University of Southern California Dissertations and Theses

Rationality and the primacy of the occurrent

(USC Thesis Other)

Rationality and the primacy of the occurrent

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Rationality and the Primacy of the Occurrent

by

Woo Ram Lee

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHILOSOPHY)

August 2018

Copyright 2018 Woo Ram Lee

ii

to Jin Yi, for love and support
and to my parents, for always being proud of me

iii
Acknowledgments

Being admitted to USC was definitely one of the best things that happened to me. I have been
extremely lucky to have had the opportunity to interact with and learn from so many brilliant
people in this community, without which this dissertation would not have come into existence.
First, I’d like to thank my cohort: Mike Ashfield, Renee Bolinger, Maegan Fairchild and
Jonathan Wright. The first few years at USC were especially tough for me, and I wouldn’t have
been able to make it through the rough time without their help and support. Being around these
outstanding people often made me feel bad about myself at first, but I quickly came to fully
appreciate how lucky I’ve been to be around them and how much I’ve benefited from them. I’d
also like to thank other grad students at USC who have read my work and contributed to my
research in some way or other: Rima Basu, Stephen Bero, Alexander Dietz, Sean Donahue, Erik
Encarnacion, Nathan Howard, Tanya Kostochka, Nicholas Laskowski, Michael Milona, Caleb
Perl, and Abelard Podgorski. I also owe thanks to philosophers outside of USC for helpful
comments and conversations: Garrett Bredeson, Bryson Brown, Louis Gularte, Brian Knab, John
Simpson, Patrick Skeels, Marek Twarzinsky, and Alex Worsnip.
I’m grateful to Janet Levin and Jake Ross, who helped me with my qualifying exam.
Jake’s forceful objections to my sample chapter made me realize that the initial direction my
project took was hopelessly ambitious, and that I needed to reshape the entire dissertation. Janet
also raised some important concerns that I tried to address in the dissertation. She also gave me
helpful comments on an earlier version of Chapter 2.
I’m especially grateful to my committee members: Steve Finlay, Mark Schroeder, and
Ralph Wedgwood. It is Ralph’s seminars on rationality and epistemology that got me interested

iv
in some central issues in the recent literature on rationality. Everything I say about rational
capacities, dispositions, and the nature of belief and intention, is influenced in some way or other
by those seminars and the conversations I had with Ralph, although I ended up disagreeing with
him on many points. As a result, Ralph has been probably the best critic of my ideas, helping me
clearly see and address some major worries about my view, which I deeply appreciate.
Ironically, it was also Ralph who suggested to me some really effective ways to make the points
I wanted to make, which I hope to have incorporated in my dissertation.
Mark has been incredibly helpful in making me see the big picture (and everyone knows
that I’m not implying here that he is not helpful in working through the details). He probably
knows better than me what contributions my work can make to the field, and what the main
strengths and weaknesses of my work are. I know he, like Ralph, gets off the boat at some point,
but he has consistently encouraged me to develop my ideas further in ways that could make them
much more interesting and philosophically worthwhile. Finally, like many others, I can’t thank
Mark enough for being patient with me, for insisting that I believe in myself, and for being there
for me whenever I had no more than a vague sense of what to do at important stages of my life in
graduate school.
My greatest philosophical debt is to Steve. It is Steve’s paper, “Against all Reason?:
Skepticism about the Instrumental Norm”, which totally shifted my way of thinking about the
nature of the rational requirements, and led me to take seriously the idea that principles of
coherence might be constitutive principles about belief and intention. The unorthodox
interpretation of Kant I offer in Chapter 1 would not have even been conceived had I not read the
paper. Chapter 3 can be seen as an attempt to establish, through a different line of arguments, a
similar view about coherence requirements on intention in general. Steve has been such a

v
dedicated advisor, too. He devoted an incredible amount of time to reading, giving comments on,
and discussing basically everything I have written since I started preparing for the area exam.
Steve can be very critical, but it never let me down because he has always made clear how
excited he is about my project. And it is not an exaggeration to say that my philosophical
writings can be classified into two categories: those written before I started working with Steve,
which consist of some random, unintelligible thoughts floating around without structure; and
those written after I started working with Steve, which are noticeably clearer, more organized,
and more philosophically worthwhile. I’m not sure if I’ll ever be able to come up with something
that meets the high standard Steve sets, but I’m sure that I have become a better philosopher and
a better writer, thanks to his feedback and guidance.
I’d also like to thank my friends outside of the department, for the time they spent with
me and the emotional support they gave me: Ronan Fu, Ben Johnson, Bokkyu Kim, and Sung-
June Pyun. The memories I’ve built with them are an important part of my life in LA. I must
thank Jeonggyu Lee, who started his Ph.D. at UCSB in the same year as I did, for countless pep
talks and philosophical discussions. He knows, perhaps more than anyone else, about the
struggles I had to go through as an (typical, I’d say) Asian student in philosophy.
Finally, I thank my family: my parents, who have always had an unconditional trust in
me and supported my decisions; my wife Jin Yi Han and my son Sean Lee, who always keep me
going. Thank you for standing by me in weal and woe, and for being the meaning of my life.

vi

Table of Contents

Acknowledgments.......................................................................................................................... iii

Introduction .................................................................................................................................. viii

Chapter 1: An Overlooked Reading of Kant................................................................................... 1
1. The Normative Reading ...................................................................................................... 2
2. Variations on the Instrumental Principle ............................................................................ 5
3. Problems with the Normative Readings ............................................................................. 8
4. The Resolute Reading ....................................................................................................... 16
5. The Resolute Reading and the Possibility of Hypothetical Imperativess ......................... 19
6. Worries about the Resolute Reading................................................................................. 24
7. Concluding Remarks ......................................................................................................... 32

Chapter 2: Reasoning, Rational Requirements, and Occurrent Attitudes ..................................... 34
1. Setting the Stage ............................................................................................................... 35
2. Satisfying Rational Requirements Directly through Reasoning: Broome’s Case ............ 36
3. Directness under Threat: Broome’s Dispositional Conception of Belief ......................... 42
4. The Threat Generalized..................................................................................................... 49
5. The Solution: Rational Requirements Govern Occurrent Beliefs..................................... 55
6. Objections and Replies ..................................................................................................... 59
7. Concluding Remarks ......................................................................................................... 61

Chapter 3: The Real Myth of Coherence ...................................................................................... 64
1. Preliminaries: Coherence Requirements and Violability .................................................. 65
2. Constitutive Principles about Intention ............................................................................. 68
3. Compartmentalization and Irrationality ............................................................................ 75
4. Violability to the Rescue?: Going Dispositional .............................................................. 82
5. You Can’t Go Dispositional.............................................................................................. 85
6. Conclusion: Doing without Coherence “Requirements” .................................................. 92

Chapter 4: Settledness and the Rationality of Occurrent Attitudes .............................................. 94
1. Belief as a Settling Attitude: Friedman’s Account ........................................................... 95
2. Incoherence as a Descriptive Relation: Against Friedman ............................................. 103
3. Settledness and Rationality ............................................................................................. 110
4. Revisiting the Rationality of Occurrent Intentions ......................................................... 116
5. The Picture: Principles of Coherence as Descriptive Principles ..................................... 119

vii

Chapter 5: A New Way Out of the Preface Paradox .................................................................. 121
1. The Paradox .................................................................................................................... 122
2. The Standard Solution and its Problems ......................................................................... 126
3. Interlude: Size, Rational Capacities, and Demandingness.............................................. 133
4. Towards the Solution: Occurrent and Dispositional Beliefs ........................................... 135
5. The Solution: Deductive Constraints Govern Occurrent Beliefs .................................... 140
6. Objections and Replies ................................................................................................... 144

Chapter 6: A New Solution to the Problem of Misleading Higher-Order Evidence .................. 148
1. Preliminaries: Misleading Higher-Order Evidence and Conflicting Norms on Belief ... 149
2. The Solution: Being Cautious ......................................................................................... 155
3. The Appeals of ILBC and Occurrent Attitudes .............................................................. 157
4. Responding to Evidence Through Dispositional Belief.................................................. 162
5. Scorecard......................................................................................................................... 168

Concluding Chapter: Doing without Local Coherence ............................................................... 172

References ................................................................................................................................... 182

viii

Introduction

Suppose I assert, in all sincerity, that God created the earth. Suppose I then assert, no less
sincerely, that God did not create the earth. You ask me, ‘so you changed your mind?’, and I say
no. You will probably think that something has gone wrong. Probably I have failed normatively,
in having a pair of contradictory beliefs, which rationality tells us not to have. I breached a rule
of rationality, and that is what has gone wrong. Or so you might think.
But perhaps you should not stop there. You might legitimately wonder if I really have
any belief as to whether God created the earth. The sense that something has gone wrong might
have to do with the unintelligibility of conceptualizing my attitude towards the earth’s origin as a
belief. You might be puzzled because you have naturally assumed that I have expressed my
beliefs in sincerely asserting the propositions but have not been able to make sense of someone
as consciously believing a proposition and believe its negation at the same time. On this way of
seeing the situation, something has gone wrong not because I violated a rule of rationality, but
because I simply cannot have the attitudes that I claim to have.
My dissertation develops this second approach towards a particular set of principles that
have received wide attention in the literature on rationality: the principles of coherence, which
specify the combinations of attitudes one is rationally required not to have. I argue that such
principles are not norms one can fall short of, but rather descriptive (or constitutive) principles
about the attitudes they are seen as governing. It is part of what it is to believe, for example, that
one cannot believe P and believe not-P, which explains your puzzlement in the above case. More

ix
generally, incoherence understood as a violation of a rational requirement is impossible, and that
is why we have hard time making sense of the attitudes of someone who is said to be incoherent.
This is a striking feature of incoherence. Other kinds of normative failure don’t pose any
serious problem for intelligible attitude-ascriptions. Consider immorality. We are enraged at
Hitler’s deeds, but we don’t ever doubt that he intended to kill innocent people for his personal
ambitions. Rather, we blame him precisely because he had such a heinous intention. Or consider
imprudence. We might have pity for someone who only seeks momentary pleasure at the price of
their greater future good, or indulges in self-destructive activities, but we hardly doubt that they
intend to do such things. This suggests that principles of coherence, unlike other normative
principles, have something to do the very conditions for having the attitudes.
The question of whether coherence requirements are really violable is seldom explicitly
addressed in the literature.
1
However, the general idea that principles of coherence are in some
sense constitutive of having belief or intention (or being an agent) has been around for a while. It
dates back at least to Donald Davidson (1985), who argued that the standards of rationality
impose a general constraint on interpreting an agent as having psychological states in such a way
that if the agent deviates from the standards to a significant degree, he cannot be intelligibly
interpreted as having propositional attitudes at all.
2
Christine Korsgaard, in a series of works on
constitutive principles of agency, has defended the idea that complying with some principles of
coherence, such as the principle of instrumental rationality and the Kantian categorical
imperative, is an essential part of being an agent (Korsgaard 1996, 1997, 2009). Ralph
Wedgwood, in the context of providing an interpretation of the idea “the intentional is essentially
normative”, has argued that it is constitutive part of possessing a concept that we have particular

1
Finlay (2009), who argues that means-end incoherence is impossible, is a notable exception.
2
See also Dennett (1987) and Blackburn (1998) for similar views.

x
dispositions to use the concept rationally. With regard to the concept if, for example, the general
disposition to accept instances of modus ponens argument is essential to your having the concept,
such that you wouldn’t count as possessing the concept if, were you not at all disposed to comply
with the modus ponens rule in drawing inferences (Wedgwood 2007, 2009). More recently, Alex
Worsnip (forthcoming) has developed a unifying account of incoherence (or coherence
requirements), according to which a combination of attitudes is incoherent if and only if it is part
of what it is to have those attitudes that for any agent who holds those attitudes, the agent is
disposed to give up at least one of the attitudes under certain conditions.
My dissertation advocates a stronger version of this constitutivist idea. I argue that
principles of coherence are inviolable. They describe how occurrent (or activated) beliefs and
intentions work: one cannot have what people call incoherent combinations of attitudes when the
attitudes are all occurrent, because satisfying such requirements is part of what it is to
occurrently intend (or to believe). This, I argue, gives us the best explanation of the
unintelligibility of, or the puzzlement that accompanies, attributing what people think is an
incoherent combination of attitudes. Moreover, this view captures the respect in which belief
differs from other cognitive attitudes, and intention with other conative attitudes.
One might wonder how this alone can establish that coherence requirements are
inviolable. Surely beliefs and intentions are widely understood as dispositional states, and it
might seem that we are talking about such states when discussing the rationality of belief and
intention. But dispositional states can exist without being activated, as when one is asleep. If so,
beliefs and intentions might fail to be coherent precisely when they are non-occurrent, and so it
might still be possible for the principles of coherence to be violated. In fact, this is the kind of

xi
reasoning which led some philosophers, such as John Broome (2013), Peter Railton (1997), Tim
Scanlon (2007), to quickly dismiss the claim that rational requirements are inviolable.
This question brings us to yet another major theme of my dissertation: coherence
requirements, insofar as they are rational requirements, are plausibly seen as governing
occurrent attitudes. This has to do with the intuitive conditions under which we would assess an
agent as irrational, on grounds of incoherence. The gist of the idea, which will recur throughout
the dissertation, is that the intuition about irrationality of incoherence would be manifest only
when the attitudes are all occurrent. While incoherence between one’s dispositional attitudes is
surely possible, either it doesn’t seem irrational at all, or it seems to be a far less blatant and
unobvious form of irrationality. Why should it be so? Here’s the simple answer I propose:
because coherence requirements govern occurrent attitudes rather than dispositional attitudes.
Once we put this thesis together with the view that occurrent attitudes necessarily satisfy
coherence requirements, we can conclude that they are inviolable.
The reader should be informed from the outset that many arguments in this dissertation
sit well with a particular view of irrationality: irrationality is akin to moral culpability. As Derek
Parfit (2011: 163) notes, “questions about rationality are, in several ways, like questions about
blameworthiness”, which I think has to do with the widely endorsed idea that irrationality is an
especially blatant and severe form of normative failure.
3
In fact, this view explains the ways in
which irrationality is standardly understood in the philosophical literature. It is widely accepted
that, for example, just as one can perform a morally wrong action without being blameworthy (as
when they are mistaken about relevant facts), one can do what one has reason not to do without
being irrational. Just as it is plausible that one is blameworthy for A-ing only if it is within one’s

3
See, for example, Scanlon (1998), Parfit (2011), and Way (forthcoming).

xii
capacity to avoid A-ing, it is plausible that one is irrational for A-ing only if it is within one’s
capacity to avoid A-ing. This view of irrationality seems to be the one that many theorists
implicitly invoke when they insist that rational requirements should not be too demanding for
cognitively limited creatures like us, beings with limited logical skills and memory capacities
(Broome 2013, Harman 1986, Schroeder 2004). This idea has also been explicitly fleshed out in
terms of rational capacities and opportunities: one is rationally required to A only if one has both
the relevant capacity to A and the opportunities to exercise it (Setiya 2004, Wedgwood 2013).
The reason for this clarification is to make sure that the reader doesn’t get off the boat too
quickly. The reader might balk at my intuitive judgments about particular cases or have opposing
judgments. But now that I have clarified how I’ll conceive of irrationality in this dissertation, the
reader will still be able to see where I am coming from, or even to think that I am right as far as
this particular conception of irrationality is concerned.
Here is the plan. In the first chapter, I show how Kant is an important historical precedent
of the constitutivist position I advocate, by offering a novel interpretation of Kant’s claim that
whoever wills the end also wills the indispensably necessary means to it. The orthodox reading
has it that it is a claim about a normative principle of practical reason. I consider both the wide-
scope and the narrow-scope reading of the alleged normative principle and show that neither can
fit into Kant’s view. Then I provide my alternative reading, according to which the proposition
expresses a descriptive principle about how the will works, and my reconstruction of Kant’s
argument for it. The neglected reading of Kant I present motivates one major thesis of the
dissertation: instrumental irrationality, understood as incoherence between one’s attitude towards
an end and one’s attitude towards the believed means it, is impossible.

xiii
In the second chapter, I tackle Broome’s (2013) recent view of reasoning and rational
(coherence) requirements and provide one important reason why rational requirements on
attitudes should be understood as governing occurrent attitudes, which is another main thesis of
my dissertation. Broome plausibly holds, first, that we can satisfy rational requirements directly
through reasoning (when we satisfy them through reasoning). He also holds that the attitudes
targeted by such requirements are dispositional (or standing) states, which initially seems natural.
However, I argue that these views cannot go together, since dispositional states aren’t the right
kind of things to be directly affected by reasoning. As a solution, I propose that we should see
the attitudes governed by rational requirements as occurrent attitudes.
The third chapter combines these two ideas and argues that some widely accepted
coherence requirements on intention are in fact descriptive principles about how (occurrent)
attitudes work. On the basis of the kind of unintelligibility of ascribing incoherent combinations
of attitudes, I first argue that the said requirements are necessarily satisfied when one’s attitudes
are all occurrent and one’s mind isn’t compartmentalized. Then I argue that when some of one’s
relevant attitudes are non-occurrent, or when one’s mind is compartmentalized, one isn’t subject
to rational requirements. I provide further support for the view that coherence requirements
govern occurrent attitudes, on the grounds that it best explains our intuitions about the
irrationality of violating such requirements. The view that compartmentalization puts one outside
of the scope of rational requirements relies on a particular view about the relationship between
rationality and rational capacities, which I develop in detail also in this chapter.
The fourth chapter extends the same approach to some prominent coherence requirements
on belief. In doing so, it explains what is special about occurrent beliefs and intentions. In
particular, I argue that occurrent belief and intention each involves being settled on a relevant

xiv
question: believing p involves being settled on the question of whether p is true; intending to A
involves being settled on the question of whether to A. It is precisely this feature occurrent
beliefs and intentions, I suggest, that makes it plausible that one cannot violate the said
requirements with occurrent attitudes.
The fifth and the sixth chapter bracket the question of violability and focus on
strengthening the view that coherence requirements govern occurrent attitudes. In particular, I
show how this view can offer novel and neat solutions to some puzzles in epistemology: the
preface paradox; and the puzzle about misleading higher-order evidence. While the preface
paradox has the standard, probabilistic solution, I point out that the standard solution cannot
handle some intuitive judgments about consistency and closure requirements on beliefs, and, the
better solution lies in setting the jurisdiction of such requirements right: such requirements only
govern occurrent beliefs. The puzzle of misleading higher-order evidence admits of the same
kind of solution. While cases of misleading higher-order evidence are often seen as giving rise to
a dilemma between the requirement to believe in accordance with evidence and the requirement
that one’s higher- and first-order beliefs cohere with each other, I argue that we can avoid the
dilemma again by clarifying the proper jurisdiction of each: the former governs dispositional
beliefs; and the latter governs occurrent beliefs.

1

Chapter 1: An Overlooked Reading of Kant

In his Groundwork of the Metaphysics of Morals, Kant famously calls the following proposition
“analytic”: “whoever wills the end also wills (in so far as reason has decisive influence on his
actions) the indispensably necessary means to it that is in his control” (4:417).
4
Read naïvely,
with little attention to the caveat in the parenthesis, this proposition is most straightforwardly
interpreted as specifying a descriptive relation between willing an end and willing the necessary
means to it. It simply tells us what we do in willing an end, not what we ought to do.
However, contemporary commentators near-unanimously reject the naïve reading of the
analytic proposition. Their basic idea is that the analytic proposition specifies a requirement of
practical reason telling us that we rationally ought to will the necessary means to the ends we
will. Call this the normative reading of the proposition. This wide acceptance of the normative
reading is hardly surprising, since Kant repeatedly describes hypothetical imperatives as
imperatives, that is, as claims about what we ought to do, as opposed to what we do, which can
be hard to understand if the analytic proposition is a descriptive claim about willing an end.
This chapter argues that this orthodox consensus is mistaken, despite the initial
attractions of the normative reading. To this end, I consider two salient readings of the analytic
proposition that have appeared in the literature, and argue that each is incompatible with what
Kant obviously commits himself to in the text (Sections 2 and 3). When the normative reading
comes into conflict with Kant’s more obvious philosophical commitments, it is the normative
reading that has to go. Having rejected the normative readings, I defend a version of the naïve

4
Page references are to the Akademie edition pagination. I use the translations that appear in the bibliography.

2
reading of the analytic proposition, which I call the resolute reading. This preserves the basic
idea of the naïve reading that Kant is identifying a descriptive relation, while at the same time
giving due regard to the caveat in the parenthesis. The crux of the resolute reading lies in
interpreting the caveat as referring to the conditions under which you have the knowledge of the
means: if you know the means, then willing the end entails willing the means. I shall defend the
resolute reading by showing how it sits well with what Kant obviously commits himself to
(Section 4), how nicely it explains the possibility of hypothetical imperatives (Section 5), and by
responding to the objection that this view is too implausible to be Kant’s (Section 6).

1. The Normative Reading
According to the normative reading of the analytic proposition, it expresses a normative principle
of practical reason, which tells us what we ought to do. It finds its explicit expression at least as
early as 1947 in H. J. Paton’s The Categorical Imperative:

First of all, we have the objective principle of practical reason, ‘Any rational agent who
wills the end will necessarily—so far as reason has a decisive influence over his
actions—will the means which are in his power.’ This proposition, which is still analytic,
appears as an imperative to us because reason, though present in us, has no such decisive
influence. It then takes the form ‘If any rational agent wills the end, he ought to will the
means. (Paton 1947: 124)

3
The normative reading is explicitly endorsed in a large number of writings on Kant’s view of
hypothetical imperatives.
5
Indeed, strong reasons have been offered in the literature to favor the
normative reading over the naïve, descriptive reading.
To begin with, it is widely agreed that the analytic proposition seems simply false unless
qualified by the caveat in the parenthesis. First, you might fail to will the necessary means to
your ends if you don’t know what they are. As Christine Korsgaard (1997: 236) points out, “It is
not true that if someone wills to be healthy, then he necessarily wills to exercise. He must also
know that exercise is the cause of health.” Second, it has been taken to be obvious that you can
sometimes fail to will the necessary means to the ends you will even when you know the means,
owing to your irrationality: you might will the end of building muscle, know that taking exercise
is a necessary means to it, and yet irrationally stay in bed, due to your laziness.
6

It is also widely agreed that the naïve reading is difficult to reconcile with Kant’s
conception of an imperative. Kant states the proposition in the context of explaining the
possibility of hypothetical imperatives, which, according to Kant, are expressible by an ‘ought’,
representing the “practical necessity” of a particular action for a presupposed end. Such necessity
counts as an “imperative”, and hence normative, only for an imperfectly rational being whose
will is not always in line with reason (4:413). But if the naïve reading were true, it would be
impossible for one to fail to will the necessary means to the ends one wills, and the sense in

5
Besides Paton (1947), see Beck (1960), Hill (1973; 1989), Korsgaard (1986; 1997; 2009), O’Neill (1989),
Hampton (1998), Wood (1999; 2013), Rawls (2000), Reath (2006), and Galvin (2009).
6
See, for example, Beck (1960: 87): “Since beings like us do not, in fact always will the means necessary to their
ends, even when they know the means, the hypothetical imperative expresses a constraint of reason upon impulse.”
Hampton (1998: 131-2) dismisses the naïve reading as “highly implausible”, because “not only does introspection
controvert it, more importantly, it is a psychological thesis that would make irrational action impossible.” This point
is made repeatedly in Korsgaard (1986; 1997; 2009), in her critique of the Humean account of instrumental
rationality. See also Wood (1999; 2013).

4
which a particular hypothetical imperative like ‘if you want to build muscle, you ought to take
regular exercise’ is an imperative would be unclear.
7

The normative reading is reached by paying closer attention to the caveat in the
parenthesis, insofar as reason has decisive influence, which seems to imply that only a fully
rational person, whose will is decisively influenced by reason, would always will the necessary
means to her willed ends. If we took the claim to be that if you will an end you will the means in
so far as you are fully rational, the analytic proposition would plausibly be equivalent to: if you
will an end, you rationally ought to will the necessary means to it.
8

The normative reading is immune from the problems. First, it is consistent with the fact
that we can fail to will the necessary means to the ends we will, and therefore doesn’t conflict
with Kant’s view that ordinary hypothetical imperatives are imperatives for imperfectly rational
beings like us. Second, the normative reading specifies one way in which the analytic
proposition, as Kant intended, explains the possibility of hypothetical imperatives: the analytic
proposition expresses a general principle of practical reason that justifies or grounds ordinary
hypothetical imperatives like ‘if you want to build muscle, you ought to take regular exercise’,
which we might call the Instrumental Principle (‘IP’, henceforth).
9

The normative reading, then, is the view that the analytic proposition is a version of IP.
The influence of the normative reading can also be shown by considering how it has been
presupposed in some recent philosophical debates. First, consider the dispute about whether the

7
See, for example, Paton (1947: 124): “If to will the end is to will the means, how can it be said that a rational agent
ought to will the means? Does not the latter statement imply that in fact it is possible to will the end and yet not to
will the means?”, and Hill (1973: 431): “Kant explicitly refers to ‘the imperative which commands him who wills
the end to will the means,’ and there could be no such command or imperative, by Kant’s doctrines, if failure to
conform were impossible.” See also Korsgaard (1997) and Wood (1999; 2013).
8
This reasoning is most explicit in Korsgaard (1997: 240)
9
Hill (1973: 431) takes IP to be a general premise in practical reasoning, which justifies particular premises.
Schwartz (2008) disagrees, arguing that it is rather a rule of inference which doesn’t itself figure as a premise in
reasoning.

5
‘ought’ in IP, ‘if you will an end, you ought to will the necessary means to it’, logically takes
wide or narrow scope over the conditional. All participants in the debate take it for granted that
the analytic proposition is equivalent to some version of IP, and disagree over which version
correctly represents Kant’s view.
10
The whole debate, however, would turn out misconceived if
Kant didn’t put forward any normative principle like IP.
Second, it is also presupposed in so-called “neo-Kantian” arguments against the neo-
Humean view on which instrumental rationality is all there is to practical rationality.
11
The gist
of the arguments is that there is no principled reason not to accept other rational norms like the
Kantian categorical imperative, once we accept IP as a principle of rationality: since IP must be a
requirement which is valid for all rational beings as such, anyone who accepts IP as a principle
of practical reason is committed to accepting the existence of a non-hypothetical imperative,
whose validity does not depend upon our contingent ends. Such arguments, whether successful
or not, would not be Kantian, and lack the authority of Kant which is usually claimed for them, if
Kant himself did not accept a principle like IP.

2. Variations on the Instrumental Principle
Supposing that Kant accepted IP, how exactly should it be formulated? It would posit a
normative relation between willing an end and willing the necessary means to it. Since Kant
thought that imperatives are requirements of reason, expressed with ‘ought’, I will use the
concept of rational ought, although nothing substantial hinges on my choice of the specific
normative relation. All that matters is that it is a concept of normative necessity, and the

10
See, e.g., Greenspan (1972), Hill (1973, 1989); Darwall (1983), Schroeder (2005, 2015), Siyar (2013), and Rippon
(2014).
11
See, e.g., Korsgaard (1986, 1997, 2009), Hampton (1998), and Wood (1999).

6
necessity has its source in reason or rationality. Since Kant did not make all of the fine-grained
distinctions between normative concepts familiar to contemporary writers, one might make do
with other normative concepts, such as: is obligated to, is rationally required to.
12
With this in
mind, let’s formulate IP as follows:

(Bare) Instrumental Principle: If an agent wills an end, then they rationally ought to
will the necessary means to it that are within their control.
13

As recognized in the literature, IP admits of the following readings, depending on the logical
scope of ‘ought’ in IP:

IP-Wide: An agent rationally ought to be such that if they will an end, they will the
necessary means to it.
14

IP-Narrow: If an agent wills an end, then they rationally ought to will the necessary
means to it.
15

Let me formalize each in a way that makes explicit the subject-predicate structure in it. This is
because Kant famously held that a proposition is analytic just in case the predicate concept is

12
I shall not split hairs over the normative word itself, and simply assume that ‘ought’, in Kant’s own usage,
expresses a concept of necessity. But the idea that ‘ought’ expresses necessity could be controversial. See, e.g.,
Sloman (1970). Paradigm words for necessity are ‘must’ and ‘have to’, rather than ‘ought’. Kant even uses ‘good’,
which obviously does not express a concept of necessity: “[imperatives] say that to do or to omit something would
be good…” (4:413). Again, I simply assume that it expresses the same concept as ‘ought’ in Kant’s usage.
13
For the sake of simplicity, I shall deliberately omit Kant’s phrase ‘that is within your control’, unless it becomes
particularly relevant.
14
See Greenspan (1972), Hill (1973), Darwall (1983) and Rippon (2014).
15
See Schroeder (2005).

7
“contained” in the subject concept, and one must be able to identify a subject-predicate structure
in order to establish either IP-Wide or IP-narrow as the correct interpretation of the analytic
proposition. If we let ‘A’ stand for being an agent, ‘We’ for willing an end, and ‘Wm’ for willing
the necessary means to the end, IP-Wide and IP-Narrow can be formalized as follows:

IP-Wide: A → Ought: (We → Wm)
IP-Narrow: We → Ought: Wm

In IP-Wide, ‘ought’ takes wide scope over the whole conditional, whereas it takes narrow scope
over the consequent of the conditional in IP-Narrow. IP-Wide and IP-Narrow differ in two
important respects: (i) the jurisdiction of the requirement, or when an agent is under the
requirement; (ii) the object of the requirement, or what the agent who is under it is required to
do. Every agent is under IP-Wide, and is rationally required to satisfy a disjunction, with respect
to any end: either not have the end or will the necessary means to it. Since you are required to
satisfy the disjunction, you can comply with IP-Wide in two ways: either by giving up the end or
by willing the means. On IP-Narrow, only those who will an end are under the requirement to
will the necessary means to it. IP-Narrow itself does not require anything of someone who
doesn’t will the end. Moreover, the object of the requirement is not disjunctive: once you are
under a requirement, you can only satisfy it by willing the means.
16

16
It should be noted that IP is, in a crucial way, distinct from the following principle, which the recent debate on the
scope of instrumental rationality concerns: you are rationally required to intend m, if you intend e and believe that m
is a necessary means to e. Here, the relevant normative relation is taken to hold between intending an end and
intending what is believed to be, rather than what in fact is, a necessary means to it. While such a principle merits a
discussion in its own rights, and will be discussed in detail in Chapter 3, it is at best unobvious how it could serve as
the correct interpretation of the analytic proposition, since no condition about belief is present in the proposition. For
a variety of wide-scope interpretations of this belief-relative instrumental principle, see Way (2010), Brunero
(2012), and Broome (2013). Prominent defenses of the narrow scope reading include Kolodny (2005), Bedke
(2009), and Schroeder (2009).

8

3. Problems with the Normative Readings
This section argues that neither IP-narrow nor IP-Wide can be a plausible interpretation of the
analytic proposition, because each runs contrary to Kant’s more obvious philosophical
commitments. For each reading, I shall survey some standard objections to it, consider existing
responses, and argue that they are not satisfactory.

3.1. IP-Narrow and the Problem of Immoral Means
There is the notorious problem of immoral means for IP-Narrow, which has been the main
reason for rejecting the narrow-scope interpretation of IP.
17
The basic assumption generating the
problem is Kant’s ethical rationalism: an act is wrong only if it is rationally impermissible; and
this means that the maxim on which the act was performed fails the test of the categorical
imperative. On Kant’s view, therefore, it is implausible that for any end you might happen to
will, you rationally ought to will the necessary means to it.
To see this, note first that the end itself can be immoral. Suppose your end is to kill an
innocent person grotesquely, and a necessary means to doing so is to use the electric saw near
you. IP-Narrow would predict that you rationally ought to will that you use the electric saw (to
kill the person), but this is clearly unacceptable from the standpoint of Kant’s ethical rationalism.
Second, even when the end itself is permissible, a necessary means to it might be immoral. To
cite a familiar example, if your end is to save dying patients but the only available way to do so
is to pull out organs from a healthy, innocent person around you, IP-Narrow would predict that
you rationally ought to will that you do so. In either case, you would be rationally required to

17
See Hill (1974, 1989), Darwall (1983), Korsgaard (1997). A more general form of this problem has been called
the bootstrapping problem in recent literature on rational requirements. See Broome (1999) and Kolodny (2005).

9
will the immoral means, which is, at the same time, prohibited by rationality. But the idea that
rationality could make such conflicting demands is dubious.
18

There have been attempts to defend IP-Narrow from the problem of immoral means. The
general tactic is to give up the idea that hypothetical imperatives apply to agents independently
of the categorical imperative and see hypothetical imperatives as rationally binding an agent only
when their ends meet the constraint given by the categorical imperative. For example, Mark
Schroeder (2005) argues that the categorical imperative, for Kant, is a constitutive principle of
willing. Thus, immoral ends failing its test cannot be willed. Jamsheed Siyar (2013) argues that
while it is possible for you to will immoral ends, the consequent of a narrow-scope requirement
detaches only if you permissibly will the end. In any case, the solution lies in reinforcing the
antecedent of IP-Narrow so that the consequent does not detach when the end is immoral. As a
result, we get:

Sophisticated IP-Narrow: If an agent permissibly wills an end, then they rationally
ought to will the necessary means to it.

IP-Narrow, however, hardly solves the problem of immoral means.
19
When introducing the
problem, I considered two kinds of cases. In the first, the end itself is immoral. In the second, the
end is permissible but the necessary means is immoral. Sophisticated IP-Narrow deals with cases
of the first kind by not allowing the consequent to detach when the end is impermissibly willed,
but it cannot handle cases of the second kind.

18
See, for example, Hill (1973: 436).
19
See Brunero and Kolodny (2013) and Rippon (2014) for other independent problems for Sophisticated IP-Narrow.

10
To see why, we only need to suppose that when you (permissibly) will an end, you need
not be aware of all of the necessary means to it. This assumption should not be controversial,
since we often set up an end first without having all of the means in view, and then go on to
deliberate about how to achieve the end. If this is right, it seems perfectly possible for you to will
an end, some of the necessary means to which is, unbeknownst to you, clearly immoral. In such a
case, Sophisticated IP-Narrow requires that you take what is in fact a necessary means to the end,
which is immoral. So the problem of immoral means remains.
One might reply that an end cannot be permissible if what is in fact a necessary means to
it is impermissible, and so IP-Narrow does not imply that you are required to will the immoral
means in such a case. This response is problematic, however. Even when M-ing is clearly
immoral and M-ing happens to be a necessary means to E-ing in a particular situation S, and it is
therefore in fact impermissible for you to E in S, you can still permissibly will E-ing, if you are
ignorant of the fact that M-ing is the only way to E in S. Suppose my end is to get my child’s
favorite toy for her birthday and I am not aware of any immoral means to doing so. It clearly
seems that I can still permissibly will the end. Suppose, however, that the toy is in fact out of
stock, and the only way to get it is to steal it from someone who already has it. Sophisticated IP-
Narrow implies that I ought to will that I steal the toy, which is immoral.

3.2. IP-Wide and the Problem of False Categoricity
IP-Wide avoids the problem of the immoral means. Even when the necessary means to your end
is immoral, IP-Wide wouldn’t imply that you ought to will the immoral means, because you have
the other option: giving up the end. However, it faces problems of its own.

11
One important problem for IP-Wide, already recognized in the literature, is that IP-Wide
seems to imply that it is a categorical imperative, rather than a hypothetical imperative.
20
To see
why it has such an implication, recall that IP-Wide is a requirement applying to all agents,
irrespective of their particular desires or ends: all rational agents, regardless of their contingent
ends, are required to satisfy a disjunction. Moreover, even ordinary hypothetical imperatives
must share this feature, since they are supposed to be grounded by or derived from IP. But this
just is what Kant takes to be the essential feature of a categorical imperative.
This implication is problematic in two ways. First, it obscures the sense in which
hypothetical imperatives are hypothetical. Kant makes it clear that a hypothetical imperative is
an imperative only for those who will the relevant end. After classifying hypothetical imperatives
into rules of skill and counsels of prudence, he writes,

[Since] both merely command the means to that which one presupposes one wills as an
end, the imperative that commands willing the means for someone who wills the end is in
both cases analytic. (4: 419, emphases added)

Second, this is inconsistent with Kant’s claim that there is only a “single categorical imperative”,
which is his famous Formula of Universal Law: act only according to the maxim through which
you can at the same time will that it become a universal law. (4: 421, emphasis added) Surely,
however, the Formula of Universal Law is not equivalent to IP, since IP is in no way a principle
about the universalizability of your maxim. If they cannot be equivalent, then Kant’s claim that

20
See Schroeder (2005) and Schwartz (2008).

12
there is only one categorical imperative must be false. Call this the problem of false
categoricity.
21

Rippon (2014: 786-7) replies to the problem of false categoricity by identifying a
different sense in which an imperative can be hypothetical.
22
A wide-scope imperative like ‘you
rationally ought to be such that you will that you take regular exercise if you will the end of
building muscle’, on his view, is still hypothetical in the sense that it provides normative
guidance only for those who satisfy the antecedent of the conditional, that is, who will the end of
building muscle. If you lacked this particular end, for example, this wide-scope requirement
would be entirely irrelevant to your practical deliberation.
This response to the problem of false categoricity is problematic, because the wide-scope
interpretation still fails to capture the sense in which a hypothetical imperative is supposed to
provide normative guidance only for those who satisfy the antecedent of the conditional. To be
sure, wide-scope imperatives can be guiding, in the way Rippon suggests. Someone who wills an
end might, on the basis of her recognition of a relevant wide-scope imperative, conclude that
either she ought to will the means or give up the end, and decide to satisfy a particular disjunct,
perhaps by reference to further normative principles. The wide-scope requirement plays a

21
One might worry that this is not a significant problem, on the grounds that Kant in fact offers three formulations
of the moral law in Groundwork. One might argue that these formulations might not be equivalent, as a number of
writers (Wood 1999; Kerstein 2002; Formosa 2017) argue, and if so, we should not take Kant’s claim that there is a
single categorical imperative seriously. What matters in this context, however, is that Kant himself thinks that the
three formulations “are at bottom only so many formulae of the very same law, and any one of them or itself unites
the other two in it” (4: 436), and it is obvious that he does not regard anything like IP, which has nothing to do with
morality, as an expression of that law.
22
Hill (1973: 443) also defends IP-Wide from the problem of false categoricity by offering a different sense in
which it can be hypothetical: “Imperatives are hypothetical if either they support particular prescriptions for a
person only in conjunction with premises describing that person’s ends or they cannot themselves be supported
without premises describing the ends of the person to whom they are directed.” Hill thinks that IP-Wide satisfies the
first disjunct and therefore is hypothetical. But I think this is vulnerable to the same objection I raise against
Rippon’s suggestion.

13
guiding role in her deliberation, in the sense that her recognition of it leads her to deliberate
further about which disjunct to choose, and eventually leads her to satisfy it.
The problem is that this kind of guidance is not offered exclusively to those who will the
end. Imagine someone who hates taking exercise and firmly believes that he will never take
regular exercise in his life. Suppose he comes to recognize the wide-scope hypothetical
imperative, ‘you ought to be such that if you will the end of building muscle, you will that you
take regular exercise’, and take measures to prevent himself from willing the end, in order to
comply with the requirement. In this case, the person never satisfies the antecedent: he only
considers the end without willing it. Still, the wide-scope requirement seems to provide him with
normative guidance of the same kind: his recognition of the requirement leads him to deliberate
further about which disjunct to choose, which eventually leads him to satisfy it. Thus, there is no
clear sense in which wide-scope requirements are normatively guiding only for those who will
the relevant ends, and so the sense in which they are hypothetical is still obscure.

3.3. IP-Wide and the Problem of No Subject (or No Predicate)
Another problem for IP-Wide, posed trenchantly by Schroeder (2005: 362-7), is that it doesn’t
seem to be what we can derive straightforwardly from the analytic proposition, if we assume
Kant’s own notion of analyticity. Recall that a proposition is analytic just in case the predicate
concept is “contained” in the subject concept. The target proposition, whoever wills the end also
wills the means (in so far as reason has decisive influence on his actions), has the form ‘all Fs
are Gs’, which analyzes the subject concept of being F into the predicate concept of being G. So
understood, the analytic proposition is clearly supposed to be an analysis of willing an end.

14
As Schroeder (2005: 363) points out, however, IP-Wide does not derive the ought-claim
from the concept of willing at all. It says that all agents have the following property: they
rationally ought to satisfy the relevant disjunction. But this is an analysis of the concept of an
agent, rather than the concept of willing.
23
Indeed, some wide-scopers interpret the analytic
proposition as the analysis of the concept of rational agent. For example, Hill (1989: 365)
casually recasts the analytic proposition as follows: “any fully rational agent (necessarily) wills
the necessary (and available) means to his ends.”
24

This slide from willing to agency or rational agency, however, puts pressure on IP-Wide,
since Kant asserts outright that the analytic proposition is an analysis of the concept of willing.
We can of course grant that IP-Wide might be a successful analysis of (rational) agency. It could
be argued, for example, that since (i) rationality (analytically) requires us to do what a fully
rational agent would do
25
and (ii) it is analytically true that a fully rational agent wills the
necessary means to the ends she wills, IP-Wide must also be analytically true. The problem is
just that this is inconsistent with Kant’s claim, since Kant’s analysandum is not the concept of
rational agent. Call this the problem of no subject.
In response to the problem of no subject, Rippon (2014) concedes that it is hard to make
sense of IP-Wide as an analysis of the concept of willing, but argues that we need not narrowly
restrict ourselves to the subject-predicate conception of analyticity. According to the alternative
view of analyticity he attributes to Kant, “a claim is analytic just if its negation would be a
contradiction when its conceptual containments are fully spelled out” (Rippon 2014: 787).

23
See Shaver (2006) for a similar point.
24
Korsgaard (1997) also suggests that Kant derives the normative principle in the same way, but that is a slip into
his “pre-critical rationalism” for Kant to do so.
25
One might think, however, this is false, let alone analytically. See, e.g., Smith (1995).

15
However, this response at best gives us reason to think that Kant might have multiple
conceptions of analyticity, not that this alternative conception of analyticity is invoked in the
particular context in which Kant states the analytic proposition. The textual evidence strongly
suggests that it is the subject-predicate conception of analyticity that Kant has in mind, as he
makes it explicit that the proposition is analytic “as far as willing is concerned” and also that a
hypothetical imperative “extracts the concept of actions necessary to this end merely from the
concept of a volition of this end.” (4:417) So the problem of no subject remains.
One might justify taking the proposition as an analysis of agency, arguing that there is no
deep distinction between the project of analyzing the concept of willing and the project of
analyzing the concept of agency: after all, it is only agents who will in the first place, and it
might seem quite natural to think that Kant’s analytic proposition is an analysis of the concept of
an agent as qualified by the concept of willing.
But this response fails, for it falls prey to what I call the problem of no predicate. This is
just the same problem seen from the other side: so long as one removes the concept of willing an
end from the subject-place, one cannot but find a place for it in the predicate-place, which leads
to an implausible reading of the proposition. To see this, note that Kant is just as explicit on the
analysans as he is explicit on the analysandum. On the most straightforward reading of the
analytic proposition, the predicate concept he wants to derive from the subject-concept is the
concept of willing the necessary means that is in his control. The same is true of another
proposition which he calls “analytic”: “When I know that only by [an action] can the proposed
effect take place, then it is an analytic proposition that if I fully will the effect I also will the
action requisite to it” (4:417, emphasis added).

16
According to IP-Wide, however, the predicate concept must be (being rationally
required) either not to will the end or to will the necessary means to it. This would be a
surprisingly artificial way of identifying the predicate in the sentence, “whoever wills the end
also wills … the indispensably necessary means to it.” Moreover, the disjunction under the scope
of the requirement is logically weaker than willing the means, simpliciter. So if IP-Wide were the
correct interpretation of the analytic proposition, it would remain unexplained why Kant would
make the stronger claim that the predicate concept is the concept of willing the necessary means,
instead of either giving up the end or willing the necessary means.

4. The Resolute Reading
Here is an interim report. The available normative readings of the analytic proposition, as we
have seen, confront serious interpretive problems, because of the assumption that the analytic
proposition expresses a principle about rational ought. But we have also seen that the naïve
reading, in leaving no gap between willing an end and willing the necessary means, flies in the
face of the obvious fact that we often fail to will the necessary means to our ends. The task,
therefore, is to find a reading which does not turn the analytic proposition into a normative claim
about rational ought, but still leaves a gap between willing an end and willing the necessary
means to it, a gap sufficient to avoid denying the obvious.
In this section, I outline such a reading of the analytic proposition. I shall call it the
resolute reading, not just because I need a label that distinguishes it from the obviously false
naïve reading, but also because it remains most faithful to what Kant says. The orthodox view
fills the gap with a normative condition, but as we have seen, this is what leads it into problems.
The resolute reading fills this gap, instead, with an epistemic condition: when you are ignorant

17
of, or mistaken about what the means are, you can fail to will the means. On this reading, the
caveat in the analytic proposition, in so far as reason has decisive influence on your action,
refers to the conditions under which you have full knowledge of the means. This is consistent
with Kant’s view of the roles of reason, because Kant is explicit that it is reason’s job to figure
out the means to our ends: “appraisal of the relation of means to ends certainly belongs to
reason” (5: 59).
26
But the resolute reading has it that, when the epistemic condition is met, and
you have the knowledge of the means, you cannot fail to will them. This reading leaves a gap
between willing the end and willing the means, but captures the essence of the naïve reading: the
idea that the analytic proposition expresses a descriptive relation.
The resolute reading focuses on Kant’s remarks immediately following his statement of
the analytic proposition, on which commentators have been surprisingly silent:

In the volition of an object as my effect, my causality as acting cause, that is, the use of
means, is already thought, and the imperative extracts the concept of actions necessary to
this end merely from the concept of a volition of this end. […] When I know that only by
[an action] can [the effect] take place, then it is an analytic proposition that if I fully will
the effect I also will the action requisite to it. It is one and the same thing to represent
something as an effect possible by me in a certain way and to represent myself as acting
in this way with respect to it. (4: 417, emphases added)

26
See also [5:61]: “The human being is a being with needs, insofar as he belongs to the sensible world, and to this
extent his reason certainly has a commission from the side of his sensibility which it cannot refuse, to attend to its
interest and to form practical maxims with a view to happiness in this life.”

18
Kant’s intriguing argument in this passage proceeds as follows. When you will an end, you
represent an object as something to be brought about, that is, as an end. Since it is through your
action that it is to be brought about, to represent it as an end just is to represent it as a possible
effect to be brought about by your action, which is its cause. This is the sense in which the “use
of means”, which is your action, is already thought in willing an object. What this means is that
when you will an end, the content of your willing is, strictly speaking, your acting so as to bring
about the end, where your action is conceived of as using a means to the end.
The sentence ‘you will that you take what is a necessary means to your end E’ is
ambiguous, however. On the first, de re reading, it means that for any object that is in fact a
necessary means to E, you will that you take it. On the second, de dicto reading, it means that the
content of your willing is: to do whatever is necessary to E.
27
Kant’s claim up to this point is that
when you will an end, you will, de dicto, that you take the necessary means to it. On the de dicto
reading, it is true, without any reservation, that in willing an end you also will the necessary
means to it. But you can still fail to will the necessary means de re. You can will, de dicto, that
you take whatever means is necessary for your end, without having any idea about what the
means are: you might be ignorant of, or mistaken about them. On this reading, therefore, it is still
possible for you to fail to will the necessary means to your end de re, because you might not
have the right beliefs about the necessary means.
This is why Kant needs to add the epistemic condition “when I know”. Willing the end
implies willing the means de re only when you know what the necessary means are. Suppose that
you come to have the knowledge of the necessary means to your end, which combines with your
de dicto willing of the end. Since the content of your willing is that you take whatever is

27
Here I adopt Dreier’s explanation of de dicto and de re readings of a desire-ascription sentence. (Dreier 2000)

19
necessary to your end, willing the end and willing the means come down to “one and the same
thing”. Given the content of your willing of the end and your right belief about the objects that
are the necessary means, you will the necessary means to your end, both de dicto and de re, so
long as you will the end. This implies that if you do not will the necessary means to the end, in
the knowledge of the necessary means, you do not count as willing the end.
28

One virtue of the resolute reading is that it avoids all of the exegetical problems for IP.
On the resolute reading, Kant is making no claim about what you rationally ought to do, and for
good reason: nowhere does Kant explicitly assert the existence of a requirement of practical
reason to will the means to your ends. It avoids the problem of false categoricity: since there is
no rational requirement that Kant is identifying, there is nothing to function as a categorical
imperative besides the moral law. Since it is explicitly about the concept of willing an end, and
derives directly the concept of willing the means, the problem of no subject does not arise. It also
evades the problem of immoral means: Kant’s claim is not about what we rationally ought in
virtue of having ends, but rather about what is constitutively involved in willing an end. To see
this, return to the example of willing the end of killing an innocent person grotesquely. The
resolute reading of the analytic proposition doesn’t say that you rationally ought to use the
electric saw. It only says that you will also will the use of the saw, if you know that it is a
necessary means, without saying that it is what you rationally ought to do.

5. The Resolute Reading and the Possibility of Hypothetical Imperativess
Recall one of the main reasons for accepting the normative reading. (Section 1) The task Kant
sets for himself in [4:417] is to explain the possibility of hypothetical imperatives, and the

28
This argument bears striking resemblance to Finlay’s argument that instrumental irrationality (with respect to
intention) is impossible. See Finlay (2009).

20
normative reading suggests one way it can be carried out: since IP is a normative principle about
what we rationally ought to do, we can explain how the normative ought-claims made by
ordinary hypothetical imperatives can be true in term of IP. On the resolute reading, however, the
analytic proposition is a descriptive claim about the will. So it remains to be shown how
hypothetical imperatives are possible. If this can be done without invoking IP, one of the major
reasons cited for accepting the normative reading will no longer have force.

5.1. Hypothetical Imperative as “Theoretical Principles”
What is a hypothetical imperative, the possibility of which is what Kant wants to explain? It is a
proposition representing an action as practically necessary, but necessary as a means to some
presupposed end. What matters is the kind of necessity it expresses. Proponents of IP assume that
IP, which explains the possibility of hypothetical imperatives, expresses rational necessity and
conclude that particular hypothetical imperatives must also express rational necessity.
However, an imperative like ‘you ought to take regular exercise in order to build muscle’
is most naturally seen as a factual statement about what is needed for a possible end, despite its
use of an apparently normative word ‘ought’. Call this sort of necessity instrumental necessity.
Kant’s remark on rules of skill suggests that they express instrumental necessity, as opposed to
rational necessity:

There is no question here whether the end is rational and good, but only what one must
do in order to attain it. The prescriptions for the physician thoroughly to cure his man,
and for a poisoner reliably to kill him, are of equal worth, insofar as each serves to effect
its purpose perfectly. (4: 415)

21

Since the poisoner’s end is rationally impermissible, in the sense that it fails the test of the
categorical imperative, using the prescriptions for this end cannot be rationally willed. In that
case, the sentence ‘the poisoner ought to use the prescriptions’ is false if it is taken to express
rational necessity. Kant’s point is that the sentence is true because it, as a rule of skill, expresses
instrumental necessity: he ought to use the prescriptions in order to reliably kill his man. A rule
of skill merely tells us that a certain result is caused (only) by a certain action.
Moreover, Kant explicitly commits himself to the idea that rules of skill (or “technical
imperatives”) are not normative propositions but rather theoretical propositions in his later
works. See his remarks in Critique of Practical Reason, for example: “Principles of self-love can
indeed contain universal rules of skill, … but in that case they are only theoretical principles
(such as, e.g., how someone who would like to eat bread needs to construct a mill)” (5: 26,
emphasis added). A similar thought appears in his Critique of Judgment: “all precepts of skill
belong to the technic of nature, and hence to our theoretical knowledge of nature” (20: 200).
Counsels of prudence, hypothetical imperatives of the other kind, are no different in this
regard: they are theoretical principles about how to attain happiness. Indeed, Kant later denies
that there is a significant distinction between the two kinds of imperative. In Groundwork, Kant
calls rules of skill “problematic imperatives”, which presuppose merely possible ends that need
not be had by any actual agent, and calls counsels of prudence “assertoric imperatives”, which
presuppose ends actually had by agents. (4:416) In Critique of Judgment, he dismisses this
distinction as mistaken, and includes the latter under the former. This makes perfect sense if we
take hypothetical imperatives to be theoretical truths, because whether you have the end is
irrelevant to the truth of a theoretical proposition, such as you ought to take exercise in order to

22
build muscle. Thus, understanding hypothetical imperatives as theoretical truths not only has
intuitive grounds, but renders Kant’s view in Groundwork continuous with Kant’s later works.

5.2. Why Call them “Imperatives”?
But the deeper interpretive question is the following: if a hypothetical imperative is in itself a
factual judgment, which expresses instrumental necessity rather than the kind of rational
necessity invoked by IP, in what sense is it an “imperative”? If we heed our ordinary practice, for
instance, there is the fact that we issue hypothetical imperatives to others in the forms of
command or advice, and also the fact that we, as deliberating agents, are guided by hypothetical
imperatives in our deliberation about what to do.
My answer is that ordinary hypothetical imperatives can play such guiding roles because
such imperatives are, on Kant’s view, addressed to an imperfect will, which seeks an end but
lacks perfect knowledge of the necessary means to it. Consider such truths from the deliberative
standpoint of an agent who wills an end E, or who is set to bring about E. From this standpoint, E
presents itself as the thing to be brought about and theoretical truths can play a guiding role in
her deliberation about what to do, in the sense that they lead her to make further choices in such
a way as to bring about E. That is, they can inform or correct her decision about what to do when
she is ignorant of, or is mistaken about, the necessary means to E. In this way, theoretical truths
about the means, when addressed to an imperfect agent, can be an imperative.
29

Kant never explicitly puts forward an account of how hypothetical imperatives can play
such a guiding role in practical deliberation. But the following passage from Critique of
Practical Reason strongly suggests that Kant holds a view like this about counsels of prudence:

29
Finlay (2009, 2014: Ch. 3) proposes a similar account of the role ordinary hypothetical imperatives play in our
practical life.

23

If the determination of his will rests on the feeling of agreeableness or disagreeableness
that he expects from some cause, it is all the same to him by what kind of representation
he is affected. The only thing that concerns him, in order to decide upon a choice, is how
intense, how long, how easily acquired, and how often repeated this agreeableness is. (5:
23)

Here, Kant is discussing how someone whose will is determined by his desire for happiness
would deliberate, as he defines happiness as “a rational being’s consciousness of the
agreeableness of life uninterruptedly accompanying his whole existence” (5: 22), or “a maximum
of well-being in [his] present and every future condition” (4:418). Kant’s claim seems to be that,
so long as his will is determined to bring about happiness, his sole concern in practical
deliberation is to find out the means to it. For this person, counsels of prudence, despite being
theoretical propositions about how to attain happiness, can provide normative guidance, leading
him to “decide upon a choice”.
Kant’s view on how counsels of prudence can provide guidance can be generalized to
cover all kinds of hypothetical imperatives. So long as you will an end E in your practical
deliberation, you are determined to bring E about, and your sole concern in deliberation is to find
out the means to E. Then hypothetical imperatives can engage your deliberation, leading you to
reach an informed decision about what to do.
This is the point where the crucial explanatory role of the analytic proposition comes on
the scene: it is possible for theoretical propositions to play the role of an imperative for an
imperfect agent, precisely because Kant’s analytic proposition is true. If it is true, then my

24
willing of an end constitutively involves the willing of its necessary means (de dicto): from my
deliberative standpoint, whatever is necessary to build muscle, as well as the end of building
muscle itself, will at the same time present itself as an object to be brought about through my
action. Since I will be pursuing the necessary means in virtue of pursuing the end, the fact that
taking exercise is a necessary means can lead me to decide to take exercise.
By contrast, suppose that the analytic proposition were false, and the act of willing the
necessary means were a separate act of volition from the act of willing the end. Then it would be
conceptually possible that building muscle presents itself as an object to be brought about, but
the necessary means does not, from my deliberative standpoint. In such a case, it would be
mysterious how facts about the means can engage my deliberation, for the means might not be
represented as the objects to be brought about my action.
In sum, the fact that willing an end conceptually entails willing whatever is the necessary
means to it (de dicto) explains the sense in which theoretical facts about the means can be
imperatives for those who will the end. The resolute reading offers a clear way the analytic
proposition explains the possibility of hypothetical imperatives without problematically turning
the analytic proposition into any rational principle.

6. Worries about the Resolute Reading
One problem for the naïve reading of the analytic proposition (in section 1), is that it denies the
obvious: you can fail to will the necessary means to the ends you will. One way in which this can
happen is through your ignorance of the means. The resolute reading avoids this problem by
explaining the possibility of failing to will the necessary means to your willed end in terms of the
epistemic condition. As we have seen, however, commentators often have in mind failure of

25
some other kind, which they regard as distinctively practical: failing to will the means in the
knowledge of the means. Korsgaard (1997: 229) most explicitly argues that this can happen due
to impeding psychological factors, such as “terror, idleness, shyness, or depression”, which can
make you practically irrational. She argues that any theory of means-end rationality should allow
room for failure of such kind, which she calls “true irrationality”, since denying its possibility is
intuitively unacceptable. (Korsgaard 1986: 12)
Thus, one might worry that the resolute reading also denies the obvious in denying the
possibility of what Korsgaard calls “true irrationality”. This is an important interpretive
objection, because the charge is that the analytic proposition, on the resolute reading, is too
implausible to be accepted by anyone. I meet this objection by arguing (i) that there is no place
for (allegedly) “true” irrationality in Kant’s picture of reasons’ failure to determine the will and
(ii) that its possibility is at best unobvious, given Kant’s concept of willing.

6.1. Did Kant Acknowledge (Allegedly) “True Irrationality”?
Kant identifies two ways in which reason might fail to have decisive influence on actions:

[An] imperative thus says which action possible by me would be good, and represent a
practical rule in relation to a will that does not straightaway do an action just because it is
good, partly because the subject does not always know that it is good, partly because,
even if he knows this, his maxim could still be opposed to the objective principles of
practical reason. (4:414)

26
So there are two kinds of failure: the first is failure of knowledge, which is nothing more than the
failure to meet the epistemic condition, and is already accommodated by the resolute reading.
The second concerns failure in the presence of the knowledge of the good, which is characterized
in more detail in another passage where Kant discusses reason’s failure to determine the will:

If, however, reason all by itself does not sufficiently determine the will, if it is also
subject to subjective conditions (to certain incentives) that are not always in agreement
with the objective ones …, then actions objectively recognized as necessary are
subjectively contingent. (4: 412)

So the question is how we should interpret the second kind of failure, which occurs when one’s
subjective principle of action fails to accord with the objective principles of a practical reason,
even when one knows what is good. If it consisted in what Korsgaard calls “true irrationality”,
we would have reason to rethink the resolute reading.
But it does not. Kant famously distinguishes between two kinds of principles: (i)
objective principles of action that are valid for all rational beings as such and not just
contingently upon their individuals ends or desires, which he calls “laws”; and (ii) subjective
principles that are not so valid, and only contingently upon the agents’ particular ends or desires,
which he calls “maxims” (5:19). As Kant makes it clearer in Critique of Practical Reason,
whether a principle is objective depends on the determining ground of the will that adopts the
principle. A principle is objective only if the will is determined by the pure form of a rule,
without reference to any object of desire, and subjective if otherwise.

27
On this view, no principle grounded in an object of desire can afford us a practical “law”,
even if the desire itself is universally shared. For example, what he calls the “principle of self-
love”, cannot serve as a “practical law”, despite its universal acceptance by all possible (finite)
rational beings who cannot but seek happiness due to their biological nature, since the will is still
determined by an object of desire, that is, happiness. Only when the will determined by pure
reason, by “the mere form of a practical rule” can the principle serve as a practical law, or an
objective principle (5:26).
Now, there is every reason to think that moral duties deriving from the categorical
imperatives are clearly one species of such “objective” principles, on Kant’s scheme. The
categorical imperative, recall, tells you only to adopt a maxim that you can at the same time will
that it become a universal law, without any reference to any particular object of desire. If this is
the only sense in which one can go against the objective principles of practical reason, it is bad
news for the defenders of true irrationality, since immorality is not equivalent to true
irrationality.
30
An amoralist who, in the knowledge of the immorality of her action, flouts the
duties of morality and pursues only her self-interest would be acting contrary to reason on Kant’s
view. But she need not be someone who is “truly” irrational in the sense Korsgaard specifies.
Things would be different if hypothetical imperatives, like moral duties, were objectively
valid. But Kant denies that hypothetical imperatives are valid in this way. It is part of the
definition of a hypothetical imperative that it is valid only contingently for agents who have the
end it presupposes. That is, hypothetical imperatives are only addressed to an agent whose will is

30
It is telling, in this regard, that Kant’s example (in the opening chapter of Critique of Practical Reason) of a
maxim opposed to an objective principle of practical reason concerns immorality, or a maxim that is not
universalizable: “someone can make it his maxim to let no insult pass unavenged and yet at the same time see that
this is no practical law but only his maxim—that, on the contrary, as being one and the same maxim a rule for the
will of every rational being it could not harmonize with itself.” (5:19)

28
already determined by objects of desire. This is precisely why Kant states that “the categorical
imperative alone expresses a practical law, and that the others can indeed one and all be called
principles of the will, but not laws” (4:420). Since a hypothetical imperative is not a law, or an
objective principle, true irrationality cannot consist in failing to will what it prescribes.
Therefore, if true irrationality could count as violating a principle valid for all rational
beings as such, there must be, independently of individual hypothetical imperatives, an objective
principle of practical reason saying that you rationally ought to will what you believe to be the
necessary means to your ends. Such an objective principle, applying to all rational beings, would
amount to the following, wide-scope principle:

Belief-Relative IP-Wide: An agent rationally ought to be such that they will m, if they
will an end e and believe that m is a necessary means to e.

However, Belief-Relative IP-Wide is not derivable from IP-Wide in any obvious way, since the
analytic proposition does not contain any clause about belief. Moreover, it inherits all of the
problems facing IP-Wide: the problem of false categoricity and the problem of no subject (and
no predicate). So it is implausible that Belief-Relative IP-Wide is Kant’s own principle.
The above discussion suggests that what Kant means by “reasons’ failure to fully
determine the will” is either ignorance or immorality.
31
If so, you might fail to will the necessary
means to your end because you do not know what the means is, or you might fail to act in

31
This view ties in nicely with Kant’s other remark on the two roles reason plays in a human being: “no doubt once
[the] arrangement of nature has been made for him he needs reason in order to take into consideration at all times his
well-being and woe; but besides this he has it for a higher purpose: namely, not only to reflect upon what is good or
evil in itself as well—about which only pure reason, not sensibly interested at all, can judge—but also to distinguish
the latter appraisal from the former and to make it the supreme condition of the former” (5: 62).

29
accordance with the categorical imperative, but nowhere in Kant’s scheme would “true
irrationality” fit, and there is no reason to ascribe Kant the view that it is possible.
32

Admittedly, Kant criticizes agents not only on moral but also on prudential grounds,
which might be taken to suggest that imprudence is another, independent way in which reason
can fail to fully determine the will. For example, Kant defines a passion as “[an] inclination that
prevents reason from comparing it with the sum of all inclinations in respect to a certain choice”
and claims that passions lead one to violate the principle telling one not to “please one
inclination by placing all the rest in the shade or in a dark corner, but rather to see to it that it can
exist together with the totality of all inclinations.” (7: 266) Simply put, this principle tells one to
be happy, given Kant’s definition of happiness (from section 5.1).
But on Kant’s view, imprudence is not a failure independent of ignorance. In the context
of establishing the supreme authority of moral law over precepts of happiness, he writes,

A command that everyone should seek to make himself happy would be foolish, for one
never commands of someone what he unavoidably wants already. One would have to
command him only the measure—or better, provide him with them, since he cannot do all
that he wants to do. But to command morality under the name of duty is quite reasonable;
for […] it is not the case that everyone willingly obeys its precepts when it is in conflict
with his inclinations. (5: 37)

32
This reading of Kant, which is consistent with denying the possibility of “true irrationality”, has been suggested
(in passing) by Lavin (2004). Lavin refers to the following sentence as favoring the possibility of true irrationality:
“[imperatives] say that to do or to omit something would be good, but they say it to a will that does not always do
something just because it is represented to it that it would be good to do it” (4: 413), which seems to suggest that,
when applied to hypothetical imperatives, you can fail to will the means even when you know that it is good for the
end you will. But the very next sentence makes it clear that Kant is here referring not to a hypothetical imperative,
but to the categorical imperative, or what is objectively good: “practically good, however, is what determines the
will by means of representations of reason, hence not from subjective causes, but objectively, i.e., from grounds that
are valid for every rational being, as such.”

30

Kant’s point is that precepts of happiness give us only “the measure”, because we are often
ignorant of how to attain happiness by resolving the conflicts between our individual desires.
Once we know how to harmonize our desires in a way that leads to happiness, no further
guidance of reason is required, and we “willingly obey” the precepts of happiness. Going against
a principle of reason in full knowledge of the principle is possible only in the moral case.
33

6.2. How Plausible is the Analytic Proposition?
The only remaining worry for the resolute reading, then, must be the following: “Isn’t it still
obvious that you can exhibit ‘true irrationality’? If so, wouldn’t the resolute reading, in denying
the possibility of ‘true irrationality’, uncharitably make Kant’s view false?”
Offering a sustained argument for the impossibility of Korsgaard’s “true irrationality”
will be a task in Chapter 3, and my points will only be defensive in this chapter.
34
Still, we can
blunt the force of this worry, once we attend to what is constitutively involved in willing an end,
as opposed to merely desiring an end. The will, according to Kant, is the capacity to act in
accordance with a principle of action, a maxim, which is the capacity distinctive to rational
beings. (4: 412) Thus, to will an end is to adopt a principle to act in such a way as to bring it
about. This is plausibly seen as committing oneself to a course of action: a person who adopts a
principle is someone who commits herself to living up to it, in the sense of being ready to resist

33
A full account of rational failure would plausibly make a distinction between ignorance that is criticizable (e.g.
when one neglects available evidence) and ignorance that is blameless (e.g. when the available evidence is
misleading), since one need not be irrational at all in exhibiting the latter. Developing such an account is a task for
another occasion. My main point here is that even the former would be a failure in the theoretical use of reason, far
from being a distinctively practical failure that is alleged to be involved in “true irrationality.”
34
See, however, Finlay (2009).

31
any temptations to deviate from it and being ready to take what she regards as effective steps to
bring about what it prescribes.
35

It is clear that not any old desires involve this element of commitment. My strong,
momentary urge to smoke a cigarette, for example, does not involve my commitment to
smoking. When extremely stressed out with work, I sometimes wish that someone else write a
paper for me, but it falls short of committing myself to the end of getting someone to do so. I can
hope that no one in the world suffers from poverty without being ready to take any effective
steps towards reducing poverty, but it is only when I am so ready that I can be said to have
committed myself to reducing poverty.
So understood, Kant’s analytic proposition is not at all implausible. On any plausible
account of willing, there must be some element, such as being committed, which sets willing
apart from other conative attitudes like urge, wish, hope, etc. But then it is hard to see how one
could recognize something as a necessary means to achieving the end one wills and yet not will
it. For example, you would be able to intelligibly attribute the attitudes I claim myself to have if I
said, ‘I want to build muscle. I know that getting exercise is a necessary means to it, but I just
don’t want to get exercise—I’m too lazy.’ By contrast, it would be difficult to intelligibly
conceptualize my attitude towards building muscle as that of willing if I said, ‘I shall build
muscle. I know I need to take regular exercise in order to do so. But I don’t feel like it, so I
shan’t do it. I didn’t say I would take whatever it takes to do so. But you know—I shall build
muscle!’
36
Kant’s analysis of willing, reconstructed along the lines of the resolute reading, offers
a nice explanation of why: what is distinctive about willing is that if you know the necessary

35
See, e.g., Hill (1973), O’Neill (1989) and Korsgaard (1997).
36
See Worsnip (forthcoming) for a similar point about intention. This is not a coincidence, because intention is
commonly regarded as involving distinctive commitments that distinguish it from ordinary desires. See Bratman
(1987).

32
means to your end, willing the end entails willing the means. Since you do not will the known
necessary means to your end in this case, you do not count as willing the end. This gives us at
least a prima facie case for the resolute reading of the analytic proposition.
Finally, it is important to note that the resolute reading of the analytic proposition is
entirely consistent with another prominent form of practical weakness, akrasia, or acting against
one’s normative judgment about what one ought to do. For example, you can judge that you
ought to build muscle, believe that taking regular exercise is necessary, but fail to will the
exercise. This is because the resolute reading is silent on the possibility of the conflict between
one’s cognitive state (such as belief) and one’s conative state. What the resolute reading does
deny is the claim that your willing of the end (building muscle) is still present in such a case,
which, as we have seen, is at best unobvious.

7. Concluding Remarks
On the resolute reading defended in this chapter, Kant takes it as a conceptual truth about willing
an end that you will what you believe to be the necessary means to the ends you will, rather than
a rational requirement on willing. The genuine norms of rationality governing the imperfect will
are norms of morality through and through. The categorical imperative, the only objective law of
practical reason, gets our ends right, and hypothetical imperatives tell us the right means to
attaining our ends, but there is no further requirement telling us to will the necessary means to
our ends: there is no such normative principle as IP.
Let me conclude with upshots. First, so-called “neo-Kantian” arguments against
Humeanism aren’t Kantian at all. These arguments all begin by deriving a general normative
principle from Kant’s analytic proposition and proceed to argue that it is no less mysterious than

33
the categorical imperative itself, or that IP is incoherent unless backed up by some categorical
norm of reason. The resolute reading has it, however, that the very first step of such a project is
not Kantian. Whatever the merits of such neo-Kantian arguments might be, they do not reflect on
Kant’s actual argument for the possibility of the categorical imperative.
Secondly, and more relevantly for the purpose of the dissertation, Kant leaves no room
for “true (instrumental) irrationality”. He even offers a serious argument for its impossibility
from the analysis of willing and paves a way towards the view that it is constitutive of willing
that one’s willing of the end coheres the willing of the believed necessary means. Moreover, the
problems for the IP-Narrow and IP-Wide considered in this chapter have their analogues in the
contemporary debates on the scope of a requirement Belief-Relative Instrumental Principle. The
problem of immoral means, for example, is often seen to be an instance of the bootstrapping
problem for the narrow-scope reading of the principle: one cannot come under the requirement to
take a necessary means to some arbitrary end simply by intending the end (Broome 1999). The
problem of no subject, for example, corresponds to an important objection to the wide-scope
reading: it implausibly implies that everyone is under the same requirement of instrumental
rationality, regardless of their particular ends, and fails to capture the sense in which it is a
hypothetical requirement (Lord 2014a, Schroeder 2015). Kant’s view offers us a way to avoid
the dilemma: give up the initial assumption that the requirement is a normative law.

34
Chapter 2: Reasoning, Rational Requirements, and Occurrent Attitudes

This chapter argues that rational requirements on our attitudes like belief and intention govern
occurrent, rather than dispositional attitudes. I first identify a tension between two plausible and
popular claims, both of which are well manifested in the recent work of John Broome, but can
also be found in many other philosophers’ writings. One claim is that our attitudes can be
directly modified through processes of reasoning. The other is that the relevant attitudes are
standing states of mind. The tension arises from the fact that our standing states are not in the
direct control of our reasoning processes in the requisite way. I first argue that the tension can be
resolved by distinguishing between occurrent and dispositional attitudes: only our occurrent
attitudes, as opposed to dispositional attitudes, can be directly affected by reasoning. I then argue
that this solution has a significant implication about the proper jurisdiction of rational
requirements on attitudes: they should be understood as governing only occurrent attitudes, given
the way we satisfy requirements of rationality through reasoning.
Section 1 reviews some relevant platitudes about rationality and reasoning. Section 2
explains the plausibility of Broome’s idea that we can directly satisfy rational requirements by
reasoning. Section 3 identifies a tension between this plausible idea and Broome’s own,
dispositional conception of belief. Section 4 generalizes the tension, showing that the problem
goes deeper than any specifics of Broom’s view: the real problem is that belief as a standing
state isn’t the right kind of thing to be directly influenced by reasoning. Section 5 proposes a
solution: we should restrict the kind of beliefs rational requirements govern to occurrent belief.
Section 6 examines some implications of the solution and addresses possible objections to it.

35
1. Setting the Stage
The rational requirements featuring in this dissertation are coherence requirements, requirements
governing combinations or patterns of attitudes like belief and intention. It is standardly assumed
that coherence requirements underlie our ordinary rational assessment, on the grounds that any
breach of the following requirements would, intuitively, constitute irrationality: the requirement
not to believe p and believe not-p at the same time (No Contradiction); the requirement to
believe q if you believe p and believe if p then q (Modus Ponens); the requirement to intend to B
if you intend to A and believe that B-ing is a necessary means to A-ing (Means-End Coherence);
the requirement to intend to A if you believe that you ought to A (Enkrasia).
37

Such requirements are agreed to be local rather than global, in the sense that they specify
specific conflicts between your attitudes that you are rationally required to avoid, as opposed to
what rationality requires of you in light of your entire set of attitudes. The idea is that you would
be irrational if you violated a local requirement of coherence, regardless of your overall
psychology. Thus, the satisfaction of such requirements guarantees local coherence between
specific attitudes, rather than the coherence of one’s attitude-states as a whole.
38

It is also standardly assumed that such local requirements are violable: they are normative
principles we might fail to live up to, rather than descriptive principles about how we in fact
believe and intend. At least apparently, people are often criticized for holding contradictory
beliefs, for failing to believe logical consequences of their beliefs, or for failing to intend what
they believe they ought to do, due to such impeding psychological factors as laziness,

37
These requirements are more precisely formulated in Broome (2013: Ch. 9).
38
For the notion of local requirements of rationality, see Kolodny (2005), Broome (2007b), Brunero (2010), Way
(2011). Nomy Arpaly (2003) offers a case purporting to show that violating Enkrasia, i.e., not intending to do what
one believes one ought to do, coheres better with one’s overall beliefs and desire, and is therefore more rational,
than satisfying Enkrasia by intending what one believes one ought to do. This can be seen as a case in which the
satisfaction of a local requirement decreases global coherence between one’s overall attitudes.

36
depression, fear, etc.
39
Once it is established that rational requirements govern occurrent, rather
than dispositional attitudes, however, this assumption will no longer be obvious, which I will
show at the end of this chapter.
Finally, ordinary talk of rationality recognizes a close relationship between reasoning and
rationality. We are under rational requirements because we are rational, in the sense that we,
unlike fish or stones, have the capacity to comply with them. It seems that the relevant capacity,
when it comes to requirements on belief and intention, is the capacity for reasoning, an activity
through which we modify these attitudes. A person with inconsistent beliefs can come to drop
some of their beliefs by correctly reasoning about the matter, a person who fails to see a logical
consequence of their beliefs can be led to do so by reasoning correctly, and so on.

2. Satisfying Rational Requirements Directly through Reasoning: Broome’s Case
This section explains a plausible thesis about the way in which we satisfy a rational requirement
through reasoning, drawing on the account of reasoning proposed by Broome: reasoning, unlike
other ways of satisfying a rational requirement, is a direct way of doing so.
For Broome, reasoning is a prominent way of satisfying rational requirements on
attitudes. He distinguishes between two distinct ways of improving your rationality, where
improving your rationality is a matter of satisfying the rational requirements you didn’t
previously satisfy. He says, ‘you have both an indirect and a direct means of improving your
rationality: cultivating your rational disposition on the one hand, and reasoning on the other’
(Broome 2013: 207). The former relies on subconscious mechanisms you are equipped with,

39
See Broome (2013) for the possibility of having contradictory beliefs and intentions, which will be discussed in
detail. Kolodny (2005), in passing, dismisses the worry about the impossibility of violating some coherence
requirements. Korsgaard (1997) argues in detail for the violability of Means-End Coherence.

37
which often automatically bring you to satisfy the requirements of rationality. Take the
requirement not to hold contradictory beliefs simultaneously. According to Broome, people often
satisfy this requirement through sub-personal, automatic processes:

Suppose you believe platypuses are not mammals, but then you learn from a radio
programme that actually platypuses are mammals, so now you believe that platypuses are
mammals. Normally, at that point you will automatically lose your belief that platypuses
are not mammals. A great deal of our rationality is given us by automatic processes.
(Broome 2013: 206-7)

You can take a variety of means to facilitate the workings of such sub-personal mechanisms,
including both mental and physical actions.
40
You might, for example, take a nap or drink coffee
to relieve fatigue. You might also try to get your thoughts together or try not to be distracted. In
each case, however, it is still the automatic processes that are directly responsible for your
compliance with the requirements. Your acts only affect your rationality by causing the relevant
mechanisms to operate more effectively, and its effect on your rationality is therefore mediated.
If those sub-personal mechanisms accidentally break down, for example, you might fail to
change your attitudes in the right way, despite having performed such acts.
By contrast, reasoning is a direct means to satisfying the requirements of rationality.
Broome’s focus is on active reasoning, a conscious, person-level process through which you
form an attitude (or what he calls the ‘conclusion-attitude’) on the basis of some of your existing
attitudes (the ‘premise-attitudes’), by following a relevant rule. In active reasoning, you are

40
See, for example, Hedden (2016).

38
conscious of the contents of the premise-attitudes, which you operate on in accordance with
some rule (Broome 2013: 221-5). This rule-governed operation is what makes reasoning active,
or what you do as an agent, rather than what merely happens to you. As Broome notes, there are
automatic, causal processes through which your attitudes might be formed or dropped, but such
processes are passive and are not attributable to you as an agent.
41

To illustrate, consider the following, putative requirement on belief, which Broome calls
‘Modus Ponens’:

Modus Ponens: Rationality requires of N that, if N believes at t that p, and N believes at t
that if p then q, and if N cares at t whether q, then N believes at t that q. (Broome 2013:
157)

Suppose you, at t1, have the following combination of attitudes: (i) believing that it is raining; (ii)
believing that if it is raining then the ground is wet; and (iii) caring about whether the ground is
wet, but you don’t yet believe that the ground is wet. Then you are irrational at t1, for violating
Modus Ponens. One way you can bring yourself to satisfy Modus Ponens at some later time t2
through reasoning is to come to believe, at t2, that the ground is wet. On Broome’s view, you can

41
For conceptions of reasoning as a conscious, person-level activity, see McHugh and Way (2018) and Valaris
(2017). Broome remains agnostic about the existence of automatic reasoning, although he does think that automatic
reasoning would mostly be passive and wouldn’t be an activity that could be attributable to you (Broome, 2013:
232). One might, however, argue that this is too narrow a conception of what counts as active reasoning. Staffel
(2013: 3549-3550), for example, argues that while speaking and driving counts as active processes, much of what
we do when we speak and drive is automatic, and suggests the possibility of reasoning that is both active and
automatic. While I don’t deny the possibility, I’ll nonetheless restrict my points about reasoning to the kind of
reasoning that is both conscious and directed by the agent in the way Broome describes. This is because reasoning
itself is a rationally assessable (rather than arational) activity, and irrationality in reasoning seems to be ‘a failing of
the person, and not of some subpersonal system’, as Kolodny (2005: 552) puts it. The sense in which failure of an
automatic process could be a person-level failure, however, is unclear.

39
do so by operating on the contents of (i) and (ii), which you are conscious of, following the rule
of modus ponens.
42
When you do so, you satisfy Modus Ponens directly.
In sum, Broome’s idea can be formulated as follows:

Directness: For any local coherence requirement R, if reasoning is a way to satisfy R, it is
a direct way to satisfy R.

The precise sense in which reasoning is a direct way of satisfying the requirement of rationality
is given by a crucial point Broome makes regarding such a process: when you reason to an
attitude, your acquisition of the attitude must be a part of the reasoning process itself. It must not
be the result of a separate causal process that follows the reasoning. As he puts it:

We cannot split reasoning into two processes: first forming a conclusion and then coming
to believe it. […] When you reason your way to a new belief, your reasoning is an act.
Since the reasoning is an act, and is the forming of a belief, the forming of the belief this
way is an act. Acts of this sort are sometimes called ‘judgments’. (Broome 2013: 234-5)

Broome’s claim is that to judge p is just is to form a belief in p. His rationale is that a deviant
causal chain could slip in if there were any causal gap between your forming a conclusion and
your belief in that conclusion. Broome claims, however, that there is no such room for a deviant
causal chain, since the act of judgment is identical with the act of belief formation. This depends

42
I shall sidestep the vexed issue of articulating what following a rule consist in. Broome himself understands it as a
matter of manifesting a disposition to act in accordance with a rule while it seems right to you. See Boghossian
(2014) and Valaris (forthcoming) for criticism.

40
on there not being further causal steps between concluding the reasoning process and the
formation of the belief. In sum, reasoning can modify your attitudes directly, in the sense that the
change in your attitudes results, without any further causal processes, from concluding
reasoning.
The basic idea is also endorsed by Pamela Hieronymi, who takes the directness of the
way reasoning affects our attitudes to be a fundamental fact about the distinctive kind of control
we have over our attitudes. She distinguishes between two kinds of control we have over
attitudes: ‘evaluative control’ and ‘managerial control’ (Hieronymi 2006, 2009). In exercising
managerial control over belief, for example, you change beliefs in the same way you change
ordinary objects as you desire them to be: you use a means, such as hypnosis or brain surgery, or
taking a belief-pill, which subsequently causes some change in your mental states. In such cases,
there are causally intermediate steps between the completion of your act and the acquisition of
belief, and so the act might not succeed in bringing about the desired change.
By contrast, you exercise evaluative control over beliefs by settling a question for
yourself.
43
In the case of belief about a proposition P, the relevant question is whether P. Since
the question is whether P, you evaluate considerations bearing on the truth of P. If you see the
considerations as showing P to be true (or false), and you answer the question in the positive (or
negative), then you believe (or disbelieve) P. The crucial point is that settling the question
whether P by reasoning of this kind involves forming or revising one’s belief in P directly: the
completion of reasoning, without any intermediary causal steps, modifies your belief(s).
44

Hieronymi writes, ‘the immediacy of evaluative control is thus not temporal or causal but rather

43
For articulations of the idea of settling questions, see Hieronymi (2005, 2006).
44
For a similar account of doxastic deliberation, see Shah and Velleman (2005).

41
a consequence of the constitutive relation between the commitment to p as true and the belief’
(Hieronymi 2006: 54).
The above discussion has concerned reasoning with belief, but the general point carries
neatly over to practical reasoning, or reasoning with intention. Hieronymi, for example,
explicitly draws a parallel between theoretical and practical reasoning. You exercise evaluative
control over intention, again by settling a question: the question of whether to do A (Hieronymi
2006: 56-7). When you conclude reasoning by evaluating reasons for and against A-ing and
answering the question in the positive, you directly form the intention to perform A. While
Broome does not explicitly discuss how we can form intentions directly as a result of reasoning,
a natural parallel can be drawn: we could say that, just as you directly form a belief by
performing the mental act of judging, you can directly form an intention by performing the
mental act of decision. Consider, for example, an instance of instrumental reasoning: if you
intend to build muscle and you believe that exercising is a means implied by building muscle,
you might, on the basis of these attitudes, conclude reasoning by deciding to take exercise.
45
In
so concluding reasoning, we could say, you directly form the new intention to exercise.
To see the plausibility of Directness, consider again the above case of modus ponens
reasoning. If you come to judge that the ground is wet in the way Broome describes, there is a
clear sense in which you believe that the ground is wet, and thereby satisfy Modus Ponens. That
is, there is a sense of ‘belief’ in which the question of the form, ‘You have judged that p, through
modus ponens reasoning, but have you formed the belief that p?’, is unintelligible. In so forming
the attitude, you seem to satisfy the relevant local requirement of rationality: once you complete
the activity of reasoning in this way, there seems to be no further causal processes that needs to

45
For Broome’s own account of instrumental reasoning, see Broome (2013: 259-60)

42
occur for you to count as having satisfied Modus Ponens. This is in stark contrast to the various
indirect ways you might cause beliefs in yourself. Even if you perform the act of, say, taking a
belief pill with a view to inducing a particular belief in yourself, there is no guarantee that you
thereby have the belief, for the causal process leading from the ingestion of the pill and the
acquisition of the belief might be thwarted. Thus, one might intelligibly ask, ‘You have taken the
pill, but have you formed the belief you wanted to have?’.
Moreover, Directness affords a nice way of fleshing out the intuitive platitude about the
close relationship between reasoning and rationality (from section 1). The capacity for reasoning,
recall, is one of the most basic capacities you have as a rational being. Directness expresses the
distinctive way in which you can improve your (local) rationality through the exercise of this
basic rational capacity: the successful exercise of the capacity, or performing the act of reasoning
correctly, is sufficient for satisfying a relevant requirement on your attitudes. One implication of
this, which will be discussed in further detail (section 5), is that so long as you successfully
exercise your reasoning capacity, other causal processes lying beyond the compass of the
capacity are irrelevant to your rationality.

3. Directness under Threat: Broome’s Dispositional Conception of Belief
Directness offers us a plausible picture of how we change our attitudes and thereby improve our
rationality through reasoning. In what follows, I shall argue that the truth of Directness crucially
depends on the nature of attitudes that are subject to the rational requirements. I do so by
identifying a tension within Broome’s view, which, as we have seen, embraces Directness. In
particular, I argue that Broome’s point about the kind of belief governed by a rational
requirement is inconsistent with Directness.

43
When we say that there are rational requirements on belief, it is important first to get
clear on the nature of belief governed by such requirements, because there are different
conceptions of belief and we get entirely different accounts of what it is to violate or satisfy
rational requirements, depending on the conception of belief we plug into those requirements.
Since it makes a difference to what it is to violate the requirements, it also pertains to the
plausibility of the assumption that rational requirements on belief are violable (Section 1). To see
this, consider the following requirement, which Broome formulates as a consistency requirement
on our beliefs:

No Contradictory Beliefs (‘NCB’): Rationality requires of N that N does not believe at t
that p and also believe at t that not-p. (Broome 2013: 155)

Depending on how we understand ‘belief’ figuring in NCB, it might turn out be impossible to
violate NCB. Consider, for example, a theory of belief according to which you believe at t that p
just in case you, at t, consciously judge (or take it to be true) that p. On this view, beliefs whose
contents are not currently present in your consciousness wouldn’t be subject to NCB. It would be
only when you consciously judge that p and simultaneously judge that not-p that you count as
violating NCB. But it seems at least prima facie difficult to understand how a person competent
with the concept of negation could judge a proposition to be true and judge its negation to be true
at the same time. It is equally unclear why we should conceptualize such a person to be
genuinely judging that p in the first place. Once we adopt this conscious judgment conception of
belief, therefore, we can legitimately call the violability of NCB into question.

44
Broome explicitly addresses such a worry about NCB.
46
He finds it clear that NCB is
violable and puts forward the following as a case in which it is violated:

Whisky. Suppose you used to keep the whisky under the bed, but recently started keeping
it behind the fridge. When you instinctively go for whisky, you head for the bedroom, but
if you first think about it, you go to the kitchen. (Broome 2013: 155)

Again, however, the plausibility of Broome’s claim that you are violating NCB in Whisky
depends crucially on which view of belief we accept. Let’s again consider the conscious
judgment view. Plausibly, you wouldn’t consciously judge any proposition when you
‘instinctively’ head for the bedroom. After recalling where you put it, of course, you would
consciously judge that it is not under the bed. But it is implausible to think that, at the same time,
you would judge that it is under the bed. On the conscious judgment view of belief, therefore, it
is at best unclear how Whisky could be a case in which you violate NCB.
This means that Broome should adopt a different view of belief to make his diagnosis of
Whisky compelling, and to defend the claim that NCB is a violable requirement, which is indeed
what he does. He argues that you violate NCB in this case, because you hold contradictory
beliefs, understood as a cluster of relevant dispositions: ‘A belief is a bundle of dispositions, and
your conflicting dispositions may be enough to determine both that you believe the whisky is
under the bed and that you believe the whisky is not under the bed’ (Broome 2013: 156). On

46
See Broome (2013: 156): ‘Lars Bergström and Kent Hurtig have separately put it to me that some of my putative
requirements of rationality are ones you cannot fail to satisfy. For example, it may be impossible for you to believe a
contradiction. But rationality cannot require you to satisfy a contradiction that you cannot fail to satisfy.’

45
Broome’s view, then, beliefs that are subject to a coherence requirement like NCB are a bundle
of disposition.
Dispositional accounts of belief, which Broome invokes here, have a venerable tradition
in the philosophy of mind.
47
Broadly, to believe p is to be disposed to act (or react) as if p is true.
Such a view makes Broome’s diagnosis of Whisky at least initially plausible. For it is arguable
that in Whisky, you have the behavioral dispositions to, upon acquiring a desire to drink the
whisky, act as if it is under the bed (manifested in your act of heading for the bedroom), as well
as the disposition to act as if it is in the fridge (manifested in your act of going to the kitchen).
The problem is that what Broome’s view of belief implies about the satisfaction (and
violation) conditions of rational requirements spells trouble for his commitment to Directness.
As we have seen, Broome claims that we can satisfy Modus Ponens through forming a belief
through reasoning. Given Directness, Broome is committed to accepting that one can satisfy it
directly by forming a belief through reasoning. However, it is implausible that you can form a
bundle of dispositions to act as if p directly through reasoning by judging p through reasoning.
48

To see this, consider the following case, presented by Christopher Peacocke:

Someone may judge that undergraduate degrees from countries other than her own are of
an equal standard to her own, and excellent reasons may be operative in her assertions to
that effect. All the same, it may be quite clear, in decisions she makes on hiring, or in

47
See, for example, Braithwaite (1932/1933); Ryle (1949); Audi (1994); Schwitzgebel (2002, 2010, 2013); Gertler
(2011).
48
In the remainder of this chapter, I shall focus on reasoning that concludes in the formation of an attitude. This is
partly for the sake of simplicity, and partly not to beg question against those who think that it isn’t possible to
conclude reasoning by giving up one’s attitude(s). Broome (2013: 278-9), for example, finds it unclear what the
conclusion-attitude of reasoning would have to be in such a case, and concludes that we can rely on automatic
processes when it comes to satisfying requirements like NCB.

46
making recommendations, that she does not really have this belief at all. (Peacocke 1998:
90)

We can assume that the reviewer has already reasoned herself towards judgment about the equal
worth of the foreign degrees, on the basis of adequate reasons. Still, she does not seem to count
as believing that the foreign degrees are of equal standard to her own, because of the way she is
disposed to act (or react): she might, for example, feel it unfair when an applicant with a degree
from her own country is treated equally with an applicant with a foreign degree, even when they
have otherwise the same qualifications; she might be unduly fastidious when evaluating
applicants with foreign degrees, etc. That is, the intuitive judgment that she does not really
believe the proposition she judges or asserts is supported by the fact that she is not appropriately
disposed to act as if foreign degrees are of an equal standard to her own.
There are ample examples with the same flavor. Peacocke’s case plausibly is an instance
of ‘belief-discordant behavior’ or ‘dissonant belief’, which is well discussed in the philosophy of
mind. There is a case of a person who used to keep her trashcan from the cabinet beneath the
sink, replaces it with a large garbage bin next to the stove, but keeps walking over to the sink
with her trash in hand (Zimmerman 2007). More dramatic cases grouped together with these
cases include: an agent who judges, with unwavering confidence, that the Skywalk of the Grand
Canyon is safe whilst sweating in fear and panic, refraining from stepping towards the center of
the walkway; an agent who clearly judges that eating soup served in brand new bedpans is
harmless, but avoids eating it (Gendler 2008); an implicit racist who affirms that all races are
intellectually equal, but feels surprise whenever a black student raises a good question and acts
as if black students were intellectually inferior (Schwitzgebel 2010). In each case, there is a

47
discrepancy between what the agent judges and how they are disposed to act (or react), which
inclines us to deny that they believe what they judge.
The intuitive diagnosis of such cases is that the agents never forms a belief (understood
as having a state which disposes them to act in appropriate ways) in the proposition they judge.
As Peacocke himself says regarding his own case, although the reviewer has made the judgment
by correctly evaluating reasons, it has not resulted in ‘a stored belief which has the proper
influence on other judgments and action’ (Peacocke 1998: 90). Quassim Cassam concurs: ‘[in] a
given case, the judgement that P might fail to lead to the belief that P because belief-formation is
also influenced by non-rational factors such as self-deception, prejudice and phobias’ (Cassam
2010: 83).
49

One might resist such a diagnosis, on the grounds that one judges that p if and only if one
forms the belief that p, where believing involves being disposed to act as if it is the case that p.
There are two ways to do so. First, one could deny that the reviewer has the belief and deny, on
that basis, that the reviewer has made the judgment. Second, one could accept that the reviewer
has made the judgment and accept, on that basis, that the reviewer has the belief.
Neither option, however, is plausible. The first implies that the reviewer never has
completed reasoning. This is implausible, since she clearly concludes reasoning by judging the
proposition. It is not as if she is wavering in confidence, holding out for further evidence, or
thinking herself to be lacking relevant premises from which to draw the conclusion. She has, in
Hieronymi’s words, settled for herself the question whether the proposition is true. There seems
nothing further she has to do to complete reasoning.
50

49
See also Shah and Velleman (2005: 503) and Schwitzgebel (2010), for how there can be a gap between judgments
and dispositional beliefs.
50
Buckwalter et al. (2015) argues for a distinction between thick belief and thin belief, which is highly relevant
here. On their view, to have a thin belief that p is simply to have a bare cognitive pro-attitude towards p, which

48
The second fails to do justice to the intuition that the reviewer lacks the relevant belief.
We are stipulating that she lacks some crucial dispositions characteristic of belief with respect to
the proposition: for example, she is not disposed to act or make decisions as if the proposition is
true. So, the objector needs to explain the sense in which she has the belief, as well as explaining
why there is a difference between this particular case and normal cases in which judgments lead
to stored beliefs that properly play their characteristic functional roles.
Thus, it is plausible that there is some causal or temporal gap between making a
conscious judgment and its resulting in the relevant dispositions to act as if its content is true in
ways characteristic of belief. This is not to deny that judging p leads to the dispositions to act as
if p is true in normal cases: such judgments, when all goes well, successfully result in
corresponding dispositions. Nor is this to deny that a judgment might directly initiate a causal
process that leads to the acquisition of a relevant disposition. My point is that initiating a causal
process that would lead to a disposition, however, isn’t identical with having the disposition
itself, since it is possible for one to have initiated the process without having completed it, and it
is only the latter which would warrant the corresponding belief-ascription, if we adopt the
dispositional conception of belief.
But this spells trouble for Directness, because the concept of judgment Broome needs is
the one that doesn’t allow any causal gap between it and forming the belief. Take Modus Ponens,
for example, which is intuitively a requirement we can satisfy through reasoning by consciously
judging a proposition on the basis of one’s antecedent beliefs, as in Broome’s account of modus

consists in taking it true that p, whereas having a thick belief, in addition to having a thin belief, requires relevant
emotion or conation. The following are some ways (they suggest) of ‘thickening’ a thin belief in a proposition P: to
emotionally endorse the truth of P; to actively promote the agenda that makes sense given P. On the basis of this
distinction, we could say that having the belief-constituting dispositions requires forming a think belief, whereas
judgment one reaches through reasoning itself leads one to form only a thin belief.

49
ponens reasoning (section 2). If the judgment completes reasoning, but there is a causal gap
between the judgment and the formation of a corresponding belief (understood as a bundle of
dispositions), you cannot form the belief directly through reasoning. This means that you cannot
satisfy Modus Ponens, for example, directly through reasoning.
In sum, accepting Broome’s dispositional conception of belief forces a choice upon us:
either to conclude that Modus Ponens isn’t a requirement that can be satisfied by reasoning or to
deny Directness. The first option is problematic, simply because Modus Ponens intuitively can
be satisfied by reasoning, and the second option is problematic because if it were true, a person
who comes to judge a conclusion to be true through correct, modus ponens reasoning wouldn’t
satisfy Modus Ponens unless some further, causal condition obtained.

4. The Threat Generalized
My ultimate goal is to establish that Directness conflicts with any requirement on beliefs that
could be satisfied by reasoning, if we interpret belief as a standing state. The preceding section
illustrates such a conflict by combining a rational requirement on belief with a dispositional
conception of belief. I haven’t, however, yet argued that reasoning cannot directly change any
standing state whatsoever. Even if one grants that it is impossible to directly bring about or
eliminate dispositions through reasoning, one could legitimately wonder why all standing states
must be like the kind of behavioral dispositions Broome refers to. If there were a kind of
standing state that could be directly affected by reasoning, and belief were such a state, then
there might not be any genuine conflict. All we need to reconcile Directness and a standing-state
conception of belief would be to replace Broome’s conception with a more suitable one. This
section seeks to generalize the lesson in response to this worry. The conflict identified in the

50
previous section doesn’t depend on any specifics of Broome’s view. There is a principled reason
why one has to deny Directness so long as one upholds a standing-state conception of belief.
There are differing views on the nature of belief as a standing state. Some, as we have
seen, regard belief as a purely dispositional state: to have a belief just is to have a bundle of
relevant dispositions. On such a view, there are no further constraints on the internal structure of
your cognitive system. You count as believing a proposition so long as certain counterfactuals
supported by the relevant dispositions are true of you. On another view, belief is identified rather
with a categorical state of the agent, or what might be called the ‘causal basis’ of such
dispositions, which are causally responsible for the manifestations of the various dispositions
characteristic of belief. As Frankish puts it, whereas dispositional states are states of the whole
cognitive system, defined by its typical inputs and responses, categorical states are sub-states of
the system, defined also by their internal relations to each other and other mental states.
51

Whichever view we take, the crucial point is that a standing belief has the property of
being robust. By calling a state ‘robust’, I mean that whether an agent has a standing state is a
matter of how she would be caused to respond in a wide range of circumstances, rather than what
she is like at a particular moment on a particular occasion. Robustness is what distinguishes the
concept of standing state from the concept of momentary occurrences: a standing state disposes
the agents in various ways, and whether they are disposed in such ways is a matter of how they
would be caused to respond in a wide range of circumstances, rather than what they are like at a
particular moment on a particular occasion. This feature makes a standing state a relatively stable

51
See Frankish (2004: 35). For further, general contrasts between a dispositionalist view and a categorical state
view, see Frankish (2004: 33-38). One of the most prominent theories about such categorical states is
representationalism, a representative sample of which is Jerry Fodor’s view: ‘to believe that such and such is to
have a mental symbol that means that such and such tokened in your head in a certain way; it’s to have such a token
in your belief box.’ (Fodor 1987: 17) Some arguments for the former can be found in Schwitzgebel (2013), and
some arguments for the latter can be found in Armstrong (1973) and Carruthers (2013).

51
feature of mind, which means that a standing state persists in our mind independently of its
particular manifestations, which is why we can truly ascribe it to agents even when it is not
activated or manifested (for example, even when the agents are asleep or non-conscious).
52

Once we accept the robustness of a standing belief, however, there is a plausible
argument that reasoning cannot affect standing beliefs directly, which runs as follows:

P1. A standing state is robust.
P2. A standing state is robust only if it is stored in one’s long-term memory.
P3. If a standing state is an item stored in one’s long-term memory, a result of reasoning
can modify one’s standing state only if it affects one’s long-term memory.
P4. A result of reasoning can modify one’s standing states only if it affects one’s long-
term memory. [from P1, P2, P3]
P5. Reasoning proceeds at the level of working memory.
P6. The process by which items in working memory affect one’s long-term memory is
causal.
P7. If P5 and P6 are true, a result of reasoning affects one’s long-term memory only if it
undergoes a further causal process.
C. A result of reasoning can modify one’s standing belief only if it undergoes a further
causal process. [from P4, P5, P6, P7]

52
This, of course, is not to deny that a standing state can be short-lived. One could, for example, acquire a standing
belief but lose it in just a few seconds, due to some accident that destroys its causal basis. My point is simply that a
standing state, unlike an occurrent event, must have the potential to be activated across multiple occasions.

52
Given the definition of directness, however, the conclusion is equivalent to the claim that
reasoning cannot affect one’s standing beliefs directly.
Let me elaborate on some of the key premises. P2 invokes the notion of long-term
memory, which is often distinguished from working memory and short-term memory. Long-term
memories are what, on various occasions, can be retrieved and put into use in one’s working
memory, which, when it comes to reasoning, involves being activated in one’s performance of
theoretical or practical reasoning. A standard view of working memory is that it is a combination
of short-term memory and other process mechanisms that manage and organize items in the
short-term memory, enabling the effective use of them (Cowan 2008).
There are two reasons why having an item, p, stored in one’s long-term memory (as
opposed to short-term memory) is necessary. The first is the limitation on the capacity of short-
term memory: the amount of information activated in short-term memory is highly limited. The
second is its short duration: the items are kept in short-term memory only within the time
interval in which they are activated, which is said to be between 15-30 seconds (Atkinson and
Schiffrin 1971). Since a standing state is distinct from momentary occurrences in a single
episode of reasoning and are manifested across multiple occasions, p must be incorporated in
long-term memory, which suffers from no such limitations on capacity and duration, and can be
activated (or retrieved) on a range of circumstances.
P5 states that any reasoning is performed at the level of working memory. As Christopher
Cherniak notes, the distinction between short-term and long-term memory maps onto the
distinction between what is currently activated, exerting influence on thought and action, and
what is relatively inactive. Only the items in short-term memory ‘can influence the choice of
actions, and be “logically processed”—used as premises for inferences and or compared for

53
inconsistency’ (Cherniak 1983: 169). Recall, for example, the modus ponens reasoning from
section 2. You first acquire new information that it is raining (now), which initially enters your
short-term storage to be processed. To reach the conclusion that the ground is wet, you must first
retrieve the general proposition that if it is raining then the ground is wet from your long-term
memory. Then you operate on the content of the new information and the memory, and draw the
conclusion. P4 has it that the entire process occurs at the level of working memory.
P6 is a well-established empirical fact about long-term memory. The process in which
working memory is stored in long-term memory is a time-dependent, causal process, called
‘consolidation’. It is widely agreed that hippocampus is responsible for the consolidation
process, which transforms temporary alterations in synaptic transmission into persistent
modifications of synaptic architecture.
53
There is strong evidence that items in working memory
are only selectively transferred into long-term memory, which indicates an intermediate step of
selection between one’s acquiring a piece of information and its being stored in long-term
memory: more significant experiences are more likely to have a place in long-term memory, and
therefore be more well remembered.
54

Let me further defend the argument by addressing some immediate objections. One might
object to this argument by rejecting P5 in the following way: ‘Reasoning is a process that
concludes in an attitude, which is said to be a standing state. If so, however, reasoning is not
complete until a standing state results. Thus, reasoning cannot proceed at the level of working
memory. If so, Directness is not threatened by the standing-state conception of belief.’
55

53
For an overview of the consolidation process, see Paller (2009).
54
For selection involved in the consolidation process, see McGaugh (2000).
55
I thank John Broome for raising this objection.

54
This objection secures Directness by defining reasoning to be a process which directly
results in some change in one’s standing state. But it seems that reasoning can be complete even
if it doesn’t result in such a change. Imagine someone who can recall his past memories, has the
capacity to perform modus ponens reasoning, but lacks, due to some hippocampal damage, the
ability to transfer the items in working memory to long-term memory, and thus immediately
forgets everything he learns.
56
Intuitively, even such a person can perform and complete modus
ponens reasoning so long as he can, at the level of working memory, consciously believe that p,
and that if p then q, and come to judge that q, despite his inability to form a standing belief that q.
Another objection that could be raised against the argument is the following: ‘if the
intervening causal steps between concluding reasoning and acquiring (or losing) a standing state
are, or capable of being, almost immediate, then this difference doesn’t matter. After all, if
something is a judgment, it would seem to be a belief in the relevant sense regardless of whether
an extra step or two is needed for it to be stored in long-term memory.’
There is indeed is a clear sense in which one believes that p if one judges that p, on which
I shall elaborate in the following section: one occurrently believes that p. But the distinction
between occurrent belief and standing belief is significant, so long as the revision of the latter
(but not the former) requires some intervening causal steps lying outside of one’s rational
capacity. One implication of Directness, as we have seen in section 2, is that so long as you
successfully exercise your reasoning capacity in satisfying a relevant requirement, other causal
processes lying beyond the compass of the capacity is irrelevant to your rationality.
To see this, compare the following two agents performing the modus ponens reasoning:

56
Clive Wearing, whose memory lasts between 7-30 seconds because of severe anterograde amnesia, can be a real-
life example.

55
Jack and Jill. Jack and Jill are internal duplicates looking out of the window. Seeing that
it is raining, they both get interested in whether the ground is wet. All facts about their
mind are the same: they have exactly the same background beliefs, perceptual
experiences, memories, etc., and also reason in exactly the same way, on the basis of the
newly acquired belief that it is raining, and the belief that if it is raining then the ground
is wet. Both conclude by judging that the ground is wet. Immediately after Jill concludes
her reasoning, however, an evil neuroscientist secretly paralyzes Jill’s neural activities.
As a result, only Jack manages to store the content in his long-term memory.

Intuitively, there is no relevant difference between Jack and Jill’s rationality, despite Jill’s failure
to form the standing belief that the ground is wet. More specifically, it seems that if Jack satisfies
Modus Ponens by forming the belief that the ground is wet, then Jill also satisfies the same
requirement by forming the same belief (in exactly the same way). There is first the process of
reasoning in which both Jack and Jill engage. Then there is some sub-personal process, distinct
from the process of reasoning, which only Jack undergoes. The judgment about Jack and Jill’s
equal rationality makes it clear that only the outcome of the former process is relevant to whether
we satisfy a requirement. But if the completion of reasoning required the formation of a standing
belief, even the latter process would be relevant, and there would be a difference between Jack’s
and Jill’s rationality vis-à-vis Modus Ponens, which is counterintuitive.

5. The Solution: Rational Requirements Govern Occurrent Beliefs
The dialectic up to this point puts us in a quandary. On one hand, both Directness and the related
idea that we can improve our rationality through reasoning are highly plausible. We fail to

56
capture the intuitive, intimate connection between reasoning and rationality unless we respect
them. On the other hand, we cannot simply do away with the standing-state conception of belief:
belief as a standing state, which disposes us to respond in some characteristic ways, seems to be
the subject matter of both our ordinary belief-attributions.
This section proposes a solution to the problem. I argue that Directness and Broome’s
idea that reasoning is a direct means to satisfying requirements of rationality need not be given
up, because there is a sense in which you directly form or revise beliefs by reasoning, which is
given by the concept of occurrent belief. Things have gone awry because we assumed that
rational requirements should be formulated with the standing-state concept of belief. The first
step of the solution introduces the distinction between dispositional and occurrent beliefs, which
is explicitly employed in philosophy of mind and implicitly employed folk psychology. The
second step is to recognize the belief governed by rational requirements as occurrent belief.
I shall juxtapose two subtly different ways of drawing the distinction between
dispositional and occurrent beliefs, without adjudicating between them, since the argument in
this section doesn’t turn on any particular one. On the first, you occurrently believe p just in case
you consciously judge p, that is, accept p as true. Dispositionally believing p per se is a matter of
having p as an available item in your representational system, although you can be conscious of
p by recalling it on relevant occasions.
57
Alternatively, one might take the basic contrast to be
between what is activated and unactivated (or dormant), which doesn’t coincide with the
conscious/non-conscious distinction: an attitude can be activated in the sense of playing an
explanatory role in one’s thought and action, without being conscious.
58
If an attitude is

57
See Frankish (2004); Price (1969); Rose and Schaffer (2013).
58
See Harman (1986: 14), for this broader conception: ‘a belief is occurrent if it is either currently before one’s
consciousness or in some other ways operative in guiding what one is thinking or doing.’

57
conscious, it counts as occurrent, since it is operative in one’s thought, but not vice versa.
59
On
this distinction, if a dispositional belief is activated or manifested for a time period, influencing
your thought and action in ways characteristic of belief, it is occurrent. So an activated
dispositional belief is occurrent (at the moment it is activated).
In Section 4, I suggested that belief as a standing state can be characterized either as a
categorical state or as a dispositional state. Whichever way the distinction is drawn, both fall
under the heading of a ‘dispositional belief’ in this sense.
60
Finally, we can observe a parallel
between the dispositional vs. occurrent distinction and the distinction between what is in one’s
working memory and what is in one’s long-term memory (section 4). Just as dispositional beliefs
can only affect action and reasoning by being manifested, the items in your long-term memory
can only influence them by being retrieved in your working memory.
We can now resolve the tension between the standing-state conception of belief and
Directness. Return to the way we satisfy Modus Ponens through reasoning. Here, (i) the belief
that it is raining and (ii) the belief that if it is raining then the ground is wet are all activated in
your working memory, so long as you are reasoning with them, are occurrent in that sense. In
general, mental states or events that serve as the inputs to the process are all occurrent, in the
sense that they are operative or activated in one’s thought. Moreover, the outputs of such process
are occurrent. Suppose, again, you conclude reasoning by judging that the ground is wet. Since

59
An attitude’s being conscious can be taken to mean either that the subject is consciously aware of the fact that she
has the attitude, or that she is aware of the content of the attitude. If a person judges p, or engages in reasoning
which employs p, they can be aware of the content (p) without being aware of the fact that they believe p. I adopt the
second reading for two reasons. First, conscious judgments and beliefs activated in conscious reasoning are main
examples of occurrent belief this chapter focuses on. Second, crucial to my use of ‘occurrent’ is that the attitude is
operative, but it doesn’t follow from one’s awareness of an attitude that the attitude itself is operative.
60
Schwitzgebel (2006) suggests that a representationalist can offer the following gloss of dispositional belief: ‘a
subject dispositionally believes P if a representation with the content P is stored in her memory or “belief box”’.

58
judgment is a species of occurrent belief, you count as having formed the occurrent judgment
that the ground is wet. Reasoning therefore directly changes your occurrent belief.
We can also see how we can preserve Broome’s idea that we satisfy rational requirements
directly through reasoning. The trick is to restrict the jurisdiction of rational requirements: while
there is a coherent notion of a dispositional belief which might be a useful subject matter of the
philosophy of mind, rational requirements like Modus Ponens and NCB govern occurrent beliefs,
not dispositional beliefs. Accordingly, Broome’s requirements, Modus Ponens and NCB, would
have to be reformulated in terms of occurrent beliefs. For example:

Occurrent Modus Ponens. Rationality requires of N that, if N occurrently believes at t
that p, and N occurrently believes at t that if p then q, and if N occurrently cares at t
whether q, then N occurrently believes at t that q.
61

No Contradictory Occurrent Beliefs. Rationality requires of N that N does not occurrently
believe at t that p and also occurrently believe at t that not p.

Now that the basic targets of a rational requirement are taken to be occurrent beliefs, there is a
straightforward sense in which reasoning enables you to immediately form or revise your beliefs,
and thereby to satisfy the rational requirements on belief. In the above, modus ponens reasoning,
you satisfy Occurrent Modus Ponens directly. This is because believing that the ground is wet no
longer requires the acquisition of relevant behavioral dispositions: coming to judge, and to

61
While I make do with this version of Modus Ponens requirement for the purpose of this chapter, I don’t take this
version to be ultimately correct, for once we restrict the relevant attitudes to occurrent ones, it becomes unclear what
role the caring condition should be playing.

59
occurrently believe, that the ground is wet suffices for the satisfaction of the requirement. We
can also make sense of the idea that Jack and Jill (section 4) are equally rational, satisfying the
same requirement of rationality by forming the same belief.
The internal conflict within Broome’s view is thus resolved. Rationality plausibly
requires you to comply with Modus Ponens and NCB, but only with respect to your occurrent
beliefs. Reasoning is a process in which you transition from occurrent (premise) attitudes to an
occurrent (conclusion) attitude. This secures the directness of forming the attitude as a result of
reasoning: there is no gap between your forming a judgment and your coming to believe it, since
for you to judge p is for you to occurrently believe p, and for you to suspend judgment about p is
for you to stop occurrently believing p. We need not factor in any further causal processes to
explain how your judgment leads to produce other dispositional beliefs.

6. Objections and Replies
This section responds to an important objection to my solution: if we take rational requirements
to govern occurrent rather than dispositional attitudes, what are we to make of our rational
assessment of people whose attitudes aren’t occurrent, such as those who are asleep or thinking
about something else?
My response is that the violation of a rational requirement entails irrationality, and we
would have a clear case of irrationality only if all of the involved attitudes were occurrent. To see
why this is true of Modus Ponens, suppose Jim dispositionally believes that his kindergarten
teachers are human beings although he hasn’t thought about them ever since he left the
kindergarten. Jim also has various dispositional beliefs about human beings, e.g., that they aren’t
mammals, that they aren’t reptiles, that they have organs, and so on. But it doesn’t seem at all

60
irrational for Jim not to have formed any beliefs about whether his teachers are a reptile, whether
they have organs, and so on. A requirement would be unreasonably demanding if it required of
you that you believe everything that follows by modus ponens from what you believe.
62

Broome’s formulation of Modus Ponens, which adds the condition about caring, is partly
a response to this problem (Broome 2013: 158).
63
But ‘caring’, like ‘belief’, also admits of both
dispositional and occurrent readings: we care about many things that aren’t currently occupying
our attention. If we understand ‘care’ in Modus Ponens as a dispositional state, however, we face
the same problem. Suppose Jane likes Joe, and dispositionally cares about whether Joe will go on
a date with her. Jane believes that Joe won’t if it rains today, that being what he has told her.
Today, however, Jane gets an emergency call from the hospital saying that his father got into a
car accident and runs to the hospital. On her way, she sees that it is raining, but doesn’t form the
belief that Joe won’t go on a date, since she is worried only about her father at the moment.
Again, it doesn’t seem at all irrational for Jane not to form the belief that Joe won’t go on a date,
even if Jane still dispositionally cares about whether he will and has the relevant beliefs.
By contrast, we would have a clear case of irrationality if, for example, Jane were
currently attending to the question whether Joe will go on a date with her, judged both that he
won’t if it rains and that it is raining. But this would be a case in which all of Jane’s relevant
attitudes are occurrent, which Occurrent Modus Ponens rightly forbids. Thus, there doesn’t seem
to be any significant rational criticism we would miss out by restricting the jurisdiction of Modus
Ponens to occurrent attitudes.
We can make the same point about Occurrent No Contradictory Beliefs. We have a vast
range of dispositional beliefs we aren’t aware of, and we might, from time to time, acquire a new

62
This is noted, for example, by Harman (1986), Schroeder (2004), and Broome (2013).
63
Broome (2013: 158).

61
belief that contradicts some of them. It isn’t clear, however, that it is irrational for us to have
contradictory dispositional beliefs in such a way. In Broome’s Whisky, for example, there seems
nothing rationally criticizable about you, so long as you, upon recalling the location of Whisky,
give up the old belief and act consistently, although you can be said to have held contradictory
beliefs for a while. Our cognitive resources are too limited to enable us to notice and prevent
every such conflict, which means that a putative requirement not to have any contradictory
dispositional beliefs at a given time would be unreasonably demanding. By contrast, we would
have a clear case of irrationality if you, for example, consciously judged or took it to be true that
p and also that not-p, which is forbidden by No Contradictory Occurrent Beliefs.
It is also important that my points concern only coherence requirements that are local, as
I have noted from the outset. My view is consistent with the idea that there are rational
requirements of other kinds that govern our dispositional attitudes, in light of which we could
make sense of the rational assessment of attitudes that aren’t occurrent. For example, there might
be other global coherence requirements on attitudes, or requirements to respond correctly to the
reasons you have for a particular attitude (as opposed to a combination of attitudes)
64
, which
target both occurrent and dispositional attitudes. My point, therefore, is that restricting the target
of local coherence requirements to occurrent attitudes doesn’t objectionably limit the range of
criticism we would be able to make.

7. Concluding Remarks
By examining Broome’s case, I have argued that we should conceive of beliefs governed by
rational requirements on belief as occurrent beliefs, if we are to preserve the appealing idea that

64
See, for example, Schroeder (2009) and Lord (2014b) for reasons-responsiveness conception of rationality.

62
we satisfy them directly when we do so. Although I have exclusively focused on the case of
belief and theoretical reasoning, the main thesis of this chapter is neatly generalizable to
intention and practical reasoning. (Section 2) Directness is no less plausible about rational
requirements on intention. Since there is an exactly parallel distinction between occurrent and
dispositional intention,
65
we can expect the same kind of tension to arise between any standing
conception of intention and Directness, to which the same solution can be offered: since we can
directly modify our occurrent intentions as a result of practical reasoning, rational requirements
on intention should be understood as governing occurrent intentions.
I end by revisiting the issue of violability of such requirements, as I did in Chapter 1. As
we have seen, Broome takes it to be obvious that NCB is violable, because he thinks it governs
dispositional beliefs. Now that we have restricted the target of requirements like NCB and
Modus Ponens, however, it becomes hard to imagine how one might violate them. Suppose
someone says, ‘I believe that the whisky is under the bed, but I also believe that it is not.’, or
‘I’m wondering whether the ground is wet. I believe that it is raining and believe that if it is
raining then the ground is wet, but I don’t believe that the ground is wet.’ It is difficult to
intelligibly attribute the attitudes they claim themselves to have, so long as we assume that they
are sincere and they are using the same concept not or the concept if that we use.
One plausible explanation of this difficulty, as we’ve observed in Chapter 1, is that it is
constitutive of belief that beliefs satisfy local requirements of coherence, so long as they are
occurrent: part of what sets belief apart from other cognitive attitudes, such as supposing,
imagining, having credence 0.5 in, is the fact that one cannot intelligibly hold occurrent beliefs
towards contradictory contents. For example, one could intelligibly suppose P and suppose not P

65
See, for example, Mele (2003), which I’ll introduce in the following chapter.

63
for the sake of reasoning. One could simultaneously imagine that one will win a lottery and
imagine that one won’t, thinking about what one will do in each case. Likewise, one could
intelligibly have credence 0.5 towards contradictory contents. The same, I believe, goes for
Modus Ponens. This important asymmetry is well explained by the constitutivist thesis at issue.
The view that one counts as occurrently believing only if one doesn’t violate any local
requirements, together with the thesis that such requirements govern only occurrent beliefs,
entails that they are inviolable requirements. If so, however, they would be better seen as
principles describing what it is involved in (occurently) believing rather than as norms we might
fail to live up to. Of course, much more would need to be done for a full defense of such a view.
My point is simply that recognizing the jurisdiction of the requirements as occurrent attitudes
gives us grounds for taking such a view seriously.

64
Chapter 3: The Real Myth of Coherence

In Chapter 1, I explained Kant’s argument that the principle of instrumental rationality is a
descriptive (rather than a normative) principle about how the will works: it is part of what it is to
will an end that you will what you believe to be a necessary means to it, which means that it is
impossible to fail to will what you believe to be a necessary means to your willed end. In
Chapter 2, I argued that the primary target of coherence requirements should be seen as
occurrent attitudes.
This chapter combines and further develops the two ideas. The goal is to offer a neglected
view
66
about the nature of the widely accepted coherence requirements on intention. I argue that
the idea that there are normative principles of coherence on intention is a myth. In saying this, I
mean to contrast ‘normative’ with ‘descriptive’. Once we articulate what is involved in
intending, and what is involved in the kind of irrationality that would be entailed by the violation
of a coherence requirement, the requirements turn out inviolable. They should rather be seen as
constitutive principles of intention, which describe the ways in which you believe and intend
under certain conditions.
This view is distinct from the view advanced by Niko Kolodny (2007a, 2008a, 2008b).
67

On his view, rationality conceived of as coherence is a “myth”, in the sense that there aren’t
really independent requirements of coherence. On his view, there are only requirements of
reasons, which support having or lacking particular attitudes, rather than combinations of

66
Such a view has been put forward by Stephen Finlay (2009) regarding instrumental rationality. My argument has a
broader target, i.e., coherence requirements on intention in general. Although I don’t argue for it in this chapter, my
argument even extends to requirements of epistemic rationality.
67
Raz (2005) advances a similar view about the requirement of means-end coherence, as Kolodny himself
acknowledges.

65
attitudes. We are only misled into the illusion of coherence requirements because complying
with the requirements of reasons generally guarantees having coherent patterns of attitudes. On
my view, however, coherence requirements have a genuine subject matter: they are the standards
for determining whether you are intending an object. The real myth is the idea that they are
normative requirements you can violate.
I shall proceed by arguing for two theses. The first is a thesis about what is distinctive to
the attitude of intention: occurrent intentions (and beliefs) necessarily satisfy the coherence
requirements on intention unless your mind is compartmentalized. The second concerns the
jurisdiction of the requirements. I argue that you are subject to a coherence requirement on
intention only if you are in those conditions under which your intentions necessarily satisfy the
requirements, that is, the conditions under which your attitudes are all occurrent and your mind
isn’t compartmentalized. The requirements are inviolable because you satisfy them whenever
you are subject to them.

1. Preliminaries: Coherence Requirements and Violability
This section introduces the coherence requirements this chapter focuses on, lays out their basic
characteristics, and explains why they must be violable in order to qualify as a normative
requirement. As explained in Chapter 2, an important strand of recent work on rationality
understands rationality as a matter of coherence between one’s attitudes. The basic idea is that
there are certain combinations of attitudes it would be intuitively irrational for you to have. We
can explain why there would be something rationally amiss if you had any one such combination
of attitudes, by positing a corresponding requirement of rationality forbidding you from having

66
it. With regard to the attitude of intention, for example, the following requirements have been
proposed:

Means-End Coherence: Rationality requires of you that you intend M if you intend an
end E and believe that M-ing is a necessary means to E.
68

Intention Consistency: Rationality requires of you that you do not, at the same time,
intend to A, intend to B, and believe that A-ing and B-ing are not co-possible.

Strong Consistency: Rationality requires of you that you do not, at the same time, intend
to A and believe that you will not A.
69

The above requirements govern combinations of attitudes, without requiring you to have or lack
a particular attitude. Take Means-End Coherence, for instance. There can be a possible scenario
in which there is good reason for you to do each of the following, taken individually: (i) intend
to visit your friend; (ii) believe that getting a cab now is a necessary means to it; (iii) not intend
to get a cab. Still, you would go wrong, or would be subject to the charge of irrationality, if you

68
The nature of Means-End Coherence has been hotly debated, on the assumption that there would be something
rationally amiss with the combination of attitudes it prohibits. The controversy is over what the irrationality of the
combination consists in. Wide-scopers, who think that ‘requires’ takes wide scope over the whole conditional, argue
that the irrationality consists in the failure to satisfy the following disjunction: (not intend E or not believe that M is
a necessary means to E, or intend M). See Broome (2013), Brunero (2012), Wallace (2001), Way (2010), and
Worsnip (2015). Narrow-scopers, who think that ‘requires’ takes narrow scope over the consequent, hold that the
irrationality lies in failing to intend M, despite the intention for E and the belief that M is a necessary means. See, for
example, Kolodny (2005) and Schroeder (2004, 2009). Finally, Setiya (2007) offers yet another option, according to
which it is a principle of theoretical rationality. On this “cognitivist” view, the problem with violating Means-End
Coherence is that the beliefs alleged to be involved in intentions are incoherent. See Harman (1976) and Velleman
(2007) for similar ideas.
69
Bratman (1987) introduces both Strong Consistency and Intention Consistency. See Holton (2009) for discussion.

67
lacked the intention to get a cab, while also intending to visit your friend and believing that
getting a cab now is a necessary means to doing so. By postulating Means-End Coherence as a
requirement of rationality, we can explain why you would be irrational in such a case. The
acceptance of Intention Consistency and Strong Consistency can be justified in the same way.
70

A coherence requirement is said to be strict, in the sense that you would always go wrong
if you violated it. In this respect, coherence contrasts with the concept of having a reason.
71

Unlike being incoherent, failing to do what you have a reason to do doesn’t always invite
normative criticism. You might, for example, have a reason to A, and yet have a stronger reason
not to A, in which case failing to A would not amount to any normative failure. This is not so
with incoherence: necessarily, you go wrong if you are incoherent.
Another distinguishing mark of a coherence requirement has to do with the charge of
irrationality, which expresses a distinctive kind of criticism, the kind of criticism that could be
also expressed by calling someone ‘stupid’, ‘senseless’ or ‘crazy’ (Scanlon 1998, Parfit 2011).
As Jonathan Way (forthcoming) puts it, incoherence seems to be “blatant and severe form of
irrationality”, distinct from the kind of criticism that “we might express by saying ‘he made a
mistake’, ‘he did the wrong thing’, or ‘he lacked sufficient reason for what he did’”, as someone
might fail to do the right thing or what they ought to do without being incoherent, if they are, for
example, ignorant.
In what follows, I argue that the above requirements aren’t normative, by calling their
violability into question. The relevant contrast, to repeat, is between ‘normative’ principles

70
Scanlon (1998, 2007) also distinguishes between structural claims about the relations between one’s attitudes that
must hold if one is to be rational and substantive claims about reasons for one’s attitudes, and suggests that we
restrict the charge of irrationality to the failure to meet the structural requirements. This distinction is implicitly
invoked by Davidson (2004: Ch.12).
71
Broome (1999) was the first to explicitly draw the contrast, calling the requiring relation ‘strict’ and the having a
reason relation ‘slack’. This distinction is accepted by, for example, Setiya (2007), Schroeder (2009), Way
(forthcoming).

68
telling us what ought to be the case and ‘descriptive’ principles telling us what is the case. It is
plausible in this context that if a principle is to qualify as a normative requirement for an agent, it
must be possible for them to fail to do what it says they ought to do.
72

2. Constitutive Principles about Intention
This section argues that there are conditions under which the coherence requirements identified
in the previous section are inviolable, and the satisfaction of them is a necessary condition for
intending. It is one thing to simply gesture at the possibility of violation, and another to describe
a concrete scenario in which someone can be said to be violating one of the requirements. Once
we start to do the latter, it will be clear that there are limits to the range of possibilities in which
we can intelligibly attribute the combinations of attitudes prohibited by the above requirements.
Reconsider the case from section 1. Suppose I tell you that I intend to visit my friend at
3pm and that I believe that the only available way to do so is to intend to get a cab now.
Suppose, however, that I do not get a cab. You ask me, ‘Is anyone getting a cab for you?’ and I
answer, ‘No—I don’t intend to get one. I hate spending money on a cab.’ It seems that you can,
assuming my sincerity, cast doubt on the intelligibility of my attitude-reports: either I do not
really intend to visit my friend, or I do not really believe that I need to intend to get a cab now, or
actually I do intend to get one, in which case I would not violate Means-End Coherence.
73

72
While the sense of ‘possibility’ could be disputed, it is plausible that a principle must at least be conceptually
possible for you to violate it, if it is to qualify as a norm. See Hubin (2001) and Lavin (2004) for discussions.
73
We could of course continue further. You might ask, ‘but didn’t you say that you intend to visit him at 3pm? How
are you going to accomplish that?’, and none of the following possible answers would be intelligible: (i) ‘I don’t
think I will, but I still intend to’; (ii) ‘I don’t know how, but I still intend to’; (iii) ‘I’m going to take a cab, of course,
but I don’t intend to do so.’ As Worsnip (forthcoming) notes regarding a similar case, “the most natural way to hear
[my] speech is as a joke.” See Wallace (2001) and Finlay (2009) for a similar doubt as to whether you can be
intelligibly seen as intending your end in such a case.

69
By contrast, you would not be entitled to deny that I really have the attitudes I claim to
have, if my attitude towards visiting my friend were some other conative attitude. Suppose I tell
you that, instead of intending, I hope, wish, or want to visit my friend. Suppose I have exactly the
same beliefs as above, and again fail to get a cab, saying that I dislike spending money on cabs.
This time, the claim about the combination of attitudes I have seems perfectly intelligible, which
would be clearer if I backed up my claim in the following way: ‘Look, I really want to visit my
friend, but I haven’t actually made my mind up about it. You know, I need to get myself on a cab
now, but I really don’t want to spend that much money, too.’
Here we have striking data calling for explanation. First, there are conditions under which
it is unintelligible to ascribe one a combination of attitudes that violates Means-End Coherence.
Second, the unintelligibility is removed when the intention towards the end is replaced by an
ordinary desire. It is removed, in particular, by emphasizing that my mind is not made up about
the object of the desire. Finally, this asymmetry between the intelligible attributability of
intention and ordinary desires is not specific to Means-End Coherence. Consider Intention
Consistency. Suppose I tell you that I intend to go shopping this evening, that I intend to go
swimming, and also that I believe I cannot do both. Assuming the sincerity of my utterances, you
can legitimately doubt whether I really have the attitudes that I claim to have: I might not really
intend either, or not really believe that they are incompatible. Clearly, this is not so with desires
of other kinds. There would be nothing puzzling about intelligibly attributing the attitudes I
claim myself to have, if I told you that I really want to go shopping this evening, that I really
want to go swimming, but that I believe I cannot do both. The same goes for Strong Consistency:
while it would be difficult to attribute both the intention and the belief I claim myself to have, if I
sincerely asserted, ‘I intend to go shopping this evening, but I believe that I won’t go shopping’,

70
there would be no comparable difficulty if I asserted, ‘although I won’t go shopping this
evening, I want to do so’.
74

I argue that the following thesis gives the best explanation of the data at hand:

Constitutivism: It is partly constitutive of intention that intentions, unlike ordinary
desires, necessarily comply with the coherence requirements on intention if they (and the
relevant beliefs) are occurrent and one’s mind is not compartmentalized.

Consitutivism is a particular version of the idea that compliance with rational requirements is in
some way constitutive of having relevant attitudes.
75
I take the necessity in Constitutivism to be a
conceptual one, since it is simply reflection on the concept of intention, rather than any
substantive inquiry into metaphysical or psychological truths about intention, that disallows
attributing intention to me in the above cases: the most natural response to my attitude-reports
seems to be that I am misusing, or mistaken about, the concept of intention. Mastery of the
concept of intention involves grasping some platitudes about the distinctive roles it plays in our
mental life, which we might call its job description.
Second, it is important, in relation to my argument, to understand the way the thesis
allows for qualifications such as compartmentalization. My goal is not to boldly deny, on the
basis of the above observation, that there are possible scenarios in which one can violate the
requirements in question. Constitutivism states that the conditions under which it is unintelligible
to attribute an incoherent combination of attitudes are the conditions under which: (i) the

74
In fact, Mele (1992) and Brunero (2014) argue that Strong Consistency is simply a constitutive fact about
intention: you cannot intend to A while believing you will not A.
75
See, for example, Davidson (2004), Korsgaard (2009), and Wedgwood (2009).

71
attitudes are occurrent; and (ii) one’s mind isn’t compartmentalized. We can say that an attitude
is occurrent if it is doing its job, that is, if it is currently playing a causal or explanatory role in
one’s thought or action in a way that is characteristic of the attitude. We have considered
Harman’s characterization of occurrent belief in Chapter 2: “a belief is occurrent if it is either
currently before one’s consciousness or in some other way currently operative in guiding what
one is thinking or doing.” (Harman 1986: 14) In fact, Alfred Mele (2009: 4) offers a parallel
characterization of occurrent intention: an intention is occurrent if either “it is suitably at work at
that time in producing relevant intentional actions or in producing items appropriate for the
production of relevant intentional actions”, or “it is a conscious intention at that time”.
Compartmentalization (or fragmentation) of the mind is often characterized
metaphorically as having multiple mental states independently without failing to “putting them
together” or fitting them “together into a single coherent system”.
76
There are two ways of
characterizing compartmentalization. One way is to say that the compartments (or fragments) of
your mind are dispositional states, but those dispositional states never become occurrent or
activated at the same time, and so you fail to form or revise an attitude on the basis of those
attitudes, which you would if they were all occurrent.
77
Someone might, for example, learn P,
and (perhaps after a good amount of time) learn if P then Q, but fail to believe Q because they
fail to consider whether Q, even if they would reason themselves towards believing Q if both
beliefs were activated at the same time.
The other way to characterize compartmentalization is to say that independent occurrent
(activated) attitudes are simultaneously influencing your reasoning or action in different ways.

76
See, for example, Davidson (2004: 198), Stalnaker (1984: 83-4).
77
This seems to be how Worsnip (forthcoming) seems to understand compartmentalization or fragmentation of the
mind.

72
The following example of a split-brain patient, whose left and right hemispheres are
disconnected to some degree, illustrates such a possibility:

When one split-brain patient dressed himself, he sometimes pulled his pants up with one
hand (that side of his brain wanted to get dressed) and down with the other (this side
didn’t). Also, once he grabbed his wife with his left hand and shook her violently, so his
right hand came to her aid and grabbed the aggressive left hand.
78

It seems okay to ascribe him a pair of contradictory occurrent intentions, generated by different
compartments of his mind, which influence his actions in different ways. Since I already have a
separate clause about all of the relevant attitudes’ being occurrent in Constitutivism, I’ll hereafter
refer to this second kind of compartmentalization by ‘compartmentalization’.
So understood, Constitutivism is a descriptive thesis about what it is to intend: our
occurrent intentions, together with relevant occurrent beliefs, would (rather than should) be in
conformity with the requirements under certain conditions. This explains our data: those
conditions under which it is unintelligible to ascribe me intentions are the conditions in which
they are occurrent, and my mind isn’t compartmentalized. In the above cases, it is assumed that
the alleged intentions and beliefs are conscious, and therefore occurrent, and also that my mind
isn’t compartmentalized in any way. This is part of the reason why there is intuitive resistance to
the idea that I really have intentions. Constitutivism predicts that the same resistance will be

78
https://en.wikipedia.org/wiki/Split-brain. It is arguable that the patient doesn’t have contradictory intentions, since
he will eventually either end up getting dressed or end up not getting dressed, and one could argue that he only
intended whatever he ended up with, and not both. So far as I know, however, intuitions about such cases are at best
evenly divided, and I will simply grant that compartmentalization of this kind can be a source of incoherence.

73
found about non-conscious but occurrent intentions, which seems plausible. Suppose I am
driving to campus, which is caused and sustained by an intention to arrive on campus as soon as
possible. Suppose that while driving, I am so preoccupied with the thought of my upcoming
presentation that the intention is not conscious. Still, if the intention were properly at work and I
came to occurrently believe that I have arrived at the intersection at which turning left is the
quickest possible way to campus, I would come to intend to turn left. If I did not intend to turn
left, you would intuitively deny that my intention to arrive on campus as soon as possible is
occurrent, assuming, of course, that my mind wasn’t compartmentalized at the moment.
So much for what Constitutivism is, and how it accounts for the unintelligibility result. I
shall end this section by giving reason to doubt that anything weaker than Constitutivism would
do the job. Any view that allows for the possibility of violating the coherence requirements on
intention even under the conditions in which the relevant intentions (and beliefs) are occurrent
within a unified, non-compartmentalized mind is unlikely to accommodate the data.
First, consider global constitutivism, the kind of constitutivism famously proposed by
Davidson (2004), according to which the standards of rationality impose general constraints on
interpreting an agent as having psychological states in such a way that if the agent deviates from
the standards to a significant degree, they cannot be intelligibly interpreted as having
propositional attitudes at all.
79
This view is consistent with the falsity of Constitutivism, since, on
this view, you only need to comply with the requirements in sufficiently many cases in order to
count as having the attitudes. You don’t need to comply with them on every particular occasion
where the relevant attitudes are all occurrent and your mind is unified.

79
See also Dennett (1989) and Blackburn (1998: Ch.3). Korsgaard (2009) also seems to hold the same position
about instrumental rationality.

74
However, this exactly is the reason why global constitutivism is in a poor position to
explain the data: in the above cases where my self-reported attitudes aren’t intelligible, we are
told nothing about whether I have done well in sufficiently many cases in terms of satisfying the
coherence requirements, and yet we are inclined to refuse to ascribe intention to me. Regardless
of how well I have done in general, I seem to lack intentions in those particular cases, which
global constitutivism cannot explain. Moreover, since it is a thesis about propositional attitudes
in general and not specifically about intention, it fails to account for the observed asymmetry
between intention and ordinary desires.
Second, one might think that it is only because of the rarity or abnormality of violating
such requirements in a situation where one’s attitudes are all occurrent and one’s mind isn’t
compartmentalized (and not because of its impossibility) that my attitudes aren’t intelligible in
the above cases. The idea, on this proposal, is that it is just so unlikely for someone to violate
coherence requirements on intention under such conditions.
But this proposal also fails. There are apparently other kinds of normative failure distinct
from incoherence that would be rare or unusual, which don’t at all incline us to doubt the
intelligibility of attributing intentions. Think about people who intend to engage in some crazy,
imprudent, or pointless activities that you could not imagine any normal person would intend.
We would criticize them precisely on the grounds that they intend what they have enormous
reason not to do, even if we generally expect ordinary agents not to intend such objects:
intentions to do the unthinkable still count as a full-blown intention.
80
Thus, the mere fact
incoherence goes against what is generally expected of an ordinary agent cannot explain why the

80
This is not to deny Anscombe’s point that you might be highly suspicious of my claim if I told you that I want a
saucer of mud, unless I give some further “desirability characterization” of it (Anscombe 1963: 72). My point is
simply that, in principle, any object could be given such a characterization and be desired.

75
distinctive pull to deny the intelligibility of intention-attribution is present in the cases of
incoherence, but is absent in other cases of extremely unusual normative failure.
In this section, I have justified Constitutivism by inferring to the best explanation. In
identifying possible sources of incoherence, however, it might seem that I have only made it
harder for myself to establish my main thesis. That is, one might think that the requirements can
be violated precisely (i) when one’s mind is compartmentalized; or (ii) when one of the attitudes
remain dispositional or unactivated. To complete my argument, therefore, I’ll argue that this
move to rescue the violability of the requirements fails, because when either (i) or (ii) doesn’t
hold, you are not subject to a coherence requirement.

3. Compartmentalization and Irrationality
As we’ve seen in the previous section, it is consistent with Constitutivism that one can fail to
comply with a coherence requirement when one’s mind is compartmentalized, even if all of
one’s attitudes are occurrent. For some, the possibility of compartmentalization has been vital to
explaining the possibility of incoherence. Davidson, for example, offers compartmentalization as
a solution to the difficulty of explaining how incoherence is possible at all:

Synchronic inconsistency requires that all the beliefs, desires, intentions, and principles
of the agent that create the inconsistency are present at once and are in some sense in
operation—are live psychic forces. It is by no means easy to conceive how a single mind
can be described in this way. … [It] is only by postulating a kind of compartmentalization

76
of the mind that we can understand, and begin to explain, irrationality. (Davidson 2004:
197-198)
81

This section argues that compartmentalization cannot secure the violability of the requirements,
because you are subject to a rational requirement only if your mind isn’t compartmentalized.
This is because, first, you are subject to a rational requirement if you have the specific ability to
satisfy the requirement, and, second, you lack the specific ability to satisfy the requirement if
your mind is compartmentalized. Let me first outline the main argument:

P1. You are subject to a rational requirement R only if you have relevant general abilities.
P2. The reason why P1 is true is that R must be such that you could be legitimately
expected to satisfy it.
P3. If P2 is true, then you are subject to R in a situation S only if you have the specific
ability to satisfy R in S (as well as having the relevant capacities).
C1. You are subject to R in S only if you have the specific ability to satisfy R in S. [from
P2 and P3]
P4. If your mind is compartmentalized in S, you lack the specific ability to satisfy R in S.
C2. If your mind is compartmentalized in S, you aren’t subject to R. [from C1 and C4]
P5. You violate R in S only if you are subject to R in S.
C. If your mind is compartmentalized in S, you don’t violate R. [from C2 and P5]

81
Stalnaker (1984: Ch. 5) likewise appeals to compartmentalization/fragmentation to explain the possibility of
contradictory beliefs.

77
Let me elaborate on each premise (besides P5, which is conceptually true). P1 expresses the
uncontroversial idea that only a rational being, or a being with a relevant set of general abilities
or capacities, is subject to rational requirements. This is why, for example, plants and animals
aren’t subject to such requirements. Obviously, you must be capable of having beliefs and
intentions to be subject to the requirements at issue. Similarly, you must be competent with the
relevant concepts. It is plausible, for example, you are subject to the consistency requirements
only if you are competent with the concept of negation, and you are subject to Means-End
Coherence only if you are competent with the concept of necessary means, and so on. Finally, it
seems that you must also be able to perform a basic kind of reasoning, by which I mean a
reasoned process through which you form or revise attitudes, in response to some mental events
or antecedent mental states of yours.
82

P2 is a claim about why is P1 true: the explanation for P1 is that the requirement must be,
as Kieran Setiya (2004) puts it, something you could be legitimately expected to comply with,
and you could be legitimately expected to do so only if you have such general capacities. Setiya,
for example, argues that even when babies or toddlers do apparently stupid things, they are “a-
rational” rather than irrational, because they fall outside the scope of rational assessment, in
lacking the capacity to engage in effective practical reasoning. He also draws the same
conclusion about an agent who has an irresistible, compulsive desire to wash his hand: at least
with respect to handwashing, he isn’t irrational because he couldn’t be legitimately expected to
refrain from it.
83
(Setiya 2004: 274-5)

82
See Fink (2014) for a relevant discussion.
83
See also Finlay (2009) for the claim that requirements of practical reason must meet the condition of reasonable
expectability.

78
This is roughly equivalent to the widely accepted idea rational requirements we are under
should not be too demanding for ordinary agents like us, since a requirement is too demanding
just in case we could not be legitimately expected, on the basis of our limited cognitive
capacities and skills, to comply with it. This idea has been utilized in a variety of ways, such as
in rejecting the (putative) requirement that all of your beliefs be logically consistent, the
requirement to believe every logical consequence of what you believe.
84

But if this is the reason why P1 is true, it seems that if you are to be subject to the
requirement, you must also possess the specific ability to satisfy the requirement, as well as
relevant general capacities to do so,
85
where having the specific ability to A in a situation S
implies both: (i) that you have an opportunity to exercise the relevant general capacities; and (ii)
that if you exercised the capacities in the right way, you would A. To borrow David Lewis’
(1973) semantics for counterfactuals, (ii) is true if and only if you A in all of the nearby worlds in
which you exercise the general capacities in the right way.
P3 is true partly because judgments about whether one could be legitimately expected to
A track our judgments about one’s specific ability to A: if either (i) or (ii) fails to hold in S, you
could not be legitimately expected to satisfy the requirement in S. To see this, suppose first that
(i) fails to obtain: although you have the general capacity to A, you might be temporarily
prevented from exercising the capacity. You might have the general ability to speak French, for
example, but there is a sense in which you cannot, at the moment, speak French if you are

84
See, for example, Harman (1986), Schroeder (2004), Broome (2013), Titelbaum (2015), and Way (forthcoming).
See also Podgorski (2017) for an elegant argument that attempts to reject state-oriented norms (as opposed to
process-oriented norms) of all kinds.
85
For the distinction between general and specific ability, see Maier (2013) and Clarke (2015). I take the notion of
exercising a (general) capacity and the notion of having an opportunity as primitive. See Wedgwood (2013) for a
particular analysis of these notions.

79
currently intoxicated, anesthetized, hypnotized, etc. It seems that you could not, in this situation,
be legitimately expected to speak French.
Suppose now that (ii) fails to obtain. This means that you could be prevented from A-ing
despite your exercise of the general capacity to A: there would be a possibility in which you
exercise your relevant capacity in the right ways and yet fail to satisfy the relevant requirement.
But it seems that you could not be legitimately expected to do A if even the flawless exercise of
your relevant capacity would not be sufficient for A-ing. Therefore, if the fact that you could be
legitimately expected to satisfy a requirement is a necessary condition for your being subject to
it, then so is having the specific ability to satisfy it.
Against P3, one might point out that we still call those who are incapacitated ‘irrational’:
it seems that the very factors preventing us from exercising general rational capacities, such as
intoxication, hypnosis, etc., are precisely what make us irrational. If this were right, being
subject to rational requirements wouldn’t require having specific abilities even if it requires
having relevant general capacities, and P3 would be false.
This objection conflates what we might call the capacity sense and the normative sense of
‘irrational’.
86
The first sense is the one invoked when one calls oneself irrational, saying, for
example, ‘I was out of my mind’, ‘I wasn’t able to think clearly’. It is true that those who are
incapacitated due to the aforementioned factors are not rational and could be called ‘irrational’ in
this sense. This, however, isn’t the sense of ‘irrational’ relevant to rational failure, which
warrants rational criticism. It is irrationality in this second, normative sense, that is relevant to
P3. Moreover, being irrational in the capacity sense doesn’t entail being irrational in the second
sense. Suppose there is a benevolent demon who takes a liking to an incapacitated agent, and

86
See, for example, Ridge (2014) for a similar distinction.

80
causes their attitudes to conform to every coherence requirement. The agent would still count as
irrational in the first sense, but this doesn’t mean that they are irrational in the second sense. The
fact that incapacitation entails irrationality in the first sense, therefore, doesn’t threaten P3.
P4 is plausible, given this understanding of what it is to have a specific ability. Consider
the split-brain patient from the previous section. He lacks the specific ability to satisfy Intention
Consistency, for he fails to meet either (i) or (ii) with respect to satisfying it. One could simply
argue that he, due to the brain damage, is prevented from exercising his general rational capacity:
his brain damage incapacitates the patient, just as intoxication, anesthetization, etc., do.
Alternatively, one could grant that he possesses all of the general capacities that are necessary for
his being subject to the requirement: he is capable of having beliefs and intentions, possesses the
requisite concepts, he is able to engage in reasoning, etc. Still, due to the disconnection between
the two hemispheres of the brain, he lacks the specific ability to comply with the requirement: it
isn’t the case that if he exercised his rational capacities he would succeed in satisfying the
requirement. Either way, the patient lacks the specific ability to satisfy the requirement.
One might object that (ii) is too strong a as a necessary condition for specific ability. For
example, Ralph Wedgwood (2013: 82) agrees that being rationally required (at t) to A implies
having the specific ability (at t) to A, but argues that this doesn’t require its being the case that
you would A if you exercised the general capacity. He argues, for example, even if there is some
nearby world in which you exercise your rudimentary darts-playing capacities but you fail to hit
the bottom half of the darts board, because (say, something causes the board to slip) it is true that
you can intentionally hit the bottom half of the board, provided that it is one of the normal
consequences of your exercising your dart-playing capacities that you hit the bottom half. If
Wedgwood’s proposal is right, then P4 could be false: so long as it is one of the normal

81
consequences of exercising your rational capacities that you do what you are required to do,
compartmentalization doesn’t deprive you of the specific ability to do so.
I have two responses. First, Wedgwood’s proposal is more appropriate as an analysis of
general ability, rather than specific ability. The intuitive idea behind Wedgwood’s argument is
that your ability might not be 100% reliable, and this is surely true of ascriptions of general
abilities: for you to count as having the general capacity to hit the bottom-half of the board, it is
sufficient that your exercise of the capacity guarantees hitting the bottom-half in sufficiently
normal circumstances. However, ascriptions of specific abilities, unlike ascriptions of general
abilities, are sensitive to the particular features of the situation in which you exercise the
capacity: if the particular situation in which you find yourself is abnormal enough to prevent you
from A-ing, despite the exercise of your general capacity to A, you lack the specific ability to A,
even if you would succeed in normal possibilities.
Second, Wedgwood’s analysis of specific ability, if understood as a precondition for
being subject to rational requirements, allows for what we might call rational luck, which seems
implausible. Consider again the case of Jack and Jill from Chapter 2, who possess exactly the
same rational capacities (and the opportunities to exercise them) and exercise those capacities in
the same way in the same situation. For example, the inputs they receive, the ways they process
them, the ways in which they employ relevant concepts and the attitudes they revise or form,
etc., are identical. Given this, it seems that Jack and Jill must be equally rational.
On Wedgwood’s account, however, it could turn out that only Jack is rational and Jill
isn’t. This is so if there is some normal nearby world in which Jill exercises her capacity and
satisfy the requirements, and some abnormal nearby world in which Jack violates the same
requirement although he exercises his capacities exactly in the same way as Jill. This means that

82
two identically situated agents who exercise their rational capacities in exactly the same way,
could differ in their rationality by sheer luck, which is counterintuitive. My account, which
implies that Jack and Jill must be equally rational, avoids this implication.
This completes my defense of the argument: although compartmentalization might be
regarded as an importance source of incoherence, the conditions under which your mind is
compartmentalized are the conditions under which you aren’t subject to a rational requirement.

4. Violability to the Rescue?: Going Dispositional
Constitutivism, recall, states that you necessarily satisfy the coherence requirements on intention
unless your mind is compartmentalized or some of your relevant attitudes aren’t occurrent. The
previous section undermined the attempt to secure the violability of the requirements by an
appeal to compartmentalization. The remaining option is to appeal to non-occurrent, or
dispositional states, which has been a prominent strategy to reconcile the basic idea of
constitutivism with violability. This section surveys some such attempts.
One intuitive notion of attitudes like belief and intention is that they are dispositional
states, which would be activated or become occurrent under their typical stimulus conditions,
but can be truly ascribed to you even when they are unactivated, or you are not engaging in any
physical or mental activities. The key to saving the violability of the requirements consistently
with Constitutivism is to recognize that dispositional states can exist without being activated.
Constitutivism states that there is a conceptual bar to attributing a combination of attitudes that
violates a coherence requirement only under certain conditions, i.e., when the attitudes are
occurrent. So even if believing that p or intending an end E involves being disposed to
occurrently believe that p or to occurrently intend E, and therefore to satisfy the coherence

83
requirements, you might still fail to satisfy some requirement when those conditions fail to
obtain. Since a single failure doesn’t show that you lack the disposition to do so, it is consistent
with attributing the relevant attitudes to you. So it is open for one to argue for the possibility of
an incoherent combination of dispositional attitudes, while conceding that you cannot be so
incoherent when those attitudes are occurrent.
The strategy of going dispositional has been an attractive option for making room for the
violability of rational constraints we have a general tendency to meet in virtue of having the
relevant attitudes.
87
We have already seen Broome’s take on the possibility of contradictory
beliefs in Chapter 2. In fact, Broome also considers the same worry about the violability of the
requirement not to have contradictory intentions simultaneously, and addresses it too by going
dispositional: holding contradictory intentions is possible, so long as we understand each
intention as a dispositional state. In Broome’s own example, you form the intention to be in
London on 23 February, to go to a seminar. You later learn that there will be a tube strike on 23
February and form the intention not to be in London on the same day, without noticing the
coincidence of the dates. Each intention is a dispositional state. For example, you are disposed,
when thinking about the seminar, to (sincerely) say that you will be in London, but you are also
disposed, when thinking about the strike, to say that you will not be there. This, on Broome’s
view, is a clear case of contradictory intentions, which surely seems possible.

87
For example, both Railton (1997: 312-3) and Scanlon (2007: 61) argue that while intending an end E surely
disposes one to give some weight to E in one’s subsequent practical reasoning, this is consistent with one’s giving
no weight to E on a particular occasion, because the disposition can fail to be activated. Scanlon makes the same
point about believing a proposition P and using P in one’s subsequent reasoning. Similarly, Neil Sinhababu (2011)
offers a Humean response to Korsgaard’s argument that Humean account of motivation makes instrumental
irrationality impossible by invoking the notion of “dispositional desire strength”. What Korsgaard calls “true
irrationality” can occur, on his view, when one’s dispositional desire to A is stronger than one’s dispositional desire
to B, but one chooses to B because vivid representations of B, or the nearness of B in space or time, increases the
violence of the desire for B, and motivates one to pursue B rather than A.

84
As before, it would be difficult to make sense of someone who sincerely asserts that she
intends to be in London and also that she intends not to be in London at the same time as
genuinely intending to be in London. Indeed, it is this kind of difficulty that motivates the worry
about the possibility of having contradictory intentions in the first place. But it is open for
Broome to argue, to capture the data about the unintelligibility of having such a combination of
attitudes, that one cannot have contradictory occurrent intentions. (Broome 2013: 156)
This approach has been most directly applied to one of the coherence requirements at
issue, Means-End Coherence, in Alex Worsnip’s treatment of coherence requirements in general.
On the one hand, Worsnip explicitly advocates a thesis similar to Constitutivism, arguing that it
is constitutive of the attitudes governed by a coherence requirement that you would give up or
revise one of the attitudes that jointly violate the requirement, were you under the “conditions of
full transparency”, which he defines as: “the conditions under which the agent knows, explicitly
and consciously, that she has the state in question, without, say, self-deception, mental
fragmentation, or any failure of self-knowledge.” With regard to Means-End Coherence, this
view implies that it is constitutive of intention that if you were correctly aware that you intend an
end E, that you believe that M is a necessary and available means to E, and that you do not
intend M, you would revise your in attitudes in one of the following ways, bringing your
attitudes in line with Means-End Coherence: (i) stop intending E; (ii) giving up the means-end
belief; or (iii) forming the intention to M. So his view implies that fully conscious violations of
Means-End Coherence (and other coherence requirements) are impossible. This nicely captures
the data supporting Constitutivism.
On the other hand, he still holds that Means-End Coherence is violable. It is possible that
even when you have an intention for an end and a relevant means-end belief,

85

[You] never put these two mental states together or reflect on what they jointly commit
you to. By putting at least one of the two states out of your occurrent reach … you never
come to intend [the believed means]. … You just avoided simultaneous and conscious
consideration of the fact that you had both of those states.

So violating Means-End Coherence crucially depends on at least one of your mental states’
remaining dispositional (or being “out of your occurrent reach”).
We now have a family of views that not only allow for the violability of coherence
requirements, but also capture the conceptual link between intention and the satisfaction of the
coherence requirements. Although it is conceptually difficult to imagine a person violating a
coherence requirement with occurrent attitudes, it seems easy to describe a scenario in which
they have incoherent dispositional attitudes. So one might think that dispositional attitudes can
make room for the violability of the requirements, and occurrent attitudes can explain the data
supporting the conceptual link between intention and the satisfaction of coherence requirements.

5. You Can’t Go Dispositional
This section argues that going dispositional confronts a deep problem of its own.
88
The crux of
the problem is that coherent requirements on intention govern occurrent attitudes rather than
dispositional attitudes. If having combinations of beliefs and intentions that violate or satisfy
them is to count as irrationality or rationality, the relevant attitudes must all be occurrent. I’ll

88
Finlay (2009), who denies the violability of Means-End Coherence, anticipates the dispositionalist response
considered in the previous section, but rather quickly dismisses it. This section is devoted to showing in detail why
this response cannot be made to work.

86
restrict my focus to Means-End Coherence for convenience, assuming plausibly that the point is
perfectly generalizable to other requirements.
Recall the points from Section 1: incoherence is a “blatant and severe form of
irrationality”, and is distinguishable from other normative failures in this respect. With this in
mind, let’s reconsider the cases of means-end incoherence involving dispositional attitudes and
ask if such cases involve any irrationality in this sense. Assuming that you fail to intend what
you believe to be a necessary means, there are four possibilities, depending on whether your
intention for the end and your means-end belief are occurrent or dispositional:

Dispositional Intention Occurrent Intention
Dispositional Belief DBDI DBOI
Occurrent Belief OBDI OBOI

I shall proceed by first considering OBOI, where your irrationality would be manifest, and
comparing it with other cases, to establish that you are irrational only in OBOI.
OBOI, where your means-end belief and your intention for the end are both supposed to
be occurrent, is the kind of case in which your attitudes would exhibit a blatant and severe form
of irrationality. Suppose it’s about lunchtime. You occurrently intend to cook an omelet, judge
that you need to break eggs now in order to do so. But you don’t intend to break eggs: you could
be simply lazy, or it could be that you became afraid of burning the pan, etc. The discussion from
Section 2 already casts serious doubt on the intelligibility of the claim that you still intend to
cook an omelet for brunch in this case. But let’s assume, for the sake of argument, that it is a
possibility. Then we could agree that you would be clearly irrational in not intending to break

87
eggs, assuming that you still intend to cook it: you clearly have your end in view, and are fully
aware of the necessary and available means you can intend, but you do not intend it.
With this in mind, consider other cases, starting with DBDI. Suppose you dispositionally
intend to cook an omelet for brunch today, and also dispositionally believe that breaking eggs is
the only available means. You made up your mind to cook an omelet the night before, and it
successfully resulted in the dispositional means-end belief and the dispositional omelet-cooking
intention since then: you have characteristic dispositions associated with each attitude. For
example, you would avow your end of cooking an omelet, if asked. If asked about the recipe,
you would confidently assert that you would need to break eggs, and so on. But you are tied up
with an urgent work all morning. You are too busy to give a thought about getting something to
eat, and so skip your meal. As a result, you fail to occurrently intend to break eggs. You only
realize, way after lunchtime, that you missed cooking an omelet. None of the attitudes become
occurrent, and you never manage to put them together.
We can give only a slight twist to this case in order to generate a case that falls under
OBDI, where only your means-end belief is occurrent. It’s about lunchtime, you know it, and
you are again all into the work. But this time, your roommate comes and asks, ‘Hey, you need to
break eggs in order to cook an omelet, right?’ Irritated, you yell at him: ‘Of course! Didn’t I tell
you that I’m busy?’ and continue with your work immediately, again without thinking about
brunch at all. As before, your intention to cook an omelet for brunch does not become activated.
The only difference is that your belief is briefly activated when you respond to your roommate.
Still, it doesn’t seem you are irrational in this case.
Finally, consider case DBOI, where only your intention is occurrent and the means-end
belief remains dispositional: such a case could come about because you fail to recall what the

88
necessary means is, although you dispositionally believe what it is. Suppose you know that your
child couldn’t be happier if he could get, for his birthday gift, the (special-edition) sneakers he
always has had his eye on. So you come to believe that getting the sneakers is the only available
way you can please him the most on his birthday. Months later, you notice that his birthday is
tomorrow and form the intention to get him the gift that he would be most fond of. But you fail
to recall what it is. Then the thought of disappointing your child suddenly looms large and, as a
result, you go blank. Since you are in a rush, you eventually get something else instead, failing to
intend to get the sneakers. Only the next day does your nervousness fade away and it occurs to
you that it was the sneakers.
In this case, you seem to have a stored belief that the necessary means to your end is to
get the sneakers. It is commonplace to fail to retrieve a piece of information at the moment we
desperately need it, only to unexpectedly remember it at a later time, even when the information
remains in your mind, in the form of stored knowledge. If so, it is plausible that you have the
means-end belief, at least dispositionally. This is so even if, on this particular occasion, this
dispositional belief fails to be manifested, and therefore fails to affect your reasoning or action at
the right time. Although it is generally true that if you dispositionally believe p, you would recall
p were you to wonder whether p, this is not always the case. For example, you might still fail to
recall p, if you are, e.g., drunk, exhausted, or panicky.
Now for the comparison: in none of the above cases do you seem to be irrational in the
way you are in OBOI. Even if there were some failure involved in the other cases, say, it seems
to lack the property that makes incoherence a blatant and severe form of irrationality. This
already gives us a reason to deny that DBDI, OBDI, and DBOI are a genuine instance of
incoherence. This might not convince a reader who rejects this particular conception of

89
incoherence. But even if one takes issue with this diagnosis, one could at least agree with the
weaker claim that there is a further problem or failure in OBOI, which is absent in other cases.
But anyone who accepts this, but also holds that in each of the above cases you violate Means-
End Coherence should explain why there is a further problem: if you are violating the same
requirement in each case, why does it seem that there is a further problem in OBOI, which
makes it clearly proper to call you ‘incoherent’ or ‘irrational’?
The most obvious explanation, which I propose, rejects the idea that you violate Means-
End Coherence in each case: Means-End Coherence governs only occurrent intentions and
beliefs. In cases other than OBOI, either your belief or your intention is unactivated, and you fail
to put them together as a result. If Means-End Coherence governs only occurrent attitudes,
unactivated dispositional attitudes do not fall under its jurisdiction in the first place, so you do
not violate it. It is only in OBOI that the relevant intention and belief are both occurrent but you
fail to intend the means, which explains why you are irrational: you are means-end incoherent
only in OBOI, and that is the only blatant form of irrationality we can identify.
How might one attempt to explain the intuition that there is a further problem in OBOI,
as well as maintaining that you are incoherent in all of the above cases? Three possible tactics
suggest themselves, which I shall reject in turn. First, one might argue that although you violate
Means-End Coherence in each case, there is a further requirement of rationality you violate in
OBOI, which makes you more irrational. The problem with this move is that it is difficult to
identify any rational requirement you would violate only in OBOI, other than Means-End
Coherence. It cannot be that there is an independent rational requirement forbidding you from
having the means-end belief, or from having the omelet-intention, since it would equally forbid
you from having them in other cases. The only relevant difference is that you did manage to

90
recall the end-intention and the means-end belief, and therefore have all of them activated in
OBOI, whereas you fail to do so in other cases. But it is hardly plausible that having your
dispositional attitudes activated amounts to rational failure of any sort. If anything, you are
negligent or inattentive in forgetting your intention and/or belief in other cases, but not in OBOI.
So one cannot explain why you seem more irrational in OBOI by positing a further requirement
of rationality that you violate only in OBOI.
The second strategy would be to argue that there are in fact two distinct requirements of
means-end coherence, one governing dispositional attitudes and the other governing occurrent
attitudes, and that it is more irrational to violate the latter. If this proposal were right and
rationality had different principles for dispositional and occurrent attitudes, there would be a
version of Means-End Coherence whose satisfaction and violation depends on dispositional
attitudes, and dispositional attitudes only.
But this has implausible consequences. Suppose you form, on the basis of your intention
to cook an omelet and your belief that you need to break eggs, the occurrent intention to break
eggs. Then it seems that you have satisfied Means-End Coherence, regardless of whether you
also form a dispositional intention to break eggs. To see this, suppose you immediately come to
hate the slippery feeling of a raw egg, as soon as you break one, and decide never to break an egg
yourself again. Then it would seem that you do not, at any point, form the dispositional intention
to break eggs. Still, it does not alter the fact that you satisfied Means-End Coherence.
But if there were an independent version of Means-End Coherence that only governs
dispositional attitudes, you would still be incoherent in not forming a dispositional intention to
break egg, because you would still have the dispositional intention to cook an omelet (at least
until the cooking is over), and you would still retain the dispositional means-end belief. This

91
seems an extremely bizarre criticism to make, as you have succeeded in doing what is necessary
for the execution of the omelet-intention. So there is no reason to posit a version of Means-End
Coherence, which only governs dispositional attitudes.
Finally, one might explain the difference between OBOI and the other cases in the
following way: although you violate the same requirement in each case, and is irrational, your
irrationality is mitigated in the cases in which some of your attitudes is dispositional. The idea
would be that the fact that at least one of your attitudes remains dispositional, in making it more
difficult for you to satisfy Means-End Coherence, counts as a mitigating factor when assessing
your irrationality. By contrast, when all of the attitudes are occurrent or activated, it is much
easier to satisfy the requirement, and you cannot plead mitigation.
This argument is also problematic, however. I grant that in DBOI, it is relatively difficult
to satisfy the requirement because the means-end belief is not, at that moment, accessible to you:
you try hard to recall what your child wants the most, but fail nonetheless. So it makes sense to
think that your irrationality is mitigated, to a significant degree, in DBOI. But it is hard to
explain DBDI and OBDI in the same way, because the attitudes are accessible: if you had
simply given a thought about getting a meal, your omelet-intention and means-end belief would
have been activated. So either there is no difficulty at all, or there is a difficulty that is not
significant enough to explain the difference between the two cases and OBOI. So the
explanation in terms of mitigated irrationality only has limited applicability.
One might reply that the relevant difficulty in DBDI and OBDI has to do, not with the
accessibility of the attitudes, but with the fact you are busy: the fact that you are busy with some
other task makes it difficult for you to attend to the thought of cooking an omelet. But this
response only touches upon an accidental feature of the case. Suppose instead that you fail to

92
recall the omelet-intention not because you are busy, but simply because you spend all morning
staring off into space, forgetting all about the meal. There is nothing that makes it difficult for
you to activate the omelet-intention, but it still doesn’t seem that you exhibit the kind of
irrationality comparable to the kind of irrationality that you would exhibit in OBOI.
I have considered some apparently promising attempts to explain why there seems to be a
further problem in OBOI while at the same time maintaining that you violate Means-End
Coherence in all four cases: (a) by positing a further requirement of rationality you might be
violating in OBOI; (b) by arguing that you violate a different requirement in OBOI; or (c) by
treating the fact that one of your attitudes remains dispositional as a mitigating factor. Each
attempt, however, fares worse than my proposal. The best way to accommodate the data is to
posit that only occurrent attitudes are subject to Means-End Coherence.

6. Conclusion: Doing without Coherence “Requirements”
Let me sum up the argument up to this point: it is constitutive of intending that you satisfy the
coherence requirements under certain conditions. If you are under those conditions, you cannot
violate the requirements. If you are not under such conditions, you aren’t subject to them and
therefore cannot violate them. Thus, they are inviolable. But given that violability is a criterion
that divides normative principles and descriptive principle (section 1), the “requirements” should
not be seen as norms, but rather as principles describing how our attitudes would be under
certain conditions, so far as we are intending.
This has an important implication about the recent debate on whether there is reason to
avoid having incoherent combinations of attitudes. Both Kolodny (2005, 2008a, 2008b), who
seeks to explain away the appearance of the requirements of formal coherence as such, and

93
Broome (2005, 2007, 2013), a prominent defender of coherence requirements, raise serious
skepticism about the normativity of rationality understood as coherence. Both assume that if
rationality itself is to be normative at all, it must always give you at least a reason to satisfy its
requirements that apply to you, but find it unclear what such a reason could be.
It seems that the question of whether there is reason to be coherent presupposes that the
coherence requirements are norms: As Niko Kolodny puts it, prominent accounts of practical
rationality regard coherence as “the only sure thing in the domain of normativity.” (Kolodny
2008b: 366) For example, many contemporary Humeans regard instrumental rationality, the
central idea of which can be seen as being expressed by Means-End Coherence, as the sole
principle of practical reason. Even rationalists who recognize requirements of other kinds, such
as requirements of morality and prudence, attempt to vindicate their normativity by reducing
them into practical consistency of some kind.
89
If the view advanced in this chapter is right,
however, the question rests on a false presupposition: coherence requirements on intention,
unlike the demands of morality and prudence, aren’t norms. Asking whether there is reason for
us to follow them is akin to asking whether there is reason for us to follow the laws of nature.

89
Nagel (1970), Darwall (1983), Korsgaard (1997, 2009) are the locus classicus of this view.

94
Chapter 4: Settledness and the Rationality of Occurrent Attitudes

In Chapter 3, I advanced an argument that it is impossible for one to violate the coherence
requirements on intention because one, as a matter of conceptual necessity, satisfies the
requirements whenever one is subject to them. In doing so, I have largely relied on an important
asymmetry between dispositional and occurrent attitudes: ascribing one the combinations of
occurrent attitudes forbidden by such requirements (assuming the non-compartmentalization
condition) is unintelligible, while it is easy to imagine a scenario in which one has the same
combinations of dispositional attitudes. We have seen that this asymmetry is supported by
intuitive judgments about some problematic first-person assertions about one’s attitudes, and is
tacitly assumed by many philosophers who choose to “go dispositional” to make room for the
violability of the requirements at issue.
The previous chapter was about requirements of practical rationality, that govern both
intentions and beliefs. The aim of this chapter is to offer a more unifying picture of coherence
requirements by extending this approach to similar requirements of epistemic or theoretical
rationality: these requirements too, at bottom, are descriptive principles which occurrent beliefs
(but dispositional belief do not) necessarily meet. The key to understanding this essential
rationality of occurrent beliefs, I argue, is the settledness involved in occurrent beliefs. So I begin
by elucidating the idea that belief involves being being settled on (or one’s mind being made up
about) relevant questions.
In doing so, I draw upon Jane Friedman’s (forthcoming) recent account of belief and
“interrogative attitudes”, according to which belief involves settledness in the sense that
believing a complete answer to a question Q fails to cohere with one’s ongoing inquiry into Q.

95
While Friedman claims that the coherence relation at issue is a normative relation between belief
and inquiry, I argue that it is a descriptive relation between occurrent belief and inquiry (into a
question). For occurrent beliefs, the incoherence means impossibility: occurrently believing p is
incompatible with ongoing inquiry into questions to which p is a complete answer. If we instead
focus on dispositional beliefs, it is possible for one to have them while inquiring into relevant
questions, but there is no plausible norm that necessarily prohibits being in such a state. Either
way, the coherence relation cannot plausibly be a normative one.
Once it is established that settledness is a descriptive relation between occurrent beliefs
and ongoing inquiry into a question, it becomes plausible that (alleged) coherence requirements
on belief are also descriptive principles about occurrent beliefs. For example, if (occurrently)
believing p involves being settled on the question whether p in a particular way, it is difficult to
see how one could, at the same time, simultaneously believe not-p, believe if p then q and yet not
believe q, or how one could believe that one lacks sufficient evidence for p.
To be clear on the ambitions of this chapter from the outset, the goal is not to present a
deductive proof of the impossibility of incoherent combinations of occurrent beliefs from the
point about the settledness of occurrent belief. The goal is rather to show how the view can be
made plausible by reflecting on what it is for an agent who believes p to be settled on the
question whether p, and how this, together with the arguments from Chapter 3, can give us a
unifying view of the nature of the coherence “requirements” on belief and intention.

1. Belief as a Settling Attitude: Friedman’s Account
The idea that belief involves some sort of settledness is not new. For example, Pamela
Hieronymi (2005, 2006, 2009) has argued that for some attitude(s), settling a relevant question

96
just is to form the attitude, and belief is one such attitude in the sense that settling for oneself
positively the question of whether p just is to believe p.
90
Moreover, a similar point has been
made in the context of contrasting believing p with having a high credence in p (that falls short
of certainty). The intuitive idea is that if you believe p, you take the truth of p for granted, in the
sense that the question of whether p is not a matter for consideration for you. Jeremy Fantl and
Matthew McGrath nicely put this point:

If you outright believe that p, your mind is made up that p. […] Consider a person who is
searching hard for evidence, given just how much is at stake for her, and who stays up
late at night worrying. Is this person’s mind made up? Not in any natural sense. But,
consistent with all this, the person might have a Lockean credence. (Fantl and McGrath
2009: 142)
91

As yet, the idea that belief involves settledness remains a vague slogan. At the concluding
section of Chapter 2, I have noted a way in which belief crucially differs from other cognitive
attitudes, but I haven’t connected it up with the feature of settledness. In any case, little has been
said on what exactly it is to be settled on a question, how beliefs are different from other
attitudes which lack settledness, and how beliefs interact with other attitudes or activities in
virtue of its settledness.

90
Hieronymi (2005; 2006; 2008) draws a parallel between belief and intention and argues that the fact we can form
beliefs and intentions by settling relevant questions can shed light on “the wrong kind of reasons” problem for belief
and intention, as well as on the nature of our responsibility for these attitudes.
91
For closely related remarks, see Nagel (2008; 2010), Clarke (2013), Wedgwood (2014), Ross and Schroeder
(2014), and Friedman (forthcoming). Of course, this might not be a fully accurate way to put the point, since a
person can search hard for evidence as to whether p, simply in order to convince someone who doesn’t believe p,
which is consistent with their believing p.

97
One major goal of this chapter is to elucidate the idea that belief involves being settled on
a question. Instead of speculating on every possible suggestion on the matter and considering the
(de)merits of each, I’ll proceed by considering Friedman’s account of the settledness of belief,
which seems to provide an initially promising way of fleshing out the concept of settledness.
This section explains her view that the settledness of belief consists in the normative relation
belief bears with what she calls “interrogative attitudes” or “inquiry” (which I will argue against
in what follows).
On Friedman’s view, inquiring at a given time (or over a period of time) is a matter of
being in a particular state of mind, or having what she calls “interrogative attitudes”, which
include inquiry-related states or processes, such as curiosity, wondering, suspension of judgment,
investigation, contemplation, and deliberation (Friedman 2013, 2017, forthcoming).
92

On her view, there are two key features that justify grouping such attitudes together.
First, an interrogative attitude has a question as its content, in the sense that the verbs and
phrases picking out interrogative attitudes, such as ‘is wondering’, ‘is curious about’, ‘is
investigating into’, ‘suspends judgment about’, ‘deliberating about’, etc., embed interrogative
complements, such as ‘whether he took the cookie’, ‘who took the cookie’, ‘when he took the
cookie’, etc.
93
Second, an interrogative attitude is a goal-directed attitude, in the sense that it
aims to resolve the question that it has as its content. Someone with an interrogative attitude has

92
Many items on this list do not seem to count as attitudes in any sense. For example, investigating, contemplating,
deliberating, all seem to be occurrent processes or activities. Friedman is aware of the oddity of calling these
‘attitudes’ and admits that she is using the term ‘attitude’ loosely. She notes, however, that they are activities that “a
subject is engaged only in virtue of being in a particular sort of goal-directed state of mind” (Friedman 2013: 169),
and so they all involve some sort of attitude in this sense.
93
See Friedman (2013) for some linguistic considerations supporting this point. This linguistic condition by itself
does not give us a criterion for deciding whether an attitude is interrogative, however. Even ‘know’, for example,
can embed interrogative complements, as in ‘I know whether he took the cookie from the cookie jar’, but knowledge
does not seem to count as an interrogative attitude. Moreover, as Friedman herself notes (2013: 148), there are some
nuanced differences in the embedding patterns of the phrases that pick out interrogative attitudes.

98
a distinctively epistemic aim of answering their question, which involves seeking or being at
least minimally sensitive to information that bears on the question, and in this sense can be
characterized as “wanting to know” (Friedman 2017: 309; forthcoming). This gives a gloss of
what it is to have a question open in thought: you have a question open in thought when you
have an inquiring attitude that has the question as its content.
In what sense does a belief settle a question, then? In answering this question, Friedman
argues that a distinctive relation holds between beliefs and interrogative attitudes: necessarily,
inquiring into or having an interrogative attitude towards a question Q at t fails to cohere with
believing p
Q
at t, where ‘p
Q
’ represents the proposition p as a complete answer to Q (Friedman,
forthcoming: 8).
94
She illustrates this point by considering a version of British detective
Inspector Morse:

Morse: Morse wakes up one morning with thoughts of his killing a doctor the night
before, sees that his flat is covered with blood and becomes convinced that he murdered
the doctor.
95
Out of a desire to cover up his “crime”, Morse goes to the crime scene and
tries to appear as usual: he searches the scene; talks to potential witnesses, and so on.

The intuitive verdict about this case, she argues, is that Morse does not count as genuinely
inquiring into who killed the doctor and does not count as having any interrogative attitudes
towards the question. To be sure, she might behave just like someone who is wondering, curious,

94
Friedman distinguishes between complete and partial answers to a question by way of an example: “the question,
Who in this café is drinking tea? May be partially answered by the proposition, The person sitting at the corner
table, but that answer doesn’t fully settle the question—it doesn’t say of everyone in the café whether they are
drinking tea or not” (Friedman, forthcoming: 8).
95
In order to ensure that it is belief rather than knowledge that is at stake, Friedman stipulates that Morse’s thoughts
are only pseudo-memories, implanted in his mind by his arch-nemesis, and that his flat has been set up to confirm
these pseudo-memories.

99
deliberating, etc., about who killed the doctor and also take part in the investigation. Still,
describing him as inquiring into who killed the doctor, or as being curious, wondering,
deliberating, etc., about who killed the doctor, is hardly intelligible. At best, he is rightly
described as pretending to be inquiring, wondering, deliberating, etc.
This judgment about Morse lends support to the view that necessarily, you would be in a
conflicted or incoherent state if you believed p
Q
and at the same time inquired into Q (or have an
interrogative attitude towards Q), where p
Q
is a complete answer to Q.
96
This gives a sense in
which belief necessarily involves being settled on a question: the fact that you cannot coherently
inquire into Q and simultaneously believe p
Q
means that in believing the answer, you will drop
the interrogative attitude towards Q, and move out from an inquiring state of mind, if you are
relevantly coherent.
Moreover, this view captures the idea that settledness is a distinguishing property of
belief, which can be seen by considering the relation between inquiry and other cognitive
attitudes: intuitively, there is no conflict or incoherence in, for example, supposing, hoping,
imagining, or having high credence in p, and simultaneously wondering, being curious about, or
deliberating about whether p: one could assume (or suppose) a scientific hypothesis in the
context of testing it against evidence, but still wonder whether the hypothesis is true; one could
hope (or imagine, or be highly confident) that one will recover soon from one’s disease while
genuinely wondering if one will recover soon.

96
One might worry that believing a proposition that is a complete answer to a question isn’t really sufficient for you
to be settled on the question, if you don’t also take the proposition to be a complete answer. I think this worry is
pressing for wh- questions, such as Who in this café is drinking tea?: you might believe that Ralph, Mark, and Steve
are drinking tea, which in fact is a complete answer, but you might still wonder whether if there are other people
who are drinking. In such a case, we don’t seem to have settled the question. Fortunately for my purposes, this
problem doesn’t seem to arise for polar questions. For example, if you’re wondering whether Steve is drinking tea
and believe that Steve is (or is not) drinking tea, then you are settled on the question. As we shall see, what I need
for my argument is only that believing a proposition involves being settled on such whether- questions.

100
How exactly should we understand the claim about the incoherence between settling
attitudes and inquiry? Since incoherence is often taken to be an important form of normative
failure, it is tempting to construe the claim as a normative one, which Friedman does. On her
view, to say that believing p
Q
is incoherent with inquiring into Q is to say that the following,
wide-scope requirement on belief and inquiry holds:

Don’t Believe and Inquire (DBI): One ought not inquire into/have an interrogative
attitude towards Q at t and believe p
Q
at t. (Friedman, forthcoming: 8)

However, as Friedman herself notes about Morse, the intuitive judgment about Morse is that he
does not count as genuinely inquiring into or as having an interrogative attitude towards Q. More
generally, the relevant datum seems to be the following:

Unintelligibility Result: It is difficult to make sense of someone as believing p
Q
at t and
inquiring into Q at t.

But it seems that Unintelligibility Result directly supports the following principle, rather than
DBI, as our judgment is that Morse cannot be inquiring, given his belief:

Can’t Believe and Inquire (CBI): One cannot inquire into/have an interrogative attitude
towards Q at t and believe p
Q
at t.

101
Why not then accept CBI rather than DBI? This is because Friedman thinks that there are cases
in which one does seem to inquire into a question while believing an answer to it. She writes,

I think it’s fairly easy to get cases in which one knows or believes the answer to some
question at a time and has an [interrogative attitude] towards that question at that same
time. Specifically when the known or believed answer is hidden from conscious
awareness it’s easy to imagine subjects wondering or curious about or contemplating the
questions those beliefs answer. (Friedman, forthcoming: 7)

The following case serves as a good example:

Lapse: Getting ready to go out, you look for your keys. For couple minutes, you fail to
recall where you put them, and seriously wonder where they are. You manage to recall
that you have put them on the shelf only after getting your thoughts together.

It might seem that even at the moment you are wondering you still have the belief about (and
even knowledge of) their whereabouts: what you undergo is simply a momentary memory lapse.
So a case like Lapse might be seen as a counterexample to CBI.
Moreover, Friedman (2017: 310) argues that there is something epistemically unhappy or
unfortunate about your state of mind in such cases. For example, you might feel like a fool upon
recalling where the keys are, judge that you were confused, or that you should not have been
wondering, etc.. On Friedman’s view, this the kind of normative failure forbidden by DBI, which
CBI does not capture.

102
What about Unintelligibility Result, then? Friedman’s explanation is that it is a mark of
having an incoherent combination of attitudes that it is difficult to make sense of someone as
jointly holding the involved attitudes in full awareness. For example, it is difficult to intelligibly
describe someone as believing P and believing not-P or preferring A to B and B to C and C to A
in full awareness. She argues that this explains why assertions like ‘Morse killed the doctor, but I
am wondering/curious about/deliberating about, who killed the doctor (or whether Morse killed
the doctor)!’ sound confused: such assertions imply that one is fully aware that one has an
incoherent combination of attitudes, which we expect agents not to hold in full awareness.
97

Friedman’s final defense of DBI is that it nicely explains our intuitive judgment about
“high stakes” cases in epistemology, such as the following:

Bank: You are driving past a bank on a Friday afternoon and you have a paycheck to
deposit. You have visited the bank on Saturdays several times in the past and believe that
the bank will be open tomorrow. But it suddenly occurs to you that you have an
impending bill due tomorrow, which, if unpaid, would impose a huge penalty. You are
also aware that some banks change hours. (DeRose 1992; Stanley 2005)

Friedman argues that the standard intuition about cases like Bank, which is that you do not count
as knowing and ought not believe that the bank will be open, can be nicely explained in terms of
DBI: DBI implies that, in Bank, you ought not both inquire into whether the bank will be open
and believe that it will be open; but you ought to inquire into whether the bank will be open in
Bank; thus, you ought not believe that it will be open.

97
This, of course, is Worsnip’s view which I discussed in Chapter 3.

103

2. Incoherence as a Descriptive Relation: Against Friedman
This section argues, against Friedman, that the claim about incoherence between settling
attitudes and inquiry should be understood as a descriptive one. When it comes to occurrent
beliefs, it is more plausible to regard it as a descriptive claim: you cannot believe p or intend to A
while at the same time inquiring into relevant questions. When it comes to dispositional beliefs,
it is possible to have them while inquiring into relevant questions, but it is simply false that you
(always) go wrong in some way whenever you inquire into the questions without dropping the
dispositional attitudes. Either way, the claim about incoherence cannot plausibly be a normative
one. It is rather a descriptive principle about how occurrent beliefs relate to inquiry.
To see why it is plausible to take CBI to hold for occurrent beliefs, note first that there is
a perfect sense in which you do not believe that the keys are on the shelf when you are genuinely
wondering where they are: you would not have any confidence at all about the matter; you would
be reluctant to make any judgments or assertions about where they are; you would seek further
information or evidence, etc., just as someone who does not have any relevant belief would.
To see that this is a sense of ‘belief’ people do latch onto, consider the following case,
put forward by Colin Radford as a counterexample to the thesis that knowledge entails belief:

Unconfident Examinee: You have studied hours for the history test, and are now facing
the final question, ‘what year did Queen Elizabeth die?’, the answer to which you have
repeatedly memorized. However, your proctor reminds you that you have only a minute

104
left, and you suddenly go blank. Totally flustered, you fail to recall the answer. At last,
you (correctly) write down “1603”, feeling as if you are guessing.
98

Radford’s diagnosis is that the examinee does not believe (although knows) that Queen Elizabeth
died in 1603. This diagnosis has been bolstered by the empirical studies conducted by Blake
Myers-Schultz and Eric Schwitzgebel (2013), in which ordinary English-speaking participants
tended to attribute knowledge to the examinee (87%) but without attributing belief (37%).
This is significant because Unconfident Examinee is similar in crucial respects to Lapse,
and it is plausible to think that they admit of the same analysis: you have a piece of information
as a stored item in your representational system; you are seriously wondering about some
question to which it is relevant; you try hard to recall the information but fail, due to some
memory lapse, etc. Moreover, it seems reasonable to expect that we can, by making these
respects more salient in describing Lapse, elicit the judgment that you do not believe that the
keys are on the shelf.
This is not to deny that there is an equally good sense in which you do believe the
answers in both Lapse and Unconfident Examinee. Rather, the point is that we need a principled
distinction that enables us to capture our ambivalence towards each case. This I take to be the
distinction between occurrent and dispositional belief, which I explained in Chapter 2: in each
case, you dispositionally believe the answer, but fail to occurrently believe it. This in fact is the
distinction David Rose and Jonathan Schaffer (2013) also invoke in defending the thesis that
knowledge entails belief in the face of an example like Unconfident Examinee: although you fail
to occurrently believe that Queen Elizabeth died in 1603, you still dispositionally believe that she

98
This is a modified version of Radford (1966)’s original example, due to Myers-Schultz and Schwitzgebel (2013).

105
died in 1603, in the sense that you have learned and stored the information in your memory. On
their view, this dispositional belief is what is entailed by your knowledge in Unconfident
Examinee, which is nonetheless prevented from manifesting itself (“masked”) by your
momentary fear on this particular occasion, resulting in the failure to occurrently believe the
proposition.
99

This means that we can defend CBI as a principle about occurrent beliefs and one’s
ongoing inquiry. Lapse fails to be a counterexample to CBI thus understood, since it only
illustrates the possibility of inquiring into Q at t while dispositionally believing p
Q
at t. It is
highly implausible that you occurrently believe that the keys are on the shelf in Lapse, because
your belief about the keys is not activated in any relevant sense: you have no idea about their
location at the moment; you are not, mentally or overtly, acting as if the keys are on the shelf,
etc. Moreover, CBI offers the most straightforward explanation of Unintelligibility Result: we
cannot make sense of Morse as genuinely inquiring into who killed the doctor because (i) CBI
holds and (ii) Morse occurrently believes that he killed the doctor. This completes my defense of
CBI as a descriptive principle about occurrent beliefs.
So it seems that DBI should be a claim about dispositional beliefs, if anything. As I’ll
argue, however, DBI is false, understood as a claim about dispositional beliefs: there are cases in
which it is permissible for one to inquire into a question (or even ought to do so) while
dispositionally believing an answer to the question.
To see this, suppose you proceed as follows in Bank: you become aware of the
impending bill, and decide to double-check whether the bank will be open tomorrow and begin
to seek further evidence as to whether the bank will be open. It seems your response in this

99
Rose and Schaffer (2013: 25).

106
situation is perfectly acceptable, epistemically speaking. But if DBI were right, and you also
believed that the bank will be open, something would go wrong with you.
So what should we say about your belief about whether the bank will be open tomorrow?
There is again some ambivalence in attributing the belief that the bank will be open tomorrow.
There is again a sense in which you fail to believe that it will: you seriously wonder whether it
will be open and seek further evidence; you worry that you might be wrong and fail to pay bill,
etc. As Jennifer Nagel (2010: 414) argues, perceived high stakes can trigger what she calls
epistemic anxiety, or one’s inclination or desire for increased (or more thorough) cognitive
activities. Perceived high stakes, on her view, can deprive an agent of the belief she would have
in a low-stakes situation, because “it’s psychologically realistic to read her as needing more
evidence either to make up her mind at all or to attain the same level of subjective confidence in
[the high stakes situation]” (Nagel 2008: 286).
100

However, as John Hawthorne points out, “there seems to be a perfectly reasonable sense
of ‘belief’ in which [you] believe … even when the possibility of error is salient in the relevant
sense” (Hawthorne 2004: 169). Nagel concurs:

[You] would be surprised to find out that the hours had in fact changed. If [you] were
forced to choose, on way or the other, [you] would say that the bank would be open rather
than closed. […] Because [your] high credence in the proposition that the bank will be
open would ordinarily support the outright judgment that the bank will be open, [you] can

100
Nagel refers to Arie Kruglanski and Dona Webster, who introduced the concept of cognitive closure, “the
juncture at which a belief crystallizes and turns from hesitant conjecture to a subjectively firm ‘fact’” (Kruglanski
and Webster 1996: 266). This is noteworthy, since the notion of cognitive closure is similar to the characterization
of settledness we have been interested in. They argue that need-for-closure can vary within an individual, depending
on their desires and circumstances. Nagel’ diagnosis is that subjects in high stakes situation are typically have
relatively low need-for-closure, and therefore consider more competing hypotheses and search relevant information
more thoroughly than subjects with relatively higher need for closure.

107
be fairly described as having a general tendency to judge that the bank will be open.
(Nagel 2010: 423, emphases added)

In order to do justice to both judgments, we again need the distinction between occurrent and
dispositional belief: the above counterfactuals about how you would feel, assert, or judge, for
example, strongly suggest that you dispositionally believe that the bank will be open, but this
dispositional beliefs of yours fails to be manifested due to your heightened epistemic anxiety on
this particular occasion, just as your belief fails to be occurrent in Unconfident Examinee.
Moreover, when you eventually gain further evidence and conclude that it will be open, it does
not seem correct to describe you as having lost the whole set of dispositions entirely, as soon as
you start seeking further evidence, and regained them, as soon as you reaffirm that the bank will
be open. As noted in Chapter 2, such dispositions are plausibly based on a “categorical basis”,
which is realized in relatively stable neural configurations. So a better description seems that the
dispositions are momentarily masked throughout the reaffirmation process.
If this is right, however, inquiring into whether the bank will be open while
dispositionally believing that it will be open does not involve any normative failure, since it
seems that double-checking, which is a form of inquiry, is a perfectly reasonable response in
Bank. Unlike in Lapse, you would not feel like a fool, or judge that you were confused, or that
you should not have double-checked, after reaffirming your dispositional but nonoccurrent belief
that the bank will be open. Moreover, it is implausible that you rationally ought to give up your
dispositional belief about the bank altogether. For example, it is not as if you are faced with
strong considerations that defeat your justification for the belief that the bank will be open, and it
seems permissible, if not required, for you to retain the dispositional belief, provided that you do

108
not act on the basis of this belief and hold off judgments on the matter. Since it is plausible that
retaining your dispositional belief in this way is permissible in Bank, DBI is false as a principle
about dispositional belief.
But what about the intuition that something goes wrong in Lapse, which Friedman cites
in favor of DBI? I think that such an intuition (if there is any) can be explained away without
positing DBI. The real failure (again, if there is any) lies in the failure to make effective use of
one’s epistemic resources, which isn’t obviously tied to the sense of irrationality that is focal in
my dissertation.
First, it is implausible that your failure in Lapse is a rational failure. The relevant failure
seems to consist in momentary forgetfulness, but it is implausible that such forgetfulness always
is a form of irrationality.
101
As a creature with limited cognitive capacities, it is simply
impossible for us to hold everything we learn in our mind. We might even have good epistemic
reasons to be selective in storing and maintaining information one acquires. Plausibly, in
responding to such reasons, one would inevitably forget information that is trivial. Forgetfulness
in general can be seen as part of making an effective use of one’s limited cognitive resources in
the pursuit of one’s epistemic goals and, as such, seems to be rationally permissible in many
cases.
102

Second, what goes wrong in Lapse in fact is better explained by this point about making
the most of one’s cognitive resources gives us the key to explaining what goes wrong in Lapse: if
making the best use of one’s limited cognitive resources is an epistemic ideal, wasting your

101
See Archer (forthcoming) for the point that it seems to be simply a psychological failure rather than a normative
one.
102
This point is continuous with Harman’s (1986) point about reasons not to “clutter one’s mind” with trivialities.
See also Foley (1987) and Svavarsdóttir (2006) for a conception of epistemic rationality as using one’s cognitive
resources as well as possible, given one’s epistemic limitations.

109
limited cognitive resources should be epistemically objectionable. In Lapse, you do not
encounter any considerations that might improve or undermine the current epistemic standing of
your belief. For example, you do not acquire any evidence against it, or you do not become
aware of the cost of being wrong. You simply waste your time and energy pointlessly wondering
where the keys are, when you already have the answer you could have recalled. This seems to be
a distinctively epistemic vice, which can be seen as a violation of the following, prima facie duty
not to inquire when one believes:

Don’t Waste Resources: Other things being equal, if one believes p
Q
, then one ought not
waste one’s cognitive resources by inquiring into Q.

With Don’t Waste Resources, we can explain the failure in Lapse as a failure to conform to your
(prima facie) epistemic obligation, without invoking a problematic principle like DBI.
Moreover, we can understand the asymmetry between Lapse and Bank in terms of Don’t
Waste Resources. In Bank, it seems okay to (dispositionally) believe and inquire at the same
time, precisely because other things are not equal, given other features of the situation. For
example, you come to realize that a great deal hangs on your being right that the bank will be
open tomorrow, which overrides the costs of seeking further evidence as to whether it will be.
Moreover, you do not yet currently have evidence that would rule out the possibility that the
bank will not be open, which means that a further inquiry is necessary.
This completes my argument that settledness of belief should be understood in terms of
CBI, which is a descriptive principle about occurrent beliefs and inquiry. On the view I propose,
occurrent beliefs are the primary bearers of settledness. Being settled is, strictly speaking, a

110
property of occurrent beliefs, and dispositional beliefs involve settledness only in a derivative
sense: to dispositionally believe is to be disposed to occurrently believe under appropriate
triggering conditions, and therefore to be settled on relevant questions. This is consistent with
such dispositions being masked in abnormal circumstances (as in Unconfident Examinee and
Bank), in which case one might fail to be settled as one would be in normal circumstances.

3. Settledness and Rationality
Occurrent beliefs are settled in the sense that one cannot occurrently believe while at the same
time inquire into relevant questions. On the basis of this distinctive dimension of occurrent
beliefs, this section argues that some prominent coherence requirements on beliefs are also
descriptive principles about occurrent beliefs. More specifically, the settledness of occurrent
beliefs makes it plausible that occurrent beliefs necessarily satisfy the following principles (I will
drop ‘occurrent’ for the remainder of this section):

No Contradictory Belief: One ought not believe that p and believe that not-p at the same
time.

Modus Ponens: One ought to believe that q, if one believes that p, believes that if p then
q, and attends to whether q.

Inter-Level Coherence: One ought (i) to believe that p if one believes that one has
sufficient evidence that p; and (ii) not believe that p if one believes that one lacks
sufficient evidence that p. (Worsnip 2018a)

111

The first step of the argument is that settling (or being settled on) a question involves ruling out
salient answers to the question. In our ordinary reasoning and deliberation, we typically consider
concrete scenarios or possibilities as potential answers to the questions we are trying to solve.
Such questions can take a variety of forms. For example, you might wake up in the morning,
remembering the weather forecast saying it’s likely to rain today. So you look outside the
window, and come to believe that it is raining. As you get ready to leave home, you look for your
key, and see it lying on the shoe rack. On your way to campus, you check the updated class
roster, and notice that the number of registered students went down from 25 to 24. You compare
it with the previous version and figure out who dropped the course. In forming beliefs about
these matters and bringing your inquiry to a close, you settle a series of questions: Is it raining?;
Where is the key?; Who dropped the course?.
What is it for an answer to be salient to you? I suggest that an answer is salient to you in
case you are taking the answer seriously, that is, as a potential answer to the question. More
formally, a set of salient answers to a question are possibilities that divide up your epistemic
space that are mutually exclusive and jointly exhaustive with respect to the question. With that
said, settling a question can be identified with ruling out alternative answers (or other epistemic
possibilities). Let’s apply this idea to the case we are considering, where you settle a series of
questions. The set of salient answers can be, respectively: {it is raining, it is not raining}; {it is
on the shoe rack, it is on the bed, it is on the shelf, it is next to the coffee machine}; {it is Student
1, it is Student 2, …, it is Student 25}. In each case, you answer the question by ruling out other
potential answers, thereby judging the answer to be true.

112
That our belief-formation is sensitive to the range of salient alternatives and our ability to
rule them out can be seen by comparing Unconfident Examinee with its variant. In Unconfident
examinee, we can reasonably assume that the answer Queen Elizabeth (‘QE’) died in 1603 is one
of the salient answers, since you correctly guess the answer. Your failure to believe the
proposition, and to answer the question ‘Did QE die in 1603?’ can be traced back to the failure to
rule out the salient alternatives, the set of which could be: {1601, 1602, 1604, 1605}. Each
answer is a potential answer to the question ‘When did QE die?’, representing some concrete
possibilities in which it is not the case that QE died in 1603. Since she fails to rule out those
possibilities in which QE didn’t die in 1603, she fails to believe that QE died in 1603. But if the
question took a different form, and (accordingly) the salient alternatives were different, you
could have formed the belief. For example, if our examinee were faced with the following
question, then she might have easily judged that QE died in 1603: ‘Among the famous historical
figures we learned in class, who died in 1603?’. To that question, the set of salient alternative
answers could have been: {Henry III, Queen Mary, Queen Elizabeth, James I}, and she could
have more easily judged that QE died in 1603, by ruling out the salient alternative answers.
103

The set of salient answers is determined by relevant background beliefs and desires. Your
background beliefs about where you usually put the key, for instance, disqualify numerous far-
fetched possibilities from being a salient answer, such as: it is on Mars; it is in the toilet bowl.
Moreover, the set of salient answers can shift as the inquiry goes on. If it had turned out that
none of its members answer the question, you would have begun considering an entirely
different set of answers, and the above answers would no longer be salient. Similarly, in the third

103
This observation is made by Blaauw (2013), in the context of defending contrastivism about belief ascription.

113
judgment, if you had some background beliefs about who recently expressed dissatisfaction with
their grades, the set of salient answers could have been narrowed down to those students.
Similarly, your background desires, goals, or values, can also function to constrain the
range of salient answers. Kruglanski and Webster, for example, argue that if you have strong
need for cognitive closure, that is, if you have a “desire for a firm answer to a question and an
aversion toward ambiguity”, you might consider “fewer competing hypothesis or suppress
attention to information inconsistent with [your] hypothesis.”
104
This suggests an inverse
correlation between the strength of one’s desire for a settled answer to a given question. If, due
to time pressure, fatigue, or desire for consensus, etc., you strongly desire closure, you would
consider less alternatives.
Still, there is a sense in which in believing p, one is necessarily settled on single kind of
question: the question whether p (or: whether p is true; whether p is the case). Although you
settle different forms of questions in each instance (whether it is raining, where the key is, who
dropped the course) by forming a belief, when you come to believe something that is a complete
answer to the question, you thereby settle a whether- question: in believing that it is raining, you
settle the question whether it is raining; in believing that the key is on the shelf, you settle the
question whether it is on the shelf; in believing that Student 1 dropped the course, you settle the
question whether they dropped the course. Since settling a question involves, as we have seen,
ruling out salient alternatives, we can maintain that to believe p, in general, is to be settled on the
question whether p, by ruling out the alternative answer: not-p.
With this in mind, let us consider whether it is possible to violate No Contradictory
Belief. Suppose you believe that Trump is a man. This, on my view, means that you are settled

104
Kruglanski and Webster (1996: 264). See fn. 10 for the concept of cognitive closure.

114
on the question whether Trump is a man by having a particular answer to it, thereby ruling out
the salient alternative answer, Trump is not a man. But for you to violate No Contradictory
Beliefs, you must, at the same time, have ruled out Trump is a man as an answer to the question.
The state of mind you must be in, therefore, must be the state of having an answer to the question
that you have ruled out at the same time. Assuming that your mind is not compartmentalized (in
the way characterized in Chapter 3), however, it is unclear how it is possible to be in such a state.
But one of the main arguments from Chapter 3 has it that if your mind is compartmentalized,
either you lack the general capacity to satisfy the relevant requirement or it is not the case that
you would satisfy the requirement if you fully exercised the capacity. In either case, you are not
subject to the requirement, since you lack the specific ability to satisfy it and could not be
legitimately expected to satisfy it. The lesson is that it is difficult to come up with a possibility in
which someone who is genuinely under the requirement violates it.
It is equally difficult to conceive of a situation in which you violate Modus Ponens.
Suppose you simultaneously believe that if Trump is a man then Trump is a human being,
believe that Trump is a human being, and attend to the question of whether Trump is a human
being. This means that you have settled positively the following questions: ‘Is Trump a man?’;
‘Is Trump a human being if he is a man?’, and have ruled out the salient answers. It then
becomes difficult to see how you can fail to believe that Trump is a human being, while
attending to the question of whether Trump is a human being.
This can happen, of course, if you find it dubious that Trump is a human being, and (at
least momentarily) reconsider at least one of the above questions. But this contradicts the
assumption that you believe both premises, because reconsideration is plausibly a way of
inquiring into the question. CBI, which states that you cannot both inquire into a question and

115
believe a complete answer to it, implies that you are not, at the moment, settled on whether
Trump is a man or whether Trump is a human being if he is a man, and the alternative answers
have become salient. This means that you fail to believe either that Trump is a man or that if he
is a man he is a human being, and do not violate Modus Ponens.
Alternatively, this can happen because you are, at the moment, conceptually incompetent,
in the sense that you are unable to correctly apply the concept if to the contents of the antecedent
beliefs and conclude that Trump is a human being. This might be because you simply lack the
general ability to use the concept correctly, or because you lack the specific ability to do so, due
to some interfering factors that might prevent you from applying the concept correctly or drive
you out from your “epistemic safety zone”. However, as I have argued in Chapter 3, the
conditions under which you lack the specific ability to satisfy a requirement of rationality are the
conditions under which you are not subject to the requirement. As in the case of No
Contradictory Beliefs, it is difficult to see how anyone who is genuinely subject to the
requirement could violate it.
Finally, consider Inter-Level Coherence, which demands coherence between one’s beliefs
about evidence and one’s first-order beliefs. Suppose you are competent with the concept of
evidence. Then you grasp that your evidence as to whether Trump is a man consists precisely in
those considerations that bear on whether Trump is a man, and that for your evidence to be
(in)sufficient is for your evidence to be (in)sufficient to establish that Trump is a man. If so,
however, it is then unclear how one could violate Inter-Level Coherence: to believe that you
have sufficient evidence for the truth of a proposition P is, assuming that you are competently
using the concept of (sufficient) evidence, to believe that there are considerations that establish P
as true, which means that you are settled (positively) on the question of whether P is true. If so,

116
you also believe P and satisfy (the first clause of) Inter-Level Coherence. Likewise, to believe
that you lack sufficient evidence for P is to believe that there are not yet considerations that
establish P as true, which means that you are not settled on the question whether P. This means
that you do not count as believing P, and again satisfy (the second clause of) Inter-Level
Coherence.
The lesson from this section is that, so long as we restrict our attention to situations in
which one is genuinely under a coherence requirement, the settledness of belief makes it
extremely difficult, if not impossible, for one to violate the requirement.

4. Revisiting the Rationality of Occurrent Intentions
Let me conclude this chapter by drawing connection between it and Chapter 3 and offering a
more unified picture of the coherence “requirements”. In Chapter 3, I have only briefly suggested
that the reason why we cannot intelligibly attribute incoherent combinations of occurrent beliefs
and intentions might have to do with the fact that intentions, unlike other conative attitudes like
wish, hope, urge, etc., involves being settled on the question of what to do. This section fleshes
out this idea in detail by showing that we can draw a perfect parallel between the settledness of
belief and the settledness of intention.
The first thing to note is that it has also been widely accepted that if you intend to A, you
take it for granted that you will (or shall) A, in the sense that the question of whether to A is not a
matter for consideration for you. Indeed, the settledness of intention has been taken to be a
distinguishing mark of intention, which sets it apart from intention from ordinary desires.
105

Alfred Mele puts this contrast succinctly:

105
For an explicit endorsement of this point, see Harman (1976), Bratman (1987), Velleman (1989; 1997), McCann
(1991), Mele (1992), Finlay (2007), Holton (2009), Alonso (2017). It should be clear from the outset that picking

117

To have the intending attitude toward a plan is to be settled ... on executing it. The
intending and desiring attitudes towards plans differ in that the former alone entails this
settledness. Someone who desires to A, or to execute a certain plan for A-ing … may still
be deliberating about whether to A, or about whether to execute the plan, in which case
he is not settled on A-ing, or not settled on executing the plan. (Mele 2003: 28)

Importantly, we can cash out the settledness of intention in the same way we did for the
settledness of belief. Recall the case of Morse, from which we got the Unintelligibility Result: it
is difficult to make sense of someone as believing p
Q
at t and inquiring into Q at t. We can in fact
tell a relevantly similar story about a person who is intending to A and inquiring into a question
to which A-ing (or the proposition that they will A) is a complete answer.
To see this, consider the following scenario which is relevantly similar to Morse’s case:
Jill, out of her passion for philosophy, decides and forms the intention to pursue a Ph.D. in
philosophy, but nonetheless wants to keep it secret from her parents until the right time comes, as
they would otherwise oppose her decision and hurt her feelings. When discussing which career
to pursue, she tries to appear as serious as she can, goes through several options with her parents,
participates in tallying up the pros and cons for each option, etc.
The intuitive judgment about Jill, again, is that she does not count as inquiring into, or
having an interrogative attitude towards, the question of which career to pursue, although she

out being settled as the distinguishing mark of intention does not cast reductionism about intention in an unfavorable
light, the view according to which intentions can be analyzed into belief-desire pairs. The point is simply that there
is something to be accounted for by any adequate theory of intention, which is widely acknowledged to give a
distinctive dimension to intention, not that intention must therefore be a psychological state independent from desire.
Indeed, reductionists about intention have taken the settledness in intention seriously in their reductive analyses of
intention. See Ridge (1998) and Sinhababu (2013).

118
does what a typical student would do when they genuinely inquire what to do for their future.
Describing her as wondering about or deliberating about the question would seem confused,
assuming that she keeps intending to pursue a Ph.D. in philosophy throughout the story.
This, again, supports the view that necessarily, you would be in a conflicted or incoherent
state if you intended to A and at the same time inquired into Q (or had an interrogative attitude
towards Q), where A-ing (or that you will A) constitutes a complete answer to Q. Your coming to
intend to A settles the question to which A-ing (or that you will A) is a complete answer, in the
sense that in intending to A, you will drop a relevant inquiring attitude and close inquiry, so long
as you are relevantly coherent. Moreover, this captures the idea that intention is distinguishable
from other conative attitudes by its settledness. Jill might, without incoherence, wonder or
deliberate about which career to pursue while hoping to, wishing to, being strongly inclined to,
having a strong urge to, pursue a Ph.D. in philosophy.
And just as we explained Unintelligibility Result for belief in terms of CBI, we can
explain the datum in terms of the following principle, which is a descriptive principle about
occurrent intentions and inquiry:

Can’t Intend and Inquire (CII): One cannot inquire into/have an interrogative attitude
towards Q at t and occurrently intend p
Q
at t.

Again, the incoherence between intention and inquiry should not be understood as a normative
relation, since it is false that one ought not to inquire into a question and (dispositionally) intend
a complete answer to the question. This can be seen by considering the following case:

119
Calendar: You set up a detailed plan to learn several languages starting from April, and
write down in your calendar which one to study for each day. You wake up and see that
it’s the first day of April, but you don’t remember which language you planned to study.
You look into the calendar and recall that it was Japanese, which you immediately come
to intend to study.

In Calendar, you have a dispositional intention to learn Japanese, presumably since you wrote it
down in the calendar, but fail to recall that you were going to do so, and genuinely wonder which
one to study today. As in the case of belief, there is a sense in which you fail to occurrently
intend to study Japanese, at least at the moment when you wonder about the question of which
language you will study, to which it is a complete answer. But there seems nothing amiss with
you in Calendar: no sensible normative theory of intention should forbid using calendars for the
purpose of keeping track of and organizing our complex plans. There might be a sense of
epistemic failure, which we observed in Lapse. But that could again be explained in terms of the
violation of Don’t Waste Resources, not by positing a principle telling us that you go wrong in
some way whenever you wonder what to do when you dispositionally intend an answer to that
question. The settledness of intention means the incompatibility between occurrent intention and
one’s ongoing inquiry.

5. The Picture: Principles of Coherence as Descriptive Principles
On the picture I propose in Chapter 3 and 4, there are no normative requirements of coherence on
belief and intention. They are descriptive principles about how we believe and intend under
normal conditions, that is, the conditions under which one’s mind is not incapacitated by some

120
aberrant features. In chapter 3, I have argued that such conditions might prevent one’s mind from
satisfying the principles at issue, they are at the same time the conditions under which one is not
properly evaluated as irrational. This chapter has been an attempt to identify a feature that belief
and intention share, being settled on a question, which makes it plausible that one’s occurrent
beliefs and intentions cannot be (irrationally) incoherent.

121
Chapter 5: A New Way Out of the Preface Paradox

Chapter 5 and Chapter 6 apply the idea about the proper jurisdiction of coherence requirements
to some well-known puzzles in epistemology. In each chapter, the central focus will be on
arguing that recognizing the proper target of coherence requirements on belief as occurrent
beliefs provides us with a neat solution to the puzzle, thereby giving further support to the idea
that coherence requirements govern occurrent attitudes. For this reason, I’ll deliberately set aside
the question of whether such requirements are really violable and speak as if they were violable
requirements.
This chapter offers a new way out of the preface paradox. The paradox is widely taken to
pose a challenge to the idea that one’s beliefs rationally ought to be deductively cogent, that is,
the idea that one’s beliefs rationally ought to be logically consistent and one’s beliefs rationally
ought to be deductively closed. However, the preface paradox is commonly taken to present us
with a case in which a subject who doesn’t meet such requirements appears to be fully rational.
So we are apparently forced to deny either that they are genuine requirements of rationality or
that the subject described in the paradox can be fully rational.
This chapter argues that the alleged choice isn’t forced upon us: we can retain the
deductive constraints on beliefs, consistently with holding that the author can be rational.
Whatever independent motivation there can be for giving up the deductive constraints at issue,
the preface paradox itself doesn’t give us any reason to do so, since the rationality of the author’s
beliefs can be adequately explained without compromising any of the deductive constraints.
Once we recognize the deductive constraints on belief as being imposed upon occurrent rather
than dispositional beliefs, we can see how the author can be perfectly rational. Moreover, I argue

122
that this solution is superior to the standard, probabilistic solution to the paradox, which rejects
deductive constraints at issue. This is because wholesale rejection of such constraints on belief
has some troubling consequences for our argumentative and deliberative practice, which my
solution avoids.

1. The Paradox
This section pinpoints the set of beliefs and the rational constraints that together generate the
preface paradox. In the original preface paradox, presented by David Makinson (1965), a
historian (call him ‘Eric’) rationally believes every claim he makes in the body of the book. Call
them body claims. Let’s suppose that, for each body claim, he forms a belief in it on the basis of
adequate evidence. But he is also aware of compelling inductive evidence that his book contains
some error: no one has ever managed to produce a book as lengthy as this one that is completely
error-free. So he believes, and claims that some error will be found in the book. Call it the
preface claim. So far, Eric seems to have proceeded in a way that is entirely rational.
But the following principles intuitively seem to be important deductive constraints on
rational belief:

Consistency: Rationality requires of you that your beliefs be logically consistent.

Conjunctive Closure: Rationality requires of you that you believe p-and-q if you believe
p and you believe q.

123
It is easy to see their initial appeal. Having a logically consistent set of beliefs is often taken to be
a necessary condition for having a coherent set of beliefs. So long as one thinks that if there are
any coherence requirements on belief at all, it is natural to think that Consistency must be one of
them. Conjunctive Closure is an instance of the more general view that one’s beliefs rationally
ought to be deductively closed, that is, one rationally ought to believe what is logically entailed
by what one believes. To see the appeal of Conjunctive Closure, consider a person who says, ‘I
believe that the earth is round, and I believe that the earth revolves around the sun. But if you ask
me if the earth is round and it revolves around the sun, I don’t believe that’. It seems that
something has gone wrong with the person, and we can explain what it is by positing a constraint
like Conjunctive Closure.
However, accepting either Consistency or Conjunctive Closure seems to commit us to
accepting that Eric could not be rational in believing all of the body claims and also believing the
preface claim. To see this, let {P1, P2, ..., Pn} be the set of body claims and let ‘Bel [ ]’
designate the belief in the proposition in the square bracket. Eric’s beliefs in the body claims can
then be represented as follows:

(1) Bel [P1], Bel [P2], …, Bel [Pn].

Eric’s belief in the preface claim can be represented in such a way as to make its content
logically inconsistent with the contents of the beliefs in (1):

(2) Bel [~(P1&P2&…&Pn)]

124
Represented as such, it is easy to see how Eric’s belief-set can cause trouble for both
Consistency and Conjunctive Closure. The beliefs in (1) and (2) are logically inconsistent. So if
Consistency is true, it cannot be rational for Eric to believe both (1) and (2). This runs contrary to
our initial judgment that Eric can be rational. Moreover, Eric is required, by Conjunctive
Closure, to believe the conjunction of the body claims, so long as he believes each body claim.
And Eric is required, if he believes the conjunction of the body claims, to conjoin that
conjunction with the content of (2), which means that he is required to believe a contradiction:
((P1&P2&…&Pn)&~(P1&P2&…&Pn)). So accepting Conjunctive Closure seems to commit us
to denying that
So there is the puzzle. It seems perfectly rational for Eric to believe each claim in the
book, including the preface claim. But if we accept either Consistency or Conjunctive Closure, it
cannot be rational for Eric to believe every claim in the book, which runs contrary to the intuitive
verdict about Eric’s rationality.
Before proceeding further, it is worth considering whether the paradox could be evaded
simply by adopting a different way of representing Eric’s belief-set. Indeed, some
writers have pointed out that (2) is not an accurate representation of Eric’s belief about the
preface claim, which might suggest a simple dissolution of the paradox. What Eric believes and
writes in the preface is a second-order claim about the (first-order) claims made in the book, and
not the negation of the conjunction of all of the body claims. That is:

(3) Bel [At least some claim I make in this book is false]

125
With that said, one could question whether (1) is inconsistent with (3). Indeed, Evnine tries to
resolve the paradox by (rightly) pointing out that beliefs in (1) and (3) aren’t strictly inconsistent.
For (1) consists of beliefs in the first-order claims in the book, whereas (3) is a belief in a
second-order proposition about the claims made in the book, and it is logically possible for the
contents in (1) and the content in (3) to be true together.
106

This attempt to evade the paradox by itself isn’t fully convincing, however, since the
paradox can be generated with the aid of some reasonable auxiliary assumptions.
107
Consider the
following assumptions, for example: (i) for every claim in the book, if Eric believes it, then Eric
also believes that it is a true claim in the book; (ii) Eric believes that there are no other body
claims than P1, P2, …, Pn. If we combine (i) and (ii) with (1), the following would also be part
of Eric’s system of beliefs:

(4) Bel [P1 is a true claim in the book]; Bel [P2 is a true claim in the book]; … Bel [Pn is
a true claim in the book]; Bel [there are no claims other than P1, …, Pn in the book].

What this shows is that, even if (1) and (3) aren’t strictly inconsistent, we can easily reinstate the
paradox by helping ourselves to some fairly minimal assumptions about Eric’s beliefs. So one
cannot simply avoid the paradox by representing his belief in a different way.
There is something revealing about this proposed solution, however. The main reason
why people think that Eric’s belief should be represented as (3) rather than (2), is that it seems
impossible for Eric to ever form the belief represented in (2). To form the belief in the negation
of the conjunction of all of the body claim, perhaps Eric must consider the long conjunction of

106
See also Weatherson (2005: 429).
107
For similar criticisms on Evnine’s solution, see Christensen (2004), Clarke (forthcoming), Kim (2015).

126
the body claims and then negate it. But ordinary authors like Eric lack the capacity to entertain
such a long conjunction. The same worry might be raised for the final belief represented in (4): it
is simply implausible that Eric has (or can form) a single belief about exactly which propositions
his book contains, which seems as psychologically demanding as forming a belief in the long
conjunction of the claims in the book.
This point about the cognitive load involved in forming beliefs like (2) and (4) is an
important one, crucial to my own solution to the paradox (I’ll return to it in Section 3). Still, we
aren’t yet told why our limited cognitive capacities should matter for the solution. We can still
conceive of creatures who don’t suffer from the same kind of limitations and are fully able to
form beliefs like (2) and (4). If such a being could rationally believe the preface claim and
believe each body claim at the same time, our contingent cognitive limitations would not be
relevant to the puzzle at all. Moreover, even if such limitations turn out relevant, we are still in
need of a principled story about how exactly they are relevant.

2. The Standard Solution and its Problems
This section explains the standard solution to the paradox and shows why it isn’t satisfactory
with respect to respecting our ordinary argumentative and deliberative practice. The standard
story begins with the following principle:

Lockean Thesis: It is rational to believe p if and only if it is rational to have a degree of
belief in p higher than some threshold value T (greater than zero). (Foley 1993)

127
It is a short step from Lockean Thesis to seeing how Eric could be rational in believing both the
preface claim and the body claims at the same time. All it takes is for his credence in each body
claim to be greater than or equal to T (which is assumed to be lower than one) and for his
credence in the preface claim to be greater than or equal to T. For example, if T=0.99 and there
are at least 99 body claims, all of Eric’s beliefs can be rational. Surely it doesn’t seem rationally
required for Eric to invest credence 1 (or absolute certainty) in every claim in the book, so it is
always possible to imagine a sufficiently lengthy book which can render Eric’s belief-set
rational: all we need is just to increase the number of the body claims, as T gets higher. There
will be a point at which the probability of the conjunction of the body claims falls below 0.01
and the probability of its negation passes T, which will make it rational for Eric to believe the
preface claim.

2.1. Conjunctive Closure
Since the standard solution rationally permits Eric to believe each body claim and the preface
claim without conjoining them, it is committed to rejecting Conjunctive Closure. But it seems
that it would be irrational to violate Conjunctive Closure if the number of claims were
sufficiently small, which calls for explanation.
108

To see this, consider the objection Mark Kaplan presses against the denial of Conjunctive
Closure entailed by the standard solution.
109
It is part of our argumentative practice that, for any
propositions p, q, r, s, you can persuade a person who believes p, q, r to accept s (or to give up

108
There might be some other conditions, too. If you believe two, totally unrelated propositions, it might not be
irrational for you not to bother conjoining them. I’ll get to this issue when I formulate my own versions of
Consistency and Conjunctive Closure.
109
Kaplan (1996: 96-7, 2013) Foley (1993) addresses the same worries in justifying the denial of conjunctive
closure.

128
their belief in p, q, or r) by showing that p, q, and r jointly (but not individually) entail s. If
Conjunctive Closure were to be given up, however, the person could happily refuse to accept s
without revising any of their beliefs, saying: ‘Sure, I believe P, believe Q, and believe R. None of
these individually entails S. Sure, S follows from (P&Q&R). But so what? I don’t believe that!’
But this seems incoherent.
Christensen meets Kaplan’s objection by making sense of our argumentative practice in
terms of the norms of probabilistic coherence (Christensen 2004: Ch. 4). For example,
probabilistic coherence requires that if a set of premises logically entails a conclusion, one’s
credence in the conclusion should be as high as one’s credence in their conjunction. Insofar as it
is rational to have a degree of belief in the conjunction of the premises of an argument that
passes the threshold value, therefore, one rationally should accept the conclusion. This,
Christensen argues, accounts for the force of ordinary deductive arguments with relatively small
number of premises.
Still, this doesn’t seem to be the right kind of explanation of our ordinary argumentative
practice. The problem is that even with respect to ordinary deductive arguments with a small
number of premises, rejecting conjunctive closure yields counterintuitive results: for any number
(>1) of premises and any threshold value T, it is possible to construct a case in which you
rationally have credence equal to T in all of the premises, each of which is probabilistically
independent from each other, so that the credence in their conjunction mandated by probabilistic
coherence is less than T. In any such case, Lockean thesis implies that it can be irrational for you
to accept the conclusion jointly entailed by the premises, which is counterintuitive. To illustrate,
suppose you believe the following propositions:

129
P: There will be dancing at the end-of-semester party tomorrow.
Q: It will rain all day tomorrow.
R: Jon doesn’t come to any dance party on a rainy day.

Let the Lockean threshold T be 0.9, for convenience. Then we can easily conceive of a scenario
in which you believe each proposition rationally, with credence 0.9 in each. Every time you
attended an end-of-semester party in the past, there was dancing. You heard about the rain from
the nation’s most accurate weather forecast. As an old friend of Jon, you know about his quirky
disposition to stay away from seeing people dance on a rainy day. So even if you don’t assign
credence 1 to each proposition, you can rationally believe each, by the standards of Lockean
Thesis.
Intuitively, it is perfectly rational for you to conclude that Jon won’t come to the party
(call this proposition ‘S’). If we were to give up Conjunctive Closure in favor of norms of
probabilistic coherence, however, it would be difficult to see how you could come to rationally
believe S. P, Q, and R are probabilistically independent from each other, so the credence it is
rational for you to have in the conjunction (P&Q&R) is 0.729, which falls below the stipulated
threshold value (=0.9). From Lockean Thesis it follows that it is irrational for you to believe
(P&Q&R), which entails S. There is a weak probabilistic constraint on what your credence in S
must be: in general, if p entails q, then your credence in q rationally should at least be as high as
credence in p. Thus, the credence it is rational for you to assign to S is at least as high as 0.729.
There seem no further probabilistic constraints that you should meet in this case.
Suppose now that your credence in S is such that it is above 0.729 but less than 0.9. From
Lockean Thesis it follows that it is irrational for you to believe S. According to the standard

130
solution, even if you accept P, Q, and R, and someone correctly point out to you that S follows
from them, you are able to respond in the following way: ‘Sure. I do believe P, Q, and, R, and of
course, I clearly know that they jointly entail S. Still, I don’t believe S—it is irrational for me to
believe S!’
But it is implausible to say that you can be irrational in either asserting or believing that
Jon won’t come to the party in this scenario. Not only does the standard solution entail that there
is no need to accept the conclusion, it also entails that you can be rationally forbidden from
accepting it. This means, however, that the standard solution simply fails to capture Conjunctive
Closure plays in our ordinary argumentative practice.

2.2. Consistency
Since the body claims cannot all be true if the preface claim is true, the standard solution is also
committed to denying Consistency. Not imposing any consistency constraint on our beliefs,
however, also has disturbing consequences for our argumentative practice. Suppose I believed P,
Q, R, and not-S. It seems perfectly legitimate for you to point out that they cannot all be true,
putting pressure on me to revise at least one of the beliefs. But if there were no principle like
Consistency, however, I might rationally refuse to revise any of my beliefs. ‘I know P, Q, R, and
not-S can’t all be true, but so what? I believe all of them!’ seems incoherent.
Again, proponents of the standard solution might seek recourse in the norms of
probabilistic coherence. For example, Christensen argues that we can explain why ordinary
reductio arguments with relatively small number of premises generally carry weight, in terms of
probabilistic coherence. Suppose again your credence in each claim is 0.9 so that the rational
credence in the conjunction (P&Q&R) is 0.729. By the norms of probabilistic coherence, your

131
credence in S rationally must be at least as high as 0.729, and therefore your credence in not-S
should be below 0.271. Holding fixed the Lockean threshold as 0.9, it would then be irrational
for you to believe not-S. So it might seem that ordinary argumentative practices involving a
small number of propositions aren’t under any threat. Only when the number of relevant claims
is extremely large, as in the preface case, will reductio arguments lose force: for any threshold
value T and number n of body claims, you can be rational in believing all of them and the
preface claim, so long as n ≥ (T/1-T).
110
Again, this seems to offer a nice explanation of why
Consistency appears to be a requirement governing our ordinary argumentative practice.
But this solution faces problems. Consider ideal agents with limitless memory capacities,
who can survey all of the claims in Eric’s book instantaneously and form their beliefs in the
following way: ‘Okay, I have now brought all of the claims to my mind, and considered each. I
can see that P1 is true, P2 is true, …, and Pn is true. So I believe each body claim. Nonetheless, I
don’t believe that every claim is true—actually, it must be false that every claim in the book is
true!’ This seems to involve the same irrationality Eric would exhibit if he believed that there
must be an error whilst believing a book with a very few claims (say, P, Q, R, S), given that the
cognitive load of surveying all of the body claims, for such ideal agents, would be, in the least, as
negligible as the cognitive load of surveying P, Q, R, S for ordinary agents like Eric.
If we agree that these ideal agents would indeed be irrational in this case, then it shows
that the account given by the standard solution doesn’t explain the appeal of Consistency in the
right way. For this is a case in which one would be irrational in believing both the preface claim

110
Foley (1993) makes similar points about the force of reductio arguments: “in proving that the conjunction is
false, reductios provide a potentially powerful argument against any given premise, but the strength of this argument
is a matter of how closely the truth of this premise is tied to the truth of the conjunction.” If there are a huge number
of logically independent premises, the reductio wouldn’t have force, as the preface cases illustrates.

132
and the body claims, although the number of the body claims is sufficiently large that it is
rational for such agents to believe the preface claim and all of the body claims, by the standards
of Lockean Thesis.
There can be a dissenting voice. For example, Christensen writes,

It is undoubtedly true that ordinary humans cannot entertain book-length conjunctions.
But surely, agents who do not share this fairly superficial limitation are easily conceived.
And it seems just as wrong to say of such agents that they are rationally required to
believe in the inerrancy of the books they write. (Christensen 2004: 38)

But Christensen’s verdict seems dubious at best, since our limited cognitive capacity does not
seem to be a “fairly superficial limitation” at all.6 It rather seems to be a general fact about our
assessment of an agent’s rationality that it seems to depend crucially on the agent’s relevant
capacities. Consider, for example, logical skills. While it would not be irrational for infants, who
simply lack logical skills, to fail to detect simple inconsistencies (such as the inconsistency
between not-P or Q, P, and not-Q), it would be irrational for a student who has just taken a
course in introductory logic to fail to do so. And a logic TA, who has studied logic for years, will
naturally be subject to more stringent requirements than this lay student, in virtue of his better
logical skills.
Likewise, facts about an agent’s memory capacity also seem relevant to determining the
agent’s rationality. For whether it is rational for the agent to believe a proposition is partly a
function of how much information or evidence the agent can consider (or assess), which is
determined by the agents’ memory capacity. Suppose that the following pieces of evidence

133
together decisively support a proposition P, but only when taken as a whole: E1, E2, ..., En. But
suppose that although she has access to each piece of evidence, she can, due to her limited
memory, only consider (proper) parts of her evidence, none of which supports believing P. In
such a case, it doesn’t seem irrational for her to fail to believe P. However, if she had a better
memory capacity, and were capable of considering all of her evidence together in her situation, it
would indeed be irrational for her to fail to believe P. This asymmetry is explained partly by the
facts about her memory capacity. If so, there is every reason to expect that abstracting away from
the cognitive limitations of ordinary agents would change the intuition about the case.
To be sure, Christensen might reply that ideally rational agents would only operate with
degrees of belief rather than outright beliefs, and that requirements governing outright beliefs
would not gain purchase on such agents: all that matters for them would be how one’s degree of
belief in the conjunction of the body claims fares vis-à-vis Lockean Thesis. I agree with this. As
we have seen, however, a wholesale rejection of the concept of outright belief and the rational
constraints associated with it has radically revisionary consequences about our ordinary
discourse and argumentative practice. So long as we find it desirable to preserve the rational
constraints underlying our ordinary argumentative practices, we are justified in seeking a
solution to the paradox that does better on this front.

3. Interlude: Size, Rational Capacities, and Demandingness
One lesson from the previous section is that the paradox depends crucially on the fact that the
number of the body claims is sufficiently large. On the one hand, Eric is rational despite his
violation of Consistency and Conjunctive Closure in the original case, where there are large
number of claims. This led us to to call the requirements into question. On the other hand, if

134
Eric’s book contained only, say, three simple claims (P, Q, and R), it would seem clear that Eric
could not rationally believe, at the same time, that there must be some error in his book in this
case. If he simply claimed, for example, ‘P, Q, R, but at least one of them must be false’ or ‘P, Q,
R. But I don’t believe P-and-Q-and-R’, we would have a case in which Eric does seem irrational,
rather than a paradoxical case in which Eric doesn’t seem irrational at all despite his violation of
the putative requirements. In such a case, the “solution” would be to simply view Eric as
irrational, and there would be no pressure to deny either Consistency or Conjunctive Closure.
So a satisfactory solution to the paradox should explain why the rationality of the
author’s believing the preface claim together with the body claims depends on the size of the
book: the author seems rational only when the number of the body claims is sufficiently large
and seems irrational otherwise. The arguments from the previous section shows that the standard
solution fails to meet this desideratum in a satisfying way: no matter how we set the Lockean
threshold that determines the rationality of belief, it is bound to make wrong predictions about
the rationality of the author, in discarding the above constraints altogether.
Here is my alternative proposal as to what the right solution to the paradox should look
like: it is the cognitive load involved in holding all of the claims in the author’s mind that
explains why our intuitions about the rationality of the author are sensitive to the size of the
book: it is because consciously considering or surveying each claim in an ordinary-length history
book is too demanding for ordinary, cognitively limited agents like us that Eric is not irrational
in believing the preface claim, and not because of the link between rational credence and rational
belief that Lockean Thesis posits.
The idea that rational requirements should not be too demanding for ordinary agents like
us is to be understood in terms of our rational capacities: a requirement would be too demanding

135
for us just in case we could not be legitimately expected, given our limited cognitive capacities
and skills, to comply with it. In the literature, this idea has been utilized in properly formulating
rational requirements.7 Broome, as we have seen, argues that Consistency should be given up in
favor of a weaker consistency requirement, on the grounds that it would be too demanding to
require that all of one’s beliefs be logically consistent (Broome 2013: 154-155). Harman also
(1986: 12) argues against the idea that one is rationally required to believe the logical
consequences of what one believes, on the grounds that it would be pointless to clutter one’s
mind with trivialities, given the limited amount of information one can store in one’s mind.
My suggestion is that we appropriately restrict the conditions under which we are subject
to rational constraints like Consistency and Conjunctive closure in a way that is sensitive to our
limited capacities, without of abandoning the constraints altogether. By doing so, we can better
explain why the size of the book matters: in the preface case, the size of the book makes it too
demanding for us to comply with the constraints so that they fail to apply to us. But such
constraints do apply to us when the size of the book is sufficiently small so that we could be
reasonably expected to comply with them.

4. Towards the Solution: Occurrent and Dispositional Beliefs
This section outlines the alternative solution, which preserves the legitimacy of Conjunctive
Closure and Consistency while explaining the rationality of Eric’s belief in terms of the cognitive
load involved in satisfying those requirements. The paradox begins with the observation that an
author like Eric (rationally) believes all of the claims in the body of their book. But in what sense
does Eric believe them? The question is important because there is a sense in which Eric doesn’t
believe all of them. If the thought were that Eric currently has every claim before his

136
consciousness and accepts each, then it would be clearly false. Rather, Eric believes all the
claims, in the sense that he has a standing attitude towards each claim in his cognitive system,
which need not be activated at the moment the beliefs are truly ascribed to him. We believe a
vast number of things in this sense, say, while we are asleep or unconscious.
The distinction between the two senses in which Eric might believe the body claims is an
instance of the distinction between occurrent and dispositional beliefs, which will (again) play a
central role in this chapter.
111
I shall work with the distinction drawn in Chapter 2. Recall that
there were two ways in which can understand the distinction, about which we can remain neutral
for the sake of argument. On the first, you occurrently believe p just in case you consciously
judge p, that is, accept p as true. Dispositionally believing p per se is a matter of having p as an
available item in your representational system, although you can be conscious of p by recalling it
on relevant occasions.
112
The second way of drawing the distinction takes the basic contrast to be
between what is activated and unactivated (or dormant), which doesn’t coincide with the
conscious/non-conscious distinction: an attitude can be activated in the sense of playing an
explanatory role in one’s thought and action, without being conscious. On this distinction, if a
dispositional belief is activated or manifested for a time period, influencing your thought and
action in ways characteristic of belief, it is occurrent.
Let’s now see how the distinction applies to the paradox. Eric’s beliefs in the body claims
are plausibly regarded only as dispositional beliefs rather than occurrent beliefs. It pre-
theoretically seems impossible that any ordinary human being can have all of the claims in an
ordinary-size history book activated in one’s consciousness, reasoning, or other actions. This

111
See Schwitzgebel (2010) for an overview of the distinction. In fact, we can draw the same distinction about other
attitudes. See Mele (2003) for a distinction between occurrent and dispositional desires and intentions.
112
See Frankish (2004); Price (1969); Rose and Schaffer (2013); Shah and Velleman (2005).

137
pre-theoretic judgment is empirically bolstered if, as I’ve suggested in Chapter 2, we make a
plausible conjecture about the way the conceptual distinction between dispositional and
occurrent belief maps onto a well-established psychological distinction. Consider the contrast
between working memory, “a central domain-general resource that enables representations, to be
actively sustained, rehearsed, and manipulated for purposes of reasoning and problem solving”,
and long-term memory, which consists of stored representations, which can be activated and
affect reasoning and other activities only by being called onto the workspace of working memory
(Carruthers 2015: 12).
113
It is easy to see how this general characterization provides an empirical
basis for the distinction between occurrent and dispositional beliefs. Since occurrent beliefs were
defined to be beliefs operative in one’s thought and action, we can plausibly conjecture that they
are items in working memory, whereas dispositional beliefs are items in long-term memory.
It is uncontested that working memory is highly limited in its capacity, whereas long-
term memory suffers from no such limitation. By any reasonable measure, it is impossible for all
of the claims in Eric’s book to be activated in his working memory at a given time, assuming that
Eric is a normal human being. It follows that Eric cannot occurrently believe all the claims in his
book. Eric ipso facto cannot occurrently believe the conjunction of those claims.
With respect to particular body claims, Eric can only occurrently believe some proper
subset of the set of body claims, the number of which depends on his working memory capacity.
Take an arbitrary set S of m claims (where m < n) which consists of the members of the set {P1,
P2, ..., Pn}. Suppose Eric is rational in occurrently believing every member in S.

113
See also Cherniak (1983).

138
With respect to the preface claim, Eric can only occurrently believe that at least some
claim in his book is false. Recall that when we set up the paradox in Section 1, we considered the
following, two alternative ways to represent Eric’s preface belief:

(2) Bel [~(P1&P2&…&Pn)]
(3) Bel [At least some claim I make in this book is false]

In section 1, we didn’t yet have a principled story about why Eric’s cognitive limitations should
matter for the paradox, and about why we should represent Eric’s beliefs as (3) instead of (2).
But now we have seen how such limitations could matter for his rationality. Now that we are
considering only Eric’s occurrent beliefs, and he cannot occurrently entertain the conjunction of
all of the body claims while he can entertain the content of (3), his belief in the preface claim
should be represented as (3).
Once we think of Eric’s occurrent beliefs in the body claims and the preface claim in this
way, it is clear that the combination of the preface claim and the body claims doesn’t give rise to
any paradox, since no deductive constraints are threatened: the preface claim that there must be
some error in his book is perfectly consistent with his occurrent beliefs in the members of S,
which is a small, proper subset of {P1, P2, ..., Pn}. Moreover, occurrently believing the
conjunction of every member in S together with occurrently believing the preface claim doesn’t
commit him to believing a contradiction. So neither consistency nor conjunctive closure would
be under threat.
Given the structure of the paradox, we cannot single out particular members from {P1,
P2, …, Pn} which Eric rationally ought not believe. For we are supposing that there isn’t any

139
particular body claim that his evidence for the preface claim undermines. For any S, therefore,
Eric could be rational in occurrently believing its members and occurrently believing the preface
claim at the same time, without violating any of the deductive constraints at issue.
Before moving on, let me address Christensen’s (2004) objection, which he thinks could
apply to any solution to the paradox invoking the cognitive limitations of ordinary human beings.
He thinks that even if Eric can’t entertain all of the body claims, Eric is still able to gradually
reach the belief in their conjunction. For example, Eric can proceed as follows: he entertains the
first two claims in the book, marks them off as true, concludes that the first two claims are true.
He then turns his attention to the next two claims, again marks off them as true and concludes, by
conjoining it with first result, that the first four claims in the book are true, and so on. He can
repeat the procedure until he concludes that every (body) claim in the book is true.
However, the occurrent belief Eric can reach as a result of this procedure, is the belief in
the proposition that every claim I make in this book is true, which is different from occurrently
believing the conjunction of all the body claims in the book. And if this were what Eric believes
together with the proposition that at least some of the claims in this book is false, this would
simply be a case of holding occurrently contradictory beliefs, which is a clear case of
irrationality. In this case, the fault is on Eric and not the deductive constraints. But the paradox
arose precisely because Eric’s believing the body claims and the preface claim at the same time
didn’t seem irrational, which is the reason why we were inclined to doubt the legitimacy of the
deductive constraints. So the belief that results from the above procedure cannot preserve what
makes Eric’s case paradoxical in the first place: it turns a paradox into a case of manifest
irrationality.

140
5. The Solution: Deductive Constraints Govern Occurrent Beliefs
The paradox doesn’t arise if only occurrent beliefs are, and dispositional beliefs aren’t, subject to
such deductive constraints as Consistency and Conjunctive Closure: if we assume that
consistency and conjunctive closure govern only occurrent beliefs, dispositional beliefs don’t fall
under their jurisdictions, and so there is no violation of rational requirements. The point is not
that Eric’s dispositional beliefs would be rational despite their inconsistency, but rather that the
inconsistency among dispositional beliefs wouldn’t constitute irrationality, because dispositional
beliefs aren’t subject to such requirements. This suggests that we concretize the relevant
constraints in the following way:

No Occurrent Inconsistent Beliefs: Rationality requires of you that your occurrent
beliefs be logically consistent.

Occurrent Conjunctive Closure: Rationality requires of you that you believe p-and-q if
you occurrently believe p, occurrently believe q, and attend to both p and q.

So why should we accept the idea that deductive constraints on belief govern only occurrent
beliefs? My argument in support of this thesis runs as follows: if deductive constraints were
genuine rational constraints on dispositional beliefs, one would be irrational in violating them.
But one is not irrational in holding dispositional beliefs running afoul of such deductive
constraints. So deductive constraints don’t impose rational demands on dispositional beliefs.
To see why one wouldn’t be irrational in holding inconsistent beliefs, return to the
following propositions from Section 2:

141

P: There will be dance at the end-of-semester party tomorrow.
Q: It will rain all day tomorrow.
R: Jon doesn’t come to any dance party on a rainy day.
S: Jon will not come to the end-of-semester party.

Suppose you learned R a few years ago. But suppose that Jon, his quirky disposition
notwithstanding, is also a renowned partygoer who otherwise loves dancing and you also know
it. Suppose you learn P two weeks prior to the party, let your friend Jon know about it, and
immediately come to believe not-S, i.e., that Jon will come to the party. The day before the party,
however, you learn Q from a highly reliable weather forecast. At that time, however, you are at
pains to finish your work so that you can enjoy the party without feeling guilty. Accordingly, you
never give a thought about whether Jon will come to the party tomorrow. You end up with
dispositionally believing P, Q, R and not-S, which are inconsistent.
Intuitively, however, you don’t seem irrational at any point in this version of the story. At
each moment of belief formation, at which you form a dispositional belief in a particular
proposition, you need not be violating any requirement of rationality (we are assuming that your
belief is responsive to epistemic reasons). But if each formation of belief has been rationally
innocuous, then it is difficult to see why you should count as irrational upon dispositionally
believing P, Q, R, and not-S. We might put the same point by saying that when you form each
belief, you satisfy every requirement which applies to you then. But if dispositionally believing
P, Q, R, and not-S, is simply a result of satisfying every requirement which applies you at the

142
time you form each belief, then it is difficult to see how you could be irrational in ending up
dispositionally believing each.
One might object that you haven’t really satisfied every requirement that applies to you
simply because you have failed to maintain a consistent set of beliefs, although having satisfied
the requirement to align your belief to available evidence. But what would this requirement be,
which demands maintaining consistency between your beliefs? It cannot simply be Consistency,
because its status as a rational requirement on belief is precisely what is called into question by
cases like the preface paradox. Alternatively, it might be the requirement that you are rationally
required, whenever you form a belief, to check how it fits with the rest of your beliefs, and to
revise some of your beliefs if you find out that they cannot all be true. But such a requirement,
just like Consistency, seems too demanding to be a plausible requirement of rationality.
More plausibly, one might reply that you violate an epistemic duty in being forgetful, or
inattentive, in not recalling relevant beliefs upon forming the belief that it will rain tomorrow,
e.g. the belief that Jon will come to the party. But this response also fails. While you might well
be rationally required to recall or attend to propositions relevant to your situation, your situation
does not seem to be the one in which you are rationally required to attend to the question whether
Jon will come to the party. To see this, suppose, at the moment you hear the weather forecast,
you care about finishing the work much more than whether Jon will come to the party, and you
do have more practical reason to finish the work than to find out whether Jon will come. It seems
clear that you are not irrational in continuing to work without thinking about whether Jon will
come or recalling your belief.
This gives us a reason to conclude that you are not irrational in the above scenario. If it
were rationally required of you that your dispositional beliefs be logically consistent, however,

143
you would count as irrational. For it is plausible that you, in this case, dispositionally believe P,
Q, R and not-S individually: for example, you are disposed to affirm each proposition, when
asked (individually) whether each is true.
The general point concerns the precise point at which you can be legitimately accused of
irrationality. Merely having a set of dispositional beliefs that fall afoul of deductive constraints
doesn’t entail that you are irrational: you may not have put those beliefs together, or may not
have had the opportunity to consider all of them together, because some of them are simply out
of your occurrent reach. Although you form each belief at intervals, the beliefs are never
activated at the same time, and the mere fact that they are inconsistent doesn’t in itself seem to
impugn your rationality. Of course, you would clearly be irrational, if you reconsidered whether
Jon will come to the party and considered all the propositions at the same time, and yet did not
give up any of the beliefs. But this is simply because your beliefs are all occurrent in this case,
and my view correctly predicts that you would be irrational.
The same goes for conjunctive closure. Begin again with R. Suppose you found this fact
amusing when you first learned it (a few years ago) but it no longer occupies your attention in
any way. That said, it doesn’t seem to be any kind of rational defect if you didn’t form, upon
recently learning P, the belief in their conjunction P-and-R. By contrast, not believing the
conjunction would count as irrational when each belief is occurrent: if you judge P and judge R
at a given time, for example, it seems that you rationally should also judge P-and-R, so long as
you are attending to whether P-and-R. Again, this is because conjunctive closure governs
occurrent beliefs. To sum up, if deductive constraints are to serve as rational demands on beliefs,
they should be seen as targeting only occurrent beliefs. This, as we have seen, is the thesis that
affords us a neat way of solving the preface paradox.

144

6. Objections and Replies
In this section, I address some objections to my proposal. Even if one agrees with the intuition
that there is nothing rationally amiss with your beliefs in the case discussed in section 4, one
might still disagree with my explanation as to why you are not irrational. That is, one might take
issue with the idea that the reason why you are not irrational is that Consistency and Conjunctive
Closure govern occurrent beliefs. I shall deal with two competing explanations. According to one
view, rational assessment fundamentally concerns processes like reasoning rather than states.
114

On this view, the reason why there is nothing irrational about you is that you don’t violate any
rational norms on processes in forming beliefs in P, Q, R, not-S, taken individually: we are
assuming that you rationally form each belief, respecting evidence (and other reasons).
This explanation overreaches. The reason why you aren’t irrational, on this explanation,
is that consistency isn’t a genuine rational constraint on belief at all, since consistency
requirement on belief is a synchronic requirement on states, not a requirement on processes.
Thus, although this view might explain why you aren’t irrational, it doesn’t explain why you
would be irrational if your beliefs were all occurrent. Moreover, it implies that even if someone
occurrently held contradictory beliefs they wouldn’t be irrational. True, some flaw could be
found in the process that led to the contradictory beliefs in some cases, but there is no guarantee
that holding contradictory occurrent beliefs must be irrational.
Defenders of process-norms happily admit that even someone who judges or sincerely
asserts a proposition and its negation at the same time (assuming its possibility) need not be
irrational. But this seems a hard to bullet to bite. Alternatively, one could further attempt to

114
See Podgorski (2017) for a thorough development of this view, although Kolodny (2007b) surely is an important
precursor.

145
establish that whenever one holds contradictory beliefs, there must be some rational flaw in the
process through which one formed those beliefs. While this is an open possibility, it seems
unlikely to succeed. Inconsistent occurrent beliefs could come about through all sorts of
processes, such as accidentally getting an electric shock, or getting brain damage, which seem
simply arational. In such cases, it would still be irrational to hold the beliefs occurrently, even if
you never made an error in reasoning.
Another explanation might invoke the consideration of clutter avoidance. Harman argues
against the idea that one is rationally required to believe the logical consequences of what one
believes, on the grounds that it would implausibly require one to clutter one’s mind with
trivialities.
115
John Broome, in response to this problem, suggests the following as a replacement
for the closure principle:

Modus Ponens: Rationality requires of you that, if you believe that p, and you believe
that if p then q, and if you care whether q, then you believe that q.
116

By ‘caring’, Broome means to include not just having practical interests but also theoretical
interests in a proposition, such as wondering or being curious about its truth. In light of this, one
might think the relevant requirement of conjunctive closure should formulated in such a way as
to incorporate an analogous clause on caring:

Caring Conjunctive Closure: Rationality requires of you that if you believe p and if you
believe q, and if you care whether p-and-q, then you believe p-and-q.

115
Harman (1986: 12).
116
Broome (2013: 157)

146

Supposing that you aren’t caring about whether P-and-R is true at the moment you are working,
Caring Conjunctive Closure can explain why you aren’t irrational for not conjoining P and R:
you are not irrational, because you don’t care about whether there will be dance at the party
tomorrow and Jon doesn’t come to the party.
‘Caring’, like ‘belief’, also admits of both dispositional and occurrent readings: we care
about many things that aren’t currently occupying our attention. If we understand ‘care’ in
Caring Conjunctive Closure as a dispositional state, however, the principle is too demanding.
Let’s add to the above scenario that you love seeing Jane in company with Jon at parties, and in
this sense care about whether both Jon and Jane will come to the party. Suppose you also let Jane
know about the party two weeks prior to the party, but she had to see if she could come until the
day before the party. Suppose Jane finally calls and tells you that she will come, while you are
concentrating on your work. You form the belief that Jane will come. At that time, however, you
are too busy to think about whether both Jon and Jane will come, and so fail to conjoin the belief
that Jon will come and the belief that Jane will come. Even then you dispositionally care about
whether both will come: you would, if you were free from work, deeply wonder whether they
will. So you violate the dispositional version of Caring Conjunctive Closure. Again, however,
there seems nothing clearly irrational about the beliefs you have.
One might therefore adopt the occurrent reading of ‘care’. But it would be ad hoc to
maintain that only caring should be given the occurrent reading, if one didn’t hold the same view
about belief, too. But if both caring and belief in Caring Conjunctive Closure were given
occurrent readings, the principle would be a near-equivalent of Occurrent Conjunctive Closure.
Even then, Occurrent Conjunctive Closure would be more plausible. Caring Conjunctive Closure

147
would implausibly imply that even if you occurrently believed p and occurrently believed q, and
you were attending to both p and q, you could rationally refuse to conjoin them if you simply
didn’t care about them. But it seems you cannot be rationally off the hook so easily. Suppose you
recalled P, Q, R, not-S, and you confidently assert each. If someone points out to you that P, Q,
R entails S, the rational response seems to admit your irrationality and stop occurrently believing
at least one of the propositions. The requirement at issue would entitle you to hold onto your
beliefs, so long as you didn’t have any relevant question you care about, which is implausible.
By contrast, No Occurrent Inconsistent Beliefs doesn’t imply that you could be off the
hook so easily: it correctly predicts that you would be irrational were you to occurrently believe
P, Q, R and not-S. Nor does it suffer from the problem of fitting the consideration about caring
into a consistency requirement. Moreover, it captures, in a sense, Harman and Broome’s basic
insight. There is a role your practical or theoretical interests play in the background: if there is a
question you are interested in, your interests typically cause your relevant beliefs to be occurrent,
say, by being retrieved into your working memory. The proposal is to recognize rationality as
governing all and only those beliefs, in order not to make the relevant requirements implausibly
demanding. It is precisely considerations of this sort that motivates Harman’s insistence on
clutter avoidance and Broome’s caring clause, which No Occurrent Inconsistent Beliefs capture
without confronting the difficult problems of incorporating conditions about interest/caring into
the relevant requirements themselves.

148
Chapter 6: A Solution to the Problem of Misleading Higher-Order Evidence

A burgeoning literature in epistemology examines the question of how to respond to a case of
misleading higher-order evidence, a case in which you have compelling evidence that a
proposition is true, but also have compelling (but misleading) evidence that your evidence
doesn’t sufficiently support its truth. Such a case is often seen to reveal a fundamental tension
between two important requirements on belief: the requirement to have beliefs supported by your
evidence, and the requirement to avoid incoherence between first-order and higher-order beliefs.
It is widely agreed that such a case would pose a dilemma, because you would not be able
to satisfy the two requirements jointly.
117
The aim of this final chapter is to shake this cross-party
consensus, relying on one of the central themes of my dissertation: coherence requirements
govern occurrent attitudes rather than dispositional attitudes. Once we recognize the requirement
of coherence as governing occurrent doxastic attitudes, we can carve up a conceptual space for
the joint satisfaction of the requirements in cases of misleading higher-order evidence. Suppose,
for example, your total evidence requires you to believe that your evidence doesn’t support
believing P, but also requires you to believe P. You can satisfy the requirement to believe in
accordance with your evidence, by dispositionally believing both. But this doesn’t lead to an
incoherent combination of beliefs. If the incoherence in question consists only in occurrently
believing both at the same time, and you can refrain from occurrently believing both while
dispositionally believing both, you can avoid being incoherent, by not occurrently believing
both.

117
Christensen (2007, 2010) takes it to create a rational dilemma, where one is bound to fall short of some rational
ideals no matter how one responds. Hazlett (2012) concurs. Worsnip (2015) argues that it illustrates an irresolvable
conflict between two norms that have different sources. Coates (2012), Lasonen-Aarnio (2014, forthcoming) and
Wedgwood (2012), suggest that we reject the requirement of inter-level coherence.

149
We could characterize such a state as being cautious: you have a dispositional belief, a
standing mental state towards a proposition P. Due to some relevant features of your situation,
however, you nonetheless refrain from using the belief in your thought and action. For example,
you refrain from judging P as true, asserting P, using P as a premise in your reasoning, etc. I
argue that this concept of cautiousness affords us a solution to the puzzle of misleading higher-
order evidence that is superior to the existing solutions.

1. Preliminaries: Misleading Higher-Order Evidence and Conflicting Norms on Belief
This section introduces the puzzle by characterizing the phenomenon of misleading higher-order
evidence, and precisely formulating the two norms alleged to make conflicting demands in a case
of misleading higher-order evidence.

Misleading Higher-Order Evidence (MHOE): Your total evidence requires you to
believe a proposition p, and also requires you to believe that your evidence does not
sufficiently support believing p.

Here is a case of MHOE. Suppose you are at a crime scene. You obtain sufficient evidence that
Smith committed the murder: the weapon is the one Smith was often carrying with him; it is
highly likely, given what the witnesses say, that Smith was at the scene at the estimated time of
murder, etc. So you come to believe that Smith is the murderer. Then Holmes comes along and
asks, ‘who do you think is the murderer?’ You answer, confidently, ‘It’s Smith’, and tell him
about the evidence you have obtained so far. But Holmes responds, shaking his head, ‘Well,
maybe Smith is, maybe he isn’t, but you haven’t evaluated your evidence correctly—your

150
evidence doesn’t establish that he is’. Then Holmes hurriedly leaves the scene. Holmes is an
extremely reliable, world-renowned detective, whom you have every reason to trust.
Unbeknownst to you, however, Holmes has made a mistake on this particular occasion: your
evidence does decisively support the proposition that Smith is the murderer. This is a case of
misleading higher-order evidence: you have sufficient evidence that Smith is the murderer, but
you also have excellent higher-order (albeit misleading) evidence about the quality of your
evidence: the evidence that you don’t have sufficient evidence that Smith is the murderer.
118

Higher-order evidence can be defined as “evidence about the existence, merits, or
significance of a body of evidence.” (Feldman 2009: 304). While the higher-order evidence in
this case consists in testimony from your epistemic superior, there is a variety of higher-order
evidence discussed in the literature. First, there is higher-order evidence that comes from
skeptical arguments about the external world, as Feldman (2005) illustrates: suppose a student,
on their way to their first epistemology class, sees an oak tree on the quad, and forms the belief
that there is an oak tree on the quad. Suppose, however, that their professor presents a skeptical
argument in such a convincing way that they realize their inability to rule out the possibility that
their perceptual experiences aren’t good enough evidence that external objects exist. In such a
case, the student seems to have good evidence that their perceptual evidence isn’t sufficient
establish that there is an oak tree in the quad.
119

Peer disagreement is also treated as an important kind of higher-order evidence. If you
take up a doxastic attitude towards a proposition P by evaluating a body of evidence, but learn
that your epistemic peer has carefully considered the same (or comparable) body of evidence and

118
This example is based on Coates (2012: 114). Horowitz (2014) and Worsnip (2018a) focus on similar example in
the context of discussing misleading higher-order evidence.
119
See also Greco (2014).

151
has taken up a different doxastic attitude, this gives you some reason to think that you have
misevaluated the evidence.
120
There are also cases involving drugs: Suppose you solve a logic
puzzle through competent deductive reasoning, and form a belief in the conclusion, but later
learn that you have taken a special drug, which has 80% chance of significantly impairing a
person’s logic-puzzle reasoning abilities, without causing them to feel anything abnormal or to
notice any difference. This seems to be higher-order evidence that your competent reasoning, on
this particular occasion, is not good evidence that your conclusion is correct.
121

What should you believe after learning about Holmes’ assessment? The following
principle entails that you ought to believe both that Smith is the murderer and that you don’t
have sufficient evidence that he is the murderer:

Evidence Requirement (ER): You ought to believe p if your total evidence requires you
to believe p, and ought not believe p if your total evidence requires you not to believe p.

‘Evidence’ is used here in the most neutral possible way, so that ER is consistent with both
internalism and externalism about epistemic rationality (or justification). Any plausible theory of
epistemic rationality (or justification), internalist or externalist, would spell out the conditions
under which you ought, or ought not, to believe a proposition, in virtue of the evidence you
have.
122
This is all ER is intended to capture.

120
See Kelly (2005), Feldman (2009), Christensen (2010).
121
Christensen (2010) deals with such cases at length. See also Wedgwood (2012) and Schechter (2013).
122
For some prominent (and competing) accounts of what it is to have reasons, of which having evidence is an
instance, see Schroeder (2008, 2011) and Lord (2010). All parties agree that the reasons you have are propositions to
which you stand in some privileged relation, such as belief, which is all that is needed for the purpose of this
chapter.

152
ER is plausibly a narrow-scope requirement: ‘ought’ in ER takes narrow scope over the
consequent of the conditional, and the consequent detaches if the antecedent holds. This is
because ER expresses the relation of reasons: evidence is an important (if not the only) class of
epistemic reason, and ER can be seen as a special case of a more general requirement to respond
to the epistemic reasons you have.
123
To say that you have reason to have an attitude is to say
that there is some consideration that supports or justifies your having that particular attitude, and
you can respond to your reason only by taking up that attitude. Thus, with respect to a case of
MHOE, ER implies that you ought to have a (first-order) belief that your evidence requires you
to have, and also that you ought to believe that that attitude is not sufficiently supported by your
evidence.
If you followed ER in the case of Holmes, therefore, you would believe that Smith is the
murderer while at the same time believing that your evidence doesn’t sufficiently support having
the belief. But it has been widely accepted that having such a combination of beliefs, if
possible,
124
would be deeply incoherent, and would count as an epistemic analogue of
“akrasia”.
125
That is, it is argued that you are bound to violate the following requirement of
coherence in believing both:

123
If evidentialism is true, that is, if all epistemic reasons are evidence, the broader requirement is equivalent to ER.
See Kelly (2003) and Shah (2006) for defenses of evidentialism. But if there are non-evidential epistemic reasons,
they are not equivalent, and the proper requirement should rather say: you ought to have D towards p if your
evidence, together with all of the relevant non-evidential reasons, requires you to have D. I shall hold onto ER,
however, simply for the continuity with the literature.
124
Hurley (1993), Pettit and Smith (1996), and Adler (2002) argue that epistemic akrasia is impossible. While I
agree that epistemic akrasia is impossible, I shall set aside such a view for the purpose of this chapter.
125
For the view that it would be always irrational to have such a combination of beliefs, see Feldman (2005, 2009),
Smithies (2012), Horowitz (2014), Greco (2014), Worsnip (2018a).

153
Inter-Level Belief Coherence (ILBC): If you believe that your evidence does not
sufficiently support believing p, you ought not believe p.
126

ILBC is most plausibly seen as a principle that prohibits certain combinations of higher-order
beliefs and first-order beliefs: regardless of the reasons you have for or against the particular
belief, you would exhibit incoherence if you had a combination of beliefs ILBC prohibits. Unlike
in the case of ER, whether ILBC is a narrow-scope requirement to lack or take the particular
attitude in the consequent is at best controversial, because the narrow-scope reading faces the
familiar bootstrapping objection: when your higher-order belief is irrational or insane, for
example, it does not seem plausible that you ought to revise your belief in accordance with it, as
the narrow-scope reading of ‘ought’ would imply. I shall remain neutral on the issue of scope,
however, since what matters for the purpose of this chapter is that having the combinations of
attitudes ILBC prohibits would be irrational, which is consistent with either construal.
ILBC refers to higher-order belief
127
about how good or strong your evidence is. One
might wonder why we could not instead take the relevant kind of belief to be a belief about
reasons, justification, or ought. This is because of the different concepts that could be expressed

126
The label ‘Inter-Level Coherence’ is due to Worsnip (2018a). ILBC is a special case of the more general
requirement Worsnip formulates, which is supposed to govern doxastic attitudes of all kinds including belief:
“Rationality requires of S that: (i) S believes that her evidence supports D(p) →she takes D(p); (ii) S believes that her
evidence does not support D(p) →she does not take D(p).” ILBC, which limits its scope to belief, focuses only on the
second clause of Worsnip’s requirement. Since the existing literature almost exclusively focuses on cases in which
the second clause is violated, and since discussing all doxastic attitudes will unnecessarily complicate the matter, I
shall restrict my attention to ILBC.
127
Some think that belief is not the only relevant higher-order attitude. They think that suspending judgment whether
one has sufficient evidence or justification for the first-order attitude is rationally incompatible with having that
first-order attitude. See, for example, Bergmann (2005), Feldman (2005, 2009), Huemer (2011), and Smithies
(2012). Hazlett (2012) objects, however: while it seems that there is something problematic about believing that you
don’t have sufficient evidence or justification as to p and yet believing p, suspending judgment about whether one’s
belief in p is justified is rationally compatible with believing p, and indeed can be a manifestation of the epistemic
virtue he calls “intellectual humility”. So I shall deliberately focus on belief, which is incontrovertibly taken to be an
attitude that could give rise to the kind of irrationality ILBC is supposed to forbid.

154
by ‘ought’, ‘justified’, or ‘reason’ in different contexts. Suppose that someone, while thinking
that there is no evidence as to the existence of God, is nonetheless persuaded by the Pascalian
argument for believing that God exists. While one could ascribe them the belief that they ought,
in some sense, to believe that God exists, they would not be seem incoherent or epistemically
irrational, in failing to believe that God exists.
128
Rather, it is when they believe that God exists,
while also believing they have no evidence that God exists that they would count as
epistemically irrational. The same goes for ‘justification’ and ‘reasons’.
129

So here is the structure of the apparent dilemma posed by misleading higher-order
evidence. Suppose you are in a situation that falls under MHOE. It follows from ER that you
ought to believe P and also that you ought to believe that believing P lacks sufficient evidential
support, since (by hypothesis) your total evidence requires you to have believe P and it also
requires you to believe that your evidence does not sufficiently support your believing P. But if
you satisfy ER by believing P and also believing that your evidence doesn’t sufficiently support
believing P at the same time, you are bound to violate ILBC.
The existing solutions to the puzzle rule out one of the elements that together generate the
puzzle. That is, they deny either (i) ER; (ii) ILBC; or (iii) the possibility of MHOE.
130
Each
option, however, incurs significant theoretical burdens. In what follows, I shall offer my own
solution to the puzzle, which doesn’t require giving up any of the element, and show how it fares
better than other solutions in avoiding controversial theoretical commitments.

128
See Broome (2013: 94-5).
129
To be sure, non-evidentialists can work with the concept of epistemic reasons, or epistemic justification, rather
than evidence, in formulating ILBC (as noted in fn. 12), and I have no objection to that. Again, however, I shall use
the concept of evidence, simply for the sake of convenience.
130
I am aware of no solution that rejects ER entirely, although Worsnip (2018a) does deny that it is a requirement of
rationality, the fundamental subject matter of which, on his view, is coherence between attitudes. Coates (2012),
Lasonen-Aarnio (2014) and Wedgwood (2012) deny ILBC (as noted in fn. 1). Bergmann (2005), Feldman (2005),
and Titelbaum (2015) can be read as denying the possibility of misleading higher-order evidence. See also
Wedgwood’s (2012) remarks on the iterative failure of epistemic justification.

155

2. The Solution: Being Cautious
Here is the solution: in a case of MHOE, you should be cautious. To get the intuitive idea,
suppose you respond in the following way in the case of Holmes (section 1), after learning about
Holmes’ assessment: ‘Hmm… Smith definitely seems to be the murderer, as far as I can tell. But
I also have excellent evidence that I have made some mistake in evaluating my evidence—how
could I doubt what Holmes says? He’s been always right. So I’d better be careful here. It would
be rash for me now to treat it as a fact that Smith is the murderer.’
In such a case, it seems completely fine for you to self-ascribe the belief that Smith is the
murderer. It is not as if you have no attitude whatsoever towards the proposition that Smith is the
murderer. After all, you had concluded, on the basis of your solid evidence, that he is the
murderer, and you haven’t been presented with any evidence indicating his innocence. And
while you trust Holmes, his testimony doesn’t indicate that Smith is innocent, because what he
points out is just that I have made error in evaluating the evidence, which is consistent with
Smith’s being the murderer. That is, you have a standing belief towards the proposition. It is
rather that you aren’t using the belief in your further reasoning, due to some salient possibility of
error you might have made in the situation. In such a case, there is an intuitive sense in which
you are being cautious.
Having introduced the intuitive idea of cautiousness, let’s turn to my gloss of it. My
proposal is that being cautious with respect to P consists in refraining from occurrently believing
P (or from using the belief in P) while at the same time dispositionally believing P. To see
clearly what this proposal amounts to, recall the distinction between dispositional and occurrent
belief I have been employing in the dissertation: dispositional beliefs are items stored in your

156
representational system that normally endure over time. They, under appropriate triggering
conditions, manifest themselves in one’s thought or action in ways characteristic of beliefs. By
contrast, occurrent beliefs are those activated in your thought or action by either being conscious
or being non-consciously operative in one’s thought or action. For example, consciously
affirming P as true (i.e. judging P), taking the truth of P as settled in your reasoning, and acting
on the basis of one’s belief, all count as a way of occurrently believing P.
131

This paves the way for satisfying both ER and ILBC in a case of MHOE. Suppose your
total evidence requires you to believe that your evidence doesn’t support believing P, but also
requires you to believe P. In such a case of MHOE, you ought, given ER, to believe each.
Suppose ER governs dispositional beliefs. Then you can satisfy ER by dispositionally believing
that your evidence doesn’t sufficiently support believing P, and by also dispositionally believing
P. If you were cautious with respect to either P or the content of your higher-order belief,
however, you wouldn’t occurrently believe both at the same time. If we suppose that ILBC,
unlike ER, governs only occurrent attitudes, then you can satisfy ILBC by lacking the pair of
incoherent occurrent beliefs forbidden by ILBC. If so, we are not forced to give up any of the
requirements even if we grant the possibility of MHOE.
To apply this idea to a concrete scenario, consider again the case of Holmes. You acquire,
upon reflecting on your evidence, the belief that Smith is the murderer, which becomes a stored,
dispositional belief. This explains why you recall, and confidently tell Holmes, that Smith is the
murderer. Through Holmes’ testimony, however, you also acquire the belief about the

131
In the previous chapter, I have also suggested that this conceptual distinction could plausibly be mapped onto the
distinction between items in one’s long-term memory and working memory, where the former consists of stored
representations available to be used in one’s activities like reasoning, and the latter consists in “a general domain-
general resource that enables representations to be actively sustained, rehearsed and manipulated for purposes of
reasoning and problem solving.” (Carruthers 2015: 12)

157
insufficient evidential support for your belief that Smith is the murderer. Since we are assuming
that your evidence requires you to have both beliefs, you thereby satisfy ER. But suppose you
become cautious, and refrain from occurrently believing that Smith is the murderer. For
example, you might not consciously affirm it as true, or take it as settled, that he is the murderer
in your further reasoning (at least at the moment you are consciously accepting Holmes’
testimony). So you satisfy ILBC, on the assumption that ILBC governs occurrent attitudes, since
at no point are your higher- and first-order beliefs both occurrent.
The success of this solution hinges on the two crucial assumptions identified above: (i)
ILBC governs occurrent beliefs; and (ii) ER governs dispositional beliefs. I shall justify each
assumption in what follows.

3. The Appeals of ILBC and Occurrent Attitudes
This section argues that ILBC, as a coherence requirement, governs only occurrent attitudes. I
shall first survey the main reasons cited for accepting ILBC, and point out that they all concern
the irrationality you would exhibit if your occurrent higher-order belief failed to cohere with
your occurrent first-order belief. This means that the appeals of ILBC can be adequately
captured consistently with restricting its target to occurrent doxastic attitudes. Moreover, when
either of the beliefs remains dispositional, you intuitively don’t count as irrational. Thus, ILBC
governs occurrent attitudes.

3.1. A Problematic Conjunction
One prominent argument for accepting ILBC relies on the observation that having the attitudes it
prohibits commits you to believing what is intuitively paradoxical. For example, if you believed

158
that your total evidence doesn’t support believing P, but also believed P, you would be in a
position to conjoin the content of each to believe the following:

P, but my overall evidence doesn’t support believing P.

Despite the logical consistency between the two conjuncts, believing the conjunction would be
deeply problematic. Feldman (2005: 108-9) argues that such a conjunction would be
unknowable, on the following grounds: (i) you know the conjunction only if the conjunction is
true and you are justified in believing the conjunction; (ii) you are justified in believing P only if
P is sufficiently supported by your evidence; (iii) if the second conjunct is true, however, your
evidence doesn’t support P. Therefore, either you lack the justification for P or the second
conjunction is false. In neither case do you know the conjunction.
132
Another explanation of the
oddity of believing such a conjunction is given by Worsnip (2018a), on the basis of some
minimal assumptions about: (i) what it is to believe something; and (ii) the concept of evidence.
To believe P is to take P to be true, and your evidence consists precisely in those considerations
that bear on how likely it is that P is true. So believing such a conjunction would amount to
taking P to be true, while at the same time thinking that P isn’t likely to be true. This seems to be
a kind of incoherence that would not be intelligible without further explanation.

3.2. Problematic Reasoning and Action

132
Huemer (2011) makes a similar point about believing P and disbelieving that one knows P at the same time,
pointing out that it puts one in a position to assert a Moore-paradoxical sentence of the form: ‘p, but I don’t know
that p’. Smithies (2012: 284) similarly argues that there is a conflict between simultaneously believing both
conjuncts of ‘P, but I don’t have justification to believe P’, since “believing a proposition rationally commits one to
believing that one has justification to believe it and vice versa.”

159
Sophie Horowitz (2014: 725-7) offers yet another reason to accept ILBC: if you believed P and
also believed that your evidence doesn’t sufficiently support P, and used both beliefs as premises
in reasoning, you would engage in “patently bad reasoning”. The first kind of reasoning concerns
the way you should think about your first-order belief. One of your premises, recall, is the
proposition that your evidence doesn’t support P. But since you are also using P as a premise in
reasoning, and you know that you believe P, it seems that from these premises you can conclude
that you must have got lucky in correctly believing P! But this seems to be a bad form of
reasoning: given the (higher-order) belief that your belief lacks evidential support, the correct
response to your situation seems to be to stop believing P.
The second kind of reasoning concerns the way you should think about your first-order
evidence. Suppose you solve a logic puzzle through competent deductive reasoning and conclude
that P is the correct answer. Suppose, however, you come to learn that, prior to solving the
puzzle, you have taken a drug that has 90% chance of leading you to a wrong conclusion by
impairing your logical abilities. So you come to believe that the quality of the evidence you have
for P is extremely poor. But if you used both (i) the proposition that your evidence doesn’t
support P and (ii) P, as premises in your reasoning, you would be able to reason in the following
way: ‘P is true, but my evidence supports low confidence in P, which means that it supports high
confidence in not-P. But not-P is false. So my evidence is misleading.’ As Horowitz stresses,
however, this intuitively isn’t a right way to conclude that one’s evidence is misleading.
133

Finally, Horowitz points out that if you were to act upon both beliefs, you would be
exhibiting highly irrational behavior. You would be willing to give fairly strong, say, 9:1 odds
that P is the right answer to the logic puzzle, but at the same time, would be willing to offer only

133
See also Littlejohn (2015) for the unintelligible series of assertions would a violation of ILBC would lead to.

160
1:9 odds that your evidence supports that P is, treating the two bets as entirely independent from
each other. You even might assert, at the very moment you offer such strong odds that P is true,
that your evidence doesn’t support P. But it seems irrational to treat the two bets so separately.

3.3. Satisfying ILBC Through Occurrent Attitudes
There is a common thread in the above arguments for accepting ILBC: they all concern
conditions under which the incoherent combinations of beliefs are both occurrent. To see this,
note first that the worry about the problematic conjunction ‘P, but my total evidence doesn’t
support believing P’ is a worry about having an incoherent combination of occurrent beliefs.
Believing the conjunction on the basis of the beliefs you already have requires forming a belief
in the conjunction by conjoining the contents of those beliefs. But conjoining the contents of
your antecedent beliefs involves using each belief as a premise in your reasoning. That is,
whenever you come to believe a conjunction, on the basis of your antecedent beliefs, your beliefs
must both be occurrent, in the sense of being activated in reasoning. The same goes for the
problematic kinds of reasoning and action that, as Horowitz argues, you would exhibit if you had
an incoherent pair of first-order and higher-order beliefs. In all of the cases Horowitz imagines,
the beliefs that lead to worrisome forms of reasoning or betting behavior are all occurrent, in the
sense that they are activated in your reasoning or action.
It is crucial that both beliefs would be occurrent whenever you would count as irrational.
In section 2, we have already seen a case in which your first-order dispositional belief isn’t
activated: when you become cautious after deferring to Holmes’ testimony, you refrain from
occurrently believing that Smith is the murderer, while having your dispositional belief intact. It

161
doesn’t seem irrational for you to respond in such a way. Rather, it seems to be the right kind of
way to respond to a case of MHOE.
We can also consider a variant of the logical puzzle case in which your higher-order
belief remains dispositional without being activated. Suppose you conclude that P as a result of
competently solving the logic puzzle, and you acquire the dispositional belief that P is the right
answer. But you also learn that you have been drugged when you solved the logic puzzle and
form the belief that your evidence doesn’t support believing P, which also gets stored in your
memory and remain a dispositional belief. After several years, your friend asks you about the
logic puzzle you solved. This triggers the first dispositional belief, and you manage to recall that
P is the answer. Since you are attending only to the correct answer to the puzzle and not to its
evidence, the dispositional belief about insufficient evidential support doesn’t become activated
at all. At this point, you would not form any belief in a problematic conjunction, would not
reason or act in obviously unreasonable ways, since the other, higher-order belief about the lack
of sufficient evidential support remains unactivated. So the problems that led us to posit ILBC
wouldn’t occur.
This doesn’t mean that you lack the dispositional higher-order belief altogether. Were
someone to say, for example, ‘Hey, but I thought you said you were drugged at that time?’, your
belief about the insufficient evidential support for believing P would then be activated. If you
were then to use both higher- and first-order beliefs as premises in your reasoning, you would
indeed engage in manifestly irrational reasoning or behavior.
Again, however, this just means that the problems supposed to arise from having a pair of
beliefs forbidden by ILBC have only to do with the conditions under which the beliefs are both

162
occurrent. The troubling consequences of having an incoherent pair of beliefs can be avoided by
positing the following, a restricted version of ILBC:

Inter-Level Occurrent Belief Coherence: If you occurrently believe that your evidence
does not sufficiently support believing p, you ought not occurrently believe p.

4. Responding to Evidence Through Dispositional Belief
This section justifies the second major assumption made in section 2, i.e. the assumption that ER
targets dispositional beliefs. Here is the rough structure of the argument: (P1) ER, unlike ILBC,
specifies a rule you should follow to gain knowledge, in the sense that complying with ER,
together with other conditions on knowledge, will make it the case that you know p; (P2) If this
is right, it is plausible that the kind of belief you come to possess by complying with ER is the
kind of belief that could constitute knowledge, or the kind of belief entailed by knowledge; and
(P3) knowledge entails dispositional belief. So ER governs dispositional beliefs.
I say that the point of ER has to do with knowledge, because the satisfaction of ER is
partly constitutive of knowledge. Knowledge makes a demand upon the subject, in the sense that
in order for a subject to know, they must do their part. And satisfying ER is plausibly the
demand knowledge makes on the subject. As I pointed out in section 1, ER requires you to
respond to the (epistemic) reasons you have, where the reasons you have consists of propositions
to which you have privileged epistemic access. If you form beliefs by complying with ER,
therefore, you succeed in believing what you have sufficient subjective reason to believe.
Crucially, believing what you have sufficient subjective reason to believe is partly constitutive of
knowledge, since a belief acquired in such a way amounts to knowledge, if other, objective

163
conditions for knowledge are met (such as the belief’s being true, the subjective reasons’ also
being objective reasons, etc.).
134
Complying with ER, to wit, is doing your part in coming to
know p.
To see that this point is distinctive to ER, note that we cannot say the same about ILBC:
satisfying ILBC is not a necessary condition for knowledge. Suppose you hold the higher-order
belief that your evidence doesn’t sufficiently support your first-order belief, when in fact you
have decisive evidence against the higher-order belief and your first-order belief is supported by
your reasons. In any such case, if you gave up the first-order belief, you would comply with
ILBC, regardless of how we take the scope of ‘ought’ to be. But you would then fail to have
knowledge. What you ought to do in this situation, in order to have knowledge, is to give up the
second-order belief and to hold onto the first-order belief. So the satisfaction of ILBC itself is
unnecessary for knowledge. It can even be incompatible with it.
Of course, it is true that if you do what you ought to do (in order to have knowledge) by
discarding the higher-order belief and maintaining the first-order belief, you would avoid having
a combination of attitudes prohibited by ILBC. But this is simply a byproduct of your complying
with ER in this situation. And while it might turn out that you are guaranteed to satisfy ILBC in
some particular way by doing something else you ought to do, viz., satisfying ER, it doesn’t
follow from this that you ought to satisfy ILBC, or that the satisfaction of ILBC itself is a
necessary condition for knowledge.
135

134
Stewart Cohen’s famous “new evil demon” case best illustrates the possibility in which the subject perfectly does
their part in responding to their subjective reasons, but fails to know because the external conditions for knowledge
aren’t met. Cohen says: “if S’s belief is appropriate to available evidence, he is not to be held responsible for
circumstances beyond his ken.” (Cohen 1984: 282)
135
This, of course, is an application of the point that satisfying the requirements of formal coherence on belief and
intention is unnecessary for believing and intending what we ought to believe and intend, which is made most
forcefully by Kolodny (2008a).

164
If the point of ER is to serve as a rule that subjects ought to follow in doing their part,
however, it is extremely plausible that the kind of belief you come to have by complying with
ER is the kind of belief that having knowledge entails. For if knowledge is an attitude you come
to possess by following ER, together with the satisfaction of the objective conditions for
knowledge, it must be the same attitude of the same kind as the attitude ER refers to.
The kind of belief entailed by knowledge, I submit, is dispositional belief. What is crucial
is the fact that our ordinary knowledge-ascriptions are based on the concept of a stored mental
state. When we ascribe knowledge about something to a person, the object of the knowledge
need not be activated in their thought at all. Rather, knowledge about various things we ascribe
them, such as their knowledge of the capital of Uzbekistan or the name of their first elementary
school teacher, only rarely, if at all, occupy their mind. Still, the knowledge remains a
dispositional state of mind, stored in long-term memory, available to be retrieved on relevant
occasions.
This conception of knowledge as a stored mental state is best illustrated by the
phenomenon of evidence detachment that has interested epistemologists: people don’t remember,
or keep track of, the evidential grounds for their knowledge. As Alvin Goldman observes:

Many justified beliefs are ones for which an agent once had adequate evidence that she
subsequently forgot. […] Last year, Sally read a story about the health benefits of
broccoli in the “Science” section of the New York Times. She then justifiably formed a
belief a belief in broccoli’s beneficial effects. She still retains this belief but no longer
recalls her original evidential source. […] Nonetheless, her broccoli belief is still
justified, and if true, qualifies as a case of knowledge. (Goldman 1999: 280)

165

Clearly, the concept of belief invoked when Goldman describes Sally as retaining her broccoli
belief is that of dispositional belief, items stored in one’s representational system. The
phenomenon of evidence detachment is, according to Gilbert Harman, best explained by our
need to avoid cluttering our minds: “there is a limit to what one can remember, a limit to the
number of things one can put into long-term storage, and a limit to what one can retrieve”
(Harman 1986: 41). Given such limitations on our cognitive capacity, it is enough to put into
one’s long-term memory what one’s evidence shows to be true, and not the evidence itself.
136
It
is such beliefs stored in long-term memory, if things go well, that constitutes the stock of
knowledge we have about the world.
Let me address some immediate objections to this solution. First, one might well wonder
why ER could not apply, in addition to dispositional beliefs, to occurrent beliefs. That is, one
might wonder why cannot we have the following version of ER:

? Evidence Requirement on Occurrent Beliefs: You ought to occurrently believe p if
your total evidence requires you to believe p, and ought not occurrently believe p if your
total evidence requires you not to believe p.

For it seems that even when you occurrently believe (or judge) P, your belief ought to be
supported by evidence. But if we allow that ER also governs occurrent beliefs, it might seem that
the dilemma we are trying to avoid might simply recur at the level of occurrent beliefs.

136
See also Adler and Hicks (2013: 150) on the need to economize: “Once the evidence has served its purpose of
establishing that p, once p is accepted, the epistemic value of the evidence is absorbed or used up. No continuing
support is required for its truth.”

166
My response is that ER, if understood as a requirement on occurrent beliefs, would not be
a norm that you could comply with. To see this, note that ER refers to your total evidence, rather
than the evidence you are currently attending to. Your total evidence as to P would consist in the
totality of your mental states that pertains to the question whether P, which would include
dispositional states that aren’t occurrent, such as your stored knowledge. With that said, there is a
wide range of propositions your total evidence (now) supports, e.g. Trump is a man, The earth is
round, Seoul is the capital of South Korea, etc. If ER were a requirement on your occurrent
beliefs, however, you would be required (now) to occurrently believe all of those propositions at
Surely this would be a norm you can hardly live up to. You are able to meet the demands of ER
only through your dispositional beliefs.
This is not to say that occurrent beliefs cannot be evaluated in light of one’s total
evidence. Perhaps the following is a relevant norm that links one’s total evidence with one’s
occurrent beliefs:

Evidence Requirement on Occurrent Beliefs: You ought to occurrently believe p only
if your total evidence either requires or permits you to believe p.

This version of Evidence Requirement on Occurrent Beliefs doesn’t say that whenever your
evidence requires you to believe a proposition, you should occurrently believe it. It only requires
(plausibly) that if you occurrently believe it, it should be licensed by your total evidence. So this
principle doesn’t come into conflict with Inter-Level Occurrent Belief Coherence in cases of
MHOE. You can still satisfy both by not occurrently believing both the first-order and the
higher-order proposition.

167
Second, one might worry that my solution still rules out the possibility of knowing both
the second-order proposition and the first-order proposition in a case of MHOE. Such an
objection would appeal to the following principle:

Knowledge-Occurrent Belief Link: If S knows p, then it is epistemically rational for S
to occurrently believe p (cf. Ross and Schroeder 2014: 272).

The appeal of such a principle is obvious. If you know a proposition P in a given situation, you
should be epistemically justified in using your knowledge: judging P; using P as a premise in
theoretical or practical reasoning; asserting P, etc. On my solution, however, you are rationally
forbidden from occurrently believing both the first- and second-order propositions in such ways,
in cases of MHOE. Together with Knowledge-Occurrent Belief Link, my solution implies that
you can’t know them both in such cases.
My reply is that we should reject Knowledge-Occurrent Belief Link. First, recall
Goldman’s point about stored knowledge. Even if you have you have a mental state that amounts
to stored knowledge in P, you might momentarily fail to recall the original evidence for it.
Indeed, you might momentarily fail to recall any evidence. In such a case, it would be irrational
for you to judge P, assert P, or use P in your reasoning, even if you knew P and your total
evidence supported P. What it is rational for you to occurrently believe seems to depend not just
on your total evidence, but also on the evidence you can currently muster. Second, consider
again the case of the unconfident examinee from Chapter 4. Even if we agree that the examinee
knows that Queen Elizabeth died in 1603, it wouldn’t be rational for her to occurrently believe

168
that Queen Elizabeth died in 1603 in her situation. Thus, it can sometimes be rationally
impermissible for you to occurrently believe a proposition while knowing it.
This completes my argument for the thesis that ER targets dispositional beliefs, which
has been the second major assumption of my solution: we can satisfy both ILC and ER in a case
of MHOE by being cautious, which consists in dispositionally holding the beliefs required by ER
but refraining from using at least one of the beliefs.

5. Scorecard
So far, I have presented a solution to the puzzle generated by the possibility of MHOE, ILBC,
and ER, and justified its main assumptions. This section offers reasons to prefer this solution to
the existing solutions, by examining the theoretical costs of denying at least one of the following:
(i) ER; (ii) ILBC; (iii) the possibility of MHOE.
While no existing solution to the puzzle denies that ER is a genuine requirement on
belief, Worsnip (forthcoming) denies that ER is a requirement of epistemic rationality. On his
view, ER and ILBC are norms with fundamentally distinct sources, the former being the
requirement to respond to one’s reasons, and the latter being the requirement of coherence.
Worsnip claims that epistemic rationality is strictly a matter of having coherent patterns of
doxastic attitudes, such as having consistent beliefs, probabilistically coherent patterns of
credences, etc., and ILBC specifies one such coherent pattern. On this picture, the possibility of
MHOE suggests precisely that even ideally rational agents can fail to believe in accordance with
one’s evidence.
Such a view avoids the consequence that there can be situations in which rationality
makes conflicting demands on our doxastic attitudes and we are bound to fall short of some

169
rational ideal. But it has two costs. First, so long as one accepts ER as an important norm that
one should comply with, regardless of whether it is rational norm, one should prefer a view
according to which the (alleged) conflict between ER and ILBC can be avoided. Consider an
analogue of such a view in the practical domain. Suppose what has been regarded as a conflict
within morality is shown to be a conflict between morality and prudence. While it would
eliminate a moral dilemma, we would still be left without guidance as to what we ought to do
when this conflict between morality and prudence arises.
Secondly, and more importantly, ER seems to be borne out by our pre-theoretic talk of
rational belief: we criticize a person when they fail to believe or give up their beliefs in
accordance with the evidence they have, pointing out that it is rational for them to believe such-
and-such, given their evidence. And when we make such an assessment, we are evaluating the
particular belief one has and their evidence for it, apart from considerations of coherence. Of
course, it might be possible to construct a systematic error theory about the ordinary assessment
of rational belief.
137
Other things being equal, however, it would be better to have a solution that
preserves the phenomenon.
My solution, by contrast, allows for the joint satisfaction of the requirements. And it is
theoretically less committal: one doesn’t need to take a specific view about the nature of the each
requirement: it is consistent with the hypothesis that ER and ILBC is a requirement of rationality,
but also with the hypothesis that only ILBC is a rational requirement.
Rejecting ILBC entirely isn’t an attractive option, either. Without a requirement like
ILBC, there would be no grounds for rationally criticizing believing the paradoxical conjunction,
problematic forms of reasoning and action, discussed in section 3. The major trick of my solution

137
In fact, Worsnip (2016) offers such a theory.

170
consists in endorsing ILBC as a genuine requirement on belief, but making it clear that the
jurisdiction of ILBC is limited to occurrent beliefs. In doing so, I make room for the rational
criticism of the problematic reasoning and action, without thereby ruling out the possibility of
satisfying ER.
Finally, what are the costs of denying the possibility of MHOE? The first reason is that
its possibility seems to be a default position: nothing in the concept of evidence seems to rule out
a case like Holmes’: if you can form evidentially supported beliefs about various matters through
testimony from excellent sources, it seems just as possible for you to form evidentially supported
higher-order belief through Holmes’ testimony.
138

Second, the most prominent strategy of arguing for the impossibility of MHOE, which
has been to appeal to the notion of defeat, lacks a firm footing. The general idea is that your
higher-order evidence about the lack of sufficient evidential support for your first-order belief
can defeat the justification for the first-order belief.
139

However, this strategy faces problems, as there are formidable reasons to doubt that
higher-order evidence can plausibly count as a defeater.
140
Allen Coates (2012: 117-120), for
example, argues that the idea that such higher-order evidence can defeat the justification for
one’s belief has an extremely odd consequence. First, the following is plausibly a definition of
defeater: a consideration c is a defeater for one’s belief that p just in case if it is rational for one
to believe that p on the basis of one’s evidence e, but it is not rational for one to believe that p, on

138
See also Worsnip (2018a) for a thorough defense of the possibility of MHOE.
139
See, most notably, Bergmann (2005) and Feldman (2005).
140
Feldman (2005: 104) himself considers the intuitive oddity of treating higher-order evidence as a defeater.
Wedgwood’s (2012: 292) point is also noteworthy: “it is the real nature of [the] internal process… and not the
higher-order beliefs that the thinker has, or even the beliefs that the thinker is justified in having, about the nature of
that process, that is crucial to the rationality of the mental event that results from that process.”

171
the basis of e and d. Second, the kind of evidence we are considering is the evidence that
believing that p on the basis of e is not rational. Now, suppose that a consideration is both a
defeater and higher-order evidence that one’s believing that p on evidence e is not rational. It
follows from the above definition of defeater that one’s believing that p on the basis of e must be
otherwise rational. But this means that the content of higher-order evidence, which is that one’s
believing that p on the basis of e is not rational, is false, which means that the higher-order
evidence is misleading evidence. Thus, such higher-order evidence can be a defeater for one’s
first-order belief only if it is misleading.
This consequence is odd enough. Consider yourself as deliberating about what to believe
in knowledge of this consequence. That is, suppose you are aware that higher-order evidence is a
defeater only if it is misleading. Then it becomes implausible that when you face higher-order
evidence, it is rational for you give up your first-order belief, because, by hypothesis, you know
that it is misleading, and your first-order belief is in fact sufficiently supported by evidence,
which seems to make it rational for you to hold onto your first-order belief. By definition,
however, a defeater for belief is a consideration that makes it irrational for you to hold onto the
belief. So it is unclear how higher-order evidence can be a defeater.
In sum, denying the possibility of MHOE, just as denying other elements of the puzzle,
brings in some controversial theoretical commitments, which too are avoided by my solution.
Thus, I conclude that the solution presented in this chapter offers the best available solution to
the puzzle of misleading higher-order evidence.

172
Concluding Chapter: Doing without Local Coherence

This dissertation has advanced two main theses. One is that coherence requirements are plausibly
seen as governing occurrent, rather than dispositional attitudes (or attitudes understood as
standing states). The first argument in support of this thesis is that such a restriction on the
jurisdiction of the requirements allows us to respect the plausible idea that one can satisfy the
requirement directly through reasoning (Chapter 2). The second is that the irrationality of having
an incoherent combination of attitudes would be manifest only if the involved attitudes were all
occurrent, and there would be no clear cases of irrationality when some of the attitudes is
dispositional (Chapter 3). The third is that this view offers a neat solution to some prominent
puzzles in epistemology (Chapters 5 and 6).
The other thesis is that once we accept the first thesis, the alleged requirements are better
seen as descriptive principles about how beliefs and intentions work or operate in our activities
involving them. The main argument started from the intuitive unintelligibility of ascribing an
incoherent combination of attitudes to one under certain conditions: when one’s attitudes are all
occurrent and when one’s mind is not compartmentalized. I have argued that if such conditions
are not met, one is not subject to a rational requirement, given some philosophical assumptions
about the conditions under which one is subject to those requirements. The requirements are
satisfied, as a matter of conceptual necessity, whenever one is subject to them (Chapter 3). I have
also shown how the plausible idea that occurrent belief and intention each involves settledness
can bolster this view (Chapter 4).
Let me now sum up some major upshots of my view. There have been two prominent,
distinguishable approaches to rationality in the philosophical literature. According to one,

173
rationality is a matter of coherence between one’s attitudes: rationality requires that some
particular relations hold between your propositional attitudes, without requiring that you take up
or give up a particular attitude. Rationality so understood is sometimes called structural
rationality, and this is the conception of rationality that has been the main focus of this
dissertation. According to the other approach, rationality is a matter of responsiveness to reasons:
your rationality is determined by how well you respond to reasons you have, or how well your
individual attitudes are supported or justified by your reasons.
141
Rationality so understood is
sometimes called substantive rationality.
142

Plausibly, it is this distinction between structural and substantive rationality that has
given rise to the apparent puzzles we have considered. In cases of misleading higher-order
evidence (Chapter 6), for example, Evidence Requirement (a requirement of substantive
rationality) and Inter-Level Coherence (a requirement of structural rationality) seem to yield
conflicting verdicts about what to believe. The preface paradox, discussed in Chapter 5, can also
be seen as resulting from the conflict between substantive and structural rationality: Eric the
historian seems to be required by structural rationality to adhere to Consistency and Conjunctive
Closure, but at the same time seems to be required by substantive rationality to believe each
claim in the body of the book, as well as the preface claim.
More generally, this distinction gives rise to a puzzle about the possibility of a unifying
account of rationality: why do such distinct (and seemingly conflicting!) “normative” properties
go under the same heading of “rationality”? Is there an interesting relationship between the two,
such that one can be explained in terms of the other? Or should we simply equate rationality with

141
Just what it is to have a reason (or what it is for a consideration to be one’s subjective reason) is a matter of
controversy. For some prominent accounts, see Kiesewetter (2017), Lord (2010, 2017), Schroeder (2008, 2011).
142
The distinction between structural and substantive rationality is due to Scanlon (2007), and ise explained in detail
in Fogal (2018) and Worsnip (2018b).

174
one of them and classify the other into a distinct category? There is little consensus on these
issues. For example, Broome (2013) and Worsnip (2016, 2017) identify rationality with
structural rationality and deny the rational significance of reasons. Kolodny (2008a, 2008b)
argues that rationality understood as coherence is a myth, and opts for a conception of rationality
on which rationality consist in responding to one’s beliefs about reasons. Others, such as Lord
(2014b, 2017) and Schroeder (2009), seem to take substantive rationality theoretically primary
and try to explain structural requirements in terms of responsiveness to subjective reasons.
143

My view suggests a simple dissolution of the puzzle: we should drop the assumption that
there are two, potentially competing sources of normativity, since structural and substantive
rationality differ in nature. Structural requirements are, as I’ve argued, descriptive principles of
belief and intention, whereas substantive rationality genuinely normative, in the sense that you
can fail to correctly respond to your reasons. It is on the basis of substantive rationality, which is
the only genuinely normative domain, that we can make sense of our ordinary criticism of our
beliefs, intentions, and other attitudes.
To see how one can necessarily satisfy the requirements of coherence without thereby
responding to reasons one has, consider what people might naturally see as a paradigm case of
practical irrationality. You know you must stop binge watching and get to work in order to do
important things you wanted to get done by tonight. But you just keep going and fail to get
things done. This might be taken to be a perfect example of incoherence which people are said to
often exhibit, perhaps due to their laziness or temptation. So one might be led to think that you
violate Means-End Coherence here, and that it explains why you are criticizable in this case.

143
See Way (2013, forthcoming), however, for some important objections to such attempts.

175
On my view, however, you never violate Means-End Coherence in such a case. In
believing that you need to get to work in order to get things done but not forming the intention
do so, you simply fail to occurrently intend to get things done by tonight. What you occurrently
intend to do at that moment is to keep on watching, and you intend what you believe to be a
necessary means to doing so—sitting in front of the screen. There is no violation of Means-End
Coherence, so far as your occurrent attitudes are concerned.
However, this leaves room for criticizing your attitudes in terms of reasons. We might
say that you are unreasonable if and only if you fail to have attitudes that you have strong
reasons to have, and if you have attitudes that you have strong reasons not to have. Then there is
a clear sense in which you have proceeded unreasonably in the above scenario. Recall that
Means-End Coherence specifies a combination of attitudes it would be irrational for you to have,
and there are multiple ways of avoiding such a combination in this case: (i) not intending to get
things done; (ii) not believing that starting to work is a necessary means; and (iii) intending to
start working. Importantly, these options are not equally reasonable. In this case, (iii) is
intuitively a reasonable thing for you to do, while neither (i) nor (ii) seems reasonable. That is
because you seem to have good reasons to start working in this situation, whereas you seem to
lack good enough reason to give up your means-end belief or your intention to get things done.
The fact that your attitudes are necessarily means-end coherent doesn’t mean that you always
correctly respond to reasons you have.
I’ve deliberately called the failure to respond to subjective reasons unreasonableness,
rather than irrationality. This is because it is not clear if unreasonableness is equivalent to the
concept of irrationality I’ve been working with in this dissertation. First, the charge of
unreasonableness differs from the charge of irrationality in that you could be unreasonable in

176
having an attitude A in a situation S, even if you lacked the specific ability not to A in S and so
could not be legitimately expected not to A in S. (Chapter 3). To see this, suppose you have
sufficient evidence that P is false, but you believe P nonetheless, because your epistemic reasons
(or evidence) regarding P might be too complicated for you to evaluate, or the range of relevant
evidence is too vast for you to consider. Even then we could properly call your belief
unreasonable, on the grounds that it goes against your evidence. In such a case, however, you
lack the specific ability to respond to your epistemic reasons.
Moreover, consider other attitudes (including desires and emotions) which don’t seem to
be susceptible to rational control, or the attitudes which we don’t seem to be able to affect by
exercising our rational capacities alone. There are cases in which we cannot simply be rid of our
fear of an insect even when we are fully aware of the fact that it is totally harmless (and the
evidence thereof). There are also cases in which we simply cannot resist having strong urges,
lusts, depression, sadness, etc. even when we fully recognize that they are groundless. While it
seems perfectly okay to call such attitudes unreasonable in the sense that they fail to be
responsive to our reasons, it doesn’t seem that we have the specific ability to be rid of such
attitudes whenever we have them.
Second, irrationality in this dissertation has been understood as a matter of breaching a
rule or a principle. Indeed, rationality on this model can be conceived of as a system of rules,
each of which prohibits a combination of attitudes, which is similar to a legal system which
forbids various courses of actions. And just as whether you are obeying the speed limit when
driving on Freeway 110 is an “all-or-nothing” affair, whether you are rational with respect to a
rational rule is an all-or-nothing affair: assuming that you are subject to the rule, you either
breach it or obey it. By contrast, (un)reasonableness comes in degrees, since one’s reasons for

177
attitudes can vary in their strength. Plausibly, one could have more (or stronger) reasons to have
an attitude A than to have an attitude B, and the degree to which one’s failure to respond to one’s
reasons is unreasonable could be evaluated in terms of the strength of the reasons.
In any case, my view offers a unified picture of “rational” discourse. Contrary to
appearances, structural and substantive rationality aren’t really competing normative conceptions
of rationality, since structural rationality consists of descriptive principles. It is only the demands
of substantive rationality that we could fail to meet, and the only relevant normative failure we
can identify in our rational discourse is unreasonableness.
Another puzzle my view dissolves is the puzzle about reasons to satisfy the coherence
requirements (briefly touched upon in Chapter 3). We are inclined to think that we ought to be
rational or there is reason to be rational. Rules of rationality, it is said, differ from rules of
etiquette or rules of chess, in that they are authoritative for all rational beings. Moreover, this
judgment about the fundamental authority of rationality seems to be what explains some ethical
rationalists’ attempt to vindicate the authority of morality by establishing that being moral is a
part of what it is to be rational (Nagel 1970, Darwall 1983, Korsgaard 2009).
On careful reflection, however, it is far from clear why we ought to satisfy individual
coherence requirements. Indeed, both Kolodny (2007a, 2008a, 2008b) and Broome (2005,
2007b, 2013) doubt that a satisfying answer can be given to this question. Both assume that if
rationality itself is to be normative at all, it must always give you at least a reason to satisfy its
requirements that apply to you, but find it unclear what such a reason could be.
For example, one might argue that complying with coherence requirements is always a
means to doing what you ought to do, and see this fact as a reason to be rational. The problem is
that there are many ways of satisfying a particular coherence requirement that are not such

178
means. Consider Intention Consistency, for example. Suppose you falsely believe that you
cannot both finish writing a paper and go on a date this weekend, when in fact you can do both,
and you ought to do both this weekend. You would satisfy Internal Consistency if you intended
only one of the options or intended neither, none of which is a means to doing what you ought to
do. So rationality doesn’t always give us an instrumental reason to satisfy its requirements. Nor
is it clear that rationality provides you with any intrinsic reasons. As Kolodny (2008b: 386) puts
it, “it seems outlandish that the kind of psychic tidiness that [a coherence requirement] enjoins
should be set alongside such final ends as pleasure, friendship, and knowledge.”
This raises a puzzle: on its face, rationality seems to be an important source of
normativity, which determines what we ought to do or what we have reason to do. Recall the
main reason why coherence requirements have been widely accepted. It is the intuitive judgment
that necessarily something would go wrong with you if you had an incoherent combination of
attitudes, which we might have naturally expressed by saying that you ought always to have a
coherent combination of attitudes. But now we are being told that in many cases there is simply
no reason to meet a particular requirement of rationality that applies to you.
If my view is correct, the question about reasons to satisfy individual coherence
requirements is ill-posed. Asking whether we have any reason to follow a principle seems to
presuppose that complying with the principle is up to us, and also that such reason, if it exists,
can guide us to do what we otherwise would not do. Indeed, the presupposition is met in many
cases. If I were to recognize that I have strong reason to do what morality requires, for example,
I would choose to follow the requirements of morality, whereas I would choose simply to ignore
them if no convincing reasons to follow them could be found.

179
Things are different in the case of (structural) rationality. Once we see the requirements
of coherence as descriptive principles about attitudes, asking whether there is any reason for our
attitudes to be in accordance with the requirements of coherence becomes similar to asking
whether there is reason to act in accordance with the laws of nature. While such a question might
make sense in some context, it would surely be an odd question to ask: just as it would be odd to
seek justification for the way the world is, it would be odd to seek justification for the way our
attitudes operate; just as it would be odd, if not confused, to deliberate about whether to follow
the laws of nature, it would be odd to deliberate about whether to follow coherence requirements.
This suggests that the question of whether we have any reason to satisfy the coherence
requirements doesn’t make good sense.
I should make it clear that my view is distinct from the view which Kolodny (2005,
2008a) considers (and rejects) as an attempt to vindicate the authority of rationality, according to
which we ought to, or have reason to, comply with the requirements of coherence because they
are constitutive principles of belief and intention, or more generally, constitutive principles of
agency (Korsgaard 2009, Velleman 2000). While I agree that they are constitutive principles of
belief and intention, I also think that it is not normative precisely for that reason. That is, my
view says that principles of coherence cannot really be appropriately said to be something that
we ought to comply with, because they are not norms in the first place. They aren’t the right kind
of principles to have authority over us.
Let me conclude by setting an agenda for future research. My dissertation has dealt with
local, synchronic coherence requirements, which specify conflicts between specific attitudes, in
particular, belief and intention. It is silent, for example, on diachronic, or process-oriented
conceptions of rationality, on which rational assessment concerns how well one performs

180
rational activities (such as reasoning) over time, rather than how one’s attitudes should fit
together at a given time. (Podgorski 2017). My point about local, synchronic requirements of
coherence leaves diachronic process requirements out of the picture, and questions about the
nature of such requirements, e.g., whether they really are rational requirements, whether they
might similarly turn out to be constitutive principles, should be answered in the future in the
course of developing a complete picture of rationality.
Moreover, I have remained silent on coherence requirements that govern attitudes other
than belief and intention. Belief and intention, as I’ve argued, are settled attitudes, and my
working hypothesis has been that this isthe reason why they are subject to coherence
requirements. For example, it seems to be because you are settled on the question of whether p is
true when you believe p that you are unlike when you have other, unsettled cognitive attitudes
towards p, required not to believe not-p at the same time. It seems to be because you are settled
on the question of whether to A when you intend to A that you would be irrational to
simultaneously intend not to A, unlike when you have other, unsettled conative attitudes.
But there seem to be coherence requirements that govern attitudes that are not so settled.
For example, there are requirements of probabilistic coherence that govern our credences (or
degrees of belief). There are, of course, interesting questions my dissertation raises for such
requirements. If probabilistic coherence is simply understood as one’s conformity with
Kolmogorov’s axioms, it could not plausibly be a rational requirement, since that would be too
demanding: you might fail to assign credence one (or certainty) to some mathematical or logical
truths, because they are simply too complex for you to grasp; as a result, you might fail to assign
credence to a proposition Q that is as high as the credence you assign to P, even when in fact P
entails Q. If this is a problem, we might wonder how we should formulate the relevant

181
requirement on degrees of belief. Moreover, a full account of the rationality of our degrees of
belief investigating into what the relevant rational capacity for satisfying such requirements on
our degrees of beliefs is, and how exactly we might exercise such a capacity (Staffel 2013,
forthcoming). Again, such questions should be adequately answered in order to arrive at a
complete picture of rationality understood as coherence.

182
References

Adler, Jonathan (2002). Akratic Believing?. Philosophical Studies, 110(1), 1-27.
Adler, Jonathan and Hicks, Michael (2013). Non-Evidential Reasons to Believe. In Timothy
Chan (Ed.), The Aim of Belief (140-166). Oxford: Oxford University Press.
Atkinson, Richard C. and Schiffrin, Richard M. (1971). The Control of Short-Term Memory.
Scientific American, 225(2), 82-90.
Alonso, Facundo. (2017). Intending, Settling, and Relying. In David Shoemaker (Ed.), Oxford
Studies in Agency and Responsibility Vol. 2 (50-74). Oxford: Oxford University Press.
Anscombe, G. E. M. (1963). Intention (2
nd
Edition). Oxford: Blackwell.
Archer, Avery (forthcoming). Wondering about What You Know. Analysis.
Armstrong, David M. (1973). Belief, Truth and Knowledge. Cambridge: Cambridge University
Press.
Audi, Robert. (1994). Dispositional Beliefs and Dispositions to Believe. Noûs, 28(4), 419-434.
Beck, Lewis W. (1960). A Commentary on Kant’s Critique of Practical Reason. The University
of Chicago Press.
Bedke, Matthew (2009). The Iffiest Oughts: A Guise of Reasons Account of End-Given
Conditionals. Ethics, 119(4), 672-698.
Bergmann, Michael (2005). Defeaters and Higher-Level Requirements. Philosophical Quarterly,
55(220), 419-436.
Blackburn, Simon (1998). Ruling Passions. Oxford: Oxford University Press.
Blaauw, Martijn (2013). Contrastive Belief. In Martijin Blaauw (Ed.), Contrastivism in
Philosophy (88-100), New York: Routledge.
Boghossian, Paul (2014). What is Inference?. Philosophical Studies, 169(1), 1-18.
Braithwaite, R.B. (1932/1933). The Nature of Believing. Proceedings of the Aristotelian Society,
33, 129-146.
Bratman, Michael (1987). Intention, Plans, and Practical Reason. Harvard University Press.
Broome, John (1999). Normative Requirements. Ratio, 12(4), 398-419.
Broome, John (2005). Does Rationality Give Us Reasons?. Philosophical Issues, 15(1), 321-337.
Broome, John (2007a). Wide or Narrow Scope?. Mind, 116(462), 359-370.
Broome, John (2007b). Is Rationality Normative?. Disputatio, 2(23), 161-178.

183
Broome, John (2013). Rationality Through Reasoning. Wiley-Blackwell.
Brunero, John (2010). The Scope of Rational Requirements. Philosophical Quarterly, 60(238),
28-49.
Brunero, John (2012). Instrumental Rationality, Symmetry and Scope. Philosophical Studies,
157(1), 125-140.
Brunero, John (2014). Against Cognitivism about Practical Rationality. In Russ Shafer-Landau
(Ed.), Oxford Studies in Metaethics Vol. 9 (18-44). Oxford: Oxford University Press.
Brunero, John and Kolodny, Niko (2013). Instrumental Rationality. Stanford Encyclopedia of
Philosophy. (http://plato.stanford.edu/entries/rationality-instrumental/)
Buckwalter, Wesley et al. (2015). Belief through Think and Thin. Noûs, 49(4), 748-775.
Carruthers, Peter. (2013). On Knowing Your Own Beliefs: A Representationalist Account. In
Nikolaj Nottelmann. (Ed.), New Essays on Belief: Constitution, Content, and Structure
(145-165). Palgrave: MacMillan.
Carruthers, Peter (2015). The Centered Mind. Oxford: Oxford University Press.
Cassam, Quassim. (2010). Judging, Believing, and Thinking. Philosophical Issues, 20(1), 80-95.
Cherniak, Christopher (1983). Rationality and the Structure of Memory. Synthese, 57, 163-186.
Christensen, David (2004). Putting Logic in its Place: Formal Constraints on Rational Belief.
Oxford: Clarendon Press.
Christensen, David (2007). Does Murphy’s Law Apply in Epistemology?: Self-Doubt and
Rational Ideals. In Tamar Gendler and John Hawthorne (Eds.), Oxford Studies in
Epistemology Vol. 5 (3-31). Oxford: Oxford University Press.
Christensen, David (2010). Higher-Order Evidence. Philosophy and Phenomenological
Research, 81(1), 185-215.
Clarke, Randolph (2015). Abilities to Act. Philosophy Compass, 10(12), 893-904.
Clarke, Roger (2013). Belief Is Credence One (in Context). Philosopher’s Imprint, 13, 1-18.
Clarke, Roger (forthcoming). Preface Writers Are Consistent. Pacific Philosophical Quarterly.
Coates, Allen (2013). Rational Epistemic Akrasia. American Philosophical Quarterly, 49(2),
113-124.
Cohen, Stewart (1984). Justification and Truth. Philosophical Studies, 46(3), 279-295.
Cowan, Nelson (2008). What Are the Differences between Long-Term, Short-Term, and Working
Memory?. Prog. Brain Res, 169, 323-338.

184
Darwall, Stephen (1983). Impartial Reason. Ithaca: Cornell University Press.
Davidson, Donald (2004). Problems of Rationality. Oxford: Clarendon Press.
Dennett, Daniel (1989). The Intentional Stance. Cambridge, MA: MIT Press.
DeRose, Keith (1992). Contextualism and Knowledge Attributions. Philosophy and
Phenomenological Research, 52(4), 913-929.
Dreier, James (2000). Dispositions and Fetishes: Externalist Models of Moral Motivation.
Philosophy and Phenomenological Research, 66(1), 159-161.
Evnine, Simon (1999). Believing Conjunctions. Synthese, 118(2), 201-227.
Fantl, Jemery and McGrath, Matthew (2009). Knowledge in an Uncertain World. Oxford:
Oxford University Press.
Feldman, Richard (2005). Respecting the Evidence. Philosophical Perspectives, 19(1), 95-119.
Feldman, Richard (2009). Evidentialism, Higher-Order Evidence, and Disagreement, Episteme
6(3), 294-312.
Fink, Julian (2014). A Constitutive Account of ‘Rationality Requires’. Erkenntnis, 79(4), 909-
941.
Finlay, Stephen (2007). Responding to Normativity. In Russ Shafer-Landau (Ed.), Oxford Studies
in Metaethics Vol. 2 (220-239). Oxford: Oxford University Press.
Finlay, Stephen (2009). Against All Reason?: Skepticism about the Instrumental Norm. In
Charles R. Pigden (Ed.), Hume on Motivation and Virtue (155-178). Palgrave MacMillan.
Finlay, Stephen (2014). Confusion of Tongues: A Theory of Normative Language. New York:
Oxford University Press.
Fodor, Jerry (1987). Psychosemantics. Cambridge, MA: MIT Press.
Fogal, Daniel (2018). On the Scope, Jurisdiction, and Application of Rationality and the Law.
Problema, 12, 21-57.
Foley, Richard (1987). The Theory of Epistemic Rationality. Cambridge, MA: Harvard
University Press.
Foley, Richard (1993). Working Without a Net: A Study of Egocentric Epistemology. New York:
Oxford University Press.
Formosa, Paul (2017). Kantian Ethics, Dignity and Perfection. Cambridge University Press.
Frankish, Keith (2004). Mind and Supermind. Cambridge: Cambridge University Press.
Friedman, Jane (2013). Question-Directed Attitudes. Philosophical Perspectives, 27(1), 145-174.

185
Friedman, Jane (2017). Why Suspend Judging?. Noûs, 51(2), 302-326.
Friedman, Jane (forthcoming). Belief and Inquiry. Noûs.
Galvin, Richard (2009). The Universal Law Formulas. In Thomas E. Hill. (Ed.), The Blackwell
Guide to Kant’s Ethics (52-82). Wiley-Blackwell.
Gendler, Tamar (2008). Alief and Belief. Journal of Philosophy, 105(10): 634-663.
Gertler, Brie (2011). Self-Knowledge and the Transparency of Belief. In Anthony Hatzimoysis
(Ed.), Self-Knowledge (125-45). Oxford: Oxford University Press.
Goldman, Alvin (1999). Internalism Exposed. Journal of Philosophy, 96(6), 271-293.
Greco, Daniel (2014). A Puzzle about Epistemic Akrasia. Philosophical Studies, 167(2), 201-
219.
Greenspan, Patricia (1972). Conditional Oughts and Hypothetical Imperatives. Journal of
Philosophy, 72(10), 259-276.
Hampton, Jean (1998). The Authority of Reason. Cambridge University Press.
Harman, Gilbert (1976). Practical Reasoning. Review of Metaphysics, 29(3), 431-463.
Harman, Gilbert (1986). Change in View. Cambridge, MA: MIT Press.
Hawthorne, John (2004). Knowledge and Lotteries. Oxford: Oxford University Press.
Hazlette, Allan (2012). Higher-Order Epistemic Attitudes and Intellectual Humility. Episteme,
9(3), 205-223.
Hedden, Brian (2016). Mental Processes and Synchronicity. Mind, 125(499), 873-888.
Hieronymi, Pamela (2005). The Wrong Kind of Reason. Journal of Philosophy, 102(9), 437-457.
Hieronymi, Pamela (2006). Controlling Attitudes. Pacific Philosophical Quarterly, 87(1), 45-74.
Hieronymi, Pamela (2008). Responsibility for Believing. Synthese, 161(3), 357-373.
Hieronymi, Pamela (2009). Two Kinds of Agency. In Lucy O’Brien & Matthew Soteriou. (Eds.),
Mental Action (138-162). Oxford: Oxford University Press.
Hill, Thomas E. (1973). The Hypothetical Imperative. Philosophical Review, 82(4), 429-450.
Hill, Thomas E. (1989). Kant’s Theory of Practical Reason. Monist, 72(3), 363-83.
Holton, Richard (2009). Willing, Wanting, Waiting. Oxford: Oxford University Press.
Horowitz, Sophie (2014). Epistemic Akrasia. Noûs, 48(4), 718-744.
Hubin, Donald (2001). The Groundless Normativity of Instrumental Rationality. Journal of
Philosophy, 98(9), 445-468.
Huemer, Michael (2011). The Puzzle of Metacoherence. Philosophy and Phenomenological

186
Research, 82(1), 1-21.
Hurley, Susan (1993). Natural Reasons. Oxford: Oxford University Press.
Kant, Immanuel (1987). Critique of Judgment Trans. Werner. S. Pluhar. Hackett.
Kant, Immanuel (1996). Practical Philosophy, Ed. and trans. Mary Gregor. Cambridge
University Press.
Kant, Immanuel (2007). Anthropology, History, and Education. Ed. Günter Zöller and Robert B.
Louden. Cambridge University Press.
Kaplan, Mark (1996). Decision Theory as Philosophy. Cambridge: Cambridge University Press.
Kaplan, Mark (2013). Coming to Terms with our Human Fallibility: Christensen on the Preface.
Philosophy and Phenomenological Research, 87(1), 1-35.
Kelly, Thomas (2003). Epistemic Rationality as Instrumental Rationality: A Critique. Philosophy
and Phenomenological Research, 66(3), 612-640.
Kelly, Thomas (2005). The Epistemic Significance of Disagreement. In Tamar Gendler and John
Hawthorne (Eds.), Oxford Studies in Epistemology Vol. 1 (167-196). Oxford: Oxford
University Press.
Kerstein, Samuel J. (2002). Kant’s Search for the Supreme Principle of Morality. Cambridge
University Press.
Kiesewetter, Benjamin (2017). The Normativity of Rationality. Oxford: Oxford University Press.
Kim, Brian (2015). This Paper Surely Contains Some Errors. Philosophical Studies, 172(4),
1013-1029.
Kolodny, Niko (2005). Why Be Rational?. Mind, 114(455), 509-563.
Kolodny, NIko (2007a). How Does Coherence Matter?. Proceedings of the Aristotelian Society,
107, 229-63.
Kolodny, Niko (2007b). State or Process Requirements?. Mind, 116(462), 371-385.
Kolodny, Niko (2008a). Why Be Disposed to Be Coherent?. Ethics, 118(3), 437-463.
Kolodny, Niko (2008b). The Myth of Practical Consistency. European Journal of Philosophy,
16(3), 366-402.
Korsgaard, Christine M. (1986). Skepticism about Practical Reason. Journal of Philosophy,
83(1), 5-25.
Korsgaard, Christine M. (1997). The Normativity of Instrumental Reason. In Garrett Cullity and
Berys Gaut (Eds.), Ethics and Practical Reason (215-254). Clarendon Press.
Korsgaard, Christine M. (2009). Self-Constitution: Action, Identity and Integrity. Oxford

187
University Press.
Kruglanski, Arie W. and Webster, Dona M. (1996). Motivated Closing of the Mind: “Seizing”
and “Freezing”. Psychological Review, 103(2), 263-283.
Lasonen-Aarnio, Maria (2014). Higher-Order Evidence and the Limits of Defeat. Philosophy and
Phenomenological Research, 88(2), 314-345.
Lavin, Douglas (2004). Practical Reason and the Possibility of Error. Ethics, 114(3), 424-457.
Littlejohn, Clayton (2018). Stop Making Sense?: On a Puzzle about Rationality. Philosophy and
Phenomenological Research, 96(2), 257-272.
Lord, Errol (2010). Having Reasons and the Factoring Account. Philosophical Studies, 149(3),
283-296.
Lord, Errol (2014a). The Real Symmetry Problem(s) for Wide-Scope Accounts of Rationality.
Philosophical Studies, 170(3), 443-464.
Lord, Errol (2014b). The Coherent and the Rational. Analytic Philosophy, 55(2), 151-175.
Lord, Errol (2017). What You Are Rationally Required to Do Is What You Ought to Do. Mind,
126(504), 1109-1154.
Maier, John (2013). The Agentive Modalities. Philosophy and Phenomenological Research,
87(3), 113-134.
Makinson, David C. (1965). The Paradox of the Preface. Analysis, 25(6), 205-207.
McCann, Hugh J. (1991). Settled Objectives and Rational Constraints. American Philosophical
Quarterly, 28(1), 25-36.
McGaugh, James (2000). Memory—A Century of Consolidation. Science, 287(5451), 248-251.
McHugh, Conor and Way, Jonathan. (2018). What is Reasoning?. Mind, 127(505), 167-196.
Mele, Alfred (1992). Springs of Action. Oxford: Oxford University Press.
Mele, Alfred (2003). Motivation and Agency. New York: Oxford University Press.
Mele, Alfred (2009). Effective Intentions: The Power of Conscious Will. New York: Oxford
University Press.
Myers-Schulz, Blake and Schwitzgebel, Eric (2013). Knowing that P without Believing that P.
Noûs, 47(2), 371-384.
Nagel, Jennifer (2008). Knowledge Ascriptions and the Psychological Consequences of
Changing Stakes. Australasian Journal of Philosophy, 86(2), 279-294.
Nagel, Jennifer (2010). Epistemic Anxiety and Adaptive Invariantism. Philosophical

188
Perspectives, 24(1), 407-435.
Nagel, Thomas (1970). The Possibility of Altruism. Oxford: Clarendon Press.
O’Neill, Onora (1989). Constructions of Reason. Cambridge University Press.
Parfit, Derek (2011). On What Matters, Vol 1. Oxford: Oxford University Press.
Paton, Herbert. J. (1947). Categorical Imperatives: A Study in Kant’s Moral Philosophy.
Hutchinson’s University Library.
Peacocke, Christopher (1998). Conscious Attitudes, Attention, and Self-Knowledge. In C. J. G.
Wright et al. (Eds.), Knowing Our Own Minds, Oxford: Oxford University Press.
Pettit, Philip and Smith, Michael (1996). Freedom in Belief and Desire. Journal of Philosophy,
93(9), 429-449.
Podgorski, Abelard (2017). Rational Delay. Philosopher’s Imprint, 17(5), 1-19.
Price, Henry H. (1969). Belief. London: Allen & Unwin.
Radford, Colin (1966). Knowledge—By Examples. Analysis, 27(1), 1-11.
Railton, Peter (1997). On the Hypothetical and Non-Hypothetical in Reasoning about Belief and
Action. In Garrett Cullity and Berys Gaut (Eds.), Ethics and Practical Reason (53-79).
Oxford: Oxford University Press.
Rawls, John (2000). Lectures in the History of Moral Philosophy. Ed. Barbara Herman, Harvard
University Press.
Raz, Joseph (2005). The Myth of Instrumental Rationality. Journal of Ethics and Social
Philosophy, 1(1), 1-28.
Reath, Andrews (2006). Agency and Automony in Kant’s Moral Theory. Oxford University
Press.
Ridge, Michael (1998). Humean Intentions. American Philosophical Quarterly, 35(2), 157-178.
Ridge, Michael (2014). Impassioned Belief. Oxford: Oxford University Press.
Rippon, Simon (2014). Were Kant’s Hypothetical Imperatives Wide-Scope Oughts?.
Australasian Journal of Philosophy, 83(3), 783-788.
Rose, David and Schaffer, Jonathan (2013). Knowledge Entails Dispositional Beliefs.
Philosophical Studies, 166(S1), 19-50.
Ross, Jacob and Schroeder, Mark (2014). Belief, Credence, and Pragmatic Encroachment.
Philosophy and Phenomenological Research, 88(2), 259-288.
Ryle, Gilbert (1949). The Concept of Mind. New York: Barnes & Noble.

189
Scanlon, Thomas (1998). What We Owe to Each Other. Cambridge, MA: Harvard University
Press.
Scanlon, Thomas (2007). Structural Irrationality. In Geoffrey Brennan (Ed.), Common Minds:
Themes from the Philosophy of Philip Pettit (84-103). Oxford: Clarendon Press.
Schechter, Joshua (2013). Rational Self-Doubt and the Failure of Closure. Philosophical Studies,
163(2), 428-452.
Schroeder, Mark (2004). The Scope of Instrumental Reason. Philosophical Perspectives, 18(1),
337-364.
Schroeder, Mark (2005). The Hypothetical Imperative?. Australasian Journal of Philosophy,
92(4), 357-372.
Schroeder, Mark (2008). Having Reasons. Philosophical Studies, 139(1), 57-71.
Schroeder, Mark (2009). Means-End Coherence, Stringency, and Subjective Reasons.
Philosophical Studies, 143(2), 223-248.
Schroeder, Mark (2011). What Does It Take to “Have” a Reason?. In Andrew Reisner and
Asbjørn Steglich-Petersen (Eds.), Reasons for Belief (201-222). Cambridge: Cambridge
University Press.
Schroeder, Mark (2015). Hypothetical Imperatives, Scope, and Jurisdiction. In Robert N.
Johnson and Mark Timmons (Eds.), Reason, V alue, and Respect (89-100). Oxford
University Press.
Schwartz, Jeremy (2008). Do Hypothetical Imperatives Require Categorical Imperatives?.
European Journal of Philosophy, 18(1), 84-107.
Schwitzgebel, Eric (2002). A Phenomenal, Dispositional Account of Belief. Noûs, 36(2), 249-
275.
Schwitzgebel, Eric (2010). Belief. Stanford Encyclopedia of Philosophy.
(http://plato.stanford.edu/entries/belief/)
Schwitzgebel, Eric (2013). A Dispositional Approach to Attitudes: Thinking Outside of the Belief
Box, In Nikolaj Nottelmann. (Ed.), New Essays on Belief: Constitution, Content, and
Structure (75-99). Palgrave: MacMillan.
Setiya, Kieran (2007). Cognitivism about Instrumental Reason. Ethics, 117(4), 649-67.
Shah, Nishi (2006). A New Argument for Evidentialism. Philosophical Quarterly, 56(225), 481-
498.

190
Shah, Nishi and Velleman, David (2005). Doxastic Deliberation. Philosophical Review, 114(4),
497-534.
Shaver, Robert (2006). Korsgaard on Hypothetical Imperatives. Philosophical Studies, 129(2),
335-347.
Sinhababu, Neil (2011). The Humean Theory of Practical Irrationality. Journal of Ethics and
Social Philosophy, 6(1), 1-13.
Sinhababu, Neil (2013). The Desire-Belief Account of Intention Explains Everything. Noûs,
47(4), 680-696.
Siyar, Jamsheed (2013). The Conditionality of Hypothetical Imperatives. Kantian Review, 18(3),
439-460.
Sloman, Aaron (1970). ‘Ought’ and ‘Better’. Mind, 37(315), 385-394.
Smith, Michael (1995). Internal Reasons. Philosophy and Phenomenological Research, 55(1),
109-131.
Smithies, Declan (2012). Moore’s Paradox and the Accessibility of Justification. Philosophy and
Phenomenological Research, 85(2), 273-300.
Staffel, Julia (2013). Can There Be Reasoning with Degrees of Belief?. Synthese, 190(16), 3535-
3551.
Staffel, Julia (forthcoming). Attitudes in Active Reasoning. In Magdalena B. Jackson and
Brendan B. Jackson (Eds.), New Essays on Theoretical and Practical Reasoning. Oxford:
Oxford University Press.
Stalnaker, Robert (1984). Inquiry. Cambridge, MA: MIT Press.
Stanley, Jason (2005). Knowledge and Practical Interests. Oxford: Oxford University Press.
Svavarsdóttir, Sigrun (1997). Evaluations of Rationality. In Terry Horgan and Mark Timmons
(Eds.), Metaethics After Moore (121-136). Oxford: Oxford University Press.
Titelbaum, Michael (2015). Rationality’s Fixed Point. In Tamar Gendler and John Hawthorne
(Eds.), Oxford Studies in Epistemology Vol. 5 (253-294). Oxford: Oxford University
Press.
Valaris, Markos (2017). What Reasoning Might Be. Synthese, 194(6), 2007-2024.
Velleman, David (1989). Practical Reflection. NJ: Princeton University Press.
Vellman, David (2000). The Possibility of Practical Reason. Oxford: Oxford University Press.
Velleman, David (2007). What Good Is a Will?. In Anton Leist and Holger Baumann (Eds.),

191
Actions in Context (193-215). De Gruyter/Mouton.
Wallace, R. Jay (2001). Normativity, Commitment, and Instrumental Reason. Philosopher’s
Imprint, 1(4), 1-26.
Way, Jonathan (2010). Defending the Wide-Scope Approach to Instrumental Reason.
Philosophical Studies, 147(2), 213-233.
Way, Jonathan (2011). The Symmetry of Rational Requirements. Philosophical Studies, 155(2),
227-239.
Way, Jonathan (forthcoming). Reasons and Rationality. In Daniel Star (Ed.) Oxford Handbook of
Reasons and Rationality. Oxford: Oxford University Press.
Weatherson, Brian (2005). Can We Do Without Pragmatic Encroachment?. Philosophical
Perspectives, 19(1), 417-443.
Wedgwood, Ralph (2009). The Normativity of the Intentional. In Ansgar Beckermann and Brian
P. McLaughlin (Eds.), The Oxford Handbook of the Philosophy of Mind (421-436).
Oxford: Clarendon Press.
Wedgwood, Ralph (2012). Justified Inference. Synthese, 189(2), 273-295.
Wedgwood, Ralph (2013). Rational ‘Ought’ Implies ‘Can’. Philosophical Issues, 23(1), 70-92.
Wedgwood, Ralph (2014). Outright Belief. Dialectica, 66(3), 309-329.
Wood, Allen (1999). Kant’s Ethical Thought. Cambridge University Press.
Wood, Allen (2013). Kant on Practical Reason. In Mark Timmons and Sorin Baiasu (Eds.), Kant
on Practical Justification (57-86). Oxford University Press.
Worsnip, Alexander (2015). Narrow-Scoping for Wide-Scopers. Synthese, 192(8), 2617-2646.
Worsnip, Alexander (2016). Moral Reasons, Epistemic Reasons, and Rationality. Philosophical
Quarterly, 66(26), 341-361.
Worsnip, Alexander (2018a). The Conflict of Evidence and Coherence. Philosophy and
Phenomenological Research, 96(1), 3-44.
Worsnip, Alexander (2018b). Reasons, Rationality, Reasoning: How Much Pulling Apart?,
Problema, 12, 59-93.
Worsnip, Alexander (forthcoming). What is (In)coherence?. In Russ Shafer-Landau (Ed.), Oxford
Studies in Metaethics (Vol. 13). Oxford University Press.
Zimmerman, Aaron. (2007). The Nature of Belief. Journal of Consciousness Studies, 14(11), 61-
82.

Abstract (if available)

Abstract Rationality, in an important sense, is about the coherence between attitudes. There are certain incoherent patterns of attitudes that would be clearly irrational, such as holding contradictory beliefs or intentions, failing to intend the necessary means to the ends that one intends, failing to believe what logically follows from one’s beliefs, and so on. Naturally, many philosophers have come to hold that there are coherence requirements of rationality, which permit or forbid certain combinations of attitudes (and absences of attitudes), applying primarily to beliefs and intentions. ❧ My dissertation develops two novel theses about the nature of such coherence requirements. Its first main thesis is that the coherence requirements should be properly seen as governing occurrent beliefs and intentions, rather than beliefs and intentions understood as dispositional (or standing) mental states. Roughly, occurrent attitudes are attitudes currently activated in one’s mind, influencing one’s thought or action in ways characteristic of those attitudes, whereas dispositional attitudes are items stored in one’s representational system, which we can attribute to one even when they are not playing any role in one’s mind. ❧ The second thesis of the dissertation is that once we restrict the target of the requirements to occurrent attitudes, we should accept that each coherence requirement is not really a normative requirement we can violate, but rather a descriptive principle about the relations that necessarily hold between beliefs and intentions. For example, it is extremely difficult to imagine a concrete scenario in which someone occurrently holds contradictory beliefs or intentions, or someone occurrently intends an end, occurrently believes that something is a necessary means to the end, and yet fails to intend it. I argue that the best explanation of the unintelligibility of attributing such incoherent combinations of occurrent attitudes is given by the view that it is constitutive of occurrent belief and intention that they meet the coherence requirements. Since only occurrent attitudes are subject to the coherence requirements, and occurrent attitudes necessarily meet the requirements, the requirements turn out to be inviolable. ❧ What is the philosophical significance of the overall view? It dissolves some puzzles raised by the natural idea that there are coherence requirements of rationality. First, rationality seems to be an important source of normativity, which partly determines what we ought to do, unlike rules of chess or etiquette. However, some writers, such as John Broome and Niko Kolodny, have forcefully argued that rationality understood as coherence between attitudes is not normative, in the sense that there isn’t necessarily a reason to comply with its demands. But how could rationality be a source of normativity in the absence of good reason to satisfy its requirements? My view simply dissolves this question, since there are no coherence requirements in the first place. There are only descriptive principles about how our attitudes combine together, and no normative demands that our attitudes ought to meet. ❧ Chapter 1 uses Kant’s view on instrumental rationality as a test case, encapsulated in his claim that whoever wills the end also wills the indispensably necessary means to it that is in his control. While many commentators read it as a normative claim to the effect that there is a requirement of reason which dictates one to take the necessary means to one’s ends, I argue that no version of the normative reading can be plausibly squared with Kant’s more explicit philosophical commitments. On my alternative interpretation of Kant, it is simply part of what it is to will an end that if one wills an end and has knowledge of the necessary means, one also wills those means. This means that on Kant’s view, the principle of instrumental rationality is not a principle we ought to comply with, but rather a principle that we in fact follow. ❧ Chapter 2 offers a reason for taking rational requirements to apply to occurrent attitudes, invoking the intimate connection between reasoning and rationality. It seems that we can directly modify our attitudes, and thereby come to meet the demands of rationality, through performing reasoning. I argue that this intuitive idea turns out false if we conceive of the basic targets of rational requirements as dispositional attitudes. Reasoning proceeds at the level of working memory, whereas dispositional attitudes are items stored in one’s long-term memory, but a causal process called “consolidation” is involved in transferring items in working memory to one’s long-term memory, which means that the process in which reasoning affects our dispositional attitudes are at best indirect. By contrast, we can modify our occurrent attitudes directly through reasoning. Therefore, we should recognize the proper targets of rational requirements as occurrent attitudes if we want to preserve the attractive idea that we can directly change our attitudes and thereby improve our rationality through reasoning. ❧ Chapter 3 provides a general argument that occurrent attitudes are necessarily coherent. I appeal to the difficulty of making sense of someone who claims to have a pattern of attitudes that goes against a coherence requirement. For example, if a person says, ‘I intend to go hiking, but, at the same time, I intend not to go hiking’ or ‘I intend to go hiking, I believe that I can do so only by preparing a pair of shoes, but I don’t intend to prepare any shoes’, it is difficult to attribute the intentions they claim themselves to have. I argue that the best explanation of the unintelligibility of ascribing someone an incoherent combination of attitudes is given by the view that occurrent attitudes necessarily comply with rational requirements. I also offer another main argument for the view that coherence requirements apply only to occurrent attitudes, which is that our intuitions about the (ir)rationality of having an incoherent combination of attitudes are sensitive to the distinction, and that we would have a case of criticizable irrationality only if the involved attitudes were all occurrent. Since only occurrent attitudes are subject to requirements, but occurrent attitudes necessarily satisfy the requirements, the requirements turn out inviolable. ❧ Chapter 4 aims to connect the impossibility of incoherence with the nature of belief and intention. I explore the idea that they involve being settled on a question. It is this very feature, I argue, that guarantees the necessity of the coherence of occurrent belief and intention. ❧ Chapters 5 and 6 give further support to the view that rational requirements govern only occurrent attitudes, by showing how some prominent philosophical puzzles about rational belief can be solved by the view. Chapter 5 discusses Makinson’s well-known paradox of the preface. I point out that it will be impossible for any human author to have occurrent beliefs in every proposition in which the book expresses belief. So, in believing ‘Some of the beliefs expressed in this book are false’, the author does not guarantee that she has an inconsistent set of occurrent beliefs. At most, the author has an inconsistent set of dispositional beliefs. But inconsistent dispositional beliefs need not violate any local coherence requirement. Chapter 6 deals with the puzzle about misleading higher-order evidence, which is often taken to reveal the conflict between demands of coherence and the demands of reasons. I argue that this (alleged) conflict can be resolved by meeting the former with occurrent beliefs, and the latter with dispositional beliefs.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Contrastive reasons

PDF

Process-oriented rationality

PDF

Reasoning with degrees of belief

PDF

The virtue of reasonableness: on the normative epistemic implications of public reason liberalism

PDF

Cognitive boundaries for rational coherence requirements

PDF

Beliefs that wrong

PDF

A perceptual model of evaluative knowledge

PDF

A deontological explanation of accessibilism

PDF

Units of agency in ethics

PDF

Reasons, obligations, and the structure of good reasoning

PDF

The case for moral skepticism

PDF

Conceptually permissive attitudes

PDF

Positivist realism

PDF

Public justification beyond legitimacy

PDF

Feeling good and living well: on the nature of pleasure and its role in well-being

PDF

Reasoning with uncertainty and epistemic modals

PDF

The mental states first theory of promising

PDF

Aggregating complaints

PDF

Responsibility and the emotional structure of relationships

PDF

Capturing moods: a philosophical exploration

Asset Metadata

Creator Lee, Woo Ram (author)

Core Title Rationality and the primacy of the occurrent

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Philosophy

Publication Date 08/06/2018

Defense Date 06/14/2018

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Belief,coherence,dispositional attitudes,intention,OAI-PMH Harvest,occurrent attitudes,rational requirements,rationality,reasoning,reasons,structural requirements

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Finlay, Stephen (committee chair), Schroeder, Mark (committee member), Wedgwood, Ralph (committee member)

Creator Email lwr0109@gmail.com,wooramle@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-53435

Unique identifier UC11670818

Identifier etd-LeeWooRam-6653.pdf (filename),usctheses-c89-53435 (legacy record id)

Legacy Identifier etd-LeeWooRam-6653.pdf

Dmrecord 53435

Document Type Dissertation

Format application/pdf (imt)

Rights Lee, Woo Ram

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