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
/
Telling each other what to do: on imperative language
(USC Thesis Other)
Telling each other what to do: on imperative language
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Telling Each Other What To Do:
On Imperative Language
Matthew Babb
A dissertation presented to:
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirments for the Degree:
DOCTOR OF PHILOSOHY
(PHILOSOPHY)
May 2017
Contents
Acknowledgements 3
1 The Syntax of Imperatives 6
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 What are Imperatives? . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Syntactic Features of Imperatives . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Imperative Subjects . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Imperatives and Vocatives . . . . . . . . . . . . . . . . . . . . 17
1.3.3 Imperative Predicates . . . . . . . . . . . . . . . . . . . . . . 20
1.3.4 Imperatives, Negation, and Do-Support . . . . . . . . . . . . . 25
1.3.5 Speciers in Imperatives: Subjects and Adverbials . . . . . . . 33
1.3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.4 Three Theories about the Syntactic Nature of Imperatives . . . . . . 41
1.4.1 The \Impoverished Clause" Theory . . . . . . . . . . . . . . . 41
1.4.2 The Future Tense Theory . . . . . . . . . . . . . . . . . . . . 45
1.4.3 The Special Mood Theory . . . . . . . . . . . . . . . . . . . . 49
2 Two Theories of Imperative Meaning 54
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2.1 Principle of Full Interpretation . . . . . . . . . . . . . . . . . 55
2.2.2 Imperatives, Truth, and Representation . . . . . . . . . . . . . 57
2.2.3 Imperatives and Action . . . . . . . . . . . . . . . . . . . . . . 59
2.2.4 Core vs. Peripheral Meaning . . . . . . . . . . . . . . . . . . . 61
2.2.5 Conversational Backgrounds . . . . . . . . . . . . . . . . . . . 62
2.3 The Property Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.3.1 The Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.3.2 Interpreting Imperatives . . . . . . . . . . . . . . . . . . . . . 69
1
2.3.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.4 The Deontic Proposition Theory . . . . . . . . . . . . . . . . . . . . . 78
2.4.1 The Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.4.2 Interpreting Imperatives . . . . . . . . . . . . . . . . . . . . . 82
2.4.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.5 What We've Learned and Going Forward . . . . . . . . . . . . . . . . 91
3 A Na ve Satisfactionist View of Imperative Meaning 95
3.1 Recap and the Proposed View . . . . . . . . . . . . . . . . . . . . . . 95
3.2 Semantics of Imperatives: A Selective History . . . . . . . . . . . . . 98
3.3 Na ve Satisfactionism . . . . . . . . . . . . . . . . . . . . . . . . 102
3.4 The Role of Sentential Mood . . . . . . . . . . . . . . . . . . . . . . . 108
3.5 Reporting Imperative Use . . . . . . . . . . . . . . . . . . . . . . . . 113
3.6 Summary and Remaining Issues . . . . . . . . . . . . . . . . . . . . . 116
4 The Problem of Mixed Compound Sentences 119
4.1 The Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.2 Solving the Problem: Strong Na ve Satisfactionism . . . . . . . 122
4.3 Some Consequences of Strong Na ve Satisfactionism . . . . . . 125
4.4 Irrational Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.5 Why Not Go Dynamic? . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5 The Pragmatics of Imperatives 136
5.1 Recap: Imperatives and Intentions . . . . . . . . . . . . . . . . . . . . 136
5.2 Imperatives and the 2
nd
-Person . . . . . . . . . . . . . . . . . . . . . 138
5.2.1 A Syntactic Explanation? . . . . . . . . . . . . . . . . . . . . 139
5.2.2 The Essential 1
st
-Personality of Intentions . . . . . . . . . . . 139
5.2.3 The Non-Dominance of 1
st
- and 3
rd
-Person Imperatives . . . . 141
5.2.4 Problem Cases and Norms of Imperative Use . . . . . . . . . . 142
5.3 The Normativity of Directives . . . . . . . . . . . . . . . . . . . . . . 143
5.3.1 Reasons: Motivating v. Normative; First-Order v. Second-Order145
5.3.2 Directives and Proerings of Reasons . . . . . . . . . . . . . . 147
5.3.3 What Kind of Reasons? . . . . . . . . . . . . . . . . . . . . . 149
5.3.4 From Reasons to Act to Reasons to Intend . . . . . . . . . . . 152
5.4 Bringing It All Together: Or, How to Tell Each Other What to Do . . 153
Bibliography 155
2
Acknowledgements
The path to this dissertation has been a long one. Along the way I have had a lot
of help and I owe thanks to many. Thanks are rst due to those I met at Northern
Illinois University. William \Bill" Tolhurst (1947-2011; may he rest in peace) helped
me write my rst serious philosophy paper, an undergraduate honors thesis, and was
perhaps the kindest and most nurturing person I have ever met. During my time
in the masters program at NIU I had the pleasure of working with other amazing
members of the faculty: David Buller, Mylan Engel, Valia Alori, and Carl Gillett.
I'm especially grateful to Tomis \TK" Kapitan (1949{2016; may he rest in peace)
during these highly formative years. He taught me about topics that I still work on
to this day, supported me even when I needed it most, and showed me what a mentor
should be like.
Thanks are next due to my fellow masters students at NIU: Ben Rohrs, Jonah
Nagashima, Danae Patterson, Mike Tolhurst, Paul Bowman, Brian Chambliss, Brock
Rough, Olav Vassend, Dominick DeLorenzo-Breed, Drew Freeman, Anthony Kelley,
Hope Sample, Tony Chiarito, Dylan Black, Anna Moorman, Patrick Campbell, An-
drew Morgan, Ty O'Grady, Tim Kakos, Justin Dealy, David Jacobs, and Joe Haney.
The two years I spent with these individuals intellectually invigorating and let me
know how great it is to be a part of a philosophical community.
I am very fortunate to have come to USC for my PhD. There are so many
people to thank! First are the members of my dissertation committee. Robin Jeshion
has been my biggest supporter at USC. She has been involved in one way or another
in all of my milestones, and has been both critical and yet fair when it comes to my
work. She has also encouraged me to be a systematic thinker, relating ideas I have in
one area to others. Gabriel Uzquiano has taught me the importance of motivating my
ideas and trying to make them appealing to broader audiences. He has also read more
of my written work than perhaps anyone else; for that I am especially grateful. Scott
Soames has been my biggest critic, never letting me slip by with sloppy analyses or
incomplete arguments. When I rst arrived at USC, Scott was the person who scared
me the most, particularly because my philosophical views tended to be contrary to
3
his. Over the years, as I grew philosophically, he became a generous mentor and an
encouraging ally. Barry Schein taught me syntax and formal semantics. The rst two
chapters of this dissertation would not have been possible without his guidance and
insights. He also taught me that the complexity of natural languages is something to
appreciate. Steve Finlay joined my committee later than the rest, but his impact on
the dissertation has been substantial. He made me think much harder about some
of the ner points of my syntactic and semantic analyses of imperatives, and the
last chapter in particular was completely written for the better because of Steve's
observations. Though I still cannot say with certainty what a reason is, with Steve's
help my thoughts on the subject are much, much clearer. I would also like to thank
Edwin McCann here. Ed has been my teaching mentor at USC and I have learned
so much from him about what it means to be a good teacher.
The next people to thank at USC are the faculty. I've benetted hugely from
numerous conversations and seminars with them, and they have supported me so
much. Thanks, in particular, to Jim Higginbotham (1941-2014; may he rest in peace)
who rst introduced me to the intersection of philosophy and linguistics and showed
me why it's important. Thanks to Mark Schroeder and John Hawthorne for their
invaluable support and guidance. Thanks also to faculty no longer at USC who
read my work and guided my thoughts: Kenny Easwaran, Karen Lewis, and Andrei
Marmor. Thanks to you all for your encouragement and your support.
The USC graduate and postdoc community has been a constant source of
engaging philosophical conversations, encouragement, and fun. I know I am going
to miss some folks but thanks to (in no particular order): Aness Webster, Greg
Ackerman, Michael Milona, Jen Liderth, Michael Hatcher, Abelard Podgorski, Ara
Astourian, Michael Pressman, Nick Laskowski, Kenneth Sliver, Aaron Veek, Ryan
Walsh, Keith Hall, Ellie Neufeld, Renee Bolinger, Caleb Pearl, Justin Dallmann, Kory
DeClark, Alex Dietz, Maegan Fairchild, Marina Folescu, Indrek Reiland, Johannes
Schmitt, Sam Shpall, Justin Snedegar, Jonathan Wright, Ben Lennertz, Shyam Nair,
Kenny Pearce, Matt Lutz, Alex Sarch, Bryan Roberts, Nathan Howard, Erick En-
carnacion, Steve Bero, Tiany Chang, Cecila Stepp, and Robert Shanklin. Thanks
also goes to some students of UCLA: Steve Lookner, Lauren Schaeer, and especially
Eliot Michaelson. Though our schools were cross-town rivals, I very much enjoyed
our conversations and learned a lot from them.
The next round of thanks goes to the amazing friends I made in LA: AJ
Moorehead, Myles Webster, Lauren Weindling, Keri Smith, Marcy Lascano, Jason
Raibley, Clo Pazera, Beth Snyder, Hannah Lau, and Ilana Winter. When I rst
moved to LA I wasn't terribly excited about the city { they made my time fun and
often provided a much needed distraction from philosophy.
4
My family has kept me sane and grounded throughout the years, always re-
minding me that I am loved. To my parents, Connie Balles and Michael Babb, and
my stepparents, Kathleen Babb and Richard Balles, thank you for believing in me
and showing me how to be a good person. To my brothers, Daniel Babb and Christo-
pher Babb, and my step siblings, Ian Hutchinson, Anne Helbig, and Tim Balles, I
know I'm sometime a ghost, but thank you for not holding it against me (at least not
too much). To the Smudde Clan { Tom Smudde, Marsha Smudde, Stacey Paolini,
Jason Paolini, Robert Smudde, Katie Smudde, Kevin Smudde, Hillary Smudde, Ja-
son Wold, Joy Wold, and Robyn Smudde { thanks for reminding me of the value of
enjoying life and always being there for me when I need you. To the newest members
of my family { Irene Zieba, Jerry Zieba, Ania Zieba, and Anna Zieba { thank you
for accepting me into your lives in such a caring way. Thanks to you all for never
giving up on me and supporting the life path I've chosen.
I would like to thank and give remembrance to some members of my family no
longer with us. Jim Smudde (1959-2011) was the sweetest man I have ever met and I
am honored and humbled to be able to call him my uncle and friend. Lillian Fox (1921-
2006), my maternal grandmother, took care of my brothers and I after school during
elementary and middle school. Helen Babb (1917-2011), my paternal grandmother,
was a matriarch who taught us how to be generous, caring, and forgiving. Finally,
and most importantly, Ryan Babb (1981-2006) was my brother and without whom
I would not be the person I am { I love you, bro, and I miss you dearly.
This dissertation is dedicated to Ryan and his memory.
And last, but certainly not least, I thank Jennifer Zieba. Thank you for all
the little and big things that you?ve done to support me. I know you know how
much you mean to me, and how much your presence in my life has changed it for
the better. You are my best friend, my closest condant, and an amazing individual.
I look forward to the next phase of our lives together!
5
Chapter 1
The Syntax of Imperatives
1.1 Introduction
The subject matter of my dissertation is imperative sentences, such as:
(1) a. Jack, go to the store.
b. Jill, fetch some water.
c. Jerry, eat your vegetables.
These are to be contrasted with:
(2) a. Jack went to the store.
b. Jill will fetch some water.
c. Jerry ate his vegetables.
(3) a. Has Jack gone to the store?
b. Will Jill fetch some water?
c. Did Jerry eat his vegetables?
The constructions in (2) are declarative sentences, while those in (3) are interrog-
ative sentences. Both of these types of sentence are distinct from imperatives. And
they are distinct in two ways. First, they are syntactically distinct, in that they
are dierent types of syntactic, or grammatical, constructions. Imperative sentences
are neither declaratives nor interrogatives. Second, they serve distinct conversational
roles. Imperatives are used in conversation to tell others to do things. By contrast,
declaratives are used to tell others the way things are, while interrogatives are used
to ask others how things are.
This latter point raises an interesting question that lies at the boundary be-
6
tween the philosophy of language and linguistics. Namely: What is the meaning of
an imperative sentence? This question lies at this border because, on the one hand,
asking what is the meaning of some linguistic construction is a question possessing
more of a philosophical
avor, and on the other hand, in order to answer this ques-
tion one needs to rst answer questions about what imperative sentences are, which
requires looking closely at their syntax and trying to nd there what distinguishes
them from other types of linguistic constructions. So answering the question of what
is the meaning of an imperative requires delving into both linguistics and philosophy.
What is the meaning of an imperative sentence? This is the main question
of my dissertation. And it is not an easy question to answer. It is not because the
primary focus of philosophers and linguists over the last century has been on declar-
ative sentences, and as it happens, trying to take what we know about declarative
meaning and extend it to imperatives is no easy task. In fact, it may be an impossible
task, because a putatively key semantic feature of declaratives is that they are true
or false, a feature not shared with imperatives. So it seems that answering questions
about the meanings of imperatives requires starting at the beginning.
In this opening chapter I address two fundamental questions. First, what is
an imperative sentence? Second, what makes a particular linguistic construction an
imperative sentence? My answer to the rst question will be that imperatives are
sentences like any other, that is, they are syntactic constructions formed out of a
noun subject phrase (NP) and a predicate verb phrase (VP), among other structures.
As for the second question, rst I will examine the features of the subjects and
predicates of imperatives. Then I will look at three dierent views of the syntax of
imperatives, each of which attempting to explain what makes imperatives unique
constructions. The dierence between the three views lies in what each says occupies
the left periphery of the syntactic structure of imperatives. In the end, I will endorse
the view that says that in the left periphery of the syntax of imperatives, there is a
dening \mood" morpheme: IMP.
The importance of this rst chapter cannot be stressed enough. Few who have
worked on the semantics and meanings of imperatives have taken the time to rst
examine the syntax of these sentences.
1
. And this has led to a number of confusions
and misunderstandings about just what imperatives are, and by extension, what they
mean. So by starting with an investigation into what exactly imperative sentences
are, we place ourselves in an ideal position to avoid mistakes and make progress.
What's more, this chapter will provide a syntactic framework against which we can
ask questions about the meanings of imperatives. More specically, the upshot of
1
Notable exceptions are Chung-Hye Han and Paul Portner. See Han (2000), Portner (2004), and
Portner and Zanuttini (2003).
7
this chapter will be to restrict the possible theories of the meanings of imperatives:
any adequate theory will have to be able to accommodate the syntactic data, and
in particular it will have to explain the semantic roles and interactions of each of
the syntactic elements of imperatives contributing to their meanings. As we'll see in
the next chapter, prominent theories currently on the market strive to adhere to this
requirement.
1.2 What are Imperatives?
Perhaps the most common confusion found in the literature (and in personal discus-
sions) on the meanings of imperatives involves just what they are. Often one reads
that imperatives are the proclamations of those in authority, as when a boss tells
his employees to take fewer breaks or a general tells his subordinates to clean the
barracks. Other times one read that imperatives are any directive whatsoever, such
that a child asking his mother for a new bike is an imperative. The trouble is that
none of these things are imperatives. They are all rather speech acts that are per-
formed using imperatives. In particular, they are directive speech acts, which are
to be distinguished from assertive and inquisitive speech acts (which are performed
using declarative sentences and interrogative sentences, respectively).
What we need is an old but familiar distinction between speech acts and sen-
tences (both of which also need to be distinguished from meaning or content; more on
this later). Speech acts are a certain class of action-types that can be performed only
by way of using linguistic constructions. To perform a speech act is to do something
with the items of a language, however rudimentary. A general might command his
subordinates to clean the barracks, for instances, by forcefully uttering (4).
(4) Clean the barracks.
(4) is an English imperative and it is being used by the general to issue a command.
More importantly, (4) is a linguistic construction { it is a sentence, which is a type
of linguistic construction { that is formed using the words and rules of English. Not
only (4), but all imperatives are merely linguistic constructions { no more, no less.
That imperatives are sentences and that they need to be distinguished from
certain sorts of speech acts (commanding, demanding, advising, etc.) may seem like
obvious points not needing to be stressed, but they do. For ignoring them might
lead one to think that the way to theorize about the meanings of imperatives is
through analyzing directive speech acts. But this puts the cart before the horse.
Since directives are performed using imperatives, understanding what a directive is
8
requires rst understanding what imperatives are and what they mean. This is of
course not to say that directives can't tell us something about imperatives and their
meanings { they certainly can. In particular, they tell us that they have to be the
sort of thing that can be used in directives. That is, whatever imperatives and their
meanings are, they have to make directive speech acts possible, and any theory of
the meanings of imperatives should explain this.
Another role for directive speech acts to play is in helping us identify which
linguistic constructions are imperatives. The idea here is that relationship between
imperatives and directives provide us with a criterion by which we can use the latter
to identify the former:
Imperative Identifying Criterion A construction is an imperative i it can
be used in any kind of directive speech act
whatsoever.
2
To see why this is so, take a construction that though it can be used in some kinds
of directives, it certainly can't be used in all of them:
(5) Can you pass the salt?
Anyone familiar with English and its customs knows, (5) can be used to issue a
request: I can request that you pass the salt by uttering (5) to you in the right
circumstance. It can also be used to make a demand. This might happen when I
utter (5) to you with a particular tone and the belief that your not passing the salt
would warrant censure. But using (5) is no way to (e.g.) issue a command or provide
advice. A mother who wanted to command her son to pass the salt would not use
(5). Nor would someone who wanted to advise a friend about how to behave at a
dinner party use (5). Consider:
A: What am I to do at the dinner party tomorrow?
B: #Can you dance?
In general, (5) cannot be used to issue commands or provide advice. And this is not
surprising, as we know that (5) is an interrogative sentence, not an imperative, and
interrogatives are designed for inquisitive speech acts, not directive ones. That (5)
can sometimes be used in a request is not due to the type of sentence it is, but rather
to the customs of English speakers.
Unlike (5), (6) can be used in any type of directive speech act.
2
This criterion would be called into question if we could nd a language in which imperatives
contained morphemes that marked them for use in only a particular kind of directive speech act.
My investigations have thus far turned up no such language.
9
(6) Pass the salt.
This can be seen, in the rst instance from the fact that in certain circumstances,
(6) would be a perfectly acceptable thing for B to say in response to A's question:
A: I need advice.
B: Dance!
And (6) can also be used in commands, unlike (5){ and so on for all other kinds of
directive speech acts. Moreover, (6) is a paradigm example of an imperative sentence:
if any linguistic construction in English is an imperative sentence, (6) is. And the
same goes for the constructions in (1) as well as:
(7) a. Do your homework.
b. You watch your step.
c. You be nice.
And so on. The thing to note about these is that, like (6), each of them can be
used in any type of directive speech act whatsoever. (7-a) can be used to command,
demand, advise, or request someone to do their homework. The same goes for (7-b),
(7-c), and any other imperative sentence.
The reason for this is due to the fact that whereas interrogatives are designed
for the performance of inquisitive speech acts, imperatives are designed for the di-
rective speech acts. Put another way, the reason why there are imperative sentences
at all is because they are needed for the production of directives. So though im-
peratives are not directives, we should nonetheless expect that all and only those
constructions that can be used in any type of directive whatsoever are imperatives.
We might say that all and only imperatives are apt for the production of directives.
This aptness relationship between imperatives and directives is what grounds the
Imperative Identifying Criterion introduced above.
So imperatives explain why directive speech acts are possible, and because of
this, directive speech acts can be used to identify which constructions are imperatives
and which are not. There is of course much more to say about directives and their
relationship with imperatives, but before we can do that we need to know more
about imperatives and their meanings. That being the case, I will have nothing more
to say about directive speech acts until the nal chapter, where I will explain how
my theory of the meanings of imperatives makes them possible, and explains the
mechanism by which they are produced.
10
1.3 Syntactic Features of Imperatives
In order to address our next question { What makes a linguistic construction an
imperative { we need to take a closer look at the syntactic features of imperative
sentences. In particular, we need to examine their subjects and predicates, to see
what they can tell us about these constructions. It has often been thought that
what's distinctive about imperatives is that there are certain grammatical restric-
tions on their subjects and predicates, or that the syntactic structure of imperatives
dramatically diverges from that of the other major clause types. As we shall see,
however, the evidence for these claims is lacking.
3
1.3.1 Imperative Subjects
One of the most striking features of imperative sentences (in English at least) is they
can, though they need not, be formulated without any explicit subject NP. Some
simple examples:
(8) a. Go to the store.
b. Eat your vegetables.
c. Mix together the wet and dry ingredients.
On the surface these appear to be completely missing a subject; there is no overt word
or morpheme serving to indicate a subject. This is quite dierent from what we nd
with other sentence types. Omitting a subject NP or morpheme from a declarative
or interrogative renders them ungrammatical:
(9) a. *Went to the store.
b. *Ate his vegetables.
c. *Is mixing together the wet and dry ingredients.
(10) a. *Did go to the store?
b. *Ate his vegetables?
c. *Mixed together the wet and dry ingredients?
We might call this observation Optionality of the Imperative Subject, or just Subject
Optionality for short.
3
This conclusion { that there is nothing exceptional about the syntactic structure of imperatives
and the restrictions on their subjects and predicates { is also drawn by Potsdam (1996), though
I approach it in a somewhat dierent way, mostly because of a lack of space. Those interested in
topics discussed in this chapter are strongly urged to read Potsdam's work, as he addresses them
much more carefully and in more detail than I am able.
11
From this observation it is tempting to conclude that imperative do not really
have subjects. After all, if an imperative can be formulated without a subject NP,
then doesn't this mean these NPs make no substantive contribution, semantic or
otherwise, to the imperative? That is, if what can be said and done with (11-a) can
be just as well said and done with (11-b), then what is the NP `you' contributing to
(11-a)?
(11) a. You step back three paces.
b. Step back three paces.
Seems like the `you' in (11-a) is contributing nothing at all, and so there is no sense
in saying that it is the \subject" of (11-a) { and so (11-a) and other imperatives
don't in fact have subjects, despite appearances. If this is correct, then it seems we
have found a way to syntactically distinguish imperatives from other sentence types:
imperatives don't have true subjects, while other sentence types do.
4
The trouble with the forgoing is that there is evidence showing that even
imperatives lacking an overt subject NP nevertheless must possess morphemes in-
dicating a subject. The thing about these morphemes is that, for whatever reason,
they are not phonologically realized; they are so-called null morphemes, in the sense
that they correspond to no { null { phonemes. The primary evidence for the presence
of null morphemes in sentences like those in (8) comes from what we know about
re
exive pronouns and that they can occur in imperatives lacking overt subject NPs
(Perlmutter and Soames 1979). Consider:
(12) a. Teach yourself some self-control.
b. Move yourself over three paces.
c. Mix together the wet and dry ingredients yourself.
Here we have \subjectless" imperatives also containing the re
exive pronoun `your-
self', as direct object in (12-a) and (12-b) and as an emphatic in (12-c). The impor-
tant thing about re
exive pronouns is that they must be syntactically bound and
governed by the subject of the sentence in which they occur, else their presence will
render a sentence ungrammatical, as in:
(13) a. *He taught themselves some self-control.
b. *You will move himself over three paces.
c. *I am mixing the ingredients together yourself.
The reason these constructions are ungrammatical is because the re
exives `them-
4
Cf. Platzack and Rosengren (1997).
12
selves', `himself', and `yourself' are bound by the subject NPs `He', `You', and `I',
respectively, and in being so bound they must agree in at least person, number, and
gender (Radford 2004). Put another way, the syntactic rules require that a re
ex-
ive pronoun is \checked" by a sentence's subject to see if they are in agreement in
terms of the appropriate grammatical properties (person, number, etc.). The sen-
tence can be grammatical only if there is such agreement between re
exive pronoun
and subject.
But this means that sentences like those in (13) must have subjects that bind
and govern the re
exive `yourself'. Specically, since `yourself' is second person and
singular, the subjects of these sentences must be second person and singular as well.
And since these sentences lack overt subject NPs, they must instead possess subject
null morphs that, in English, are marked for at least grammatical person, number,
and gender. Following one tradition we will represent these null morphs by using the
`?' symbol and superscripts. So, for example, the subject morph for the sentences in
(14) is `?
2nd; Singular; Neutral
', which is semantically equivalent to `you' in functioning
to refer to an addressee (more on this later), and these sentences can be rewritten as
follows:
(12
0
) a. ?
2nd; Singular; Neutral
teach yourself some self-control.
b. ?
2nd; Singular; Neutral
move yourself over three paces.
c. ?
2nd; Singular; Neutral
mix together the wet and dry ingredients yourself.
What's more, if imperatives with re
exive pronouns but lacking an over subject NP
nonetheless have subject morphs, then there is nothing stopping us from extrapolat-
ing and maintaining that all apparently subjectless imperative in fact have subject
null morphs.
Re
exive pronouns aside, another reason for thinking apparently subjectless
imperatives have subject null morphs, hence subjects, is that presumably the dening
feature of sentences is that they are composed of a subject element and a predicate
element. That is, what separates sentences from (other
5
) phrases, words, morphemes,
and other grammatical units is that only sentences have a subject NP or morpheme
and a predicate VP. This is a very old and deeply entrenched view, and for good
5
Some linguists seek to classify sentences and clauses as types of phrases. For example, one
popular idea in the generative grammar literature is that sentences (declaratives, interrogatives,
imperatives, etc.) are actually in
ectional phrases. Though it won't be made explicit, I will essen-
tially be agreeing with this idea later in this chapter. For now, though, I just want to point out that
whether or not the general idea that sentences are in
ectional phrases is right, not only is it still
the case the sentences have subjects and predicates, but our question of what makes a construction
an imperative sentence also still remains.
13
reason, given the both the linguistic and especially the cross-linguistic data.
6
That
being the case, we should hang on to this view, and only let it go if we nd evidence
with which it is irreconcilable. The way to reconcile it with apparently subjectless
imperatives is by postulating the presence of a null subject morpheme.
Questions remain, however, about the subjects of imperatives. Must the sub-
jects of imperatives be singular, or can they be plural as well? And must the subjects
of imperatives be 2nd person, or are 1st and 3rd person subjects also possible? How
imperatives with overt subjects and imperatives with non-overt subjects related? I
take each of these questions in turn.
Imperatives can have either singular or plural subjects. We have already seen
some examples involving imperatives with singular subjects (e.g. (7), (8), and (12)).
Others include, with overt subjects:
(14) a. You go to the store.
b. You take your shoes o.
c. You do your homework.
7
That the subjects of these sentences are singular is due to the singularity of `you'.
But imperatives can also have `you-all' in subject position:
(15) a. You-all go to the store.
b. You-all take your shoes o.
c. You-all do your homework.
While `you' is primarily singular, `you-all' is primarily plural.
8
Other obvious exam-
ples of imperatives with plural subjects are:
(16) a. You two/three/four/etc. go to the store.
b. You folks take your shoes o.
c. You kids do your homework.
Even non-overt subjects can be plural, as is evidenced by the presence of the 2nd
person plural re
exive pronoun `yourselves' in:
6
The data coming from in
ected languages (especially the Romance languages) is particularly
compelling. See Matthews (1991).
7
Schmerling (1975) calls these emphatic imperatives, because `you' is normally only present
when a speaker is emphasizing to whom the imperative is directed.
8
There is much debate about whether `you' is exclusively singular and whether `you-all' is exclu-
sively plural. I don't which to enter these debates. Fortunately, for my purposes, all that matters is
that the `you's in (14) are singular and the `you-all's in (15) are plural, which, by stipulation, they
are.
14
(17) a. Teach yourselves some self-control.
b. Move yourselves over three paces.
c. Complete your homework yourselves.
These examples are sucient for showing that imperatives can have both singular
and plural subjects.
The fact that all of our example imperatives so far have had 2nd person
subjects might lead one to think imperatives can only have such subjects { that 1st
and 3rd person subject are not possible in imperatives. However, it is not hard to
establish that imperatives can have 3rd person subjects. Consider:
(18) a. Everybody drive themselves. (Imagine said to group deciding who drives
who)
b. Everyone eat their vegetables. (Imagine said to students in a school
cafeteria)
c. Nobody do their homework. (Imagine said to students in detention)
The subject NPs of these sentences must be 3rd person because of the presence of
`themselves' and `their'. Recall that re
exive pronouns must agree in person with the
subject of a sentence. Since `themselves' is a 3rd person re
exive pronoun, and since
(18-a) is grammatical (though it may not often be felicitous to use), it follows that
the `everybody' in (18-a) must be 3rd person as well. Similarly, the pronoun `their',
when it is re
exive and not deictic, as in (18-b) and (18-c), must also agree in person
with a sentence's subject. So the `everyone' in (18-b) and the `nobody' in (18-c) are
15
also 3rd person.
9;10
Unfortunately, re
exive pronouns can't help establish that imperatives can
have 1st person subjects in English. This is because the grammaticality of imperative
constructions like the following is questionable:
(19) a. ?Go to the store myself.
b. ?Eat my vegetables myself.
c. ?Mix together the wet and dry ingredients myself.
As is the grammaticality of imperatives with explicit 1st person subjects:
(20) a. ?We go to the store.
b. ?I go to the store.
c. ?You and I go to the store.
11
We don't want to rest our arguments on problematic data like this. Interestingly,
however, there are a number of languages in which imperatives do have 1st person
subjects. French is a one example of a language with clear 1st person imperatives
9
Bolinger (1967) suggests that \subjectless" imperatives might also have 3rd person subjects.
(i) a. ?Teach himself some self-respect.
b. ?Move herself over three paces.
c. ?Mix together the wet and dry ingredients himself.
For my part, I'm not sure about the grammaticality of sentences like these, which is why I have
marked them as questionable. The argument for their being grammatical imperatives comes from
the fact that they can be used as answers to questions like What is Jack to do?, asked by someone
who wants (e.g.) advice about what to tell Jack to do. One worry with this argument, however,
is that it might be counter-argued that when used in response to such questions, (i-a)-(i-b) are
actually elliptical for:
(i
0
) a. Jack is to go to the store himself.
b. Jack is to eat his vegetables himself.
c. Jack is to mix together the wet and dry ingredients himself.
If so, then uses of (18-a)-(18-c) as answers toWhat is so-and-so to do? questions are not imperatives
unless (i
0
-a)-(i
0
-c) are, and though an argument might be made for (i
0
-a)-(i
0
-c) being imperatives,
it will unavoidably rely on a number of contestable assumptions about the syntactic structure and
possible uses of these sentences.
10
It's worth noting that imperatives with 3rd person subjects are more common in other lan-
guages, such as Greek, Latin, and Finnish.
11
A number of people have reported to me that they nd the constructions in (20) to be more
acceptable than those in (19), and though I share their judgments, it's not clear why the (20)
constructions should be okay while the (19) constructions are unacceptable, if they are.
16
(Hawkins and Towell 2015, Ch. 11), and Irish is another (Stenson 2008, p. 168). So
whether or not such imperatives are possible in English may not be the best way
to put the question we are interested in. Rather, what we want to know is whether
imperatives with 1st person imperatives are possible in general. The cross-linguistic
data points towards a positive answer.
However, it does appear that even English once had 1st person imperatives. In
older versions of English, especially those in which the subject NP and predicate VP
were often inverted in imperatives (e.g. Modern English), the following constructions
were grammatical:
(21) a. Be I wary of new comers.
b. Give we to the less fortunate.
(Think about the writings of Chaucer and Shakespeare, in which sentences like these
were very common.) That `I' and `we' are the subjects of these sentences (and not
objects) is evidenced by their case: in new and old versions of English, `I' is a subject
case 1st person singular pronoun, while `we' a subject case 1st person plural pronoun.
It's a curious question why such sentence have fallen out of fashion.
In sum, imperatives can have 3rd person and also 1st person subjects, so
what's distinctive of imperatives, qua clause type, is not that they can only have
2nd person subjects. But this does not mean that there isn't a question of why,
in terms of everyday usage, imperatives with 2nd person subjects are signicantly
more prevalent than imperatives with 3rd or 1st person subjects. Potsdam (1996),
Zwicky (1988) and others have suggested that the answer to this question likely has
something to do with pragmatic and more broadly sociological factors, instead of
grammatical ones. What these factors are exactly is not something I wish to dwell
on yet. (I will return to this in Chapter 5)
1.3.2 Imperatives and Vocatives
Here I address a worry that may have occurred to some readers, namely that what
appear to be subject NPs of imperatives are really just vocatives.
12
First a little setup.
A vocative is a particular occurrence of an NP that essentially serves a particular
function. The function in question is one in which it is meant to identify who the
addressee is. For example, `Anne', `Bill', and `Chris' all occur as vocatives in the
sentences in (22), and as such they x, or identify, who the addressee of an utterance
of one of these sentences is.
12
Thanks to Steve Finlay for bringing this worry to my attention and pressing me to address it.
17
(22) a. Anne, why is grass green?
b. Grass is green, Bill.
c. Chlorophyll, Chris, is why grass is green.
Importantly, vocatives are not the subjects of sentences. That this is so is revealed by
the fact that the subject NP of (22-b), for instance, is `Grass' and not `Bob' (similarly,
the subject NPs of (22-a) and (22-c) are not `Anne' and `Chris'). Of course, this is
not to say that a vocative and a subject cannot co-refer. In fact, sometimes a subject
and a vocative appear to stand in an anaphoric relation:
(23) a. Danny
1
, you
1
forgot to turn o the stove yesterday.
b. You
2
two are going to hurt yourselves, Erica and Elise
2
.
c. You
3
, Gary
3
, haven't learned a thing.
In (23-a), for example, the subject NP is `you' and the vocative NP is `Danny', and
it seems that the former is anaphoric on the latter as indicated by the subscripts.
But anaphoric relations are semantic, not syntactic, phenomena (though sometimes
they are established syntactically, as with re
exive pronouns). Whether vocatives are
integrated into the syntactic structure of a sentence, and how, is a tricky question I
won't be considering.
On more than a few occasions it has been suggested to me that, despite
appearances, NPs appearing to be the subjects of imperatives might instead merely
be vocatives, thus drawing us back to the idea that imperatives lack subjects. So, for
example, on this view, the `you's in (14-a)-(14-b) are vocative NPs, and really there
should be a comma or other intonational break in these sentences, such as:
(16
0
) a. You, go to the store.
b. You, take your shoes o.
c. You, do your homework.
Thorne (1966) argues for this view. Since some of what I will say and argue for in
succeeding chapters depends on imperatives having true subjects, let me take a brief
moment to show why this view is mistaken.
Potsdam (1996) points out that since vocatives necessarily function to identify
an addressee, in order to establish that imperatives can have subject NPs all that is
needed is to show that imperatives can have preverbal NPs that do not refer to the
addressee (Potsdam 1996, pp. 188-198).
13
The examples provided are these, each of
13
It's important to note that this can't be established by showing that imperatives can have
preverbal NPs that are 3rd person, because 3rd person NPs can be used to refer to an addressee.
Consider `everyone' used to refer to all the people listening to the speaker.
18
which is intuitively grammatical:
(24) a. You and Fred stay out of the construction zone.
b. Your men guard the front while we creep round the back.
c. Nobody ride o with my bicycle.
Though the `you' in (24-a) will pick out an addressee, `Fred' need not. Consider a
case where you and I are having a conversation and Fred is not around. At some
point you tell me that you and Fred have been going into a construction zone, to
which I respond with (24-a). This seems like a perfectly apt and felicitous response
on my part; there is nothing odd about it in this situation. But note that since Fred
is not in the conversation, he can't be an addressee. So `Fred' and, by extension,
the larger construction `You and Fred' cannot be vocatives. They instead must be
subjects.
(24-b) and (24-c) are even more compelling. Regarding (24-b), the NP `your
men' can't be used to refer to the addressee. It rather has to refer to something
else that stands in a particular relation to the addressee (the being the men of
relation). And regarding (24-c), the NP `nobody' can't refer to the addressee, because
\the set of people that `nobody' picks out is empty" (Potsdam 1996, p. 192). The
preverbal NPs in the imperatives (24-b) and (24-c) therefore also can't be vocatives.
By elimination, they too must be subjects.
Potsdam's other argument against the view that all apparent imperative sub-
ject NPs are really just vocatives once again turns on anaphoric pronouns, and how
they relate to vocatives. An interesting feature of such pronouns is that they must
be 2nd person when they are anaphoric on a vocative.
(25) a. Helen
4
, x the leak yourself
4
.
b. Ian
5
, clean your
5
mess up.
c. Jen and Kevin
6
, be kinder to your
6
neighbors.
Compare these to:
(26) a. *Helen
7
, x the leak herself
7
.
b. *Ian
8
, clean his
8
mess up.
c. *Jen and Kevin
9
, be kinder to their
9
neighbors.
The dierence in grammaticality of the constructions in (25) and (26) is entirely
due to the fact that pronouns that are anaphorically related to a vocative must
be 2nd person. This means if there are imperatives with non-2nd person anaphoric
pronouns, then the preverbal NPs in these sentences aren't vocatives. As it happens,
19
we have already seen one example of an imperative with a 3rd person re
exive (hence
anaphoric) pronoun with (20-a). Others include, again from Potsdam:
(27) a. Somebody
10
lend me his
10
handkerchief. (Said indiscriminately to a
crowd)
b. Everyone
11
pick up his
11
bill on the way out. (Said by host to guests)
c. One of the boys
12
test himself
12
while I wait. (Said by a teacher to a
class)
d. Nobody
13
forget her
13
lunch for the picnic. (Said to a group of girls)
Given the presence of the 3rd person anaphoric pronouns in these sentences, the
preverbal NPs can't be vocatives. They are subjects. This together with the ndings
of the last section is sucient for establishing that imperatives do have subjects, just
like any other clause type.
1.3.3 Imperative Predicates
In English and many other languages, perhaps the most recognizable thing about
imperative sentences is that their verbs are typically in the bare innitive form. This
fact is readily seen by looking back at all of the example imperatives of the preceding
sections. Consider the sentences in (8), repeated here:
(8) a. Go to the store
b. Eat your vegetables.
c. Mix together the wet and dry ingredients.
The verbs of these sentences are `go', `eat', and `mix', respectively. That these verbs
are in the bare innitival form is exhibited by their surface morphology, as well as
the fact that trying to replace them with any of the verbs' nite forms results in
ungrammatical constructions:
(8
0
) a. *Went to the store.
b. *Ate your vegetables.
c. *Mixes together the wet and dry ingredients.
14
14
Of course these constructions might be used as elliptical declaratives. For example, (8
0
-a) could
be used as a response to the question What did I do last night?, in which case it is elliptical for
`You went to the store'. Interesting as this is, it is beside the present point, which is just about
imperatives and that these constructions cannot be imperatives of any sort in English, elliptical or
otherwise.
20
Since the only relevant dierence between (8-a)-(8-c) and (8
0
-a)-(8
0
-c) is that the
verbs are innitive in the former and nite in the latter (`went' and `ate' are past
tense nite forms of the verbs GO and EAT, and `mixes' is a present tense nite
form of MIX ), the ungrammaticality of (8
0
-a)-(8
0
-c) must be due to the niteness of
their verbs.
From this observation we might try to generalize and hold that all imperatives
must have innitive verbs. But this would be too quick, as there appear to be imper-
atives containing the auxiliary verbs BE and HAVE in which the other, participial
verbs are in
ected for tense or aspect, respectively:
(28) a. Be cleaning the laundry when I return.
b. Be opening the shop at 9am.
c. Have eaten the casserole by next Tuesday.
d. Have changed the sheets before the guests arrive.
The verbs CLEAN and OPEN in (28-a) and (28-b), respectively, are both morpho-
logically marked for progressive aspect by the sux -ing, while EAT and CHANGE
in (28-c) and (28-d) are marked for perfect aspect by the suxes -en and -ed, re-
spectively.
15
Thus, assuming that a dening feature of a verb being innitive is that
it is not marked for grammatical aspect, the verbs of the sentences in (28) are not
innitive. They are, however, still nonnite, because they are in participial form,
and participial verb forms are generally considered nonnite. What's more, it does
appear the verbs of English imperatives must be nonnite (innitive verb forms are
nonnite) given that there are no known examples of English imperatives with nite
verbs.
Is the fact that the verbs of imperatives must be nonnite what distinguishes
them from other clause types (in at least English and other in
ectional languages)?
Unfortunately, no. Innitive to-clauses and especially subjunctive clauses also obli-
gatorily have nonnite verbs.
(29) a. Jack decided to go to the store.
b. To win the race is all Jill wanted.
c. It would be surprising for the bus to arrive now.
15
The time references needed for the grammatical aspects of these sentences to \latch onto" is
provided by the adjunct PPs `when I return', `by 9am', `by next Tuesday', and `before the guests
arrive'. Interestingly, however, such temporal PPs are not necessary for an imperative to have a
participle verb. So where does the time reference needed for the aspect of these sentences come
from? Later (inx4) I'll argue that imperatives have tense, and it is this tense that is being latched
onto.
21
The italicized constructions in (29) are innitive to-clauses. The dening feature
of such clauses, unsurprisingly, is that their verbs are in the to-innitive form (`to
go', `to win', `to arrive'). Since their verbs are obligatorily in innitive form, all
to-clauses must have nonnite verbs, just like imperatives. But unlike imperatives,
to-clauses are dependent clauses, meaning that they need to be embedded in a larger
construction to avoid ungrammaticality. Compare (29-a)-(29-c), in which the to-
clauses are embedded, to (30-a)-(30-c), in which they are not.
(30) a. *(Jack) To go to the store.
b. *(Jill) To win the race.
c. *The bus to arrive now.
The constructions in (30) are ungrammatical because to-clauses cannot be used
as standalone clauses. So though we cannot distinguish imperatives from innitive
clauses in terms of the niteness of their verbs
16
, they are distinguishable along the
dependent clause/independent clause dimension.
Like imperatives and innitive clauses, subjunctive clauses also obligatorily
have nonnite verbs. Moreover, like imperatives and unlike innitive clauses, they
are also independent clauses. Though use of subjunctives as standalone clauses (as
opposed to embedded uses) has largely fallen out of fashion in present-day English
17
,
they are nevertheless still grammatical constructions.
(31) a. The rock be falling o the cli.
b. Logan consider being mayor.
c. Maya run in circles.
The morphological similarity between the verb forms of subjunctives and impera-
16
We could try to do this by saying that while the verbs of imperatives can take participle
nonnite forms, the verbs of innitive clauses cannot. But this move won't work, since the verbs of
to-clauses can also be in participle form:
(i) a. Jill told Jack to be cleaning the laundry when she returns.
b. The owner seemed to be opening the shop at 9am.
c. To have eaten the casserole by next Tuesday is a good idea.
d. Jerry decided to have changed the sheets before the guests arrive.
This of course raises interesting questions about the relation between imperatives and innitive
clauses: Why is it that a use (e.g. (i-a)) or acceptance (e.g. (i-d)) of an imperative is reportable by
way of an embedded to-clause? Is it solely because of their syntactic similarity vis- a-vis their verbs
being nonnite? Or does it perhaps have something to so with the semantic side of things? I will
return to these questions later, in Chapter 3.
17
See Harsh (1968) for a possible explanation.
22
tives is obvious. They are so similar, in fact, that some have gone so far as to hold
that imperatives and subjunctives should be categorized under a single grammatical
category (Isac 2015), a possibility which can't be rejected out of hand.
The connection between imperatives and subjunctives is further strengthened
by certain sociolinguistic facts. For instance, Huntley (1984, p. 109) points out that
not only are imperatives and subjunctives morphologically similar, but one of the
most common uses of subjunctives in English is in reporting a directive use of an
imperative. Suppose Bill uses (32-a) to issue a demand to Mary, (32-b) to issue a
request to Joe, and (32-c) to issue an order to Ted.
(32) a. Mary, invite Joe.
b. Joe, allow Mary to drive.
c. Ted, keep an eye out for Mary.
The best way to report these uses of (32-a)-(32-c) by Bill is with:
(33) a. Bill demanded that Mary invite Joe.
b. Bill requested that Mary be allowed to drive.
c. Bill ordered that Ted keep an eye out for Mary.
18
Here we have that-clauses embedded in larger declarative constructions having verbs
that semantically correspond to dierent directive speech acts. And as it happens,
the that-clauses contain the subjunctives `Mary invite Joe', `Mary be allowed to
drive', and `Ted keep an eye out for Mary'. Given certain prevalent assumptions
about true speech act reports
19
, this provides prima facie reason for thinking that
imperatives and subjunctives are semantically quite similar as well (a point I will
return to Chapter 3).
Even if imperatives and subjunctives fall under a single grammatical category
due to the forms that their verbs take, and even if they are semantic kin, there is still
reason to distinguish them syntactically. One striking dierence between them is that
while subjunctives can occur embedded just about anywhere that a declarative can,
the same does not go for imperatives. In particular, subjunctives can, but imperatives
cannot, embed under the complementizer IF. For example, if the dependent clause
of (34) is taken to be a subjunctive, it is grammatical. But if it is taken to be an
imperative, it is ungrammatical.
(34) (*)If Jack be happy, then he be asleep. [no * if subjunctive, * if imperative]
18
The rst two examples are borrowed from Huntley (1984); the third is mine.
19
In particular, that such reports are true only if the content of the embedded clause is the same
as the content of what's being reported as having been said.
23
That this restriction on embedding is a syntactic dierence between imperatives
and subjunctives, and not, say, a semantic or pragmatic one, is something that I
will merely assume for now (returning to argue for it in chapter 4). So there must
be some dierence in the syntax of imperatives and subjunctives that explains this
restriction. And as we've seen, that dierence is not in the verb or VP.
Another reason the fact that English imperatives obligatorily have nonnite
verbs doesn't really help in distinguishing them qua clause-type is due to there be-
ing non-in
ectional languages, in which verbs only take one form and so the nite
verb/non-nite verb distinction does not apply. If imperatives are to form a univer-
sal clause-type, as is commonly assumed, then whatever is distinctive of them had
better be present in all the languages in which they occur. But most non-in
ectional
languages do have identiable imperative sentences; Cantonese is one example, in
which imperatives are characteristically formed with the particle `l a' (Matthews and
Yip 2013, Ch. 19). So verb-form can't be the distinguishing syntactic feature of
imperatives. Something else is doing the work.
Setting verb-form aside, another not uncommon thought about the predicates
of imperatives is that their verbs must be dynamic, as opposed to stative (Barker
2012). The dynamic/stative verb distinction is a semantic one, in contrast to mor-
phosyntactic distinctions like the nite/nonnite distinction between dierent verb
forms. Dynamic verbs are those that refer to or describe action-types or a type of
event that unfolds over a period of time. RUN, MOVE, HIT, and CHANGE are
some examples of dynamic verbs. Stative verbs, by contrast, are verbs that refer to
or describe types of states or properties of things. Included here are verbs like copular
BE, LOVE, HATE, KNOW, and SIT.
20
More often than not, the imperatives we use in everyday conversation do have
dynamic verbs in their matrices. We have seen many examples of such commonly used
imperatives over the course of this chapter. In fact, nearly all of the matrix verbs
of the imperatives we have looked at are dynamic, lending some weight to the idea
that the verbs of imperatives must be dynamic. But we have also seen an imperative
with a stative verb. Namely, (27-c), repeated here:
(27) c. Jen and Kevin, be kinder to your neighbors.
And it's not hard to come up with other examples of imperatives with stative verbs:
(35) a. Love yourself.
b. Hate those who have harmed you.
c. Sit down.
20
See Jackendo (1972); Dowty (1979); Binnick (1991) for more on stative and dynamic verbs.
24
d. Desire to be more open-minded.
Thus, like other clause-types, the verbs of imperatives can be either stative or dy-
namic, which means the stative/dynamic distinction also doesn't help us distinguish
imperatives qua clause type.
We could go on examining the verbs and VPs of imperatives attempting to
nd some way of putting them to use in dening what's distinctive about impera-
tives, but doing so will likely be a waste of time. This is because there is just nothing
fundamentally distinctive about these. Whatever features the verbs and VPs of im-
peratives might have, or whatever restrictions might be placed on them, will be
shared by other clause types. That of course doesn't mean that these features and
restrictions aren't of interest. They certainly are. It's just that they won't help us
dene { or will be of too limited help in dening { what imperatives are.
1.3.4 Imperatives, Negation, and Do-Support
We can learn more about the syntactic structure of imperatives by looking at how
certain morphosyntactic elements behave in them. Of particular interest are the aux-
iliary verb DO, negation, and adverbs because their presence and interactions suggest
that, like other clause-types, the syntactic structure of the left-periphery of impera-
tives is rather ne-grained (Pollock 1989; Rizzi 1997; Cinque and Rizzi 2008, and the
references therein). In this section I look at just negation and DO, and conclude (with
many others) that imperatives have NegPs and IPs in their syntactic structures. That
imperatives can have morphemes marking them for negation probably won't come
as a surprise to most, but that they have in
ectional morphemes might, especially
given what was already said above about the verbs and VPs of imperatives being
nonnite, which suggests that imperatives themselves are not in
ected. Suggestive
as this is, to capture the data imperatives need to have in
ectional morphology, or
so I will maintain.
Imperatives can be negated. An interesting feature of negated imperatives,
however, is that the auxiliary verb DO must be present. Consider the sentences in
(8) again:
(8) a. Go to the store
b. Eat your vegetables.
c. Mix together the wet and dry ingredients.
We cannot just throw a `not' into these constructions in order to form their negated
counterparts:
25
(36) a. *Go not to the store.
b. *Eat not your vegetables.
c. *Mix not together the wet and dry ingredients.
Rather, a preverbal `do' (the bare innitive form of supportive DO) needs to be
added and then `not' needs to be placed between this and the main verb:
(37) a. Do not go to the store.
b. Do not eat your vegetables.
c. Do not mix together the wet and dry ingredients.
(Compare:
(38) a. *Do go not to the store.
b. *Do eat not your vegetables.
c. *Do mix not together the wet and dry ingredients.)
This is quite dierent from non-negated imperatives, in which a preverbal occurrence
of DO is grammatically optional.
Why this grammatical need for an occurrence of supportive DO in negated
imperatives, and why must `not' be between this and the main verb? It is generally
accepted that the answer has something to do with the pre-VP syntax of clauses.
\The pre-VP syntax of clauses" refers to a particular region of the syntactic struc-
tures of clauses: the region to the left of the VP. Most are familiar with the idea
that phrases are decomposable into two basic parts: a head and a complement. The
head of a phrase is that word which determines the type of phrase that it is, and
the complement of a phrase is any word, phrase, or clause modifying the head. To
illustrate, since we know `eat at 7pm' is a VP, its head must be the verb EAT (oth-
erwise it wouldn't be a VP) and the prepositional phrase (PP) `at 7pm' must be a
complement. A tree is normally used to make this syntactic structure concrete:
VP
PP
NP
7pm
P
at
V
eat
Not only are phrases decomposable into their syntactic constituents, but so
26
also are clauses. But with clauses things are complicated by the fact that they are
often morphologically marked for things like tense, aspect, case, person, and the
like. In in
ected languages these features are often represented in the morphology
of a sentence's verb, but there is good reason not to build them into the syntactic
structure of VPs. Perhaps the simplest reason is we want to say that, despite their
verbs carrying dierent in
ections, sentences like the following all have the same
main VP:
(39) a. Jack went to the store.
b. Jack goes to the store.
c. Jack will go to the store.
Where these sentences dier is in their tense. (39-a) is past tense, (39-b) is present,
and (39-c) is future. But if we say that tense is part of a VP, then these sentences
don't have the same VPs, since each will have a structural element not shared by
the others. So, to avoid this, we need to separate tense from the VP.
21
The most straightforward way of separating tense from the main VP and still
acknowledging its role as a sentential constituent is by maintaining that in addition
to VPs (and subjects of course; more on these later) sentences have In
ection Phrases
(IPs), which are partly made up of in
ectional morphemes encoding things like tense,
aspect, and perhaps mood and they are located \above and to the left" of the VP in a
syntactic structure. To take a simple example, consider the phrase `goes to the store',
which is in
ected for at least present tense. According to the present approach, this
is not a VP. Instead, it is an IP, with its head being the axal morpheme -es and
its complement being the VP `go to the store':
21
(39-c) is particularly persuasive, because it's utilizing the future tense auxiliary verb WILL
which is clearly not a part of the VP. Compare:
(i) *Jack go will to the store.
27
IP
VP
PP
the store
NP P
to
V
go
I
-es
How do we go from this structure to the familiar `goes to the store'?
22
By a process
called V-to-I movement. In V-to-I movement a verb is \raised" from its position as
the head of a VP to a higher position in the in
ectional structure of a clause, leaving
behind what's know as a trace. Applying the process to the above structure results
in:
IP
VP
PP
the store
NP P
to
V
t
1
I
go
1
+ -es
Reading this left-to-right, V-to-I movement thus gives us the ordering of words and
morphemes we are familiar with: `goes to the store'.
We can now turn to the morphosyntactic role DO plays in clausal structures.
To get an idea of what this role is consider the following a-b pairs of declaratives.
22
Admittedly, it would be better to utilize a more sophisticated theory of phrase structure, such
as
X Theory (see Jackendo (1977); Kornai and Pullum (1990) than the simplistic one I'm using.
But its simplicity is why I'm using it. Since the points I'm making can be made equally well using
this very basic approach to phrase structure, bringing in a more sophisticated machinery would,
at this point, only introduce unnecessary complexity into the discussion and thereby increase the
likelihood of confusion. In the next section, however, when I turn to look at the syntactic roles of
adverbs in imperatives, I will have to complicate things. But we'll cross that bridge when we come
to it.
28
(40) a. Jack goes to the store.
b. Jack does go to the store.
(41) a. Jack eats his vegetables.
b. Jack does eat his vegetables.
(42) a. Jill mixed together the wet and dry ingredients.
b. Jill did mix together the wet and dry ingredients.
(43) a. Jill changed her mind.
b. Jill did change her mind.
The main verb of each a-sentence is in
ected for a particular tense, agreement, etc.
But in the b-sentences the main verbs are not in
ected; they are rather in the bare
innitive form. But these sentences still possess in
ectional tense, agreement, etc.
thanks to DO. In the b-sentences, DO is acting as an auxiliary verb and is itself
in
ected for the tenses, agreement, etc. of the sentences. In (40-b) and (41-b), DO
is in
ected for (at least) present tense by the morpheme -es, while in (42-b) and
(43-b) it is in
ected for (at least) past tense by the taking on the form `did'. Oddly,
however, DO itself doesn't appear to add anything to the meanings of these sentences.
Arguably, each a-b pair means exactly the same thing. All DO seems to be doing
is (i) allowing the main verbs to remain in their original positions in VP, instead of
undergoing movement to I
0
(the head of IP), by (ii) giving the stranded in
ectional
morphemes something to hang on to. DO thus seems to be a dummy verb serving
a purely { and in this case, optional { supportive role (hence the name, supportive
DO).
23
It is not, however, always the case that the appearance of DO is optional like
it is above. Suppose we want to negate the a-sentences in (40){(43). To do so, we
cannot just toss a preverbal or post-verbal `not' into these constructions.
(44) a. *Jack goes not to the store.
b. *Jack not goes to the store.
c. *Jill mixed not together the wet and dry ingredients.
d. *Jill not mixed together the wet and dry ingredients.
23
So why use the b-sentences when they are semantically equivalent to the a-sentences? One
common observation is that the b-sentences are used for emphasis. If I really want to get the point
across that Jill in fact mixed together the ingredients, then I will use (42-b) and place focused stress
on `did', such that \Jill DID mix together the wet and dry ingredients." It's worth noting that some
have argued that this emphasis should be semantically encoded into DO, and thus that DO does
contribute to the meanings of the b-sentences (Krifka 1995). Though I nd these arguments suspect,
nothing I say ultimately turns on DO being semantically void. So those who are inclined towards
the opposite view can feel free to ignore what I say about the semantics of DO.
29
Rather, it is obligatory that a form of DO be added and `not' be placed between this
and the main verb:
(45) a. Jack does not go to the store.
b. Jill did not mix together the wet and dry ingredients.
A similar thing happens when constructing interrogative sentences lacking any other
auxiliary verb or copular BE. Compare:
(46) a. Will he go to the store?
b. Can he go to the store?
c. *Goes he to the store?
d. *Goes to the store he?
e. Does he go to the store?
(47) a. Should she eat her vegetables?
b. May she eat her vegetables?
c. *Ate she her vegetables?
d. *Ate her vegetables she?
e. Did she eat her vegetables?
Just like when trying to construct certain negated declarative sentences, the only
way to construct interrogatives without using another auxiliary verb is by utilizing
supportive DO. The question is, Why?
The answer has to do with in
ectional morphemes and V-to-I movement. One
feature of V-to-I movement not mentioned above is that it is a strictly local process
in English: a verb can only undergo V-to-I if there is nothing between V
0
and I
0
(Radford 2006, pp. 106-109). If there is something between these heads, the verb
cannot make the move. As it happens, negation gets in the way. In the syntactic
structure of a negated sentence, the negation phrase (NegP) occurs above VP but
below IP, which is where TP resides, such that:
IP
NegP
VP Neg
I
30
Because of interference by NegP, the verb in V
0
is not able to move to I
0
and any in-
ectional morphemes in I
0
will be left to their own devices. But as we know, in
ection
morphemes are not standalone words; they need to be bound to another morpheme
or word (`Jack -es not go to the store' is not grammatical). This is why DO must bef
brought in, to support the in
ectional morphemes that are left stranded by NegP
(`Jack do+-es not go to the store'). Similarly for the aforementioned interrogatives:
the reason DO must be present is because something in the syntactic structure of
these interrogatives is blocking V-to-I movement, resulting in the need for DO.
24
Since DO is playing an obligatory support role in certain negated declara-
tives and interrogatives (and an optional support role in other constructions), the
phenomenon under discussion is often called DO-support (or DO-insertion). Now to
bring the discussion full-circle, our question is: Is what's going on with negated and
non-negated imperatives also an instance of DO-support? Reason to think that it is
comes from the fact that, like with indicatives, DO must precede the main verb, and
in negated imperatives the `not' must be situated between DO and the main verb,
as we saw in the sentences in (37), repeated here.
(37) a. Do not go to the store.
b. Do not eat your vegetables.
c. Do not mix together the wet and dry ingredients.
The occurrences of DO in these sentences is obligatory, and it's plausible the reason
for this is that an in
ectional morpheme has been stranded due to `not' and NegP
blocking V-to-I movement of the verbs.
What's more, appealing to DO-support can also help us explain the dierence
between the following a-b pairs of imperatives.
(48) a. Go to the store.
b. Do go to the store.
(49) a. Eat your vegetables.
b. Do eat your vegetables.
(50) a. Mix together the wet and dry ingredients.
b. Do mix together the wet and dry ingredients.
Like with the pairs of declaratives in (40)-(43), in the a-sentences here the verbs have
undergone V-to-I movement and taken up host to the morpheme that is the head
of IP. The dierence, of course, is that this morpheme is null: like with subjects of
24
This way of explaining the need for DO is undoubtedly oversimplied. For a more detailed
explanation, see Radford (2006, Ch. 4).
31
these sentences, the in
ectional morpheme is not phonologically realized. But we can
represent it with `?
Infl
', so that the a-sentences might instead be written:
(48
0
) a. Go+?
Infl
to the store.
(49
0
) a. Eat+?
Infl
your vegetables.
(50
0
) a. Mix+?
Infl
together the wet and dry ingredients.
(Or, if we wish to also take account of their subjects:
(48
00
) a.?
Subj
j go+?
Infl
to the store.
(49
00
) a.?
Subj
eat+?
Infl
your vegetables.
(50
00
) a.?
Subj
mix+?
Infl
together the wet and dry ingredients.)
As for the b-sentences in (48)-(50), the presence of DO is again explained by the fact
that V-to-I movement is optional and that without it DO must step in to support
the stranded in
ectional morpheme.
25
None of what I have said or argued in this section is new. Others have made
the same argument and/or come to the same conclusion
26
. Nevertheless, it may be
surprising to some that English imperatives must carry in
ectional morphology and
that imperatives in general must have IPs. But is it really so surprising? Just looking
at the surface structure of imperatives certainly does give the impression that they
lack all the in
ectional properties possessed by their indicative brethren. But once we
remember that imperatives are themselves clauses { sentences { the surprise begins
to wear o. After all, without an IP there would be almost nothing to distinguish
an imperative from a VP. Adding further support is the fact that some languages do
have special morphemes and/or in
ections for imperative sentences.
27
25
The presence of DO in a sentence can also function to mark emphasis. For example, in using
(48-b) a speaker might put extra stress on the pronunciation of `do', and do this in order to emphasis
that they want the addressee to carry out the act that this sentence species. Similarly for (40-b): a
speaker might put extra stress on `does' in order to emphasis that they believe that Jack preforms
the act that this sentence species. When used in this way it is called the emphatic DO. Though I
have characterized things in terms of certain uses of DO, which suggests that the emphatic DO is a
strictly pragmatic phenomena, it should be noted that some maintain that this emphasis is in fact
semantically encoded by an underlying Arm(ation)P in the syntactic structure of these sentences.
Perhaps my biggest objection to this view is that it requires us to make DO ambiguous between a
non-emphatic support meaning and an emphatic one. Like most, I believe we should avoid positing
ambiguities whenever we can, and since the emphatic DO can be understood purely pragmatically,
this means we should not incorporate it into our semantics.
26
E.g. Han (2000).
27
See, for example, Zanuttini (1994).
32
One question raised by saying that imperatives have in
ectional morphosyntax
is: Just what are their in
ectional heads? This question will be answer in the next
and nal section of this chapter, in which I look at three dierent accounts of what
makes an imperative an imperative. But before we can turn to that question there
are a few other matters regarding the syntactic properties of imperatives to attend
to.
1.3.5 Speciers in Imperatives: Subjects and Adverbials
The curious reader may at this point be wondering where the subjects of impera-
tives reside in their syntactic structure and, more generally, what the full syntactic
structure of imperatives looks like. In the previous subsection I did not address these
questions, instead only arguing that imperatives have IPs and that these dominate
imperative VP in syntactic structure. In this section I take on these questions. My
approach will be to extend the simplistic view of phrase structure I've been using
to the more sophisticated
X theory (pronounced \X-bar theory"), which adds sev-
eral additional layers of complexity to the basic structure and structural elements
of a phrase. One of these is the category of speciers, which are neither heads nor
complements in phrase structure.
28
What I wish to maintain { following again in the
footsteps of Potsdam (1996) { is that the subjects of imperatives, whether overt or
null, occupy the specier position of IP.
I begin with a quick overview of the traditional theory of phrase structure.
According to the traditional theory, a phrase is composed of at most a head and a
complement, following this schema:
XP
(complement) X
0
The `X' is a stand-in for any grammatical category whatsoever (lexical or functional;
e.g. V(erb), N(oun) I(n
ection), P(reposition), etc.), and the left branch is where the
complement of the phrase goes. XP is said to dominate both the head, X, and the
complement, while the head and the complement are said to be sisters, since they
share a common mother { XP { and are on the same structural level. Because X
0
is
the head of the phrase, and so denes what kind of phrase XP is, it cannot itself be a
phrase; it must be a lexical or functional morpheme (free or bound). Complements,
28
Another category that
X introduces is that of an adjunct, which for the sake of space I won't
be discussing here. But see Jackendo (1977).
33
by contrast, can be either morphemes or phrases. By allowing the complement to
be a phrase, the traditional theory is both recursive and productive: More complex
phrases (i.e. phrases containing other phases as constituents) are constructed out of,
and are therefore denable in terms of, simpler phrases, as we saw inx3.4.
Though the traditional theory is useful for many purposes, Chomsky (1970)
and others showed it to be ultimately inadequate. We don't need to get into the
details of the arguments, but the gist is that the traditional view has diculty ac-
counting for derived nominal phrases and their relation to the clauses from which
they are derived. For example, the constructions in (51) are derived nominal phrases
and those in (52) are the clauses they are derived from.
(51) a. Jack's criticism of the new law
b. Jill's ability to program computers
c. Jerry's fall into the well
(52) a. Jack criticized the new law.
b. Jill is able to program computers.
c. Jerry fell into the well.
For various reasons, however, the traditional view of phrase structure prevents any
straightforward, non-ad hoc derivation of the nominals in (51) from the clauses in
(52), and is therefore inadequate.
X theory is a generalization of the traditional theory that gets around the
problem. What it says, in schematic form, is that the basic structure of a phrase is
rather this:
XP (aka X
00
)
X
0
(complement) X
(specier)
Though now lower down in the structure, X
0
is still the head of the phrase,
and it is said to project up through the structure to X
0
and then to XP, which is
sometimes referred to as X
00
(\X-bar-bar"). Because of its position between X and
XP, X
0
is an intermediate projection of X, while XP is the maximal projection.
29
29
Phrase structures can have more than one intermediate projection, but only one maximal. Ad-
ditional intermediate projections arise when a phrase contains adjunctual material, that is, material
that can be deleted from the phrase structure without aecting the grammaticality of the phrase.
34
Importantly, now there are two structural positions for material that qualies an
X to occur: the complement, as before, and the specier. The specier of an XP is
conventionally referred to as SpecXP. Like complements, speciers can be thought
of as something that restricts { or further species, hence the name { the category of
an X. In the VP `quickly run', for example, the adverb `quickly' species the kind of
running involved, and in the NP `the barn' the denite article species the quantity
of barns under discussion.
30
Placing `quickly' in SpecVP and `the' in SpecNP, gives
us:
VP
V
0
(compliment) V
0
run
SpecVP
soon
NP
N
0
(compliment) N
0
barn
SpecNP
the
There can be only one complement and only one specier in an XP. So what distin-
guishes complements and speciers? In
X theory, the main dierence between them
is their respective positions in XP. Speciers are daughters of XP and sisters of X
0
,
whereas compliments are daughters of X
0
and sisters of X
0
(Chomsky 1970).
A simple example of this is the VP `eat vegetables at noon', which has the following
X phrase
structure with two intermediate projections.
VP
V
0
NP
N
0
at noon
PP vegetables
Spec NP
V
0
eat
SpecVP
30
See Cinque (1999) for adverbs; Chomsky (1970) for articles.
35
Now recall that imperatives and other clause types have IPs which dominate
VP. Using
X theory, this implies the following structure:
IP
I
0
VP
V
0
(compliment) V
0
SpecVP
I
0
DO/?
Infl
SpecIP
On this picture, the null in
ectional morpheme?
Infl
and DO will be base-generated
in I
0
. But what about the subjects of imperatives { where do they t into this
structure? There are two options. They could go in SpecIP or in SpecVP (Potsdam
1995).
31
SpecIP Hypothesis
IP
I
0
(compliment) I
0
SpecIP
(subject)
31
A third option would be to say that an imperative subject is in the specier position of an XP
that dominates IP. As we will see in a moment, the best candidate for this XP is C(omplementizer)P,
such that:
CP
C
0
IP C
0
SpecCP
(subject)
Since I know of no one in the literature who even suggests this view, I pass it by without comment.
36
SpecVP Hypothesis
VP
V
0
(compliment) V
0
SpecVP
(subject)
The SpecIP Hypothesis roughly corresponds to the traditional view that holds
that all sentences/clauses are decomposable into subjects and predicates (assuming
that auxiliary verbs are base generated in I
0
and not a position internal to VP), and
so on this view imperatives share the same basic clausal structure as declaratives and
interrogatives. The SpecVP Hypothesis, on the other hand, corresponds to the
idea that imperatives are essentially just VPs, where what makes a VP distinctively
imperative is not its structure, but rather its use in directive speech acts (Bolinger
1967; Downes 1977; Portner and Zanuttini 2003).
Potsdam (1995, 1996) argues that the SpecVP Hypothesis (he calls it the
VP-Internal Hypothesis) is fundamentally
awed. First, it predicts that in im-
peratives with DO or DON'T and another auxiliary, the subject can occur immedi-
ately to the left of the main verb. But such constructions are ungrammatical:
(53) a. *Do have someone nished by the time I return.
b. *Don't be you staying up too late.
By contrast, assuming that the auxiliary verbs BE and HAVE are the heads of
substructures of IP
32
, the SpecIP Hypothesis only allows a subject to occur to
the immediate left of auxiliary verbs.
(54) a. Do someone have nished by the time I return.
b. Someone do have nished by the time I return.
c. Don't you be staying up too late.
32
This assumption makes it necessary to break up { \split" { IPs into several hierarchically
ordered functional projections (e.g. T(ense)P, Asp(ect)P, Agr(eement)P, etc). Some may nd this
suspect. Fortunately, the assumption can be dispensed with and replaced by the hypothesis that
IPs are unied structures (Chomsky 1995) together with the either the assumption that auxiliary
verbs head position of an additional VP, sometimes called vP (pronounced \little v" phrase) that
is dominated by IP and dominates the main VP (as well as NegP when it occurs), or that DO is
base-generated in either I
0
or C
0
(see below regarding CPs). What's more, the only major wrinkle
added by IPs being split into more ne-grained phrase structures would be that we would have to
determine which specier position of which substructure the subject belongs in, something which
Potsdam (1995) discusses.
37
d. You don't be staying up too late.
Second, the SpecVP Hypothesis predicts that quantier
oat can't occur in im-
peratives, a predication which doesn't pan out, as shown by
33
:
(55) a. You all go to the store.
b. You both go to the store.
The SpecIP Hypothesis, however, can accommodate quantier
oat. Lastly, the
SpecVP Hypothesis also wrongly predicts that so-called E-adverbs, which modify
the VP (Jackendo 1972), can occur to the left of an imperative's subject, as in the
following b-sentences:
(56) a. Somebody just go to the store.
b. *Just somebody go to the store.
(57) a. Nobody quickly run the race.
b. *Quickly nobody run the race.
(58) a. Everyone simply turn over one card.
b. *Simply everyone turn over one card.
The SpecIP Hypothesis only allows E-adverbs to occur to the right of the subject,
as in the a-sentences. Given this (and much more I'm not able to discuss), the
SpecIP Hypothesis is to be preferred; like with other clause types, the subjects
of imperatives are speciers of IP.
So here is the view of the structures of imperatives that we have arrived at:
33
Quantier
oat is what happens when a quantier moves { \
oats" { to the right of the NP
that it is quantifying over. In (i-a) the quantifying word `both' is in its original position to the left
of the NP `the dogs', and in (i-b) and (i-c) `both' has \
oated" to positions to the right of `the
dogs'.
(i) a. Both of the dogs are hungry.
b. The dogs both are hungry.
c. The dogs are both hungry.
See Bobaljik (2003) and the references therein for further discussion of quantier
oat.
38
IP
I
0
VP
V
0
(compliment) V
0
SpecVP
I
0
DO/?
Infl
SpecIP
(subject)
But we are not done. One more layer needs to be added to this picture. Specically,
we need to add a Complementizer Phrase (CP) layer to the left-periphery, so the
nal structure of imperatives looks like:
CP
C
0
IP
I
0
VP
V
0
(compliment) V
0
SpecVP
I
0
DO/?
Infl
SpecIP
(subject)
C
0
SpecCP
The main need for adding a CP layer comes from the occurrence of certain sentence-
initial and sentence-nal adverbs and adverbial NPs in imperatives (collectively:
adverbials). Consider:
(59) a. Frankly (do) go to the store.
b. Denitely (do) go to the store.
(60) a. Your book, (do) give it to me.
b. That cookie, don't eat it.
(61) a. (Do) move that box, not the other one.
39
b. Everyone (do) hand in yesterday's assignment, not today's.
The italicized adverbials in these sentences aren't modifying DO or VP. Rather,
they are somehow modifying the features of the entire sentence. In (59), `frankly'
and `denitely' appear to be qualifying the force that the sentences can be used with,
while in (60) `your book' and `that cookie' are qualifying the sentences' topics, and
in (61) `not the other one' and `not today's' are qualifying the sentences' focuses.
Importantly, however, in order for these adverbials to have the roles they do in
these sentences they must occupy a structural position external to IP (Rizzi 1997).
Which means we need to add an additional IP-dominating layer to the structures of
imperatives { and that layer is CP.
1.3.6 Summary
A lot has happen inx3, so let's pause to sum up our ndings. Inx3.1 andx3.2 we found
that imperative have subjects, that these subjects are sometimes morphologically null
(i.e.?
Subj
), and that there are no distinctive grammatical restrictions on what the
subjects of imperatives can be (though there does appear to be a number of non-
grammatical restrictions on them; more on this in later chapters). Inx3.3 we found
that the VPs of imperatives are closely related to those of innitive and subjunctive
clauses, in possessing nonnite morphology. We also found inx3.3 that there are
no substantive restrictions on what the main verbs of imperatives can be (stative
or dynamic). Then inx3.4 we found that imperatives have IPs that structurally
dominate their VPs, that they optionally have NegP that dominate VP and are
dominated by IP, and that a null morpheme ?
Infl
occupies I
0
. Finally, inx3.5 we
found that the subjects of imperative occupy SpecIP, and also that imperatives have
CPs that dominate IP.
These ndings were earlier represented by this schematic tree:
40
CP
C
0
IP
I
0
VP
V
0
(compliment) V
0
SpecVP
I
0
DO/?
Infl
SpecIP
(subject)
C
0
SpecCP
Unfortunately, however, none of these structural features and elements of impera-
tives, nor all of them taken together, is sucient for distinguishing imperatives from
other clause types. The exact same schematic tree can be equally well used to repre-
sent the syntactic structures of declaratives, interrogatives, subjunctives, and other
clause types. So despite all that's been said, we are still left with the question of
what makes an imperative an imperative. But that doesn't mean progress hasn't
been made. For now we can identify where in their structures they must dier. Since
we know (fromxx1.3.1-1.3.3) that imperative VPs and subjects aren't grammatically
distinguishable from the VPs and subjects of other clause types, the dierentiating
factor must be either?
Infl
or something in CP. That being so, I now turn to three
dierent theories.
1.4 Three Theories about the Syntactic Nature of
Imperatives
1.4.1 The \Impoverished Clause" Theory
The rst theory I wish to look at says imperatives are impoverished clauses because
?
Infl
fails to encode certain functional features (e.g. time reference, person, number,
etc.) that are encoded by other clause types. To see what this amounts to consider the
fact that the in
ectional heads of declaratives in a number of languages are marked
for a number of functional features, including tense/time reference, aspect, number,
person, gender, and others. Let's use (62) to illustrate.
41
(62) Jack was going to the store.
This sentence has the following syntactic structure (note that I have purposefully
omitted CP, since we are not concerned with it at the moment):
IP
I
0
VP
V
0
to the store
PP V
0
go
SpecVP
I
0
BE[+Tense][+Aspect][+Number][+Person]
SpecIP
Jack
The important thing to notice here is that we have made explicitly marked I
0
for the
functional features of (62), in this case that it has tense (Past), aspect (Progressive),
number (Singular), and person (3rd). What's more, we know that all declarative
sentences need to be somehow marked for certain functional features.
34
The most
obvious of these is time reference. In English and most other in
ecting languages,
for example, time reference is grammaticalized by tense morphemes that ax to the
main and auxiliary verbs of a sentence. What's more, a construction lacking tense
morphology is not a declarative in languages like English. Compare the following,
where each construction is taken to be a declarative:
(63) a. Jack is happy.
b. *Jack be happy.
(64) a. Jack did go to the store.
b. *Jack do go to the store.
(65) a. Jill has eaten her vegetables.
34
Of course the way a sentence is marked will depend on the language. In
ected languages mark a
sentence by placing in
ections on verbs. Non-in
ected languages do so either by having designated
words (i.e. free morphemes) carrying the features, or by using variations in word order to indicate
the presence of these features, or by leaving it up to language users to infer it from a context. See
Matthews (1991).
42
b. *Jill have eaten her vegetables.
Importantly, even if a declarative is not positively marked for a particular functional
feature there is good reason to think that it is negatively marked for it. Take aspect
for instance and consider:
(66) Jack is hungry.
Since (66) carries present tense on it's main verb { the copular BE { we say it
is positively marked for tense. But there is nothing about the morphology of (66)
that indicates it carries some sort of aspect. Despite this, because we know that
declaratives can be marked for aspect, instead of saying that aspect is in no way
present in the syntactic structure of (66), we say that it is negatively marked for it.
IP
I
0
VP
V
0
to the store
PP V
0
go
SpecVP
I
0
BE[+Tense][Aspect]
SpecIP
Jack
So, structurally, there is still a place in (66) for aspect to occur in IP; it's just that
it doesn't occur.
35
We can now more clearly state what it means for?
Infl
to be \impoverished".
It means that it fails to even provide a place in the syntactic structures of imperatives
for certain functional features that are present in other clause types. That is, the
35
Another way to make this point would be to `explode" IP into a T(ense)P, an Asp(ect)P,
and an Agr(eement)P, and then maintain that all declaratives have TP, AspP, and AgrP in their
syntactic structures. Saying that a sentence is positively marked for tense, aspect, person, etc., will
then amount to saying that there is a morpheme occupying T
0
, Asp
0
, and Agr
0
, respectively. And
saying that a sentence is negatively marked for one of these functional features will mean that T
0
,
Asp
0
, and/or Agr
0
is empty. Insofar as the point I'm driving at is concerned, these two approaches {
exploding IP and not exploding it { are merely notational variants; both can do the job equally well.
The only reason I've opted for the non-exploded IP approach is because of its relative simplicity.
43
feature is neither positively marked nor negatively marked { it is essentially absent
from an imperative's structure.
Which functional feature is it that?
Infl
fails to possess? There are a number
of proposals, but perhaps the most popular is it fails to mark for tense. Zanuttini
(1991) is a representative of this view:
Zanuttini (1991) proposes that the crucial dierence between true im-
peratives and other verbal forms (even those which might happen to
be supercially identical to the true imperatives [e.g. subjunctives]) is
that imperatives lack certain functional projections [ functional fea-
tures] which are associated with verbal forms expressing other tenses and
moods. In particular, following the assumption that the expression of
tense has as its syntactic counterpart a functional projection TP, the
proposal is that the projection TP is missing from the syntactic repre-
sentation of true imperatives. . . . It is argued that verbal forms from the
indicative, the subjunctive and the innitive have a projection TP. . . ,
whereas true imperatives do not have a projection TP, neither lexically
realized nor abstract. (Zanuttini 1994, pp. 125-126)
Though Zanuttini expresses things in terms of the absence of a functional projec-
tion, TP, the view described here is at its core just another way of stating the view
that?
Infl
fails to encode tense, a functional feature possessed by other clause types.
Hence, on this view, imperatives are defective in the sense that they lack a crucial
feature possessed by the other clause types (declaratives, interrogatives, subjunc-
tives, etc.). But tense is not the only functional feature that has been argued that
imperatives completely lack. Other candidates include: agreement features, mood,
niteness, and at the extreme, all of these. Fortunately, however, the proponents of
the impoverished clause view don't have to say exactly which functional features
?
Infl
fails to encode. Rather, following Zanuttini (1994, p. 131), they just have to
say that it fails to encode some one or more of these features. Which feature exactly
is another question entirely, one that Zanuttini showed doesn't need to be answered
in order to advance the impoverished clause theory of imperatives.
The usual arguments for the impoverished clause theory of imperatives (some-
times instead called the defective clause theory or the primitive clause theory) often
focus on the meager in
ectional form of imperative verbs, something which we noted
inx3.3. Schmerling (1982) and Huntley (1980) both argue that this meager verb form
gives us prima facie reason to believe the impoverished clause theory. This is because
there must be some reason why, despite possessing the in
ectional morpheme?
Infl
,
44
the main verbs of imperatives still only take nonnite forms, and that reason seems
to have to be due to?
Infl
lacking certain important features.
Tempting as it is, this argument is quite weak. The meagerness of a word- or
morpheme-form does not necessarily correlate to a meagerness of functional features
of a sentence. Non-in
ected languages are a case in point. For obvious reasons, the
main verbs of sentences in these languages only take one form. Nevertheless, these
sentences can be marked for any number of functional features. To put it another way,
if the meager form of the verb is what establishes the impoverished clause theory of
imperatives, then it should equally establish that all clause types in all non-in
ected
languages are impoverished. Surely this is a conclusion we want to avoid.
Another problem for the impoverished clause theory is that for any functional
feature that a proponent may say imperatives lack, a convincing counterargument for
its presence can be found. Take tense for example, a prime candidate for what must
be missing from imperatives. It has been noted that a number of languages do overtly
mark imperatives for tense of some sort. Latin and Bengali are two such examples
(Van der Wur 2007,x5.2). In addition to the hard to ignore cross-linguistic data,
more theoretical arguments are also available. For instance, it has been argued that
VP-ellipsis is licensed only when an auxiliary verb with a [+Tense] feature is present
(Lobeck 1999). Accordingly, if imperatives can undergo VP-ellipsis, then they must
possess [+Tense]. As it happens, it's not hard to show that VP-ellipsis can occur in
imperatives:
(67) a. You
1
go to the store, but don't you
2
[go to the store] { you
2
are to mow
the lawn.
b. Either eat your vegetables or don't [eat your vegetables] { I don't care
which.
Therefore, imperatives have [+Tense].
36
1.4.2 The Future Tense Theory
That imperatives have tense may be surprising to some. To others, however, it is
what sets imperatives apart. More specically, the idea is that what is special about
imperatives is they are necessarily marked for future tense. Fitting this into the
current syntactic picture we are working with, the claim is that ?
Infl
is a future
tense morpheme. Call this the Future Tense Theory.
36
As we already saw inxx3.1 and 3.3, neither cross-linguistic data nor theory-driven arguments
are needed to show that imperatives can have the features [+Aspect], [+Number], and [+Person].
The surface facts of English imperatives were enough to do the job.
45
Proponents of this view, such as Katz and Postal (1964), argue for it on the
basis of tag-questions. One way to tag any imperative for a question is by using `will
you?':
(68) a. Go to the store, will you?
b. Eat your vegetables, will you?
If we assume that the WILL involved in these tags is a future-tense modal, then
that imperatives are necessarily marked for future tense seem to immediately follow.
Moreover, this means?
Inlf
has the same grammatical status as WILL. Though this
doesn't mean that ?
Infl
= WILL (for obvious reasons, including that WILL but
not ?
Infl
is phonologically realized), one consequence of this way of spelling out
the Future Tense Theory is that the following declarative-imperative pairs should be
semantically equivalent.
(69) a. You go to the store. (imperative)
b. You will go to the store. (declarative)
(70) a. Everyone eat their vegetables. (imperative)
b. Everyone will eat their vegetables. (declaratives)
(71) a. Don't you mix together the wet and dry ingredients. (imperative)
b. You will not mix together the wet and dry ingredients. (declarative)
For many reasons, however, the Future Tense Theory can't be right. First,
though it may be true that the declarative b-sentences in (68)-(70) can be used
in certain directive speech acts like orders and commands, they can't be used in
directives like advice, and so unlike imperatives, declaratives with WILL don't meet
the Imperative Identifying Criterion set out above. Second, there are cases
in which the imperative a-sentences in (68)-(70) are satised while the b-sentences
are not. For example, if a mother uses (68-a) to tell her son to go to the store, it
turns out that he already has, and it's true that he will never again go to the store
in his lifetime, then relative to this context, (68-a) is satised but (68-b) is not.
Lastly, there are languages with imperatives that appear to be marked specically
for present tense, such as Classical Greek (Rijksbaron 2002), and also languages with
imperatives marked for past tense, such as Dutch (Proeme 1984) and Frisian (Wolf
2013).
Regarding the tag data, it's not clear how much weight to place on this. WILL
regularly occurs in constructions like (72) (Sarkar 1998).
(72) Jack will go to the store now.
46
Because of the occurrence of `now', this suggests that in (72) WILL is a present-tense
or some other sort of modal, not a future-tensed one. If this is right (I think it is),
then that imperatives are tagged with `will you?' doesn't support the future-tense
theory.
Rather than say ?
Infl
is obligatorily marked for a specic tense like past,
present or future, which the cross-linguistic data alone shows is untenable, an alter-
native approach is to say it is marked for a more nonspecic tense, such as the simple
present. This not only requires distinguishing specic and nonspecic tenses, but also
requires postulating the existence of nonspecic tense, in particular the nonspecic
simple present. What exactly is nonspecic simple present tense? It's best to think
of the specic/nonspecic tense distinction in terms of restrictions on a reference
time/time of evaluation. Specic tenses place particular restrictions on the points in
time that a sentence makes reference to. Past tense, for instance, restricts the refer-
ence time to a point prior to the time at which a sentence is used. The nonspecic
tense, by contrast, places open-ended restrictions on reference time { it is left rela-
tively open or undetermined what the boundaries of the time of evaluation are. The
nonspecic simple present, in particular, encompasses points of time in the present
as well as points in both the past and the future, but does not strictly determine
how far into the past or the future the points extend.
37
It is important not to confused nonspecic tenses like the simple present with
the so-called universal tense (Comrie 1985, p. 40). To say a clause has universal tense
is to say it references all times { past, present, and future, inclusive. To illustrate,
some have thought certain generic declaratives (i.e. ones not given a \habitual"
reading) are one example of clauses with universal tense (Dahl 1995, The Generic
Book). Consider (73).
(73) Cats are animals.
Arguably, it would seem that (73) is true if and only if every cat that has, does, or
will exist { that is, every cat that exists at any time { is an animal, and in this way
the tense of (73) is universal. The nonspecic tense, by contrast, only references one
time, but when, relative to the context of use, that time is is left open. So it is, very
roughly, more like the sort of tense had by (74).
(74) Cats were, are, or will be playful.
In many ways, then, the nonspecic simple present tense is just like the aorist tense
37
There are, however, restrictions on how far into the past or future one can look. Two months
after the time of use might not be too late for an imperative to be satised; twenty years surely is.
47
of certain Indo-European languages like Ancient Greek and Sanskrit. If we wished
to go out on a limb, we might even say that nonspecic tense is an aorist tense, or
that aorist tense is a nonspecic tense, or that they are identical.
Nonspecic simple present tense allows us to explain why the satisfaction of
an imperative need not happen at a point in time strictly after the time it is used,
because if imperatives have nonspecic simple present tense marking, then this only
requires satisfaction to occur at some time close enough to the present. What's more,
we can also account for past tense imperatives by maintaining that the nonspecic
simple present tense of an imperative can be modied or restricted by additional
tense morphology. This works in much the same way that temporal adverbials can
modify or restrict a sentence's reference time
38
, as in:
(75) a. Jack is now going to the store.
b. Jack is today going to the store.
c. Jack is tomorrow going to the store.
d. Jack is next year going to the store.
(75-a)-(75-d) are each marked, by the morphology of `is', with the same basic tense:
the present. But due to the occurrence of dierent temporal adverbials in each, they
don't have the same reference times. (75-a) refers to right now, i.e. the exact time
that it is used. (75-b) refers to any time within the day that the sentence is used.
(75-c) refers to any time within the day after the day that the sentence is used. And
(75-d) refers to any time within the year after the year that the sentence is used.
In order to account for past tense imperatives, all we need to do is treat the past
indicating morphemes in them in the same way we treat the temporal adverbials in
(75), i.e. as restricting the nonspecic time reference of the sentences and not as the
primary tense indicators of the sentences.
As promising as saying?
Infl
is marked for nonspecic simple present tense is,
the major drawback is that it doesn't help distinguish imperatives from other clause
types. In particular, it doesn't help distinguish imperatives from subjunctive and in-
nitive clauses. This is because all the reasons and arguments for imperatives having
this tense can be, mutatis mutandis, used to show that subjunctives and innitives
have it as well. First, there is cross-linguistic data showing that subjunctives and
innitives can be tensed marked, most notably in Latin. Second, subjunctives and
innitives can undergo VP-ellipsis:
(76) a. Jack be going to the store and Jill be [going to the store] too.
38
Assuming of course that such adverbials do function to modify time-reference. This assumption
has been questioned by Larson (2003).
48
b. Jill wants to eat the vegetables and Jack wants to [eat the vegetables]
as well.
Third, from a semantic point of view, subjunctives and innitives can also be satised
at times prior to, at, or after the time-point of use.
(77) a. Jack be going to the store yesterday.
b. Jack be going to the store tomorrow.
c. Jill wants to have eaten the vegetables yesterday.
d. Jill wants to eat the vegetables tomorrow.
All in all, then, this strongly suggests that whatever tense features imperatives might
have, subjunctives and innitives will have them as well.
39
So neither specic (e.g. future) nor nonspecic tense helps set imperatives
apart from other clause types. This is bad news for any theory that wishes to distin-
guish imperatives in terms of tense.
1.4.3 The Special Mood Theory
The nal theory we'll look at, and the one I'll ultimately adopt, eschews the idea
that?
Infl
is where the dierence between imperatives and other clauses lies. Instead,
the dierence is located in a separate morpheme that occupies SpecCP, and which
carries a special imperative mood feature: [+Imperative]. Let's call this morpheme
IMP. Since IMP is what distinguishes imperatives according to this theory, the view
must hold that it can only occur in SpecCP of imperatives.
Like?
Infl
, in English IMP is always phonologically null; there are no phonemes
that it corresponds to. This may make the postulation of IMP appear ad hoc and
unsubstantiated. But as it happens there are languages in which IMP is articulated.
Regarding in
ectional realizations, the verbs of Latin imperatives, for example, have
a special form not found in other clause types, and some verbs in Spanish have an
irregular form that is unique to imperatives, such as the verb TENER, which is ren-
dered as `ten', `ten e' and `tened' only in imperatives. Regarding non-in
ectional real-
39
Some have argued, contra what I've just said, that the tense feature of innitive clauses is
necessarily future, and in this respect we might be able to distinguish imperatives and innitives
on the basis of tense. See Stowell (1982). For my part I nd these arguments, which are almost
invariably based on the \intuition" that innitives are future oriented, wholly suspect. Moreover,
such account will have a dicult { if not impossible { time providing a non-ad hoc explanation of
the cross-linguistic data, which shows us that in some languages innitives can be marked for past
tense and for present tense. For additional arguments against the future tense analysis of innitives,
see Wurmbrand (2007).
49
izations, languages that contain a specic imperative particle include: Greek (Anttila
2000), Korean (Pak 2004), Bodiotto (Poletto and Zanuttini 2003), and Hungarian
(P eteri 2012).
Some authors have also noted that IMP can be used to explain some of the
more confounding syntactic aspects of imperatives in certain languages, such as the
diculty of formulating English imperatives with 1st person subjects (see section
x3.1). Suppose that in English (but not necessarily in other languages) IMP carries
the feature [1st person] and that the subject of an imperative has to be checked
against this. Then since subjects that are [+2nd person] and [+3rd person] are also
[1st person], they will be in agreement with IMP and are thus permissible in im-
peratives in English, as the data shows. But any subjects that are [+1st person], like
1st personal pronouns, when checked against IMP, will fail to agree, thus producing
ungrammaticality, and this is why they are not permissible in English. Evidence for
IMP carrying additional features like [1st person] in some languages is provided
by the above-mentioned languages containing either unique imperative in
ectional
morphology or unique imperative particles. It appears that in many of these lan-
guages, when the imperative in
ection or particle is present the subject must be
[1st person] or carry some other functional feature.
40
This theory, that there is a unique mood morpheme in SpecCP has a lot
going for it. It also raises a dicult question that needs answering if the theory is to
have any plausibility. The question stems from the fact that the theory appeals to the
vague notion of sentential \mood." So here's the question: What is sentential mood?
A number of answers have been given. According to Palmer (2001), the notion of
mood goes hand in hand with that of modality, in that mood can be thought of as a
kind of grammaticalized modality in much the same way that tense can be thought
of as grammaticalized time reference (see previous section). Moreover, Palmer thinks
that the imperative mood represents the basic or primitive or \unmarked" modality
of the system of deontic modalities, which also includes the (non-basic) modalities
of permissible and obligatory.
41
To see what this means, consider the following sentences possessing the declar-
ative mood:
(78) a. Jack is going to go to the store.
b. Jill did eat her vegetables.
c. Jerry mixes together the wet and dry ingredients.
40
See Rupp (1999) for further discussion.
41
For an excellent introduction to deontic modalities, in the context of a discussion about deontic
logic, see McNamara (2014).
50
According to many, the mood/modality of declaratives like these is the basic modality
of the alethic modal system. Huntley (1980, 1984) says that the key feature of this
modality is that it is somehow anchored in the actual world or situation in which
a sentence is used. This does not mean that actuality must be part of the meaning
or content of these sentences, nor do we want it to be, since this would mean that
every declarative sentence is necessarily true, if true at all. Instead, to say that
alethic modality anchors things in the actual world can mean that we are forced
to evaluate the truth of declarative sentences at the actual world. It's kind of like
declaratives sentences are screaming \Believe I'm true here in the actual world!" This
sort of modality involved in declaratives is often called realis, which in turn is not
inappropriate to call actuality.
In addition to the basic actuality modality of the sentences in (78), there are
two non-basic, i.e. derived, alethic modalities: necessity and possibility. The latter
modality is had by the sentences in (79), while the former is had by those in (80).
(79) a. Jack may be going to the store.
b. Jill possibly ate her vegetables.
(80) a. Jack must be going to the store.
b. Jill necessarily ate her vegetables.
Note that the modality of a sentence can be changed either by an auxiliary verb,
as in (79-a) and (80-a), or by an adverbial, as in (79-b) and (80-b). What these
auxiliaries and adverbials do is modify { or operate on { the modality possessed by
the sentence's main verbs. In (79), the auxiliary verbs and adverbial are modifying
the actuality modality carried by the main verbs, such that the nal modality of these
sentences is one of actual possibility, or possible actuality. And in (80) the auxiliaries
and adverbials are modifying things so that the nal modality of these sentences in
one of actual necessity, or necessary actuality. Thus, on the view being entertained
here, actuality is the basic modality of declarative sentences, and (actual) possibility
and (actual) necessity are modalities derived from actuality.
Analogously, to say that the imperative mood, i.e. the mood encoded by IMP,
represents the basic deontic modality is to say that the other deontic modalities, e.g.
permissible and obligatory, represented in (81) and (82), respectively, are somehow
explained in terms of it.
(81) a. Jack may go to the store.
b. Jill is permitted to eat her vegetables.
c. It is permissible for Jerry to mix together the wet and dry ingredients.
51
(82) a. Jack must go to the store.
b. Jill is obligated to eat her vegetables.
c. It is obligatory for Jerry to mix together the wet and dry ingredients.
Exactly how the modalities in (81) and (82) are derivative of the modality of im-
peratives, however, is not a question Palmer answers, though he does say that each
of these modalities is irrealis (i.e. non-actuality), thereby distinct from the modality
of declaratives, and also suggests that these modalities have something to do with
what's desirable or preferable (Palmer 2001, pp. 29-30). This is as it should be, since
questions about what particular modality the imperative mood is are questions that
move us away from syntax and towards semantics. After all, these questions are ask-
ing about the particular semantic contribution of IMP, the carrier of the imperative
mood, to a sentence as a whole. Such questions are out of place in a chapter on the
syntax of imperatives.
So, to repeat, according to the Special Mood Theory of imperatives, imper-
atives contain a unique, often unarticulated, morpheme IMP and this morpheme
encodes some sort of sentential modality that is unique to imperatives. What's more,
given the data, it would seem that the Special Mood Theory is the most plausible
of the three we have looked at. Assuming that to be the case, the nal picture that
emerges of the syntactic structure of imperatives is this:
CP
C
0
IP
I
0
VP
V
0
(compliment) V
0
SpecVP
I
0
DO/?
Infl
SpecIP
(subject)
C
0
IMP
SpecCP
Interestingly, this theory is just another way of stating the commonsense view that
imperatives, and other clause types, are morphosyntactically marked for their own
unique sentential moods. But neither commonsense nor what we have said here tells
52
us how IMP contributes to the meaning of an imperative. So how, if at all, does IMP
semantically interact with the subject and predicate? Does it semantically correspond
to an operator? A function? Something else entirely? And what is the result of
combining all these elements, i.e. the semantic values of the subject, predicate, and
IMP? In the next chapter we turn from syntax to semantics and begin to answer
these questions. In particular, in the next chapter, I will survey and evaluate two
general theories of the semantics of imperatives. Then in Chapter 3 I will put forth
my own positive theory, one which says that imperatives are semantically equivalent
(in a certain sense) to simple-present tense declaratives, and that IMP, while essential
to the syntax of imperatives, makes no semantic contribution.
53
Chapter 2
Two Theories of Imperative
Meaning
2.1 Introduction
In this chapter we begin our investigation into the meanings of imperatives. The
plan is to look, separately, at two types of theories of the meanings of imperatives,
show how they can compositionally emerge from the syntactic results of the previous
chapter, and then critically evaluate them. To give them names, let's call the rst
theory we will look at The Property Theory and call the second The Deontic
Proposition Theory. The upshot of our discussion of these two theories will
be twofold. First, in addition to the requirement that a theory respects a suitably
strong compositionality principle (more on this below), we will uncover several other
requirements a theory of the meanings of imperatives needs to meet if it is to be
adequate. Second, we will nd that neither the Property Theory nor the Deontic
Proposition Theory meet these requirements, and so are not adequate. This will set
the stage for a third theory of the meanings of imperatives, which will be, in one way
or another, the topic of succeeding chapters.
The chapter will proceed as follows. The next section sets out some prelim-
inaries that will frame the rest of the discussion.x2 focuses on The Property
Theory of imperative meaning as found in Portner (2004). I will rst sketch the
relevant features of the theory, and show how it can be made to t the syntactic
results of Chapter 1, in particular by showing how the theory can interpret the im-
perative mood morpheme IMP. Then I will show that The Property Theory has
several drawbacks { drawbacks which push us in the direction of something more like
The Deontic Proposition Theory. That being so, The Deontic Proposi-
54
tion Theory will be the topic ofx3. The version of The Deontic Proposition
Theory I will focus on is that of Kaufmann (2012), because this is at present the
most worked out version. Like before, after saying what the key features of the view
are, I will show how it can be made to t the syntactic facts. And then, like before,
I will argue that The Deontic Proposition View, in any form, faces signicant
challenges. Lastly, inx4, I will summarize our { largely negative { ndings about the
meanings of imperatives and put them in terms of further restrictions on what an
adequate theory needs to look like. This will set the stage for succeeding chapters,
in which a third theory is introduced and defended.
2.2 Preliminaries
2.2.1 Principle of Full Interpretation
The reason we are looking at only two types of theories of the meanings of imperatives
in this chapter { and only three overall { is because the syntactic ndings of the
previous chapter, together with a Chomskian principle about the syntax-semantics
interface, rules other theories out. The principle is this:
Full Interpretation (FI) (i) Every functional and lexical element in the syn-
tactic structure of a construction corresponds to
something in 's interpretation, and (ii) The struc-
tural relations between the functional and lexical ele-
ments of correspond to semantic relations between
them in 's interpretation.
This principle is defended by Chomsky (1995). The thing to notice about FI is that
it that things are restricted in only one direction: from syntax structure to semantic
structure. What it roughly says is that if a syntactic element has an interpretation,
then that interpretation contributions to the interpretation of any larger construction
of which it is a part. But what it doesn't say is anything about the other direction:
from semantics to syntax. It doesn't say that if something is part of a construction's
interpretation, then it is the interpretation of some syntactic element that is part
of the construction. Importantly, it also doesn't say that absolutely every syntactic
element in a construction contributes to the construction's interpretation. This leaves
open the possibility that that there some syntactic elements with are semantically
inert. FI is thus rather weak, at least compared to the more robust assumption that
syntax and semantics stand in a perfect one-to-one correspondence to each other. FI
55
allows both that the semantics of a construction can \outrun" its syntax, and also
that constructions can have more syntactic complexity than semantic. And this is as
it should be if we want to avoid having to deal with issues surrounding things like
anaphoric relations established beyond the limits of a sentence, quantier domain
restrictions not grounded in syntactic structure, and whether emphasis words like
'frankly' has semantic import { among other things.
Despite leaving open the possibility of syntactic-semantic asymmetries, FI
still places a signicant restriction on semantic theorizing. Namely, it doesn't allow a
theory about the semantics of some construction or other to ignore the interpretable
part of the construction's syntax. To illustrate, consider an older view of the mean-
ings of imperatives that says imperatives are semantically equivalent to declaratives
containing the future tense auxiliary verb WILL, such that (1-a) and (1-b) have the
same interpretations/mean the same thing.
(1) a. You go to the store.
b. You will
future
go to the store.
This theory thus says all imperatives make reference to a time after the time of use.
But recall from the previous chapter that imperatives are not obligatorily marked
for future tense. So this theory, that all imperatives make future reference, is stuck
saying that imperatives have a syntactic feature (i.e. optional future tense) which
doesn't match up with a semantic one (i.e. obligatory future time reference). This
puts the theory at odds with FI, and as such, it should be ruled out as an adequate
theory.
1
Other theories of imperative meaning ruled out by the results of the previous
1
This of course does not rule out theories which say imperatives have the same meaning as
declaratives with the so-called volitional WILL, which is synonymous with SHALL:
(i) a. You go to the store.
b. You will
volition
go to the store.
c. You shall go to the store.
Since the volitional WILL is a modal auxiliary, as opposed to a tense auxiliary, this theory can
say that the modality of IMP just is volitional modality. This, however, raises a question. Do
imperatives have the same kind of meaning as other declaratives (i.e. propositions), or do sentences
like (i-b) and (i-c) have the same kind of meanings as imperatives, whatever those are? As we will
see imperatives can't have propositions as their meaning, because they don't have truth-values, so
if this theory is right, then declaratives with volitionalWILL andSHALL don't have propositions
as their meaning. They rather have whatever are the meaning of imperatives as their meaning.
Though I am sympathetic to this take on the semantics of declaratives with the volitional WILL
and SHALL, some will think it to be a reductio of the theory in question, maintaining instead
that sentences like (i-b) and (i-c) must have propositions as their meanings. What's important for
our purposes, however, is that this view doesn't actually answer the question of what the meanings
56
chapter together with FI are those that say the subject of an imperative, which is
interpretable, makes no contribution to its interpretation (Thorne 1966).
Critically, given FI, if IMP is interpretable, then any adequate theory of imper-
ative meaning must account for the semantic role of the mood morpheme IMP, since,
we argued, this is obligatory element in an imperative's syntax. More specically, the
theory must be able to say what sort of modality IMP gives an imperative's mean-
ing. According to the Property Theory of the meanings of imperatives, IMP is to be
construed as a function from the meanings of an imperative's subject and predicate
to a special kind of property, thus making that property the imperative's meaning.
According to the Deontic Proposition Theory, IMP semantically corresponds to an
operator which maps propositions onto deontic propositions, thus making the latter
propositions the meanings of imperatives. As I'll show more fully, both of these the-
ories respect { or at least are capable of respecting { FI, that is, I'll show how what
they say about the meanings of imperatives is fully constrained by their syntax, as
dened in chapter 1, which is partially why these are the theories I have decided to
focus on.
2
Nevertheless, I will also show that both theories have serious problems
arising from theoretical and pre-theoretical considerations about the semantics of
imperatives.
2.2.2 Imperatives, Truth, and Representation
One of the most recognizable things about imperative sentences is how counterintu-
itive it is to say of them that they are the kind of thing that can be true or false.
So counterintuitive that responding to a use of an imperative with \That's true" or
\That's false" is never appropriate:
A: Tell me what happened.
B: #That's true/#That's false.
This of course marks an interesting dierence between imperatives and declaratives,
since responding to a use of a declarative with \That's true" or \That's false" is
rarely unacceptable:
A: Jack is such a klutz.
B: That's true/That's false.
of imperatives are, but instead just raises the question in another form. Namely, in the form of {
what is the meaning of sentences like (i-b) and (i-c)?
2
Another reason why I have chosen to focus on just these two theory types is because they have
been extensively argued for and defended in the recent literature.
57
Given that the unacceptability of ascribing truth-values to imperatives is universal,
holding across all situations, we should expect the explanation to have something to
do with their meanings. Something about the meanings of imperatives is such that
ascribing truth or falsehood of them is always unacceptable.
Does this mean that imperatives don't and can't have truth-values? Not nec-
essarily. Notice that we have been careful to put things in terms of ascribing truth-
values, and not in terms of possessing them. To those who think that imperatives do
in fact have truth-values, the dierence between ascribing and possessing is critical,
as they will say that though it is inappropriate to ascribe truth-values of impera-
tives, these constructions nevertheless have truth-values. Lewis (1970, pp. 54-61), for
example, argued that imperatives can, without loss or change of meaning, be para-
phrased as performative declarative sentences. So (2a), for example, can in writing
or conversation and without semantic harm be paraphrased away by a declarative
like (2b), because for Lewis they have the same meaning.
(2) a. You be more careful.
b. It is commanded that you are more careful.
And since, arguably, performative declaratives have truth-values, imperatives do as
well.
3
The method of paraphrased performatives calls for the ascription of truth-
values to non-declarative sentences. The truth-value assigned is not that
of the embedded sentence (corresponding to the sentence radical [e.g.
the that-clause in (2-b)]), however, but rather that of the paraphrased
performative. If I say to you `Be late!' and you are not late, the embedded
sentence is false, but the paraphrased performative is true because I do
command that you be late. I see no problem in letting non-declaratives
have the truth-values of the performatives they paraphrase; after all, we
need not ever mention their truth values if we would rather not. (Lewis
1970, pp. 58-59)
The last sentence of this quote is important. Essentially, what Lewis is saying in there
is that though on his account imperatives have truth-values, there may be reason to
never speak of these truth-values.
The question of course is why we never ascribe truth-values of imperatives if
they in fact have them. What reason is there for this state of aairs? Although Lewis
3
Lewis acknowledges that J. L. Austin and many of his followers reject this. But he rightly, I
think, points out that neither Austin nor anyone else has yet to provide a compelling argument for
why performatives can't be true or false, thus leaving it open that they can.
58
acknowledged that we are deeply adverse to ascribing truth-values to imperatives,
he never answers this question, which allows critics to respond by saying that the
only possible explanation is that imperatives really do lack truth-values, and that the
only reason Lewis wouldn't admit this was because he was wedded to a particular
approach to semantic theorizing, one which made truth the fundamental semantic
notion. In any case, the point is that any theory of the meanings of imperatives has
to be able to explain why ascribing truth-values of imperatives is never acceptable.
This task will be more or less dicult for dierent theories. Theories that aren't
committed to imperatives possessing truth-values will obviously have it much easier
than theories that are committed to this, as we will see in the coming sections.
2.2.3 Imperatives and Action
Suppose Jill says (1-a), repeated here, to Jack, thereby making Jack the subject of
this use of (1-a), and suppose that Jack commits to doing what (1-a) says.
(1) a. You go to the store.
It is likely that the immediate result of this is that Jack turns and goes to the store.
So it would seem that Jill's use of (1-a) somehow caused Jack to go to the store.
More specically, Jill's use of (1-a) together with Jack's understanding of what (1-a)
means causes him to go to the store. What's more, there is nothing special about
this scenario; in the typically case there is a causal connection between the use of
an imperative and some action. But what explains this phenomenon? How does the
chain beginning with a use of an imperative and ending with an action work exactly?
And what does this tell us, if anything, about the meanings of imperatives?
It's hard not to think that an imperative's meaning plays an essential role
in this chain. For if the subject fails to understand the imperative's meaning, the
chain will be broken and no action will result from a use of the imperative. But
comprehension, while necessary, is not enough to complete the chain. To see why,
suppose that Jill uses (1-a) and Jack perfectly well understands (1-a)'s meaning, but
because he and Jill haven't been getting along lately he refuses to do anything Jill
says { hence he doesn't go to the store. Acceptance of an imperative therefore also
seems necessary. No acceptance of an understood imperative, no resulting action.
4
4
This is actually not quite right. Since not going to the store appears to be as much an action
as going, we can say an action does in a sense result. But the question to ask is what this action
is the result of. Is it the result of Jill's use of (1-a)? In one sense it is: we can trace a causal link
between the issuance and Jack's not going to the store. But this link will be indirect, making a
detour through another imperative Jack does accept { namely, the imperative `(Jack,) you do not
59
And adding acceptance to the chain appears to complete it, as once one accepts an
imperative the natural next step is to do what the imperative says.
Schematically, then, the causal chain between imperatives and actions looks
like this:
Use of imperative! Understanding of meaning! Acceptance! Action
We have already seen two ways that this chain can break: when there is a failure of
understanding and when there is failure of acceptance. The rst failure corresponds
to a break in the use-understanding link. The second failure corresponds to a break
in the understanding-acceptance link. There is also a third way that break can oc-
cur. This break occurs in the acceptance-action link, and is usually the result of an
agent running into physical or psychological barriers and being unable to perform
the appropriate action. Sometimes there are insurmountable obstacles to performing
some action, despite one's best intentions. Jack might fully understand and accept
Jill's use of (1-a), but if he is locked up in a jail cell, there is little chance that he
will be going to the store. In such cases the link between acceptance and action is
what is severed.
Though the foregoing schema depicts the causal chain between uses of im-
peratives and subsequent actions, it does not explain it. It does not explain why,
for example, understanding an imperative and accepting what is says leads, in the
absence of barriers, to action. All it does is portray that use, understanding, and
acceptance are together necessary and, in normal cases, sucient to bring about an
action. But this picture by itself falls short of an explanation. Why after all is it that
use, understanding, and acceptance cause action? We know that they are necessary
because the absence or failure of any one will break the chain. But why are they usu-
ally sucient as well? Since what one understands and accepts is not the imperative,
which is just a grammatical construction, but rather its meaning, the real question
to ask is this: What is it about an imperative's meaning that explains why under-
standing and accepting it suces to lead to action in cases without insurmountable
barriers?
The schema also does little to help us get a grip on each of its links. All it tells
us about imperative use, understanding, and acceptance is that these are somehow
causally linked. We know that each of these is some sort of act or process and that
do what Jill says'. And it is this imperative that is doing the critical work. For if Jack didn't accept
it then we would be hard pressed to explain why Jack did what he did (assuming, of course, that
there are no other physical or psychological barriers preventing Jack from doing what Jill said).
So what the above slogan should instead say is: \No acceptance of an understood imperative, no
directly resulting action." Or something similar.
60
they involve the meaning of an imperative, since that's what is used, understood,
and accepted. But what is it to use, understand, or accept an imperative? So in
order to answer the question about why imperatives lead to actions a theorist will
have to say something about each link in the chain. Relating this back to theories
of the meanings of imperatives, what we want, then, is a theory that is, together
with supplemental theories about issuance, understanding, and acceptance, able to
explain the causal link between imperatives and actions.
2.2.4 Core vs. Peripheral Meaning
It has become common in the literature on meaning to distinguish, in one way or
another, between two dierent parts of a clause's meaning: the core part and the
peripheral part.
5
The core of a clause's meaning is that part which enters into com-
positional relations when it is embedded in larger constructions (e.g. conjunctions
and conditionals), and which also explains its logical properties and relations (e.g.
that the clause logically entails some other clause). For example, only the core of the
meaning of the declarative embedded in the consequent of (3) contributes to (3)'s
meaning; no other part needs to be taken into account.
(3) If you don't study, you will fail the exam.
And only the core meaning of (4-b) needs to be considered if we want to explain why
it is logically entailed by (4-a).
(4) a. You study and you don't fail the exam.
b. You don't fail the exam.
This part of a clause's meaning is its core meaning.
Unlike the core part, the peripheral part of a clause's meaning (peripheral
meaning) does not contribute to the meanings of larger constructions, nor does it
serve to explain the logical properties and relations of the clause. Instead, it merely
serves to explain some of the communicational features of a clause that cannot be
explained by its core meaning alone. In the case of imperatives, peripheral meaning
is often appealed to in order to help semantically distinguish imperatives from other
types of constructions that might share the same core meanings, and also to help
explain some of the roles that imperatives play in conversation. For example, the
peripheral meanings of imperatives (and other clause types) typically have something
5
Another way to label these two parts is: theatissue part vs. the not at-issue part. See Potts
(2015).
61
to do with their discourse roles, and in particular the fact that imperatives are the
preferred vehicles of directive speech acts across languages.
6
An important distinction between core and peripheral meaning is that while
the core meaning of a clause is always present, it peripheral meaning need not be.
Put another way, the peripheral, but not core, meaning of a clause is default: though
it is expressed in many contexts, there are other contexts in which it fails to be
expressed. To capture this feature, peripheral meanings are usually characterized
either as default conventions or as default semantic presuppositions. As we will see,
in the case of imperatives, some theorists say that the peripheral meaning of an
imperatives is a default convention that requires the imperative's core meanings to
be somehow added to the conversational background if the imperative is accepted by
the conversational participants, while other theorists say that the peripheral meaning
is a default presupposition restricting the sorts of speech acts in which the imperative
can be used. Both of these options will be discussed in what follows.
2.2.5 Conversational Backgrounds
A conversational background, as Kratzer (1981) tells us, has at least two parts: a
modal base and an ordering source. Informally, the modal base is that set of propo-
sitions which are taken to be true in a conversation and against which declaratives
containing a modal auxiliary verb in matrix position are interpreted. Suppose that at
the start of their conversation, Jack and Jill take the following declaratives as true,
and only these.
(5) a. Jerry is not at home
b. Jerry is not at work
c. Jerry does not have his car
The propositions expressed by these declaratives thus form the modal base of Jack
and Jill's conversation. Now suppose Jack says (6).
(6) Jerry might be at work.
To determine whether what Jack has said is true, we need to interpret it against the
modal base. Since included in the modal base is the proposition that Jerry is not at
work, and since this rules out the possibility that Jerry is at work, the proposition
that semantically corresponds to (6), and thus what Jack has said, is false in this
conversation.
6
See Searle (1969).
62
A bit more formally, if we treat propositions as sets of possible worlds (some-
thing we will call into question later), then we can model the modal base as a set of
sets of possible worlds. Moreover, since the propositions taken to be true will vary
from one world to another, we can say that the modal base is a function from worlds
to sets of propositions.
Modal Base A function f such that if w is a world, then f(w) =fp: p is
taken to be true in wg
Next we can use the modal base to determine a set of accessible worlds, i.e. those
worlds which are relevant to the interpretation of declaratives containing deontic
modal verbs.
Accessible Worlds If w is a world, then\f(w) contains those worlds which are
accessible from w.
Returning to Jack's use of (6), what we do now is interpret this against the set of
worlds that are accessible to the worlds that are supposed to be true in context,
namely,\f(w
J&J
). What Jack says will be true if and only if there is at least one
world in\f(w
J&J
) in which Jerry is at work. But since Jack and Jill take as true the
proposition that Jerry is not at work (= the complement of the set of possible worlds
in which Jerry is at work), there will be no worlds in\f(w
J&J
) in which Jerry is at
work. Thus, what Jack has said is false. But suppose Jack had instead said (7).
(7) Jerry might have his bike.
This will be true if and only if there is some world in\f(w
J&J
) in which Jerry has
his bike. And since such a world is compatible with what Jack and Jill take to be
true (i.e.9w(Jerry has his bike in w^ w2\f(w
J&J
)), what Jack has said this time
is true. Similarly for other modal verbs: they are all to be interpreted against those
worlds which are accessible from the world of use, as dened in terms of the modal
base. For instance, suppose we replace the MIGHT s in (6) and (7) with MUSTs.
(8) a. Jerry must be at work.
b. Jerry must have his bike.
For any world w, (8-a) will be true in w if and only if in all the worlds in\f(w)
Jerry is at work, while (8-b) will be true in w if and only if in all the worlds in\f(w)
Jerry has his bike.
As for the second part of a conversational background { the ordering source {
this is also a set of propositions against which declaratives containing modal auxiliary
63
verbs are interpreted. But unlike the modal base, the propositions in the ordering
source are neither taken as true in the conversation, nor do they play a role in de-
termining which worlds are accessible. These are instead propositions that someone
wants, hopes, prefers, or demands be made true. For example, suppose the govern-
ment of the country in which Jack and Jill reside demands all its citizens to make
the following declaratives true, and only these.
(9) a. Nobody murders.
b. Nobody steals.
c. Everybody respects the property of others.
Taken together, these propositions form an ordering source. Like with modal bases,
ordering sources can vary across worlds, so it's best to similarly treat the ordering
source as a function from worlds to sets of sets of worlds (i.e. propositions).
Ordering Source A function g such that if w is a world, then g(w) =fp: p is
desired/hoped/ demanded/required/etc. to be true in wg
The role of the ordering source is to impose a ranking on the worlds in\f(w), for
any w. For example, ifg(w
J&J
) =fThat nobody murders, That nobody steals, That
everybody respects the property of othersg, then those worlds in\f(w
J&J
) in which
it's true that nobody murders, nobody steals, and everybody respects the property
of other will be ranked higher than those in which only two of these propositions are
true, and these latter worlds will be ranked higher than worlds in which only one is
true, which in turn will be ranked higher than worlds in which all three propositions
are false. (Of course, since there will be multiple worlds in which all three propositions
are true, multiple worlds in which just two are, and so on, the ordering imposed by
the ordering source will be a partial one).
The semantic role of the ordering source can be seen in the interpretation of
declaratives containing deontic modal verbs.
(10) a. Jack ought to call his mother.
b. Jill may have desert.
According to a simplistic Kratzerian analysis, for any world w, (10-a) is true in w
if and only if Jack calls his mother in all of the highest ranked worlds in\f(w) as
determined byg(w), and (10-b) is true in w if and only if Jill has desert in at least one
of the highest ranked in world\f(w) as determined byg(w). Thus what the ordering
source does semantically is allow us to interpret declaratives that are concerned not
with the background facts of a conversation, but rather with the background norms,
such as those in (10).
64
2.3 The Property Theory
2.3.1 The Basics
According to the Property Theory, the core meanings of imperatives are properties
of a certain sort.
7
On a nave version of this theory, an imperative's core meaning
is fully determined by its predicate VP, and the meaning of a predicate VP is just
a property. If we represent properties in terms of lambda operators, then what the
na ve version of the Property Theory essentially says is this:
Na ve Property Theory If is an imperative, F is the predicate of , c is
a context of utterance (i.e. a tuple containing at
least a time, a possible world, a speaker, and the
addressees), and w is a possible world
8
, then:
(i) JFK
c;g
= wx : x is F in w, and
(ii) JK
c;g
= JFK
c;g
One immediate problem with this is it violates FI, because it accords no semantic
role to an imperative's subject. Portner (2004) recognizes this problem and amends
the Property Theory as follows:
Portner's Property Theory
If is an imperative, F is the predicate of , is
the subject of , c is a context of utterance, and
w is a possible world, then:
(i) JK
c;g
= the addressee of c,
(ii) JFK
c;g
= wx : x is F at t in w, and
(iii) JK
c;g
= wx : x = JK
c;g
. x is F in w
7
In what follows I will be leaving ambiguous what is meant by `a property'. From the perspective
of formal semantics, a property is just a certain kind of intension. Specically, on this understanding
a property is a function from a world to a set of individuals. Call this the Formal Notion of
Properties. In contrast to this there is the Robust Notion of Properties. The Robust Notion is at
home in metaphysics, and according to it a property is a metaphysical entity that can be instantiated
or possessed by objects and persons. Since robust properties are had by individuals, given a few
uncontroversial assumptions about how functions from worlds to individuals can be dened, the
Robust Notion implies the Formal Notion. If P is a robust property P, then we can use P to dene a
function from worlds to individuals that have P in the worlds. It's because of this that equivocating
between these two notions is fairly harmless, which is why I will be doing so for the time being.
8
We should also throw times into the mix, since we learned in Chapter 1 that imperatives carry
tense, but for convenience I won't. Nothing rests on this choice.
65
On this version of the Property Theory, the way the subject of an imperative con-
tributes to its core meaning is by restricting the domain of individuals having the
property denoted by an imperative's predicate to just those individuals denoted by
the imperative's subject, which if Portner is right is the addressee or addressees of a
context, as codied in clause (i) of the theory.
Why does the subject of an imperative denote the addressees? At rst glance,
this seems unassailable, but it does face some diculties. For one thing, we already
saw in chapter 1 (xx2.1.-2-2.) that there are imperatives in English with subjects that
don't denote an addressee:
(11) a. You and John take the 11am train. (Said by anyone to anyone)
b. Your men
ank the enemy from the east. (Said by a colonel to a lieu-
tenant)
c. Those students go home at noon. (Said by a principle to a teacher)
d. The dogs don't sit on the couch. (Said by a parent to a child)
For another, there are other languages in which it is clear that the subjects of imper-
atives needn't denote an addressee. Irish and Finnish are two such languages. The
imperatives of Irish are particularly striking because there appear to be no grammat-
ical restrictions on who or what the subject can denote. Given this, I will maintain
that the subjects of imperatives should be interpreted in exactly the same way as the
subjects of other clause types are interpreted, which means that what's denoted by
the subjects of imperatives will not be xed or restricted, at least not by anything
in their syntax or semantics.
910
In order to capture this in the theory, all we will do is eliminate the rst clause
of Portner's original Property Theory, thus giving us:
General Property Theory If is an imperative, F is the predicate of , is
the subject of , c is a context of utterance, w is
a possible world, and t is a time, then:
(i) JK
c;g
= ???
(ii) JFK
c;g
= wx : x is F in w, and
(iii) JK
c;g
= wx : x = JK
c;g
. x is F in w
9
If is a 2nd person singular pronoun (e.g. `you'), then JK
c;g
= the addressee of c. If is a
proper name for individual X, then JK
c;g
= X. If is a demonstrative, then JK
c;g
= g(), where
g is an assignment function that is somehow metasemantically selected (e.g. by the intentions of
the speaker); see Kaplan (1977, 1989). And so on.
10
For a dissenting view see pakportnerandzanuttini2012, who argue that imperative subjects
are obligatorily 2nd person (though there are also other kinds of clauses that are very much like
imperatives but which can have non-2nd person subjects).
66
In what follows, this is what I will be referring to as The Property Theory.
Since the way we arrived at this statement of the theory was by generalizing away
from a particular detail of Portner's version of it, Portner's version and this one
are fully compatible. So, strictly speaking, we have not (yet) ruled out Portner's
theory; rather, we have just moved a step away from it in order to avoid some of the
diculties that attend Portner's theory.
11
That being the case, I will follow Portner
in lling in the rest of the details of The Property Theory, and only deviate
in those instances in which what he says relies on the claim that the subject of an
imperative must be an addressee.
To summarize, on The Property Theory the core meanings of imperatives
are properties the domain of which are restricted to the imperative's subject. One
consequence of this is that imperatives will share the same core meanings as VPs
in which a re
exive pronoun, which is essentially anaphoric on the subject of any
sentence (seexx2.1.-2.2. of Chapter 1), occurs, such as the VPs in (12).
(12) a. You
1
will go to the store yourself
1
.
b. You
2
might eat your vegetables yourself
2
.
c. Will you
3
mix together the ingredients yourself
3
?
The most plausible core meanings of the VPs in these sentences will be such that they
are semantically equivalent to the imperatives in (13), if The Property Theory
is right.
(13) a. Go to the store.
b. Eat your vegetables.
c. Mix together the ingredients.
11
But what if we do want to reject Portner's claim that the subjects of imperatives always
denote the addressees? What will we say to the question of why it is often the case that imperative
subjects often denote addressees? I supplied the makings of a syntactic answer to this question in
the previous chapter,x3.3. Another possibility is to answer in pragmatic or perhaps metasemantic
terms. It might be maintained that because imperatives are the preferred vehicles of directive speech
acts, and because the whole point of a directives is to direct someone to do something, it would be
fruitless to issue a directive to someone who cannot receive it. And who better to receive a directive
than an addressee? Moreover, in those case where the subject does not denote the addressee, it's
often the case that the addressee is able to \pass the message along" to whoever or whatever is
denoted by the subject. This is evident in the examples in (11). In these examples, the addressee
is someone who is presumed to know and be able to communicate with the subject and so is in a
position to ensure that the subject receives the directive. In short, there are good ways of explaining
why the subjects of imperatives are more often than not addressees, besides saying the semantics
requires that the subject to be the addressee.
67
But clearly there is something dierent about the meanings of the VPs in (12) and
the imperatives in (13). So what distinguishes them?
12
Enter the second part of the meanings of imperatives: their peripheral mean-
ing. According to Portner (2004, 2007), Charlow (2014) and others, the peripheral
meaning of an imperative has something to do with the imperative's eect on a con-
versational background. In particular, Portner (2004) seeks to amend the traditional
picture of conversational backgrounds (as presented inx1.5.) in one crucial way: not
only can the modal base of a conversation C (i.e.\f(w
C
)) be ordered by the ordering
source of C (i.e.g(w
C
)), but it can also be ordered by something called a To-Do List,
where a To-Do List is to be thought of as a set of properties, not propositions, that
an individual is committed to possessing { or required to possess { in a given world.
Roughly:
To-Do List A functionT such that if w is a world and x is an individual,
thenT (hx;wi) =fP(roperty): P is required to be true of x in
wg
Suppose, for example, Jill has the following To-Do List in a world w:
Training to run 5k
Eating plenty of protein and complex carbs
Drinking plenty of water
So,T (hJill;wi) =fTraining to run 5k, Eating plenty of protein and complex carbs,
Drinking plenty of waterg. According to Portner this list contributes to interpreting
declaratives that have Jill as subject and a deontic auxiliary verb in matrix in the
same way that the original ordering source contributed. Take the declaratives in (14).
(14) a. Jill ought to run 5k.
b. Jill may eat nothing but simple carbs.
Given the right \
avor" of `ought' and `may', for any world w, (14-a) is true in
w i Jill runs 5k in all of the highest ranked worlds in\f(w) as determined by
T (hJill;wi), and (14-b) is true in w i Jill eats nothing but simple carbs in at least
one of the highest ranked in world\f(w) as determined byT (hJill;wi). Thus, on this
12
Note that claiming that imperatives have tense (of some sort) won't help here. It will just
push the problem back. For then the core meanings of imperatives will be what we might call
time-indexed properties, which are properties that can only be possessed at a certain time (e.g. the
property of being hungry at 2pm), and time-indexed properties are the core meanings of TPs. So
the question now becomes: what distinguishes an imperative from a semantically equivalent TP?
68
interpretation, (14-a) is true, because all of the best worlds according toT (hJill;wi)
are ones in which Jill has the property of running 5k, but (14-b) is false, because none
of the best worlds according to T (hJill;wi) are ones in which Jill has the property
of eating nothing but simple carbs.
Returning to The Property Theory, one way to get properties onto some-
one's To-Do List is by way of imperatives. This is because the conventional discourse
function of imperatives (perhaps encoded in IMP) is to update To-Do lists. When
someone uses an imperative not only do they communicate the property that is its
core meaning, but they also attempt to get the subject of the imperative to add that
property to their To-Do list. A use of an imperative is successful only if the subject
so updates (Portner 2016). What's more, this discourse function of imperatives is
a conventional { or in Portner's terms, canonical { one: convention dictates that
any use of an imperative is an attempt to update the To-Do list of the imperative's
subject. To put this in our terminology, this dynamic component of imperatives is
their peripheral meaning. Importantly, this conventional association with To-Do list
updating is what sets imperatives apart from VPs containing re
exive pronouns.
Since VPs are not conventionally associated with a particular discourse functions
(they are just VPs after all, not clauses), they cannot have the peripheral meanings
imperatives have, thus marking the dierence between them.
2.3.2 Interpreting Imperatives
How do we get a domain-restricted property out of an imperative? Since we know
that imperatives are essentially composed of at least a subject NP, a predicate VP,
and IMP, this question reduces to asking how we need to interpret those syntactic
elements so that for any imperative , has this interpretation:
JK
c;g
= wx : x = JK
c;g
. x is F in w
And since we have already established that the subject NPs and predicate VPs of
imperatives get their usual interpretations, it all comes down to how we interpret
IMP. So what sort of interpretation does IMP need in order for imperatives to have
the foregoing interpretations?
Portner and Zanuttini (2003) give us the beginnings of an answer to this
question. They start by saying IMP is semantically an operator. They then go on:
The element IMP has modal-like syntactic properties, binding a world
and individual variable. However, we do not want IMP to have an or-
dinary modal's meaning, because in that case the form would have a
69
declarative-type meaning. Actually, we do not attribute any meaning to
IMP. The only semantic eect of the operator is via its binding of the
world and individual variables (and it is thus like relative clause operators
like who, such that, and that in the system of Heim & Kratzer). Thus,
the ultimate interpretation of the imperative is a property. (Perhaps the
semantic vacuousness of IMP can help explain why it is typically realized
as just a feature or phonetically null head.)
(15) JIMP [wx [w x leave]]K = wx . x=H[earer]. x leaves in w
13
This is our target meaning.
(I have inserted IMP where appropriate, changed their (9) to (14), and removed their
superscripts, as they are not important here.) Portner and Zanuttini's idea is thus
that IMP semantically corresponds to an operator, but it is a rather insubstantial and
odd operator because it doesn't do anything except bind the world and individual
variables that are part of the interpretation of an imperative's predicate. This is of
course ne as far as it goes; it doesn't cause a violation of FI. But there is something
else about Portner account that makes it unsatisfactory. It (once again) doesn't take
account of an imperative's subject. The only elements represented in (15) are IMP
and the predicate. As such, this rst pass at computing the meanings of imperatives
ultimately does violate FI, and so is inadequate.
A simple way to x this problem, suggested by Zanuttini et al. (2012), is to
have IMP denote an operator that abstracts, binds, and then restricts. First the
subject and predicate of an imperative are combined in the usual way. Take (16), for
example.
(16) You go to bed.
The subject of (16) is `You', while the predicate is `go to bed'. These are combined
as follows:
Jgo to bedK
c;g
(JYouK
c;g
) = w. Addressee
c
goes to the store in w
Obviously if we stop here then (16) will have the same interpretation as a simple-
present tense declarative (an option I'll return to in the next chapter). But The
Property Theory doesn't stop here. Instead, it says that IMP semantically func-
tion to rst abstract out the subject of an imperative and bind the resulting variable,
13
This is meant to be the core meaning of the one-word imperative `Leave'. The `H' is meant to
denote the hearer or addressee of a context.
70
leaving us with the run-of-the-mill property denoting by the imperative's predicate.
Then IMP functions to introduce a restriction on the domain of this property, namely,
it restricts the property's domain to just whatever the subject denotes (if anything),
which gives us the desired interpretation of (16):
J(16)K
c;g
= JIMPK
c;g
(Jgo to bedK
c;g
(JYouK
c;g
)) = wx : x = Addressee
c
.
x goes to bed in w
Importantly, this means that IMP is not just a functional element of an imperative.
It is also syncategorematic. This of course is not a problem, as other functional words
appear to be similarly syncategorematic (Heim and Kratzer 1998).
Nevertheless, if one prefers IMP to be categorematic, then an alternative ap-
proach is to treat IMP not as a modal-like operator morpheme, but rather as an
adverbial-like morpheme that functions to impose a restriction on a predicate's do-
main, with the following interpretation:
JIMPK
c;g
= yFwx : x = y. x is F in w
To see how this proposal works, consider this imperative:
(17) Go to bed.
Deriving the meaning of an imperative from this is straightforward. First we stipulate
that the subject of (17) is ?
2nd;singular
, which we said (in Chapter 1) denotes the
addressee of the context. Combining this with JIMPK give us:
JIMPK
c;g
(J?
2nd;singular
K
c;g
) = Fwx : x = the addressee
c
. x F s in w
Next, the imperative's predicate is `go to bed', which denotes the property of going
to bed:
Jgo to bedK
c;g
= wx : x goes to bed in w
So:
J(17) K
c;g
=
JIMPK
c;g
(J?
2nd;singular
K
c;g
)(Jgo to bedK
c;g
)
= wx : x = the addressee
c
. x goes to bed in w
71
As desired. Thus, according to this analysis JIMPK is in eect looking for two argu-
ments. Moreover, this analysis says that IMP is of the very unusual typehe;hhe;ti;he;tiii,
which succeeds in lexically dierentiating it from virtually all other morphemes.
14
Lastly, this analysis is also able to bring The Property Theory in line with FI,
since it takes account of each and all of the syntactic elements of an imperative, and
shows how they are related in the interpretation of an imperative.
15
Whether one goes with the syncategorematic or the categorematic approach
to IMP is not all that important here. Both proposals amount to saying the same
thing: the semantic function of IMP is ensure that imperatives express properties.
Moreover, the point is simply that IMP has to have this function if the The Prop-
erty Theory is to work.
2.3.3 Evaluation
Properties, Truth, and Action
The Property Theory has no problems explaining why it's always inappropriate
to ascribe truth-values to imperatives. Because their meanings are properties and
properties don't have truth-values, imperatives can't have truth-values. And since
it's only appropriate to ascribe truth or falsity of something else if that thing can
be true or false, it follows that it's never appropriate to ascribe truth-values to
imperatives.
As for the action-producing nature of imperatives, Property Theorists can ap-
peal to To-Do lists to explain this. The reason that an action results more often than
not when an imperative is understood and accepted, according to The Property
Theory, is because to understand an imperative is to grasp the property that is its
meaning, and to accept it is to place that property on one's To-Do list. From there,
an action results because what a To-Do list represents as far as the psychology of
a language user goes is the commitments { or as Portner calls them, the require-
ments { of the person whose list it is, and commitments are the right sort of thing
to move people to action. So if I have the property of going to the store myself on
my personal To-Do lists, then I am committed to coming to possess this property
and as such am motivated, to some degree, to doing what it takes to possess it. This
places a lot of weight on To-Do lists and sets them apart from other elements of the
14
For an overview of semantic type theory, see Dowty (1979).
15
I do not know how happy Portner and other Property Theorists will be with this way of inter-
preting imperatives. But given the need to be in accord with FI and the need to be able to lexically
dierentiate IMP, as Portner himself tries to do in saying that IMP semantically corresponds to a
vacuous operator, it's not at all clear what other options are available.
72
conversational background (i.e. the modal base and the ordering source), because
now they cannot just be formal devices used to interpret and evaluate declaratives
containing modal auxiliary verbs. They must now also be part of or model the basic
psychological states of language users. This (or something very much like it) is not
only what Portner and others have in mind when they talk about To-Do lists, but
it is also for the best since it grounds To-Do lists in our psychologies and thereby
makes them much less mysterious.
16
The Status of To-Do Lists
There are, however, problems with the status of To-Do lists themselves. One prob-
lem is that using them to order the worlds in a modal base in order to interpret
declaratives with deontic auxiliary verbs seems to lead to false predictions. Consider:
(18) a. Though you're committed to not exercising, you must exercise
b. Though not exercising is on your To-Do list, you must exercise.
c. Though you intend to not exercise, you must exercise.
Holding xed the meaning of MUST in each of these sentences, and assuming it
doesn't (or at least needn't) scope out over the entire complex sentence, on Portner's
view (18-a) and (18-b) are roughly semantically equivalent: they have the same inter-
pretations. And if one thinks that you can't be committed to doing something unless
you intend to do it, then (18-a) analytically entails (18-c), but probably not vice
versa. So if (18-a) and (18-b) are true, then so is (18-c). Moreover, it seems perfectly
possible for (18-a)-(18-c) to be true: cases can be constructed in which someone must
do something even though they haven't committed to or aren't planning on doing it.
Now here's the thing to notice about Portner's account of To-Do lists: if the
MUSTs in (18-a)-(18-c) are interpreted in terms of an individual's To-Do list, then
(18-a)-(18-c) are necessarily false. To see why take (18-a). If the rst conjunct of
this sentence { `you're not committed to exercising' { is true, then the property of
exercising is not on the addressee's To-Do List, nor is it entailed by any thing that
is on it. Therefore, the ordering that this To-Do list imposes on the modal base will
be such that not all of the highest ranked worlds are worlds in which the addressee
exercises; some will be worlds in which the addressee does not exercise and perhaps
instead sits on the couch eating potato chips. But this means that the second conjunct
of (18-a) is false. For on Portner's analysis of declaratives containing MUST, when
16
I will return to discuss To-Do Lists further in chapter 6, where I will try to say more about
which psychological states they represent and about how the psychological reality of To-Do Lists
provides an explanation of the link between imperatives and action.
73
interpreted in terms of To-Do lists, `you must exercise' is true in a world w i all of
worlds in the modal base of w that are ranked highest by the addressee's To-Do list
are worlds in which the addressee exercises. Given these truth-conditions, `you must
exercise' is false, if `you're not committed to exercising' is true. And the reverse holds
as well: if `you must exercise' is true, then `you're not committed to exercising' is
false. Therefore, because (18-a) is a conjunction, and is true i both of its conjuncts
are true, it follows that (18-a) cannot be true { the meanings of the conjuncts prevent
it. But this is absurd, or at least it calls out for a substantive defense.
One might respond by saying that To-Do lists are the ordering sources only
for certain \
avors" of MUST, and though it's right that for these
avors of MUST
(18-a)-(18-b) cannot be true, they can be for those
avors that don't rely on To-Do
lists as ordering sources. Fair enough. But the question to ask now is: What
avor(s)
of MUST are the ones for which To-Do lists play a semantic role? At a glance it
seems like very few and very uninteresting ones. For surely they won't play a role in
any broadly norm-based
avors of MUST (e.g. moral-
avored and legal-
avored), for
norms are not properties on To-Do lists. It thus seems that if we accept this response
to the problem, then To-Do lists will only have a semantic role to play in relatively
few and uninteresting { not to mention hard to recognize! {
avors of MUST (and
by extension, other deontic verbs as well). Surely Property Theorists don't want to
limit the semantic roles of To-Do lists so severely.
An alternative response would be to backtrack a bit and instead of saying
that To-Do lists are to be used by themselves to interpret declaratives containing
deontic auxiliary verbs, we might change this so that To-Do lists are rather added to
or a part of the normal ordering source in a conversation, and it is in terms of this
combined structure that declaratives with deontic auxiliary verbs are interpreted.
This is actually Portner's later position on the matter:
[I]n terms of the theory of modality of Kratzer (1981) we can say that
To-Do Lists help determine the ordering source for the interpretation of
a class of modals which includes deontic modals. For now, we may think
of this relationship as follows: at any point in a conversation, there is a
contextually salient deontic conversational background O such that, for
each propertyP on the addressee's To-Do List and each w compatible
with the Common Ground, P() is in O(w). Moreover, O is very likely
to be used in the interpretation of any deontic modal in the local stretch
of discourse. (Portner 2007, p. 353)
By \deontic conversational background" Portner just means the usual Kratzerian
ordering source. This appears to x the problem. For now sentences like `you must
74
exercise' will be interpreted against an individual's personal To-Do list in addition
to whatever else we can throw into the ordering source, and this will make it possible
for the sentences in (18) to be true on Portner's analysis.
The foregoing modication to the semantic roles of To-Do Lists does get
around the problem, but it comes at a cost: it makes specialized To-Do Lists ap-
pear completely unnecessary. Look at it like this. If what a To-Do List contributes to
the ordering source is not the properties that are on it, but rather the propositions
that result from supplying the properties on a To-Do List with the List's owner as an
argument, as Portner says in the above quote, then why not just make To-Do Lists
lists of propositions, not properties, thereby treating them in the exact same way we
treat the other things that contribute to the ordering source? Moreover, if a To-Do
List is meant to represent the personal commitments of the individual whose List it
is, why can't these commitments be just as well represented by lists of propositions?
Treating To-Do Lists as lists of properties, and thereby dierently from other things
that contribute to the ordering source, just seems pointless given that the roles they
play can be played just as well { if not better { by lists of propositions.
17
To be clear,
this is not an objection to The Property Theory, especially since To-Do Lists
being lists of propositions is consistent with and in no way undermines the main
claims of the theory, which are that imperatives have properties as core meanings
and that the peripheral meaning of an imperative is a convention saying that a use
of it is an attempt to update someone's To-Do List. Nevertheless, it does mean that
To-Do Lists are not exclusive to The Property Theory. In the next chapter I will
incorporate the idea that To-Do Lists are lists of propositions into my own theory of
imperative meaning.
Properties, Mood, and the Prevalence of Imperatives
To-Do Lists aside, the next question you may be asking yourself about The Prop-
erty Theory is just what it takes the mood or modality of an imperative sentence
to be. One option, seemingly adopted by Portner (2011), is to say that sentential
mood is what's contributed by the peripheral, not core, meanings of imperatives,
namely, the convention which makes any use of an imperative an attempt to update
someone's To-Do List. But this option is problematic because not all uses of impera-
tives are attempts to update someone's To-Do List. In particular, uses of embedded
atomic imperatives like those in (19) do not amount to attempts to get someone't to
alter their To-Do List.
17
I am not the rst to notice this; see Charlow (2011). But I may be the rst to press it as a
problem.
75
(19) a. If you are hungry, eat an apple.
b. Take out the trash or clean the bathroom.
c. Have a beer only if your doctor says it's okay.
A use of (19-b) is not an attempt to get one to take out the trash, nor an attempt
to get one to clean the bathroom. Similarly for the atomic imperatives in (19-a)
and (19-c). (Obvious uses of the complex sentences themselves might amount to
attempts to update someone's To-Do List, but that's beside the point.) So on this
way of understanding sentential mood, imperatives lose their mood when embedded.
But mood, whatever it is, is essential to the life of a sentence. A sentence that doesn't
have the imperative mood is not an imperative.
Perhaps a better option is to say that properties themselves (in some sense)
possess mood or modality, or that they constitute a sentential mood. The fact that
imperatives have properties as their core meanings is at the end of the day what
semantically distinguishes them from other clause-types. This of course would require
us to say that VPs also have sentential moods, but that might not be too high a
price to pay. We could after all just drop the `sentential' adjective and replace it with
something more like `linguistic' or even `phrasal', since sentences are themselves just
a type of phrase. And it's certainly not counterintuitive to say that VPs have phrasal
mood { at least not on its face.
There is something strange about saying properties possess or constitute a
mood however. Assume The Property Theory is right. What then is the point
of there being a unique type of construction called imperative sentences? Can't all the
work done by this supposed type of clausal phrase equally well be done by VPs with
non-argument re
exive pronouns? After all, not only can these VPs serve the exact
same logical and compositional roles as imperatives. But there is no reason why
they can't also serve the same pragmatic and more broadly communicative roles.
Recall that on The Property Theory really the only thing, vis--vis meaning,
separating imperatives and VPs with non-argument re
exive pronouns is the fact
that imperatives are conventionally marked for use in attempts to update someone's
To-Do List. But given that To-Do Lists are just sets of properties, why can't it be
the case that there is a convention to the eect that using these VPs outside of any
sentential context amounts to an attempt to update a To-Do List? The point is that
if The Property Theory is right, then it would seem that imperatives { and more
specically IMP { are entirely unnecessary: all the work they do can be just as well
done by the relevant sort of VPs. This makes it a mystery as to why virtually every
language has a distinct class of imperative clausal constructions even though they
already have { or easily could have { VPs with non-argument re
exive pronouns.
76
Property Theorists might want to chalk this up as just another case of syn-
onymy, a phenomenon that also appears to be universal across languages. But that
won't work. First, though imperatives are universal across languages, and so is the
phenomenon of synonymy, there is no evidence to support the claim that there are
systematic universal synonyms of the sort that would need to exist if imperatives and
VPs with non-argument re
exive pronouns are to mean the same things across lan-
guages. Second, it's incredibly doubtful that competent native speakers of languages
would recognize or acknowledge imperatives and these VPs as synonyms, unless they
are primed for it beforehand of course, as a Property Theorist might be. This makes
the sort of synonymy going on between imperatives and VPs very dierent from
the sort between, say, the noun `bachelor' and the phrase `unmarried adult male',
which more than a few competent speakers of English will readily acknowledge as
synonymous. And it's not like we can say that `the core meanings of imperatives'
and `the core meanings of VPs with non-argument re
exive pronouns' are like `water'
and `H2O', in that the their shared meanings is something we discover empirically.
We can't go out and discover, through observation and experiment, that the core
meanings of imperatives = the core meanings of VPs, like we did in discovering that
water = H2O. Meanings just aren't that sort of thing.
These considerations give us reason not to say that properties constitute a
mood. So what then is mood? I'll return to this question in the next chapter, where
I'll argue that instead of governing use, sentential moods governing acceptance.
Properties and Mixed Complex Clauses
The nal, and most pressing, problem is that at present The Property Theory
has not provided an adequate analysis of mixed conjunctions, disjunctions, condition-
als, and biconditionals (Starr ming). Consider, for example, the mixed constructions
in (20).
(20) a. I'll watch the stairs and you cover the exit.
b. Eat your vegetables or your sister will eat them for you.
c. If the grass is green, then don't put fertilizer on it.
d. If the grass is brown, then I'll water it and you put fertilizer on it.
e. Go to the store only if you are hungry.
f. Go to the store if and only if you are hungry.
If The Property Theory is correct, then the connectives in these sentences are
semantically coordinating two very dierent things { properties on the one hand and
propositions on the other { and at present there is no semantic analysis of connectives
77
that allows for this. Sure, there are analyses of sentence conjunction and analyses
of predicate conjunction, but to date there are no analyses of sentence-predicate
conjunction, and the reason there is not is because conjunction is thought to be an
operation on two things of the same semantic type.
Relatedly, The Property Theory also has a problem with negation. On
the standard analysis, negation is an operation on full-blown contents, and not on
properties. For example, to negate a declarative sentence is to say that its truth-
conditions do not hold, that it is not true. But what does it means to negate a prop-
erty? I'm not even sure I can make sense of this idea, but perhaps one option is that
to negate a property is to express a property with the complement extension. But if
that or something like it is right, then the word `not' is ambiguous: in declaratives it
expresses an operation on truth-conditions, while in imperatives it expresses an op-
eration on property extensions. But this doesn't seem to be the case; intuitively `not'
expresses the same sort of operator in both declaratives and imperatives, namely, an
operator on full-blown contents conditions, whether these be truth-conditional (i.e.
propositions) or some-other-sort-of-conditional contents. (More on this in the next
chapter.)
In eect, the overarching challenge to the Property Theorist is to provide an
adequate semantic analysis of all the sentential connectives (negation included) that
allows them to operate on things of mixed semantic types while at the same time
not positing any deep ambiguity. Until this challenge is met, it is reasonable to be
skeptical of The Property Theory.
18
2.4 The Deontic Proposition Theory
2.4.1 The Basics
The problems with the Property Theory seems to push us towards a theory that
(i) avoids things like To-Do Lists, (ii) doesn't make the universal prevalence of im-
perative clauses and their associated sentential mood a mystery, and (iii) is able to
account for mixed complex sentences as well as the role of negation in imperatives. As
it happens, The Deontic Proposition Theory of the meanings of imperatives
meets each of these requirements.
According to The Deontic Proposition Theory, imperatives have de-
ontic propositions as their core meanings. A deontic proposition is to be understood
18
I return to the problem of mixed compound sentences in Chapter 4. Importantly, the solution
I defend is not open to the Property Theorist precisely because she denies that imperatives have
satisfaction conditions.
78
as any proposition that is the output of a deontic propositional operator. A deontic
operator is one having what we above called a particular deontic
avor, that is, it
expresses not how the world is or might be, but rather how it is permitted, obligated,
required, desired, etc., to be. So a deontic proposition, being the output of a deontic
operator, represents some possible state of the world as being permitted, obligated,
required, desired, or whatever. Suppose p is a proposition and O is a deontic oper-
ator. Then O(p) is a deontic proposition. And if O is of the right
avor, then O(p)
may be the meaning of an imperative. (More on this in a moment.)
Proponents of The Deontic Proposition Theory motivate it by appeal-
ing to a very simple argument. The argument begins by re
ecting on pairs of sentences
like the pair in (21).
(21) a. (You) run 5k today.
b. You should run 5k today.
The rst sentence of this pair is an imperative, while the second is a declarative sen-
tence containing the modal verb SHOULD. Call the latter a SHOULD-declarative.
One of the most striking things about pairs like this is any situation in which the im-
perative can be used is a situation in which the corresponding SHOULD-declarative
can equally well be used - and vice versa. Suppose Jack asks Jill for her advice about
whether to run 5k today. One way for Jill to respond is with a use of (21-a). But
another, equally good option is for her to use (21-b). The only catch here is the
use of the SHOULD-declarative (21-b) must be a performative one. If it is instead
a descriptive use, that is, a use in which something is described as being some way
or other, then it would be an inappropriate replacement for an imperative, and an
imperative would be an inappropriate replacement for it.
Let us say two types of sentence have the same discourse roles if they can
serve the same purposes in a conversation. Thus, imperatives and performatives uses
of SHOULD-declaratives appear to have the same discourse roles. This observation
is the rst part of the argument for The Deontic Proposition Theory. The
rest goes like this:
P1 Imperatives and performatives uses of SHOULD-declaratives have the same
discourse roles.
P2 In order to explain P1, it must be that imperatives and performative uses of
SHOULD-declaratives have the same type of meanings.
P3 The most plausible accounts of the meanings of SHOULD-declaratives main-
tain that these constructions have deontic propositions of some particular
79
type as their meanings.
C Therefore, imperatives must have deontic propositions of the same type as
their meanings.
I will have more to say about this argument when I turn to evaluate the theory. For
now the thing to notice is P2. As we will see later, the key assumption behind this
premise { that sameness of discourse roles involves sameness of meaning { causes
problems for The Deontic Proposition Theory.
There are many types of deontic modality to choose from (e.g. being permit-
ted, obligatory, desirable, etc.). So which modality are imperatives? For our purposes,
the answer to this question doesn't matter. What matters is that imperatives possess
some particular deontic modality. Which particular one is another question.
19
So, to
keep things moving forward, let's just assume imperatives possess weak deontic ne-
cessity, which is roughly the modality of the phrase `It is required that'. Moreover,
let's use `O
!
' to represent the operator corresponding to the deontic modality of im-
peratives. So O
!
(p) is the proposition that p is required, and which can serve as the
meaning of an imperative.
The proposition O
!
operates on is the one resulting from combining the im-
perative's subject and predicate, hence it is the proposition that the imperative's
subject instantiates the imperative's predicate. Take (22) for example.
(22) Brush your teeth.
Because of the presence of an anaphoric instance of the pronoun `your' in this sen-
tence, the subject is?
2nd;singular
, which denotes the addressee. And the predicate of
(22), i.e. the VP `brush your teeth', denotes the property brushes their teeth.
J?
2nd;singular
K
c;g
= addressee
c
Jbrush your teethK
c;g
= wx. x brushes their teeth in w
Putting these two pieces together gives us: the proposition that the addressee brushes
their teeth.
Jbrush your teethK
c;g
(J?
2nd;singular
K
c;g
) =w. the addressee
c
brushes their
teeth in w
19
But see Han (2000), Kaufmann (2012), and Ninan (2005) for discussion.
80
This is the proposition being operated on byO
!
to give us the nal meaning of (22):
namely, that the proposition that the addressee brushes their teeth is required, which
has this logical form:
(23) O
!
(the addressee brushes their teeth)
Importantly, according to The Deontic Proposition Theory not only must
the subject and predicate be interpreted before bringing in O
!
, but to ensure all
imperatives have the same deontic modality,O
!
also has to take wide scope over any
quantiers and any other operators present in an imperative's meaning (Han 2000,
Chapter 5). Thus, on The Deontic Proposition Theory, the LFs in (25) do
not represent possible interpretations of the imperatives in (24), respectively:
(24) a. Someone help Jimmy up.
b. Be nicer to everyone.
(25) a. 9x(O
!
(x helps Jimmy up))
b. 8x(O
!
(addressee is nicer to x))
The only possible LFs for the imperatives in (24) are:
(26) a. O
!
(9x(x helps Jimmy up))
b. O
!
(8x(addressee is nicer to x))
And this is as it should be, sinceO
!
is supposedly what's providing imperatives with
their distinctive character. (More on this below.)
As it stands, Deontic Proposition Theory is unsatisfactory. The problem
is we know deontic propositions can be used in both performative and descriptive
speech acts. This means that if a deontic proposition is all there is to the meanings of
imperatives, the theory should predict imperatives are likewise useable in descriptive
speech acts and thereby that we should sometimes evaluate them for truth. But
as anyone who's tried to make an assertion with an imperative knows, successfully
executing a descriptive speech act using an imperative is an impossible task in normal
situations.
To explain this, Deontic Proposition Theorists, in particular Kaufmann (2012),
have proposed that imperatives also have a peripheral meaning, and included in this
peripheral meaning is a default semantic presupposition (or convention) saying the
imperative can only be used in performative, never descriptive, speech acts. That
this presupposition is default means in certain circumstances, it is overridden and so
not carried by the imperative. For example, an imperative embedded in a complex
sentence typically does not carry the presupposition. Consider:
81
(27) a. Wash the dishes.
b. Wash the dishes or go to your room.
c. If you are done eating, then wash the dishes.
It is plausible any use of (27-a) directs the addressee to wash the dishes. But when
embedded in (27-b) and (27-c) this is no longer the case. (27-b) is not directing anyone
to wash the dishes, and nor is (27-c). The reason, then, that imperatives are the
preferred vehicles of directive speech acts is because they are specically marked for
such uses. And any attempt to use an imperative in a descriptive speech act will result
in a violation of the presupposition, thus rendering the use infelicitous (unless the
presupposition is somehow overridden beforehand). Moreover, the reason it is odd to
say of an imperative that it is true or false is that the aim of a directive/performative
speech act is not to describe something as being some way or other, but to make
it some way, to change it. We will return to these claims below, when we turn to
evaluate The Deontic Proposition Theory.
2.4.2 Interpreting Imperatives
The Deontic Proposition Theory is readymade for interpreting and bringing
together all of the syntactic elements of imperatives that we identied in Chapter 1.
First, the subject of an imperative is given its usual interpretation. If it is a proper
name, then it has an object from the domain as its meaning. If it's `you', then its
meaning will be the addressee of a context. If it is a quantier phrase (QP; e.g.
`someone'), then it will have as its meaning as an appropriate quantier+property
combination. And so on. The same goes for the predicate of an imperative: it too will
get its usual interpretation. To this extent, interpreting imperatives on The Deon-
tic Proposition Theory is the same as doing so on The Property Theory.
One place where the theories might diverge is in what they say about how
the subject and predicate are put together. Recall that on one way The Property
Theory might interpret imperatives, the subject of an imperative is not an argu-
ment of the imperative's predicate (seex2.2). Rather, both subject and predicate are
arguments of an operator associated with IMP. On The Deontic Proposition
Theory, by contrast, an imperative's subject is always treated as an argument of
the imperative's predicate. Let's look at (16) again, repeated here.
(16) You go to bed.
The subject of (16) is the addressee of the context:
JYouK
c;g
= addressee
c
82
And recall the interpretation of the predicate of (16):
Jgo to bedK
c;g
= wx : x goes to bed in w
Because JYouK is of semantic type e, it can be made an argument of Jgo to bedK, so
that:
Jgo to bedK
c;g
(JYouK
c;g
)
= Jgo to bedK
c;g
(addressee
c
)
= wx : x goes to bed in w(addressee
c
)
= w : addresee
c
goes to bed in w
20
This last line of the derivation is the proposition that the addressee goes to bed, as
desired.
But what about IMP? Well clearly The Deontic Proposition Theory
is going to interpret it as the operator O
!
. Since O
!
is a propositional operator it is
of typehhs;ti;ti, that is, it takes a proposition as input and returns a truth-value.
Keeping with our previous simplifying assumptions about the modality of O
!
, in
particular that it represent weak deontic necessity, we can give IMP the following
interpretation:
JIMPK
c;g
= p : for most of the highest ranked worlds w
0
in\f(w
c
) as
ordered by g(w
c
), p(w
0
) = 1
where f and g are again the modal base and ordering source, respectively, w
c
is the
world of the context of use, and w0 is variable ranging over worlds. Now let's combine
this with the proposition that resulted from combining the subject and predicate of
(16):
JIMPK
c;g
(Jgo to bedK
c;g
(JYouK
c;g
))
= JIMPK
c;g
(w : addressee
c
goes to bed in w)
20
Obviously we continue to follow one tradition here in assuming that propositions are sets of
possible worlds, or equivalently, functions from worlds to truth-values. If we didn't assume this,
the last line of the foregoing derivation would make no sense. See Von Fintel and Heim (2002) for
further discussion.
83
= p : for most of the highest ranked worlds w
0
in\f(w
c
) as ordered by
g(w
c
), p(w
0
) = 1(w : addressee
c
goes to bed in w)
= For most of the highest ranked worlds w
0
in\f(w
c
) as ordered by
g(w
c
), w : addressee
c
goes to bed in w(w
0
) = 1
Hence:
J(16)K
c;g
= 1 i for most of the highest ranked worlds w
0
in\f(w
c
) as
ordered by g(w
c
), addressee
c
goes to bed in w
0
= 1
This then is (one example of) how The Deontic Proposition Theory interprets
imperatives.
But notice something about the last line of the foregoing derivation. It does
not represent a proposition, at least not on the conception of propositions that we
are currently working with. It rather represents a truth-value. True if the necessary
and sucient conditions are met; false otherwise. This, however, is not what The
Deontic Proposition Theory says the meanings of imperatives are { it says
that they are deontic propositions. There is an easy way to x this. Instead of the
above interpretation of IMP, we can interpret it as follows:
JIMPK
c;g
= wp : for most of the highest ranked worlds w
0
in\f(w) as
ordered by g(w) , p(w
0
) = 1
where everything is as before expect that the world of the context of use is no longer
given as the argument for f and g. As a result:
J(16)K
c;g
= w : for most of the highest ranked worlds w in\f(w) as
ordered by g(w), addressee
c
goes to bed in w
0
= 1
This denes a set of worlds, hence a proposition, as desired.
21
21
This is a simplied version of Kaufmann's interpretation of imperatives, and especially of IMP,
which she instead calls OP
Imp
. I have omitted many of the complicating details of Kaufmann's
analysis is because they don't concern us here, and because many of them are questionable. For her
nal analysis of IMP, see Kaufmann (2012, p. 162).
84
2.4.3 Evaluation
Deontic Propositions, Truth, and Action
As we've already seen, The Deontic Proposition Theory has a tidy explanation
for why it is inappropriate to ascribe truth-values to imperatives: because part of
their peripheral meaning is a presupposition requiring that they only be used in
performative speech acts, and the point of such speech acts is to change the ways
things are, not describe how they in fact are.
It's important to note that Kaufmann's presupposition must be pragmatic,
not semantic.
Pragmatic Presupposition
Sentence A pragmatically presupposes proposition
B i it is the case that A can be felicitously uttered
only in contexts which entail B.
Semantic Presupposition Sentence A semantically presupposes proposition B
i it is the case that A can be semantically processed
(i.e. interpreted) only if B is true.
(Karttunen 1974)
Were the presupposition semantic, then none of the example imperatives we have
looked at should be interpretable. They all should experience a semantic crash. But
obviously this is not the case. Despite not actually being used, and in some cases
merely mentioned, all of the imperatives we have so far considered are perfectly
interpretable. So Kaufmann's presupposition can't be semantic, else her view makes
obviously false predictions. But now here's the kicker: since the presupposition is
merely pragmatic, imperatives that are not being used in conversation should be
evaluable for truth. For example, each of the following should be truth-evaluable.
(28) a. Magdalena Kaufmann, watch out for scam artists.
b. David Lewis, stop bickering with Quine.
c. Hector-Neri Castaneda, write a book on personal identity.
22
But even here claims of representation and ascriptions of truth or falsity are out of
place. So Kaufmann only avoids ascribing truth to an imperative in cases where the
22
I use vocatives here to ensure these sentences don't lack subject-referents and for that reason
fail to be interpretable.
85
imperative is in fact being used. This is not avoided when an imperative is merely
being considered or merely being semantically evaluated. So despite what proponents
of The Deontic Proposition Theory may say, their theory does predict that it
is appropriate (in some cases) to evaluate imperatives for truth. But commonsense
tells us otherwise: it is never appropriate to evaluate imperatives for truth.
Perhaps this dispute over whether imperatives can have truth-values is only
verbal. Kaufmann treats propositions as sets of possible worlds, which are formal
objects. As formal objects, they don't have any substantive representational proper-
ties. There is no interesting sense in which a set represents something as being some
way. We can of course say that sets of worlds \represent" in the sense that a world
is a member of one of these sets i certain facts obtain in that world. But notice
how hollow this kind of representationality is. It's certainly not like the sort of rep-
resentationality that photographs and memories are said to have. It thus may not be
appropriate to use `true' to describe the logical values of imperatives. A better term
is `satised' and `satisfaction-values': an imperative is satised or not depending on
whether the actual world is in the set of worlds expressed by the imperative. So if
what Kaufmann is saying is that imperatives represent in a substantial way and can
be true or not, then I think her view is untenable. But if her claim is only that im-
peratives \represent" in an insubstantial sense and are satised or not, then the only
thing worth disputing here is her use of the phrases `represents' and `is true'. In the
interests of charity, let's assume Kaufmann is using these terms in the insubstantial
sense. (I return to this point in Chapter 4, where I argue that 'true' and 'satised'
denote the same thing but under dierent modes of presentation.)
Regarding the link to action, The Deontic Proposition Theory has
an adequate explanation available. To understand an imperative involves grasping
the deontic proposition that is its core meaning. As for what it means to accept
an imperative, there are a few possible stories to tell. Here's one. According to a
so-called internalist picture, there is a necessary connection between accepting a
deontic necessity proposition and being motivated to act on it, such that (roughly):
necessarily, if I accept deontic proposition O(p), where O is some sort of deontic
necessity operator, then I am to some degree motivated to make p true. Given this
picture, the link between imperatives and action is straightforward: accepting an
imperative leads, in the absence of physical and psychological barriers
23
, to action
because one does not really accept an imperative unless the are motivated to act on
it, and this motivation will lead to action, again provided there are no barriers.
23
Because the motivation for acting on an accepted deontic proposition might be overridden by
the motivation one has to do something else, this latter psychological fact will likely have to count
as a barrier on this picture of things.
86
Deontic Propositions and Negated Imperatives
One question to ask when evaluating a theory of the meanings of some linguistic
construction is whether it makes the right predictions about the construction's logical
properties. In this section I show that The Deontic Proposition Theory makes
the wrong predictions about negated imperatives.
To set things up, consider the following two imperatives:
(29) a. Open the window.
b. Don't open the window.
Intuitively, (29-b) is the negation of (29-a), such that if (29-b) is satised/true out
(at a given time and world), then (29-b) is not { and vice versa. Put another way,
(29-a) and (29-b) are contradictories. What's more, since sentences by themselves are
merely strings of symbols, two sentences can be contradictory only if their meanings
are. So what explains the contradictoriness of (29-a) and (29-b) is that the meaning
of (29-a) is being negated in the meaning of (29-b). Schematically, the meaning of
(29-a) is p and the meaning of (29-b) isp, and we know from classical logic that
(p &p) is a contradiction.
But now recall that on The Deontic Proposition Theory the modal
operator O
!
always takes wide scope over any other operators or quantiers in the
LF representation of the proposition being operated on (x2.2). Consequently, if The
Deontic Proposition Theory is correct, the meaning of (29-b) cannot be the
negation of the meaning of (29-b). Rather, the meaning of (29-b) has to be a propo-
sition with this LF:
(30) O
!
((addressee opens the window))
But any proposition with this LF is not the negation of the proposition that is the
meaning of (29-a), which roughly has the LF in (31).
(31) O
!
(addressee opens the window)
So, since O
!
must always take widest scope, one consequence of The Deontic
Proposition Theory is that there is really no such thing as a negated imperative.
There are imperatives with negation operators, sure. This we see in (30). But these
operators can never negate the deontic proposition that is the imperative's meaning.
The upshot: something has gone wrong with The Deontic Proposition The-
ory.
24
24
Of course it might be that (30) and (31) are inconsistent, depending on how exactly we dene
87
(Those familiar with debates in metaethics over expressivist theories of the
meaning of normative language will recognize the foregoing as an instance of the
negation version of the Frege-Geach Problem, adapted to apply to a theory of imper-
ative meaning. For those not familiar, see Schroeder (2010).)
Since the root of the problem is the claim that O
!
always takes widest scope
in an imperative, the obvious solution is to weaken this, instead holding that at least
the negation operator can take scope over O
!
. This would allow the core meaning of
(29-b) to have the LF in (32).
(32) (O
!
(addressee opens the window))
This is the negation of the meaning of (29-a), as desired.
A consequence of this solution is that negated imperatives will have one de-
ontic modality while non-negated imperatives have a dierent one. This is because
a negated necessity modal is semantically equivalent to a non-negated possibility
modal, and a negated possibility modal is equivalent to a non-negated necessity
modal. Accordingly, the pairs in (33) and (34) are semantically equivalent.
(33) a. It is not necessary that the snow is white.
b. It is possible that snow is not white.
(34) a. It is not permissible for you to steal.
b. It is obligatory for you to not steal.
This means ifO
!
expresses deontic necessity, then the result of negating it is deontic
possibility, and so according to the proposal under consideration, negated impera-
tives will express deontic possibility, while non-negated imperatives will continue to
express deontic necessity. Similarly, if non-negated imperatives instead express de-
ontic possibility, negated ones will express deontic necessity. So the price Deontic
Proposition Theorists have to pay for this solution to the problem is rejecting that
all imperatives have the same deontic modality.
That price is much too high. If it's true imperatives carry deontic
avor { and
that's still a big if { then it's more plausible than not that all imperatives, negated
and non-negated alike, express the same modality. After all, on the face of it, non-
negated and negated imperatives have the exact same feel, so to speak. What's more,
they also have the same discourse roles. Anywhere a negated imperative can be used,
a non-negated imperative can equally be used, and vice versa. For example:
O
!
. But this is beside the present point, which is that an imperative can be negated and the resulting
imperative will be contradictory, not merely inconsistent with, to the original.
88
A: What am I to do? B: Do your homework/Don't do your homework.
But this is not so for dierent deontic modalities. There are situations in which one
modality is acceptable, but the other is inappropriate:
A: What may I do? B: You may do your homework/#You must do your
homework.
This suggests all imperatives carry the same deontic modality.
A more serious problem with saying O
!
is what's negated in negated imper-
atives is that since negated and non-negated imperatives have the same discourse
roles, the key assumption of the main argument for The Deontic Proposition
Theory doesn't allow them to have the dierent deontic meanings. The assumption
in question, brie
y mentioned inx2.1, is:
DR-Meaning Link If two clause-types have the same discourse roles,
then they have the same type of meaning.
An instance of this assumption (i.e. P2) and the observation that performative uses of
SHOULD-declaratives and imperatives have the same discourse roles is what led us to
The Deontic Proposition Theory in the rst place. But given DR-Meaning
Link, if non-negated and negated imperatives also have the same discourse roles,
and it's clear they do, it follows that non-negated and negated imperatives have the
same type of meaning. But this is exactly what the proposed solution to the negation
problem has to reject. By saying O
!
is what's negated in a negated imperative,
the theory has to say negated and non-negated imperatives have dierent types of
meaning.
This puts the Deontic Proposition Theorist in a bind. They have to either (i)
reject the very assumption they rely on in motivating their theory { DR-Meaning
Link { or (ii) embrace the problem and reject that imperatives can be negated.
Option (ii) is, I think, unacceptable, and (i) is self-undermining.
25
25
It's worth noting Han (2000) seems to bite the bullet on this, defending an analysis in which the
negation operator never takes scope over the deontic operator in a imperative's meanings and so to
say that an imperative is \negated" does not mean that the imperative's meaning is negated, but
rather means that the proposition that is being operated on in an imperative's meaning is what's
negated.
89
Deontic Propositions and Reporting Imperatives
The last issue with The Deontic Proposition Theory I will discuss involves
the way uses of imperatives are reported. Consider the following conversation:
A: I need your orders for the day, lieutenant. B: Don't engage the enemy,
have your troops back here at 1300 for resupply, and scout the grounds
after lights out for soldiers out of the barracks.
Suppose that a third individual, C, has overheard what B said to A and wants
to report this to D. How can he go about this? Bolinger (1967) points out that
we do sometimes { though not often { report uses of imperatives with SHOULD-
declaratives, such that C might say:
C: B said A shouldn't engage the enemy, should have his troops back
at 1300 for resupply, and should scout the grounds after lights out for
soldiers out of the barracks.
But he also points out that this is a rather fast and loose way of reporting uses of
imperatives, because the person being reported could competently and reasonably
reject that that's what they said. B could, upon hearing C's report, retort:
B: That's not what I said! I didn't say A should do anything. I gave him
orders to be followed.
And the same goes for other deontic auxiliary verbs C might try to use to report what
B said; in each case B can competently and reasonably reject that the declaratives
containing those verbs accurately report what she said. This points up a disconnect
between the meanings of imperatives and the meanings of declaratives containing
deontic auxiliary verbs.
Bolinger presses the disconnect even further by next pointing out that one
way to accurately report a use of an imperative is with an innitive clause, not a
declarative one.
C: B told/ordered A to not engage the enemy, to have his troops back at
1300 for resupply, and to scout the grounds after lights out for soldiers
out of the barracks.
Adding to this, Huntley (1984) notes that an equally good way of accurately reporting
uses of imperatives is with subjunctive clauses:
90
C: B said/ordered that A not engage the enemy, A have his troops back
at 1300 for resupply, and A scout the grounds after lights out for soldiers
out of the barracks.
The thing to notice about both innitive and subjunctive clauses is that they in
no way encapsulate a deontic
avor, nor is there any (obvious) syntactic element
in their matrix that could semantically correspond to a deontic operator. This of
course strongly suggests that imperatives are closer, if not identical, in meaning to
innitives and subjunctives than they are to declaratives with deontic expressions.
So not only do declaratives with deontic terms fail to accurately report the meaning
of an imperative, but the types of clauses that do accurately report this are neither
declarative nor deontic. Given that innitives and subjunctives don't have deontic
propositions as their core meanings, we should therefore conclude imperatives don't
either.
Admittedly, this argument against The Deontic Proposition Theory
relies on a number of assumptions about both the nature and rules of indirect speech
reports as well as the semantics of innitive and subjunctive clauses, assumptions
which some may nd suspicious. I will return to these assumptions in Chapter 3
to spell them out more explicitly and to defend them, and then to show how they
should lead us to the conclusion, shared by Huntley, that despite being independent
clause-types, imperatives, innitives, and subjunctives nevertheless have the same
core meanings.
2.5 What We've Learned and Going Forward
The Property Theory and The Deontic Proposition Theory each have
their own advantages and shortfalls. The Property Theory has the advantage
of being able to denitively say that it's never acceptable to say an imperative has
a truth-value, while The Deontic Proposition Theory more work to do on
this front. Thus, all else being equal, we might lay down the following adequacy
conditions:
No Truth A theory of the meaning of imperatives should not say that
these meanings are the sort of thing that represent and can
be true or false in a substantial way.
Most will nd this condition uncontroversial, and some might even wonder why I
didn't just institute it at the get-go. I didn't because it might have led some to
91
reject The Deontic Proposition Theory out-of-hand, before we had a chance
to discuss it. Moreover, the preceding discussion has revealed that what's problematic
is not saying that imperatives are \true". Rather, what's problematic is saying that
imperatives are true because they represent the world as being some way. Imperatives
do not represent, directly or otherwise, and any view that implies that they do is
inadequate for that reason.
As for the relation between imperatives and action, The Property Theory
and The Deontic Proposition Theory are pretty much on a par. Both have
the makings of an explanation why understanding and accepting an imperative leads
to action, but these explanations are still very much partial and needing completion
before they can be fully evaluated. Regarding The Property Theory, the unan-
swered question is why putting a property on one's To-Do List moves them to do
what it takes to possess that property. Since there's nothing motivational about a
property, are To-Do Lists themselves somehow inherently motivational? Or is some
third, as yet unidentied factor at work here? As for The Deontic Proposition
Theory, it has to say why deontic propositions lead to action. Is it because ac-
cepting one of these propositions necessarily means being motivated to do what it
says? Or again, is something else doing the work of linking deontic propositions and
action? Let's encapsulate this in an adequacy condition:
Action Link A theory of the meaning of imperatives should provide the re-
sources for fully explaining why understanding and accepting
an imperative leads to action in the absence of barriers.
As things stand, while The Property Theory and The Deontic Proposition
Theory can meet this condition, both still have work to do.
One thing The Deontic Proposition Theory does much better than The
Property Theory is explain the distinctive sentential mood of imperatives. For
The Deontic Proposition Theory, the mood of imperatives is just whatever
deontic modality is expressed by O
!
(e.g. weak deontic necessity). The Property
Theory, by contrast, seems to have to either con
ate mood with conventional il-
locutionary force, which is the force of attempting to update someone's To-Do List,
or deny that imperatives have any characteristic mood at all. Neither of these op-
tions is desirable. The rst forces us to say that imperatives only have mood when
unembedded. And the second goes against the traditional way of thinking about the
dierent clause-types. Hence, all else equal:
Mood6=Force A theory of the meanings of imperatives shouldn't con
ate the
sentential mood of imperatives with their conventional illocu-
92
tionary force, nor should it eectively deny that imperatives
have a characteristic mood.
Finally, when it comes to embedding imperatives, The Property Theory
falls
at when it comes to the semantics of mixed compound sentences. The Deon-
tic Proposition Theory, however, has no problem with such sentences: mixed
compound sentences have exactly the same semantics as non-mixed ones, that is,
they combine two syntactic constructions of the same semantic type (i.e. type t).
Moreover, The Property Theory also seems forced to deny that uses of impera-
tives can be reported, since properties are not the right sort of objects for speech or
attitude reports, as can be seen in (35).
(35) a. *Jack said [
VP
go to the store].
b. *Jack told [
VP
go to the store].
c. *Jack accepted [
VP
go to the store].
Compare this to instead usingSHOULD-declaratives, which express deontic propo-
sitions:
(36) a. Jack said [
CP
that you should go to the store].
b. Jack told you [
CP
that you should go to the store].
c. Jack accepted [
CP
that he should go to the store].
But though The Deontic Proposition Theory allows for the meanings of im-
peratives to embedded under speech and attitude verbs, what they say embeds {
deontic propositions { is hard to square with actual reporting behavior, as accurately
reporting a use of an imperative requires employing an innitive or subjunctive, not
a modal declarative:
(37) a. Jack said for you [
CP
PRO to go to the store].
b. Jack told you [
CP
PRO to go to the store].
c. Jack accepted [
CP
that he go to the store].
(The embedded clauses in (37-a) and (37-b) are innitives and the one in (37-c) is a
subjunctive. The PRO element is the unarticulated subject of the innitive clauses.)
In addition, The Deontic Proposition Theory incorrectly predicts that imper-
atives cannot be negated, because it says that their meanings cannot embed under
the negation operator. All in all, then, neither The Property Theory nor The
Deontic Proposition Theory meet this adequacy condition:
Embedding (i) A theory of the meaning of imperatives should be able to
93
explain why imperatives can embed in both mixed and non-
mixed complex sentences as well as under negation, and
(ii) What the theory says is the meaning of imperatives should
be able to enter into accurate reports of people's speech and
attitudes.
The Property Theory fails to meet clause (i) because it can't account for mixed
compound sentences, and The Deontic Proposition Theory fails to meet it
because it doesn't allow the meaning of imperatives to be negated. As for clause (ii),
The Property Theory fails this because properties aren't the objects of speech
and attitude verbs, and The Deontic Proposition Theory fails it because any-
one being reported can deny that what they said or thought with an imperative was
a deontic proposition.
In the next chapter I will present and begin to develop a theory that meets all
of these adequacy conditions (it will take the rest of the dissertation to nish devel-
oping it). The theory has features in common with both The Property Theory
and The Deontic Proposition Theory. But it also diverges from them is one
dramatic way: it maintains that imperatives express regular old atomic propositions
(i.e. sets of possible worlds).
94
Chapter 3
A Na ve Satisfactionist View of
Imperative Meaning
3.1 Recap and the Proposed View
A lot has happened over the last two chapters, so a recap is in order. In Chapter 1, I
argued that imperatives are sentences, should not be confused with directive speech
acts, and are distinguished by containing in their syntactic structures a morpheme,
IMP , which is base-generated in C
0
and marks the construction for a distinctive
sentential mood or modality. Examples of imperative sentences I used to reach this
conclusion include:
(1) a. Go to the store.
b. You go to the store.
c. Do go to the store.
d. Don't go to the store.
e. Don't you go to the store.
What I argued was that imperatives like (1-a){(1-e) have exactly the same basic
syntactic structure and features as other clause types, including declaratives, and so
distinguishing imperatives requires positing the presence of the special mood mor-
phemeIMP . Moreover, I pointed out that though IMP is unarticulated in English
{ it's a morpheme with no phonological realization { this is not the case univer-
sally, as there are some languages in whichIMP is articulated. And I also proposed
that IMP can be used to explain some of the more confounding morphosyntactic
features of imperatives. For example, perhaps the reason imperatives cannot be the
argument of verbs likeSAY is due to an mismatch between the features of such verbs
95
and IMP .
According to the syntax I argued for, (1-a), for example, has this syntactic
structure:
CP
C
0
IP
I
0
VP
V
0
to the store
PP V
0
SpecVP
I
0
go+?
Infl
SpecIP
?
2nd;singular
C
0
IMP
SpecCP
I did not, however, say much about IMP itself in Chapter 1, other than that its
presence marks a construction for a particular mood or modality, and not tense,
aspect, or anything of that nature. This was because once we start asking questions
about which mood or modality it marks for, or even just what mood or modality is,
we have left syntax and entered the realm of semantics and meaning. The semantics
of imperatives was the topic of the previous chapter and this one, and is the main
question of this dissertation.
In Chapters 2, I presented two distinct theories of the meanings of imperatives,
and argued that both are inadequate, ultimately because they fail to meet one or
more of the following adequacy conditions:
Full Interpretation (i) Every functional and lexical element in the syntactic
structure of a construction corresponds to something in
's interpretation, and (ii) The structural relations between
the functional and lexical elements of correspond to se-
mantic relations between them in 's interpretation.
No Truth A theory of the meaning of imperatives should not say that
these meanings are the sort of thing that represent and can
be true or false in a substantial way.
96
Action Link A theory of the meaning of imperatives should provide the
resources for fully explaining why understanding and ac-
cepting an imperative leads to action in the absence of bar-
riers.
Mood6=Force A theory of the meanings of imperatives shouldn't con
ate
the sentential mood of imperatives with their conventional
illocutionary force, nor should it eectively deny that im-
peratives have a characteristic mood.
Embedding (i) A theory of the meaning of imperatives should be able
to explain why imperatives can embed in both mixed and
non-mixed compound sentences as well as under negation,
and
(ii) What the theory says is the meaning of imperatives
should be able to enter into accurate reports of people's
speech and attitudes.
The general view I now want to defend, and which is capable of meeting the
foregoing conditions, maintains that the core meanings of imperatives are run-of-
the-mill propositions, understood as sets of worlds, and are what are understood and
accepted when one accepts an imperative. Particular versions of this view { that
imperatives express run-of-the-mill propositions { will of course disagree on exactly
which run-of-the-mill propositions are the meanings of imperatives. The view I wish
to advance says that imperatives express the same propositions as simple present
tense declaratives. I call this view Na ve Satisfactionism. It is called \satisfac-
tionism" because it holds that the logical values of imperatives are satisfaction-values.
And it is a \na ve" view because it holds that there is no more to the meanings of
imperatives than what we can glean from basic observations about their syntax (i.e.
that they have subjects, predicates, and simple present tense). Obviously such a view
appears ot immediately face diculties when it comes to the above desiderata. How
does it satisfy FI, No Truth, Action Link, Mood6=Force, and Embedding?
I will answer this question in due course. But rst we need to get view on the
table. To that end, I will begin with a selective overview of Satisfactionism's prede-
cessors, showing where they go wrong and how correcting them pushes us towards
not just Satisfactionism in general, but towards Na ve Satisfactionism. After the
larger dialectical context has been xed and the view has been laid out, I will then
add a further element to the story. This is a view of the role of sentential mood in
conversation. What I will argue is that (i) IMP makes no semantic contribution to
97
the interpretations of imperatives, but rather (ii) what it does is determine a conver-
sational rule or norm for what's involved in accepting an imperative. By combining
this view of sentential mood together with Na ve Satisfactionism, FI, No Truth,
and Mood6=Force can be met. However, still more is required for Embedding and
Action Link. So signicant room is dedicated to these. In particular, meeting Em-
bedding is the aim of the end of this chapter as well as all of Chapters 4, while
meeting Action Link is the topic of Chapter 5. As we will see, meeting these two
desiderata involves rethinking some of the basic assumptions of semantic theorizing
as well as looking much more closely at linguistic phenomena that is usually given
short shrift.
3.2 Semantics of Imperatives: A Selective History
It's fair to say that work on the semantics of imperative began in earnest with
Jrgensen (1937). Jrgensen gave us what amounts to a puzzle. The puzzle is this.
Intuitively, imperatives are neither true nor false. So whatever the logical values (i.e.
extensions) of imperatives are, they are not truth-values. Yet imperatives also appear
to enter into valid inferences.
(A) If you are hungry, then eat something.
You are hungry.
Therefore, eat something.
(B) Put all of the toys you can see in the box.
This
1
is a toy you can see.
Therefore, put this
1
in the box.
This must mean that the logical values of imperatives are such that a notion of
validity can be dened in terms of them. But what could these values be if not
truth-values?
Jrgensen's solution to this puzzle was to deny that imperatives have any logi-
cal values to call their own, and then maintain that the validity of inferences involving
imperatives should be reduced to the validity of inferences involving corresponding
declaratives: an inference involving imperatives is valid if and only if replacing all the
imperatives with their declarative counterparts results in a declarative-only inference
that is valid. Strictly speaking, then, 'imperative inference' is a misnomer. There are
no imperative inferences on Jrgensen's understanding of things, nor can there be
given that imperatives don't have their own logical values. There are just declarative
98
inferences and we call something involving imperatives an "inference" because we
are confused about what is being inferred from what, or because we're being lazy.
This proposal is radical. Jrgensen is not just denying that imperatives stand
in logical relations to themselves and other sentences. But the reason he denies this is
because he also denies that there is anything to be called a semantics of imperatives.
According to him, the only sentences that are meaningful, hence appropriate for
semantic theorizing, are declaratives:
The derived indicative sentence [= the declarative counterpart of an im-
perative] is capable of being true or false { in Wittgenstein's terminology,
it "pictures" a (possible) fact { and therefore it is meaningful in contradis-
tinction to the imperative sentences which as such are not testiable and
seem to acquire a meaning through their indicative derivations only. Or
more strictly expressed: An imperative sentence has a meaning if and
only if the corresponding indicative sentence which may be derived from
it and which describes its contents is meaningful. (Jrgensen 1937, p. 291)
Jrgensen's argument against imperatives being meaningful in their own right rests
on the logical positivists' principle that a sentence is meaningful only if it is veri-
able or \testiable". Given that today this principle is almost universally rejected,
Jrgensen's argument will hardly seem persuasive. But if imperatives are meaningful
in their own right, what are their semantics?
Hare (1952), taking up an idea inchoate in Jrgensen, thought that while it's
true the contents of imperatives share a component with the contents of declara-
tives, imperatives nevertheless have their own unique contents. The basic idea is
that imperatives have two semantic components: the content they share with their
declarative counterparts and the distinctive operator, which I label as `!'.
1
Suppose
the content shared by a declaratives and an imperative is p. Then the content of the
imperative is !(p). The shared content is supplied by an imperative's subject and
predicate, which are also the subject and predicate of the declarative counterpart,
while ! is supplied by the so-called imperative mood. Correspondingly, the content
of the declarative counterpart is D(p), where D is the distinctive operator supplied
by the declarative mood. One of the main semantic dierences between imperatives
and their declarative counterparts is while D(p) = p, !(p)6= p. If p is a proposition,
then this amounts to saying both that the meaning of a declarative is no more than
the proposition it expresses and that the meaning of an imperative is more than just
this.
1
Hare called the shared content the phrastic and the distinctive mood component the neustic.
99
What sort of operator does the imperative mood contribute? Hare made the
problematic claim that the imperative mood contributes directive illocutionary force.
This makes ! an illocutionary force operator. So the meaning of an imperative is
a proposition presented with directive force. The trouble with such a proposal is
that it obviously violates the desideratum Mood6=Force. The motivation behind
Mood6=Force is the Frege-Geach Problem. In brief, the problem is that illocutionary
force does not embed. For example, while the imperative (2) directs whoever is
addressed to wash the dishes, (3) does not.
(2) Wash the dishes.
(3) If you are done eating, wash the dishes.
We might say that (3) conditionally directs one to wash the dishes, in the sense that
! takes scope over the entire conditional and not just the consequent, giving it the
logical form !(if p, then q). But now (2) must have one meaning when unembed-
ded { one with the form !(q) { and another when embedded. Not only does this
mean imperatives are systematically ambiguous, but also that inferences involving
imperatives like (A) and (B) are invalid. But if any inferences involving imperatives
are valid, surely those are. Thus illocutionary force can't be part of an imperative's
meaning, and so, contra Hare, ! can't be an illocutionary force operator.
2
Another closely related issue with Hare's semantics is that it appears to be
fundamentally incomplete. Hare seems to accept that the logical values of declara-
tives are truth-values. But what are the logical values (i.e. extensions) of imperatives?
Without an answer to this question, it will be dicult to provide denitions of logical
notions like validity and consistency. What's more, it will also make it dicult to
formulate a compositional semantics for imperatives. This is because on the standard
way of doing semantics, logical values are determined, compositionally, by an inter-
pretation function (relative to a contextnindex and assignment function). Granted,
we can say that the meaning of an imperative is a proposition presented with direc-
tive force. But directive force is not the sort of thing which can be the output of an
interpretation function. Jimperative moodK is therefore ill-dened and by extension
so is any construction with the imperative mood as a part.
It's worth noting, as Hare does, that the encoding of directive force does not
2
Sometimes Hare seems to go full expressivist and say that the meaning of an imperative is
(the expression of) a volitional attitude: roughly, when one uses an imperative they express their
willing that a proposition be made the case, and this attitude of willing is thus the meaning of the
imperative (Hare 1999, p. 21). Since I'm not concerned with expressivist semantics for imperatives,
and also since the move to expressivism in many ways makes the Frege-Geach problem worse (see
Schroeder (2010)), I won't try to gure out whether Hare really was an expressivist.
100
prevent an imperative from being used in non-directive ways, as in a plea or excla-
mation. To accommodate this
exibility, Searle (1962) points out it would be better
to characterize the Harean meanings of imperatives as propositions that are made
suitable, by way of the imperative mood, for use in directive speech acts. Importantly,
this way of characterizing Hare's view helps to avoid violating Mood6=Force, pro-
vided we also hold that compound sentences, like (3), don't also have directive force
operators taking scope over them, in which case inferences involving imperatives
like (A) above will still come out invalid (having the invalid form: !(p! !q), pnn
therefore, !q). But this move doesn't help with the compositionality problem, since
\suitability for use in directive speech acts" is still not the right sort of thing to be
the output of an interpretation function.
Casta~ neda (1975) addresses the compositionality problem by arguing that
the logical values of imperatives are what he calls legitimacy-values. According to
Casta~ neda, while declaratives express propositions, imperatives express prescriptions,
and every prescription is either legitimate or illegitimate, just as every proposition
is either true or false. This allows imperatives to stand in logical relations to each
other. A set of imperatives is inconsistent i there is no interpretation on which each
member of the set expresses a legitimate prescription. And an argument consisting
of nothing but imperatives is valid i the conclusion-imperative must, in the logical
sense, express a legitimate prescription if the premise-imperatives do, while an argu-
ment consisting of both imperatives and declaratives is valid i the conclusion must
express a legitimate prescription or true proposition if the premises do.
To get a compositional semantics out of Casta~ neda's idea we need to hold that
sentential moods are semantic type determiners. The idea is roughly this. To every
sentence type there corresponds a basic semantic type in a Montagovian framework.
3
Declaratives are of type t; imperatives are of type i; interrogatives of type n; etc.
Sentences are built up in the usual way from words and phrases. But nothing is a
sentence unless it is of one of the basic types (except for basic type e, which is the
type of referring words and phrases), and what sentential mood does is determine
which type a sentence is.
4
To keep things simple we can say that the types of words
and phrases are built up from the basic types e and t.
5
So, as usual, names are type
e, intransitive verbs are typehe;ti, transitive verbshe;he;tii, and so on. Then the
3
See Dowty et al. (1981) for an overview of Montague Grammar.
4
cf. Boisvert and Ludwig (2006), who make a similar proposal without utilizing Montagovian
type theory.
5
Nothing of present concern turns on this choice. We can just as easily build up the semantic
types of words and phrases from e andi, or frome andn. It would be an empirical question whether
words and phrases are built up from t, i, n, or the semantic type of some other kind of sentence.
101
semantic function of the imperative mood is to convert things of type t into type i.
Strictly speaking, this makes the imperative mood typeht;ii, which is why I call it
a type determiner, since it determines the ultimate semantic type of a sentence.
6
There are two problems with the view we have arrived at. First, Casta~ neda's
legitimacy-values need to be relativized not just to worlds and times, as truth-values
are, but also to an agent's ends. Second, the view forces us to say that the logical
connectives are systematically ambiguous. In the next section I will argue that the
rst problem with legitimacy-values is enough to push us towards Na ve Satis-
factionism, the view that the logical values of imperatives are satisfaction-values.
As for the second problem, an entire Chapter is required. So we will hold of discus-
sion until Chapter 4. To preview coming events, I will argue that the problem of
mixed compound sentences pushes us one step further, towards what I call Strong
Na ve Satisfactionism, which is Na ve Satisfactionism plus the view that
satisfaction-values = truth-values. Perhaps the most surprising upshot of this view
is that it means there is no real semantic role for sentential mood to play, because
it makes imperatives, declaratives, and the other sentence types semantically iden-
tical. This of course does not mean there is no communicative dierence between
imperatives and declaratives. There is, but as I will maintain inx4 of this chapter,
this dierence is in what it means to accept these dierent sentences. To accept an
imperative is to change one's intentions in the appropriate way, while to accept a
declarative is to change one's beliefs, and to accept an interrogative is to change
something like one's state of wonder.
3.3 Na
ve Satisfactionism
While I agree imperatives must have binary logical values, I'm skeptical about
Casta~ neda's legitimacy-values.
7
The crux of the problem is that unlike truth-values,
legitimacy-values need to be relativized to the ends of the subject of the imperative.
Consider (4).
6
Similarly for the interrogative mood. It will be of typeht;ni. Had we chosen to build up the
semantic types of words and phrases from i instead of t, then the imperative mood would be of
typehi;ii, making it semantically vacuous.
7
Vranas (2008) argues that imperatives are in fact three-valued. Unfortunately, his arguments
for this are less than persuasive. In particular, they seem to rest on a con
ation of imperative
sentences and directive speech acts. His idea is that in compound sentences like (i), intuition pulls
us to the thought that if the antecedent is not true, then the entire conditional is neither satised
nor not satised { it is \avoided".
(i) If you are hungry, eat something.
102
(4) Clean the bathrooms.
Suppose Alice says (4) to Bob. With Bob as the subject, is (4) legitimate? According
to Casta~ neda (1975, p. 134), it depends. If cleaning the bathrooms supports Bob's
ends, it is legitimate. But if it doesn't support Bob's ends or if it is merely consistent
with them, it is illegitimate. So relative to some sets of ends (4) is legitimate, but
relative to other sets it is illegitimate. As Bratman (1983) points out, the problem
with this is that (i) people can have dierent ends in dierent situations, and (ii) a
person's ends in a given situation can con
ict. Suppose Bob is a janitor and currently
at work. Then in this situation (4) is legitimate because one of Bob's ends is doing
his job and cleaning the bathrooms supports this end. But once Bob leaves work
his ends change. No longer is he concerned with doing his job. Moreover, outside of
work, Bob has the end to never clean a single bathroom. So now (4) has gone from
legitimate to illegitimate just by virtue of Bob clocking out. This marks a sharp
divergence between legitimacy-values and truth-values. Relativization to things like
ends is not required for declaratives and truth-values. A declarative will be true or
not (in a context) regardless of what agents want, believe, etc.
Bratman points out we can take this divergence one step further. We all know
that we can have ends that may con
ict with each other. For example, Alice may
have the following two ends: to do everything her wife says and to do everything her
mother says. Now suppose, as is plausible, that on Monday Alice's wife says (5-a) to
Alice and then on Tuesday her mother says (5-b) to her.
(5) a. Clean the garage on Saturday.
b. Don't clean the garage on Saturday.
Only one of these can be legitimate. But which one? Given the ends that Alice has
it's hard to say. Cleaning the garage supports her end to do what her wife says,
but undermines her end to do what her mother says. And not cleaning the garage
supports her end to do what her mother says, but undermines her end to do what
her wife says. If an imperative is legitimate if it supports at least some an agents
ends, then both (5-a) and (5-b) will be legitimate here, which is not what Casta~ neda
I don't share this intuition. It seems to me that if the antecedent of (i) is not true, then the whole
conditional is satised. Now I agree that if (i) isused in a conditional order, such that one is ordered
to eat something if one is hungry and one is in fact not hungry, then there is a serious question
about what to say about status of the order. It does seem that the order is avoided in such a
case. But does that mean (i) itself has being avoided as its logical value? Clearly not. An order can
be avoided while the sentence used to issue the order is satised (or unsatised). Avoidance is a
property of speech acts, and in particular conditional speech acts. It is not, though, a property of
sentences. Once we pull these apart, Vranas's case for imperatives being three-valued collapses.
103
wants, since he takes pairs like this to be contradictories, as certainly appears to be
the case. But if, on the other hand, an imperative is legitimate if it supports all of
an agents ends, then neither (5-a) nor (5-b) will be legitimate in our scenario, which
we also don't want, because it will require us to take up a non-classical logic for
imperatives.
It is because legitimacy-values are relativized to the ends of agents that these
problems arise.
8
This suggests that the logical values of imperatives shouldn't be
relativized in this way. More strongly: I take it to mean that the logical values of im-
peratives should be relativized to no more than what truth-values are relativized to,
which are just worlds and times. Whatever the positive logical values of imperatives
are, whether an imperative has this value shouldn't depend on the mental states or
aims of the person to whom it is addressed, nor those of the person who uses it, in
case one thought that might solve the problems (it won't). All it should depend on
is the way the world is, was, or will be.
One idea that this line of reasoning seems to lead to is that imperatives have
truth-values after all. Lewis (1970), following Stenius (1967), argued that in order
to accommodate imperatives in our formal semantics, we need to treat !, which is
supplied by the imperative mood, as a propositional operator. On his proposal, what
! does is take us from a proposition to a proposition stating that the latter proposition
is demanded or ordered. Roughly, then, imperatives are semantically equivalent to
certain performative declarative sentences.
(6) a. JGo to the storeK = JIt is demanded that you go to the storeK
b. JClean the garage on SaturdayK = JIt is demanded that you clean the
garage on SaturdayK
Thus an imperative will be true just in case the agent has in fact been demanded
to perform the relevant act. Arguably, the simple act of using an imperative in
a directive speech act is enough to make it true, enough for the demand to have
occurred. So it is a rarity to encounter a false imperative in our day-to-day lives, as
virtually all imperatives we encounter are being used in directives.
9
One obvious problem with Lewis's proposal is that not all imperatives are used
demandingly. Often we use imperative merely to suggest that someone do something
8
To the readers who are still wondering what exactly legitimacy values are: you are not alone.
Casta~ neda himself struggled to say something substantive about legitimacy (Casta~ neda 1975, ch.
5). In the end he only really commits to legitimacy being the imperative analogue to truth, which
is for the most part all I have said about it as well.
9
Interestingly, this means the example imperatives I have produced are all false, because they
are not actually being used { we are just pretending they are.
104
or to advise them to do it. Here is a natural conversation with an imperative being
used to advise.
Don: Do you have any suggestions about what to do about my job situation?
Helen: Yeah, quit. That job has caused you nothing but stress.
Don: Thanks, I'll think about it.
Here Helen is not demanding that Don quit, far from it. She is merely advising it.
Kaufmann (2012), whose view we discussed in the previous Chapter, concludes from
the fact that imperatives can be used in a wide variety speech acts, including speech
acts that don't have an obvious directive character (e.g. wishes), that ! is better
thought of as a deontic propositional operator, similar to the operator expressed by
the verb SHOULD. So instead of the equivalences in (6), on Kaufmann's view we
have:
(7) a. JGo to the storeK = JYou should go to the storeK
b. JClean the garage on SaturdayK = JYou should clean the garage on
SaturdayK
Accordingly, an imperative is true just in case the subject should perform the relevant
act.
As re
ected in the desideratum No Truth, the idea that imperatives can be
or are true or false is deeply unnatural. But as discussed in the previous chapter
(x4.3.1) and also re
ected in No Truth, there is are two ways of understanding
`truth': one problematic, the other not. The unproblematic understanding is the one
according to which truth is merely aformal property of sentences, in the de
ationary
sense that a sentence is true just in case the world is some way. This understanding
is unproblematic because it doesn't require saying anything substantive about the
meanings of imperatives, only that an imperative is true just in case such-and-such
is the case.
The problematic understanding of `truth' is the one where we treat truth as a
property belonging to things that represent in a substantial committal way. Beliefs,
for example, can be true on this sense because the function of belief is to represent
the world as being certain ways, such that if you believe p, then your are committed
to the world being the way you believe it to be. A belief is true if the world is the way
the belief represents it. The story with declaratives is a bit more complicated. Declar-
atives don't represent directly. They represent only because if someone understand
and accepts a declarative, then the result will be them representing the world in
some way and being so committed. Despite this indirectness it is still apt to say that
105
the function of declaratives in communication is to represent. Imperatives, by con-
trast, don't function to represent the world in a committal way, not even indirectly,
because imperatives are not the sort of thing that represent in a substantial sense.
Rather, imperatives function to change the world. If an imperative is understood
and accepted, then the result will be someone doing something, thereby changing
something from one state into another (e.g. from someone not having cleaned the
garage to them having clean the garage).
For Kaufmann (and equally for Lewis), propositions are merely sets of possible
worlds, which are formal objects. As formal objects, they don't have any substantive
representational properties. There is no interesting sense in which a set represents,
in a committal sense, something as being some way. We can of course say that sets of
worlds \represent" in the sense that a world is a member of one of these sets i certain
facts obtain in that world. But notice how hollow this kind of representationality is.
It's certainly not like the sort of representationality that pictures and memories
are said to have. It would thus be appropriate not to use `true' to describe the
logical values of imperatives. A better term is `satised' and `satisfaction-values': an
imperative is satised or not depending on whether the actual world is in the set of
worlds expressed by the imperative. (In the next chapter I argue truth=satisfaction.)
I have already argued that, overall, Kaufmann's view is inadequate (x4.3 of
previous chapter). The crux of the issue is that her view accords imperatives more
logical structure than they in fact have (obviously the same issue applies to Hare's
and Lewis's views). This is due to her claim that in the semantics of imperatives there
is an operator that operates on the proposition expressed by an imperative's subject
and predicate. We need to remove this operator, and keep what remains. Thus, what
I wish to propose is that the logical values of imperatives are satisfaction-values (as
described above) and that the set of possible worlds that imperatives express is solely
determined by their subjects and predicates. (4), for example, expresses the set of
worlds in which the addressee sweep the bathrooms, and it is satised i the actual
world is a member of that set. In the terminology of the previous Chapter (x2.4), the
proposition determined by an imperative's subject and predicate is its core meaning.
I call this view Na ve Satisfactionism. What makes it \na ve" is the satis-
faction conditions it says imperatives have. It adds no bells and whistles, so to speak.
We observe that imperatives have subjects and predicates and that the main verb
is morphologically meager. From this and this alone we conclude that an imperative
is satised i the subject performs the act specied by the predicate. We don't say
that imperatives semantically encode directive force or deontic modality. Nor do we
say they encode future tense. We don't even say that the subject must perform the
106
act intentionally.
10
At the end of the day, all we say is that an imperative is at least
intensionally equivalent to its simple declarative counterpart, such that:
(8) a. JGo to the storeKu JYou go to the storeK
b. JClean the garage on SaturdayKu JYou clean the garage on SaturdayK
Adding anything else to the satisfaction conditions of imperatives would take us away
from Na ve Satisfactionism towards a more sophisticated and yet less plausible
version of the theory.
11
Satisfactionism in general gives us an easy way of dening when an imperative-
involving argument is valid. An argument is valid i the conclusion must be true or
satised if the premises are (i.e. there is no interpretation on which the premises
come out true or satised while the conclusion comes out false or unsatised). Given
this denition, the arguments we previously looked at, repeated here, come out valid,
as desired.
(A) If you are hungry, then eat something.
You are hungry.
Therefore, eat something.
(B) Put all of the toys you can see in the box.
This
1
is a toy you can see.
Therefore, put this
1
in the box.
10
Boisvert and Ludwig (2006) argue for a more sophisticated version of satisfactionism according
to which an imperative is satised (in their terminology obeyed) only if the subject performs the
appropriate act intentionally. I think the impetus to hold that the satisfaction of an imperative
requires intentional act performance rests on a common confusion between imperatives and the
directive speech acts they are used in. I fully grant that to obey or comply with, say, an order
requires intentional performance of some act. But that doesn't mean the satisfaction of imperatives
used in such speech acts requires intentional act performance as well. An imperative is satised or
not regardless of how it came to be satised, just as a declarative is true or not regardless of how
it came to be true.
11
One diculty in stating the view is that in English 2nd-person simple-present-tensed declara-
tives have the same morphology as 2nd-person imperatives in which the subject is made explicit.
Because of this, it may seem that what's being equated in (18-a) and (18-b) is the imperative to it-
self. A better way to make clear that what's being equated is an imperative and a simple declarative
is to use 3rd person subjects:
(i) a. JEveryone go to the storeKu JEveryone goes to the storeK
b. JSomeone clean the garage on SaturdayKu JSomeone cleans the garage on SaturdayK
107
For now let's assume a compound sentence made up of a declarative and an imper-
ative is an imperative (we will eventually reject this in Chapter 4). Now suppose
that the premises of the rst argument are satised and true, respectively. If the
rst premise is satised then either you are not hungry or you eat something. But
if the second premise is true, then you are hungry. Therefore, it must be the case
that the conclusion is satised, because it must be the case that you eat something.
Similarly, suppose that the premises of the second argument are satised and true,
respectively. Then you have put all of the toys in the box and a particular thing (the
referent of the demonstrative `this') is a toy. From this it follows that you must have
put that particular thing in the box.
Na ve Satisfactionism thus has distinct advantages over its competitors,
foremost amongst which is its simplicity. But issues remain. Not least of which is
this question: what of the imperative mood? What about the morpheme IMP? Does
it have no role to play in the life of an imperative? As I understand the view, Na ve
Satisfactionism denies there is any semantic role to be played by IMP and sen-
tential moods; moods are uninterpreted features much like grammatical gender. But
that doesn't mean sentential mood serves no communicative function. In the next
section, I will argue that the sentential moods { and more specically, mood mor-
phemes { non-semantically function to determine what one should do when accepting
a sentence. The imperative mood, for instance, determines that accepting an impera-
tive requires intending to make the world a certain way. By contrast, the declarative
mood functions to determine that accepting a declarative requires believing the world
is a certain way. And the interrogative mood functions, perhaps, to determine that
accepting it requires wondering whether the world is some way.
3.4 The Role of Sentential Mood
There's no denying sentences have moods. If they didn't then there would be no
reason why languages have sentences with distinctive yet identiable morphologies
and constituents, no reason for their to be imperatives, declaratives, and the rest,
rather than just one general nondescript type. Mood therefore must play some im-
portant role in language and communication (a point re
ected in the desideratum
Mood6=Force). I maintain that mood does not play a semantic role: whatever mood
is, it does not enter into a sentence's interpretation. Accordingly, IMP and other
mood morphemes are uninterpretable features of sentences. In other words, mood
does not contribute to the core meanings of imperatives.
An alternative idea is that mood morphemes function to determine the rules
or conventions of use governing a sentence, which are part of a sentence's peripheral
108
meaning.
12
In the literature on imperatives we often nd people saying things like
\an imperative is used to express a desiderative attitude."
13
Clearly such a state-
ment is factually inaccurate. Imperatives are regularly used without expressing any
desiderative attitude. For instance, I don't care one way or the other about whether
David Lewis stops bickering with Quine, but I can still use (9).
(9) David Lewis, stop bickering with Quine.
So this statement { \an imperative is used to express a desiderative attitude" { must
be a normative one, expressing a rule that imperatives are to be used to express
desiderative attitudes. That this rule of use is unique to imperatives is due to IMP
and the imperative mood: only sentences with the imperative mood are governed by
this rule. Declaratives by contrast have their own rule, determined by the declar-
ative mood, as do interrogatives. The rule for declaratives seems to be something
like a declarative is to be used to express an attitude of belief. While the rule for
interrogatives is an interrogative is to be used to express an inquisitive attitude.
It's useful to compare this approach to sentential mood to a view in Metaethics
that is growing in popularity. Hybrid Expressivism is the view that normative
declarative sentences like (10) express two things.
(10) Jill ought to help Jack.
Not only do they express propositions like other declaratives (perhaps because they
express beliefs in those propositions; I ignore this complication). But they also express
certain non-cognitive attitudes (e.g. desires).
14
One way to put this is by saying
that normative declaratives have two meanings, one propositional and the other
attitudinal. The rules of use approach to sentential mood is saying very much the
same thing, but in a dierent domain. When dealing with moods, we are concerned
with dierent types of sentences. Hybrid Expressivism, by contrast, in concerned
with dierent subtypes of the declarative type. Nevertheless, one way of thinking of
the Hybrid Expressivist project is as a project in how certain words and phrases,
such as `ought', can add to the rules of use governing a sentence. (10) is arguably a
declarative. But if Hybrid Expressivism is right, the rules governing (10)'s use are
not exhausted by the rule governing declaratives. The declarative use-rule still stands,
but there is another use-rule governing normative declaratives, the one determined by
12
In what follows I focus on rules of use. Nothing turns on this choice. Everything I say applies,
mutatis mutandis, to a view that replaces rules with conventions.
13
Searle (1969) and especially Bach and Harnish (1979)
14
Proponents of this view, in one form or anther, include Copp (2001), Ridge (2006), and Boisvert
(2008).
109
normative words and phrases. Thus there are two use-rules for normative declaratives.
Moreover, the use-rule that `ought' and other such words introduce is very much like
the one introduced by the imperative mood. It's therefore natural to think of Hybrid
Expressivism as something like an extension to the proposed view about the role
of sentential moods, because it holds (i) that other sentential elements besides mood
can determine use-rules, and (ii) that use-rules can be compounded.
15
Although I too endorse a rule-based account of sentential mood, rules govern-
ing use make me wary. For one thing, how does the rule apply when a sentence is
embedded? This is just the Frege-Geach Problem arising in a dierent form. Clearly,
when I utter sentences like (11-a)-(11-c), I am using the embedded imperatives.
(11) a. If there are dishes in the sink, clean them.
b. Take the food out of the freezer only if it will defrost in time for dinner.
c. Sweep the
oors or clean the bathroom.
But in this context I am not taken to be expressing any sort of attitude towards
the contents of these imperatives. So when embedded, the use-rule seems not to
apply. Why? It can't be because in these contexts imperatives somehow lose their
mood. Mood arises out of syntax, and an imperative's syntax does not change when
embedded. So maybe embedding instead causes the rule to be cancelled.
16
Then in
what sense are these rules of use, if there are wide classes of uses for which the rules
no longer hold? Moreover, what about embedding causes the cancellation?
17
It seems
implausible and ad hoc to say it's due to the connective words `and' and the rest.
None of this is to suggest that proponents of use-rules cannot answer these
questions. Many have tried, to varying degrees of success.
18
But the diculties con-
15
Pure Expressivism agrees with (i), but not (ii). According to Hare's early view, which is
a direct predecessor to Pure Expressivism, normative words and phrases function to shift the
use-rule of a sentence from the one determined by the declarative mood to the one determined by
the imperative mood. Importantly, Hare was careful not to say that normative words and phrases
transformed a declarative into an imperative. Rather, what they do is something like override the
rule determined by the declarative mood and put in its place the rule determined by the imperative
mood. Thus, on this way of viewing things, which is shared by Pure Expressivism, use-rules do
not compound: a sentence has one single use-rule { period. But it's also worth noting that Hare
thought a sentence's use-rules were semantically encoded. As we saw when looking at Hare's view
of the semantics of imperatives, sentential moods express specialized operators which contribute
to a sentence's interpretation. The same goes for normative words and phrases. However, we have
rejected this and instead maintained that mood makes no semantic contribution.
16
Cf. Hanks (2007)
17
Cf. Hom and Schwartz (2013)
18
Barker (2004) has, I think, made the most progress on this front; Hanks (2007, 2015) comes in
close second.
110
fronting use-rules give us reason to look for an alternative account of sentential mood.
The alternative I prefer involves rules of acceptance. Sentential moods determine the
rules governing the acceptance of a sentence. Here are the rules for the three main
moods:
Imperative Mood To accept a sentence in the imperative mood is to intend its
content.
Declarative Mood To accept a sentence in the declarative mood is to believe
its content.
Interrogative Mood To accept a sentence in the interrogative mood is to wonder
about its content, or to intend to provide an answer, or to
form some other attitude.
19
In short, the role of a mood morpheme and the corresponding sentential mood is
to tell us which region of our mental lives we should be engaging when considering
the proposition expressed by a sentence; they tells us which mental \box" to put a
proposition in. Imperative mood tells us to be engaging the intention region, to put
a proposition in the \intention box". Declarative mood tells us to be engaging the
belief region, to put a proposition in the \belief box". And so on. It's in this way
that sentential mood contributes to the peripheral meanings of sentences (but not
the core meanings).
There are a number of advantages to this account of sentential mood. Not
least of which is that acceptance-rules don't face Frege-Geach. Acceptance occurs, if
at all, at the back-end of a communicative exchange. Acceptance-rules say nothing
about what is involved in the use or mere production of a sentence, and it is only at
these earlier stages that Frege-Geach Problems can take hold. When an imperative is
embedded, as in (11-a)-(11-c), the acceptance-rule associated with imperative mood
still applies in full force. To accept one of the imperatives embedded in (11-a)-(11-c)
is, according to the rule, to intend the propositions these imperatives express.
20
19
I take no rm stance on what attitude is involved in accepting an interrogative. I lean towards
the view that it's a state of wonder, which is distinct from both belief and intention. But this
view requires argument. So I leave it open that accepting an interrogative might instead involve an
intention to answer, or some other attitude entirely.
20
There may of course be strictly pragmatic dierences between uses of embedded and unembed-
ded imperatives. The pragmatic story I prefer is that while a use of an unembedded imperative in
a directive speech act can provide one with direct and immediate reason to accept it, this is not so
when the imperative is embedded. Suppose Jill says (i) to Jack in an order.
111
Another advantage of the rules of acceptance approach is it ties in nicely with
the work of Stalnaker (1978), Kratzer (1991), and especially Portner (2004, 2011).
Portner argues that verbal mood, which is another name for what I call sentential
mood, determines how a sentence is supposed to update the information structure
of a conversation. Sentences in the declarative mood, for example, are supposed to
update the Conversational Background (i.e. the set of propositions taken for granted
in a conversation). Recall from the previous Chapter (x3.1) that in addition to the
Conversational Background, there is an Ordering Source (the set of propositions used
in a conversation to rank-order the set of possible worlds which are consistent with
the Conversational Background) and each conversational participant is assigned a
To-Do List. Recall also from the previous Chapter that, on Portner's account, To-Do
Lists are sets of properties. But they can just as well be sets of propositions instead,
as I argued (x2.3.3).
21
. Importantly, sentences in the imperative mood conversa-
tionally function to update To-Do Lists. While the primary role of Conversational
Backgrounds, Ordering Sources, and To-Do Lists is in the interpretation of modal
declaratives, what I want to point out is that To-Do Lists and Conversational Back-
grounds can be thought of as modeling the intentions and shared beliefs, respectively,
of conversational participants. So sentences in the declarative mood are supposed to
update the beliefs of conversational participants, while sentences in the imperative
(i) Sweep the
oor.
Jill's speech act immediately and directly provides Jack reason to intend to sweep the
oor. By
'immediate and direct' I mean the speech act itself gives or presents one with a reason for accepting
the sentence used in such a way that the reason can be grasped without need for re
ection (i.e. non-
inferentially). At this stage I don't want to commit to what exactly the character of these reasons
is, but it might be something as simple as because Jill said so. Compare this to (11-c). Suppose Jill
instead says this to Jack. I think it's clear that this speech act does not give Jack an immediate
and direct reason to sweep the
oor. Nor does it give him an immediate and direct reason to clean
the bathroom. But suppose Jack knows that Jill generally prefers that he not sweep the
oors,
because he always does a lousy job. In light of this, it now does appear that Jill's ordering of (11-c)
gives him reason to clean the bathroom. This is all well and good, but notice that this reason is
neither immediate nor direct: it came from Jack re
ecting on the speech act and its relation to
his background knowledge. In any case, it's important to note that what we are discussing here is
the role of imperatives in speech acts. We are not saying anything of substance about imperatives
themselves or sentential mood; the theory is not that the imperative mood is what provides reasons.
The theory is that uses of unembedded imperatives in directive speech acts { orders, commands,
advice, etc. { immediately and directly provide reason to accept the imperative. A similar story
will go for unembedded declaratives used in assertive speech acts and unembedded interrogatives
used in inquisitive speech acts. I further discuss the relation between imperatives and reasons in
Chapter 5.
21
Cf. Charlow (2011, Ch. 2)
112
mood are supposed to update their intentions. We model this process with Conver-
sational Backgrounds and To-Do List. If all goes well, the models will match reality,
though they need not. It's therefore not a stretch to think of acceptance-rules and
modeling of conversations as two sides of a single coin.
Finally, this view is capable of explaining the non-representationality or non-
factuality (in the sense of Palmer (2001)) of the imperative mood. Imperatives aren't
take to represent or be true in any substantive sense not because of their core mean-
ings, but because intentions, the mental states involved in accepting imperatives,
are not representation mental states. Intentions don't represent the world as in fact
being one way or the other. Rather, an intention leads the agent whose intention
it is to make the world some way; it causes the agent to change the world. That
just is part of the nature of intentions, and what distinguishes intentions from rep-
resentational mental states like belief.
22
In the terminology of Searle (1983), while
beliefs have mind-to-world direction of t, intentions have world-to-mind t. Ar-
guably, mind-to-world t is necessary for a mental state being representational in
a substantial sense. So having world-to-mind t means intentions are substantially
non-representational, and it is in virtue of this that imperatives are (derivatively)
substantially non-representational as well.
In terms of syntax, that this non-representational aspect of imperatives is
the (non-semantic) product of the imperative mood, hence of IMP, means IMP is
marked for something like [{factuality], that is, it is marked as not carrying any
representational/factual feature. Like (arguably) gender features (e.g. [+feminine]),
however, neither [+factuality] nor [{factuality] is an interpreted feature, meaning
they do not contribute to a construction's interpretation. Rather, they serve extra-
semantic functions, such as indicating to a hearer to engage their intentional system.
That IMP marks a construction for [{factuality] is important for the nal topic of
this chapter: reporting uses of imperatives.
3.5 Reporting Imperative Use
In English, imperatives cannot embed as the objects of verbs that take clauses as
objects, that is, they cannot embed as subordinate clauses.
(12) a. *Jack said [
CP
go to the store].
b. *Jill accepts [
CP
look for the dog].
22
There is a hard question here about exactly how intentions cause action. There is a sizable
literature on this topic, and I will not be able to do wade into it here. For those interested, a good
place to start is Harman (1976, 1986), followed by Mele (1992) and Kapitan (1995).
113
c. *Jerry fears [
CP
change the tire].
But in other languages it appears imperatives do embed. See Portner (2007) for
Korean and Rus (2005) for Slovenian.
If imperatives don't embed in a given language, does that mean uses of them
cannot be reported? Yes and no. When reporting a use of an imperative in English,
the imperative itself is not used in the report { another kind of clause has to be
used. But we have to be careful to distinguish reporting a use of an imperative from
reporting what someone said in using an imperative. Importantly, what we care about
in reports is what was said, not what sentence was uttered. Moreover, we can report
an imperative having been used
(13) Jill said the imperative `Go to the store' to Jack.
But such a report is ambiguous between what is being report. (13) either is merely
saying that Jill mentioned the imperative to Jack, or is saying that actually used the
imperative in, say, the issuance of an order or suggestion. It is this latter reading
that we care about, that is, in reporting uses of imperatives what we care about is
what was said, what was expressed by its semantics.
Setting aside constructions like (13), the question to ask is, can what was said
in a use of an imperative be accurately reported in languages in which imperatives
don't embed? Any competent speaker of English will think that it can.
23
Typically
when we want to say that someone has used an imperative, in a command or some
other directive speech act, we report this with innitive or subjunctive clauses. Sup-
pose Jack's mom says (14) to him in an order. A natural and accurate way to report
this is with (15-a) or (15-b).
(14) Don't waste time on the internet.
(15) a. Jack's mom told him not to waste time on the Internet.
b. Jack's mom ordered that he not waste time on the Internet.
The subordinate clause in (15-a) is an innitive, since the main verb of the clause is
preceded by `to'. And we can tell that the subordinate clause in (15-b) is a subjunctive
because of the lack of a supportive occurrence of the verb DO, which would be
grammatically required were the clause a declarative, as can be seen by comparing
the grammaticality of (16-a) and (16-b).
(16) a. Jack's mom believed that he did not waste time on the Internet.
23
See alsox2.4.3 of the previous chapter.
114
b. *Jack's mom believed that he not waste time on the Internet.
In order for (15-a) and (15-b) to be accurate and true reports, the subordinate
innitive and subjunctive, respectively, must express what (14) expresses. That is,
they have to have the same satisfaction-values and express the same propositions
that (14) has and does. If Na ve Satisfactionism is right, then this means that
innitives and subjunctives also express those sets of worlds in which the subject
performs the act (or instantiates the property) specied by the predicate.
That Na ve Satisfactionism plausibly extends to innitives and subjunc-
tives can be argued for as follows. Consider again (17) and (18).
(17) Jill wants Jack to go to the store.
(18) Were Jack to fall, he would break his crown.
What (17) roughly says is that Jill wants it to be the case that Jack goes to the
store, while (18) says that if it were the case that Jack falls, then it would be the
case that he breaks his crown. Put another way, (17) is true i Jill stands in the
desiring relation to the set of worlds in which Jack goes to the store. Thus Jill's
desire is sated i the actual world is a member of this set { thus i the innitive
clause is satised. Similarly, (18) is true i (for some appropriately restricted set of
worlds
24
) every world is either not a member of the set of worlds in which Jack falls
or it is a member of the set of worlds in which Jack breaks his crown. Thus (18)
is true i every world is such that either the antecedent-subjunctive is not satised
in it or the consequent-subjunctive is satised in it. In this way we see why Na ve
Satisfactionism is perhaps even more plausible when it comes to innitives and
subjunctives than when it comes to imperatives.
Despite imperatives, innitives, and subjunctives all being semantically iden-
tical, we can still distinguish them syntactically, and these syntactic dierences are
what explain why innitives and subjunctives can be embedded while imperatives
can't. The critical factor (in English) is IMP, which imperatives have but innitives
and subjunctives lack. Something about IMP prevents imperatives from being the
direct objects of speech or attitude verbs like SAY, TELL, WANT, and the like.
It can't be the [{factuality] feature, because innitives and subjunctives are them-
selves non-representational and non-factual, as is drawn out by the fact that you
can't make assertions with either innitives or subjunctives. A speaker has to use
declaratives if they want to represent something as being some way to their hearers.
So there has to be some further feature that IMP carries and which explains why
24
See Lewis (1973).
115
(English) imperatives can't be the objects of speech or attitude verbs. I'm not sure
what this feature is, but one idea is that clauses are marked according to the mental
states they are supposed to engage. If that's right, then imperatives are marked for
[+Intention]. Innitives and subjunctives, by contrast, are not marked for this. In
fact, these clause-types don't appear to correspond to any particular mental state {
not intention, desire, belief, or wonder. This would explain why in many languages
subjunctives and especially innitives cannot be used as independent clauses; by and
large they are only usable as dependent clauses.
There are of course qualications that need to be made here (e.g. in order to
accommodate constructions like (18)), especially considering that in some languages,
like Latin, subjunctives are regularly used as independent clauses. But the general
idea holds: imperatives, innitives, and subjunctives are semantically identical, so in-
nitives and subjunctives can be used in accurate reports of uses of the former, and
what explains the asymmetries in the embeddability of these dierent clause-types
will ultimately be dierences in their syntactic features. Extending Na ve Satis-
factionism to innitives and subjunctives, this means that these clauses are also
satised i their subjects perform the acts (or instantiate the properties) specied
by their predicates { i their simple declarative counterparts are true.
3.6 Summary and Remaining Issues
I have argued for a compound view that so far has two parts (to give it a name, let's
call the proposed view of sentential mood Mood Acceptancism):
Na ve Satisfactionism (i) Imperatives (and innitives and subjunctives) are ei-
ther satised or unsatised, and (ii) the conditions under
which an imperative is satised are the same conditions
under which its simple declarative counterpart is true.
Mood Acceptancism (i) Sentential moods do not contribute to the interpre-
tations of sentence, but rather (ii) they determine what
one should do in accepting a given sentence (e.g. believe,
intend, or wonder about the proposition the sentence ex-
presses).
As it stands, this compound view is able to meet the desiderata FI, No Truth,
Mood6=Force, Action Link and one part of Embedding
It meets FI because it denies that mood morphemes are lexical or functional
elements of sentences { it denies that morphemes like IMP are interpretable, and
116
instead holds that the only interpretable elements of imperatives are their subjects
and predicates. Since Na ve Satisfactionism does fully interpret the subjects and
predicates imperatives, FI is thereby met.
No Truth is met due to our denying that imperatives represent in a substan-
tial way and thereby maintaining that they merely have satisfaction-values. There
is of course a serious question about the relation between the satisfaction-values of
imperatives and the truth-values of declaratives. Are these the same thing at the
end of the day? Or can we uphold a distinction between them? I will return to an-
swer this in Chapter 5, where I will argue for the surprising thesis that there is no
substantive semantic dierence between imperatives and declaratives (or any other
sentence-type for that matter).
By linking imperatives up to intentions, Action Link is met. Imperatives
lead to action because uses of them give rise to intentions. Intentions are paradigm
motivational mental states. I intend that p only if I am moved to bring it about that
p. And provided that this motivation is suciently strong and there are no barriers
preventing me, I will bring it about that p. This is just how intentions work. There
are, however, some lingering questions. First, why one would form an intention that p
on the basis of some someone uttering an imperatives with p as its content, especially
in those cases where the utterer lacks any kind of authority? Second, why is it only
ever the addressee that forms or entertains forming the intention (exception in cases
of confusion or vague address)? In the case of declarative use, it seems appropriate
for anyone to form a belief in its content? So why is it only appropriate for the
addressee to form an intention about the content of an imperative? I will return to
these and related questions in the nal chapter, when we turn to look more closely
at the pragmatics of imperatives.
As for Mood6=Force, this is met because we have neither con
ated sentential
mood and illocutionary force, nor denied that sentences have moods. Illocutionary
force, as the name suggests, is a feature of speech acts, of uses of linguistic construc-
tions. Mood Acceptancism rejects that mood determines or is otherwise involved
in how a sentence is used. It instead looks further downstream to what one does {
or should do { in accepting (a use of) a sentence. Thus mood cannot be con
ated
with illocutionary force on this view. Importantly, by focusing on acceptance, we
don't have to worry about how or where a sentence is used in order to account for its
mood. The acceptance-rules determined by a sentences mood will stand regardless
of the kind of speech act it is used in, if it is used in any speech act at all, and of
whether the sentence is embedded or not. Mood Acceptancism thus makes it easy
to defend the existence of sentential mood.
Next is clause (ii) of Embedding, which says that what a theory says is
117
the meaning of imperatives should be able to enter into accurate reports of people's
speech and attitudes. The way this condition is met is by extending Na ve Satis-
factionism to innitives and subjunctives. Innitives and subjunctives can be the
objects of speech and attitude verbs, and more importantly, these are the clause-types
we use when reporting uses of imperatives, as in (15). So insofar as innitives and
subjunctives are semantically identical to imperatives, as I argued they are, clause
(ii) of Embedding is met.
Meeting clause (i) of Embedding, however, is more dicult. Na ve Satis-
factionism and Mood Acceptancism are not sucient for meeting these desider-
ata. Clause (i) of Embedding isn't met because we don't yet have an adequate way
to account for mixed compound sentences. And Action Link isn't met because
though we have connected imperatives up with intentions (via sentential mood), we
still have not explained why imperatives lead to intentions and thereby action.
Thus, meeting clause (i) of Embedding and Action Link will require further
complicating the view. These complications are the subject matter of Chapters 4
and 5. Chapter 4 takes the semantic equivalence between imperatives, innitives,
and subjunctives one step further and argues that in order to solve the problem
of mixed compound sentences, we need to identify the truth-values of declaratives
with the satisfaction-values of imperatives (and innitives and subjunctives). Finally,
Chapter 5 argues that the link between imperatives and action is secured not through
semantic or even pragmatic mechanisms, narrowly conceived, but rather through
imperatives engaging one's agency and bringing to the fore the reasons one has to
perform the acts imperatives specify; in other words, two necessary conditions on
accepting an imperative and thereby forming an intention are having one's agency
engaged and having reasons to perform the relevant act. It is through these channels
that acceptance unfold and the link to action is established.
One important result of this and the next two chapters is a comprehensive
view according to which there is absolutely no substantive semantic dierence between
declaratives and imperatives (or any other clause-types). All the dierences between
these clause-types reside either in their syntax or in their conversational functions.
This of course means that attempts to dierentiate declaratives and imperatives at
a semantic level (as Portner (2004) attempts to do) are misguided. That there are
no substantive semantic dierences between imperatives and other sentence-types,
especially declaratives, is ultimately the main thesis I am arguing for in this disser-
tation.
118
Chapter 4
The Problem of Mixed Compound
Sentences
4.1 The Problem
It's well known that imperatives can enter into compound sentences (i.e. sentences
formed using words and phrases like `and', `or', `only if', etc.) not just with other
imperatives, but with declaratives as well. Here are some examples of what we might
call pure compound sentences involving imperatives:
(1) a. Go to the store and pick up some milk.
b. Look through the catalogue, but don't order anything.
c. Eat your vegetables or don't eat them.
d. Clean your room or clean the garage.
And here are some compound sentences that are a mix of a declarative and an
imperative:
(2) a. If you are hungry, have an apple.
b. Drink soda pop only if your doctor says it is okay.
c. Go around the back and I will go around the front.
d. Either it is raining or go outside.
Pure compound sentences do not cause too much trouble for theories of imperative
meaning, though they may cause issues for the semantics of connective words and
phrases like `and' and `or'. For if imperatives are of a dierent semantic type than
declaratives, as many theorist suppose, then it might be the case that connective
119
terms are ambiguous between a logical connective that conjoins the meanings of
declaratives and one that conjoins the meanings of imperatives. This amount of
ambiguity doesn't seem too problematic.
The real problem for theories of imperative meaning is mixed compound sen-
tences. The root of the problem is that imperatives appear to be of one semantic
type (call it type i) while declaratives are of another (type t).
1
So what then are the
semantic types of compound sentences like those in (2)? Call this The Problem
of Mixed Compound Sentences.
To draw the problem out, let's consider Casta~ neda's semantics of mixed com-
pound sentence, which holds that imperatives and declaratives are of two distinct
semantic types. The rst question to answer is: what is the semantic type of a mixed
compound sentence. Casta~ neda seems to endorse the following principle.
Imperative Priority A compound sentence is of the imperative type if at least
one sentence making it up is an imperative.
(Casta~ neda 1975, p. 115)
Given this principle, all of the sentences in (2) are imperatives. In order for this to be
so, however, words like `and', `or', `only if', etc. must express yet other logical connec-
tives in compound sentences like those in (2) than what they express in compound
sentences made up of sentences of the same type. This is because like all words and
phrases, those which express logical connectives have a certain semantic type. Tradi-
tionally, this type is taken to behht; ti; ti. What this says is that if you give a logical
connective an order-pair of sentences of type t, the result will be a sentence of type
t. Because Casta~ neda accepts Imperative Priority he has to deny that this holds
for all compound sentences. He may grant that compound sentences made up of only
declaratives is of type t. But when imperatives are involved, compound sentences are
of type i. Because of this it cannot be the case that the words and phrases used to
form compound sentences are of typehht; ti; ti, at least not necessarily. They will be
of this type in declarative-only compound sentences, sure. But in mixed compound
1
Some recent theorists who are committed to imperatives and declaratives being of distinct
semantic types include Portner (2004), Boisvert and Ludwig (2006), Vranas (2008), and Mastop
(2011), and so each of their views face this problem. It's worth recalling that for Portner, unlike
declaratives, imperatives are not a basic semantic type. They are rather the derived typehe;ti.
Mastop and the rest, by contrast, hold that imperatives are of one basic semantic type while while
declaratives are of another. Importantly, the problem arises no matter what the distinct types are;
all that matter for generating it is that the semantics types for declaratives and imperatives are
distinct.
120
sentences they will have to be of a type that outputs something of type i. Thus `and',
`or', `only if', etc. will be systematically ambiguous across pure and mixed compound
sentences.
The problem is worse than it may at rst appear. One may think that the
ambiguity involved isn't so bad. After all, all we have to do is say that `and' and
the rest are of typehht; ti; ti in the declarative-only case, and of typehht; ii; ii in
the mixed case, right? Wrong. We also need a type for imperative-only compound
sentences:hhi; ii; ii. And we need a dierent type for mixed compound sentences in
which the imperative occurs rst, such as (2-c) and (2-d): namely, typehhi; ti; ii.
All in all, then, `and' and the rest will be ambiguous between logical connectives
of at least four dierent types:hht; ti; ti,hhi; ii; ii,hht; ii; ii, andhhi; ti; ii. And once
we throw interrogatives into the mix this number will balloon to as large as nine
dierent types. This much ambiguity is too much.
One might try to mitigate the damage to Casta~ neda-like views by holding
that, although the sentences in (2) appear to be formed out of a declarative and an
imperative, they are really formed out of two sentences of the same type, which in
eect reduces mixed compound sentences to pure compound sentences. The most
plausible way of implementing this idea is to maintain that `and', `or', and the rest
are ambiguous between only two types:hht; ti; ti andhhi; ii; ii. Implementing this will
require expanding on an idea from Partee and Rooth (1983). It goes like this. When
two sentences being conjoined by `and', `or', and the rest are not of the same type, a
semantic type shift is triggered: the conjoined subsentence of type t is shifted to one
of type i. Roughly, this means that in mixed compound sentences, the declarative
is forced to become an imperative. The result of this type-shifting is that mixed
compound sentences are not really mixed at all: despite outward appearances to
the contrary, sentences like those in (2) are made up of two imperatives. Moreover,
this means that in such sentences `and' and the rest are of typehhi; ii; ii, which
reduces the amount of ambiguity these terms must be subjected to by at least half.
(If we think compound sentences can also be of typehhn;ni;ni, then the amount of
ambiguity will be reduced by two-thirds.)
This strategy is not without costs. It requires some perfectly valid arguments
to be counted as invalid. To see why, consider the following argument which should
be unimpeachable.
(C) Jack will go around front and you, Jill, go around back.
Therefore, Jack will go around front.
If we follow the type-shifting strategy, this argument rests on an equivocation. In
the premise, the sentence `Jack will go around front' is an imperative and of type i.
121
But in the conclusion, since it is not being conjoined with an imperative, `Jack will
go around front' is a declarative and of type t. Thus this sentence has one meaning
in the premise and a dierent meaning in the conclusion. Surely any argument that
equivocates in this way is invalid, in which case the type-shifting strategy makes the
wrong prediction about the validity of (C).
4.2 Solving the Problem: Strong Na
ve Satisfac-
tionism
By itself Na ve Satisfactionism is not confronted by the Problem of Mixed
Compound Sentences. Recall that the problem is a problem for views which say
imperatives and declaratives are of distinct semantic types. Na ve Satisfaction-
ism, however, is silent about the semantic type of declaratives; it is just a theory
about the semantics of imperatives. The problem becomes a problem only when we
combine Na ve Satisfactionism with a view according to which the logical values
of declaratives are not satisfaction-values. Call this combined view Weak Na ve
Satisfactionism.
Weak Na ve Satisfactionism (i) Imperatives are either satised or unsatis-
ed, (ii) the conditions under which an imper-
ative is satised are the same conditions under
which its simple declarative counterpart is true,
and (iii) truth-values6= satisfaction-values.
This is to be contrasted with:
Strong Na ve Satisfactionism (i) Imperatives are either satised or unsatis-
ed, (ii) the conditions under which an imper-
ative is satised are the same conditions under
which its simple declarative counterpart is true,
and (iii) truth-values = satisfaction-values.
In eect, Weak Na ve Satisfactionism says that the interpretations of impera-
tives and declaratives are almost but not quite identical, such that (3) is true but (4)
is not.
(3) JGo to the storeK JYou go to the storeK
(4) JGo to the storeK = JYou go to the storeK
122
Strong Na ve Satisfactionism, by contrast, says that the interpretations of
imperatives and declaratives are identical, thereby making both (3) and (4) true.
Avoiding the Problem of Mixed Compound Sentences requires iden-
tifying truth-values and satisfaction-values. Thus, given Na ve Satisfactionism,
this means accepting Strong Na ve Satisfactionism. With this we have type t
= type i and now we can say that the type of `and', `or', and the rest ishht; ti; ti {
or, if one prefers, typehhi; ii; ii, since these now amount to the same thing.
But wait! Isn't it enough that satisfaction-values and truth-values fall under
the same general type and so we don't need to identify them? Suppose this is right
and satisfaction-values and truth-values are both kinds of fulllment-values (I borrow
this terminology from Boisvert and Ludwig (2006), who propose exactly this view).
Then what I would point out is that, as far as semantics is concerned, only fulllment-
values matter and (4) will still be true. This is because if having satisfaction-values
versus having truth-values makes any semantic dierence at all, the Problem of
Mixed Compound Sentences will return in full force. For the only dierence we
have made is the introduction of yet another semantic type { call it f { which will
be the type of at least mixed compound sentences and which will be distinct from
types t and i. From here it's not hard to see how all the same issues we have already
discussed will arise. So the answer to the original question is no, satisfaction-values
and truth-values both falling under the same general type is not enough to solve
the problem. Solving the problem demands that imperatives and declaratives have
identical logical values. It demands Strong Na ve Satisfactionism.
So why does commonsense pull us to the thought that truth-values6=satisfaction-
values? My diagnosis is that this is just another Frege Puzzle. Frege Puzzle's arise
in cases where a person has two (or more) ways of thinking about a single thing,
but they don't know or aren't otherwise aware of this identity. For example, I think
of a single individual rst under the name `Clark Kent' and then under the name
`Superman', but because of this individual's clever dressing patterns I don't believe
that Clark Kent is Superman. To put it in terms of sentence-patterns, (5-a) and (5-b)
are true while (5-c) is false.
(5) a. Clark Kent is Superman.
b. I believe that Superman can
y.
c. I believe that Clark Kent cannot
y.
This is a puzzle because if (5-a) is true, as it is, then I believe of one and the
same object that it both can and cannot
y. But I don't take myself to believe a
contradiction!
There are a number of proposed solutions to Frege Puzzles. Frege proposed
123
that the objects of belief { what we believe { are Thoughts, and dierent Thoughts
have dierent cognitive signicance. So the Thought that Superman can
y and the
Thought that Clark Kent can
y, being dierent Thoughts, have dierent cognitive
signicance. This is why I can believe the one Thought while believing the negation
of the other without realizing that they are both about the same individual (Frege
1948). A separate yet related view is that proper names have descriptive contents
and, importantly, dierent names have dierent descriptive contents. So `Superman'
describes an individual in one way and picks out the individual through this de-
scription, while `Clark Kent' describes the same individual in a dierent way and
picks out the individual through this other description (Russell 1910). More recent
versions of this view hold that the description is a metalinguistic one (e.g. being
the bearer of the name `Superman') (Russell 1919; Kneale 1962; Burge 1973; Bach
1981; Geurts 1997). Importantly, I might not realize that the two descriptions pick
out the same individual. This is how I can accept the sentence `Superman can
y'
while denying the sentence `Clark Kent can
y', because I don't realize that `Super-
man' and `Clark Kent' involve distinct descriptions of one and the same individual.
Both Frege's solution and the descriptivist one share the feature that I don't strictly
speaking believe a contradiction when I believe that Superman can
y but Clark
Kent cannot. A still other view, sometimes called Millianism, holds that I do in
fact believe a contradiction. This view holds that the content of `Superman' and of
`Clark Kent' just is the individual, and so I literally believe of that individual that he
can
y and that he cannot. To explain how this can happen, defenders of this view
argue that while we do directly think of individuals when we use proper names, we
nevertheless do so under a way of thinking (Salmon 1986; Soames 1989). I can think
of the same individual in a Superman-way and in a Clark Kent-way. Importantly,
however, these ways in which we think of things do not enter into the contents of
our thoughts. Again, the content of `Superman' and of `Clark Kent' is all and only
the individual. Ways of thinking about an individual are modes of thought, not the
contents of thought.
I do not wish to take a stance about the proper way to solve Frege Puzzles.
My aim is instead to point out that there can also be dierent ways of thinking about
satisfaction-values. One way to think about them is under the way associated with
the terms `satised' and `unsatised'. But another way is that associated with the
terms `true' and `false'. (Similarly for whatever the terms are that we use when talking
about interrogatives.) These dierent ways arise because the conversational functions
of imperatives and declaratives dier (chapter 3,x3.4). Imperatives primarily function
to tell others what (to intend) to do; declaratives to tell others what to believe.
Thus when dealing with imperatives we think of satisfaction-values in a practical,
124
non-representational way, but when dealing with declaratives we think of them in a
representational, non-practical way. Despite these dierent ways of thinking, like in
the `Superman'/`Clark Kent' case, we are thinking about one and the same thing.
Also like with `Superman'/`Clark Kent' case, even when we learn of this identity
through trying to solve The Problem of Mixed Compound Sentences, that
does stop us from continuing to think of satisfaction-values in dierent ways { nor
should it. But don't let these dierent ways of thinking fool you. Imperatives and
declaratives have the same logical values.
4.3 Some Consequences of Strong Na
ve Satis-
factionism
There are some important consequences of the view we have arrived at. First, sentence
types cannot be distinguished semantically. Imperatives and declaratives without a
doubt are distinct sentence types. However, semantically they are indistinguishable.
So what then is the distinguishing factor? It is what we discussed in Chapter 1: each
has in its syntax a distinct mood morpheme or feature that makes it the type of
sentence it is but which has no semantic function. In other words, sentences are syn-
tactically marked for distinct moods, but contrary to popular belief, these mood mor-
phemes or features don't contribute to a sentence's interpretation. Moreover, since
these morphemes/features are what distinguish sentence types, it make the claim
(also made in Chapter 1) that they are located in the highest level of a sentence's
syntactic structure, which is the C(omplementizer)P(hrase) layer, more plausible.
2
CP
C
0
I(n
ection)P
I
0
V(erb)P I
0
SpecIP
(Subject)
C
0
(Mood)
SpecCP
2
cf. Han (2000, ch. 4), who also locates mood features in C
0
.
125
Thus we are holding on to the idea that all sentences share the same basic syntactic
structure, only diering in what occupies certain nodes. Moreover, as I argued in
the last chapter, though mood morphemes make no semantic contribution, they do
play a conversational role in determining the acceptance-rules of a sentence, which
in turn determines the dierent conversational functions of dierent sentence types
(which in turn leads us to think of satisfaction-values in dierent ways, giving rise
to Frege Puzzles).
Another, more surprising consequence of Strong Na ve Satisfactionism
is that, together with a plausible denition of validity, it implies that arguments like
the following are valid.
(D) Jack, if you are hungry, eat something.
Jack is hungry.
Therefore, Jack eats something.
(E) Jack, go to the store.
Therefore, Jack goes to the store.
(F) Jack goes to to the store.
Therefore, Jack, go to the store.
The denition of validity being relied on here is: an argument is valid i the conclusion
must be satised if the premises are (i.e. there is no interpretation on which the
premises are satised and the conclusion unsatised).
I suspect many will have the impression that the foregoing arguments are
invalid. This impression has its source in the fact that these arguments involve swap-
ping an imperative-premise for a declarative-conclusion, or a declarative-premise for
an imperative-conclusion, and so the conclusion is not \contained in the premises".
To correct this we might try to add a clause to our denition of validity:
Validity
S
An argument is valid
S
i (i) the conclusion must be satised if the
premises are, and (ii) any non-compound sentence in the conclusion
occurs somewhere in the premises.
Since imperatives and declaratives are dierent types of sentences, none of the forego-
ing arguments will be valid
S
. Unfortunately, neither will arguments like the following:
(G) Jack is hungry.
Therefore, Jack is hungry or Jack is tired.
126
This argument has an atomic sentence in its conclusion that does not occur in the
premise. Clause (ii) of Validity
S
thus invalidates the classical inference rule Dis-
junction Introduction. Worse still, it also invalidates the classical inference rules
for quantiers { Existential Introduction, Existential Elimination, Universal
Introduction, and Universal Elimination { because a sentence with a quanti-
er phrase in subject position is dierent from a sentence with a name in subject
position. Consider:
(H) Jack ate the cake.
Therefore, someone ate the cake.
(I) Everyone ate the cake.
Therefore, Jack ate the cake.
For both of these arguments, the sentence in the conclusion does not occur in the
premises. The upshot: making validity sensitive to what sentences occur in an argu-
ment, and where, is not the way to go.
3
But if validity is not sensitive to what sentences are in an argument, and it is
certainly not sensitive to conversational function, what is preventing the arguments
Strong Na ve Satisfactionism predicts to be valid from being valid? Look at it
this way. Often discussions of logic begin by rejecting the idea that arguments are, in
the rst instance, composed of sentences. Instead, they are thought to be composed of
propositions, or at the very least Logical Forms (LFs), which are clothed in sentential
garb (or ways of thinking). According to this understanding, our alternative denition
of validity rests on a fundamental confusion. It tries to (partly) ground validity
in the sentences occurring in an argument. But this makes no sense if arguments
are made up of propositions or LFs. So we have reason to reject this denition
independent of the poor predictions it makes. But if arguments are made up of
propositions or LFs, then before we can evaluate the arguments in question we need
to make their propositional/logical structure clear. According to Strong Na ve
Satisfactionism, argument (D) has the LF
If P, then Q
P
Therefore, Q
3
Cf. Geach (1958), who objects to Hare's attempt to make validity sensitive to what sentences
occur in an argument.
127
which is just an instance of Modus Ponens. And the arguments (E) and (F) both
have the form
P
Therefore, P
which is unassailable. Viewing things in this way, that arguments like (D)-(F) are
valid is not surprising. (For convenience, in what follows I will sometimes continue
to speak as if arguments are made up of sentences. The reader is free to replace all
sentence-talk with the appropriate proposition- or LF-talk.)
We are, however, still owed an explanation of why intuition pulls us towards
judgments of invalidity. Frege Puzzle considerations can't help us here. Whether or
not truth=satisfaction, anyone can readily see that `Go to the store' is satised i
`You go to the store' is true, in which case (D){(F) are valid (assuming an argument
is valid i its conclusion must be satised or true if its premises are). But despite
this we still think that there is something wrong with arguments (D){(F).
The explanation I prefer is an epistemic one. It is largely agreed that imper-
atives seem to be governed by a norm which says one is to use an imperative only
if it is not known whether the subject is going to perform the relevant act or not
(Kaufmann 2012, pp. 155-157).
No-Knowledge Norm Use imperative (at a time t) only if you don't know or
are otherwise uncertain (at t) whether the subject of
is going to perform the act specied by 's predicate.
That a norm like this governs imperatives can be seen by comparing it to the Knowl-
edge Norm governing uses of declaratives.
Knowledge Norm Use declarative (at a time t) only if you know or are
otherwise certain (at t) that the subject of has the
property specied by 's predicate.
The argument for this norm involves re
ecting on Moorean Paradoxical constructions
like:
(6) Jack went to the store, but I don't know that he did.
This sentence is strange, to be sure. And it's strangeness doesn't appear to have
anything to do with it meaning. After all, both the conjuncts can be true. What's
more, the strangeness vanishes when we replace the indexical `I' with a third-person
term.
128
(7) Jack went to the store, but Matt doesn't know that he did.
One explanation for this is that the strangeness is purely pragmatic: though there
is nothing semantically or logically wrong with (6), it is inappropriate to use it in
conversation. The question then is why it is inappropriate to use. A common answer,
which I think is largely right, is that a use of it literally communicates to a hearer
that the speaker { referenced by the use of `I' { is violating Knowledge Norm. I
am using `Jack went to the store' and at the same time telling any hearers that I
don't know that the proposition expressed by this sentence is true. If this is true, I
have ipso facto violated Knowledge Norm.
4
Thus, this norm must be governing
uses of declaratives.
A similar argument can be used to show No-Knowledge Norm governs
imperatives. First we need a Moorean paradoxical construction:
(8) Go to the store, even though I know you won't.
This is just as odd and weird as (7). And just like (7) the weirdness can be removed
by replacing the `I':
(9) Go to the store, even though Matt knows you won't.
Importantly, however, the weirdness also dissipates when knowledge is traded for
ignorance:
(10) Go to the store, even though I don't know whether you will.
So unlike declaratives, the norm governing imperatives doesn't concern what one
knows. It concerns what one doesn't know, i.e. what one is ignorant of. Hence No-
Knowledge Norm. Why is it inappropriate to use (8) in conversation? It is because
a use of it literally communicates to a hearer that the speaker is violating No-
Knowledge Norm. A speaker uses `Go to the store' and at the same time tells
any hearers that they, the speaker, knows that the proposition expressed by the
imperative is false. The hearer takes this at face value and they register a violation
of No-Knowledge Norm. This violation explains the weirdness of constructions
like (8).
I suspect that No-Knowledge Norm { or something like it { is also behind
4
Things are actually a little more complicated than this. Strictly speaking, I have not used a
declarative. What I've used is a conjunction of declaratives. So we need a bridge principle or norm
that says: a use of a conjunction of declaratives amounts to uses of both conjuncts. Accordingly,
in using (6), I thereby use `Jack went to the store'. From here the derivation of the violation of
Knowledge Norm is the same as before.
129
our intuitions about the validity of arguments like (D){(F). The reason we want to
say these arguments are invalid is because when we see the imperatives, we immedi-
ately assume that their satisfaction-values are not known. And if it's not known, for
example, whether (11-a) is satised, then it also can't be known whether (11-b) is.
(11) a. Jack, go to the store.
b. Jack goes to the store.
But (11-b) is governed by Knowledge Norm. So when we see it we assume the
proposition expressed is known. But if it is known that (11-b) is satised, then it
can't be unknown whether (11-a) is. So arguments like (D){(F) seem to be deriving
knowns from unknowns and unknowns from knowns, which pulls us to think of these
arguments as being logically bad. But known-ness is not what's preserved by validity.
What's preserved is positive logical status, i.e. satisfaction/truth. And if (11-a) is
satised, then it must be the case that (11-b) is also satised { and vice versa. The
lesson is that we need to be careful to separate what's expressed by a sentence from
what it might communicate when used. A failure to do so makes us liable to close
o possibilities prematurely { such as the possibility that (D){(F) are valid after all.
4.4 Irrational Validity
Still, one might insist it is not rational to accept a declarative just because one has
accepted its imperative counterpart, even if the declarative is entailed by the imper-
ative (and vice versa). With this I completely agree. That the proposition expressed
by one sentence logically follows from the proposition expressed by another does not
guarantee it is rational to accept the second sentence on the basis of accepting the
rst. For example, it is not rational for one to accept (11-b) just because one has
accepted (11-a). I am therefore endorsing a controversial thesis.
Irrational Validity There are some valid arguments for which it is irrational
to accept the sentence expressing the conclusion solely
on the basis of having accepted the sentences expressing
the premises.
For some, this will be the most interesting claim made in this chapter, because I'm
in eect denying a thesis that is commonly thought to be correct, namely, that there
are no such arguments. In order to see why Irrational Validity holds, we need
to think back on the conversational function of sentential mood.
130
Recall from the previous chapter that the mood of a sentence determines
which region of one's mental life one is to engage in accepting the sentence. Roughly,
the imperative mood tells an agent to engage the intentional region of their mind
and thereby that accepting a sentence in the imperative mood means forming an
intention; and the declarative mood tells an agent to engage the doxastic region of
their mind and thereby that accepting a sentence in the declarative mood means
forming a belief. And so on for other moods that have cognitive import.
With this in mind, Irrational Validity holds for arguments like (E)-(F)
because the premises tell you to engage one region of your mental life, the conclusion
tells you to engage a dierent region, and the connection between the two regions is
not sucient for rationality. Take (E) for example, repeated here.
(E) Jack, go to the store.
Therefore, Jack goes to the store.
The premise of this argument is (expressed by) an imperative, whereas the conclu-
sion is (expressed by) a declarative. Thus accepting the premise means forming an
intention that Jack goes to the store. But accepting the conclusion means forming a
belief that Jack goes to the store. This is why it is irrational to accept the conclusion
solely on the basis of accepting the conclusion, because by itself intending a propo-
sition is not enough to make a belief in it rational.
5
For example, I might intend to
win the next lottery. Even with this intention, however, unless I have some other
information (e.g. that the lottery is rigged in my favor), it is not rational for me to
believe that I win the next lottery. Similarly, Jack might intend to go to the store,
but this alone does not make it rational for him to believe that he goes to the store.
He needs also to know where the store is and how to get there and he needs to know
of a means of getting there. Without at least this much additional information, it
would be irrational for Jack to believe that he goes to the store.
Compare this to valid arguments constructed out of sentences of a single type.
(J) Jack goes to the store or Jill goes to the store.
Jill doesn't go to the store.
Therefore, Jack goes to the store.
(K) Jack go the the store or Jill go to the store.
Jack don't go to the store.
Therefore, Jill go to the store.
5
This is not the same as, nor does it require, denying the thesis that if it is rational to intend p,
then it is rational to believe p. I'm not sure what to think of this later thesis, though I am inclined
to nd it suspect. Cf. Ross (2009).
131
Since it is irrational to believe an inconsistent set of propositions, if one believes
the propositions expressed by the premises of (J), then one had better also believe
the proposition expressed by the conclusion. It therefore must be rational to accept
the conclusion of (J) solely on the basis of accepting the premises. Similarly for (K).
Presumably, it is irrational to intend an inconsistent set of propositions. So it must be
rational to accept the conclusion of (K) solely on the basis of accepting the premises.
Another way of putting this point is by saying that one must, on pain of irrationality,
believe (intend) those propositions that logically follow from the propositions they
already believe (intend). Thus, in the case of valid arguments constructed out of a
single sentence type, not only is it rational to accept the conclusion solely based
on accepting the premises, but one is required to accept the conclusion if one has
accepted the premises.
It's worth stressing that though it is irrational to accept the conclusion of
an argument like (E) solely due to accepting the premises, this doesn't mean it is
irrational to do so partly in virtue of having accepted the premises. I mentioned
cases like this above. Here's another one. Suppose I accept the premise of (E), and
thereby intend for Jack to go to the store. Moreover, suppose I know this intention is
achievable, I know how to achieve it, and I know that I am incredibly reliable when
it comes to seeing that my intentions are achieved. Knowing all this together with
my intention for Jack to go to the store makes accepting the conclusion of (E) {
hence a belief that Jack goes to the store { rationally justiable. The same will go
for (F) if one thinks that more is required for an intention to be rationally justied
than a belief in its content. For example, following Part (1997), we might holds that
intending A on the basis of beliefs B
1
{B
n
is rational if were B
1
{B
n
true they would
provide sucient normative reason for intending to A and then Aing. But a mere
belief that you will A, were it true, does not provide sucient normative reason for
Aing. After all, A might be jumping out of a 4th story window of a building, and
surely the belief that you will jump out of a 4th story window, were it true, is not
sucient normative reason for intending to jump out of a 4th story window. More
is needed, such as the belief that there is something to break your fall and the belief
that you can jump to whatever it is. What this something more is we needn't decide
here. The view is simply meant to illustrate why it might not be rational to intend
a proposition p solely on the basis of believing p.
4.5 Why Not Go Dynamic?
I have argued that imperatives and declaratives must share the same logical values
if we are to solve The Problem of Mixed Compound Sentences. The view
132
resulting from this and the previous chapter is Strong Na ve Satisfactionism.
I have provided a diagnosis for why people have been led to thinking that impera-
tives and declaratives have distinct logical values, namely, that it's a Frege Puzzle.
And I have defended some of the more surprising consequences of Strong Na ve
Satifactionism. In particular, I have defended the consequence that arguments
like (E){(F) are in fact valid, which is what led many to think that imperatives and
declaratives must be semantically distinct.
6
The core of this defense is a view of
sentential mood, defended in the previous chapter, according to which the mood of
a sentence determines the mental conditions for accepting the proposition expressed
by the sentence.
The total view we have arrived at (Strong Na ve Satisfactionism+Mood
Acceptancism) shares features with dynamic accounts of meaning. An ever-growing
number of philosophers and linguists are recently turning to dynamic semantics. Ac-
cording to these accounts, the meaning of a sentence is an update potential, which
is a function either from one discourse structure to another (Kamp 1981), or from
one mental state to another (Heim 1983). Regarding imperatives in particular, Starr
(ming) puts forth a dynamic account. For Starr, imperatives function to update
preference orderings, such that an imperative like (12) results in conversational par-
ticipants mutually preferring states of the world in which the addressee drinks her
milk over states in which she doesn't.
(12) Drink your milk.
A somewhat dierent proposal, suggested by Charlow (2014), is that imperatives
result in an addressee updating her To-Do List. For example, (12) expresses a function
that takes an addressee from her original To-Do List to one with the proposition that
she drinks her milk on it (or the property of drinking one's milk).
The obvious dierence between my view and dynamic views is that on my
view the eects of a sentence on a discourse structure or the mental lives of the
conversational participants are not part of the semantics. They are, however, still
critically important to communication. There is no communication without these
eects, but rather just a bunch of noises or scribbles thrown out into the universe.
But isn't this enough to go in for a dynamic view? If the eects of language use
are essential to communication, why shouldn't they be captured in the semantics?
The simple answer is, once again, that we won't be able to solve The Problem
of Mixed Compound Sentences. The trouble this time is that declaratives and
imperatives (and other sentence types) each have dierent eects in a conversation.
6
See, for example, Sosa (1967, pp. 59-60).
133
The eect of a declarative is a new way of representing the world (i.e. a new discourse
structure) or a new belief. The eect of an imperative is a new way plan for changing
the world or a new intention. Dierent eects means dierent semantic types. Thus,
declaratives and imperatives have dierent dynamic semantic types { and we are
once again confronted with The Problem of Mixed Compound Sentences.
Starr (ming) recognizes this, and argues that the problem can be solved by
interpreting conjunction as a sequential update function and disjunction as a parallel
update function. Since it's simpler, let's focus on conjunction. Consider an unmixed
case rst:
(13) You changed the sheets and I washed them.
On Starr's view, what this roughly says is: rst update your beliefs so you believe
that you changed the sheets, then update your beliefs so you believe that I washed
them. The same will go for a mixed case, except we replace a belief update with a
preference update.
(14) Change the sheets and I'll wash them.
This roughly says: rst update your preferences so you prefer changing the sheets
over not, then update your beliefs so you believe that I will wash the sheets. In
this way we seem to be able to give conjunction a uniform dynamic interpretation:
conjunction just means rst update according to the rst conjunct (whatever it is),
then update according to the second (whatever it is).
It's important not to lose sight of an important feature of Starr's account,
which is shared by other dynamic accounts: the update functions expressed by declar-
atives are still of a dierent kind from those expressed by imperatives. On a Dynamic
Montague Grammar, this means assigning imperatives one dynamic semantic type
(the type of preference updates) and declaratives a dierent semantic type (the type
of belief updates). To keep things simple, let's uset andi again and say that declar-
atives are of dynamic type [t]![t], while imperatives are of dynamic type [i]![i].
In expressing a sequential update function, conjunction then will have a type of
the form [?]![?]![?]. The questions is, what to replace the `?'s by? Put another
way, what is the unied type of conjunction given Starr's account? It's not hard
to see that this question doesn't have an answer, because conjunction doesn't have
a unied semantics on Starr's account. In constructions like (13), conjunction will
be of type [t]![t]![t]. But in constructions like (14), it will be something like type
[i]![i]![t]. Mixed compound sentences in which the declarative comes rst will have
yet another type, and pure compound sentence with only imperatives will have still
134
another. Starr's account thus does not avoid the problem after all.
7
Can't the dynamic semanticist solve the problem in the same way that Strong
Na ve Satisfactionism does, by holding that imperatives and declaratives are of
the same basic semantic type? They cannot. The problem, as I've already men-
tioned, is that on the dynamic account semantic type is sensitive to the features of
the discourse context that a sentence functions to update. So saying that impera-
tives and declaratives are of the same type amounts to saying that they update the
same features of a discourse context: if imperatives update feature X, then declar-
atives update X as well { and vice versa. But this is precisely what theorists like
Starr deny, and for good reason. Imperatives do not function to update represen-
tational discourse structures or representational mental states, and declaratives do
not function to update preferential discourse structure or preferential mental states.
Both of these claims match our commonsense intuitions, but saying imperatives and
declaratives are of the same dynamic type requires rejecting one (or both) of them.
This is why we should not go dynamic, because solving The Problem of Mixed
Compound Sentences ultimately involves either rejecting that imperatives do not
result in representational mental states or rejecting that declaratives do.
7
Views that try to solve the problem by holding that (some) mixed compound sentences are
actually compound speech acts (Russell 2007; Kaufmann 2012) will fail for the same reasons. In
order for such views to work, they need to invoke certain Gricean reasoning: when we encounter a
mixed compound sentence, we use Gricean principles and background knowledge to infer that the
speaker is make a special sort of speech act. In order for such an inference to be possible, however,
the mixed compound sentence has to have a semantic content all its own from which the inference
proceeds. Hence, we need to gure out what the semantic contents of mixed compound sentence are
before the inference to the performance of a special sort of speech act can be made { at which point
we are right back to where we started, confronted with The Problem of Mixed Compound
Sentences.
135
Chapter 5
The Pragmatics of Imperatives
5.1 Recap: Imperatives and Intentions
This nal chapter is about the pragmatics of imperatives and will tie up remaining
loose ends. I have two chief aims. First, I want to argue that the dominance of
the 2
nd
-person imperatives is due to the pragmatics of imperatives. Second, I want
to argue that use of an imperative in a directive speech act gives a reason for the
addressee to intend the proposition expressed by the imperative. My starting point
for both of these arguments is the same: Mood Acceptancism and the claim that
to accept an imperative is to form an intention. The path to the rst claim { that the
dominance of 2
nd
-person imperatives is due to pragmatics { rests on a view about the
nature of intentions, namely, that they are essentially 1
st
personal. The path to the
second claim { that use of an imperative creates a reason to intend a proposition {
rests on the idea that directive speech acts are action-guiding. Taken together, these
arguments lay the groundwork for an explanation of why uses of imperatives lead to
intentions and thereby action (Action Link). But before we get to that, let's pause
to recap Mood Acceptancism and how it connects imperatives and intentions.
In chapter 3 I argued that sentential mood makes no semantic contribution
to a sentence and that its role is to determine what sort of mental state a hearer
should form in accepting it. This is Mood Acceptancism.
Mood Acceptancism (i) Sentential moods do not contribute to the contents of
sentences, but rather (ii) they indicate what one should
do in accepting a given sentence (e.g. believe, intend, or
wonder about the proposition the sentence expresses).
What this basically says is that sentential mood plays a strictly pragmatic or extra-
136
linguistic role in communication, in the broad sense of `pragmatic'.
1
One way to
think about this view is in terms of \tagging" a proposition. The lexical components
of a sentence determine the proposition that is expressed. Once that proposition is
determined, sentential mood steps in and puts a \tag" on the proposition indicating
whether it goes in one's belief-box, intention-box, or wonder-box. These tags have no
content in themselves. They do not have written on them \This goes in intention-
box" or anything like that. Rather, they are mere mental sorting devices, like the
bar-codes shipping companies use to automate the sorting of packages.
Mood Acceptancism is easy to misconstrue. One might think it is just
Hare's view about sentential mood repackaged.
2
. But this is incorrect. Hare held
that sentential mood does contribute to a sentence's content. Because of this he held
that sentences express truth-conditions (= propositions) together with something like
illocutionary force. This led him straight into the Frege-Geach Problem (see chapter
3,x3.2). By contrast, Mood Acceptancism explicitly rejects that sentential mood
makes any contribution to a sentence's content. Now one might wonder how tagging
a proposition is any dierent from expressing it together with illocutionary force.
After all, wouldn't the tag encode a certain \feel" just as much as illocutionary force
would? Not at all. Again, we need to think of the tags as serving the same function
as the barcodes used by automatic sorting machines, in that they are mere sorting
devices. When a sentence is uttered, (competent) language users scan the sentence's
1
The broad sense of `pragmatic' is the one concerning what happens after or what is done with
the content communicated { or `said', to use Grice's term { by a sentence in a a conversation.
Implicatures are pragmatic in the broad sense because they are derived from the content actually
communicated together with other contextually available information and norms (Grice 1975). By
contrast, the narrow sense of `pragmatic' is the one concerning indexicals and other context-sensitive
terms. These terms have a content only relative to a context and so in this way their content is
pragmatically determined. Thus narrow, or \near-side", pragmatics is plays a role in the information
a sentence communicates and to that extent is unlike broad, or \far-side", pragmatics, since the
latter has no role to play in content-determination. To be sure, this way of conceiving broad and
narrow pragmatics commits me to a certain way of conceiving the more general distinction between
semantics and pragmatics. As I am understanding the distinction, semantics is only concerned with
the content encoded into a sentence, while pragmatics is concerned with anything that goes beyond
this basic content, including the determination of the content of context-sensitive terms. I am aware
that this way of carving things up is controversial. Fortunately, nothing I say turns on understanding
the broad pragmatics/narrow pragmatics and semantics/pragmatics distinctions in the way that
I have. One could equally treat semantics as the study of the conventional associations of words,
grammar, etc. and pragmatics as the study of what can be inferred from the use of a linguistic item.
Granted, this would mean that the contribution of sentential mood is not purely \pragmatic", but
rather \semantic", in which case simply change the basic claim to: \The communicative role of
sentential mood is purely to indicate what one should do with the content of a sentence."
2
Hare (1952, 1971)
137
tag and thereby know that accepting the sentences means coming to form the ap-
propriate mental state about the proposition expressed. The contents of a package
do not become something dierent or gain anything just because a barcode has been
attached to the package. Similarly, the overall content of the sentence get enriched
beyond the proposition so expressed. This is why I said that the communicative
role of sentential mood is purely pragmatic, since it only concerns what we do with
communicated propositions.
With Mood Acceptancism in hand, the next question to ask is what sort of
mental states are involved in the acceptance of imperatives. I maintained in chapter 3
that to accept an imperative is to form an intention whose content is the proposition
expressed by the imperative. My reasoning is fairly simple. It goes like this. Uses of
imperatives lead to action. If the doctor says `Stick out your tongue' to me and I
accept this imperative (responding with an honest \Ok" or \Sure"), the result will be
me sticking out my tongue (provided there are no barriers, physical or psychological,
to my doing so). Moreover, my sticking out my tongue will be an intentional action:
I purposefully stick out my tongue. So the mental states involved in the acceptance
of an imperative have to be suitable related to intentional action. The most natural
candidates are intentions. Uses of imperatives lead to intentional action because to
accept an imperative is to form an intention, which in turn leads to the action in
question.
3
Connecting imperatives to intentions in this way is the rst step to explaining
certain interesting facts about how imperatives are commonly used in conversation.
I turn now to the rst of these facts.
5.2 Imperatives and the 2
nd
-Person
One of the most notable facts about the imperatives used in conversation is that,
across languages, imperatives with 2
nd
-person subjects are the norm while imper-
atives with 3
rd
-person and 1
st
-person subjects are rarely, if ever, encountered. 1
st
-
person imperatives in particular seem to exist in only very few languages (see chapter
1,x1.3.1). 3
rd
-person imperatives are more common, but the frequency of their use
still pales in comparison to 2
nd
-person imperatives. By contrast, the declaratives,
interrogatives, and other sentence-types used in conversation do not give preferential
treatment to any of the grammatical persons. So why are things dierent for imper-
atives? We rst encountered this question all the way back in chapter 1 (x1.3.1), but
had to set it aside. I now wish to answer it.
3
Casta~ neda (1975); Portner (2004); Starr (ming).
138
5.2.1 A Syntactic Explanation?
One might want to explain the dominance of 2
nd
-person imperatives in terms of
imperative syntax. Perhaps, this line of thinking goes, imperatives are marked for
the 2
nd
. This is the explanation of Zanuttini et al. (2012). They maintain that all
imperatives contain a functional head carrying a 2
nd
-person feature that the subject
of the clause must be checked against.
4
If the subject is not 2
nd
-person the sentence
is rendered ungrammatical. But the existence of both 1
st
-person and 3
rd
-person im-
peratives rules out a such an explanation. Put another way, Zanuttini et al. (2012)'s
view predicts that there are no 1
st
-person or 3
rd
-person imperatives, a predication
not supported by the cross-linguistic data.
Zanuttini et al. are aware of this data and try to accommodate it by saying
that these problem sentences are not in fact imperatives { they are some other, related
clause-type, which they label promissives and exhortatives, respectively. While they
do bring some interesting data to the table in support of this view, such as the
particles involved in Korean imperatives, it is hard to be convinced. Crucially, they
give no reason why imperatives cannot be anything but 2
nd
-person. Even if they
are right that imperatives are syntactically marked for 2
nd
-person, the immediate
question to ask should be why this is so. Given that imperatives are artifacts just
as much as any other piece of language, why would we make it such that they
are obligatorily 2
nd
-person, if they are? Couldn't we just as easily have placed no
restrictions on the subjects of imperatives, as is the case with other clause-types?
The syntactic approach to explaining the dominance of 2
nd
-person imperatives thus
only succeeds in pushing the question back.
5.2.2 The Essential 1
st
-Personality of Intentions
So what explains the dominance of 2
nd
-person imperatives? I think the answer lies
not on the language side of things, but the mind side. The imperatives used in con-
versation predominately have 2
nd
-person subjects because intentions are essentially
1
st
-personal.
Imperatives are primarily used to get others to do things. The doctor uses
`You stick out your tongue' or just `Stick out your tongue' to get you to stick out
your tongue. What the doctor wants is for you to perform an intentional action; he
wants you to intentionally stick out your tongue. Intentional actions are the products
of agency. To intentionally do something is to have an intention that causes you to
4
This head and the corresponding projection will be located somewhere between IP and CP in
an imperative's syntactic structure; see chapter 1,x1.3.6.
139
do it, and forming an intention requires employing one's agency. To intend to A (or
to intend to B which practically requires Aing) is to intend myself to A (or B). My
agency has no direct control over anyone else's actions but my own. I can intend for
you to A, but what this means is that I intend myself to see to it that you A. There
is no such thing as an intention that determines the agency of another. There are
intentions that have as their goal the guidance or manipulation of another's agency,
sure, but this is quite dierent from an intention that directly controls someone
else's agency.
5
. This is mostly because the process of intention formation necessarily
unfolds through one's own agency. As I argue elsewhere
6
, this means that intentions
themselves are essentially de se attitudes. Because of this, if the doctor's attempt
is to succeed, she will have to engage your agency, and in so doing lead you to
intentionally stick out your tongue (either by forming an intention to or by doing so
in order to satisfy some other intention you have, e.g. an intention to do whatever
the doctor says).
Suppose the doctor attempts to get me to stick out my tongue by instead
using a 3
rd
-person imperative, `Donald stick out his tongue', where `Donald' refers
to me.
7
Upon hearing this I will be confused. Is the doctor addressing me? Is there
someone else in earshot she might be addressing instead? If she is addressing me,
why is she doing so 3
rd
-personally? Now clearly this confusion can easily be resolved
by me asking if I'm the Donald she's talking to. But importantly the confusion will
be cleared up only after I either ask this follow-up question or at least infer that she
is addressing me (which let's assume I don't do in the present case). Moreover, in
order to clear up the confusion the doctor will have to respond to my question with
something like \Yes, I'm talking to you." The reason for this is the same as before: if
she wants me to do as she says { stick out my tongue { she needs to engage my agency.
And a way to do that is by using the 2
nd
-person. In using a 2
nd
-person pronoun like
`you' the doctor engages my agency. Once my agency is engaged, I can come to
intend to stick out my tongue and thereby intentionally do so. The 3
rd
-person, by
contrast, fails to engage one's agency (in normal circumstances), which is why use of
3
rd
-person imperatives breeds confusion amongst conversational participants. Thus,
in this way we can see that uses of 2
nd
-person words and phrases serve an important
conversational function not served by uses of non-2
nd
-person words and phrases.
So, according to the explanation being put forth, this is why use of 2
nd
-person
imperatives is so dominant: imperatives are used to get others to do things; to get
others to do things requires engaging their agency; and a primary function of uses
5
Cf. Ferrero (2013)
6
Babb (2016)
7
`Donald' not my real name.
140
of 2
nd
-person words and phrases is to engage another's agency. Since non-2
nd
-person
words and phrases don't function to engage one's agency, using these to get someone
to do something will require more work on the part of either the speaker or the
hearer. The speaker might have to make a follow-up statement directly speaking to
the addressee's agency (\I mean you"), or the addressee might have to work out that
she herself is being told to act.
5.2.3 The Non-Dominance of 1
st
- and 3
rd
-Person Imperatives
The foregoing gives us the workings of an explanation for why use of 1
st
-person
imperatives is even rarer than 3
rd
-person imperatives, and, moreover, why it is that
only very few languages have 1
st
-person imperatives. First, in order to get oneself
to do something one needn't issue an imperative to oneself. My wanting or already
intending to do something will in most cases be sucient for getting myself to do it.
Second, in those cases where one does issue an imperative to oneself, the goal is the
same as before: to get someone { in this case one's self { to do something. But the fact
that the person one wants to do something is oneself does not change that in order to
get this person (oneself) to do it, one must engage the person's (one's own) agency.
So the reason I say `You take out the trash!' to myself is because I do not have an
existing intention to take out the trash, but I want myself to have this intention. So
I have to engage my own agency and get it to produce an intention to take out the
trash. And as before, the primary way to engage someone's agency is through use of
the 2nd-person. 1
st
-person imperatives are therefore completely unnecessary. There
is no purpose they might serve that is not already served by 2
nd
-person imperatives.
8
The same doesn't go for 3
rd
-person imperatives. There are purposes 3
rd
-person
imperatives serve that cannot be served by 2
nd
-person imperatives. For example,
when one captain wants another captain's men to do something, he might use a 3
rd
-
person imperative like `Your men go around the right'.
9
So while there is no need for
1
st
-person imperatives in languages, there is some need for 3
rd
-person ones. There
is a catch, however, when it comes to 3
rd
-person imperatives that don't refer to the
addressee, of which `Your men go around the right' is an example; I discuss this catch
presently.
8
Which is not to say that 1
st
-person words and phrases couldn't or don't serve the same function
as 2
nd
-person ones when it comes to speaking to oneself. The point is merely one of simplicity: given
that 1
st
-person imperatives serve the same function as 2
nd
-person imperatives, language users can,
and clearly do, do without 1
st
-person imperatives.
9
Potsdam (1996)
141
5.2.4 Problem Cases and Norms of Imperative Use
According to the explanation being put forth, the link between imperatives and in-
tentions is all there is to the dominance of 2
nd
-person imperatives. If that's right then
there is no need to posit syntactic restrictions on imperative subjects. Zanuttini et
al. have an objection to such purely pragmatic explanations worth examining. Their
objection is that pragmatic approaches cannot explain the oddness of imperatives
like (1) and (2).
10
(1)
?
The table be clean!
(2)
?
John kiss you!
Recall that according to Na ve Satisfactionism an imperative expresses a simple-
present tense proposition, determined by the imperative's interpretable features. Now
suppose there are no syntactic restrictions on the subjects of imperatives. If that's
right, then according to Na ve Satisfactionism there is no reason why (1) and
(2) are not possible imperatives. Moreover, (2) should express the same proposition
as (3), which should imply that (2) can be used in any situation that (3) can.
(3) You be kissed by John!
But it is clear (2) is a poor replacement for (3) in conversation. What gives?
The problem with imperatives like (1) and (2) can be explained in terms of
agency-engagement and the nature of intentions. I begin with (2). The main issue
with (2), and the reason it is a poor substitute for (3), is that while it does represent
the addressee, it is not suitable for engaging the addressee's agency. It's not precisely
because `you' is not occurring in subject position. Thus the key dierence between
(2) and (3) is that the latter is suitable for agency-engagement whereas the former is
not. Now of course that doesn't mean (2) is completely unusable. But it does place
the burden on the addressee to engage her own agency, perhaps through an inference
such as: \So-and-so said (2) to me, so they must want me to do something { namely,
(to intend to) be kissed by John." We can thus say that the following conversational
norm governs directives, the speech acts in which imperatives are used:
MQD-I When issuing directives make sure to speak in a way that engages the
addressee's agency.
In Gricean fashion, MQD-I is short for Maxim of Quality for Directives I. (2)
prima facie violates this norm, which explains its oddness.
10
Zanuttini et al. (2012, p. 4)
142
The explanation of the oddness of (1) is dierent, relying instead on the na-
ture of intentions. In being essentially rst-personal, intentions take as their objects
only propositions about the individual whose intention it is. (1), however, expresses
a proposition that can't be intended by any agent, because it isn't about any agents
{ only tables. What an agent can do with such an imperative is go through some
Gricean reasoning. Suppose A says (1) to B. Upon hearing this B may reason as fol-
lows: \A has said (1) to me. No one would say such an imperative unless they wanted
the proposition expressed to be true. Moreover, since A has said this imperative to
me, he must want me to be the one to make it true. So what A must be trying to
communicate, as a directive, is the proposition that I clean the table. This I can
intend." A similar line of reasoning would be applied to (4), which was mentioned
earlier, since it too fails to express a proposition about the addressee.
(4) Your men go around the right!
Here the addressee would have to infer that the speaker is trying to communicate,
as a directive, something like the proposition that I, the addressee, have my men go
around the right.
What makes (1) and (4) odd, then, is that uses of these imperatives prima
facie seem to violate the Maxim of Quality for Directives II, or MQD-II for
short:
MQD-II When issuing directives make sure to communicate propositions that
can be the objects of the hearer's intentions.
Given that the contents of intentions must be about the agent whose intention it is,
it follows that one is to make the contents of their directives about the addressee.
Thus, a directive involving (1) or (4) will prima facie violate MQD-II { which is
why we nd uses of it odd. In going through the forgoing Gricean reasoning, what
an addressee is doing is saving a speaker from such a violation, by determining that
he has communicated an intendable proposition.
5.3 The Normativity of Directives
There are a number of dierent kinds of directive speech acts (i.e. the speech acts
imperatives are used in). The list includes: ORDERS, COMMANDS, DEMANDS,
SUGGESTIONS, ADVICE, REQUESTS, INSTRUCTIONS, and RECIPES. De-
spite this diversity, a distinguishing feature of directives in general { and imperatives
143
used in directives in particular { is that they are action-guiding. The point of a di-
rective, whether it be a command, request, instruction, etc., is to get someone to do
something, such that if the directive is accepted (and there are no barriers), the act
in question will be performed. The doctor says `Stick out your tongue' to me in a
directive and in accepting this directive I am thereby led to stick out my tongue.
The strength of a directive speech act depends on what kind of directive it is.
Some directives are stronger than others, in the sense that they carry more action-
guiding force. We can classify the dierent kinds of directives in terms of whether they
are strong, weak, or neutral. Strong directives include ORDERS and COMMANDS.
Weak directives include SUGGESTIONS and ADVICE. Neutral directives include
INSTRUCTIONS and RECIPES. This gives us the following taxonomy:
Strong Directives: Commands, Orders, Demands
Weak Directives: Suggestions, Advice, Requests
Neutral Directives: Instructions, Recipes
The dierence between strong and weak directives is that strong directives
guide action by attempting to impose a requirement that the action be performed,
thus not permitting that it not be performed, but with weak directives no such
requirement imposition is attempted, leaving it permitted that the action not be
performed.
11
Neutral directives also don't attempts to impose requirements. The dif-
ference between neutral directives and both strong and weak ones is that strong
and weak but not neutral attempt to make performance of an act better than non-
performance. Whether or not a neutral directive promotes action will depend on
the interests and ends of individuals. The instruction manual for a bookshelf will be
action-guiding only to the extent that I want to build a bookshelf. Thus, while strong
and weak directives are essentially action-guiding (these directives essentially involve
creating requirements and/or making act performance better than non-performance),
neutral directives are only contingently so (these directives bring about requirements
or make act performance better only to the extent that one has certain antecedent
desires, ends, etc.).
Suppose A orders B to put gas in the lawn mower and B acknowledges this
order (more on what it means to \acknowledge" a directive later). Then all else
being equal it would be better for B to put gas in the mower than not to. Same
11
This is essentially a variation of Lewis (1979)'s view about the function of imperatives used
in directives, but without taking on board Lewis's semantics of imperatives, according to which
imperatives express deontic propositions.
144
goes for a request. A requesting of B that he put gas in the mower, together with B
acknowledging this request, makes it such that it would be better for B to put gas in
the mower than not. But now suppose that A is merely providing B with instructions
for how to start the mower, and one of these instructions is that he put gas in it.
Would this make it better for B to put gas in the mower, all else being equal? No.
A mere instruction to does not by itself make it better to than not. The same
goes for the lines of a recipe, which are also a form of directive. The mere fact that
a recipe directs you to whisk 3 eggs with a cup of
our does not make it such that
it would be better for you to whisk 3 eggs with a cup of
our.
Directives are thus dividable along two normative dimensions: (i) whether or
not they attempt to create requirements and (ii) whether or not they attempt to make
performance of an act all else equal better than nonperformance. Strong directives
both attempt to create a requirement and attempt to make performance better than
nonperformance. Weak directives don't create a requirement but do make perfor-
mance better than nonperformance. Neutral directives do neither. Hence the names.
Strong directives are so-called because they are normatively strong: creating require-
ments and making act performance good. Weak directives are so-called because they
are normatively weak, merely making act performance better than nonperformance
but not obliging it. And neutral directives are so-called because they are normatively
neutral, not even making act performance better than nonperformance.
How to explain these normative dierences? A natural suggestion { one previ-
ously oered by Bach and Harnish (1979) { is that the explanation has something to
do with reasons. Certain directive speech acts by their nature involve proering one
reasons for action. More specically, by their very natures certain directives involve
proering addressees reasons to make true the propositions communicated. If A says
`Mow the lawn!' to B in an order or request, then this ipso facto involve proering
some form of reason for B to mow the lawn. Moreover, the key dierence between
strong, weak, and neutral directives involves a dierence in the kind of reasons that
each attempts to provide.
5.3.1 Reasons: Motivating v. Normative; First-Order v. Second-
Order
Reasons for action are standardly divided into two basic kinds. Motivating reasons
are those reasons for which one performs a given action. If I ate the pie because
I was hungry, then my being hungry was my motivating reason for eating the pie.
Normative reasons, by contrast, are considerations which favor or lend support to the
performance of a given action. That there is gasoline in the glass is a normative reason
145
which favors my not taking a drink.
12
While it is often true that a directive is one's
motivating reason for performing an action, the reasons that directives essentially
attempt to provide are, in the rst instance, clearly normative. They are, after all,
the sorts of reasons that can give rise to requirements and make action performance
better than nonperformance { which are normative notions through and through.
In order to account for the normative dierences between strong, weak, and
neutral directives we need to distinguish between two functionally dierent kinds of
normative reasons for action. Normative reasons can be rst-order or second-order.
A rst-order reason is one which bears directly on an action. That there is gasoline in
the glass is a rst-order reason for not drinking. Second-order reasons are ones that
bear on rst-order reasons. Raz (1975, 1999) argues that laws proer second-order
reasons. Typically when deciding what to do we take into account our (rst-order)
reasons for and against certain actions. Then, provided we are rational, we ultimately
decide what to do on the balance of these reasons. When deciding whether to cross
the street on red, I consider the risks involved, such as that I might get hit or cause
a car to crash, as well as the rewards, such as that I will get to my destination a
bit quicker. Whether or not I cross on red will depend on how these reasons play
out against each other. Laws cut short this reasoning process. If I know that the
law says never to cross on red, then no matter what my reasons for crossing are, it
is not the case that I ought to cross. According to Raz, this is because laws proer
exclusionary reasons: they proer second-order reasons that exclude certain rst-
order reasons from consideration.
13
In the case of crossing on red, the law excludes
my rst-order reasons for crossing on red from consideration and in so doing leaves
not crossing as the only rational option. But laws also proer rst-order reasons
in favor of the actions they prescribe.
14
Not only does the no-crossing-on-red law
exclude reasons for crossing on red from consideration, but it also gives me positive
reason not to cross on red in any given situation. If it didn't do this, it would be
dicult to explain why it is rational to follow the law. Suppose the law proers no
rst-order reasons. Then though the law makes it such that it is not the case that
I ought to cross, it does not make it the case that I ought not cross. Put another
way, if laws only proer exclusionary reasons, then laws will establish that one is
12
Dancy (2000) thinks that all reasons are normative reasons, and that motivating reasons are
merely normative reasons that have been acted on. Nothing I have to say turns on whether there
is a fundamental dierence between motivating and normative reasons.
13
This is not the same as saying that these reasons eliminate or erase the normative force of the
rst-order reasons. First-order reasons keep there normative force even in the face of exclusionary
reasons. Rather, all these reasons do is exclude the relevant rst-order reasons from consideration.
That is, in the face of an exclusionary reason, a rst-order reason is left out of practical deliberation.
14
(Raz 1975)
146
permitted to follow them but not that one ought to. But clearly one ought to follow
the law (assuming it is legitimate). So laws must also proer rst-order reasons in
favor of following them.
15
5.3.2 Directives and Proerings of Reasons
Bringing things back to directives, it's important to recognize that laws are com-
mands of a government addressed to its citizens.
16
The law never to cross on red is
a command of the government to its citizens not to cross on red at any time. This
raises a question about the relation between commands and the reasons they prof-
fer. Do laws proer reasons because they are commands, in that there is something
about commands in general that is reason-proering? Or is it because they are com-
mands of an authoritative government, such that it is only commands of this kind,
those issued by political authorities, that are reason-proering? Suppose we take the
second option. To simplify things, let's generalize and say that, on this view, only
the commands of recognized authorities of any kind (political, familial, etc.) proer
reasons; commands of those not recognized as authorities do not proer reasons. The
motivation for this view is that authority relations can ground the reasons proered
by the commands. Command C proers reasons R
1
...R
n
because it was issued by
an authority. On the other hand, the trouble with this view is that commands of
non-authorities can also carry normative weight. Even though he has no authority
over such matters, my neighbor's command to water my lawn more can have bearing
on whether I ought to. I can certainly take it into account when considering whether
to water my lawn, and it can certainly cut short my deliberations, such that on the
basis of his command I jump right to the decision to water my lawn, ignoring the
reasons I have not to. So, taken by itself, independent of other considerations, his
command does seem to provide a reason not only for me to water my lawn more, but
also a reason not consider reasons not to water my lawn. Of course these reasons can
15
The existence of second-order reasons has been challenged by many. See, for example, Clarke
(1977); Mian (2002); Piller (2005); Whiting (2016), amongst others. None of these are challenges
to claim that second-order reason are reasons; rather they are all challenges to the claim that
they are second-order. While I do think that second-order reasons exist and that Raz has rightly
identied some of their sources, everything I say is compatible with \second-order" reasons merely
being particularly weighty rst-order reasons. However, the reason I think this is wrong and so will
continue to assume that second-order reason exist is that it's not clear { and many opponents of
second-order reasons acknowlege this { that weighty rst-order reason really can play all the roles
that second-order reasons play, such as the role of cutting practical deliberation short even before
one has taken into account all the (rst-order) reasons they have for or against some act.
16
What, if anything, makes such commands legitimate is not something I wish to entertain here.
147
be defeated, and quite easily so. Both might be defeated by my desire not to waste
water. But being easily outmatched does not mean they are not reasons.
This pushes us towards the other view, according to which, in general, to
command is to proer reasons. More specically, a speech act is a command to
just in case it proers (i) a rst-order reason to and (ii) a second-order reason not
to consider rst-order reasons not to . Since only agents can perform speech acts
and proer reasons, this view might ground the proering of reasons in commands in
the agency of others: command C proers reasons R
1
...R
n
because it was issued by an
agent (oneself inclusive).
17
The role of authority, then, is to augment these reasons.
It takes more to defeat the reasons proered by the commands of authorities than
it does the reasons proered by the commands of non-authorities. It is easier for
me to reject my neighbor's command than, say, the command of a city ocial. But
despite this dierence in strength, on the present view the commands of authorities
and non-authorities alike proer rst-order and exclusionary second-order reasons
for performing certain acts.
Commands exemplify strong directives. In fact, it makes sense to say that all
strong directives are commands.
18
Take orders and demands. Orders can be analyzed
as ocial or institutional commands. And demands are coercive commands. When
a colonel says `Clean the barracks' to a private in an order what he is doing is com-
manding the private to clean the barracks under threat of institutional reprimand
if he doesn't. Importantly, that the private will be reprimanded for noncompliance
is an additional rst-order reason for the private to clean the barracks: this reason
doesn't replace the reasons constituted by the command itself, but rather supple-
ments them. Similarly, when a mugger says `Give me your wallet' in a demand what
he is doing is commanding you to give him your wallet on threat of physical harm if
you don't, and this threat of harm is an additional reason to do as the mugger says.
Thus, orders and demands are simply commands that involve additional reasons { in
the form of consequences { to perform a given act, which in turn characterize these
speech acts. In the case of orders, the additional reasons which characterize them are
supplied by the threat of institutionalized consequences. In the case of demands, the
additional characterizing reasons are supplied by the threat of non-institutionalized
consequences, often taking the form of physical harm. What unies the class of strong
directives is that they all proer a rst-order reason to perform an act as well as an
17
Darwall (2006) argues that to recognize the agency of another is ipso facto to recognize them
as having a kind of second-personal authority. I see no reason to disagree with this, but it's worth
noting that this is a much broader notion of `authority' than what I am using here.
18
Rescher (1966); Hamblin (1988)
148
exclusionary reason not to consider rst-order reasons against performing the act.
19
The forgoing analysis of strong directives gives us a way to account for the nor-
mative dierences between strong, weak, and neutral directives. If strong directives
proer both a rst-order and an exclusionary reason, weak directives only proer a
rst-order reason, while neutral directives don't proer either. A neighbor's request
that you turn the music down, for example, proers a rst-order reason to turn the
music down. But it does not proer an exclusionary reason for you to do this { which
is why the request doesn't create a requirement. Nevertheless, the rst-order reason
the request proers explains why turning down the music is all else equal better
than not. By contrast, the fact that neutral directives don't proer reasons at all
explains why these directives neither create requirements nor make act performance
better than nonperformance. This also explains why when it comes to things like
cookbook recipes and assembly manuals one must have their own reasons or motives
for carrying them out, such as a desire to cook a certain dish or a desire to build a
certain piece of furniture. To be sure, this view is weaker than other reason-based
analyses of directives. Bach and Harnish (1979), for example, hold that all directives
are reason-proering: directives \express the speaker's intention (desire, wish) that
his utterance or the attitude it expresses be taken as (a) reason for the hearer to act."
Thus, as I understand it, on their view being reason-proering is an essential feature
of directives. The class of neutral directives, however, is a counterexample to this
view, since such directives need not involve any intention on the part of the speaker
that their utterance or the attitude it expresses be taken as a reason; I can give my
friend instructions for how to cook an omelette with no intention that these instruc-
tions be taken as reasons to cook an omelette. On my view, rather than being an
essential feature of directives in general, reason-proering is essential only of strong
and weak directives, with these being distinguished by the reasons they proer.
5.3.3 What Kind of Reasons?
This general view { that a dening feature of strong and weak directives (i.e. strong
and weak) is the proering of reasons to perform the specied act { is consistent with
most ways of understanding what normative reasons are and where they come from.
It is consistent, for instance, with both the view that reasons are facts about the world
with no necessary and the view that reasons are mental states. On the conception of
reasons as mental states, `proering reasons' would come out as meaning an attempt
to get one to form certain mental states { in Williams's terms, it would be to \add
to their subjective motivation set" certain mental states. On the reasons-as-facts
19
Cf. MacIntyre (1965)
149
conception, `proering reasons' would mean trying to get one to recognize, at least
subconsciously, the relevant facts, such as the Kantian fact that the speaker is an
autonomous person.
20
The view is also consistent with both reasons internalism and externalism.
According to Finlay and Schroeder (2015), internalism is the general view that every
reason to act is in some way related to a motivational fact. Dierent versions of
internalism vary according to (i) what they say the relation is in question is and (ii)
what they say the relevant motivational fact is. But if any version of internalism is
true, then the reasons proered by directives could not fail to be related in some
way to motivational facts. Externalism, by contrast, is the denial of internalism,
hence it is the view that not every reason is somehow related to a motivational fact.
Nevertheless, many externalists do think reasons are typically related to motivational
facts, but unlike the internalist, for the externalist these relations need not exist
for the reason to be a reason.
21
So if externalism is true, then though the reasons
proered by directives tend to motivate a hearer to act in the specied way, these
reasons could fail to motivate or even be related to any motivational facts about the
hearer.
It would seem, however, that on my view the reasons involved in directives
must be agent-relative rather than agent-neutral.
22
After all, a natural thought is that
the reasons being proered have a form like so-and-so told me to do such-and-such
or so-and-so wills or wants me to do such-and-such, both of which are distinctively
agent-relative (on at least one way of understanding the agent-neutral/agent-relative
distinction; see Ridge (2011)). But this is not a foregone conclusion; the view is
also consistent with these reasons being agent-neutral. The reasons involved in A
commanding B to sweep the patio, for example, might be reasons for anyone to
sweep the patio. Of course A is only proering these reasons to B, and, what's more,
since B is presumably the subject of the proposition expressed by A's command, this
command can be satised only if B sweeps the patio { but it doesn't follow from
either of these facts that the reasons involved are not agent-neutral.
An objection to directives proering agent-neutral reasons is that this will
over-generate reasons. Every strong or weak directive will bring about new reasons
for anyone to perform the act specied by the directive. But a vast number of strong
and weak directives are issued in a single day { let alone a single hour. If the reasons
involved in directives are agent-neutral then any given person is being constantly
20
Cf. Korsgaard (1996) and Darwall (2006)
21
See for instance Nagel (1978), Part (1997), Scanlon (1998).
22
Is internalism compatible with agent-neutral reasons? Schroeder (2007) argues for a positive
answer.
150
bombarded with reasons to perform all varieties of acts. This is deeply implausible.
Neither a mother in Poland commanding her daughter to wash their family's clothes
nor a boss in Japan ordering his subordinates to work overtime involve reasons to
perform the relevant acts that apply to me! The only way to avoid this result, this
over-generation of reasons that apply to each of us, is for the reasons proered in
directives to be necessarily agent-relative.
So who do these reasons apply to? Given the existence of 3
rd
-person imper-
atives, we need to remember that sometimes the subject and the addressee of a
directive can come apart. In the case of one captain uttering `Your men go around
the right' to another captain in an order, the subject of the directive is the second
captain's men whereas the addressee is the captain herself. So we seem to have three
options for who the reasons apply to: the subject, the addressee, or both. The best
option, I think, is both, with an important qualication: the reasons to perform the
act specied in the directive only apply to the subject, but the reasons that apply
to the addressee are instead reasons to get the subject to perform the specied act.
I take each of these points in turn.
In a directive a speaker is proering reasons for the subject of the directive
to perform the act specied. But why not the addressee rather than the subject?
Directives are satised or carried out only if the subject of directive performs the
relevant act. The rst captain's order is satised only if the second captain's men go
around the right; the second captain herself going around the right will not satisfy
this directive. True, the rst captain is implicating that the second captain is to make
her men go around the right (seex5.2.4 above), which if done will ipso facto satisfy
the rst captain's order. But the order can also be satised if the men, overhearing the
conversation, immediately move to go around the right without the second captain's
intervention. Moreover, the order (arguably) creates a requirement for the men to go
around the right and also makes it better, all else being equal, that they do so than
not { which the men know and so go around the right. This makes for a prima facie
case for the claim that even though the order was addressed to the second captain,
it proers reasons for the men, the subjects of the directive, to go around the right.
Now this doesn't mean there are no reasons that apply to the addressees in
cases of directives with 3
rd
-person subjects. Korsgaard (1993, 1996) argues that a
reason for A to can also be a reason for others to promote A's ing (Intersub-
jectivity). If this is right, then one idea is that in proering reasons for the second
captain's men to go around the right the rst captain is equally proering reasons
for the second captain, who is the addressee, to promote or see to it that her men go
around the right. Supporting this idea are the observations (i) that the command is
capable of cutting short the second captain's deliberations and leading her to decide
151
to have her men go around the right, and (ii) that if the second captain doesn't see to
it that her men go around the right, she will be held accountable. These observations
suggest the command imposed a requirement on the second captain to see to it that
her men go around the right (and she acknowledged this), in which case it involves
reasons for her to.
5.3.4 From Reasons to Act to Reasons to Intend
We are now in a position to link this account of directive speech acts up to Mood
Acceptancism. To do so we need to bring intentions into the mix. More speci-
cally, we need to note that in proering normative reasons to act, strong and weak
directives proer normative reasons to intend, either because reasons to act are rea-
sons to intend or because reasons to intend are derivative of reasons to act.
23
This
links the analysis of directives back up to the imperative mood. Why do we accept
the imperatives used in conversation (= form intentions about the contents of used
imperatives), or at least why is it acceptable to do so? Because we acknowledge that
those uses, insofar as they are strong or weak directives, proer reasons to intend
the contents of the imperatives (or, in the case of 3
rd
-person imperatives, reasons to
intend to make that content true). I intend to nish the report by 5pm because my
boss said `Finish the report by 5pm' to me in a command. This command proers,
and I acknowledge this, reason to nish the report by 5pm and reason not to consider
reasons not to do so. These reasons enter into practical deliberation as premises sup-
porting the formation of an intention to nish the report by 5pm. Similarly, I intend
to mow the law because my wife said `Mow the lawn' to me in request. I acknowledge
that this request proers rst-order reason to mow the law, and provided I have no
suitably strong reasons against doing so, this reason will suce for me deciding to
(= forming an intention to) mow the lawn. One last example: I intend to make the
proposition that my room is clean true because my dad said `Your room clean, now!'
in an order, which proered me exclusionary reason to clean my room as well as the
reason stemming from the consequences of not doing so.
23
For the view that reasons to act are reasons to intend, or vice versa, see Davidson (1978), Scanlon
(1998), Kolodny (2005), Setiya (2014), and Way (2015), amongst others. For a nice discussion of
issues involved in metaphysically relating reasons to act and reasons to intend see Heuer (2017).
152
5.4 Bringing It All Together: Or, How to Tell Each
Other What to Do
A speaker proering reasons is one thing, an addressee acknowledging them is an-
other. It is here in the acknowledging of reasons that our two topics come together.
For in order for one to acknowledge the reasons proered by directives, one's agency
must rst be engaged. Suppose someone says `Water the plants' to me in a request.
In being a request, this speech act proers (rst-order) reasons for me to water the
plant. However, if I do not recognize that this request is directed to me, I will not
consider watering the plants and so these reasons (if they actually exist) won't be
brought to bear in my practical deliberations. Not just intentions, but practical de-
liberation as well is essentially rst-personal; it essentially involved an agent deciding
what she herself is to do.
24;25
So one's agency must rst be engaged in order to get
practical deliberation up and running. Since reasons can only be brought to bear
(i.e. acknowledged) in practical deliberation if one's agency is engaged and practical
deliberation is up and running, prior agency engagement is a necessary condition on
acknowledging the reasons proered by directives. This brings us back to the impor-
tance of the 2
nd
person and 2
nd
personal address, as I argued above that the primary
function of the 2
nd
person just is agency-engagement. So it's through the 2
nd
person
that the normativity of directives is felt, because the 2
nd
person enables the reasons
proered by a directive to be brought to bear in practical deliberation, and in being
brought to bear one recognizes them as reasons.
This then is the picture of imperative pragmatics we have arrived at. Imper-
atives are used in directive speech acts. The basic role of directive speech acts in
communication is to get people to do things. They do this by proering reasons to
perform the relevant acts or to promote someone else performing the act (or, in the
case of neutral directives, by bringing to bear the antecedent reasons and motives
one has to perform the act in question). However, proering reasons is not enough.
These reasons (again, if they exist) also need to be brought to bear in one's deciding
of what to do. Since practical deliberation is essentially rst-personal, this means
that the reasons proered by a directive can be brought to bear only if one's agency
is engaged. Agency-engagement is achieved by way of the 2
nd
person and 2
nd
personal
address, which in turn explains why the imperatives used in discourse are predom-
inantly if not exclusively 2
nd
personal. Once one's agency is engaged, and provided
that all else goes accordingly in the deliberation process, the result will be an in-
24
Casta~ neda (1975) See also Burge (2000)
25
`herself' is a quasi-indicator here; see Casta~ neda (1967).
153
tention the contents of which will be the proposition expressed by the imperative
used in the directive. This intention can then be translated into action. So though
we often take it for granted in day-to-day conversation, telling each other what to
do is not a simple matter. There are many places where it can break down, such as
one's agency not being engaged, or the reasons proered not being acknowledge or
not being weighty enough, or the world simply not cooperating in the translation of
the intention into action. But if all does go well, we can succeed in telling each other
what to do.
154
Bibliography
Anttila, R. (2000). Greek and Indo-European etymology in action: proto-Indo-
European. John Benjamins Publishing.
Babb, M. (2016). The essential indexicality of intentional action. Philosophical
Quarterly, forthcoming.
Bach, K. (1981). What's in a name? Australasian Journal of Philosophy, 59:371{86.
Bach, K. and Harnish, R. (1979). Linguistic communication and speech acts. MIT
Press, Cambridge, MA.
Barker, C. (2012). Imperatives denote actions. In Proceedings of sinn und bedeutung,
volume 16, pages 57{70.
Barker, S. J. (2004). Renewing meaning: A speech-act theoretic approach. Oxford
University Press, Oxford.
Binnick, R. (1991). Time and the verb: A guide to tense and aspect. Oxford University
Press, Oxford.
Bobaljik, J. (2003). Floating quantiers: Handle with care. In Cheng, L. and
Sybesma, R., editors, The second Glot International state-of-the-article book: The
latest in linguistics, pages 107{148. Mouton de Gruyter Berlin.
Boisvert, D. and Ludwig, K. (2006). Semantics for nondeclaratives. In Smith, B.
and Lepore, E., editors, The Oxford handbook of the philosophy of language, pages
864{892. Oxford University Press, Oxford.
Boisvert, D. R. (2008). Expressive-assertivism. Pacic Philosophical Quarterly,
89(2):169{203.
Bolinger, D. (1967). The imperative in english. To honor Roman Jakobson: essays
on the occasion of his 70 birthday, pages 335{362.
155
Bratman, M. (1983). Casta~ neda's theory of practical reason. In Tomberlin, J., editor,
Agent, Language and the Structure of the World: Essays Presented to Hector-Ner
Casta~ neda with His Replies. Hacket, Indianapolis, IN.
Burge, T. (1973). Reference and proper names. The Journal of Philosophy,
70(14):425{439.
Burge, T. (2000). Reason and the rst person. In Wright, C., Smith, B., and Mac-
donald, C., editors, Knowing our own minds, pages 243{270. Oxford University
Press.
Casta~ neda, H.-N. (1967). Indicators and quasi-indicators. American philosophical
quarterly, 4(2):85{100.
Casta~ neda, H.-N. (1975). Thinking and doing. D. Reidel Publishing Company, Dor-
drecht, Holland.
Charlow, N. (2011). Practical language: Its meaning and use. PhD thesis, University
of Michigan.
Charlow, N. (2014). Logic and semantics for imperatives. Journal of Philosophical
Logic, 43(4):617{664.
Chomsky, N. (1970). Remarks on nominalization. In Jacobs, R. and Rosenbaum,
P., editors, Readings in English transformational grammar. Ginn and Company,
Waltham, Massachusetts.
Chomsky, N. (1995). The minimalist program. Cambridge University Press, Cam-
bridge.
Cinque, G. (1999). Adverbs and the universal hierarchy of functional projections.
Oxford University Press, Oxford.
Cinque, G. and Rizzi, L. (2008). The cartography of syntactic structures. Studies in
linguistics, 2:42{58.
Clarke, D. (1977). Exclusionary reasons. Mind, 86(342):252{255.
Comrie, B. (1985). Tense. Cambridge University Press, Cambridge.
Copp, D. (2001). Realist-expressivism: A neglected option for moral realism. Social
Philosophy and Policy, 18(02):1{43.
156
Dahl, O. (1995). The marking of the episodic/generic distinction in tense-aspect
systems. In Carlson, G., editor, The generic book, pages 412{426. University of
Chicago Press, Chicago, IL.
Dancy, J. (2000). Practical reality. Clarendon Press.
Darwall, S. L. (2006). The second-person standpoint: Morality, respect, and account-
ability. Harvard University Press.
Davidson, D. (1978). Intending. In Philosophy of history and action, pages 41{60.
Springer.
Downes, W. (1977). The imperative and pragmatics. Journal of linguistics,
13(01):77{97.
Dowty, D. (1979). Word meaning and Montague grammar: The semantics of verbs
and times in generative semantics and in Montague's PTQ, D. Reidel, Dordrecht.
Dowty, D. R., Wall, R., and Peters, S. (1981). Introduction to montague semantics.
Reidel, Dordrecht, Holland.
Ferrero, L. (2013). Can i only intend my own actions? In Shoemaker, D., editor, Ox-
ford Studies in Action and Responsibility Oxford, pages 70{94. Oxford University
Press.
Finlay, S. and Schroeder, M. (2015). Reasons for action: Internal vs. external. In
Zalta, E. N., editor, The Stanford Encyclopedia of Philosophy. Metaphysics Re-
search Lab, Stanford University, winter 2015 edition.
Frege, G. (1948). Sense and reference. The Philosophical Review, 57(3):209{230.
Geach, P. T. (1958). Imperative and deontic logic. Analysis, 18(3):49{56.
Geurts, B. (1997). Good news about the description theory of names. Journal of
semantics, 14(4):319{348.
Grice, H. P. (1975). Logic and conversation. In Ezcurdia, M. and Stainton, R. J.,
editors, The Semantics-Pragmatics Boundary in Philosophy, page 47. Broadview
Press.
Hamblin, C. L. (1988). Imperatives. Basil Blackwell Ltd.
157
Han, C.-H. (2000). The structure and interpretation of imperatives: mood and force
in Universal Grammar. Outstanding Dissertations in Linguistics. Garland Pub-
lishing, New York, NY.
Hanks, P. (2007). The content-force distinction. Philosophical Studies, 134(2):141{
164.
Hanks, P. (2015). Propositional Content. Oxford University Press, Oxford.
Hare, R. M. (1952). The language of morals. Clarendon Press, Oxford.
Hare, R. M. (1971). Some alleged dierences between imperatives and indicatives.
In Practical Inferences, pages 25{43. Springer.
Hare, R. M. (1999). Objective prescriptions, and other essays. Oxford University
Press.
Harman, G. (1976). Practical reasoning. The Review of Metaphysics, pages 431{463.
Harman, G. (1986). Willing and intending. In Philosophical grounds of rationality,
pages 363{380. Oxford: Oxford University Press.
Harsh, W. (1968). The subjunctive in English. University of Alabama Press.
Hawkins, R. and Towell, R. (2015). French grammar and usage. Routledge.
Heim, I. (1983). File change semantics and the familiarity theory of deniteness.
In B auerle, R., Schwarze, C., and von Stechow, A., editors, Meaning, use and
interpretation of language, pages 164{189. de Gruyter, Berlin.
Heim, I. and Kratzer, A. (1998). Semantics in generative grammar. Blackwell,
Oxford.
Heuer, U. (2017). Reasons to intend. In Starr, D., editor, The Oxford Handbook of
Reasons and Normativity. Oxford University Press.
Hom, C. and Schwartz, J. (2013). Unity and the frege-geach problem. Philosophical
studies, 163(1):15{24.
Huntley, M. (1980). Propositions and the imperative. Synthese, 45(2):281{310.
Huntley, M. (1984). The semantics of english imperatives. Linguistics and Philosophy,
7(2):103{133.
158
Isac, D. (2015). The morphosyntax of imperatives. Oxford University Press.
Jackendo, R. (1972). Semantic interpretation in generative grammar. MIT Press,
Cambridge, MA.
Jackendo, R. (1977). X-bar syntax. MIT Press, Cambridge, MA.
Jrgensen, J. (1937). Imperatives and logic. Erkenntnis, 7(1):288{296.
Kamp, H. (1981). A theory of truth and semantic representation. In Groenendijk, J.,
Janssen, T., and Stokhof, M., editors, Formal Methods in the Study of Language,
pages 277{322. Mathematisch Centrum, Amsterdam.
Kapitan, T. (1995). Intentions and self-referential content. Philosophical papers,
24(3):151{166.
Kaplan, D. (1977). Demonstratives. In Almog, J., Perry, J., and Wettstein, H. K.,
editors, Themes from Kaplan, pages 481{564. Oxford University Press.
Kaplan, D. (1989). Afterthoughts. In Almog, J., Perry, J., and Wettstein, H. K.,
editors, Themes from Kaplan, pages 565{614. Oxford University Press.
Karttunen, L. (1974). Presupposition and linguistic context. Theoretical linguistics,
1(1-3):181{194.
Katz, J. J. and Postal, P. M. (1964). An integrated theory of linguistic descriptions.
Mit Press.
Kaufmann, M. (2012). Interpreting imperatives. Springer, Berlin.
Kneale, W. (1962). Modality de dicto and de re. In Nagel, E., Suppes, P., and Tarski,
A., editors, Logic, Methodology and Philosophy of Science, Proceedings of the 1960
International Congress, pages 622{633. Stanford University Press, Stanford, CA.
Kolodny, N. (2005). Why be rational? Mind, 114(455):509{563.
Kornai, A. and Pullum, G. (1990). The x-bar theory of phrase structure. Language,
66:24{50.
Korsgaard, C. M. (1993). The reasons we can share: an attack on the distinction
between agent-relative and agent-neutral values. Social Philosophy and Policy,
10(01):24{51.
159
Korsgaard, C. M. (1996). The sources of normativity. Cambridge University Press.
Kratzer, A. (1981). The notional category of modality. In Eikmeyer, H.-J. and Rieser,
H., editors, Words, worlds, and contexts, pages 38{74. de Gruyter, Berlin.
Kratzer, A. (1991). Modality. In von Stechow, A. and Wunderlich, D., editors,
Semantics: An International Handbook of Contemporary Research, pages 639{650.
de Gruyter, Berlin.
Krifka, M. (1995). The semantics and pragmatics of polarity items. Linguistic
analysis, 25(3-4):209{257.
Larson, R. (2003). Time and event measure. Philosophical perspectives, 17(1):247{
258.
Lewis, D. (1970). General semantics. Synthese, 22(1):18{67.
Lewis, D. (1973). Counterfactuals. Basil Blackwell Ltd., Oxford.
Lewis, D. (1979). A problem about permission. In Saarinen, E., Hilpinen, R., Ni-
iniluoto, I., and Hintikka, M. B. P., editors, Essays in honour of Jaakko Hintikka,
pages 163{175. D. Reidel Publishing Company, London.
Lobeck, A. (1999). Vp ellipsis and the minimalist program: Some speculations and
proposals. In Lappin, S., editor, Fragments: Studies in ellipsis and gapping, pages
98{123. Oxford University Press, Oxford.
MacIntyre, A. (1965). Imperatives, reasons for action, and morals. The Journal of
Philosophy, 62(19):513{524.
Mastop, R. (2011). Imperatives as semantic primitives. Linguistics and philosophy,
34(4):305{340.
Matthews, P. H. (1991). Morphology. Cambridge Textbooks in Linguistics. Cam-
bridge: Cambridge University Press.
Matthews, S. and Yip, V. (2013). Cantonese: A comprehensive grammar. Routledge.
McNamara, P. (2014). Deontic logic. The Stanford Encyclopedia of Philosophy.
Mele, A. R. (1992). Springs of action: Understanding intentional behavior. Oxford
University Press.
160
Mian, E. (2002). The curious case of exclusionary reasons. Canadian Journal of Law
& Jurisprudence, 15(1):99{124.
Nagel, T. (1978). The possibility of altruism. Princeton University Press.
Ninan, D. (2005). Two puzzles about deontic necessity. In Gajewski, J., Hacquard,
V., Nickel, B., and Yalcin, S., editors, MIT Working Papers in Linguistics 51,
pages 149{178. MIT Press.
Pak, M. (2004). Korean particles and clause types. Ms., Georgetown University.
Palmer, F. R. (2001). Mood and modality. Cambridge University Press, Cambridge,
second edition.
Part, D. (1997). Reasons and motivation. Proceedings of the Aristotelian Society,
71:99{130.
Partee, B. and Rooth, M. (1983). Generalized conjunction and type ambiguity.
In B auerle, R., Schwarze, C., and von Stechow, A., editors, Meaning, Use, and
Interpretation of Language, pages 334{356. de Gruyter, Berlin.
Perlmutter, D. and Soames, S. (1979). Syntactic argumentation and the structure of
English. University of California Press, Oakland, CA.
P eteri, A. (2012). The hungarian imperative particle hadd: A contrastive look at the
development of analogous markers of sentence mood in some european languages.
Acta Linguistica Hungarica, 59(4):439{463.
Piller, C. (2005). Kinds of practical reasons: Attitude-related reasons and exclusion-
ary reasons. Coordena c~ ao, page 98.
Platzack, C. and Rosengren, I. (1997). On the subject of imperatives: A minimalist
account of the imperative clause. Journal of comparative germanic linguistics,
1(3):177{224.
Poletto, C. and Zanuttini, R. (2003). Making imperatives: evidence from central
rhaetoromance. In Tortora, C., editor, The syntax of Italian dialects, pages 175{
206. Oxford University Press.
Pollock, J.-Y. (1989). Verb movement, universal grammar, and the structure of ip.
Linguistic inquiry, 20(3):365{424.
161
Portner, P. (2004). The semantics of imperatives within a theory of clause types. In
Watanabe, K. and Young, R. B., editors, Proceedings of semantics and linguistic
theory 14, volume 14, pages 235{252. CLC Publications.
Portner, P. (2007). Imperatives and modals. Natural Language Semantics, 15(4):351{
383.
Portner, P. (2011). Verbal mood. In Maienborn, C., Heusinger, K. V., and Portner,
P., editors, Semantics: An international handbook of natural language meaning,
pages 1262{1291. de Gruyter, Berlin.
Portner, P. (2016). Imperatives. In Aloni, M. and Dekker, P., editors, The cambridge
handbook of formal semantics. Cambridge University Press, Cambridge.
Portner, P. and Zanuttini, R. (2003). Decomposing imperatives. In IX Giornata di
Dialettologia, Padova.
Potsdam, E. (1995). Phrase structure of the english imperative. In Gabriele, L.,
Hardison, D., and Westmoreland, R., editors, Proceedings of the sixth annual meet-
ing of the Formal Linguistics Society of Midamerica, pages 143{54, Bloomington,
IN. Indiana University Linguistics Club.
Potsdam, E. (1996). Syntactic issues in the English imperative. PhD thesis, Univer-
sity of California - Santa Cruz.
Potts, C. (2015). Presupposition and implicature. In Lappin, S. and Fox, C., editors,
The handbook of contemporary semantic theory, pages 168{202. John Wiley and
Sons, Ltd., Oxford, second edition.
Proeme, H. (1984). Over de nederlandse imperativus. Forum der Letteren La Haye,
25(4):241{258.
Radford, A. (2004). English syntax: An introduction. Cambridge University Press.
Radford, A. (2006). Minimalist Syntax Revisited. Citeseer.
Raz, J. (1975). Reasons for action, decisions and norms. Mind, 84(1):481{499.
Raz, J. (1999). Practical reason and norms. OUP, Oxford.
Rescher, N. (1966). The logic of commands. Dover Publications, New York, NY.
Ridge, M. (2006). Ecumenical expressivism: Finessing frege. Ethics, 116(2):302{336.
162
Ridge, M. (2011). Reasons for action: Agent-neutral vs. agent-relative. In Zalta,
E. N., editor, The Stanford Encyclopedia of Philosophy. Metaphysics Research
Lab, Stanford University, winter 2011 edition.
Rijksbaron, A. (2002). The syntax and semantics of the verb in classical Greek: an
introduction. University of Chicago Press, Chicago, IL.
Rizzi, L. (1997). The ne structure of the left periphery. In Haegeman, L., edi-
tor, Elements of grammar: handbook of generative syntax, pages 281{337. Kluwer
Academic Publishers, The Netherlands.
Ross, J. (2009). How to be a cognitivist about practical reason. In Russ, S.-L., editor,
Oxford studies in metaethics, volume 4, pages 243{281. Oxford University Press,
Oxford.
Rupp, L. M. (1999). Aspects of the syntax of English imperatives. PhD thesis,
University of Essex.
Rus, D. (2005). Slovenian embedded imperatives. Georgetown Working Papers in
Theoretical Linguistics, 4:153{183.
Russell, B. (1910). Knowledge by acquaintance and knowledge by description. In
Proceedings of the Aristotelian Society, volume 11, pages 108{128.
Russell, B. (1919). The philosophy of logical atomism. The Monist, 29(3):345{380.
Russell, B. (2007). Imperatives in conditional conjunction. Natural Language Se-
mantics, 15(2):131{166.
Salmon, N. (1986). Frege's puzzle. Cambridge University Press, Cambridge, MA.
Sarkar, A. (1998). The con
ict between future tense and modality: The case of will
in english. University of Pennsylvania Working Papers in Linguistics, 5(2):6{18.
Scanlon, T. (1998). What we owe to each other. Harvard University Press.
Schmerling, S. (1975). Imperative subject deletion and some related matters. Lin-
guistic Inquiry, 6(3):501{511.
Schmerling, S. (1982). How imperatives are special, and how they aren't. Papers
from the Parasession on Nondeclaratives, Chicago Linguistic Society, 202:218.
Schroeder, M. (2007). Slaves of the Passions. Oxford University Press.
163
Schroeder, M. (2010). Noncognitivism in ethics. Routledge, New York, NY.
Searle, J. (1969). Speech acts: An essay in the philosophy of language, volume 626.
Cambridge University Press, Cambridge.
Searle, J. R. (1962). Meaning and speech acts. The philosophical review, 7:423{432.
Searle, J. R. (1983). Intentionality: An essay in the philosophy of mind. Cambridge
University Press, Cambridge.
Setiya, K. (2014). What is a reason to act? Philosophical Studies, 167(2):221{235.
Soames, S. (1989). Direct reference and propositional attitudes. In Almog, J., Perry,
J., and Wettstein, H., editors, Themes from Kaplan, pages 393{419. Oxford Uni-
versity Press, Oxford.
Sosa, E. (1967). The semantics of imperatives. American Philosophical Quarterly,
4(1):57{64.
Stalnaker, R. (1978). Assertion. Syntax and Semantics, 9:315{332.
Starr, W. B. (forthcoming). A preference semantics for imperatives. Semantics and
Pragmatics.
Stenius, E. (1967). Mood and language-game. Synthese, 17(1):254{274.
Stenson, N. (2008). Intermediate Irish. Routledge, London.
Stowell, T. (1982). The tense of innitives. Linguistic Inquiry, 13(3):561{570.
Thorne, J. P. (1966). English imperative sentences. Journal of Linguistics, 2(01):69{
78.
Van der Wur, W. (2007). Imperative clauses in generative grammar: An introduc-
tion. In Van der Wur, W., editor, Imperative clauses in generative grammar:
Studies in honour of Frits Beukema, pages 1{94. John Benjamins Publishing.
Von Fintel, K. and Heim, I. (2002). Lecture notes on intensional semantics. Mas-
sachusetts Institute of Technology.
Vranas, P. (2008). New foundations for imperative logic i: Logical connectives, con-
sistency, and quantiers. No^ us, 42(4):529{572.
164
Way, J. (2015). Reasons as premises of good reasoning. Pacic Philosophical Quar-
terly.
Whiting, D. (2016). Against second-order reasons*. No^ us, pages n/a{n/a.
Wolf, H. (2013). Imperatieven in de verleden tijd. Nederlandse Taalkunde, 18(3):252{
266.
Wurmbrand, S. (2007). Innitives are tenseless. University of Pennsylvania Working
Papers in Linguistics, 13(1{14):31.
Zanuttini, R. (1991). Syntactic properties of sentential negation. A comparative study
of Romance languages. PhD thesis, University of Pennsylvania.
Zanuttini, R. (1994). Speculations on negative imperatives. Rivista di linguistica,
6(1):67{89.
Zanuttini, R., Pak, M., and Portner, P. (2012). A syntactic analysis of interpretive
restrictions on imperative, promissive, and exhortative subjects. Natural Language
and Linguistic Theory, 30(4):1231{1274.
Zwicky, A. M. (1988). On the subject of bare imperatives in english. In Duncan-Rose,
C. and Vennemann, T., editors, On language: Rhetorica, phonoligica, syntactica: A
festschrift for Robert P. Stockwell from his friends and colleagues, pages 437{450.
Routledge, London.
165
Abstract (if available)
Abstract
This dissertation explores the syntax, semantics, and pragmatics of imperative sentences. Chapter 1 is on the syntax of imperatives and argues that there are no substantive structural differences between imperatives and declaratives. The differentiating syntactic factor, I argue, is the presence of a unique morpheme, IMP. Chapter 2 looks at two popular and conservative theories of the semantics of imperatives, arguing that neither is adequate. Chapter 3 proposes a different theory called Naive Satisfactionism. According to Naive Satisfactionism, imperatives semantically express satisfaction conditions indexed to a time range centered on the time of utterance. Chapter 4 takes Naive Satisfactionism one step further. In it I argue that in order to solve the problem of Mixed Compound Sentences, the semantics of imperatives must be identified with the semantics of simple present tense declaratives. I call this enhanced view Strong Naive Satisfactionism. The surprising upshot is that while there is a syntactic difference between imperatives and declarative (because of IMP), there is no substantive semantic difference. Chapter 5 turns to the pragmatics of imperatives and aims to explain how imperatives function to get people to do things. Part of the explanation involves the role of 2nd-person pronouns: they function to engage agency. The other part of the explanation involves reasons: certain (but not all) directive speech acts essentially involve attempting to give someone a reason to perform the act specified by an imperative. If in using an imperative a speaker succeeds in both engaging an addressees agency and giving them a reason to perform the relevant act (i.e. to intend the imperative's satisfaction conditions), then the speaker will have succeeded in telling the addressee what to do.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Constraining assertion: an account of context-sensitivity
PDF
Comparative iIlusions at the syntax-semantics interface
PDF
Reasoning with uncertainty and epistemic modals
PDF
Representation, truth, and the metaphysics of propositions
PDF
Meaningfulness, rules, and use-conditional semantics
PDF
Conceptually permissive attitudes
PDF
Perspective in Turkish complementation
PDF
The grammar of individuation, number and measurement
PDF
Instantial terms, donkey anaphora, and individual concepts
PDF
Superlative ambiguities: a comparative perspective
PDF
Positivist realism
PDF
The morphosyntax of states: deriving aspect and event roles from argument structure
PDF
What "ought" ought to mean
PDF
Subjectivity, commitments and degrees: on Mandarin hen
PDF
Building phrase structure from items and contexts
PDF
Towards a correlational law of language: three factors constraining judgement variation
PDF
Decomposing Slavic aspect: the role of aspectual morphology in Polish and other Slavic languages
PDF
Binding and scope dependencies with 'floating quantifiers' in Japanese
PDF
The vulnerable corpus of Propertius
PDF
Processing the dynamicity of events in language
Asset Metadata
Creator
Babb, Matthew W.
(author)
Core Title
Telling each other what to do: on imperative language
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Philosophy
Publication Date
02/15/2017
Defense Date
11/08/2016
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
imperatives,Linguistics,OAI-PMH Harvest,philosophy of language,pragmatics,propositions,semantics,syntax
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Jeshion, Robin (
committee chair
), Finlay, Stephen (
committee member
), Schein, Barry (
committee member
), Soames, Scott (
committee member
), Uzquiano, Gabriel (
committee member
)
Creator Email
Babbmw@usc.edu,Matthew.w.babb@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-337912
Unique identifier
UC11258072
Identifier
etd-BabbMatthe-5022.pdf (filename),usctheses-c40-337912 (legacy record id)
Legacy Identifier
etd-BabbMatthe-5022.pdf
Dmrecord
337912
Document Type
Dissertation
Rights
Babb, Matthew W.
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
imperatives
philosophy of language
pragmatics
propositions
semantics
syntax