Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Conceptually permissive attitudes
(USC Thesis Other)
Conceptually permissive attitudes
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
.
Conceptually Permissive Attitudes
by
Jonathan Wright
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHILOSOPHY)
May 2020
Table of Contents
Acknowledgments.......................................v
1 Introduction.........................................1
1.1 Conceptually Permissive Attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Potential Applications of Conceptually Permissive Attitudes . . . . . . . . . . 10
1.3 Brief Summary of the Rest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Permissive Inconstant Reports and the Strict View . . . . . . . . . . . . . . 17
2.1 Introducing the Strict View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Sensitive Contents Accounts of Permissive Reports . . . . . . . . . . . . . . . . 18
2.3 Some Initial Objections to the Strict View . . . . . . . . . . . . . . . . . . . . . 20
2.3.1 Controversial Permissive Reports . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.2 The Cognitive Significance of Object-Directed Thoughts . . . . . . . . . 23
2.4 Permissive Inconstant Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4.1 The Simple Plurality Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.2 The Problem of Permissive Inconstant Reports . . . . . . . . . . . . . . . 30
2.4.3 A Functional Defense of the Plurality View . . . . . . . . . . . . . . . . . 34
2.5 Permissive Inconstant Reports and Substitution Theories . . . . . . . . . . . . 49
2.6 The Sensitive Attitudes Account of Permissive Reports . . . . . . . . . . . . . 55
3 Contextual Augmentation Accounts of Permissive Reports . . . . . . . . 60
3.1 Permissive Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.2 Contextual Augmentation Theories . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.3 The Sensitive Contextual Augmentation View . . . . . . . . . . . . . . . . . . . 66
3.4 Irrelevant Entailment Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.4.1 Cheesy Irrelevant Entailment Problems . . . . . . . . . . . . . . . . . . . 68
3.4.2 Harder Problems of Irrelevant Entailment . . . . . . . . . . . . . . . . . . 70
3.5 Acquaintance Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.6 Asymmetry Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7 Attitude Variation Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.8 Erroneous Assumption Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
ii
4 Permissive Dispositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2 More Di culties for Loaded-First Accounts . . . . . . . . . . . . . . . . . . . . . 81
4.3 Dispositional Attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.1 In-between Attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.2 Sensitive Dispositional Attitudes . . . . . . . . . . . . . . . . . . . . . . . 92
4.4 Permissive Dispositional Attitudes . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.5 Using Permissive Dispositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.5.1 Permissive Dispositions and Fully Permissive Cases . . . . . . . . . . . 101
4.5.2 Permissive Dispositions and Partially Permissive Reports . . . . . . . 105
4.5.3 Permissive Dispositions and Contextual Augmentation Problems . . . 107
4.6 Worries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.6.1 Permissive Speech Act Reports . . . . . . . . . . . . . . . . . . . . . . . . 112
4.6.2 Overgeneration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5 Question Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.1 Questions Under Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.2 Universal Instantiation Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.3 Starting to Apply QUD Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.4 A First Problem for QUD Sensitivity . . . . . . . . . . . . . . . . . . . . . . . 125
5.5 Refining and Rejecting QUD Sensitivity . . . . . . . . . . . . . . . . . . . . . . 128
5.5.1 Verb Phrase Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.5.2 Focus, Questions, and Verb Phrases . . . . . . . . . . . . . . . . . . . . . 130
5.5.3 Motivating a Broader Pragmatic Picture . . . . . . . . . . . . . . . . . . 135
5.5.3.1 Multiple Choice Questions and Answers Under Discussion . . 137
5.5.3.2 Questions about Attitudes . . . . . . . . . . . . . . . . . . . . . . 140
5.6 Question Sensitivity and the Truth-Conditional Question . . . . . . . . . . . 145
Appendix: Various Variations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
A.1 More Kinds of UI Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
A.2 More Switched Contexts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
A.3 Non-UI Asymmetry Problems . . . . . . . . . . . . . . . . . . . . . . . . 117
A.4 Bolstering Observations about Switched Contexts . . . . . . . . . . . . . 159
iii
6 Acquaintance Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.1 The Problem of Acquaintance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.2 The Problem with “Acquaintance” . . . . . . . . . . . . . . . . . . . . . . . . . 166
6.3 Extending the Principles to Non-Predicates: Determiners . . . . . . . . . . . 172
6.4 Extending the Principles to Non-Predicates: Names . . . . . . . . . . . . . . 179
6.5 Revising Specific Inaccurate Thinking . . . . . . . . . . . . . . . . . . . . . . . 185
6.6 The Acquaintance Problem for Contextual Augmentation Views . . . . . . . 191
6.7 Inaccurate Concept Application and Contextual Augmentation Views . . . . 195
7 Wrapping Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
7.1 Addressing Remaining Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
7.1.1 The Attitude Variation Problem . . . . . . . . . . . . . . . . . . . . . . . 207
7.1.2 The Irrelevant Entailments Problem . . . . . . . . . . . . . . . . . . . . 213
7.2 Final Recap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
iv
Acknowledgments
First I want to thank my advisor, Scott Soames. Scott was the reason why I came to
USC. He has mentored me throughout my time in the program. He has met with me about
manypapersandprojects, alwaysprovidinggenerouscommentsandinstructivediscussion.
He has been supportive of me as a philosopher, has been an advocate of my work, and has
generally had my back. He makes me feel like I am good at philosophy and that I have
important things to contribute to the field. My time with Scott counts among the most
gratifying experiences of my life. I am lucky to have been his student. Thanks, Scott!
I also owe special thanks to the other members of my committee.
John Hawthorne has been incredibly influential on this dissertation and on my thinking
abouttheissuesdiscussedwithin. ManyofmycasesareinspiredbytalkswithJohn; pretty
much every interesting case in this dissertation is a cool idea I stole from him and then
made convoluted and wordy. I’ve learned a lot from talking with John and have had a lot
of fun struggling with his tricky examples. Mark Schroeder was helpful to me in ways I
didn’t realize I needed. In addition to his helpful criticisms about the big ideas driving
my dissertation, throughout the process he has been full of insightful guidance on how to
presentaspectsofmyviewsandhasbeenaninvaluablesourceofgeneralencouragementand
advice. I am grateful to Andrew Bacon for his helpful questions and comments about this
dissertation, for helping me to develop ideas in other philosophical work, and for teaching
me so much exciting stu↵ in courses I have taken at USC. I am grateful to Barry Schein
for his challenging feedback. His comments have pushed me to think through di cult
examples and powerful theoretical alternatives to the views I try out in this dissertation.
Indeed, I’m still working through them. Thanks, committee members!
I’d like to thank Caleb Perl for countless conversations about work of mine over the
years. Caleb has taught me much, and I cherish his friendship. Thanks, Chip!
SpecialthanksarealsoduetoNatalieSchaad,theprogramadministratorfortheSchool
of Philosophy, who has helped me out of many a jam. Thanks, Natalie!
I’d also like to thank others whom I have learned from during my time at USC, includ-
ing Mike Ashfield, Renee Bolinger, Alex Dietz, Sean Donahue, Kenny Easwaran, Maegan
Fairchild, Steve Finlay, Jen Foster, Joe Horton, Nathan Howard, Nicola Kemp, Tanya Kos-
tochka,WooRamLee,ElliNeufeld,ShyamNair,DanielPallies,MichaelSechman,Kenneth
Silver, Douglas Wadle, Aness Webster, and Ralph Wedgwood. Thanks, philosophers!
I’d also like to thank my close friends, who have been supportive in so many ways
throughout my time in grad school. I am especially grateful to Daniel and Shelli, Ted and
Jenny and Claire, Louis, David, Devin, Kenny and Marlo, and Max. Thanks, buddies!
Finally, I want to thank my parents for their love and encouragement. My parents
taught me to question and to learn and to love questioning and learning, for which I
cannot thank them enough. Thanks, mom and dad!
v
Chapter 1. Introduction
This dissertation is about verbs like ‘believe’, ‘know’, ‘doubt’, ‘hope’, and ‘fear’, what
such verbs mean, and how sentences in which such verbs occur can be used to say di↵ erent
things in di↵ erent conversational contexts. It is also about what we learn about mental
states like belief, knowledge, doubt, hope, and fear, by learning what the meanings of these
verbs must be like, given certain ways that these verbs can be used to say di↵ erent things
in di↵ erent conversational contexts.
Consider a sentence involving the word ‘believes’:
(1) Mrs. Peacock believes that every forgery in the gallery is an original Renoir.
Sentence (1) surely can be used to say many di↵ erent things in di↵ erent conversational
situations, since it’s full of words that we know can be used to pick out, or designate,
di↵ erent things in di↵ erent conversational situations. Depending on the conversational
situations that (1) is used in, the names ‘Mrs. Peacock’ and ‘Renoir’ might be used to
refer to di↵ erent people, ‘the gallery’ might be used to refer to di↵ erent places, ‘every
forgery in the gallery’ might be used to speak of every forgery of any kind in the gallery
or to speak of every forgery that is a painting in the gallery, and so on. Pretty much every
sentence of English is in some sense context-sensitive, since so many words and ways of
combining words are context-sensitive; this much is not news.
In this dissertation, I am concerned with another particular way in which sentences
like (1) can be used to say di↵ erent things in di↵ erent conversational contexts. To get a
feel for this kind of context-sensitivity, consider two particular conversational contexts in
which one might use (1), which I’ll call Forgeries
1
and Forgeries
2
.
Forgeries
1
: Detective Black and detective Grey know some secrets: there are forgeries
inthegallery, andofthepaintingsinthegallery, allandonlytheforgeriesareimpressionist
paintings. By speaking with Mrs. Peacock about the gallery, Black learns that Peacock
believes that every impressionist painting in the gallery is an original Renoir. Later, in
discussing the gallery with Grey, Black assertively utters
(1) Mrs. Peacock believes that every forgery in the gallery is an original Renoir.
Contrast this context with
Forgeries
2
: Detective Black and detective Grey know that Peacock has some strange
opinions about art history. They know that Peacock believes that both Renoir and Monet
wereprolificartforgers. IncontrasttoForgeries
1
,itiswidelyknownthatthereareforgeries
of Caravaggio paintings in the gallery. By speaking with Mrs. Peacock about the gallery,
BlacklearnsthatPeacockthinksthateveryforgeryofaCaravaggiowaspaintedbyRenoir.
Later, in discussing the gallery with Grey, Black assertively utters
(1) Mrs. Peacock believes that every forgery in the gallery is an original Renoir.
1
The single sentence (1) apparently is used to say di↵ erent things in these di↵ erent
contexts. The di↵ erence I am interested is this. In Forgeries
2
, intuitively what detective
Black asserts with her use of (1) in some sense represents Peacock as thinking of things as
forgeries; by contrast, in Forgeries
1
, intuitively what Black asserts with her use of (1) does
not represent Peacock as thinking of things as forgeries. For instance, were Grey to accept
what Black says with (1) in Forgeries
1
, he would not expect to get much information by
asking Peacock “which particular paintings in the gallery are forgeries?”, but were Grey to
accept what Black says with (1) in Forgeries
2
, he could reasonably expect that in response
to this same question Peacock would direct him to particular paintings on the wall.
This kind of context-sensitivity is not itself really news either, since philosophers and
linguistics have been aware of this sort of context-sensitivity for over a hundred years.
1
But there are things we might say about this kind of context-sensitivity, things that I do
say in this dissertation, which are new and which I think have new things to teach us.
The di↵ erence in what (1) can be used to say in these contexts raises a question, a kind
of puzzle. How should we describe the di↵ erence in what (1) says across these contexts?
What about the meanings of the words in (1), how these words are interpreted, and how
these words combine, gives rise to the di↵ erence in what (1) can serve to assert?
Our attention immediately is drawn to the phrase ‘forgery in the gallery’. After all, the
di↵ erence in what (1) can be used to say across these contexts concerns whether Peacock
is represented as “thinking in terms of” the words in this phrase, or whether Peacock
is represented as using certain concepts associated with this phrase. So perhaps the felt
di↵ erence in this pair of cases is due to our interpreting this phrase in di↵ erent ways across
these two contexts. It is a tempting thought, and has been developed in various ways by
linguists and philosophers both. I consider such views in chapter 2.
Another thought is that this di↵ erence in interpretation instead has something to do
with the meaning of the verb ‘believes’ and how we interpret that word across these con-
texts. Perhaps, one might think, we simply use verbs like ‘believes’ in di↵ erent ways, such
that there are di↵ erent kinds of beliefs, or di↵ erent ways of believing. Sometimes, perhaps,
whenyousaythatsomeonebelievessomethingorotheraboutforgeries, youascribetothat
person a kind of belief that involves thinking of things as forgeries. Other times when you
say that someone believes something or other about forgeries, you ascribe to that person
another kind of belief that does not involve thinking of things as forgeries. This is the view
I defend and develop in this dissertation. In Forgeries
1
and Forgeries
2
, on my view, it is
literally true that Peacock believes that every forgery in the gallery is an original Renoir.
What di↵ ers across these cases is just what sort of belief Peacock is reported to possess.
In the next section, I elaborate on how I develop this kind of view and point to ways in
which this view, together with other views I adopt concerning issues in the philosophy of
language and mind, diverges from certain prevalent ideas about ‘believes’ and beliefs, and
other such verbs and attitudes.
1
Russell (1905).
2
1. Conceptually Permissive Attitudes
Philosophers and linguists call the class of verbs I reference above “attitude verbs”,
call certain speech acts made with sentences in which such verbs occur “attitude reports”,
and call the mental states that such verbs serve to express “propositional attitudes”.
2
In
the course of this dissertation I defend, or perhaps it would be better to say, I try on for
size, various linguistic principles about the semantics and pragmatics of attitude verbs and
philosophical principles about the metaphysics of propositional attitudes. Many of the
specific principles which I try out along the way turn out to be false or look like they may
meet counter-examples soon enough. But there are a few ideas central to the dissertation
that I stand by and which I think have special things to teach us.
I conceive of these views as a package deal; my main goal is to persuade the reader that
this package of views is a promising way of making sense of the kind of context-sensitivity
observed above. The package includes three main views. The sensitive attitudes view is
a broadly semantic view, concerning the nature of the conventional meanings of attitude
verbs. The conceptually permissive attitudes view is a view in the metaphysics of mind,
concerningwhatcertainkindsofpropositionalattitudesarelike. Thepragmatic speech acts
view, assumed but not explicitly defended in the dissertation, is a view about pragmatics
and direct speech acts, concerning how what is asserted or asked by a use of a sentence
involving attitudes verbs depends on features of conversational contexts in a certain way.
Accordingtothesensitiveattitudesview, attitudeverbsaresemantically underspecified
in such a way that di↵ erent uses of a single attitude verb can express di↵ erent things in
di↵ erent contexts. Simply put, on the sensitive attitudes view, di↵ erent uses of a partic-
ular attitude verb like ‘believes’ can designate di↵ erent propositional attitude relations,
believes
1
, believes
2
, and so on.
3
I call the class of attitude relations that uses of an
attitude verb can designate that verb’s “designation candidates”. The sensitive attitudes
view stands in contrast to a view according to which, given the conventions of English,
a verb like ‘believes’ always serves to express a single relation, regardless of what conver-
sational context it is used in. On the insensitive attitudes view, there is just one single
relation that is a designation candidate for ‘believes’, and so too for any other attitude
verb.
4
Ifthesensitiveattitudesviewistoprovideanadequatetreatmentofthekindofcontext-
2
Such verbs can be used to express attitudes towards things other than propositions as well, but in this
dissertation I focus on uses of attitude verbs that express relations between subjects and propositions.
3
Throughout most of the dissertation I assume that attitude verbs express relations between subjects
and propositions, though in chapter 4 I consider an alternative view on which attitude verbs, together with
clauses with which they combine, express dispositions. The sensitive attitudes view is compatible with
other views of the semantic type of attitude verbs, but I don’t explore these alternatives in detail here.
4
One who generally favors an insensitive attitudes view may still adopt a kind of sensitive attitudes
view for individual verbs to handle philosophical or linguistic puzzles idiosyncratic to that verb. Con-
textualist theories about ‘knows’ such as those discussed in Hawthorne (2004), adopted to resolve certain
epistemological puzzles, are an example.
3
sensitivityobservedabove,theclassofdesignationcandidatesforaverblike‘believes’must
be vast. As we will see, to account for the kind of context-sensitivity I am interested in,
among the designation candidates for any particular attitude verb there will be many
relations varying in what restrictions they place on how subjects must be thinking of, or
conceptualizing, di↵ erent objects and relations.
5
These relations, of course, must share
certain features in common, to distinguish the designation candidates of ‘believes’ from,
say, those of ‘hopes’.
To shed some light on the version of the sensitive attitudes view that I adopt in this
dissertation,Iwillpointtoatheoreticalchoicepointconcerninghowthestandingmeanings
of attitude verbs give rise the variety of relations that they can designate.
On one view, any attitude verb like ‘believe’ that gives rise to the kind of context-
sensitivity observed above is conventionally associated with a semantic rule that somehow
makesreferencetocertaindiscretefeaturesofcontexts. Onthisview,theobservedcontext-
sensitivity of attitude verbs is accounted for by postulating linguistic meanings for such
expressions similar to those traditionally associated with indexical expressions like ‘I’, ‘yes-
terday’, and ‘nearby’. Traditionally, the linguistic meanings of these indexical expressions
explicitly include open variables that are saturated by contextually determined parame-
ters that can vary across contexts. According to the linguistic rule associated with ‘I’,
for example, a use of ‘I’ serves to contribute the speaker in the context of use; according
to the rule associated with ‘yesterday’, a use of ‘yesterday’ serves to contribute the day
before the day of the context of use; according to the rule associated with ‘nearby’, a use
of ‘nearby’ serves to contribute the property of being near to the location of the context
of use. Given these rules, the conventional linguistic meaning of ‘I met Green at a nearby
bar yesterday’ is semantically complete, in a way; we might represent its meaning, or its
“semantic content” as ‘[x met Green at a bar near y on the day before the day including
z]’, where the variables ‘x’, ‘y’, and ‘z’ are to be saturated, as a matter of conventional
rule, by the speaker in thecontext, thelocation of the context, and thetime of thecontext,
respectively.
6
Similarly, on one way of developing the sensitive attitudes view, the di↵ erent read-
ings exhibited by (1) above are accounted for by claiming that the conventional linguistic
meaning of ‘believes’ is explicitly context-sensitive in such a way that it includes, as a
matter of conventional rule, certain open variables that take as values certain contextually
determined parameters. What these parameters are and how they would account for the
di↵ erenceininterpretationexhibitedinForgeries
1
andForgeries
2
issomethinganadvocate
5
They plausibly will vary further in how their application conditions relate to those of other designation
candidatesof‘believes’andvarioussetsofpropositions. Ifsomethinglikethefinal“contextualaugmentation
view” I propose in chapter 7 is correct, then the application conditions of many believes relations that
are designated by ‘believes’ in certain contexts supervene on how the application conditions of distinct
believes relations themselves relate to certain propositions that are conversationally assumed in those
reporting contexts.
6
The simple indexical semantics listed here for ‘nearby’ is too simplistic, as we will see in chapter 2.
4
of this style of view would have to develop.
A contrasting view, which I adopt in this dissertation, is that attitude verbs are simply
semantically underspecified in such a way that their conventional linguistic meanings are
notrigidrulesthatdeterminedi↵ erentattituderelationsindi↵ erentcontextsinaccordance
with di↵ erent potential values for conventionally encoded variables. Instead, the linguistic
meanings of attitude verbs are open-ended constraints on which relations can be desig-
nated by di↵ erent uses of ‘believes’. On this view, all there is to the linguistic meaning
of ‘believes’ is a conventional constraint on what ordinary, sincere uses of ‘believes’ can
contribute; we might represent its meaning as a rule of use that says that one may use
this verb to designate one of the elements in B,where B is a certain class of attitude
relations including believes
1
, believes
2
, and so on—the class of designation candidates
of ‘believes’. We don’t calculate, as theorists or conversationalists, which attitude relation
is contributed by a particular use of ‘believes’ in terms of a context-sensitive semantic rule,
or a function from certain contextually given parameters to attitude relations. Rather,
semantic competence with ‘believes’ involves an ability to use the verb to designate just
relationsinB. Ascertainingwhichrelationaparticularuseof‘believes’servestodesignate,
or contribute to a proposition asserted by a use of a sentence involving this verb, is a mat-
ter of reasoning to which designation candidate of ‘believes’ makes the most sense given
the goals and assumptions characterizing the conversational situation in which the verb
is used. Below, in presenting the pragmatic speech acts view, I say more about this and
point to some reasons to favor this underspecification view over the indexical alternative.
At this point, we begin to see how the sensitive attitudes view might provide a solu-
tion to our puzzle about the di↵ erent readings that (1) takes in Forgeries
1
and Forgeries
2
.
On the sensitive attitudes view, in Forgeries
1
Black’s use of (1) serves to assert a propo-
sition like Peacock-believes
1
-that-every-forgery-is-an-original-Renoir, whereas,
inForgeries
2
,Black’suseof(1)servestoassertapropositionlikePeacock-believes
2
-that-
every-forgery-is-an-original-Renoir.Thefeltdi↵ erence in what Black says across
these contexts is due to her expressing di↵ erent relations with her uses of ‘believes’ in these
di↵ erent contexts.
This leads me to the next view central to this dissertation, the conceptually permissive
attitudes view. This view says something about what certain of the things that attitude
verbs can express are like.
I use the lingo of ‘permissive’ and ‘loaded’ throughout the dissertation in two main
ways. I use these words in the first instance to classify di↵ erent ways of interpreting atti-
tude reports. I say that in Forgeries
1
, (1) takes a “permissive reading”. In particular, it
takes a permissive reading “with respect to” the phrase ‘forgery in the gallery’ in (1).
7
In
7
As I develop the ideology of permissiveness throughout this dissertation, I end up relativizing this way
of using ‘permissive’ further to be sensitive to certain occurrences of the relevant phrase in the complement
of the report sentence and to the concepts which are associated in a certain way with that occurrence by
the conversationalists in the reporting context. Throughout this introductory chapter I pass over these
complications.
5
speaking this way, I am merely characterizing the way that we interpret (1) in Forgeries
1
:
as we interpret this report, we don’t think that Green has to think in terms of, or concep-
tualize any particular things as, forgeries in the gallery. Similarly, I say that (1) takes a
“loaded reading” in Forgeries
2
. In particular, it takes a loaded reading with respect to the
phrase ‘forgery in the gallery’. Again, in speaking this way I am simply characterizing how
we interpret (1) in Forgeries
2
: as we interpret that report, we do think that for the report
to be true Green must, in some way to be clarified, be thinking in terms of, or conceptu-
alizing certain things as, forgeries in the gallery. This is the primary way throughout this
dissertation that I use ‘loaded’ and ‘permissive’: to characterize our way of interpreting
attitude reports as carrying or not carrying this kind of requirement on how the subject of
a report must think of certain things.
I use the terms ‘permissive’ and ‘loaded’ in a second way, to characterize what the
di↵ erent attitude relations are like. What is it about believes
1
and believes
2
that
account for the di↵ erence in interpretation in Forgeries
1
and Forgeries
2
? The short answer
is that believes
1
is what I call a “conceptually permissive attitude”. It is conceptually
permissive in that it is possible that one bears believes
1
towards the proposition that
every forgery in the gallery is an original Renoir despite the fact that one does not think
of things as forgeries in the gallery. By contrast, believes
2
is what I call a “conceptually
loadedattitude”. Itisconceptuallyloadedinthatnecessarilyonebearsbelieves
2
towards
the proposition that every forgery in the gallery is an original Renoir only if one thinks of
things as forgeries in the gallery.
8
Onecanagreethat(1)takesapermissivereadingwithrespectto‘forgeryinthegallery’
in Forgeries
1
without accepting that ‘believes’ contributes a conceptually permissive atti-
tude relation in that context. One might instead adopt the idea floated above, that the
best account of the permissive reading of (1) in Forgeries
1
is that in that context we in-
terpret ‘forgery in the gallery’ in a di↵ erent way from how we interpret that phrase in
Forgeries
2
. Perhaps in Forgeries
1
we interpret this phrase as contributing a property other
than forgery-in-the-gallery, such as the property impressionist-painting-in-the-
gallery, so that in fact Black asserts in that context just that Green thinks that every
impressionist painting in the gallery is an original Renoir.
9
This could account for why (1)
takesapermissivereadingwithrespectto‘forgeryinthegallery’. OrperhapsinForgeries
1
we interpret ‘forgery in the gallery’ as outside the scope of ‘believes’, so that what Black
asserts in that context is that every forgery in the gallery is such that Green believes that
it is an original Renoir.
10
This also could account for the permissive reading of (1) in that
context. In chapter 2 I discuss such views, arguing that this sort of view cannot account
for the full range of permissive readings; we need to adopt the sensitive attitudes view
8
Again, this “thinks of things as” ideology is a placeholder for something more involved, to be developed
throughout the dissertation.
9
I call these views “substitution theories” in chapter 2. See Schwager (2009) and Sudo (2014).
10
Russell (1905) o↵ ers an explanation along these lines. This general kind of strategy can be developed
in intricate and powerful ways. We see several in chapter 2, but also see Schein (2006).
6
together with the conceptually permissive attitudes view.
The conceptually permissive attitudes view supplements the sensitive attitudes view’s
account of the di↵ erent readings that (1) takes in Forgeries
1
and Forgeries
2
.In Forg-
eries
1
Black’suseof(1)servestoasserttheproposition Peacock-believes
1
-that-every-
forgery-is-an-original-Renoir,where believes
1
is conceptually permissive with re-
spect to forgery-in-the-gallery. For this reason, (1) can take the true reading it does
in this context even though Green does not think of the things that are in fact forg-
eries “as forgeries”. In Forgeries
2
, Black’s use of (1) serves to assert a proposition like
Peacock-believes
2
-that-every-forgery-is-an-original-Renoir,wherebelieves
2
is
conceptually loaded with respect to forgery-in-the-gallery. For this reason, for the
proposition that (1) is used to assert in this context to be true, Green must think of things
that are in fact forgeries as forgeries. These consequences of the sensitive attitudes view
andtheconceptuallypermissiveattitudesviewgosomewaytoaccountingforthedi↵ erence
between the readings that (1) takes in Forgeries
1
and Forgeries
2
. Surely, though, more
needstobesaidaboutwhatittakestobearapermissiveattituderelationtowardsapropo-
sition; to this end, throughout this dissertation I try on for size various views concerning
the application conditions of permissive attitude relations. It gets tricky.
The final view central to this dissertation is the pragmatic speech acts view, according
to which what one asserts by uttering an attitude report sentence is intimately related
to pragmatic features of conversational situations. The contrast felt in Forgeries
1
and
Forgeries
2
is due to di↵ erences in communication that are not fully determined by the
semantics of attitude report sentences and are also not merely di↵ erences in what uses of
(1) conversationally implicates or presupposes. Rather, this interpretative contrast is a
matter of Black’s (literally) asserting di↵ erent propositions in these contexts.
In chapters 5 and 6, I motivate the view that whether a report takes certain permissive
readings depends in delicate ways both on what conversational goals and what conversa-
tional assumptions are made in a reporting context. My discussion in those chapters shows
that there appear to be no simple pragmatic principles that provide a comprehensive char-
acterization of the conditions under which a report takes a permissive reading. Rather,
severaldefeasibleprinciplesaboutconversationalassumptionsandgoalsappeartoworkto-
gether in a delicate confluence, balancing against each other in di↵ erent ways to determine
whether a report takes a permissive reading. These pragmatic complexities are interest-
ing in their own right, but they also support the combination of the sensitive attitudes
view and the pragmatic speech acts view: attitude verbs are semantically underspecified
in such a way that they can be used to designate di↵ erent relations in di↵ erent contexts,
this context-sensitivity is not traceable to a rigid indexical linguistic rule associated with
attitude verbs, and which relation is in fact designated by a use of such a verb depends on
7
norms of pragmatic reasoning of the same kind involved in interpretation generally.
11
Inchapter5,Iregimenttalkofconversationalgoalsintermsof“questionsunderdiscus-
sion”. By priming di↵ erent uses of attitude report sentences with di↵ erent questions that
the conversationalists in a reporting context are trying to resolve, we can toggle whether
uses of a particular report sentence serve to make permissive or loaded reports. In chapter
6, Iraiseexamplesthatshowthatwhetherareporttakesapermissivereadingwithrespect
to a certain concepts depends on conversational assumptions about whether and how the
subject uses that concept.
12
That whether a report takes a permissive reading depends
somehow on these two conversational features already puts some pressure on the version
of the sensitive attitudes view on which the linguistic meanings of ‘believes’ is explicitly
context-sensitive in such a way that it includes open variables that take conversationally
given parameters as values. For on that view the linguistic meaning of ‘believes’ would
apparently at least have to include variables that are to be saturated by certain questions
underdiscussionandcertainconversationalassumptionsabouthowthesubjectusescertain
concepts. The idea that linguistic conventions about a certain class of verbs can develop
to explicitly track such features of conversations is itself dubious. The idea is shown to be
more dubious once we take on board the idea, defended in chapters 5 and 6, that there
is no way of stating simply exactly how permissive readings are dependent on questions
under discussion and certain conversational assumptions. Or, at least, there is no way of
stating how permissive readings depend on such features of conversational situations short
of stating a comprehensive theory of the pragmatic reasoning involved in the practice of
interpretation generally. Better to leave these complications in the realm of pragmatics
and say that what a speaker asserts constitutively depends on pragmatic influence than to
build such complexities explicitly into the linguistic meanings of certain verbs.
13
This package of views has some surprising consequences in the philosophy of mind,
which I think bear mentioning up front. I dig into these consequences further in chapters
11
While I do think that general pragmatic reasoning is involved in determining whether a report takes a
permissive reading, I do not think that the interpretative di↵ erence exhibited by Forgeries1 and Forgeries2
is simply a matter of what the speaker conversationally implicates or presupposes. Although there are
potentially ways of using pragmatics to account for this interpretative di↵ erence in such terms, I don’t
spend much time in the dissertation arguing in favor of the pragmatic speech acts view against such
alternatives. One interesting alternative that I do not much discuss but that I intend to address in future
work is the account given by Romoli and Sudo (2009), which accounts for permissive readings (for them,
“transparent” readings), in terms of presupposition projection.
12
Throughout the dissertation I use the lingo of “concepts” in a minimal, non-committal way. I stand
by no particular theory of concepts, but rather here take concepts simply to be ways of representing, or of
thinking about, or of cognizing, di↵ erent properties and relations.
13
The literature on the relationship between semantics and pragmatics is far more vast and complicated
than suggested by this simple introduction, where I make as if a hidden indexical view and a constraint-
based underspecification view are the only options for how to make sense of the influence of pragmatics on
what one asserts in such cases. An important kind of alternative to consider are views on which pragmatic
influence serves to add “unarticulated constituents” to a sentence’s conventional semantic content. See
Bach (1994), Perry (2000, chapter 10), and Carston (2004).
8
2 and 4. Consider the following sort of case, which I discuss throughout the dissertation:
Blackmail: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called in to discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet: Green doesn’t know any name for Scarlet, could not identify her in a line-up,
and so on. By questioning Green, Detective Black learns that Green thinks that every
socialist guest was being blackmailed by Mr. Body. Green doesn’t, however, know the
names of any of the socialist guests: he claims “not to know who the socialist guests are”.
The detectives are exchanging information, trying to find out who had a motive for killing
Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
In Blackmail, intuitively (2) is true and takes a permissive reading with respect to
‘Scarlet’. Black does not represent Green as thinking of anything or anyone “as Scarlet”,
or ashavingany particular awarenessof Scarlet’sexistence. On our packageof views, what
Black asserts with (1) in this context is literally true. Black’s use of ‘thinks’ designates a
relation thinks
Blackmail
, and the proposition she asserts predicates thinks
Blackmail
of the
pair of Green and the proposition that Scarlet was being blackmailed by Mr. Body. Green
surely must satisfy certain conditions to count as a bearing thinks
Blackmail
towards this
proposition (exactly what conditions a permissive attitude relations place on pairs of sub-
jects and propositions is a matter taken up in this dissertation). But since thinks
Blackmail
is conceptually permissive with respect to what ‘Scarlet’ designates in (2), one thing Green
need not do to verify (2), given our package of views, is cognize Scarlet in certain ways,
or in any way whatsoever. Green can count as literally believing that Scarlet was being
blackmailed even though Green is unaware of Scarlet’s existence, so lacks any fully loaded
attitudes towards propositions directly about Scarlet.
14
This result may be surprising to some. Surely, one might object, Green doesn’t really
think that Scarlet was being blackmailed. You can’t think that Scarlet is one way or
another if you can’t even think of Scarlet in the first place! Along these lines, Recanti
(2012, 123) says that inferring from examples like these that one possesses a “singular
thought”, a thought the content of which is directly about an individual, such as Scarlet,
14
Throughout the dissertation I use the label “fully loaded attitude relations” which are loaded with
respect to every constituent of a proposition (and associated concepts).
9
is an “extraordinary conclusion” and is “clearly unacceptable”.
15
If by saying that Green doesn’t “really” think that Scarlet was being blackmailed,
one means that Green doesn’t bear a fully loaded thinks relation to the proposition that
Scarlet was being blackmailed, then I agree. But it seems clear that one who raises this
kindofcomplaintdoesnotmeanjustthis. Onethingsuchacomplainermightmeanisthat
(2) is simply not literally true. With this I disagree, and it is a task of this dissertation
to motivate the idea that any adequate semantic and pragmatic account of permissive
reporting predicts the literal truth of (1) in Blackmail. Another thing the complainer
might mean is that even if (2) is literally true in Blackmail, this should not be taken as
evidence about what “genuine” propositional attitudes (that is, I take it, propositional
attitudes answering to certain theoretical goals in the philosophy of mind) are like. With
this I might actually agree, depending on how these “genuine” attitudes are understood; it
may be that only certain of the designation candidates of ‘believes’ answer to traditional
theoretical roles associated with belief.
But for those who take this line, I insist that it is important to be vigilant in clarifying
what these special theoretical roles are, and also to be vigilant in attending to how one’s
sense of their importance may be informed by intuitions about di↵ erent readings of atti-
tude reports. If our package of views is correct, then the attitude relations designated by
attitude verbs can vary diversely and can do so for diverse and delicate pragmatic reasons.
So we should expect that those drawing from intuitions about attitudes, which are often
mediatedbyintuitionsaboutattitudereports,mayeasilyconflatedistinctkindsofproposi-
tional attitudes, so may draw theoretical generalizations about the nature of belief, desire,
knowledge, or whatever attitude, that hold of no particular attitude relation in particular.
Thisleadsmetothenextsection, whereIdescribesomewaysinwhichtheconceptually
permissive attitudes view might be used to transform certain debates in the philosophy of
language and mind.
2. Potential Applications of Conceptually Permissive Attitudes
I said above that the package of principles which I stand by in this dissertation have
special things to teach us. Here I point to some ways that this picture of propositional
attitudes may inform philosophical debates in the philosophy of mind and language. The
generalmoraltobedrawnfromthesensitiveattitudesviewisthatsincethereisaplentitude
of attitude relations associated with any particular attitude verb, there is a plenitude
of philosophical questions or views that may be asked or expressed using any particular
attitude verb. I expect that some confusion in certain philosophical debates has been
15
Specifically, Recanati claims that this is unacceptable for “singularists”, those who believe that our
thoughts can be directly about individual objects rather than just being directly about properties and
relations. On the sensitive attitudes view together with the conceptually permissive attitudes view, one
can, it seems to me, be such a singularist. Green has a conceptually permissive thought the content of which
is directly about Scarlet. So the unacceptability here presumably has more to do with certain philosophical
assumptions about certain theoretical roles one might associate with singular thought in the philosophy of
mind.
10
wrought by failing to appreciate this. Given the natural assumption that ‘believes’, say,
is univocal, always expressing a single attitude relation, theorists may take themselves to
be addressing a single philosophical question about the nature of belief where in fact there
are many.
First we can address questions in the philosophy of mind about which agents are in a
position to bear propositional attitudes at all. Do simple animals, computer programs, or
groups of individuals have beliefs? Given our package of views, there are several questions
that one might be asking here. In light of this, one idea that becomes available on the
conceptuallypermissiveattitudesviewisthatalthoughthesekindsofagentsdonotpossess
fully loaded beliefs, this does not itself tell against their having any beliefs worth the name.
Consider the following case, discussed in chapter 4:
Chess: ProfessorPlumisplayingchessagainstthecomputerprogramLeela. Detectives
Black and Grey are looking on. After Plum moves his bishop to d4, Leela moves a bishop
to d7. Black turns to Grey and sincerely utters:
(3) Leela is boxing out Plum’s knight because it thinks that Plum wants to fork the
queen and rook.
What Black asserts with (3) sounds true and appropriate. But it is as intuitive that
the chess-playing program Leela lacks a concept of Plum. Indeed, one could argue that
Leela has no concepts whatsoever, so has no loaded attitudes whatsoever. One beholden
to an insensitive attitudes view might be tempted to say therefore that (3) is not literally
true, and that it has to be understood as somehow metaphorically or non-literally true.
But on the conceptually permissive attitudes view, even if the program Leela has no
loaded belief that Plum wants to fork its queen and rook, it may yet bear a conceptually
permissive belief relation towards that proposition. On this view, then, we potentially can
take (3) as literally true, as literally true as (1) and (2) in the contexts discussed above.
Similarly, as discussed in chapter 4, we might say on the conceptually permissive atti-
tudes view that simple animals, simple machines, corporations, baseball teams, and even
neurons have genuine propositional attitudes. Resistance to such claims, where they are
found, may be diagnosed with our package of views as downstream from a myopic assump-
tion that there is but one kind of belief (or desire, or whatever attitude) such that one
believes a proposition only if possesses and deploys concepts of every object and prop-
erty which that proposition is directly about. In chapter 2 I call this the “strict view” of
attitudes, and there work to discredit it.
Other questions in the philosophy of mind concern what it takes for one to have certain
kinds of thoughts. One such issue is what it takes for one to have “singular thoughts”,
whereasingularthoughtisathoughtwhosecontentcanbeexpressedbysentencesinvolving
directly referential terms or by sentences with open variables relative to assignments. A
singular thought is one whose content is directly about objects, rather than a thought
which is directly about only properties and relations.
11
One debate concerning singular thoughts is about whether one must be “acquainted”
with an object o in order to have a singular thought about o. According to latitudinar-
ianism, or liberalism, about singular thought, no such acquaintance constraint holds for
singularthought.
16
Onthesensitiveattitudesview,weseethattherearemanyviewswhich
can adopt this slogan, as long as having a singular thought consists in bearing a thinks
relation towards a singular content. For certain permissive attitude relations which are
permissive with respect to concepts of particular objects, liberalism looks definitionally
true. Reconsider Blackmail; in that case, I claimed that Green bears a thinks relation
that is conceptually permissive with respect to Scarlet (and, as I will say later, a particular
conceptofScarlet). Iftherearesuchcasesinwhichoneisreportedtobearacertainthinks
relation that is permissive with respect to Scarlet, say, and any concept of Scarlet, then
liberalism will clearly hold for singular thoughts corresponding to this thinks relation.
But this does not close the issue on whether liberalism holds for all attitude relations.
It may be that those who have toiled after di↵ erent kinds of acquaintance constraints on
singular thought have been attending to certain kinds of loaded thinks relations, and their
toil may not be in vain. There is still a question of whether acquaintance constraints
hold for certain loaded attitude relations, even if such constraints clearly do not hold for
certain permissive ones. The debate between those who uphold acquaintance constraints
on singular thought and those who deny such constraints may su↵ er from a failure to
distinguish between permissive and loaded thinks relations and corresponding kinds of
thoughts.
We can get a feel for how the conceptually permissive attitudes view may inform this
debate in particular by noticing certain new questions that it raises. Consider Evans
(1979)’s case of ‘Julius’ as a name given to whoever was the inventor of “the zip”. In
Evans’s example, we introduce the name ‘Julius’ to refer to whoever invented the zipper,
without knowing who exactly that is. It seems that once we go through this linguistic
stipulation, we can go on to speak truly using the name ‘Julius’, saying, for instance,
that Julius was born before 1986, and in so doing refer to Julius, whoever they may be.
Likewise, we can assert something true in uttering ‘Jonathan believes that Julius was born
before 1986’; if the heuristic mentioned above concerning how to detect singular thoughts
is correct, in so believing, I apparently bear a singular thought towards the proposition
o-was-born-before-1986,where o is the person who in fact invented the zipper.
One way that one might respond to this example is to say that I do indeed have a
singular thought in a proposition involving o, and that that I do have this singular belief is
evidence that it is easier than what one might have otherwise thought to acquire singular,
non-descriptiveconceptsofindividualobjects. Onthisview,Icanpossessanon-descriptive
concept of o by leveraging the semantic mechanism of introducing names by description.
But once we accept the conceptually permissive attitudes view, di↵ erent options, and
16
Hawthorne&Manley(2012)usethelabel“liberalism”forthedenialoftheclaimthattohaveasingular
thought about an object one must be acquainted with it. See Sosa (1970) for an earlier latitudinarian view.
12
corresponding theoretical questions, open up. It might be that that our attitude report
sentence ‘Jonathan believes that Julius was born before 1986’ takes a loaded reading with
respect to ‘Julius’ and this name directly refers to o.
17
In this case, the truth of reports
madewiththissentenceisevidencethatonecanpossessasingularconceptofanindividual
just by leveraging the linguistic mechanism of introducing names by description. But it
might also be that such attitude reports in fact regularly take permissive readings, so
that what we would assert with ‘Jonathan believes that Julius was born before 1986’
is permissive with respect to ‘Julius’. If that is so, then such examples may show that
singular thoughts are easier to come by than one might have thought without showing that
singular concepts are easier to come by than one might have thought. On the conceptually
permissive attitudes view, we should distinguish and explore what are two issues: first,
whether to bear some attitude relation or other towards a singular proposition about o
one must be acquainted with o; second, whether to possess a concept of o one must be
acquainted with o. Is the Julius case to be explained in terms of conceptually permissive
attitudes, or does this case show that singular concepts are easier to come by than some
might have otherwise thought? More work is needed to decide. New questions open up in
this discussion, given our package of views.
Finally, the conceptually permissive attitudes view may inform debate surrounding
influential arguments for the thesis that names and demonstratives are directly referen-
tial expressions and the related philosophical issue of externalism about thought content.
Consider the simple report
(4) Mrs. White believes that Scarlet is a thief.
A widely accepted view in the philosophy of language is that names like ‘Scarlet’ are
directlyreferential. Thatis, whatanyuseofanamelike‘Scarlet’servestodesignateisjust
theindividualtowhichthatuserefers;inthiscase,Scarletherself. Thisstandsincontrastto
viewsonwhichusesofnamessuchas‘Scarlet’servetodesignatepropositionalconstituents
involving just properties and relations, such as what is designated by a description like ‘the
personcalled‘Scarlet’. Ifnameswerenotdirectlyreferentialexpressions,thenitseemsthat
(4) would report a purely descriptive belief to White, the content of which would involve
just properties and relations. ‘Scarlet’ would contribute a property uniquely instantiated
by Scarlet to White’s reported thought content.
As argued by Kripke (1972), for any purely descriptive belief which could plausibly
be ascribed to White by (4), intuitively for White to believe such a purely descriptive
proposition, such as the proposition that the person called ‘Scarlet’ is a thief, White would
have to think of things which intuitively she need not think of to verify (4). For intuitively
there are contexts in which (1) is true in which White does not know that Scarlet’s name is
‘Scarlet’, so does not think of Scarlet as one so named. Since White can verify (4) without
17
See below for further complications that the conceptually permissive attitudes view raises for issues of
direct reference.
13
cognizing Scarlet as the person named ‘Scarlet’, or without cognizing Scarlet via any other
purely descriptive concept, according to one argument for the view that names are directly
referential, the linguistic meaning of ‘Scarlet’ cannot be such that any ordinary, sincere use
of ‘Scarlet’ designates purely descriptive content.
But with the sensitive attitudes view, things are not so straightforward. If the kind of
beliefsreported inthecontextsrelevanttothiskind of Kripkeanargumentare conceptually
permissive beliefs, then (4) could indeed serve to report a purely descriptive belief. White
can bear a believes relation which is conceptually permissive with respect to the purely
descriptive contribution of ‘Scarlet’, the-person-named-‘Scarlet’ and an associated de-
scriptive concept the-person-named-‘Scarlet’ without possessing or deploying this concept.
On the conceptually permissive attitudes view, a use of (4) could even ascribe to White a
belief that the unique instance of person-named-‘Scarlet’ in the actual world is a thief,
without requiring that White possess any concept of the name ‘Scarlet’ or of the actual
world.
18
This way of applying the conceptually permissive attitudes view to the issue of directly
referential expressions has further potential consequences for the question of whether ex-
ternalism about thought contents is correct. According to externalism, given the idea that
‘Scarlet’isdirectlyreferential,ratherthandesignatingapurelydescriptivepropertyorrela-
tion, thetruthof(4)supervenesonmorethaninternalfactsaboutWhite. Rather, whether
White believes this proposition about Scarlet must supervene on something further about
how White relates to her environment, such as how she is perceptually or socially related
to Scarlet. In this way, assuming a version of the insensitive attitudes view according to
which ‘believes’ univocally designates an attitude relation that is fully loaded, one can
argue from the fact that names like ‘Scarlet’ are directly referential to the conclusion that
one’s thought contents supervene on more than one’s inner goings-on—that beliefs ain’t in
the head.
The sensitive attitudes view is some ways friendly and in some ways unfriendly to
this idea. For, on the sensitive attitudes view, there is a plenitude of externalist views,
corresponding to di↵ erent attitude relations one might be externalist about. Some versions
of externalism opened up by the sensitive attitudes view are apparently more extreme than
standard externalism. Since to bear a permissive attitude relation to a proposition does
not require one to possess and deploy concepts of constituents of that proposition, but
rather requires that certain facts hold in the context of attribution, beliefs not only ain’t in
the head, they ain’t in the environment either. For, as in the permissive reading of (2) in
Blackmail,therecanbetruepermissivereadingsof(4)onwhichWhiteneednotbecausally
or epistemically related to Scarlet at all, but instead which would just require that White’s
fully loaded attitudes, together with certain assumptions made in a conversation towards
which she need not bear any special causal, epistemic, or conceptual relation, themselves
18
The conceptually permissive attitudes view thus threatens, for instance, Soames (2002, 39-50)’s actu-
ality argument against descriptivism about names and some of Braun (2008)’s arguments against quantifi-
cational accounts of complex demonstratives.
14
relate in a certain way to the proposition that Scarlet is a thief.
With respect to other, loaded, attitude relations, the sensitive attitudes view throws
externalismintoquestion: eventhoughwhetherPeacockbearsthebeliefrelationwhichshe
is said to by (4) does not supervene on intrinsic facts about Peacock, we cannot conclude
that there are not other theoretically important, fully loaded belief relations, for which
externalism is false. If there are fully loaded believes relations which are such that
necessarilyonebearssucharelationtowardsapropositiononlyifonepossessesanddeploys
concepts associated with that proposition’s constituents in a certain way, then Kripke’s
semantic argument does not itself show that the contents of such attitudes cannot be given
a purely internalist treatment.
19
Perhaps then the debate over externalism and interalism
about thought contents, too, su↵ ers from a failure to distinguish between the variety of
attitude relations which attitude verbs can serve to designate. Some of these relations may
abide by internalist constraints, others may abide by standard externalist constraints, and
others yet may swing free in yet more liberal ways of externalist constraints. This kind of
thought,thatattitudereportingpracticesareprofligateinawaythatdoesnotreliablytrack
features of certain kinds of attitudes that play special theoretical roles in the philosophy
of mind, is not new, but we’ve seen here that the conceptually permissive attitudes view,
insofarasitaccountsforourpracticesofpermissiveandloadedreportinggenerally,provides
independent reason to adopt a picture on which this thought is exceedingly natural in its
own right.
20
These then are some potential applications of the conceptually permissive attitudes
view. I think there may be more besides. Intuitions about proposition attitudes, mediated
by intuitions about attitude reports, are widely relied upon in the philosophy of language
and mind. Anywhere such intuitions motivate philosophical views is a place to look for a
possible clarificatory role for our package of views. I expect that these views may further
transform discussions about functionalism in the philosophy of mind, about the problem
of logical omniscience for certain metaphysical views about propositions, about eavesdrop-
pingobjectionsagainstcontextualismaboutconditionalsandmodals, aboutcontextualism
about knowledge, about Frege puzzles, about linguistic observations concerning the distri-
bution of permissive, or “transparent”, reports, and maybe more. If the work in this
dissertation of developing the conceptually permissive attitudes view is successful, there is
more work to do.
3. Brief Summary of the Rest
I have introduced the main views defended in this dissertation in general terms, have
pointedoutsomeoffeaturesoftheseviewsthatcontrastwithcertainwidespreadphilosoph-
19
This is of course is not the only argument for externalism; but the sensitive attitudes view may have
consequences for other arguments as well, such as those drawing on Putnam (1975) and Burge (1979)’s
thought experiments involving intrinsic duplicates.
20
Forpreviousargumentsthatbeliefreportsdonotprovidedataaboutattitudeswhichareofsuchinterest
in the philosophy of mind, see Dummett (1996) and Cummins (1991).
15
ical and linguistic assumptions, and have pointed to some ongoing debates in philosophy
that these views could transform if they are correct. Now I describe how the rest of the
dissertation unfolds.
In chapter 2, I argue for the sensitive attitudes view. My argument proceeds by using a
variety of examples of permissive reports to discredit the main alternative to the sensitive
attitudes view, according to which permissive readings are not to be accounted for in
terms of a context-sensitivity in which attitude relation a report sentence serves to ascribe,
but are instead to be accounted for in terms of context-sensitivity in what propositions
are ascribed as attitude contents. In chapter 3, I consider one way of developing the
sensitive attitudes view, which I call the “contextual augmentation view”. I raise several
kinds of problems with simple ways of articulating this view. These problems inform the
remainder of the dissertation: any adequate account of permissive attitudes should be
able to resolve these di↵ erent apparent problems. (Accordingly, one thing I try to do in
chapters 5 through 7 is to show how a version of the contextual augmentation view that
is appropriately sensitive to the pragmatics of permissive reporting might handle these
problems.) Chapter 4 is a detour. Given the problems facing the contextual augmentation
view, I consider an alternative way of developing the sensitive attitudes view, which I
call the “permissive dispositions view”, according to which for one to bear a propositional
attitudetowardsapropositionconsistsinone’spossessingcertaindispositions,wherewhich
dispositions a subject must possess for a particular attitude report sentence to be true may
vary across with the contexts in which the sentence is used. Chapter 5 and 6 address the
pragmatics of permissive reports. In these chapters I do two things. First, I illustrate some
of the many ways in which whether a report sentence serves to make a permissive report
depends on certain features of conversational contexts; chapter 5 focuses on the pragmatic
importance of conversational goals, and chapter 6 focuses on the pragmatic importance of
conversational assumptions about what concepts subjects possess and how they use those
concepts. Second, in these chapters I try to show how one can develop a version of the
contextual augmentation view that avoids certain of the problems raised in chapter 3, by
framing such a view in a way that is sensitive to such pragmatic phenomena. Chapter
7 concludes the dissertation. There I further develop the contextual augmentation view,
drawing on lessons learned along the way, and speculate on how such a view might handle
some of the open problems left over from chapter 3.
16
Chapter 2. Permissive Inconstant Reports and the Strict View
1. Introducing the Strict View
MissScarletisunawarethatsomewhereinthemanortherearehiddenthreetwo-pound
statues made from moonstone of the purest blue adularesence. Not only does she lack any
evidence that there are precious artifacts hidden in the manor, but she also lacks any con-
cept of the property of being a moonstone or of the natural kind moonstone. She has never
encountered a moonstone, has never encountered the word ‘moonstone’ in speech or print
(nor has she encountered any non-English word designating these properties and kinds),
and in general she lacks any means of discriminating moonstones from non-moonstones.
She also lacks a concept of the property of adularesence, as did I before I started Googling
about moonstones. Intuitively, then, given Scarlet’s conceptual impoverishment, not only
is she not in a position to know that there are three artifacts made from moonstone of
pure blue adularesence hidden in the manor, but she is not in a position to believe this,
or, indeed, to bear any propositional attitude towards this proposition at all. Intuitively,
to believe a proposition one has to be able to think about the things that proposition is
directly about.
IwillregimentthisintuitivethoughtasaprinciplewhichIcall“thestrictview”ofbelief:
necessarily one believes a proposition only if one possesses concepts of that proposition’s
constituents.
21
While there has been some debate over such conceptual acquaintance con-
straintsforcertainspecialclassesofattitudeattributions,theguidingideabehindthestrict
viewhasremainedattractive, especiallyforpurely“general”beliefcontents, whichareonly
directly about properties and relations.
22
Jeshion (2010), for example, cites “widespread
agreement” in the philosophy of mind on this point: to bear an attitude towards a gen-
eral proposition ‘one must possess and grasp those constituent concepts in the general
proposition, must do so in the way they are structured in the proposition.’
This widespread agreement has given the strict view something like the status of a
conceptual constraint on theorizing about the attitudes. Part of the paradigm guiding
research in the philosophy of mind which is framed in terms of concepts and structured
thought contents is the view that one entertains (that is: believes, knows, hopes, fears,
doubts, etc.) a particular proposition only if one is in a position to deploy, hence possesses,
concepts of the objects or properties which that proposition is directly about. Indeed, as I
discussedintheintroductorychapter,thestrictviewimplicitlyfiguresincentralarguments
forlandmarkviewsinthephilosophyofmindandlanguage. Notonlywoulditbesurprising
if the strict view turned out false, but it would prompt a reappraisal of some prevalent
philosophical ideas.
21
So understood, the strict view is committed to the view that propositions are structured entities, things
with constituents. But the guiding idea behind the strict view remains intuitive even if we deny that view.
Where relevant I make mention how the arguments I give in this chapter relate to non-structured views of
propositions.
22
For debate about whether such conceptual constraints govern singular thought, see Jeshion (2002),
Recanati (2010), and Hawthorne & Manley (2012).
17
I argue that the strict view is false. In general, bearing an attitude which is ordinarily
expressible by attitude verbs like ‘believe’ towards some proposition does not require one
to use concepts of constituents of that proposition. The ways we attribute attitudes to
agents cannot be made consistent with the strict view. In particular, using a range of
examples, which I call “permissive inconstant reports”, I show that the most promising
accounts of these readings cannot be used to defend any general version of the strict view
which is both empirically adequate and of theoretical interest. What looks at first like a
conceptual constraint on empirical theorizing about attitudes in the end commits one to
linguistic theories that cannot be reconciled with a range of empirical data.
My central claim, then, is that we cannot hold onto this attractive idea on the most
natural ways of spelling it out. This is admittedly hard to accept, since the idea just
seems so right. I conclude by trying to show what is right about it, sketching alternative
ways of spelling out the idea which both are empirically adequate and respect the original
idea’s appeal. In place of the strict view, I defend a picture consisting of two compatible
ideas about attitudes, one linguistic and one metaphysical. The first idea is the sensitive
attitudes view, according to which attitude verbs are semantically underspecified in such
a way that distinct uses of a single attitude verb can serve to express distinct relations.
The second idea I call the conceptually permissive attitudes view, according to which some
relations which can be expressed by a use of ‘believe’ are conceptually permissive, where
an attitude relation R is conceptually permissive if for some agent a and proposition p it
is possible that a bears R to p but a lacks a concept of a constituent of p. Together these
views explain the problematic data I raise here. As I show in the concluding section, they
can be used to respect versions of the intuitive conceptual constraint which motivates the
strict view. For all I say here, these alternative ways of explicating our starting intuition
about concepts and attitudes could turn out true, but such principles di↵ er from the strict
view in ways which raise new questions and prompt us to revisit old assumptions about
the work which such conceptual principles can do for us in the philosophy of language and
mind.
2. Sensitive Contents Accounts of Permissive Reports
I begin by sketching a naive version of the style of linguistic argument I raise against
the strict view. Once the general argumentative strategy is clear, I move on to the paper’s
central data.
Our ordinary ways of attributing attitudes can seem not to respect the strict view.
Consider the following example:
Passage: Detective Black has been observing Plum, so is aware that he has pursued
some shadowy figures into the conservatory. Black knows both that these shadowy figures
are suspects in a murder investigation and that they have escaped into the secret passage.
Detectives Black and Grey are conversing; they know that Plum is unaware that there is a
secret passage, but Grey doesn’t know that Plum is searching around in the conservatory,
18
so Black assertively utters the following to Grey:
(5) Plum believes that every suspect in the secret passage is in the conservatory.
Black’s assertion is true even though Plum is known not to relevantly deploy a con-
cept of the property suspect-in-the-secret-passage. This report takes what I call
a “permissive reading” with respect to ‘secret passage’: the proposition which Black as-
serts with (5) is such that it is possible that it be true and that Plum lack a concept of
the property suspect-in-the-secret-passage.
23
But on a flatfooted interpretation of
Black’s report, she attributes a belief to Plum in a proposition containing the property
suspect-in-the-secret-passage as a constituent. Given this flatfooted way of inter-
preting Black’s report, the strict view of belief seems false.
This little case, of course, falls far short of a knockdown argument against the strict
view. We need not rest easy with the flatfooted interpretation; the strict view has an
illustrious history of support, which has motivated alternative ways of interpreting reports
like Black’s.
Russell (1910: 117) gives voice to something like the strict view when he says that
‘every proposition which we can understand must be composed wholly of constituents
with which we are acquainted’. This acquaintance principle was a central motivation
for Russell’s elegant and now-famous semantics for attitude reports, which accounts for
permissive readings like that of (5) in terms of relative scope. Russell’s scope theory is a
paradigmatic example of what I call “sensitive contents” views (which stand in contrast
to the sensitive attitudes view which I shall defend). Sensitive contents views make sense
of (5) by claiming, contrary to the flatfooted interpretation, that Black’s report does not
ascribe to Plum a belief in the proposition that every suspect in the secret passage in the
conservatory. Rather, according to sensitive contents views, di↵ erent uses of (5) can serve
to ascribe beliefs with di↵ erent contents to Plum, and this fact reconciles the example
Passage with the strict view.
Here’s how Russell’s version of the sensitive contents view works. First observe that
(5) admits of multiple interpretations. In addition to the interpretation adduced by the
naive argument, there is another on which Plum is represented as thinking of something
as a secret passage. Such a reading is more natural in di↵ erent contexts, such as one in
which Plum is known by the conversationalists to be aware that there is a secret passage
and to believe that parts of the secret passage are in the conservatory. These two readings
have gone by di↵ erent labels; sometimes they are called “de re” and “de dicto” reports,
23
Of course, in this scenario Plum may actually possess a concept of this property. My claim is that
since Plum can verify this report without relevantly deploying such a concept, intuitively there is a possible
circumstanceinwhichthereportmadewiththissentenceinthiscontextistrueyetPlumaltogetherlackthe
relevant concept. Despite my speaking in this way here, however, we will see later in the dissertation that
this way of characterizing permissive readings is too coarse. I don’t hang too much on this characterization
of permissiveness and neither should you; my attempts at giving an accurate yet general characterization
of this phenomenon lead eventually to persnickety mouthfuls that I prefer to avoid while I can.
19
sometimes they are called “transparent” and “opaque”.
24
For various reasons, some of
which we will encounter below, I call the reading of (5) in Passage “permissive” with
respect to suspect-in-the-secret-passage and the other reading “loaded” with respect
to suspect-in-the-secret-passage.
With this ambiguity in focus, an adherent to the strict view can account for the per-
missive reading of (5) by following Russell (1905) in explaining the permissive and loaded
readings of ‘every suspect in the secret passage’ in terms of that phrase’s relative scope.
Roughly, relative scope is a matter of interpretive order: on one reading, we interpret ‘sus-
pect in the secret passage’ before ‘believes’; on the other, we interpret that phrase after
‘believes’. We can represent these readings as
(5W) [every suspect in the secret passage]
x
[Plum believes [x is in the conservatory]]
(5N) Plum believes [[every suspect in the secret passage]
x
[x is in the conservatory]]
In (5W) we interpret ‘suspect in the secret passage’ before ‘believes’; the corresponding
proposition represents every suspect in the secret passage as being believed by Plum to be
in the conservatory. So Plum isn’t represented as believing one proposition about every
suspect in the secret passage, but as believing several propositions, one for every suspect.
None of these propositions contain suspect-in-the-secret-passage as a constituent, so
that Plum can verify such a reading of (5) without possessing or deploying a concept of
that property does not violate the strict view.
In (5N), we interpret ‘suspect in the secret passage’ after ‘believes’. The corresponding
proposition represents Plum as believing one proposition about every suspect in the secret
passage, namely that every suspect in the secret passage is in the conservatory. This
situates suspect-in-the-secret-passage inside the proposition which Plum is said to
believe; so, given to the strict view, requires that Plum be conceptually acquainted with
suspect-in-the-secret-passage. The defender of the strict view can admit as much
by associating (5N) with the loaded reading. On Russell’s sensitive contents view, which
content a use of a report sentence serves to ascribe to a subject is context-sensitive and
this context-sensitivity explains the variation between permissive and loaded readings.
The naive way of pressing this linguistic argument, then, does not itself discredit the
strict view. Now let’s turn to more troubling ways of pressing this style of argument.
3. Some Initial Objections to the Strict View
My primary aim in this chapter is to discredit the strict view. My strategy is to show
that permissive readings cannot be accounted for by sensitive contents views, according to
whichthedi↵ erencebetweenpermissiveandloadedreadingsofattitudereportsisexplained
24
This use of ‘transparent’ and ‘opaque’, from Fodor (1970) and Percus (2000), is related to but distinct
from the notion opaque sentential contexts derived from Quine (1956). Even if ‘that’-clauses of attitude
reports are opaque in Quine’s sense, in Fodor’s sense there are still transparent readings of some uses of
expressions in some reports. To avoid terminological confusion, I stick with “permissive” and “loaded”.
20
by context-sensitivity in which attitude contents are ascribed by di↵ erent uses of a single
attitude report sentence. The central examples I focus on (starting in §4) are ones in
which some descriptive linguistic material occurring within the ‘that’-clause of a report
sentence cannot be interpreted outside the scope of the sentence’s main attitude verb
without predicting erroneous truth-conditions for the permissive report made.
Inadditiontomycentralexamples,thereareotherswhichalsoprovideevidenceagainst
the strict view in the presence of certain assumptions about the nature of thinkers and
thought contents. Since these extra assumptions may be and have been disputed, I won’t
rest my case on these extra examples, but they nevertheless are worth raising up front.
Even should I fail to achieve the ambitious goals of the rest of the chapter, the examples
in this section shift some burden of proof onto certain ways of holding to the strict view.
These extra examples fall into two categories. The first category of example, which I
call “controversial permissive reports” involve liberal uses of attitude reports on which a
subject is reported to bear an attitude towards some proposition which bears no obvious
resemblancetothesubject’sloadedthoughtcontents,eitherbecausetheystructurallydi↵ er
fromthosecontentsorbecausethesubjectintuitivelylacksloadedattitudesaltogether. The
second category of example concerns the cognitive significance of object-directed thoughts
based on perceptual or causal acquaintance.
3.1. Controversial Permissive Reports
My first category of example cannot be accounted for using the scope theory of per-
missiveness, nor indeed using any sensitive contents view that we will encounter in this
chapter. I call these examples “controversial permissive reports” because, in the face of
such problematic examples, some philosophers choose to reject that they are literally true
or that they should count as data about the “nature of propositional attitudes”. I present
theseupfronttoillustratethatdefendingthestrictviewatleastrequiresonetobeselective
about which attitude reports are to be respected as data about which attitudes subjects
possess. While I think that it is di cult to ignore these cases in a principled way, I won’t
rest my criticism of the strict view on them.
Ball: Mrs. Peacock is organizing a ball at the manor, she has delegated the assigned
seating to Mrs.White. Peacock informs Detective Black that she believes that every musi-
cian at the ball is seated at table nine. Detectives Black and Grey both know that Peacock
is unacquainted with Scarlet: Peacock doesn’t know Scarlet by any name, couldn’t pick
her out in a line up, etc. The detectives also know that Scarlet is a musician. During the
ball, Grey asks Black where Scarlet is sitting. Black responds by sincerely uttering:
(6) Peacock believes that Scarlet is seated at table nine.
What Black asserts in uttering (6) sounds true and appropriate. But it is as intuitive
that Peacock does not deploy a concept of Scarlet in this scenario. Given this, it is possible
that the proposition Black asserts with (6) be true even though Peacock altogether lacks
21
a concept of Scarlet.
25,26
So apparently the strict view is false.
The scope theory of permissiveness also appears to be of little help in this case. The
best chance for the scope defense here is to interpret ‘Scarlet’ as scoping wide of ‘believes’,
as in
(6W) [Scarlet]
x
[Peacock believes [x is seated at table nine]]
But this hardly delays the problem. For the proposition corresponding to (6W) still
attributesasingularbelieftoPeacockwithScarletasaconstituent. Theproblemhereisn’t
just that Peacock doesn’t have a loaded belief requiring a particular concept of Scarlet,
but that, given the setup of Ball, Peacock lacks, or fails to deploy, any concept of Scarlet
whatsoever.
Next consider a classic kind of case involving chess-playing programs:
Chess: ProfessorPlumisplayingchessagainstthecomputerprogramLeela. Detectives
Black and Grey are looking on. After Plum moves his bishop to d4, Leela moves a bishop
to d7. Black turns to Grey and sincerely utters:
(7) Leela is boxing out Plum’s knight because it thinks that Plum wants to fork the
queen and rook.
WhatBlackassertswith (7) soundstrueand appropriate. Butitisasintuitivethat the
chess-playing program Leela lacks a concept of Plum. Indeed, one could argue that Leela
lacksanyloadedattitudeswhatsoever. Onthemoststraightforwardwayoftakingthestrict
view, it makes the false prediction that the truth of (7) requires Leela to possess a concept
of Plum (and concepts of the relation wanting, of the queen and of the rook). The scope
theory of permissiveness also is of no use here; interpreting any or all of the constituents
of the ‘that’-clause of (7) as scoping wide of ‘thinks’ still would serve to attribute a belief
to Leela in a proposition containing constituents for which Leela lacks concepts. So Chess
appears to show that the strict view is false.
27
Both of these examples tell against the strict view, but both might plausibly be dis-
missed as abnormal in ways that render them theoretically irrelevant. In debates about
25
Specifying such a possibility may be a non-trivial a↵ air. For example, in the scenario as described, we
can suppose that Scarlet is the third tallest musician. But presumably Peacock possesses the concept the
third tallest musician, which is a concept “of” Scarlet. Finding a scenario where Peacock has no concept of
Scarlet, in this sense, would require special care. I take it, however, that even should this di culty prove
insurmountable, in some cases of permissive reports it would be of no consolation to the defender of the
strict view. What matters in this kind of case is that Peacock intuitively does not need to useaconceptof
Scarlet, or does not bear some attitude relation the instantiation of which requires that Peacock possess a
concept of Scarlet.
26
An example of this general shape is given by Recanati (2012) and is discussed in Blumberg & Holgu´ ın
(2018).
27
In chapter 4 I will present further examples of this kind which illustrate that we can make apparently
truly ascribe attitudes to simple animals, neurons, and groups.
22
singular or “de re” thought, some already defend that one can possess a thought which
is singular with respect to some object while lacking “acquaintance” with that object.
28
Others insist upon a distinction between “genuine” and “pseudo” de re reports, taking
examples like Ball to be instances of the latter.
29
Given the incompatibility of so-called
“pseudo de re” reports with the strict view, such theorists choose to disregard such reports
as evidence of singular thought.
30
Once one takes this selective posture, it’s no stretch
to further disregard examples like Chess as providing evidence about what things have
“genuine” thought contents. Belief attributions like (7), one could say, are just an example
of some kind of “loose talk”, so, the thought goes, can be safely ignored by advocates of
the strict view.
In what follows, I’m going to make things harder on myself by spotting the defender of
thestrictviewthisdismissalofsuchcontroversialcases. Evenifwefocusonmorestandard
instances of permissive reports, there is a range of permissive reports that tells against the
strict view.
3.2. The Cognitive Significance of Object-Directed Thoughts
My main argument against the strict view, which begins with example (5) above and
which I will pursue in §3 and beyond, focuses on permissive readings of general or de-
scriptive thoughts. It looks at first like Black’s use of (5) serves to ascribe a thought
content to Plum which is directly about the property suspect-in-the-secret-passage,
even though intuitively Plum lacks or fails to deploy a concept of that property, contrary
to the strict view. But there are also examples available concerning non-descriptive, or
“object-directed” thoughts, which tell against a certain version of the strict view.
31
To
these I now turn.
My first example contrasts a report sentence as used in two contexts.
32
First consider:
Game Show
1
: There are four doors to the ballroom, which the guests at the manor
are using as part of a simple game show. Placed behind some of the doors are prizes. A
contestant in the game show picks one of the four doors to open and so wins whatever
stands just behind. It’s Scarlet’s turn, and she deliberating over her choice. She tells
Black that, given her evidence about the structure of the game, she thinks that whatever is
behind door #3 might be a shiny new Trek bicycle. Black and Grey, having looked behind
each door, know that the thing behind door #3 in fact is a shiny case of Coca Cola. In
discussing Scarlet’s choice with Grey, Black assertively utters:
28
See Jeshion (2002) and Hawthorne & Manley (2012).
29
Kaplan (1989) introduces this terminology of the “pseudo de re”, claiming that such reports fail to
“pose any issues of su cient theoretical interest”. I’ll note here that even the sentence to which this
footnote is attached is plausibly an example of such a “pseudo” de re report; this reporting is so natural
and ubiquitous that I think it foolhardy to presume they lack theoretical interest.
30
Recanati (2012) adopts this methodological approach of ignoring such reports.
31
I take the label ‘object-directed thoughts’ from Azzouni (2011).
32
Thanks to Barry Schein for this example.
23
(8) Scarlet thinks that a case of Coca Cola might be a shiny new Trek bicycle.
Intuitively Black’s report takes a permissive reading with respect to the indefinite ‘a
case of Coca Cola’.
Now contrast Game Show
1
with:
Game Show
2
: The situation is just as in Game Show
1
, save for that before Scarlet’s
turn, she sneaks a peak from a distance at the prizes behind the various doors. All she
gets from her fleeting and distant glimpse at the prize behind door #3 is that the prize lets
o↵ a glimmering shine. Knowing a shiny new Trek bicycle is one of the potential prizes,
she tells Black that she thinks that the thing behind door #3 might be a shiny new Trek
bicycle. Black and Grey, having looked behind each door, know that the thing behind door
#3 in fact is a shiny case of Coca Cola. In discussing Scarlet’s choice with Grey, Black
assertively utters:
(8) Scarlet thinks that a case of Coca Cola might be a shiny new Trek bicycle.
As before, Black’s report takes a permissive reading with respect to ‘a case of Coca
Cola’ here.
Now, this pair of reports exhibits an apparent contrast in the cognitive significance
of Scarlet’s loaded thoughts. For those who take perceptual or causal acquaintance as a
condition on object-directed thoughts, in Game Show
2
Scarlet has an object-directed, or
de re, loaded thought, whereas in Game Show
1
, she does not.
33
Regarding Game Show
2
,
such a strict attitudes theorist who accepts that objects may be proper constituents of
propositions thought, will want to say that (8) reports a singular object-directed thought
towards a proposition containing a case of Coca Cola as a constituent. Black is reporting
a loaded thought directly about a (certain) case of Coca Cola. But apparently this kind of
strict theorist will not say this is so in Game Show
1
: Scarlet is not perceptually acquainted
with, so is not in a position to have thoughts directly about, any particular case of Coca
Cola.
For Game Show
2
, the defender of the strict view may well resort to a wide scope
interpretationoftheindefinite‘acaseofCocaCola’inspecifyingScarlet’sthoughtcontents:
(8W) [a case of Coca Cola]
x
[Scarlet thinks that x might be a shiny new Trek bicycle]
According to (8W), there is a case of Coca Cola such that Scarlet thinks (in a loaded,
object-directed way) that it might be a Trek bike. This regimentation appropriately repre-
sents Scarlet as having a singular object-directed thought about a particular case of Coca
Cola.
What is this species of strict theorist to say, however, about Game Show
1
? Clearly
Scarlet does not have any “de dicto” thought about something which is “a case of Coca
33
For such versions of the strict view, see for example Donnellan (1977), Bach (1987), Salmon (1987),
Recanati (1993), and Recanati (2010).
24
Cola”. So this strict theorist seems forced to give the same regimentation, (8W), of the
report made about Scarlet in Game Show
1
. But, on the going version of the strict view
which accepts a causal/perceptual constraint on object-directed concepts, this is hard to
accept. Scarlet apparently lacks any direct, non-descriptive concept, of the case of Coca
Cola. So she is no position to have a singular object-directed attitude about that case of
Coca Cola, contrary to the regimentation given by (8W).
Atthispoint,onecommittedtothiskindofperceptualconstraintonattitudepossession
apparentlyseemsforcedtosaythattheremustbetwokindsofthinking: onewhichanswers
to the strict view and one which doesn’t. But this in e↵ ect is to give up the strict view as a
general constraint on attitude possession and instead to endorse a version of the “sensitive
attitudes” view which I propose below in §5. On that style of view, there are di↵ erent
attitude relations contributed by a use of an attitude verb like ‘thinks’, some abiding by
the strict view and some not.
In the face of this pair of examples, an alternative approach for the strict theorist is
insteadtorelaxtheconstraintstheyacceptonobject-directedconcepts. Onthisalternative
way of adopting the strict view, one allows that a subject may possess a singular concept
of an object, such as a case of Coca Cola, without being perceptually or otherwise causally
acquainted with that object.
34
Onceonetakesthisstep,(8W)maywellserveasaregimentationofthetruthconditions
of both uses of (8) in Game Show
1
and Game Show
2
. In each case, Scarlet has a singular
thought about a certain case of Coca Cola; the di↵ erence between these two examples lies
in the kind of singular concept which Scarlet possesses of the case of Coca Cola across
the examples. There is, of course, still an empirically relevant di↵ erence between the two
examples that needs to be accounted for. But the more conceptually liberal version of
the strict view can plausibly punt here to pragmatic explanation. Plato, you, and the
green undergraduate who is only aware that someone wrote Nicomachean Ethics may each
count, onthisview, asbelievingsingularlyofAristotlethatheisaphilosopher, thoughyou
all may di↵ er in the conceptual means by which you possess this singular thought. That
in itself does not discredit a regimentation of the common content of these thoughts as
including Aristotle himself as a propositional constituent. This conceptually liberal strict
view can insist that whatever account of any empirically relevant di↵ erences between these
three singular thoughts will extend to Game Show
1
and Game Show
2
.Sowhilethese
examples put pressure on certain ways of upholding the strict view, they do not directly
discredit the view as such.
4. Permissive Inconstant Reports
InthissectionIraisethecentraldataofthischapter,whichIcall“permissiveinconstant
reports”. In building towards this data, I first consider a sensitive contents account of
permissive reports, which I call the “plurality theory” of permissiveness, which apparently
34
See Hawthorne & Manley (2012) for a defense of this approach to singular thought. See also Sosa
(1970) and Jeshion (2002).
25
can handle some problems facing the simple scope theory of permissiveness. I consider and
reject various ways of articulating the plurality theory of permissiveness.
4.1. The Simple Plurality Theory
Even if there are principled grounds which can be used to set §3’s examples aside,
there are less controversial examples of permissive readings which the scope theory of
permissiveness still cannot be used to avoid. I’ll start with a simple example which directly
discredits the scope theory of permissiveness, then move on to what I call “permissive
inconstant reports”, which further discredit a natural modification of the scope theory.
35
These examples don’t involve typical singular thought attributions or subjects who lack
concepts altogether.
Innocence: Professor Plum is agnostic about who committed the murder: he neither
believes that Green is the murderer nor believes that Green is innocent, and similarly for
everyotherguest. Butheknowsthatonlyonepersoncommittedthemurder,sohebelieves
that most of Green, White, and Scarlet are innocent. Detectives Black and Grey know
something that Plum doesn’t: Green, White, and Scarlet are all and only the members
of a conspiracy to blow up police headquarters. Having just interrogated Plum about
his opinions concerning who the murderer might be, detective Black sincerely utters the
following to detective Grey as they are discussing the conspirators:
(9) Plum believes that most conspirators are innocent.
On this use of (9), Black’s report is permissive with respect to the concept associated
with ‘conspirators’. Intuitively Plum can verify Black’s assertion without thinking of any-
thing as a conspirator. But we cannot capture this reading by scoping ‘conspirators’ wide
of ‘believes’ as in
(9’) [Most conspirators]
x
[Plum believes [x is innocent]].
Thepropositioncorrespondingto(9’)representsmostconspiratorsxasbeingsuchthat
Plum believes x is innocent; but Plum doesn’t believe of any conspirator in particular that
they are innocent. Plum’sbelief concerns theplurality of conspirators together; hebelieves
that most of them are innocent. But we cannot capture Plum’s belief by interpreting the
‘most’ quantifier as narrow to ‘believes’ without including conspirator as a constituent
of what Plum is said to believe, which, given the strict view, would erroneously entail that
Plum must possess a concept of that property in verifying the report. So the simple scope
treatment of this case is incorrect.
Consider another kind of example which directly discredits the scope theory of permis-
siveness which involves an attitude verb other than ‘believe’:
35
Fodor (1970)’s famous examples of non-specific transparent readings show that the transparency and
specificity of indefinites cannot both be accounts for in terms of scope. The Innocence example below
directly shows that transparent (permissive) readings cannot be explained in terms of scope.
26
Bridge: The detectives know that Colonel Mustard is hosting his weekly bridge game
and has invited every guest at the manor who plays bridge to the game. Mustard tells
detective Black that he fears that every invitee will want to play bridge, in which case
there will be an extra player, making for an awkward bridge night. He hopes that no more
than three invitees will want to play. The detectives know something Mustard doesn’t: all
and only the invitees are undercover cops. In discussing the bridge game, Black sincerely
utters the following to Grey:
(10) Mustard fears that every undercover cop will want to play bridge.
What Black asserts here sounds true and appropriate. But intuitively Mustard does
not deploy a concept of undercover-cop in this scenario. It is possible that the propo-
sition Black asserts with (10) be true even though Mustard altogether lacked a concept
of undercover-cop. So on the scope theory, to account for the permissive reading of
‘undercover cop’, this phrase should scope wide of ‘fears’ as in:
(10’) [Every undercover cop]
x
[Mustard fears [x will want to play bridge]].
But this delivers the wrong reading. The proposition corresponding to (10’) represents
every undercover cop x as being such that Mustard fears that x will want to play bridge.
But Mustard has no such set of fears: he does not fear of any individual invitee that they
will want to play bridge, but rather has a fear concerning all of the invitees (undercover
cops) taken together. So again, we have a simple case which illustrates that the simple
scope treatment cannot account for all permissive readings.
This isn’t the end of the road for sensitive contents accounts of permissive readings. In
these examples, there is a felt tension between the determiner and the bare noun phrases
with which they combine.
36
For example, in Innocence we intuitively want to interpret
the determiner ‘most’ as scoping narrow to ‘believes’, but want to interpret the bare noun
phrase ‘conspirators’ as scoping wide of it. Plum does think in terms of the higher-order
relation most,but doesn’t think in terms of conspirator; given the strict view, the former
is a constituent of Plum’s reported belief but the latter is not. But if we assume that the
entire quantifier phrase must be interpreted as a whole either before or after the attitude
verb, we are apparently forced to give up on explaining permissive readings in terms of
scope.
A response suggests itself: deny that a quantifier phrase must be interpreted in its
entirety, with both its determiner and bare noun phrase, at one location in logical form.
37
Instead, one can allow that the determiner and bare noun phrase may scope wherever we
36
Whereabarenounphraseisanounphraselackingadeterminer: ‘everyundercovercop’and‘undercover
cop’ are both noun phrases, but the latter is a “bare noun phrase”.
37
The first suggestion of something like this I’ve found is Hellan (1980). A similar approach is discussed
by von Fintel & Heim (2011). Szabo (2010) adopts something like it to account for specific opaque reports,
but he allows the determiner phrase alone to scope wide of attitude verbs.
27
need them independently. To see how this might work, contrast the typical wide scope
interpretation (9’) with an unorthodox regimentation on which the bare noun phrase alone
scopes above ‘believes’ without the determiner:
(9’) [Most conspirators]
x
[Plum believes [x is innocent]]
(9”)9 Y[conspirators](Y) [Plum believes [[most of the Ys]
x
[x is innocent]]]
The underlined segment of (9’) corresponds to the property had by any object o such
that Plum believes o is innocent; the quantifier phrase in (9’) contributes the higher order
property which holds of any property which is instantiated by most conspirators. (9’)
predicates the latter higher order property of the former first-order property. We saw that
this delivers the wrong reading for (9) in Innocence.
By contrast, since the variable ‘Y’ in (9”) cannot range over individual objects, the
underlinedsegmentin(9”)determinesadi↵ erenttypeofproperty. IsaidbeforethatPlum’s
reportedbeliefconcernsthepluralityofconspiratorsratherthananyparticularconspirator.
Given this, we could take ‘Y’ to range over pluralities.
38
If it does, the underlined segment
of (9”) determines the property had by any plurality F such that Plum believes that most
of the Fs are innocent. Since ‘conspirators’ takes widest scope in (9”), we can say that it is
evaluatedatthecontextworld,@,hencecontributestheplurality conspirators-in-@ofthings
which actually have the conventionally associated property conspirator.Theresulting
truth conditions for (9”) are just what we want: roughly, the plurality conspirators-in-@
is such that Plum believes that most elements of it are innocent.
39
Since cognizing this
plurality need not involve cognizing the property conspirator, there is no violation of
the strict view. To account for loaded readings of attitude reports, one can claim that a
bare noun phrase which scopes narrow to an attitude verb contributes its conventionally
associated property. Of course, in a fully developed version of this plurality theory more
would need to be said about how these readings arise compositionally, but if this could
be done, then something like the plurality theory could account for Innocence.Permissive
readings don’t yet give us reason to abandon the strict view; one can advance a sensitive
38
Itwon’tdotolet‘Y’rangeoverproperties,sincetheresultinginterpretationwouldincludeconspirator
as a constituent of what Plum is represented as believing, requiring Plum to cognize that property, which
is what we’re trying to avoid. Nor will it help in the present setting to say that scoping ‘conspirator’ wide
of ‘believes’ has it contribute a world-relativized property, conspirator-in-@. In the presence of semantic
assumptions like those of von Fintel and Heim (2011)’s intensional semantics, that kind of suggestion is
more promising; but given the strict view, it is as hopeless as the idea of taking ‘conspirators’ to contribute
conspirator. For just as Plum can verify (9) without possessing a concept of conspirator,hecandoso
without possessing a concept of conspirator-in-@.
39
Other semantic accounts of permissiveness have been developed in non-structured frameworks. Percus
(2000) explains permissive reports by indexing expressions to situation variables bound by attitude verbs.
I discuss this approach below in §4.3. Winter (1997) explains permissive readings in terms of choice func-
tions. Neither is obviously amenable to a structured approach, given the strict view, since each apparently
would require that subjects cognize propositional constituents which intuitively they need not: certain
situations/worlds or functions.
28
contents explanation of permissiveness using semantic machinery other than simple scope
ambiguities of quantifier phrases taken as whole linguistic chunks.
Before raising my special data against the plurality theory, it’s worth noting that this
particular way of spelling out the basic plurality-theoretic idea won’t even handle simple
variations on Innocence or Bridge. The analysis can be shown to be too weak with a slight
tweak on Innocence:
Innocence
2
: As before, Plum is agnostic about who committed a particular type of
crime. But in this case, for each individual guest at the manor, he suspends judgment
about whether they are a thief. He neither believes that Green is a thief nor believes that
Green is innocent of being a thief, and similarly for every other guest. He tells detective
Black that he believes that most people in the billiard room are thieves but also believes
that most women in the billiard room are innocent. Detectives Black and Grey know that
all and only the guests in the billiard room are conspirators; they further know that Plum
is unaware that there is any conspiracy. After this discussion, Black sincerely utters the
following to detective Grey as they are discussing the conspirators:
(9) Plum believes that most conspirators are innocent.
This case supports the subtle judgment that Black’s report is permissive with respect
to ‘conspirators’ yet is false. Plum is in no way represented by Black’s report as thinking
of anything as a conspirator, but Black’s assertion is nevertheless fails to be true. Plum
doesn’thaveapermissivebeliefthatmostconspiratorsareinnocent,butapermissivebelief
that most women conspirators are innocent.
To capture the permissive reading here, the plurality theory just sketched will interpret
‘conspirators’ as scoping wide of ‘believes’, as before:
(9”)9 Y[conspirators](Y) [Plum believes [[most of the Ys]
x
[x is innocent]]]
But this fails to capture the false reading of (9) in Innocence
2
. For in the scenario
described, there is a plurality of conspirators such that Plum believes that most of them
is innocent: the plurality of women conspirators. So even the simple plurality theory
presentedneedsfurtherrefinementtomakesenseofsimpleexamplesofpermissivereadings.
I’ll consider one possible tweak to the plurality view which may avoid this problem.
One might hope here that appealing to the general phenomenon of nominal restriction,
as described in Stanley (2002), can resolve the problem. Black is not making an assertion
about most conspirators in the universe, but rather most conspirators who are party to the
conspiracy to blow up police headquarters and who are staying at the manor, or something
like that. But restricting the application of ‘conspirators’ to a plurality meeting that
description runs into the same problem: there is a plurality of such conspirators such that
Plumbelievesthatmostofthemisinnocent: thewomenconspirators. Socovertrestriction
in itself won’t immediately help the plurality theory.
To handle this problem, once we invoke covert restriction of this kind, we can instead
29
e↵ ectively cast the plurality theory as a demonstrative theory: the regimentation of (9)
would become something like [Those conspirators]
Y
[Plum believes [[most of the Ys]
x
[x is
innocent]]]. To do this, we can follow King (2008)’s theory of demonstratives in allowing
that covert restriction can involve directly referential content, so that ‘conspirators’ is
covertly restricted by an accompanying directly referential expression. This tweak requires
somespecialassumptionsabouthowtherelevantcovertrestrictionofthebarenounphrase
is to work, but it may get around the problem raised by Innocence
2
. I won’t bother
investigating this particular way of cashing out the plurality view here, since the basic idea
faces more direct problems, to which I now turn.
4.2. The Problem of Permissive Inconstant Reports
In this section I present the central examples of this chapter, versions of which I think
cannot be accommodated by any sensitive contents view. I start by working through
severalexamplesofwhatIcall“permissiveinconstantreports”, showinghowtheydiscredit
the plurality theory. I give several of these examples both to show the generality of the
phenomenon and to make vivid for the reader how the examples work. The rest of the
chapteraimstodiscreditalternativesensitivecontentsviewsbybuildingo↵ ofthesecentral
examples.
Permissiveinconstantreportsinvolveanotherinterpretativefeatureofbarenounphrases
which manifests in sentence involving modal and temporal expressions. Consider
Marked: Yesterday Scarlet found several stacks of bills, twenties and hundreds, in
the library. This morning she returned to the library to find that the bills are all gone.
But this afternoon she saw Mr. Green carrying two briefcases, one brown and one black.
She thinks that the bills are all in the briefcases; she’s not sure which bills are in which
briefcase—perhaps the hundreds are all in the black briefcase, perhaps the twenties and
hundreds are mixed between the two briefcases. Scarlet informs detective Black of all this.
The detectives know something Scarlet doesn’t: a cop cracked into the black briefcase and
marked all of the bills inside. So they know that whichever bills ended up in the black
briefcase are all and only the marked bills.
Detective Black sincerely utters the following to detective Grey:
(11) Scarlet believes that every marked bill might be a hundred.
Here ‘believes’ is permissive with respect to ‘marked bill’; Scarlet isn’t represented as
thinking of the bills as marked. So according to the plurality theory, ‘marked bill’ scopes
wide of ‘believes’:
(11W)9 Y[marked bill](Y) [Scarlet believes [might [every Y]
x
[x is a hundred]]]
But this delivers the wrong reading. For across the possibilities relevant to Black’s use
of‘might’, di↵ erentbillsareintheblackbriefcase. Thereisonepossibilitycompatiblewith
Scarlet’s evidencein which thebills in the black briefcase areall twenties, one in which half
30
are hundreds, etc. Intuitively, what Scarlet is reported to believe by (11) is true because
there is a possibility compatible with her evidence in which the things which are bills in
the black briefcase at that possible state are all hundreds. If it turns out that in fact only
half of the marked bills (those in the black briefcase) are hundreds, the belief attributed to
Scarlet is not thereby falsified: it remains true that it might be (given Scarlet’s evidence)
that every marked bill was a hundred, even if it is false that it might be that every member
ofthepluralityofbillswhichactually aremarkedbillsisahundred. ThisisbecauseScarlet
is represented as believing something about what might happen to be bills in the black
briefcase (the marked bills), not as believing something about how things might stand
with respect to those things which actually satisfy the descriptive condition associated
with ‘marked bill’.
40
I’llsaythat‘markedbills’is“modallyconstant’withrespecttoanoccurrenceof‘might’
if it applies to the same things at every world relevant to the interpretation of that occur-
rence of the modal expression; ‘marked bills’ is “modally inconstant” if it doesn’t apply
to same things across those possibilities. Since Scarlet is reported to believe something
which is true if there is some world w compatible with her evidence such that every thing
which is a marked bill in w is a hundred in w, ‘marked bill’ is modally inconstant as used
in Marked. But since the plurality theory secures the permissiveness of ‘believes’ with re-
spect to ‘marked bill’ by making Scarlet’s belief about one particular plurality, this forces
a modally constant reading on which ‘marked bill’ applies to the same things, those which
are among the plurality of actually marked bills, across the possibilities relevant to ‘might’
in (11). So the plurality theory is false.
Permissive inconstant readings are quite general; the phenomenon isn’t due to some
particular quirk of epistemic modality. To illustrate the breadth of the phenomenon, I’ll
give a few more examples which concern di↵ erent kinds of inconstancy.
Bugged: Professor Plum has noticed that for the last few nights certain statues have
beenmovedfromoneroomtoanother. Thebustsofcomposersinthemanorkeepchanging
and ending up in di↵ erent rooms: two nights ago Bach was in the study, then last night
he was nowhere to be found, and tonight he is in the billiard room. Plum thinks that the
40
There is a response here which may work against cases like Marked which involve inconstancy with
respect to epistemic modals. One could insist that (11W) does indeed capture the correct truth conditions
for (11) in Marked by claiming that there is an epistemic possibility for Scarlet at which the plurality of
marked bills, which may not all be hundreds, are yet hundreds. The way one may do so is by looking
to parallel examples involving directly referential expressions. Suppose Ms. Scarlet sees the back of Mr.
Green, but she can’t tell if it is Green or Mustard; one could assert something true with ‘Scarlet believes
that Green might be Mustard’. But this apparently requires that there be an epistemic possibility for
Scarlet at which Green is Mustard. Whatever analysis makes sense of this could presumably also make
sense of there being an epistemic possibility for Scarlet at which every element of plurality of non-hundreds
is a hundred. I myself doubt that this line could make sense of all permissive inconstant reports involving
epistemic modals, but I won’t debate the issue here. The response won’t extend to the other examples of
permissive inconstant reports below.
31
movements of these busts follow a pattern: he thinks that every Tuesday every bust of a
composer on display in the manor is in the east wing. Detective Black knows all of this.
ButthedetectivesknowsomethingPlumdoesn’t: eachnight, thedetectivesselectdi↵ erent
from a vast collection of busts, bug some of them, and then place the busts, both bugged
and non-bugged, throughout the manor. Every Tuesday, they put microphones only inside
busts of composers on display in the manor. They don’t always pick the same busts of
composerstoplacearoundthemanoronTuesdays, andonnightsotherthanTuesdaysthey
bug other busts (Wednesdays is poets, Thursdays is presidents, etc.) The only rule the
detectives follow is that on Tuesdays only busts of composers are bugged. They both know
that Plum is unaware that any busts are bugged, and are discussing Plum’s suspicions
about the busts’ movements.
Detective Black sincerely utters one of the following to detective Grey:
(12) Plum thinks that on Tuesdays every bugged bust is in the east wing.
(13) Plum thinks that on every Tuesday every bust that is bugged on that day is in the
east wing on that day.
Here ‘thinks’ is permissive with respect to ‘bugged bust’; Plum isn’t represented as
thinkingofanythingasabuggedbust, sincethedetectivesknowthatPlumisunawarethat
any busts contain microphones. Intuitively, he could verify (12) without even possessing a
concept of bugged-bust. So on the plurality theory, ‘bugged bust’ scopes wide of ‘thinks’:
(12W)9 Y[bugged bust](Y) [Plum thinks [on Tuesdays [every Y]
x
[x is in the east
wing]]]
But this delivers the wrong reading. For ‘bugged bust’ is also temporally inconstant
with respect to ‘on Tuesdays’; the truth of what Plum is said to believe depends on how
things stand with respect to whatever on Tuesdays are bugged busts, not on how things
stand on Tuesdays with respect to what are now bugged rooms. Were Black’s assertion
true, we would not count Plum as having a false thought just if it were false that on
Tuesdays everything which is a bugged bust at the time of detective Black’s use of (12)
isn’t in the east wing. But the plurality theory secures the permissive reading of ‘thinks’
withrespectto‘buggedbust’byforcingthelatterphrasetoapplytothesamethingsacross
the times relevant to ‘on Tuesdays’. The proposition corresponding to (12W) represents
Plum as suspecting something about what on Tuesday be the case concerning the plurality
of things which are bugged busts at the time of utterance. So the plurality theory cannot
account for this example of a permissive inconstant reading.
We can also get permissive inconstant reports with temporal expressions using singu-
lar definite descriptions. The phenomenon is not inherently related to many-membered
pluralities.
Debate: There is a regular debate competition at the manor between two teams, the
Stars and the Stripes. At each competition, di↵ erent individuals are elected to be the
32
captains of the Stars and of the Stripes. Green tells detective Black that he believes that
whenever there is a debate, the captain of the Stripes team (whoever it is that week)
makes the best argument. Detectives Black and Grey know something Green doesn’t:
each competition, the captain of the Stripes (whoever it is that week) is bribed to throw
that competition. The detectives know that Green is unaware that anyone is ever bribed
to throw any competition. In discussing the regular debate competition, Black assertively
utters the following to Grey:
(14) Green believes that whenever there is a debate, the debater bribed to throw the
competition makes the best argument.
What Black asserts here is both true and appropriate, but Green intuitively does not
possessordeployaconceptofanypropertyrelevantlylikedebater-bribed-to-throw-the-
competition. Inaccountingforthispermissivereadingofthedefinite, thepluralitytheory
interprets its bare noun phrase as scoping wide of ‘believes’, as in:
(14W)9 Y[DBTC](Y) [Green believes [whenever there is a debate [the Y]
x
[x makes
the best argument]]]
But this delivers the wrong reading. Green does not believe of the single-membered
plurality which presently is the debater bribed to throw the competition that, whenever
there is a debate, they make the best argument. The definite here is temporally inconstant
with respect to ‘whenever there is a debate’; there is no one person whom Green believes
makes the best argument whenever there is a debate. Whether the relevant determiner
phrase concerns an individual or mutli-membered plurality, the plurality theory gets the
wrong result for permissive temporally inconstant readings.
One last example shows that we can get permissive inconstant reports with metaphys-
ical modality:
Masquerade: Most guests who actually were invited to the masquerade are democrats.
Some have complained that this is the result of political bias, but Mustard disagrees. He
thinks that the invitees were chosen by a fair lottery, so believes that most guests invited
to the masquerade could have been republicans. Detective Black knows all this. But the
detectives know something Mustard doesn’t: cops were instructed to tail whoever was
invited to the masquerade. No matter who was invited to the masquerade or for whatever
reasons, all and only those people would later be tailed. Detective Grey asks Black: is
there a special reason we’re tailing so many democrats?
Detective Black sincerely utters the following to Grey:
(15) Mustard believes that most guests we’re tailing could have been republicans.
Here ‘believes’ is permissive with respect to ‘guests we’re tailing’; Mustard isn’t repre-
sented as thinking of anyone as being tailed, since the detectives know that he is unaware
33
cops are tailing anyone. He could verify (15) while lacking concepts of the conversational-
ists, hence lacking a concept of the property expressed by ‘guests we’re tailing’. So on the
plurality theory, ‘guests we’re tailing’ scopes wide of ‘believes’:
(15W)9 Y[guests we’re tailing](Y)[Mustard believes [could have been [most Y]
x
[x are
republicans]]]
But this delivers the wrong reading. For the proposition corresponding to (15) rep-
resents the plurality guests-cops-are-tailing-in-@ as being such that Mustard believes that
it could have been that most members of that plurality were republicans. But Mustard
doesn’tbelievethatmostofthoseactuallyinvited(thosewhoactuallyhappentobetailed)
could have been republican; rather, he believes something about those who could have been
invited. In short, ‘guests we’re tailing’ is inconstant across the worlds relevant to ‘could
have been’. But again, since the plurality theory secures permissive readings by interpret-
ing the relevant report as being about a single plurality, it fails to predict the inconstancy
of ‘guests we’re tailing’.
41
4.3. A Functional Defense of the Plurality View
Before leaving the plurality theory behind, I want to consider a possible line of defense.
This defense begins by adverting to examples in which a noun phrase apparently outside of
the scope of another intuitively covaries in its interpretation with that of the phrase in its
scope. The defense then insists that whatever semantic account works for those examples
canalsobeextendedtomakesenseofplurality-designatingphrasescovaryingininterpreta-
tionwithexpressionswithintheir scope, includingmodalandtemporalexpressionsintheir
scope. The plurality theory, then, can be re-written along these lines to preserve the basic
explanatory strategy of this sensitive contents view in the face of permissive inconstant
reports.
I should say that this section is somewhat of a detour from the main argument of
this chapter. It is a detail-mongering examination of one plausible way of developing this
functional defense of the plurality view, in which I try to show that whatever its merits,
it cannot be used to defend a version of the strict view. For the semantic machinery used
to account for how a plurality-designating expression can covary with modal or temporal
expressions within its scope will be forced to handle permissive inconstant reports by
includingthingsasconstituentsofasubject’sreportedthoughtcontentswhichthatsubject
need not cognize in verifying the relevant permissive report. (The main thread of this
chapter picks up again on page 54.)
41
It may be worth noting that we can get permissive inconstant readings on the metaphysical flavor of
modality using counterfactuals. A slight variation on Masquerade will allow for a use of the following to
issue such a report: ‘Mustard believes that if Peacock had sent out the invitations then every tailed guest
would be a republican.’ The setup would just need to stipulate that the actually tailed guests are not all
republicans, but that Mustard believes that Peacock would have invited only republicans.
34
The kind of example which motivates this defensive strategy involves a determiner
phrase containing a relative clause which itself contains a quantifier phrase. Here is the
classic example from Geach (1964):
(16) The woman whom every true Englishman most reveres is his mother.
The most natural interpretation of (16) is one on which the definite description ‘the
woman whom every true Englishman most reveres’ does not serve to designate just one
woman who is a mother, since presumably it is not the case that all true Englishman share
onemother,butratherservestopickoutdi↵ erentmothersrelativetodi↵ erentEnglishmen.
This stands in contrast with an example like
(17) The woman whom every true Englishman most reveres is the queen.
In (17), unlike in (16), the definite does intuitively serve to designate a single woman.
The contrast between these cases seems to occasion special treatment. One cannot capture
the truth-conditions of (16) in a straightforward compositional way just by interpreting
‘every true Englishman’ as scoping wide of ‘the woman’, as in
(16R) [every true Englishman]
x
[the woman whom x most reveres is x’s mother]
Though this regimentation does seem to reflect the intuitive truth-conditions of (16),
it is unclear how to generate (16R) in a systematic way. Since ‘every true Englishman’ is
embedded in a relative clause within the complex quantifier phrase involving ‘woman’, this
approach apparently requires the wide-scoped phrase to bind the ‘his’ which lies outside
of its scope.
42
So we seem forced instead to develop some alternative semantic treatment
whichcanaccountforthe“multipleindividual”readingofthecomplexquantifierphrase.
43
This phenomenon is by no means restricted to definite descriptions, as the following
example sentences illustrate:
(18) Each weapon that every suspect tried to hide was one bearing their fingerprints.
(19) Two cops whom every suspect lied to were the cops who first interrogated them.
(20) Mostroomswhicheverycopsearchedwereroomswherehefoundmuddyfootprints.
(21) A person whom each suspect was keen to talk to was their lawyer.
To my ear, the most natural interpretations of these sentences are not ones on which
the fronted quantifier phrase serves to make reference to each, two, most or one of, a single
42
On the standard approach to quantifier movement given in Heim & Kratzer (1998), as every true
Englishman’“moves”fromitsoriginalpositionatsurfacestructure,itleavesbehindatraceandtheresulting
embedding phrase ‘the woman whom t most reveres’ is modified to designate the function mapping an
individual i to the woman whom i most reveres (or, the property of being the woman whom one most
reveres). But since the ‘his’ occurring in ‘his mother’ is not c-commanded by ‘every true Englishman’ it is
not modified to contain a trace which can be bound by that phrase.
43
I take this label of this reading from Sharvit (1999).
35
plurality. In(18),forinstance,themostnaturalreadingisnotoneaccordingtowhichthere
is a single plurality P of weapons that every suspect tried to hide such that each element of
P bore every suspect’s (or certain suspects’) fingerprints. Rather, the best interpretation is
one on which for every suspect x there is a plurality Y of weapons which that suspect tried
to hide and each element of Y borex’s fingerprints. Similar observations hold for (19)-(21).
Again, special treatment of these examples seems needed. As with (16), we apparently
cannot secure the readings we want simply by appealing to an inverse scope interpretation
on which the quantifier phrase embedded in the relevant relative clause scopes wide of the
quantifier phrase in which that relative clause is embedded. For in these examples the
quantifier phrases embedded within the relative clause of the fronted quantifier phrase do
not take scope over the phrases intuitively anaphoric on them which live to the right of
the relevant identity verb.
We have just seen that a complex determiner phrase, such as ‘the woman whom every
true Englishman most reveres’ can covary in its interpretation with respect to expressions
which don’t scope wide of that phrase. This observation opens up a line of defense against
the potential counter-examples raised by permissive inconstant reports. The simple plu-
rality view, recall, foundered on permissive inconstant reports for attempting to interpret
a bare noun phrase, like the ‘bugged bust’ in ‘every bugged bust’ in (12), as alone scop-
ing wide of the attitude verb with respect to which it takes a permissive reading. But
since in permissive inconstant reports there are temporal or modal phrases with respect to
which the relevant bare noun phrase intuitively covaries in interpretation but which also
take narrow scope to the relevant attitude verb, this wide-scoping maneuver forced the
bare noun phrase to take a constant reading with respect to the embedded temporal or
modal phrase. But once we acknowledge that already are in need of a semantic treatment
which can make sense of a quantifier phrase’s covarying in interpretation with respect to
phrases within their scope, we apparently are committed to semantic resources which can
be used to interpret a bare noun phrase as scoping wide of a temporal or modal phrase yet
covarying in interpretation with respect to that phrase.
Just as the most natural reading of (16) is one on which the relevant women covary
with true Englishman, we might understand the pluralities invoked by the plurality view’s
analysis of permissive inconstant reports as covarying with times or worlds relevant to
interpretation of some temporal or modal phrase, even though the plurality-designating
expression is interpreted as scoping wide of that temporal or modal phrase. Suggestively,
thereadingsofpermissiveinconstantreportswilltakeaformlikeoneofthefollowing, with
some di↵ erent predicative material replacing ‘most reveres’:
(22) The plurality which every epistemic possibility most reveres is the plurality of
marked bills in the black briefcase at that possibility.
(23) The plurality of bugged busts which every Tuesday most reveres is the plurality of
composer busts at that Tuesday.
36
(24) The plurality of debaters bribed to throw the competition which some time most
reveres is the plurality of team captains of the Stripes at that time.
(25) The plurality of guests we’re tailing which some alethic possibility most reveres is
the plurality of invitees at that possibility.
In each of these examples, we have a definite description serving to designate di↵ erent
pluralities at di↵ erent worlds or times that are relevant to the interpretation of a temporal
ormodalphrase, eventhoughthedescriptionitselfisnotwithinthescopeofthattemporal
or modal phrase.
Since we need a special semantic treatment to account for examples such as these,
there is hope that the plurality view can extend whatever resources are used to treat these
examples to the permissive inconstant examples raised in §2.3.
This defensive strategy can be seen as involving two key steps: first, determine the
phrase to replace for ‘most reveres’ in the examples above; second, adopt a particular
semantic treatment for the “multipleindividual”reading of thedeterminer phrases in (22)-
(25). If these two steps are taken in a way that successfully captures the truth-conditions
of §2.3’s permissive inconstant reports, then the sensitive contents view is saved, as is the
strict view along with it.
Let’s try to develop this approach for Bugged. The report sentence central to that
example was
(12) Plum thinks that on Tuesdays every bugged bust is in the east wing.
The simple plurality view seemed forced to give the following rendering of (12)’s truth-
conditions:
(12W)9 Y[bugged bust](Y) [Plum thinks [on Tuesdays [every Y]
x
[x is in the east
wing]]]
Let’strytore-writethisalongthelinesof(23)togetasenseofhowourspecialsemantic
treatment has to go. The basic idea is this. In (16), ‘woman’ covaries in interpretation
with ‘Englishman’ even though ‘woman’ is in not in the scope of ‘Englishman’, and the
sentence serves to make a claim about those women. Similarly, we are to find a re-write
of (12) on which ‘bugged bust’ covaries with ‘Tuesdays’ even though ‘bugged bust’ is not
in the scope of ‘Tuesdays’, which serves to make a claim about those pluralities of bugged
busts.
As I see it, we have two main choice points in constructing our re-write of (12). First,
does ‘Tuesday’ scope wide of ‘thinks’ or not? Second, is ‘bugged busts’ used in the spec-
ification of pluralities which is given in terms of Plum’s thinking, or is it instead used
to specify certain pluralities which (12) identifies with those specified in terms of Plum’s
thinking? Corresponding to these choice points, we have four options:
37
1.
(12WP) The plurality of bugged busts to which every Tuesday bears the relation of
being such that Plum thinks that on that day every element of that plurality
is in the east wing is the plurality of composer busts on that day.
(‘Tuesdays’ scopes Wide to ‘thinks’; ‘bugged busts’ is used in the specification of
Pluralities given in terms of Plum’s thoughts)
2.
(12NP) The plurality of bugged busts which Plum thinks on Tuesdays is in the east
wing is the plurality of composer busts on that day.
(‘Tuesdays’ scopes Narrow to ‘thinks’; ‘bugged busts’ is used in the specification of
Pluralities given in terms of Plum’s thoughts)
3.
(12WI) The plurality to which every Tuesday bears the relation of being such that
Plumthinksthatonthatdayeveryelementofthatpluralityisintheeastwing
is the plurality of bugged busts on that day.
(‘Tuesdays’ scopes Wide to ‘thinks’; ‘bugged busts’ is used in the specification of plu-
ralities Identified with those specified in terms of Plum’s thoughts)
4.
(12NI) The plurality which Plum thinks on Tuesdays is in the east wing is the plurality
of bugged busts on that day.
(‘Tuesdays’ scopes Narrow to ‘thinks’; ‘bugged busts’ is used in the specification of
pluralities Identified with those specified in terms of Plum’s thoughts)
None of these re-writes of (12) is fun to read, but using them as a guide we can try to
specify truth conditions for (12) along the lines of whatever account we use for (16). This
may serve to reconcile the plurality theory with permissive inconstant reports. I’ll consider
each option in turn.
2.4.1 Option 1
Option 1 seems most closely to match the structure of (16).
(12WP) The plurality of bugged busts to which every Tuesday bears the relation of
being such that Plum thinks that on that day every element of that plurality is in the east
wing is the plurality of composer busts on that day.
I say that this option best matches the form of (16) since ‘bugged busts’ and ‘Tuesdays’
occur in (12WP) in basically the same positions that ‘women’ and ‘Englishman’ occur in
(16): ‘bugged busts’ occurs in the head of a determiner phrase which picks out di↵ erent
things relative to di↵ erent interpretations of another determiner phrase ‘every Tuesday ...’
that occurs within a relative clause embedded in the first phrase.
38
I will first use Option 1 to illustrate the semantic proposal of the functional defense of
the plurality view. Then I will raise two problems with Option 1, which also pose problems
for Options 2 and 3.
With (12WP) as a working re-write of (12), let’s lay out a semantic proposal which
can handle multiple individual readings of examples like (16). To simplify my discussion,
I focus on one of the many accounts of these readings. I’ll adopt a rough version of the
account found in Sharvit (1999), which develops an account from Jacobson (1994).
44
On
this proposal, multiple individual readings are accounted for by quantifying over functions.
Simplifying away from some of the particular details of Sharvit’s semantic system, the
functional analysis of (16) is roughly:
(16F) ◆ f[Nat(f)&8 x[x2 Dom(f)! woman(f(x))]&8 x[[true-Englishman(x)! most-
reveres(x,f(x))]]] = x◆ z[mother-of(z,x)]
An English translation: the “natural” function which maps every true Englishman to
thewomanhemostreveresisthemother-of function, thefunctionthatmapsanindividual
to their mother. Sharvit (1999), building from Jacobson (1994), restricts the function
quantifiedovertobea“natural”function(“Nat(f)”),onewhichisconversationallysalient,
since there may be many functions from Englishman to women answering to the given
descriptive condition. The chief advantage of this style of account is that by ascending
to talk of functions, we avoid the problem of having to construe a wide-scoped ‘every
Englishman’ as binding beyond its scope.
With this style of semantic treatment on the table, we can apply it to our suggested
re-write of (23), (12WP):
(12WPF) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on-Tuesday(t)! [thinks(Plum,[everyelementoff(t)isintheeastwing)]]]]
= t◆ z[plurality-of-composer-busts(z,t)]
In something like English: the natural function which maps Tuesday times to the
plurality of bugged busts at that time such that Plum thinks that every element of that
plurality is in the east wing is the function mapping times to the plurality of busts of
composers at that time. With this we successfully avoid rendering the predicted truth-
conditions of (12) as depending on how things stand with respect to the single plurality
of things which are bugged busts at the context-time (securing its temporally inconstant
reading) while also keeping the relevant pluralities of bugged busts from being interpreted
within the scope of ‘thinks’ (securing its permissive reading).
I see two problems with Option 1. The first concerns our choice to use ‘bugged busts’
in specifying the function which is also given in terms of Plum’s thinking. The second
44
These theories themselves build from Groenendijk & Stokhof (1984), which o↵ ers an account for func-
tional readings of questions like ‘which woman does every man most revere?’, which license the answer ‘his
mother’.
39
concerns our choice to interpret ‘Tuesdays’ as scoping wide of ‘thinks’.
The first problem I raise for Option 1 concerns its explicitly identifying the function
at issue with one that maps times to busts of composers at that time. Not only is this
not intuitively relevant to the truth-conditions of (23), but this aspect of this approach is
hard to square with certain examples of coordinated reports, in which multiple subjects
are reported as believing some proposition. Consider:
Bugged
2
: As in Bugged, Black knows that Plum thinks that on Tuesdays every bust of
a composer on display in the manor is in the east wing. In this scenario, Scarlet also has
beliefsaboutthebusts: shehasbeenpayingattentiontowhatthebustsaremadeof. Some
busts are marble, some are terracotta, some are wood. Black knows that Scarlet thinks
that on Tuesdays every bust made of marble is in the east wing. The detectives know that
on Tuesdays only marble busts of composers are bugged (but some marble busts are not
bugged and some composer busts are not bugged). They also know that not all of the
marble busts in the east wing are busts of composers and that not all busts of composers
in the east wing are marble. The detectives know that Plum and Scarlet are each unaware
that any busts are bugged. In discussing Plum and Scarlet’s suspicions about the busts’
movements, Black says to Grey:
(26) Plum and Scarlet think that on Tuesdays every bugged bust is in the east wing.
As with (12), ‘bugged bust’ is permissive yet is also inconstant with respect to ‘on
Tuesdays’.
45
On Option 1, this would have to be regimented as something like:
(26Fa) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on-Tuesday(t)! [thinks(p&s,[every element of f(t) is in the east wing)]]]]
= t◆ z[plurality-of-composer-busts(z,t)]
(26Fb) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on-Tuesday(t)! [thinks(p&s,[every element of f(t) is in the east wing)]]]]
= t◆ z[plurality-of-marble-busts(z,t)]
But neither of these captures the correct truth conditions of (26). For the function
from Tuesday-times to pluralities of composer busts is not identical with the function from
Tuesday-times to marble busts. So neither of (26Fa) or (26Fb) works.
In response to this problem, an advocate of Option 1 might instead appeal to the
intersection of the function from Tuesday-times to composer-busts and the function from
Tuesday-times to marble busts. Why not use the function from Tuesday-times to marble
busts of composers? This gives us:
45
If you find this reading more di cult to access, it may help to add that there is another guest under
discussion who is known by Black to think that on Tuesdays every bust made of marble is in the west wing.
Black could preface her remark with ‘some guests thinks that on Tuesdays the bugged busts are in the east
wing, others think that on Tuesdays the bugged busts are in the west wing...’
40
(26Fc) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on-Tuesday(t)! [thinks(p&s,[every element of f(t) is in the east wing]]]]
= t◆ z[plurality-of-marble-composer-busts(z,t)]
Call this style of analysis of permissive coordinated reports the “intersectional ap-
proach”. The intersectional approach seems to do better than (26Fa) or (26Fb). Since
Plum believes of the plurality of composer busts that on Tuesdays every element of that
plurality is in the east wing, he presumably believes of the plurality of marble composer
busts that every element of that plurality is in the east wing. Similar comments apply to
Scarlet. SothefunctionfromTuesday-timestopluralitiesofmarblecomposerbustsatthat
time is the function from Tuesday-times to pluralities such that Plum and Scarlet think at
that time every element of that plurality is in the east wing. There is, of course, a worry
of how the truth-conditions given by (26Fc) are to be derived (it will have to involve some
kind of special contributions from the conversational context), but it seems to capture the
truth-conditions of (26)) in a way consistent with the strict view, without rendering the
report loaded or inconstant in ways it isn’t.
There are, however, other coordinated reports for which the intersectional approach
cannot work:
Bugged
3
: Plum thinks that on Tuesdays the three non-marble composer busts in the
manorareintheeastwing. ScarletthinksthatonTuesdaysthethreemarblenon-composer
busts in the manor are in the east wing. The detectives know that on Tuesdays all six
of these busts are bugged: on Tuesdays, the three non-marble composer busts and the
three marble non-composer busts are all bugged. The detectives also know that Plum and
Scarlet are unaware that there are any bugged busts in the manor on Tuesdays.
(27) Plum and Scarlet think that on Tuesdays three bugged busts are in the east wing.
As with (26), this is permissive with respect to ‘bugged busts’, which is also inconstant
with respect to ‘on Tuesdays’. If we apply the intersectional approach, we get:
(27Fa) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on-Tuesday(t)! [thinks(p&s,[three elements off(t) are in the east wing)]]]]
= t◆ z[plurality-of-marble-composer-busts(z,t)]
This yields an incorrect prediction. Neither Plum nor Scarlet believe of the plurality of
things which are both marble busts and composer busts on a Tuesday that three elements
of it are in the east wing. So the intersectional approach will not serve to account for all
permissive coordinated reports.
This does not yet show that Option 1 cannot work. To handle Bugged
3
, we might look
for yet another way of combining the two relevant functions. Intuitively in this case we
need the union of these functions:
(27Fa) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
41
8 t[on-Tuesday(t)! [thinks(p&s,[three elements off(t) are in the east wing)]]]]
= t◆ z[plurality-of-marble-busts-or-composer-busts(z,t)]
It does seem true that, in the scenario of Bugged
3
, Plum and Scarlet each believe of
each plurality of things which are either marble busts or composer busts at a Tuesday
time that three of that plurality are in the east wing (at that time). So if we allow that
this particular kind of functional approach may appeal to intersections or unions of certain
functionsatwill,advertingtoinfluencefromtheconversationalcontext,thenitcanaccount
for permissive coordinated reports.
I’ll raise one more related example which I think puts some pressure on Option 1 along
these lines, then I’ll move to my second problem with this Option. The example combines
features from both of Bugged
2
and Bugged
3
:
Bugged
4
: Monthsaredividedintooddweeksandevenweeks. Thedetectivesknowthat
on odd Tuesdays the bugged busts are all and only the marble composer busts (though
therearesomenon-marblecomposerbustsandsomenon-composermarblebusts)andthat
on even Tuesdays the bugged busts consist of three non-marble composer busts and three
non-composer marble busts (and that these six exhaust the composer and marble busts).
BlackknowsthatPlumthinksthatonoddTuesdaysallcomposerbustsareintheeastwing
and that Scarlet thinks that on odd Tuesdays all marble busts are in the east wing. Black
also knows that Plum thinks that on even Tuesdays the three non-marble composer busts
are in the east wing and that Scarlet thinks that on even Tuesdays the three non-composer
marble busts are in the east wing. In discussing the busts, Black says to Grey:
(28) Plum and Scarlet think that on Tuesdays either exactly three or all bugged busts
are in the east wing.
Thecaseiscomplex,butitseemslike(28)cantakeatruepermissiveinconstantreading
with respect to ‘bugged busts’. On the going version of Option 1, we are to adopt either
the intersectional or unional approach:
(28FI) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on-Tuesday(t)! [thinks(p&s,[exactly three or all elements of f(t)isin
the east wing)]]]] = t◆ z[plurality-of-marble-composer-busts(z,t)]
(28FU) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
8 t[on Tuesday(t)! [thinks(p&s,[exactly three or all elements of f(t)isin
the east wing])]]] = t◆ z[plurality-of-marble-busts-or-composer-busts(z,t)]
The problem is that neither the intersectional (28FI) nor the unional (28FU) can ac-
countforbothPlum’sandScarlet’sthoughtsaboutoddTuesdaysandtheirthoughtsabout
even Tuesdays. Although Plum and Scarlet think of the plurality of things which are both
marbleandcomposerbustsonanoddTuesdaythatallelementsofthatpluralityareinthe
east wing on that Tuesday, neither thinks of the plurality of things which are both marble
42
and composer busts on an even Tuesday that either exactly three or all elements of that
plurality are in the east wing on that Tuesday. Similarly, although Plum and Scarlet think
of the plurality of things which are either marble busts or composer busts on an even Tues-
day that exactly three of that plurality are in the east wing, neither thinks of the plurality
of things which are either marble busts or composer busts on an odd Tuesday that exactly
three or all elements of that plurality are in the east wing on that Tuesday. (Recall that
on odd Tuesdays there are non-composer marble busts and non-marble composer busts.)
The second problem I raise for Option 1 concerns its interpreting ‘Tuesday’ as scoping
wide of ‘thinks’. By interpreting the restriction on times, ‘on-Tuesday’, as outside of the
scope of ‘thinks’, (12WP) renders the report made with (12) as permissive with respect to
‘on Tuesdays’ and the concept Tuesday via which one cognizes a time as taking place on a
Tuesday. This leads to false predictions in cases in which Plum conceives of non-Tuesday-
times as occurring on Tuesday, as in
Bugged
5
: Black knows that Plum thinks that every Tuesday every bust of a composer
on display in the manor is in the east wing. The detectives know that Plum is confused
aboutwhichdaysareTuesdays; hereliesonafaultycalendarwhichrepresentsWednesdays
asTuesdays. SotheyknowthatPlumtakeswhathappensonWednesdaysashappeningon
Tuesdays: whenPlumsaysthat acurator willvisit thegallery next Tuesday, thedetectives
plan on a curator arriving next Wednesday. The detectives also know that on Wednesdays
only busts of composers are bugged. They both know that Plum is unaware that any busts
are ever bugged, and are discussing Plum’s suspicions about the busts’ movements. Black
says to Grey:
(29) Plum thinks that on Tuesdays every bugged bust is in the east wing.
ThereportwhichBlackmakeswith(29)isintuitivelypermissivewithrespectto‘bugged
bust’, as in (12), but is loaded with respect to ‘on Tuesdays’. Option 1 captures the truth-
conditions of (29) with:
(29F) ◆ f[Nat(f)&8 t[t2 Dom(f)! plurality(f(t))]&
8 t[[on-Tuesday(t)! [thinks(p,[at t every element of f(t) is in the east wing)]]]]]
= t◆ z[plurality-of-bugged-busts(z,t)]
This delivers incorrect truth-conditions by constraining the relevant function to output
pluralities specified in terms of times which in fact fall on Tuesday. This in e↵ ect renders
thereportmadeby(29)aspermissivewithrespectto‘onTuesdays’,whichitisn’t. Butthis
also delivers incorrect truth conditions simply because in the case Plum need not possess
any beliefs at all about pluralities of things which happen (unbeknownst to him) to be
busts on Tuesdays.
These two problems, then, tell against Option 1. Let’s turn to the other options.
2.4.2 Option 2
43
Option 2 is like Option 1, but interprets ‘on Tuesdays’ as scoping narrow to ‘thinks’:
(12NP) The plurality of bugged busts which Plum thinks on Tuesdays is in the east
wing is the plurality of composer busts on that day.
The Sharvit-style regimentation answering to this English re-write is something like:
(12NPF) ◆ f[Nat(f)&8 x[x2 Dom(f)! plurality-of-bugged-busts(f(x))]&
[thinks(p,[8 t[on-Tuesday(t)! [atteveryelementoff(t)isintheeastwing)]]]]
= t◆ z[plurality-of-composer-busts(z,t)]
I won’t bother here to try to interpret this regimentation in English; as we’ll see in my
discussion of Option 4, I have my doubts that interpretations which interpret ‘Tuesday’ as
scoping narrow to ‘thinks’ are intelligible on the functional approach.
Butwedon’tneedtotarryhere,since,giventhestructureofthisoption,wecanalready
see that although it may avoid the second problem raised against Option 1, it will fall prey
to the first problem raised against Option 1. Including ‘the plurality of composers’ in
characterizing the truth conditions of (12) leads to false predictions in cases of coordinated
reports, and there seems to be no general substitute for this description which can account
for the full range of coordinate permissive inconstant reports considered in the previous
subsection.
2.4.3 Option 3
Option 3 is like Option 1, but instead of using ‘bugged busts’ in specifying the plurality
which is also specified in terms of Plum’s thinking, it uses ‘bugged busts’ in specifying the
plurality which is identified with that specified in terms of Plum’s thoughts:
(12WI) The plurality to which every Tuesday bears the relation of being such that
Plum thinks that on that day every element of that plurality is in the east wing is the
plurality of bugged busts on that day.
The Sharvit-style regimentation of (12WI) is:
(12WIF) ◆ f[Nat(f)&8 t[t2 Dom(f)! plurality(f(t))]&
8 t[on-Tuesday(t)! [thinks(Plum,[everyelementoff(t)isintheeastwing)]]]] =
t◆ z[plurality-of-bugged-busts(z,t)]
In English-ish: the unique natural function which maps Tuesday times to the plurality
of things such that Plum thinks that every element of that plurality is in the east wing is
the function mapping times to the plurality of bugged busts at that time.
By avoiding reference to composer busts, marble busts, or other busts known by the
conversationalists to be bugged busts as di↵ erent Tuesday times, (12WIF) avoids the first
problemraisedin §2.4.1againstOption1. Butbyinterpreting‘Tuesday’asscopingwideof
‘thinks’, Option 3 runs into the second problem raised by Bugged
5
, in which Plum confuses
44
Tuesday times with Wednesday times. So we’ve made little progress.
ArelatedshortcomingforOption3(andOption1)concernsreportsinvolvingtemporal
ormodalexpressionswhichdonot,roughlyspeaking,playtheroleofuniversalorexistential
quantifiers over index parameters. Consider:
Bugged
6
: Black knows that Plum thinks that on most Tuesdays every bust of a com-
poser in the manor is in the east wing, though Black knows that there are no particular
Tuesdays, past, present, orfuture, suchthatPlumbelievesthatonthatparticularTuesday
every bust of a composter is in the east wing. The detectives know that on Tuesdays only
busts of composers are bugged. They both know that Plum is unaware that any busts are
ever bugged, and are discussing Plum’s suspicions about the busts’ movements. Black says
to Grey:
(30) Plum thinks that on most Tuesdays every bugged bust is in the east wing.
This report is permissive with respect to ‘bugged bust’. Option 3 regiments this as:
(29F) ◆ f[Nat(f)&8 t[t2 Dom(f)! plurality(f(t))]&
[most[on-Tuesday]
t
[thinks(p,[at t every element of f(t) is in the east wing)]]]] =
t◆ z[plurality-of-bugged-busts(z,t)]
In English-ish: the unique natural function which maps most Tuesday-times to the
plurality of things such that Plum thinks that every element of that plurality is in the east
wing at that time is the function mapping times to the plurality of bugged busts at that
time.
Option 3 falters here because, in Bugged
6
, there is no particular plurality of busts such
that Plum believes of that plurality that on any particular Tuesday that it is in the east
wing. For Plum to believe that on most Tuesdays every bust of a composer is in the
east wing does not require that with respect to any particular Tuesday Plum believe of
some particular plurality that every element of it is in the east wing. Plum can remain
totally agnostic about which pluralities in particular are in the east wing on any particular
Tuesday while persisting in the general belief that on most Tuesdays pluralities of some
kind are in the east wing.
2.4.4 Option 4
We arrive finally at Option 4:
(12NI) The plurality which Plum thinks on Tuesdays is in the east wing is the plurality
of bugged busts on that Tuesday.
This option departs most from the guiding example of (16): ‘bugged busts’, which
we wanted to play a role similar to ‘woman’ in (16), does not figure in the head of the
determiner phrase which covaries in interpretation with that of a relative clause embedded
within that phrase. And ‘Tuesdays’, the temporal expression with respect to which in
45
the interpretation of ‘bugged busts’ is to covary, is not the head bare phrase within the
determiner phrase embedded in the relevant relative clause. But this option avoids the
problems with Option 1, so is worthy of consideration.
The Sharvit-style regimentation of (12NI) is:
(12NIF) ◆ f[Nat(f)&8 t[t2 Dom(f)! plurality(f(t))]&
[thinks(p,[8 t[on-Tuesday(t)! [atteveryelementoff(t)isintheeastwing)]]]]]
= t◆ z[plurality-of-bugged-busts(z,t)]
How should we understand the way of specifying a function here in terms of Plum’s
thinking? That is, how should we understand what it takes for a function f to be such
that
(12NIFa) [Nat(f)&8 t[t2 Dom(f)! plurality(f(t))]&[thinks(p,[8 t[on-Tuesday(t)!
[at t every element of f(t) is in the east wing)]]]]]
Under what conditions does a function satisfy the condition laid down by (12NIFa)?
Supposeafunctionfromtimestopluralitiesf
1
issuchthat[thinks(p,[8 t[on Tuesday(t)!
[at t every element of f
1
(t) is in the east wing]]])]. It does not look like f
1
must be such
that for Tuesday times t
T1
, t
T2
, ..., Plum thinks that at t
T1
every element of f
1
(t
T1
)isin
the east wing, Plum thinks that at t
T2
every element of f
1
(t
T2
) is in the east wing, and so
on. For ‘on-Tuesday’ scopes narrow, not wide, of ‘thinks’.
Similarly, there seems to be no requirement for any particular time t that f
1
be such
thatPlumthinksthatatteveryelementoff
1
(t)isintheeastwing. ForwhatitisforPlum
to have a general belief in [8 t[on-Tuesday(t) ! [at t every element of f(t) is in the east
wing]]] is not for Plum to have a collection of singular beliefs about individual times t
1
, t
2
,
..., of the form; [thinks(p,[at t
1
every element of f
1
(t
1
) is in the east wing])], [thinks(p,[at
t
2
every element of f
1
(t
2
) is in the east wing])], and so on. One can have a general belief
that Every F is G without having a collection of singular beliefs about Fs a, b, ..., that a
is G, b is G, and so on.
And, of course, there is no requirement that f
1
map any time t⇤ to a single plurality
f
1
(t⇤ ) such that [thinks(p,[8 t[on-Tuesday(t)! [at t every element of f
1
(t⇤ ) is in the east
wing]]])]; thisisthekindoftruth-conditiondeliveredbytheoriginalpluralitytheory(which
we’ve seen founders on permissive inconstant readings).
Giventheseconsiderations,itlooksasthoughtheoccurrenceof‘f(t)’in(12NIFa)isnot
to be interpreted as designating pluralities which are constituents of any particular Plum
thoughts. This in itself may be fine, but this suggests already that Option 4 is departing
from the plurality theory in some way. On this account, the truth conditions of permissive
inconstant readings are not captured by representing Plum as bearing attitudes towards
propositions involving pluralities. Rather, the occurrence of ‘f(t)’ (12NIFa) apparently
must be interpreted as contributing a complex propositional constituent consisting in part
of a function designated by ‘f’. On this view, then, the truth-conditions of (12) represent
46
Plum as possessing one thought about how things stand with respect to the pluralities
pickedoutbyfunctionf onTuesdays. Wecantheno↵ eranEnglishtranslationof(12NIF):
the unique natural function which maps times to pluralities such that Plum thinks that
every Tuesday is such that every element of the value of that function on that day is, on
that day, in the east wing, is the function which maps a time to the bugged busts at that
time.
This at least, according to the strict view, requires that Plum possess a concept of
this function f. Instead of construing the truth-conditions of (12) as representing Plum as
thinking about individuals which happen to be bugged busts or pluralities which happen
to consist of bugged busts, (12NIF) represents Plum as thinking about a function which
happens to range over pluralities of bugged busts. For some this may go down easy, but it
at least is worth pausing to note that this may come at some intuitive cost. It is not hard
to stoke the intuition that for Plum to verify (12) Plum need not possess a concept of any
function, or a concept of a function from times to pluralities in particular. To make this
vivid, consider the following example:
Bugged
7
: BlacklearnsfromconversationwithPlumthatPlumthinksthatonevery day
everybustofacomposerisintheeastwing. Asbefore,thedetectivesknowsomethingPlum
doesn’t: every Tuesday, the detectives put microphones only inside busts of composers on
display in themanor, soon Tuesdaysonly bustsof composersarebugged. They both know
thatPlumisunawarethatanybustsarebugged,andarediscussingPlum’ssuspicionsabout
the busts’ movements. Black sincerely utters
(12) Plum thinks that on Tuesdays every bugged bust is in the east wing.
To my ear, this still goes through as an appropriate and true permissive report. But
it is especially hard here to accept that in order for Plum to verify (12) in this context he
must use a concept of a function that maps Tuesday times to certain pluralities. This is
my first worry for Option 4: Plum doesn’t have to possess a concept of any function from
times, or Tuesday times in particular, to pluralities. The approach may otherwise work,
but there are reasons up front to think that an advocate of the strict view cannot easily
adopt Option 4.
I cannot confidently o↵ er any truth-conditional counter-examples against Option 4,
but I’ll mention one concern. My concern concerns the di↵ erence between functions from
indices to pluralities and properties in hyper-intensional contexts. Consider:
Tri-Forgeries: Detectives Black and Grey think there are two kinds of forgeries in the
gallery on any given day: triangular paintings which are forgeries and trilateral paintings
which are forgeries. But the detectives are confused; they don’t believe that there are
any forgeries which are both triangular and trilateral. They commonly assume that on
Sundays all and only the triangular forgeries in the gallery are impressionist paintings in
the gallery and that on Sundays all and only the trilateral forgeries in the gallery are
47
surrealist paintings. Black knows that Peacock thinks that on Sundays every surrealist
painting in the gallery is an original Dali but that no impressionist painting in the gallery
is an original Dali. Black and Grey know that Peacock is unaware that there are every any
forgeries in the gallery. Black says to Grey:
(31) Peacock thinks that on Sundays every trilateral forgery is an original Dali but that
no triangular forgery is an original Dali.
In this context intuitively the report made with (31) take a permissive inconstant
reading with resepct to ‘trilateral forgery’ and ‘triangular forgery’. On Option 4, we are
to regiment this as something like:
(31F) ◆ f[Nat(f)&8 t[t2 Dom(f)! plurality(f(t))]&
[thinks(p,[8 t[on-Sundays(t)! [at t every element of f(t) is a Dali)]]]]]
= t◆ z[plurality-of-trilateral-forgeries(z,t)]
&
◆ f
0
[Nat(f
0
)&8 t[t2 Dom(f)! plurality(f
0
(t))]&
[thinks(p,[8 t[on-Sundays(t)! [at t no element of f
0
(t) is a Dali)]]]]]
= t◆ z[plurality-of-triangular-forgeries(z,t)]
We can see right away what the problem is supposed to be. The function from times to
pluralities of trilateral forgeries is the function from times to triangular forgeries. So the
truth-conditionsgivenby(31F)representPeacockasbelievingofthisfunctionthatonevery
Sunday every element of the value of that function on that day is a Dali and as believing of
this function that on every Sunday no element of the value of that function on that day is
a Dali. The detectives intuitively do not take the report made by (31) as attributing such
contradictory beliefs to Peacock, so this may be viewed as a counter-example to Option 4.
I think that this argument against Option 4, as posed, doesn’t do much. Option 4’s
treatment of this example can be understood as an instance of well-known puzzles with
direct reference. Just as one can believe of an object o that o is F and that o is not F,so
our detectives can belief of a function f that f is F and that f is not F. The intuitions of
what is and what is not appropriate for the confused detectives to report can be explained
in whatever terms one uses to explain why one can appropriately report in some contexts
that Peacock believes that Hesperus is bright but cannot appropriately report in those
contexts that Peacock believes that Phosphorus is bright. Perhaps some version of this
worry can be pressed against Option 4, but I don’t see how this should go.
A last point to raise concerning all of the foregoing options is methodological, but
retreats from some of the ambitious goals of this chapter. We’ve seen various reports
throughout §3 which intuitively are instances of a single phenomenon. This includes the
reports from §3.1, which I called “controversial permissive reports”. I take it that the
functional approach given by Option 4 cannot be used to account for the full range of all
of these examples. So, if we can find an alternative approach which can make sense of all
48
of these examples in terms of a single semantic phenomenon, I take it that that approach
should be preferred. In the rest of this dissertation I will be concerned with developing
such a general account.
But again, retreating to this kind of consideration moves away from the goals of this
chapter, which are to show that even when we focuses on the limited set of permissive
reportsconcerningtheinterpretationofbarenounphraseconstituentsofquantifierphrases,
still the strict view cannot be defended. So it’s worth reminding the reader that at this
point, the ambitious claim of this chapter rests on the judgment raised above, that Plum
need not possess a concept of any particular functions to verify (12). If this judgment can
be resisted, then Option 4 may serve as a legitimate defense of the strict view. Without
passing a final judgment on Option 4, in the next section I move on to defenses of the strict
view that depart from the guiding ideas behind the plurality theory.
5. Permissive Inconstant Reports and Substitution Theories
We’ve just seen that neither the scope theory of permissiveness nor various versions
of the plurality theory of permissiveness can account for permissive inconstant reports in
general. The defender of the strict view is left wanting an alternative explanation which
can respect their favored conceptual constraint on attitudes.
We can trace the problem facing the foregoing views to two assumptions. The first
assumption is that the linguistic elements occurring in the ‘that’-clause of a belief report
sentence contribute conventionally associated properties which partially determine what
the subject is said to believe. The second assumption is the strict view of attitudes: if
the linguistic elements in a true belief report sentence contribute conventionally associated
properties as constituents of the proposition said to be believed, then the subject must be
able to think in terms of those properties, hence possess concepts of them. Given these
assumptions, the bare noun phrases from §2 have to contribute conventionally associated
properties as constituents of the subjects’ reported beliefs, since they are inconstant. But
they also cannot contribute those properties as constituents of what is said to be believed,
since they are permissive.
If one wants to hold onto the strict view, we should amend the first assumption. Ac-
cordingly, one diagnosis of the failures of the sensitive contents accounts we have seen is
that they are too wedded to using properties which are conventionally associated with
the relevant bare noun phrases in giving the objects of subjects’ attitudes.
46
Perhaps to
account for each example from §2 we should instead take some property other than that
associated with the relevant bare noun phrase to be contributed to the proposition enter-
46
I do not mean to commit myself here to the view that there is one particular property conventionally
associated with any given bare noun phrase: bare noun phrases may in general be associated with linguistic
conventions which are somehow underspecified or vague so that multiple di↵ erent properties or relations
are conventionally associated with a particular bare noun phrase. To be sensitive to this possibility, we can
state this point by diagnosing the failure of of sensitive contents views as being committed to using the
properties which are contextually associated with the relevant bare noun phrases.
49
tained. A natural way to carry out this idea would be to specify for each property F and
each contextc a set of properties which a bare noun phraseF semantically associated with
F can contribute instead of F in c. With this set of “contextual analogues” of F, one could
say that a report is permissive with respect to an occurrence of phrase F if the use of that
occurrenceofthephraseinvolvedinthereportservestocontributeacontextualanalogueof
F rather than F itself. I call this kind of explanation of permissive readings a “substitution
theory”; these are yet more sensitive contents views. They avoid the foregoing problem
cases but still founder on variations on those cases, as I now show.
To see how the substitution theory can explain permissive inconstant readings, suppose
that in Marked the property bills-in-the-black-briefcase is a contextual analogue
of marked-bill. Then according to the substitution theory, in uttering (11) in Marked
detective Black can assert the proposition corresponding to
(11’) Scarlet believes [might [[every bill in the black briefcase]
x
[x is a hundred]]]
ThispropositionrepresentsScarletasbelievingapropositionwhichcontainsbills-in-
the-black-briefcase as a constituent. The truth of (11’), given the strict view of belief,
requires that Scarlet possess a concept of this property but not of marked-bill.This
preserves the intuition that ‘marked bill’ is permissive on detective’s Black’s use of (11)
in Marked while also preserving the intuition that it is inconstant across the worlds rele-
vant to ‘might’. Furthermore, the substitution theory is compatible with the strict view,
since according to the strict view subjects must possess concepts only of properties which
are constituents of the propositions they entertain. As a sensitive contents account, the
substitution theory is compatible with this.
I’ll discuss two general problems with the substitution theory, which can be brought to
light by considering two explications of contextual analogues on o↵ er in the literature.
To account for various permissive readings, Sudo (2014) develops a substitution the-
ory according to which two properties are contextual analogues if they are “contextually
equivalent”. Roughly, two properties are contextually equivalent in a context if they are
materially equivalent in every world which is compatible with everything the conversa-
tionalists in the reporting context commonly assume.
47
Since in the context for Marked
the detectives commonly assume that whichever bills ended up in the black briefcase are
exactly the marked bills, bill-in-the-black-briefcase and marked-bill are contextu-
ally equivalent in that context. With the idea that two properties are contextual analogues
if they are contextually equivalent, the substitution theory predicts that ‘marked bill’
can contribute bill-in-the-black-briefcase as a constituent of the proposition which
Scarlet is said to believe, which seems to get what we want. I’ll call this the “equivalence
substitution theory”.
A modified version of Marked raises a problem for the equivalence substitution theory.
47
Sudo’s notion of contextual equivalence defined for expressions, not propositional constituents, and is
given in terms of Stalnaker (1978)’s notion of “common ground”.
50
I call it the “problem of erroneous assumptions”.
Marked
2
: The context is like that for Marked but with three important changes. First,
detective Black knows that Scarlet has just seen Green take the black briefcase into the
lounge. So Black knows that Scarlet knows that every bill in the black briefcase is in
the lounge. Second, although the detectives still privately assume that whichever bills
ended up in the black briefcase are all and only the marked bills, this assumption is false.
Unbeknownst to the detectives, the cop who was supposed to mark the bills from the black
briefcase mistakenly marked some other bills instead. Finally, although the marked bills
are not in the lounge, the bills in the black briefcase are.
Black sincerely utters the following to detective Grey:
(32) Scarlet knows that every marked bill is in the lounge.
As in Marked, ‘marked bill’ takes a permissive reading, since the detectives know that
Scarlet is unaware that any bills have been marked. But since the marked bills are not in
thelounge,intuitivelywhatdetectiveBlackassertsisfalse: Scarletdoesn’tknowthatevery
marked bill is in the lounge. The equivalence substitution theory, however, predicts that
detective Black here asserts something true. For this theory explains the transparency of
‘marked bill’ by claiming that it contributes the property bill-in-the-black-briefcase
as a constituent of what Scarlet is said to know. And intuitively Scarlet does know the
resulting proposition: Scarlet knows that every bill in the black briefcase is in the lounge.
Soaccordingtotheequivalencesubstitutiontheorythereshouldbesomepermissivereading
of (32) in context Marked
2
which is true, but apparently there is none.
48
What happens if we adopt an alternative explication of contextual analogues which
doesn’t saythatbill-in-the-black-briefcaseisacontextualanalogueof marked-bills
in Marked
2
? To do this we could turn to Schwager (2009)’s substitution theory, which uses
aslightlydi↵ erentnotionofcontextualanalogues. Schwagerappealstoaclosenessordering
over worlds like that used in Lewis (2013)’s semantics for conditionals. Using this closeness
ordering, we say that G is a contextual analogue of F if the closest world in which F is
instantiated is a world in which G is instantiated and every instance of G is an instance of
F. The resulting theory, which I’ll call the “closeness substitution theory”, says that if G is
a contextual analogue of F, and bare noun phrases G and F are conventionally associated
with G and F, respectively, then one can truly report thataVs that S(F) if it is true that
aVs that S(G).
49
48
It won’t help for the equivalence substitution theorist to insist (implausibly) that there is at least some
true reading of (32). For were detective Black to utter ‘Every marked bill is in the lounge’ in this context,
she would assert something false and nothing true. So the substitution theorist must claim that in some
context c there is a sentence S such that psomeone knows that Sq is true in c but S is false in c.Sothey
must deny that in some contexts it is false that if a knows that S then it is true that S.
49
Where S(F)isaclause S in which phrase F occurs and S(G)isaclause S structurally similar to S(F)
in which phrase G occurs in the place F occurs in S(F).
51
The closeness substitution theory explain the examples from §2 just as the equivalence
theory does, and also apparently avoids the problem of erroneous assumptions. For in
Marked
2
, marked-bill is actually instantiated, but it isn’t the case that every instance of
bill-in-black-briefcase is an instance of marked-bill, though the detectives assume
as much. So bill-in-black-briefcase isn’t a contextual analogue for marked-bill in
that scenario, and therefore the closeness substitution theory doesn’t predict that Black’s
utterance of (32) serves to report that Scarlet knows that every bill in the black briefcase is
in the lounge. This theory accordingly avoids any prediction that (32) takes a true reading
in Marked
2
.
The problem, however, is that the closeness substitution theory avoids the problem
raised by Marked
2
by failing to predict that (32) can take a permissive reading in that
context at all. This raises the question of what is asserted by Black? As we observed,
‘marked bill’ is permissive on detective Black’s use of (32) in Marked
2
. But the closeness
substitutiontheorycannotpredictanysuchreading,trueorfalse,sincetherearenorelevant
contextual analogues of marked-bill in that context according to the closeness theory. So
in the face of the detectives’ erroneous assumption, the closeness substitution theory faces
the empirical problem of being unable to predict the permissive reading of ‘marked bill’ in
the first place, which is what it is designed to do.
50
Given the problems with these two substitution theories, we can see that Marked
2
poses a structural problem for any substitution theory whatsoever. For whatever expli-
cation of contextual analogues one adopts, the substitution theory either will entail that
bill-in-black-briefcase is a contextual analogue of marked-bill or will entail that it
isn’t. If it is a contextual analogue, then the theory is committed to the incorrect true
reading delivered by the equivalence theory; if it isn’t, then the theory fails to predict the
transparency of ‘marked bill’ in the false report made by Black, like the closeness theory.
Either way, the substitution theory cannot accurately account for the permissive reading
exhibited in Marked
2
.
A second problem with the substitution theory concerns reports involving multiple
subjects. Icallthisthe“problemofuncoordinatedagreement”. Toappreciatethisproblem,
consider another modified version of Marked:
Marked
3
: Scarlet thinks that the bills which were in the library were put into the black
and brown briefcases earlier this afternoon. As in Marked, detective Black knows that she
50
There are also truth-conditional problems with the closeness substitution theory, related to its use of
closeness orderings. Given the way that the closeness substitution theory explicates contextual analogues,
therelation ofbeingacontextual analogue isnotasymmetricrelation: Gmay beacontextualanalogueof F
without Fbeing a contextual analogueof G.Let a“demondog”bea dog whichexists inaworldwherethere
is a demon. Now suppose that actually White believes that every dog is cute. It should therefore always
be true and appropriate, on the closeness substitution theory, for one to attribute to White a belief with
‘White believes that every demon dog is cute’. For in the closest world in which demon-dog is instantiated,
dog is instantiated and every instance of dog is an instance of demon-dog. This result is quite implausible.
Thanks to John Hawthorne for discussion here.
52
thinks that every bill in the black briefcase might be a hundred. This morning White also
happened upon the bills in the library which Scarlet saw. Later, White saw Green putting
the bills into two bags, one red and one blue. She’s not sure which bills are in which
bag—perhaps the hundreds are all in the blue bag, perhaps the twenties and hundreds are
mixed between the two bags. White informs Black of all this, so Black comes to learn that
White thinks that every bill in the blue bag might be a hundred. But the detectives know
something which White and Scarlet don’t: Green moved every bill from the blue bag into
the black briefcase, after which, as before, a cop marked those bills. The detectives don’t
know which bills are marked, but just that whichever were in the blue bag are all and only
those which were in the black briefcase, which are all and only the marked bills. Black
knows that White is aware of the bags but not the briefcases, and Scarlet is aware of the
briefcases but not the bags. Detective Black says to detective Grey:
(33) Scarlet and White agree that every marked bill might be a hundred.
As in Marked, ‘marked bill’ is permissive but inconstant with respect to ‘might’; the
detectives know that neither Scarlet nor White is aware that any bills are marked, hence
don’t think in terms of marked bills. But still ‘marked bills’ is inconstant with respect to
‘might’: which bills are marked varies across the worlds relevant to ‘might’.
51
For neither
WhitenorScarletthinkofthebillswhichactuallyaremarkedthatthey mightbehundreds.
The problem for the substitution theory is that neither of the properties which intu-
itively are contextual analogues of marked-bill in the context of Marked
3
can be con-
tributed by ‘marked bill’ without determining a false interpretation of (33). For both of
the following sentences are false if the substituted bare noun phrase is opaque with respect
to ‘believes’:
(34) Scarlet and White agree that every bill in the black briefcase might be a hundred.
(35) Scarlet and White agree that every bill in the blue bag might be a hundred.
Since Scarlet need not be disposed to think of anything as a bill in the blue bag and
White need not be disposed to think of anything as a bill in the black briefcase, neither of
these is true given the strict view and the substitution theory.
52
Atemptingfixistogeneralize oversubstitutions. Insteadofrequiringapermissivebare
noun phrase F to contribute a particular contextual analogue of the property semantically
51
Recall that the reading of interest is one on which ‘might’ scopes wide of ‘every marked bill’. A clunkier
but more perspicuous way of putting the reading is ‘Scarlet and White agree that it might be that every
marked bill is a hundred.’.
52
One might hope to avoid this problem by using something disjunctive like ‘bill which is in the black
briefcase or in the blue bag’ as a contextual analogue for ‘marked bill’ here. But this immediately faces
problems. The most pressing is that we can construct the case so that neither Scarlet nor White possesses
the relevant loaded attitude. Just let Scarlet have a loaded belief that some bill in the green bag is a twenty
and let White have a loaded belief that some bill in the black briefcase is a twenty. So neither believes that
every bill which is in the black briefcase or in the blue bag might be a hundred.
53
associated withF, we could say instead that the attitude report involvingF is true if there
is some contextual analogue or other of that property which, if substituted for it, would
yield a true report.
53
Consider how this generalized substitution theory might handle Marked
3
. First note
that it won’t do to capture the permissive reading of ‘marked bill’ in (33) with
(33R)9 G.R(G,[marked bill]) [Scarlet and White agree [Might [every G is a hundred]]]
For there is no one contextual analogue of marked-bill which figures in both Scarlet’s
and White’s relevant beliefs. The generalized substitution theory avoids this problem,
however, by allowing the existential quantifier to scope elsewhere, as in
(33R’) [Scarlet and White]
x
[9 G.R(G,[marked bill]) [x agrees [Might [every G is a
hundred]]]]
Here contextual analogues covary with subjects. Since Scarlet and White are each such
thatthereissomecontextualanalogueof marked-billwhichfiguresintheirrelevantbelief,
(33R’) accounts for the permissive reading of ‘marked bill’ without requiring a particular
contextual analogue to figure in both Scarlet and White’s reported attitudes.
54
The generalized substitution theory is as compatible with the strict view as is the orig-
inal substitution theory. On the permissive reading of ‘marked bill’ in (33) in the context
of Marked
3
, for example, the property marked-bill is not a constituent of what Scarlet
and White are said to believe. So there is no violation of the strict view. Furthermore,
as with the original substitution theory, this explanation of permissive readings relies on
the truth of the strict view. For although what detective Black asserts doesn’t represent
Scarlet and White as cognizing some property in particular, that assertion is true only if
there are such properties.
Evenso, aproblemfacesthegeneralizedsubstitutiontheoryonceweattendtoacertain
featureofagreementreports. Whenstandardlycontext-sensitiveexpressionsareembedded
in agreement reports, their interpretation cannot covary with the subjects of the report.
Consider
(36) Scarlet thinks that a nearby bar has a great happy hour.
53
Schwager (2009)’s theory has this feature.
54
Regimentation (33R’) itself is too charitable towards substitution theories. For ‘agree’ is non-
distributive: for White and Scarlet to agree that the sun is shining is not for Scarlet alone to “agree”
that the sun is shining and for White alone to “agree” that the sun is shining. Agreeing is a group a↵ air.
(33) should be regimented with a plural subject, as in
[Scarlet and White]Y [9 G.R(G,[marked bill]) [Y agree [Might [every G is a hundred]]]]
That is, roughly, the plurality Y consisting of Scarlet and White is such that there is a contextual
analogueG of ‘marked bill’ and theYs agree that it might be that everyG is a hundred. But this just gives
us one contextual analogue which is substitution within the object of Y’s agreement, losing the varying of
contextual analogues with di↵ erent members of Y.
54
Giventhecontext-sensitiveexpression‘nearby’, (36)admitsoftworeadings. According
to one, Scarlet that thinks a bar near to her has a great happy hour; on another, Scarlet
thinks that a bar near to the speaker has a great happy hour. Now suppose that Scarlet
thinks a bar near Scarlet has a great happy hour and that White, far from Scarlet, thinks
that a bar near White has a great happy hour. Detective Black cannot report these
attitudes with
(37) Scarlet and White agree that a nearby bar has a great happy hour.
There is no reading of (37) on which the location relevant to ‘nearby’ covaries with
the subjects Scarlet and White. The only available reading is one on which there is one
locationrelevantto‘nearby’,thelocationofutterance.
55
Thatis,(37)canonlybecorrectly
regimented by (37R), not by (37R’):
(37R)9 l. [Scarlet and White agree [a bar near l has a great happy hour]]
(37R’) [Scarlet and White]
x
9 l. [x agrees [a bar near l has a great happy hour]]
This raises a troubling question for the generalized substitution theorist: what entitles
us to allow an intermediate scope for the quantification over contextual analogues but not
for other context-sensitive expressions? To obtain the desired reading of (33) at best re-
quiresad hoc allowancesregardinghowagreementreportsinteractwithcontext-sensitivity.
So again, though contextual analogues may help in explaining permissive inconstant bare
noun phrases, that explanation cannot be a matter of substituting analogues for the prop-
erties conventionally associated with those phrases. In the next section we’ll see how the
machinery of contextual analogues might be used to explain permissive inconstant reports
outside of the substitution theory.
6. The Sensitive Attitudes Account of Permissive Reports
At the end of §2, permissive inconstant readings left us at a crossroads. To account
for such readings we had to take one of two paths: amend the assumption that bare noun
phrases occurring in a ‘that’-clause contribute conventionally associated properties to the
proposition designated by that clause, or amend the assumption that these constituents
constrain whether the subject can entertain that proposition in accordance with the strict
view. Takingthefirstpathbymakinguseofcontextualanaloguesledustotheproblemsof
erroneous assumptions and uncoordinated agreement. Now we should see where the other
path takes us.
A diagnosis of the problems facing the substitution theory is that the pragmatic ma-
chinery which it deploys is built into the propositions asserted by speakers who use bare
noun phrases in a transparent way. That is, the substitution theory takes contextual ana-
logues themselves to figure as constituents of the propositions said to be believed. The
final theory I consider makes recourse to similar pragmatic machinery while casting o↵ this
55
SeeHawthorneandCappelen(2009)foradiscussionofagreementreportsasdataforcontext-sensitivity.
55
commitment. In contrast to sensitive contents accounts, which explain the distribution of
permissive readings in terms of context-sensitivity in what propositions a report ascribes
as attitude contents, “sensitive attitudes accounts” explain permissive readings in terms of
context-sensitivity in what attitude relations a report represents one as bearing.
Sensitive attitude accounts consists of two main ideas. First, which relation is con-
tributed by an attitude verb is context-sensitive. Following Dorr (2014), the sensitive
attitude theory claims that a use of an attitude verb like ‘believes’ can serve to contribute
distinct relations in distinct contexts.
56
There is no one belief relation believes,but
several: believes
c
1
, believes
c
2
, etc., one for each context c
1
, c
2
,etc.
57
Second, which
relation V
c
is contributed is by an attitude verb V in a context c is constrained at least by
assumptions made by the conversationalists of c.
We’ve already seen variation among sensitive contents accounts of permissiveness con-
cerning the particular nature of the context-sensitive mechanisms at work in permissive
readings: the scope theory, plurality theory, and substitution theories each made distinct
particular proposals about what proposition a permissive reading of some report sentence
was. Similarly there is variation among potential sensitive attitudes accounts, given what
eachsaysabouttheapplicationconditionsofthecontextually-determinedattituderelations
contributed by verbs like ‘believe’. I am less interested here in defending a full sensitive
attitudes account than I am in showing that this view has the right shape to make sense
of the range of permissive readings we have heretofore encountered.
Since I don’t want to be distracted by the details here, I’ll just partially sketch one
way of developing the sensitive attitudes account of permissiveness which I take to be
empirically superior to the competing substitution theory (even if it cannot account for all
cases of permissiveness, such as those set aside in §2.1). On this sensitive attitude account,
wemakeuseofthenotionofcontextualanalogues,butinsteadofclaimingthatacontextual
analogueof property Fmay be substituted for Fin what asubject isclaimed tobelieve, this
view claims that contextual analogues play a role in determining the application conditions
of the belief relation contributed by ‘believes’ in the context. I’ll call this illustrative view
“sensitive attitudes analogues”:
SAA:Asubject a bears relation V
c
to proposition p(F) containing property F as a
constituent only if there is a contextual analogue of F in c, G, such that a
possesses a concept of G and bears V
L
to p(G).
Let’s unpack some of symbols here. First, V
c
is whatever attitude relation contributed
by a use of an attitude verb V in context c. Next, p(F) is a proposition in which the
56
See Goodman (2018) for a similar view as it relates to attributions of singular thoughts. Schwitzgebel
(2002)’s dispositional view of belief can also be understood as a sensitive attitudes view, although it is
unclear if he takes it this way.
57
Ultimately, contexts should be individuated in quite a fine-grained way, so that multiple belief relations
might be contributed by di↵ erent occurrences of the same verb in a single sentence. I ignore this subtlety
in presenting sensitive attitude accounts here.
56
property F occurs and p(G) is a proposition structurally just like p(F) save that property G
occurs in the place of F. And last, V
L
is the loaded V-relation: it is an attitude relation for
which the strict view does hold. For any agent a and proposition p, necessarily a bears V
L
to p only if a possess concepts of every constituent of p.
58
With all this, we can say that according to SAA, the report made by (11) in Marked is
trueonlyifScarletbearsbelieves
Marked
tothepropositionthateverymarkedbillmightbe
ahundred,whichshedoesonlyifshepossessaconceptofapropertyGwhichisacontextual
analogue of marked-bill in Marked and bears believes
L
to the proposition that every G
might be a hundred. Assuming that bill-in-black-briefcase is a contextual analogue
of marked-billinMarked,sinceScarletpossessesaconceptof bill-in-black-briefcase
and does have a loaded belief that every bill in the black briefcase might be a hundred,
SAA allows that (11) may be true in Marked. The “may be” here shows that SAA falls
short of a full truth-conditional account of permissiveness; I state SAA just as a necessary
condition on the applications of contextually-determined attitude relations because there
may be further conditions associated with di↵ erent attitude verbs, which we’ll see in a
moment. But for the case of ‘believe’ we can adopt a version of the sensitive attitudes
account on which nothing else is required for the truth of (11).
We appeal to further verb-specific conditions in handling the problems facing the sub-
stitution theory. Recall the problem of erroneous assumptions raised by Marked
2
,inwhich
the detectives’ assumption that the marked bills are all and only the bills in the black
briefcase is false. On SAA, detective Black is said to assert a proposition corresponding to
(32**) Scarlet knows
Marked
2
[every marked bill]
x
[x is in the lounge].
On the assumption that bill-in-the-black-briefcase is a contextual analogue of
marked-billinMarked
2
, thepropositioncorrespondingto(32**)istrue(givenSAA)only
if Scarlet bears the loaded knowledge relation knows
L
to the proposition that every bill in
the black is in the lounge. She meets both of these requirements. However, we should want
that a general condition on any contextually-determined knowledge relation, permissive
or not, is factivity: for any knowledge relation knows
c
and agent a and proposition p,
necessarily a bears knows
c
to p only if p is true. The version of the sensitive attitudes
account which we are adopting, then, will take this principle on board as constraining the
application conditions of any knowledge relation.
But now recall that in Marked
2
it isn’t true that every marked bill is in the lounge. So
the factivity constraint entails that no person bears knows
Marked
2
, the relation contributed
58
It is worth noting that there may be more than one totally loaded relation associated with a particular
attitude verb. Dorr (2014)’s version of the sensitive attitudes view is this way. Scarlet can have a loaded
belief that Clark Kent flies under a Clark Kent concept and White can have a loaded belief that Clark
Kent flies under a Superman guise. They each are loaded in that their truth requires the subject to possess
concepts of every constituent of the proposition believed, but di↵ er in which concepts are required. I
ignore such complications here, so the present sensitive attitudes view just shared with Dorr’s that multiple
attitude relations are associated with any given attitude verb.
57
by ‘knows’ here, to the proposition that every marked bill is in the lounge. Accordingly,
Scarletdoesnotbearthisrelationtothatproposition,andsothesensitiveattitudesaccount
predicts that the report made by (32) is false. In this way, the sensitive attitudes theory is
abletocapturethefalsepermissivereadingof(32)inMarked
2
eventhoughtheproposition
that every bill in the black briefcase is in the lounge, which Scarlet does know and which
is erroneously assumed to be equivalent with what Scarlet is said to know, is true.
Theproblemofuncoordinatedagreementisalsoeasilyhandledbythesensitiveattitude
theory. RecallthatinMarked
3
ScarletandWhitethinkofthethingswhicharethemarked
bills in di↵ erent ways. Scarlet thinks of them as bills in the black briefcase, White thinks
of them as things in the blue bag. This licenses Black’s report:
(15) Scarlet and White agree that every marked bill might be a hundred.
The sensitive attitude theory captures the transparent inconstant reading of ‘marked
bill’ here by claiming that Black asserts the proposition corresponding to
(15**) Scarlet and White agree
Marked
3
[Might [every marked bill]
x
[x is a hundred]]
Since both bill-in-the-black-briefcase and bill-in-the-blue-bag are contex-
tual analogues of marked-bill in Marked
3
, SAA allows that both Scarlet and White bear
believes
Marked
3
to the proposition that every marked bill is a hundred. On the (surely
over-simplified) verb-specific assumption that for any agents a and b necessarily if a bears
believes
c
to p and b bears believes
c
to p then the pair of a and b bears agrees
c
to
p, we have that the sensitive attitudes account predicts that (15) is true in this context.
This treatment avoids the problem facing the substitution theory just because it doesn’t
representthesubjectsasagreeinginvirtueofentertainingdistinctpropositions. Theinter-
pretationofthecontext-sensitive‘believes’isuniform, buttheapplicationconditionsofthe
relation that verb contributes in Marked
3
allow Scarlet and White to cognize marked-bill
in virtue of di↵ erent concepts. So an answer to the problem of uncoordinated agreement
simply falls out of the sensitive attitudes view given by SAA. Even if SAA should ulti-
mately fail to account for the full range of permissive reports (as I argue it does in the
next chapter), we’ve seen how the resources of the sensitive attitudes view can be used to
account for permissive inconstant reports without running into the main problems facing
sensitive contents views. The sensitive attitudes view has the right shape to make sense of
permissiveness, so we should consider building on this foundation going forward.
We began with the compelling idea that if one believes a proposition then one must be
abletothinkaboutthepropertieswhichthatpropositionisdirectlyabout. Iexpressedthis
asthestrictview: necessarilyonebearsapropositionalattitudetoapropositiononlyifone
possesses concepts of every constituent of that proposition. I have made a case against the
strict view by arguing in favor of the sensitive attitudes view, which seems incompatible
with the strict view (insofar as among the designation candidate of attitude verbs there
58
are conceptually permissive attitudes) and against sensitive contents views, which look like
the only views that can be used to make sense of permissive readings on the strict view.
Now, despite the range of cases which put pressure on the strict view, the argument
of this paper doesn’t constitute an airtight case against that view. Perhaps some yet
unconsidered refinement of a sensitive contents account could make sense of permissive
inconstant readings; such an account could be used to protect the strict view. But even
if you are unconvinced that the strict view is false, I hope to have convinced you that it’s
far less than obviously true. What looked at first like an obvious conceptual constraint on
empirical theorizing about attitudes now is in need of supporting argument. Given this,
it’s worth entertaining the hypothesis that philosophers have made a mistake in adopting
such general acquaintance constraints on attitudes and seeing where the alternative, the
conceptually permissive attitudes view, takes us.
59
Chapter 3. Contextual Augmentation Accounts of Permissive Reports
Permissive readings of bare noun phrases seem to be a special case of a more general
linguisticphenomenonwhichextendstoadjectives, determiners, andnames. Thisfactputs
considerable pressure on the sensitive attitudes view as it was developed in the previous
chapter. In this chapter I consider another way of developing the basic thought behind
the sensitive attitudes view, which I call “contextual augmentation views”. This family
of views account for the conditions in which an attitude report is permissive roughly in
terms of how the attitude content which it serves to ascribe relates to the assumptions
made in the reporting context. In particular, such views account for permissive reports in
terms of whether the report’s complement is entailed (in some sense) by the contents of
the report’s subject’s fully loaded attitudes together with certain assumptions which are
made in the reporting context. The sensitive attitudes analogues view, from the end of
chapter 2, is one of many ways of spelling out the contextual augmentation view. Though
it doesn’t work in full generality, other ways of spelling out this view do somewhat better.
But we will see that even these other contextual augmentation views aren’t perfect either.
After introducing and refining such views, I raise five broad problems with contextual
augmentation views in general. My discussion suggests that to account for the conditions
in which a report is or may be permissive, we must take into account further features of
conversational contexts than just what the conversationalists commonly assume. These
di↵ erent problems will serve as data to guide discussion in later chapters.
1. Permissive Reports
In chapter 2 introduced permissive reports more or less by ostension. Let me now
characterize them a little more carefully. As a first pass, I’ll say that a use of a clause R
of the formAVsthatS in a context c takes a permissive reading if it is possible that the
proposition asserted by (or designated by)R inc is true but a subject designated byA inc
lacksaconceptconversationallyassociatedwithalinguisticconstituentofS. We’llseethat
this characterization needs to be refined in a few ways, but it does well in classifying the
examples discussed in the previous chapter. For example, reconsider the following report
from chapter 2:
(10) Mustard fears that every undercover cop will want to play bridge.
In the context Bridge from §1.2.2, Mustard was said to be unaware that there were
any undercover cops in the manor. If it turned out, somewhat bizarrely, that he altogether
lacked a concept of the property undercover-cop, which is associated by the conversation-
alists in Bridge with ‘undercover cop’, the report made by Black’s use of (10) nevertheless
would have been true. So we have it that it is possible that the subject designated by
‘Mustard’ lacks a concept associated with a constituent of the ‘that’-clause in (10), namely
‘undercover cop’, yet the report made by (10) is true, so Black’s report takes a permissive
reading.
60
The central examples of permissive reports from chapter 2 similarly answer to this
characterization. But there are other kinds of reports which have not been discussed in
detail which also answer to this characterization. Consider the following three examples:
Blackmail: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called into discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet: Green doesn’t know any name for Scarlet, could not identify her in a line-up,
and so on. By questioning Green, detective Black learns that Green thinks that every
socialist guest was being blackmailed by Mr. Body. Green doesn’t, however, know the
names of any of the socialist guests: he claims “not to know who the socialist guests are”.
The detectives are exchanging information, trying to find out who had a motive for killing
Body. Black assertively utters the following to Grey:
(38) Green thinks that Scarlet was being blackmailed by Mr. Body.
What Black asserts in uttering (38) is true, even though Green apparently lacks a con-
cept associated in Blackmail with ‘Scarlet’. (In thiscase, it could even bethat helacks any
concept of Scarlet, the individual designated by ‘Scarlet’, but our current characterization
doesn’t require anything so strong.)
59
What our going characterization requires for a per-
missive reading is that it be possible that the subject verifies the report yet lacks a concept
associated with a constituent of the ‘that’-clause which characterizes the content of their
reported attitude. We will get further into this in chapter 6, but for now can say that
Black and Grey associate a concept with ‘Scarlet’ which picks out Scarlet by her name.
Green lacks any such concept, so (38) counts as a permissive report.
Consider another example:
Society: Mrs. White has informed detective Black that she has it on good authority
thateverywomaninthestudyisamemberofasecretsociety. Shedoesn’tknowhowmany
women are in the study or who in particular they are, nor does she know if there are more
people in the study than women. But she believes that every woman in the study is in a
secret society. The detectives know that in addition to the three women in the study there
are two men. In discussing the people in the study, Black assertively utters the following
to Grey:
(39) White thinks that most people in the study are members of a secret society.
59
It can prove di cult to establish that a subject like Green lacks any concept whatsoever of Scarlet.
Suppose that Scarlet is the fourth tallest person in the manor; it’s plausible that Green possess a concept
of the property of being the fourth tallest person the manor, in which case he arguably does possess a
concept of Scarlet (a concept of the fourth tallest person in the manor), even though intuitively this does
not undermine the relevant permissiveness of (38). What matters for whether a report takes a permissive
reading ultimately is whether the subject must use, or use in certain ways, a concept associated with
material related to the ‘that’-clause of a report.
61
The report Black makes with (39) is permissive with respect to ‘most’: White does not
have a loaded thought that most people in the study are in a secret society. For all White
knows,thewomeninthestudymaybeoutnumberedbythemeninthestudy. Nevertheless,
Black’s report intuitively is true and appropriate. Though it would be bizarre for White
to altogether lack a concept of the relation contributed by ‘most’, intuitively it is possible
that she lack such a concept yet the proposition asserted by Black’s use of (39) be true.
One last example:
Scotch: Colonel Mustard is sipping di↵ erent brands of scotch and bourbon in the
lounge. For each kind of alcohol he sips, he declares whether he thinks it is peaty. He
growsprogressivelydrunk, sodoesn’tkeepcountofhowmanyscotcheshethinksarepeaty.
DetectiveBlack hasbeen watchingandkeepingtrackof which kindsof scotch andbourbon
Mustard declares to be peaty. After Mustard has tasted all the scotch and bourbon in the
lounge, Black consults her record to find that Mustard has declared fifteen scotches to be
peaty. Black assertively utters the following to detective Grey:
(40) Mustard thinks that fifteen scotches are peaty.
Black asserts something true in uttering (40). She does so even though Mustard is
in no position to tell how many scotches he thinks are peaty. We can imagine that were
Mustard asked about any particular scotch he tasted, he would reproduce his judgment
from before. But were he directly asked about how many scotches in the lounge were
peaty, he wouldn’t readily claim that fifteen are. In his state, he might even deny that
fifteen scotches are peaty, guessing instead that more than fifteen are peaty. However,
were Mustard to consult Black’s record in a sober moment, he would be in a position to
sum his previous judgments together, so could rationally conclude that he thinks fifteen
scotches in the lounge are peaty. Now, Scotch does not constituent strong evidence that
one can truthfully ascribe an attitude to a subject who lacks concepts of constituents of
the thought content thereby ascribed to them. For it is hard to imagine such a case in
which Mustard lacks a concept of fifteen (though perhaps such a case is possible). Still,
intuitively Mustard need not deploy any concept of fifteen in having any token thoughts
for (40) to be true as uttered in Scotch. So it seems possible that Black’s report be true
yet Mustard lacks the concept associated with ‘fifteen’.
According to our going characterization, these are all permissive reports. They all
share in common that for the subject of the report to bear the reported attitude does not
require that subject to possess a concept associated with a linguistic constituent of the
‘that’-clause of the report. But the sensitive attitudes analogues view from chapter 2 has
trouble accounting for some of these reports. On that view, there must be a contextual
analogue of ‘Scarlet’, ‘most’, and ‘fifteen’, in the relevant conversational contexts, such
that the relevant subject bears a loaded attitude towards a proposition by substitution of
analogues in the relevant scenario. In each of these contexts, there seems to be no ready
contextual analogue to serve such a role.
62
InthenextsectionIconsiderotherwaysinwhichonemighttrytodevelopthesensitive
attitudes view to make sense of these and other cases of permissive reports.
2. Contextual Augmentation Theories
I’d like to start by considering and dismissing one natural thought about what is going
on in the casesfrom §1. You might think that what is goingon in these reportsis that they
are true because were the relevant subject to learn some auxiliary assumption made by
the conversationalists, then they would bear the reported attitude to the reported attitude
content. For example, were Green to learn that Scarlet is a socialist guest, then Green
would come to believe that Scarlet was being blackmailed by Body. Similarly, if White
were to learn that there are three women and two men in the study, then she would come
to think that most people in the study are members of a secret society. And if Mustard
were to learn that fifteen scotches had been described by him as peaty, then he would come
to think that fifteen scotches in the lounge are peaty. We can regiment this as the
Counterfactual Attitudes View:
A report of the formAVsthatS made in context c is true if and only if there are
contextually assumed propositions p
1
...p
n
in context c such that if were A to learn
p
1
...p
n
then A would bear V
L
to the proposition designated by ‘that S’in c.
HereV
L
isthe“fullyloaded”V relation,suchasthefullyloadedbelieves relation,which
one bears to a proposition necessarily only if they possess concepts of every constituent of
that proposition. This is a little rough, but it’s easy to see the basic idea’s appeal. For
example, in Scotch it looks like were Mustard to learn that the scotches he thought were
peaty, s
1
, s
2
,... s
15
, were fifteen in number, then we would readily come to believe that
fifteen scotches in the lounge are peaty. Similarly, if Green were to learn that Scarlet is a
socialist,thengivenhisgeneralbeliefaboutsocialists,hewouldreadilycometobelievethat
Scarlet was being blackmailed by Body. The thought, then, is that to bear the permissive
belief relation designated by ‘believes’ in context c towards a proposition p is to be such
that you would believe p (in a fully loaded way) were you to learn certain contextual
assumptions of c.
60
Despite its initial appeal, the Counterfactual Attitudes View won’t do. To see this,
consider the following example of a permissive report:
Forgeries: DetectivesBlackandGreyknowthatallandonlytheimpressionistpaintings
inthegalleryareforgeries, butthatthisfactisawellkeptsecret, sothatnoneoftheguests
inthemanorknowthatthereareanyforgeriesinthegallery. Peacock,aguestinthemanor,
has told detective Black that she believes that every impressionist painting in the gallery
is an original Renoir. Later, in discussing matters related to the gallery, Black assertively
60
As we will see with certain counterexamples facing contextual augmentation views, the counterfactual
attitudes view which is stated in terms of “any/a contextual assumption” rather than “certain contextual
assumptions” would be too weak.
63
utters the following to Grey:
(41) Peacock believes that every forgery in the gallery is an original Renoir.
Here Black makes a permissive report: it is possible that her assertion be true yet
Peacock lack the concept of forgery contextually associated with Black’s use of ‘forgery’.
HowdoesCounterfactualAttitudesView handlesuchacase? Thetheoryrequiresthatthere
be some contextual assumption such that were Peacock to come to believe that proposition
then she would come to have a loaded belief that every forgery in the gallery is an original
Renoir. But there seems to be no such contextual assumption available. Intuitively the
relevant contextual assumption bridging Peacock’s loaded attitudes to the belief content
reported by (41) is the assumption that all and only the impressionist paintings in the
gallery are forgeries. But presumably were Peacock to come to learn this proposition, then
she would not maintain her prior belief that every impressionist painting in the gallery
is an original Renoir. She would, accordingly, not come to believe that every forgery
in the gallery is an original Renoir; instead she would just give up her belief that every
impressionist in the gallery is a Renoir. It is furthermore intuitive that there is no other
contextual assumption which the counterfactual attitudes view could use to generate the
correct prediction in Forgeries.
As I work through the problems in this chapter, I will point out other cases that
constitute counter-examples to the Counterfactual Attitudes View. Many of the cases that
tellagainstcontextualentailmentviewsalsotellagainstitscounterfactual-theoreticcousin.
AlthoughthesortofconditiongivenbyCounterfactual Attitudes View doesnotcapture
the relevance of contextual assumptions to the truth conditions of permissive reports,
contextual assumptions intuitively must be taken into account. One thing the permissive
reports from §1 share in common is that each intuitively exploits special conversational
assumptions. Uses of (38)-(41) in di↵ erent contexts may fail to assert truths, depending
on what the conversationalists do or do not commonly assume. For instance, imagine a
context just like Forgeries except in that the detectives do not assume that all and only
the impressionist paintings in the gallery are forgeries. In such a context, Black would not
assert something true in uttering (41). Similar observations can be made for Blackmail,
Society, and Scotch.
It looks like a report may be permissive if objects of the report’s subjects loaded atti-
tudes together with assumptions made in the reporting context simply entail the propo-
sition contributed by the complement of the relevant report. In Blackmail, for example,
Greenhasaloadedbeliefinthepropositionthateverysocialistguestwasbeingblackmailed
by Mr. Body. This proposition together with the detectives’ assumption that Scarlet is a
socialist entails the proposition contributed by the complement of (38). An alternative to
the counterfactual attitudes view, then, is that whenever this kind of entailment relation-
ship holds between the objects of a subject A’s loaded attitudes, assumptions made in the
reporting context, and a proposition p, a report made in that context that A bears the
64
relevant attitude to p made is (or may be) permissive. I will call any view of this general
shape a “contextual augmentation” view. I’ll give a general statement of this view as:
Contextual Augmentation View:
A report of the formAVsthatS made in context c is true if and only if there are
contextually assumed propositions p
1
...p
n
in context c and propositions q
1
...q
m
to
whichA bears V
L
such that {p
1
...p
n
, q
1
...q
m
} entails the proposition designated by
that S in c.
Tomakethecontextualaugmentationviewintoasensitiveattitudesaccount,weassume
that for each reporting context c and attitude verb V, there is a particular relation V
c
which a use of V contributes in c. V
c
need not be the “fully loaded” attitude relation. The
contextual augmentation view thus indirectly characterizes the application conditions of
V
c
, for any context c.
One need not, however, accept a sensitive attitudes view in accepting the contextual
augmentation view. One who likes the contextual augmentation view could, for example,
endorsean“insensitiveattitudes”view,accordingtowhicheachattitudeverbV issuchthat
a use ofV only ever contributes a single relation V which takes three arguments: a subject,
asetofcontextualassumptions, andaproposition. Onthisinsensitiveattitudescontextual
augmentationview, usesofaverblike‘believe’alwayscontributeasinglerelation believe,
which relates a subject to a belief content and to assumptions provided by the context.
Let’s contrast this view with the sensitive attitudes contextual augmentation view:
Sensitive Contextual Augmentation View:
AbearsV
c
topropositionr ifandonlytherearecontextuallyassumedpropositions
p
1
...p
n
incontextcandpropositionsq
1
...q
m
towhichAbearsV
L
suchthat{p
1
...p
n
,
q
1
...q
m
} entails r.
Insensitive Contextual Augmentation View:
A bears V to proposition r and contextual assumptions p
1
...p
n
if and only if there
are propositions q
1
...q
m
to which A bears V
L
such that {p
1
...p
n
, q
1
...q
m
} entails r.
On the first view, which relation is contributed by a use of an attitude verb V varies
across contexts. On the second view, all uses of an attitude verb V make a uniform
contribution to propositions asserted, but that relation is three-placed rather than two-
placed, so there is no question of whether one bears V to a single proposition r.
Whether to prefer the sensitive or insensitive version of the contextual augmentation
view is a purely semantic question. In the rest of this chapter, I will focus mainly on
the truth-conditional question of whether a contextual augmentation view of either sort is
adequate. The truth-conditional question of present interest is
TCQ: Under what conditions is a report of the formAVsthatS made in context c
true?
65
It looks at first as though the sensitive and insensitive versions of the contextual aug-
mentationviewwillgivethesameanswertothistruth-conditionalquestion,merelyarriving
at that answer in di↵ erent ways. For this reason I will ignore the di↵ erences between these
two contextual augmentation views in the rest of the chapter, focusing just on my favored
version, the sensitive contextual augmentation view.
3. The Sensitive Contextual Augmentation View
On the sensitive contextual augmentation view, we begin with the idea that for every
agent a, world w, and time t, there is a set of propositions which in some privileged,
context-independent sense, are a’s beliefs in w at t. (We might identify this set with the
set of propositions which are objects of a’s fully loaded beliefs in w at t, but we need not in
principle.) AlthoughIdon’tlikethiswayoftalking,onemightviewthepropositionsinthis
set (which I’ll call a’s “belief set” in w at t, or ‘B
w,t
a
’) as those which a “really” believes.
It is open question whether ‘believes’ ever contributes this fundamental belief relation
believes
L
; perhaps this verb never contributes this relation, given the pervasiveness of
contextual influence on assertion. Still, I shall assume that such a fundamental relation
exists.
61
Next we allow that there may be contexts in which a speaker can “get away with”
claiming that a subject a “believes” (in world w at time t) some proposition p not in B
w,t
a
.
Such reports count as true not because a bears the fully loaded belief relation believes
L
to p, but because there are some propositions in B
w,t
a
which bear a special relationship to
p relative to the given context.
To fill out contextual augmentation views, we have to say what this special contextual
relationship is. The first proposal I consider, motivated by my rough comments above,
is that this relation is one of “contextual entailment”. I assume that at any point in a
conversation context c, the conversationalists in c commonly assume some propositions;
I call these the “contextual assumptions” in c and write ‘CA
c
’ for the set of contextual
assumptions in c. I’ll say that a set of propositions {q
1
,...,q
n
} “contextually entails” a
propositionp relative to contextc if propositionsq
1
,...,q
n
together with some propositions
fromCA
c
entail p.
62
With this terminology, we can specify a particular theory concerning
the conditions under which a subject a bears an attitude relation V
c
to a proposition p.
The basic idea is thata bears attitude V
c
top ifV
a
contextually entailsp inc.
63
Where
61
In later chapters, I argue that there are several fully loaded believes relations, and similarly for other
attitude types.
62
Contextual entailment relations thus relate sets of propositions to propositions, but I will often speak of
an individual propositions p contextually entailing a proposition q instead of speaking of {p} contextually
entailing q.
63
On van Fraassen (1979)’s way of developing the idea, a bears Vc to p roughly if some one proposition q
in attitude set Va contextually entails p in c (unless it’s relevant, going forward I’ll ignore the relativization
of attitude sets to worlds and times). Parikh (1998) extends this idea by claiming that a bears Vc to p if
every proposition in Va together with every proposition assumed to be necessary in c,entails p. Versions
of my complaints against (CE) below carry over to these proposals.
66
)
c
is the relation of contextual entailment in c and V
a,c
is the set of propositions to which
a bears V
c
, I regiment the proposal as:
(CE) V
a,c
= {p :V
a
)
c
p}
Applying (CE) to the case of belief, we say thata bears believes
c
top (so,p is in B
a,c
)
if a’s belief set contextually entails p in c. The contextual entailment theory accordingly
claims that if a’s belief set contextually entails p in c then what is asserted by a use of a
report sentence of the form A believes that S in c,where A designates a in c and that S
designates p in c,istrue.
The Sensitive Contextual Entailment View, supplemented with (CE), straightforwardly
handlestheexamplesfrom§1. InBlackmail, sincethepropositionthateverysocialistguest
was being blackmailed by Mr. Body is in Green’s “thinks set”, T
Green
, and the proposition
that Scarlet is a socialist is in CA
Blackmail
, it follows that B
Green
)
Blackmail
b,where b is
the proposition that Scarlet was being blackmailed by Body. Hence (38) is true as uttered
by Black in the context of Blackmail given (CE).
In Society, since the proposition that every woman in the study is in a secret society
in White’s belief set, B
White
, and the proposition that three women and two men are in
the study is inCA
Society
, it follows that B
White
)
Society
r,where r is the proposition that
most people in the study are members of a secret society. Hence (39) is true as uttered by
Black in the context of Society given (CE).
In Scotch, there are fifteen di↵ erent propositions in Mustard’s “thinks set”, T
Mustard
:
that scotch s
1
is peaty, that scotch s
2
is peaty, ..., and that scotch s
15
is peaty. Since the
proposition that the scotches s
1
,...s
15
are fifteen in number is in CA
Scotch
, it follows that
T
Mustard
)
Scotch
p,where p is the proposition that fifteen scotches are peaty. Hence (40)
is true as uttered by Black in the context of Scotch given (CE).
In Forgeries, since the proposition that most impressionist paintings in the gallery are
original Renoirs is in Peacock’s belief set, B
Peacock
, and the proposition that all and only
the impressionist paintings in the gallery are forgeries in the gallery is in CA
Forgeries
,it
follows that B
Peacock
)
Forgeries
r,where r is the proposition that most forgeries in the
gallery are original Renoirs. Hence (41) is true as uttered by Black in the context of
Forgeries given (CE).
This explanatory success of the simple contextual entailment view speaks in its favor.
Framing contextual augmentation views in terms of the sensitive attitudes theory, one can
give a unified explanation of permissive reporting by claiming that a subjecta bears a per-
missive attitude relation V
c
contributed by an attitude verbV in contextc to a proposition
p if and only ifa bears the corresponding loaded attitude relation V
L
to propositionsq
1
...q
n
such that {q
1
,...,q
n
} contextually entail p in c. This tidy view accounts for the permissive
interpretations of (38), (39), (40), and (41) in §1.
Even so, contextual entailment views face a variety of counterexamples. In the rest of
this chapter I work through five di↵ erent problems facing such views.
67
4. Irrelevant Entailment Problems
Despite its ability to handle the examples from §1, the contextual entailment theory
faces a variety of over-generation problems. In this section I present examples which show
that propositions in an agent’s attitude set can contextually entail propositions which that
agent cannot be truly said to bear the relevant attitude toward. I call these problems of
“irrelevant entailment”. Some of these problems can be handled by slightly tweaking the
contextual entailment theory; others are more troubling.
4.1. Cheesy Irrelevant Entailment Problems
An immediately worrying irrelevant entailment problem is that in some cases a propo-
sition is in a subject’s attitude set but the negation of that proposition is contextually
assumed in a reporting context. Consider
Pipe: Mrs. Peacock believes that the pipe is the murder weapon. The detectives
know two things which Peacock doesn’t: first, the pipe is the only weapon hidden in
the conservatory; second, the pipe is not the murder weapon. Detective Black knows of
Peacock’s belief about the pipe, and assertively utters each of the following to Grey:
(42) Peacockbelievesthattheweaponhiddenintheconservatoryisthemurderweapon.
(43) Peacock believes that the candlestick is the murder weapon.
Black asserts something true by uttering (42), but asserts something false by uttering
(43). But according to the contextual augmentation theory given by (CE), she should
assert (or be able to be interpreted as asserting) something true in uttering either. For the
proposition p that the pipe is the murder weapon together with the proposition ¬p that it
is not the case that the pipe is the murder weapon entails every proposition, including the
proposition q that the candlestick is the murder weapon. But since p is in B
Peacock
and ¬p
is in CA
Pipe
, p )
Pipe
q,so q should be in B
Peacock,Pipe
, the set of propositions to which
Peacock bears believes
Pipe
.
Toavoidthisirrelevantentailmentproblem, thecontextualaugmentationtheoristcan’t
just choose to ignore p or claim that p isn’t really in Peacock’s belief set. That p is in
B
Peacock
is a crucial part of the contextual entailment theorist’s explanation of the truth
of (42).
Instead,onecouldreplacetherelationofcontextualentailmentwithatweakedrelation.
Say that a set of propositions S “non-trivially contextually entails” a proposition p in
context c if there is some subset S
of S and some subset CA
c
of CA
c
such that the
propositions in S
[ CA
c
together entail p but there is some proposition q which they do
not entail. Where )
⇤ c
is the relation of non-trivial contextual entailment, the contextual
augmentation theorist then can replace (CE) with
(NCE) V
a,c
= {p :V
a
)
⇤ c
p}
68
I call the resulting view the “non trivial contextual entailment view”. Developing a
contextualaugmentationviewwith(NCE)insteadof(CE)avoidstheirrelevantentailment
problem raised by (15) in Pipe. For p does not non-trivially contextually entail q in the
context of Pipe,since p and ¬p together classically entail every proposition. So the non-
trivial contextual entailment view isn’t forced to erroneously predict that (43) is true in
Pipe. But the theory still can predict the truth of (42) in Pipe,since p, together with
the proposition that the pipe is the only weapon hidden in the conservatory, entails the
propositionh, thattheweaponhiddenintheconservatoryisthemurderweapon, butthere
is something they don’t together entail, such as q. So Peacock’s belief set non-trivially
contextually entails h, hence the contextual entailment theory with (NCE) predicts (42)
to be true.
The non-trivial contextual entailment view faces other counter-examples. For not ev-
erything which follows from certain contextual assumptions alone can be truly claimed in
that context to be believed by any arbitrary agent. Consider
Forgeries
2
: The context is just like that of Forgeries, but detective Black sincerely
utters one of the following to detective Grey:
(44) Peacock believes that we are having this conversation.
(45) Peacock believes that all and only the forgeries in the gallery are impressionist
paintings in the gallery.
Detective Black asserts something false in uttering each of (44) and (45). But accord-
ing to the contextual entailment theory as supplemented by (NCE), utterances of these
should assert true propositions in this context. Since the proposition that Black and Grey
are having a conversation is commonly assumed between Black and Grey, it is among the
contextual assumptions in Forgeries
2
,so any set of proposition from Peacock’s belief set
non-trivially contextually entails in Forgeries
2
that they are having a conversation. Simi-
larly, since Black and Grey commonly assume that all and only the forgeries in the gallery
are impressionist paintings in the gallery, (45) should be true in Forgeries
2
as well given
(NCE).
To avoid this irrelevant entailment problem, the contextual augmentation theorist can
nextreplacetherelationofnon-trivialcontextualentailmentwithanothertweakedrelation.
Say that a set of propositions S “distinctively contextually entails” a proposition p in c if
there is some subset S
of S and some subset CA
c
of CA
c
such that the propositions in
S
[ CA
c
together entail p but the propositions in CA
c
alone do not entail p.
64
64
Note that we cannot simply say that a set of propositions S distinctively contextually entails a propo-
sition p in c if S contextually entails p in c but the propositions in CAc do not alone entail p.Thatwould
be too strong, for sometimes a subject can be said in a context c to believe a proposition which is not
in their belief set but which is indeed a contextual assumption in c,aswhen a thinks the F is G,the
conversationalists of c assume both that the F is the H and that the H is G.Incanbethatin c one may
truly say that a believes that the H is G,butthisisentailedbythecontextualassumptionsof c alone.
69
WhereV
c
istherelationofdistinctive,non-trivialcontextualentailment,thecontextual
augmentation theorist then can replace (NCE) with
(DCE) V
a,c
= {p :V
a
V
c
p}
I call the resulting view the “distinctive non-trivial contextual entailment view”. Sup-
plementing the contextual augmentation theory with (DCE) avoids the irrelevant entail-
mentproblemraisedbyForgeries
2
. FornopropositionsinPeacock’sbeliefsetdistinctively
non-trivially contextually entail that Black and Grey are having a conversation or that all
and only the impressionist paintings are forgeries. So this version of the contextual entail-
ment theory need not erroneously predict that (44) is true in the context of Forgeries
2
.
4.2. Harder Problems of Irrelevant Entailment
This concludes my friendly modifications of the contextual entailment theory. The
following instances of the irrelevant entailment problem are not so easily avoided. I’ll
present three problems of this kind. The first serves to introduce a feature of context-
sensitive theories which will be important in later chapters. Consider
Forgeries
3
: The context is just like that of Forgeries, but detective Black assertively
utters the following to detective Grey:
(46) Peacock believes that most forgeries in the gallery are original Renoirs and that we
are having this conversation.
Detective Black asserts something false in uttering (46). But the distinctive non-trivial
contextual entailment view predicts that Black asserts something true. For the contents of
Peacock’s loaded beliefs together with the detectives’ assumptions entail the proposition p
expressedbythecomplementof(46)withoutentailingeverypropositionandthedetectives’
assumptions do not alone entail p.
In the face of the potential counter-example raised by Forgeries
3
, a contextual aug-
mentation theorist may try to adopt further constraints on which relations can be con-
tributed by ‘believes’. For example, one can adopt a distributive constraint on any relation
believes
c
which can be contributed by ‘believes’ according to which an agent a bears
believes
c
top&q onlyifabearsbelieves
c
topandabearsbelieves
c
toq. Asbefore,for
atobear believes
c
toq,a’sbeliefsetmustdistinctivelynon-triviallycontextuallyentailq
relative toc. So for detective Black to assert something true by uttering (46) in Forgeries
3
,
Peacock’sbeliefsetmustdistinctivelynon-triviallycontextuallyentailthatBlackandGrey
are having the conversation they are. But as before, it doesn’t: so, Peacock doesn’t bear
believes
Forgeries
3
to the proposition that Black and Grey are having their conversation,
hence, given the proposed distributive constraint, doesn’t bear believes
Forgeries
3
to the
conjunction attributed to her by (46) in Forgeries
3
.
Itseems,then,thatthecontextualaugmentationviewmightavoidthepotentialcounter-
example raised by Forgeries
3
by adopting a special constraint on what relations can be
70
contributed by ‘believes’. This may invoke a new explanatory burden of justifying such a
constraint, but plausibly some such justification could be given.
The prospects for using this style of explanation for similar examples involving disjunc-
tion, however, are less promising. Consider
Protest: There are rumors that some guests at the manor are secretly anarchists, that
others are secretly socialists, and that others still are secretly fascists. From interrogating
Scarlet, detective Black knows that Scarlet believes that some anarchists or socialists will
protest tonight. Scarlet isn’t confident whether anarchists, socialists, or both will hold a
protest; she just believes that some people from those groups will put on some kind of
political demonstration tonight. The detectives know that the rumor that there are fascist
guests hasn’t gotten around to Scarlet. So they know that Scarlet neither believes nor
disbelieves that there are fascist guests; they don’t assume that she has any particular
fascist-related beliefs. The detectives further know that the rumors about anarchists and
fascists are unfounded: no guest is fascist or anarchist. But, as the detectives know, some
guestsaresocialists. AwareofScarlet’sbeliefs,BlacksincerelyuttersthefollowingtoGrey:
(47) Scarlet believes that some fascist or socialist guests will protest.
Detective Black asserts something false in uttering (47). Since Scarlet doesn’t believe
anything about fascists, she doesn’t believe that some guests who are fascists or socialists
will protest (even though she does have such beliefs about socialists alone). Now, the
detectives commonly assume that no guests are fascists or anarchists but that some guests
are socialists. Being reasonable, they also assume that every guest is a fascist or socialist if
andonlyiftheyareananarchistorsocialist. This,incombinationwiththatsomeanarchist
or socialist guests will protest, which is in Scarlet’s belief set, entails that some fascist or
socialist guests will protest. So Scarlet’s belief set distinctively non-trivially contextually
entails that some fascist or socialist guest will protest, so the contextual entailment theory
with (DCE) predicts (47) is used to assert something true, contrary to observed fact.
There seems to be no plausible constraint on the relation believes
Protest
which the
contextual entailment theorist could adopt to avoid this counter-example. For instance,
there is no hope for the kind of distributive constraint which helped answer the potential
counter-example raised by (46). For belief (or rational belief) does not distribute over
disjunction; indeed, believes
c
can relate a subject a to the disjunction p_ q without
relating a either to p or to q. For example, it isn’t true that one believes that a coin will
come up heads or tails only if one believes that the coin will come up heads or believes
that the coin will come up tails. So just because Scarlet’s belief set doesn’t distinctively
contextually entail that some fascist guests will protest isn’t enough to block the erroneous
prediction which the contextual entailment theory makes regarding (47).
Althoughthecontextualaugmentationviewcanavoidsomeirrelevantentailmentprob-
lems by appealing to the notion of distinctive non-trivial contextual entailment, it can’t
avoid all of them. Some further constraints on which contextual entailments are relevant
71
to attitude reports is required.
Toillustratefurthertheneedforspecialconstraintsoncontextualentailment,Icontinue
my criticism of the contextual augmentation theory with some further counter-examples.
Each involves simple kinds of entailment which appear not to give rise to true permissive
readings.
Chase: Professor Plum is chasing Mr. Green through the manor. Given the trajectory
which he was just running along, Plum believes that Green is either in the billiard room or
in the library. The detectives know that Green didn’t go into the library. Aware of Plum’s
belief, detective Black sincerely utters the following to Grey:
(48) Plum believes that Green is in the billiard room.
Detective Black asserts something false in uttering (48). In Chase,Plumintuitively
suspends judgment about whether Green is in the billiard room, hence does not believe as
much. But the contextual entailment theory predicts otherwise, given that the proposition
b_ l, that Green is either in the billiard room or the library, is in B
Plum
, and ¬l is in
CA
Chase
.Since b_ l and ¬l entail b without entailing every proposition, and l alone does
not entail b, B
Plum
distinctively non-trivially contextually entails b. So according to the
distinctive non-trivial contextual entailment view, Black asserts something true in uttering
(48), contrary to observed fact.
Knife: Professor Plum has been conducting his own investigation about the murder.
Heisn’tconfidentyetwhichweaponisthemurderweapon,buthethinksthatifthemurder
happened in the ballroom then the knife is the murder weapon. The detectives know that
the murder happened in the ballroom. Aware of Plum’s beliefs, Black sincerely utters the
following to Grey:
(49) Plum believes that the knife is the murder weapon.
DetectiveBlackassertssomethingfalseinuttering(49). SincePlumsuspendsjudgment
about which weapon is the murder weapon, it is very di cult to hear Black’s use of (49)
in Knife as asserting something true. But Plum’s belief set distinctively non-trivially
contextually entails that the knife is the murder weapon. So the contextual entailment
theory here too issues a false positive, predicting that (49) should be true in Knife.
Marbles: The detectives know that there is an urn in the library; they know that there
are exactly ten red marbles in in the urn and that the urn contains only red and blue
marbles, but they are not sure how many marbles total are in the urn. Detective Black has
just spoken to professor Plum, who told Black that he thinks there are exactly five blue
marbles in the urn. The detectives are discussing how many marbles are in the urn, and
Black sincerely utters the following to Grey:
(50) Plum thinks that there are (exactly) fifteen marbles in the urn.
72
It is di cult to hear (50) as permissive and true in this Marbles.ButPlum’sbeliefset
in this context distinctively non-trivially contextually entails that there are exactly fifteen
marbles in the urn. So again the contextual entailment theory appears to issue a false
positive, predicting that (50) should be true in Marbles.
To summarize: even if we modify the contextual entailment theory by using a relation
of distinctive non-trivial contextual entailment rather than simple contextual entailment,
the theory makes false positive predictions, claiming that various intuitively false attitude
reports should have true readings. In predicting the permissive reports from §1 with the
help of distinctive non-trivial contextual entailment, the theory thus gives speakers too
much latitude in what permissive reports they can truly make. This kind of theory appar-
ently needs further resources than contextual assumptions to account of the conditions in
which a report is or may be interpreted as permissive.
5. Acquaintance Problems with Contextual Entailment
Another simple problem for contextual augmentation views is that it is di cult for a
report of the formAVsthatS(F) (whereS(F) is a clause containing linguistic constituent
F)madeinacontextctotakeatruepermissivereadingunlessgiventheconversationalists’
assumptions in c, A is not disposed to think of things as F. I bring up this issue both
to pose another problem for contextual augmentation views and to get a particular data
point on the table. This data point is something which positive theories in chapters to
come must take into account. To see an example of the problem I have in mind, consider:
Blackmail
2
: The context is just like the original Blackmail, save for that the detectives
donot commonlyassumethatGreenisunacquaintedwithScarlet. Rather, theyknowthat
GreeniswellacquaintedwithScarlet: GreenandScarletareclosefriends. AsinBlackmail,
the detectives know that that Scarlet is a socialist, and Black knows that Green thinks
that every socialist was being blackmailed by Mr. Body. The detectives are exchanging
information, trying to find out who had a motive for killing Body. Black assertively utters
the following to Grey:
(38) Green thinks that Scarlet was being blackmailed by Mr. Body.
In this context, (38) is less readily interpreted as a permissive report. This is so even
though the proposition contributed by the complement of (38) is distinctively non-trivially
contextuallyentailedbytheobjectofGreen’sloadedthoughtandthecommonassumptions
inthecontext. Perhapsapermissivereadingof(38)insuchacontextispossible,butitisfar
from the most preferred reading. This, then, at least shows that contextual augmentation
views must be put in terms of the conditions in which a permissive reading of some report
may be accessible, rather than in terms of the conditions in which a report does in fact
take a permissive reading.
The acquaintance problem for contextual augmentation views highlights that whatever
entailment relation one chooses (as long as entailment is a relation between sets of proposi-
73
tionsandpropositions)istoocoarsegrainedofamechanismfortrackingtheconversational
vicissitudes that guide the availability of permissive readings of reports.
I take up the issue of how to account for the acquaintance problem in chapter 6.
6. Asymmetry Problems with Contextual Entailment
Inresponsetotheexamplesfrom §3, acontextualaugmentationtheoristmightsaythat
these examples just tell against contextual entailment views which invoke the wrong kind
of entailment relation. These examples don’t conclusively show that contextual entailment
views cannot work in principle; for all I’ve said, there may be some ingenious contextual
entailment relation which can account for §1’s permissive reports without ruling in §3’s
examplesaspermissiveaswell. InthissectionIpresentaslightlydi↵ erentclassofproblems
for the contextual augmentation view. These problems discredit a wider range of views
which account for the conditions in which a report is permissive just in terms of the
conversationalists’ common assumptions.
First contrast our original example Blackmail with
Switched Blackmail: Mr. Body has been murdered; detectives Black and Grey are
investigating the crime. The detectives have private information that every socialist guest
was being blackmailed by Mr. Body. They know that Green is unaware that Body was
blackmailing any guests at all. Having just questioned Green, Black knows that Green
believes that Scarlet (with whom Green is well acquainted) is a socialist. Although the
detectives commonly know that Scarlet is a guest, they do not know that she is a socialist.
They are exchanging information, trying to find out who had a motive for killing Body;
in particular, in this scenario they are trying to figure out who was being blackmailed by
Body. Black assertively utters the following to Grey:
(38) Green thinks that Scarlet was being blackmailed by Mr. Body.
It is considerably more di cult to interpret what Black asserts by uttering (38) in
Switched Blackmail as true than it is in the original Blackmail. In the present context,
Black is most naturally interpreted here as making a false, loaded report, since it is known
that Green is unaware of any blackmail going on.
Themaindi↵ erencebetweenBlackmail andSwitchedBlackmail isthatwehaveswitched
contextual assumptions with objects of Green’s loaded thoughts. In Blackmail,thedetec-
tives assume that Scarlet is a socialist and Green has a loaded thought that every socialist
was being blackmailed by Body. By contrast, in Switched Blackmail, the detectives assume
that every socialist was being blackmailed by Body and Green has a loaded thought that
Scarlet is a socialist. In each of these cases, Green’s loaded thoughts distinctively non-
trivially contextually entails the proposition ascribed to him, so both should should count
as true permissive reports if either does. Since one but not the other is a true permissive
report, then, even the refined view of the contextual entailment view cannot be correct.
This kind of case seems to show that accounting for permissive readings requires taking
74
further contextual factors into account than contextual assumptions alone.
I take up the issue of how to account for this asymmetry problem in chapter 5.
7. Attitude Variation Problems with Contextual Entailment
Another sort of problem facing contextual augmentation views concerns reports which
ascribe attitudes other than belief. One can make permissive reports in ascribing attitudes
other than belief, and such permissive reports cannot always be accounted for in terms of
contextual entailment. To make this point, consider schematic versions of the first case,
which lays its structure bare:
Context Schema 1:
(i) The conversationalists A and B commonly assume that o is F.
(ii) A knows that subjectSVs that every F is G.
(iii) A assertively utters the following to conversationalist B:
(S1)SVs that o is G.
The original Blackmail case is an instance of Context Schema 1. By considering exam-
ples which di↵ er from Blackmail just in which verb replaces ‘V’ in Context Schema 1, we
find further evidence against simple contextual augmentation views. Consider scenarios
corresponding to the following context structure:
Blackmail Fear:
(i) Black and Grey commonly assume that Scarlet is a socialist.
(ii) Black knows that Green fears that every socialist is being blackmailed by Body.
(iii) Black assertively utters the following to conversationalist Grey:
(38F) Green fears that Scarlet is being blackmailed by Mr. Body.
Suppose we fill out Blackmail Fear by saying that Green made a bet that not all
socialistswerebeingblackmailedbyBody. Forthisreason,heisafraidthateverysocialistis
beingblackmailedbyBody. Asbefore,heisknownnottobeacquaintedwithScarlet. With
these assumptions in the background, it is very di cult for (38F) to take a true permissive
reading in Blackmail Fear. But whatever entailment relation the contextual augmentation
theorist uses to predict the permissive reading of (38) in the original Blackmail will hold
between the detectives’ assumptions and Green’s loaded attitude contents here, since these
sets are the same as those in Blackmail.
Consider also
Blackmail Doubt:
(i) Black and Grey commonly assume that Scarlet is a socialist.
(ii) Black knows that Green doubts that every socialist is being blackmailed by Body.
(iii) Black assertively utters the following to conversationalist Grey:
75
(38D) Green doubts that Scarlet is being blackmailed by Mr. Body.
As with the original Blackmail, we assume that Green is known not to be acquainted
with Scarlet. With this assumption in the background, it is very di cult for (38D) to
take a true permissive reading in Blackmail Doubt. But again, the assumptions and loaded
attitude contents involved in this context are the same as in Blackmail.
Similar problems can be raised using variations of examples like Forgeries and Scotch.
For instance, the context Forgeries is an instance of
Context Schema 2:
(i) The conversationalists A and B commonly assume that all and only Fs are Gs.
(ii) A knows that subjectSVs thatdF are H
(iii) A assertively utters the following to conversationalist B:
(S2)SVs thatdG are H
Here ‘d’ is to be replaced by a determiner like ‘every’, ‘the’, ‘some’, etc. In Forgeries it
is replaced with ‘most’ and ‘V’ is replaced with ‘believe’. Contexts can be specified which
are just like Forgeries except for that ‘V’ is instead replaced with a di↵ erent verb and so
fail to give rise to permissive readings. Consider
Forgeries Suspect:
(i) The conversationalists Black and Grey commonly assume that all and only the
impressionist paintings in the galleries are forgeries in the gallery.
(ii) Black knows that Peacock suspects that most impressionist paintings in the gallery
are original Renoirs.
(iii) Black assertively utters the following to conversationalist Grey:
(41R) Peacock suspects that most forgeries in the gallery are original Renoirs.
The contextual augmentation view predicts that in a scenario like Forgeries Suspect
(41R) should take a true and permissive reading, but intuitively it more readily takes a
false loaded reading.
If the contextual augmentation view is to be preserved, then, it will have to adopt
di↵ erent kinds of entailment relations in explaining permissive reports involving di↵ erent
attitude verbs. I have no argument that a view like this couldn’t work in principle, but
I take it that such a heterogenous account is less desirable than one o↵ ering a unified
explanation.
In chapter 7 I discuss a treatment of the attitude variation problem which o↵ ers a
unified explanation of cases like Forgeries Suspect and Blackmail Doubt together with the
initial examples of permissive readings given in §1.
8. Erroneous Assumption Problems
An overlooked issue up to this point is whether the attitude of assuming is the correct
76
one to use in explaining permissive attitude reports. There is reason to doubt that con-
textual assumption will do, given that conversationalists can make false assumptions in
context, and this has an e↵ ect on whether one can issue true permissive reports. Consider
Gamblers: DetectiveBlackknowsthatPlumknowsthateverygamblerisinthebilliard
room. The detectives erroneously assume that Green is a gambler. Unbeknownst to the
detectives, Green is neither a gambler nor in the billiard room. In discussing who is in the
billiard room, Black assertively utters the following to detective Grey:
(51) Plum knows that Green is in the billiard room.
What Black asserts here intuitively is false. But according to the given distinctive
non-trivial contextual entailment view, this should be true. For Plum’s loaded knowledge
distinctively non-trivially contextually entails that Green is billiard room, given the de-
tectives’ (erroneous) assumption that Green is a gambler. So it looks like the contextual
augmentation view cannot say that a report is permissive if and only if the subject’s rele-
vant loaded attitudes together with what is contextually assumed in the reporting context
entail (even in some suitably restricted sense) the proposition ascribed to the subject by
the relevant report.
It is tempting to think that this problem of erroneous assumptions doesn’t derive from
using the wrong notion of entailment, but instead derives from using the notion of con-
textual assumptions rather than something factive, like contextual knowledge. If the con-
textual augmentation view used the set of contextually known propositions in place of the
set of contextually assumed propositions, Gamblers wouldn’t pose the problem it does,
since the proposition said to be known in (51) does not follow from Plum’s loaded beliefs
together with what the conversationalists know. So it may be worth moving to such a
version of the contextual augmentation theory to avoid this kind of problem.
There is, however, another way to avoid this problem, without moving to contextual
knowledge. Instead, one could try to explain this example by invoking constraints which
coordinate various contextually determined permissive relations. First one can say that
in Gamblers the report made with (51) ascribes a contextually determined permissive
knowledge relation to Plum. So Plum is represented as bearing knows
Gamblers
to the
proposition that Green is in the billiard room. To bear this relation to this proposition, at
least two things must hold. First, Plum must bear believes
Gamblers
to this proposition.
The contextual augmentation view can insist that this requires Plum’s belief set together
with what is contextually assumed, not necessarily known, to entail (in some restricted
sense) that Green is in the billiard room. This it does. Second, the proposition that Green
is in the billiard room, which is said to be known, must be true.Whetherthisissodoes
not depend on the detectives’ contextual assumptions. And indeed it is false that Green is
in the billiard room, even though what Plum known entails as much together with what
the detectives take themselves to know. This suggests that the idea to move to contextual
knowledge in place of contextual assumptions may have been too quick. The contextual
77
augmentation theory seems to deal with this example of erroneous assumptions by using
contextual assumptions rather than contextual knowledge.
Unfortunately, things are more complicated. There are other examples of erroneous
assumptions for which such addition constraints between permissive knowledge and per-
missive belief are of no help. Consider the following examples:
Study: Plum is in the lounge. Detective Black knows that Mustard believes that Plum
isintheroomwhichisconnectedtotheconservatorybysecretpassage. ButdetectiveBlack
and Grey falsely believe that the study, not the lounge, is connected to the conservatory by
secret passage. They also assume that Mustard is unaware, or does not believe, that the
study is connected to the conservatory by the secret passage. Detectives Black sincerely
utters the following to Grey:
(52) Mustard believes that Plum is in the study.
ThepropositionthatPlumisintheroomconnectedtotheconservatoryisinMustard’s
belief set. Detective Black believes a proposition which she might express by uttering ‘the
room connected to the conservatory is the study’. This proposition and the proposition in
Plum’s believe set distinctively non-trivially contextually entail that Plum is in the study,
which Black could express by uttering ‘Plum is in the study’ to Grey. So according to the
distinctive non-trivial contextual entailment view, (52) should be true in this context, but
intuitively it is not.
Scotch
2
: The context is just like of Scotch except for that detective Black has mis-
counted the number of scotches which Mustard has declared to be peaty. Detective Black
has mixed up her records of Mustard’s evaluations of scotch and his evaluations of bour-
bon. So she has recorded that Mustard declared fifteen scotches to be peaty, but in fact he
claimed that fifteen bourbons are peaty and that 25 di↵ erent scotches he tasted are peaty.
Detective Black assertively utters the following to detective Grey:
(40) Mustard thinks that fifteen scotches are peaty.
What Black asserts here is false. But still her report intuitively is permissive. She
does not represent Mustard, who is drunk and is known to have lost count of his peaty-
evaluations, as thinking in terms of “fifteen”. Her erroneous assumption renders her report
false, though not as loaded with respect to ‘fifteen’.
These two examples cannot be explained just by insisting that contextually determined
permissive knowledge relations are coordinated with permissive belief/thinking relations,
sinceneitherattributeknowledgetotherelevantsubject. Theconversationalists’erroneous
assumptions render permissive belief reports false.
In the face of examples Study and Scotch
2
, the contextual augmentation view seems
forced to embrace a notion of contextual entailment defined in terms of contextual knowl-
edge rather than in terms of contextual assumptions. To avoid falsely predicting that the
78
reports in these examples are true, the contextual augmentation view has to say that an
assertion made with a sentence of the formAVsthatS in a context c (where A designates
a in c and that S designates p in c) is permissive and true if and only if V
a
(agent a’s rel-
evant attitude set) bears R to p,where R is some suitably contextual entailment relation
defined in terms of the set CK
c
of commonly known propositions in context c.
Notice, though, that though each of the example reports in this section are false, they
still are intuitively permissive. Explaining the falsity of these reports by invoking contex-
tual knowledge can respect this contrast. For according to a knowledge-based contextual
augmentation view, a bears a permissive relation V
c
to proposition p if and only if p is
entailed (in some sense) by propositions to which a bears V
L
(the fully loaded V relation)
and what the conversationalists in c commonly know. The reports in this sections are pre-
dicted to be false since the propositions said to be believed or known are not entailed by
the relevant subjects’ attitudes together with the conversationalists’ common knowledge.
But the reports all still count as permissive since the conditions in which they are true
are not conditions in which the subject possesses concepts of each of the constituents of
the reported attitude content. In the case of (51) in Gamblers, for example, the propo-
sition said to be known does not count as known
Gamblers
, since it does not follow from
Plum’s loaded knowledge and what the detectives commonly know. But the contextual
augmentation view can still account for the permissive reading of (51) by saying that the
the relation which would be contributed by ‘knows’ in this context, were the detectives’
common assumption that Green is a gambler true, would relate Plum to the proposition
that Green is in the billiard room even if Plum lacked a concept of Green.
Finally, it is worth emphasizing that this contextual knowledge view still faces several
of the problems of irrelevant consequences from above. Reconsider Protest for example;
in that context, the relevant contextual assumptions were stipulated to be known by the
detectives. So by appealing to contextual knowledge rather than contextual assumptions
will be of no help in avoiding that erroneous conclusion. Similar comments apply to Chase
and Knife. So although something like a knowledge constraint is apparently needed to
constrain which assumptions are to be used in determining contextual entailments, that
alone won’t address the variety of problems raised above. More work is needed.
In summary, I have raised five di↵ erent problems with contextual augmentation views:
(i) The problem of irrelevant entailments
(ii) The problem of acquaintance
(iii) The problem of asymmetry
(iv) The problem of attitude variation
(v) The problem of erroneous assumptions
Each of these poses a problem for simple contextual entailment views which cannot
be resolved by appealing to a more discriminating kind of contextual entailment relation.
Going forward, I test proposed accounts of permissiveness against these kinds of cases.
79
chapter 4. Permissive Dispositions
1. Introduction
Given the variety of problems with contextual augmentation views, we should consider
whether there is some other way that the sensitive attitudes view can make sense of the
truth conditions of permissive reports. I consider such an alternative here.
Let me start by pointing to a structural feature shared by the views we’ve considered
so far: each begins with fully loaded attitudes as somehow fundamental. On these views,
all permissive attitudes which a use of an attitude verb can express are to be understood
in terms of corresponding loaded attitudes; we ascribe permissive attitudes as a kind of
shorthand, perhaps to highlight just those features of a subject’s fully loaded attitudes
which are conversationally relevant while ignoring others. The contextual augmentation
view basically treats the practice of ascribing permissive attitudes as derivative on the
allegedly central cases of loaded attitude attributions.
We’ve seen that this style of account faces a variety of problems. In later chapters I
will consider how we might defend contextual augmentation views against these apparent
counter-examples by distinguishing di↵ erent theoretical questions which such views could
be taken to answer. But I think it will be fruitful also to have an alternative way of
understanding the truth-conditional question of what conditions an attitude report is true
in. ThekindofalternativeIwouldlikeexplorerejectsthestructuralfeatureoftheaccounts
we’ve taken for granted so far. Instead of trying to account for the application conditions
of permissive attitude relations in terms of loaded attitudes, this alternative starts from
the other end. First we develop a framework for explaining the application conditions
of permissive attitude relations generally, then we extend this framework to make sense
of loaded attitude relations. On this style of view, not only is a permissive report that a
subjectbelievessomepropositionliterallytrue, ratherthanbeingamatterofmetaphorical
attribution, but what the attitude it expresses is as metaphysically fundamental as fully
loaded attitudes are. On this view, it isn’t that one counts as bearing a permissive attitude
relation towards some proposition just because that subject bears loaded attitude relations
towards certain related propositions. This style of view, then, is in an important sense
egalitarian with respect to loaded and permissive attitude relations: we are to account for
both kinds of attitude relation in broadly the same way.
The sort of alternative approach I develop is one according to which bearing a partic-
ular propositional attitude is just a matter of possessing certain dispositions.Thiskind
of approach to attitudes has a long history within behaviorism in the philosophy of mind,
but recently a non-behaviorist version of this approach has been defended by Schwitzgebel
(2002, 2012). After raising some further relevant data points in §2, I sketch Schwitzgebel’s
non-behavioralist dispositional account of attitudes in §3 and work it into a sensitive atti-
tudesview,onwhichdi↵ erentusesofasingleattitudeverbcancontributedi↵ erentattitude
relations, understood as dispositional profiles.In §4 I use this view to defend what I call
the “permissive dispositions” view. In §5 I show how the permissive dispositions view can
80
make sense of several troubling cases of permissive reports seen so far. In §6 I show how
we can make sense of some the intuitions behind the “loaded-first” approach which is ex-
hibited by the contextual augmentation view. In §6 I address some potential problems for
the sensitive dispositional attitudes view.
2. More Di culties for Loaded-First Accounts
Let’s start with some examples of attitude reports which could, from our working char-
acterization of permissiveness, be understood as permissive reports. If these are admitted
as genuine permissive reports, then they constitute a sixth species of counter-example to
contextual augmentation views in addition to the five raised in chapter 3.
Reconsider the example of the chess-playing program Leela, from chapter 2:
Chess: ProfessorPlumisplayingchessagainstthecomputerprogramLeela. Detectives
Black and Grey are looking on. After Plum moves his bishop to d4, Leela moves a bishop
to d7. Black turns to Grey and sincerely utters:
(3) Leela is boxing out Plum’s knight because it thinks that Plum wants to fork the
queen and rook.
In chapter 1 I observed that although what Black asserts here sounds true and appro-
priate, intuitively the program Leela lacks a concept of Plum, perhaps lacks concepts of
queens and rooks, and may lack any concepts whatsoever. In chapter 3 I characterized
permissiveness as follows: a use of a clause R of the formAVsthatS in a context c takes
a permissive reading if it is possible that the proposition asserted by (or designated by) R
in c is true but a subject designated by A in c lacks a concept conversationally associated
with a linguistic constituent of S. Given this characterization, together with the truth of
(3), it looks as though Black successfully makes a permissive report in Chess: not only is it
possible that what Black asserts with (3) is true yet Leela lacks a concept associated with
linguistic constituents of the complement of (3), but actually this is so.
It is di cult to see how a contextual augmentation view could be used to account for
this permissive report without running into problems. Assume the best version of the
contextual augmentation view which we’ve encountered so far, (DCE), according to which
a subject a bears a contextually determined attitude relation V
c
towards a proposition p
if and only if a’s V set distinctively, non-trivially contextually entails p. So: if and only if
there is some subset A of a’s V set and some subset B of the set of contextual assumption
in c, CA
c
, such that A\ B entails p but B does not entail p. This version of the view
clearly won’t do; for in cases like Chess,thesubjecta’sV set is empty (in this case, Leela’s
think set), hence A is empty, hence A\ B = B, so the view boils down to: a bears V
c
towards p if and only if B entails p but B doesn’t entail p. Not too good. Stepping
back, intuitively there must be something about the subject which grounds permissive
attributions in cases like Chess, but that something cannot be the subject’s fully loaded
attitudes. Since contextual augmentation views take fully loaded attitudes as crucial in
81
specifying the application conditions of permissive attitude relations, such views cannot be
correct if examples like (3) are genuine permissive reports.
Next consider a series of examples concerning other non-human subjects.
Lecture: Plum is giving a wide-ranging and rambling scientific lecture in the study.
Among the various claims he makes are:
(53) A resonator neuron prefers inputs having frequencies that resonate with the fre-
quency of its subthreshold oscillations.
65
(54) Once a clam believes there is danger about, it will not relax its grip on its closed
shell until it is convinced that the danger has passed.
66
We can assume that Plum means to be speaking truly in uttering (53) and (54). But
presumably neurons and clams lack concepts associated with constituents of the comple-
ments of these reports. One might try to dismiss these uses of ‘prefers’ and ‘believes’ as
non-literalandsoforsomereasonnotoftheoreticalinterestforatheoryofpropositionalat-
titudes, but if we take the reports at face value they seem to constitute permissive reports.
Once we’ve adopted the sensitive attitudes view, it’s unclear why we shouldn’t accept this
data at face value and design our theory of permissiveness to account for it.
Another example:
Party: Scarlet, Green, and Mustard are members of the party-planning committee;
they have been asked by the detectives to pick a party theme for the holiday party which
they think will attract the most attendees. Scarlet thinks a pirate theme will be best,
Green thinks a dinosaur theme will be best, and Mustard thinks that a sci-fi theme will be
best. Given this disagreement, they resort to a Condorcet voting scheme to choose; since
each ranks a fairytale theme as second best, they choose a fairytale theme, though none
of them individually thinks this theme is best for attracting attendees. They submit a
proposal for party funding for a fairytale-themed party. In discussing the party proposal,
Black assertively utters the following to Grey:
(55) The party-planning committee thinks that a fairytale theme will be best at attract-
ing party attendees.
WhatBlackassertswith(55)intuitivelyistrue. Thoughitishardtosaythattheparty-
planning committee, the group, itself possess a concept associated with ‘fairytale theme’.
Certainly the individuals do possess such concepts, it would be strained to say that the
committee itself does. Now, there are proposals which analyze attributions of attitudes to
groups in terms of attitudes held by the group’s members, and I do not mean to discredit
65
From Izhikevich (2007), cited in Figdor (2017).
66
From Dennett (1989).
82
such views with this quick example.
67
All I need for my purposes is the observation that
if we take this report at face value, it apparently constitutes a permissive report with a
conceptless group as subject. And again, with the sensitive attitudes view to hand, there
seems to be no special reason to seek some special local analysis of this attitude report;
whatever sensitive attitudes treatment we can make to work for other permissive reports
plausibly should extend to such group attitude reports.
In each of these examples, we have an intuitively true attitude report about a subject
whom we can reasonably take to lack concepts. If these are admitted as genuine permissive
reports, then no cleverly specified contextual entailment relation, even one which could
handle the five problems raised in the previous chapter, can be correct. For there are no
loaded attitudes from which we can derive the attributed thought contents as contextual
entailments. As long as chess-playing computer algorithms, neurons, clams, and groups
lack concepts, they don’t have loaded attitudes. So we have further reason for looking
for an alternative way of developing the sensitive attitudes view, a way which doesn’t just
treat a subject’s permissive attitudes as derivative on the subject’s loaded attitudes. In
the next section I begin to develop just such an alternative.
Before I dive into this alternative, I will digress briefly here to consider the prospects
for defending the contextual entailment view against this style of example. As mentioned
above, I argue in later chapters that attending to the pragmatics of permissive reports
a↵ ords one a way of defending the contextual entailment view against several of the prob-
lemsraisedinchapter3. Ifthatdefensivestrategysucceeds,thenthecontextualentailment
theorist would just need some way of making sense of the kind of example presented in
this section.
The best way I can see of doing this is to insist that in fact the reports made with (7),
(53),(54),and(55)areloaded,orareindeedsimplecontextualentailmentsofrelatedloaded
attitudes which the relevant subject possesses. To do this, we also insist that entities such
as computer programs, neurons, clams, and groups, can possess concepts of some kind.
This defense requires us to say more than I have about what concepts are. Up to this
point, I have simply spotted myself that there are such things as concepts and that we can
characterize the di↵ erence between permissive and loaded readings of a report sentence
in terms of which concepts the subject has to possess and deploy to verify these di↵ erent
kinds of report.
Distinguish two prominent approaches to concepts. On one approach, concepts are
mental representations: to possess a concept of some thing is for there to be a particular
psychological entity or symbol which can figure in one’s internal representations of the
world which represent that thing.
68
On the other approach, concepts are a certain kind of
67
Clearly an analysis which says that a group believes p if and only if all or most of its members believe
p would not survive the Party example. But there are other accounts which make sense of group attitudes
in terms of more complicated attitudes of its members; see Tuomela (1993) and Gilbert (1996)
68
This view has its home in the representationalist theory of mind. See Fodor (1975) and Margolis &
Laurence (2007).
83
ability: forsome, topossessaconceptconsistsinanabilitytouseawordorsymbol, butwe
also can adopt a more general approach on which to possess a concept of some thing is to
be able to discriminate between instances or presentations of that thing and non-instances
of that thing.
69
Dummett (1996, 97) gets closer to this conception of concepts: “What is it
to grasp the concept square, say? At the very least, it is to be able to discriminate between
things that are square and those that are not. Such an ability can be ascribed only to one
who will, on occasion, treat square things di↵ erently from things that are not square; one
way, among many other possible ways, of doing this is to apply the word ‘square’ to square
things and not to others.”
The intuitive notion of concept assumed above, according to which clams don’t have
concepts of danger, seems to be that of a mental representation. Clams don’t have a
‘danger’ word in any language of thought which the clam manipulates in syntactically
complicated inferential operations. Clams don’t have “mental images” of danger, or any
Lockean “ideas” of danger; they don’t intuitively think “in terms of” ‘danger’.
Clams could plausibly, however, have an ability to discriminate dangerous situations
from non-dangerous situations. To have such an ability, a clam would merely need to be
able to enter in certain internal states which reliably enough co-vary with the nearby pres-
ence of dangerous situations. A clam, on this view, would count as possessing a concept of
danger just if they enter into certain internal states relevant to the functional organization
of their behavior which are appropriately sensitive to dangerous situations.
70
Presumably
such states are part of the explanation for the presumed truth that clams clam up when
they “think” that danger is present, even if we choose to claim that this kind of attribution
of attitudes is merely metaphorical. Taking the idea seriously that these states underly an
ability to discriminate dangerous situations from non-dangerous situations, which consti-
tutes a concept of danger, the contextual entailment theorist can insist here that a report
such as (54) is loaded with respect to ‘danger’ and the clam’s concept of danger. Or, at
69
Kenny (2010) defends this kind of view, though his way of characterizing the abilities in which concepts
consistinvolvedlinguisticabilities. BennettandHacker(2008)toocharacterizeconceptsasabilities,though
again as linguistic symbol-manipulation abilities.
70
Theseinternalstatesmaynotneedtoreliablycovarywithdangeroussituations,inthesenseofcovarying
with dangerous situations more often than not. This follows from a general observation about abilities,
made by Millikan (2000, 53), that one can possesses an ability to do something without possessing the
disposition tosucceed atdoingthatthingifonetries: “Manypeoplewiththeabilitytoswimhavedrowned,
presumably when they were trying to swim”. Similarly, one may have a concept, understood as an ability
to discriminate between instances of some kind and non-instances of that kind even though one may fail,
and even fail often, to successfully discriminate between such instances and non-instances in exercising this
ability. It may be a matter of the evolutionary design history of clams that their entering certain states
is responsive to dangerous situations. For a simplified example, assume that there were some clams which
would enter into certain “danger states” such that it would be 20% probable that there was danger nearby
conditionalontheirbeinginadangerstateand80%probablethattheyareintheirdangerstateconditional
on there being danger nearby. As long as this success rate allows those clams to survive and proliferate
more than present competitors, then they may thrive due to their faulty ability to track and response to
dangerous situations.
84
least, that the clam possesses some fully loaded attitudes which can serve as a base from
which contextual entailments may be drawn.
For some motivation in favor of this idea, consider a variant of Chess:
Chess
2
: Professor Plum is playing chess against the computer program Leela. Detec-
tivesBlackandGreyarelookingon. GreyhasplacedabetthatPlum’sdark-squarebishop
willbetakeninthisgamebeforehislight-squarebishop. AfterPlummoveshisdark-square
bishop to d4, Leela moves a bishop to d7. Black turns to Grey and sincerely utters:
(56) Leela is boxing out Plum’s knight because it thinks that Plum wants to use the
bishop you bet on to fork the queen and rook.
In contrast to the original report in Chess, this report has somewhat more of a permis-
sive flavor: I can work up the intuition that Black’s report with (56) is permissive with
respect to ‘the bishop you bet on’ in a way that the original report about Leela is not
permissive with respect to the expressions ‘fork’, ‘queen’, ‘rook’, and even ‘Plum’. One
way to make sense of this feeling is that Leela lacks and so fails to deploy a concept of
the property bishop-Grey-bet-on in a way that Leela does not lack a concept of rooks,
queens, forking, or Plum.
If this intuition bears any weight, the contextual entailment theorist, with their going
defense against the examples above, is in a position to account for the di↵ erence. Leela
does have a loaded thought that Plum wants to use his dark-square bishop to fork the
queen and rook, and this proposition in Leela’s thinks set contextually entails in Chess
2
that Plum wants to use the bishop Grey bet on to fork the queen and rook. So, the
contextually determined permissive relation contributed by ‘thinks’, T
Chess
2
, relates Leela
to that proposition.
Why say that Leela possesses a loaded thought relation to the proposition that Plum
wantstousehisdark-squaredbishoptoforkthequeenandrook? BecauseLeelahasabilities
to discriminate, in her way, between rooks and non-rooks, queens and non-queens, forking
situations and non-forking situations, and opponents and non-opponents. (Presumably
LeeladoesnotpossesscertainabilitiesofdiscriminatingbetweenPlum andthingswhichare
not Plum, which the conversationalists may associate with ‘Plum’ in the context, in which
case the report would also be permissive with respect to Plum and those concepts.) Leela
possesses these abilities in that it can enter into certain states relevant to the functional
organization of its behavior which reliably co-vary with di↵ erent facts about rooks, queens,
forkingsituations,andopponents. Giventheseabilities,Leeladoesindeedpossessconcepts,
understood as abilities to discriminate, between these things, so is in a position to have
a fully loaded thought that its opponent wants to use its dark-squared bishop to fork
its queen and rook. With this way of understanding concepts, the contextual entailment
theory might account for the examples raised above.
Now, even if the contextual entailment view can be defended against the objections
fromchapter3andtheconcepts-as-abilitiesdefenseagainstthepotentialcounter-examples
85
presented in this section does well enough, this defense does hitch its wagon to the ability-
theoretic understanding of concepts. So at least for those who disfavor such a theory of
concepts, it is worth considering what alternatives there are for making sense of permissive
reports. In the rest of the chapter I focus on what I take to be a promising alternative.
3. Dispositional Attitudes
To make sense of the cases from §2, apparently we need a di↵ erent way of accounting
for the application conditions of permissive attitude relations. The place to look, I suggest,
is accounts of attitudes in terms of dispositions. On such views, possessing concept-laden
representations is only incidental to possessing attitudes like belief, preferring, hoping,
etc.; rather, what is fundamentally required to possess such attitudes is to be disposed
to behave in certain ways, disposed to feel certain sensations, or to disposed enter into
certain mental states. The reason this approach to attitudes is potentially useful for the
sensitive attitudes view is that, depending on how these dispositions are specified, these
dispositionalstatesmaybethekindsofstatesthatchess-playingprograms,simpleanimals,
or groups, in addition to normal human thinkers, could have.
Schwitzgebel (2002, 2012) o↵ ers a dispositionalist account of attitudes which I take to
be particularly friendly to the sensitive attitudes view. In the rest of this section, I present
Schwitzgebel’s view and work it into a sensitive attitudes account of the kind I favor. As
will become clear, although I am using Schwitzgebel’s general view as a guide, I depart
from his particular way of developing this kind of view at several points. With my revised
version of a dispositional view in place, I spend the rest of the chapter using the view to
explain the problematic permissive reports raised above in addition to showing how the
view can account for loaded attitude attributions.
According to Schwitzgebel’s dispositional account of attitudes, what it is to have a
particular attitude, such as belief, is just to possess certain dispositions. As he puts his
view, “to believe that P...is nothing more than to match to an appropriate degree and
in appropriate respects the dispositional stereotype for believing that P” (Schwitzgebel
2002, 253, emphasis mine). For Schwitzgebel, a “dispositional stereotype” is a cluster of
dispositional properties which “we are apt to associate with a thing”. Schwitzgebel gives
the example of being an a↵ able person. We are apt to associate with the condition of being
an a↵ able person certain dispositions, such as “the disposition to greet people warmly, the
disposition to put others at ease in conversation, the dispositions not easily to become
o↵ ended, ...” and so on.
71
It is plausible that all there is to being an a↵ able person is
exhibiting all, most, or certain salient instances of this class of dispositions which we are
apt to associate with being an a↵ able person.
Schwitzgebel argues that likewise all there is to believing a proposition is to match
a dispositional stereotype associated with believing that proposition. The dispositional
stereotype associated with believing that the Dodgers will win the World Series, for ex-
71
Schwitzgebel (2001).
86
ample, involves the disposition to assertively utter “the Dodgers will win the series”, the
disposition to agree when others say similar things, the disposition to take bets on the
Dodgers’s winning, the disposition to use the proposition that the Dodgers will win in
reasoning, the disposition to be surprised upon learning that the Dodgers were knocked
out in the NLCS, and so on. If one’s own dispositions match this profile “to a su cient
degree”, then one is appropriately described as believing that the Dodgers will win the
World Series. Moreover, on Schwitzgebel’s view, there is nothing further to believing this
proposition: what it is to believe that the Dodgers will win the World Series just is to
match this dispositional profile in the right way.
Notice that the dispositional stereotype associated with believing that the Dodgers will
wintheWorldSeriesincludesdispositionswhichabehavioralistversionofadispositionalist
theory could not make use of. On that style of dispositionalist theory, the dispositions in
whichbearingapropositionalattitudeistoconsistarejustdispositionstobehave incertain
ways.
72
So a constraint on such theories is that the operative dispositions can be specified
in purely non-mental terms. But the dispositional stereotype listed above includes such
dispositions as to use a certain proposition in reasoning or to enter into other mental states
such as being surprised. Along these lines, other dispositions in the stereotype may include
the disposition to believe that the team playing against the Dodgers won’t win the World
Series. It is for this reason that Schwitzgebel calls his view a “phenomenal dispositional
view”. This latitude concerning which dispositions can constitute the state of bearing a
propositional attitude is something which I wish to preserve; indeed, this feature will play
an important role in making sense of loaded attitude reports.
In a moment I will begin to work this style of theory into a version of the sensitive atti-
tudes view. But before I do, it’s worth attending to a phenomenon raised by Schwitzgebel
in support of his view, which he calls “in-between believing”. It’s worth focusing on this
issue both to appreciate independent reasons in favor of the dispositional view and also
to draw attention to a potential source of trouble for the view: namely, how much must
one match a dispositional stereotype to count as believing a proposition? What general
specification can we give of the application conditions of a given use of ‘believe’? I will
first present and discuss Schwitzgebel’s approach to in-between believing, then will address
potential problems with his particular approach. I reject Schwitzgebel’s particular way of
making sense of in-between believing and instead o↵ er a di↵ erent way of understanding
what is going on with this phenomenon. This will lead us to a sensitive attitudes version
of the dispositional approach.
3.1. In-between Attitudes
In cases of in-between believing, a subject exhibits some but not all of the dispositions
in a dispositional stereotype for a certain attitude. Given this partial match with the
stereotype, we feel a natural hesitance either to assert or to deny that the subject possesses
72
The classic example is Ryle (1949).
87
the given attitude. Schwitzgebel takes such cases to lend support to the dispositional
approachoverarepresentationalapproach,sinceonrepresentationaltheories,suchasFodor
(1975)’s “language of thought” view of attitudes, Schwitzgebel claims that it is di cult to
make sense of what is going on in cases of in-between belief. Consider the following case:
Forgetting: Long ago, Mustard was a bridge player. As an avid bridge player, he was
familiar with many of the standard bidding conventions, including the famous Blackwood
convention. Back in his heyday, Mustard knew that in the Blackwood convention one
player bids 4 no trump to solicit a response from their partner that will indicate how many
aces they hold: depending on the partner’s response, 5 clubs, 5 diamonds, 5 hearts, or 5
spades, the partner indicates that they hold 0 or 4, 1, 2, or 3 aces. It’s been quite some
time since he’s played any bridge, so Mustard is now hazy on the details: he knows the
Blackwood convention starts with one player’s bidding 4 no trump, but he forgets which
response indicates what. If presented with certain multiple-choice quizzes on the topic,
he would be able to give the right response, though he could not produce the details of
the convention o↵ the top of his head. Finally, assuming that he will continue not to play
bridge, we can assume that eventually Mustard will forget how the Blackwood convention
works at all, or even that such a convention exists.
73
Given this case, we are invited to consider these facts: at one time, Mustard knew that
bidding 5 diamonds indicates that one has 2 aces according to the Blackwood convention,
but at some time in the future he will not know this. The transition from his early state
of knowing to his later state of not knowing, however, is gradual. As time passes, Mustard
forgets di↵ erent details of the convention, including what bids represent what, or the order
inwhichthebidsaretocome,whattheconventionisdesignedtoreveal,howbiddingworks
in bridge, what bidding conventions are, and eventually forgetting the name ‘Blackwood
convention’ altogether. Representationalist approaches to attitudes are faced with the
question: at what point does Mustard cease to know what how the Blackwood convention
works? At what point does Mustard cease to know that bidding the five of diamonds
indicates that one has two aces according to the Blackwood convention?
Mustard’s current state is particularly pressing: he presently recalls the gist of what
the convention is designed to do, but could only say which bids indicate what if primed in
certain ways. Does Mustard now know that bidding five diamonds indicates that one has
two aces according to the Blackwood convention? Schwitzgebel’s claim is that a simple yes
or no to such a question would be forced and awkward. Rather than insist on a simple
answer here, Schwitzgebel says we should adopt a view which can explain our hesitance
either to assert or deny that Mustard knows this proposition.
The dispositional account seems to do just this: there is a cluster of dispositions which
we are apt to associate with knowing that bidding five diamonds indicates that one has
two aces according to the Blackwood convention. This cluster includes the disposition to
73
This is based on an example from Schwitzgebel (2001).
88
be able to make such a bid, the disposition to describe that bidding so indicates that one
has two aces if asked to, the disposition to correctly answer what bidding five diamonds
indicates on a multiple choice quiz, and so on. Mustard exhibits some but not all of the
salient instances of this dispositional stereotype, hence our hesitancy. He matches the
stereotype pretty well, but not perfectly, so his situation constitutes a borderline case
of knowing that bidding five diamonds indicates that one has two aces according to the
Blackwood convention. Schwitzgebel takes this as an account of such in-between cases of
knowing.
Non-dispositional accounts of attitudes, which claim that to possess an attitude is to
be in a particular representational state, such as having a sentence in the language of
thought in one’s “belief box”, seem forced to say definitively that Mustard believes that
bidding five diamonds indicates that one holds two aces on the Blackwood convention or
that he does not. Insofar as a clear cut verdict of this kind fails to account for our feeling
of hesitancy regarding how to describe Mustard’s attitudes towards this proposition in
Forgetting, Schwitzgebel takes this to be a shortcoming of representational accounts. In a
moment I will critically consider whether this is so, but before I do let’s consider another
example of in-between believing:
Full House: Scarletisnewtopoker. She’sbeentaughttherelativevaluesofthedi↵ erent
hands, including that a full house is better than a flush. When asked, she faithfully asserts
that a full house beats a flush. But as she plays poker she seems to value flushes more
than full houses. She bets more aggressively when she has a flush than when she has a full
house. On one particular hand, Scarlet had three nines and two aces, four of which were
spades, and she drew to get a fifth spade: she gave up a full house to try to get a flush.
On another occasion after Green won a hand with a full house, she was asked what hand
Green had played, and simply described it as “just a three of a kind and a pair”, rather
than as a “full house”.
74
Now we are asked to assess whether Scarlet believes that a full house is better than a
flush. Given the scenario, it is natural to feel some hesitancy to give a simple yes or no
answer. She clearly exhibits certain important dispositions associated with this attitude:
she would straightforwardly assert that a full house is better than a flush if asked, and
she would agree when others say as much. But much of her poker-playing behavior fails
to match the dispositional stereotype associated with this attitude: she intuitively doesn’t
bet or play as though she believes that a full house beats a flush.
As Schwitzgebel (2002, 257) says in discussing a similar case, whether we feel comfort-
able saying that Scarlet does believe this proposition seems to depend on additional facts
about the context of attribution. If we were discussing whether Scarlet would pass a quiz
about poker hands, we would feel comfortable claiming she does believe that a full house
beats a flush. But if we were discussing whether she will play poker well on a particular
74
This example is also based on one from Schwitzgebel (2002).
89
hand, we may instead claim that she doesn’t believe that a full house beats a flush. Indeed,
we may say that she believes that a full house doesn’t beat a flush. Absent such contextual
priming, however, Schwitzgebel claims that we are hard-pressed to ascribe to Scarlet the
attitude of believing or disbelieving the proposition that a full house beats a flush.
This, according to Schwitzgebel, is another case of in-between believing. We can char-
acterize what is going on using the dispositional account of attitudes, saying that Scarlet
imperfectly matches the stereotype associated with believing that a full house is better
than a flush. On a representational approach to attitudes, however, we seem forced to
say definitively one way or another that Scarlet believes the proposition or disbelieves it.
Schwitzgebel again takes this to be a count against representational accounts.
IwanttoannouncerightawaythatIamnotconvincedthatcasesofin-betweenattitudes
aregoodevidenceagainstrepresentationalaccounts. Onecouldmakesenseofthesecaseson
arepresentational viewbydescribingthemascasesin which itissimply vague whether the
subject possesses the relevant attitude. Just as it is no contradiction to say that Mustard
isbaldbutitisvaguewhetherMustardisbald, itisnocontradictionthatMustardbelieves
p but it is vague whether Mustard believesp. A representational theorist can say that, yes,
definitively Mustard believes or doesn’t believe that bidding the five of diamonds indicates
that one has two aces, but it is vague whether he believes that bidding the five of diamonds
indicates that one has two aces. I discuss this further below in contrasting my favored way
of developing the dispositionalist view with Schwitzgebel’s.
The dispositional approach seems compatible with a sensitive attitudes view, given the
context-sensitivity of whether it is “appropriate” to describe either Mustard or Scarlet as
having the relevant attitudes in these cases. It is a subtle question, however, whether to
characterize Schwitzgebel’s own view as a sensitive attitudes view.
Some things which Schwitzgebel says seem to conflict with the sensitive attitudes ac-
count. For example, in discussing a case like Full House, he says that he is not saying
that the “mental state” of the subject varies with the context. He says that rather, “given
that the [subject] deviates from the stereotype in some respects but not in others, how
best to describe her mental state will depend on the practical demands of the moment”
(Schwitzgebel 2002, 257). He also says that the greater the proportion of stereotypical
dispositions one possesses and the more “central” those dispositions are to the relevant
stereotype, the more appropriate it is to describe one as possessing the relevant attitude.
75
Schwitzgebel frequently emphasizes how the assumptions and conversational goals in
a particular context will influence whether it is “appropriate” to “describe” someone as
having a particular attitude, but he resists semantically descending from these charac-
terizations about our attribution practices to saying outright that whether one possess a
particularkindofattitudeiscontext-dependent. Ononewayofinterpretinghisdiscussion,
Schwitzgebel seems content to bracket the question of whether Mustard or Scarlet “really”
believe certain propositions, instead o↵ ering an account merely of our reporting practices.
75
This language patterns after Schwitzgebel (2001, 81), emphasis mine.
90
But he also says, in discussing how the dispositional account allows for indeterminacy, that
“the metaphysically inclined may wish to ask whether, useful or not, the present account
accurately describes what belief really is. I must admit that I fail to feel the impulse that
drives questions such as this”.
76
This could be taken as claiming that there is no meta-
physical fact of the matter about whether Scarlet believes that a full house beats a flush;
there are only facts about how we are disposed to ascribe this attitude.
77
On a truth-conditional approach to attitude reports, this way of speaking invites the
question of how these considerations of “appropriateness” relate to whether a particular
attitude report is true, hence whether a subject in fact possess a particular attitude. How
are we to regiment the truth conditions of any given report if we take on board the basic
framework of dispositional stereotypes?
It admittedly does sound odd to say that the mental state of a subject “varies with the
context”. But I think that it is a false step to avoid this oddness by embracing metaphys-
ical quietism about whether Mustard or Scarlet bear particular attitudes in the foregoing
scenarios. Once we are clear on the relation between attitude reports and attitudes them-
selves, given the sensitive attitudes view’s context-sensitive approach, we can avoid the
oddness here easily enough. To show this, I’ll raise two reasons to disfavor Schwitzgebel’s
wayofdeployingthedispositionalaccounttomakesenseofin-betweenattitudes, according
to which there is no metaphysical fact of the matter about whether Scarlet believes that a
full house beats a flush.
First, insofar as we are committed to the law of excluded middle, we are committed to
theclaimthatScarleteitherbelievesthatafullhousebeatsaflushordoesn’tbelievethata
full house beats a flush. That belief attributions are context-sensitive does not undermine
the force of this consideration. Uses of ‘every bottle is empty’, ‘You should wash your
hands before you eat’, or ‘I am here’ can assert di↵ erent propositions depending on the
contextinwhichtheyareused. Butoncewefix acontextofuse,thelawofexcludedmiddle
still holds: either every bottle is empty or it is not the case that every bottle is empty.
As long as the context influences the interpretation of both occurrences of ‘every bottle
is empty’ in a uniform way, ‘either every bottle is empty or it is not the case that every
bottle is empty’ must come out true. Even though it is hard to say in Full House whether
Scarlet believes or doesn’t, it must be true that Scarlet believes or doesn’t believe that a
full house beats a flush. We should not so easily retreat to an account which invalidates
the law of excluded middle; rather, we should start by looking for an account which can
respect excluded middle while making sense of our hesitancy to endorse either disjunct of
this claim.
A related, second reason to avoid Schwitzgebel’s way of using the dispositional account
ties into a methodological point from the previous chapter. In that chapter, I expressed
76
Schwitzgebel (2002, 270).
77
Of course, given my sensitive attitudes view, I also say there is no interesting sense in which one belief
relation (or dispositional profile) is the “real” kind of belief. Perhaps Schwitzgebel’s hesitance here is just
an indication that his view is best interpreted as a sensitive attitudes view after all.
91
suspicion about attempts to divorce the issue of whether one has some attitude from the
issue of whether one is rightly described as having some attitude. Recall van Fraassen’s
claim that we should “drive a wedge between what x’s beliefs are and what the true
attributions of beliefs to x are”.
78
To my ear, this is as odd as saying that there is a
di↵ erence between whether x is a door and whether it is true to attribute being a door
to x. Central to the approach I favor is the following schematic principle: it is true that
a believes that S i↵ pa believes that Sq is true. Once we accept this kind of simple
disquotational principle, we cannot bracket o↵ questions about what attitudes one has
from what attitude reports about one are true.
79
If in some context it is correct to ascribe
an attitude to some agent towards some proposition, then the agent bears that attitude to
that proposition.
3.2. Sensitive Dispositional Attitudes
The sensitive attitudes view a↵ ords us a way of accepting the context-dependence of
attitude attributions without having to accept any metaphysically objectionable version of
the claim that one’s “mental state” “varies with the context”.
On the sensitive attitude view, there is a plentitude of attitude relations associated
with any given attitude verb: which of various attitude relations is expressed by a use of
a particular attitude verb varies across contexts.
80
With this view, we can boldly say that
for any belief relation whether one bears that belief relation to a particular proposition
does not vary across contexts: what varies across context is which belief relation from a
plentitude of belief relations a use of ‘believes’ serves to express. Similarly, if we take uses
of attitude verbs to designate clusters of dispositions, we can say that there is a plenitude
of clusters of dispositions associated, or associable, with any given attitude verb; which
cluster is expressed by a use of a particular attitude verb varies across contexts. Whether
one matches the cluster associated with an attitude verb on some use does not vary “across
contexts’; rather, what varies is what cluster one is represented as matching.
So on the sensitive attitudes view, we can embrace both the law of excluded middle
and a version of the disquotational principle above, as long as these principles are suitably
relativized to contexts. For any context c,subject a, and proposition p, pa believes p or a
doesnotbelievepqistrue,aslongasbothoccurrencesof‘believe’areinterpreteduniformly
in c. And for any context c, a believes
c
that S i↵ pa believes that Sq is true in c,where
‘believes
c
’ expresses the relation which ‘believes’ expresses in c. We do, of course, have to
give up the context-invariant versions of these principles: we cannot say that a believes
c
0
that S i↵ pa believes that Sq is true in c. But this is no surprise on the sensitive attitudes
78
van Fraassen (1979, 373).
79
Disquotational principles are admittedly trickier to defend when context-sensitive expressions are on
the loose. I address this complication further below.
80
There is also the view floated in chapter 3 that uses of a particular attitude verb always serve to con-
tribute asingle relation, but one which takes additional arguments in addition to a subject and proposition,
such as an extra argument for contextual assumptions relevant to the truth-conditions of the report.
92
view, for believes
c
0 maybealoadedbeliefrelationwhereas believes
c
, therelationwhich
one expresses with ‘believes’ in c, may be permissive. The sensitive attitudes view lets us
admit that attitude verbs are context-sensitive in what they express while respecting these
broader constraints on truth.
The sensitive attitudes view also gives us a way of making sense of cases of in-between
attitudes. Themainthingtobeaccountedforinsuchcasesisourhesitancytoassertordeny
that a subject possesses a particular attitude. It is apparently implicit in Schwitzgebel’s
discussion of such cases that for any attitude verb V there is but one attitude relation V
which a use of V serves to express (or one cluster of dispositions which V “is associated
with”). On this view, our hesitancy to assert or deny that a subject possess a particular
attitude V towards some proposition may be evidence that there is no metaphysical fact
of the matter whether the subject bears V to the relevant proposition. But the sensitive
attitudesviewgivesusanalternativeexplanationofourhesitancy: wearesimplyuncertain
of which proposition is being asserted. Are we being asked to pass judgment on whether
Scarletbearsbelieves
1
tothepropositionthateveryfullhousebeatsaflushoronwhether
Scarlet bears believes
2
to the proposition that every full house beats a flush? Are we
being asked to pass judgment on whether Scarlet possess the dispositions in cluster B
1
or cluster B
2
? Given the underspecified scenario, we are unsure of which belief relation
is under discussion and so are unsure of which attitude report the speaker is committing
themselves to. This then is an alternative explanation of our hesitancy to confidently utter
‘Scarlet believes that a full house beats a flush’.
Combining the sensitive attitudes view with the dispositional account of attitudes
gives an explication of what it takes to bear di↵ erent belief relations, such as believes
1
,
believes
2
, etc., towards a proposition. It also provides a picture of wherein these belief
relations di↵ er. In short, these are relations of possessing di↵ erent dispositional profiles,
where a dispositional profile is a set of dispositions. Whereas Schwitzgebel talks in terms
of the dispositional stereotype associated with believing a proposition, on the sensitive at-
titudes view we allow that di↵ erent dispositional profiles may be associated with di↵ erent
uses of ‘believes’ in di↵ erent conversational scenarios. I don’t think this is far from what
Schwitzgebel has in mind, since he allows that in di↵ erent reporting contexts the kind of
required “match” with the relevant dispositional stereotype varies. One way to understand
my sensitive attitudes version of the dispositional approach is to take it as explicating this
contextual variation of matching a dispositional stereotype by associating particular uses
of attitude verbs with relations of possessing dispositional profiles which include di↵ erent
subsets of the dispositions included in the standing stereotype. I say more about this in
the next section, where I spell out the sensitive attitudes version of the dispositional view
in more detail.
Fornow,let’sconsideranexampleofhowthisalternativeexplanationroughlyproceeds.
IncertaincontextsinwhichwearediscussingwhetherScarletwouldpassaquizaboutpoker
hands, we say that the occurrence of ‘believes’ in ‘Scarlet believes that a full house beats
a flush’ is associated with a dispositional profile including the disposition of assertively
93
uttering ‘a full house beats a flush’ when asked, but not including any disposition to
bet in a particular way in the course of a poker game. In another kind of context, in
which the conversationalists are discussing how Scarlet will play a particular hand, we
say that the occurrence of ‘believes’ in ‘Scarlet believes that a full house beats a flush’
is associated with a dispositional profile including the disposition to, ceteris paribus, fold
when holding a flush if you think another player is holding a full house, but not including
the disposition to assertively utter ‘a full house beats a flush’ if asked which hand is
better. For conversational scenarios with underspecified conversational assumptions and
goals, however, we are simply uncertain of which attitude is expressed by ‘believes’ and so
are uncertain of what proposition is asserted by ‘Scarlet believes that a full house beats a
flush’.
Similarly, in the case of Forgetting, we say that our hesitance to assertively utter ‘Mus-
tard believes that bidding the five of diamonds indicates that one holds two aces’ is due to
our uncertainty about which proposition we would be committing ourselves to in uttering
this sentence. We hold back because, roughly, on certain ways of interpreting ‘believes’,
we ascribe a dispositional profile to Mustard involving the disposition, say, to assert as
much if asked without priming, but on other ways of interpreting ‘believes’, we ascribe a
dispositional profile to Mustard not including this disposition.
It’s worth noticing that other examples of unquestionably context-sensitive expressions
follow the same pattern. Consider:
Checkmate: Mrs. White is playing a game of Chess with Mr. Green. White is a
seasoned chess player, but knows that Green is just learning the game. White wants Green
to want to continue learn how the play the game, so is worried that if she wins too easily
he will be discouraged. If White takes Green’s knight on h2, then it will be mate in 1 for
White.
Now let me ask you: should White take the knight on h2? It is natural to feel some
hesitancy here to say that she should take the knight, and it is natural feel some hesitancy
to say that she shouldn’t take the knight. The obvious explanation for our hesitancy in
this case is that it’s unclear what I’m asking. Am I asking whether White should, in view
of the goals of chess, take the knight? Or am I asking whether White should, in view of
her desire to encourage Green to learn the game, take the knight? Once this is clarified,
our hesitancy falls away. One interpretation of ‘should’ licenses a “yes” answer, another
licenses a “no” answer. On the sensitive attitudes version of the dispositional account
of attitudes, our hesitancy to say that Scarlet believes that a full house beats a flush is
similarly downstream from uncertainty about what we would be saying.
The dispositional account of attitudes helps to make sense of cases of in-between be-
lieving, but the best way to spell out this account is with the sensitive attitudes view. To
bear an attitude expressed by an attitude verb V towards some proposition is ultimately
a matter of matching a dispositional profile, but which profile one must match depends
94
on the context in which V is used. I hope that now the general shape of the sensitive
attitudes dispositional view is clear; in the next section I’ll develop the view to account for
permissive readings of attitude verbs.
4. Permissive Dispositional Attitudes
To work this sensitive attitudes dispositional account into an account of permissive
readings, we have to say which kinds of dispositional profiles are associated with permis-
sive readings. To do so, I will distinguish two kinds of dispositions: conceptually free
dispositions and conceptually laden dispositions. My choice of terms here is meant to pat-
ternwith“conceptuallypermissive”and“conceptuallyloaded”; Iusethisnewlingosimply
to avoid confusion. With this distinction in place I propose
The Permissive Dispositions View:
A use u of a clause of the formAVs that S(F) is permissive with respect to an
occurrence o of linguistic constituent F in S(F) and a concept associated with
o if and only if the use u
0
of V involved in u serves to contribute a dispositional
profile V
u
0 which is conceptually free with respect to o and .
I’ll explain conceptually free dispositional profiles in a moment, but first notice a few
details of the permissive dispositions view.
Instead of speaking just in terms of whether a use of a particular report clause is
permissive per se, I here speak in terms of whether a use of a report clause is permissive
with respect to anoccurrenceofalinguisticconstituentofthecomplementofthereportand
an associated concept. I do this because there can be multiple occurrences of a particular
linguistic expression or phrase occurring in a particular ‘that’-clause, with the relevant
reportbeingpermissivewithrespecttosomebutnotallofthoseoccurrences. Forexample,
consider the following case:
Sister: Green sees Scarlet from afar. Scarlet is wearing her sister Jade’s green dress,
rather than Scarlet’s usual red dress. Green tells detective Black that the person he sees
from afar is Jade, Scarlet’s sister. Black and Grey both know that the person in the green
dress is Scarlet, not Scarlet’s sister. In discussing the situation, Black assertively utters
the following:
(57) Green thinks that Scarlet is Scarlet’s sister
What Black reports with (57) intuitively is permissive with respect to the first but
not the second occurrence of ‘Scarlet’. So it would be too rough to simply describe the
report made with (57) as permissive per se or even permissive with respect to ‘Scarlet’: it
95
is permissive with respect to the first occurrence of ‘Scarlet’ in (57).
81
Similarly, (57) is permissive with respect to a particular concept associated with ‘Scar-
let’. Arguably Green is represented by Black as thinking of Scarlet via some concept of
Scarlet, such as a descriptive demonstrative concept like that woman in the green dress.
So for a report to be permissive isn’t always for it to represent the subject as lacking
any concept whatsoever of the relevant object or property. For this reason, the permissive
dispositions view is expressed in terms of a report’s being permissive with respect to a par-
ticular concept associated with some phrase. Sometimes I lapse into speaking of whether
or not a given report is permissive per se, but the notion of permissiveness of fundamen-
tal importance is one relativized to particular concepts and occurrences of propositional
constituents.
Now, to make sense of the talk of “conceptually free dispositional profiles” let’s turn
to the issue of which dispositions are conceptually free and which are conceptually laden.
Simply put, a disposition is conceptually laden with respect to a concept if necessarily
one possesses that disposition only if one possesses . One such disposition would be
the disposition to cognize things via . Again, given the discussion above about how
Schwitzgebel’s and my dispositional account contrast with behavorialist dispositionalist
accounts, nothing prohibits us from allowing that certain dispositional profiles associated
with a particular attitude includes conceptually laden dispositions. We are free to use
dispositions such as the disposition to deploy certain concepts or the disposition to come
to possess other related loaded attitudes. Only those who burden themselves with some
reductionist or behaviorist project of accounting for the mental in terms of purely non-
mental dispositions have to avoid such “phenomenal” dispositions.
With this freedom to appeal to phenomenal dispositions in characterizing conceptually
laden dispositions, our dispositional account of attitudes can reap some of the benefits
associated with representational accounts. One of the many dispositions associated with
believing that the Dodgers will win the World Series may be the disposition to enter a
representational state whereby one cognizes the Dodgers via a concept The Dodgers and
to predicate the property will-win-the-World-Series, cognized via some concept, of the
baseball team so cognized. This, to me, seems pretty much all there is to having a fully
loaded attitude that the Dodgers will win the World Series. The sensitive attitudes version
of the dispositional account thus has the resources to make sense of fully loaded readings
of attitude reports.
A disposition which is conceptually free with respect to a concept is a disposition
81
In characterizing permissive readings I say that a report is permissive with respect to a particular
occurrence of a linguistic constituent, such as ‘Scarlet’; in characterizing whatever is expressed by attitude
verbs, relations or dispositional profiles, I will say that an attitude is permissive with respect to which
an occurrence of a constituent of a proposition, certain objects and properties, such as Scarlet herself. In
this case, I take the attitude that Black designates to be permissive with respect to the first occurrence of
Scarlet in the proposition said to be thought. These two ways of talking about permissiveness are useful
for talking both about certain uses of report sentences being permissive or loaded and for characterizing
certain dispositions which come to be associated with permissive or loaded reports.
96
which is not conceptually laden with respect to . Plenty of dispositions should be uncon-
troversially free: the dispositions which we are apt to associate with being fragile, such as
being disposed to break if struck, is conceptually free: an object, like a vase, can possess
this disposition without having any concept whatsoever. I will say that such dispositions
are “fully” conceptually free, since they are not conceptually laden with respect to any
concepts.
Theconceptuallyfreedispositionswhichcandoworkforusinaccountingforpermissive
readingsofattitudereportsareoneswhichweareapttoassociatewithparticularattitudes.
Reconsider the attitude of believing that the Dodgers will win the World Series. If the
Dodgers are playing the Yankees, then one disposition which one might associate with
this attitude is the disposition of betting that the Yankees will lose the World Series.
While not fully conceptually free, this disposition is free with respect to any concept of the
Dodgers. In some situations, the dispositional profile associated with the belief that the
DodgerswillwintheWorldSeriesmayincludethisconceptuallyfreedispositionalongwith
dispositions which are conceptually laden with respect to various concepts of the Dodgers.
But perhaps there are contexts in which the dispositional profile which is associated with
a use of ‘believes that the Dodgers will win the World Series’ contains only dispositions
which are conceptually free with respect to some particular concept D of the Dodgers.
Such a dispositional would be a profile which is conceptually free with respect to D.On
the permissive dispositions view, if there were such a case, then a use of ‘believes that the
Dodgers will win the World Series’ in that scenario would take a permissive reading with
respect to ‘Dodgers’ and D.
Here is an example:
Baseball: The detectives know that Mrs. White grew up in New York and is a Yankees
fan, though she doesn’t follow baseball too closely. She knows about the Yankees and the
Mets, but would be hard-pressed to name any other major league baseball team. Having
learned that the Yankees are in the World Series, White tells Black that they’re bound to
win, no matter who the opponent is. She takes a bet at long odds that the Yankees will
win the series. The detectives know that the Yankees are playing the Dodgers in the world
series, and also know that White is unaware of this. Later in discussing baseball, Black
assertively utters the following to Grey:
(58) White believes that the Dodgers are going to lose the World Series.
WhatBlackassertswith(58)soundstrueandappropriate,thoughsheisnotrepresented
as thinking in terms of ‘the Dodgers’: Black does not represent White as possessing or
deploying a linguistic concept of the Dodgers whereby she cognizes the Dodgers via the
phrase ‘the Dodgers’. Furthermore, in this scenario intuitively White is not represented
as possessing or deploying any concept of the Dodgers, save perhaps some descriptive
concept like the team the Yankees are playing. So the report made with (58) in Baseball
is permissive with respect to ‘the Dodgers’ and the concept which the detectives associate
97
with ‘the Dodgers’, which presumably is a concept D which one uses to cognizes the
Dodgers partly via the name ‘the Dodgers’.
Howthepermissivedispositionsviewtogetherwiththesensitiveattitudesviewaccounts
for this case is clear: given the conversational setup, Black and Grey associate a disposi-
tional profile with ‘believes that the Dodgers are going to lose the World Series’ which is
conceptually free with respect to D. This profile involves the disposition to bet that the
Yankees will win, the disposition to bet against whoever the opponent of the Yankees is,
to assertively utter that ‘the Yankees will win the World Series’, and perhaps others. The
report counts as true on this theory because White matches (perfectly, we might as well
say) this dispositional profile.
82
Since matching this dispositional profile does not require
possessing the concept D, the reading of (58) is also accounted as permissive with respect
to ‘the Dodgers’ and D.
Before moving on to apply the permissive dispositions view to various problem cases,
I’ll make some passing comments about the nature of conceptually laden and conceptually
free dispositions.
First note that in some contexts one can issue a report which is permissive with re-
spect to some occurrence of a propositional constituent F and concept but which still is
associated with dispositions that are not conceptually free with respect to every concept
of F (or with respect to every concept of certain pluralities or objects which are Fs). For
example, consider a variant of Forgeries from chapter 3:
Forgeries
1
: Detectives Black and Grey know that all and only the impressionist paint-
ings in the gallery are forgeries and that all and only the landscape paintings in the gallery
are forgeries. (So all and only the impressionist paintings are landscape paintings.) Both
factsarewellkeptsecrets; noneoftheguestsinthemanorknowthatthereareanyforgeries
in the gallery. Peacock, a guest in the manor, has told detective Black that she believes
that every impressionist painting in the gallery is an original Renoir. Later, in discussing
matters related to the gallery, Black assertively utters the following to Grey:
(1) Peacock believes that every forgery in the gallery is an original Renoir.
In this context, Black’s report is permissive with respect to ‘forgery in the gallery’
and the concept forgery-in-the-gallery. But intuitively Black does represent Peacock as
thinking of the things which in fact are forgeries in the gallery via some concept or other.
Some candidate dispositions which may be associated with this report are the disposition
to cognize the plurality F of things which in fact are forgeries via some concept or other
82
Note again that Schwitzgebel talks about matching a dispositional stereotype “to a su cient degree”,
where this degree is contextually determined. On the sensitive attitudes version of the dispositional view,
we instead associate di↵ erent dispositional profiles with di↵ erent uses of an attitude report sentence, so are
free to say that the attitude relation contributed by a use of a particular attitude report sentence is the
relation of perfectly matching the contextually associated profile, of possessing all of the dispositions in
that profile
98
and to predicate the property being-a-Renoir, cognized via the concept being-a-Renoir,
of F so cognized, or the disposition cognize the plurality F of things which in fact are
forgeries via impressionist-painting or landscape-painting and to predicate the property
being-a-Renoir, cognized via the concept being-a-Renoir, of F so cognized. These dispo-
sitions are conceptually free with respect to the occurrence of the property forgery in the
proposition said to be believed and the concept forgery, but are not conceptually free to
every concept of the plurality F.
I will say that these dispositions are conceptually “demanding” with respect to the oc-
currence of the property forgery-in-the-gallery in the proposition said to be believed.
A dispositiond which is conceptually demanding with respect to an occurrence of a propo-
sitional constituent o of proposition p is such that necessarily one possesses d only if one is
disposed to cognize o as it occurs in p via some concept or other. Similarly, a disposition
d which is conceptually demanding with respect to an occurrence of a propositional con-
stituent o of proposition p and with respect to a set of concepts C is such that necessarily
one possesses d only if one is disposed to cognize o as it occurs in p via some concept in
C. Given this, we can view a disposition which is conceptually laden with respect to an
occurrence of a propositional constituent o of proposition p and a concept as a special
case of conceptual demandingness: such a disposition is conceptually demanding with re-
spect to the relevant occurrence ofo inp and the set of concepts { }. The thing of interest
to note here is that a permissive report can be associated with dispositions which are not
conceptually laden with respect to an occurrence of a propositional constituent (and some
concept ) but which also are not fully conceptually free with respect to that constituent’s
occurrence.
Another thing to note is that a report can take a loaded reading with respect to an
occurrence of a propositional constituent expressed by a phrase F and a concept even
though is not associated with F by any linguistic convention. To see this, reconsider
Sister: Green sees Scarlet from afar. Scarlet is wearing her sister Jade’s green dress,
rather than Scarlet’s usual red dress. Green tells detective Black that the person he sees
from afar is Jade, Scarlet’s sister. Black and Grey both know that the person in the green
dress is Scarlet, not Scarlet’s sister. In discussing the situation, gesturing towards Scarlet
dressed in green, Black assertively utters the following:
(57) Green thinks that Scarlet is Scarlet’s sister.
We observed above that this report is permissive with respect to the first occurrence of
‘Scarlet’ in the ‘that’-clause of (57) (and so expresses an attitude that is permissive with
respect to the first occurrence of Scarlet in the proposition said to be thought) and the
concept Scarlet (by which one cognizes something via the name ‘Scarlet’). But intuitively
it is loaded with respect to the first occurrence of ‘Scarlet’ and a perceptual concept S
by which one cognizes something via certain visual means (something like “looking like
Scarlet” in certain conditions). On the permissive dispositions view, then, we say that
99
Black’s report in Sister is associated with a disposition which is conceptually laden with
respect to the first occurrence of Scarlet in the proposition said to be thought and the
concept S. In this case, S is not associated with ‘Scarlet’ by any linguistic convention,
but rather is associated with the first occurrence of that expression just given the relevant
conversational scenario.
Similarly, a report can be conceptually demanding with respect to an occurrence of a
propositional constituent and a set of concepts which aren’t even concepts of that propo-
sitional constituent; indeed, they need not be concepts of constituents of the same type.
To see this, reconsider the following example from chapter 3:
Blackmail: The detectives know that Scarlet, a guest at the manor, is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet: Green doesn’t know any name for Scarlet, could not identify her in a line-up,
and so on. Through questioning Green, Detective Black learns that Green thinks that
every socialist guest was being blackmailed by Mr. Body. Green doesn’t, however, know
the names of any of the socialist guests: he claims “not to know who the socialist guests
are”. The detectives are exchanging information, trying to find out who had a motive for
killing Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
The report Black makes with (2) is permissive with respect to ‘Scarlet’ and the concept
Scarlet in uttering (2) and is loaded with respect to ‘being blackmailed by Mr. Body’ and
the concept blackmailed-by-Body. But in this conversational scenario, it is plausible that
Black represents Green as possessing the concept socialist. So whatever dispositions we
use to characterize Black’s report in this context, it should include dispositions which are
conceptually demanding with respect to some things (perhaps propositional constituents
of propositions other than that which is ascribed as thought here) and the concept social-
ist. With this we see that the general phenomenon of permissiveness and loadedness not
only is not tightly linked to concepts which are associated by linguistic convention with
the constituents of a given report sentence, but neither are they tightly linked with the
particular propositions which are ascribed as thought contents by a given report. We are
beginningtosee,then,thatthephenomenonofpermissiveandloadedreadingisthoroughly
pragmatic; a full account of permissive and loaded reports cannot be myopically semantic.
Understanding permissive and loaded reporting requires careful attention to pragmatics;
this will be a central theme in later chapters.
5. Using Permissive Dispositions
To account for the cases raised in §2 involving subjects who plausibly lack any concepts
whatsoever, we need only identify dispositions which both plausibly could be associated
with certain uses of attitude verbs and which could be possessed by non-human subjects
like chess programs, simple animals, or groups.
100
To do this, I suggest that we consider cases similar to those in §2 but in which the
relevant attitude is ascribed instead to a normal human thinker, and determine what fully
conceptually free dispositions are plausibly part of the dispositional stereotype normally
associated with the attitudes ascribed in those parallel cases. Once we’ve identified these
conceptually free dispositions, we then ask whether it is plausible that the abnormal non-
human subjects could possess them. We then claim that the permissive reports from §2
involve dispositional profiles including such conceptually free dispositions.
5.1. Permissive Dispositions and Fully Permissive Cases
Let’s start with Chess. Consider a parallel case:
Chess
2
: Professor Plum is playing chess against Mrs. Peacock, who is a well-ranked
chess player. Detectives Black and Grey are looking on. After Plum moves his bishop to
d4, Peacock moves a bishop to d7. Black turns to Grey and sincerely utters:
(59) Peacock is boxing out Plum’s knight because she thinks that Plum wants to fork
the queen and rook.
The report which Black makes with (59) takes a loaded reading with respect to ‘Plum’,
‘fork’,‘queen’,and‘rook’. Onthesensitivedispositionalattitudesview,thisisexplainedby
Black’s use of ‘believes’ being associated with certain dispositions which are conceptually
laden with respect to concepts associated with these linguistic constituents.
Presumably Black and Grey also associate other dispositions with this attitude which
are not so laden: for instance, the disposition to move pieces around the chess board in
such a way that would prevent a situation in which both a queen and a rook were either
two squares away horizontally and one square vertically from an opposing knight or two
squares vertically and one square horizontally from that opposing knight (in which case
the queen and rook would be “forked” by that knight). This is distinct from a disposition
to move pieces around the chess board in such a way that would prevent a situation which
one would conceive in these terms; rather, this is a disposition which any simple piece-
moving machine could possess if it were appropriately designed. We could program such
a machine to maneuver pieces on a chess board to prevent such situations without that
machine’s possessing concepts which we may deploy in characterizing those situations: the
disposition is fully conceptually free. Nevertheless, it is still a disposition which Black and
Grey plausibly take Peacock also to possess. Similar dispositions are the disposition to
move pieces of one color around a chess board to bring about a situation in which the
king piece of the other color is threatened with capture and there is no way to remove the
threat, the disposition to move pieces of one color around a chess board so that according
to the rules of chess as few pieces of that color could be captured without recapture, and
so on. These are also dispositions which Black and Grey plausibly associate with Peacock
in Chess.
101
ThisclusteroffullyconceptuallyfreedispositionsarealsodispositionswhichLeelapos-
sesses. The computer program has its own ways of moving pieces, di↵ erent from Peacock’s
way of moving pieces, but this does not prevent the program from being disposed to move
pieces, in its way, to prevent situations in which a queen and rook of a particular color are
forked by a knight of the opposing color. The permissive dispositions view then can say
that in the original Chess case Black’s use of ‘believes’ is associated with a dispositional
profileconsistingjustofsuchfullyconceptuallyfreedispositions. Theprofileincludesmany
but not all of the dispositions which are associated with ‘thinks that Plum wants to fork
the queen and rook’ in Chess
2
. Leela matches this dispositional profile despite lacking any
concepts whatsoever. So, ignoring the important details of how a particular dispositional
profile comes to be associated with a particular use of an attitude-reporting clause, the
permissive dispositions view has the right shape to account for the truth of (3) in Chess.
The other examples from §2 are similarly accounted for on the permissive dispositions
view. The group case exhibited in Party is similar enough. For a normal human subject,
such as Mrs. Peacock, to think that a fairytale party theme will be best at attracting
partyattendees(inascenariolikeParty)wouldbeassociatedwiththedispositiontosolicit
funding for the upcoming party which happened to be organized around that theme. This
is a conceptually free disposition with respect to the concept fairytale-themed-party, one
which Peacock could have without possessing that concept.
Similar dispositions are the disposition to announce that the fairytale theme would
be best at attracting party attendees or the disposition to orchestrate a party around a
fairytale theme. While it might seem that being disposed to announce that a fairytale-
themed party is best is a conceptually laden disposition, this need not be so.
Here a version of the permissive attitudes view extended to speech act reports can
help. While in some cases announcing p may be associated with possessing concepts of
the constituents of p, in other cases announcing p may itself be conceptually permissive.
Consider clocks which “announce” the time, and then think of a clock which instead of
announcing the time plays an automated message to the e↵ ect that di↵ erent themes would
be best at attracting attendees. Suppose that at noon the clock plays a message that a
sci-fi theme would be best for the holiday party. It makes sense to say that the clock
announces at noon that a sci-fi theme would be best for the holiday party, but of course
the clock itself yet lacks a concept of the property sci-fi-themed-party.Sotheclock’s
fully conceptually free dispositions ground an ‘announces’ report about the clock. On
the permissive dispositions view, we associate a conceptually free dispositional profile with
announcingasmuchwiththeclocktomakesenseofsuchspeechactattributions. Asimilar
disposition may be similarly associated with the party-planning committee.
As long as a group like the party-planning committee can itself solicit funding, issue
announcements about party themes, orchestrate a party around a certain theme, and so
on, then such groups can be disposed to perform such actions. So there are clusters of
dispositions which may be associated with the attitude of thinking that a fairytale theme
will be best at attracting attendees which are conceptually free with respect to the concept
102
fairytale-themed-party and can be possessed by the party-planning committee. With the
permissive dispositions view, we can say that Black’s use of ‘believe’ involved in her use
of (55) is associated with a dispositional profile including only such fully conceptually free
dispositions. This account of the truth of (55) in Party shops out the important question
of what it is for groups to perform certain kinds of actions, such as soliciting funding
or announcing things, but this seems to be a need which could be met. The permissive
dispositions view is no worse o↵ for not o↵ ering a full-blown account of group action.
At last, let’s turn to the examples from Lecture. These cannot be directly handled by
thesamestrategyasthatusedinChess orParty, sincepresumablynormalhumanthinkers
do not have subthreshold oscillations or closed shells. We would have to strain ourselves
in finding parallel cases involving normal human thinkers and exactly the ‘that’-clauses of
(53) and (54). But there seem to be types of fully permissive dispositions associated with
preferring and believing generally which could be used to make sense of these cases.
Consider Smith (1987)’s proposal that a desire to perform an action A consists in a
dispositiontoAincertainconditions. Todesiretosmokeacigaretteconsistsinaclusterof
dispositions including at least dispositions to smoke a cigarette in various conditions, such
as the condition of having access to a cigarette and believing that smoking is permitted.
It may involve further dispositions, such as one’s disposition to walk to a local 7 Eleven
in the condition that one believes cigarettes are sold at reasonable price at that 7 Eleven.
Similarly, we can say that to preferAing toBing involves a disposition toA rather than to
B in certain conditions in which both actions are available. So to prefer smoking American
Spirits to smoking Marlboros consists in a disposition to buy a pack of American Spirits
in certain conditions in which American Spirits and Marlboros are both available to buy,
a disposition to smoke an American Spirit in certain conditions in which one could either
smoke an American Spirit cigarette or a Marlboro cigarette, and so on.
Extending this kind of approach to examples like (53) in Lecture is not immediately
straightforward,sincethatexampleinvolvesapreferencereportinwhichthesubjectprefers
being in a particular kind of state or prefers one kind of object over another. To handle
this kind of case, I argue first that there is not really any thing such as a preference per se
foranobject: atleasttypically, whenoneassertivelyuttersasentenceoftheforma prefers
o to o
0
, for some object-designating noun phrases a, o, and o
0
that subject a prefers object
o to object o
0
, I take it that this is somehow pragmatically enriched to assert a proposition
which could be more explicitly asserted by a sentence of the form a prefers A(o)ing to
B(o
0
)ing,where A(o) and B(o
0
) are complex verb phrases including occurrences of o and
o
0
, respectively. For example, to assertively utter that one ‘prefers chocolate ice cream
to strawberry ice cream’ is, in many contexts, in e↵ ect to assert that one prefers tasting
chocolate ice cream to tasting strawberry ice cream.
83
In other contexts, uttering such a
sentencemayinsteadservetoassertthatoneprefersmaking chocolateicecreamtomaking
strawberryicecream. Giventhegoingapproachtopreferenceattributions,tosayincertain
83
See Je↵ rey (1965) for this kind of idea.
103
contexts that one prefers chocolate ice cream to strawberry ice cream is represent one as
possessing a cluster of dispositions including the disposition to eat chocolate ice cream in
certain conditions in which one could either eat chocolate ice cream or eat strawberry ice
cream.
We should say something similar for the case in which we say that a neuron prefers
one kind of input over another. What we presumably are asserting in such a case is that
the neuron prefers to fire in response to one kind of input rather than another kind of
input. So to say in certain contexts that a neuron prefers one kind of input over another
is to represent the neuron as possessing the disposition to fire in response to one kind of
input in certain conditions when firing in response to either kind of input is available, or
as possessing the disposition to fire in response to one kind of input but not possessing the
disposition to fire in response to other kinds of inputs, or as possessing the disposition to
fire in response to one kind of input in conditions C
1
...C
n
and possessing the disposition
to fire in response to another kind of input in conditions C
m
...C
k
such that |C
1
...C
n
| >
|C
m
...C
k
|⇥ f, for some factor f.
Consider the following case, which is similar in relevant respects to Lecture:
Neuron: Plum is playing around with a toy model of neurons. He is monitoring the
input frequencies given to a synthetic resonator neuron. When di↵ erent input frequencies
stimulate the synthetic neuron, Plum presses a button that causes the synthetic neuron to
fire. Currently he is interested in testing how his model neural system behaves when the
synthetic neuron mostly fires in response to frequencies which resonate with the neuron’s
subthreshold oscillations. Black knows all of this. In discussing Plum’s experiment with
Grey, Black assertively utters
(60) Plum prefers inputs having frequencies that resonate with the frequency of its
subthreshold oscillations.
Firstlet’smaketwoobservationsaboutBlack’sreport. First,itisloadedwithrespectto
‘frequencies’. Second,intuitivelysheassertssomethingtothee↵ ectthatPlumprefersfiring
his synthetic neuron in response to inputshavingcertainresonatingfrequenciesrather than
doing so in response to non-resonating frequencies. Given these observations, a disposition
intuitively associated with Black’s use of ‘prefers’ is the disposition to fire the synthetic
neuron in response to inputs which Plum cognizes as having frequencies resonating with
the synthetic neuron’s subthreshold oscillations. Were, however, a report made with (60)
to take a permissive reading with respect to ‘frequencies’, intuitively a di↵ erent disposition
would be associated with Black’s use of ‘prefers’. In such a case, intuitively we would
associate the use of ‘prefers’ with dispositions such as the disposition to fire the synthetic
neuroninconditionsinwhichtheinputinfact hasfrequenciesresonatingwiththesynthetic
neuron’ssubthresholdoscillations. Thisdispositionisconceptuallyfreewithrespecttoany
concept of frequencies resonating with this or that. A concept-free machine designed to
press the button which fires the synthetic neuron could also possess this disposition.
104
Looking back on Plum’s original use of (53), then, we can say that similar disposi-
tions are associated with Plum’s use of ‘prefers’ in Lecture. These may include a neuron’s
disposition to fire in response to input frequencies resonating with its subthreshold oscil-
lations or the disposition to not fire in response to input frequencies not resonating with
its subthreshold oscillations.
5.2. Permissive Dispositions and Partially Permissive Reports
I hope now that the way that the permissive dispositions view handles fully permissive
reports like those in §2 is clear enough. Now I’d like to show how the permissive disposi-
tions view can handle the cases of partially permissive reports from chapter 3, which raised
problems for the sensitive attitudes account from chapter 2 and the contextual augmenta-
tions views from chapter 3. These cases involve reports which are permissive with respect
to some but not all of the linguistic constituents of a report sentence. Recall:
Blackmail: The detectives know that Scarlet, a guest at the manor, is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet: Green doesn’t know any name for Scarlet, could not identify her in a line-up,
and so on. Through questioning Green, Detective Black learns that Green thinks that
every socialist guest was being blackmailed by Mr. Body. Green doesn’t, however, know
the names of any of the socialist guests: he claims “not to know who the socialist guests
are”. The detectives are exchanging information, trying to find out who had a motive for
killing Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
Again, Black’s report is permissive with respect to ‘Scarlet’ and the concept Scarlet in
uttering(2)andisloadedwithrespectto‘beingblackmailedbyMr. Body’andtheconcept
blackmailed-by-Body. Are there any partially permissive dispositions which could plausibly
beassociatedwiththereportedattitudeinBlackmail whichGreencouldplausiblypossess?
HereIsuggestthatwedrawontheintuitivethoughtbehindthecounterfactualattitudes
view, which we met early in chapter 3. On that view, all permissive reports are to be
understood in terms what loaded attitudes a subject would have were they to learn certain
information which the conversationalists commonly assume. So, were Green to come to
learn that Scarlet is a socialist, Green would have a loaded belief that Scarlet was being
blackmailed by Body, given his loaded general belief that every socialist was blackmailed
by Body.
Even though the counterfactual attitudes view cannot make sense of all permissive
reports, we can use the basic thought behind that view to account for this case with
the permissive dispositions view. Intuitively, in Blackmail the conversationalists associate
the attitude of thinking that Scarlet was being blackmailed by Body with the disposition
to predicate the property blackmailed-by-Body of Scarlet in the condition of gaining a
concept of Scarlet and learning what the conversationalist’s know about her: that she is
105
a socialist. This disposition is conceptually free with respect to concepts of Scarlet; one
can be disposed to come to have certain loaded attitudes involving a concept of Scarlet
without antecedently possessing such a concept.
To motivate the idea that this sort of disposition is indeed associated with Black’s use
of (2), consider the following modified version of Blackmail:
Blackmail
3
: ThedetectivesknowthatScarlet,aguestatthemanor,isasocialist. They
also know that she is proudly and openly a socialist: Scarlet often announces that she is a
socialist to others at the manor and easily gets into heated political debates in which she
loudly defends socialist political ideas. The detectives also commonly know that Green,
another guest at the manor, is unacquainted with Scarlet: Green doesn’t know any name
for Scarlet, could not identify her in a line-up, and so on. Through questioning Green,
detective Black learns that Green thinks that every socialist guest was being blackmailed
by Mr. Body. This belief of Green’s is based on his belief that every socialist guest at
the manor desperately wants to keep their political views a secret, for fear of becoming
involved in some kind of scandal. Green doesn’t, however, know the names of any of the
socialist guests: he claims “not to know who the socialist guests are”. The detectives
are exchanging information, trying to find out who had a motive for killing Body. Black
assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
Inthiscase,intuitivelywhatBlackassertswith(2)ispermissivewithrespectto‘Scarlet’
but it also seems intuitive that this report is false. It is at least much more di cult to
interpret the assertion as straightforwardly true. Our way of applying the permissive
dispositions view to the original Blackmail case o↵ ers an explanation for the false reading
of (2) in Blackmail
3
. For in Blackmail
3
Green does not have the disposition to predicate
the property blackmailed-by-Body of Scarlet in the condition of gaining a concept of
Scarlet and learning what the conversationalist’s know about her. Were Green to learn
that Scarlet makes no e↵ ort to hide her socialist views, he would presumably abandon his
general belief that every socialist guest was being blackmailed by Body. So the permissive
dispositions view can account for both the permissiveness and the falsity of Black’s report
using this particular kind of conceptually free disposition.
Howdoesthepermissivedispositionsviewavoidtheuntowardconsequencesofthecoun-
terfactual attitudes view? Recall that the counterfactual attitudes view faced a problem
with the following case:
Forgeries
1
: Detectives Black and Grey know that all and only the impressionist paint-
ings in the gallery are forgeries, but that this fact is a well kept secret, so that none of the
guests in the manor know that there are any forgeries in the gallery. Peacock, a guest in
the manor, has told detective Black that she believes that every impressionist painting in
the gallery is an original Renoir. Later, in discussing matters related to the gallery, Black
106
assertively utters the following to Grey:
(1) Peacock believes that every forgery in the gallery is an original Renoir.
Thecounterfactualattitudesviewstruggledwiththisexample,sincePeacockwouldnot
come to have a loaded belief that every forgery in the gallery is an original Renoir were she
to learn what the detectives know. Were Peacock to learn that all and only the forgeries
are impressionist paintings, she would instead abandon her belief that every impressionist
painting in the gallery is an original Renoir.
The permissive dispositions view, however, has the resources to avoid wrongly pre-
dicting a false reading of (1) in Forgeries
1
. For in accounting for a permissive report
that a subject believes a proposition p, the permissive dispositions view is not wedded
to always appealing to the disposition to come to have a loaded belief in p in the con-
dition of learning what the conversationalists know. There are other conceptually free
dispositions which plausibly are associated with Black’s use of ‘believes that every forgery
in the gallery is an original Renoir’ in Forgeries
1
. For instance, one disposition which
arguably is associated with Black’s report is the disposition to predicate the property
being-an-original-Renoir of the things which are forgeries in the gallery. This disposi-
tionisconceptuallyfreewithrespecttotheconceptforgery-in-the-gallery: indeed, Peacock
possesses this disposition despite her lacking, or failing to deploy in the relevant scenario,
this concept. As long as we can say that only conceptually free dispositions of this kind are
associated with Black’s report, we can account for the permissive reading of (1) without
having to say that (1) is false in Forgeries
1
.
Here we see that the flexibility of the permissive dispositions view can avail itself of
the advantages of the counterfactual attitudes view without succumbing to that view’s
shortcomings. In this way, the permissive dispositions view can make sense of other prob-
lematic examples from chapter 3. The examples of Society and Scotch from §1 of chapter
3 receive a similar treatment. In each, the detectives know some propositions such that,
were the subject to come to learn those propositions, they would be disposed to have a
loaded attitude expressible by the report sentence uttered by the speaker. In both of those
cases, as long as the conversationalists associate with the speaker’s report a disposition
to come to possess the relevant loaded attitude upon coming to learn certain contextually
known propositions, then the permissive dispositions view predicts the permissive readings
we find in those cases.
5.3. Permissive Dispositions and Problems with Contextual Augmentation
Now let’s turn to the data which raised problems for contextual augmentation views of
permissiveness. We identified five major classes of problems, which I’ll address here.
Icalledthefirstclassofproblems“irrelevantentailment”cases. Theseexamplesshowed
that propositions in an agent’s attitude set can contextually entail propositions which that
agent cannot be truly said to bear the relevant attitude toward. I’ll start with the simpler
107
irrelevant entailment problems; the general strategy which I’ll advocate for these problems
is pretty much the same for the more complicated cases.
Pipe: Mrs. Peacock believes that the pipe is the murder weapon. The detectives
know two things which Peacock doesn’t: first, the pipe is the only weapon hidden in
the conservatory; second, the pipe is not the murder weapon. Detective Black knows of
Peacock’s belief about the pipe, and assertively utters each of the following to Grey:
(43) Peacock believes that the candlestick is the murder weapon.
The report Black makes with (43) is false. Notice that intuitively it is also fully loaded.
Simple contextual augmentation views predicted a true permissive reading should be avail-
ablehere, sincePeacock’sbeliefsettogetherwiththecontextualassumptionsinPipe entail
all propositions, including that the candlestick is the murder weapon.
On the permissive dispositions view we can account for the falsehood of (43) by saying
thattheconversationalistsareassociatingfullyconceptuallyladendispositionswithBlack’s
reportsentence,suchasthedispositiontopredicatetheproperty murder-weapon,cognized
via some descriptive concept, of the candlestick, cognized via a descriptive concept like the
candlestick. Since Peacock lacks these dispositions, Black’s report is false. That’s all there
istothepermissivedisposition’sexplanationofthisexample. Noticethattheapproachhere
is to point out that the false reading also is loaded, and then to find standard conceptually
laden dispositions to associate with the relevant report sentence.
Now reconsider a more complicated case:
Marbles: The detectives know that there is an urn in the library; they know that there
are exactly ten red marbles in in the urn, but are not sure how many marbles total are in
the urn. Detective Black has just spoken to professor Plum, who told Black that he thinks
there are exactly five blue marbles in the urn. The detectives are discussing how many
marbles are in the urn, and Black sincerely utters the following to Grey:
(50) Plum thinks that there are (exactly) fifteen marbles in the urn.
What Black asserts here is false, but notice that her report also intuitively is loaded
with respect to ‘fifteen’. Given this loaded reading, the permissive dispositions view has to
include a disposition which is conceptually laden with respect to the concept contextually
associated with ‘fifteen’ in the dispositional profile associated with Black’s use of ‘thinks
that there are exactly fifteen marbles in the urn’. An example of such a disposition would
be the disposition to say ‘fifteen’ when asked how many marbles are in the urn. Plum lacks
such dispositions, so fails to match the dispositional profile associated with Black’s report,
so the report is predicted rightly to be false.
The other irrelevant entailment problems are handled in this way. Something which
standsouthereisthat, unlikesimplecontextualaugmentationviews, thepermissivedispo-
sitions view implicitly distinguishes the question of whether a report is true from whether
108
a report is permissive. The permissive dispositions view begins by providing an expla-
nation of the loaded reading of the relevant reports and then extending that explanation
to account for the reading’s falsehood. One thing I do in the next three chapters is to
try to take this lesson learned from the permissive dispositions view and apply it to more
sophisticated versions of the contextual augmentation view.
Next let’s turn to acquaintance problems. These problems show that it is di cult
for a report of the formAVsthatS(F) to take a permissive reading unless given the
conversationalists’ assumptions A is not disposed to think of things as F. The following
variation on the Blackmail case illustrates this:
Blackmail
2
: The context is just like the original Blackmail, save for that the detectives
donot commonlyassumethatGreenisunacquaintedwithScarlet. Rather, theyknowthat
GreeniswellacquaintedwithScarlet: GreenandScarletareclosefriends. AsinBlackmail,
the detectives know that that Scarlet is a socialist, and Black knows that Green thinks
that every socialist was being blackmailed by Mr. Body. The detectives are exchanging
information, trying to find out who had a motive for killing Body. Black assertively utters
the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
Here (2) is di cult to interpret as either true or permissive. Given the contextual as-
sumption that Green is well acquainted with Scarlet, it is natural for the conversationalists
to associate more than just permissive dispositions with respect to Scarlet with Black’s re-
port. With the permissive dispositions view we then can say that such a loaded disposition
is in the dispositional profile associated with Black’s report. In particular, we can include
the disposition to predicate the property being-blackmailed-by-Body, cognized via a
contextually associated concept, of Scarlet, cognized via a contextually associated concept
of Scarlet. Since Green lacks such a disposition, we can account for the false reading of
(2).
The asymmetry problem gets a similar treatment. Reconsider:
Switched Blackmail: Mr. Body has been murdered; detectives Black and Grey are
investigating the crime. The detectives have private information that every socialist guest
was being blackmailed by Mr. Body. They know that Green is unaware that Body was
blackmailing any guests at all. Having just questioned Green, Black knows that Green
believes that Scarlet (with whom Green is well acquainted) is a socialist. Although the
detectives commonly know that Scarlet is a guest, they do not know that she is a socialist.
They are exchanging information, trying to find out who had a motive for killing Body;
in particular, in this scenario they are trying to figure out who was being blackmailed by
Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
109
Here intuitively Black makes a false loaded report. The problem raised by this example
forcontextualaugmentationviewsisofcoursethatthepropositiondesignatedbythe‘that’-
clause of (2) is entailed by the proposition that Scarlet is a socialist and the proposition
that every socialist was being blackmailed by Body. These are the same propositions at
work in the original Blackmail case; what has changed is which proposition is contextually
assumed and which is a content of Green’s loaded attitudes.
The permissive dispositions view can be used to make sense of this case once again by
first pointing out that Black’s report is loaded, specifically with respect to ‘being black-
mailed by Mr. Body’. So the permissive dispositions view requires that the dispositional
profileassociatedwithBlack’sreportsentenceinthiscontextincludesadispositionwhichis
conceptuallyladenwithrespecttoaconceptofthepropertybeing-blackmailed-by-Body.
An obvious disposition to appeal to is, again, the disposition to predicate the property
being-blackmailed-by-Body, cognized via a contextually associated concept, of Scarlet.
Green lacks this disposition, so the permissive dispositions view can deliver a false reading
here. The way context-sensitivity is built into the permissive dispositions view, with its
focus on what parts of a report sentence’s ‘that’-clause a report made with that sentence
is loaded with respect to,a↵ ords us the flexibility to predict that a use of (2) is false in
Switched Blackmail but true in Blackmail. Again, this is an important lesson, which I
take to heart in developing di↵ erent versions of the contextual augmentation view in later
chapters.
Now let’s turn to the problem of erroneous assumptions, which illustrates that in cases
where the conversationalists make a false assumption a report which they take to be per-
missive and true is instead permissive and false. Reconsider:
Study: Plum is in the lounge. Detective Black knows that Mustard believes that Plum
isintheroomwhichisconnectedtotheconservatorybysecretpassage. ButdetectiveBlack
and Grey falsely believe that the study, not the lounge, is connected to the conservatory by
secret passage. They also assume that Mustard is unaware, or does not believe, that the
study is connected to the conservatory by the secret passage. Detectives Black sincerely
utters the following to Grey:
(52) Mustard believes that Plum is in the study.
In this context, Black’s report takes a permissive reading with respect to ‘the study’.
Were the detectives’ contextual assumption true, intuitively the report would be true,
though given their error we instead judge it to be false. So the permissive dispositions view
needs to say that the detectives associate a disposition which is permissive with respect
to the ‘the study’ with Black’s report sentence, but that this permissive disposition is one
which Mustard lacks. We have some di↵ erent options about how to achieve this. One is to
includeintherelevantdispositionalprofileadispositiontopredicatebeing-in-the-study,
cognized some way or other (but not necessarily via the concept the study which the
detectives conversationally associate with ‘the study’) of Plum. While this disposition
110
might serve in this case, this basic strategy won’t work in general. For consider
Study
1
: Plumisin thelounge. DetectiveBlack knowsthat Mustard believesthat Plum
isintheroomwhichisconnectedtotheconservatorybysecretpassage. Blackfurtherknows
that Mustard believes that the room connected to the conservatory by secret passage is
the room with the revolver in it, so that Mustard also believes that Plum is in the room
with the revolver in it. Unbeknownst to Black, Grey, and Mustard, the room with the
revolver in it is the study. Black and Grey falsely believe that the study, not the lounge,
is connected to the conservatory by secret passage. They also assume that Mustard is
unaware, or does not believe, that the study is connected to the conservatory by the secret
passage. Detective Black sincerely utters the following to Grey:
(52) Mustard believes that Plum is in the study.
The judgment here is subtle. The report intuitively is permissive, as before, though to
myearthereportremainsfalse,eventhoughMustarddoespredicatetherelationbeing-in
ofPlumandthestudy,cognizingthestudyviatheconceptthe-room-with-the-revolver-in-it.
Insofar as this is right, then the approach sketched above won’t work here: Mustard does
possess a disposition to predicate being-in-the-study, cognized some way or other (but
not necessarily via the concept the study which the detectives conversationally associate
with ‘the study’) of Scarlet. Namely, he predicates that property, cognized via the concept
the-room-with-the-revolver-in-it,ofScarlet. Accordingtothatsuggestion,then,(52)should
take a straightforwardly true reading.
In both Study and Study
1
, then, the permissive dispositions view needs to appeal to
some other disposition which is permissive with respect to ‘the study’ but which Mustard
lacks. Analternativedispositiontouseisthedispositiontopredicatetherelationbeing-in
of Plum and the study, where the study is cognized via some concept other than the-study,
which the conversationalists primarily associate with ‘the study’. Intuitively in both Study
and Study
1
the detectives do take Mustard to have such a disposition, and this disposition
isconceptuallypermissivewithrespecttotheconceptwhichtheconversationalistsMustard
lacks this disposition, allowing us to predict that (52) is permissive but false in both Study
and Study
1
.
Regardless of the judgment one has regarding the truth value of (52) in Study
1
,the
permissive dispositions view has the resources to account for the uses of (52) in both Study
and Study
1
. The flexibility of which permissive dispositions we use lets us account for the
former as false but the latter as true or to account for the former as false but the latter as
false. The flexibility of the permissive dispositions view, once again, is a powerful tool in
being able to account for di↵ erent permissive reports.
6. Worries
ToconcludeIwillconsideracoupleofworrieswiththepermissivedispositionsaccount.
111
6.1. Permissive Speech Act Reports
The first worry concerns the fact that not only do attitude reports a↵ ord for permis-
sive and loaded readings, but so do speech act reports made with verbs like ‘say’, ‘assert’,
‘ask’, ‘command’, and so on. The worry is that to perform a speech act such as saying,
asserting, asking, or commanding, cannot just be a matter of possessing certain disposi-
tions. Presumablyperformingaspeechactalsoinvolvesbehavingincertainwaysatcertain
times.
Consider some examples of permissive speech act reports:
Form: White has told Black that every person in the dining room wants to eat salmon
for dinner and every person in the lounge wants chicken for dinner. Black and Grey both
know that Scarlet is in the dining room, but also that White is unaware that Scarlet is in
the dining room. They know that White is simply informed by the butler and the maids
how many dishes to prepare for those dining in the di↵ erent rooms. Indeed, Black and
Grey know that White is unaware of Scarlet’s existence. White doesn’t know the name
‘Scarlet’, couldn’t identify Scarlet in a line-up, and so on. Discussing what Scarlet is to eat
what for dinner, Black assertively utters the following to Grey:
(61) White said that Scarlet wants the salmon for dinner.
In this context, Black’s speech act report intuitively seems permissive with respect to
‘Scarlet’. It would be odd for Grey to respond, given his knowledge of the scenario, with
‘Wait, White knows who Scarlet is?’. Given what Black and Grey know about White’s
methods for determining which kinds of guests want which kind of dinner, it is clear that
Black is not representing White as having made an assertion about Scarlet in particular.
Another case:
Forgeries
4
: Detectives Black and Grey know that all and only the impressionist paint-
ingsinthegalleryareforgeries; theyalsoknowthatallandonlytheimpressionistpaintings
in the hall are forgeries. These are well kept secrets; none of the guests in the manor know
that there are any forgeries in the gallery or in the hall. Peacock, a guest in the manor, has
asked detective Black whether every impressionist painting in the gallery is a landscape.
Later, in discussing matters related to the paintings in the manor, Black assertively utters
the following to Grey:
(62) Peacock asked whether every forgery in the gallery is a landscape.
In this context, Black’s speech act report intuitively seems permissive with respect to
‘forgery in the gallery’. It would be odd for Grey to respond, given his knowledge of the
scenario, with ‘Wait, Peacock knows that there are forgeries in the gallery?’. Given what
Black and Grey know about Peacock’s knowledge of forgeries, it is clear that Black is not
representing Peacock as having asked a question explicitly about “forgeries”.
One more:
112
Meters: Green is overseeing the construction of trellises in the conservatory. Green has
demandedthatMustardbuildeverytrellistobeatleasttenfeettall. BlackandGreyknow
that Green is an unwashed American, ignorant of the metric system. But the sophisticated
Black and Grey tra c mainly in metric. Black assertively utters the following to Grey:
(63) Green demanded that Plum build every trellis to be at least three meters tall.
In this context, Black’s speech act report intuitively is permissive with respect to ‘three
meters tall’. Given what Grey knows about Green, it would be odd for Green to respond
with ‘Wait, Green is familiar with the metric system?’ Given what Black and Grey know
about Green, it is clear that Black is not representing Green as having made a demand
explicitly about how tall the trellises stand in meters.
With these examples, we see plainly that speech act reports take something like per-
missive readings. One can imagine contexts in which each of the foregoing example report
sentences could also take something like loaded readings.
Now the worry. In each of these cases, to say that the permissive speech act reports
made in these cases just serve to represent di↵ erent subjects as possessing a cluster of
dispositions would apparently fail to account for the truth conditions of these reports.
Saying that something is so, asking whether something is so, or demanding that something
be so, involves more than merely being disposed to behave in certain ways, to feel certain
feelings, or to enter certain mental states. Rather, performing a speech act involves at
least executing certain behaviors at certain points in time. So, the worry is, the permissive
dispositions account has the wrong shape to make sense of permissive speech act reports,
since on that view a report’s being permissive consists in its just representing a subject as
possessing a certain cluster of dispositions.
Inresponse, itfirstisimportanttonotethatalthoughperformingaspeechacttypically
involves the uttering or inscribing of words, the speech act thereby performed is not to be
identified with such an utterance or inscription. The illocutionary acts of asserting, asking,
commanding, and so on, are best viewed as distinct from the locutionary acts of uttering
and inscribing by which one can perform such speech acts. Holding this distinction in
mind, we now note that it is generally thought that what makes someone’s utterance of a
particularsentencelike‘Youwillarriveontime’servetomakeaparticularspeechact, such
asanassertion,prediction,command,orrequest,isthattheuttereralsohas,orisrationally
assumedtohave,certainbeliefs,desires,orintentions.
84
Accordingly,torepresentsomeone
as asserting p not only is to represent them as having uttered or inscribed certain words
at a time, but to have been in certain accompanying mental states.
84
As in Bach & Harnish (1979)’s elaborate theory and taxonomy of speech acts, which distinguishes be-
tweenvariouspotentialspeechacttypesintermsofwhatattitudesare“expressed”incertaincircumstances
by one’s performing some locutionary act. In asserting p one (at least) intends for a hearer to take one’s
utterance as reason to think that one believes p;inaskingwhether p, one (at least) intends for a hearer to
take one’s utterance as reason to think that one desires that they tell one whether or not p.
113
Given that being represented as performing certain speech acts, over and above per-
forming mere locutionary acts, involves being represented as having certain beliefs, desires,
or intentions, on the permissive dispositions view the truth conditions of a speech act re-
port will involve the subject’s possessing certain dispositions. Let’s see how this goes with
the examples above.
The structure of the conversational scenario in Form is like that of Blackmail: in each,
the subject is known by the speaker to bear a relevant loaded attitude or speech act rela-
tion towards a general proposition of the form Every F is G, the conversationalists know a
proposition of the formaisF, so in context the speaker is in a position to ascribe a related
permissive attitude relation or permissive speech act relation to the subject and a proposi-
tion of the formaisG. Given this similarity, the permissive dispositions view here should
adverttothekindofdispositionsusedinaccountingforthereportinBlackmail: Blackrep-
resents White as possessing the disposition to predicate the property wants-the-salmon
of Scarlet in the condition of gaining a concept of Scarlet and learning what the conversa-
tionalist’s know about her: that she is in the dining room. That is, White is represented
as having the disposition to come to have a loaded belief that Scarlet wants the salmon in
the condition of learning that Scarlet is in the dining room.
85
This does not exhaust how Black represents White in making her report with (62):
Black also represents White as having performed some locutionary acts or other. Notice,
Black does not represent White as having perform a locutionary act of uttering ‘Scarlet
wants the salmon for dinner’; rather, Black represents White as having performed a locu-
tionary act of uttering some sentence or other which served to express her loaded belief in
a related proposition. Just as some permissive reports are conceptually demanding with
respect to a set of concepts without being conceptually laden with respect to any of those
concepts in particular, so too is this speech act report is, we might say, “locutionarily
demanding” with respect to a set of utterances without definitely representing White as
having performed any locutionary act in particular.
The way that the permissive dispositions view handles the other speech act reports
above is similar: the speaker represents the subject as having performed a locutionary
act of a certain kind (one which served to express certain loaded attitudes) and as having
possessed dispositions which are conceptually free with respect to certain expressions in
the ‘that’-clause of the report and certain associated concepts. While there are certainly
complications to explore here, I take it that this general style of accounting for permissive
speech act reports has the right shape to forestall any general worries of this kind about
the permissive dispositions view.
6.2. Overgeneration
85
Since one may assert a proposition one doesn’t believe, the relevant disposition may be somewhat more
complicated. The dispositions which Black represents White as having had may rather be something like
the disposition to take on a commitment to defend the truth of the proposition that Scarlet wants the
salmon in the condition of learning that Scarlet is in the dining room.
114
An obvious worry with the permissive dispositions view proposed here is that it is
too flexible. Is there any rigorous, systematic way by which we can determine which
dispositional profile is associated with a particular use of an attitude verb? So far, in
each case we have wanted to apply the permissive dispositions view to, we have looked
for, and have been able to find, particular dispositions which can both account for the
permissive or loaded reading of a report as well as accounting for its truth value. But what
entitles us to claim that the dispositions to which we have adverted are indeed included
in the dispositional profiles for these various cases? Permissive dispositions seem easy to
come by, so how is the permissive dispositions view to avoid over-generation problems in a
systematic way?
I think that this question is best approached by thinking about the truth-conditional
theory given by the permissive dispositions view interfaces with the pragmatic question of
under what conditions does a report take a permissive reading. What we need to do is first
to provide a theory of which dispositions in general may be associated with an arbitrary
report sentence. Then we need to develop a theory of what conversational factors influence
whether a use of a report sentence serves to make a permissive report. Using these two
theories, onemightbeabletodevelopageneraltheoreticalpicturethattellsusunderwhat
conversational conditions dispositions of a certain sort can, in context, come to be included
in the dispositional profile associated with a given report sentence. Anyway, that is how I
would develop the permissive dispositions view to address the worry that it is too powerful
to be theoretically informative.
But that is not what I do in the rest of the dissertation. Instead, I want to take some
of the helpful distinctions and observations made in discussing the permissive dispositions
view and apply these insights to to the contextual augmentation view.
115
Chapter 5. Question Sensitivity
In chapter 3, I presented an account of the truth conditions of permissive reports,
the contextual augmentation view. We saw that this view faced problems which suggest
that a full account of permissive reporting has to consider the pragmatics of our reporting
practices. In the remaining chapters of the dissertation, I turn to these concerns. My goals
are twofold.
First, I investigate di↵ erent pragmatic principles that bear on the pragmatic question
of under what conditions a report takes a permissive reading (with respect to certain
occurrences of phrases used in the report and certain concepts).
86
I argue that simple
principles that attempt to state necessary conditions on permissive readings are bound to
fail given the flexibility of our practices of permissive reporting; a fully adequate answer
to the pragmatic question of permissive reports will have to wait for a complete theory of
pragmatic interpretation.
Second, I apply these pragmatic considerations to the issue of which truth-conditional
account of permissiveness to adopt. My goal here is to determine whether, by attending to
the details of the pragmatics of permissive reporting we can address any, many, or all, of
theapparentobjectionsraisedagainstcontextualaugmentationviewsattheendofchapter
3. I take those objections raised against contextual augmentation views to show, at least,
thatanaccountofpermissivereportsstatedjustintermsofcontextualassumptionscannot
be correct. Such views end up being too permissive, as it were, about which reports count
as permissive. But, as I argue in this chapter and the next, these objections don’t tell
against the basic insight behind contextual augmentation views. In this chapter I focus in
particular on the asymmetry problem from §6 of chapter 3. I show how appreciating that
whether a report takes a permissive reading is question-sensitive can help to dissolve the
problem.
With respect to the first, purely theoretical goal of developing a pragmatic account of
permissiveness, in this chapter I argue that we need to take account of conversational ends
inadditiontoconversationalassumptions. Whichreportstakepermissivereadingsdepends
in part on what the conversationalists are trying to do. Following Roberts (2012), I regi-
ment such talk about conversational ends in terms of questions under discussion (QUDs).
Afterconsideringarangeofpermissivereports,itbecomesevidentthatpermissivereadings
are indeed sensitive to QUDs, but the exact ways in which they are sensitive proves di -
cult to express with discrete predictive principles. I show that several tempting pragmatic
principles concerning permissive readings face counter-examples given the complexity of
pragmatic interpretation; I take this as evidence that although whether a report takes a
permissive reading is sensitive to conversational goals, we cannot express how permissive
readings are sensitive to QUDs with any simple principles stating necessary conditions on
86
Answering this question informs can help us determine why permissive reports depend on context. Is
permissivenessamatterofimplicature, ofpresupposition, somekindofpragmaticenrichment, orsomething
else? Given the sensitive attitudes view, I generally favor the pragmatic enrichment option.
116
permissive readings.
In place of adopting simple necessary conditions on permissive readings, I instead favor
a looser picture. According to the sensitive attitudes view, attitude verbs like ‘believes’
are semantically underspecified in such a way that attitude report sentences can be used
to assert di↵ erent assertion candidates, varying in the conceptual requirements they place
on subjects. Given this view, we should expect from the outset that simple pragmatic
principles of the kind examined in this chapter will fail. In the presence of semantic under-
specification, utterance interpretation is a matter of arriving at which of various assertion
candidates is the most plausible for what the speaker intends to commit herself to in
uttering what she does. This process involves attending to a range of di↵ erent factors,
importantly including, as I show here, conversational goals. But the process includes much
more, includingwhatobjectsandrelationsaresalient, whatisassumedaboutthesubject’s
conceptual repertoire, and so on. The very idea that we might isolate any one of these
factors and write down a tidy rule that captures the way that permissive readings depends
on that lone factor is naive. Rather, we should expect that the phenomena tracked by
the di↵ erent pragmatic principles discussed in this chapter and the next balance against
one another in interpretation, together informing what reading a sentence takes in a del-
icate confluence. While I do suggest somewhat more specific generalizations about how
this interpretative process manifests in the case of permissive reporting, these suggestions
are ultimately subordinate to this loose picture. The strongest answer to the pragmatic
question of permissiveness I can commit myself to is just that a use of an attitude report
sentence takes a permissive reading when an assertion candidate involving a permissive
attitude relation is, given the goals and assumptions in the reporting context, the most
plausible candidate for what the speaker means to commit herself to in uttering what she
does.
1. Questions Under Discussion
I begin by proposing a question-relative constraint on permissive reports. I ask the
reader’s patience as I lay out some of the ideas behind this proposal in some detail before
I get to applying the constraint to particular cases. I do this both to ward o↵ confusion
about the nature of the proposal and to frame the reader’s focus in processing through
cases to come throughout the chapter.
The main problem facing contextual augmentation views is that they overgenerate, in
the sense that the minimal constraints they place on which relation is contributed by a
use of an attitude verb yield false predictions in certain cases. To constrain which kinds
of contextual augmentations can ground permissive reports, I propose that we appeal to
other aspects of conversational contexts than merely what the conversationalists assume.
Justasatanypointinaconversationalexchangetherearecertaincommonassumptions
(orpiecesofcommonknowledge)amongtheconversationalists,therearecertaincommonly
adoptedconversationalgoals. Thesegiveconversationalistsreasonsformakingassertionsin
the first place, and, in particular, for ascribing attitudes. In line with this, Grice’s famous
117
Cooperative Principle exhorts one to make one’s addition to a conversational exchange
such as is required given “the accepted purpose or direction of the talk exchange”.
87
On
the proposal Idevelop here, these conversational purposes arecrucial todetermining which
relation some use of an attitude verb contributes to a particular attitude report.
To regiment the talk of conversational goals, I follow Roberts (2012) in understanding
these goals in terms of which questions are under discussion at some point in a conver-
sational exchange. Following Stalnaker (1978)’s theory of assertion, Roberts takes the
primary goal of conversational exchanges to be shared inquiry: the goal of sharing infor-
mation together to find out together “what way the world is”. For Stalnaker, at any point
in a conversation there is a set of possible worlds, called the “context set”, which includes
exactly those worlds which are live possibilities about which way the world is given what
theconversationalistscommonlypresuppose. Theroleofassertion,then,istonarrowdown
these live possibilities: in asserting p, one rules out every world in which ¬p is true as a
live possibility. We may then view Stalnaker’s theory of assertion as saying that the main
goal of any conversation is to answer the question of which world is actual.
Now, this question, which world is actual?, sounds pretty grand as a character-
ization of what ordinary conversationalists are trying to accomplish.
88
Were ordinary
conversationalists asked to describe the purpose or direction of their conversation, they
presumably would not first o↵ er up that they were trying to discover which world is ac-
tual. Roberts (2012), however, generalizes Stalnaker’s picture so that conversations are
governed by more immediate questions, questions like who committed the murder?, or
what should we have for dinner?.
89
The basic role of an assertion on Roberts’s the-
ory, then, is to advance inquiry with respect to whatever questions are under discussion
at some point in a conversation. I take the notion of being under discussion as primitive,
though presumably it may be analyzed in terms of the beliefs, desires, and intentions of
the conversationalists.
90
There are di↵ erent options here for how to model questions for our purposes. First,
we could adopt a semantic model of questions, representing conversation-guiding questions
by whatever entities uses of interrogative sentences are normally used to express, in some
sense. A compelling approach to the semantics of questions pioneered by Hamblin (1973)
is that a question is identified with a set of propositions, understood as the “those proposi-
tions that count as answers to [the question]”. For example, the question who committed
the murder? can be identified with the set of propositions {Scarlet committed the
murder, Plum committed the murder,...}. Generally: {for all relevant individuals in the
87
Grice (1975).
88
Just as I use the type font to indicate propositions, the “semantic value” of indicative sentences, I use
the type font with a question mark to indicate questions understood as semantic objects, the “semantic
values” of interrogative sentences.
89
Here she follows Carlson (1982) in assuming that discourse is structured around a set of questions the
answering of which constitutes the principal goals in a discourse.
90
Another assumption I make is that, following Roberts (2012), multiple questions may be under discus-
sion at a given time. Conversational goals can be complex, involving multiple related lines of inquiry.
118
discourse’s domain, i | i committed the murder}.
It is standard to call the propositions in this set of alternatives the “complete” an-
swers to the question who committed the murder?, since learning any of these alterna-
tives would completely resolve this question. Discovering the truth of one of the complete
answerssuccessfullyterminatesinquiryintotherelevantquestion. Withthisnotionofcom-
pleteanswers,wecan,followingRoberts(2012),definea“partial”answertoaquestionasa
proposition equivalent to a disjunction of complete answers. So, one partial answer to the
question who committed the murder? is the proposition Scarlet or Plum committed
the murder. Discovering the truth of a partial answer to a question advances inquiry into
that question without resolving the inquiry altogether. There is more to say in developing
an adequate semantics for questions, but for our purposes identifying a question with its
set of alternatives should su ce.
91,92
Next I introduce some notation. Since I aim to develop constraints on whether a
given report takes a permissive reading with respect to some phrase in terms of which
questions that report serves to address, it will be helpful to be able to specify a question
in terms of a particular clause. For example, a use of a sentence ‘Scarlet stole the pearls’
can be used at least to answer one of the following questions: who stole the pearls?,
what did Scarlet steal?, what did Scarlet do to the pearls?, who stole what?,
or did Scarlet steal the pearls?.
93
Relatedly, a use of ‘White believes that Scarlet
stole the pearls’ may or may not take permissive readings with respect to various linguistic
constituents of its ‘that’-clause in contexts in which di↵ erent of these questions are under
discussion.
To allow us to characterize di↵ erent questions in terms of the structure of a report’s
‘that’-clause, I introduce an operator, ‘?’, modified from Aloni (2005).
91
Roberts (2012) adopts a version of the semantic representation of the alternatives corresponding to
aquestiono↵ ered by Groenendijk & Stokhof (1989), according to which questions are represented by
partitions of logical space, each element of which corresponds to the worlds in which a corresponding
complete answer is true. This is a helpfully evocative model to bare in mind, though since I am assuming a
viewofpropositionsasstructuredentitiesonwhichdistinctpropositionsmayberepresentationallyidentical,
I won’t assume the partition representation going forward.
92
A second model of questions we might use individuates questions in terms of the linguistic means by
which they are asked. This would be to represent questions in terms of either the interrogative sentences
or the interrogative speech acts used to ask questions (understood semantically). On this way of modeling
questions, forexample, ‘DidScarletcommitthemurder, didMustardcommitthemurder, didPlumcommit
the murder, or did White commit the murder?’, as used in a particular conversational context in which
Scarlet, Mustard, Plum, and White are conversationally salient, asks a distinct question from what a use
of ‘Who committed the murder?’ would ask in the same context. This stands in contrast to certain ways
of individuating questions semantically, according to which the questions asked by uses of these sentences
in such a context may turn out to be the same.
93
In §5belowIdiscusshowtohandlequestionsaboutthesubject’sattitudes, like who does Green think
was blackmailing Scarlet?.
119
?~ x[S]
g
= {8 p|8 ~
d2 D:[S]
g[~ x/
~
d]
=p}.
94
That is, where [S]
g
is the proposition asserted by a use of a clause with logical form
S relative to assignment g,?~ x[S]
g
is the question corresponding to the set of alternatives
obtained by replacing the sequence of variables ~ x in S by sequences of relevant individuals
from the discourse’s domain.
95
For example, ?x[x stole the pearls] is the question corresponding to the set of alterna-
tives {8 p|8 d2 D:[x stole the pearls]
g[x/d]
=p}, namely: {Scarlet stole the pearls,
Plum stole the pearls,...}.
Similarly, ?x[Scarlet stole x] is the question corresponding to the set of alternatives
{8 p|8 d2 D : [Scarlet stole x]
g[x/d]
= p}, namely: {Scarlet stole the pearls,Scarlet
stole the knife,...}.
Similarly, ?xYz[PAST[xYz]] is the question corresponding to the set of alternatives
{8 p|8h d,F,ii2 D :[PAST[xYz]]
g[hx,Y,zi/hd,F,ii]
= p}, or, {Scarlet stole the pearls,
Scarlet stole the knife, Plum stole the pearls, Plum hid the revolver,...} (we
can understand this question as that corresponding to a sincere use of something like ‘who
did what with what?’).
Now, since I am assuming that propositions are structured entities, Aloni (2005)’s orig-
inal definition of ‘?’ cannot straightforwardly be used to characterize yes/no questions,
sometimes called “polar” questions. In Aloni’s intensional semantic system, a question
like did Scarlet steal the pearls? is given by ?x[Scarlet stole the pearls] and deter-
mines the set of alternatives {Scarlet stole the pearls,Scarlet did not steal the
pearls}, with the alternative propositions here understood as sets of worlds. I’m not sure
how to re-write the definition of ‘?’ in a way consistent with the structured propositions
approach which can make sense of polar questions while still characterizing ‘wh’-questions,
like who stole the pearls?, in terms of sets of complete answers.
96
Since this result of
Aloni’s definition is desirable, I’ll simply stipulate that when no variable in sequence ~ x
does not occur freely in S,?~ x[S]= {[S],[¬S]}. Since I’m just using this notation for the
purposes of stating QUD Sensitivity, rather than for proposing some elegant semantics for
questions, I won’t bother with trying to define it in a tidy way.
That, then, is the ‘?’ operator, which I will use in specifying question-sensitive con-
straints on permissive reports.
94
This di↵ ers from Aloni’s own definition, since I am assuming that propositions are structured enti-
ties rather than sets of worlds. Her definition, cast in a standard world-based intensional semantics, is:
?~ x[S]M,w,g = {8 v 2 W|8 ~
d 2 D:[S]
M,v,g[~ x/
~
d]
=[S]
M,w,g[~ x/
~
d]
}. This renders the meaning of a question as
function from a world w to the alternative of the question which is true at w. This naturally gives rise to
a partition semantics like that of Groenendijk & Stokhof (1989).
95
I leave this use of “relevant” at an intuitive level: the knife may be in the discourse’s domain yet fail to
be a relevant individual of the domain in specifying the alternatives in {8 ~
d2 D:[x stole the pearls]
g[~ x,
~
d]
}
96
One option which won’t apparently work is ?~ x[S]g = {8 p|8 ~
d2 D:[S]
g[~ x/
~
d]
=p or [S]
g[~ x/
~
d]
= ¬p},since
this would include a proposition like Scarlet did not steal the pearls as an alternative for who stole
the pearls?.
120
On to the first proposed constraint. Roberts (2012) appeals to questions under discus-
sion (QUDs) in the first instance to explain the conditions in which it is felicitous to make
some assertion. My aims are slightly di↵ erent. Instead of explaining which propositions
are assertible in a context, I use QUDs to explain which propositions are in fact asserted
in a context.
97
Myquestionsensitiveconstraintadoptsthefollowingideas. First,Iassumethesensitive
attitudes view: each attitude verb V is associated with a set [V] of attitude relations,
varying in conceptual permissiveness. Corresponding to this, each attitude report sentence
R is associated with a set of propositions [R], which includes just those propositions that a
normal sincere use of R can be used to assert. Corresponding to the multiple relations in
[V], there are multiple “assertion candidates” in [R]; which assertion candidate is in fact
asserted by a use of R in a conversational context depends on features of that context. On
simple contextual augmentation views, which assertion candidate is asserted depends just
on the conversational assumptions made in the reporting context; my account extends the
factors on which which candidate is asserted depends.
Second, I propose that which candidate in [R] is asserted by a use of R in a context c
depends on at least three factors: (i) which propositions are in [R], (ii) which assumptions
are in CA
c
, and (iii) which questions are under discussion in c. My first pass (which we
will see needs to be revised, supplemented, and rejected for being too simplistic) is:
QUD Sensitivity:
An attitude report sentence of the formaVsthat S(F) serves to make a report
which is permissive in contextc with respect toF inc only if ?x[S(x)] is not under
discussion in c.
I’ll make a few comments about QUD Sensitivity before applying the view to some
examples.
First, QUD Sensitivity isaprincipleabouttheconditionsinwhichanattitudereportis
permissive,notabouttheconditionsinwhichanattitudereportisfelicitousorappropriate.
It is compatible with QUD Sensitivity that there are contexts in which one may appropri-
ately issue a report using a sentence of the formAVsthat S even though ?x[S(x)] is under
discussion in c. Indeed, there are many contexts of this kind. What QUD Sensitivity says
about such cases is simply that the corresponding reports would not be permissive, but
rather would represent the subject as bearing a conceptually loaded attitude towards the
proposition contributed by that S for relevant clause S.Theprincipleisintroducedjust
as a constraint on which reports are permissive.
Second, QUD Sensitivity only states a necessary condition on permissive reports, not
a necessary and su cient condition. There may be a contextc in which a use of a sentence
of the formAVsthat S serves to make a report which is not permissive even though
?x[S(x)] is not under discussion. Indeed, there are many such contexts. The modest idea
97
For a general theory of assertion along these lines, see Schoubye & Stokke (2016).
121
behind QUD Sensitivity is to point to one factor constraining permissive readings, rather
than to provide an exhaustive characterization of all such factors.
Next I begin to apply QUD Sensitivity to some of the problems raised in the previous
chapters. In §2 I will remind the reader of a class of permissive reports which I call
“universal instantiation reports”, then in §3 will rehearse how these examples can be used
to press the asymmetry problems raised in chapter 3. Once this is all presented, I will
show how QUD Sensitivity looks like it helps to avoid those problems. It will not take
long, however, to show that the principle cannot be correct as stated. After my darnedest
to make something like QUD Sensitivity work, I propose a broader, more suggestive view
of how and why permissive reporting is sensitive to QUDs.
2. Universal Instantiation Reports
Recall the permissive report which I presented in chapter 3:
Blackmail: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called in to discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet: Green doesn’t know any name for Scarlet, could not identify her in a line-up,
and so on. By questioning Green, Detective Black learns that Green thinks that every
socialist guest was being blackmailed by Mr. Body. Green doesn’t, however, know the
names of any of the socialist guests: he claims “not to know who the socialist guests are”.
The detectives are exchanging information, trying to find out who had a motive for killing
Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
What Black asserts in uttering (2) is true, even though Green is totally unacquainted
with Scarlet. Given a natural constraint on what it is to possess a concept of an individual,
Blackmail seemstoshowthatsometimesonecantruthfullyascribeanattitudetoasubject
who lacks concepts of constituents of the thought content thereby ascribed to them.
Inthisexample,thesubjectofthereportisknownbythespeakertohaveaconceptually
loaded attitude in a universally quantified proposition. This proposition together with
assumptions made by the conversationalists entails a further proposition which the speaker
reportsthesubjectasbearinganattitudetowards. Butthisreporttakesatruereadingeven
though the subject lacks a loaded attitude in that proposition, hence is permissive. I call
attitude reports of this structure “Universal Instantiation Reports”, or “UI Reports”.
98
In the face of UI reports, one might be tempted by the view presented in chapter 3:
whenever the contents of a subject’s loaded attitudes together with assumptions made
98
It is worth noting that names are inessential: one can get similar UI reports that are permissive with
respect to descriptions and other quantifier phrases. To streamline my discussion I relegate examples which
show this to the appendix to this chapter.
122
within a conversational context entail a proposition p, one can issue a permissive report
in that conversation claiming that the subject bears a permissive attitude towards p.But
modified versions of UI reports show that this simple contextual entailment idea cannot be
correct. I call these modified cases “switched UI reports”.
ExamplesinvolvingpermissiveUIreportsarecharacterizedbyacertainstructure. Each
is an instance of
UI Context Schema:
(i) The conversationalists A and B commonly assume thatNP is F.
(ii) A knows that subjectSVs that every F is G.
(iii) A assertively utters the following to conversationalist B:
(UI)SVs thatNP is G.
To obtain an instance of UI Context Schema, ‘NP’ is replaced with a noun phrase, like
‘Scarlet’, ‘some room with a secret passage’, ‘the cat burglar, etc.; ‘V’ is replaced with an
attitude verb like ‘believe’ or ‘think’; ‘F’ and ‘G’ are replaced with predicate phrases.
To obtain a switched UI report, simply swap the relevant contextual assumption for
the object of the relevant loaded attitude, as in
Switched UI Context Schema:
(i) The conversationalists A and B commonly assume that every F is G.
(ii) A knows that subjectSVs thatNP is F.
(iii) A assertively utters the following to conversationalist B:
(UI)SVs thatNP is G.
When we consider an instance of Switched UI Context Schema corresponding to Black-
mail, we find that it is surprisingly more di cult for a use of (2) to make a true permissive
report. Thistellsagainstthecontextualentailmentview;forifallittookforanutteranceto
makeatruepermissivereportwereforthecontextualassumptionstogetherwiththeloaded
attitudes of the subject to entail the content ascribed by the report, then it shouldn’t mat-
ter whether the universally quantified proposition is among the subject’s loaded attitudes
or among the conversational assumptions. That switching around contextual assumptions
and the subject’s loaded thought contents su ces for a truth-conditional di↵ erence is a
direct counter-example to any contextual entailment view. To see this, consider:
Switched Blackmail: Mr. Body has been murdered; detectives Black and Grey are
investigating the crime. The detectives have private information that every socialist guest
was being blackmailed by Mr. Body. They know that Green is unaware that Body was
blackmailing any guests at all. Having just questioned Green, Black knows that Green
believes that Scarlet (with whom Green is well acquainted) is a socialist. Although the
detectives commonly know that Scarlet is a guest, they do not know that she is a socialist.
They are exchanging information, trying to find out who had a motive for killing Body;
in particular, in this scenario they are trying to figure out who was being blackmailed by
123
Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
It is considerably more di cult to interpret what Black asserts by uttering (2) in
Switched Blackmail as true than it is in the original Blackmail. In the present context,
Black is most naturally interpreted here as making a false, conceptually loaded report,
since it is known that Green is unaware of any blackmail going on.
99
3. Starting to Apply QUD Sensitivity
Whydoesswitchingcontextualassumptionswiththesubject’sloadedthoughtcontents
in UI reports alter the intuitive truth value of (2)? One promising way to explain this felt
di↵ erence between UI reports and switched UI reports makes use of QUD Sensitivity, or
something like it.
Firstlet’srepeatourfirstpassataquestion-sensitiveconstraintonpermissivereadings:
QUD Sensitivity:
An attitude report sentence of the formaVsthat S serves to make a report
which is permissive in contextc with respect toF inc only if ?x[S(x)] is not under
discussion in c.
Again, thisprinciplelaysdownanecessaryconditiononwhetheranutterancecanserve
to make a permissive report. So it will tell us when a report sentence cannot be interpreted
permissively, but not when it can be interpreted so.
An important way in which the original Blackmail context di↵ ers from its switched
counterpart is in what kinds of questions could reasonably be under discussion in the re-
portingcontext. ItisquitenaturalintheoriginalSwitched Blackmail totakethedetectives
to be interested in answering
(Q1) Which guests were being blackmailed by Body?
In Switched Blackmail, the detectives assume that every socialist was blackmailed by
Body and are actively trying to find out what motives one might have for killing Body.
In my statement of the case, I claimed that the detectives “are trying to figure out who
was being blackmailed by Body”. So the question which Grey would ask with (Q1) is
intuitively under discussion. By contrast, the following questions are not under discussion
in Switched Blackmail:
(Q2) Who was blackmailing Scarlet?
(Q3) Scarlet was being blackmailed by whom?
99
In the appendix I show how switched contexts can be used to show that examples of UI reports that are
permissive with respect to descriptions and quantifier phrases similarly fall prey to the asymmetry problem.
I also show how switched contexts can be used to pose the asymmetry problem for permissive readings of
reports that are not UI reports.
124
For the detectives make no assumption that Scarlet was being blackmailed by anyone
at all. It is not a goal of their conversational exchange in Switched Blackmail to resolve
who was blackmailing Scarlet.
Contrast this with the original Blackmail context. (Q1) is not under discussion for the
detectives in Blackmail, since in that context it is Green who believes that every socialist
wasbeingblackmailed,notthedetectives. InBlackmail thedetectivesmakenoassumption
that any guests were being blackmailed, so it would not even be felicitous for Grey to ask a
questionswith(Q1)withoutfurtherado. Withoutfirstintroducingthepresuppositionthat
some were being blackmailed by Body, were Grey to utter (Q1) in Blackmail, Black might
wellrespondinba✏ ement,askingforclarification: ‘Wait,Bodywasblackmailingsomeone?’
So (Q1) is not under discussion in Blackmail. Intuitively in this context, in o↵ ering up
the report of Green’s blackmail-related thought, Black is introducing the possibility of
blackmail as a potential motive, addressing the question which I explicitly said was under
discussion in stipulating Blackmail, who had a motive for killing Body?.
With this contrast in mind, we can begin to apply QUD Sensitivity.In Switched
Blackmail, (Q1) is under discussion. Since the guests are the relevant individuals in the
discourse’s domain, in e↵ ect this is for (Q4) to be under discussion:
(Q4) Who was being blackmailed by Body?
Using our notation, we can represent this question with ?x[x was being blackmailed by
Body]. So according to QUD Sensitivity, (2) should be able to take a permissive reading
withrespecttotheoccurrenceof‘beingblackmailedbyBody’but(2)shouldnot beableto
take a permissive reading with respect to ‘Scarlet’. Since Green is known by the detectives
tobeunawareofScarlet’sexistence,lackinganyconversationallysalientconceptsofScarlet,
QUD Sensitivity predicts that (2) should take a false reading which is loaded with respect
to ‘Scarlet’. This is just as we observed.
Next turn to the original Blackmail, in which (Q1) is not under discussion. In that
context the only question which I suggested was under discussion among the detectives
(who again, do not assume that anyone was being blackmailed), was what Grey might ask
by uttering
(Q5) Who had a motive for killing Body?
First, neither (Q4) nor (Q1) are under discussion in Blackmail, since for either to be
under discussion would require that the detectives commonly assume that someone was
(or may have been) blackmailed by Body, which they don’t in the original Blackmail.So
according to QUD Sensitivity, (2) should be able to take a permissive reading with respect
to ‘Scarlet’, as we observe it does. So far, then QUD Sensitivity serves to account for the
contrast raised by the asymmetry problem.
4. A First Problem for QUD Sensitivity
QUD Sensitivity has some promise, but we don’t have to look far to find cases that
125
suggest that this particular way of stating a question-sensitive constraint on permissive
readings cannot be the full story. In this section I present some variations on the cases
considered above that undermine QUD Sensitivity as stated. In the next section I make
various observations about how permissive reports behave in asymmetry cases to build
towards a revision of QUD Sensitivity.
If we add further details to the switched context Switched Blackmail, we find that (2)
can more easily be used to issue a true permissive report. Consider first a modified version
of Switched Blackmail in which multiple parties are assumed to have been blackmailing
guests.
Switched Blackmail
2
: The detectives have private information both that every socialist
was being blackmailed by Mr. Body and that every adulterer was being blackmailed by
Ms. Mind. The detectives know that Scarlet was being blackmailed by someone, but are
unsure of whether she was being blackmailed by Body or Mind. As in Switched Blackmail,
detective Black questions Green and learns that he thinks that Scarlet (with whom he is
acquainted) is a socialist, though Green is known to be unaware that any blackmailing
was going on. The detectives are exchanging information, trying to figure out who was
blackmailed by whom. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
Here(2)iseasilyinterpretedasatruepermissivereport. ButasintheoriginalSwitched
Blackmail, it is the detectives who assume the relevant universally quantified proposition.
So again, the asymmetry observed in §2 does not just depend on whether the universally
quantified proposition is a contextual assumption or an object of the subject’s loaded
attitudes.
Other modifications of the original Switched Blackmail context lead to a similar result.
Consider
Switched Blackmail
3
: The detectives have private information that every socialist was
being blackmailed by Mr. Body. They also know that sometime last night Scarlet caught
a flight to South America. As in Switched Blackmail, detective Black questions Green
and learns that he thinks that Scarlet (with whom he is acquainted) is a socialist, though
Green is known to be unaware that any blackmailing was going on. The detectives are
exchanging information, trying to figure out why Scarlet might have absconded to South
America. Black assertively utters to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
To my ear, here (2) also can be interpreted as making a true permissive report. This is
somewhat surprising, since the main thing which has changed from the original Switched
Blackmail context is just the addition of an extra contextual assumption about Scarlet.
126
But, looking ahead, we can say that since this assumption makes room for a new question
tobenaturally under discussion, Why did Scarlet abscond?, Black’s useof (2) can helpfully
contribute to the conversation without having to issue a loaded report.
100
These cases are at least interesting in that they provide further evidence that whether
a report can take a permissive reading does not just depend on which propositions are
contextually assumed and which are the objects of the subject’s loaded attitudes. The
original Switched Blackmail case matches the variations on that case considered in this
chapter in terms of what is contextually assumed and what Green’s loaded attitude are,
yet in the original Switched Blackmail case (2) takes a false reading that is loaded with
respect to ‘blackmailed by Mr. Body’, whereas it takes a true permissive reading in the
cases considered in this section.
But this observation apparently refutes QUD Sensitivity as stated. While Switched
Blackmail
2
and Switched Blackmail
3
suggest that changing the question under discussion
may be relevant to whether a given report can take a given permissive reading, Switched
Blackmail
2
serves as a counter-examples to our going version of QUD Sensitivity.
In Switched Blackmail
2
, recall, the detectives are trying to resolve a question asked by
something like:
(Q6) Who was blackmailed by Body and who was blackmailed by Mind?
(Q7) Was Scarlet blackmailed by Body or was she blackmailed by Mind?
The judgment in this case was that with (2) Black can issue a true report that is per-
missive with respect to ‘being blackmailed by Mr. Body’ even though Green is unaware
that there is any blackmailing going on. To use QUD Sensitivity to account for this obser-
vation, it seems at first su cient to point out that in this conversational context ?x[Scarlet
was x] is not a question under discussion; (2) instead serves to address (Q7). But notice
that (Q7), in this context, could as well be asked by Grey with ‘Who was blackmailing
Scarlet?’, which we can represent in our notation as ?x[Scarlet was blackmailed by x]. So
according to QUD Sensitivity, (2) should take a loaded reading with respect to ‘Body’,
whom Green does not cognize via a concept Body in holding his fully loaded thought in
this case. We do not detect this loaded reading in this case, which would intuitively result
in a false reading of (2) were it present. So the letter of QUD Sensitivity makes a false
prediction in this case.
Which questions are under discussion in the reporting context clearly seem somehow
relevant to whether a report can take certain permissive readings, but the particular way
that I have stated QUD Sensitivity is false. In the next section, I build towards a refine-
ment of the principle that might handle this problem. By drawing from some theoretical
observations about questions we develop a question-sensitive principle that avoids some
100
In the appendix I show that by similarly complicating switched contexts corresponding to other exam-
ples of permissive reports, involving descriptions and other quantifier phrases, we see similar results.
127
problems. But in time we will see that even such refinements cannot be the whole story.
This will lead me to sketch a more suggestive picture of what is going on and to use
this picture to argue that our practice of interpreting permissive reports inherently resists
any simple characterization in simple pragmatic principles stated just in terms of discrete
features of conversational contexts.
5. Refining and Rejecting QUD Sensitivity
I start by making an observation that has previously been made in the linguistics
literature, that it is, in general, more di cult to interpret a report as permissive with
respect to the main verb phrase of the ‘that’-clause of a report sentence. I then present
some common assumptions made in the linguistics literature about connections between
questionsandfocus,whichhelpstomakesomesenseoftheconnectionbetweenverbphrases
and loaded readings. This discussion leads to a question-sensitive principle that is better,
but which is still not good enough. Just as there is no tidy semantic rule that serves to
specify the truth conditions of permissive reports, there appears to be no tidy pragmatic
rulesthatservetospecifytheconversationalconditionsinwhichareporttakesapermissive
reading.
5.1. Verb Phrase Sensitivity
Whenthetargetphrase‘blackmailed’in(2)occursintheverbphrase ofthecomplement
clause rather than in its subject, a permissive report becomes immediately more accessible
in Switched Blackmail. Observe how in that context there is a contrast in how we naturally
interprettheoriginal(2)andhowweinterpretthesimilarsentence(2V)inwhichthetarget
material involving ‘blackmailed’ is fronted in the report’s complement’s main noun phrase:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
(2V) Green thinks that someone Mr. Body was blackmailing is Scarlet.
Recall that in Switched Blackmail the detectives have private information that every
socialist was being blackmailed by Body, and Black knows that Green both is acquainted
with Scarlet and thinks she is a socialist. Green, again, is unaware that anyone was being
blackmailed, so if either of (2) or (2V) can be used to make a true report, that report
must be permissive with respect to ‘blackmailed’. And for whatever reason, (2V) can be
readily interpreted as not representing Green as thinking “in terms of” blackmail; (2V)
more readily takes a permissive reading with respect to ‘blackmail’ in Switched Blackmail.
This observation itself already counts in favor of QUD Sensitivity or something like
it. For (2) and (2V) intuitively di↵ er in how they relate to the conversational goals in
Switched Blackmail. In that context, the detectives are trying to answer a question which
Grey might ask by uttering one of:
(Q4) Who was being blackmailed by Mr. Body?
(Q1) Which guests were being blackmailed by Mr. Body?
128
The detectives assume that every socialist is being blackmailed and are actively trying
to find a motive one might have for murdering Body. So resolving either of (Q4) or (Q1)
directly would help to resolve another of their shared goals, given what they commonly
know.
In response to these questions, (2V) sounds quite natural, even given the background
assumption that Green is unaware that there was any blackmail going on. By contrast,
(2), for reasons to be explained, sounds slightly less natural in this case. It isn’t terrible
as a response to (Q4) or (Q1), by any means, but it’s touchy. I probe this phenomenon in
more depth below.
Another thing worth observing about (2) and (2V) is that their intuitive truth values
flip when we evaluate them in the original Blackmail context. In that context, recall, it
is Green who believes that every socialist was being blackmailed and it is the detectives
who assume that Scarlet is socialist. So if (2) or (2V) can serve to make true reports in
Blackmail, they must be permissive with respect to ‘Scarlet’. Now, as before, we observed
that (2) does indeed serve to make a permissive report in Blackmail. But with that context
in place, it also is true that Black’s assertion made by (2V) is most naturally interpreted
as a false loaded report rather than as a true permissive report. That ‘Scarlet’ occurs in
the verb phrase makes it sound as though Black represents Green as deploying a concept
of Scarlet in thinking what he is reported to think. Accordingly, a natural response from
Grey to (2V) in Blackmail would be to ask for clarification: “wait, Green is familiar with
Scarlet? I thought he was unaware of her existence!”
Despite this contrast between (2) and (2V), still the complement clause of (2) is true
if and only if the complement clause of (2) is.
101
Therefore, by the way, this contrast
constitutes further evidence that we cannot state the conditions under which reports can
be permissive just in terms of what the subject’s loaded attitudes contextually entail, as
the contextual augmentation view suggests. For were that so, it would follow that in both
Blackmail and Switched Blackmail (2) and (2V) should both or neither take permissive
readings, contrary to observed fact. This contrast also shows us that we cannot account
for the variation in accessibility of permissive reports just by making the initially tempt-
ing prediction that permissive UI reports are possible only when the relevant universally
quantified claim is a loaded thought content of the subject rather than a contextual as-
sumption.
102
This observation about verb phrases already seems to provide part of an explanation
of the interpretive asymmetry observed in §2: it is not that permissive reports are possible
only when the subject bears a loaded attitude towards a universally quantified proposition.
101
In this case, this requires us to ignore the past tense in ‘was being blackmailed’; to make this point
we could as easily use a version of the case in which the report sentences are ‘Green thinks that Scarlet is
being blackmailed by Mr. Body’ and ‘Green thinks that someone Mr. Body is blackmailing is Scarlet.’
102
In the appendix I illustrate that the observations I’ve made here about the connection between verb
phrases and loaded readings do not depend just on peculiarities about the Switched Blackmail case. They
hold in the other switched cases as well.
129
Rather, what the switched contexts appear to do is render certain sentences as suboptimal
ways of issuing a permissive report, given the kinds of questions which plausibly are or
could reasonably be under discussion in the switched context.
Now, although something like QUD Sensitivity may be able to explain asymmetry
pairs, the observations made in this section shows that it may not be the only possible
explanation. Instead, one might try to explain this section’s observation by claiming that
there is a universal prohibition against attitude verbs being permissive with respect to
material in the main verb phrase of the complement of an attitude reporting sentence.
103
However, whileIdothinkthatthisisarobusttrendandgoessomewaytowardsexplaining
what is going on in the foregoing examples, insisting upon such a universal prohibition
would go too far.
To see as much, we need only reconsider our problem case, Switched Blackmail
2
. For in
that context, in which the detectives are concerned with answering the question of whether
ScarletwasblackmailedbyBodyorbyMind, wedetectedatruereadingof(2)withrespect
to the verb phrase ‘blackmailed by Mr. Body’. So whether a report takes a permissive
readingwithrespecttosomelinguisticmaterial(andaconversationallyassociatedconcept)
does not depend on whether that linguistic material occurs in the verb phrase. This is
enough to see that the suggestion that a report cannot be permissive with respect to verb
phrase constituents is too simple to be correct.
With these observations about verb phrases on board, next I present some widely
held linguistic assumptions that help to account for these observations. These linguistic
assumptions will play a central role in my proposed revision of QUD Sensitivity.
5.2. Focus, Questions, and Verb Phrases
Thesimpleobservationfrom §5.1isthatwhetherareportcantakeapermissivereading
with respect to some phrase is often sensitive to whether that phrase occurs in the report
sentence’s complement’s verb phrase. By attending to standard assumptions made in the
literature on questions and focus, we can start to see how something like QUD Sensitivity
can make sense of the foregoing observations about verb phrases, and, more importantly
for our purposes, make sense of what is going on in Switched Blackmail
2
.
The basic idea behind using something like QUD Sensitivity to make sense of the con-
trast between (2) and (2V) in Switched Blackmail is that a use of a sentence with a certain
structure can, in certain contexts, pragmatically suggest that the speaker intends to ad-
dress a question which that sentence, given its structure, is best suited to answer. Which
question a sentence intuitively serves to directly address in a context can be influenced
by that sentence’s structure. If a sentence could in principle be used to answer di↵ erent
questions but, given its structure, is most felicitously used to address a particular one of
those questions, Q, then, all else being equal, the conversationalists will assume that the
103
This is Percus (2000)’s approach. He proposes a constraint on transparent readings of noun phrases,
according to which, roughly, noun phrases constituents of a sentence’s main predicate cannot be interpreted
as transparent.
130
sentence is being used to addressQ. As long asQ is not inconsistent with standing conver-
sational assumptions, accords with standing conversational goals, and does not presuppose
anything conversationally implausible, Q is accommodated as being under discussion.
By appealing to some standard linguistic assumptions, we can say that (2) and (2V),
given their di↵ erent structures, are best suited to answer di↵ erent questions, hence in
Blackmail and Switched Blackmail serve to introduce as under discussion and address dif-
ferent questions. With this, we apply something like QUD Sensitivity to generate di↵ erent
predictions of which permissive readings are available for these report sentences in these
di↵ erent contexts.
Here are three standard assumptions found in the literature on focus and questions.
First, it is standard to assume that any sentence, as used on some occasion, has some
constituent which is focused, even if it is the whole sentence.
104
Second, it is standard to assume that the focus of a sentence, as used in some context,
is intimately related to what question the use of that sentence serves to directly address.
Specifically, for a sentence S(F) in which a linguistic constituent F occurs, if on some use
of S(F), F is focused, then that that use of the sentence introduces a presupposition that
the question ?X[S(X)] is under discussion.
105
For example, assertively uttering ‘Scarlet is
in the lounge’ carries a presupposition that the question ?x[x is in the lounge], asked by
‘Who is in the lounge?’, is under discussion; assertively uttering ‘Scarlet is in the lounge’
carries a presupposition that the question ?x[Scarlet is in x], asked by ‘Where is Scarlet?’
or ‘What room is Scarlet in?’, is under discussion.
106
If the presupposed question is not
obviouslyunderdiscussionpriortotheuseofS(F),thenitisaccommodatedasbeingunder
discussion.
107
So, what question a sentence serves to directly address may be distinct from
the questions which antecedently are established as under discussion in the context. The
notionof“underdiscussion”here, then, istobeunderstoodassomewhatflexible; thebasic
thought is that focusing di↵ erent constituents of a sentence pragmatically indicates that
one means to address a particular question with one’s remark.
Third, in English, the end of a clause (typically the clause’s verb phrase) is typically
focused, unless another part of the larger clause enjoys special intonational prominence.
This phenomenon is sometimes called “end-focus”.
108
For example, Greenbaum & Nelson
(2015) provide the following pair:
(64) On guard stood a man with a gun in each hand.
(65) A man with a gun in each hand stood on guard.
104
See Gundel & Fretheim (2004) and Roberts (2012, 2018).
105
Roberts (2012). A similar principle is argued for by Geurts & van der Sandt (2004): they argue that
whenever focusing gives rise to a “background” x. (x), there is a presupposition to the e↵ ect that x. (x)
holds of some individual.
106
Here and below I use italics to indicate intonational prominence.
107
Roberts (2012).
108
Nelson & Greenbaum (2015).
131
The italicized parts of (64) and (65) are understood as typically focused given their
end-position. Absent special pragmatic factors, the verb phrase of a clause is typically
focused.
Putting these three ideas together, we predict that a use of a clause of the form [[d
F]
NP
[is G]
VP
] typically focuses the verb phrase is G, so, given QUD Sensitivity, typically
serves to address the question ?X [[d F] [X]]. For example: a use of ‘Green smokes’ will
typically serve to address the question ?X [Green X] (asked by ‘What does Green do?’ or
‘What is Green like?’); by contrast, a use of ‘One smoker is Green’ will typically serve to
address the question ?x[One smoker is x] (asked by ‘Who is one smoker?’). Similarly, a
use of ‘Green is in the dining room’ will typically serve to address the question ?x[Green
is in x] (asked by ‘Where is Green?’ or ‘Which room is Green in?’); by contrast, a use of
‘One person in the dining room is Green’ will typically serve to address the question ?x
[One person in the dining room is x] (asked by ‘Who is one person in the dining room?’
or perhaps ‘Who is in the dining room?’). Of course, ‘Green is in the dining room’ can be
used to answer a question asked by ‘Who is in the dining room?’; the point here is that,
absent interfering extra pragmatic factors, this sentence will tend to serve to address a
question like ‘Where is Green?’.
Note that these principles seem to hold in some cases when a clause is embedded in a
larger sentence.
(66) Scarlet is in the lounge and Green is in the kitchen.
(67) If Scarlet is in the lounge then Green is in the kitchen.
(68) It’s possible that Scarlet is in the lounge.
(69) Green believes that Scarlet is in the lounge.
In these examples, intuitively the focus induced by intonational prominence can serve
either to introduce a presupposition that a question corresponding to the whole sentence
is under discussion or to introduce a presupposition that a question corresponding just
to the embedded clause is under discussion. Take (66): depending on the conversational
scenario, this may serve to address the question ?x[x is the lounge and Green is in the
kitchen] or the question ?x[x is in the lounge]. Similarly, (67) can serve just to address the
question ?x[Green is in x] or instead the question ?x[If Scarlet is in the lounge then Green
is in x] (asked by ‘Where is Green if Scarlet is in the kitchen?’). More relevant to our
purposes is that (68) and (69) can serve to directly address just a question corresponding
to the embedded ‘that’ clause. While (68) can serve to directly address the question ?x[It’s
possible that x is in the lounge], it is as natural as providing a tentative answer to the
simple question ?x[x is in the lounge]. Similarly, (69) can serve to directly address the
question ?x[Green believes that x is in the lounge] (asked by ‘Who does Green believe is in
the lounge?’), but it can also serve to directly address the question ?x[x is in the lounge].
When it does, QUD Sensitivity tells us that the report made takes a loaded reading with
respect to ‘Scarlet’.
132
Now then we can begin to apply these considerations about focus to the observed
di↵ erence between report sentences such as (2) and (2V):
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
(2V) Green thinks that someone Mr. Body was blackmailing is Scarlet.
Given the phenomenon of end-focus, we should expect that ‘being blackmailed by Mr.
Body’ or ‘Mr. Body’ is focused in (2) and that ‘Scarlet’ is focused in (2V). So, (2) can
serve to address the question ?x[Scarlet was x] or ?x[Scarlet was being blackmailed by
x]. According to QUD Sensitivity, then, this report should take a loaded reading with
respect to ‘being blackmailed by Mr. Body’ or ‘Mr. Body’ (and relevant conversationally
associatedconcepts). Bycontrast,QUD Sensitivity predictsthat(2V)shouldtakealoaded
reading with respect to ‘Scarlet’, since the report most naturally is used to address the
question ?x[Mr. Body was blackmailing x].
These predictions help us account for the di↵ erent readings these report sentences tend
to take in conversational scenarios like Blackmail and Switched Blackmail.In Blackmail,
it is conversationally assumed that Green is unacquainted with Scarlet via any relevant
concept. So (2) can take a permissive reading with respect to Scarlet (and a relevant
conversationally associated concept), so may take a true reading. But since (2V) has to
takealoadedreadingwithrespectto‘Scarlet’,thisreporttakesafalsereadinginBlackmail.
In Switched Blackmail, by contrast, it is conversationally assumed that Green is relevantly
acquainted with Scarlet but is unaware that anyone was being blackmailed by Body, hence
is not disposed to deploy a concept of blackmailed-by-Body. So (2V), which may take a
permissive reading with respect to ‘someone Mr. Body was blackmailing’, can take a true
reading in Switched Blackmail according to QUD Sensitivity. But (2), which is predicted
to take a loaded reading with respect to ‘being blackmailed by Mr. Body’, is accordingly
predicted to take a false reading in Switched Blackmail.
To emphasize just how directly this phenomenon tells against simple contextual entail-
ment views, I want here to draw the reader’s attention to another pair of this kind.
109
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
(2V’) Green thinks that Mr. Body was blackmailing Scarlet.
Tomyear,thispairexhibitsthesamecontrastsexhibitedby(2)and(2V).InBlackmail,
(2) takes a permissive reading with respect to ‘Scarlet’ and is true, whereas (2V’) takes
a loaded reading with respect to ‘Scarlet’ and is false. In Switched Blackmail, (2) takes a
loaded reading with respect to ‘Scarlet’ and is false, whereas (2V’) can take a permissive
reading with respect to ‘Scarlet’ and so read true. Since the complements of (2) and
(2V’) are clearly equivalent, this pair is a particularly telling counter-example to simple
109
I present more pairs of sentences that support the observations just made with (2) and (2V) in the
appendix.
133
contextual entailment views.
Of course, end-focus is defeasible. When linguistic material at the front of a clause
is uttered with special intonational prominence, it can be focused rather than the end of
the clause. So QUD Sensitivity predicts that the contrast observed with (2) and (2V) is
likewise defeasible. These predictions seem to be borne out. Consider uses of the following
in contexts like Blackmail and Switched Blackmail:
(2F) Green thinks that Scarlet was being blackmailed by Mr. Body.
(2FV) Green thinks that someone Mr. Body was blackmailing is Scarlet.
Were Black to assertively utter (2F) in Blackmail, in which the detectives are trying
to find out who had a motive for killing Body, it would be natural for Grey to respond in
confusion: ‘Wait, I thought Green was unacquainted with Scarlet?’. This reflects that (2F)
naturally takes a loaded reading with respect to ‘Scarlet’, which can be explained by the
focus of ‘Scarlet’ introducing a presupposition that ?x[x was being blackmailed by Body]
is under discussion. To utter (2FV) in Blackmail, by contrast, is less prone to take a false,
loadedreadingwithrespectto‘Scarlet’. Thisisn’ttosaythatitwouldbeexactlyfelicitous
for Black to assert in that context; presumably Grey could respond in confusion: ‘Wait,
was someone other than Body also blackmailing guests?’. But this does not prevent the
report from taking a true, permissive reading with respect to ‘Scarlet’.
In Switched Blackmail things once again are reversed. In that context, recall, the
detectives are trying to figure out who was blackmailed by Body and assume that Green
is unaware of any blackmailing. It is easier for (2F) to take a true, permissive reading
with respect to ‘being blackmailed by Mr. Body’, given that ‘Scarlet’ is focused. And it
is easier for (2FV) to take a false, loaded reading with respect to ‘someone Mr. Body was
blackmailing’, given the focus on ‘Mr. Body’.
Having made these observations about the connection between focus, questions, and
verb phrases, I first propose the following refinement of QUD Sensitivity:
Direct QUD Sensitivity:
An attitude report sentence R of the form A Vsthat S(F) serves to make a
report which is permissive in context c with respect to F in c only if ?x[S(x)] is
not directly addressed by R in c.
This principle di↵ ers from the original QUD Sensitivity in replacing the notion of a
question’s being “under discussion” in a context with a question’s being “directly ad-
dressed” in a context. Sometimes a question Q may not be antecedently under discussion
in a context, yet, given a sentence’s structure, the intonation with which it is used in a
context, and how Q relates to questions already under discussion, that sentence may serve
134
to directly address Q.
110
Let’s contrast Direct QUD Sensitivity with the original QUD Sensitivity as they apply
to uses of (2) and (2V) in Blackmail. We observed that in that context, (2) naturally
takes a true reading that is permissive with respect to ‘Scarlet’ but (2V) naturally takes a
false reading that is loaded with respect to ‘Scarlet’. Naively applying the original QUD
Sensitivity, we would fail to detect this di↵ erence. For in the context of Blackmail the
question antecedently under discussion is something like Who had a motive for killing
Body?, or What is Scarlet’s motive?, or Scarlet was what?, in which case, both (2)
and (2V) should take permissive readings with respect to ‘Scarlet’ (and a conversationally
associated concept) in such a context. But Direct QUD Sensitivity, together with the
previously discussed linguistic assumptions about end-focus, allows that the structural
di↵ erencebetween(2)and(2V)caninfluencewhichquestioneachservestodirectlyaddress.
In particular, assuming that there are no intervening pragmatic factors in the case, a
use of (2V) typically will lay focus on ‘Scarlet’, thereby pragmatically suggesting that,
contrary to antecedent contextual priming, it serves to directly address the question Who
was Body blackmailing?, whereas there is no such pragmatic suggestion with (2), given
its structure. So Direct QUD Sensitivity successfully predicts that (2V) takes a loaded
readingwithrespectto‘Scarlet’, whereasitpredictsthat(2)maytakeapermissivereading
with respect to ‘Scarlet’, accounting for the detected truth-conditional di↵ erence between
the two, given the conversational assumptions in the context, Green does not possess or
deploy any concept conversationally associated with Scarlet.
There seems to be something right about Direct QUD Sensitivity,butinthenext
section I present other classes of permissive reports that show that this principle cannot be
the complete story about how and why permissive reporting is sensitive to conversational
goals. The examples discussed in the next section all concern ways in which conversational
goals can influence whether a permissive reading of a report sentence is most plausible, but
none can be accounted for by any straightforward revision of Direct QUD Sensitivity.In
light of such examples, I suggest instead that the rough generalization expressed by this
principle reflects but one facet of a broader pragmatic practice involved in interpreting
attitude reports.
5.3. Motivating a Broader Pragmatic Picture
AlthoughIultimatelyrejectQUDSensitivity,thereissomethingrightaboutit. Whatis
it? Why should anything like Direct QUD Sensitivity accurately constrain the distribution
of permissive readings?
110
As reflected in some of the discussion above, one might, with Roberts (2012), say that in such a case
Q becomes under discussion via accommodation. But to make explicit the importance of the distinction
between a question’s being antecedentlyunder discussion in a context and itsbeing accommodated as being
under discussion, and also not to do violence to our ordinary notion of what it is for something to be “under
discussion”, I choose to state an additional principle using the lingo of what question a report sentence
serves to “directly address”.
135
One tempting answer is this: to introduce content that is conversationally novel by
embedding it within an attitude report pragmatically conveys that this novel content is
provided, in some sense, by the fully loaded thought life of the subject whose attitude is
being reported. So, against a conversational scenario characterized by a question under
discussion such as ?x[x is in the ballroom], asked by ‘Who is in the ballroom?’, when Black
assertively utters ‘Green thinks that Scarlet is in the ballroom’, the embedded ‘Scarlet’
is conversationally novel, in contrast with the rest of the report’s complement, so she
pragmatically conveys that Green has a thought in which a concept of Scarlet essentially
figures. Conversely, since across the set of complete answers to this question, {Scarlet
is in the ballroom, Plum is in the ballroom, White is in the ballroom,...},the
content in the ballroom is shared in common, so is not conversationally novel, Black
need not, depending on the scenario, represent Green as possessing or deploying a concept
of the ballroom. The content (the properties, relations, and objects) shared in common as
constituents of the complete answers to the question being addressed by an attitude report
are, we can say, among the things under discussion, the things concepts of which the
conversationalists are already using in conceiving of the issue being discussed, but which
the subject of the report may or may not use in conceiving of any particular issue. This
picture suggests one way of articulating the guiding idea behind Direct QUD Sensitivity:
a report is permissive with respect to a linguistic constituent F of an attitude report S(F)
only if the thing contributed by F to the report proposition expressed is not among the
things under discussion in the reporting context.
This goes part of the way towards a general statement of why permissive readings are
sensitive to QUDs, but as we’ll see, even this perspective is too myopic. We can situate the
idea in a yet broader pragmatic picture of how and why permissive readings are sensitive
to discourse goals that does better.
Why should an object or property’s being under discussion pragmatically suggest that
a report takes a permissive reading with respect to linguistic material that designates that
object or property? Speaking quite generally, whether a report takes a permissive reading
withrespecttoaphraseF (andaconversationallyassociatedconcept )intuitivelydepends
on whether in using F the speaker intends to represent the subject as using in certain
of her fully loaded attitudes. In certain circumstances in which what F designates is not
under discussion, a speaker’s introducing F in the complement of an attitude report can
suggest that the reason the speaker used this linguistic material is to represent the subject
as conceiving of certain things via . But in some contexts in which what F designates is
antecedently under discussion, there may be other pragmatically available explanations for
why the speaker would use F in the complement of an attitude report sentence; they may
use F just to draw the hearer’s attention to certain objects or properties already being
discussed, which the subject’s fully loaded thoughts happen to pertain to in some way.
In reporting contexts, the hearer can generally expect that there has to be some reason
why the speaker uses the words she does in making a report. Given the idea that attitude
verbs are semantically underspecified in such a way that they can designate conceptually
136
permissive attitude relations, there is a possibility that the speaker uses certain words to
indicate that she intends to ascribe an attitude relation that is loaded with respect to
what is designated by those words. But in a reporting context there is also the possibility
that she instead is using those words for some other reason and intends to ascribe an
attitude relation that is permissive with respect to what is designated by those words.
On the sensitive attitudes view, there are many designation candidates for any particular
attitude verb; one mechanism a speaker can exploit to convey which candidate she intends
to designate is to use certain words (or to put words in a certain order) in issuing her
report. Which words carry what pragmatic significance will depend in part on questions
are under discussion and what things are under discussion. This much explains the limited
success Direct QUD Sensitivity enjoys as a pragmatic constraint on permissive readings.
A yet broader pragmatic picture, then, is that a report takes a loaded reading with
respect to some phrase and an associated concept if, given the assumptions and goals
of the conversation, the most plausible reason that the speaker uses that phrase is to
represent the subject as using an associated concept in holding certain of their fully loaded
attitudes. In cases where there are other more plausible reasons as to why the speaker
would use some phrase, di↵ erent conceptually permissive attitude relations may be favored
as the designated candidate of the attitude verb used in a report. This more general
picture suggests that even in conversational scenarios in which one uses a report sentence
of the formAVsthatS(F) to directly address the question ?x[S(x)], the most plausible
reason that the speaker uses F may not be to represent A as using any particular concept
associated with F in her fully loaded thoughts. Below I present cases of this kind, which
discredit Direct QUD Sensitivity and support our more expansive sketch of how and why
permissive readings are sensitive to conversational goals.
5.3.1. Multiple Choice Questions and Answers Under Discussion
The first class of examples I consider build from the problem raised by Switched Black-
mail
2
. Itakethisexampleandacertainclassofexamplesofwhichitisaninstanceasdirect
evidence against Direct QUD Sensitivity. That principle is best construed as expressing
just part of what is going on with the pragmatics of permissive reporting.
I begin with examples that are like Switched Blackmail
2
in that either the context’s
QUD or certain contextual assumptions make certain alternatives salient. I call them
“multiple choice questions”. What we find by considering di↵ erent kinds of multiple choice
questions is that introducing alternatives that correspond to certain linguistic constituents
of the report’s complement induces a permissive reading with respect to ‘blackmailed by
Mr. Body’ in a way that allows that phrase’s sub-constituents to inherit this permissive
reading. By contrast, introducing alternatives that don’t correspond to certain linguistic
constituents of the report’s complement fails to induce this kind of permissive reading.
Each of these cases is like Switched Blackmail in that the detectives assume that every
socialist was being blackmailed by Body and in that Black is aware, from interrogating
Green, that Green thinks (in a fully loaded way) that Scarlet is a socialist.
137
Switched Blackmail
3
: The detectives have private information that every socialist was
being blackmailed by Mr. Body. They further know that either Plum or Scarlet was
being blackmailed by Body. Detective Black questions Green and learns that he thinks
that Scarlet (with whom he is acquainted) is a socialist, though the detectives know that
Green is unaware that any blackmailing was going on. The detectives are exchanging
information, trying to figure out whether Plum or Scarlet was being blackmailed by Body.
Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
To my ear, in this context (2) more easily takes a permissive reading with respect to
‘blackmailed by Mr. Body.’. If that’s right, I think that Direct QUD Sensitivity alone can
explain as much. The QUD in this context is was Plum being blackmailed by Body
or was Scarlet being blackmailed by Body?, which, in the context, given the detec-
tives’ contextual assumption that either Plum or Scarlet was being blackmailed by Body,
is equivalent to who was being blackmailed by Body?. Given the connection between
focus and questions under discussion observed above in §5.2, it is natural for us to expect
‘Scarlet’ to be focused in Black’s use of (2) in this context, which is borne out: it is most
natural to read Black’s report as focusing ‘Scarlet’. Since it is ?x[x was being blackmailed
by Body], and not ?x[Scarlet was x] that is directly addressed by (2) in this context, we
should expect on Direct QUD Sensitivity that (2) takes a permissive reading with respect
to ‘blackmailed by Mr. Body’ in this context, which is what we seem to observe.
Let’s consider some other variations:
Switched Blackmail
4
: The detectives have private information that every socialist was
being blackmailed by Body, every fascist was being bribed by Body, and every anarchist
was being framed by Body. Detective Black questions Green and learns that he thinks
that Scarlet (with whom he is acquainted) is a socialist, though the detectives know that
Green is unaware that any nefarious doings were being done: he is unaware of any black-
mailing, bribing, or framing. The detectives are exchanging information, trying to figure
out whether Scarlet was being blackmailed, bribed, or framed by Body. Black assertively
utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
In this context, (2) intuitively takes a permissive reading with respect to ‘blackmailed
by Mr. Body’. This raises a problem for Direct QUD Sensitivity similar to that raised
by Switched Blackmail
2
. In this context (2) is used to directly address whether Scarlet
was being blackmailed, bribed, or framed by Body, which in the context is equivalent to
?x[Scarlet was being x by Mr. Body], which one might ask with ‘what was Body doing to
Scarlet?’. Giventhis,accordingtoDirect QUD Sensitivity,(2)shouldtakealoadedreading
with respect to ‘blackmailed’ in Switched Blackmail
4
, contrary to what we observe.
138
AlthoughDirect QUD Sensitivity deliversthewrongresultsinthiscaseandinSwitched
Blackmail
2
,thebroaderpragmaticpicture-thinkingsketchedabovecansaysomethinghere.
FirstnotethatinbothSwitched Blackmail
2
andSwitched Blackmail
4
thequestionsthat
(2) is used to directly address intuitively are “multiple choice” questions. The question
in Switched Blackmail
4
, for example, is not in fact best represented just as ?x[Scarlet was
being x by Mr. Body], per se. Only against a background in which it is obvious that the
conversationalistsarejustfocusingonthepossibleanswers Scarlet was blackmailed by
Body, Scarlet was bribed by Body, and Scarlet was framed by Body, can we under-
stand the interrogation ‘what was Body doing to Scarlet?’ as giving voice to what question
Black addresses with (2). We can represent this question as ?x[Scarlet was being x by
Mr. Body] just if we understand ‘x’ as ranging over a quite restricted domain of relations,
{blackmailed, bribed, framed}. We can contrast this with an “open-ended” reading of
‘what was Body doing to Scarlet?’, in which the relevant domain of properties is left unre-
stricted (or much less restricted). With multiple choice readings of questions, we can say
that certain answers are under discussion, whereas in open-ended readings, no answers in
particular are salient or understood as the only appropriate responses.
Likewise, in Switched Blackmail
2
, (2) is used to address a multiple choice question that
centers on ‘Body’, with the answers Scarlet was blackmailed by Body, and Scarlet
was blackmailed by Mind being contextually understood as the only admissible answers
to the question that Black addresses with (2). This contrasts with an open-ended reading
of the question ‘who was blackmailing Scarlet?’, which it would take in a context where
there were no particular potential blackmailers antecedently under discussion.
IpointtothemultiplechoicereadingsofthequestionsaddressedinSwitchedBlackmail
2
and Switched Blackmail
4
to develop the suggestive pragmatic picture floated above. In
contextsinwhichanattitudereportsentenceoftheformAVsthatS(F) isusedtodirectly
address a question ?x[S(x)] which takes an open-ended reading, one plausible reason why
thespeakerwoulduseF inthecomplementofanattitudereportinaddressingthisquestion
is to represent A as “thinking of certain things” as F, that is to indicate that a concept
associated withF is used as part of certain ofA’s fully loaded thoughts. In multiple choice
cases, however, the answer involving F is already under discussion, so the speaker’s using
F in the complement of an attitude report is not introducing F (or what F designates or
any concept associated with F) embedded within an attitude report. So there are other
reasons that may be more plausible as to why the speaker uses this linguistic material than
to represent aspects of A’s fully loaded thought life. Although Direct QUD Sensitivity
doesn’t by itself make sense of these cases, the general picture-thinking that helped to
roughly rationalize why the principle should hold as a general rule in certain circumstances
also gives us a rough rationalization of why in multiple choice cases the principle should
often fail.
For a final little contrast, the next example introduces alternatives that don’t obviously
raise a contrast for particular linguistic constituents of (2)’s complement:
139
Switched Blackmail
5
: The detectives have private information that every socialist, fas-
cist, and counterfeiter was being blackmailed by Mr. Body. Detective Black questions
Green and learns that he thinks that Scarlet (with whom he is acquainted) is a socialist,
though the detectives know that Green is unaware that any blackmailing was going on.
As in the original Switched Blackmail, the detectives are exchanging information, trying to
figure out who in particular was being blackmailed by Body. Black assertively utters the
following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
Inthiscontext,tomyear,itisonceagaindi culttointerpret(2)astakingapermissive
reading with respect to ‘blackmailed by Mr. Body’. Although the detectives’ contextual
assumptions do raise certain alternatives to salience, that assumption alone doesn’t make
any question come to be under discussion other than who-was-blackmailed-by-Body?.
In particular, the particular “things” that might be understood as being under discussion
in this case, the properties socialist, fascist, and counterfeiter, do not neatly cor-
respond to any multiple choice question that centers on a particular constituent of (2).
Since the contextual assumptions in this case do not suggest that Black directly addresses
a multiple choice question, this case is basically just like the original Switched Blackmail.
As before, this can be made sense of by noting that the phenomenon of end-focus naturally
focuses the complement’s entire verb phrase, ‘being blackmailed by Mr. Body’. So even
though the question that is antecedently under discussion is ?x[x was being blackmailed
by Body], we can say that this end-focus makes us interpret (2) as directly addressing
?x[Scarlet was x]. Even though ?x[x was being blackmailed by Body] is under discussion
in the context, it is not directly addressed by Black’s use of (2), so we predict that Black’s
report cannot take a permissive reading with respect to ‘blackmailed by Mr. Body’.
We see then that the mere presence of conversational assumptions that correspond
to relevant alternatives is not what makes the pragmatic di↵ erence in cases like Switched
Blackmail; rather, it apparently is whether those relevant alternatives correspond to multi-
plechoicequestionscenteredonaconstituentoftherelevantreportsentence’scomplement.
5.3.2. Questions about Attitudes
Here I present another class of examples that show that Direct QUD Sensitivity fails,
but which support the general kind of picture-thinking floated above.
Uses of attitude report sentences sometimes directly address questions that cannot be
specifiedintermsofour‘?’ notationandjustthereportsentence’scomplement,butinstead
directly address a question that is most naturally specified with our ‘?’ notation and the
entire report sentence. For example, consider
Blackmail
4
: The detectives know that Scarlet, a guest at the manor, is a socialist.
They also know that she was being blackmailed, but do not know who was blackmailing
Scarlet. The detectives also commonly know that Green, another guest at the manor, is
140
unacquainted with Scarlet: Green doesn’t know any name for Scarlet, could not identify
her in a line-up, and so on. By questioning Green, Detective Black learns that Green
thinks that every socialist guest was being blackmailed by Mr. Body. Green doesn’t,
however, know the names of any of the socialist guests: he claims “not to know who the
socialist guests are”. The detectives are exchanging information, trying to find out who
GreenthinksScarletwasbeingblackmailedby. Indeed, indiscussingeventsgoingoninthe
manor,GreyasksBlack: whodoesGreenthinkwasblackmailingScarlet? Blackassertively
utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
To my ear, (2) takes a loaded reading with respect to ‘Scarlet’ in this context. What
does Direct QUD Sensitivity say about this case? Well, ?x[x was being blackmailed by
Body] does not seem to be directly addressed by (2) in this context. So according to Direct
QUD Sensitivity this should be able to take a permissive reading with respect to ‘Scarlet’,
contrary to my judgment.
What question is directly addressed by (2) in this context? Intuitively it is ?x[Green
thinks Scarlet was being blackmailed by x], which might be asked in this context by ‘Who
does Green think Scarlet was being blackmailed by?’. Direct QUD Sensitivity appears to
be basically silent on whether this question’s being under discussion should have any e↵ ect
on what kinds of permissive or loaded readings (2) can take.
Contrast this case with
Blackmail
5
: The detectives know that Scarlet, a guest at the manor, is a socialist.
They also know that she was being blackmailed, but do not know who was blackmailing
Scarlet. The detectives also commonly know that Green, another guest at the manor, is
unacquainted with Scarlet: Green doesn’t know any name for Scarlet, could not identify
herinaline-up,andsoon. ByquestioningGreen,DetectiveBlacklearnsthatGreenthinks
that every socialist guest was being blackmailed by Mr. Body. Green doesn’t, however,
know the names of any of the socialist guests: he claims “not to know who the socialist
guests are”. The detectives are exchanging information, trying to find out who Scarlet was
being blackmailed by. Indeed, in discussing events going on in the manor, Grey asks Black:
who was blackmailing Scarlet? Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
To my ear, (2) takes a permissive reading with respect to ‘Scarlet’ in this context. The
questiondirectlyaddressedby(2)intuitivelyis?x[Scarletwasbeingblackmailedbyx]; this
is quite similar to the question directly addressed by (2) in Blackmail
4
,?x[Green thinks
Scarlet was being blackmailed by x]. In both cases, ?x[x was being blackmailed by Body]
is not directly addressed; according to Direct QUD Sensitivity, therefore, both Blackmail
4
and Blackmail
3
should take permissive readings with respect to ‘Scarlet’.
141
Direct QUD Sensitivity is stated too narrowly to handle this case, but our more general
pragmaticpicturecanbeusedtomakesenseofwhatisgoingon. Let’sconsidertwocontexts
in which the question which the relevant report directly addresses varies itself in how it is
interpreted. First:
Forgeries
5
: The detectives know that all and only impressionist paintings in the gallery
areforgeriesinthegallery. AfterinterrogatingPeacock,detectiveBlackknowsthatPeacock
believes that every impressionist painting in the gallery is on the west wall of the gallery.
ThedetectivesknowthatPeacockisunawarethatthereareanyforgeriesinthegallery, but
for all they know she could think that there are forgeries hidden elsewhere in the manor.
Detectives Black and Grey are discussing their investigation with their superior, and Grey
says: ‘Somepeoplehavebeenspreadingrumorsaboutwheretheforgeriesarehiddeninthe
manor and we’d like to figure out who these people are. We’ve been trying to figure out
who has been tipped o↵ about the forgeries by asking them where they think the forgeries
are. Black,youinterrogatedPeacock: wheredoesshethinktheforgeriesare?’. Inresponse,
Black assertively utters
(70) Peacock believes that every forgery in the gallery is on the west wall.
To my ear, (70) takes a loaded reading with respect to ‘forgery in the gallery’. Like
Blackmail
2
, the report in this context directly addresses a question explicitly about the
subject’s attitudes. In this case Black’s report intuitively directly addresses ?x[Peacock
thinks that the forgeries are x], with the variable ‘x’ ranging over locations in the manor.
I’ll note that, given this loaded reading, (70) intuitively reads false as well; in response, it
would not be inappropriate for Grey to respond with ‘Wait, I thought that Peacock was
unaware that the gallery had any forgeries in it. Was I wrong?’.
Now contrast this case with
Forgeries
6
: The detectives know that all and only impressionist paintings in the gallery
areforgeriesinthegallery. AfterinterrogatingPeacock,detectiveBlackknowsthatPeacock
believes that every impressionist painting in the gallery is on the west wall of the gallery.
ThedetectivesknowthatPeacockisunawarethatthereareanyforgeriesinthegallery, but
thatshemaythinkthatthereareforgerieshiddenelsewhereinthemanor. DetectivesBlack
and Grey are discussing their investigation with their superior, and Grey says: ‘We know
thattheforgeriesinthegalleryareimpressionists,butwedon’tknowwhichofthepaintings
are impressionists. None of the guests is aware that the forgeries are impressionists or that
the impressionists are forgeries. We’d like to figure out which paintings in particular
are the forgeries, so we have been interrogating the guests about the paintings. Black,
you interrogated Peacock: where does she think the forgeries are?’. In response, Black
assertively utters
(70) Peacock believes that every forgery in the gallery is on the west wall.
142
To my ear, (70) takes a permissive reading with respect to ‘forgery in the gallery’. But
like Blackmail
4
and Forgeries
5
, the report in this context directly addresses a question
explicitly about the subject’s attitudes. Indeed, it serves to directly address what appears
to be the same question it addresses in Forgeries
5
,?x[Peacock thinks that the forgeries are
x], with the variable ‘x’ ranging over locations in the manor.
With these examples we gather some interesting data. First, contrary what might
have been at first suspected, a report can take a permissive reading with respect to some
linguistic constituent of its complement even when the report directly addresses a ques-
tion explicitly about the subject’s attitudes. Second, a single report sentence can vary
in whether it takes a permissive or loaded reading with respect to some linguistic con-
stituent of its complement across distinct contexts even though in each context it directly
addresses a question that can be asked by the same interrogative sentence in each. In
Forgeries
5
and Forgeries
6
both, (70) directly addresses a question that can be asked by
‘where does Peacock think the forgeries are?’.
Building on these observations, I’ll further note that questions explicitly about a sub-
ject’s attitudes themselves can take permissive readings with respect to linguistic con-
stituents of the complement of the embedded attitude report. This is just what happens
in Forgeries
6
: when Grey asks Black a question with ‘where does she think the forgeries
are?’, speaking of Peacock, intuitively we interpret ‘think’ as taking a permissive reading
with respect to ‘forgeries’. In that context, Grey is not asking anything like the question
of where does Peacock think the things she conceptualizes as forgeries are. In contrast, in
Forgeries
5
, when Grey asks black a question with ‘where does she think the forgeries are?’,
speaking of Peacock, intuitively we interpret ‘think’ as taking a loaded reading with re-
spectto‘forgeries’. Greyisindeedaskingsomethinglikethequestionwhere does Peacock
think the things she thinks of as forgeries are?.
At first we can simply hypothesize that when a report of the formAVsthatS(F,G)
directlyaddressesaquestionoftheform?x[AVsthatS(x,G)]thatitselftakesapermissive
reading with respect to G, then the report too may take a permissive reading with respect
to G, whereas if the report addresses a question of the form ?x[A Vs that S(x,G)] that
itself takes a loaded reading with respect toG, then the report takes a loaded reading with
respect to G.
111
This hypothesis alone can make sense of the observations made in this
section.
How does this fit in with the general sketch of how pragmatics informs permissive
reporting from above? Roughly put, in cases in which one uses a report of the form A
Vs that S(F,G) to address a question that is loaded with respect to G and an associated
concept , what is being asked already concerns aspects of A’s thought life in which A
uses . For the proposition asserted by the speaker’s use of this report sentence to serve to
provideananswertothequestionaddressedatall,itapparentlymustatleasttakealoaded
reading with respect to G and . When the question addressed,AVsthatS(F,G), takes
111
Here “S(F,G)” stands for a clause S in which both phrases F and G occur.
143
a permissive reading with respect to G, however, reasoning like that involved in multiple
choicescenarioscankickin: Gisalreadyinsomewayunderdiscussion,sothespeaker’suse
ofG is not an introduction ofG as embedded in an attitude report. Furthermore, there are
no considerations of general discourse coherence that require that G take a loaded reading
for the speaker’s report to be relevant to the question addressed. Our general picture can
be used to make some sense of the phenomenon of questions about attitudes, then, even
though Direct QUD Sensitivity itself cannot.
Finally,giventhevarietyofwayswehaveseenalreadythataspeaker’susingsomewords
in issuing a report may or may not pragmatically indicate that they intend to represent
certain aspects of the subject’s fully loaded thought life in doing so, I speculate that it
is too heroic to attempt to bring all of these pragmatic subtleties under one principle
stated just in terms of QUDs. Rather, I take it that what is going on here is that the
observed sensitivity of permissive readings to QUDs is one facet of a larger pragmatic
practice in which whether a use ofAVsthatS(F) takes a loaded reading with respect toF
depends ultimately on whether, given the assumptions and goals in the context, the most
plausible reason that the speaker used phrase F associated with a concept is that they
intend to represent A as using in certain ways in certain of their fully loaded thoughts.
This, clearly, is not so much a special pragmatic theory specific to the phenomenon of
permissive reporting as much as it is a roughly stated consequence of the sort of semantic
underspecification postulated by the sensitive attitudes view and conceptually permissive
attitudes view, given our general pragmatic practices involved in utterance interpretation
across the board.
6. Question Sensitivity and the Truth-Conditional Question
Question-sensitive principles, despite their limitations, appear to account for the con-
trastraisedbyexamplesbehindtheasymmetryproblemforcontextualaugmentationviews.
If we can cast the contextual augmentation view in a way that takes to heart the question-
sensitivity of permissive readings, then the resulting view may skirt the asymmetry prob-
lem. Further, I think that once we reframe contextual augmentation views to be sensitive
to questions, we have the resources for a general defense of contextual augmentation views
against over-generation worries.
The first step towards reframing the contextual augmentation view to be compatible
with question-sensitive pragmatic principles is to take care in distinguishing which theo-
retical question contextual augmentation views are meant to answer. At this point, we
can distinguish two theoretical questions concerning permissive reports. One we can call
the “truth-conditional question of permissive reports”: under what conditions is a permis-
sive attitude report true? The other we can call the “pragmatic question of permissive
reports”: under what conversational conditions does a use of a report sentence serve to
make a permissive report? (or: ...a report that is permissive with respect to certain lin-
guistic constituents of the sentence’s ‘that’-clause and certain conversationally associated
concepts).
144
Once we have distinguished these two theoretical questions, we can o↵ er a diagnosis of
why the original contextual augmentation view goes wrong: namely, it conflates these two
questions. The contextual augmentation view was put like this:
Contextual Augmentation View:
A report of the formAVsthatS made in context c is true if and only there are
contextuallyassumptionsp
1
,...,p
n
incontextcandpropositionsq
1
,...,q
m
towhich
A bears V
L
such that {p
1
,...,p
n
,q
1
,...,q
m
} entails the proposition designated by
‘that S’ in c.
Putthisway, thecontextualaugmentationviewisimplicitlyaimedatprovidinggeneral
conditionsforwhenonecanissueatruepermissivereport. Theconditionofconversational
entailment is designed to account for the conditions in which a report is permissive, but
thisisconflatedwiththeissueofunderwhatconditionsareportistrue. Toposeacounter-
example to this theory, one therefore need only find a context in which a report is either
not permissive or not true even though the proposition designated by its ‘that’-clause is
contextually entailed by the subject’s loaded attitudes.
Some of the over-generation worries raised against contextual augmentation views took
this shape. What these potential counter-examples do is illustrate that some contextual
entailments of a subject’s loaded attitudes cannot be both truly and permissively ascribed
as thought contents to the subject. Similarly, the asymmetry problem tells against this
version of the contextual augmentation view by posing pairs of contexts such that some
proposition is contextually entailed in each, but can only be reported as the content of a
permissive attitude in one. Since the context in which the relevant report takes a loaded
readingisoneinwhichitalsotakesafalsereading,wehaveacounter-exampletocontextual
augmentation views so posed.
But once we have distinguished the truth-conditional question from the pragmatic
questionofpermissiveness, thereappearstobesomethingillicitaboutthisstyleofcounter-
example. Intuitively what we want from the contextual augmentation view is an account
of the conditions under which a permissive report is true; that we can find conversational
scenarios in which a report of the formAVsthatS takes a loaded reading and is also false
should be irrelevant to any contextual augmentation view understood strictly as an answer
to the truth-conditional question of permissiveness.
A first pass at a revised version of the contextual augmentation view, then, might be
something like:
Permissive Augmentation View:
A report of the formAVsthatS, that takes a permissive reading in context c,
istrueifandonlyiftherearecontextuallyassumedpropositionsp
1
,...,p
n
incontext
c and propositions q
1
,...,q
m
to which A bears V
L
such that {p
1
,...,p
n
,q
1
,...,q
m
}
entails the proposition designated by ‘that S’ in c.
145
This is a start, but we already have reasons to be unsatisfied with this simple fix,
given some observed subtleties of permissiveness. Specifically, we have observed that a
report can be permissive with respect to some but not all of the linguistic constituents
of a report’s ‘that’-clause (and concepts that are conversationally associated with those
linguistic constituents). Since the above version of the Permissive Augmentation View
runs roughshod over this important detail, we should adopt an alternative.
IfinditinstructivetoseehowsomewaysofmakingthePermissive Augmentation View
sensitive to this phenomenon leads to inaccurate results, so I work through some tempting
false starts in building toward my proposed revision of this principle. Consider
Concept-Possession Augmentation View:
A report R made with a sentence of the formAVsthatS(F) in context c that
takes a permissive reading with respect to F (and a conversationally associated
concept) is true i↵ V
A
non-trivially and distinctively contextually entails the
propositionpdesignatedby‘thatS’incandApossesseseveryconceptwithrespect
to which R is loaded in c.
This version of the augmentation view at least allows that a report can be permissive
with respect to some but not all constituents of its ‘that’-clause. This, together with a
sensitivity to the distinction between the truth-conditional question and the pragmatic
question of permissiveness, serves to account for some of the counter-examples to the
original contextual augmentation views, as I now show.
Recall our main instance of the asymmetry problem: in Switched Blackmail the re-
port made with (2) takes a loaded reading with respect to ‘blackmailed by Mr. Body’.
Although the proposition designated by the complement of (2) is non-trivially and distinc-
tively contextually entailed by Green’s belief set, since Green lacks a concept of Body in
that contextual scenario, hence lacks a concept of the property blackmailed-by-Body,we
predictthatthereporttakesafalsereading, asitdoes. Similarly, intheoriginalBlackmail,
the report made with (2V) takes a loaded reading with respect to ‘Scarlet’; so, although
this report’s complement is non-trivially and distinctively contextually entailed by Green’s
belief set, we predict a false reading since in that scenario Green lacks a concept of Scarlet.
One does not have to look far to find slight variants of Blackmail and Switched Black-
mail that underly clear counter-examples to Concept-Possession Augmentation View. For
example, consider contexts just like Blackmail and Switched Blackmail in which Green
does not have a loaded belief that Scarlet is being blackmailed by Body but he does pos-
sess concepts of Scarlet and of Body. In such contexts we still detect a false reading of
(2), but Concept-Possession Augmentation View does not predict this false reading. A
subject’s merely possessing a concept with respect to which a given report is loaded is not
enough for the subject to verify the report.
What extra restriction is required to get the right results for these variant cases? We
want at least to require that the subject use the concepts they are required to possess;
146
though it won’t su ce to require that the subject use the relevant concepts in entertaining
any ole proposition. Green could have a loaded belief in, for example, the proposition
that Scarlet is taller than Mr. Body, but he would not thereby verify the reports that
Concept-Possession Augmentation View founders on. Intuitively what is required is that
the subject use those concepts with respect to which a report is loaded as part of whatever
loaded attitudes are the “basis” for the permissive report. The idea behind the contextual
augmentation view, after all, is that one has a permissive attitude in some proposition in
virtueofhavingarelatedloadedattitudeinsomeotherpropositions; intuitively, ifareport
is loaded with respect to some constituent F of its complement, then, the subject must
deploy the concepts conversationally associated withF in holding the loaded attitude from
which the reported permissive attitude is derived. This idea is reflected in the following
principle:
Concept-Use Augmentation View:
A report R made with a sentence of the formAVsthatS(F) in context c that
takes a permissive reading with respect to F is true i↵ for every concept with
respect to which R is loaded, A uses in entertaining a proposition p expressible
by a clause of the form S(G), and p is in a subset V
of V
A
, such that V
non-trivially and distinctively contextually entails the proposition designated by
‘that S’ in c.
Roughly, this says that a report which is permissive with respect to some constituent
F (and a conversationally associated concept) occurring in the report’s complement, that
S(F), is true if and only if the subject has a related loaded attitude in a proposition p
which is of the same structure as that designated by that S(F) save for the constituents in
the place of F in S(F), p (perhaps with other propositions which the subject has a loaded
attitude in) contextually entails the proposition designated by that S(F), and the subject
uses every concept with respect to which the report is loaded in entertaining p.
For example, since in Blackmail the report made with (2) is permissive with respect to
‘Scarlet’ and is loaded with respect to ‘being blackmailed by Mr. Body’, Concept-Use Aug-
mentation View says that (2) is true if and only if Green has a loaded attitude in a propo-
sition p of the form NP was being blackmailed by Mr. Body, where some constituent
expressiblebyanounphraseisintheplaceof NP,pcontextuallyentailsthatScarletwasbe-
ing blackmailed by Mr. Body, and Green uses a concept of being-blackmailed-by-Body
in entertaining p. The report is predicted to be true since Green has a loaded belief in
the proposition that every socialist was being blackmailed by Mr. Body, this proposition
contextually entails that Scarlet was being blackmailed by Mr. Body in Blackmail, and
Green uses a concept of being-blackmailed-by-Body in entertaining that every socialist
was being blackmailed by Mr. Body.
Concept-Use Augmentation View successfully avoids the counter-examples to Concept-
Possession Augmentation View. For example, reconsider a context like Switched Blackmail
147
inwhich Greenpossessesaconceptof Body but doesnotusethatconceptin entertaininga
proposition which he has a loaded belief in, which is of the form Scarlet-was-F and which
contextually entails that Scarlet was being blackmailed by Mr. Body. Whereas Concept-
Possession Augmentation View predictsthatareportmadewith(2)shouldbetrueinsuch
a context, Concept-Use Augmentation View predicts that is should be false, since Green
does not use his concept of Body (or his concept of blackmailed-by-Body) as part of a
loadedbeliefthecontentofwhichcontextuallyentailsthatScarletwasbeingblackmailedby
Body. He does have a loaded belief in some proposition of the form Scarlet-was-F which
contextually entails that Scarlet was being blackmailed by Body; namely, the proposition
that Scarlet was a socialist. But in entertaining this proposition he does not use the
concepts with respect to which (2) is loaded in this variant of Switched Blackmail,sowe
correctly predict a false reading.
This is almost what we want. But at least one last issue prevents Concept-Use Aug-
mentation View from being quite right. This can be illustrated with an example from
chapter 4:
Sister: Green sees Scarlet from afar. Scarlet is wearing her sister Jade’s green dress,
rather than Scarlet’s usual red dress. Green tells detective Black that the person he sees
from afar is Jade, Scarlet’s sister. Black and Grey both know that the person in the green
dress is Scarlet, not Scarlet’s sister. In discussing the situation, Black assertively utters
the following:
(57) Green thinks that Scarlet is Scarlet’s sister.
What Black reports with (57) intuitively is permissive with respect to the first but not
the second occurrence of ‘Scarlet’. Concept-Use Augmentation View therefore says that
this is true if and only if Green uses a concept of Scarlet in entertaining a proposition of
the form NP-is-Scarlet’s-sister, to which he bears the fully loaded thinks relation,
and this proposition contextually entails that Scarlet is Scarlet’s sister. Since Green has a
loadedbeliefinthepropositionthatScarletisScarlet’ssister, cognizingthefirstoccurrence
of Scarlet in this proposition via a visually mediated concept (or, depending on how we
spell out the case, a linguistic ‘Jade’ concept, or a descriptive concept the woman in the
green dress) and cognizing the second occurrence of Scarlet via the conversationally asso-
ciated linguistically mediated ‘Scarlet’ concept, and this contextually entails that Scarlet
is Scarlet’s sister, we correctly predict a true reading. So far so good.
But notice that in applying Concept-Use Augmentation View to this example we did
not have to make reference to which constituent of the proposition that Scarlet is Scarlet’s
sisterGreencognizesviathelinguistic‘Scarlet’conceptwithrespecttowhich(57)isloaded.
For this reason, we can pose a counter-example to Concept-Use Augmentation View which
exploits this oversight. Consider the following variant of Switched Blackmail:
Switched Blackmail
6
: The detectives have private information that every guest in the
148
manor who was blackmailing someone who was being blackmailed by Body was themselves
beingblackmailedbyBody. HavingjustquestionedGreen,BlackknowsthatGreenbelieves
thatScarlet(withwhomGreeniswellacquainted)wasblackmailingWhiteandalsobelieves
that White was blackmailed by Body. Although the detectives know that Scarlet was
blackmailing White, the detectives do not know that White was, as Green believes, being
blackmailed by Body. The detectives further know that Scarlet is Jade’s sister; they know
that Green does not know this. They are exchanging information, trying to find out who
had a motive for killing Body; in particular, in this scenario they are trying to figure out
who in particular was being blackmailed by Body. Black assertively utters the following
to Grey:
(71) Green believes that Jade’s sister was being blackmailed by Mr. Body.
In this context, (71) intuitively takes a false reading which is loaded with respect to
‘being blackmailed by Mr. Body’ but is permissive with respect to ‘Jade’s sister’. Concept-
Use Augmentation View predicts in this case that, since the report is permissive with
respect to ‘Jade’s sister’, (71) is true if and only if Green possesses a fully loaded belief
in a proposition of the form NP was being blackmailed by Mr. Body, this proposition
contextually entails that Jade’s sister was being blackmailed by Mr. Body, and Green uses
a concept of blackmailed-by-Mr.-Body in holding this fully loaded belief. Green does
possess a fully loaded belief in the proposition that White was blackmailed by Body, which
is of the form NP was being blackmailed by Mr. Body. This proposition contextually
entails that Jade’s sister was being blackmailed by Mr. Body (since, together with the
proposition that Scarlet was blackmailing White, which is a contextual assumption, it en-
tails that Scarlet was blackmailing someone who was blackmailed by Mr. Body. This,
together with the contextual assumptions that everyone who was blackmailing someone
who was blackmailed by Body was themselves blackmailed by Body and the contextual
assumption that Scarlet is Jade’s sister, entails that Jade’s sister was being blackmailed
by Mr. Body.) And Green uses a concept of blackmailed-by-Mr.-Body in holding this
fully loaded belief (that White was being blackmailed by Body).
112
So Concept-Use Aug-
mentation View predicts that (71) is true in this context, contrary to observed fact.
TheproblemseemsatleastdueinparttothefailureofConcept-UseAugmentationView
to discriminate between which propositional constituents are cognized by which concepts.
In Switched Blackmail
6
Green does not use a concept of blackmailed-by-Body as part
of a fully loaded belief which predicates this property of Scarlet. A correct version of this
contextual augmentation view has to require that the subject of a permissive report which
is loaded with respect to certain occurrences of linguistic constituents and conversationally
associated concepts uses those concepts to cognize particular constituents of their fully
loaded attitudes.
112
This presupposes a certain notion of concept deployment, similar to assumptions invoked before, ac-
cording to which in cognizing a complex propositional constituent containing a constituent F one uses a
concept of F.
149
To frame a question-sensitive contextual augmentation view in a way that is sensitive
to how permissive readings are occurrence-specific, I think that first we should to be more
explicit about the structure of propositions.
113
For our purposes, it will serve to con-
ceive of propositions as ordered sequences propositional constituents.
114
So, for example,
we will model the proposition that Scarlet is being blackmailed simply with hScarlet,
being-blackmailedi, representing that this proposition predicates the property of be-
ing blackmailed of Scarlet (ignoring tense). With this simple model, we can distinguish
between di↵ erence occurrences of a single proposition in terms of positions within the as-
sociated ordered sequence. Consider Russell’s proposition that di↵ erence is di↵ erent from
identity.
115
We can model this with the sequence, hdifference,hdifference,identityii,
representing that this proposition predicates the relation of difference of the relations
difference and identity. By distinguishing between di↵ erent occurrences of a proposi-
tional constituent we can say whether an attitude expressed by some report is permissive
withrespecttoaparticularoccurrence ofsomepropositionalconstituentandsomeconcept
associate with that occurrence.
116
With this distinction, I now pose a version of a contextual augmentation view as a
truth-conditional rule for loaded reports. To be specified are the truth conditions for a
reportR made by a use of a sentence of the formAVsthatS(F) in contextc that is loaded
with respect to a concept and a linguistic expression F that contributes a propositional
constituentftoaparticularpositionointhepropositionpdesignatedbythereport’s‘that’-
clause. The specification then proceeds in two parts. First, we say that R is true only if
there is a proposition q structurally similar to p such that f occurs in q in a structurally
analogous position to o in p, and A uses to cognize f as it occurs in that position in q
in bearing V
L
to q.
117
Second, as before, we say that R is true only if q non-trivially and
distinctively contextually entails p in c. This gives us:
Sophisticated Concept-Use Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) in context
c that is loaded with respect to concept and linguistic expression F that
contributespropositionalconstituentftopositionointhepropositionpdesignated
113
Those who favor an unstructured approach to propositions have to make due by theorizing in terms of
occurrences of linguistic expressions in ‘that’-reports.
114
Where for us we can distinguish atomic constituents, which are objects and relations, from complex
constituents, which are ordered sequences of constituents (atomic or complex).
115
From Russell (1903)’s discussion on the problem of the unity of the proposition.
116
Forreasonsdiscussedinchapter4, Idon’twanttosaythattheoccurrenceofapropositionalconstituent
with respect to which some report is loaded occur in the proposition which the subject said to bear an
attitude towards.
117
As in chapter 3 and above, VL is a fully loaded V relation. I avoid saying “the” fully loaded V relation
given observations made in this chapter: our ideology of “loaded” should be designed so that one report
may be loaded with respect to an occurrence of f in p and concept while another report may be loaded
with respect to an occurrence of f in p and a concept distinct from . Once we allow for this, we can
expect that there will be many many fully loaded V relations.
150
by the report’s ‘that’-clause is true i↵ there is a proposition q structurally similar
to p such that f occurs in q in a structurally analogous position to o in p, and A
uses to cognize f as it occurs in that position in q in bearing V
L
to q and q
non-trivially and distinctively contextually entails p in c.
To seeSophisticatedConcept-Use Augmentation View atwork,let’sapplyittoSwitched
Blackmail
6
, which posed a problem for the simpler Concept-Use Augmentation View.In
this context, Green bears a fully loaded belief relation to the proposition that White was
blackmailed by Body, so uses the concept blackmailed-by-Body in cognizing the property
blackmailed-by-Body as it occurs in that proposition in bearing the fully loaded belief
relation towards that proposition. This proposition,hblackmailed-by-Body,Whitei intu-
itivelyisstructurallysimilartothepropositionhblackmailed-by-Body,Jade’s sisteri.
118
Since the detectives commonly assume (i) everyone in the manor who was blackmailing
someone who was blackmailed by Body was blackmailed by Body, (ii) Scarlet was black-
mailing White, and (iii) Scarlet is Jade’s sister, that White was blackmailed by Body
non-trivially and distinctively contextually entails that Jade’s sister was blackmailed by
Body. Soifthereportmadewith(71)inSwitched Blackmail
6
isindeedloadedwithrespect
to ‘blackmailed by Mr. Body’ but is not loaded with respect to ‘Jade’s sister’, then even
Sophisticated Concept-Use Augmentation View predicts a true reading in this case.
Intuitively (71) takes a false reading in Switched Blackmail
6
. Now, according to QUD
Sensitivity,sinceneither?x[x was blackmailed by Body] nor ?x[Jade’s sister was x]is
under discussion, QUD Sensitivity is silent as to whether Black’s report takes a permissive
readingwithrespecttoeither‘Jade’ssister’or‘blackmailedbyBody’. Aswewillseeinthe
next chapter, there are general pragmatic reasons to expect that since Green is aware that
some were being blackmailed by Body, we should expect that (71) takes a loaded reading
with respect to ‘being blackmailed by Body’, and since Green is unaware that Scarlet is
Jade’s sister, and in general is not relevantly deploying a concept Jade’s-sister, we should
expect that (71) takes a permissive reading with respect to ‘Jade’s sister’.
Assuming, then, that (71) is loaded with respect to ‘blackmailed by Body’ but per-
missive with respect to ‘Jade’s sister’, Sophisticated Concept-Use Augmentation View does
predict a true reading in this case, contrary to observed fact. So more work is needed.
Sophisticated Concept-Use Augmentation View does seem to improve on its predeces-
sors, despite the problem it faces with Switched Blackmail
6
. At this point, rather than
continue to tweak Sophisticated Concept-Use Augmentation View to get the right results, I
will point to a few factors that may be driving the problem raised by Switched Blackmail
6
and will make some methodological remarks as to why we should expect that we may be
in a position to see how to modify Sophisticated Concept-Use Augmentation View after we
have addressed certain other puzzling features of permissive reports.
118
Clearly I am ignoring complications to do with the structure of tensed propositions and the structure
of the contributions of possessive phrases.
151
There seem to be at least three factors that may be responsible for the problem that
Sophisticated Concept-Use Augmentation View encounters in Switched Blackmail
6
.
First, in this context the loaded attitude of Green’s that is supposed to ground Black’s
reportintuitivelyfailstoconnectinsomerelevantwaytothepropositionascribedtoGreen
by that report.
When a speaker issues an attitude report with a sentence of the formAVsthat S
that takes a loaded reading with respect to some occurrence o of a phrase F in S(F) and
an associated concept , a speaker intuitively is committed at least to the claim that the
subject cognizes F via as part of holding a certain related fully loaded propositional
attitude. That is, the speaker does not just represent the subject as possessing a concept
, and furthermore does not just represent the subject as using to cognize F per se;but
the speaker commits herself to the subject’s using to cognize F as part of entertaining a
particularproposition(orperhapssomepropositionfromaparticularclassofpropositions).
We have seen several times over that not any attitude will do here: the truth of such a
report requires more than that the subject just use in entertaining some proposition or
other. The case of Switched Blackmail
6
seems to indicate that this certain attitude in this
particular proposition cannot just be a fully loaded attitude towards a proposition q such
that F occurs in q in a structurally similar position to o in that S(F) and such that q
contextually entails that S(F).
It is di cult to say exactly what more is needed by way of a connection between
Green’s fully loaded attitude and the proposition ascribed to him by Black’s report. It
may have something to do with a divergence in what Green and the detectives apply the
blackmailed-by-Body conceptto. Greendoesindeeddeploytheconceptblackmailed-by-Body
in cognizing the property blackmailed-by-Body in a structurally similar position in the
proposition that White was blackmailed by Body to the position of blackmailed-by-Body
inthepropositionthatJade’ssisterwasblackmailedbyBody,whichthedetectivesassociate
with blackmailed-by-Body. But Green does not apply this concept to the same things that
the detectives do in their entertaining of the proposition that Jade’s sister was blackmailed
by Body. Green applies this concept to White, but is not disposed to apply it to Scarlet
under any guise. Perhaps it is for this reason that there is some sense in which Green’s
loadedbeliefdoesnot“count”towardshavingthereportedpermissivebeliefinthiscontext.
This, then, is one potential factor that could be responsible for the failure of Sophisticated
Concept-Use Augmentation View in Switched Blackmail
6
: there are further constraints on
which propositions can “ground” a true permissive report, perhaps having to do with how
the subject and the conversationalists apply concepts with respect to which the report is
loaded.
A second potential factor that may be responsible for the problem facing Sophisticated
Concept-Use Augmentation View concerns the number and complexity of contextual as-
sumptions involved in the contextual entailment in this case. The contextual entailment
in this case goes by way of several contextual assumptions rather than just one. Unlike
the original Blackmail scenario, in which the contextual entailment just relies on the de-
152
tectives’ single assumption that Scarlet is a socialist, the contextual entailment in this
case relies on the assumptions that Scarlet was blackmailing White, that Scarlet is Jade’s
sister, and the peculiar assumption that everyone blackmailing someone blackmailed by
Body was blackmailed by Body. One possibility is that it is di cult to hear a true report
that is permissive with respect to ‘Jade’s sister’ in this case just because the contextual
entailment depends on conversational assumptions that are many and strange. Without
trying here to give any theoretical explanation for why this might be a relevant factor, I’ll
simply note that without exploring this dimension of permissive reporting, we cannot be
sure that some version of Sophisticated Concept-Use Augmentation View that is sensitive
to this issue won’t account for Switched Blackmail
6
.
Athird, related, factorthatmayberesponsibleforthefailureofSophisticated Concept-
Use Augmentation View is simply that contextual entailment, even non-trivial and dis-
tinctive contextual entailment, may be the wrong relation to use in characterizing how
a subject’s fully loaded attitudes relate to a proposition ascribed as entertained by them
by a true permissive report. Indeed, if the number or nature of the contextual assump-
tions involved in a given case can influence whether a given report takes a true reading,
then contextual entailment itself, which is insensitive to such di↵ erences, cannot be the
right relation to use. Even once we have distinguished the truth-conditional question of
permissiveness from the pragmatic question of permissiveness, there are still reasons to
question whether a contextual entailment relation is the right relation of augmentation to
use in answering the truth-conditional question. For instance, we still have not addressed
the attitude variation problem raised against contextual augmentation views in chapter 3.
Perhaps this problem can be handled just by distinguishing the truth-conditional question
ofpermissivenessfromthepragmaticquestionofpermissiveness, butifitcannot, itmaybe
reason to look for some more general augmentation relation than contextual entailment.
And if that is so, then perhaps whatever refinement of contextual augmentation views
can handle the attitude variation problem may also serve to avoid the problem raised by
Switched Blackmail
6
. Again, then, it would be methodologically hasty at this point to dis-
card the apparent progress made by moving to Sophisticated Concept-Use Augmentation
View just because it cannot account for this particular complicated case.
I suggest then that we put a pin in this problem as we press on to investigate further
pragmatic aspects of permissive reporting. By exercising some methodological patience
here we can appreciate that there is something right about the Sophisticated Concept-Use
Augmentation View even if it itself is not wholly right. In the next chapter I address the
pragmatic significance of mismatches in how the conversationalists and the subject apply
a target concept. In the concluding chapter, I speculate about how the lessons learned in
this chapter and the next can be used to address the attitude variation problem.
153
Appendix: Various Variations
Here I present various examples of UI reports for various reasons to show that the main
points made in this chapter do not crucially depend on peculiarities about the Blackmail
and Switched Blackmail cases.
A.1. More Kinds of UI Reports
At the end of §2 I claimed that one can find UI reports that are permissive with respect
to phrases other than names, including indefinite descriptions, definite descriptions, and
other quantifier phrases. I produce examples of such reports here.
Some Passage: Continuing their investigation, the detectives learn that some room
with a secret passage is in the east wing, though they are unsure which room (or rooms) in
the east wing have secret passages. They know, however, that White is unaware that there
are any secret passages in the manor. After investigating White, Black learns that White
mopped up muddy footprints which were tracking through the east wing. On the basis
of this, White believes that every room in the east wing had the murderers footprints in
them. The detectives are exchanging information, trying to find out whether the murderer
used a secret passage in the course of committing the crime. Black assertively utters the
following to Grey:
(72) White believes that some room with a secret passage had the murderer’s footprints
in it.
What Black asserts in uttering (72) is true, but since White is known to be unaware
that there are any secret passages, this use of (72) constitutes a permissive report.
Now, it is not clear whether the indefinite ‘some room’ should be interpreted as specific
or non-specific. Insofar as the report is permissive, it isn’t obvious that there is a reason
against interpreting ‘some room’ as specific. But it’s also reasonable to think that here
‘someroom’isnon-specific, sincethere’snospecialreasontotakeBlack(orWhite)tohave
a certain room “in mind”. It’s worth pointing out that there are versions of this kind of
example on which the target indefinite is definitely interpreted as specific. Consider:
A Certain Passage: The detectives have discovered the secret passage leading from the
conservatoryintheeastwingtothestudyinthewestwing. Sotheyknowthatthereareat
leasttworoomswithsecretpassages. Theyrecentlyhaveacquiredevidencesuggestingthat
the murder took place in the conservatory, so they’ve been focusing their investigation on
the conservatory. As in Some Passage, Black knows that White believes that every room
in the east wing had the murderer’s footprints in it. Black assertively utters the following
to Grey:
(73) White believes that a certain room with a secret passage had the murderer’s in it.
154
Here the indefinite ‘a certain room’ naturally takes a specific reading: intuitively, Black
is making reference to the conservatory, as it is salient to the detectives ongoing investiga-
tion. (73) takes a permissive reading similarly to (72) in Some Passage. Apparently, then,
a UI report can be permissive with respect to an indefinite description whether or not it
is specific or non-specific.
Burglar: Thedetectiveshaveprivateinformationthatlastnighttherewasacatburglar
afoot in the manor, though they don’t know the burglar’s name, couldn’t identify the
burglar in a line-up, and so on. The detectives have further private information that the
cat burglar was in the ballroom at the time of the murder. They also know that Peacock
is unaware that there was a cat burglar in the manor and indeed that she is unacquainted
with whoever the cat burglar is. From investigating Peacock, detective Black learns that
Peacockthinksthateveryonewhowasintheballroomatthetimeofthemurdercouldhear
Body scream. The detectives are exchanging information, trying to find out why the cat
burglar didn’t steal anything of importance last night. Detective Black assertively utters
the following to Grey:
(74) Peacock thinks that the cat burglar could hear Body scream.
Black’s assertion here readily takes a permissive reading. Notice also that there isn’t
much promise in denying that the report is permissive by resorting to a wide scope inter-
pretation of the definite, as in:
(74R) [The cat burglar]
x
[Peacock believes [x could hear Body scream]]
If the proposition asserted by Black corresponding to regimentation (3R), then on a
scopal account of transparent readings, Black does not use ‘cat burglar’ in a permissive
way. Still, though, the truth of this assertion would require Peacock to have a singular
belief about cb,where cb is the cat burglar. But Peacock is also known to be unacquainted
with that person. Thus whether the definite in (74) is interpreted as scoping narrow or
wide to ‘believes’, (74) counts as a permissive report in Burglar.
Safe: The detectives have private information that every weapon with Green’s finger-
prints on them is in the safe (in addition to some other weapons). They further know that
Plum is acquainted with the weapons in the safe only via that very description: the sole
way Plum has of thinking of these objects is “as weapons in the safe”. Given this, the
detectives know that Plum is unaware of which weapons have Green’s prints on them, and
indeed is unaware of whether Green has touched any weapons at all. After questioning
Plum, Black learns that Plum believes that every weapon locked in the safe was used to
commit a crime. The detectives are exchanging information, still trying to figure out who
killed Body. Black assertively utters the following to Grey:
(75) Plum thinks that every weapon with Green’s prints on them was used to commit
a crime.
155
To my ear, what Black asserts in uttering (75) here is true. If that’s correct, then it is
a permissive report, since Plum is not relevantly disposed to think of any weapons as those
“with Green’s prints on them”.
A.2. More Switched Contexts
NextIshowthattheSwitched UI Context Schema from§2canbeusedtoposeinstances
of the asymmetry problem for some of the UI reports just presented above.
Switched Passage: Thedetectiveshavefoundmuddytracksleadingthrougheveryroom
in the east wing. These match the boots which the detectives take to be the murderer’s,
so the detectives commonly assume that every room in the east wing has the murderer’s
footprints in it. From questioning White, Black learns that White has been told by the
butler that there is a secret passage leading from a room in the east wing. So Black
knows that White believes that some room with a secret passage is in the east wing. The
detectives know, however, that she is unaware that there was a murder last night. The
detectives are exchanging information, trying to figure out how the murderer escaped from
the scene of the crime without being seen. Black assertively utters the following to Grey:
(72)Whitebelievesthatsomeroomwithasecretpassagehadthemurderer’sfootprints
in it.
Though the proposition contributed by the complement clause of (72) is entailed by
the contextual assumptions and White’s loaded beliefs, it is distinctively more di cult to
interpret (72) in Switched Passage as a true permissive report than it is in Some Passage.
Instead, inthiscontextBlackismostnaturallyinterpretedasmakingafalseloadedreport.
Switched Burglar: The detectives have worked out among themselves that whoever
who was in the ballroom at midnight last night was in a position to hear Body scream.
The detectives know that Peacock is unaware of this; Black and Grey commonly know that
PeacockisunacquaintedwithBodyandisunawarethattherewasmurderlastnight. From
investigating Peacock, Black comes to learn that she believes that the cat burglar was in
the ballroom at midnight. The detectives are exchanging information, now trying to figure
out who in particular could hear Body scream. Black assertively utters the following to
Grey:
(74) Peacock believes that the cat burglar could hear Body scream.
Here it is di cult to interpret (74) as a true permissive report. Intuitively, Black’s use
of (74) in Switched Burglar represents Peacock as thinking of the cat burglar as one who
“could hear Body scream”. So it makes a false loaded report.
Now, it’s reasonable to expect that modifying the context of Safe by swapping its con-
textual assumptions with the objects of Plum’s loaded attitudes will make no di↵ erence in
156
whether (74) can serve to make a true permissive report. For in this case both the primary
contextual assumption and the primary relevant object of Plum’s loaded attitudes are uni-
versally quantified propositions. Still, we find that switching the contextual assumptions
and Plum’s loaded thought contents in Safe makes for a di↵ erence in the accessibility of a
permissive report.
Switched Safe: Thedetectivescommonlyknowthateveryweaponwhichislockedinthe
safe was used to commit acrime. The detectives further knowthat Plum is unawareof this
fact; he isn’t even aware that any crimes have been committed around the manor. After
questioning Plum, Black learns that Plum has dusted all of the weapons in the manor for
fingerprints; both those which are now in the safe and those which aren’t. After dusting,
Plum concluded that every weapon with Green’s prints on it is currently locked in the safe.
Black assertively utters the following to Grey:
(75) Plum thinks that every weapon with Green’s prints on them was used to commit
a crime.
It is more di cult to interpret (75) as true in Switched Safe than it is to interpret it as
true in the original Safe. To my ear, Black’s assertion in Switched Safe represents Plum as
“thinking in terms of crimes”. It would be natural for Grey to respond to Black’s assertion
here by questioning: “wait, Plum knows about the crimes? I thought he was unaware that
anycrimeswerecommitted!” Thisreflectshow(75)apparentlymostnaturallytakesafalse
loaded reading in Switched Safe. If this is correct, it is somewhat surprising. For, again,
both the object of the subject’s relevant loaded attitude and the relevant assumption made
by conversationalists is a universally quantified proposition.
A.3. Non-UI Asymmetry Problems
UI reports are, on my view, but one species of permissive report. I point out here that
the asymmetry problem does not just arise for UI reports, but is a general phenomenon
for all sorts of permissive reports.
Reconsider, for example, the following example from chapter 2:
Innocence: Professor Plum is agnostic about who committed the murder: he neither
believes that Green is the murderer nor believes that Green is innocent, and similarly
for every other guest. But he knows that only one person committed the murder, so
he believes that most of Green, White, and Scarlet are innocent. Detective Black and
Grey know something that Plum doesn’t: Green, White, and Scarlet are all and only the
members of a conspiracy to blow up police headquarters. Having just interrogated Plum
about his opinions concerning who the murderer might be, detective Black sincerely utters
the following to detective Grey as they are discussing the conspirators:
(76) Plum believes that most conspirators are innocent.
157
As noted before, the report made with (76) in this context is permissive with respect
to ‘conspirators’.
Now consider a switched version of this context:
Switched Innocence: The detectives together assume that only one person committed
the murder, so assume that most of Green, White, and Scarlet, are innocent. By inter-
rogating Plum, Black has learned that Plum believes that Green, White, and Scarlet, are
all and only the member of a conspiracy to blow up police headquarters. Black sincerely
utters (76) to detective Grey as they are discussing the conspirators.
In Switched Innocence, the report made by Black intuitively is unfounded and false,
takingaloadedreadingwithrespectto‘innocent’. InSwitched Innocence,intuitively,Plum
has neither loaded nor permissive beliefs towards the proposition that most conspirators
are innocent.
Now consider a permissive inconstant report from chapter 2:
Debate: There is a regular debate competition at the manor between two teams, the
Stars and the Stripes. At each competition, di↵ erent individuals are elected to be the
captains of the Stars and of the Stripes. Green tells detective Black that he believes that
whenever there is a debate, the captain of the Stripes team (whoever it is that week)
makes the best argument. Detectives Black and Grey know something Green doesn’t:
each competition, the captain of the Stripes (whoever it is that week) is bribed to throw
that competition. The detectives know that Green is unaware that anyone is ever bribed
to throw any competition. In discussing the regular debate competition, Black assertively
utters the following to Grey:
(77) Green believes that whenever there is a debate, the debater bribed to throw the
competition makes the best argument.
Asnotedbefore, thereportmadewith(77)inthiscontextintuitivelyispermissivewith
respect to ‘debater bribed to throw the competition’.
Now consider a switched version of this context:
Switched Debate: There is a regular debate competition at the manor between two
teams, the Stars and the Stripes. At each competition, di↵ erent individuals are elected
to be the captains of the Stars and of the Stripes. The detectives together assume that
wheneverthereisadebate, thecaptainoftheStripesteam(whoeveritisthatweek)makes
the best argument. From interrogating Green, Black knows that Green believes that for
each competition, the captain of the Stripes (whoever it is that week) is bribed to throw
that week’s competition.In discussing the regular debate competition, Black assertively
utters (77) to Grey.
In Switched Debate, the assertion Black makes with (77) is intuitively unwarranted and
false, taking a loaded reading with respect to ‘makes the best argument’. Again we find
158
that switching which propositions are contextually assumed to be true with those believed
by the subject of a report can flip our intuitions on whether a single report sentence takes
a true permissive reading or a false loaded reading. The question-sensitivity of permissive
readings is not peculiar to universal instantiation reports, but seems to be a feature of
permissive reports in general.
A.4. Bolstering Observations about Switched Contexts
The observations I made in §5 about switching around the contents of the subject’s
loaded attitudes and the conversationalist’s assumptions do not depend just on peculiar-
ities about the Switched Blackmail case. They hold in the other switched cases as well.
For example, contrast the natural way of interpreting the following sentences relative to
Switched Passage:
(72)Whitebelievesthatsomeroomwithasecretpassagehadthemurderer’sfootprints
in it.
(72V) White believes that some room with the murderer’s footprints in it has a secret
passage.
RecallthatinSwitchedPassage thedetectivesassumethatthemurdererwalkedthrough
everyroomintheeastwing,anditisWhitewhobelievesthatsomeroomwithasecretpas-
sage is in the east wing. So if either of (72) or (72V) serve to make true reports in Switched
Passage, those reports are permissive with respect to the phrases containing ‘murderer’.
In that context, the detectives were discussing how the murderer escaped from the scene
of the crime without being seen. As already discussed, (72) resists such a reading in this
scenario, but (72V) sounds much better.
Next consider the following report sentences relative to Switched Burglar:
(74) Peacock believes that the cat burglar could hear Body scream.
(74V) Peacock believes that someone who could hear Body scream was the cat burglar.
RecallthatinSwitched Burglar thedetectivesassumethateverypersonintheballroom
could hear Body scream, and it is Peacock who believes that the cat burglar was in the
ballroom. Given this background, (74V) is more easily interpreted as true permissive
reports than is (74). This is especially so if we suppose that the detectives are concerned
with resolving a question which Grey might ask by uttering:
(Q9) Who in particular could hear Body scream?
So again, it looks like one thing controlling whether a sentence can serve to make a true
permissive report is what kinds of questions which could be under discussion in given the
assumptions made by the conversationalists.
Finally reconsider Switched Safe, in which the detectives assume that every weapon in
159
the safe was used to commit a crime and Plum has dusted every weapon in the manor for
prints, so has a loaded belief that every weapon with Green’s prints on it is in the safe.
(75) Plum thinks that every weapon with Green’s prints on them was used to commit
a crime.
(75V) Plum thinks that only weapons which were used to commit a crime have Green’s
prints on them.
Truth value judgments seem to pattern here as in the three previous cases. Plum’s
being unaware in Switched Safe that any crime was committed makes (75) di cult to hear
as a true permissive report. But (75V) more readily takes such a reading.
Next I turn to other examples that bolster the observations made in §5.2 about com-
plicating switched contexts. Consider the following modification of Switched Passage:
Switched Passage
2
: The detectives have found muddy tracks leading through every
room in the east wing. These match the boots which the detectives take to be the mur-
derer’s, so the detectives commonly assume that the murderer walked through every room
in the east wing. They also have discovered muddy tracks leading through the west wing;
these match the shoes which they take to be the cat burglar’s. By questioning White,
Black learns that White has been told by the butler that there is a secret passage leading
from a room in the east wing. So Black knows that White believes that some room with
a secret passage is in the east wing. The detectives know, however, that White is unaware
that there is a murderer or a cat burglar on the loose. The detectives are exchanging
information, trying to figure out whether the murderer or the cat burglar used a secret
passage. Black assertively utters one of the following to Grey:
(72) White believes that the murderer walked through a room with a secret passage.
Thereportmadeusing(72)inSwitched Passage
2
moreeasilytakesapermissivereading
than does the original assertion made in Switched Passage. This is especially so if we add
an extra detail that the detectives assume that either the burglar or the murderer walked
through a room with a secret passage. If the detectives are in the course of trying to
figure out which of the two did so, then (72) is an appropriate way to report White’s
attitudes, even though White is unacquainted with the murderer and is unaware there was
any murder.
Next consider a complicated version of Switched Burglar in which a report can be
permissivewithrespecttomaterialintheverbphraseofthereportsentence’scomplement:
Switched Burglar
2
: The detectives know that both Mr. Body and Ms. Mind were
murdered at the same time last night. They have worked out among themselves that
whoever who was in the ballroom or the kitchen at midnight last night was in a position to
hear Body scream. They have also worked out that whoever was in the hall or the study
160
at midnight was in a position to hear Mind scream. The detectives know that Peacock
is unaware of all of this; in particular, Black and Grey commonly know that Peacock is
unacquainted with Body and is unaware that there were any murder-induced screams last
night. From investigating Peacock, Black comes to learn that she believes that the cat
burglar was in the ballroom at midnight. The detectives are exchanging information: they
know the cat burglar didn’t steal anything last night, and now are considering whether
this was because because she heard Body scream or because she heard Mind scream. Black
assertively utters the following to Grey:
(74) Peacock believes that the cat burglar could hear Body scream.
UnliketheoriginalSwitchedBurglar,inthiscontext(74)morereadilytakesapermissive
reading with respect to ‘could hear Body scream’. The main di↵ erence here, again, is the
introduction of additional contextual assumptions which make new questions available.
‘The cat burglar could hear Body scream’ is a fine way to answer the question:
(Q10) Could the cat burglar hear Body scream or could she hear Mind scream?
And (Q10) is available in Switched Burglar
2
but not Switched Burglar. In the original
Switched Burglar case, furthermore, it was not laid down that the detectives commonly
assumed that there was a cat burglar. Instead, they were simply interested in the question
of who in particular could hear Body scream. Following the discussion from §4.1 above,
although ‘The cat burglar could hear Body scream’ is an appropriate enough way of an-
swering this question, there are more appropriate alternatives, such as ‘One person who
could hear Body scream was the cat burglar’, or ‘The cat burglar could’, or simply ‘The
cat burglar’.
161
Chapter 6. Acquaintance Sensitivity
The pragmatic question of permissiveness asks: under what conditions does a use of
an attitude report sentence serve to make a permissive report? In the previous chap-
ter I considered principles like the following partial answer to the pragmatic question of
permissiveness:
Direct QUD Sensitivity:
An attitude report sentence R of the formaVsthatS(F) serves to make a
report which is permissive in context c with respect to F in c only if ?x[S(x)] is
not directly addressed by R in c.
Direct QUD Sensitivity states a necessary condition on whether a report takes a per-
missive reading in terms of what question the speaker is addressing in giving the report.
In this chapter I show that although something like Direct QUD Sensitivity (as appro-
priately supplemented by other pragmatic principles) seems to capture an important part
of the pragmatics of permissive reporting, such a condition cannot provide us with a full
answer to the pragmatic question of permissiveness (even when supplemented by certain
other pragmatic principles or taken as a rough and defeasible generalization). That is, the
following principle stating necessary and su cient conditions on permissiveness is false:
Full Direct QUD Sensitivity:
An attitude report sentence R of the formAVsthatS(F) serves to make a report
which is permissive in context c with respect to a concept c-associated with F if
and only if ?x[S(x)] is not directly addressed by R in c.
Even if we ignore problems with Direct QUD Sensitivity raised in chapter 5, focusing
just on the cases in which Direct QUD Sensitivity works well, we still can show that Full
Direct QUD Sensitivity is false due to a phenomenon encountered in chapter 3, which I
called the “problem of acquaintance” for contextual augmentation views.
The problem is, roughly speaking, that when it is not conversationally assumed that
the subject of a report is unacquainted with what is designated by F,thenitisdi cult
for that report to be permissive with respect to F. This phenomenon not only shows
that simple contextual augmentation views are incorrect, but further shows that what
questions a speaker directly addresses with a report (even in circumstances where Direct
QUD Sensitivity otherwise does well) does not by itself determine whether that report
is permissive, contrary to Full Direct QUD Sensitivity. Given complications addressed
in §5 of chapter 5, this should be expected; assumptions about how a subject cognizes
certain things plays a crucial role in determining whether a report takes a permissive
reading with respect to certain linguistic occurrences, their propositional contributions,
and conversationally associated concepts. Attending to whether a report is relevant to
conversational goals is not enough.
I begin in the next section by presenting the problem of acquaintance, illustrating how
it tells against Full Direct QUD Sensitivity.Thenin §2-§4 I critically evaluate further
162
conditions one might adopt in order to avoid the problem of acquaintance. In §5 and
§6 I return to connections between the pragmatics of permissive reports and the truth-
conditional problem of permissiveness, picking up open problems from the end of chapter
5.
1. The Problem of Acquaintance
Full Direct QUD Sensitivity cannot be correct, since, as we’ve seen several times, when
it is not conversationally assumed that the subject of a report doesn’t “think in terms of”
F (lacks a concept which is conversationally associated with F, or fails to be disposed
to use in certain ways), then it is di cult for that report to be permissive with respect
to F. In chapter 3 I called this the “problem of acquaintance”.
The contextual augmentation theories discussed in chapter 3 fall afoul of the problem
ofacquaintance, sincetheyaccountforwhetherareportispermissivejustintermsofwhat
the subject’s fully loaded attitude contents contextually entail. In a context c in which
the objects of a subject’s loaded attitudes contextually entail a proposition designated by
the complement of a report sentence R, but the conversationalists in c do not assume that
the subject is unfamiliar with something designated by a linguistic constituent of that
complement, simple contextual augmentation theories predict a permissive reading of the
report made by a use of R even though such a reading is hard to access or is altogether
unavailable.
Reconsider the example of this problem given in chapter 3:
Blackmail
2
: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called into discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. The
detectives know that Green is well acquainted with Scarlet: Green and Scarlet are close
friends. ByquestioningGreen,DetectiveBlacklearnsthatGreenthinksthateverysocialist
guest was being blackmailed by Mr. Body. The detectives are exchanging information,
trying to find out who had a motive for killing Body. Black assertively utters the following
to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
In this context it is di cult to interpret the report made with (2) as a permissive
with respect to the target expression ‘Scarlet’, since it is conversationally assumed that
Green is well acquainted with Scarlet. This was a problem for contextual augmentation
theories, since the proposition contributed by the complement of (2) is distinctively non-
trivially contextually entailed by the object of Green’s loaded thought and the common
assumptions in the context.
FullDirectQUDSensitivity facesasimilarprobleminaccountingforthiscase. Foreven
if Black is directly addressing question ?x[Scarlet was x] in uttering (2), the assumption of
Green’s being acquainted with Scarlet apparently blocks a permissive reading of (2). To
163
see this, consider a context just like Blackmail
2
in which Grey asks a question of Black
with
(Q2) Who was blackmailing Scarlet?
Inresponse, Blackutters(2). Thereporttherebymadestillseemstoresistapermissive
reading with respect to ‘Scarlet’ in such a context, even though, given (Q2), intuitively
BlackpresupposesthatScarletwasbeingblackmailedbysomeone, andsodoesnot address
the question ?x[x was being blackmailed by Mr. Body]. Given the assumption that Green
is well acquainted with Scarlet, the preferred reading is a loaded one on which Black
represents Green as thinking in terms of ‘Scarlet’ (or deploys a certain conversationally
associated concept of Scarlet in believing that Scarlet was being blackmailed by Body).
This suggests that whether the speaker is directly addressing the question ?x[S(x)]
in uttering a sentence of the formAVsthatS(F) does not by itself determine whether
the report thereby made is permissive with respect to F, contrary to Full Direct QUD
Sensitivity.
This problem is also raised by a version of Marbles:
Marbles: The detectives know that there is an urn in the library; they know that there
are exactly ten red marbles in in the urn, but are not sure how many marbles total are in
the urn. Detective Black has just spoken to professor Plum, who told Black that he thinks
there are exactly five blue marbles in the urn. The detectives are discussing how many
marbles are in the urn, and Black sincerely utters the following to Grey:
(50) Plum thinks that there are (exactly) fifteen marbles in the urn.
With this minimal setup, it is quite di cult to hear (50) as making either a true or
permissive report in Marbles. This was a problem for contextual augmentation theories,
since the proposition designated by the complement of (50) is distinctively non-trivially
contextually entailed in Marbles by a loaded thought of Plum’s. Notice that in this case
there is no conversational assumption that Plum is unfamiliar with the relevant marbles
or the relevant urn. Indeed: what Black knows about Plum’s loaded beliefs requires that
Plum be disposed to think of things as marbles and urns.
Suppose that Grey asks a question of Black in such a context with
(Q11) How many marbles are in the urn?
In response, Black utters:
(50) Plum thinks that there are (exactly) fifteen marbles in the urn.
This report fails to be permissive with respect to both ‘marbles’ and ‘in the urn’. This
is so despite the fact that the question ?x[There are x marbles in the urn] is clearly under
discussionandthequestion?F[TherearefifteenFs]isnot directlyaddressedbyBlack’suse
of(50). Giventhis, Full Direct QUD Sensitivity predictsthat(50)shouldtakeapermissive
164
readingwithrespectto‘marblesintheurn’. But(50)doesn’ttakesuchareading. Soagain
we see that, despite the explanatory success of question-sensitive views, a fully adequate
answer to the pragmatic question of permissiveness must make reference to more than just
what question is or isn’t addressed by the speaker of an attitude report.
This shouldn’t come as a surprise on the sensitive attitudes view; on this view, attitude
verbs can contribute di↵ erent relations as used in di↵ erent conversational circumstances.
We should expect that, like all cases of interpretation in the presence of ambiguity or se-
mantic underspecification, which of multiple possible interpretations is most appropriate
is subject to a wide array of pragmatic factors. In general, the process of interpretation in-
volves recovering the speaker’s intended meaning on the basis of what they utter together
with available facts about why they uttered what they did. This process of recovery,
of course, is informed by the apparent conversational goals which their speech serves to
address,butitalsomaybeinformedbycertaincontextualassumptions. Fromthisperspec-
tive, we should expect that any principle like Full Direct QUD Sensitivity that just focuses
on conversational goals will fall as a fully adequate answer to the pragmatic question of
permissiveness.
To account for the problem of acquaintance, then, we at least need to adopt a further
necessaryconditiononpermissivenessinadditiontoFullDirectQUDSensitivity. Anatural
first proposal is something like
Non-Acquaintance:
An attitude report R sentence of the formAVsthatS(F) serves to make a report
which is permissive in context c with respect to phrase F occurring in S only if it
is conversationally assumed in c that A is not acquainted with what F designates.
Accepting this principle immediately rules out any prediction that (2) or (50) should
takepermissivereadingsinBlackmail
2
orMarbles. Thisprincipleisalsothesortofprinciple
which was assumed in §5 of chapter 5.
Combining Non-Acquaintance together with the original Direct QUD Sensitivity yields
a revised version of Full Direct QUD Sensitivity, which I’ll call “Non-Acquaintance QUD
Sensitivity”, or:
NAQS:
An attitude report R sentence of the formAVsthatS(F) serves to make a report
which is permissive in context c with respect to phrase F if and only if it is
conversationallyassumedincthataisnotacquaintedwithwhatF designatesand
165
?x[S(x)] is not directly addressed by R in c.
119
The problem with NAQS is that in ruling out permissive readings in cases like Black-
mail
2
and Marbles, it also rules out permissive readings for cases in which permissive
readings are possible. In the next section, I illustrate this problem and begin to develop
better ways of capturing the guiding idea behind Non-Acquaintance.
2. The Problem with “Acquaintance”
Thetroublesstartwhenwetrytosaywhatexactlyismeantby“acquainted”in NAQS.
Itisdi culttoexplicatethisnotionwithouteitherinaccuratelyrulingoutpermissiveread-
ings in some cases or rendering the notion of acquaintance theoretically unilluminating. In
response to this di culty, I consider two possible replacements of NAQS, which attempt
in di↵ erent ways to capture the basic idea behind our initial intuitive notion of “acquain-
tance”. I call them “Inaccurate Thinking Question Sensitivity” and “Implausibly Loaded
Question Sensitivity”. Once these two regimentations of the guiding idea behind NAQS
havebeenpresented, Ispendmostoftherestofthechapterconsideringcomplicationswith
each and attempting to evaluate whether one or the other does a better job at providing a
partial answer to the pragmatic question of permissiveness.
How are we to understand the notion of “acquainted with” which figures in NAQS?
A first thought is to understand “acquainted with” just as “has a concept of”. But this
won’t work. Consider another case involving paintings:
Stolen
1
: The detectives know that Peacock is responsible for curating the gallery; she
hates art theft and so assiduously tries to make sure that no paintings in the gallery
are stolen works. The detectives know, however, that every impressionist painting in the
gallery is a stolen painting and that every stolen painting in the gallery is an impressionist
painting. Black has spoken to Peacock and so has learned that Peacock thinks that every
impressionist painting in the gallery is taking up too much wall space. Black sincerely
utters the following to Grey:
(78) Peacock thinks that every stolen painting is taking up too much wall space.
IntuitivelythereportwhichBlackmakeswith(78)inStolen
1
ispermissivewithrespect
to ‘stolen painting’. But in Stolen
1
it is common knowledge that Peacock has a concept of
stolen-painting: Peacock is known to work hard to assure there are no stolen paintings
in the gallery (cognizing the property stolen-painting via the relevant conversationally
119
Throughout this chapter, to get clear on how best to develop pragmatic acquaintance constraints on
permissive readings, I simply assume that Direct QUD Sensitivity is correct as stated. Of course, my final
view is that this principle just reflects a part of our general practices of interpretation in cases of permissive
reporting. Indeed, I expect even the final versions of the acquaintance-based principles I consider here at
best will be defeasible, rough generalizations, to be situated in a broader pragmatic picture. But it helps
us to see the many and various ways that di↵ erent conversational factors bear on permissive reporting by
approaching these issues with tidy pragmatic principles.
166
associated concept stolen-painting). So we have a report which is permissive with respect
to a phrase F even though it is not conversationally assumed that the subject lacks a
concept of the semantic contribution of F. So the relevant notion of “acquainted” in
Non-Acquaintance cannot simply be “has a concept of”.
In Stolen
1
, Peacock does not herself assume that there are any stolen paintings (in the
gallery). But even if she is conversationally known to make such an assumption, still a
report made with (78) can be permissive with respect to ‘stolen painting’. Consider
Stolen
2
: Thedetectivesknowthatallandonlytheimpressionistpaintingsinthegallery
are stolen paintings in the gallery. They also know that on every wall in the gallery there
are both impressionist and non-impressionist paintings. Finally, they know that Peacock
erroneously thinksthat every paintingon thewestwallof thegallery (impressionist ornot)
is a stolen painting. Black has just spoken with Peacock and thereby has come to know
that she also thinks that every impressionist painting in the gallery is taking up too much
wall space. Grey asks a question of Black with:
(Q12) Is Peacock going to keep the stolen paintings in the gallery?
In response, Black sincerely utters
(78) Peacock thinks that every stolen painting is taking up too much wall space.
Although the detectives commonly assume that Peacock has a false belief involving
the concept stolen-painting, hence possesses that concept, intuitively the report made with
(78) in Stolen
2
is permissive with respect to ‘stolen paintings’.
120
WhatevernotionofacquaintanceisinvolvedinNon-Acquaintance,wewantittohelpus
rule out certain permissive readings for (50) in Marbles without forcing us to rule out such
a reading in Stolen
1
.In Marbles, the subject is not assumed to lack a concept of properties
designatedbythetargetphrases‘marbles’and‘intheurn’;similarly,inStolen
1
,thesubject
is not assumed to lack a concept of the property designated by the target phrase ‘stolen
painting’. Despite this, the report in Marbles is loaded with respect to the target phrases
‘marbles’ and ‘urn’, whereas the reports in Stolen
1
is permissive with respect to the target
phrase ‘stolen painting’.
121
Nevertheless, it does feel that there is some sense in which
Peacock is not assumed to be “thinking in terms of” ‘stolen painting’ in Stolen
1
but on
which Plum is assumed to be thinking in terms of ‘marbles’ and ‘urn’ in Marbles.
What is the theoretically relevant di↵ erence between Marbles and Stolen
1
, here? Al-
though the subjects in both Marbles and Stolen
1
are not assumed to lack a concept asso-
ciated with the relevant target phrase, one di↵ erence between these two contexts is that in
120
The use of a definite description here isn’t necessary for a permissive reading in such a case. For
instance, consider a version of the case in which ‘the’ is replaced with ‘ten’, or a version in which ‘the stolen
paintings’ is replaced with ‘every stolen painting on the east wall’.
121
The key di↵ erence between these cases isn’t just that the target phrase is at issue in Marbles but is not
at issue in Stolen; given the QUDs in both contexts, neither target phrase is at issue.
167
onebutnottheotherthesubjectisassumedtomisapply thatconcept. InMarbles,Plumis
not only assumed to have a marbles concept, but intuitively the detectives assume that he
applies (or would be disposed to apply) that concept to roughly the same objects as those
to which the detectives apply their marbles concept. By contrast, in Stolen
1
and Stolen
2
,
although Peacock is assumed to have a stolen-paintings concept, she is not assumed to be
disposed to apply that concept to the same objects as those to which the detectives are
disposed to apply their stolen paintings concept. Rather, in the Stolen examples, Peacock
is assumed not to apply the concept stolen paintings to some thing x such that the detec-
tivesdoapplystolen paintings tox. Thissuggestsadi↵ erentexplicationof“acquaintance”
that might track a theoretically relevant di↵ erence between these cases, which we might
capture with the following principle:
Inaccurate Thinking:
An attitude report R sentence of the formAVsthatS(F) serves to make a report
which is permissive in context c with respect to phrase F occurring in S only if
thespeakerassumesincthatAdoesnotaccuratelyapplytheconcept associated
with F in c.
The rough idea here is that a report of the formAVsthatS(F) is permissive with
respect to F in context c only if it is commonly assumed in c that A does not think
of the things which are commonly thought of as F by the conversationalists in c as F.
Inaccurate Thinking is phrased in terms of the concept “associated” with a target phrase
in the context; this “associated” concept, again, intuitively, is that with respect to which
the given report is permissive. In the Stolen examples we can take the concept associated
with ‘stolen painting’ to be the concept of stolen-painting-in-the-gallery.
What is it to ‘accurately apply’ a concept? The way I am thinking of it, one fails to
accurately apply the concept chair in some instance if one conceives of something which is
not a chair as a chair (as does one who exclaims ‘damn chair!’ after bumping into a co↵ ee
table in the dark) or if one conceives of a chair as not a chair (as does one who exclaims
‘damn co↵ ee table!’ after bumping into a chair in the dark).
ReplacingNon-Acquaintance withInaccurate Thinking seemslikeanimprovement. Let
me make two clarificatory comments about Inaccurate Thinking. First, notice that Inaccu-
rate Thinking doesn’tclaimthatareportispermissivewithrespecttoF onlyifthesubject
in fact inaccurately applies a concept conventionally associated withF. It just claims that
the subject is conversationally assumed to do so. Second, notice that examples in which
the subject is assumed to lack a concept conventionally or conversationally associated with
a target phrase will typically be cases in which the subject is assumed not to accurately
apply that concept (as long as the conversationalists apply the concept to something; if
even to something which merely possibly exists). For in lacking the concept altogether,
the subject will fail to apply that concept to whatever the conversationalists do.
By combining Inaccurate Thinking together with Direct QUD Sensitivity, we obtain
the following principle:
168
ITQS:
A use of a sentence R of the formAVsthatS(F) serves to make a report which
is permissive in context c with respect to phrase F occurring in S if and only if it
is conversationally assumed in c that A does not accurately apply the concept conversationally associated with F and ?x[S(x)] is not directly addressed by R in
c.
This principle seems to handle Marbles and the Stolen examples, which involve reports
which are evaluated for permissiveness with respect to a property-designating phrase F,
such as ‘marble’ or ‘forgery’. Since in Marbles it is not conversationally assumed that
Plum does not accurately apply the concept marble, ITQS predicts that (50) is loaded
with respect to ‘marbles’. But since in the Stolen examples Peacock is conversationally
assumed not to accurately apply the concept stolen paintings in the gallery, ITQS predicts
that (78) is loaded with respect to ‘stolen painting’ in those contexts.
ITQS replaces the notion of conceptual acquaintance with a notion of concept misap-
plication. Note again that typical contexts in which a subject is assumed to lack a target
concept are de facto contexts in which the subject is assumed not to accurately apply the
target concept. So cases in which Non-Acquaintance correctly rules out as loaded will also
be ruled out as loaded by ITQS. But since there are cases in which a subject is assumed
not to accurately apply a target concept but in which it isn’t assumed that the subject
lacks the target concept, there will be some cases, like the Stolen contexts, in which ITQS
predicts a permissive reading where NAQS incorrectly predicts a loaded reading. So ITQS
rightly rules out the cases which NAQS rightly rules out without wrongly ruling out some
cases which NAQS wrongly rules out.
Why should the conversational assumption that the subject does not accurately apply
the target concept give rise to a permissive reading? Here’s one explanation. The con-
versationalists’ goals in the Stolen contexts are given by a question concerning how things
stand with respect to what the detectives conceive of as stolen paintings. Given this, for
a fully loaded report made with (78) to help the conversationalists answer the QUD in
their context, Peacock must accurately apply ‘stolen paintings’ with respect to the stolen
paintings referenced by the detectives’ QUD. Since it’s conversationally assumed that Pea-
cock inaccurately applies the concept stolen painting in the gallery, her having a loaded
belief involving that concept provides poor evidence in favor of either answer to (Q12). For
detective Grey to learn that Peacock has a fully loaded belief that every stolen painting
is taking up too much space would do little to increase his confidence that Peacock isn’t
going to keep what the detectives take to be stolen paintings in the gallery.
Incontrast,tolearnthatPeacockhasabeliefwhichispermissivewithrespectto‘stolen
painting’ provides much better evidence in favor of a negative answer to (Q12). Roughly
speaking, if Grey recognizes that Black issues a permissive report using (78), Grey is in
a position to know that Peacock has a fully loaded belief which contextually entails a
proposition concerning what the detectives together take to be stolen paintings. This at
169
least is something which contextual augmentation views seem to get right. Grey is in a
position to know that for some G such that Black knows a proposition assertible by a
sentence of the form every G is a stolen painting in the gallery, Peacock has a fully loaded
belief assertible by a sentence of the form every G is taking up too much wall space.This
providesevidencethat Peacock isn’t going to keep what thedetectivesconceive of asstolen
paintings in the gallery.
I’ve just given a general pragmatic explanation of why a report resists a loaded reading
with respect to a phrase F if it conversationally assumed that the subject does not accu-
rately apply the concept contextually associated with F. The proposition which would be
asserted if the report were loaded with respect to F is less relevant to QUDs in the con-
text than is the proposition which would be asserted if the report were permissive. This
explanation suggests a more general necessary constraint which we could adopt in place of
either Non-Acquaintance or Inaccurate Thinking:
Implausibly Loaded:
An attitude report sentence R of the formAVsthatS(F) serves to make a report
which is permissive in context c with respect to phrase F occurring in S only if,
given what the speaker is commonly assumed to believe, no assertion candidate of
R which is loaded with respect to F is both true and relevant to the QUDs in c.
This principle is stated in terms of the “assertion candidates” of an attitude report
sentence. We’ve seen this lingo before, but for clarity, let me say that as I am using
this phrase the “assertion candidates” of a sentence S are the propositions which a normal
sincereuseofS canservetoassertinsomecontext. ForanyreportsentenceRwhichcanbe
used to make either a permissive report or a loaded report, there will be multiple assertion
candidates of R. When we evaluate whether in some context a particular sentence takes
a permissive or loaded reading with respect to some phrase, we are trying to determine
which assertion candidate of that sentence is in fact asserted in the context in question.
Implausibly Loaded does as well as Inaccurate Thinking in ruling out a permissive
reading in Marbles while allowing permissive readings in the Stolen examples. In Marbles,
it isn’t the case that given what Black is assumed to believe that no assertion candidate of
(50) which is loaded with respect to ‘marbles’ is true and relevant to (Q11). Since Plum
is assumed to accurately apply the concept marbles, that Plum has a loaded belief that
there are (exactly) fifteen marbles in the urn does provide evidence in favor of answers
to (Q11). So the loaded assertion candidate of (50) is indeed relevant to the QUD in the
context Marbles. So the permissive reading is ruled out by Implausibly Loaded (and with
QUD Sensitivity, the loaded reading of (50) is in fact predicted, which is what we want).
In contrast, in both Stolen examples, since Peacock is assumed not to accurately apply
the concept stolen paintings in the gallery, associated with ‘stolen paintings’, the loaded
assertion candidate of (78) is not evidentially relevant to the QUDs in the Stolen contexts.
So a permissive reading of (78) is allowed, which is what we want.
Before moving on, I’d like to point out certain features of Implausibly Loaded.First,
170
Implausibly Loaded is stated in terms of QUDs, so this principle is connected to the same
conversationalgoalsthatDirect QUD Sensitivity is. SothepicturewegetwithImplausibly
Loaded is, superficially at least, more theoretically unified with other plausible pragmatic
principles of permissiveness than is Inaccurate Thinking.
Second, notice that just as ITQS correctly rules out the permissive readings which
NAQS correctly rules out, the combination of Implausibly Loaded and QUD Sensitivity
seems to correctly rule out the permissive readings which are correctly ruled out by ITQS.
For when it is conversationally assumed that the subject does not accurately apply the
targetconcept, aloadedreportinvolvingthatsubjectandtargetconceptwilltendtofailto
beasconversationallyrelevanttoaQUDthanthecorrespondingpermissivereport(unless,
of course, the QUD asks about the subject’s loaded attitudes). Furthermore, in principle
Implausibly Loaded may allow permissive readings where ITQS rules them out. For any
contexts in which a loaded report would not be relevant to the QUDs in the context but
in which the subject is not assumed not to accurately apply the target concept, Inaccurate
Thinking will rule out a permissive reading of the relevant report while Implausibly Loaded
will allow a permissive reading. This di↵ erence in what these proposed conditions predict
gives us a way to contrast the two.
Third, note that Implausibly Loaded is phrased in terms of “what the speaker is com-
monly assumed to believe”. Omitting this phrase would give us a di↵ erent necessary
condition on permissive readings: a report sentence R takes a permissive reading with
respect to F only if (in fact) no assertion candidate of R which is loaded with respect to
F is both true and relevant to the QUDs in c. Right away we should expect that this
alternative to Implausibly Loaded should make incorrect predictions, since it divorces the
process of interpretation, which is fundamentally a matter of recovering which assertion
candidate the speaker intends to commit herself to with her utterance, from whether a
particular assertion candidate is true and relevant to the conversation’s goals. Speakers
can of course be mistaken about what is true and relevant, so tying interpretation to the
issue of which assertion candidates are in fact true and relevant directs our attention as
interpreters to factors that may not track a speaker’s communicative intentions. This is
why it is important to include the phrase ‘given what the speaker is commonly assumed
to believe’.
Going forward I’d like to contrast ITQS specifically with
ILQS:
A use of a sentence R of the formAVsthatS(F) serves to make a report which
is permissive in context c with respect to phrase F occurring in S if and only if
given what the speaker is commonly assumed to believe, no assertion candidate of
R which is loaded with respect to F is both true and relevant to the QUDs in c
and ?x[S(x)] is not directly addressed by R in c.
In the next section, I’ll evaluate how well ITQS and ILQS do with respect to reports
which are permissive with respect to expressions other than predicates and bare noun
171
phrases. In the last few chapters we’ve seen several examples of permissive reports which
are permissive with respect to other expression types, such as names and determiners. I’ll
use such cases as a first test of the relative merits of ITQS and ILQS.
3. Extending the Principles to Non-Predicates: Determiners
Both Inaccurate Thinking and Implausibly Loaded e↵ ectively rule out examples of the
problem of acquaintance which involve predicates or bare noun phrases, such as ‘stolen
painting’. Now let’s see how they do for permissive readings of other expression types.
First reconsider an example of a report which is permissive with respect to a numerical
determiner:
Scotch: ColonelMustardissippingdi↵ erentbrandsofscotchandbourboninthelounge.
For each kind of alcohol he sips, he declares whether he thinks it is peaty. The detectives
commonly know that he has grown drunk, so he hasn’t kept count of how many scotches
he has had or how many scotches he thinks are peaty. Detective Black, however, has been
watching and keeping track of which kinds of scotch and bourbon Mustard declares to be
peaty. After Mustard has tasted all the scotch and bourbon in the lounge, Black consults
her record to find that Mustard has declared fifteen scotches to be peaty. Black assertively
utters the following to detective Grey:
(79) Mustard thinks that (exactly) fifteen scotches are peaty.
Detective Black asserts something true in uttering (79). She does so even though
Mustard is in no position to tell how many scotches he thinks are peaty. We can imagine
that, were Mustard asked about any particular scotch he tasted, he would reproduce his
judgment from before. But were he directly asked about how many scotches in the lounge
were peaty, he wouldn’t readily claim that fifteen are. For this reason I take the assertion
which Black makes with (79) in Scotch to be permissive with respect to ‘fifteen’.
Note that if we do take this report as permissive in this way, then NAQS,withits
notion of “acquaintance”, is particularly implausible. For it isn’t plausible that Mustard
lacksaconcept, say, ofwhat‘fifteen’servestodesignate. PresumablyMustardhasafifteen
concept. Forinstance, thepermissivenessof(79)doesnotrequirethatMustardfailtohave
any attitude towards the propositions, say, that he has tasted more than fifteen beverages,
that fifteen dollars are in his pocket, or that there are fifteen birds on the wire.
For this reason, ITQS does somewhat better with this example. The detectives com-
monly know that Mustard hasn’t kept count of how many scotches he has thinks are peaty.
Presumably, therefore, it is reasonable to take the detectives also to commonly know that
for no natural number n does Mustard have a fully loaded belief expressible by a clause of
the form exactly n scotches are peaty. Likewise, it is reasonable to take the detectives to
commonly assume that Mustard fails to accurately apply the concept conventionally asso-
ciated with ‘fifteen’. Given this, as long as in Scotch it is not at issue how many scotches
arepeaty, thenITQS predictsthat(79)takesapermissivereadingwithrespectto‘fifteen’.
172
ILQS similarly handles this example. Given the contextual assumption that Mustard
hasn’t kept count of how many scotches he has tasted or has declared to be peaty, given
what is conversationally assumed about what Black believes, no assertion candidate asso-
ciated with (79) which is loaded with respect to ‘fifteen’ is both true and conversationally
relevant. In particular, none of these candidates are true. Given what is conversationally
assumed, a reading of (79) which is loaded with respect to ‘fifteen’ is false. Thus Implau-
sibly Loaded does not rule out the reading which is permissive with respect to ‘fifteen’.
Here we see that it is important that Implausibly Loaded is phrased in terms of whether
certain loaded reports are both true and relevant. For apparently a proposition asserted
on a reading of (79) which is loaded with respect to ‘fifteen’ could be relevant to the con-
versationalists’ shared goals. But even if it were, it would not be an appropriate thing
to assert in the context, given that it is false given the conversational assumptions in the
context.
The style of account o↵ ered by ITQS of reports which are permissive with respect to
determinerssuggestsothercaseswhichputpressureonthenotionofinaccurateapplication
invoked by ITQS. To see this, first consider
Scotch
2
:Thesetupisjustlike Scotch save for two changes. First, the detectives com-
monly know that of the drinks in the lounge, there are exactly fifteen scotches and fifteen
bourbons. Second, they also commonly know that Mustard can’t reliably discriminate be-
tween scotch and bourbon: he just declares di↵ erent drinks as peaty, without concern for
whether those drinks are scotch or bourbon. Since Black know that Mustard has declared
fifteenscotchestobepeaty, Blackknowsthateveryscotchintheloungehasbeendescribed
as peaty by Mustard. Black assertively utters the following to detective Grey:
(80) Mustard thinks that every scotch is peaty.
Similarto(79)inScotch,intuitively(80)ispermissivewithrespectto‘every’inScotch
2
.
GivenITQS’sexplanationof(79)’spermissivenessinScotch, itisconversationallyassumed
that Mustard inaccurately applies ‘every’. This is reasonable, since the detectives take
Mustard to have lost track of how many bourbons and scotches he has tasted and not to
discriminate between scotch and bourbon. So Scotch
2
is no problem for ITQS.Butnow
consider
Scotch
3
: Colonel Mustard is sipping di↵ erent brands of scotch and bourbon in the
lounge. For each kind of alcohol he sips, he declares whether he thinks it is peaty. Unlike
Scotch and Scotch
2
, the detectives know that Mustard hasn’t swallowed any of the scotch
or bourbon, but has only tasted them and spat them back out. So throughout the tasting,
Mustard is stone cold sober. The detectives know that Mustard can reliably discriminate
between scotch and bourbon, and has been keeping a going count of how many scotches
and bourbons are peaty. After the tasting, Black consults her record to find that Mustard
has declared fifteen scotches to be peaty; in fact this corresponds to his running count,
though Grey hasn’t heard about Mustard’s final count. Finally: the detectives know that
173
Mustard both (i) has a loaded belief that there are fifteen marbles in an empty urn in the
library, and (ii) has a loaded belief that every guest in the manor is a democrat, where in
fact only half of the guests in the manor are democrats. Black assertively utters one of the
following to detective Grey:
(79) Mustard thinks that (exactly) fifteen scotches are peaty.
(80) Mustard thinks that every scotch is peaty.
Intuitively in Scotch
3
these reports are not permissive with respect to ‘fifteen’ and ‘ev-
ery’. ButITQS apparentlyappearscommittedtothesereportstakingpermissivereadings.
For the detectives commonly assume that Mustard inaccurately applies both the concepts
associated with ‘fifteen’ and ‘every’, given Mustard’s loaded beliefs about marbles and
democrats.
ILQS, however, is not committed to these incorrect predictions. For according to Im-
plausibly Loaded, (79) or (80) can take a permissive reading with respect to ‘fifteen’ or
‘every’ only if readings of (79) and (80) which are loaded with respect to these determin-
ers are not true and relevant to the conversationalists’ goals, given what the speaker is
assumed to believe. But since Black does not assume that Mustard does not accurately
applyafifteen orevery conceptwithrespecttothescotcheshehastasted, thepropositions
asserted when (79) and (80) take loaded readings, may be, given what Black is assumed
to believe, true and relevant. Accordingly, Implausibly Loaded predicts that (79) and (80)
take loaded readings with respect to ‘fifteen’ and ‘every’ in Scotch
3
.
The relative success of ILQS here is not yet reason to reject ITQS, or something like
it, outright. For we can adopt a revision of ITQS which restricts the range of objects over
which the subject is assumed to inaccurately apply the target concept. Just as NAQS is
incorrect for requiring that the subject be assumed not to possess the target concept per
se, ITQS is incorrect for requiring that the subject be assumed to inaccurately apply the
target concept per se. But this doesn’t mean that a principle similar to ITQS is doomed
to failure. To give something like ITQS a fair shake, we should consider how we might
restrict the range of objects over which the subject is assumed to inaccurately apply the
target concept.
Intuitively, we want such a restriction to relate just to the proposition ascribed to the
subject by the report which we are evaluating for permissiveness. In the case of Scotch,
for instance, we care about whether Mustard accurately applies the concept associated
with ‘fifteen’ with respect to the properties scotch-in-the-lounge and peaty.Todo
this, first let’s consider more closely what it is to apply concepts typically associated with
determiners.
I’llassumeastandardgeneralizedquantifierapproachtodeterminersaccordingtowhich
such expressions semantically contribute relations between properties. For example, con-
sider
174
(80C) Every scotch is peaty.
(79C) Exactly fifteen scotches are peaty.
The simple extensional treatment of (80C) and (79C) o↵ ered by the theory of general-
ized quantifiers calculates the truth-conditions for these sentences by taking ‘scotch’ and
‘peaty’ to contribute sets and ‘every’ to contribute a relation between sets, where such
relations are roughly regimented as functions from pairs of sets to truth values.
122
In
particular, ‘every’ and ‘exactly fifteen’ contribute relations regimented by
(8 ) F. G.F ✓ G
(=15) F. G.|F \ G| = 15
Conceiving of determiners as contributing relations, we have that ‘every’ contributes
the relation which holds between two properties F and G if and only if every instance of F
is an instance of G. likewise, ‘exactly fifteen’ contributes the relation which holds between
two properties F and G if and only if there are fifteen things which are instances of both
F and G. Given this, I assume that a concept conventionally associated with ‘fifteen’ or
‘every’ in English is a concept of a higher-order relation (of course, this is not to say that
anyone who possesses such concepts conceives of them as such).
Nowwewanttoknowwhatitistoaccuratelyorinaccuratelyapplyaconceptassociated
with a determiner. Before, I characterized what it is to fail to accurately apply a property-
designating concept in terms of what one thinks using that concept. One inaccurately
applies the concept chair if one uses that concept to predicate the property chair of an
object which lacks that property, or if one doesn’t use that concept to predicate chair
to an object which bears that property. Similarly, then, to inaccurately apply a concept
✏ which represents relation (8 ) with respect to properties F and G is to erroneously think
that F bears (8 )to G,where(8 )isrepresentedby ✏ within that thought, or to erroneously
think that F doesn’t bear (8 )to G,where(8 )isrepresentedby ✏ within that thought. That
is: to inaccurately apply (8 )withrespectto F and G is tantamount to erroneously thinking
that every F is G or erroneously thinking that not every F is G. Finally, one inaccurately
applies ✏ per se if there is some pair of properties with respect to which one inaccurately
applies ✏.
We now have a working characterization of what it is to inaccurately apply a concept
associatedwithadeterminerwithrespecttoapairofproperties. TohandleScotch
3
givena
principle like ITQS, I suggest that we revise ITQS so that a condition on permissiveness is
not just that it is conversationally assumed that the subject inaccurately applies concepts
associated with ‘fifteen’ or ‘every’ per se, but rather that it is conversationally assumed
that the subject inaccurately applies those concepts with respect to the properties involved
in the relevant report.
122
See Westerst˚ ahl & Westerst˚ ahl (2006).
175
IwillsaythatinassertingwhatBlackdoesinuttering(80),Blackpredicatesthehigher-
order relation (8 ) to the properties scotch and peaty. In doing so, Black applies the
concept ✏ to those properties. Accordingly, my suggestion is that Black’s report made with
(80) is permissive with respect to ‘every’ in Scotch
3
only if it is conversationally assumed
in Scotch
3
that Mustard doesn’t accurately apply the concept ✏ associated with ‘every’
with respect to the pair of properties of which Black predicates the property represented
by ✏ in uttering (80). That is: Black’s report is permissive with respect to ‘every’ only if
it is conversationally assumed that Mustard doesn’t accurately apply ✏ with respect to the
pair hscotch,peatyi.Sincein Scotch
3
this is not conversationally assumed, we correctly
predict that (80) is not permissive with respect to ‘every’ in Scotch
3
.
By contrast, since in Scotch
2
it is conversationally assumed that Mustard does not
accurately apply ✏ with respect to hscotch,peatyi, as long as how many scotches are
peaty is not at issue, we predict that (80) is permissive with respect to ‘every’ in Scotch
2
,
just as we observed.
I’ll express my proposed revision as a principle which I’ll call “Specific Inaccurate
Thinking”:
Specific Inaccurate Thinking (SIT):
An utterance of an attitude report sentence R of the formAVsthatS(F) by
Speaker S in context c serves to make a report which is permissive with respect
to expression e occurring in R only if it is conversationally assumed in c that A
does not accurately apply the concept conversationally associated with e with
respect to what S is conversationally assumed to apply to in making her report.
SIT states a necessary condition on permissiveness in terms of what the speaker is
conversationallyassumedtoapplyaconcepttoinmakingthereportshedoes. Forinstance,
suppose in a context c that Black utters
(80) Mustard thinks that every scotch is peaty.
In making this report, Black is conversationally assumed to apply the concept ✏ asso-
ciated with ‘every’ to the pair of properties hscotch,peatyi, since in uttering (6C) in c,
Black predicates (8 ) of that pair of properties. (In Scotch, it would be more accurate to
say that Black applies ✏ tohscotch-in-the-lounge,peatyi.) This fits with the proposed
explanation of the permissive reading of (80) in Scotch
2
and the loaded reading of (80) in
Scotch
3
.
How does SIT handle the Stolen examples, in which we evaluate reports for permis-
siveness with respect to the property-designating phrase ‘stolen painting’? Reconsider
(78) Peacock thinks that every stolen painting is taking up too much wall space.
Thissentencealonedoesnotdeterminewhatonewhouttersitappliestheconceptstolen
painting to. For ‘stolen painting’ occurs as the restrictor of a noun phrase which combines
176
with some verb phrase. This stands in contrast to an occurrence of ‘stolen painting’ in a
simple verb phrase, as in
(81) Starry Night is a stolen painting.
This sentence by itself (partly) determines what one who utters it would apply the con-
cept stolen painting to in uttering the sentence; namely, Starry Night. To perform an act
of assertion in uttering (81), one thereby performs the act of predicating stolen-painting
of Starry Night. This is unlike what is involved in performing the act of assertion in ut-
tering (78), since one can perform the latter act of assertion without thereby predicating
stolen-painting of any particular painting or of any paintings satisfying certain con-
straints.
In light of this, does it make sense to talk of what Black is conversationally assumed
to apply stolen painting to in uttering (78) in Stolen? I think it does. In each Stolen case,
there are objects (or presumed objects) to which Black is conversationally assumed to
apply the concept associated with ‘stolen painting’ in saying what she does in the context
that she does, even though this act of application may not be an essential part of the act
of asserting what she does in uttering (78).
Intuitively speaking, Black applies stolen painting to those paintings which she and
Grey commonly know to be stolen paintings in the gallery; namely, the impressionist
paintings in the gallery. She does not, in making an assertion by uttering (78), predicate
stolen-painting of any paintings in the Louvre. This is so, I suggest, even if it is
conversationally assumed that Black may in fact think some paintings in the Louvre are
stolen paintings. Notice further that this doesn’t require that the conversationalists be in a
position to identify which paintings in particular are stolen paintings. All that is required
is that the conversationalists believe that every x which is an impressionist painting is
somethingtowhichBlackwouldapplystolen painting; alternatively: theconversationalists
believe that there is some plurality X consisting of impressionist paintings such that Black
applies stolen painting to each element of X. This does not require that Black be able to
discriminate by visual means, say, which paintings are elements of X.
I suggest then that we think of SIT’s notion of what a speaker’s is assumed to apply
a concept to somewhat broadly, encompassing whatever that speaker represents herself as
applying that concept to in saying what she does, given the assumptions and goals of the
conversational context, regardless of whether an arbitrary auditor could ascertain that the
speaker predicates that property of those things just by being a competent English speaker
and hearing that the speaker utters what she does.
Ifweembracethisbroadconceptionofpredicationtargets,then,withrespecttocontext
Stolen
1
, we will take Black as predicating ‘stolen painting’ of all and only the impressionist
paintings in the gallery in saying what she does with (78) (even if the detectives commonly
know that there are stolen paintings elsewhere in the universe). Since the detectives com-
monly assume that Peacock does not accurately apply the concept stolen painting with
177
respect to the impressionist paintings in the gallery, a permissive reading is therefore pre-
dicted to be possible by SIT.
One last word about SIT before moving on. This principle says that it must be con-
versationally assumed that the subject does not accurately apply a certain concept with
respect to the things which the speaker is conversationally assumed to apply that concept
to in making her report. The way to read this is: it is conversationally assumed that
there are things X to which the speaker applies concept in making her report, and it
is conversationally assumed that the subject does not accurately apply with respect to
X.Theway not to read this is: there are things X which it is conversationally assumed
that the speaker applies , and it is assumed the subject does not accurately apply with
respect to X. On this latter way of understanding SIT, there must be certain objects in
existence which the speaker and the subject apply di↵ erently with respect to. This way
of understanding SIT, however, would face di culties in the following kind of example:
Stolen
3
: The detectives know that Peacock is responsible for curating the gallery; she
hates art theft and so assiduously tries to make sure that no paintings in the gallery are
stolen works. The detectives commonly assume that every impressionist painting in the
gallery is a stolen painting and that every stolen painting in the gallery is an impressionist
painting. Unbeknownst to them, however, there are neither impressionist paintings nor
stolen paintings in the gallery. So one of their conversational assumptions is false. Black
hasspokentoPeacockandsolearnedthatPeacockthinksthateveryimpressionistpainting
inthegalleryistakinguptoomuchwallspace. BlacksincerelyuttersthefollowingtoGrey:
(78) Peacock thinks that every stolen painting is taking up too much wall space.
Despite the detectives’ false conversational assumption, (78) still takes a permissive
reading with respect to ‘stolen painting’. This is so even though there are no objects to
which the speaker applies stolen painting and to which Peacock does not. So we cannot
understandSIT asclaimingthattherearethingsX towhichitisconversationallyassumed
that the speaker applies , and it is assumed the subject does not accurately apply with
respect to X. The other way of reading SIT, however, faces no problem with respect to
Stolen
3
. For in this context it is the case that it is conversationally assumed that there are
thingsX which the speaker applies stolen painting to but to which Peacock does not apply
stolen painting.So SIT, so understood, successfully predicts that a permissive reading is
possible for (78) in Stolen
3
.
Tofurtherevaluate SIT, consider another exampleof reportswhicharepermissivewith
respect to a property-designating phrase:
Stolen
4
: There are two galleries: the fancy gallery and the plain gallery. The detectives
knowthatineachgalleryallandonlytheimpressionistpaintingsarestolenpaintings. They
know that there are impressionist stolen paintings in each gallery. The detectives know
that Scarlet is responsible for curating the fancy gallery, but that she doesn’t care whether
178
the paintings are stolen; they know that she knows that all and only the impressionist
paintings in the fancy gallery are stolen. The detectives also know that Scarlet thinks that
no painting (including the impressionist paintings) in the plain gallery, which she does not
curate, is a stolen painting. So they commonly know that Scarlet erroneously believes the
impressionist paintings in the plain gallery aren’t stolen. Black has spoken to Scarlet and
so has learned that Scarlet thinks that every impressionist painting in the fancy gallery is
taking up too much wall space. Grey asks a question of Black with
(Q13) Is Scarlet going to keep the stolen paintings in the fancy gallery?
Black sincerely utters the following to Grey:
(82) Scarlet thinks that every stolen painting is taking up too much wall space.
In this context, intuitively (82) takes a loaded reading with respect to ‘stolen painting’.
ThisissodespitetheconversationalassumptioninStolen
4
thatScarletdoesnotaccurately
apply the concept stolen painting, given her erroneous assumption that no painting in the
plain gallery is stolen. Stolen
4
thus constitutes a counter-example to our earlier princi-
ple ITQS, since (82) is not permissive with respect to ‘stolen painting’ even though it is
commonly assumed that Scarlet inaccurately applies the concept associated with ‘stolen
painting’ and Black presupposes that there are stolen paintings in Stolen
4
.
OurworkingprincipleSIT,however,e↵ ectivelyrulesoutapermissivereadinginStolen
4
.
For, given our broad conception of what a speaker is assumed to apply a concept to in
making her report, the impressionist paintings in the fancy gallery, but not those in plain
gallery, are among the things to which Black is assumed to apply stolen painting in issuing
a report with (82). For in responding to (Q13) Black is applying ‘stolen painting’ just to
those stolen paintings in the fancy gallery; that is, to the impressionist paintings in the
fancy gallery. And the detectives do not commonly assume in Stolen
4
that Scarlet inaccu-
ratelyappliestheconceptassociatedwith‘stolenpainting’totheimpressionistpaintingsin
the fancy gallery. So SIT correctly predicts that (82) should not take a permissive reading
with respect to ‘stolen painting’.
At this point then, both SIT and Implausibly Loaded are viable extra necessary condi-
tions on permissiveness. Next let’s turn to other examples of reports which are permissive
with respect to non-predicates.
4. Extending the Principles to Non-Predicates: Names
At this point it’s worth considering how SIT and Implausibly Loaded, or something like
them,canhandlereportswhicharepermissivewithrespecttoclassicallydirectlyreferential
expressions, such as names. To this end, let’s reconsider an example with which we began:
Blackmail
2
: The detectives know that Scarlet, a guest at the manor, is a socialist. The
detectives know that Green is well acquainted with Scarlet: Green and Scarlet are close
179
friends. ByquestioningGreen,DetectiveBlacklearnsthatGreenthinksthateverysocialist
guest was being blackmailed by Mr. Body. The detectives are exchanging information,
trying to find out who had a motive for killing Body. Black assertively utters the following
to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
As noted above, (2) naturally takes a loaded reading with respect to ‘Scarlet’ in Black-
mail
2
. Intuitively this is so because the detectives commonly know that Green is well
acquainted with Scarlet. Accordingly, we’d like for some version of SIT, which is designed
to handle this kind of phenomenon, to be able to account for the failure of a permissive
reading in this case.
First consider what ILQS has to say about this case. To evaluate this case, we first
need to say what concept is associated with ‘Scarlet’ in this context. Presumably it is an
individualconcept whichservestorepresentScarletbycertainordinarymeans: meansof
cognizingScarletviathename‘Scarlet’, meansofcognizingviawaysshenormallypresents
visually, and so forth. So to determine what ILQS predicts we need to ask whether, given
what Black is contextually assumed to believe, any assertion candidate of (2) which is
loaded with respect to ‘Scarlet’ and is both true and relevant to the QUD in the context.
First we can note that it is contextually assumed that Green is familiar with Scarlet.
So given what is assumed about Black’s beliefs, an assertion candidate associated with (2)
whichisloadedwithrespectto‘Scarlet’, whichrequiresforitstruththatGreenpossessthe
concept of Scarlet, can be true. Further, were that assertion candidate true, given that
the conversationalists are trying to find out who had a motive for killing Body, it would
also be conversationally relevant. So the necessary conditions invoked by ILQS are not
satisfied,(itisnot conversationallyassumedthatno loadedcandidateistrueandrelevant).
So ILQS correctly predicts that Black’s use of (2) in Blackmail
2
takes a loaded reading
with respect to ‘Scarlet’.
Now let’s consider what SIT has to say about Blackmail
2
. To do so, first we need
to determine what concept is associated with ‘Scarlet’ in this context. As before, we can
safelyassumethatitisanindividualconcept whichservestorepresentScarletbycertain
ordinary means: means of cognizing Scarlet via the name ‘Scarlet’, means of cognizing by
how she normally presents visually, and so forth. Given the way that SIT is phrased, we
first need to spot ourselves the assumption that it is conversationally assumed that Green
applies this very concept in cognizing Scarlet. If we adopt this assumption, then we can
further say that it is not conversationally assumed that Green does not accurately apply
the concept with respect to what Black is assumed to apply to in making her report.
For Black is assumed to apply this concept to Scarlet, and it is also assumed that Green
applies this concept to Scarlet. Furthermore, there is no other thing x which it is assumed
that Black applies this concept to such that it is assumed that Green fails to apply to
x. So the necessary condition laid down by SIT is not satisfied, so according to SIT (2)
180
should not be able to take a permissive reading with respect to ‘Scarlet’, hence takes a
loaded reading.
Next consider
Blackmail: Detectives Black and Grey know that Scarlet is a socialist. The detectives
alsocommonlyknowthatGreen, anotherguestatthemanor, isunacquaintedwithScarlet.
Green has never met Scarlet nor is he competent with any name for Scarlet. Black knows
that Green thinks that every socialist guest was being blackmailed by Mr. Body. The
detectives are exchanging information, trying to find out who had a motive for killing
Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
As has been noted several times already, (2) can take a permissive reading with respect
to ‘Scarlet’ in Blackmail. By this, again, I mean that Black does not represent Green as
possessing and deploying any conversationally relevant concept of Scarlet.
ILQS readily explains this, since the conversationalists’ commonly assume that Green
is unacquainted with Scarlet. So given what Black is commonly assumed to believe, no
assertioncandidateof(2)whichisloadedwithrespectto‘Scarlet’istrue. Sotheconditions
laid down by ILQS are satisfied in this case (it is conversationally assumed that no loaded
assertions candidate is both true and relevant). So ILQS correctly predicts that (2) should
take a permissive reading with respect to ‘Scarlet’ in this context.
SIT similarly explains the permissive reading of (2) in Blackmail. Since it is conversa-
tionally assumed both that Green is unacquainted with Scarlet in this context (hence lacks
the ordinary concept of Scarlet, , and thereby fails to apply to Scarlet) and that Black
applies the concept to Scarlet in making her report, we can say that in this context it
is conversationally assumed that Green does not accurately apply with respect to what
Black applies to in making her report. So the necessary condition laid down by SIT is
satisfied, and so according to that principle (2) should be able to take a permissive reading
with respect to ‘Scarlet’ in Blackmail.
Something tricky about readings which are permissive with respect to names arises in
cases involving identity confusion. In such cases, an individual typically possesses multiple
individualconcepts,atleastoneofwhichtheindividualdoesnotaccuratelyapply. Consider
Shadow: Detectives Black and Grey know that White is well acquainted with Scarlet:
they are close friends, and White is competent with the name ‘Scarlet’ for Scarlet. One
darknight, thedetectivesaretailingWhite, whoistailingashadowyfigure. Thedetectives
know that this shadowy figure is Scarlet, but further know that White is not aware of this:
they know that White believes that Scarlet is asleep in bed. At some point, the detectives
losetrackofScarlet,soBlackapproachesWhitetoaskher‘Wheredidshego?’. Inresponse
White says, ‘She took the secret passage’. Black returns to Grey and assertively utters:
(83) White believes that Scarlet took the secret passage.
181
This report does not represent White as “thinking in terms of” ‘Scarlet’. It is also
natural to say that in Shadow (83) is not loaded with respect to ‘Scarlet’, if by this we
mean that White is not represented as having any particular fully loaded beliefs the having
of which involves deploying a concept which White usually uses to store information about
Scarlet, such as . In this sense it is tempting to characterize (83) as permissive with
respect to ‘Scarlet’ and .
Some delicacy is needed here, however. For the report made with (83) in Shadow
intuitively is not permissive with respect to Scarlet in quite the same way that the report
made with (2) in Blackmail is. In the latter case, Green is not represented as deploying
any concept of Scarlet: the report could be true even if Green altogether lacked a concept
of Scarlet. By contrast, in Shadow intuitively White is not represented as deploying a
particular concept of Scarlet, that she would, roughly speaking, associate with the name
‘Scarlet’. But in Shadow it is not the case that White is not represented as deploying any
concept whatsoever of Scarlet: she is represented as having a belief involving one concept
of Scarlet: the concept she has been using to store information about Scarlet by detecting
her shadowy figure. So it is too flatfooted to merely characterize (83) as permissive per
se; rather, in line with observations made in chapter 4, we should not only relativize
permissiveness to particular linguistic constituents of a report sentence’s clause, but we
should further relativize permissiveness to particular concepts which are conventionally or
conversationally associated with constituents of report sentences. Thinking in this way, we
can say that (83) is permissive with respect to ‘Scarlet’ relative to one concept of Scarlet
(such as ) but is loaded with respect to ‘Scarlet’ relative to another concept of Scarlet.
Insofar as we choose to extend the notion of permissiveness in this way to this case, we
should hope that something like ILQS or SIT can faithfully account for the conditions in
whichareportispermissivewithrespecttoaparticularexpressionerelativetoaparticular
concept associated with e.
Let’s begin with the reading of (83) which is permissive with respect to Scarlet relative
to White’s “ordinary” Scarlet concept, that concept which White intuitively associates
with the name ‘Scarlet’.
ToevaluateILQS andSIT,giventhiswayofthinkingofpermissivenessandloadedness,
wehavetodistinguishassertioncandidatesof(83)whichareloadedwithrespecttoWhite’s
ordinary Scarlet-concept (that concept associated with means of cognizing Scarlet via the
name ‘Scarlet’ and with ways that Scarlet presents perceptually in normal circumstances)
from those assertion candidates of (83) which are loaded with respect to White “shadow-
concept” of Scarlet, her concept which is associated with means of representing Scarlet
via the way she appears visually as she stalks through the shadows. So in characterizing
loaded readings, we cannot just describe a particular report as being loaded with respect
to a particular expression; to distinguish between the di↵ erent kinds of loaded readings,
we have to describe reports in terms being loaded with respect to a particular expression
182
and a particular concept.
123
With this distinction, we can distinguish four classes of
possible readings of (83): those which are loaded for both the ordinary-Scarlet concept and
the shadow-Scarlet concept, those which are loaded with respect to the ordinary-Scarlet
concept but not the shadow-Scarlet concept, those which are loaded with respect to the
shadow-Scarlet concept but not the ordinary-Shadow concept, and those which are loaded
with respect to neither.
Now we can understand ILQS as saying that a report is permissive with respect to an
expression e and a concept only if, given what the speaker is contextually assumed to
believe, no assertion candidate of the relevant report sentence which is loaded with respect
to e and is both true and relevant to the QUDs of the context. In Shadow,sinceit
is conversationally assumed that White had a belief that Scarlet is asleep in bed which
is loaded with respect to ‘Scarlet’ and her ordinary Scarlet-concept, given what Black is
assumedtobelieve, anyassertioncandidateof(83)whichisloadedwithrespectto‘Scarlet’
and White’s ordinary Scarlet-concept is not true, hence not true and relevant to the QUDs
in the context. So according to ILQS, (83) should take a permissive reading with respect
to the ordinary-Scarlet concept.
By contrast, it is not the case that given what Black is assumed to believe in Shadow,
some assertion candidate of (83) which is loaded with respect to ‘Scarlet’ and White’s
shadow-concept of Scarlet is not true. But it is the case that given what Black is assumed
to believe that some such assertion candidate is relevant to the conversational goals in
Shadow. The detectives are actively trying to resolve the question of where the shadowy
figure (whom they know to be Scarlet) is. Either loaded interpretation of (83) would be
relevant to this question, but only the interpretation which is loaded with respect to the
shadow-concept of Scarlet is true given the conversational assumptions about what White
believes. So the necessary conditions laid down by ILQS are satisfied in this case, and our
reinterpreted version of ILQS predicts that (83) should take a loaded reading with respect
to ‘Scarlet’ and White’s shadow-concept of Scarlet.
Now let’s consider how SIT handles this case. Something first to note is that applying
SIT to this example forces us to make the further assumption that distinct persons can
possess a single concept. For according to SIT, to determine whether (83) takes a read-
ing which is loaded with respect to ‘Scarlet’ and a particular concept requires that we
determine whether White applies concept to the same things which Black is conversa-
tionally assumed to apply to in making her report. In order for White and detective
Black to successfully apply to the same things demands that both possess the concept .
Specifically, when is White’s ordinary-Scarlet concept, by which she cognizes Scarlet via
the name ‘Scarlet’, we have to assume that Black possesses that very concept in order to
claim that (83) takes a loaded reading with respect to ‘Scarlet’ and that concept in some
context.
In the case of Shadow, we can safely assume that White has two distinct concepts of
123
This accords with observations made about conceptually laden dispositions in chapter 4.
183
Scarlet: herordinary-Scarletconceptandhershadow-Scarletconcept. FordetectiveBlack,
wecanthenmakeoneoftwochoices: first,thatshepossesstwodistinctconceptsofScarlet,
corresponding to White’s two concepts, or else that she possess one concept of Scarlet,
which involves both means of cognizing which are involved in White’s two concepts of
Scarlet. In the former case, we may assume that White’s ordinary-Scarlet concept and her
shadow-Scarlet concept are identical with Black’s ordinary-Scarlet concept and Blacks’s
shadow-concept, respectively. If, however, we take Black to possess a single concept of
Scarlet which she deploys in cognizing Scarlet both in “ordinary” ways or in “shadowy”
ways, then we would have to rephrase SIT explicitly in terms of means of cognizing rather
than in terms of concepts.
For now I will make the first choice: the detective possesses two distinct concepts of
Scarlet, an “ordinary” Scarlet concept and a shadow-concept. Even with this simplifying
assumption, SIT has a di cult time predicting the right permissive reading of (83) in
Shadow. Inorderfor(83)totakeapermissivereadingwithrespectto‘Scarlet’andWhite’s
ordinaryconceptofScarlet, ,itmustbethatitisconversationallyassumedincthatWhite
doesnotaccurateapply withrespecttowhatdetectiveBlackisconversationallyassumed
to apply to in making her report. In making her report, intuitively Black is assumed to
apply to Scarlet, whom the detectives know is the shadowy figure stalking through the
manor. But White also applies to Scarlet. That very concept is a constituent of a belief
of White’s which we can report with ‘White believes that Scarlet is asleep in bed’. If we
understand the notion of “not accurately applying” a concept “with respect to” certain
things simply as failing to apply the concept to the right objects or applying the concept to
the wrong objects, per se, then White is not assumed not to accurately apply in Shadow.
Still, intuitively there is some sense in which White does not accurately apply . For
White does not actively apply to Scarlet when White sees Scarlet stalking through the
manor. In short, we can say that White does not identify that which is picked out by
with that which is picked out by her shadow-concept of Scarlet. Detective Black, however,
does identify that which is picked out by with that which is picked out by the shadow-
concept of Scarlet. It is in this sense that White intuitively does not accurately use ,her
ordinary concept of Scarlet.
Let’s then understand what it is to apply a concept as involving two concepts, in the
following way. White applies her ordinary concept to the object of her concept, since
she identifies that picked out by with that picked out by . White does not apply her
ordinary concept to the object of her shadow-concept, since she does not identify that
picked out by with that picked out by her shadow-concept. One fails to accurately apply
a concept if one identifies that picked out by with that picked out by a concept ⇢ ,
though those things are distinct. Likewise, one fails to accurately apply a concept if one
does not identify that picked out by with that picked out by a concept ⇢ , though those
things are the same. I will say that one applies a concept to an object o if one identifies
that picked out by with that picked out by a concept ⇢ , and ⇢ in fact picks out o.In
the simplest case, if an individual concept picks out an object o, then one who possesses
184
this concept applies to o as long as they identify that which is picked out by with that
which is picked out by itself.
With this way of talking, we can say that White does apply her ordinary concept to
Scarlet even though White does not apply to Scarlet when she detects her via the way
she presents visually as she stalks through the manor at night. Since White fails to apply
to Scarlet as detected by her shadowy figure, whereas detective Black does apply to
Scarlet as detected by her shadowy figure, and furthermore Black is assumed to apply to
Scarlet so in making her report, White fails to accurately apply “with respect to what”
Black is assumed to apply in Black’s making her report. Since the detectives commonly
assume White is not accurately applying in this way, a suitably reinterpreted version of
the necessary condition of SIT is met, and so SIT allows that (83) may take a permissive
reading with respect to White’s ordinary concept of Scarlet.
At this point, then, we have two principles which capture the spirit behind Non-
Acquaintance in di↵ erent ways. Both seem to make sense of the various cases which
illustrate the problem of acquaintance discussed in chapter 3. ILQS lays down a restric-
tion on whether a report can take a particular permissive reading in terms of whether
the corresponding loaded report is true and relevant given what the speaker is conversa-
tionally assumed to believe. SIT lays down a restriction on whether a report can take a
permissive reading with respect to a particular concept in terms of whether the subject is
conversationally assumed to accurately apply that concept. Both ways of cashing out the
idea behind Non-Acquaintance seem promising, and at this point I can see no particular
grounds for favoring one over the other.
5. Revising Specific Inaccurate Thinking
Before moving on from the pragmatic question of permissiveness, I want to revisit
how SIT accounts for some original cases in which a report takes a permissive reading
with respect to a bare noun phrase, such as ‘stolen painting’. The subtleties observed in
this section suggest some cases that constitute counter-examples to SIT as it is originally
worded. But some of the discussion in the previous section about what it is to accurately
apply concepts associated with names can be used to revise SIT in a way that preserves
its good qualities and also avoids these new problems. This revised principle will serve as
the “suitably reinterpreted version” of SIT mentioned above.
First let’s consider a variation on Shadow:
Shadows: Detectives Black and Grey know that White is well acquainted with the
members of the jazz club: at the jazz club’s weekly meetings, White serves the club hors
d’oeuvres and drinks as they listen to jazz records in the lounge. One dark night, the
detectives are tailing White, who is tailing a group of shadowy figures. The detectives
know that these shadowy figures are all and only the members of the jazz club, but further
know that White is not aware of this: they know that White believes that the members of
the jazz club are all listening to jazz records in the lounge. At some point, the detectives
185
lose track of the jazz club, so Black approaches White to ask her ‘Where did they go?’. In
response White says, ‘They took the secret passage’. Black returns to Grey and assertively
utters:
(84) White believes that the members of the jazz club took the secret passage.
A similar problem arises in this case involving the plural definite description ‘the mem-
bers of the jazz club’ as arises in the original Shadow case involving the name ‘Scarlet’. In
this case, intuitively (84) takes a permissive reading with respect to ‘members of the jazz
club’ and the conversationally associated concept members-of-the-jazz-club. According to
our working version of SIT, this requires that it is conversationally assumed that White
doesnot accuratelyapply theconcept members-of-the-jazz-club withrespect towhat Black
is conversationally assumed to apply this concept to in making her report. As in Shadow,
there intuitively is a sense in which White does apply this concept to what Black is as-
sumed to apply the concept to in making her report: White does apply this concept to
the plurality of jazz club members under ordinary concepts she has of that plurality. But
there is also intuitively a sense in which White does not apply this concept to what Black
is assumed to apply the concept to in making her report: White does not apply this con-
cept to the plurality of jazz club members in any thoughts by which she conceives of this
plurality as shadowy figures stalking through the manor.
Now, although White does misapply the concept members-of-the-jazz-club in at least
one sense, Shadows does not yet constitute a counter-example to SIT. But there are other
variationsonShadows thatdoseemtoposeacounter-exampletoSIT;somecanbehandled
by resources already developed, whereas others require a revision of SIT. Let’s start with:
Shadows
2
: Detectives Black and Grey know that White is well acquainted with the
members of the jazz club: at the jazz club’s weekly meetings, White serves the club hors
d’oeuvres and drinks as they listen to jazz records in the lounge. One dark night, the
detectives are tailing White, who is tailing a group of shadowy figures. The detectives
know that these shadowy figures are all and only the members of the jazz club, but further
know that White is not aware of this: they know that White believes that the members
of the jazz club are all listening to jazz records in the lounge. The detectives believe that
White knows where the jazz club was planning to go later in the evening. At some point,
the detectives lose track of the jazz club, so Grey instructs Black to ask White where the
jazz club members are headed. Black approaches White and asks her ‘Where is the jazz
club headed later this evening?’. In response White says, ‘The jazz club members were
planning to take the secret passage to the kitchen to practice playing some swing songs’.
Black returns to Grey and assertively utters:
(85) White believes that the members of the jazz club were going to take the secret
passage to the kitchen.
186
In this context, in contrast to Shadows, Black’s report takes a loaded reading with
respectto‘membersofthejazzclub’andtheconversationallyassociatedconceptmembers-
of-the-jazz-club. But this loaded reading does not at first appear to be predicted by the
version of ITQS that corresponds to replacing Inaccurate Thinking with SIT. For, first,
Direct QUD Sensitivity apparently allows that (85) can take a permissive reading with
respect to ‘the members of the jazz club’, since Black seems to be directly addressing the
question ?x[the members of the jazz club arex] in this context, which in this context might
beaskedby‘Wherearethemembersofthejazzclubgoing?’. And, second, SIT allowsthat
(85) can take a permissive reading with respect to ‘the members of the jazz club’, since
it is conversationally assumed in the context that White does not accurately apply the
concept members-of-the-jazz-club with respect to what Black is conversationally assumed
toapplythisconcepttoinmakingherreport. For, asbefore, itisconversationallyassumed
that White does not apply this concept towards the plurality of jazz club members as she
cognizes them as shadowy figures stalking through the manor. We made use of this fact
in the original Shadows to use SIT to account for the permissive reading of ‘members of
the jazz club’ in that case; given the way that our going version of SIT is worded, there is
nothing to prevent applying it in the same way in this case. So we should expect, insofar
as we are using these di↵ erent necessary conditions on permissiveness together to pose a
necessary and su cient condition, that (85) takes a permissive reading with respect to
‘members of the jazz club’ in this case, contrary to observed fact.
But this argument that (85) in Shadows
2
poses a counter-example to our working
version of SIT is too quick. Given some discussion from chapter 5, we can see that the
problem here can be explained in terms of what question (85) serves to directly address in
this context. In §5.5 of chapter 5, we observed that one way that permissive reporting is
question-sensitive concerns questions explicitly about what the subject thinks. There we
observed that if a speaker makes a report by assertively uttering ‘Peacock believes that
every forgery in the gallery is on the west wall’ to directly address a question asked by
‘Where does Peacock believe that the forgeries are?’, whether the report made takes a
permissive reading with respect to ‘forgery in the gallery’ depends on whether the question
the report addresses itself takes a permissive reading with respect to the corresponding
phrase ‘forgeries’. This observation alone can be used to account for the loaded reading of
‘members of the jazz club’ in Shadows
2
. For since Grey has instructed Black to ask White
where the jazz club members are headed, intuitively in this section (85) serves to directly
address the question ?x[White thinks the jazz club members are headed to x], with ‘x’
rangingoverlocations. Recallthatwhichquestionareportservestodirectlyaddresscomes
apart from what questions are explicitly under discussion in a reporting context. Although
the main QUD in Shadows
2
is something like where-is-the-jazz-club-headed-later?,
it remains intuitive that, given the prior discourse, Black directly addresses instead a
question like where-does-White-think-the-jazz-club-is-headed-later?. It also is
intuitive that this question that (85) serves to directly address takes a loaded reading with
respecttotheconceptmembers-of-the-jazz-cub andthecorrespondingquestionconstituent.
187
Given this, our hypothesis madeabout questionsabout attitudesin chapter 5predictsthat
(85) should take a loaded reading with respect to ‘members of the jazz club’ in Shadows
2
.
So some of our question-sensitive resources su ce to protect our working version of
SIT against Stolen
2
. But there are yet other variations on Shadows that show that SIT is
indeed in need of revision. Consider
Shadows
3
: Detectives Black and Grey know that White is well acquainted with the
members of the jazz club: at the jazz club’s weekly meetings, White serves the club hors
d’oeuvres and drinks as they listen to jazz records in the lounge. One dark night, the
detectives are tailing White, who is tailing a group of shadowy figures. The detectives
know that these shadowy figures are all and only the members of the jazz club; they also
know that these shadowy figures are all and only the members of the anarchist secret
society. The detectives know that White is aware that she is tailing the members of the
jazz club, but that she is unaware that the shadowy figures of jazz club members she is
tailingaremembersoftheanarchistsecretsociety;theyknowshehasheardoftheanarchist
society, but thinks that none of the jazz club members are anarchists. At some point, the
detectives lose track of the anarchist jazz club members in the dark, so Black approaches
White to ask her ‘Where did they go?’. In response White says, ‘They took the secret
passage’. Black returns to Grey and assertively utters:
(84) White believes that the members of the jazz club took the secret passage.
In this case, intuitively (84) takes a true reading that is loaded with respect to ‘mem-
bers of the jazz club’. But it is conversationally assumed that White inaccurately applies
theconcept members-of-the-jazz-club with respect totheindividualstowhomBlack iscon-
versationally assumed to apply that concept. For the detectives assume that White does
not apply members-of-the-jazz-club to the plurality of members of the jazz club as she cog-
nizes that plurality as members of the anarchist society. Although White does accurately
apply members-of-the-jazz-club with respect to the plurality of jazz club members as she
cognizes that plurality as jazz club members and as shadowy figures in the manor, she
still is conversationally assumed to misapply that concept with respect to this plurality,
just as in the original Shadows case White is assumed to inaccurately apply this concept
towards some plurality as she cognizes that plurality in one way though she is assumed to
accuratelyapplythisconcepttowardsthatsamepluralityasshecognizesitinanotherway.
In using SIT to account for the permissive reading in the original Shadows, we made use of
the fact that there was at least one way in which White inaccurately applied the concept
members-of-the-jazz-club; the same kind of reasoning therefore should apply here, and so
this way of applying SIT should predict a permissive reading with respect to ‘members of
the jazz club’ in (84) in Shadows
3
, contrary to observed fact.
To confirm that this case does indeed raise a problem for SIT, contrast this version of
Shadows
3
with
188
Shadows
4
: Detectives Black and Grey know that White is well acquainted with the
members of the jazz club: at the jazz club’s weekly meetings, White serves the club hors
d’oeuvres and drinks as they listen to jazz records in the lounge. One dark night, the
detectives are tailing White, who is tailing a group of shadowy figures. The detectives
know that these shadowy figures are all and only the members of the jazz club; they also
know that these shadowy figures are all and only the members of the anarchist secret
society; they also know that these shadowy figures are all and only the real estate owners
staying in the manor. The detectives know that White is aware that she is tailing the
members of the jazz club, but that she is unaware that the shadowy figures of jazz club
memberssheistailingaremembersoftheanarchistsecretsociety; theyknowshehasheard
oftheanarchistsociety,butthinksthatnoneofthejazzclubmembersareanarchists. They
know, however, that she does know that the members of the anarchist society are exactly
the real estate owners staying in the manor. At some point, the detectives lose track of the
anarchist jazz club members in the dark, so Black approaches White to ask her ‘Where did
they go?’. In response White says, ‘They took the secret passage’. Black returns to Grey
and assertively utters:
(86) White believes that the anarchists took the secret passage.
Black’s use of (86) in this context intuitively takes a true permissive reading with re-
spect to ‘anarchists’. This is so despite the fact that in Shadows
4
White is conversationally
assumed both to accurately apply the target concept anarchist, in one way, and to inaccu-
ratelyapplythisconcept, inanotherway, withrespecttothesamepluralitytowhichBlack
is conversationally assumed to apply this concept in issuing her report with (86). We saw
in the Shadows
2
case that White is similarly conversationally assumed both to accurately
and to inaccurately apply the target concept members-of-the-jazz-club with respect to a
single plurality, but in that context we detect a loaded reading with respect to its target
phrase, in contrast to the permissive reading with respect to the target phrase here. Given
this similarity between these two cases and the similarity in what sort of question Black
directly addresses in issuing her reports in these cases, we should expect on our version
of SIT that either both (84) in Shadows
3
and (86) in Shadows
4
take permissive readings
with respect to ‘members of the jazz club’ and ‘anarchists’, respectively, or both should
take corresponding loaded readings. Since these reports di↵ er in this respect, the problem
does seem to lie with our stated version of SIT.
Now, intuitively the sense in which White is conversationally assumed to inaccurately
apply the concept members-of-the-jazz-club in Shadows
3
should not be relevant to whether
the report in that case takes a permissive reading with respect to that concept, whereas in
the original Shadows it is relevant. As suggested by our previous discussion of permissive
readings with respect to names, what apparently is needed is that we relativize the central
notion of accurate concept application to the concepts by which the objects to which the
relevant concept is applied are themselves cognized. Intuitively what we want to do is to
189
reframeSIT tobesensitivetohowtheobjectstowhichatargetconceptisappliedarecog-
nized in the subject’s applying that concept to those objects, so that White’s inaccurately
applying members-of-the-jazz-club to those she cognizes as shadowy figures is relevant to
whether it takes a permissive reading with respect to that concept in Shadows but White’s
inaccurately applying this concept with respect to those she cognizes as anarchists is not
relevant in Shadows
3
.
In performing an act of applying a concept to an object o, one must cognize o in
some way or other. Likewise, insofar as a speaker is conversationally assumed to apply
a concept to some object o in issuing a report involving linguistic material F conver-
sationally associated with concept , that speaker may be assumed to apply to o as
o is cognized in some particular way or ways. Now, in both Shadows
3
and Shadows
4
,
since both Black and Grey are tracking the group of shadowy figures, cognizing them as
shadowy figures (or simply via some demonstrative concept those people)intuitivelyitis
conversationally assumed in both contexts that in making a report with (84) Black applies
the concept members-of-the-jazz-club to the plurality of members of the jazz club as they
are cognized as shadowy figures (or demonstratively as those people). Similarly, in (86) it
is conversationally assumed that Black applies the concept anarchists to the plurality of
anarchists as they are cognized as shadowy figures (or demonstratively). By contrast, in
Shadows
3
it does not seem contextually assumed that in making her report Black applies
members-of-the-jazz-club to the plurality of members of the jazz club as they are cognized
as anarchists, even though it happens to be contextually assumed that the members of the
jazz club are the anarchists. Likewise, in Shadows
4
it does not seem contextually assumed
that in making her report Black applies anarchists to the plurality of anarchists as they
are cognized as real estate owners staying in the manor, even though it happens to be
contextually assumed that the anarchists in the manor are the real estate owners in the
manor. I realize that it may not always be vividly clear when a speaker is conversationally
assumed to apply a concept to some object or plurality cognized in some way, but going
forward I am going to take the foregoing claims for granted even though more work is
needed to sharpen this ideology.
With these assumptions about Shadows
3
and Shadows
4
, we can account for the di↵ er-
ences between them using the following revision of SIT:
Specific Inaccurate Conceptualization (SIC):
An utterance of an attitude report sentence R of the formAVsthatS(F) by
Speaker S in context c serves to make a report which is permissive with respect
to expression e occurring in R only if it is conversationally assumed in c that A
does not accurately apply the concept conversationally associated with e to
object o as o is cognized via concept ,where o is any object and any concept
such that it is conversationally assumed in c that in using R, S applies to o as
o is cognized via .
The simple way that SIC di↵ ers from SIT is that in place of requiring that A be
190
conversationally assumed not to apply to whatever objects S is assumed to apply to
in issuing her report, per se, we use our relativized notion of concept application, requiring
that A be conversationally assumed not to apply to whatever objects S is assumed to
apply to in issuing her report, in such a way that A cognizes those objects in applying
to them just as S cognizes those objects in applying to them.
To apply SIC to Shadows, Shadows
3
, and Shadows
4
, it is enough to recall our assump-
tions from before that in both contexts Black is assumed to apply the target concept to the
relevantpluralityasthatpluralityiscognizedviashadowy-figures (ordemonstrativelyviaa
those-people concept); I will assume that this holds in the original Shadows context as well.
In Shadows, recall, White is conversationally assumed not to apply members-of-the-jazz-
club to the plurality of members of the jazz club as she cognizes them via shadowy-figures
or demonstratively. So SIC correctly allows a permissive reading with respect to ‘members
ofthejazzclub’intheoriginalShadows. Bycontrast,inShadows
3
itisassumedthatWhite
doesaccuratelyapplymembers-of-the-jazz-club tothepluralityofmembersofthejazzclub
as she cognizes them via shadowy-figures; the detectives assume that she knows that the
shadowy figures (those she is tracking demonstratively) are the jazz club members. So SIC
correctly predicts a loaded reading of (84) with respect to ‘members of the jazz club’, even
though it is conversationally assumed that White misapplies members-of-the-jazz-club to
the plurality of jazz club members in some way or another (as they are cognized as anar-
chists). Finally, in Shadows
4
, White is assumed not to apply anarchists to the plurality of
anarchists as she cognizes them via shadowy-figures, or demonstratively. So SIC allows a
permissive reading of ‘anarchists’ in Shadows
4
. It appears, then, that SIC is an advance
over SIT.
6. Addressing the Acquaintance Problem for Contextual Augmentation Views
In chapter 3 I observed that the acquaintance phenomenon can be used to pose ap-
parent counter-examples to contextual augmentation views. Now that we have developed
two pragmatic principles to account for this phenomenon, it is time to see whether these
principlesprovideanyhelpinprotectingcontextualaugmentationviewsfromthisapparent
problem.
I presented this problem in §4 of chapter 3 with Blackmail
2
from §1 of this chapter,
which I repeat here:
Blackmail
2
: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called into discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. The
detectives know that Green is well acquainted with Scarlet: Green and Scarlet are close
friends. ByquestioningGreen,DetectiveBlacklearnsthatGreenthinksthateverysocialist
guest was being blackmailed by Mr. Body. The detectives are exchanging information,
trying to find out who had a motive for killing Body. Black assertively utters the following
to Grey:
191
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
The potential problem posed by this example for contextual augmentation views is
this. Although (2) resists a permissive reading (with respect to ‘Scarlet’, but also with
respect to the rest of the constituents of the report sentence), unlike (2) in the original
Blackmail, the proposition contributed by the complement of (2) is distinctively and non-
trivially contextually entailed by the object of Green’s loaded thought and the common
assumptions of the context. On the simplest versions of the contextual augmentation view,
(2) should therefore take a true permissive reading.
The discussion from §6 of chapter 5 already provides a way to dissolve this problem.
In that section I insisted that the way to resolve the asymmetry problem for contextual
augmentation views is to draw a distinction between the truth-conditional question of
permissiveness and the pragmatic problem of permissiveness. The original contextual aug-
mentationviewimplicitlyconflatedthesetwoquestionsbyaccountingatonceforthetruth
conditions and the permissive-reading conditions of a given report. In developing the con-
textual augmentation to be sensitive to the distinction between these theoretical questions,
in chapter 5 I arrived at:
Sophisticated Concept-Use Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) in context
c that is loaded with respect to concept and linguistic expression F that
contributespropositionalconstituentftopositionointhepropositionpdesignated
by the report’s ‘that’-clause is true i↵ there is a proposition q structurally similar
to p such that f occurs in q in a structurally analogous position to o in p, and A
uses to cognize f as it occurs in that position in q in bearing V
L
to q and q
non-trivially and distinctively contextually entails p in c.
Wesawinthelastchapterthatthisprinciplehitstroubleonsomecomplicatedcounter-
examples, but at least it is sensitive to the distinction between the truth-conditional ques-
tion of permissiveness and the pragmatic question of permissiveness. This may be enough
to handle the original acquaintance problem for contextual augmentation views.
Intuitively (2) is loaded with respect to both ‘Scarlet’ and ‘being blackmailed by Body’
in Blackmail
2
. According to Sophisticated Concept-Use Augmentation View, then, for (2)
to take a true reading in this context, there must be a proposition q in which Scarlet
occurs in a structurally similar position to the occurrence of Scarlet in the proposition
designated by the ‘that’ clause of (2), and Green must use the Scarlet concept that is
conversationally associated with ‘Scarlet’ in (2) in cognizing the occurrence of Scarlet in
bearing a fully loaded thinks relation towards q.In Blackmail
2
, Green is not said to have
anysuchfullyloadedattitude. TheonlypropositionthatGreenisrepresentedasbearinga
fully loaded thinks relation towards is the proposition that every socialist guest was being
blackmailed by Mr. Body. While this proposition does contextually entail the proposition
designated by the ‘that’ clause of (2) and the occurrence of blackmailed-by-Body occurs
192
in a structurally similar position in this proposition to the occurrence of that property in
the proposition designated by the ‘that’ clause of (2), this proposition has no occurrence of
Scarlet at all. This is enough for Sophisticated Concept-Use Augmentation View to rule
out a true reading of (2) in Blackmail
2
.
The way that Sophisticated Concept-Use Augmentation View handles the acquaintance
problem in this case suggests, however, some other problematic instances of the acquain-
tance problem. Given how Sophisticated Concept-Use Augmentation View is stated, if an
attitudereportR isloadedwithrespectto andpropositionalconstituent f, thatreportis
true if and only if the subject both bears a fully loaded attitude towards an appropriately
relatedpropositioninwhich foccursanduses incognizing f. Forcasesinwhichareport
takes a loaded reading with respect to multiple constituents of the proposition ascribed as
entertained by the report, Sophisticated Concept-Use Augmentation View just says that
for each target constituent with respect to which the report is loaded, the subject possesses
some appropriately related fully loaded attitude. But this sort of condition is too weak.
To see an example of why, consider
Blackmail
3
: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called into discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. They
also learn that every counterfeiter in the manor was being blackmailed by Mr. Body.
The detectives know that Green is well acquainted with Scarlet: Green and Scarlet are
close friends. By questioning Green, Detective Black learns that Green thinks that every
socialist guest was being blackmailed by Mr. Body. Black also learns that Green thinks
that Scarlet is a counterfeiter. The detectives are exchanging information, trying to find
out who had a motive for killing Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
As before, this intuitively takes a false reading that is loaded with respect to both
‘Scarlet’ and ‘blackmailed by Mr. Body’. Since this is loaded with respect to multiple
constituents of the ‘that’-clause of (2), according to Sophisticated Concept-Use Augmen-
tation View, (2) is true if and only if Green bears a fully loaded thinks relation towards
(i) a proposition p that contextually entails that Scarlet was blackmailed by Body and is
structurally similar to that proposition such that Scarlet occurs in p a structurally anal-
ogous position to Scarlet’s occurrence in the proposition ascribed as thought by (2), and
Green uses the concept Scarlet in cognizing that occurrence of Scarlet, and (ii) a propo-
sition q that contextually entails that Scarlet was blackmailed by Body and is structurally
similar to that proposition such that blackmailed-by-Body occurs in q a structurally
analogous position to blackmailed-by-Body’s occurrence in the proposition ascribed as
thought by (2), and Green uses the concept blackmailed-by-Body in cognizing that occur-
rence of blackmailed-by-Body. Both of these conditions are met, for Green bears a fully
loaded thinks relation towards both the proposition that every socialist guest was being
193
blackmailed by Body and the proposition that Scarlet is a counterfeiter; these attitudes
satisfy (i) and (ii), respectively. So according to Sophisticated Concept-Use Augmentation
View (2) should take a true multiply loaded reading in Blackmail
3
, contrary to observed
fact.
The problem here apparently is that Sophisticated Concept-Use Augmentation View
allows that (2) can be true just if, for each loaded constituent of the proposition designated
by the ‘that’-clause of (2), there is some appropriately related fully loaded attitude of
Green’s. WhatintuitivelyisneededinsteadisthatGreenhaveasingle fullyloadedattitude
that appropriately relates to each of the loaded constituents of the proposition designated
by the complement of (2). Intuitively, when a speaker issues a multiply loaded report, one
represents the subject as having some one fully loaded attitude in a proposition that is
structurally similar to the ascribed thought content for every constituent with respect to
whichthereporttakesaloadedreading,andthesubjectcognizesthoseconstituentsintheir
fully loaded thought via the concepts conversationally associated with those constituents
by the report in the context.
To state a refinement of Sophisticated Concept-Use Augmentation View along these
lines, I will first propose some new shorthand. For a reportR which takes a loaded reading
withrespecttoaconcept andanoccurrenceoofpropositionalconstituentfinanascribed
propositionp, I will say thatR takes a loaded reading with respect tohp, ,f,oi.Thisway
of speaking will allow us to more simply describe reports which takes loaded readings with
respect to multiple propositional constituents, by simply referencing multiple quadruples.
Refined Concept-Use Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) in context c,
where that S(F) designates proposition p in c,istruei↵ there is a proposition q
suchthatforeveryquadruplehp, ,f,oiwithrespecttowhichRisloaded, foccurs
in q in a structurally analogous position to o in p, and A uses to cognize f as it
occurs in that position inq in bearing V
L
toq, andq non-trivially and distinctively
contextually entails p in c.
This refined augmentation view handles the problem raised by Blackmail
3
. For in
that context, there is not one proposition q that contextually entails that Scarlet was
being blackmailed by Mr. Body that meets this view’s stated condition. Call the propo-
sition designated by the complement of (2) “p”, and call the occurrences of Scarlet and
blackmailed-by-Bodyinp,“a”and“b”, respectively. InBlackmail
3
, (2)isloadedwithre-
specttobothhp,Scarlet,Scarlet,aiandhp,blackmailed-by-Body,blackmailed-by-Body,bi.
So according to Refined Concept-Use Augmentation View, the report made by (2) in this
context is true only if Green has a fully loaded thought in the entertaining of which
he uses both Scarlet and blackmailed-by-Body in cognizing occurrences of Scarlet and
blackmailed-by-Bodyinasingleappropriatelyrelatedproposition;giventheseconstraints,
this fully loaded thought would have the proposition p itself as its content: that Scarlet
194
was being blackmailed by Body. Since Green does not bear a fully loaded thinks relation
to this proposition, on this view (2) is false in Blackmail
3
. So it looks like Refined Concept-
Use Augmentation View handles the acquaintance problem for contextual augmentation
views well enough.
7. Inaccurate Concept Application and Contextual Augmentation Views
The previous chapter ended in optimistic aporia. The Sophisticated Concept-Use Aug-
mentation View seemed like a promising revision of earlier contextual augmentation views
for dissolving the asymmetry problem raised in chapter 3. But that view also foundered
on a tricky case; despite the principle’s promise, the way forward was unclear. I concluded
the chapter by pointing to a few potential factors that might be responsible for the view’s
failure; one potential factor concerned the phenomenon of a mismatch in how the conver-
sationalists and the subject apply a concept with respect to which a partially permissive
report is loaded. Having discussed the phenomenon of misapplied concepts in developing
SIC in this chapter, I now return to the problem from the end of chapter 5, to see whether
we can gain further optimism and overcome our aporia.
First let’s recall what the problem was. It was driven by the following complicated
case:
Switched Blackmail
6
: The detectives have private information that every guest in the
manor who was blackmailing someone who was being blackmailed by Body was themselves
beingblackmailedbyBody. ThedetectivesalsoknowthatScarletwasblackmailingWhite.
Having just questioned Green, Black knows that Green believes that White was being
blackmailed by Body. Although the detectives know that Scarlet was blackmailing White,
the detectives do not know that White was, as Green believes, being blackmailed by Body.
The detectives further know that Scarlet is Jade’s sister; they know that Green does not
know this. They are exchanging information, trying to find out who had a motive for
killing Body; in particular, in this scenario they are trying to figure out who in particular
was being blackmailed by Body. Black assertively utters the following to Grey:
(71) Green believes that Jade’s sister was being blackmailed by Mr. Body.
Ourworkingassumptionattheendofchapter5wasthat(71)takesfalsereadingthatis
loaded with respect to ‘blackmailed by Mr. Body’ and is permissive with respect to ‘Jade’s
sister’. If this assumption is correct, then Sophisticated Concept-Use Augmentation View
andournewRefined Concept-Use Augmentation View seemtodeliverincorrectpredictions
for this case.
Before recapitulating these failures, however, I will note that if this assumption is
incorrect and (71) is multiply loaded with respect to both ‘Jade’s sister’ and ‘blackmailed
byMr. Body’, thatbothoftheseprinciplescorrectlypredictafalsereadinghere. Theclear
reason why is that in this case Green does not possess any fully loaded attitudes towards
195
any proposition containing Jade’s-sister as a constituent.
124
Even if we built into the
casethatGreendidhavesuchfullyloadedattitudes, saybybelievingthatJade’ssisterwas
blackmailing White, Refined Concept-Use Augmentation View would still predict a false
reading in this case, since Green does not have a single fully loaded attitude by which he
cognizes both ‘Jade’s sister’ and ‘blackmailed by Mr. Body’ via concepts Jade’s-sister and
blackmailed-by-Body.
If, however, the assumption from chapter 5 does stand and (71) is permissive with
respectto‘Jade’ssister’inSwitchedBlackmail
6
,thenbothoftheseprinciplesissueincorrect
predictions for this case. We can illustrate as much by focusing on what Refined Concept-
UseAugmentationView hastosay. Assumethat(71)isloadedinSwitchedBlackmail
6
with
respect to hp,blackmailed-by-Body,blackmailed-by-Body,oi,where p is the proposition
designated by the complement of (71) and o is the occurrence of blackmailed-by-Body in
p. According to Refined Concept-Use Augmentation View, then, (71) is true if and only if
Greenbearsafullyloadedbeliefrelationtowardapropositionqinwhichblackmailed-by
-Body occurs in a structurally analogous position to the occurrence of blackmailed-by-
Bodyinthepropositiondesignatedbythe‘that’-clauseof(71)suchthatq non-triviallyand
distinctively contextually entails the proposition designated by the ‘that’-clause of (71) in
Switched Blackmail
6
, and in holding this fully loaded attitude Green cognizes the relevant
occurrence of blackmailed-by-Body via the concept blackmailed-by-Body associated with
that property in (71).
In Switched Blackmail
6
, the proposition that White was being blackmailed by Body
contextually entails that Jade’s sister was being blackmailed by Body, given the detectives’
contextual assumptions that Scarlet is Jade’s sister, that Scarlet was blackmailing White,
and that every guest in the manor who was blackmailing someone who was being black-
mailed by Body was themselves being blackmailed by Body. The proposition that White
was being blackmailed by Body also is structurally similar to the proposition that Jade’s
sister was being blackmailed by Body, and the occurrence of blackmailed-by-Body in the
former sits in a structurally analogous position to the occurrence of that property in the
latter. And Green cognizes blackmailed-by-Body via the concept blackmailed-by-Body
that is associated with that property in (71). So, according to Refined Concept-Use Aug-
mentation View, the use of (71) in Switched Blackmail
6
should assert a truth, contrary to
our judgment for that case.
I speculated at the end of chapter 5 that one factor that may be responsible for the
problem here is that, in holding his fully loaded belief in that White was blackmailed by
Body,Greendoesnotapplytheconceptblackmailed-by-Body tothe“samethings”towhich
Black applies that concept in issuing a report with (71). It is tempting to say that with
the partially permissive report made by (71), one thing that Black does is to represent
Green as cognizing blackmailed-by-Body via the concept blackmailed-by-Body in holding
124
As before, I’m ignoring complications about the propositional contribution of possessive phrases in
representing the propositional contribution of ‘Jade’s sister’.
196
some fully loaded attitude whereby Green applies blackmailed-by-Body to the thing that
the conversationalists in fact cognize via the concept Jade’s-sister. Since the report is
permissive with respect to ‘Jade’s sister’ and the concept Jade’s-sister, of course Black
does not represent Green as cognizing Jade’s-sister, or the thing which is in fact Jade’s
sister, Scarlet, via the conversationally associated concept Jade’s-sister.Buttherestillis
some intuition that Black represents Green as applying blackmailed-by-Body to Scarlet or
at least as being somehow disposed to do so.
Let’s run with this idea, to see whether it can help us to resolve the problem raised by
Switched Blackmail
6
. According to this chapter’s SIC, whether a report takes a permis-
sive reading with respect to some constituent of an attitude report and a conversationally
associated concept depends on whether, according to the assumptions made by the con-
versationalists in the reporting context, the subject of the report does not accurately apply
with respect to those things the speaker applies in issuing her report (cognizing those
thingsasthespeakercognizingtheminapplying totheminissuingherreport). Wemight
re-purpose something like this condition in specifying the truth-conditions of partially per-
missive reports. To motivate this idea, we can note that according to SIC, if according
to the contextual assumptions in the reporting context the subject does accurately apply
with respect to those things the speaker is conversationally assumed to apply to in
making her report (and cognizes them in the right way in so doing), then the report does
nottakealoadedreadingwithrespectto . Itisnotafarjumpfromheretospeculatethat
if a report does take a loaded reading with respect to some constituent of its ‘that’-clause
and an associated concept , that the subject is represented by that report as accurately
applying with respect to what the speaker is conversationally assumed to apply to in
making her report. So, then, on this idea, there is this kind of conceptual constraint on the
truth conditions of a partially permissive report. We might build this kind of constraint in
to Refined Concept-Use Augmentation View as follows:
Accurate Concept-Application Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) made by
speaker S in context c,where that S(F) designates proposition p in c,istruei↵ there is a proposition q such that for every quadruplehp, ,f,oi with respect to
which R is loaded, f occurs in q in a structurally analogous position to o in p, and
A uses to cognize f as it occurs in that position in q in bearing V
L
to q and in
so doing applies with respect to object o as o is cognized via concept ,where o
is any object and any concept such that it is conversationally assumed in c that
in using R, S applies to o as o is cognized via .
What Accurate Concept-Application Augmentation View adds is the condition that the
subject of a report that is loaded with respect to some concept apply towards what
the speaker is conversationally assumed to apply to in issuing her report (cognizing those
thingsindoingsoasthespeakercognizestheminissuingherreport). So,applyingAccurate
197
Concept-Application Augmentation View to Switched Blackmail
6
, for (71) to be used to
assert a truth in that context, we require not only that Green bear a fully loaded belief
relation towards a proposition that contextually entails the ascribed proposition and is
structurally similar to that proposition in the right ways, but also that Green apply the
target loaded concept blackmailed-by-Body to what Black is conversationally assumed to
apply that concept to in making her report: namely, to Scarlet, whom the detectives know
to be Jade’s sister. Green fails to do this, despite bearing a fully loaded belief relation to
thepropositionthatWhitewasbeingblackmailedbyBody. Greenisnowherenearapplying
the concept blackmailed-by-Body to Scarlet in having the fully loaded attitudes he has in
this scenario, so Accurate Concept-Application Augmentation View correctly predicts that
(71) takes a false reading in Switched Blackmail
6
.
But with the added constraint to this new augmentation view, we should look to see
whether this principle makes erroneous predictions of false permissive reports elsewhere.
More constraints on true permissive readings means more potential for accidentally ruling
out true permissive readings for reports that in fact read true. There seem to be some such
erroneous predictions on this new view, which we can see by reconsidering our original
permissive UI report:
Blackmail: Mr. Body was murdered last night somewhere in the manor; detectives
Black and Grey have been called into discover who killed Body. In the course of their
investigation, the detectives learn that Scarlet, a guest at the manor, is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet: Green doesn’t know any name for Scarlet, could not identify her in a line-up,
and so on. By questioning Green, Detective Black learns that Green thinks that every
socialist guest was being blackmailed by Mr. Body. Green doesn’t, however, know the
names of any of the socialist guests: he claims “not to know who the socialist guests are”.
The detectives are exchanging information, trying to find out who had a motive for killing
Body. Black assertively utters the following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
Once again, we detect a true reading of (2) in Blackmail. The reading is partially
permissive: permissive with respect to ‘Scarlet’ and loaded with respect to ‘blackmailed by
Mr. Body’. So let’s apply Accurate Concept-Application Augmentation View to this case
and see what we get. This contextual augmentation view requires, in addition to plenty
else, that Green apply the concept blackmailed-by-Body to those things to which Black
applies that concept in issuing her report with (2). In so reporting, Black clearly applies
blackmailed-by-Body to Scarlet. Does Green do so in entertaining his fully loaded thought
that every socialist guest was being blackmailed by Body? It sounds desperate to say yes.
Green is known by the detectives not to know who Scarlet is under any description or
guise. So how on earth could he be accurately applying blackmailed-by-Body to Scarlet
in any of his conversationally relevant fully loaded thoughts? Given that Green fails to
198
apply blackmailed-by-Body to Scarlet in holding his fully loaded belief that every socialist
was being blackmailed by Body, then, Accurate Concept-Application Augmentation View
incorrectly predicts that (2) takes a false reading in Blackmail.
It is tempting here to take a page out of the book of the permissive dispositions view
from chapter 4. Perhaps the truth-conditional constraint we want is not that the subject
in fact applies the target loaded concept with respect to whatever the speaker is conver-
sationally assumed to apply that concept towards in issuing her report, but instead that
the subject be disposed under certain conditions to apply that concept so. The kind of
disposition we would want for this case, intuitively, is something like that the subject be
disposed to apply the target loaded concept to whatever the speaker is conversationally
assumed to apply that concept towards in issuing her report on the condition of learn-
ing the conversational assumptions that underlie the relevant contextual entailment in the
context.
Although rewriting Accurate Concept-Application Augmentation View with this kind
of dispositional constraint would successfully account for UI Reports like (2) in Blackmail,
it would not serve us as a general account of the truth-conditions of partially permissive
reports. Forinstance, intheproblematicSwitched Blackmail
6
discussedabove, sinceGreen
is indeed disposed to apply the target loaded concept blackmailed-by-Body to what Black
is conversationally assumed to apply that concept to in issuing her report with (71) on the
condition of learning the multiple conversational assumptions relevant to the contextual
entailment in Switched Blackmail
6
. So this revised view would erroneously predict a true
reading of (71) in that context. Using this kind of disposition as part of a general truth-
conditiononpartiallypermissivereports,then,basicallyrunsintothesamestyleofproblem
thatfacedthecounterfactualattitudesviewdiscussedearlyoninchapter3. Butinsteadto
refine Accurate Concept-Application Augmentation View by leaving it open-ended which
kinds of dispositions to apply target loaded concepts are involved in the truth-conditions
of partially permissive reports would, it seems, turn this contextual augmentation view
instead into a complicated version of a permissive dispositions view. Such a view may
be what we ultimately want, but embracing that view at this point would be to abandon
our current goal of developing a contextual augmentation view to handle present apparent
counter-examples.
Before giving up on the project of defending contextual augmentation views here, I
want to consider another approach to avoiding the problem raised by Switched Blackmail
6
.
In the final section of the previous chapter, I briefly discussed another potential factor that
may be driving this problem: that contextual entailment, even non-trivial and distinctive
contextualentailment, isthewrongrelationtouseinaccountingforthetruth-conditionsof
partially permissive reports. Attending to plausible alternative relations to use in place of
contextual entailment suggests a di↵ erent way of developing Accurate Concept-Application
Augmentation View to avoid the problem raised by Switched Blackmail
6
.I’llbuildupto
this new way of handling that problem by discussing the attitude variation problem from
199
chapter 3, which puts a distinctive kind of pressure on the use contextual entailment to
answer the truth-conditional problem of permissiveness.
As mentioned in chapter 5, the attitude variation problem raised in chapter 3 seems to
give independent support to the idea that contextual entailment is too restrictive a version
of contextual augmentation to use in accounting for the truth-conditions of permissive
reports across the board. Recall that the basic shape of the attitude variation problem is
that if we construct a context just like one involving a permissive report that a subject
“believes” or “thinks” some proposition by replacing the main attitude verb with another,
such as ‘fears’ or ‘doubts’, that the resulting context is one in which the report fails to take
the same kind of permissive reading that it does in the original. For example, consider
Blackmail Fear: Black knows that Green has made a bet that not all socialists were
beingblackmailedbyBody. BlackalsoknowsthatGreenthereforefearsthateverysocialist
is being blackmailed by Body. The detectives are aware that Green is not acquainted with
Scarlet. The detectives commonly know that Scarlet is a socialist. In discussing who had
a motive for killing Body, Black says to Grey:
(87) Green fears that Scarlet was being blackmailed by Mr. Body.
Inthiscontextitisveryhardtohear(87)astakingatruepermissivereading,butthisis
predicted by simple contextual augmentation views. For Green bears a fully loaded fears
relation towards a proposition that, together with the detectives’ contextual assumption,
non-trivially and distinctively entails that Scarlet was being blackmailed by Body.
Thereseemtobeatleastacoupleofwaysthatonecouldattempttoaddresstheattitude
variation problem raised by Blackmail Fear. One way would be to use the same strategy
we have used so far in dealing with the asymmetry problem and the acquaintance problem:
takecaretodistinguishthetruth-conditionalquestionofpermissivenessfromthepragmatic
question of permissiveness. For not only does (87) clearly take a false reading in Blackmail
Fear, it might be argued that it also takes a loaded reading with respect to the occurrence
of Scarlet contributed by ‘Scarlet’ and a conversationally associated Scarlet concept (as
well as a loaded reading with respect to the occurrence of blackmailed-by-Body and
blackmailed-by-Body). If that is right, then Accurate Concept-Application Augmentation
View already can account for the false reading of (87) in Blackmail Fear,sinceGreendoes
not apply Scarlet to what Black is conversationally assumed to apply Scarlet to in making
her report, since Green simply lacks that concept altogether.
But it is also not completely obvious that (87) does take a loaded reading with respect
to ‘Scarlet’ and Scarlet in this case. To see this, we can consider a context that is similar
to Blackmail Fear in which (87) does intuitively take a true reading that is permissive with
respect to ‘Scarlet’ and Scarlet:
Blackmail Fear
2
: Black knows that Green has made a bet that the shadowy figure he
sees stalking through the hallway in the manor this evening was not being blackmailed
200
by Body. Black also knows that Green therefore fears that the shadowy figure is being
blackmailed by Body. The detectives are aware that Green is not acquainted with Scarlet.
The detectives commonly know that the shadowy figure stalking through the hallway is in
fact Scarlet. In discussing who had a motive for killing Body, Black says to Grey:
(87) Green fears that Scarlet was being blackmailed by Mr. Body.
In this context, to my ear, (87) more easily takes a true reading that is permissive
with respect to ‘Scarlet’ and the associated concept Scarlet. Given the similarity between
Blackmail Fear and Blackmail Fear
2
, there is some reason to say that in fact (87) takes a
permissive reading with respect to ‘Scarlet’ and Scarlet in each, even though it is false in
Blackmail Fear but true in Blackmail Fear
2
.
If (87) does take a false reading that is permissive with respect to ‘Scarlet’ and Scarlet
in Blackmail Fear, we still then have reason to suspect that the problem lies with using
contextual entailment to account for the truth conditions of partially permissive reports.
Looking to what di↵ ers between Blackmail Fear and Blackmail Fear
2
suggests, however, a
secondwayofaddressingtheattitudeproblem,whichinvolvesusingamoregeneralrelation
than contextual entailment in accounting for the truth conditions of partially permissive
reports.
With respect to Blackmail Fear
2
, were Green to hold his fully loaded fears attitude
that the shadowy figure was being blackmailed by Body and also to assume that the
shadowy figure in the hallway is Scarlet, then he would be rationally committed, in some
sense, to having a fully loaded fear that Scarlet was being blackmailed by Body. Turning
to Blackmail Fear, however, were Green to have a fully loaded fear that every socialist was
being blackmailed by Body (since he hopes that not all socialists were being blackmailed
by Body) and he also assumed that Scarlet was a socialist, he would not be rationally
committed in the same intuitive sense to having a fully loaded fear that Scarlet was being
blackmailed by Body. Since in such a case he just fears that all of the socialists taken
together were being blackmailed, he is not rationally constrained to fear of each individual
socialist that they were being blackmailed: even if Scarlet were being blackmailed he could
still win his bet. But to assume that the shadowy figure is Scarlet and to fear that the
shadowy figure was being blackmailed but not to fear that Scarlet was being blackmailed
smacks of some kind of incoherence.
So, then, one might hope to save the spirit of contextual augmentation views by re-
placing the central notion of contextual entailment with some kind of contextual rational
commitment. Instead of saying that one bears a conceptually permissive V
c
relation, ex-
pressedbyverbV incontextc,towardsapropositionpifandonlyifonebearsafullyloaded
V
L
relation towards a proposition q that contextually entails p in c, we might say instead
something like that one bears a conceptually permissive V
c
relation towards a proposition
p if and only if one bears a fully loaded V
L
relation towards a proposition q such that bear-
ing V
L
to q and holding (some relevant subset of) the assumptions commonly held by the
201
conversationalists inc rationally commits one to holding a fully loaded V
L
relation towards
p. The details of this style of view would need to be worked out further, but it does seem
to successfully rule out a true permissive reading of (87) in Blackmail Fear while ruling
in a true permissive reading of (87) in the original Blackmail. For were Green to hold his
fully loaded fear in Blackmail Fear and also assume what the conversationalists do in that
context, he would not be rationally committed to a fully loaded fear hat Scarlet was being
blackmailed by Body, but where he to hold his fully loaded belief in Blackmail and also
assume what the conversationalists do in that context, he would be rationally committed
to a fully loaded belief that Scarlet was being blackmailed by Body. Similar comments
apply to Blackmail Fear
2
.
In the final chapter I will discuss the prospects of relaxing contextual augmentation
views along these lines. But I mention this style of view in this section just to raise to
salienceitscentralnotionofrationalcommitment,whichcanhelpustoaddresstheproblem
raised by Switched Blackmail
6
for Accurate Concept-Application Augmentation View.We
may be able to use this notion of rational commitment in modifying the added constraint
of this view (that the subject apply target loaded concepts to whatever the speaker is
conversationally assumed to apply those concepts to in issuing her report).
To build towards this fix of Accurate Concept-Application Augmentation View,let’s
engage in some reflection about what is going on with the rational commitment version of
contextual augmentation views. The guiding thought behind this version of the contextual
augmentation view, speaking picturesquely, is that what one does when one issues a true
permissive report is to ascribe as believed (or feared, or hoped, etc.,) what one would
rationally be committed to believing (or fearing, hoping, etc.) were they to know certain
thingstheconversationalistscommonlyknow. Wemight,then,ininvokingthephenomenon
of accurate concept application to address the problem raised by Switched Blackmail
6
for
Accurate Concept-Application Augmentation View, likewise draw on this picture-thinking
about permissive reports on which they serve to draw out the rational commitments of
the subject’s conceptually loaded cognitive state. In particular, we might say that for a
partially permissive report that is loaded with respect to an occurrence of constituent f
and associated concept , that such a report represents the subject as rationally committed
to applying to what the speaker is conversationally assumed to apply to in issuing her
report: chiefly and namely, to f. Something, say, like:
Rational Concept-Application Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) made by
speaker S in context c,where that S(F) designates proposition p in c,istruei↵ there is a proposition q such that for every quadruplehp, ,f,oi with respect to
which R is loaded, f occurs in q in a structurally analogous position to o in p, and
A uses to cognize f as it occurs in that position in q in bearing V
L
to q and in
so doing is rationally committed to applying with respect to object o as o is
cognized via concept ,where o is any object and any concept such that it is
202
conversationally assumed in c that in using R, S applies to o as o is cognized
via .
This principle di↵ ers from Accurate Concept-Application Augmentation View by re-
quiring not that the subject in fact apply target loaded concept to what the speaker is
conversationally assumed to apply to, but instead that the subject be rationally commit-
ted, given how they use as part of their relevant fully loaded attitude, they be rationally
committed to applying to what the speaker is conversationally assumed to apply to
in making her report (conceptualizing those things in doing so as the speaker conceptual-
izes those things in making her report). This might help with the set of problems facing
Accurate Concept-Application Augmentation View.
Here’s how Rational Concept-Application Augmentation View can help. The most re-
cent problem facing Accurate Concept-Application Augmentation View was that it success-
fully handled the problem raised by Switched Blackmail
6
at the cost of failing to account
for the true permissive reading of (2) in the original Blackmail case. The reason for its
failure in the latter case was intuitively that this principle required that Green apply the
target loaded concept blackmailed-by-Body towards what Black was conversationally as-
sumed to apply that concept to in reporting that Green thinks that Scarlet was being
blackmailed by Body, namely Scarlet. Green, lacking a concept of Scarlet, did not ap-
ply blackmailed-by-Body to Scarlet, leading to an erroneous false prediction from Accurate
Concept-Application Augmentation View of (2) in Blackmail.
Now,inUIreports,suchas(2)inBlackmail,thesubjecttypicallyappliesatargetloaded
concept with respect to a plurality of which the individual picked out by the permissive
concept is an element. Green applies blackmailed-by-Body to the plurality of socialists,
or at least is rationally committed to applying that concept to that plurality, in having
a fully loaded belief that every socialist was being blackmailed by Body. But when one
applies a concept that picks out a distributive predicate of a plurality, one thereby, I
claim, takes on a rational commitment to applying that concept to each element of the
plurality. When one sees a crowd from afar and cognizes them as Dodgers fans, one does,
it seems, do something similar to applying the concept Dodger-fan of each individual in
the crowd. Such a person need not, of course, in fact possess token thoughts about each
individual, whereby one cognizes each individual person in the crowd and predicates the
property Dodger-fan of each such individual individually. But one at least is rationally
committed topredicatingthatproperty, andsoapplyingtheconceptbywhichtheycognize
that property, of each individual in the crowd. Similarly, if one cognizes three individuals
individually as being Dodgers fans, and if later, unbeknownst to this cognizer, those three
gather in the Lord’s name to pray, the cognizer is rationally committed to applying the
concept Dodger-fan of the plurality of praying persons, even though the cognizer does not
cognize that plurality in such terms and indeed fails to entertain any proposition whereby
they cognize that plurality at all. One can incur rational commitments to apply a concept
towards individuals or pluralities by applying that concept in certain ways towards certain
203
metaphysically related pluralities and individuals.
So, I argue, that in having a fully loaded thought that every socialist was being black-
mailedbyBody, wherebyGreenappliestheconceptblackmailed-by-Body ofthepluralityof
socialists(orsocialistguestsinthemanor), Greentakesonarationalcommitmenttoapply
that concept to each individual of that plurality, whoever they may be, Scarlet included,
even though Green may lack any concept associated with any element of that plurality,
Scarlet included, in the Blackmail context, or whatever context. Even when an individual
tucked deep in a crowd that is plurally cognized via the concept Dodger-fan is hidden
from view, and in general is hidden from the cognizer’s powers of individual cognition,
the cognizer is in some sense rationally committed to applying the Dodger-fan concept to
that individual. This is not to say that the cognizer would in fact apply this concept to
that individual in any circumstance in which they may happen to cognize that individual
via other concepts. But insofar as one persists in applying the Dodger-fan concept of the
crowd previously beheld, one remains rationally committed to applying that concept to
the individual previously tucked away in the crowd, even if one meets that individual later
in the bar, cognizing them perceptually as they stand in plain view.
Does one violate this in-some-sense rational commitment by not applying the Dodger-
fan concept to this individual they meet in the bar when they meet them? Intuitively
not. This suggests that the rational commitment the cognizer incurs is not to apply the
Dodger-fan concept to each member of the crowd in any circumstances in which they may
cognize those members, but rather that they are rationally committed to applying this
conceptofsuchindividualsontheconditionofcognizingthemasamemberoftheplurality
cognized via that concept. So too with Green and Scarlet. And if that is right, then
Rational Concept-Application Augmentation View successfully predicts a true reading of
(2) in the original Blackmail case. For Green is therefore rationally committed to applying
blackmailed-by-Body to every element of the plurality of socialists in the manor, Scarlet
included, cognized as members of socialists in the manor, and so is rationally committed to
applying the concept blackmailed-by-Body to whom Black is conversationally assumed to
apply that concept to in issuing her report with (2) (where those individuals are cognized
as members of the socialists in the manor by Black in issuing her report). Since the
various other conditions built into Rational Concept-Application Augmentation View are
also clearly satisfied in that case, the principle does predict that (2) is true in Blackmail.
Now, one might worry that the kind of rational commitment used to derive this predic-
tion bears a troubling resemblance to the counterfactual dispositional view mentioned and
discarded above, since this kind of rational commitment seems to be in essence a commit-
ment to apply a concept applied towards a plurality to a member of that plurality on the
condition of learning that that member of the plurality is a member of the plurality. This
may seem disturbingly similar to the rejected condition that Green be disposed to apply
the blackmailed-by-Body concept to Scarlet on the condition of learning what the conver-
sationalists commonly assume, that Scarlet is a socialist, or a member of the plurality of
socialists. That condition, recall, was rejected because, when generalized and applied to
204
other cases, it led to erroneous prediction in cases like Swtiched Blackmail
4
,inwhichthe
subject would apply the target loaded concept to what the speaker applies that concept to
in issuing her report, on the condition of learning what the conversationalists commonly
assume. OnemightthereforeworrythatRational Concept-Application Augmentation View
similarly makes erroneous predictions for Swtiched Blackmail
4
.
This worry is unfounded. For Rational Concept-Application Augmentation View does
not require that the subject of a report be such that they would be rationally committed
to applying the target loaded concept to what the speaker is conversationally assumed to
apply that concept to in issuing her report on the condition of learning certain things that
the conversationalists commonly assume. Rather, this principle requires that a subject in
fact be rationally committed to applying the target loaded concept to object o cognized
via a concept for any object o and concept such that the speaker is assumed to apply
that the target concept towards o, cognized via , in issuing her report. It just so hap-
pens in the case of Blackmail that the relevant rational commitment of Green’s to apply
blackmailed-by-Body to Scarlet, cognized as a member of the socialists, whom Black is con-
versationally assumed to apply blackmailed-by-Body to in issuing her report, is tantamount
to a commitment to apply that concept on the condition of learning that Scarlet is a mem-
ber of the plurality of socialists, which just so happens to be something conversationally
assumed. The application of Rational Concept-Application Augmentation View does not
for this reason make use of this conditional rational commitment in accounting for the
permissive reading of (2) in Blackmail. Rational Concept-Application Augmentation View
does not carry any commitment to saying that in Switched Blackmail
4
Green is rationally
committed to applying blackmailed-by-Body to Jade’s sister, or Scarlet, in virtue of apply-
ing that concept to White, even though were Green to learn what the conversationalists
assume in that concept he may be so committed. For, in holding his fully loaded thought
that White was being blackmailed by Body, Green is not as such rationally committed
to applying blackmailed-by-Body to Scarlet, cognized as a member of those blackmailing
White, or as a member of those blackmailing someone.
So it looks like appealing to facts about how a subject is rationally committed to ap-
plying target loaded concepts does help to protect contextual augmentation views against
the problems that led us into aporia in the previous chapter. By distinguishing the truth-
conditional question of permissiveness from the pragmatic question of permissiveness, and
by building in added constraints about how subjects are rationally committed to apply
target loaded concepts, we are able to handle the counter-examples to contextual augmen-
tationviewsdiscussedinthischapterandinthepreviouschapter. Now, ofcourse, Rational
Concept-Application Augmentation View still crucially uses the relation of non-trivial and
distinctive contextual entailment in accounting for the truth condition of partially permis-
sive reports, so it still is subject to the problem of attitude variation, discussed briefly in
this section. But I hope that this section has given us further reason for optimism that
a suitably framed contextual augmentation view, together with plausible pragmatic prin-
ciples, can handle the truth-conditional problem of permissive reports. In the next and
205
concluding chapter I speculate in further detail about how to use the notion of rational
commitment in resolving the problem of attitude variation and other open problems with
contextual augmentation views.
206
Chapter 7. Wrapping Up
InthisconcludingchapterIconsiderhowwellourgoingpragmaticprinciplescandefend
something like the contextual augmentation view developed in chapters 5 and 6 against
problems raised in chapter 3 for contextual augmentation views. I then revisit what I take
to the most important views presented in this dissertation.
1. Addressing Remaining Problems
In chapter 3, I presented five di↵ erent problems for contextual augmentation views.
The so-called problem of erroneous assumptions was resolved in chapter 2 by claiming
that contextual augmentation must be understood in terms of conversational common
knowledge rather than conversational common assumption. In chapters 5 and 6 I argued
thatattentiontocertainpragmaticphenomenahelpsustodissolvetheasymmetryproblem
and the acquaintance problem. So two problems remain. Here I speculate about whether
our pragmatic-sensitive picture can help to resolve these open problems.
1.1. The Attitude Variation Problem
At the end of chapter 6 I discussed the attitude variation problem and a promising
way of revising contextual augmentation views to handle it. Recall that the basic shape
of the attitude variation problem is that if we construct a context just like one involving a
permissive report that a subject “believes” or “thinks” some proposition by replacing the
main attitude verb with another, such as ‘fears’ or ‘doubts’, that the resulting context is
one in which the report fails to take the same kind of permissive reading that it does in
the original. For example, contrast the following two cases once again:
Blackmail: Detectives Black and Grey know that Scarlet is a socialist. The detectives
alsocommonlyknowthatGreen, anotherguestatthemanor, isunacquaintedwithScarlet.
Green has never met Scarlet nor is he competent with any name for Scarlet. Black knows
that Green thinks that every socialist guest was being blackmailed by Mr. Body. In
discussing whether Scarlet had a motive for killing Body, Black assertively utters the
following to Grey:
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
As observed many times over, (2) takes a permissive reading with respect to ‘Scarlet’
in Blackmail. The report also reads true.
Blackmail Fear: Detectives Black and Grey know that Scarlet is a socialist. The
detectives also commonly know that Green, another guest at the manor, is unacquainted
with Scarlet. Green has never met Scarlet nor is he competent with any name for Scarlet.
Black knows that Green has made a bet that not all socialists were being blackmailed
by Body, so Black also knows that Green therefore fears that every socialist is being
blackmailed by Body. In discussing whether Scarlet had a motive for killing Body, Black
207
assertively utters the following to Grey:
(87) Green fears that Scarlet was being blackmailed by Mr. Body.
As previously observed, (87) takes a loaded reading with respect to ‘Scarlet’ in Black-
mail Fear. The report also reads false.
The main di↵ erence between these two cases is that Green’s relevant fully loaded at-
titude in Blackmail is an attitude of thinking that every socialist was being blackmailed
by Body and in Blackmail Fear it is an attitude of fearing that every socialist was being
blackmailed by Body. Corresponding to this di↵ erence, the cases di↵ er in which attitude
Black reports Green as bearing towards the proposition that Scarlet was being blackmailed
by Body. In Blackmail (2) intuitively takes a true reading, but in Blackmail Fear (87) in-
tuitively takes a false reading. So simple contextual augmentation views that account for
the truth conditions of attitude reports in terms of contextual entailment face a problem,
since they are committed to predicting true readings or false readings for both cases.
At the end of chapter 6, I began developing a version of the contextual augmentation
viewthatreplacesthecentralrelationofcontextualentailmentwitharelationofcontextual
rational commitment. In this section I pursue this revised view in further detail and see if
it can handle the attitude variation problem.
The basic idea behind the revision of contextual augmentation views briefly proposed
in chapter 6 is to take the relevant sense in which contextual assumptions can serve to
“augment” a subject’s loaded attitudes to verify a permissive report is that a permissive
report represents a subject as being such that, insofar as they are committed to their fully
loaded attitudes, were they also committed to certain contextual assumptions made in the
reporting context, the subject would be rationally committed to holding a fully loaded
attitude corresponding to the attitude ascribed to them by the permissive report.
On this view, then, permissive reporting is in e↵ ect a way of drawing out what fully
loaded attitudes a subject would be rationally committed to holding were they to be com-
mitted to what the conversationalists commonly assume in the reporting context. Recall
the final working version of the contextual augmentation view from chapter 3. I’ll state
it in two parts, consisting of the basic sensitive attitudes view together with the special
contribution of the contextual augmentation idea.
Sensitive Attitudes View:
A report made by a sentence of the formAVsthatS made in context c is true if
and only if A bears V
c
to the proposition designated by that S in c,where V
c
is
the particular attitude relation contributed by V in c (which need not be identical
to V
c2
, the attitude relation contributed by V in context c2).
Again, the basic idea behind Sensitive Attitudes View is that a report sentence of the
formAVsthatS can be used to assert di↵ erent proposition in di↵ erent contexts since
the attitude verb V can be used to contribute di↵ erent attitude relations across di↵ erent
208
contexts. WhetherareportmadewithAVsthatS inacontextcistruedependsonwhether
A bears the attitude relation contributed by V in c towards the proposition designated by
that S in c.
Together with Sensitive Attitudes View, we can express the final version of the contex-
tual augmentation view discussed in chapter 3 with
Sophisticated Contextual Entailment View:
A bears V
c
to propositionr if and only if there is a propositionp such thatA bears
V
L
to p and there are propositions q
1
...q
m
that are conversationally assumed in
c such that p together with q
1
...q
m
non-trivially and distinctively entails the
proposition designated by that S in c.
Revising this kind of principle instead in terms of rational commitment, I suggest
Rational Commitment Augmentation View:
A bears V
c
to propositionr if and only if there is a propositionp such thatA bears
V
L
to p and there are propositions q
1
...q
m
that are conversationally assumed in c
such that were A committed to bearing V
L
towards p and were also committed to
bearing believe
L
towards q
1
...q
m
,then A would be rationally committed to
bearing V
L
towards r
The idea behind Rational Commitment Augmentation View is that, for a report made
that predicates an attitude relation V
c
to a subject A and proposition r, that report is true
just in case wereA committed to both bearing V
L
to the propositions to which they in fact
do bear V
L
to and to bearing believe
L
to certain propositions contextually assumed in c,
then A would be rationally committed to bearing V
L
to r.
The notion of “commitment” here is meant to be minimal. The main feature of the
notion of commitment needed is that if one is committed to bearing attitude relation V
to proposition p, then even were one to come to learn some propositions q
1
...q
m
, then one
would retain their original attitude V towards p.
125
The idea then is that were A both to
bear V
L
towards what they do bear that fully loaded relation towards and were also at
the same time to bear believe
L
towards propositions q
1
...q
m
, which are conversationally
assumed in c,then A would be rationally required, or “rationally committed”, to bear V
L
125
This serves to distinguish Rational Commitment Augmentation View from the counterfactual attitudes
view discussed early in chapter 3. According to that view, A bears Vc towards r if and only if, given A’s
fully loaded attitudes, were one to learn certain conversational assumptions q1...qm then one would come
to bear VL towards r. This ran into problems because in some cases in which one bears VL towards a
proposition p, were one to learn propositions q1...qm such that r is entailed (in some way) by p and q1...qm,
one would not come to believe r,butinsteadwould abandon their original loaded attitude in p.
209
towards the proposition designated by that S in c.
126
Let’s try out Rational Commitment Augmentation View on various instances of the
attitude variation problem. At first we find happy results with Blackmail and Black-
mail Fear.In Blackmail, we have, on our sensitive attitudes picture, that (2) predicates
think
Blackmail
of Green and the proposition that Scarlet was being blackmailed by Body.
In that scenario, Green is contextually assumed to bear think
L
, a fully loaded “thinks”
relation, towards the proposition that every socialist was being blackmailed by Body, and
the conversationalists assume that Scarlet is a socialist. Were Green committed to bearing
thinks
L
towards the proposition that every socialist was being blackmailed by Body and
were committed to the proposition that Scarlet is a socialist, then intuitively Green would
be rationally committed to bearing a fully loaded thinks relation towards the proposition
thatScarletwasbeingblackmailedbyBody. SoRational Commitment Augmentation View
does fine with Blackmail.
In Blackmail Fear, Green bears fear
L
, a fully loaded “fears” relation, towards the
proposition that every socialist was being blackmailed by Body, and the conversationalists
assume that Scarlet is a socialist. Were Green committed to bearing fears
L
towards the
proposition that every socialist was being blackmailed by Body and were committed to
bearing believes
L
to the proposition that Scarlet is a socialist, then Green would not
intuitively be rationally committed to the proposition that Scarlet was being blackmailed
by Body. So Rational Commitment Augmentation View does not predict that (87) should
take a true reading in Blackmail Fear.
So far, so good. But there are further revision of Rational Commitment Augmentation
View needed. Looking to switched contexts like those discussed in chapter 5 is enough to
show that Rational Commitment Augmentation View as stated is not enough. Reconsider
Switched Blackmail: Mr. Body has been murdered; detectives Black and Grey are
investigating the crime. The detectives have private information that every socialist guest
was being blackmailed by Mr. Body. They know that Green is unaware that Body was
blackmailing any guests at all. Having just questioned Green, Black knows that Green
believes that Scarlet (with whom Green is well acquainted) is a socialist. Although the
detectives commonly know that Scarlet is a guest, they do not know that she is a socialist.
They are exchanging information, trying to find out who had a motive for killing Body;
in particular, in this scenario they are trying to figure out who was being blackmailed by
Body. Black assertively utters the following to Grey:
126
I leave the notion of rational commitment involved in this principle somewhat open, even though the
details will surely matter in a fully developed version of this view. For example, one question we can
ask is whether one isrationally committed to believing every logical truth, regardless of what one’s other
fully loaded attitudes are like? It may be that the notion of rational commitment needed for this kind of
contextual augmentation view to account for the application conditions of partially permissive attitudes
will di↵ er from certain idealized notions of rationality that certain epistemologists and decision theorists
might care about. The kind of rationality involved will have to be probed by investigating a wider range of
cases.
210
(2) Green thinks that Scarlet was being blackmailed by Mr. Body.
What does Rational Commitment Augmentation View say about this case? Well, were
Green committed to bearing thinks
L
towards the proposition that Scarlet is a socialist
andwerealsocommittedtobearing believes
L
towardsthepropositionthateverysocialist
was being blackmailed by body, then clearly Green would be rationally committed to the
proposition that Scarlet was being blackmailed by Body. So our simple Rational Commit-
ment Augmentation View is committed to a true reading of (2) in this case. Clearly, this
principle su↵ ers from being insensitive to subtleties observed in chapters 5 and 6 concern-
ing the important distinction between the pragmatic question of permissiveness and the
truth-conditional question of permissiveness. As is obvious given what has been discussed
throughout this dissertation, we need a version of Rational Commitment Augmentation
View that is sensitive to this distinction; something like the principles discussed at the end
of chapters 5 and 6.
At the end of chapter 6 we landed on the following view:
Rational Concept-Application Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) made by
speaker S in context c,where that S(F) designates proposition p in c,istruei↵ there is a proposition q such that for every quadruplehp, ,f,oi with respect to
which R is loaded, f occurs in q in a structurally analogous position to o in p, and
A uses to cognize f as it occurs in that position in q in bearing V
L
to q and in
so doing is rationally committed to applying with respect to object o as o is
cognized via concept ,where o is any object and any concept such that it is
conversationally assumed in c that in using R, S applies to o as o is cognized
via .
This principle served us well with respect to some of the pressing problems in chapters
5 and 6, but it noticeably does not include the central notion of contextual augmentation
views, that a permissive report is true if and only if the relevantly related loaded attitudes
of the subject, together with certain conversational assumptions, bear the right kind of
augmenting relation to the proposition designated by the complement of a report. There
is nothing in Rational Concept-Application Augmentation View that requires that the con-
tents of the subject’s loaded attitudes entail, together with contextual assumptions, p, or
instead that the the subject’s loaded attitudes themselves, together with a commitment
to certain conversational assumptions, would rationally commit the subject to bearing the
relevant attitude V
L
towards p.
The simplest way to fix this is to tack the central condition of Rational Commitment
Augmentation View onto the already elaborate condition of Rational Concept-Application
Augmentation View. Since this combination leads to a complicated principle, I break up
my representation of this principle into parts.
211
Sophisticated Rational Augmentation View:
A report R made by a use of a sentence of the formAVsthatS(F) made by
speaker S in context c,where that S(F) designates proposition p in c,istruei↵ (i. Rational Concept-Application Condition)
There is a proposition q such that for every quadruplehp, ,f,oi with respect to
which R is loaded, f occurs in q in a structurally analogous position to o in p, and
A uses to cognize f as it occurs in that position in q in bearing V
L
to q, and in
so doing is rationally committed to applying with respect to object o as o is
cognized via concept ,where o is any object and any concept such that it is
conversationally assumed in c that in using R, S applies to o as o is cognized
via ,
and
(ii. Rational Commitment Condition)
There are propositions q
1
...q
m
to which A bears V
L
such
that were A committed to bearing V
L
towards q
1
...q
m
and were also committed to
bearing believe
L
towards propositionsp
1
...p
n
that are contextually assumed inc,
then A would be rationally committed to bearing V
L
towards the proposition
designated by that S in c.
A mouthful, to be sure. But the basic idea is just that a report of the formAVsthat
S(F) made in context c is true if and only if (i) the subject conceptualizes the constituents
of the report’s complement with respect to which the report is loaded as the speaker
conceptualizesthoseconstituentsinissuingherreport, and(ii)werethesubjectcommitted
to their actual fully loaded V attitudes and also were committed to believing certain
contextual assumptions, then they would be rationally committed to holding a fully loaded
V relation towards the proposition designated by that S(F) in c. Permissive reporting,
on this view, serves, in a way, to represent how a subject should conceive of and think
about certain things in the world, given the fully loaded attitudes that they have: the
Rational Concept-Application Condition captures roughly how a subject should conceive
of certain things given the way that she uses certain concepts in her fully loaded thought
life, Rational Commitment Condition captures roughly what certain fully loaded attitudes
a subject should have, given certain ones she already has and in view of additional truths
known by the speaker. I expect that this principle could be refined even further, but I also
think that it does well enough in making sense of the attitude variation problem.
To test this principle out a little bit, I’ll use it to account for a few examples.
Blackmail Doubt: FrominterrogatingGreen,BlackknowsthatGreendoubtsthatevery
socialist was being blackmailed by Body. Green thinks that probably some of the socialists
inthemanorwerebeingblackmailedbyBody,butisfarfromconvincedthatallofthemare,
since he thinks that some of the socialists kept their political a liations hidden from Body.
The detectives know that Scarlet is a socialist and know that Green is unacquainted with
212
Scarlet. In discussing who Body was blackmailing, Black assertively utters the following
to Grey:
(88) Green doubts that Scarlet was being blackmailed by Mr. Body.
Similarly to our judgment in Blackmail Fear, in this case (88) intuitively takes a false
reading. Arguably, it takes a permissive reading with respect to ‘Scarlet’ and a loaded
reading with respect to ‘blackmailed by Mr. Body’. This is straightforwardly explained by
Sophisticated Rational Augmentation View, sincewereGreentoholdhisfullyloadeddoubt
towards the proposition that every socialist was being blackmailed by Body and were he
committed to believing that Scarlet is a socialist, this would not rationally commit him to
having a fully loaded doubt that Scarlet was being blackmailed by Mr Body.
Contrast this case with a slight alternative:
Blackmail Doubt
2
: From interrogating Green, Black knows that Green doubts that
every socialist was being blackmailed by Body. Green doubts as much because he doubts
thatanyindividualsocialistwasbeingblackmailedbyMr. Body. Hethinksthatitisjusta
rumorthatBodywasblackmailingsocialists. ThedetectivesknowthatScarletisasocialist
and that Green is unacquainted with Scarlet. In discussing who Body was blackmailing,
Black assertively utters the following to Grey:
(88) Green doubts that Scarlet was being blackmailed by Mr. Body.
In this context intuitively (88) takes a true reading that is permissive with respect to
‘Scarlet’. Sophisticated Rational Augmentation View explains as much, since in contrast to
Blackmail Doubt, inthiscaseGreen’sfullyloadeddoubtsaresuchthat, werehecommitted
to those attitudes and to the proposition that Scarlet is a socialist, then he would be
rationally committed to holding a fully loaded doubt that Scarlet was being blackmailed
by Mr. Body.
1.2. The Irrelevant Entailments Problem
Irrelevant entailment problems were the first problems that we encountered in chapter
3 for contextual augmentation views. Several of these problems can be handled just by
insisting on the importance of the distinction between the pragmatic question of permis-
siveness and the truth-conditional question of permissiveness. By drawing this distinction,
we can explain the false reading in many irrelevant entailment cases by applying the con-
dition that, roughly, for a report that is loaded with respect to some concept to be true
the subject must be rationally committed in virtue of the way they apply that concept in
their fully loaded attitudes to applying that concept to what the speaker of the report is
contextually assumed to apply it to in issuing her report. To illustrate the promise of So-
phisticated Rational Augmentation View in handling irrelevant entailment cases, I simply
review the instances of that problem which I raised in chapter 3. Given how tricky these
examples of permissive reporting can be, this section by no means serves as a conclusive
213
demonstration that our new contextual augmentation view lays the irrelevant entailments
problem to rest. But that our new principle is successful in handling the first batch of
problems for earlier contextual augmentation views does lend it some support.
Let’s begin with
Forgeries: DetectivesBlackandGreyknowthatallandonlytheimpressionistpaintings
in the gallery are forgeries, and also know that this is a secret, so that none of the guests in
the manor know that there are any forgeries in the gallery. Peacock, a guest in the manor,
has told detective Black that she believes that every impressionist painting in the gallery
is an original Renoir. Later, in discussing matters related to the gallery, Black assertively
utters the following to Grey:
(89) Peacock believes that every forgery in the gallery is an original Renoir and that we
are having this conversation.
This example posed a problem for earlier contextual augmentation views because the
complement of (89) is non-trivially and distinctively contextually entailed by Peacock’s
loaded belief that every impressionist in the gallery is an original Renoir, since it is con-
textually assumed in Forgeries that Black and Grey are having the conversation they are
having.
Sophisticated Rational Augmentation View,whichissensitivetothedistinctionbetween
the truth-conditional question of permissiveness from the pragmatic question of permis-
siveness, does better. For in Forgeries intuitively the latter clause in the complement of
(89), ‘we are having this conversation’, takes a loaded reading. Therefore, on our new
version of the contextual augmentation view, Peacock must use concepts associated with
the constituents of this clause as part of her fully loaded attitudes in order for the report
to take a true reading. But she doesn’t, so we accurately predict that (89) takes a false
reading.
Next reconsider
Protest: There are rumors that some guests at the manor are secretly anarchists, that
others are secretly socialists, and that others still are secretly fascists. From interrogating
Scarlet, detective Black knows that Scarlet believes that some anarchists or socialists will
protest tonight. Scarlet isn’t confident whether anarchists, socialists, or both will hold a
protest; she just believes that some people from those groups will put on some kind of
political demonstration tonight. The detectives know that the rumor that there are fascist
guests hasn’t gotten around to Scarlet. So they know that Scarlet neither believes nor
disbelieves that there are fascist guests; they don’t assume that she has any particular
fascist-related beliefs. The detectives further know that the rumors about anarchists and
fascists are unfounded: no guest is fascist or anarchist. But, as the detectives know, some
guestsaresocialists. AwareofScarlet’sbeliefs,BlacksincerelyuttersthefollowingtoGrey:
(47) Scarlet believes that some fascist or socialist guests will protest.
214
Thisstrangecaseposedaproblemforearliercontextualaugmentationviews,sincegiven
the contextual assumption that there are no fascists or anarchists, that some anarchist or
socialist guests will protest contextually entails that some fascist or socialist guests will
protest. So on earlier augmentation views, (47) should take a true reading in Protest,
but it doesn’t. Sophisticated Rational Augmentation View does better here again just by
being sensitive to the distinction between the pragmatic question of permissiveness and
the truth-conditional question. In Protest, Black’s use of (47) intuitively, it seems, takes
a loaded reading with respect to ‘fascist or socialist guests’, so for (47) to be true Scarlet
must use concepts associated with that phrase as part of a fully loaded attitude that would
rationallycommithertohavingafullyloadedbeliefthatsomefascistorsocialistguestswill
protest were she also to be committed to what the detectives assume. Since Scarlet does
not use the concept fascist in holding any such attitude, our new contextual augmentation
view readily and accurately predicts a false reading in this case.
Next reconsider
Chase: Professor Plum is chasing Mr. Green through the manor. Given the trajectory
which he was just running along, Plum believes that Green is either in the billiard room or
in the library. The detectives know that Green didn’t go into the library. Aware of Plum’s
belief, detective Black sincerely utters the following to Grey:
(48) Plum believes that Green is in the billiard room.
In this minimal context, intuitively ‘the billiard room’ takes a loaded reading in Black’s
useof(48). AccordingtoSophisticated Rational Augmentation View, therefore, Plummust
at least use the conversationally associated concept in-the-billiard-room in holding a fully
loaded attitude in this case. That he does. It further must be that the propositional con-
tribution of ‘in the billiard room’, in-the-billiard-room occurs in the content of Plum’s
fully loaded belief in a structurally analogous position to the occurrence of that propo-
sitional constituent in the proposition designated by the complement of (48) in Chase.
This isn’t obvious, but it may be so. But yet further, according to our new version of the
contextual augmentation view, it must be that in using the concept in-the-billiard-room
in cognizing this occurrence of in-the-billiard-room in holding his fully loaded belief,
Plum is in fact rationally committed to applying that concept to what Black does in is-
suing her report, namely, to Green. This condition does not seem to be met. Since Plum
uses the concept in-the-billiard-room in his relevant fully loaded belief just in predicating
the disjunctive property in-the-billiard-room-or-in-the-library of Green, he is not
thereby rationally committed to applying in-the-billiard-room to Green. Our new contex-
tualaugmentationviewthenaccuratelypredictsthat(48)takesafalsereadinginthiscase,
given the loaded reading of ‘in the billiard room’.
Next reconsider
Knife: Professor Plum has been conducting his own investigation about the murder.
215
Heisn’tconfidentyetwhichweaponisthemurderweapon,buthethinksthatifthemurder
happened in the ballroom then the knife is the murder weapon. The detectives know that
the murder happened in the ballroom. Aware of Plum’s beliefs, Black sincerely utters the
following to Grey:
(49) Plum believes that the knife is the murder weapon.
The explanation of the false reading of (49) in Knife is similar to the previous expla-
nation of (48) in Chase. It begins with observing that in this minimal context we detect
a loaded reading with respect to ‘is the murder weapon’. Indeed, intuitively the entire
complement of (49) takes a loaded reading in this case. Given our discussion of multiply
loaded readings in chapter 6, this makes accounting for the false reading of (49) poten-
tially easier. The report made with (49) is loaded with respect to both hp,knife,knife,ai
and hp,is-the-murder-weapon,is-the-murder-weapon,bi (where p is the proposition des-
ignated by the complement of (49) in the context and a and b are the occurrences of
knife and is-the-murder-weapon in p). According to Sophisticated Rational Augmenta-
tion View, therefore, for the report made with (49) to be true, Green must have a single
fully loaded belief in the content of which knife and is-the-murder-weapon occur in
structurally analogous positions to their occurrences in p. This arguably is not so, since
these propositional constituents occur in the consequent of a conditional proposition in
Plum’s fully loaded belief, but do not so occur inp. But we do not need to rely on a notion
of structurally analogous occurrences that lets us say this. For it remains that Plum is not
rationally committed in his use of is-the-murder-weapon in holding his fully loaded belief
in this case to applying that concept towards what Black is assumed to apply that concept
to in issuing her report, namely, the knife. For this reason alone our new contextual aug-
mentation principle accurately predicts a false reading of (49), given the loaded reading of
‘is the murder weapon’.
This concludes my brief treatment of the irrelevant entailment problem. We’ve seen
that the original instances of this problem dissolve as soon as we adopt a principle that is
suitably sensitive to the distinction between the pragmatic question of permissiveness and
the truth-conditional question of permissiveness. More work surely is needed to show that
our new contextual augmentation view does not somehow run into analogous problems of
irrelevant rational commitments, but I leave a full treatment of that issue for another time.
2. Final Recap
This dissertation has addressed questions about attitudes by focusing on the semantics
andpragmaticsofattitudeverbs. Thediscussionhasshowncertainideastobeparticularly
important. By way of conclusion, I zoom out from the details of particular examples to
summarize these ideas and say why they are important.
Mostrecently,wehaveseenthatthatitishelpfultounderstandourpermissivereporting
practices as tracking how a subject is contextually rationally committed to thinking and
216
conceiving of things given her fully loaded mental states, rather than as tracking just what
is simply contextually entailed by certain of those loaded states. The linguistic meanings
of attitude verbs, on this picture, are open-ended constraints in that a speaker can abide
by them both in using an attitude verb to directly represent how a subject is thinking and
conceiving of things in the world, but also in using such verbs to represent how a subject
rationallyshould thinkandconceptualizethingsgiventheattitudestheyholdinconceiving
of things as they do and in view of truths known by the speaker.
While more investigation is needed concerning the exact ways in which permissive
reporting can serve to draw out certain rational consequences of a subject’s fully loaded
attitudes, already our provisional version of this idea has helped us in two ways. First,
it has helped us to articulate how the attitude relations designated in partially loaded
reportsrepresentsubjectsasusingconceptswithrespecttowhichthosereportsareloaded.
The picture that permissive reporting partially serves to draw out how a subject should
conceive of things given the way they in fact do conceive of things motivates the Rational
Concept-Application Condition. We’veseenthatthisconditionhelpstounifythewaysthat
pragmatic “acquaintance” constraints bear on the truth conditions of permissive reporting
bothforreportsthatarepermissivewithrespecttobarenounphrasesandpredicatesaswell
as for universal instantiation reports that are permissive with respect to names and other
simple terms. Second, this rationality-theoretic picture has helped us reform contextual
augmentation views to handle problems raised by interesting di↵ erences between reports
made with di↵ erent kinds of attitude verbs. Permissive reporting is not specific to belief
reports, which have dominated discussion in this dissertation; the rationality-theoretic
perspective helps us to acknowledge permissive reporting as the quite general practice that
it is while retaining the intuitions underlying the basic contextual augmentation approach.
Anotherimportantideawehavecometoappreciateinlaterchaptersofthisdissertation
is that we must take care to distinguish the truth-conditional question of permissiveness
from the pragmatic question of permissiveness. Before one familiarizes oneself with some
of the complexities of permissive reporting, it might sound obvious or trivial that these
questions should be distinguished. But we have seen how natural it is to jump to certain
views that are insensitive to this distinction in response to simple examples of permissive
reporting.
I speculate that one reason is it easy to lose sight of this important distinction is that
althoughthesequestionsaretheoreticallydistinct,theyarebothsensitivetoconversational
features, and indeed, are sensitive to some of the same features of conversational contexts.
Whatever the correct view to the pragmatic question of permissiveness is, it obviously
is sensitive to conversational features. Whether a report takes a permissive reading with
respect to a certain phrase depends on what questions are being addressed, what the
conversationalistsassumeaboutwhatthesubjectknowsandwhatthesubject’sconceptual
repertoire is like, which objects and properties are salient in the context in certain ways,
and so on.
217
Insofar as the rational commitment augmentation view is on the right track, we see
that an adequate answer to the truth-conditional question must also be sensitive to sim-
ilar contextual features. In specifying the application conditions of a partially permissive
attitude relation that is designated by a use of an attitude verb in some context, we have
to say which concepts the subject must use or be rationally committed to using in certain
ways, and these of course happen to be concepts with respect to which the corresponding
attitude report takes a loaded reading. Similarly, in specifying the application conditions
of such a partially permissive attitude, we have to track whether the subject would be
rationally committed to holding certain other fully loaded attitude relations were they to
accept certain propositions assumed in the reporting context. So an adequate answer to
the truth-conditional question of permissiveness, though this question is distinct from the
pragmatic question, has to be sensitive to conversational details—often the same conversa-
tional details to which an adequate answer to the pragmatic question is sensitive. Perhaps
it is because both the truth-conditions of permissive reports and the conditions in which
a report takes a permissive reading are sensitive to many of the same contextual features
that we are inclined to conflate these questions.
But whatever the reason is that we sometimes don’t, we have seen how it important
it is that we do distinguish the truth-conditional question from the pragmatic question.
In the first instance, it is by drawing this distinction and by relativizing the ideology of
permissiveness to particular occurrences of particular phrases that we can easily charac-
terize a report as permissive (with respect to a particular phrase occurrence) yet as false,
or characterize a report as loaded (with respect to some phrase occurrence) yet as true.
This fineness of discrimination in describing reports has been indispensable in appreciat-
ing di↵ erent phenomena connected to permissive reporting. Furthermore, by drawing this
distinction we have been able to recast contextual augmentation views to be sensitive to
how di↵ erent reports are loaded with respect to di↵ erent phrase occurrences and concepts,
which has helped us avoid a host of counter-examples for simpler contextual augmentation
views. We last saw this in §1.2 above with the problems of irrelevant entailments; it is
encouraging at least these initial instances of that kind of problem all immediately dissolve
onceweoutfitourtruth-conditionalviewtodistinguishthetruth-conditionalquestionfrom
the pragmatic question.
The final views I want to highlight here are the two related ideas with which we began
the dissertation. The first is my version of the sensitive attitudes view, on which attitudes
verbs are semantically underspecified in such a way that they can be used to designate
di↵ erent attitude relations in di↵ erent contexts, where some of these are conceptually per-
missive. I’vetalkedaroundthisideainapplyingitmanytimesthroughoutthedissertation.
Bywayofcharacterizingthesensitiveattitudesviewonelasttime,Itakeaslightlydi↵ erent
approach and o↵ er the following thought experiment.
Imagine a community in which the conventional constraints on how to use attitude
verbs were quite strict, as strict as is imagined by the strict view discussed in chapter 2. In
218
suchacommunity,onecanonly(literally)ascribefully loaded attitudestosubjects. Having
seen how usefully flexible our actual reporting practices are throughout this dissertation,
we can expect that such strict conventions would be a bother. Often the reasons we have
for communicating about how one another think about and conceive of things in the world
are such that we don’t ultimately care about what one another’s di↵ erent fully loaded
mental states are like. There are diverse reasons to convey certain aspects of someone’s
fully loaded mental state but not others or to convey just certain ways in which someone
is disposed to behave given how their means of representing stu↵ in the world inform
their di↵ erent fully loaded beliefs and desires. Similarly, there are occasions in which
we have reason to communicate about how one another thinks about and conceives of
things in an approximate way, occasions in which we judge that those we have reason to
communicatewitharen’tinapositiontoappreciatethepreciseconceptualmeansdeployed
by those whose thought life we want to talk about. Given the many and diverse reasons
we have to convey partial or approximate information about one another’s fully loaded
mental states, a strict language that enforced that any attitude ascription must report a
subject’s fully loaded attitudes would simply be a hassle, forcing us to explicitly draw out
the consequences we actually care about, or forcing us to go through the work of equipping
our interlocutors with the conceptual repertoire to appreciate the fully loaded reports we
would be conventionally required to make.
We can expect that a community burdened by such strict linguistic conventions, for
reasons of optimizing e cient communication, would start to cut corners. Their practices
of attitude reporting would plausibly begin to diverge from their community’s strict rules,
with speakers flouting these strict conventional rules, leveraging reasonable general prag-
matic expectations that those they communicate with are rational and cooperative enough
to recover what they intend to convey with their remarks and to take them as committed
to those propositions rather than to those prescribed by their burdensome linguistic con-
ventions. These corner-cutting practices could be expected to become prevalent enough
in the population to ground new, looser conventions: conventions like those postulated by
the underspecification version of the sensitive attitudes view. The linguistic rules govern-
ing attitude verbs would slacken to fit our multifarious reasons for talking about how one
another think about and conceive of things in the world. We would opt for conventional
linguistic constraints that are underspecified, that legislate only as needed, and that leave
nuances of what we might want to assert using attitude verbs that could be recovered by
general pragmatic reasoning to be recovered in just that way. Our linguistic conventions
would, given a background of a population of communicators who are known ultimately
to be interested in recovering what those they talk with intend to communicate, take the
path of least resistance and allow many designation candidates for each attitude verb.
Now, even in this fanciful hypothetical example in which underspecified linguistic con-
ventions are, historically speaking, dissipated descendants of narrow linguistic conventions
abiding by the strict view of attitudes, once such open-ended conventions are in place,
when one abides by such underspecified conventions to communicate that Green has a
219
conceptually permissive belief towards the proposition that Scarlet was being blackmailed,
one literally asserts that Green believes that Scarlet was being blackmailed. And if what
one asserts is true, it is literally true.
Moreover, even given this imagined history of such underspecified linguistic meanings
for attitude verbs, there would be no distinguished designation candidates of any attitude
verb which would count as the “real” beliefs, the “real” hopes, and so on. The view, again,
is not just that attitude verbs have a primary lexically encoded meaning, on which they
designate loaded relations, and that permissive uses are literally true but derivative on
such primary readings. The sort of pragmatic enrichment I see as involved here is not
the kind sometimes discussed in the literature, where pragmatic enrichment just serves to
“develop” or “expand on” the conventional standing meaning of an uttered sentence, or
serves to add “unarticulated constituents” to a sentence’s conventional semantic content.
The way in which pragmatic influence informs the context-sensitivity of permissive atti-
tude reports, then, is di↵ erent from the pragmatic influence plausibly responsible for the
context-sensitivityofcertainreadingsofsomepolysemousexpressions. ConsiderAristotle’s
comment:
Everything which is healthy is related to health, one thing in the sense that it
preserves health, another in the sense that it produces it, another in the sense that
it is a symptom of health, another because it is capable of it.
127
Corresponding to this idea from Aristotle, on one picture, that sentences like ‘His diet
is healthy’ and ‘His urine is healthy’ are often literally true is due to uses of those sentences
serving to assert pragmatically enriched propositions that expand on the “primary” mean-
ing of health. The only designation candidate of ‘healthy’ is the property healthy,which
is instantiated by certain people and animals, but some uses of ‘healthy’ take derivative
meanings due to pragmatic expansion: one says something like that his diet is [conducive
to being one who is] healthy, and something like that his urine is [symptomatic of one who
is] healthy. What one says is literally true, but it is derivative on some primary, central
meaning of health; one property you can predicate with ‘healthy’ is the “real” kind of
healthy.
On my way of understanding the underspecification of attitude verbs, this is not the
sort of thing that is going on in cases of permissive reporting. There is not one primary (or
sole)designationcandidatedeterminedbytheconventionallinguisticmeaningof‘believes’,
which then is expanded upon or tweaked by pragmatic processes to yield the propositions
asserted by permissive reports. One reason not to think this, which is supported by sev-
eral discussions in this dissertation, is that plausibly there are many distinct fully loaded
believes relations. We have noticed, primarily in chapters 4 and chapter 6, that a report
is not simply loaded with respect to a phrase or a phrase’s propositional contribution per
se, but also is loaded with respect to certain associated concepts. Given this, we should
127
Aristotle, (1984).
220
expect that there are many distinct fully loaded relations associated with any particular
attitude verb.
I know I said I wouldn’t, but I can’t help but discuss one last example. I do so first
to reinforce the point just made, that given the way we have been applying the sensitive
attitudes view, we should expect there to be many fully loaded attitude relations. But I
also think that this example helpfully illustrates one way in which, looking beyond this
dissertation, the conceptually permissive attitudes view might be applied to debates in the
philosophy of language and mind. Consider
Mugshots: Therearethreemugshotsonthewallinthedetectives’o ce. Thedetectives
know a secret: all three are mugshots of Scarlet, even though the mugshots look to be of
very di↵ erent people. Scarlet, apparently, is a master of disguise. Black has interrogated
Green, asking him his opinions about the individuals represented in these mugshots. He
is known not to be otherwise acquainted with Scarlet; he doesn’t know her by the name
‘Scarlet’ and has never seen her not in disguise. Later in discussing things with Grey,
Black assertively utters the following sentences, back-to-back (the italicized material in
parentheses describes demonstrative gestures Black makes as she utters what she does):
(90) Green thinks that Scarlet (pointing to mugshot 1) is blackmailing Scarlet (pointing
to mugshot 2).
(91) And Green thinks that Scarlet (pointing to mugshot 3) is blackmailing Scarlet
(pointing to mugshot 2).
(92) But Green thinks that Scarlet (pointing to mugshot 1) and Scarlet (pointing to
mugshot 3) aren’t being blackmailed by anyone.
(93) Of course, given all this, Green thinks that Scarlet (pointing to mugshot 2)is
beingblackmailedbytwopeoplewhoarenotthemselvesbeingblackmailed, namely
Scarlet (pointing to mugshot 1) and Scarlet (pointing to mugshot 3).
(94) So, poor Green thinks that Scarlet is blackmailing Scarlet.
(95) Indeed, he thinks that Scarlet both is and is not blackmailing Scarlet.
It is not hard to describe a scenario in which all of these reports are true. Now, in
contrast to (94) and (95), which each intuitively take permissive readings with respect
to both occurrences of ‘Scarlet’, (90)-(93) all take di↵ erent kinds of loaded readings. Of
particular interest are (90) and (91), which, on certain ways of individuating propositions,
both ascribe to Green a thought in the same proposition. Each takes a fully loaded reading
(presumably with respect to certain descriptive or perceptual concepts related to di↵ erent
ways that Scarlet presents in di↵ erent disguises), but clearly the truth conditions of these
two reports di↵ er: we do not take Black to be repeating herself when she makes her report
with (91). What explains the di↵ erence between these reports? On our version of the
sensitive attitudes view, the explanation is found in that there are distinct fully loaded
221
attitude relations among the designation candidates of ‘thinks’. We see that, insofar as
loadedreadingsarerelativizedtodi↵ erentconcepts,therearedi↵ erentfullyloadedattitude
relations that di↵ er in being so relativized to di↵ erent concepts.
With this example we also begin to see how the conceptually permissive attitudes view
might be applied to some traditional problems associated with so-called Frege puzzles.
This view gives us an independently motivated theoretical framework that lets us say, with
some parties to the debates over Frege puzzles, that (some) instances of this phenomenon
are due to di↵ erence concepts (or senses, or guises, or whatever) being associated with
di↵ erent occurrences of names like ‘Scarlet’. But it also lets us say, with other parties to
such debates, that (some) instances of this phenomenon are due to (certain) uses of names
like ‘Scarlet’ being associated with no conceptual constraints.
Anyway,weseethattherearedistinctfullyloadedattituderelationsamongthedesigna-
tion candidates of any particular attitude verb. So even if we were to restrict the meanings
of attitude verbs in such a way that they can only designate fully loaded attitudes, still
we would be committed to an underspecification version of the sensitive attitudes view.
Attitude verbs would still be semantically underspecified, but their designation candidates
would all be fully loaded attitudes. One could try out such a view, but once one has taken
on board an underspecification version of the sensitive attitudes view, it’s unclear to me,
especially given the many ways we have seen the conceptually permissive attitudes view
can be put to use, why one would not just use this style of account for permissive and
loaded readings both.
Finally, I turn to the second idea central to this dissertation, which is intimately con-
nected with the sensitive attitudes view and which in the introduction I called the “prag-
maticspeechactsview”. Onthisview, whatone(literally)assertswithauseofanattitude
report sentence is sensitive to a wide array of features of conversational contexts, including
conversational goals and conversational assumptions about how the subject uses certain
concepts. Which relation is in fact designated by a use of an attitude verb depends on
norms of pragmatic reasoning of the same kind involved in interpretation generally, includ-
ing in determining what a use of some sentence conversationally implicates. A particular
consequence of this view for interpretation in the case of permissive reporting is that there
are no simple pragmatic constraints we can state about how whether a report takes a per-
missive or loaded reading depends on discrete features of conversational contexts. At best
we can point to rough generalizations about how permissive readings depend on features
of conversational contexts.
I’ll start by reminding the reader about some of the rough generalizations pointed to
in chapters 5 and 6. Chapter 5 began by observing that whether a report sentence serves
to make a permissive report is sensitive in some ways to what questions under discussion
that sentence is used to directly address. Toggling conversational goals across reporting
contexts can toggle whether a use of a report sentence takes a permissive reading. We
soon saw that this kind of pragmatic sensitivity could not be totally captured by principles
222
that lay down necessary constraints on permissive readings just in terms of QUDs and the
associated phenomenon of focus. Instead, the reason why toggling conversational goals
is often relevant to whether we detect certain permissive readings is part of a broader
pragmatic practice.
We found that what was of ultimate pragmatic importance was, roughly, just whether
the most plausible reason that the speaker used some phrase F associated with a concept
in using an attitude report sentence of the formAVsthatS(F) is that they intend to
represent A as using in certain ways in one or another of their fully loaded thoughts.
In many cases, that a speaker uses a report sentence of this form to directly address
the question ?x[S(x)] is evidence that in using the phrase F they intend to represent
the subject as using a concept associated with F in their fully loaded attitudes. The
rough idea here was that since F is conversationally novel relative to the question which
the speaker’s report directly addresses, since the speaker addresses this question with an
attitude report whose complement expresses an answer to the question and which involves
F in the position “targeted” by the question addressed, they convey that features of the
subject’s loadedV attitudes provide this novel material. But this is not always so. We saw
that in cases where questions take a “multiple choice” reading such that certain ways of
expressing answers to ?x[S(x)] involvingF are saliently primed, that the speaker usesF in
issuing her report is not as good evidence that she intends to represent the subject as using
in their fully loaded thoughts, allowing for permissive readings with respect to F and .
Similarly, in some cases in which a speaker uses a report sentence of the formAVsthat
S(F) to directly address a question explicitly about A’s attitudes, without addressing the
question ?x[S(x)], their report takes a loaded reading with respect toF, since the question
aboutA’s attitudes also takes a loaded reading with respect toF. There are many ways in
which attitude reports can pertain to conversational goals, such that there is no one simple
condition we can state that exhaustively captures whether the most plausible reason that
thespeakerusesareportsentenceinvolvingF involvesanintentiontorepresentthesubject
as using a concept associated with F in certain ways some of their fully loaded attitudes.
On my way of understanding the pragmatic speech acts view, the many and various
ways that permissive readings are sensitive to conversational context are manifestations
of how our practices of issuing and interpreting permissive reports constitutively rely on
general practices of rational and cooperative communication. For this reason, I expect
that any principle that can provide a full and accurate answer to the pragmatic question of
permissiveness will involve ideology like what is “the most plausible reason” that a speaker
utters whatever she utters, and whether there is “evidence that the speakers intends to
convey” certain information in making a report. Such lingo, of course, does not serve to
give an impartial, technical theoretical description or account of our reporting practices
as much as it invites us to exercise our ability to participate in such practices. Absent
an impartial, technical theory of our practices of rational, cooperative communication
in general, we cannot describe the conditions in which a report takes di↵ erent possible
permissive readings in impartial, technical terms without our description being rough,
223
defeasible, orincomplete. Thepuzzleofhowwemanagetodowhateveritiswedowhenwe
make permissive reports has to be approached from inside the general pragmatic practices
responsible for the phenomenon of interest. This methodological perspective contrasts
with one according to which to qualify as accounting for permissive reporting a view must
consist of a tidy semantic rule or a set of tidy pragmatic principles. I admit that seemingly
there are some systematic semantic and pragmatic phenomena here to be described, and I
have tried to characterize some of these along the way. But we will fail to fully appreciate
the nature of reports of conceptually permissive attitudes, along with lots of the tricky
details, if we insist on trying to articulate semantic and pragmatic rules that aren’t there.
224
References
[1] Aloni, M. (2005). A formal treatment of the pragmatics of questions and attitudes.
Linguistics and Philosophy, 28(5), 505-539.
[2] Aristotle,Metaphysics,W.D.Ross(Trans.),inBarnes,J.(Ed.).(1984).Completeworks
of Aristotle, volume 1 (Vol. 96). Princeton University Press.
[3] Azzouni, J. (2011). II–Singular thoughts (objects-directed thoughts). In Aristotelian
Society Supplementary Volume (Vol. 85, No. 1, pp. 45-61). Oxford, UK: Blackwell
Publishing Ltd.
[4] Bach, K. (1987). Thought and reference. Oxford: Oxford University Press.
[5] Bach, K. (1994). Conversational impliciture. Mind & Language, 9: 124-162.
[6] Bach, K. (1997). Do belief reports report beliefs?. Pacific Philosophical Quarterly, 78,
215-241.
[7] Block, N. (1980). Troubles with functionalism. In Block, N. Readings in philosophy of
psychology, 1, 268-305.
[8] Blumberg, K., & Holgu´ ın, B. (2018). Ultra-liberal attitude reports. Philosophical Stud-
ies, 175(8), 2043-2062.
[9] Bennett, M. & Hacker, P. (2008). History of cognitive neuroscience, Oxford: Wiley-
Blackwell.
[10] Braun, D.(2008).Persistingproblemsforaquantificationaltheoryofcomplexdemon-
stratives. Philosophical studies, 141(2), 243-262.
[11] Burge, T. (1979). Individualism and the mental, in French, Uehling, and Wettstein
(eds.) Midwest Studies in Philosophy, IV, Minneapolis: University of Minnesota Press,
73-121.
[12] Cappelen, H., &Hawthorne, J.(2010).Relativism and monadic truth.Oxford: Oxford
University Press.
[13] Carlson, L H. (1982). Dialogue games: an approach to discourse analysis. Doctoral
dissertation, MIT.
[14] Carston (2004). Truth-conditional content and conversational implicature. In C.
Bianchi (Ed.). The Semantics/Pragmatics Distinction, pp. 65-100. Stanford: CSLI.
[15] Cummins, R. (1991). Methodological reflections on belief. In Bogdan, R. J. (Ed.).
(1991).Mind and Common Sense: Philosophical Essays on Common Sense Psychology,
Cambridge: Cambridge University Press.
225
[16] Davidson, D. (1975). Thought and talk. In Inquiries into truth and interpretation,
Oxford: Oxford University Press.
[17] Davidson, D. (1991). Epistemology externalized. Dialectica, 45(2-3), 191-202.
[18] Dennett, D. C. (1989). The intentional stance. Cambridge: MIT press.
[19] Donnellan, K. (1977). The contingent a ariori and rigid designators. Midwest Studies
in Philosophy, 2: 12-25.
[20] Dorr, C. (2014). Transparency and the context-sensitivity of attitude reports. Empty
representations: reference and non-existence, 25-66.
[21] Dummett, Michael (1996). The myth of the third realm, in Frege and other philoso-
phers, Oxford: Oxford University Press.
[22] Evans, G. (1979). Reference and contingency. The Monist, 161-189.
[23] Evans, G. (1982). The varieties of reference. Oxford: Oxford University Press.
[24] Figdor, C. (2017). On the proper domain of psychological predicates. Synthese.
194:4289-4310.
[25] Fodor, J. D. (1970). The linguistic description of opaque contexts, Doctoral Disserta-
tion, MIT.
[26] Fodor, J. (1975) The language of thought, New York: Cromwell.
[27] Geach, P. T. (1964). Referring expressions again. Analysis. 24(5), 172-175.
[28] Geurts, B. & Van der sandt, R. (2004). Interpreting focus. Theoretical Linguistics. 30
(2004), 1-44.
[29] Gilbert, M. (1996). Living together: rationality, sociality, and obligation. Rowman &
Littlefield Publishers.
[30] Goodman, R. (2017). Do acquaintance theorists have an attitude problem? Aus-
tralasian Journal of Philosophy, 1-20.
[31] Grice, P. (1975). Logic and conversation. In Cole, P.; Morgan, J. (Eds.). Syntax and
semantics. 3: Speech acts. New York: Academic Press. pp. 41?58.
[32] Gundel, J. K., & Fretheim, T. (2004). Topic and focus. The handbook of pragmatics,
175.
[33] Groenendijk, J. A. G., & Stokhof, M. J. B. (1984). Studies on the semantics of ques-
tions and the pragmatics of answers. Doctoral dissertation, Univ. Amsterdam.
226
[34] Hamblin, C. L. (1976). Questions in montague english. In Montague grammar,(pp.
247-259). Academic Press.
[35] Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford University Press.
[36] Hawthorne, J., & Manley, D. (2012). The reference book. Oxford: Oxford University
Press.
[37] Hellan, L. (1980). On semantic scope. Ambiguities in intensional contexts, 47-81.
[38] Izhikevich, E. M. (2007). Dynamical systems in neuroscience: The geometry of ex-
citability and bursting. Cambridge: MIT Press.
[39] Jacobson, P. (1994). Binding connectivity in copular sentences. in Harvey M. and
Santelmann L. (Eds.), SALT IV.
[40] Je↵ rey, R. C. (1990). The logic of decision. Chicago: University of Chicago Press.
[41] Jeshion, R. (2002). Acquaintanceless de re belief. Meaning and truth, 53-78.
[42] Jeshion, R. (2010) Singular thought: Acquaintance, semantic instrumentalism, and
cognitivism, New Essays on Singular Thought, Jeshion R. (Ed.). Oxford University
Press.
[43] King, J. C. (2007). The nature and structure of content. Oxford University Press.
[44] King, J. C. (2008). Complex demonstratives as quantifiers: objections and replies.
Philosophical studies, 141(2), 209-242.
[45] Kenny, A. (2010). Concepts, brains, and behaviour. Grazer Philosophische Studien,
81(1): 105?113.
[46] Kripke, S. A. (1972). Naming and necessity (pp. 253-355). Springer Netherlands.
[47] Lewis, D. (2013). Counterfactuals. John Wiley & Sons.
[48] Margolis, E. & Laurence, S. (2007). The ontology of concepts – abstract objects or
mental representations? Noˆ us, 41(4), 561-593.
[49] Millikan, R. G. (2000). On clear and confused ideas: An essay about substance con-
cepts. Cambridge University Press.
[50] Nelson, G., & Greenbaum, S. (2015). An introduction to English grammar. Routledge.
[51] Parikh, R. (1998). Propositions, propositional attitudes and belief revision. Advances
in modal logic, 2: 381-400.
227
[52] Peacocke, C. (1989). What are concepts? Midwest Studies in Philosophy, 14: 1-28.
[53] Percus, O. (2000). Constraints on some other variables in syntax. Natural Language
Semantics, 8: 173-229.
[54] Perry, J. (2000). The problem of the essential indexical and other essays. Stanford:
CSLI Publications.
[55] Putnam, H. (1975). The meaning of meaning. Philosophical Papers, Vol. II : Mind,
Language, and Reality. Cambridge: Cambridge University Press.
[56] Quine, W. V. O. (1956). Quantifiers and propositional attitudes. The Journal of Phi-
losophy, 53: 177-187.
[57] Recanati,F.(2010).Singularthought: Indefenseofacquaintance.InJeshionR.(Ed.).
New Essays on Singular Thought. Oxford: Oxford University Press.
[58] Roberts, C. (2012). Information structure: Towards an integrated formal theory of
pragmatics. Semantics and Pragmatics, 5, 6-1.
[59] Roberts, C. (2018). Speech acts in discourse context. New Work on Speech Acts, 317.
[60] Recanati, F. (1993). Direct reference: From language to thought. Oxford: Blackwell.
[61] Recanati, F. (2012) Mental files. Oxford: Oxford University Press.
[62] Richard, M. (2013a). Context and the attitudes. Oxford: Oxford University Press.
[63] Russell, B. (1905). On denoting. Mind, 14: 479-493.
[64] Russell, B. (1910). Knowledge by acquaintance and knowledge by description. In Pro-
ceedings of the Aristotelian Society, Aristotelian Society, Wiley. 11, 108-128.
[65] Russell, B. (1914). On the nature of acquaintance. The Monist, 24: 1-16.
[66] Ryle, G. (1949). The concept of mind. London: Hutchinson.
[67] Salmon,N.(1987).Howtomeasurethestandardmetre.ProceedingsoftheAristotelian
Society, 88: 193-217.
[68] Schein, B. (2006). Plurals. Handbook of philosophy of language, 716-767.
[69] Schwager, M. (2009). Speaking of qualities. In Semantics and Linguistic Theory, 19,
395-412.
[70] Schoubye, A. J., & Stokke, A. (2016). What is said?. Noˆ us, 50(4), 759-793.
228
[71] Schwitzgebel, E. (2002). A phenomenal, dispositional account of belief. Noˆ us, 36(2),
249-275.
[72] Schwitzgebel, E. (2013). A dispositional approach to attitudes: Thinking outside of
the belief box. In New essays on belief (pp. 75-99). Palgrave Macmillan, London.
[73] Sharvit, Y. (1999). Functional relative clauses. Linguistics and Philosophy, 22(5), 447-
478.
[74] Smith, M. (1987). The Humean theory of motivation. Mind, 96: 36?61.
[75] Soames, S. (2002). Beyond rigidity: The unfinished semantic agenda of naming and
necessity. Oxford: Oxford University Press.
[76] Soames, S. (2013). Cognitive propositions. Philosophical Perspectives, 27, 479-501.
[77] Sosa, E.(1970).Propositionalattitudesdedictuanddere.The Journal of Philosophy,
Vol. 67, No. 21.
[78] Stalnaker, R. (1978). Assertion. In Ezcurdia M. & Stainton R. J. (Eds.). The
Semantics-Pragmatics Boundary in Philosophy. Broadview Press.
[79] Sudo,Y.(2014).Onderepredicates.InProceedings of WCCFL(Vol.31,pp.447-456).
[80] Stanley, J. (2002). Nominal restriction. Logical form and language, 365-388.
[81] Szab´ o, Z. G. (2010). Specific, yet opaque. Logic, language and meaning, 32-41,
Springer, Berlin, Heidelberg.
[82] Tollefsen, D. P. (2015). Groups as agents. John Wiley & Sons.
[83] Tuomela, R. (1993). Corporate intention and corporate action. Analyse & Kritik,
15(1), 11-21.
[84] VanFraassen, B.C.(1979).Prepositionalattitudesinweakpragmatics.Studia Logica,
38(4), 365-374.
[85] von Fintel, K., & Heim, I. (2011). Intensional semantics.ms.,MIT.
[86] Weaver, W. G. (1998). Corporations as intentional systems. Journal of Business
Ethics, 17(1), 87-97.
[87] Peters, S., Westerst˚ ahl, D., & Westerst˚ ahl, D. (2006). Quantifiers in language and
logic. Oxford: Oxford University Press.
[88] Winter,Y.(1997).Choicefunctionsandthescopalsemanticsofindefinites.Linguistics
and philosophy, 20: 399-467.
229
Abstract (if available)
Abstract
This dissertation explores the view that ordinary, sincere attitude reports sometimes serve to ascribe conceptually permissive attitudes to subjects, where a conceptually permissive attitude is such that it is possible that a subject bear that attitude towards a proposition without possessing and deploying concepts of things that that proposition is directly about. At first I motivate this view by raising linguistic data to support what I call the “sensitive attitudes view”, according to which attitude verbs are semantically underspecified in such a way that they can be sincerely used to designate different attitude relations in different contexts. Having argued that we need the sensitive attitudes view to account for the full range of examples of the context-sensitivity of attitude reports, I proceed to consider how best to develop the conceptually permissive attitudes view within that broad framework. By attending to ways that certain readings of attitude reports systematically depend on different factors of conversational contexts, I argue against simple ways of articulating the conceptually permissive attitudes view that specify the truth conditions of attitude reports just in terms of which propositions are contextually assumed in the reporting context. I refine such simple views to be sensitive to the semantic and pragmatic effects of other features of conversational situations, including what questions are under discussion in reporting contexts.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Constraining assertion: an account of context-sensitivity
PDF
Telling each other what to do: on imperative language
PDF
Representation, truth, and the metaphysics of propositions
PDF
Reasoning with uncertainty and epistemic modals
PDF
Meaningfulness, rules, and use-conditional semantics
PDF
Unlimited ontology and its limits
PDF
Positivist realism
PDF
Iffy confidence
PDF
Lessons From Frege's Puzzle
PDF
Instantial terms, donkey anaphora, and individual concepts
PDF
Reference time in the dynamics of temporal dependency in Korean
PDF
Rationality and the primacy of the occurrent
PDF
What "ought" ought to mean
PDF
Subjectivity, commitments and degrees: on Mandarin hen
PDF
Comparative iIlusions at the syntax-semantics interface
PDF
Process-oriented rationality
PDF
Units of agency in ethics
PDF
Health care for all? Anti-Latino and anti-immigrant attitudes, health care policy, and the Latino community
PDF
Public justification beyond legitimacy
PDF
Feeling good and living well: on the nature of pleasure and its role in well-being