that S is quoted as having said exists. The quoter
comes after this fact and quotes S as saying ‘q’, in saying ‘S said “q”.’ It is
the meaning of the latter sentence that is alternatively explained by a referential
analysis of quotation that is Fregean in all other respects referentially, but does not
dodge (customary) reference to quoted statements as existent linguistic entities by
redirection to indirect reference construed as customary sense.
If S said that ‘q’, then S is related to ‘q’ in just the way that A is related to
Z in A is taller than Z. Quotation marks protect specific propositional content
from transformations of their exact internal content involving substitutions of
coreferential terms within ‘q’. The problem of quotation context meaning that
agitates Frege is resolved by a neo-Fregean referential analysis of quotational
attributions. The solution is meant to work without drawing upon yet another
historically Fregean semantic distinction between customary and indirect reference
equated with customary sense, or any comparably counterintuitive expedient. If
there is no compulsion in Fregean semantics generally construed to switch from
customary to indirect reference (customary sense) of proper name ‘Z’, in order to
346 D. Jacquette
understand the referential meaning of ‘A is taller than Z’, then we should also not
need to do so in order to understand the meaning of ‘S said “q”’. All that is needed
referentially then is for the existent linguistic object ‘q’ to be the exact speech act
type that the quoted subject is quoted as having expressed, without adulteration
in exact content by unlicensed substitution of coreferential terms or materially
equivalent propositions within quotation marks, within a syntactically sacrosanct
quoted context.
7 Applications of the Referential Analysis
The referential analysis of quotation statements proves itself by its results in a

family of related semantic applications. The analytic challenges are each briefly
described, together with a sketch of the referential analysis of quotation resolution
of the problems posed with running commentary. Four problems are addressed.
7.1 Use-Mention Disambiguation
The first application of the referential analysis of quotation to disambiguating use

and mention is also the most general. There is a difference between using a word,
phrase or sentence to refer or to make an assertion, and mentioning the expression
used in the assertion. Reporting on the use of offensive language does not make the
reporter guilty of using the language, even if the report includes an exact mention
of the epithets other persons have used.
The connection between quotation and use-mention disambiguation is direct.
Using a word or sentence to refer to something or express a proposition means
to originate an expression, even if only in repetition of another, whereas men-
tioning the same word or sentence, usually signaled orthographically by means
of quotation marks, does not author an expression or place responsibility on the
person mentioning another’s use for the meaning or truth of what is mentioned,
effectively of what is quoted. It is again as though a photograph is taken of the use
of language in a speech act and presented when mentioned for what it is worth.
What is mentioned is not necessarily used, and like other quotation contexts its
exact internal content is inviolable to intersubstitution of coreferential proper names
or entire materially equivalent sentences. Quotation can be described therefore as a
type of use-mention disambiguation, while use-mention disambiguation is achieved
by the same orthographic conventions as those observed in logically-semantically
circumspect use of quotation-designating devices, notably proper observance of
quotation mark placement.7
7
Sources to which I am indebted prominently include Christensen (1967), Bennett (1988),
Washington (1992), and Saka (1998, 2006).
7.2 Intensionality-Extensionality Criterion
An important application of quotation is to mark the distinction between intensional

and extensional linguistic contexts. The distinction in semantic philosophy is that
intensional theories begin with constitutive properties and define particular intended
objects as, or in terms of, specific combinations of constitutive properties. Some but
not all combinatorially defined intended objects exist whereas extensional semantic
theories begin with a referential semantic domain consisting of all and only existent
objects. Extensionalism defines true and false predications of properties to existent
entities in the logic’s referential semantic domain as the inclusion in or exclusion
from the extension of a predicate representing a property and all and only the
existent entities possessing the property defining the predicate’s extension.
Frege identifies the properties in question with unsaturated concept-functions
completed by existent argument-objects. He warns (2007: 62–63) that we must be
careful not to let the once existent object Julius Caesar stumble into arithmetic by
being made the argument-object of an arithmetical function (see Macbeth 2005:
161–164). The unwanted application C1(Julius Caesar) D ______ might be treated
in ways that are historically non-Fregean but still Frege-inspired. Since Caesar
neither adds nor subtracts anything from the abstracted numerical value of the appli-
cation, in this case the natural number 1, Frege could have reasonably proposed that
C1(Julius Caesar) D 1, and more generally that 8x , n[x ¤ n$FA (x) D FA ]. Literally
in the Caesar case we can have as FA (c) D FA the further instantiation C1(c) D C 1
for the natural-number system-building arithmetical successor function C1(__) D n.
It follows by contraposition that 8x , n[x D n$FA (x) ¤ FA ]. This is as it should
be, where in the same application, N is any natural number from the restricted
quantification over all natural numbers in universal 8n , C 1(N) ¤ C 1 (or –1 for
a negative-integer series-building regressor function, FA (n) D – 1(n) D n – 1; and so
on, implying –1(1) D 1 – 1 D 0 , – 1(0) D 0 – 1 D – 1 , – 1(–1) D – 1 – 1 D – 2, etc.,
instantiating in infinitely or finite but indefinitely recursive regression the succession
of negative integers.
Frege does not raise the question, but it seems reasonable to suppose that if
a semantic theory relies in any form on the distinction between intensional and
extensional contexts, interpretations, explanations, semantic classifications, truth-
conditions or the like, then a materially correct practically useful criterion ought
to be provided to distinguish in practice between the intensional and extensional.
Several proposals for such a criterion have been made in the history of semantics.
The difficulty is only to provide a test for the difference between intensional versus
extensional contexts. A criterion does not promise an analysis or explication, right
or wrong, of the concept of intensionality or extensionality, although it should be
consistent with the greater meaning of intensionality and extensionality, and might
point the way toward a correct analysis. By ‘criterion’ I mean simply a good way
to tell the difference between intensional and extensional contexts in the wide field
of speech acts. Failure of intersubstitution salva veritate, preserving the truth of
sentences containing a term when a coreferential term is uniformly substituted,
348 D. Jacquette
or an entire sentence for another materially equivalent sentence within a more

comprehensive propositional superstructure, is taken as a rough and ready sign of
intensionality. When intersubstitutions go through without a hitch in truth-value
transition, then the relevant contexts are said sharply in contrast to be extensional,
and for exmphasis sometimes purely extensional.
There are logically puzzling features of the salva veritate criterion. Consider this
recalcitrant application. Assume that ‘F D ma’ is a law of kinematics identifying
the force of a moving projectile with the projectile’s mass times its acceleration,
and assume that E D mc2 . If, as also appears undeniable, both of these scientific
laws are true, then it follows truth-functionally that F D ma$E D mc2 . The material
equivalence of these laws of classical and relativity physics certainly does not
support their intersubstitution in most presumably purely extensional scientific
explanatory contexts. It does follow logically in the first place that therefore ‘E
D mc2 ’ is a law of kinematics identifying the force of a moving projectile with the
projectile’s mass times its acceleration. Clearly, the substitution fails. The question
is why. Whoever would have guessed in advance that the context ‘is a law of
kinematics identifying the force of a moving projectile with the projectile’s mass
times its acceleration’ is anything other than a purely extensional context? Yet such
intensional context is typical of the discursive contexts found throughout physics
and other exact natural sciences without fear of intersubstitution of the terms or
equivalent sentences salva veritate. (The problem is first presented in Jacquette
2015).
The laws of kinematics and of relativity physics mentioned are materially
equivalent if both are true, but they are not logically equivalent. ‘F D ma$E D mc2 ’
is a true biconditional, provided that ‘F D ma’ and ‘E D mc2 ’ are true. It should not
be imagined that the invalidity is solved by limiting intersubstitution in quotation
contexts to coreferentially and logically equivalent sentences rather than merely
materially equivalent ones. We cannot afford to do so without incapacitating
numerous intuitively valid substitutions. Consider the elementary inference ‘True
(q), q $ r; therefore, True(r)’, where q $ r is only materially and not logically true.
The key equivalence is material rather than logical where obviously strengthening
the requirement to logical equivalence would invalidate the substitution on which
the inference depends. That is the solution’s attraction. We seem to get into trouble
when we proceed on the basis of the material equivalence in the apparently extra-
identity predicational context, ‘is a law of kinematics identifying the force of a
moving projectile with the projectile’s mass times its acceleration’. The description
is true of the kinematic law, but not of the materially equivalent relativity physics
law.
It can be only the criterion that is at fault. What better reinforcement of the
distinction between the intensional and the purely extensional might be found
instead? A referential analysis of quotation offers a framework for a criterion
to distinguish intensional from extensional contexts. Intensional contexts notably
include propositional attitudes – belief, doubt, fear, hope, and the like – that
something exists or does not exist, is true or is not true, depending on whatever a
given attitude referentially intends. The propositional attitudes connect subject S and
substitutionally impenetrable language object ‘q’. It is that that S is said to believe

and nothing else substitutionally equivalent, the equivalence of which S might also
believe. Contexts that have such substitution-impenetrable ‘quotation’ contexts are
intensional, and all others are extensional.
Some might want to finagle a way around the question of whether Einstein’s
relativity theory strictly speaking invalidates Newton’s kinematics. If it does, then
both cannot be true, and hence not materially equivalent. The referential analysis of
quotation analysis proposes first a paraphrase of the substitution-failing predication
reduced in logical form to an identity statement, and then intensionalizing the entire
resulting identity statement so as to block anywhere within it intersubstitution of
acknowledged materially equivalents laws of classical and relativity physics. Instead
of writing that ‘F D ma’ is a law of kinematics : : : etc., we write, ignoring the
identity sign in ‘F D ma’ as irrelevant for present purposes, that ‘F D ma’ D [is iden-
tical to] a law of kinematics : : : etc. The entire identity statement is intensionalized,
removing it from the possibility of contamination by intersubstitution of material
equivalents such as Einstein’s for any occurrence of the kinematic law.
Substitution of coreferential terms and equivalent sentences is supposed to
rule clean extensional expressive contexts in scientific discourse. Except for the
kinematics inference. We can block onerous substitutions without packing a good
explanation as to what has happened in that innocent-appearing predicational
context. The schematic problem seems to be the same as that afflicting substitution
in the application where a ¤ b, from which it follows that Fa ¤ Gb. These are
different predications, different states of affairs. Now if we freely assume that Fa
and that Gb, it follows classically that Fa $ Gb, even where Fa ¤ Gb. The material
equivalence ‘Fa $ Gb’ authorizes the intersubstitution of ‘Fa’ for ‘Gb’ and vice
versa in any purely extensional context. If substitution is made on the strength of
‘Fa $ Gb’ in what may be supposed to be the non-identity statement ‘Fa ¤ Gb’,
then we can at once deduce the self-non-identities ‘Fa ¤ Fa’ and ‘Gb ¤ Gb’ that
contradict the universal reflexivity of identity. Substitution gone awry? Or should
we conclude instead that, contrary to all previous expectations, identity and non-
identity contexts ‘____ D ____’ and ‘____ ¤ ____’ are intensional rather than
purely extensional?
An approved tactic is to work backward, from identity statements to the case
involving the material equivalence of a law of kinematics and another law of
relativity theory. The idea is that if we can analyze identity statements as intensional,
then the same provisions will extend to other constructions capable of being
interpreted as identity relation statements. There is, though, a difficulty in Frege’s
proposal that should first be addressed. Frege (1903, 1972) holds that identity is a
relation between names rather than things. If a is an object, then ‘a’ in single quotes
conventionally denotes a Fregean proper name for object a. In the non-iterative
identity relation ‘a D b’, if the identity holding between proper names for objects a
and b is observed, then the Fregean proper name analysis of identity should under
the same convention be read as ‘a’ D ‘b’. Where ‘a’ and ‘b’ are different names,
graphically indicated and obvious at a glance in the choice of different alphabetical
symbols, it is always true that ‘a’ ¤ ‘b’, even when a D b. The upshot is to cast
350 D. Jacquette
serious doubt on Frege’s analysis of identity as a relation between different proper

names potentially with different senses for the same existent entity (Jacquette 2011:
51–72).
The referential analysis of quotation solution is to treat the identity and non-
identity relations as intensional according to substitution salva veritate failure. We
do not write ‘“____” D “_ _ _”’, but instead ‘____ D _ _ _’. This makes identity
statements in their entirety intensional, not just the proper names or other terms they
may contain. The entire contents of an identity statement are thereby shielded as
inviolable by substitution of coreferential terms or materially equivalent sentences.
If the identity statement is true or false, we can make use of the identity relation it
expresses in logical inference and mathematical proof. Considering the semantics of
identity statements before we prepare to put identity and non-identity statements to
work, we speak of their logical structure and Fregean sense and reference meanings
as we enclose the entire identity statement in quotations. We apply the referential
analysis of quotation to account for the meaning of identity statements, considering
them or the abstract relations for which they stand as existent objects to which
thinking subjects can appear in a variety of propositional attitudes. An entire identity
statement or the identity relation it linguistically represents is thereby made an
intended object of thought. As such, the identity statement enclosed by quotation
is not subject to internal alteration of its exact contents by the intersubstitution of
coreferential terms or equivalent sentences salva veritate contained within it that is
validity-approved in purely extensional, extra-quotational predicational contexts.
The referential analysis of identity statements trades on the distinction already
marked between intensionality and extensionality under the referential analysis
of quotation. For this reason my proposal is provocatively labeled a non-Fregean
Fregean solution. Whatever is enclosed in quotations is off-limits for internal inter-
substitutions of coreferential terms or materially equivalent sentences. The speech
act photograph enclosed in quotations must not be doctored. Where identity and
non-identity statements are concerned, the entire statement is considered quoted and
as such placed outside the purely extensional playing field of intersubstitutions of
coreferentials or equivalents salva veritate or non veritate. The provision forestalls
the paradoxical inference described above involving substitution in the non-identity
context ‘a ¤ b’, leading directly to Fa ¤ Fa and Gb ¤ Gb. Provision must be made
to block iterative non-identities that violate the reflexivity of the identity principle.
Intensionalization of an entire relational predication by explicit or invisible
enclosure in quotation marks is not a move to be taken lightly, in part for the
bad precedent it might appear to set. Where identity and non-identity statements
are concerned, the intensionalization of identity statements for safeguarding against
transformation of their exact internal content by intersubstitution is not objection-
ably ad hoc. It is independently justified if identity relations are established by
Leibnizian identity-of-indiscernibles and indiscernibility-of-identicals principles.
Leibniz’s Law (so-called) is an intensional analysis of identity in terms of shared
constitutive properties by existent entities under the same or different proper names
or definite descriptions. A property-based semantics or in this case identity theory
by the intensionality-extensionality distinction as previously characterized is by

definition intensional rather than purely extensional.
If we intensionalize identity statements in their entireties as beyond the reach
of coreferential or equivalent sentence-term intersubstitution transformations, then
we have a solution ready to hand for the property-specifications of objects under
consideration when reductive paraphrase as an identity statement is available. The
example is typical where the properties of a certain scientific law are cashed
out in terms of its properties of being a law of kinematics, along perhaps with
other distinguishing characteristics. The law is identical to exactly that thing
described as having these properties. We do not suppose it necessary within our
finite capacities to provide in every instance a complete inventory of all the
distinguishing or essential properties by which the identified thing is constituted
as a particular individual, especially across logically possible worlds. Wherever
such reductions are plausibly enacted to an identity statement formulation, then,
for purposes of avoiding challenges involving intersubstitution salva veritate, it
can be a recommended strategy to appeal to the referential analysis of quotation,
enclose the entire reduced identity statement in quotations to designate the context
as intensional and consequently seal it off from internal syntactical transformation
of exact content by substitution of any of the identity statement’s component proper
names or sentences.
Two objections to the referential analysis of quotation in application to the
intensionality-extensionality distinction are resolved. It is not a question of whether
or not there should be substitution-free intensional zones within language usage
and speech act reportage, but of how the intensionalization of certain linguistic
constructions should be conducted. The part-whole choice makes for a good, if
crude, start. Needless to say, intensionalization for semantic purposes does nothing
to hamper the use of identity statements construed as intensional contexts in
their entireties within inference and explanation and other epistemic contexts to
which uses they are otherwise expected to contribute. Quotation marks properly
inserted distinguish the boundary between the intensional and extensional that the
intensional subsumes. The paradigm case is that of ordinary quotation or indirect
discourse reporting on a speech act, or in exceptional cases reflexively in the same
self-referential speech act. The intensional marked by quotation does not permit
interference with exact internal content by the intersubstitutions permitted in purely
extensional contexts where by definition of the salva veritate criterion they are
always truth-preserving. It is only a question of how to manage the quotation marks
so that unwanted intersubstitutions are precluded. Identity is a naturally intensional
relation, and other explicative predications are reasonably construed as intensional
identity statements.
352 D. Jacquette
7.3 Syntax Item Numbering Predications
Another form of quotation has to do with the number of syntax items in coreferential
terms or materially equivalent sentences. A numbering predication true of the
syntax items in one term need not be true of a coreferential term, and similarly for
materially equivalent sentences. Mark Twain D Samuel Clemens, but the number
of letters in the name ‘Mark Twain’ ¤ the number of letters in the name ‘Samuel
Clemens’.8
Differential numbering of letters in the spellings of the two coreferential proper
names is marked by impenetrable quotations. Else it would follow that ‘the number
of letters in “Mark Twain” ¤ the number of letters in “Samuel Clemens”’ would
logically imply that the number of letters in ‘Mark Twain’ ¤ the number of letters
in ‘Mark Twain’. Such things cannot be allowed to happen. Consider the application
of the referential analysis of quotation in syntax item numbering substitution
breakdowns. The quoted items, the proper names ‘Mark Twain’ and ‘Samuel
Clemens’, are interpreted referentially as linguistic objects in their own right, each
with its own length of alphabet strings, that, as different proper names, possess these
among their other accidentally shared or distinctive properties. Substitution within
such off-limits quotation marking an intensional context prevents the unwanted
substitution-enacting inferences.
Observing the sanctity of terms enclosed in quotations referentially as an
existent linguistic entity facilitates the recognition that coreferential proper names
and materially equivalent sentences can be lexically different by consisting of
different symbol strings, and that these differences must be preserved at some
level in developed languages by a syntactical device that conventionally precludes
transformation by whatever identities or material equivalences may happen to
prevail. The referential analysis of quotation applies quotation marks to this effect,
sealing off the linguistic object enclosed in quotes as contextually closed to the kinds
of truth-preserving syntax transformations that occur without restriction in all other
remaining purely extensional contexts.
7.4 Intensionality of Alethic Modality and Identity Contexts

(Quine’s Argument)
The framework developed thus far now provides the basis for a further critical
examination of W.V.O. Quine’s famous argument for the intensionality of modal
contexts.9 The argument is reinvented occasionally as having a slightly different
8
An early form of the substitution failure for syntax item numberings appears in Plantinga (1974:
224–229).
9
Quine (1976: 158–164, 1961: 139–144). The objection is not that Quine’s inference is deductively
valid, reflecting substitution failure salva veritate, but that the invalidity is more reasonably
form and content, although the underlying structure is always to show that the
following inference is deductively invalid. The argument’s invalidity in turn is
generally taken to prove that alethic modal contexts are intensional in the sense
that they do not support intersubstitution of number expressions within the scope of
a necessity operator. The argument states:
1. The number of planets D 9.

2. 9>7
3. Therefore, The number of planets > 7
The inference is deductively invalid if we suppose that the number of planets
in our solar system being greater than 7 in number is not a matter of logical
necessity. Quine has no hesitation in blaming the manifest invalidity on assumption
(2)’s modal context, ‘___ > 7’. Substituting ‘The number of planets’ for ‘9’ in
assumption (2), on the basis of the identity between the number of planets and
9, leads inexorably from true assumptions (1) and (2) in this instance to a false
conclusion (3). Quine recommends a flight from intension to avoid such awkward
implications. There are nevertheless at least two other solutions that are less drastic
while at the same time compatible with logic making its peace with intensionality
(Quine 1960: ch. VI).
Quinean paraphrase can be performed on assumption (1). The proposition is
misleadingly expressed as an identity statement, which would seem to warrant the
intersubstitution of one term in the identity for its identical in any purely extensional
context. The fact that the substitution Quine proposes does not go through tells
Quine that the context ‘___ > 7’ is not purely extensional. If (1) is not properly
understood as intended intuitively to express an identity relation in the first place,
but a cardinality predication, then substitution breakdown is avoided because the
inference has no authorizing identity statement with which to begin. The thought
expressed in assumption (1) is strategically revised to say more plausibly, not that
the number 9 is identical to the number of planets, but that the planets have the
property of being nine or ninefold in number. The identity relation disappears in the
paraphrase, and with it any rationale for substituting 9 for ‘The number of planets’ in
assumption (2) to falsely infer conclusion (3). If (1) should actually be expressed as a
predication rather than identity, then the intersubstitution in (2) leading to (3) cannot
occur. Alternatively, as proposed in the referential analysis of quotation, if identity
statements in their entirety are intensional, needing to be enclosed in quotations
when interpreted referentially, then the substitution failure salva veritate in Quine’s
argument can be explained not as a consequence of the intensionality of the alethic
modal context in (2), but of the identity statement in (1). If (1) after reflection is
thought to express a genuine identity relation, then the entire identity statement
in (1) needs to be enclosed in quotations to mark its intensionality, preventing the
substitution in (2) that is supposed to produce a false modal assertion in (3).
attributed to the identity whereby the number of planets D 9, rather than to the modal context,
___ > 7. See Jacquette (1986: 598–608, especially 602–604).
354 D. Jacquette
8 Quotation, Sense and Reference
A semantics of quotation is needed in an unpredetermined number of applications.

Philosophy of language needs to take stock of quotation contexts and provide an
adequate theory of meaning for such applications as 7.a.–7.d. above. Frege provides
the framework for such a general semantics of quotation, although he develops the
theory of meaning in a strikingly different direction, converting the customary refer-
ence of quoted expressions to indirect reference interpreted as their customary sense.
Without joining Frege in this intuitively questionable explanation, quotations are
interpreted referentially as a relation between a quoted subject S and the linguistic
object ‘q’ that S is quoted as saying, uttering or inscribing in a speech act. There is
no need for further finesse, and the properties of ‘q’, Fregean sense and customary
reference, are preserved when existent linguistic object ‘q’ is lifted from a quotation
context and considered as corresponding in quoted ordering, matching symbol-for-
symbol with the intended expression of the existent historically actual speech act
being quoted. The referential analysis of quotation provides the basis for at least
four distinct linguistic phenomena often remarked as related to quotation or to the
intensionality of quotation contexts, syntax item numbering or modal contexts. The
solution to these problems and the perspective on the intensionality-extensionality
distinction that the referential analysis of quotation affords recommends it as an
alternative to Frege’s theory of indirect reference as reversion to customary sense.
We arrive at a simpler explanation by avoiding Frege’s distinction and allowing
quotation as a relation between a speaker who is quoted and what the speaker
is quoted as saying. Extending quotation conventions to all intensional contexts
preserves them against internal modification by extensionally permitted intersub-
stitution of coreferential proper names or by materially equivalent sentences that
they might contain. Even when no one is literally quoted, the use of quotation
marks distinguishes a substitution-proof expressive context that is to be considered
a complete linguistic entity of its own, whose internal integrity is not subject to
alteration. The distinction is enough to prevent problematic intersubstitutions of
terms in contexts that may otherwise have been thought to be purely extensional.10
References
Bennett, J. (1988). Quotation. Nous, 22, 399–418.

Capone, A., Kiefer, F., & Lo Piparo, F. (2016). Indirect reports and pragmatics. Berlin: Springer
Verlag.
Christensen, N. (1967). The alleged distinction between use and mention. Philosophical Review,
76, 358–367.
Frege, G. (1903). Grundgesetze der Arithmetik, 2 vols. Jena: Verlag Hermann Pohle.
10
It is with grief that I report the sudden death of Dale Jacquette on 22 August 2016. The present
version of this paper has been lightly edited to incorporate clarifications that emerged in my
personal communications with him, relevant updating of references, and recommendations by an
anonymous reviewer. – Paul Saka.
Frege, G. (1970). On sense and reference. In Translations from the philosophical writings of
Gottlob Frege (P. Geach, & M. Black, Trans., pp. 56–78). Oxford: Basil Blackwell.
Frege, G. (1972). Conceptual notation and related articles (Trans. and Ed: T. Bynum). Oxford
Univerity Press.
Frege, G. (1977). Thoughts. In Frege, Logical investigations (Trans. and Ed: P.T. Geach). Oxford:
Basil Blackwell.
Frege, G. (2007). The foundations of arithmetic: A logico-mathematical investigation into the
concept of number (Trans. with an Introduction and Critical Commentary by Dale Jacquette).
New York: Pearson (Longman).
Gregorimichelaki, E. (2017). Quotation in dialogue. In P. Saka & M. Johnson (Eds.), The semantics
and pragmatics of quotation (pp. 195–256). Cham: Springer.
Hodges, W. (2001). Formal features of compositionality. Journal of Logic, Language, and
Information, 10, 7–28.
Jacquette, D. (1986). Intentionality and intensionality: quotation contexts and the modal wedge.
The Monist, 69, 598–608.
Jacquette, D. (1998). Intentionality on the instalment plan. Philosophy, 73, 63–79.
Jacquette, D. (2010). Logic and how it gets that way. London: Routledge.
Jacquette, D. (2011). Frege on identity as a relation of names. Metaphysica, 12, 51–72.
Jacquette, D. (2015). Semantics and pragmatics of referentially transparent and referentially
opaque belief ascription sentences. Philosophia, 43.
Jacquette, D. (2016). Frege: A philosophical biography. Cambridge: Cambridge Univerity Press.
Janssen, T. M. V. (2001). Frege, contextuality and compositionality. Journal of Logic, Language,
and Information, 10, 115–136.
Jaszczolt, K., & Huang, M. (2017). Monsters and I: the case of mixed quotation. In P. Saka & M.
Linsky, L. (1967). Referring. London: Routledge & Kegan Paul.
Macbeth, D. (2005). Frege’s logic. Cambridge: Harvard Univerity Press.
Partee, B. (1984). Compositionality. In F. Landman & F. Veltman (Eds.), Varieties of formal
semantics: Proceedings of the fourth Amsterdam Colloquium, September 1982 (pp. 281–312).
Dordrecht: Foris Publications.
Pelletier, F. J. (1994). The Principle of semantic compositionality. Topoi, 13, 11–24.
Plantinga, A. (1974). The nature of necessity. Oxford: Oxford Univerity Press.
Quine, W. V. O. (1960). Word and object. Cambridge: MIT Press.
Quine, W. V. O. (1961). Reference and modality. In From a logical point of view (2nd ed., pp.
139–144). Cambridge: Harvard Univerity Press.
Quine, W. V. O. (1976). Three grades of modal involvement. In The ways of paradox and other
essays (2nd ed., pp. 158–164). Cambridge: Harvard Univerity Press.
Saka, P. (2006). The demonstrative and identity theories of quotation. Journal of Philosophy, 103,
452–471.
Saka, P. (2017). Blah, blah, blah: Quasi-quotation and unquotation. In P. Saka & M. Johnson (Eds.),
Wittgenstein, L. (1922). Tractatus logico-philosophicus. London: Routledge & Kegan Paul.
Monsters and I: The Case of Mixed Quotation
Kasia M. Jaszczolt and Minyao Huang
Abstract According to Kaplan (Demonstratives: An essay on the semantics, logic,

metaphysics, and epistemology of demonstratives and other indexicals. In: J Almog,
J Perry, H. Wettstein (eds) Themes from Kaplan. Oxford University Press, New
York, pp 481–563, 1989a), the semantic value of an indexical is fixed by the context
of the current speech act, except for its occurrences in quotation. However, contrary
to this thesis, now known as ‘Fixity’ (Schlenker, A plea for monsters, Linguist
Philos 26:29–120, 2003), subsequent discussions have demonstrated that the content
of pure indexicals can be fixed by the intended speech act (Predelli, Erkenntnis
74:289–303, Analysis, 2011). Cross-linguistically, Fixity also proves untenable for
languages such as Amharic where the referent of a first-person pronoun can be
drawn from the reporting context or the reported situation (Schlenker). In this paper
we focus on the variety of ways the English first-person singular pronoun is inter-
preted in quotation, focusing on mixed quotation, and argue that (i) mixed quotation
is a case of language use and, pace Kaplan, the behaviour of first-person pronouns
in mixed quotation is relevant for, and testifies against, Fixity; (ii) mixed quotation
induces context-shifts or a generalisation over contexts that can be captured in terms
of what we call ‘character-at-issue’ and ‘content-at-issue’ uses; and (iii) the extant
contextualist accounts that appear to be best suited to account for the diversity of use
do not pursue the interpretation of first-person pronouns in quotation to its logical
end. To remedy this weakness we demonstrate how the use of first-person indexicals
in these Fixity-defying contexts (Kaplan’s ‘monster’ contexts) can be accounted
for in the radical contextualist theory of Default Semantics (Jaszczolt, Default
semantics: Foundations of a compositional theory of acts of communication. Oxford
University Press, Oxford, 2005; Default semantics. In: B Heine, H Narrog (eds) The
Oxford handbook of linguistic analysis. Oxford University Press, Oxford, pp 215–
246, 2010; Meaning in linguistic interaction: semantics, metasemantics, philosophy
of language. Oxford University Press, Oxford, 2016). Overall, we conclude that
K.M. Jaszczolt ()

University of Cambridge, Cambridge, UK
e-mail: kmj21@cam.ac.uk
M. Huang
Sun Yat-Sen University, China

https://doi.org/10.1007/978-3-319-68747-6_13
358 K.M. Jaszczolt and M. Huang
‘I’ is not an indexical term and quotation is not a monster, which points to the
possibility that indexicals are a philosophers’ fiction.
1 Monsters Galore
According to Kaplan’s (1989a) direct-reference view, also called ‘the fixity thesis’
(Schlenker 2003), the semantic value of an indexical expression is always given
by the context of the speech act in which it is used. Indexicals such as ‘I’, ‘here’
and ‘now’ are devices for direct reference: their referents are exhausted by relevant
aspects of the context in which they occur, such as the speaker, the time, and the
place of utterance. Kaplan contends that no logical operators can affect the semantic
value of an indexical. He calls such putative operators ‘monsters’. From the claim
that indexicals are directly referential, it then follows that monsters cannot exist
(Kaplan 1989a: 510).
Subsequent discussions in the literature have demonstrated, however, that there
are a lot of exceptions, and even a lot of types of exceptions, to Fixity. In this
section we briefly present some problematic cases for Fixity addressed in the extant
literature. In Sect. 2 we move to the discussion of the behaviour of the first-
person pronoun in quotation, focusing on mixed quotation vis-à-vis the open-closed
distinction, proposing a reanalysis of the category of mixed quotation. In Sect. 3
we propose a distinction between the uses of indexicals we call ‘character-at-issue’
and ‘content-at-issue’, as well as discuss the advantages of a dynamic, flexible
approach to what counts as a character. Section 4 moves to the proposal of a radical
contextualist account that would capture the discussed properties of the indexical in
question, followed by an analysis of the various uses of ‘I’ in mixed quotation in the
theory of Default Semantics and an example of a representation. We conclude that,
so to speak, nothing is as it seems: (i) first-person indexicals do not display the kind
of behaviour predicted by Kaplan’s ‘unfixed character C fixed content’ definition,
and (ii) mixed quotation, demonstrated to be a case of language use rather than
mention, offers a legitimate potential counterexample to Fixity. Put together, since
pace Kaplan ‘I’ is not an ‘indexical’ (which mixed quotation helps to show), there
are no associated monster operators either (of which mixed quotation would have
been a legitimate example, pace Kaplan again).
Our argument can be summarised as follows:
(a) So-called first-person ‘indexicals’ can undergo context shift outside
quotation.
(b) So-called first-person ‘indexicals’ can undergo context shift inside quotation.
(c) There are categorically different non-Kaplanian uses of the first-person
‘indexicals’ in quotation.
(d) Quotation is a case of language use. [from (b) and (c)]
(e) Quotation qualifies as a potential monster operator. [from (d)]
(f) There are different non-Kaplanian uses of ‘indexicals’ outside quotation.
(g) ‘I’ is not an indexical term. [from (c) and (f)]
(h) Quotation is not a monster. [from (g)]
Monsters and I: The Case of Mixed Quotation 359
We end up on a more general and speculative note as follows:
(i) Indexicals are a philosophers’ fiction. [from (a)–(h)]

(j) There are no Kaplanian monsters. [from (i)]
Let us begin with some acknowledged problems for Fixity. Firstly, there are time
and location shifts. As Kaplan (1989a) himself notes in passing, one can envisage
scenarios on which (1) is true; messages recorded for future use are such a case in
point.
(1) I am not here now.

The problem has acquired a fair airing in the philosophical literature. Many of these
discussions focus on the properties of the adverbials ‘here’ and ‘now’.1 Corazza
(Corazza 2004a: 308), for example, defends Kaplan’s position by proposing that
whenever these indexicals do not behave as they ought to, the explanation lies in
their discourse-anaphoric link. The referent associated with the character is fixed
but is inherited from the antecedent:
All that needs to be spelled out are the reference-fixing rules for ‘now’ and ‘here’ and an
appeal to a theory of discourse anaphora. We do need the latter for independent reasons;
we need a theory of discourse anaphora to handle cases where, in a discourse, people keep
track of objects and individuals using anaphors.
This solution works fine for ‘here’ and ‘now’. One can easily plug in one of the
successful theories of discourse anaphora, such as for example van der Sandt’s
(1992, 2012) presupposition-as-anaphora widely espoused in DRT research,2 to
show how it could work in technical detail. Once anaphoric binding is extended to
cross-sentential binding and even to, so to speak, ‘binding to made-up antecedents’
in case of accommodation, the problem seems to be solved.
But the first-person indexical, even when we limit its discussion to the English
language alone and the pronoun ‘I’, seems to be a more stubborn character than
‘here’ and ‘now’. Although according to Kaplan’s (Kaplan 1989a: 491) famous
dictum, uttering ‘I’ and pointing at someone else is ‘irrelevance or madness’, one
does not have to search far for counterexamples. Some of its uncommon uses can
indeed be accounted for by means of straightforward anaphora. The so-called ‘fake
indexical’ I (Kratzer 2009) in (2) is a case in point. (2) is an example of a bound-
variable use where the second occurrence of ‘I’ is semantically unspecified.
(2) Nobody except me admitted what they did wrong. John didn’t, Mary didn’t,
and neither did the rest of the people. Only I admitted what I did wrong.3
Other cases are more stubborn. There are situations where the speaker utters ‘I’ in
order to point at, say, a specific ‘temporal slice’ of him-/herself, as in (3).
1
See e.g. Predelli 2011 for pertinent references.
2
Discourse Representation Theory (Kamp and Reyle 1993).
3
But note that Schlenker (2003: 89–90) points out that ‘only’ in (2) acts akin to a syntactic binder.
Cf. ‘Not only I admitted what I did wrong; Peter did too’.
(3) Scenario: The speaker and her friend watch a home video of a holiday
Look, here the goat is about to attack me but I don’t realise it is me because I
think all the time that this is a window, not a mirror. So, I am trying to find
out how to open the window to warn her until : : : .4
Written notes such as (4) exemplify the same problem of a temporal shift.
(4) I am not in. Back at 2 pm.

So, such contexts exist and seem to testify against Kaplan’s direct reference view.
Corazza (2004b) attempts an explanation pointing out the special properties of the
context on these occasions, in that it has to pertain to a conventional setting. On the
other hand, against Corazza’s conventionalism, Predelli (2011) offers intentionalism
as a way to rescue Kaplan’s Fixity. First, he addresses the alleged ‘non-automatic’
reference assignment to ‘I’, ‘here’ and ‘now’ by agreeing that the contexts in
question constitute special cases: the only contexts in which (1) is true are the so-
called improper contexts. Kaplan’s contexts are proper contexts: they are 4-tuples
<cA , cL , cT , cW > such that the speaker cA is in location cL at time cT and world cW .5
As a result, (1) has to be analytically false. But when we tweak the time and place
parameters of context in such a way that they capture the value of the time and place
of the intended speech act, then the ‘improper’ contexts can be handled as well.
Corazza’s conventionalism as well as Predelli’s intentionalism will prove impor-
tant in our later analysis of I in mixed quotation in that they constitute versions
of a pragmatic account.6 It is clear that we are dealing here with the cases of
‘tainted I’ and what these two proposals do not yet provide is a fully-fledged
linguistic pragmatic theory showing how such instances of ‘tainted’ or ‘improper’
uses of ‘I’ are possible in discourse: how they are processed and what representation
they require. As we will argue in Sect. 4, semantic representation of indexicals is
best pursued as equivalent to a radical-contextualist representation of their use in
discourse. On the one hand, this constitutes a radical theoretical shift from semantic
minimalism (aided with a pragmatic overlay of conventionalism/intentionalism)
to contextualism. On the other, it is a straightforward way of embracing the
role of conventions and intentions in the meaning of ‘I’, ‘here’ and ‘now’:
since intentions and conventions do affect reference assignment, then instead of
relegating the pertinent contexts to the wastebasket of special cases (‘conventional’
or ‘improper’), we revindicate them, affording them ordinary status, at the same time
including the output of conventions and inferences in the semantic representation
itself.
4
See also Jaszczolt (2016, chapter 5) for a discussion.
5
Note that Kaplan (1989b: 591) does not exclude the possibility that more parameters constitute
context as an index: “context provides whatever parameters are needed”. See Jaszczolt 2012 for a
discussion.
6
We use italics (I) to mark a concept and single quotes to mark a lexical item (‘I’).
The directly referential status of the first-person indexical is also problematic

when viewed from a cross-linguistic perspective. As Schlenker (2003) argues, the
fixity thesis proves to be incorrect for languages such as Amharic where the referent
for ‘I’ in examples such as (5) can be recovered either from the context of the attitude
report or from the context of the reported situation itself, allowing for two different
contents pertaining to one character.
(5) Kasia says that I am to blame.

In Amharic, propositional attitude verbs are what Kaplan thought impossible: they
are monster operators in that they can quantify over the contexts of speech or
over the contexts of thought. As such, they create monster contexts. Languages
with logophoric pronouns, i.e. pronouns that stand for the subject whose speech,
thoughts or feelings are being reported (Clements 1975: 241), such as Ewe, Yoruba,
or Gokana, give further support for the thesis that not only possible worlds but also
contexts have to be quantified over.7
Acknowledging such arguments undermining Fixity, in what follows we focus
not so much on the context shift itself and how to explain it (or how to explain it
away) but rather, taking it for granted that such contexts as the ones exemplified
above are not so exceptional after all, we address a more general (and arguably
more interesting) question of how to represent such uses of I in a theory of
meaning. But it will not suffice to found the analysis on the cases of temporal
shifts, variable-bound uses, and the fact that what appears to be a translation of
the English ‘I’ behaves differently under attitude operators in languages that allow
or necessitate context-shift. The need for the manipulation of context in the case
of first-person indexicals is particularly disquieting in non-referential uses of the
first-person pronoun in mixed quotation and this is the topic to which we now
turn. As will become clear, mixed quotation will disclose very different kinds of
monsters.
2 The Curious Behaviour of I in Mixed Quotation
2.1 The Properties of Mixed Quotation
While the term ‘quotation’ does not receive a uniform definition in the literature, we
take as typical cases of quotation those in which certain expressions are employed
to report someone else’s speech or thought, usually but not necessarily by enclosing
the expressions in a pair of quotation marks, as in (6).
7
For an extensive discussion and further examples of languages displaying ‘monstrous’ behaviour
see Roberts (2014).
(6) Miles said, “I am a perfectionist”.

Other cases of quotation include citation of an expression in order to talk about
its linguistic properties (see Saka 2013: 937), such as (7), also known as ‘pure
quotation’; and, perhaps more controversially, scare quoting which serves to indicate
non-standard uses of the quoted material, as in (8).
(7) “I” is a first-person pronoun.

(8) Reese suggests that we go out for a “drinking” on Monday.8
Our discussion below focuses on the typical cases of reportive quotation, although
where appropriate we will comment on other cases.
Within quotation, a further distinction is drawn between closed quotation and
open quotation (Recanati 2001). The former is exemplified in (9). (10) is an instance
of a sub-type of the latter, called mixed quotation (see Davidson 1979).
(9) Jimi said, “I am voting Labour”.

(10) Jimi said he’s “voting Labour”.
Closed quotation aligns with the traditional view on quotation as metalinguistic
mention of the quoted material. That is, quotation would project a singular term
in the logical form of the sentence, whose referent is the quoted material (or, rather,
some linguistic type tokened by the quoted material; see Davidson (1979), and
Cappelen and Lepore (1997), among others). Thereby, when a quotation appears
in a syntactic position that maps onto a singular term as in (9), the quotation is
closed and as the traditional view goes, only the quoted material (or a related type)
is being referred to. In other words, the quoted material is mentioned, not used.
However, when a quotation appears in a syntactic position that does not map
onto a singular term, the quotation is said to be open, in that it is simultaneously
mentioned and used. It is mentioned so that the sentence cannot be true unless
the quoted material is uttered, whereas it is used in that the referent of the quoted
material is the referent of its unquoted counterpart as intended by the quotee in the
original context (Davidson 1979; Shan 2010; Recanati 2010, among others). Hence,
a two-part analysis of (10) can be given as in (100 ):
(100 ) Mention line: The words “voting Labour” were uttered to refer to v.
Use line: Jimi said that he’s v-ing.
To paraphrase, (10) is true iff Jimi said that he’s doing what was referred to as
“voting Labour”.
Virtually all accounts of open quotation have agreed on the use line, whereas the
discrepancies lie in the analysis of the mention line, especially the relation between
the quoted material and the subject of the sentence. The default interpretation,
8
Scare quoting may be regarded as a sub-type of reportive quoting, with the extra function of
indicating non-standard uses. See De Brabanter (2010) for detailed arguments.
namely that the subject has uttered the quoted material, has been variously construed
as part of the semantic content (Cappelen and Lepore 1997, among others); a
presupposition (Geurts and Maier 2005); a conventional implicature (Potts 2007);
or a conversational implicature (Recanati 2010). Such variations are tangential to
our purpose, as we will later on take issue with the use line. To sum up, according
to the received wisdom, open quotation contributes to truth conditions by supplying
not only the quoted material (qua being mentioned), but also its original intended
referent (thus being used).9 Therefore, as compared to closed quotation, open
quotation constitutes a more interesting environment to examine the behaviour of
first-person indexicals under quotation, which will have important repercussions for
the fixity thesis.
Before addressing the connection between Fixity and quotation, especially mixed
quotation, we briefly comment on the other sub-type of open quotation which takes
the form of free-standing reported speech or thought, as in (11):
(11) Stop that John! “Nobody likes me”. “I am miserable”. Don’t you think you
exaggerate a bit? (Recanati 2000: 182)
In (11), like in (10), the quoted material does not occupy the syntactic position
of a singular term. To repeat, both mixed quotation and free-standing quotation
would count as open quotation.10 Recanati (2000) assesses the properties of mixed
quotation against the background of the open/closed distinction, where in open
quotation the quoted material is only ‘demonstrated’, as in (11), while in closed
quotation the quoted material is ‘demonstrated’ as well as ‘referred to’ as in (12).
(12) John keeps crying and saying “Nobody likes me”. (Recanati 2000: 182)
Recanati then develops a sophisticated solution whereby mixed quotation is a sub-
species of open quotation in that the material itself is not referred to, but at the same
time mixed quotation is a pragmatic phenomenon in that it pertains to the use of
words that are specific for the context at hand and, more importantly, it involves
a pretence of a context-shift by introducing a perspective in which the first-person
9
The original intended referent is usually the one intended by the quotee, which, crucially, need
not be the standard semantic referent. For example, when a name appears in quote and it is clear
from the context that the quotee has mistaken the bearer of the name with someone else, the quoted
name would refer to what the quotee intended to refer to, instead of its standard referent (Recanati
2001: 668). Nonetheless, if the quotee is in turn echoing someone else, as in the following example
from Geurts and Maier (2005: 122), the original intended referent is one that is intended by the
speaker from whom the quoted material originates.
(*) George called the ‘POlice’. : : : Tony called the ‘POlice’, too.
There is a reading of (*) in which Tony echoes George, in which case both quoted occurrences
of “POlice” mean “whatever it is [that] George referred to using the expression ‘POlice’” (ibid.
123).
10
“A quotation which is not closed is (as one might expect) open” (Recanati 2010: 402).
indexical originates (see Recanati 2001: 678). In other words, quotation facilitates
the “pre-semantic process of context-selection” (Recanati 2001: 680).
A further twist is added by immersing first-person pronouns in quantified
sentences as in (13) and (14).
(13) Every child grabbed a toy and called it “mine”.
(14) “Nobody likes me”. “I am miserable” : : : I hate people who complain about
their lives.
Context proves to be more complicated to pinpoint in these cases. Arguably, (13) and
(14) can also be analysed as examples of quantification over contexts. But all that
matters here is that, as the story continues, open quotation constitutes an instance
of a phenomenon of pragmatic intrusion into the truth-conditional content, familiar
from the post-Gricean contextualist literature. It is so because “in quotation, what
we demonstrate is a piece of verbal behaviour – a way of speaking” (Recanati 2001:
640).
However, instead of introducing a new category, one can easily subsume free-
standing quotation as a species of mixed quotation, on the grounds that the quoted
material, albeit a (string of) sentence(s) separated from the main clause by full stops,
does not by itself constitute a complete thought from the perspective of the speaker
(i.e. quoter). In (11) for example, the quoted sentences form part of the quotational
thought which expresses the speaker’s disdain over John’s self-pitying behaviour
(verbal or otherwise). Note that the quoted material is recognised as quotation only
when one comprehends the speaker’s comment, which indicates that, sentential
boundaries notwithstanding, the speaker’s quotational thought is conveyed by a
discourse unit which encompasses the quoted sentences and the rhetorical question.
Now, if we identify the quotational thought as the unit of semantic (i.e. truth-
conditional) analysis, as in dynamic approaches such as DRT (Kamp and Reyle
1993) or Default Semantics (Jaszczolt 2005), free-standing quotation would count
as mixed quotation, as (i) it forms part of the discourse unit that bears the truth
conditions of a quotational thought, and (ii) it does not serve as a singular term in
the discourse unit: its contribution to the truth conditions is twofold, pertaining to
both the mention and the use of the quoted material.
The reanalysis of free-standing quotation as mixed quotation is further supported
by the fact that one can easily conceptualise the former as the latter, allowing for
the quoted phrases to appear in appositive positions and as such, arguably, in a more
closely-knit structure, as in (110 ).11
(110 ) Stop that John – “nobody likes me, I am miserable” – don’t you think you
exaggerate a bit?
In addition, it is not uncommon for a quoted sentence to figure in a mixed quotation,
as in (15).
11
Strict syntactic arguments do not seem independently justified: the fact that the quoted material
need not be present for the sentence to be grammatical seems tangential to the discussion of the
conceptual content.
(15) Context: a professor complains about the numerous emails he receives from
students that are too informal and sometimes inappropriate.
The tone that the students would take in e-mails was pretty astounding, “I
need to know this and you need to tell me right now,” with a familiarity that
can sometimes border on imperative. (J.D. Glater, ‘To:
Professor@University.edu Subject: Why It’s All About Me’, New York
Times, 21 February 2006).
The ease with which free-standing quotation can be conceptualised as syntactically
mixed quotation indicates that the former is not (so) free-standing after all as far as
quotational thought is concerned, but only forms part of the quotational thought
which is taken here as a unit of truth-conditional analysis. Therefore, in what
follows we will retain the closed/mixed distinction, and treat cases of free-standing
quotation as mixed quotation.
Needless to say, an utterance with mixed quotation ‘belongs to the speaker’, so
to speak: it is the speaker who takes responsibility for the expressed proposition,
no matter what happens inside it with the attributions of contributing expressions.
The uttered sentence is used rather than mentioned, and, a fortiori, its elements
that partake in the composition of meaning must partake in the ‘use’ function, no
matter what they do in addition to this function. This common-sense observation
will thus equip us with a springboard for the argument in favour of consider-
ing mixed quotation as a test case for Fixity and a fortiori for the need for
monsters.
2.2 Monsters and Quotation: A Double Twist
Our argument for employing mixed quotation as a test case for Fixity relied on the
diversity of uses to which the first-person pronoun can be put there: since there
are different types of uses, and they all have different truth-conditional import,
mixed quotation cannot be a case of mention. In a sense, this goes in the opposite
direction to the reasoning used by Kaplan. At first sight, quotation appears to be a
counterexample to Fixity, as it shifts the interpretation of an indexical in its scope
away from the context of utterance. However, to repeat, Kaplan (1989a: 510–511)
argues that quotation is not a monster, for a quoted indexical is being mentioned,
rather than used:
There is a way to control an indexical, to keep it from taking primary scope, and even to
refer it to another context (this amounts to changing its character). Use quotation marks. If
we mention the indexical rather than use it, we can, of course, operate directly on it. ( : : : )
Operators like ‘In some contexts it is true that’, which attempt to meddle with character, I
call monsters. I claim that none can be expressed in English (without sneaking in a quotation
device). If they stay in the metalanguage and confine their attention to sentences as in In
some contexts ‘I am not tired now’ is true they are rendered harmless and even do socially
useful work ( : : : ).
That is, to mention an indexical, typically with the help of quotation marks, is to
“operate directly on it” by referring to the indexical expression. The reference can
be made to an indexical type, such as (16), or to an indexical token, such as (6),
repeated here.
(16) Both English ‘I’ and Amharic ‘I’ are indexical expressions (they have, if you
will, a feature ‘Ccontextual’). But in addition, English ‘I’ is specified as
‘Cactual speech act’, while Amharic ‘I’ is simply underspecified for ˙actual
speech act. (Adapted from Schlenker 2003: 74)
(6) Miles said, “I am a perfectionist”.
But it is the very diversity of ways in which ‘I’ is used – specifically in mixed
quotation, but also to some extent in quotation at large – that led us to conclude that
mixed quotation, as a case of language use, would be a good testing ground for the
semantic properties of ‘I’ and a fortiori for Kaplan’s views on indexicals, quotation,
and monsters.
To start with, consider (15) again. The quotation therein serves to display the
students’ self-centred attitudes explicitly or implicitly conveyed in their emails,
rather than referring to any of the words they used in the emails. As the truth of
(15) does not require the quoted material to be mentioned at all, the quoted material
contributes to the intuitive truth-conditional content qua being used to demonstrate
a type of de se attitude (Lewis 1979) such that, roughly, whoever is the subject of
the attitude needs to know such-and-such from one’s addressee immediately. Since
the demonstration hinges on the standard meaning of the quoted material (more
below), i.e. it is used by the speaker (quoter) to simulate the students, in this sense,
the quotation counts as a monstrous operation.
The inadequacy of the mention-based analysis is writ large when ‘I’ appears in
generalized closed quotation, as in (17) and (18).
(17) Daniel said “I am sorry”, and so did everyone else in the shop.
(18) How many times have you heard someone say this: “If I had known then
what I know now, I would never have (fill in this blank with a great story).”
As a rule, the story filling in the blank will be about some unexpected
success [sic], or some particularly spectacular failure, or just an interesting
narrative. [H. Doss, ‘Too much caution means not enough risk’, Forbes, 3
October 2013]
Intuitively, (17) doesn’t necessarily say that (i) everyone in the shop uttered the
words “I am sorry”, but is more likely to convey that (ii) Daniel did, and the rest
of them apologised in some form or another. Yet (ii) would be unaccounted for
were the truth of (17) only to require the relevant individuals to have uttered the
words. Similarly, the main point of the quotation in (18) is not to refer to the exact
wording, but to depict someone’s naive state of mind while deciding to embark on
what later unfolded as a great story in his or her life. On this score, treating closed
quotation as meta-linguistic mention, as Kaplan did, would leave certain intuitive
truth-conditional import of the quotation unexplained.
Needless to say, the argument extends to mixed quotation in that mixed quotation
is by its very nature further removed from the category of ‘mentioning’. (19) is an
example in point.
(19) Levi Foster, in fact, is the great-great-grandfather of Gov. Mike Foster of
Louisiana, who said recently on a radio program that it would be “news to
me” if anyone in his family had owned slaves. (From Recanati 2008: 3)
We discuss this example in more detail in the next section, where we propose an
analysis of mixed quotation in terms of ‘reference to character’.
3 The Taming of the Monster
3.1 Character-at-Issue and Content-at-Issue
It appears from our discussion so far that although at first glance the use of indexicals
in quotation is governed by different principles in that quotation marks signal at least
some degree of immunity to semantic assessment of the material enclosed by them
(see also Kaplan 1989a: 510) that varies with a theory of quotation adopted (see e.g.
Cappelen and Lepore 2005), there is no doubt that mixed quotation is an instance
of language use and the material in quotes contributes to the compositionally
derived proposition. In (19) above, the quoted material contributes, albeit not in
a straightforward manner, to the compositionally derived meaning of the sentence.
We attend to this variety of ways in which ‘I’ can be used as the discussion develops
in this section. We will argue that the answer to the question as to how exactly the
quoted material participates in the composition process is much simpler than the
often unnecessarily convoluted discussions in the literature make us believe. It is so
because the contextualist outlook adopted in these solutions has not been pursued to
its logical end.
The examples we have discussed so far, or will discuss at the following stages
of the argument, show that ‘I’ appears in quotation in a variety of ways. They also
show that, since mixed quotation is a case of language use, mixed quotation qualifies
as a potential monster context: it is a case of language use, and yet the referent of
‘I’ is shifted. Now, the interesting point is that monsters can be tamed, which can
be achieved in two ways: one can either demonstrate that (i) monster operators
are not a uniform phenomenon and, arguably a fortiori, they are not monster
operators at all but rather aspects of ordinary linguistic behaviour; or (ii) they are
a uniform phenomenon but only in virtue of belonging to a ‘super-phenomenon’ of
something we could call radical pragmatic malleability, called in Jaszczolt (2012)
the phenomenon of ‘fluid characters’.
Predelli (2014) goes part-way towards (i). He points out that Kaplan does not
clearly differentiate between three different cases of ‘monstrous behaviour’. There
are in fact three kinds of monsters: (a) ‘context shifters’, (b) ‘global shifters’, and
(c) ‘character shifters’. Types (a) and (b) are exemplified in two different readings
of (20).
(20) Never put off until tomorrow what you can do today. (Predelli 2014: 390)
A context-shifting interpretation of (20) amounts to proscribing putting off any
action to any day following the day of the utterance, while the globally-shifting
interpretation proscribes putting off any action in any situation, factual or counter-
factual. In other words, it generalizes the speech act to all possible worlds. Finally,
type (c), the character-shifting behaviour, is exemplified in (21).
(21) If ‘leg’ meant what ‘tail’ means, how many legs would a horse have?
(Predelli 2014: 392)
In view of this abundance of ways of manipulating meaning in context, the question
arises whether the term ‘monster operator’ is justified. Perhaps what we have here
are three examples of perfectly accountable departures from the most typical use of
natural language expressions: indexicals in (20) and non-indexicals whose character
assumes the properties of indexicals in (21)? Viewed in this light, Kaplan’s Fixity
can arguably be stripped of the status of a semantic principle.
Analogously, as one of the authors suggested elsewhere (Jaszczolt 2013a, b,
2016), the indexical/nonindexical distinction with the associated different properties
of characters on each side would be difficult to adopt as a strict semantic principle.
Linguistic expressions assume the characteristics of different categories. To repeat,
the English ‘I’ can behave as a ‘fake indexical’ in admitting non-indexical uses
(Kratzer 2009). On the other hand, non-indexicals can assume the properties of
first-person pronoun when specifically used to perform this role as for example in
child-directed speech.12 Finally, in many languages personal pronouns that ought
to be the typical realisation of the category of indexicals do not constitute a
morphosyntactic category separate from common nouns that ought to count as a
typical realisation of non-indexicals. In many South-East Asian languages, the first-
person marker has the characteristics of both a pronoun and a noun. Like nouns,
they do not form a closed class, and like nouns, they form the plural by adding a
plural morpheme. Typically, expressions that translate directly as ‘slave’, ‘servant’,
‘royal slave’, ‘lord’s servant’, ‘Buddha’s servant’, and so forth, are used for self-
reference, conveying in addition the concept of honorification, that is the sense
self-denigration in front of a person of a higher social status, but without conveying
the semantic content of the content words.13 Needless to say, we remain here on
the level of expression types rather than tokens and pursue this observation on this
level: Kaplan’s first-person indexical does not correspond to an expression type but
rather to a (fairly standard) use of an expression type.
This leads us to solution (ii). Mixed quotation is one of the environments where
this freedom of function-guided meaning goes in yet another direction: in (11),
12
See Collins and Postal 2012 on the so-called ‘imposters’.
13
As a result, these languages abound in first-person markers. For Thai, 27 such forms have been
identified (Siewierska 2004: 228; Heine and Song 2011), and for Japanese, including archaic forms,
51 (Tanaka 2012 and p.c.). See Jaszczolt (2013a, 2016) for a discussion.
‘Nobody likes me’ and ‘I am miserable’ not only dictate the apparent shift of
context but also, or even predominantly, signal that it is the character itself that
is emphasised for the purpose of the discourse at hand. It is not the referent of ‘I’
or ‘me’ that matters, and therefore it will not suffice to point out that it is John
rather than the speaker who considers himself miserable and unpopular. Rather,
what matters is the fact that someone uttered (or otherwise conveyed) the thoughts
summed up in these two sentences.14 In other words, it is the character itself that
is foregrounded in preference to the content, in that it is the fact that the addressee
represents a specific attitude to himself that is in focus. After Jaszczolt (2015), We
will call such use ‘character-at-issue’ use.
Is character-at-issue use atypical? It does not appear to be so. Saka (1998: 24)
gives three examples of quotational uses of ‘I’ where the first two pick up a token
(the speaker in (22) and the addressee in (23)), and the third one a type as in (24).
(22) “I” refers to me.

(23) “I”, as said by you, refers to you.
(24) “I” does not refer to anyone in particular; only tokens of it do.
(24) is an example of a pure character-at-issue use. (25)–(26) are typical (and
attested) cases of character-at-issue uses, in which the quoted first-person pronouns
serve to depict certain types of de se thought, in virtue of the character being
referred to.
(25) As those feelings of exclusion persist, though, the behaviors may become
even more detrimental to the organization – lying on expense forms, drinking
on the job. A sense of “my boss has mistreated me, coworkers have
mistreated me, so I’m going to get them back,” may prevail. These notions,
obviously, can lead to serious negative impacts for organizations. [L.
Grensing-Pophal, ‘Beware the excluded employee’, Human Resource
Executive Online, 14 October 2014]
(26) What makes an “I-would-never-have” story so poignant is the knowledge
that something has been lost. Just behind the narrative there is more than a
touch of nostalgia and yearning for that time when “I didn’t know,” a return
to times that were somehow more immediate, more fulfilling and perhaps
more productive. [H. Doss,‘Too much caution means not enough risk’,
Forbes, 3 October 2013]15
14
‘Otherwise conveyed’ because the material in quotes need not directly depict the speaker’s
words; it can represent types or tokens, concepts or forms (see e.g. Saka 1998 on metalinguistic
use and Wilson 2000 on metarepresenting).
15
Besides semantically foregrounding the character of the quoted material, these examples are
complex in another way, namely that the quotations in them function syntactically as modifiers, as
pointed out by a reviewer of this volume (see De Brabanter 2005; de Vries 2008; Pascual 2014 for
discussions). The fact that they are recruited to demonstrate a type of de se attitude also challenges
the traditional view of ‘mixed quotation’ in terms of deferred reference (see §2.1).
(25) and (26) are characteristic of character-at-issue uses in that the point of the
quotation is not to refer to the words of the current speaker, or to defer their
interpretations to the original speaker, for there need not be any, but rather to depict
someone else’s de se attitude by simulating it. In this respect, putting the first-person
pronoun ‘I’ in quotation seems to be the most effective way to convey the simulated
thought.
Moreover, when it is clear from the context, character-at-issue uses need not be
indicated by quotation marks. (27), for instance, is the title of an article in the New
York Times Education session which reports how university students misused emails
to bombard their professors with unreasonable requests. The title is ingeniously
framed from the students’ perspective, with the use of “Why it’s all about me”
serving to depict the students’ self-centred attitudes. In a similar vein, the unquoted
first-person pronouns in the compound adjectives in (28) vividly refer to the kinds
of de se attitude a woman may find herself in in the face of certain perhaps unwise
purchases.
(27) To: Professor@University.edu Subject: Why It’s All About Me [title of piece
by J.D. Glater, New York Times, 21 February 2006]
(28) Context: a list in a Wall Street Journal illustration entitled “The gallery of
fashion (t)errors: A look at some unworn and unloved monsters that lurk
inside many a woman’s closet”
The Is-That-Squeaking-Noise-Me leather pants.
The Where-Do-I-Wear-This cocktail dress.
The I-Thought-It-Was-My-Style biker jacket.
The Oops-I-Forgot-How-Old-I-Was romper.
The So-Avant-Garde-I-Need-An-Art-Critic-To-Understand-It Skirt [M. Mistry, ‘Five
ways to avoid shopping mistakes’, The Wall Street Journal, 11 July 2014, illustration by
Graham Roumieu]
‘Nobody likes me’ and ‘I am miserable’ in (11) fall in the same category of
character-at-issue uses; they are just slightly more complicated ones. They are
complicated in that although at first blush they constitute standardised expressions
of self-pity16 and as such generalise beyond the discourse-specific referent, they are
also accompanied by what we can call ‘content-at-background’ use: in addition to
the fact that the speaker is emphasising a particular type of self-directed thought,
the speaker is also directly demonstrating that it is the addressee (John) who has this
attitude. This latter, content-directed use is emphasised in the follow-up (‘Don’t you
think you exaggerate a bit?’) and at this point in discourse it acquires the ‘content-at-
issue status’: now it is John’s self-pity that is in focus, while the form of expressing
it is backgrounded. In the light of the discussion of Corazza’s conventionalism and
Predelli’s intentionalism pursued in Sect. 1, what we have here is a conventional
character-at-issue interpretation, followed by a content-at-issue interpretation that
16
‘Standardised’ but not ‘conventionalised’: they are still compositional, albeit have a slightly
limited syntactic penetrability, see Bach (1995).
is intention-driven and takes place later in the discourse.17 It is only when we look
at examples such as (11) in this dynamic manner, following the process of discourse
interpretation, that the correct analysis and classification of the uses of ‘I’ emerge.
To sum up, at this point of our argument we have (a) noted that mixed quotation
is a case of language use, and on these grounds (b) questioned Kaplan’s exclusion
of quotation from potential monster contexts. At the same time, we have (c)
demonstrated the diversified behaviour of ‘I’ in such contexts which calls into
question both (d) the concept of a Kaplanian indexical with invariant semantic
properties, and a fortiori (e) the concept of a monster context associated with it.
3.2 Fluid Characters
Now, the conventional (or, on Bach’s terminology standardised) status18 of the

sentences quoted in the context of (11) points to yet another complication in their
classification as character-at-issue cases. The sentences are not simply composed
of characters pertaining to words; instead, arguably, just as it is the case in other
formulaic expressions, the expression type itself affects what counts as a character.
Jaszczolt (2012) presents a proposal of a so-called ‘fluid character’ that is construed
as discourse unveils and pertains to an expression whose length (from one- to
multi-word) is established in the context at hand: sometimes it takes more than a
word to create a character. Idioms will obviously fall in this category: since the
main intended content does not match the compositionally derived content, it is
more auspicious to adopt a character that pertains to the former. But multi-word
characters generalise beyond fossilised formulae: characters can be retrieved on
the basis of shorter or longer chunks of discourse – or, as it was called there, on
the basis of a flexible inferential base, a base of varying length.19 Applying this
analysis to ‘Nobody likes me’ uttered in the context of (11), and as such undeniably
with appropriate intonation, and thus adopting the assumption that characters can be
flexible, we reach the conclusion that it is the entire expression itself that constitutes
such a character. Arguably, the formulaic nature of ‘news to me’ in (19) is relevant
for character-construction in the analogous manner.
The need to assign Kaplanian characters to a flexible base which goes beyond
standard lexical indexicals is further supported by compound adjectives contain-
ing first-person pronouns, used to depict certain types of de se attitudes, as in
(29) and (30):
17
See Sect. 4 for a contextualist semantic representation of this mini-discourse that reflects this
order of interpretation.
18
See fn. 16.
19
Still fulfilling the criterion of being types rather than tokens: they pertain to meanings of
expressions that can undergo a context-free classification.
(29) Take Herman Cain – pre-sexual harassment accusations. His self-titled book,
This is Herman Cain! My Journey to the White House, is an I-am-who-I-am
autobiography that begins as straightforward as it gets with: “My name is
Herman Cain.” [E. Lelon, ‘The fraud of authentic leadership’, The
Huffington Post, 11 September 2011]
(30) Studies by the University of Michigan show that when health care providers
are allowed to say “I’m sorry”, patients and their families are able to let go of
their anger and move on sooner. In fact, the University of Michigan Health
System estimates the average cost of lawsuits has been cut in half since it
adopted an “I’m Sorry” policy nearly a decade ago. [Office of the Governor
of Michigan Rick Snyder, ‘Snyder signs bill allowing doctors to say “I’m
sorry” to patients, families’, News release, 19 April 2011]20
In (29), the compound ‘I-am-who-I-am’ is used to convey a type of candid de se
attitude, whose meaning is partly compositional and partly formulaic, in that by
presenting oneself as one really is, one also projects a sense of candidness. Similarly,
‘an “I’m Sorry” policy’ corresponds to a policy of sincere apology, derived from the
formulaic meaning of “I’m sorry”.
All in all, it appears that evidence is mounting that the linguistic behaviour does
not support the standard view that, to invoke Davidson (1979: 81), quotation “has a
certain anomalous feature”. Quotation targets what is to be emphasised: content-
at-issue, character-at-issue, or both, as in (11) (see also Gomez-Torrente 2017).
Or, in Saka’s (2005: 208) words, quotation marks signal that “there is reference
to something other than what is customary”. It is not a monster operator not because
it ought to be excluded from consideration as Kaplan suggests, but because, on
our solution (ii), it is a perfectly natural pragmatic device superimposed on what
is already pragmatics-rich meaning of the linguistic material encased by quotation
marks. It is a case of language use – after all, as we have shown, there are so many
different kinds of the employment of ‘I’ (see also Johnson 2017; Maier 2017: sec.
3.2).21 On the flip side, arguably, indexicals do not necessitate the postulation of
monster contexts because the category of indexical expressions itself is a theoretical
concept with no clear linguistic correlates. The lack of conformity with the strict
indexical/nonindexical distinction is, as we have indicated, normal language use.
We therefore take the liberty of exorcising the concept of a ‘monster operator’ in the
interest of being open-minded (and empirically-minded) concerning what language
can and cannot do.
This is where we are with the argument. Kaplan points out that it is the
metalinguistic status of quotation that excludes it from the category of potential
monsters. We have shown that the sheer variety of types of uses of the first-person
20
Available at http://www.michigan.gov/snyder/0,4668,7-277-61409_61412_61527-254556--,
00.html. Accessed 8 August 2016.
21
We shall not be discussing here non-linguistic material in the scope of quotation marks. See e.g.
Cappelen and Lepore (2007) for a uniform minimalist account.
pronoun renders the umbrella term ‘metalinguistic’ inapplicable. What follows is

that quotation is capable of serving as a legitimate test case for Fixity. But nothing
is as it seems: it is true that quotation can be a case of language use, as the example
of mixed quotation demonstrates, and as such arguably can fall within the category
of Kaplan’s potential monsters, but at the same time since ‘I’ does not seem to
behave like Kaplan’s indexical ought to, the raison d’être for monsters is put into
question. But were ‘I’ a well-behaving indexical, quotation would be a fine monster
indeed. What remains is the representation of this diversity of uses in a suitably
contextualist theory of meaning.
4 The Need for Radical Contextualism
The examples below provide a selection of uses of the first-person pronoun in

mixed quotation that pertain to subtly differentiated referents: the concept of the
self in (31), the word in (32), a character-at-issue supplemented with the quantifier-
bound content in (13) repeated below, and, as previously discussed, standardised
and conventional expressions pertaining to entire propositions (in virtue of fluid
characters) in (11) repeated below and in (33) respectively. Listed in this way, they
all represent variations on the reference to a character – or, as we called it earlier,
character-at-issue uses.
(31) The concept of ‘I’ has exercised some of the greatest minds including Locke
and Hume.
(32) You have to add ‘I’ in lines 5 and 13 on page 7.
(13) Every child grabbed a toy and called it ‘mine’.
(11) Stop that John! “Nobody likes me”. “I am miserable”. Don’t you think you
exaggerate a bit?
(33) “I am sorry” is called for in this situation.
Reference to a character is not pure: this follows from the fact discussed in
Sect. 3.1 that ‘I’ is not a pure indexical but rather an imperfect natural-language
instantiation of what appears to be a philosophers’ fiction: an indexical expression.
To repeat, since indexical expressions also have non-indexical uses, and on the other
hand non-indexical expressions are employed in the function of pure indexicals,
the match between philosophers’ fiction and language use is imperfect, with the
degree of this imperfection varying from language to language, as we pointed
out in the brief discussion of the morphosyntactic properties of honorifics in
Sect. 3.1. Now, in order to capture the fact that the first-person singular pronoun
can be used to refer to (a) character; (b) character with additional conceptual
information; (c) content (not necessarily the speaker, as Saka’s examples (22) and
(23) discussed above demonstrate) – or even (d) to participate in a fluid character
formed with respect to the inferential bases of different length as in (11) – we need
a contextualist perspective pushed to its logical limits. What we need is a theory
in which the composition of meaning allows for maximal possible (and maximal
empirically demonstrable) pragmatic contributions to the truth-conditional content.
This can be achieved when the truth-conditional representation is not restricted
to embellishments of the logical form of the sentence but instead is guided by
the primary meaning of the produced speech act. Default Semantics offers such
a maximally contextualist account whereby what is represented is the main message
of the speaker irrespective of the manner in which it was conveyed: directly through
full articulation, directly through partial articulation, or indirectly. A logically
maximally radical contextualist account has to capture all these options.
So, let us take stock. The diversity of uses to which first-person indexicals
can be put in quotation requires a theory of meaning that would capture these
differences. This account will have to share some general assumptions with use
theories of quotation (e.g. Saka 1998; Recanati 2000, 2001) and subscribe to the
contextualist outlook on meaning whereby context has an important part to play in
supplying the unit of the truth-conditional analysis and as such in supplying truth-
conditional aspects of the meaning of an utterance.22 It will (i) interpret quotation
as representing various kinds of objects: types, tokens, or signs – or, on Wilson’s
(2000) account, public, mental, or abstract representations – and (ii) subscribe to
the view that direct quotation “must merely resemble the original to some degree”
(Wilson 2000), simply drawing attention to some formal features of the quoted
material.23 Note that ‘Nobody likes me’ in (11) licenses a use of quotation marks
even if it merely summarises the addressee’s general behaviour or a long tirade of
self-pitying statements.
However, it seems to us that most contextualist accounts of quotation do not go
far enough in that they do not take the idea of pragmatic intrusion into the truth-
conditional content to its logical conclusion. For Recanati, mixed quotation is a
type of ‘pragmatic intrusion’ into the truth-conditional content. To repeat, it is so
because “in quotation, what we demonstrate is a piece of verbal behaviour – a way of
speaking” (Recanati 2001: 640). Next, in order to execute this contribution to truth
conditions, one has to have a way of discerning what material is to be contributed.
Here Recanati points out that it has to be resolved what type is instantiated by the
quoted token. In simpler words, the speaker’s what is said has to be arrived at in
the process of interpretation that involves recognizing the speaker’s intentions (see
Recanati 2001: 641–2). Quotation marks function here as ‘pragmatic indicators’
(p. 663). Saka (e.g. 1998) and Wilson (e.g. 2000) focus instead on the status and
role that the material in quotes has in the utterance – in other words, they focus,
in their different ways, on the type of metarepresentation. We have added to it the
flexibility of what counts as character (the fluid character) and the content-at-issue
vs. the character-at-issue distinction that cuts across the overtly quantificational
and non-quantificational uses. What remains is adding a template for building a
representation of the main meaning intended in the use of sentences with first-
22
In this context see also Capone (2013) on quotation as social practice and a language game.
23
Cf. also Saka’s (1998) quotation marks as subscripts, indicating metalinguistic use.
person indexicals in mixed quotation. To accomplish this task, we will have to

go significantly beyond Recanati’s proposal based on the recognition of speaker’s
intentions. A post-Gricean theory of Default Semantics (henceforth DS, Jaszczolt,
e.g. 2005, 2010, 2016) offers such a template. In what follows we briefly summarise
the theory and then apply it to some of our earlier examples, focusing on (11) in
view of what we had said above about its being doubly problematic.
In order to account for the diversity of ways of going beyond a simple context
shift in the case of the first-person indexical, one ought to employ a contextualist
approach that will allocate the selection of the kind of representation that the
quotation marks indicate to the relevant source of information and the relevant
pragmatic process. DS postulates that compositionality of meaning ought to be
predicated not of natural language sentences but rather of speech acts in that this
is the level at which aspects of meaning coming from different sources are put
together. It is dubbed interactive compositionality. Compositionality is understood
here as a methodological assumption and as such is functionally defined. The idea of
interactive, or ‘Gestaltist’, compositionality is not new. It is adapted from Recanati
(2004) but is further modified to apply to the level of the merger of information.
Semantic representations of the primary content of such speech acts are called
merger representations. Since the primary content can have any relation to the
logical form of the uttered sentence, from complete reliance, through partial reliance
(when pragmatic processes intrude in the logical form contributing to the truth-
conditional content), to independence (when the primary message is implicated –
conveyed indirectly), merger representations do not constitute developments of
logical forms as it is the case with the Relevance-theoretic construct of an
explicature (Sperber and Wilson 1995, 2012) or Recanati’s (1989, 2010) construct
of what is said. Instead, they represent the primary intended message tout court: the
provenance of the representation can be traced to any of the contributing sources
of information, without giving preference to the logical form of the sentence.
In this respect, the perspective on compositionality resembles that of cognitive
semantics where the compositionality of language derives from the compositionality
of thought as presented for example in Embodied Construction Grammar (e.g.
Feldman 2010). Where cognitive semantics and DS differ is in the adoption by the
latter of the truth-conditional method as well as an assumed algorithmic nature of
the meaning merger.
The following sources of meaning are identified in DS: word meaning and
sentence structure (WS) – the level at which compositionality is standardly sought;
world knowledge (WK); situation of discourse (SD); stereotypes and presumptions
about society and culture (SC); and properties of the human inferential system (IS).
These in turn trigger different kinds of processes in the overall interpretation of
utterance meaning. The source-process mapping is not bi-unique but it is robust.
WS pertains to the recovery of linguistic meaning – of words, sentence fragments,
and on occasion whole sentences (when uninterrupted by other processes); SC and
IS trigger automatic, subdoxastic interpretations called respectively social, cultural
and world knowledge defaults (SCWD) and cognitive defaults (CD); SD triggers
pragmatic inference – dubbed ‘conscious pragmatic inference’ (CPI) in order to
contrast it with the automatic processes of SCWD and CD; finally WK can trigger
either SCWD or CPI in that it can be utilised subdoxastically or in an inferential
process. A merger representation is the output of these processes, or a subset of
them, with the assumption that for this account to have suitable predictive power
the interaction is to be algorithmically specified. A more detailed discussion of the
sources and processes identified in DS can be found elsewhere (e.g. Jaszczolt 2005,
2009, 2010).
Merger representations index the chunk of information with types of processes.
That is, material enclosed in square brackets is indexed with a subscript standing
for the type of processing. The representations use a revised and extended language
of DRT (Kamp and Reyle 1993), but apply it with the DS-theoretic objective of
modelling speech acts and interactive compositionality predicated of them. This
way of modelling the diversified uses of ‘I’ in mixed quotation allows a better insight
into their properties in that it foregrounds their different provenance and different
routes through which they appear in the composition process. We turn next to some
examples of such an application.
Simple referential and quantificational uses of ‘I’ can be processed directly
through WS, and so can (31) in that information that ‘I’ refers to the concept is
explicitly present in the sentence. The expression ‘news to me’ in (19) can also
be classified here, albeit with a trace of character-at-issue attained through the
pretended context-shift. As such, it is obtained via WS as well as SCWD: ‘news
to me’ is a standardised construction and its use instead of the obvious alternative
‘news to him’ conveys this character-induced focus. In other words, ‘news to me’
conveys more emphasis on the personal perspective and the intention to express
detachment. The character-at-issue is thus wheeled in, in support of, and for the
emphasis of, the content-at-issue. Quantificational use in (13) works analogously.
Next, in (32) SCWD captures the fact that the context of copy-editing triggers
by default an interpretation on which ‘I’ refers to a sign. The corresponding source
of information is that of SC. The representation thus makes use of WS as well as
SCWD.24
Finally, we have come to the fluid character in (11). Here the SCWD operates on
the entire clauses ‘Nobody likes me’ and ‘I am miserable’: [nobody likes me]SCWD ,
but at the same time we have a compositionally obtained meaning pertaining to
the ‘content-at-issue’: [nobody likes x]WS and John (x). As a result, we have the
fluid-character-at issue reading of ‘Nobody likes me’, followed by, as proposed in
Sect. 3.1, the ‘content-at-issue’ primary meaning when the interpretation process
reaches the clause ‘Don’t you think you exaggerate a bit?’ – reflecting the dynamic
nature of the composition of discourse meaning. To compare, example (33),
24
It is possible to envisage a scenario on which an addressee has to consciously process information
that the situation is that of copy-editing. On such a scenario, instead of WSCSCWD we have
WSCCPI. It is the fact of discourse that the representations will vary in this way and all that
the algorithmic account can capture is the options for arriving at an interpretation – perhaps ranked
options when the preceding context points in the direction of one or the other process. See Jaszczolt
(2010) for a discussion and fn. 25 below.
xyz
[John (x)]CD
[DIR stop (x,y)]WS
[I am miserable]SCWD
[nobody likes me]SCWD
[self-pity (y)]CPI
[DIR stop (x,y)]WS,CPI
[exaggerate (x)]WS,CPI,SCWD
Σ’ [[no z: likes x] (person (z))]ws
[ACCΔrn ├ Σ’]WS
Σ’’ [miserable (x)]WS
[ACCΔrn ├ Σ’’]WS
Fig. 1 Merger representation for example (11): ‘Stop that John! “Nobody likes me”. “I am
miserable”. Don’t you think you exaggerate a bit?’
containing a formulaic ‘I am sorry’, gives rise only to [I am sorry]SCWD or arguably

[I am sorry]WS,SCWD , without a discourse referent pertaining to ‘I’.
We present below the merger representation for (11), using the language of DS
and relying on the semantics of the relevant linguistic constructions (such as proper
names, temporality, etc.) developed there. Like its parent theory DRT, DS accounts
for the dynamic nature of discourse and thereby for its incremental processing,
so all three sentences of (11) can be represented in one merger representation. To
repeat, the subscripts following the square brackets stand for the processes involved
in producing the compositional representation of the primary meaning of the speech
act (Fig. 1).
The variables x, y, z stand for discourse referents and the items below stand for
discourse conditions that pertain to the primary meaning of the string of utterances.
The directly referential interpretation of the proper name is given by CD (reflecting
maximizing informational strength, or the maximally strong intentionality of the
corresponding mental state as it is explained in DS), as indicated in [John (x)]CD .
The following condition represents the speech act of a directive (DIR). The recovery
of the literal meaning through WS results in [DIR stop (x,y)]WS , with the variable y
dangling unresolved for the moment (to capture the cataphoric ‘that’ in ‘Stop that
John’). Next, as was discussed above, the character-at-issue meaning of ‘Nobody
likes me’ and ‘I am miserable’ is recovered via SCWD – automatically, in view
of the fact that the sentences contain standardised forms of expression. This is
represented in [I am miserable]SCWD and [nobody likes me]SCWD . At this point,
the dangling y can be resolved through inference (CPI) from these two formulaic
sentences as [self-pity (y)]CPI , resulting in the primary meaning represented in
[DIR stop (x,y)]WS,CPI . To repeat, since the primary meaning is not necessarily
communicated directly and exhaustively, the material in brackets can depart from
the content of the corresponding sentence. Hence, in the following discourse
condition we obtain a simple [exaggerate (x)]WS,CPI,SCWD as the representation of
‘Don’t you think you exaggerate a bit?’ The representation of the literal meaning
would be something on the following lines:
[IM Q DM (exaggerate (x))]WS
Q D y/n
IM D [don’t you think]WS,SCWD
DM D [a bit]WS
where IM stands for illocutionary force modifier, Q for question, and DM for degree
modifier. Instead, the meaning recovered through the combination of the processing
of the sentence (WS), the processing of the preceding context and co-text (CPI), and
the knowledge of the standardised forms of polite criticism (SCWD, in processing
IM and DM) results in the representation of the primary meaning of this utterance
which, arguably, is ‘John is exaggerating’.25
Next, following our analysis of the character-at-issue aided by the content-at-
issue reading of the first-person pronouns in Sect. 3.1, we add the content-at-issue
sense of ‘Nobody likes me’ (using standard generalised quantifiers notation) and
‘I am miserable’, both obtained via a standard processing of the sentences (WS).
Their temporal reference is represented in a standard way developed for representing
temporality in DS (see e.g. Jaszczolt 2009), namely through a modal operator
of acceptability according to the formula ACC ˙, reading ‘it is acceptable
to the degree that it is the case that †’ where † stands for a merger repre-
sentation.26 The degree is assigned here via WS in that the sentence has an
overt grammatical marker of temporality in the form of a grammatical tense (the
superscript ‘rn’ stands for ‘regular present’). We thereby obtain [ACC rn †’]WS
and [ACC rn †”]WS .
25
Note that DS aims at representing the primary meanings as intended by Model Speakers and
recovered by Model Addressees. This is a way of ensuring the predictive power of the theory and
the availability of an algorithmic account. The flip side of this perspective is, of course, that any
change in the content of the sources of information on a given occasion can result in a different
meaning. In other words, while the theory offers a model that demonstrates how pragmatic,
interactive compositionality of meaning works, in the case of made-up examples such as (11)
where there is no access to the real content of the sources, we will never be able to produce one
unique merger representation and claim it is a correct one. In other words, with gaps in information,
we will simply never know what it ought to be.
26
† captures the idea of the summation of information coming from the sources identified in DS,
and a fortiori the idea of interactive compositionality of the meanings of speech acts.
5 Conclusions
As we anticipated in Sect. 1, our overall argument can be summarised as in (a)–(h).
(a) So-called first-person ‘indexicals’ can undergo context shift outside

quotation.
(b) So-called first-person ‘indexicals’ can undergo context shift inside quotation.
(c) There are categorically different non-Kaplanian uses of the first-person
‘indexicals’ in quotation.
(d) Quotation is a case of language use. [from (b) and (c)]
(e) Quotation qualifies as a potential monster operator. [from (d)]
(f) There are different non-Kaplanian uses of ‘indexicals’ outside quotation.
(g) ‘I’ is not an indexical term. [from (c) and (f)]
(h) Quotation is not a monster. [from (g)]
By necessity we paid more attention in this paper to some premises than others. In
particular, we focused on (b) and (c), from which we derived (d) and (e). (a) and
(f) have been extensively researched elsewhere and therefore the support for these
premises was only briefly summarised here (see Jaszczolt 2013a, b, 2016). They
were necessary, however, in order to derive (g) and (h). Needless to say, this is not
the last word on the properties of the first-person ‘indexical’, nor is it a complete
word. But this line of argument allows us to put forward speculative claims in (i)–
(j) – speculative but sufficiently supported by this research to merit positing.
(i) Indexicals are a philosophers’ fiction. [from (a)–(h)]
(j) There are no Kaplanian monsters. [from (i)]
To conclude, Davidson’s (1979: 81) well mulled-over (and, in a sense, ‘doubly-
autonymous’) example (34) set the direction for decades of theorising about
properties of quotation whereby all sides in the debate partook in one big assump-
tion: they all agreed that quotation is indeed a phenomenon that requires an
explanation.
(34) Quine says that quotation “has a certain anomalous feature”.

Proposed solutions adopt different versions of the minimalist ideology or the
contextualist ideology. In this paper we investigated the behaviour of the first-person
indexical in the scope of quotation, asking the question whether mixed quotation
should be included in the category of a monster operator. We adopted a radical
contextualist approach whereby we modelled the primary meaning of the uttered
speech act, assessing the semantic and pragmatic properties of the first-person
pronoun. We pointed out that the range of meanings that are produced is not limited
to the results of context-shift or quantification over contexts but also includes what
we called a character-at-issue reading where it is the character itself that is referred
to. While observations along similar lines have been made in Saka’s use theory and
in Recanati’s contextualist theory, we have gone further. We analysed the interaction
between the character-at-issue and content-at-issue senses, using the principles for
their co-occurrence, and accounted for the process of interpretation of relevant
utterances (and strings of utterances as in (11)) using the framework of Default
Semantics. In addition, we demonstrated that our radical contextualist solution has
the advantage of accounting for situations in which the first-person indexical does
not function as an independent character, but rather constitutes part of a formulaic
expression such as ‘I am sorry’ or a standardised expression such as ‘Nobody likes
me’. The concept of a fluid character and a flexible base for drawing pragmatic
inferences account for the limited syntactic penetrability of such collocations.
The results shed light on two big topics in semantic theory: the properties of
indexical expressions on the one hand, and the properties of quotation on the other. It
is our belief that the behaviour of I in mixed quotation significantly discredits (i) the
status of mixed quotation, if not quotation in general, as a case of language mention,
and (ii) the indexical status of first-person markers, and a fortiori the concept of a
monster operator in general. As such, it contributes informed sceptical judgement
to both lines of research.27
References
Bach, K. (1995). Remark and reply. Standardization vs. conventionalization. Linguistics and
Capone, A. (2013). The pragmatics of quotation, explicatures and modularity of mind. Pragmatics
and Society, 4, 259–284.
Cappelen, H., Lepore, E. (2005). Quotation. In E. Zalta (Ed.), Stanford encyclopedia of philosophy
(Spring 2012 Edition). http://plato.stanford.edu/entries/quotation/. Accessed 8 Aug 2016.
Clements, G. N. (1975). The logophoric pronoun in Ewe: Its role in discourse. Journal of West
African Languages, 2, 141–177.
Collins, C., & Postal, P. M. (2012). Imposters: A study of pronominal agreement. Cambridge, MA:
MIT Press.
Corazza, E. (2004a). On the alleged ambiguity of ‘now’ and ‘here’. Synthese, 138, 289–313.
Corazza, E. (2004b). Reflecting the mind: Indexicality and quasi-indexicality. Oxford: Clarendon
Press.
Davidson, D. (1979). Quotation. Theory and decision, 11, 27–40. Reprinted in: Inquiries into truth
and interpretation (pp. 79–92). Oxford: Oxford University Press.
De Brabanter, P. (2005). The impact of autonymy on the lexicon. WORD, 56, 171–200.
De Brabanter, P. (2010). The semantics and pragmatics of hybrid quotation. Language and
Linguistics Compass, 4, 107–120.
27
We owe thanks to the audiences of the seminar The Semantics and Pragmatics of Quotation
at the American Philosophical Association meeting in Vancouver, the panel The Dynamics of
Self-Expression across Languages at the 14th International Pragmatics Association conference
in Antwerp, the Cambridge Linguistics Forum, and Paul Saka and Alessandro Capone for their
comments on previous drafts of this paper.
De Vries, M. (2008). The representation of language within language: A syntactico-pragmatic

typology of direct speech. Studia Linguistica, 62, 39–77.
Feldman, J. (2010). Embodied language, best-fit analysis, and formal compositionality. Physics of
Life Reviews, 7, 385–410.
Geurts, B., & Maier, E. (2005). Quotation in context. Belgian Journal of Linguistics, 17, 109–128.
Heine, B., & Song, K.-A. (2011). On the grammaticalisation of personal pronouns. Journal of
Jaszczolt, K. M. (2005). Default semantics: Foundations of a compositional theory of acts of
Jaszczolt, K. M. (2009). Representing time: An essay on temporality as modality. Oxford: Oxford
University Press.
Jaszczolt, K. M. (2010). Default Semantics. In B. Heine & H. Narrog (Eds.), The Oxford handbook
of linguistic analysis (pp. 215–246). Oxford: Oxford University Press.
Jaszczolt, K. M. (2012). “Pragmaticising” Kaplan: Flexible inferential bases and fluid characters.
Australian Journal of Linguistics, 32, 209–237.
Jaszczolt, K. M. (2013a). First-person reference in discourse: Aims and strategies. Journal of
Pragmatics, 48, 57–70.
Jaszczolt, K. M. (2013b). Contextualism and minimalism on de se belief ascription. In N. Feit & A.
Capone (Eds.), Attitudes De Se: Linguistics, epistemology, metaphysics (pp. 69–103). Stanford:
CSLI Publications.
Jaszczolt, K.M. (2015). The demise of the first person indexical? Paper presented at the Semantics
and Philosophy in Europe 8, Cambridge.
Jaszczolt, K. M. (2016). Meaning in linguistic interaction: Semantics, metasemantics, philosophy
of language. Oxford: Oxford University Press.
Johnson, M. (2017). Quotation through history: A Historical Case for the Proper Treatment of
Quotation. In P. Saka & M. Johnson (Eds.), The semantics and pragmatics of quotation (pp.
281–302). Cham: Springer.
Kamp, H., & Reyle, U. (1993). From discourse to logic: Introduction to model-theoretic semantics
of natural language, formal logic and discourse representation theory. Dordrecht: Kluwer.
Kaplan, D. (1989a). Demonstratives: An essay on the semantics, logic, metaphysics, and episte-
mology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.),
Themes from Kaplan (pp. 481–563). New York: Oxford University Press.
Kaplan, D. (1989b). Afterthoughts. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from
Kaplan (pp. 565–614). New York: Oxford University Press.
Kratzer, A. (2009). Making a pronoun: Fake indexicals as windows into the properties of pronouns.
Linguistic Inquiry, 40, 187–237.
Lewis, D. (1979). Attitudes de dicto and de se. The Philosophical Review, 88, 513–543.
Maier, E. (2017). The pragmatics of attraction. In P. Saka & M. Johnson (Eds.), The semantics and
Pascual, E. (2014). Fictive interaction: The conversation frame in thought, language, and
discourse. Amsterdam: John Benjamins.
compositionality (pp. 405–431). New York: Oxford University Press.
Predelli, S. (2011). I am still not here now. Erkenntnis, 74, 289–303.
Predelli, S. (2014). Kaplan’s three monsters. Analysis, 74, 389–393.
Recanati, F. (1989). The pragmatics of what is said. Mind and Language, 4. Reprinted in S. Davis
(Ed.), (1991) Pragmatics: A Reader (pp. 97–120). Oxford: Oxford University Press.
Recanati, F. (2000). Oratio obliqua, oratio recta: an essay on metarepresentation. Cambridge,
MA: MIT Press.
Recanati, F. (2004). Literal meaning. Cambridge: Cambridge University Press.
Recanati, F. (2008). Open quotation and context shift. Handout, Moral Sciences Club, University
of Cambridge.
Roberts, C. (2014). Indexicality: de se semantics and pragmatics. Unpublished manuscript,
Draft of April 2015, Ohio State University. http://semanticsarchive.net/Archive/Tg0ZjMyM/
Roberts.indexicality.pdf. Accessed 8 Aug 2016.
Saka, P. (2005). Quotational constructions. In P. De Brabanter (Ed.), Hybrid quotations
(pp. 187–212). Amsterdam: John Benjamins.
Saka, P. (2013). Quotation. Philosophy Compass, 8(10), 935–949.
Schlenker, P. (2003). A plea for monsters. Linguistics and Philosophy, 26, 29–120.
Shan, C. (2010). The character of quotation. Linguistics and Philosophy, 33, 417–443.
Siewierska, A. (2004). Person. Cambridge: Cambridge University Press.
Sperber, D., & Wilson, D. (1995). Relevance: Communication and cognition (2nd ed.). Oxford:
Blackwell.
Sperber, D., & Wilson, D. (2012). Meaning and relevance. Cambridge: Cambridge University
Press.
Tanaka, H. (2012). Scalar implicature in Japanese: Contrastive wa and intersubjectivity. Paper
presented at the First International Conference of the American Pragmatics Association
(AMPRA 1), Charlotte, North Carolina.
van der Sandt, R. (1992). Presupposition projection as anaphora resolution. Journal of Semantics,
9, 333–377.
van der Sandt, R. (2012). Presupposition and accommodation in discourse. In K. Allan & K.
M. Jaszczolt (Eds.), The Cambridge handbook of pragmatics (pp. 329–350). Cambridge:
Cambridge University Press.
Wilson, D. (2000). Metarepresentation in linguistic communication. In D. Sperber (Ed.), Metarep-
resentations: A multidisciplinary perspective (pp. 411–448). Oxford: Oxford University Press.

Saka &amp; Johnson - The Semantics and Pragmatics of Quotation - 2017

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Saka &amp; Johnson - The Semantics and Pragmatics of Quotation - 2017

Uploaded by

Copyright:

Available Formats

Perspectives in Pragmatics, Philosophy & Psychology 15

Editorial Board Members

More information about this series at http://www.springer.com/series/11797

The Semantics and

ISSN 2214-3807 ISSN 2214-3815 (electronic)

Library of Congress Control Number: 2017959052

© Springer International Publishing AG 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

Part I Use & Mention

Part II Quotation Unified

Part III New Directions

Quotation Through History: A Historical Case for the Proper

Abstract I explain a mechanism I call “incorporation,” that I think is at work in a

© Springer International Publishing AG 2017 3

are to be interpreted as if uttered by some other speaker in some other context,

Discussions of “scare quotation” and of “mixed quotation” are vexed by differences

Lizard As 12 million Americans “know,” the United States government is run

2.1 Terminological Digression

in which we have a reporting verb (“said”) and a “that”-clause as its complement,

The second supposition also embodies an idealization, since what is more

3.1 The Two Shifts That Incorporation Involves

David Kaplan’s logic of demonstratives—that I think works nicely. Before getting

4 Intrasentential Change in Context

as a permissible case (where quotation marks flank the part to be interpreted in

4.i/ ˆcw . ^ / iff ˆcw and ˆcw

4.i/ ˆhc1 ;c2 iw . ^ / iff ˆc1 w and ˆc2 w

2: ˆcw ˛1 : : : ˛n iff hj˛1 jcw : : : j˛jcw i 2 I .!/

2 : ˆhc1 ;:::;cn iw ˛1 : : : ˛n iff hj˛1 jc1 w ; : : : ; j˛n jcn wi 2 I .w/

4.2 Too un-Kaplanian?

12 : If c 2 C; thenW jIjcw D cA

5 Intrasentential Shift in Lexicon

6 Applying the Account

With accounts of intra-sentential context- and lexicon-shifting in hand, we are ready

Here we have many scare-quoted expressions. Let’s focus on the sentence

6.7 Range of the Account

Before moving on to comparing the incorporation account of such cases with

7 Comparison with Other Accounts

Several linguists and philosophers have undertaken to explain how scare-quoting

7.1 “Semanticizing” Accounts

scare-quoted expression is interpreted as if it were uttered by some other, is that

is made explicit in a sentence that involves a definite description formed in terms

Your friend “Quine” is not Quine, he is only McPherson.

Original: The tardigrade : : : is particularly impressive among the exhibit’s

I move on now to non-semanticizing accounts.

Predelli’s application of the multiple-propositions idea to scare-quoting is this.

(García-Carpintero concurs, writing, “I think we should deal with [such] cases : : :

The quote-marked part is interpreted by taking it (along with the marks) as a

8 Incorporation and Assertion

Abstract In saying that “A&B” is logically equivalent to “B&A”, I use strict

1 Strict Quotation vs. Quasi-quotation

The following allows at least two interpretations.

© Springer International Publishing AG 2017 35

Section 2.1 explains the need for quasi-quotation in meta-logic. Quasi-quotation is

2.1 The Need for Quasi-quotation

(4) If it is raining is a formula then it is not raining is a formula.

(5) If “R” is a formula then “R” is a formula.

(6) If “®” is a formula then “®” is a formula (in £).

(7) If ® is a formula then “”® is a formula, i.e. if ® is a formula then the

(10) If ® and § are formulas then (® v §) is a formula.

2.2 Attempts to Explicate Quasi-quotation

As an efficient way to express generalizations about the object language, Quine

(12) If p ®q is a formula then p ® q is a formula.

(15) p ®q is equivalent to p ®q .

Saka & Johnson - The Semantics and Pragmatics of Quotation - 2017

Saka & Johnson - The Semantics and Pragmatics of Quotation - 2017