University of Southern California Dissertations and Theses

Contrastive reasons

(USC Thesis Other)

Contrastive reasons

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content CONTRASTIVE REASONS
by
Justin Snedegar
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHILOSOPHY)
May 2013
Copyright 2013 Justin Snedegar
to Emmy, for love and support, and for always sharing my intuitions
and to my parents, for always being proud of me
ii
Acknowledgements
This dissertation, in its current form, was conceived about two years ago. But work
towards it really began after my rst year of graduate school at USC, after a seminar
on deontic modality with my eventual advisor, Mark Schroeder. I spent much of my
second year working with Mark, developing a contrastive view of deontic modals. I
also spent a lot of time talking to Steve Finlay about this work, and we eventually
co-authored a paper on the topic. So Mark and Steve have been with me from the
earliest stages of this dissertation.
Thanks to Steve for always being very encouraging, and for, at the same time,
oering sharp criticisms. A common experience was getting a set of comments,
which began with something like `This is great! I think you're basically right about
everything', followed by pages of very serious criticisms and suggestions. I nd
myself agreeing with Steve more than probably any other philosopher, and it was
great to have someone who thought about things from much the same angle as I
do around. Jake Ross has also been involved from early on. Back when I thought
I would write a much longer dissertation, including lots of stu on deontic modals,
he encouraged me to focus on reasons. His work also in
uenced me a great deal.
I also beneted from talking with many of the other faculty members at USC.
I will just mention a few in particular. Thanks to Scott Soames, for playing such
a big role in my development early on in my career at USC, and for discussions
about the material here in particular. Ralph Wedgwood joined the faculty at USC
after I was well into this project, but showed a lot of enthusiasm for my work, and
provided useful feedback. Thanks also to Barry Schein, both for teaching me most
of what I know about formal semantics and for serving as an outside member on
my committee.
Thanks to all of the graduate students for contributing to the great commu-
nity at USC; thanks especially to Indrek Reiland, Julia Stael, Josh Crabill, Alida
Liberman, Lewis Powell, Johannes Schmitt, and Sam Shpall have all contributed
iii
signicantly. Ben Lennertz has read most of my work, and talked with me about
philosophy, fantasy baseball, and life (in that order) frequently over the last ve
years. I have learned a lot over the last few years from talking with Shyam Nair.
And Justin Dallmann helped me with several parts of this dissertation, especially
with the probabilities in Chapter 4. Thanks to him, and to Amanda Dallmann, for
feeding Emmy and me many Friday nights and for being great friends.
Outside of USC, two people especially have been extremely encouraging and
helpful. Fabrizio Cariani has given me great feedback on much of my work, in-
cluding big chunks of material from this dissertation. Walter Sinnott-Armstrong
has been more consistently enthusiastic about my work than anyone, including me.
He has given me very insightful comments on virtually every chapter of this dis-
sertation. Besides Mark and Steve, he has probably had the biggest in
uence on
the following pages. Thanks are also due to the philosophy department at West
Virginia University, between 2005 and 2008, when I was an undergraduate there|
especially to Sharon Ryan and Andy Cullison, who taught me how to do philosophy
and encouraged me to pursue it.
My greatest professional debt is to my advisor, Mark Schroeder. I've been work-
ing very closely with him since the beginning of my second year of graduate school.
He bought into the project from the start, and his dedication to it was crucial in
helping me see it through. Mark has in
uenced the dissertation immensely; this
will be obvious from the beginning. The ideas here are mine, but I would not have
had most of them if Mark hadn't dragged them out of me. His own work has served,
at dierent times, as part of my arsenal, as a target, and as an inspiration. I can't
imagine a better or more invested advisor; thanks, Mark.
Now some personal debts of gratitude. Thanks to my parents for supporting
me, encouraging me, and being proud of me. Thanks to my dad, my work has
been read by more bikers from West Virginia than is probably the norm for work
in metaethics. And my debt to my mom is simply immeasurable. Finally, thanks
most of all to my best friend, sounding board, gourmet chef, and wife, Emmy. She
has been the best possible partner. Throughout the last ve years in Los Angeles,
she has been unwaveringly supportive and loving. Without her, this dissertation
wouldn't exist; lots of more important things would also be missing.
iv
Table of Contents
Acknowledgements iii
Abstract vii
Chapter 1 Contrastivism and Reasons 1
1.1 Reasons and Contrastivism . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Sinnott-Armstrong's Argument . . . . . . . . . . . . . . . . . . . . 7
1.3 The Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Chapter 2 Reason Claims 21
2.1 A Simple Argument . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2 A Stronger Argument . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Contrastivism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4 Negative Reason Existentials . . . . . . . . . . . . . . . . . . . . . . 32
2.5 Looking Forward . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Chapter 3 Favoring 38
3.1 Why Resist Contrastivism? . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Shallow Contrastivism . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Favoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Contrastive Reasons and Favoring . . . . . . . . . . . . . . . . . . . 48
3.5 Looking Forward . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Chapter 4 Promotion 53
4.1 A Problem for the Contrastivist . . . . . . . . . . . . . . . . . . . . 54
4.2 Promotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3 Contrastive Reasons and Promotion . . . . . . . . . . . . . . . . . . 64
4.4 Providing the Constraints . . . . . . . . . . . . . . . . . . . . . . . 67
4.5 Promotion as Probability-Raising . . . . . . . . . . . . . . . . . . . 72
4.6 Non-Promotional Reasons . . . . . . . . . . . . . . . . . . . . . . . 74
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
v
Chapter 5 Intransitivity 78
5.1 Transitivity and Reasons . . . . . . . . . . . . . . . . . . . . . . . . 79
5.2 Intransitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.3 Two Kinds of Intransitivity . . . . . . . . . . . . . . . . . . . . . . 89
5.4 Contrast-Sensitive Importance . . . . . . . . . . . . . . . . . . . . . 90
5.5 Remaining Questions . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Chapter 6 Withholding 106
6.1 Withholding Belief and Contrastive Reasons . . . . . . . . . . . . . 106
6.2 Withholding Intention . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.3 A Non-Contrastive Alternative . . . . . . . . . . . . . . . . . . . . . 122
6.4 Wrap Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
References 129
vi
Abstract
This dissertation develops and defends a contrastive view of normative reasons,
according to which they are fundamentally reasons for one thing rather than an-
other, or for one thing out of some range of alternatives. Reasons for something
relative to one set of alternatives may not be reasons for that same thing relative to
some other alternatives. This is in opposition to nearly all existing work on reasons,
which assumes without argument|even without mention|that reasons are reasons
for actions simpliciter, independently of what we're comparing the action to.
I argue that this contrastive view is supported by (i) our talk about reasons,
(ii) the idea that reasons favor the things they are reasons for, and (iii) the idea
that reasons involve the promotion of things like desires, goals, or values. These
arguments build, step by step, to a detailed version of contrastivism that can both
claim advantages over the traditional non-contrastive picture and avoid problems
that face less well-developed versions of contrastivism.
One thread running through the dissertation is an exploration of the depth of
contrastivism. Chapter 2 argues that contrastivism goes at least as deep as our use
of the word `reason'|this word (as used to ascribe normative reasons) should be
given a contrastive treatment. Chapter 3 argues that contrastivism actually goes
deeper, to the normative reason, or favoring, relation that normative philosophers
are concerned with. Chapter 4, though, argues for a somewhat moderate contrastive
theory. The normative reason relation is indeed contrastive, but we need not|in
fact, we'd better not|adopt the deepest possible form of contrastivism.
Given the central place of reasons in normative philosophy, we should expect
that contrastivism will have upshots throughout areas like ethics, practical rational-
ity, and epistemology. In the nal two chapters of this dissertation, I start to explore
this idea. This is just a start, but it suggests that exploring further applications
for contrastivism is likely to be a fruitful area for future research.
vii
Chapter 1
Contrastivism and Reasons
When I moved to Los Angeles I lived in Thai Town, so I was surrounded by good
Thai restaurants. One of my favorites is Red Corner Asia; so the fact that I'm
in the mood for Thai food seems like a reason to go to Red Corner Asia. But
Spicy BBQ also serves delicious Thai food. So is the fact that I'm in the mood
for Thai food actually a reason to go to Red Corner Asia? It's not clear|after
all, I could just as easily get Thai food by going to Spicy BBQ. But now suppose
that lots of restaurants are closed tonight, and that my only two options are Red
Corner Asia and Paru's Indian restaurant. Then it seems obvious that the fact
that I'm in the mood for Thai is a reason to go to Red Corner Asia. Here's a very
natural way to describe what's going on: the fact that I'm in the mood for Thai
food is a reason to go to Red Corner Asia rather than go to Paru's, but it's not
a reason to go to Red Corner Asia rather than go to Spicy BBQ. Of course, you
might think that this doesn't answer the original question: is it a reason to go to
Red Corner Asia, or not? But again, that question can seem hard to answer in the
situation described, in which I could also get Thai food by going to Spicy BBQ.
The contrastive question of whether it's a reason to go to Red Corner Asia rather
than go to Spicy BBQ, though, is easy to answer. This is a special case of a very
general observation: often some consideration seems to be a reason to perform an
action rather than some actions, but not a reason to perform it rather than some
other actions.
1
A theory of reasons needs to explain this.
In this dissertation, I defend a theory of reasons that takes this observation
seriously. On this theory, called contrastivism about reasons, reasons are always
reasons for one thing rather than another, or more generally reasons for one thing
1
See Sinnott-Armstrong (2006), ch. 5 and Ross (2006), ch. 9.
1
out of a certain set of alternatives, instead of reasons for things simpliciter, as has
been traditionally assumed. So the reason why the non-contrastive question in the
previous paragraph is hard to answer is that it is underspecied: to know whether
r is a reason to , we have to know `Rather than what?'. Obviously nothing I've
said yet establishes that this theory is true. Establishing that is the goal of this
dissertation. To do so, I'll develop a detailed version of contrastivism, motivating
each step along the way by arguing that contrastivism gives us the best way to
make sense of important and puzzling phenomena involving reasons. Since reasons
are so important in philosophy|talk about reasons shows up in any eld in which
normativity plays an important role|this dissertation will include discussion from
several elds, including ethics, practical reasoning, metaethics, and epistemology.
1.1 Reasons and Contrastivism
This dissertation advocates contrastivism about reasons. To get clear on just what
this view is, we need to get clear on two things: reasons and contrastivism.
1.1.1 Reasons
The theory I will develop here concerns normative reasons. These are considera-
tions, usually taken to be facts or true propositions, that count in favor of or against
doing various things: performing actions, believing propositions, and having a va-
riety of other attitudes.
2
For example, the fact that I'm in the mood for Thai food
is a reason for me to go to Red Corner Asia, and the fact that you are sleeping
in the next room is a reason against turning the volume on my stereo up high.
These reasons are what are often called pro tanto: they count in favor of or against
performing various actions (or having various attitudes) with a certain strength or
weight. There may be some reasons for me to go to Red Corner Asia|I am in the
mood for Thai food, it is nearby, and so on|and some reasons against doing so|it
is crowded at this time of day, it is expensive, and so on. These pro tanto reasons for
and against the action contribute, in some way, to what I ought to do.
3
Reasons for
2
This talk of `counting in favor' is ubiquitous, but see especially Scanlon (1998). Hieronymi
(2005) thinks that thinking of reasons primarily in these terms leads to problems, though even
she agrees that reasons do count in favor of the things they are reasons for.
3
Broome (2004), for example, uses the term `pro tanto'; Dancy (2004), on the other hand, uses
`contributory' to talk about the same kinds of reasons.
2
an action may outweigh the reasons against it, or vice versa; similarly, the reasons
for one action may outweigh the reasons for another action. Thus, contrastivism
is not a thesis, at least not primarily, about what there is \reason to do, overall",
or what there is \decisive reason to do". It is rather a thesis about these pro tanto
reasons, that contribute to what there is reason to do, overall.
Normative reasons are distinct from two other sorts of reasons. First, they are
distinct from motivating reasons|the reasons for which agents do the things they
do. These are cited in explanations of the actions of agents. Of course, it is plausible
that the facts which are an agent's motivating reasons could also be normative rea-
sons for that agent. But the motivating and normative reason relations are distinct.
Second, normative reasons are distinct from explanatory reasons, which we cite in
explaining phenomena more generally (plausibly, motivating reasons are a species
of explanatory reasons). For example, the reason why the sink is leaking is that it
needs a new washer. Motivating reasons and explanatory reasons more generally
do not count in favor of or against doing anything; they are rather explanations
why things happen. Besides a short discussion of explanation near the end of this
chapter, I will not have much more to say about these kinds of reasons; thus, when
I talk about reasons, I will usually not include the qualier, `normative'.
1.1.2 Contrastivism
I will be almost exclusively concerned with contrastivism about reasons. But
writers have proposed contrastive theories of several important philosophical con-
cepts. Contrastivism about explanation, for example, goes back|at least in explicit
form|to van Fraassen (1980).
4
Staying in the philosophy of science, philosophers
have more recently oered contrastive theories of conrmation.
5
The basic idea is
that some fact e explains some other fact p rather than another fact q, or conrms
some hypothesis h rather than some other hypothesis j, instead of explaining p
or conrming h simpliciter. More recently contrastive theories of knowledge|the
view that knowledge is always knowledge of p rather than q|have become pop-
ular.
6
Blaauw (2012) has argued that knowledge is contrastive because belief is
4
See also Lipton (1990); Hitchcock (1996).
5
Chandler (2007, 2013). See Fitelson (2012) for critical discussion.
6
See Morton and Karjalainen (2003); Schaer (2004, 2005b, 2007, 2008).
3
contrastive. Sinnott-Armstrong (2004, 2006, 2008) has argued instead that jus-
tication is contrastive. In metaphysics, Schaer (2005a, 2012) has argued that
causation is contrastive. And, perhaps most importantly for my purposes, several
writers have proposed contrastive theories of `ought'.
7
In the most general terms,
contrastivism about some property or relation R is the view that R includes an
argument place that must be lled by a set of alternatives. SoR only holds or fails
to hold of some number of things relative to sets of alternatives.
Contrastivism about some concept is often motivated by showing how judgments
involving the concept appear to be question-relative, so that the truth value of a
claim involving the concept can vary depending on a relevant question. For example,
Schaer (2007) has argued that to know thatp is to know thatp as the answer to a
particular question, salient in the context. Thus, knowing that p when the salient
question is Q does not guarantee that you know that p when the salient question
is Q
0
. And Cariani (fc) describes his contrastive view of `ought' as the view that
`ought' is relative to a deliberative question: the question of what to do. The move
from question-relativity to contrastivism is a natural one because a standard view of
questions in formal semantics treats them as partitions on (some subspace of) logical
space, or as a set of alternatives with each alternative corresponding to one cell of
the partition.
8
So on this view, question-relativity just is alternative-relativity.
There's a question about how to implement this idea: how is the set of alterna-
tives, or question, provided? Sometimes it's provided explicitly. Most obviously, the
alternatives can be provided explicitly using a `rather than' clause, as in `There's
a reason to go to RCA rather than Paru's'. Here the alternatives are clearly going
to RCA and going to Paru's. Since these kinds of ascriptions make the alternative
very obvious, I'll appeal to them frequently. Similarly, some writers have argued
that knowledge-wh ascriptions, like `I know when to hold 'em' and `I know what
you did last summer', relate an agent and a question, where the wh-clause provides
the question explicitly (`When to hold 'em?', `What did you do last summer?').
9
Of course, even if we are given the question, we need to know the possible answers:
7
See Sloman (1970); Jackson (1985); Finlay (2009, ms); Finlay and Snedegar (fc); Cariani
(2009, fc); Snedegar (2012); Kierland (ms).
8
See Hamblin (1958); Groenendijk and Stokhof (1997); Higginbotham (1993, 1996) for this sort
of view of questions.
9
See Schaer (2007); Stanley (2011). Ascriptions that explicitly include a question are very
natural for knowledge and explanation ascriptions, but not for reason, `ought', or causal ascrip-
tions.
4
which ones count as relevant, and how are we to individuate answers? Moreover,
many ascriptions do not do anything to explicitly introduce|or even let on that
there are|alternatives, as in `You know that I love you' and `The fact that I love
you is a reason for you to love me back'. What are the sets of alternatives relative
to which we should interpret these sorts of claims?
The most natural view is a version of contextualism, according to which the
argument place is lled by some contextually relevant set of alternatives. There are
various ways to make a set of alternatives relevant. For example, using intonational
stress, as in `There's a reason to drive to the store' (letting italics mark stress),
makes relevant a set of alternatives in which the alternatives dier in (the object
corresponding to the) stressed item, e.g.fdrive to the store, bike to the store, walk
to the storeg. If I say instead, `There's a reason to drive to the store', the relevant
alternatives might befdrive to the store, drive to church, drive to campusg.
10
Other features of the context can make an alternative salient, as well. For example,
Schaer (2007); Cariani (fc); Finlay (2009); Finlay and Snedegar (fc) all appeal to
questions under discussion. If I'm deliberating between going to Red Corner Asia,
going to Spicy BBQ, and going to Paru's, the natural set of alternatives relative to
which an ascription like `The fact that my guest loves Indian food is a reason to
go to Paru's' is to be evaluated isfgo to Red Corner Asia, go to Spicy BBQ, go
to Paru'sg. Thus, on this simple contextualist view, the speaker's intentions will
play a large part in determining the set of alternatives relative to which we should
evaluate reason ascriptions.
11
I will have more to say about this in the next chapter
when I develop a contrastive account of reason ascriptions.
There are two interesting features of these sets of alternatives. First, they are
not necessarily exhaustive of logical space, or of all of the things that it's possible
for the agent to do.
12
Some alternatives are left out, for one reason or another
10
See Dretske (1970); Rooth (1992); Schaer (2005b, 2007, 2008).
11
I don't think this speaker-contextualist version is necessary. For example, we might have a
view on which the set of alternatives is provided by the context of assessment, or perhaps even
by the relevant agent's (not necessarily the speaker, and not necessarily in the speaker's context)
circumstances, if we develop things in the right way. But I'll develop the contextualist version of
the view.
12
It's not quite right to talk about questions or alternatives partitioning (part of) logical space
if you treat the elements of the set as actions, as I've been doing, rather than as propositions. So
if we take this idea of question-relativity seriously, there are two options. First, we can say that,
strictly speaking, reasons are reasons for propositions (i.e. a reason for the proposition that you
) rather than for actions (i.e. a reason for you to ). (Compare Finlay (2006, ms).) Second, we
can say that reasons really are reasons for actions, but in our model, hold the agent xed so that
5
(e.g., they are irrelevant).
13
Second, the options are divided up at a more or less
ne-grained level of detail. For example, one particularly coarse-grained set would
be something likefgo to Red Corner Asia, don't go to Red Corner Asiag, while a
more ne-grained one would befgo to Red Corner Asia, go to Spicy BBQ, go to
Paru'sg. This is what Yalcin (2011) and, following him, Cariani (fc) call resolution-
sensitivity, the idea being that we partition logical space at higher or lower resolu-
tions. Either or both of these features can be exploited to solve various puzzles.
14
I'll end up making use of both of these features in my version of contrastivism about
reasons. Given the standard treatment of questions I mentioned earlier, originally
due to Hamblin (1958), it's easy to see how both of these features could come along
with question-relativity. Non-exhaustivity comes along because questions need not
partition the entirety of logical space|sometimes they partition only a subspace.
Resolution-sensitivity comes along simply because the question has to partition
this subspace at some resolution or other. Which possibilities are left out as ir-
relevant and the resolution at which the relevant possibilities are partitioned will
both depend, among other things, on context (on the contextualist version of con-
trastivism). They may also be determined explicitly by wh-complements (in the
case of `knows', for example) or by `rather than' clauses.
So contrastivism about reasons is the thesis that reasons to are always reasons
to relative to, or out of, some particular set of alternatives. The set of alternatives
need not be exhaustive of all of the things the agent could possibly do, and the
alternatives in the set must be divided up at some particular resolution. This
captures the following ideas. First, when we deliberate about what to do, we don't
always do so over every possible alternative; that's non-exhaustivity. Second, we
can only consider the relevant alternatives at some level of detail or other; that's
resolution-sensitivity. Which set is relevant can vary depending on features of the
we can treat actions as propositions (i.e. the set of worlds in which the relevant agent performs
the relevant action). See Cariani (fc) for discussion of this last strategy.
13
Compare the relevant alternatives theory of knowledge in Lewis (1996); Dretske (1970).
14
Skorupski (2010) explicitly endorses a view of reasons for action on which they are relativized
to what he calls `choice sets': \When we think of think of sucient reasons for doing something
[. . . ] we assume some exhaustive partitioning of actions, within which one and only one action
must be done. Call an exhaustive partitioning of actions a choice set; `inaction' and likewise the
action of further deliberation are members of this set" (p. 38). Skorupski's view does not count
as a contrastivist view as I've dened it above, because he does not accept non-exhaustivity. But
he does accept resolution-sensitivity, so as far as that goes, I'm sympathetic. But since I think we
need to adopt non-exhaustivity to gain lots of the advantages of contrastivism, I reject Skorupski's
view.
6
context. This set must be supplied before we can say whether r is a reason to
or not, since it can only be or fail to be a reason relative to a set of alternatives.
Arguments for contrastivism claim that as the set of alternatives shifts, the reasons
for and against a given action can vary|what was a reason to relative to one set
may not be a reason to relative to a dierent set.
1.2 Sinnott-Armstrong's Argument
I have just been discussing contrastivism as a general philosophical program which
has been applied to several important concepts like knowledge, justication, `ought',
and explanation. One thing to notice is that all (or at least almost all) of the
concepts for which contrastivism seems plausible and fruitful are intimately related
to reasons. So a tempting thought is that contrastivism about reasons is really
central, and underwrites contrastivism about all of these other concepts. In fact,
Walter Sinnott-Armstrong has exactly this thought:
In my view, epistemologists need contrasts in their analyses of knowl-
edge and justied belief because knowledge and justied belief require
reasons [. . . ], and reasons are always reasons for one thing as opposed to
another. [. . . ] The same rationale will then apply to any other kind of
reasons, and many philosophical issues concern reasons. Epistemologists
study reasons for belief. Moral philosophers investigate moral reasons
for action. Aestheticians explore reasons to like or to value certain art,
music, and so on. Philosophers of science analyze explanations, which
give reasons why events happen. And so on. Because reasons are central
to all these areas of philosophy, and all reasons are relative to contrasts,
all these areas of philosophy can benet from introducing a new place
for contrasts into the relations used in analyses.
15
This suggests a kind of argument to the best explanation: the best explanation
for why these other concepts|justication, obligation, explanation, and so on|are
contrastive is that they inherit it from the contrastivity of reasons.
While I agree both that reasons are contrastive and that this is a tempting
thought, I do not think this is a very compelling argument for contrastivism about
15
Sinnott-Armstrong (2008), pp. 257-258.
7
reasons. To show why, I will call into question both the necessity and the suciency
of contrastivism about normative reasons for contrastivism about the normative no-
tions of justication and `ought'. Moreover, the arguments will involve the introduc-
tion of plausible views that accept contrastivism about reasons without accepting
contrastivism about the other normative concept (`ought' or justication), or vice
versa. All of this should make us suspicious of the inference Sinnott-Armstrong
seems to draw.
1.2.1 Reasons and justication
Contrastivism about justication is the view that you can only be justied in be-
lieving that p out of, or relative to, a certain set of alternatives Q. Contrastivism
about justication leads naturally to contrastivism about knowledge, since knowl-
edge requires justication, and contrastivism about knowledge has become a major
player in the epistemological literature.
16
Nevertheless, I'm going to focus on con-
trastivism about justication here since justication is more directly tied to reasons
(the relationship between knowledge and reasons is plausibly mediated by justi-
cation). One motivation for adopting contrastivism about justication is that it
seems to oer a solution to the infamous closure paradox. It seems that Moore is
justied in believing that he has hands. And he's justied in believing that if he
has hands, then he's not a handless brain-in-a-vat. But it also seems that he's not
justied in believing that he's not a handless brain-in-a-vat. The contrastivist can
say that Moore is justied in believing that he has hands rather than, say,
ippers.
But he's not justied in believing that he's not a handless brain-in-a-vat rather than
that he has hands. In other words, he's justied in believing that he has hands out
of some sets of alternatives, but not out of others. In this section, I'm just going to
assume for the sake of discussion that justication is contrastive. So now what we
need to do to establish contrastivism about reasons, along the lines suggested by
Sinnott-Armstrong, is (i) show how justication is related to reasons, and (ii) show
how this relationship requires reasons to be contrastive if justication is.
So what is the relationship between justication and reasons? A very intuitive
rst answer is that you can only be justied in believing a proposition if you have
sucient reason to believe it. But (i) this is stated non-contrastively, and (ii)
16
See Morton and Karjalainen (2003); Schaer (2004, 2005b, 2007).
8
we need some idea of what counts as sucient reason. Sinnott-Armstrong oers
the following principle, inspired by the relevant alternatives theories of knowledge
oered by Dretske (1970) and Lewis (1996):
Ruling Out: s is justied in believing thatp out ofQ is has sucient reason to
rule out all the other alternatives in Q.
17
It might be a bit more natural to state things in terms of having sucient evidence
to rule out the alternatives. But since I'm talking about reasons, and since any
piece of evidence is very plausibly a reason for belief (though it might not be true
that all reasons for belief are evidence), I'm going to state things in terms of reasons.
One important thing to keep in mind is that having sucient reason to rule out
some alternative q does not require having conclusive reason to believe:q; it may
be enough if your reasons for belief make q very unlikely, for example. A second
important caveat is that it isn't enough, according to this principle, to have more
reason to believe p than any other alternative in Q, since that still might not be
enough to rule out the other alternatives.
Note that, though this principle does make justication, and so whether you
have sucient reason to believe that p relative to sets of alternatives|you might
have sucient reason to believe that p relative to Q but not relative to Q
0
|it
does not yet make reasons contrastive. So why think that contrastivism about
justication requires contrastivism about reasons? Here's a case (borrowed from
Dretske (1970)) that I think illustrates the motivating idea well. Suppose you're at
the zoo. You see a black and white striped horse-like animal in front of you. And
suppose that this visual evidence is the only relevant reason for belief. Well, relative
to the set of alternativesfbelieve it's a zebra, believe it's an elephant, believe it's
an ostrichg, it seems that you have sucient reason to believe it's a zebra|your
reasons for belief are sucient to rule out that it's an ostrich or an elephant. But
relative to the set of alternativesfbelieve it's a zebra, believe it's a cleverly disguised
muleg, it seems that you don't have sucient reason to believe that it's a zebra.
In fact, it looks like the fact that was your reason for belief in the previous case,
that you were having such-and-such visual experience, is just not even a reason in
17
Sinnott-Armstrong is also sympathetic to the idea that being justied in believing that p
relative toQ requires having some positive reason to believe p, as well as having sucient reason
to rule out the alternatives. I'll ignore that addendum here, because I don't think it aects either
his or my argument. It is also relevant, I think, that since the alternatives in Q are supposed to
be mutually exclusive, evidence for p will likely be evidence against the other alternatives.
9
this case|the fact that you see a black and white striped horse-like animal is not
a reason to believe it's a zebra rather than a cleverly disguised mule. That would
certainly explain why you aren't justied relative to this new set, since justication
requires having sucient reason for belief. So here's a principle relating contrastive
justication and contrastive reasons:
Contrastive Justication, Contrastive Reasons: s is justied in believing p
out of Q i s's reasons for belief relative to Q are sucient to rule out every
member of Q except for p.
We can explain why justication is contrastive by (i) pointing out that justica-
tion requires having sucient reason to believe, and (ii) reasons for belief are only
reasons relative to sets of alternatives.
But is this the only, or even obviously the best, way to explain why justication
is contrastive? I don't think so. Suppose that reasons aren't contrastive. Here's
a story about how justication could still be contrastive. Non-contrastive reasons
favor believing certain propositions independently of any set of alternatives. And,
importantly, some non-contrastive reasons favor ruling out some propositions|that
is, believing their negation|independently of any set of alternatives. Then we can
give the following principle:
Contrastive Justication, Non-Contrastive Reasons: s is justied in believ-
ing that p out of Q i s has sucient reason to rule out all the alternatives
in Q except for p.
18
This thesis appeals to non-contrastive reasons, but still allows justication to be
contrastive. To know whether s is justied in believing p, we still have to know
what sets of alternatives we're talking about, so we know which alternatives have
to be ruled out by s's reasons. This principle, of course, is just the original idea
captured by Ruling Out. Reasons weigh in favor of believing or disbelieving
certain propositions, but only certain propositions are relevant in a given context.
Above I appealed to the zebra/cleverly disguised mule case to motivate thinking
that reasons are contrastive. The intuition was that the fact that you're having
such-and-such visual experience is a reason to believe that it's a zebra relative
tofbelieve it's a zebra, believe it's an elephant, believe it's an ostrichg, but not
18
We could add that s needs some positive reason to believe that p.
10
relative tofbelieve it's a zebra, believe it's a cleverly disguised muleg. I think
there's something to this claim: this kind of case does at least give us prima facie
evidence that reasons are contrastive. But note that this evidence is independent
of contrastivism about justication. Further, I think there is something the non-
contrastivist about reasons can say to explain this kind of case: it's true to say that
r is a reason to believe that p when the relevant set of alternatives is Q only if r
is a stronger reason to believe p than to believe any of the other alternatives in Q.
Since the fact that you're having such-and-such visual experience is not a stronger
reason to believe that it's a zebra than to believe that it's a cleverly disguised mule,
according to this view, it's not true to say, `It's a reason to believe it's a zebra',
when believing it's a cleverly disguised mule is a relevant alternative. On the other
hand, this claim would be true if the relevant alternative were, say, believing that
it's an ostrich. (I'll say more about this idea in Chapter 3 when I discuss the view
I call shallow contrastivism.)
I've argued that contrastivism about justication does not require contrastivism
about reasons. Now I'll argue that contrastivism about reasons does not require
contrastivism about justication.
Suppose again that reasons are contrastive. Further, suppose I'm a skeptic|I
don't think that you can ever be justied in believing anything. I can still accept
that reasons are contrastive: you might have good reason to believe that the animal
is a zebra rather than an ostrich. But I just don't think this matters for justica-
tion, because to be justied in believing that it's a zebra, you have to rule out all of
the possible alternative hypotheses. So the contrastive reasons that matter in de-
termining whether you're justied in believing that it's a zebra will befit's a zebra,
it's an ostrich, it's a cleverly disguised mule, it's a hallucination, . . .g. And since
you don't|and couldn't|have sucient reason to believe that it's a zebra out of
this set, you aren't justied. Or perhaps I'm more liberal in applying justication:
I think that the set that matters in determining whether you're justied is always
just what Sinnott-Armstrong (2004, 2006, 2008) calls an \everyday" contrast class,
likefit's a zebra, it's an ostrich, it's an elephantg. That is, I think the skeptic
makes unreasonable demands on the alternatives you have to rule out to be justi-
ed. Since you do have sucient reason to believe that it's a zebra relative to this
everyday set, your belief is justied (full stop). According to either of these views,
11
then, reasons are contrastive, but justication is not|there is a particular, xed
set of alternatives that is relevant for determining whether or not you are justied.
1.2.2 Reasons and `ought'
Contrastivism about `ought' is the view that `ought' claims are always to be in-
terpreted as relative to some particular set of alternatives. Whenever an agent
ought to perform some action, there is some set out of which she ought to perform
it. Several writers have argued that `ought', or what agents ought to do (to put
it non-linguistically), is contrastive, for a variety of reasons. For the purposes of
this section, I'll assume that contrastivism about `ought' is well motivated. So now
what I need to do is (i) sketch the relationship between `ought' and reasons and (ii)
evaluate whether this relationship demands that one of the concepts is contrastive
if the other is.
The most natural thesis about the relationship between `ought' and reasons says
that you ought to i you have most reason to . This thesis ts nicely with an
attractive picture of deliberation. To decide what you ought to do, you just weigh
up all your reasons to determine which option is best supported by them; that's
the one that you ought to do.
19
The contrastivist about `ought', though, cannot
accept it as stated since she thinks that we need to relativize claims about what
agents ought to do to sets of alternatives. And since these alternatives need not be
exhaustive, we can't simply talk about what you have most reason to do simpliciter,
either, in trying to state the relationship between what you ought to do (relative
to a certain set of alternatives) and reasons. It seems that we need to look at your
reasons to do things, somehow restricted or relativized to the set of alternatives to
which the `ought' claim is relativized.
One way around this problem is to relativize reasons, as well as `ought', to sets
of alternatives|that is, to be contrastivists about reasons. This lets us retain an
attractive picture of deliberation, analogous to the standard one I mentioned above.
We begin with some relevant set of alternatives, Q, and then simply deliberate
19
There are complicated questions about how to determine what you have most reason to do|
how reasons interact with one another. See Dancy (2004); Broome (2004); Schroeder (2007),
chapter 7, and Horty (2007, 2012) for discussion about how your reasons determine what you
ought to do, or how your pro tanto reasons determine what you have most reason to do.
12
over the alternatives in Q. So it's easy to see why one might think that adopting
contrastivism about `ought' requires adopting contrastivism about reasons.
But this is too fast. It's possible to give a very plausible picture on which
`ought' is contrastive but reasons are not. Begin with a traditional non-contrastive
picture of reasons, on which they are considerations that favor certain actions with
a certain weight, some stronger or weightier than others. Collect all of the agent's
reasons and weigh them up (in some way). This will give us an overall ranking of
all the possible options open to the agent. We can even divide these options up as
nely as you want. Maybe the resolution at which we divide them is the resolution
at which the agent's intentional actions can make a dierence: I can intentionally
bring it about that I put between 10.55 and 10.56 gallons of gas in my car, but
I can't intentionally bring it about that I put between 10.555 and 10.556 gallons,
for example.
20
If that's right, then we can have one option in which I put between
10.55 gallons and 10.56 gallons in, and one in which I put between 10.56 and 10.57
gallons in. But we don't have more ne-grained options than this. This doesn't
mean that I'll always deliberate between options divided up at even this high of a
resolution; I'll say more about that shortly.
So now what we have is an overall ranking of the possible options, according to
how my reasons weigh in favor of them. Now in any particular context of deliber-
ation, only some of these options will be relevant, and I'll only deliberate between
those relevant options at some particular resolution. This will generate our set, Q.
Then we can give the following principle:
Contrastive Ought, Non-Contrastive Reasons: s ought to out of Q i s
has more reason to than to perform any other alternative in Q
We essentially pull out the alternatives inQ from our overall ranking, retaining the
order of these alternatives. Non-exhaustivity is secured, since Q need not include
all the options that are ranked in the original ranking. Still, the options that do
make it intoQ will occupy a particular place on the original ranking, and the order
between them will be retained in Q. We just appeal to this ranking, provided by
my non-contrastive reasons, to determine what I ought to do.
That's non-exhaustivity. Here's how to capture resolution-sensitivity. If the
alternatives inQ are divided up at a lower level of resolution than the options in the
20
Compare the discussion in Cariani (fc); Kierland (ms).
13
original ranking are|so that my reasons rank the options at a higher resolution than
the resolution at which I deliberate between them|we group them in the obvious
way. For example, if I'm deliberating between putting more than ten gallons of gas
in the car and putting ten gallons or less in, the alternatives will be grouped in that
way:fput in 10 gallons or less, put in more than 10 gallonsg. We then have a few
options in determining how these coarse-grained options are ranked. We could rank
them based on (i) the average position of the more ne-grained alternatives they
subsume, (ii) the rank of the highest or lowest ne-grained option they subsume,
or (iii) the ne-grained option I would actually perform, were I to decide on the
coarse-grained option. I will not try to settle which of these is best; the point
is just that there are plausible ways for a contrastivist about `ought' to retain a
non-contrastive view of reasons.
Now I will argue that contrastivism about reasons is not sucient for con-
trastivism about `ought'. So rst, suppose that reasons are contrastive: a reason
to is only such a reason relative to certain sets of alternatives. If we accept
anything like the claim that you ought to i you have most reason to , how
can we maintain a non-contrastivist view of `ought'? Here's one way. Adopt the
plausible view that you ought to i there's more reason to than to not .
21
Importing contrastivism about reasons now, even though reasons are only reasons
relative to certain sets of alternatives, the particular reasons that we're concerned
with in trying to decide whether you ought to are the reasons relative to the set
of alternatives,f,:g.
We could also make the set that matters for determining what you ought to
do more interesting: perhaps it's determined by the things that it's possible for
you to do. This doesn't amount to accepting contrastivism about `ought', since
we can just say that what you ought to do is simply the best thing you can do.
According to this view, the contrastive reasons we are concerned with are the ones
that are reasons relative to the set of all the things that it's possible for you to do.
The central idea here is just that, when we are trying to determine what someone
ought to do, we look at the reasons relative to some particular privileged set of
alternatives. (This mirrors the strategy I used in the case of justication, when I
introduced skeptical and everyday contrast classes.)
21
This is essentially the principle that Schroeder (2007), pp. 130-131 adopts.
14
Contrastivism about reasons isn't useless on this picture, though. We can still
appeal to the contrast-sensitivity of reasons to explain phenomena like the follow-
ing, observed by Sinnott-Armstrong (2004, 2006, 2008). Today is your birthday.
Suppose I'm trying to decide whether to bake you a chocolate cake, or no cake at
all. Then intuitively, the fact that today is your birthday is a reason to bake you
a chocolate cake. But now suppose I'm trying to decide whether to bake you a
chocolate cake or a lemon cake. The fact that today is your birthday seems not
to be a reason to bake you a chocolate cake in this case. The contrastivist has an
easy explanation: it's a reason to bake you a chocolate cake out offbake you a
chocolate cake, don't bake you any cakeg, but not out offbake you a chocolate
cake, bake you a lemon cakeg. So there's still some motivation for adopting con-
trastivism about reasons, since it lets us explain these kinds of cases; it's just that
when you're trying to decide if you ought to, you're only concerned with a certain
set of your contrastive reasons: those relative to the privileged set.
I don't claim to have shown that Sinnott-Armstrong's argument to the best
explanation cannot work. To salvage the argument, we would need to show that,
even though contrastivism about `ought' or justication neither requires nor is re-
quired by contrastivism about reasons, his is explanation is nevertheless the best
one. One challenge here is that the views I described which accept contrastivism
about `ought' or justication without accepting contrastivism about reasons seem
very reasonable, and I'm not sure what evidence we could marshal in favor of the
views which accept contrastivism about both `ought'/justication and reasons over
these, beyond giving direct arguments for contrastivism about reasons. Finally,
of course, for the argument to actually be compelling, we would need to establish
contrastivism about `ought' and justication. Giving direct arguments for con-
trastivism about reasons, on the other hand, will let us avoid holding that view
hostage to the fortune of these other contrastive theories.
1.2.3 Reasons and explanation
In the last two subsections, I discussed the relationship between two important nor-
mative concepts, `ought' and justication, and normative reasons. In this subsec-
tion, I'm going to talk about the relationship between explanation, and explanatory
reasons. Moral philosophers, understandably, focus mostly on normative reasons.
15
And even when they talk about explanatory reasons, they are usually most in-
terested in explanatory reasons of the motivating variety: those that explain an
agent's behavior by being considerations the agent took to be normative reasons.
Even philosophers of science and epistemologists generally talk about explanations
rather than explanatory reasons. So there's a lack of literature on the topic of ex-
planatory reasons; everyone recognizes that there are such reasons, but few writers
have had much to say about them. But it's important for my purposes to ex-
plore the topic, because Sinnott-Armstrong thinks that all reasons, normative and
explanatory, are contrastive.
Explanatory reasons are clearly closely related to explanations. And a prominent
idea in the philosophy of science is that explanation is contrastive.
22
Contrastivism
about explanation is the view that some facte explains an event, or perhaps another
fact, f, only relative to certain sets of alternatives. For example, the fact that it's
humid outside explains why it's raining rather than not precipitating, but it does
not explain why it's raining rather than snowing. And the fact that it's above the
freezing point explains why it's raining rather than snowing, but not why it's raining
rather than not precipitating. My sense of the literature is that contrastivism about
explanation is more widely accepted than contrastivism about other concepts. As
has been my policy, I'll assume that contrastivism about explanation is true.
So the next step is to gure out what the relationship is between explanation
and explanatory reasons. As I said above, there has not been much written about
this topic. The reason for this, I suggest, is that explanatory reasons just are expla-
nations. Explanatory reasons are usually glossed as answers to `Why?' questions:
the reason why the carpet is wet is that the roof leaked; the reason why I'm sun-
burnt is that I forgot to put on sunscreen. And so on. But all of these things just
look like explanations for the relevant facts. One idea is that explanatory reasons
are part of the explanation for some fact. The explanation for why I'm sunburnt
is that I forgot to put on sunscreen and I stayed in the direct sunlight for a long
time and my skin has certain properties and . . . . Each of these individual parts
of the explanation might be an explanatory reason; each explanatory reason con-
tributes to the full explanation. The problem with this, though, is that when we
cite explanatory reasons, we simply seem to be giving explanations: saying `The
reason why I'm sunburned is that I forgot to put on sunscreen' seems to be the
22
See van Fraassen (1980); Lipton (1990); Hitchcock (1996).
16
same as saying `The explanation for why I'm sunburnt is that I forgot to put on
sunscreen'. It's plausible that we're just citing parts of explanations even when we
use the word `explanation', but the point is just that, whenever we purport to cite
an explanation, using `explanation', we could just as well have used `reason'.
If explanatory reasons just are explanations, or if to give an explanation for p
is just to cite an explanatory reason why p, then obviously contrastivism about
explanation entails contrastivism about explanatory reasons. So in that sense,
Sinnott-Armstrong's argument works in this case, though it doesn't seem that he
had anything quite so trivial in mind. The important point to consider now, though,
is whether accepting contrastivism about explanatory reasons puts pressure on us
to accept contrastivism about normative reasons. I'll argue that it doesn't, at least
not without some further assumptions that can be rejected.
The rst strategy, and the one that Sinnott-Armstrong seems to adopt, is to say
that normative reasons and explanatory reasons are simply dierent species of the
same genus, reasons. There are just reasons; normative reasons and explanatory
reasons are just particular kinds. But since they're part of the same general cate-
gory, important structural features, like alternative-relativity, should be shared. So
since explanatory reasons are contrastive, reasons are contrastive. And that means,
on this view, that normative reasons are contrastive, as well. Some support for
this view comes simply from the fact that we call both normative and explanatory
reasons `reasons'. Surely this isn't an accident, and one possible explanation is that
the two categories are just species of the same genus.
One problem with this view is that it seems undermotivated. The simple fact
that we call both explanatory and normative reasons `reasons' is not enough to
establish that there's really just one thing, reasons, and two dierent kinds. This is
especially so given the apparent dierences|and seemingly structural dierences|
between explanatory and normative reasons. For example, there can be a normative
reason fors to even ifs does not, but there can't be an explanatory reason whyp
ifp is false. Another apparent structural dierence is that explanatory reasons seem
to take propositions as their complements: we give reasons why p for a proposition
p. But normative reasons appear to take actions or attitudes as their complements:
we have reasons to for an action, or reasons to form attitudeA.
23
For these and
23
One promising route here might be to follow Hieronymi (2005) and note that both explanatory
and normative reasons are considerations that bear on a question. This might be the underlying
17
other reasons, many philosophers think that there are big, important dierences
between normative reasons and explanatory reasons. Another way in which this
thesis is controversial, if it is supposed to be dierent from the next strategy that
I'll consider, is that many philosophers think that normative reasons are special
kinds of explanatory reasons. On the view under consideration now, we have two
dierent species of reasons, with neither being a special case of the other (although
they are of the same genus).
A second strategy is to say that normative reasons are simply a special kind of
explanatory reason; that is, normative reasons are explanatory reasons that explain
a special kind of, usually normative, fact. Toulmin (1950); Finlay (2001, 2006, ms);
Searle (2001); Broome (2004) all hold some variety of this view (with important
dierences, of course). If normative reasons are just a special kind of explana-
tory reasons, and explanatory reasons are contrastive, then normative reasons are
contrastive, too.
A problem with this strategy is that it takes away the possibility of reducing
the normative to reasons. This is obviously not something that everyone agrees
can be done, but many people do think so. If we analyze normative reasons as
considerations that explain some normative fact, like why you ought to, or why it
would be good if you were to, or something like that, then we cannot analyze the
notion of `ought' or of goodness in terms of reasons, on pain of circularity. Since this
strategy is incompatible with such a widely held view, it is controversial (which,
again, is what I want to establish here).
A third strategy is to argue that the correct analysis of normative reasons must
be in terms of explanation, or of explanatory reasons. For example, perhaps for r
to be a reason for you to , it has to explain why your -ing would promote your
desires, or why it would be good in some way if you were to. Then if explanation
is contrastive, perhaps the contrastivity will be passed on to normative reasons.
24
But not all views of normative reasons have it that they are correctly analyzed in
terms of explanation. Hampton (1998); Scanlon (1998); Part (2011), for example,
simply take the reason relation as primitive. It's just a (normative) fact about
unifying feature in virtue of with both categories count as reasons. I won't pursue this here,
though.
24
It is actually not at all straightforward to move from contrastivism about explanation, through
one of these kinds of analyses, to contrastivism about reasons. But illustrating this is not necessary
for the main point I want to make|that any analysis of reasons like this is controversial.
18
the world that some considerations favor some things, and explanation doesn't
enter into it. And if the correct analysis of normative reasons is not in terms of
explanation, then contrastivism about explanation will not support contrastivism
about reasons.
I don't claim to have oered knockdown arguments against any of these three
strategies for moving from contrastivism about explanation to contrastivism about
normative reasons. But I hope I have shown that all three are controversial. Just
as I do not want to rely on the truth of contrastivism about `ought' or justication
to establish contrastivism about reasons|which I think is a more defensible view,
anyway|I do not want to rely on these controversial accounts of the relationship
between normative and explanatory reasons. In the rest of the dissertation, I'm
going to try to oer more direct evidence that reasons are contrastive.
1.3 The Plan
To close this chapter, I'll brie
y preview the rest of the dissertation. In chapters two,
three, and four, I aim to develop a detailed and attractive version of contrastivism
in three steps, each of which is independently-motivated. After the full theory is on
the table, I will illustrate other interesting applications of contrastivism in ethics,
epistemology, and practical reasoning in the nal two chapters.
In chapter two, I'll argue that our talk about reasons motivates a shallow kind
of contrastivism|the view that the word `reason', as used to ascribe normative
reasons, expresses a relation with an argument place for sets of alternatives. The
central argument will be that non-contrastive theories have trouble explaining puz-
zling features of explicitly contrastive reason ascriptions which employ the phrase
`rather than'. So the rst step is to motivate and develop a contrastive account of
our talk about reasons.
But this leaves open a view, which I call shallow contrastivism, on which con-
trastivity only goes \language-deep". According to this view, though the word
`reason' expresses a relation with an argument place for sets of alternatives, that
relation is not the important normative relation for theorizing in ethics, episte-
mology, and practical reasoning. There is an underlying non-contrastive normative
favoring relation that holds between actions or attitudes and considerations which
favor them independently of the relevant alternatives, and in terms of which we can
19
analyze the contrastive relation expressed by `reason'. This is a more conservative
view than deep contrastivism, according to which the contrastive relation expressed
by `reason' is the important normative relation, and is not to be analyzed in terms
of a deeper non-contrastive normative relation. However, after sketching shallow
contrastivism, I'll argue that the normative favoring relation itself is contrastive,
so that we should adopt deep contrastivism. This is the second step in developing
contrastivism.
The theory, as it stands after chapter three, is that reasons to , out of a set
of alternatives Q, are facts that favor -ing out of Q. This is just the contrastivist
implementation of the ubiquitous, if uninformative, view that reasons are consid-
erations that favor the things they are reasons for. The third step is to try to
say something more substantive, by giving an analysis of the favoring, or reason,
relation. In fact, I'll show that we need to do this, since otherwise contrastivism
faces a serious problem. I will argue, however, that we can solve this problem by
appealing to the popular idea that reasons involve the promotion of various kinds
of objectives, like desires or values. And moreover, I'll argue that this idea itself
provides strong independent support for contrastivism, since contrastivism provides
the best way to understand how reasons could involve promotion. The result is an
attractive, detailed version of contrastivism about reasons.
In chapter ve, I'll consider the possibility of the intransitivity of `more reason
than'. Most people think this relation couldn't possibly be intransitive, but oth-
ers, like Temkin (1987, 2012); Rachels (1998, 2001); Friedman (2009), have given
compelling arguments for the possibility of intransitivity. From this perspective,
contrastivism might seem especially well-placed to accommodate and explain in-
transitivity, in virtue of the independence of reasons relative to dierent alterna-
tives. I'll explore ways we might do just that, and conclude that contrastivism
does in fact provide an especially natural framework on which to accommodate
intransitivity, but that doing so is ultimately optional for the contrastivist.
Finally, in chapter six, I'll use contrastivism to provide an attractive account of
rational withholding of both belief and intention|an important but underexplored
topic in the theory of rationality. The upshot of these nal two chapters is that
contrastivism about reasons has interesting applications in normative domains, in
which reasons are of fundamental importance.
20
Chapter 2
Reason Claims
Now that I've gotten the preliminaries on the table, and argued that to establish
contrastivism we need some direct evidence that reasons are contrastive, I'll start
trying to provide that evidence. In this chapter, I'll argue that our talk about
reasons|our use of what I'll call reason claims|supports contrastivism. There are
puzzles with various sorts of claims which are best solved by contrastivism.
2.1 A Simple Argument
Sinnott-Armstrong (2004, 2006, 2008) points out that the following reason ascrip-
tions are both intuitively true:
(1) The fact that it's your birthday is a reason for me to bake you a chocolate
cake rather than bake you no cake at all.
(2) The fact that it's your birthday is not a reason for me to bake you a chocolate
cake rather than bake you a lemon cake.
But this seems prima facie puzzling, if we think of reasons simply as considerations
that favor some action. If the fact that it's your birthday is a reason for me to
bake you a chocolate cake rather than bake you no cake at all, then it seems that
it has to favor baking you a chocolate cake. But if it favors baking you a chocolate
cake, why wouldn't it be a reason to bake you a chocolate cake rather than bake
you a lemon cake? Sinnott-Armstrong concludes that reasons are always reasons
for one thing rather than another: to know whether some fact is a reason for some
action, we have to know `Rather than what?'. That is, `reason' expresses a relation
21
with an argument place for an alternative, or set of alternatives, relative to which
the consideration is a reason for the action. This easily explains how (1) and (2)
could both be true: the `rather than' clause makes explicit the alternative relative
to which the fact that it's your birthday is or is not a reason for me to bake you a
chocolate cake.
This argument for contrastivism, however, is too fast. First, contrastivism holds
that all reasons are contrastive. But most reason ascriptions are not explicitly
contrastive, unlike (1) and (2). So one way we may resist contrastivism, in the face
of Sinnott-Armstrong's observation, is to hold that `reason' is ambiguous between
a contrastive sense and a non-contrastive sense. The contrastive sense is used in
claims like (1) and (2), while the non-contrastive sense is used in non-contrastive
ascriptions, like `The fact that it's your birthday is a reason to bake you a chocolate
cake'.
I think this view is unattractive, because `reason' fails a standard test for am-
biguity, which Schaer (2007) has called coordination across conjunction. `Reason'
fails this test for ambiguity because sentences like the following are perfectly ap-
propriate:
(3) The fact that you sprained your ankle is a reason to wear your brace, and to
lift weights rather than run.
Notice that there is just one occurrence of `reason' here; this suggests that `reason'
latches on to both conjuncts, `to wear your brace' and `to lift weights rather than
run'. The rst conjunct, though, is non-contrastive, while the second is explicitly
contrastive. If the single occurrence of `reason' latches on to both conjuncts, that is
evidence that `reason' is univocal, whether it takes a contrastive or a non-contrastive
complement. So sentences like (3) are a problem for the ambiguity view.
1
Ruling out the ambiguity view, though, does not yet establish contrastivism.
We can still try to oer non-contrastive analyses of explicitly contrastive reason
ascriptions like (1) and (2). If we can analyze these sentences in terms of non-
contrastive reasons, then they will not cause a problem for the non-contrastivist.
A tempting, but ultimately problematic, idea is to give an analysis in terms of
the strength of reasons, like one of the following:
1
See Schaer (2007), p. 396.
22
RT-1: r is a reason toA rather thanB ir is a stronger reason to A than it is to
B.
RT-2: r is a reason to A rather than B i r is either (i) a stronger reason to A
than it is to B, or (ii) r is a stronger reason not to B than it is not to A.
Either of these analyses would let us explain how (1) and (2) could both be true
without resorting to contrastivism. Since the fact that it's your birthday is a
stronger reason to bake you a chocolate cake than to bake you no cake at all,
(1) is true, according to either analysis. But since this fact is an equally strong
reason to bake you a chocolate cake and to bake you a lemon cake, (2) is also true,
according to either analysis.
While these analyses do seem natural, I think both face a serious problem. The
problem is that they require the phrase `rather than' to function very dierently in
reason ascriptions than it functions in other contexts. In ordinary uses of `rather
than' like, `I want pizza rather than salad', `Bob went to the store rather than to the
gym', and `Two plus two equals four rather than ve', `rather than' seems to mean
something along the lines of `and not', perhaps with an implicature that the two
things being contrasted are somehow especially relevant.
2
But according to RT-1
and RT-2, `rather than', as used in reason ascriptions, does not mean anything like
`and not', so they seem to rely on an ad hoc treatment of `rather than' in reason
ascriptions. This is a serious cost for these proposals.
Some writers, in objecting to contrastivism about explanation and about knowl-
edge, have given non-contrastivist analyses of explicitly contrastive claims on which
`rather than' does mean `and not'.
3
Consider the following two proposals, which
adopt this strategy, and dier only in the scope of the `not':
RT-3: r is a reason toA rather thanB ir is a reason toA andr is not a reason
to B.
2
`Rather than' as used in preference ascriptions may seem to better t with RT-1 and RT-2:
`I prefer cake rather than pie' might seem to mean that I like cake more than I like pie. But I
think a better analysis here is the following: I would choose cake and not choose pie, when those
are the alternatives. In ordinary cases, I would make this choice because I like cake more than I
like pie, but this need not be part of the analysis of the `rather than' ascription.
3
See Ruben (1987); Temple (1998) for this sort of strategy in resisting contrastivism about
explanation. See Schaer (2008) for arguments against this kind of analysis of `rather than'
knowledge ascriptions, and Rickless (fc) for further discussion.
23
RT-4: r is a reason toA rather thanB ir is a reason toA andr is a reason not
to B.
Either of these proposals lets us explain how (1) and (2) could both be true, without
resorting to contrastivism. Since the fact that it's your birthday is both a reason
to bake you a chocolate cake and a reason to bake you a vanilla cake, (2) is true
according to either proposal. And since this fact is not a reason to bake you no cake
at all, and in fact a reason not to bake you no cake at all, (1) is also true according
to either proposal.
Since there are reasonable, independently motivated non-contrastivist expla-
nations of the data provided by (1) and (2), Sinnott-Armstrong is too quick in
concluding on this basis that reasons are contrastive. What ascriptions like (1)
and (2) do show, however, is that non-contrastivists need to say something about
these explicitly contrastive ascriptions; I've argued in this section that they should
take some version of the popular non-contrastivist strategy, treating `rather than'
as meaning `and not'. But now that we've seen this, we're in a position to give a
stronger argument for contrastivism.
2.2 A Stronger Argument
In this section I'll show that both RT-3 and RT-4 are problematic. This gives us
good reason to reject non-contrastive accounts of reasons in favor of contrastivism.
At the end of the section, I'll consider an objection to my argument, and argue that
it fails.
2.2.1 The argument
Suppose that I need to get to campus sometime today, but live twenty miles away.
Further, suppose I'm out of shape and don't want to wear myself out getting there.
Now consider the following reason ascriptions:
(3) The fact that campus is twenty miles away is a reason to drive to campus
rather than bike there.
(4) The fact that campus is twenty miles away is a reason to bike to campus
rather than run there.
24
Both of these sentences are intuitively true, given the set-up of the case. But, as
I'll now show, this causes problems for the non-contrastivist.
First, notice that it follows on both RT-3 and RT-4 that if `r is a reason for
A rather than B' is true, then r is a reason for A.
4
So on both proposed analyses,
(5) follows from (4):
(5) The fact that campus is twenty miles away is a reason to bike to campus.
Since (5) follows from (4) on both proposals, and since (4) is true, the non-
contrastivist is committed to the truth of (5).
Now consider RT-3, which says that we should analyze `r is a reason toA rather
than B' as saying that r is a reason to A and r is not a reason to B. According to
this view, (6) follows from (3):
(6) The fact that campus is twenty miles away is not a reason to bike to campus.
The problem, of course, is that (5) and (6), both of which the non-contrastivist who
adopts RT-3 is committed to, are inconsistent. So this analysis of `rather than'
ascriptions fails.
Now consider RT-4, on which `r is a reason toA rather thanB' means thatr is
a reason toA andr is a reason not toB. This view also leads to a problem, though
not to an outright contradiction. For on that proposal, (7)|instead of (6)|follows
from (3):
(7) The fact that campus is twenty miles away is a reason not to bike to campus.
The non-contrastivist who adopts RT-4 is thus committed to both (5) and (7);
that is, she is committed to saying that the fact that campus is twenty miles away
is both a reason to bike to campus and a reason not to bike to campus. But this
is an implausible result. So the second proposal, while not leading to an outright
contradiction, does give us an implausible result.
4
This also plausibly follows on RT-1, since if r is a stronger reason for A than it is for B, it
is surely a reason forA. As an anonymous referee points out, this does not follow on RT-2. But,
as I argued in section 1, RT-2 (as well as RT-1) relies on an idiosyncratic treatment of `rather
than' in reason ascriptions.
25
2.2.2 Response: Denying Exclusivity
The non-contrastivist may question the implausibility of the result, that the fact
that campus is twenty miles away is both a reason to bike there and a reason not
to bike there. If this result isn't really troubling, then the argument against RT-4
is no good. So why think it is troubling?
One reason that many people will nd this result implausible is that it violates
the following principle:
Exclusivity: For all facts r, agents s, and actions A, if r is a reason for s to A,
then it's not the case that r is also a reason for s not to A.
Philosophers who seem to accept this principle include Nagel (1970); Raz (1999);
Crisp (2000). If Exclusivity is true, then the argument I've given shows that none
of the non-contrastivist analyses of `rather than' ascriptions that I've considered so
far could be correct.
But in fact there's reason to question Exclusivity. Jonathan Dancy has argued
that the principle is false by appealing to cases like the following.
5
Suppose I love
to talk to pretty girls, but hate to be snubbed. And suppose that, unfortunately,
pretty girls are likely to snub guys like me. Then the fact that the girl across the
bar is so pretty is a reason for me to go talk to her, since I love to talk to pretty
girls. But it's also a reason for me not to go talk to her, since I hate to be snubbed,
and since pretty girls are likely to snub guys like me. If this is the right way to
describe this case (and I think it is), then Exclusivity is false.
Further, common theories of reasons seem committed to the falsity of Exclu-
sivity. On a simple desire-based theory of reasons, for example, when you have a
desire thatA-ing would help promote, andr explains why this is so,r is a reason to
A.
6
So if you have two desires, such thatr explains both whyA-ing would promote
one of them and why notA-ing would promote the other, this kind of theory entails
that r is both a reason for you to A and a reason for you not to A. Similarly, on
a standard value-based theory, all we need is a case in which there are two values
such that r explains both why A-ing would promote or respect one of them and
why not A-ing would promote or respect the other.
7
5
See Dancy (1993), p. 62 for discussion.
6
See Schroeder (2007), for example.
7
See Part (2011); Scanlon (1998); Moore (1912), for example.
26
This lets us see where Exclusivity goes wrong. It is common ground between
desire-based and value-based theories that our reasons for action are provided or
explained by various objectives|desires, on the desire-based theory and values on
the value-based theory (and perhaps both on a hybrid theory). When a factr helps
explain why myA-ing would promote or respect one of these objectives,r is a reason
for me to A, which is provided by that objective.
8
The problem with Exclusivity
is that on most theories, there are multiple objectives that can provide reasons|
agents have multiple desires, and several kinds of values are worth promoting or
respecting. In the pretty girl case above, for example, I have two relevant desires:
rst, a desire to talk to pretty girls, and second, a desire not to be snubbed. The
rst desire explains why the fact that she's so pretty is a reason to go talk to her:
this fact explains why doing so would promote my desire to talk to pretty girls.
Similarly, this same fact explains why not going to talk to her would promote my
desire to not be snubbed. It is easy to construct a similar case in which multiple
values, rather than multiple desires, are involved.
What this shows is that popular theories of reasons allow for dierent objectives
to explain why one and the same fact can be a reason toA and not toA. But they
do not allow for one and the same objective to explain why one and the same fact
is both a reason to do and not to do one and the same action. That's because one
fact r can't explain both (i) why A-ing would promote or respect an objective o,
and (ii) why not A-ing would also promote or respect o.
9
So the following weaker
principle is very plausibly true, even if Exclusivity is false:
Restricted Exclusivity: For all facts r, agents s, actions A, and objectives o, o
cannot explain both why r is a reason for s to A and why r is a reason for s
not to A.
8
For discussion of objectives providing reasons, see Moore (1912) (though he talked about
rightness rather than reasons); Nagel (1970); Anderson (1993); Scanlon (1998); Finlay (2001,
2006); Schroeder (2007); Wedgwood (2009); Part (2011). For analyses of reasons in terms of
explanation, see Toulmin (1950); Finlay (2001, 2006); Searle (2001); Broome (2004); Schroeder
(2007). I will return to this idea in Chapter 4.
9
To illustrate this, consider Schroeder (2007)'s Hypotheticalism. He holds that r is a reason
for s to A i r explains why A-ing would promote p, where p is the object of one of s's desires,
and where promoting p is just making p more probable. But it would be strange to think that r
could explain both why A-ing would make p more probable, and why not A-ing would also make
p more probable. Similarly, it doesn't seem that one fact could explain both why A-ing would
respect, say, the demands of justice, and why notA-ing would also respect the demands of justice.
27
What this principle rules out is that one and the same objective can explain why
one and the same fact is both a reason to A and a reason not to A. This is
overwhelmingly plausible. The cases that seem to violate Exclusivity, like the
pretty girl case above, do not violate this principle. And, as I argued above, popular
theories of reasons, though they allow for violations of Exclusivity, do not allow
for violations of Restricted Exclusivity.
Crucially, this principle is enough for my argument, because the joint truth of
(5) and (7) does violate Restricted Exclusivity. The objective that provides,
or explains, both of them is my desire not to wear myself out getting to campus.
This explains both why the fact that campus is twenty miles away is a reason to
drive rather than bike, and why this fact is a reason to bike rather than run. And
since these `rather than' ascriptions, according to the non-contrastivist who adopts
RT-4, are to be analyzed partly in terms of the non-contrastive reasons mentioned
in (5) and (7), this single desire also provides the reasons ascribed in both (5) and
(7). Thus, RT-4 leads to an implausible violation of Restricted Exclusivity.
2.3 Contrastivism
In this section I'll develop a contrastivist account of reasons, and then use that
account to explain the problematic case from the last section.
2.3.1 A contrastive account reason claims
Contrastivism diers from traditional non-contrastive theories by claiming that the
reason relation includes an argument place for sets of alternatives. So the relation
holds between (at least) (i) some fact r which is the reason, (ii) the particular
alternative A which the fact is said to be a reason for, and (iii) the relevant set of
alternatives out of which r is said to be a reason for A.
10
In an explicit `rather than' ascription like `The fact that today is your birthday
is a reason to bake you a chocolate cake rather than not baking you a cake', the
alternatives are provided explicitly:fbake you a chocolate cake, don't bake you a
cakeg.
11
These kinds of ascriptions, according to contrastivism, provide the model
10
I'm ignoring agents here.
11
The set of alternatives might be larger than just two members, but since I've been dealing
with `rather than' ascriptions here, I'll mostly limit my discussion to two-member sets.
28
for all reason ascriptions. But most reason ascriptions aren't like this; most are
what I'll call `bare ascriptions', like `The fact that it's your birthday is a reason to
bake you a chocolate cake'. Since the relation that serves as the semantic value of
reason ascriptions|say,R(r;A;Q)|has an argument place for a set of alternatives
Q, this set must be provided somehow before we can evaluate the ascription. The
most natural way to develop the theory, and the way I'll adopt, is to let the set
of alternatives be provided by the context of utterance.
12
In a context in which a
reason ascription is made, some particular set will be relevant. One straightforward
way in which a set of alternatives might count as the relevant one is by including
the options under discussion, though there are likely other ways.
13
Since an explicit `rather than' ascription gives the set of alternatives explicitly,
these ascriptions will have a stable content across contexts (assuming there are no
other context-sensitive terms). But since bare ascriptions have the set of alterna-
tives provided by context, their content will be shifty across contexts. Here are the
semantic principles I propose:
Explicit Ascriptions: `r is a reason to rather than ' is true in context c
14
i
r is a reason to out off, g.
Bare Ascriptions: `r is a reason to ' is true in context c i r is a reason to
out of Q, where Q is the relevant set of alternatives in c.
Bare Against Ascriptions: `r is a reason not to ' is true in context c i r is a
reason not to out of Q, where Q is the relevant set of alternatives in c.
15
12
We might develop a theory on which it's provided instead by the context of assessment, or
perhaps by some features of the agent's situation (though to count as a contrastivist theory, it
should not always simply be the set of alternatives which it is possible for the agent to perform).
Sinnott-Armstrong (2004, 2006) defends a view on which there's no saying which set is relevant for
evaluating any particular bare reason ascription, which leads him to adopt Pyrrhonian skepticism,
and thus to refuse to evaluate any bare ascription as either true or false.
13
One common contrastivist idea is that intonational stress can help x the set of alternatives. If
I say `You have a reason to buy some milk', that suggests a set of alternatives likefbuy milk, buy
juiceg, whereas if I say `You have a reason to buy some milk', that suggests a set of alternatives like
fbuy milk, steal milkg. See Rooth (1992); Dretske (1970); Schaer (2005b, 2008) for discussion
of stress and the role of alternatives.
14
Note that `in context c' doesn't show up on the right-hand side of the biconditional. That's
because, as I said, explicit ascriptions are not context-sensitive.
15
If we can think of reasons against an option A simply as reason for:A, then this clause is
just a special case of Bare Ascriptions. I agree that this is an attractive view, though things
are slightly more complicated for the contrastivist, since:A will not necessarily be in the same
set of alternatives as A|sets of alternatives need not be exhaustive in this sense. And it's not
29
Now I'll show how to use this account to explain the problematic case from the last
section.
2.3.2 A contrastive solution
One way to think about the problem facing the non-contrastivist is that, no matter
what analysis of `rather than' ascriptions we considered, we always derived thatr is
a reason toA from `r is a reason toA rather thanB'.
16
This result, when combined
with claims aboutB that we could also derive (and that are independently plausible
anyway) led either to outright contradiction or to otherwise implausible results. So
the culprit seems to be this inference:
RT: If r is a reason to A rather than B, then r is a reason to A.
The trouble is that this inference seems very hard to reject for the non-contrastivist.
The contrastivist, on the other hand, can not only reject this inference, but can
explain why it seems so natural. Note rst that RT is, of course, a non-contrastivist
principle|there is no mention of contrasts on the right-hand side. So I suggest that
we replace it with this:
CRT: If r is a reason to A rather than B, then r is a reason to A out offA, Bg
and r is a reason not to B out offA, Bg.
The second half of the consequent is perhaps less obvious than the rst, but I think
it is nevertheless very compelling. Given thatA andB are mutually exclusive (why
else would we use `rather than'?), B-ing precludes A-ing. So if r is a reason to A
out offA, Bg, it is very plausible that it's also a reason not to B out of this set.
17
Importantly, given Bare Ascriptions, we can now see why RT seems so natu-
ral: it's because the default interpretation of the consequent expresses a truth. The
antecedent, `r is a reason toA rather thanB', makes salient the set of alternatives
clear how to make sense of a reason to:A out of a set of alternatives that doesn't contain:A.
I return to this point in Chapter 4. Further, some philosophers have recently given accounts on
which reasons against A cannot simply be thought of as reasons for:A. See Greenspan (2005),
for example.
16
Again, RT-2 does not validate this inference, which is an advantage of that proposal. But it
does rely on a problematic treatment of `rather than' in reason ascriptions.
17
This shows that my view does respect the fact that `rather than' means something like `and
not': relative to the set of alternativesfA, Bg, when `r is a reason to A rather than B' is true, r
is a reason to A and a reason not to B.
30
fA, Bg. And according to Bare Ascriptions, when we interpret the consequent,
`r is a reason to A' relative to this set, it means that r is a reason to A out offA,
Bg. Finally, according to CRT, this will be true as long as `r is a reason to A
rather than B'|the antecedent of RT|is true.
The mistake, though, is to move from the truth of `r is a reason to A' in the
context set up by the antecedent of RT to the truth of this claim in every context.
If the relevant set of alternatives is not the one provided by the antecedent,fA,
Bg, then we have no guarantee that `r is a reason toA' will be true, relative to this
new set.
This puts us in a position to see how the contrastivist can solve the puzzle.
First, here are the relevant sentences, re-written to make the contrasts explicit.
(3*) The fact that campus is twenty miles away is a reason to drive rather than
bike (out offdrive, bikeg).
(4*) The fact that campus is twenty miles away is a reason to bike rather than run
(out offbike, rung).
(5*) The fact that campus is twenty miles away is a reason to bike (out offbike,
rung).
(7*) The fact that campus is twenty miles away is a reason not to bike (out of
fdrive, bikeg).
The truth of the `rather than' claim (3), along with CRT, tells us that (7) is true
in a context in which the relevant set of alternatives isfdrive, bikeg, so we interpret
(7) as (7*), above. But in this context, by Bare Ascriptions, (5) would be true
only if the fact that campus is twenty miles away is a reason to bike relative to this
same set:
(5') The fact that campus is twenty miles away is a reason to bike out offdrive,
bikeg.
But this is false. So in the context in which (7) is true, (5) is false.
The truth of the `rather than' claim (4), along with CRT, tells us that (5) is
true in a context in which the relevant set of alternatives isfbike, rung, so we
interpret (5) as (5*) above. But in this context, by Bare Against Ascriptions,
(7) would be true only if the fact that campus is twenty miles away is a reason not
to bike relative to this same set:
31
(7') The fact that campus is twenty miles away is a reason not to bike out offbike,
rung.
But this is false. So in the context in which (5) is true, (7) is false.
Thus, there's no one context in which both (5) and (7) are true|at least not
one in which both reasons are provided by my desire not to wear myself out getting
to campus.
18
It's true, of course, that (3) and (4) are true in the same contexts,
but that's not puzzling|no one should deny that these `rather than' claims can be
jointly true. The key to the contrastivist solution is, essentially, treating all reason
ascriptions as (at least implicitly) `rather than' ascriptions|as relative to sets of
alternatives.
2.4 Negative Reason Existentials
So far I've shown that contrastivism can solve problems facing non-contrastive
theories having to do with `rather than' ascriptions. In this section, I'll show that
contrastivism can solve a puzzle about a dierent kind of reason claim, negative
reason existentials. These are claims like `There's no reason to cry over spilled
milk'. These sorts of claims are also very common, and play an important role in
arguments against various theories in ethics, epistemology, and practical reasoning:
some theory implies that there is a reason to, but intuitively there is no reason to
, so the theory is false. Schroeder (2007) argues, though, that our intuitions about
negative reason existentials are not to be trusted. After presenting the puzzle, I will
argue that a contrastive solution is more attractive than Schroeder's own solution.
2.4.1 The puzzle
Consider the following famous case.
The Book Thief. You see Tom Grabit run out of the library, pull a
book from under his coat, cackle gleefully, and then run away.
19
18
What is the status of Restricted Exclusivity? It's a non-contrastivist principle, so I need
to reject it. But I relied on it in my argument against the non-contrastivist. I claim that the
non-contrastivist is committed to this principle, which is why the case I presented in section 2 is
problematic for her. The contrastivist should replaceRestricted Exclusivity with a contrastive
version, which says that one and the same objective can't explain why one and the same fact is a
reason to do and not to do one and the same action relative to one and the same set of alternatives.
19
Lehrer and Paxson (1969).
32
Intuitively, you have a reason to believe that Tom Grabit stole the book. But now
consider this variation:
The Twin Book Thief. You see Tom Grabit run out of the library,
pull a book from under his coat, cackle gleefully, and then run away.
Just as you begin to form the belief that Tom stole the book, your
friend tells you about Tom's twin brother, Tim. Tom and Tim are
indistinguishable, even to their mother.
Imagine that we've hauled Tom into the police station and charged him with stealing
the book. He protests, \Wait! I have a twin brother! You have no reason to
think I'm the one who stole the book!". His protest certainly seems warranted. So
intuitively, in this case, you have no reason to believe that Tom stole the book. The
reason provided by your visual evidence is undercut. And, it seems, when a reason
is undercut, it's just not a reason anymore. But Schroeder argues that our intuitive
judgment that there is no reason for you to believe that Tom stole the book in this
case is mistaken. For consider the following variation:
TheTripletBookThief. You see Tom Grabit run out of the library,
pull a book from under his coat, cackle gleefully, and then run away.
Just as you begin to form the belief that Tom stole the book, your friend
tells you about Tom's two identical siblings, Tim and Tam. The triplets
are indistinguishable, even to their mother.
Intuitively you have even less reason to believe that Tom stole the book in this case
than you did inThe Twin Book Thief. But if you have even less reason in this
case, you couldn't have had no reason in the previous case. So our intuition that you
had no reason is mistaken. This is the puzzle about negative reason existentials:
some of them are intuitively true, though we also seem to be committed to the
existence of the very reasons that are said not to exist.
20
2.4.2 The pragmatic solution
Schroeder oers a pragmatic explanation of this puzzling data. The rst observa-
tion is that some reasons are very weak, and very weak reasons are not generally
20
See Schroeder (2007), p. 93.
33
relevant in our deliberation about what to do. The second observation is that|
following Grice (1989)|talking about irrelevant things is conversationally inappro-
priate. Thus, we are reluctant to assume that people who tell us about reasons are
talking about very weak reasons, since this would be conversationally inappropriate.
Instead, we make the presumption that they mean to be talking about relatively
weighty reasons|their claim has this implicature. This presumption is reinforced
by making a bare existential claim, which is less informative than a claim specify-
ing what the reason is. For a bare existential claim to be suciently informative,
the reason must be relatively weighty. But since the reasons in question are not
weighty, the implicature is false, and so the claim seems false. Since the reason
existential `You have a reason to believe that Tom stole the book' seems false, its
negation, the negative reason existential `You have no reason to believe that Tom
stole the book', in The Twin Book Thief, sounds true.
This explanation predicts that saying what the reason is should make it sound
less bad, since this will make the claim more informative. It also predicts that saying
that the reason is a really weak one will make it sound less bad, since that would
cancel the implicature that the reason is relatively weighty. And these predictions
are plausibly correct. To illustrate the idea, Schroeder (2007) uses the example,
`There is a reason to eat your car'. This seems obviously false. But if I say `The
fact that eating your car will give you the daily recommended amount of iron is
a reason to eat your car', that sounds better. And it sounds better still if follow
this up with `...but it's a really weak reason'. This supports the hypothesis that
there are pragmatic factors at work here. So Schroeder argues that there are in
fact reasons in these cases, and thus that the negative reason existentials are false,
but oers an explanation of why they sound true anyway.
But for this case to actually support Schroeder's pragmatic solution, the ex-
planation for why the negative reason existential initially seems true must be the
reason's low weight. But I think there is a better explanation. It's not that we
ignore the reason because it's so weak; rather, I think that we simply miss the
reason because it's so non-obvious. Now, I agree that even once we acknowledge
it, we'd do better to ignore it, since it's so lightweight. But the point is that the
weight of the reason is not the explanation for why the negative reason existential
sounds true. The amount of iron you could get from eating your car is simply not
34
something we would ever think about, unless someone pointed it out to us. Conr-
mation for this explanation over Schroeder's comes from the fact that the dierence
in acceptability between (8) and (10) is more striking than the dierence between
(8) and (9):
(8) There's a reason to eat your car.
(9) There's a reason to eat your car, but it's a really weak one.
(10) The fact that it contains your daily recommended amount of iron is a reason
to eat your car.
If Schroeder's explanation were correct, we should expect a much stronger dierence
between (8) and (9) than between (8) and (10), since (9) explicitly cancels the
implicature that supposedly makes (8) sound false. But on my non-obviousness
explanation, we should expect a stronger dierence between (8) and (10). I think
the data here conrms my explanation over Schroeder's.
21
The important point for the Tom Grabit case is that the supposed reason, that
you saw someone who looked just like Tom run out of the library with the book,
isn't at all non-obvious. So we need a dierent explanation for the Tom Grabit case.
I'll argue now that contrastivism oers an attractive explanation. The contrastivist
explanation is also attractive because it holds that the intuitively true negative
reason existential|Tom's compliant that we have no reason to think that he stole
the book|is actually true, rather than trying to explain away that intuition.
2.4.3 A contrastivist solution
The puzzling data about negative reason existentials is that there are some claims
like `There is no reason to' which are intuitively true, though we seem committed
to there actually being reasons to . Thinking about The Twin Book Thief,
`There is no reason to believe that Tom stole the book' seems true. But when we
consider The Triplet Book Thief, it seems that there is even less reason to
believe that Tom stole the book in this case than in the previous one. So there
couldn't have been no reason before, contrary to our intuitions about the negative
21
I agree that `The fact that it contains your daily recommended amount of iron is a reason to
eat your car, but it's a really weak one' sounds (a bit) better than (10). But I can explain that
by pointing out that this claim is simply more informative than (10).
35
reason existential. Once we re
ect on this fact, we seem to be committed to the
claim that there is a reason to believe that Tom stole the book, after all.
Here's how to explain this phenomenon using contrastive reasons. In the context
given by theThe Twin Book Thief, the most natural contrast to believing that
Tom stole the book is believing that Tim, Tom's twin brother, stole the book, since
this alternative has just been made relevant. And it's true that there is no reason
to believe that Tom stole the book rather than believing that Tim stole the book.
This is why it's true to say `There's no reason to believe that Tom stole the book',
when we're considering The Twin Book Thief. But when we consider The
Triplet Book Thief, and notice that there's even less reason to believe that
Tom stole the book in this case than in the previous case, and so want to accept
`There is a reason to believe that Tom stole the book', the contrast to believing
that Tom stole the book is most plausibly not believing (that is, failing to believe)
that Tom stole the book.
Here's why. We need some reason that exists in bothThe Twin Book Thief
and inTheTripletBookThief, whose weight we can compare across the cases.
Since Tam wasn't in the picture inTheTwinBookThief, we can't have in mind
any reasons involving Tam. So the only two plausible candidates, on a contrastive
framework, are a reason to believe that Tom stole the book rather than believing
that Tim stole it and a reason to believe that Tom stole the book rather than not
believing that Tom stole it. But there's no reason to believe that Tom stole the
book rather than believing that Tim stole it in either case. So that only leaves a
reason to believe that Tom stole it rather than not believing that Tom stole it. And
there very plausibly is this reason in both cases, and it's very plausibly weaker in
The Triplet Book Thief than it is in The Twin Book Thief.
Any theory of reasons has to allow that some reasons are weightier than others,
and that some consideration might be a weightier reason for one thing than it is
for another thing. Contrastivism is no dierent|in The Twin Book Thief, the
fact that you saw (what looked like) Tom run out is a weightier reason to believe
that he stole the book out offbelieve that Tom stole the book, don't believe that
Tom stole the bookg than it is in The Triplet Book Thief.
22
This solution
22
One important thing to keep in mind here is that according to contrastivism, the weight of
a reason to out of Q does not depend on the weight of any non-contrastive reasons to and
non-contrastive reasons for the other alternatives in Q. That's because, of course, there just are
no non-contrastive reasons. This does raise the question of how best to think about the weight of
36
is available to the contrastivist because she holds that reasons are relativized to
possibly non-exhaustive sets of alternatives. One fact can fail to be a reason to
believe that p relative to one set, be a reason to believe that p relative to a second
set, and be an even weightier reason to believe that p relative to a third set.
Contrastivism has a plausible explanation of why the negative reason existential
is literally true, as well as why the positive reason existential that we accept after
considering the comparative claim about the weight of reasons is true. The reason
we're saying does not exist is a reason to believe that Tom stole the book out of
fbelieve that Tom stole the book, believe that Tim stole the bookg, while the reason
whose weight we're comparing across cases is a reason to believe Tom stole the book
out offbelieve Tom stole the book, don't believe Tom stole the bookg.
23
On a non-
contrastive theory, on the other hand, these claims are incompatible. So this is
an advantage for contrastivism: we need not bite the bullet and admit that one
of our intuitive judgments is false. And moreover, contrastivism is independently
well-motivated by the puzzle about reason against ascriptions.
2.5 Looking Forward
I've argued in this chapter that contrastivism about reasons gains some support
from considerations of our talk about reasons. The solutions to the puzzles depend
on the contrastive semantics that t very naturally with contrastivism. But in the
next chapter, I'll consider how far the argument in this chapter actually takes us.
a contrastive reason, but virtually any theory needs to give some account of weight, and few do.
I'll rely on an intuitive understanding.
23
Some people deny that the negative reason existential even seems true. I think that framing
things as I did above, by imagining Tom protesting when we accuse him of stealing the book,
makes it even more intuitive that the claim is true. My hypothesis for their resistance is that they
have in mind the setfbelieve that Tom stole the book, don't believe that Tom stole the bookg.
And contrastivism predicts that, relative to this set, there is a reason to believe that Tom stole
the book, so the negative reason existential should seem false.
37
Chapter 3
Favoring
In the last chapter, I argued that certain features of our talk about reasons supports
contrastivism. Relativizing reason ascriptions to sets of alternatives solves puzzles
that face non-contrastive theories. In this chapter, though, I'll consider how far
that argument actually takes us. The puzzles concerned reason ascriptions, and the
solutions came from a contrastive account of those ascriptions. It's very natural
to think that this is strong support for the thesis that reasons are fundamentally
contrastive. If reason ascriptions are contrast-sensitive, then it seems obvious that
the reason relation, which presumably serves as the semantic content of `reason',
must be contrastive as well. But this is an apparently radical thesis, given that
reasons are traditionally taken to be reasons for things simpliciter. We might hope,
then, to avoid committing ourselves to this sort of theory prematurely.
3.1 Why Resist Contrastivism?
As I've emphasized, reasons are traditionally taken to be reasons for things sim-
pliciter, independently of the contextually relevant alternatives. So the rst reason
why we might want to resist contrastivism is simply that it's radical. Nearly every-
one who writes about reasons seems to assume a non-contrastive picture (usually
implicitly, of course). So adopting contrastivism would require rejecting|or at least
seriously reinterpreting|these writers' claims.
Second, contrastivism might seem mysterious. Scanlon (1998) says that reasons
are considerations that count in favor of the things they are reasons for, and many
other writers adopt this gloss. But if we accept this gloss, contrastivism might seem
puzzling: how could the relevant set of alternatives matter for whether or not some
38
consideration counts in favor of an action or attitude? The availability of some
alternative can arguably alter whether or not some consideration is a reason to do
something. But contrastivism makes a more surprising claim than this. Reasons
are only reasons relative to some set of alternatives, whether or not that set includes
all of the available alternatives. So it's not simply the other available alternatives
that matter in determining whether some consideration is a reason for an action:
it also matters what we're comparing the action to.
Third, many theories in ethics and epistemology appeal to normative reasons,
and some have even taken the notion of a reason to be the fundamental normative
notion, reducing all other normative concepts to reasons.
1
But if we complicate
the fundamental normative relation by adding an argument place for a set of alter-
natives, we might worry that this will complicate our analyses of other normative
concepts. Even if, as I argued in Chapter 1, contrastivism about reasons doesn't re-
quire us to be contrastivists about these other concepts, the analyses of these other
concepts might be more complicated if we adopt contrastivism about reasons.
3.2 Shallow Contrastivism
The arguments for contrastivism in Chapter 2 concerned reason ascriptions. So if
we want to avoid contrastivism in the face of those arguments, a natural strategy
is to develop an theory which can mimic the contrastivist's solution to the puzzles
I presented in the last chapter, while retaining a more traditional non-contrastive
metaphysics to avoid some of the complications I mentioned in the last section. To
do this, we need a theory that oers a contrastive account of reason ascriptions but
accepts a non-contrastive metaphysics of the important normative favoring relation
that philosophers are concerned with when they theorize about reasons, and in
terms of which we analyze other important normative concepts. Call this kind of
theory shallow contrastivism; call the more radical theory deep contrastivism.
The basic idea is simple. Rather than tracking when some consideration does or
does not favor an action simpliciter, our reason ascriptions track the degree to which
the consideration favors the action as compared to the degree to which it favors the
other relevant alternatives. So at the bottom we have a non-contrastive favoring
1
See, for example, Raz (1999); Schroeder (2007); Scanlon (1998). See, for example, Broome
(2004); V ayrynen (2010) for skepticism about this project.
39
relation: facts favor certain actions (or attitudes) to a certain degree simpliciter.
When a factr favors-ing more than in favors -ing, we can truly say `r is a reason
to rather than '. Or, if -ing is the contextually relevant alternative to -ing,
we can simply say `r is a reason to', though what this means is thatr favors-ing
more than it favors the contextually relevant alternatives.
This view arguably has some precedent in Alastair Norcross's work defending
consequentialism.
2
He says: \[O]ur (moral) reasons for choosing between alterna-
tive actions, institutions, etc. are essentially comparative, and correspond to the
comparative consequential value of the options". This quote suggests that Norcross
is actually a deep contrastivist about (moral) reasons. But he also says, in the next
sentence, \I might have a better reason for choosing to do A than to do B, and
better by a certain amount, but neither reason is either good or bad simpliciter".
3
This quote seems to t better with a shallow contrastivist view, on which consider-
ations favor alternatives with a certain weight, though we can only call them (good
or bad) reasons relative to some comparison class. Moreover, Norcross understands
his project as providing a contextualist linguistic theory for consequentialism, where
context provides some appropriate alternatives, and we can only say something is
good or bad as compared to these alternatives. But at bottom, he seems to assume
a traditional view of favoring.
Now I'll develop shallow contrastivism in more detail. First I'll outline a tradi-
tional non-contrastive account of the favoring relation, then I'll develop a contrastive
account of reason claims and pair the two to end up with shallow contrastivism.
3.2.1 A traditional theory of favoring
A traditional, non-contrastive view of reasons says that reasons are considerations
that favor certain actions with a certain weight or strength, some weightier than
others. Most people agree that it's a mistake to identify the weight of a reason
with anything as precise as a number, but for simplicity that's what I'll do here
(or at least represent it as a number). So we can think of the relation that a fact
stands in to the thing it favors as something like this: F (r;A;x). This means
that fact r favors action A with weight x. This is a relation that holds between
facts, actions, and weights independently of contextually relevant alternatives. We
2
See Norcross (1997, 2005a,b).
3
Norcross (2005a), p. 81.
40
can use this relation to obtain a ranking of alternatives based on how strongly r
favors them. So if we have F (r;A; 2) and F (r;B; 5), we would rank B above A
on the ranking of alternatives generated by r.
4
We can understand disfavoring as
favoring with a negative weight: F (r;A;5), for example.
5
This favoring relation,
according to shallow contrastivism, is the important normative relation that writers
who talk about reasons are really concerned with, and the relation in terms of which
we should try to analyze other normative concepts. That's why nearly everyone
glosses a reason as a consideration that favors some action or attitude (with a
certain weight).
3.2.2 A contrastive account of reason claims
The next step is to pair this idea with a contrastive account of reason claims. The
way deep contrastivism solves the puzzles is by making bare reason claims context-
sensitive: as the contextually relevant set of alternatives varies, the content of the
reason ascription shifts. The shallow contrastivist wants to mimic this semantics.
There's an apparent tension here. What we want is for the underlying favoring
relationF to hold independently of contextually relevant alternatives, but the con-
tent of the reason ascription to shift with those alternatives. So it looks like the
relation which serves as the semantic value of reason claims|the semantic value of
`reason'|has to include an argument place for sets of alternatives. But we want
the important normative relation, the favoring relation, not to include such an
argument place, so that it's a non-contrastive relation.
So what we have to do is reject the natural idea that instances of the underlying
favoring relation are the semantic content of reason ascriptions. Instead, I suggest
we posit an \intermediate" relation, R(r;A;Q), to serve as the semantic content
of reason ascriptions. Q is the relevant set of alternatives. I'll call R the `reason
relation', since it's the content of reason ascriptions. But all this means is that the
fundamental relation|the one we really care about in ethics, epistemology, and so
4
We can take more facts into consideration to obtain a ranking of how strongly the balance
of all the relevant considerations favor the dierent alternatives by performing some function on
the weights with which each fact favors the various alternatives. (This is essentially the picture
I sketched in Chapter 1 to show that contrastivism about `ought' does not require contrastivism
about reasons.)
5
There might be complications here, but since I don't ultimately want to defend shallow con-
trastivism, I won't focus on those here.
41
on|is not (what I'm calling) the reason relation, but is rather the favoring relation
F .
We can respect the fact that the favoring relationF is the fundamental relation
by letting R depend on F in the following way:
Shallow Reasons: R(r;A;Q) i there's some x such that F (r;A;x), and for all
B in Q not identical to A, there's some y such that (i) F (r;B;y), and (ii)
x>y.
Intuitively what this says is that r is a reason for A out of Q i r favors A more
than it favors any other alternative in Q. And then we say that `r is a reason for
A' is true in context c i R(r;A;Q), for the relevant set of alternatives Q in c.
6
So when philosophers say things like `Reasons are considerations that count in
favor of actions or attitudes', we have to do some mild reinterpretation: they mean
that reasons are considerations that favor things more than they favor the relevant
alternatives. This is not a very radical reinterpretation. When I say that r is a
reason to, this can still be relevant for determining what I'm justied in doing, or
what I ought to do. That's because my claim can only be true ifr really does favor
(non-contrastively) -ing, and does so more than it favors other relevant actions.
And our analyses of concepts like `ought' should technically proceed in terms of
favoring, not reasons. So instead of saying, for example, that you ought to when
you have most reason to , we can say that you ought to when-ing is the most
favored thing you can do.
7
6
There's one complication I want to address here. Some considerations intuitively don't stand
in the favoring relation to some actions|not even with a negative weight. For example, the fact
that 2+2=4 doesn't seem to favor or disfavor my drinking another cup of coee: there's no weight
x such that F (2 + 2 = 4, drink another cup of coee, x). One initially attractive idea is to say
that in these cases, the fact favors the action with weight zero. But this will give us odd results
in some cases. Suppose that r neither favors nor disfavors A, and slightly disfavors B. Then it
would be true, on the theory as I've developed it so far, to say `r is a reason toA', in a context in
which the set of alternatives isfA, Bg, since zero is greater than whatever negative number we
use for the weight with which r disfavorsB. But it's strange to say that r is a reason forA when
it's completely irrelevant to A.
I think there are a couple of ways to respond to this problem. We could say instead that when
r is irrelevant toA,r is just doesn't stand in the F relation toA at all. Then it wouldn't be true
to say that r is a reason to A. That would require complicating the analysis of R in terms of the
F relation to accommodate this change. A second solution is to bite the bullet and admit that, if
the salient set of alternatives isfA,Bg, andr disfavorsB, then it is a reason toA rather than B.
I don't think this is terribly implausible, and we could try to explain away the strangeness using
pragmatic considerations. For now, though, I'll just leave this issue aside.
7
The particular analysis doesn't really matter. The one I've mentioned allows us to avoid
contrastivism about `ought', since -ing has to be more favored than anything else you could do,
42
Shallow contrastivism can essentially adopt the semantic principles I gave in
Chapter 2, but state things in terms of the R relation, which can then be cashed
out in terms of the underlyingF relation. Nothing about the solutions to the puzzles
about reason claims depended on the relation which serves as the content of reason
ascriptions being the fundamental relation. So the fact that deep contrastivism can
solve puzzles is not (for all I've said so far) evidence for deep contrastivism over
shallow contrastivism.
3.2.3 Shallow contrastivism and Exclusivity
Before I evaluate shallow contrastivism, I want to point out one interesting com-
mitment of shallow contrastivism. It requires that Exclusivity is true, at least
in cases in which the relevant set of alternatives isf,:g. Here again is that
principle:
Exclusivity: If r is a reason for s to , then r is not also a reason for s not to .
Lots of people think that Exclusivity is true, so they won't see this as a problem
for shallow contrastivism. But I gave a case in the last chapter which I think is
plausibly a counterexample to this principle.
8
The fact that the girl across the bar
is so pretty is a reason to talk to her, since I love to talk to pretty girls. But it's also
a reason not to talk to her, since I hate to be snubbed, and pretty girls are likely to
snub guys like me. If this case, or one like it, is a counterexample to Exclusivity,
then shallow contrastivism has a problem.
Here's why. In the case I presented, the set of relevant alternatives is clearly
ftalk to her, don't talk to herg. For the fact that she's so pretty to be a reason to
talk to her out of this set, this fact has to favor talking to her more than it favors
not talking to her. And for it to be a reason not to talk to her out of this set,
it has to favor not talking to her more than it favors talking to her. But this is
impossible. If the shallow contrastivist's favoring relation is going to be anything
like the traditional understanding of the favoring relation, then at least this much
has to hold: if r favors -ing more than -ing, then it doesn't also favor -ing
not just more than some (potentially non-exhaustive) set of alternatives. A contrastive view of
`ought' (see Sloman (1970); Jackson (1985); Finlay (2009); Cariani (2009, fc); Snedegar (2012))
could say instead that you ought to when -ing is more favored than any of the relevant
alternatives.
8
See Dancy (1993), p. 62 for another.
43
more than-ing. So again, if Exclusivity is false, then shallow contrastivism is in
trouble.
9
But since lots of people accept Exclusivity, I'll now move on to argue
that the favoring relation itself is contrastive.
3.3 Favoring
In this section I'll assess the shallow contrastivist strategy for resisting deep con-
trastivism by discussing the favoring relation directly, rather than the reason rela-
tion.
10
I'll argue that the favoring relation itself is contrastive|facts favor actions
or attitudes only relative to sets of alternatives.
11
If the favoring relation is contrastive, that's strong support for deep contrastivism.
First, the claim that reasons are considerations that favor things is ubiquitous.
12
If considerations only favor things relative to sets of alternatives, it follows from
this claim that they're only reasons relative to sets of alternatives. Second, and
particularly relevant for this chapter, of course, is that if the favoring relation is
contrastive, then shallow contrastivism is false. We can't resist deep contrastivism
by appealing to a non-contrastive favoring relation.
3.3.1 Against non-contrastive favoring
The argument that favoring is always favoring relative to some set of alternatives,
rather than favoring simpliciter, begins from the following case. Suppose I have
three dinner invitations: Invitation A is for Armenian with Ara. Invitation B is for
burgers with Burt. Invitation C is for Chinese with Charlie. I love spending time
9
The deep contrastivist doesn't have this problem. The fact that the girl across the bar is so
pretty is a reason to talk to her out of this set, and a reason not to talk to her out of this same
set. But since the explanations for these two reason claims are dierent, we don't have a violation
of Exclusivity*.
10
Most people, including deep contrastivists, will think these are just the same relation, of
course. But this is exactly what the shallow contrastivist has to deny.
11
The argument I'll give is a development of a discussion in Ross (2006), Chapter 9. I should
note that Ross is careful to keep contrastivism about reasons at arm's length, claiming merely that
talking in terms of contrastive reasons and contrastive favoring avoids some diculties|diculties
that I'll develop into full-
edged objections|and is thus more convenient for his purposes.
12
Scanlon (1998) is particularly explicit about this, but nearly everyone writing about reasons
accepts it. Hieronymi (2005) argues that this is not the best way to think of reasons, because it
leads to the \wrong kind of reasons" problem. She argues that it's better to think of reasons as
considerations that bear on a question. But she also accepts that reasons favor the things they're
reasons for by bearing on certain questions.
44
with Ara, really like spending time with Burt, but can barely tolerate spending
time with Charlie. On the other hand, I love Chinese, really like burgers, but can
barely stomach Armenian.
The following claims are clearly true. The fact that Invitation B is for dinner
with Burt favors choosing Invitation B rather than Invitation C. And the fact that
Invitation B is for burgers favors choosing Invitation B rather than Invitation A. But
does the fact that Invitation B is for dinner with Burt favor choosing Invitation
B simpliciter? Or does the fact that Invitation B is for burgers favor choosing
Invitation B simpliciter? It might seem like the answer to both questions is clearly
`yes': after all, these two facts favor choosing Invitation B rather than Invitations
C and A, respectively. And how could they do that if they didn't favor choosing
Invitation B simpliciter?
But the following claims are also true. The fact that Invitation B is for dinner
with Burt does not favor choosing Invitation B rather than Invitation A|after
all, I like spending time with Ara more than I like spending time with Burt. And
similarly, the fact that Invitation B is for burgers does not favor choosing Invitation
B rather than Invitation C|after all, I like Chinese food more than I like burgers.
So perhaps these facts actually don't favor choosing Invitation B simpliciter. I could
do better with respect to my food preferences by not choosing Invitation B, and I
could do better with respect to my dinner companion preferences by not choosing
Invitation B.
But given the set up of the case, it's very plausible that all things considered,
I ought to choose Invitation B. If neither of the relevant facts|that Invitation B
is for burgers, and that it's for dinner with Burt|favor choosing B, then it looks
like nothing favors choosing Invitation B. But now we have an odd result: I ought
to choose Invitation B, though nothing favors choosing it. So it seems implausible,
if you're a non-contrastivist about favoring, to say that these facts don't favor
choosing Invitation B simpliciter.
At this point, we can conclude (with Ross (2006)) the following. There are
cases, like the one above, in which it's very hard to say whether some fact favors
some action simpliciter, but very easy to say whether it favors the action rather
than some other action. My suggestion is that this is because the favoring relation
is fundamentally contrastive. Cases like this one make this clear: to answer the
question `Does r favor -ing?', we have to know `Rather than what?'. In lots of
45
cases that philosophers have been interested in, the relevant consideration favors the
relevant action rather than any of the salient alternatives, so it's not so important
to specify. But in cases like the one above, it is important.
Here's one natural response. We should understand `rather than' claims as
simple comparisons about how much the consideration favors the actions. So when
I said that the fact that Invitation B is for burgers favors choosing it rather than
Invitation A, that was true. But what this means is just that this fact favors
choosing Invitation B more than it favors choosing Invitation A; we don't need a
fundamentally contrastive favoring relation to explain this.
I think this response is problematic. The fact that Invitation B is for burgers
does not favor choosing Invitation B rather than Invitation C|remember that I
like Chinese food more than burgers. If we extend the explanation oered in the
last paragraph in the natural way, we have to say that this is because the fact that
Invitation B is for burgers does not favor choosing it more than it favors choosing
Invitation C. But this fact doesn't seem to favor choosing Invitation C at all|it
doesn't have anything to do with Invitation C. So if we insist that it does favor
choosing Invitation B simpliciter, we have to say that despite this, it does not
favor choosing Invitation B rather than Invitation C, even though it doesn't favor
choosing Invitation C at all.
Here's a more sophisticated version of this response. I've actually misidentied
the relevant facts that do the favoring in this case. The relevant facts are the
following two more complicated facts:
(i) I like Chinese more than burgers and burgers more than Armenian and Invi-
tation A is for Armenian, Invitation B is for burgers, and Invitation C is for
Chinese.
(ii) I like spending time with Ara more than Burt and spending time with Burt
more than Chris and Invitation A is for dinner with Ara, Invitation B is for
dinner with Burt, and Invitation C is for dinner with Charlie.
Fact (i) favors choosing Invitation C to the highest degree, choosing B to a slightly
lower degree, and choosing A to a much lower degree. Fact (ii) favors choosing
Invitation A to the highest degree, choosing B to a slightly lower degree, and
choosing C to a much lower degree. As suggested above, `rather than' claims
simply give us comparisons between the degree to which the relevant fact favors
46
the options. The reason why I ought to choose Invitation B, all things considered,
is that it gets \the most" favoring, once we add up the amounts provided by facts
(i) and (ii).
This view does seem to avoid the problems I've presented so far, but it has
the counterintuitive consequence that strictly speaking, the fact that Invitation B
is for burgers doesn't favor choosing it rather than Invitation A. It's only the more
complicated fact (i) that favors choosing Invitation B rather than Invitation A.
Contrastivism about favoring doesn't have this consequence. We can avoid saying
that our ordinary claims and intuitions about what favors what are actually false,
or at least incomplete. The contrastivist can admit that these other facts about my
preferences serve as background conditions which enable the fact that Invitation B
is for burgers to favor choosing Invitation B rather than Invitation A.
13
There is also a sharper problem with this response. Schroeder (2007) points
out that a view analogous to this one (which he calls No Background Conditions)
has implausible consequences when combined with a very plausible Deliberative
Constraint. The Deliberative Constraint says that when we deliberate about what
to do, we should think about our reasons, or about the considerations that favor
the various options open to us. If the view under consideration now is correct, these
considerations are (i) and (ii) above. But note that (i) and (ii) involve facts about
me and my preferences, such as the fact that I like Chinese more than burgers
and burgers more than Armenian. So if the Deliberative Constraint is true, I have
to think about my own preferences in deciding which invitation to accept. But
this seems objectionable: if I prefer Armenian to burgers, then just knowing that
Invitation B is for burgers and Invitation A is for Armenian should be enough to
move me (at least some way) toward accepting Invitation B rather than Invitation
A. Thinking explicitly about my own preferences makes deliberation out to be, in
Schroeder's words, objectionably self-regarding.
There's actually a more pressing related problem. Not only do I have to think
about my own preferences in deliberating, according to this view. If what really
favors Invitation B more than, say, Invitation C is fact (ii) above, then even if I'm
only deciding between Invitations B and C, I have to think about how I feel about
13
Again, compare Schroeder's discussion of the No Background Conditions view in chapter
2 of Schroeder (2007). See also Dancy (2004)'s discussion of enablers. There Dancy argues
explicitly that some facts play an important normative role by enabling other facts to favor
certain alternatives, without themselves favoring the alternative.
47
Armenian food (if the very plausible Deliberative Constraint is true). But this is a
bizarre result.
The defender of a fundamentally non-contrastive favoring relation (the shallow
contrastivist, for example) could oer a more complicated story about these `rather
than' favoring claims, or just bite the bullet that most of our claims and intuitions
about what favors what are incomplete, or strictly speaking mistaken, and try to
explain away these odd results. But I do think I've said enough to show that
whatever the non-contrastivist about favoring says is going to have to be somewhat
complicated, or have some counterintuitive consequences. The contrastivist, on the
other hand, oers a neat explanation. The fact that Invitation B is for burgers
favors choosing Invitation B out offchoose Invitation B, choose Invitation Ag, but
not out offchoose Invitation B, choose Invitation Cg. There's just no question
about whether some consideration favors choosing Invitation B simpliciter since all
favoring is favoring out of a set of alternatives.
3.4 Contrastive Reasons and Favoring
So far, in describing contrastivism, I've remained at a pretty abstract level. In
this section, I'll develop contrastivism further and say more about what it takes for
some consideration to be a reason to perform some action, or have some attitude,
out of a set of alternatives. I'll also consider the relationship between contrastive
reasons and what an agent ought to do.
3.4.1 Contrastivism
I've already mentioned how the contrastivist about favoring can handle the case
I've been considering in passing. We can say that the fact that Invitation B is for
burgers favors choosing Invitation B out offchoose Invitation A, choose Invitation
Bg, but not out offchoose Invitation B, choose Invitation Cg. Similarly, the fact
that Invitation B is for dinner with Burt favors choosing Invitation B out offchoose
Invitation B, choose Invitation Cg, but not out offchoose Invitation A, choose
Invitation Bg. There's just no question about whether it favors choosing Invitation
B simpliciter according to contrastivism, which is good, since it's such a hard
question to answer.
48
Here is the contrastivist version of the idea that reasons are considerations that
favor the things they are reasons for:
Contrastive Reasons: A reason to out ofQ is a consideration that favors-ing
out of Q
If we adopt Contrastive Reasons, we have two options. First, we can take the
favoring relation, and so the reason relation, as primitive, like Scanlon (1998) and
Part (2011). Second, we could try to give some analysis of this relation, saying
what it takes for some consideration to be a reason for, or to favor, some alternative
out of a set of alternatives. In this section, I'll just talk in terms of favoring, without
oering an analysis. But in the next chapter I'll show that contrastivism ts nicely
with a popular idea about how to analyze this relation, in terms of promotion.
3.4.2 A problem
So far, so good. But there is a problem. Which invitation should I choose? The
relevant set is clearlyfchoose Invitation A, choose Invitation B, choose Invitation
Cg. And intuitively, I should choose Invitation B. But|just as it was hard to
say whether anything favored choosing Invitation B simpliciter|it's not clear that
anything favors choosing Invitation B relative to this set.
It isn't surprising that the same, or at least an analogous, issue should arise
here, since I stipulated in this case that I have to choose exactly one of the three
invitations. So the setfchoose Invitation A, choose Invitation B, choose Invitation
Cg is, in this case, exhaustive. Contrastivism about favoring allows us to explain
why the relevant considerations favor choosing Invitation B rather than Invitation
A, on the one hand, and rather than Invitation C, on the other hand. But I haven't
yet said anything to get around the problem of saying whether anything favors
choosing Invitation B relative to the larger set.
14
To solve this problem, I need to
say something about the relationship between reasons and `ought'.
14
This problem actually depends on a certain understanding of what it takes for a fact to favor
some action out of a set of alternatives. In the next chapter, I'll say more about this. For now, it's
enough that intuitively, the fact that Invitation B is for burgers/dinner with Burt doesn't favor
accepting Invitation B out of the larger set of alternatives, since I could do better with respect to
both my food preferences and my dinner company preferences by not accepting it.
49
3.4.3 Reasons and `ought'
The problem is that it's hard to see how we can say that I ought to accept Invitation
B out offaccept Invitation A, accept Invitation B, accept Invitation Cg, since it
seems that neither the fact that Invitation B is for burgers nor the fact that it's
for dinner with Burt is a reason to accept it out of this set. To solve this problem,
I'm going to give a (somewhat) non-obvious account of the relationship between
contrastive reasons and `ought'.
Perhaps the most natural way to understand the relationship between reasons
and `ought', for a contrastivist, is the following (`CRO' for `Contrastive Reasons
and Ought', `*' because I'm going to eventually reject it):
CRO*: s ought to out of Q i s has most reason to out of Q
Note that this principle appeals to a contrastive `ought', as well as contrastive
reasons. I argued in chapter 1 that contrastivism about reasons neither requires
nor is required by contrastivism about `ought', but it's hard to deny that the two
do t together nicely.
15
CRO* is very natural for a contrastivist about reasons,
but it generates the problem I'm trying to solve. If neither the fact that Invitation
B is for burgers nor the fact that it's for dinner with Burt is a reason to accept
it out of the larger set, then it is hard to see how there could be most reason to
accept Invitation B out of this set.
CRO* appeals to reasons out of arbitrarily large sets of alternatives. But as
we've seen, it can be hard to say whether some fact is a reason out of larger sets
of alternatives, though it's generally easy to say whether it's a reason out of two
member sets.
16
So I propose to make use of the facts about what's a reason for
what out of these two member sets in determining what an agent ought to do out
of larger sets:
CRO: s ought to out ofQ is has most reason to out off, g for all of the
other alternatives in Q.
When\wins"in all of the pairwise comparisons with other members in the set, you
ought to . CRO may seem less natural than CRO*, but it's not ad hoc. I think
15
For some arguments in favor of contrastivism about `ought', see Sloman (1970); Jackson
(1985); Cariani (2009, fc); Snedegar (2012).
16
Of course, in many cases it's easy to say whether some fact is a reason out of a larger set: in
most situations, the fact that I'm hungry is a reason to eat lunch out offeat lunch, throw all my
food in the garbage, run a mile, watch a movieg (assuming these are all mutually exclusive).
50
it's one plausible way for the contrastivist to capture the intuitive idea that you
ought to when-ing is best on balance, even if it's not the best in any particular
category.
The dinner invitation case is exactly this sort of case: Invitation B isn't best
with respect to my food preferences, nor is it the best with respect to my dinner
companion preferences. But it's best on balance, and that's why I ought to accept
it. There's most reason to accept Invitation B out offaccept Invitation A, accept
Invitation Bg because, though I slightly prefer Ara's company to Burt's, I greatly
prefer burgers to Armenian food.
17
Though the fact that Invitation A is for dinner
with Ara is a reason to accept Invitation A out of this set, it's relatively weak. On
the other hand, the fact that Invitation B is for burgers is a relatively strong reason
to accept Invitation B out of this set. Similar reasoning shows that there's most
reason to accept Invitation B out offaccept Invitation B, accept Invitation Cg.
Thus, according to CRO, I ought to accept Invitation B out of the larger set; this
is the verdict we want.
CRO will be the principle I adopt going forward about the relationship between
reasons and `ought'. This principle allows the contrastivist to avoid the problem I
raised for the shallow contrastivist in the previous section.
3.5 Looking Forward
In this chapter, I introduced a distinction between shallow contrastivism and deep
contrastivism. The arguments from reason claims in the last chapter are neutral
between these two versions. But I argued that there's good reason to prefer deep
contrastivism. This is because the shallow contrastivist has to appeal to a non-
contrastive favoring relation, which underlies the contrastive reason relation, but
the favoring relation itself is contrastive. I began to develop a contrastivist account
of favoring, and so of reasons, but ran into trouble with the relationship between
`ought' and reasons. To solve this problem, I adopted a certain view about this
relationship.
17
The idea here is that the strength of the reason corresponds in some way to the strength of
the preference. I don't mean to endorse this as a perfectly general principle, but it does seem very
plausible in cases like this one.
51
In the next chapter, I'll continue developing contrastivism. In particular, I'll
show that the contrastivist can accommodate|and in fact is in a better position
to capture|the widespread idea that reasons (and favoring) should be analyzed
in terms of promotion. The account I'll develop there isn't a competitor to the
account I sketched in this chapter in terms of favoring. Instead, it's one way of
further spelling out the view from this chapter by giving an analysis of the favoring
relation. I'll also continue to pay attention to just how radical contrastivism is.
52
Chapter 4
Promotion
In the last chapter, I argued that the favoring relation, which people writing about
reasons generally identify with the reason relation, is contrastive. This closes o
the shallow contrastivist strategy for dealing with the puzzles about reason claims
from Chapter 2. I also began to develop contrastivism about reasons. The view,
as I've stated it so far, says that some fact is a reason to out ofQ when it favors
-ing out ofQ. This is just the contrastivist implementation of the ubiquitous idea
that reasons are considerations that count in favor of the things they're reasons for.
I motivated shallow contrastivism by raising some worries about how radical
deep contrastivism is. Now that I've argued against shallow contrastivism, these
worries are live once again. In the rst section of this chapter, I'll make this concern
especially sharp by introducing a problem for contrastivism. According to con-
trastivism, reasons for and against an option can vary with the alternatives|this
is the central feature of contrastivism, and the one that lets us solve problems fac-
ing non-contrastive theories. But this variation cannot be totally unconstrained|
knowing what reasons there are relative to one set should tell us something about
the reasons relative to certain other sets. The problem facing the contrastivist is
simply that she seems unable to provide these constraints.
The remainder of the chapter will be concerned with developing a detailed ver-
sion of contrastivism that solves this problem. This will essentially involve spelling
out the view I developed in the last chapter in terms of promotion. This taps into
a widespread idea about how to analyze the reason relation. I'll argue that, in
fact, this idea itself supports contrastivism: the best way to make good on the idea
that reasons are tied to promotion is on a contrastive framework. This argument
motivates a particular version of contrastivism that straightforwardly entails just
53
the right kinds of constraints on the variation of reasons between dierent sets of
alternatives.
4.1 A Problem for the Contrastivist
The independence of reasons relative to dierent sets of alternatives posited by the
contrastivist seems like a good thing. But for all I've said so far, and for all other
contrastivists have said, learning what reasons there are relative to one set doesn't
tell us anything about what reasons there are relative to other sets. Re
ection on
this point shows that the contrastivist may have gone too far, and made reasons
relative to dierent sets objectionably independent of one another. Intuitively, there
should be constraints on this independence: reasons relative to one set should bear
some kind of relation to reasons relative to certain other sets. The problem for the
contrastivist is simply that there seems to be nothing about the theory to constrain
the independence in the right ways. Further, the contrastivist has to be careful in
trying to provide these constraints. Otherwise, she might lose the advantages over
non-contrastive theories. In this section I'll make this problem sharp by describing
a couple of dierent kinds of intuitive constraints that we should want, and that
the contrastivist seems unable to provide, for all that's been said so far.
4.1.1 Intransitivity
Suppose I want to get in shape. Then the fact that physical exertion helps one get
in shape is a reason to jog to Cafe 101 for my daily milkshake, rather than walk
there. Similarly, it's a reason to walk there rather than drive. Is it a reason to
jog rather than drive? Of course. And any plausible theory of what makes some
consideration a reason for one thing rather than another will say so.
But the important thing to notice for my purposes here is that we shouldn't
have to think about this at all, once we know it's a reason to jog rather than walk,
and a reason to walk rather than drive. It should just follow that it's a reason to
jog rather than drive. More generally, if r is a reason for A rather than B and a
reason for B rather than C, it should simply follow that r is a reason for A rather
than C. To deny this is to allow for a very troubling kind of intransitivity.
1
1
This isn't the same kind of intransitivity that receives a lot of attention in the literature, the
intransitivity of 'better than' or 'more reason than'. (See especially Temkin (1987, 2012); Rachels
54
The problem for the contrastivist is that nothing about the theory, for all that's
been said, seems to rule out this kind of intransitivity. Since reasons relative to dif-
ferent sets can be independent of one another|as they need to be for contrastivism
to gain advantages over non-contrastive theories|we have no reason to think that
the reason relation will guarantee that reasons forA rather thanB and forB rather
than C will be reasons for A rather than C. Ruling out this intransitivity is one
important constraint.
4.1.2 Other entailment relations
The fact that Bill has high blood pressure is a reason for him to order sh out of
forder sh, order pork, order beefg.
2
If the waiter comes back to the table and tells
Bill that the restaurant is out of pork, that obviously shouldn't change this: the
fact that he has high blood pressure is still a reason to order sh out offorder sh,
order beefg. And again, any plausible theory about what makes a consideration a
reason for an action out of a set of alternatives will say so.
But again, the important point is that we shouldn't have to think about this
at all. It should just follow that, in general, when r is a reason for A out offA,
B, Cg, it's also a reason for A out offA, Bg. Even more generally, reasons for an
alternative out of one set should also be reasons for that alternative out of arbitrary
subsets that include the alternative. If r is a reason for A when some other option
B is relevant, making B irrelevant should not change this.
Similarly, suppose the fact that Bill has high blood pressure is a reason against
ordering beef out offorder sh, order pork, order beefg. If the waiter comes back
and informs Bill that the restaurant also has a salad, he shouldn't have to reconsider
whether the fact that he has high blood pressure is a reason against ordering beef.
Adding a new relevant alternative shouldn't change this. In general, reasons against
(1998, 2001); Friedman (2009) for advocates; see, for example, Broome (1991) for defense of the
much more common view that intransitivity is impossible.) But if the considerations that are
reasons for A rather than B and for B rather than C need not be reasons for A rather than C,
then it seems that we might have a case in which there's more reason for A than B (out offA,
Bg) and more reason for B than for C (out offB, Cg), but not more reason for A than for C
(out offA, Cg). Further, this seems to be an independently troubling kind of intransitivity. I
will return to this issue in Chapter 5.
2
I'm assuming that these alternatives are mutually exclusive|Bill can order at most one of
the dishes.
55
an alternative out of one set are intuitively reasons against that alternative out of
supersets of that set.
These, and other, intuitive entailment relations should hold between reasons
relative to dierent but related sets.
3
In general, as we shift between contexts in
which dierent sets of alternatives are relevant, we shouldn't have to start all over
in guring out what reasons there are. But for all that has been said, contrastivism
cannot deliver these entailment relations, since reasons relative to dierent sets can
be independent of one another|and they need to be for contrastivism to have an
advantage over non-contrastive theories.
This is the problem facing the contrastivist that I'll attempt to solve. As a
preview, my strategy will be to appeal to the popular idea that reasons involve the
promotion of something|a desire or an objective value, for example. Once we see
how to give the most plausible contrastivist analyses of reasons for and against in
terms of promotion, I'll show that the right entailment relations straightforwardly
follow. First, though, I'll argue that the idea that at least many reasons involve
promotion itself provides independent support for contrastivism.
4.2 Promotion
Several writers with importantly dierent views about the nature of normative
reasons accept something like the following schema:
Promote: There is a reason for s to i s's -ing would promote some X of the
relevant kind.
Some writers, like Moore (1903, 1912); Wedgwood (2009); Part (2011) take theXs
to be objective values, like happiness or justice.
4
Others, like Schroeder (2007), take
the relevant kind ofX to be desires of the agent. And others, like Finlay (2006, ms),
take the relevant kind of X to be contextually-specied ends. Nagel (1970) also
analyzes reasons in terms of promotion, though his view is importantly dierent.
3
I'll say more about entailments between sets that dier in resolution in section 5.
4
Moore didn't talk about reasons explicitly, but rather rightness. But his view of rightness
would extend straightforwardly to this kind of view of reasons.
56
Still, there is a plausible view in the neighborhood of Nagel's which explicitly ts
Promote.
5
We might follow writers like Scanlon (1998) and Anderson (1993), and draw a
distinction between promoting values and respecting or honoring values. If we make
this distinction, we might worry that some reasons will not be in explained by, or
even involve, the promotion of anything. For example, according to this view, the
value of friendship primarily gives us reasons to respect or honor friendships, say
by treating our friends well, rather than to promote friendship, say by matching up
people we think would make good friends.
6
The argument for contrastivism that
I will give shortly is that the idea that reasons involve promotion is best captured
on a contrastive theory. If not all reasons involve promotion, then this argument
at best would seem to show that some reasons|those that involve promotion|are
contrastive. But contrastivism, as I want to defend it, is the thesis that all reasons
are contrastive.
I think we can allow that some reasons do not involve promotion, but rather
involve respecting or honoring values, without limiting the scope of contrastivism in
this way. Everyone should recognize that at least some reasons involve promotion.
7
And if I am right that these reasons are contrastive, that is some pressure to adopt
contrastivism about all reasons. It is theoretically unattractive to posit two separate
reason relations, one of which is contrastive and one of which is non-contrastive. If
there is a plausible unied view of reasons on oer, that would be preferable.
I will develop my contrastive theory in terms of promotion, since (i) everyone
should agree that some reasons involve promotion, and (ii) doing so will simplify
exposition of the view. At the end of the chapter, however, I will return to this
point and show how to extend the view to cover non-promotional reasons.
5
The complication with Nagel's view is that he thinks what we have reasons to do is to promote
valuable states of aairs, whereas the other theories hold that the fact that performing certain
actions would promote certain objectives is what explains why we have reasons to do those actions.
6
See Pettit (1991) and Pettit's contribution to Baron et al. (1997) for a defense of the view
that all reasons involve promotion.
7
For example, instrumental reasons to take the means to our ends seem to be clearly tied to
the promotion of those ends. See Kolodny (ms); Bedke (ms) for relevant discussion.
57
4.2.1 Promotion as probability-raising
The idea that reasons involve promotion is an attractive and widespread one. But
what does it take for an action to promote some objective? A natural initial idea
is that an agent's -ing promotes X when her -ing raises the probability of X.
Of course, this is not yet very precise. Promoting a desire is making it more
probable that the desire is satised|where the object of the desire is a proposition
p, promoting the desire is raising the probability of p. It's harder to see what it
would be to promote a value. I'll follow Wedgwood (2009) and say that to promote
a value is to promote a state of aairs in which the value is instantiated.
8
On the
probability-raising view, to promote a value is to make it more probable that the
value is instantiated.
Though this probability-raising view is a natural one, it is also controversial.
One important problem concerns the promotion of values like justice. It is true
that we have reasons to do things that would promote justice, in some sense. What
is more problematic, though, is the claim that promoting justice involves making it
more probable that justice is instantiated, because it seems that justice is already
instantiated, at least to some degree. If that's right, then the probability that justice
is instantiated, no matter what we do, is 1. So we can't make it more probable.
9
I
think there are things to be said in defense of the probability-raising view.
10
But for
now I want to put these worries aside, and work with this view. I do this in order
to have a concrete version of the idea that reasons involve promotion to work with.
Moreover, existing views about the relationship between reasons and objectives like
desires, goals, or values are often framed in terms of probability-raising.
11
I show in
8
Wedgwood does reject the probability-raising view of promotion, however.
9
Thanks for Walter Sinnott-Armstrong and Jacob Ross for discussion of this point.
10
For example, we may try specifying more ne-grained values, like `I increase the amount of
justice in the world' or `I do the most just thing I can'. A further reason to be careful in specifying
the objectives is the following kind of case, raised by Fabrizio Cariani in conversation. Suppose
there are two objectives, X and Y, but that promoting X without promoting Y is worthless. Since
promoting X when we also promote Y is valuable, it seems that X can provide reasons to do the
things that would promote it. But it does not seem that it should provide reasons to do the things
that would promote it in cases in which we will not also promote Y. In this kind of case, I want
to say that the reason-providing objective is something like `X when you also promote Y', rather
than just `X'.
11
See, for example, Bedke (ms); Finlay (2001, 2006, ms); Kolodny (ms); Schroeder (2007).
Bedke's and Kolodny's discussions concern instrumental reasons; Finlay's and Schroder's concern
all reasons.
58
section 4.5, though, that this probability-raising view of promotion is not necessary
for my purposes.
So, according to this view, you have reasons to when your -ing would make
some relevant objective more probable. But this leaves an important question:
more probable than what? I'll consider two attempts to answer this question, and
argue that both of them are problematic. Then I'll sketch a contrastivist version of
the view, which I'll develop in the next section, that avoids the problems.
4.2.2 Doing nothing
The rst view I'll consider is oered by Schroeder (2007), p. 113:
Doing Nothing: s's -ing promotes X i P (Xjs s) > P (Xjs does nothing)
An important question about this view is the following: what is it for me to do
nothing? Evers (2009), in a paper objecting to Schroeder's account, considers
two ways to answer this question, but shows that both suer from the same basic
problem. On the rst reading|which I think is pretty implausible even as an
interpretation of Schroeder|doing nothing is simply sitting absolutely still. The
second more plausible reading is that doing nothing is just doing what you're already
doing. For example, if I'm walking to the store for milk, and realize I forgot my
wallet, I have a reason to turn around and get it. That's because, on this account,
the probability that I get milk is higher given that I turn around and get my wallet
than the probability that I get milk given that I keep walking to the store without
my wallet.
The problem Evers points out for both of these readings, and for Doing Noth-
ing in general, is that I can have reasons to do nothing, however we understand it,
but this account rules that out. Suppose doing nothing is sitting absolutely still.
Well, suppose that I'm out in the woods and a predator has spotted me. Fortu-
nately, the predator can only see me if I move. So I have a reason to sit absolutely
still. Next, suppose that doing nothing is doing what I'm already doing. And sup-
pose that I'm waiting at a restaurant for an important client who's just informed
me that he'll be there in ve minutes. Well, then I have a reason to keep doing
what I'm doing|that is, to keep waiting. So I can have reasons to do nothing, no
matter how we understand it. But for me to have a reason to do nothing on this
account, the following must hold: P (XjI do nothing) > P (XjI do nothing). But
59
that's impossible. Since I can have reasons to do nothing, Doing Nothing is not
the right way to understand promotion, in the sense relevant for Promote.
4.2.3 Not -ing
The second view I'll consider is oered by Finlay (2006, ms):
Not : s's -ing promotes X i P (Xjs s) > P (Xjs does not )
This account avoids the problem for Doing Nothing: for you to have a reason to
do nothing on this account, doing nothing just has to raise the probability of X
more than not doing nothing would. But to evaluate this account, we need to know
what not -ing is. I'll consider three dierent versions of this view and argue that
all three are problematic.
4.2.3.1 Possibilism
The rst proposal, call it Possibilism, is that -ing promotes X whenever -ing
raises the probability of X more than anything else the agent could do that is
incompatible with -ing. The problem with this suggestion is that it makes it too
hard to have reasons to do things. For X to give you a reason to , -ing has to
make X more likely than anything else you could do. Suppose I want to eat Thai
food for dinner, and that there are two Thai places around, RCA and Thai Patio.
Suppose that there is a very slight chance that RCA will be too crowded so that I'll
have to go home without having Thai food if I go there. So I have a slightly better
chance to get Thai food if I go to Thai Patio than if I go to RCA. Thus, according
to this proposal, my desire for Thai food gives me no reason at all to go to RCA.
But this seems like a problem. After all, my desire for Thai food certainly gives
me a reason to go to RCA rather than go to Zankou Chicken, which doesn't even
serve Thai food. But how could this desire give me a reason to go to RCA rather
than Zankou Chicken if it doesn't give me any reason at all to go to RCA? So this
proposal makes it too hard to have reasons to do things.
60
4.2.3.2 Actualism
A second version of Not, call it Actualism, holds that not-ing is doing whatever
it is you would actually do if you didn't.
12
This view would seem to make it easier
to have reasons, since your -ing doesn't have to raise the probability of X more
than anything else you could do, but only more than whatever it is you'll actually
do, if you don't .
But now consider this case from Behrends and DiPaolo (2011). Debbie desires
thatp. She is given a choice to push one of three buttons, A, B, or C. If she pushes
either A or B, p is guaranteed to be true. But if she pushes C, p will not be true.
Debbie in fact pushes A, and if she hadn't pushed A, she would have pushed B.
It turns out, according to the view under consideration, that she had no reason to
push button A, sinceP (pjDebbie pushes A) =P (pjDebbie pushes B).
13
But this is
counterintuitive. After all, she certainly had a reason to push A rather than push
C. So this proposal also seems to make it too hard to have reasons to do things.
4.2.3.3 Probabilism
Finally, consider the following Probabilist view. Rather than thinking of not -ing
as doing anything incompatible with-ing, or with doing whatever it is you would
actually do, if you didn't , we can just appeal to a probability distribution over
the things you might do. So for example, if you don't go to RCA, there might be
a 60% chance you go to Thai Patio, and a 10% chance you go to Zankou Chicken.
This view will give us some value for P (you don't ), and let us calculate one for
P (Xjyou don't ) by weighting P (Xjyou ) by the probability that you , for all
dierent ways of not -ing. What we end up with is something like the expected
value, as far asX is concerned, of not-ing. Then to determine if you have a reason
to , we can straightforwardly check to see if P (Xjyou )>P (Xjyou don't ).
This theory also makes it too hard to have reasons to do things. Suppose again
that I want some Thai food, that the only two Thai places around are Thai Patio
and RCA, and that the only other dinner option is Zankou Chicken, which doesn't
12
The distinction between this version and the last version, and the corresponding names, are
familiar from the debate between actualists and possibilists about obligation. See Jackson and
Pargetter (1986).
13
Behrends and DiPaolo point out a more basic problem. It turns out, according to this view,
that Debbie didn't even promote her desire by pressing A.
61
serve Thai food. But now suppose that RCA is very likely to be so crowded that
I'll have to go home without having Thai food if I go there. In particular, suppose
that P (I get Thai foodj I go to RCA)=.2, while P (I get Thai foodj I go to Thai
Patio)=.9, and of course P (I get Thai foodj I go to Zankou)=0. And nally,
suppose P (I go to RCA)=.3, P (I go to Thai Patio)=.6, and P (I go to Zankou
Chicken)=.1. With some calculation, we end up with the result that P (I get Thai
foodj I don't go to RCA) is roughly .77 (where not going to RCA is just either
going to Thai Patio or going to Zankou Chicken), which is greater than .2. So
according to this view, my desire for Thai food doesn't give me any reason at all to
go to RCA. But as before, surely this desire gives me a reason to go to RCA rather
than Zankou Chicken. So it must give me some reason to go to RCA.
14
It's also worth pointing out, I think, that if we're going to assign probabilities
to the various ways of not-ing, we need some way of distinguishing one way from
another. That is, we need some way of dividing up the alternatives. Dierent ways
of doing this will lead to dierent answers about what you have reasons to do. And,
as I pointed out in section 1, this kind of resolution-sensitivity is characteristically
contrastivist. So spelling out this theory may require contrastivism after all.
4.2.4 Contrastive promotion
I think we can give a simpler contrastive view that avoids the problems I've discussed
so far. This view does not rule out reasons to do nothing. And it does not need
to come up with some way to understand not -ing. Instead, the idea is that X
can give you a reason to , relative to a set of alternatives Q, when your -ing
raises the probability ofX more than the all of the other alternatives inQ.
15
Thus,
14
In defending this sort of view, Finlay (2006, ms), makes the following assumption, which he
calls the `symmetry of choice': in making judgments about what reasons you have (or what's
good, or what you ought to do, on his theory), we ignore your dispositions to take one option over
another. So in guring out which reasons I have, we would just assume that I am equally likely to
go to each of the three restaurants. But we can still construct a case on which, according to this
view, my desire for Thai food gives me no reason at all to go to RCA. Let P (I go to x) = 1=3,
for all three restaurants,x, so that the symmetry of choice is satised. And, as above, let P (I get
Thai foodj I go to RCA)=.2,P (I get Thai foodj I go to Thai Patio)=.9, andP (I get Thai foodj
I go to Zankou)=0. Then by some simple calculations, P (I get Thai foodj I don't go to RCA) is
roughly .38, which is greater than P (I get Thai foodj I go to RCA)=.2. Thus, according to this
modied Probabilist view, my desire for Thai food gives me no reason at all to go to RCA. But
surely it gives me a reason to go to RCA rather than go to Zankou Chicken, and it's hard to see
how it could do this unless it gave me some reason to go to RCA.
15
I'll defend this claim in the next section.
62
in the Thai food case, my desire for Thai food gives me a reason to go to RCA
rather than Zankou Chicken, even though it doesn't give me a reason to go to
RCA rather than Thai Patio, since going to RCA raises the probability that I get
Thai food more than going to Zankou, but less than going to Thai Patio. Similar
remarks hold in the button-pressing case from Behrends and DiPaolo (2011). The
key to these solutions to the problems facing non-contrastive theories is that, for
the contrastivist, the sets of alternatives that reasons are relativized to need not be
exhaustive of all the options open to the agent.
4.2.5 Motivating resolution-sensitivity
The arguments against the non-contrastive accounts of reasons in terms of pro-
motion illustrate the advantages of non-exhaustivity. But when I introduced con-
trastivism in Chapter 1, I pointed out a second feature of the view, namely resolution-
sensitivity. The arguments I've given so far don't motivate this feature of con-
trastivism.
16
But I think we can motivate it. Once we relativize reasons to sets of
alternatives, there are cases in which whether or not something is a reason for an ac-
tion seems to depend, not just on what possibilities are covered by the alternatives
in the set, but also how we split up those possibilities.
Consider the following version of the famous Professor Procrastinate case from
Jackson (1985); Jackson and Pargetter (1986). Professor Procrastinate has been
asked to write a review an important new book, because she is the most qualied
person to write it. But she is a terrible procrastinator. If she accepts the invitation,
she is very unlikely to actually write the review. And if that happens, the author's
career and the eld at large will suer. If Procrastinate just declines the invitation,
someone else|less qualied, but more reliable|will be asked to write, and will do
so. In this case, the fact that Procrastinate is a procrastinator is a reason for her
to decline the invitation rather than accept. That is, it's a reason to decline out of
fdecline, acceptg, plausibly provided by the objective of doing what's best for the
profession. But it is intuitively not a reason to decline out of the more ne-grained
set,fdecline, accept and write, accept and don't writeg. So whether or not this fact
is a reason to decline seems to depend on the resolution at which we divide up the
16
I did mention that spelling out Probabilism may require resolution-sensitivity, but since
I rejected that view|for reasons having to do with non-exhaustivity|I can't appeal to that
argument in motivating resolution-sensitivity.
63
alternatives. Thus, once we relativize reasons to alternatives, there's good reason
to allow the sets to vary in resolution, rather than, say, making them all maximally
ne-grained.
4.3 Contrastive Reasons and Promotion
I've argued that promotion, in the sense relevant for Promote, is best understood
contrastively: your -ing promotes X only relative to some set of alternatives. It
does so when it raises the probability of X more than all the other alternatives in
the set. In this section I'll begin to develop a contrastivist account of reasons based
on this account of promotion. In developing the view, I will make the following
two simplifying assumptions, both of which I have called attention to above: (i) all
reasons involve promotion; and (ii) the relevant kind of promotion is probability-
raising. Making these assumptions will let us see very clearly how the sort of view I
want to defend can provide the constraints on how reasons can vary between sets of
alternatives. But at the end of the chapter, I will show that we can actually relax
both of them without sacricing any of the attractions of the view.
Promote tells us what has to hold for you to have a reason to , or for there
to be a reason for you to . But it would be nice to have an account of when some
fact is a reason for you to . Philosophers who appeal to promotion to analyze
the reason relation frequently appeal to explanation here. If you have a reason to
because -ing promotes some objective, then some fact is a reason for you to
when it explains why (or is part of the explanation for why) -ing promotes the
objective.
17
I have a reason to go to the store because doing so would promote the
objective of my having breakfast in the morning. The fact that I'm out of milk is
part of the explanation for why going to the store would promote this objective. So
according to this idea, the fact that I'm out of milk is a reason to go to the store.
I'll adopt this strategy here.
So what we want is a contrastivist analysis of reasons for and reasons against
in terms of promotion. So far I've said that r is a reason for you to when it
17
For analyses of reasons partly in terms of explanation, see Toulmin (1950); Broome (2004);
Searle (2001); Finlay (2001, 2006, ms); Schroeder (2007). Like these other writers (with the
possible exception of Finlay), I take the relevant sort of explanation here to be a non-epistemic
notion. For r to explain why your -ing would promote X is for r to be part of what makes it
true that your -ing would promote X.
64
explains why your -ing would raise the probability of some relevant objective X
more than other relevant alternatives. Conversely, it's natural to think that r is a
reason against-ing when it explains why your-ing would raise the probability of
X less than other relevant alternatives. So what we need to know is, how many of
the other alternatives does your -ing have to raise the probability of X more/less
than, for X to give you a reason for/against -ing?
First a quick note about reasons against. It is often thought that a reason
against some action is just a reason for not-.
18
On this view, we can just
have one kind of reason, reasons for. But the contrastivist cannot accept this view
because, given non-exhaustivity, the sets of alternatives reasons are relativized to
need not be closed under negation: just because is in a set, it does not follow that
not- is also in the set. So if we want to talk about reasons against an alternative
relative to a set Q, we can't just talk about the reasons for not- out of Q,
since not- might not even be inQ. Thus, the contrastivist needs to dene reasons
against|reasons not to , instead of reasons to not-|separately.
So start (for expositional purposes) with reasons against. I suggest the following
analysis:
Against: r is a reason for you not to out ofQ i there's someX of the relevant
kind
19
and r explains why
20
P (Xjyou ) < P (Xjyou ), for some 6= in
Q.
If you could do better, as far as X is concerned, by doing something other than
-ing, then X very plausibly gives you a reason against -ing. The most natural
alternative to Against would say that your-ing has to raise the probability ofX
less than all the other alternatives. But this misses out on some intuitive reasons
against. For example, suppose I can travel the 15 miles to campus by jogging,
biking, or driving. The fact that I want to get there in under an hour intuitively
gives me a reason against biking, even though I have a better chance of getting
18
See Nagel (1970), footnote 1 on p. 47, for example.
19
Again, by `some X of the relevant kind', I mean the following: there's some set of objectives
that are the things to be promoted, according to our theory of reasons. A desire-based theory will
say that this set contains desires of the agent; a value-based theory will say that it contains some
independent values. X is an objective of the relevant kind when it is in this set. By existentially
quantifying here, we allow that there are multiple objectives that can give you reasons in any
individual case.
20
Or is part of the explanation of why|I'll ignore this qualication from now on.
65
there in under an hour if I bike than if I jog|it's still quite unlikely that I get there
in under an hour if I bike.
Once we accept Against as an analysis of reasons against, I think we should
accept the following analysis of reasons for:
For: r is a reason for you to out ofQ i there's someX of the relevant kind and
r explains why P (Xjyou )>P (Xjyou ), for all 6= in Q.
21
This makes reasons for harder to come by than reasons against. But I think this
is a very plausible analysis of reasons for. If you could do better, as far as X is
concerned, by doing something other than-ing, it would be strange ifX gave you
a reason to out of a set of alternatives that includes that other thing. The most
natural alternative would say thatX can give you a reason to relative to some set
when your-ing would raise the probability ofX more than some other alternative
in the set. But if we accept this in addition to Against, some considerationr could
be both a reason for and a reason against all the alternatives in the middle|that
do not do the best and that do not do the worst, as far as X is concerned. For
example, in the case I introduced above, the fact that campus is fteen miles away
would be both a reason to bike and a reason not to bike, both provided by my desire
to get to campus in under an hour. But this is a violation of the very plausible
principle, Restricted Exclusivity, that I argued for in Chapter 2.
22
One potential objection to this view is that it gives too liberal a treatment of
reasons against. A reason against some alternative, you might think, has to \imply
a signicant criticism" of that alternative, in the sense of Greenspan (2005, 2007),
and merely being a reason for an incompatible alternative isn't enough. But other
philosophers seem to agree with me that a reason for one alternative is in fact a
reason against other incompatible alternatives, for some relatively weak notion of
incompatibility, like physical or psychological incompatibility.
23
I think this is very
plausible: if some consideration r is a reason for A|if it highlights some benet
21
Interestingly, though I reject Finlay (ms)'s non-contrastive account of reasons, this contrastive
account of reasons for closely mirrors his contrastive analysis of `ought'. According to that analysis,
roughly, you ought to just in case your-ing makes the contextually salient end more likely than
any other relevant alternative. By allowing multiple objectives|the analog of Finlay's ends|to
provide reasons in a single context, my analysis is better suited to the pro tanto notion of a reason.
22
This argument also tells against a revised version of Possibilism, according to which an
objective provides a reason for you to when-ing raises the probability more than something|
instead of everything|else you could do.
23
Compare the discussion of `General Substitutability' in Sinnott-Armstrong (1992), section 1.
66
for doingA|and if doingB precludes doingA, then it is very plausible thatr is a
reason against B, since doing B would keep you from getting that benet.
24
Note, though, that we can draw an interesting distinction between two types of
reasons against on this view. A reason against some alternative that does imply a
signicant criticism of that alternative|like the fact that Zankou Chicken is really
crowded at this time of day|often won't turn out to be a reason for any other
particular alternative. That's because these sorts of facts are often irrelevant for
dierences between other alternatives in the set. For example, the fact that Zankou
Chicken is really crowded doesn't really tell us anything about the relationship
between going to Thai Patio and going to Spicy Thai BBQ. So it doesn't explain
why eitherP (I don't have to waitjI go to Thai Patio)>P (I don't have to waitjI go
to Spicy Thai BBQ), or vice versa. So by For, it won't turn out to be a reason for
either of these, though it's still a reason against going to Zankou Chicken. Reasons
against an alternative that do not imply a signicant criticism, on the other hand,
will often be reasons for other particular alternatives. So, even on this view, we
can capture Greenspan's insight: some considerations merely count against certain
alternatives without counting in favor of others.
4.4 Providing the Constraints
In this section, I'll show that the theory I developed in the last section, indepen-
dently motivated by considerations about the relationship between reasons and
promotion, straightforwardly provides the constraints on the independence of rea-
sons relative to dierent sets of alternatives I motivated in section 1, as well as some
other plausible constraints.
4.4.1 Transitivity
The rst constraint we need on the independence of reasons relative to dierent sets
is one that rules out intransitivity. That is, we want it to turn out that whenever
24
Sinnott-Armstrong (2008) says that a reason \favors one thing and disfavors others" (p. 258).
Ruben (2009) says that when doing A is incompatible with doing B, \a reason to do an act of
type B must also be a reason not to do an action of type A" (p. 63). And Broome (ms) cites
nice features of one alternative|which are reasons for taking that alternative|as reasons against
incompatible alternatives. See the Montreux-Marrakesh example on p. 63.
67
r is a reason for A rather than B and for B rather than C, it is also a reason for
A rather than C. In fact, we need an importantly but subtly dierent constraint,
once we adopt a view on which we explain reasons in terms of the promotion of
various kinds of objectives or ends. We don't just want reasons for A rather than
B and for B rather than C to be reasons for A rather than C; we want them to
be provided by the same objective. A case in which my desire for Thai food gives
me a reason r for A rather than B and for B rather than C, but not for A rather
than C is counter to the transitivity intuition, even if, say, my desire to spend lots
of money explains why r is a reason for A rather than C.
So what we need is the following kind of constraint:
Transitivity: If r is a reason for A rather than B, provided by X, and a reason
for B rather than C, provided by X, then r is a reason for A rather than C,
provided by X.
Fortunately, it's easy to see that this follows straightforwardly from the framework
I developed in the last section. If r is a reason for A rather than B, and a rea-
son for B rather than C, both provided by X, then we know that r explains why
P (XjA) > P (XjB) and why P (XjB) > P (XjC). Since these facts about condi-
tional probability don't depend on the comparisons we're making, and since the
greater than relation is transitive, it follows that P (XjA) > P (XjC). And given
plausible assumptions about how explanation works in these cases|namely, that
if r explains why P (XjA)>P (XjB) and r also explains why P (XjB)>P (XjC),
then it will explain why P (XjA) > P (XjC) (at least in cases in which this is
true)|r explains this.
25
So, by For, r is a reason for A rather than C. Thus,
by appealing to the idea that reasons are to be explained in terms of promotion,
understood as probability-raising, we secure Transitivity.
4.4.2 Non-exhaustivity
The second constraint concerns entailment relations between reasons relative to
certain kinds of related sets of alternatives. First I'll discuss entailments between
25
Remember that I'm using a non-epistemic notion of explanation here. Plausibly, the epistemic
notion will not have this feature, but the sense I intend, on which x explains y when x is part of
what makes it true that y, does.
68
reasons relative to sets that dier in the alternatives they include|that is, sets that
dier along the dimension of non-exhaustivity.
First, consider reasons for an alternative relative to one set, and reasons for
that alternative relative to a subset of that initial set. I said above that intuitively,
facts which are reasons for A relative to Q should also be reasons for A relative to
subsets ofQ that containA. Imagine that this were not true. Then there might be
a case in which r is a reason for A rather than either of B or C, but not a reason
forA rather thanB. It's implausible that there could be such a case, so we should
rule it out.
Fortunately, the theory I developed in the last section|in particular For|
delivers precisely this result. If r is a reason for A out of Q, explained by end or
objective X, then by For, P (XjA) > P (XjB), for all the other alternatives B in
Q. Since the value of P (XjA) does not vary with the alternatives, P (XjA) will
obviously be higher than P (XjB) for all the alternatives B in any subset of Q.
And since r explains why P (XjA)>P (XjB) for all of the other alternatives B in
Q, r will also explain this for all of the other alternatives in a subset of Q. So r
will be a reason for A out of any subset of Q that contains A.
Next, consider reasons against. I said above that reasons against an alternative
out of one set should also be reasons against that alternative out of supersets of
the initial set. If this entailment did not hold, then we might have a case like the
following. r is a reason againstA when the alternative isB alone, but not a reason
against A when we consider C in addition to B. Again, this is implausible.
Fortunately, the theory delivers just the result we want. Suppose r is a reason
against A out of Q, explained by X. Then, by Against, P (XjA) must be lower
than P (XjB) for some other alternative B in the set. Merely adding some new
alternatives to Q won't change this fact. Thus, r will also be a reason against
A relative to the new superset of Q. So the independently motivated version of
contrastivism that analyzes reasons in terms of promotion delivers precisely the
constraints we need, on the independence of reasons relative to sets and their subsets
and supersets.
69
4.4.3 Resolution-sensitivity
I also introduced a second feature of contrastivism, resolution-sensitivity. Two sets
of alternatives may cover the same possibilities, but do so at dierent levels of
detail. For example, we might have reasons to A relative to the more ne-grained
setfA, B, Cg or relative to the more coarse-grained setfA, B_Cg. Do we want
a theory to deliver any relationships between reasons relative to sets related in this
way?
I think there are some intuitive constraints here. Suppose the fact that I don't
have any vacation days left is a reason to go to work out offgo to work, stay home
and clean, stay home and watch TV, stay home and do something elseg, explained
by my desire not to get red. It also seems to be a reason to go to work out of
the more coarse-grained setfgo to work, stay homeg. The intuitive thought here is
that if doingA is better, with respect to the objectiveX, than all of the (relevant)
ways of doingB, then doingA is obviously better than doingB, with respect toX.
And the theory delivers this result. It's a fact about conditional probability that
P (XjB_C) cannot be higher than both P (XjB) and P (XjC), and this secures
the result: if P (XjA) is higher than P (XjB) for all the dierent ne-grained ways
B of doing some more coarse-grained option D, then P (XjA)>P (XjD). That is,
reasons forA relative to more ne-grained sets entail reasons forA relative to more
coarse-grained sets that just group together other alternatives.
26
Now consider reasons against. The fact that I'm out of vacation days is intu-
itively a reason against staying home out of the coarse-grained setfgo to work, stay
homeg. And it also seems to be a reason against this alternative out of the more
ne-grained setfgo to work by car, go to work by bus, go to work some other way,
stay homeg. Again, the theory delivers the correct result here. Since P (XjB_C)
can't be higher than bothP (XjB) andP (XjC) (it also cannot be lower than both)
we know that, in general, reasons against an alternative A out of a coarse-grained
set likefA,B_Cg entail reasons against that alternative out of more ne-grained
26
We might also wonder about the relationship between reasons for A relative to one set and
reasons for more coarse-grained alternatives that subsume A (say, A_B) in more coarse-grained
sets of alternatives. Similarly, we might wonder about the relationship between reasons for a
more coarse-grained alternative (say, A_B) and reasons for the ne-grained alternatives that it
subsumes (say, A) relative to more ne-grained sets. It turns out that the theory doesn't deliver
any constraints in these cases. I think these are the intuitively correct results, though I won't
argue for that here.
70
sets likefA, B, Cg, according to the theory. If P (XjA) is lower than P (XjD), for
some coarse-grained alternative D, then by this fact about conditional probability,
P (XjA) must be lower than at least one of the more ne-grained ways of doing
D. And so by Against, X will provide a reason against A relative to the more
ne-grained set.
4.4.4 Other entailments: unions and intersections
Now I will discuss some other entailment relations delivered by the theory that
involve unions and intersections of sets of alternatives. It is independently interest-
ing that the theory delivers entailments in these cases. But it also may have some
interesting applications. For example, when two agents who have been deliberat-
ing separately, using two dierent sets of alternatives, come together to deliberate
jointly, it's plausible that the relevant alternatives in this new context of joint de-
liberation will bear some relationship to the two separate sets the agents were using
on their own. Either taking the union or the intersection of these sets would be
reasonable.
27
Moreover, this illustrates that the theory delivers not only the most
obvious constraints (e.g., the transitivity constraint, and the subset and superset
constraints) but also less obvious constraints that seem intuitively correct, once the
theory brings them into view. This provides additional conrmation for the theory.
First, note that the following two entailments follow as corollaries of the previous
ones involving subsets and supersets. (i) Reasons forA out of bothQ andQ
0
, both
provided byX, will be reasons forA out of the intersection,Q\Q
0
, provided byX.
This just follows from the result that reasons for A out of one set are reasons for
A out of subsets of that set.
28
(ii) Reasons against A out of both Q and Q
0
will be
reasons against A out of the union, Q[Q
0
. This just follows from the result that
reasons against A out of one set are reasons against A out of supersets, as well.
We also get the following new result, which is not simply a corollary of a previous
one. Supposer is a reason toA out ofQ, provided byX, and thatr is also a reason
to A out of Q
0
, provided by X. What about the union of these two sets, Q[Q
0
?
27
This will often be more complicated, of course, since the resolutions of the two separate sets
may cross-cut one another.
28
The limiting case here is whenQ\Q
0
=fAg. Technically, byFor, everything will be a reason
for A relative to this singleton set. But singleton sets will never be the relevant set|if there's
only one relevant alternative, then there's no need for deliberation, and there won't be reasons
for or against the single alternative.
71
By For, we can immediately see thatr will be a reason forA relative to the union,
as well. And this is plausible: if nothing in either set was better than A, as far as
X is concerned, then putting them all together in the same set should not make a
dierence. Interestingly, we don't get the corresponding result for reasons against
and intersections: it's not true in general that if r is a reason against A out of Q
and out of Q
0
, then it's a reason against A out of Q\Q
0
, since we may lose all of
the alternatives in either set that are ranked higher than A, when we intersect.
4.5 Promotion as Probability-Raising
Earlier I promised to reconsider my assumption of the probability-raising view of
promotion in light of apparent problems with that view. I set those problems aside,
since (i) existing discussions of these issues are often framed in terms of probability-
raising, and (ii) the probability-raising view provided a concrete version of the idea
that reasons involve promotion to work with. But now I want to show that this
particular view of promotion is not crucial for my purposes here.
I have shown that the idea that reasons involve the promotion of objectives gen-
erates a version of contrastivism that provides important intuitive constraints on the
independence of reasons relative to dierent sets of alternatives. The probability-
raising view of promotion has two features that I used in proving that these con-
straints hold. First, it provides a ranking of options for each objective, in terms of
how probable the actions make the objective. I used this ranking to generate the
constraints involving non-exhaustivity: the transitivity constraint, as well as the
constraints involving subsets and supersets. But any way of understanding pro-
motion that provides a ranking of options for each objective in terms of how well
the options promote the objective will generate these constraints, as long as the
ranking itself is transitive.
The second important feature of the probability-raising view, which provides the
resolution-sensitivity constraints, is that the rankings it provides have the following
feature:
Disjunction Boundedness: For all actions A and B and for all objectives X,
A_B is ranked somewhere between (inclusive)A andB, in terms of how well
it promotes X.
72
The fact about conditional probabilities, that whenP (XjA)P (XjB),P (XjA)
P (XjA_B)P (XjB), secures this property for the probability-raising view. But
I think that any plausible ranking of options in terms of how well they promote a
given objective will have the Disjunction Boundedness property. Doing A_B
can't do worse than the worst of A and B, and can't do better than the best of A
and B, as far as the relevant objective goes.
My arguments against Doing Nothing, Actualism, Possibilism, and Prob-
abilism motivating a contrastive account of promotion were framed in terms of the
probability-raising view. But analogous arguments can be run against more general
versions of the rst three views, stated just in terms of how well actions promote
various objectives. For example, in Behrends and DiPaolo (2011)'s button-pressing
case, Debbie surely has a reason to press button A rather than press button C. But
according to Actualism and Possibilism, Debbie has no reason at all to press
button A, since pressing button A does not better promote her desire than pressing
button B. It is hard to see how to give a non-probabilistic version of Probabilism.
But if we are worried about the probability-raising view of promotion, the diculty
with formulating a non-probabilistic version of Probabilism should make us worry
about this view, as well.
We can also see now that the probabilistic principles, For and Against, were
special cases, in the probabilistic framework, of the following more general princi-
ples:
For*: r is a reason for you to out of Q i there is some X of the relevant kind
such thatr explains why your-ing would better promoteX than any of the
other alternatives in Q.
Against*: r is a reason for you not to out ofQ i there is someX of the relevant
kind such that r explains why your -ing would not promote X as well as
some other alternative in Q.
So the probability-raising view of promotion is not crucial for the theory I've de-
veloped here. Any view of promotion which provides rankings of options in terms
of how well they promote the objectives and which has the Disjunction Bound-
edness property will work.
73
4.6 Non-Promotional Reasons
I've just argued that we need not understand the relevant kind of promotion as
probability-raising: all we need is a ranking (with certain minimal properties) of
alternatives in terms of how well they promote (in some sense) a given objective.
In this section, I will relax the assumption that all reasons involve promotion.
In section 2, I argued that even if we think that some reasons do not involve
promotion, but rather involve respecting or honoring values, my argument that the
reasons that do involve promotion are contrastive puts rational pressure on us to
adopt contrastivism about all reasons. But there is still a serious gap. I've spent the
last few sections showing how a contrastive theory of reasons cashed out in terms of
promotion straightforwardly delivers important constraints on how reasons can vary
with sets of alternatives. In doing so, I made crucial use of rankings of alternatives
in terms of how well they promote various objectives. But if some reasons do not
involve the promotion of objectives at all, how can we provide the constraints for
these reasons? Moreover, if we need to tell some other story for these reasons,
there's a worry that it will generalize to explain the constraints on how the reasons
that do involve promotion can vary. This would undermine my claim that the
idea that reasons involve promotion and contrastivism about reasons provide an
attractive mutually supporting package.
Fortunately, I think we can generalize the explanation of the constraints given
in terms of promotion. The basic idea is that dierent alternatives can do better
or worse at respecting or honoring values. For example, going into the burning
building to save someone does better at respecting the value of human life than
merely calling 911, which in turn does better than just walking by without even
calling 911. There may be various ways of being better at respecting or honoring
some value. For example, one action may respect a given value more than another
action, or one action may be closer to the ideal way of honoring a given value than
another action. The important point is that we will still be able to get a ranking of
alternatives, not in terms of how well they promote a given value, but in terms of
how well they respect or honor that value.
29
And this ranking will very plausibly
29
Some rankings will likely be pretty uninteresting, with one alternative at the top and all other
alternatives tied at the bottom, as not respecting the value at all. For example, if I promise to
A, then a ranking of the alternativesfA,B,C,Dg in terms of how well they respect the value of
promise-keeping will be trivial in this way.
74
have the Disjunction Boundedness property. Thus, we will be able to appeal to
these rankings in providing constraints on how non-promotional reasons can vary
between sets of alternatives. To take just one example, if some fact r explains why
A does a better job at respecting value V than B does, and why B does better
than C, it will likewise explain why A does better than C. So we can see how the
Transitivity constraint can be captured.
A potential dierence between promotional and non-promotional rankings is
that ties and perhaps even incommensurability may be more common in non-
promotional rankings. This is just because, for many values, there are several
dierent ways of respecting or honoring them, and these may be equally or incom-
mensurably good. This will plausibly be less common in the promotional case (but
probably not nonexistent). This is no problem for the view I am developing. Sup-
pose it is your friend's birthday. And suppose that baking your friend a chocolate
birthday cake is an equally good way of respecting the value of friendship as baking
her a lemon birthday cake, while baking her no cake at all is not a good way of
respecting this value. In a case like this, the fact that it is your friend's birthday is
a reason to bake her a chocolate cake rather than bake her no cake, and a reason
to bake her a lemon cake rather than bake her no cake, but not a reason to bake
her a chocolate cake rather than bake her a lemon cake. And we can just explain
this by appealing to the ranking of the alternatives `bake a chocolate cake', `bake
a lemon cake', and `bake no cake': the rst two are tied, the third is ranked below
them, in terms of how well they respect the value of friendship.
Let me sum up what I have done in the last two sections. In developing my
contrastivist theory of reasons, it was convenient to make two assumptions. First,
I assumed that all reasons involve promotion, and second, I assumed that the
relevant kind of promotion is to be understood as probability-raising. Both of these
assumptions are controversial. In the previous section I showed how to relax the
second assumption|all we need are rankings of alternatives with certain minimal
properties. In this section, I have shown that we can also relax the rst assumption.
This is because we can still appeal to rankings of alternatives, not in terms of how
well they promote a given objective, but rather in terms of how well they respect or
honor it. This is enough to secure the constraints on how reasons can vary between
sets of alternatives.
75
4.7 Conclusion
In this chapter, I've shown that an independent source of motivation for con-
trastivism, the relationship between reasons and promotion, leads very naturally
to a version of the theory that straightforwardly solves an important problem for
the contrastivist. By appealing to this idea, we can capture intuitive relationships
between reasons relative to dierent sets of alternatives without sacricing the inde-
pendence that lets the contrastivist solve problems facing non-contrastive theories.
Reasons relative to dierent sets of alternatives are independent of each other, but
the theory also has an important kind of structure.
The theory I developed here forms the heart of this dissertation. In the next
two chapters, I'll apply the theory to some important issues in ethics, practical
reasoning, and epistemology. I'll show, rst, that contrastivism allows us to frame
important questions in new and interesting ways, and second, that doing so lets us
make some progress on them.
Before moving on, though, I want to take stock of where we are, in terms of
the depth of contrastivism. In Chapter 3, I argued against a view I called shallow
contrastivism, which holds that contrastivism only goes \language-deep", and in
favor of deep contrastivism, which holds that the important (possibly fundamental)
normative reason|or favoring|relation is itself contrastive. Having developed the
theory further in this chapter, however, we may wonder how deep contrastivism
really goes. After all, the key to providing the constraints on how reasons can
vary between sets of alternatives was to let things bottom out in something non-
contrastive|perhaps conditional probabilities, perhaps something else that will
give us the right kind of ranking.
This is an important point. I think what the need for constraints, which I
motivated at the beginning of this chapter, shows is that we cannot adopt an
extremely deep form of contrastivism, at least not without incurring signicant
costs. So the theory I am endorsing is a somewhat moderate form of contrastivism.
A deeper version would hold that there is no non-contrastive foundation: perhaps
we can rank a given set of alternatives, but how actions are ranked against each
other itself would vary if the relevant alternatives changed. For example, A may
be ranked higher than B relative to a setfA, B, Cg, but once we introduce D,
maybe B would be ranked higher than A. Perhaps there are some arguments in
76
favor of an extremely deep form of contrastivism like this, but I am not aware of
any. And given the signicant costs this theory would take on|in particular, the
cost of losing the constraints I've been discussing in this chapter|we would have
to think hard about how strong such arguments were.
But even though this theory is somewhat moderate, it is in opposition to nearly
all existing work on reasons, which has traditionally assumed without argument that
reasons are reasons for things simpliciter. Moreover, if we really can just appeal
to something like conditional probabilities, then the non-contrastive foundation
is not anything normative. (This may not be the case if we end up having to
appeal to something like a ranking of alternatives in terms of how closely they
approximate the ideal kind of honoring of some value.) So it may turn out that
all the normative notions are contrastive, even if they bottom out in some non-
contrastive, non-normative notion like conditional probability.
77
Chapter 5
Intransitivity
In Chapter 4, I developed a version of contrastivism (that we can call moderate
contrastivism) that rules out a troubling kind of intransitivity of reasons, where r
is a reason for A rather than B and for B rather than C, but not for A rather
thanC. This is an important result because (i) contrastivism, described abstractly
as the view that reasons are only reasons relative to sets of alternatives, initially
seemed to allow for intransitivity, and (ii) allowing for this kind of intransitivity is
plausibly sucient reason to reject a theory.
In this chapter, though, I introduce some pressure from the other direction:
several writers have given powerful arguments that there really are cases of intran-
sitivity in ethics and practical reasoning.
1
If they're correct|or even if we think
they might be correct|a theory of reasons that simply rules out cases of intransi-
tivity by denition looks problematic.
Importantly, from this perspective, an initial attraction of contrastivism, again
described independently of the framework from Chapter 4, is that it seems better
suited than non-contrastive theories to accommodate intransitivity. After all, a
popular diagnosis of why intransitivity arises is that some moral factors are essen-
tially comparative: whether or not a factor matters, or provides reasons, depends
on the specic comparison being made.
2
If this thought is correct, then it seems
that whether or not a fact is a reason for some action will depend on what you're
comparing that action to, or to the relevant set of alternative actions, since the
factors or ends that provide reasons only do so relative to some alternatives. And
1
See Temkin (1987, 1996, 2012); Rachels (1998, 2001); Friedman (2009).
2
See in particular Temkin (1987, 1996, 2012).
78
that is the central contrastivist claim. So the objection that we don't even want to
rule out intransitivity is especially sharp for the contrastivist.
A more general goal of this chapter is to give some support to the claim that
contrastivism lets us make progress in important debates in normative philosophy.
I hope to have shown in the rst four chapters of the dissertation that the theory
is independently well-motivated. In this chapter and the next, I want to show that
it's also theoretically useful. If I'm right that the view I suggest in this chapter
is a promising way to accommodate intransitivity, then we can use this view to
frame important questions about that issue, since it will be a concrete example of
a theory that allows for intransitivity. I want to emphasize right from the start,
though, that it is not my goal to establish the possibility of intransitivity. Rather,
I want to show that contrastivism sheds light on the debate, by showing us how to
accommodate intransitivity, if we want to do so.
5.1 Transitivity and Reasons
First I'll say a bit about what kind of intransitivity I'll be concerned with here. It
is dierent than the sort that is most often discussed in the literature, but the two
sorts are connected.
The most frequently discussed normative relation in the literature on intransitiv-
ity is the `better than' relation.
3
Here is the principle at issue in those discussions:
Transitivity of `Better Than': If A is better than B, and B is better than C,
then A is better than C.
For the purposes of this chapter, I'll treat the things that are ranked (A,B, andC)
as outcomes or states of aairs. Arguments against this principle tend to describe
cases (often bizarre, yet possible, cases) in which the intuitive ranking of three
(or more) states of aairs does not satisfy this principle. And frequently these
arguments are based on cases devised by Part (1984), who does not, for what it's
worth, think his cases show that `better than' is intransitive. I'll present arguments
in this vein in the next section.
3
See, for example, Temkin (1987, 1996, 2012); Rachels (1998, 2001); Friedman (2009) for
arguments in support of the intransitivity of `better than'. Broome (1991) argues against this
kind of intransitivity.
79
Since I am concerned with reasons, I'll need to amend these arguments for the
possibility of intransitivity of `better than' to apply to reasons. Fortunately, this is
easy to do, given the following linking principle:
Reasons and Betterness: A is better than B i there is more reason to choose
A than to choose B, given the choice between A and B.
This principle should be amenable both to those who want to analyze value in
terms of reasons (for example, Scanlon (1998)) and to those who want to analyze
reasons in terms of value. Even if you don't think either of these sorts of analysis
can succeed, the principle is very plausible.
Many writers have been attracted to tting-attitudes accounts of value, on which
to be good is just to be a tting object of a pro-attitude, where ttingness is of-
ten, though not always, analyzed in terms of reasons.
4
Reasons and Betterness
is attractive for some of the same reasons that these accounts are attractive. In
particular, as many writers have pointed out, there are important connections be-
tween reasons and value. For example, if some outcome is valuable, then it is very
plausible that there's a reason to choose that outcome, if given the choice. Of
course, there might be stronger reasons to choose some other outcome, but if that's
so, then plausibly it is because the other outcome is more valuable|or better|
in some way.
5
Reasons and Betterness predicts that there should be such a
connection.
But importantly, Reasons and Betterness does not require the truth of any
tting-attitudes account. We might instead take it as (part of) an analysis of
reasons to choose in terms of value or betterness. Or we might think that there's
some common third normative property that aects both the value and strength
of the reasons to choose the outcome. So again, this principle is amenable to very
dierent theories about the relationship between reasons and betterness.
The clause, `given the choice between A and B' is important here. That's
because reasons to choose, like all reasons, are reasons for agents. But it is con-
troversial whether the value of an outcome is similarly connected to agents. We
might have two outcomes, A and B, such that A is intuitively better than B, even
4
See Scanlon (1998); Danielsson and Olson (2007); Rabinowicz and Rnnow-Rasmussen (2004);
Suikkanen (2005, 2009); Reisner (2009); Schroeder (2010); Way (fc) for a sampling of this litera-
ture.
5
See, in particular, Way (fc) on this point.
80
though no agent could have any reason to chooseA because no agent would ever be
in a position to make such a choice. The `given the choice' clause blocks this. And
moreover, in the cases I'll discuss, I'll simply stipulate that the agents involved face
a choice between the relevant outcomes.
So while arguments in the intransitivity literature present cases that challenge
Transitivity of `Better Than', I'll make use of these cases in challenging the
following principle:
Transitivity of `More Reason': If there is more reason to choose A than to
chooseB, and more reason to chooseB than to chooseC, then there is more
reason to choose A than to choose C.
I think this principle is very plausible, and that nearly everyone (except advocates
of intransitivity that I mentioned above) would accept it, even independently of
Reasons and Betterness.
Often these kinds of transitivity principles are simply taken for granted. Temkin
(1987) hypothesizes that Part (1984) took it for granted that `better than' (all-
things-considered) is transitive, and that this in
uenced his reaction to the cases
he presents, which I'll discuss in the next section. Broome (1991) claims that
the transitivity of `better than' is essentially a logical truth: \Whatever decisions
we make, we shall always be guided by logic to preserve the transitivity of the
comparative. [. . . ] Logic requires betterness to be transitive".
6
The phrase `more
reason' suggests a view like this. It encourages us to think in terms of a linear scale,
like the number line, so that to say that there is more reason to choose A than to
choose B is like saying that choosing A falls to the right of choosing B on some
line. And once we have this picture in mind, it's easy to see why we would think
that `more reason' is transitive, since `to the right of' is transitive. Nevertheless,
I'll rehearse arguments in the next section that call these transitivity principles into
question.
Before moving on, I should note one qualication. Some people think that two
outcomes can be incommensurable in value|that the value of the two outcomes
cannot be compared. If A and C are two such outcomes, we might have that A
is better than B and B is better than C without having that A is better than
C. This could hold if B is commensurable with both A and C, though A and C
6
Broome (1991), pp. 12, 136.
81
are incommensurable with one another. These kinds of cases, if they exist, are
not generally treated as counterexamples to transitivity principles, but rather as
exceptions. I could make this explicit by adding a clause to the principles like
`and A and C are commensurable', but I'll stick with the simpler formulation, and
simply assume that all the outcomes I'm discussing are commensurable.
5.2 Intransitivity
The most popular kind of argument for the possibility of intransitivity is based
on cases from Part (1984). Temkin (1987, 1996); Rachels (1998, 2001); Friedman
(2009) all give arguments based on Part's cases. These cases are an obvious place
to begin because (i) denying the transitivity assumption is one way to avoid the
problematic results, and (ii) other proposed solutions are unsatisfactory. That's
not to say that many people nd the solution of denying transitivity satisfactory,
of course, but it does mean that we should at least give this solution a chance.
The main argument I'll present here is essentially just what I think is the most
compelling way to present the common thread from these other arguments. So I
make no claim to originality.
First I'll show how one of Part's cases combined with Transitivity of `More
Reason' leads to what Part (1984) calls the Repugnant Conclusion. Part, and
most other people, think this conclusion just can't be true. But the common re-
sponse is to reject a particular step in the argument other than the assumption
of transitivity. But as advocates of intransitivity have shown, this is shortsighted:
there are big problems with this strategy, and moreover, it leads us right back
into a case|Part's Mere Addition Paradox|which calls Transitivity of `More
Reason' into question.
5.2.1 The Repugnant Conclusion
Consider the following case from Part (1984).
7
Suppose a powerful being asks
you to choose between twenty-six outcomes of the universe. Outcome A
8
contains
7
See also Rachels (1998, 2001); Friedman (2009). As I said above, these authors are writing
about `better than (all things considered)', while I'm writing about `more reason than'.
8
Non-italicized capital letters are just names of outcomes that I've described. Italicized capital
letters are variables ranging over outcomes.
82
some relatively large population, say ten billion, all with a very high level of well-
being. B contains a larger population|say for deniteness that it's twice as large
as A|all with a lower, but still quite high, level of well-being. This reduced level of
well-being is not due to any sort of unfairness or other moral deciency|roughly,
all the moral factors except for total and average utility are the same in the two
outcomes. I think that, intuitively, there is more reason to choose B than to choose
A. Agree with me for now, and I'll address worries about this rst step below.
Similarly, there is more reason to choose C than to choose B|it's twice as large,
and everyone has only a slightly lower level of well-being. So by Transitivity of
`More Reason', there is more reason to choose C than to choose A.
Continue this sequence until we get to Z. The population in Z is enormous|one
billion times 2
25
. Everyone in Z has a level of well-being that makes their lives barely
worth living (\muzak and potatoes", as Part would say). Well, there's more reason
to choose B than to choose A, more reason to choose C than to choose B, more
reason to choose D than to choose C, . . . , more reason to choose Y than to choose
X, and more reason to choose Z than to choose Y. So if Transitivity of `More
Reason' is true, it follows that there's more reason to choose Z than to choose A.
But this is what Part calls the Repugnant Conclusion: that for any outcome with
a population of at least ten billion people, all of whom have unbelievably awesome
lives, there is some possible outcome with a much larger population all of whose
lives are barely worth living, such that there's more reason to choose this second
outcome than the rst.
This result seems wrong. But every step in the sequence seemed correct|as we
move through the options in alphabetical order, there seemed to be more reason
to choose each option than to choose the one before it. If we want to retain both
of these judgments|that the Repugnant Conclusion is false and that there's more
reason to choose each option in the sequence than the one before it|it looks like
we have to give up on Transitivity of `More Reason'. That's the rst argument
for intransitivity.
9;10
9
Part (1984) considers and rejects several other kinds of responses, including an appeal to
a valueless level: a level of well-being below which one's life doesn't contribute to the overall
goodness of the outcome, or how much reason we would have to choose that outcome, though the
life is still valuable to the one who lives it, since it is worth living. There are big problems with
this view, as well as other attempts to avoid the Repugnant Conclusion.
10
Rachels (1998) runs a similar argument by
ipping the diagram upside down, and focusing
on just one life for extremely long stretches of time. The idea is that in outcome A, you face very
83
5.2.2 Resisting the rst step
One way to block this argument, of course, is to resist one of the steps along the
chain. I'll focus on the rst step, from A to B, but I don't see why this choice
should make a signicant dierence. Part himself actually thinks the Repugnant
Conclusion forces us to deny that there is more reason to choose B than to choose
A, because he assumes that transitivity holds.
This strategy seems to require us to say that there is in fact more reason to
choose A than to choose B. If we said merely that there is not more reason to
choose B than to choose A, then we should say, by parity of reasoning, that there's
not more reason to choose C than to choose B, and so on. Then we could say that
there's not more reason to choose Z than to choose A. But if we don't say that
there's more reason to choose A than to choose B, and so more reason to choose B
than C, and so on, then transitivity-like considerations force us to say that there's
not more reason to choose A than to choose Z. But this is also intuitively wrong|
it isn't just that we think there's not more reason to choose Z than to choose A;
we think there is more reason to choose A than to choose Z. The principle this
argument relies on is the following.
Transitivity of `Not More Reason': If there's not more reason to choose A
than to choose B, and not more reason to choose B than to choose C, then
there cannot be more reason to choose A than to choose C.
And this is very similar to Transitivity of `More Reason'.
Some writers who accept Transitivity of `More Reason' would be willing to
deny this last principle. Cases of incommensurability provide the most compelling
intense pain for one year, while in B you face still intense, but slightly less intense, pain for one
hundred years, and so on. In Z, you face very mild discomfort (with no redeeming qualities|
you'd rather be temporarily unconscious, rather than experience the discomfort) for eons and
eons. Each step|from A to B, from B to C, . . . , from Y to Z|seems to be a step for the worse.
So A is better than B is better than C . . . is better than Y is better than Z. But then if `better
than' is transitive, it follows that A is better than Z. But intuitively, Rachels thinks, Z is actually
better than A|experiencing the very mild pain in Z, even for eons and eons, is not as bad as
experiencing the most intense pain imaginable in A, even for only one year (make it 100 years
and increase everything else accordingly, if you aren't convinced). One potentially nice feature
of this example, that I don't want to get into here, is that arguably the reasons involved are all
agent-relative reasons: since you're the one that's going to suer the pain, they're reasons for
you to choose the relevant option. Of course, if I was given the choice between which of these
fates you would suer, I would have agent-neutral reasons to choose the same options. But that's
compatible with you having agent-relative reasons.
84
counterexamples. Broome (1999) presents the following sort of case. Consider three
possible careers: (i) a career as a lawyer, (ii) a career as a lawyer making $1,000
less per year, but which is otherwise identical to career (i), and (iii) a career as
an academic. If the careers as a lawyer and career as an academic are suciently
satisfying, it's plausible that you don't have more reason to choose career (i) than
career (iii), or more reason to choose career (iii) than career (ii). The intuition here
is that there a career as a lawyer and a career as an academic are simply valuable
in dierent ways, so that neither is better than the other. And $1,000 per year is
just not enough money to make a dierence here. But if all of this is right, it would
follow from the Transitivity of `Not More Reason' that you don't have more
reason to choose career (i) than career (ii). But that's clearly false|since the two
careers are otherwise identical, an extra $1,000 per year is exactly the kind of thing
to give you more reason to choose (i) than to choose (ii). So there's some reason to
think that this principle is false. This is far from uncontroversial, though.
But even if we deny the Transitivity of `Not More Reason' while retaining
Transitivity of `More Reason', we could just make B contain many more people
with levels of well-being barely below the level in A. And that sort of modication
clearly gives us some additional reason to choose B. If there wasn't more reason
to choose A than to choose B to begin with, it seems like this sort of modication
should give us more reason to choose B than to choose A. That doesn't follow
immediately, of course. But if this sort of modication doesn't give us more reason
to choose B than to choose A, we need some explanation for why.
So why think that there's more reason to choose A than to choose B? We're
assuming that the outcomes are equivalent with respect to all the moral factors that
they can be equivalent with respect to, while still exhibiting the relevant dierences
in total and average utility. So some moral factor with respect to which A is better
than B must be making the dierence, if there's more reason to choose A than to
choose B. That is, the reasons provided by one of these moral factors must tip the
scales in favor of A. The most obvious candidate is average utility.
11
Outcome A has a higher level of average utility than B, even though there's
more total utility in B. If we want to appeal to average utility to resist the rst
step|to say that there's actually more reason to choose A than to choose B|we
have to say that the dierence in average utility is somehow more important than
11
See Temkin (1987); Friedman (2009) for more thorough discussions of this sort of strategy.
85
the dierence in total utility. There are two ways to
esh out this response. First,
we could say that average utility is one very important moral factor while admitting
that total utility is also important, though less so, in determining what we have
most reason to choose. Second, we could say that any increase in average utility is
always better than any increase in total utility.
Suppose we take the rst strategy. Then we admit that the fact that B has
more total utility than A gives us some reason to choose B rather than A. But we
just insist that the fact that A has higher average utility gives us stronger reason to
choose A rather than B. Since A and B are identical with respect to other potentially
relevant moral factors|we're assuming, for instance, that any violations of duties
that take place in B are matched in A|we can thus deny that there's more reason
to choose B than to choose A.
12
But as long as we admit that total utility gives us
some reason to choose B rather than A, we can simply alter the case so that B has
much more total utility than A and very slightly lower average utility. If we think
total utility matters at all in this case, it's implausible that this sort of modication
could never
ip the scales, so that the vast increase in total utility nally outweighs
the tiny increase in level of average well-being.
This pushes us toward the second strategy: no increase in total utility can make
up for any deciency in average utility. So no matter how much more total utility
B has, and no matter how small the dierence in the level of average utility, there
will always be more reason to choose A than to choose B. The problem with this
strategy is that it's simply implausible. Consider a version of the case in which
outcome A has just one person with a very high level of well-being, and outcome B
has tens of billions of people all with this same level of well-being as the person in
outcome A, plus a single person with a very slightly lower level of well-being. We
can even imagine that this single person lives lightyears away from the others, so
that he's not even aware of anyone with higher levels of well-being than himself.
13
The advocate of the second strategy has to hold that we still have more reason to
choose outcome A than outcome B. And that is obviously implausible.
12
Temkin (1987) considers the moral factor of perfection. See that paper for arguments that
that sort of strategy fails. See that paper, as well as Friedman (2009), for more discussion of the
objections I raise here.
13
Compare Part (1984)'s Mere Addition Paradox|more on this later.
86
5.2.3 The Mere Addition Paradox
I've been arguing that we have more reason to choose B than to choose A, given the
choice. And this sort of reasoning, along with Transitivity of `More Reason',
leads to the Repugnant Conclusion. This gives us some reason to question this
transitivity principle. But I haven't considered all the possible ways one might
resist the rst step of the argument (or some other step along the chain). So in
this subsection I'm going to present an argument that, even if we nd some way to
resist the rst step, and claim that there's more reason to choose A than to choose
B, we still have reason to question Transitivity of `More Reason'.
14
The argument turns on Part (1984)'s Mere Addition Paradox. Part was forced
to face this paradox by denying that B is better than A, in order to avoid the
Repugnant Conclusion. For my purposes, of course, we face the version I'll present
by denying that there's more reason to choose B than to choose A. Imagine an
outcome A+ that contains all the same people as A, plus some additional people
whose lives are well worth living, but whose levels of well-being are lower than that
of the original group. But the two populations are totally unaware of one another.
Imagine you are forced to choose between these three outcomes. Part argues that
there is denitely not more reason to choose A than to choose A+|we can not
make an outcome worse by adding happy people to it. There is a drop in equality
and in average well-being, but we haven't introduced any inequality via any sort of
injustice. This sort of addition is what Part calls mere addition, and he claims that
mere addition can never make an outcome worse. So we don't have more reason to
choose A than to choose A+.
15
A+ contains just as many people as B. Part argues that there's more reason
to choose B than A+: I stipulate that B is better with respect to equality, total
well-being, and average well-being. Thus, it's hard to see what would motivate us
to think there's more reason to choose A+ than to choose B.
But remember that we're assuming that there's more reason to choose A than to
choose B, in order to avoid the Repugnant Conclusion while retaining Transitivity
of `More Reason'. Note again that it's not easy to see how to avoid the Repugnant
Conclusion if we say merely that there's not more reason to choose B than to
14
See Friedman (2009).
15
I think that we plausibly have more reason to choose A+ than to choose A. Part denies this.
But nothing really turns on this for my purposes here.
87
choose A|we could just double the size of B and make the level of well-being of
its inhabitants even closer to the level of the inhabitants of A. But if there's more
reason to choose A than to choose B, and more reason to choose B than to choose
A+ then it follows from Transitivity of `More Reason' that there's more reason
to choose A than to choose A+. But Part argued, and it seems very plausible that,
this is actually false. This is the Mere Addition Paradox. And we got ourselves
into it by claiming that there's more reason to choose A than to choose B|that
is, by resisting the rst step of the argument for the Repugnant Conclusion. And
we could give a version of the Mere Addition Paradox for any adjacent members of
the chain (for F and G, rather than for A and B, say). So once we try to avoid the
Repugnant Conclusion, we walk right into the Mere Addition Paradox, which also
gives us reason to question Transitivity of `More Reason'.
5.2.4 Intransitivity and contrastivism
In this section, I've argued that Part's cases give us reason to question Tran-
sitivity of `More Reason'. I've considered some ways to resist the argument,
and shown that they either lead to implausible results or else lead us right back to
problems for transitivity. So we should take seriously the idea that `more reason
than' is intransitive.
This puts some pressure on the version of contrastivism I developed in Chapter
4, which seems to simply rule out intransitivity by denition. This is even more
striking, given that an initial attraction of a contrastive theory of reasons might be
that it seems better placed to deny Transitivity of `More Reason'. The idea that
reasons relative to dierent sets of alternatives are independent of one another|
that settling what the reasons are relative to one set does not tell us what the
reasons are relative to a dierent set|might make us think contrastivism is well-
placed to accommodate intransitivity. That's because, as I emphasized in Chapter
4, transitivity seems to amount to a constraint on this independence. But I showed
that the framework I developed there actually does constrain this independence.
It cannot turn out, according to that framework, that r is a reason for A rather
than B and for B rather than C, but not a reason for A rather than C. So if the
arguments above are at least compelling enough that we don't want to rule out the
88
possibility of intransitivity by denition, there seems to be a problem for the view
I developed in the last chapter.
5.3 Two Kinds of Intransitivity
In this section I'll distinguish between two kinds of intransitivity. I'll argue that
the sort of intransitivity that the view from Chapter 4 rules out is not the sort that
the arguments from the last section purport to establish.
First, we might have what I'll call intransitivity of reasons. If this sort of in-
transitivity is possible, then there might be a case in which some fact r is a reason
forA rather thanB and forB rather thanC, but not a reason forA rather thanC.
Recall that in Chapter 4, I pointed out that the intuitive transitivity requirement
on reasons is even stronger, once we think of reasons in terms of promotion. It isn't
just that whenever r is a reason for A rather than B and for B rather than C, it
must be a reason for A rather than C. The end (whatever lls in for `X' in the
conditional probabilities) that explains why r is a reason for A rather than B and
for B rather than C (if there is a common end that explains both of these things)
must also explain why r is a reason for A rather than C. The intuition behind
a transitivity requirement wouldn't really be captured if something else explained
whyr is a reason forA rather thanC. This is the variety of intransitivity that the
framework from Chapter 4 straightforwardly rules out.
Second, we might have what I'll call intransitivity of `more reason'. If this sort
of intransitivity is possible, then there might be a case in which there is more reason
for A than for B out offA, Bg and more reason for B than for C out offB, Cg,
but not more reason for A than for C out offA, Cg.
The arguments for intransitivity from the last section, if they're successful, only
establish the intransitivity of `more reason'. What seems wrong is that there's more
reason to choose outcome Z that to choose outcome A. But it is not counterintuitive
to say that there is some reason to choose outcome Z rather than outcome A. In
particular, considerations of total utility provide a reason to choose outcome Z
rather than outcome A. The fact that, say, more utility is better than less is a
reason to choose B rather than A, a reason to choose C rather than B, . . . , and a
reason to choose Z rather than Y. It follows from the transitivity of reasons that
it's a reason to choose Z rather than A. And this seems correct. It's just that the
89
reason is outweighed by the reasons to choose A rather than choosing Z, explained
by some other factor, like average utility.
If this is right, then it's not simply that the cases don't support the intransitivity
of reasons. They actually oer strong support for the transitivity of reasons. If this
is the most compelling support for intransitivity|and the fact that advocates of
intransitivity almost invariably appeal to these cases is some evidence for this|
and it seems wrong to describe them as cases of intransitivity of reasons, that casts
serious doubt on this kind of intransitivity.
5.4 Contrast-Sensitive Importance
Even once we recognize the dierence between these two types of intransitivity,
it's still a bit mysterious how we could have intransitivity of `more reason' without
the intransitivity of reasons. In this section I'll show how to pry them apart. I'll
oer a way of accommodating the intransitivity of `more reason' in the framework
from Chapter 4, even though it clearly rules out the intransitivity of reasons. I'll
take this in two steps. First I'll observe that the strength of reasons depends on the
importance of the ends (desires, values, and so on) that provide those reasons. Then
I'll suggest that the importance of a given end might itself be a contrast-sensitive
matter: an end that matters a lot in making some comparisons might matter very
little in making other comparisons. So the resulting view is that, even if a fact is a
reason for some action relative to both of two dierent sets of alternatives, it might
be a much weightier reason relative to one of them, due to the contrast-sensitivity
of the importance of the end that provides that reason.
5.4.1 Strength of reasons and importance of ends
I argued in the last section that Part's cases don't push us towards the intransitiv-
ity of reasons, but only towards the intransitivity of `more reason'. But this seems
unstable. The intransitivity of reasons appears to be necessary for the intransitivity
of `more reason'. After all, if any fact that's a reason for A rather than B and for
B rather than C is thereby a reason for A rather than C, the \amount" of reason,
or the combined strength of the reasons, would also seem to be transitive. Thus, it
90
seems that since the view from Chapter 4 rules out the intransitivity of reasons, it
also rules out the intransitivity of `more reason'.
But in fact the intransitivity of `more reason than' does not require the intransi-
tivity of reasons. An important rst step is recognizing that the combined strength
or weight of the reasons for A doesn't depend only, or even most importantly, on
how many reasons there are. Frequently there are more reasons to do A than B,
but stronger reason to doB. The fact that I'd get my shoes muddy is a reason not
to run into the pond to save the drowning child, as is the fact that I would be late
for the movie. The fact that the child will die if I don't is a reason to save him. So
that's two reasons not to save and only one reason to save. But clearly the reason
to save is stronger than the combined strength of the reasons not to save. And the
explanation for this is that the end that explains why I have this reason|saving
the child's life|is much more important than the ends that explain why I have
the reasons not to save|keeping my shoes clean and seeing the movie. So how
much reason there is to perform an action depends|in large part, at least|on the
importance of the ends that provide those reasons.
An important upshot of this observation is that we cannot simply read o how
strong a particular reason toA is, or what the combined strength of the reasons to
A is, o of the facts about conditional probabilities that I used in Chapter 4. We
need some account of the relative importance of the various reason-providing ends.
I don't have an account of this, but there are intuitively clear cases, like the one I
mentioned above involving the drowning child.
16
5.4.2 Contrast-sensitive importance
It's still not clear, though, how to allow for the intransitivity of `more reason' once
we rule out the intransitivity of reasons. Even if the strength of an X-provided
reason forA out offA,Bg and forB out offB,Cg depends on the importance of
X, the strength of the X-provided reason to A out offA, Cg will likewise depend
on the importance of X. And if X is very important, then it's going to provide
16
It doesn't follow, though, that the facts about conditional probabilities are irrelevant. First,
and most importantly for my purposes, the conditional probabilities establish what reasons there
are, relative to a given set of alternatives. This is very important, because it lets us retain the
nice entailment relations I established in Chapter 4. I'll have a bit more to say about this shortly.
Second, facts about conditional probabilities can arguably make a dierence when the ends that
provide two reasons are equally important.
91
relatively strong reasons across all of these cases. So the transitivity of `more reason'
still seems to follow from the transitivity of reasons.
To see that this is wrong, it might be instructive to rst see how we could allow
for the intransitivity of reasons while retaining (something like) the probabilistic
framework from Chapter 4. What we would have to do is to let the ends that
provide reasons|the Xs|vary with sets of alternatives. If X provides reasons
relative tofA, Bg andfB, Cg, but not relative tofA, Cg, then we could have
cases of intransitivity of reasons. The most important cost to adopting this kind
of picture, though, is that we would lose the entailment relations between reasons
relative to dierent, but related, sets that I established in Chapter 4. These are
attractive results, so we should try to retain them if possible.
17
The framework from Chapter 4 delivered relations between what reasons there
are relative to dierent sets. Changing what ends can provide reasons between
sets of alternatives leads to a dierence in these facts. So to retain the entailment
relations, we need to hold the ends xed between sets of alternatives. No matter
what the relevant set of alternatives is, there is one class of ends that provide
reasons, though they'll provide reasons for dierent things relative to dierent sets
of alternatives, of course, depending on which alternatives in the set best promote
them.
But as I argued above, the strength of the reasons for some action depends
on more than what reasons there are: it depends on the importance of the ends
that provide those reasons. So the way to accommodate the intransitivity of `more
reason' while retaining the transitivity of reasons (and so the entailment relations)
is to let the relative importance of the ends vary between sets of alternatives. Some
endX always gives us reasons, but these reasons are more or less weighty, depending
on the set of alternatives, since the importance of the ends varies with sets of
alternatives.
18
Thus, even if every fact that is a reason for A rather than B and
for B rather than C is thereby a reason for A rather than C, the weights of these
17
One way to try to retain these entailment relations on this more radical picture would be to
place some restrictions on how the reason-providing ends could vary between sets of alternatives.
I won't explore this option here, since as I argued above, the arguments for intransitivity don't
even purport to establish intransitivity of reasons.
18
If there are some cases|though I haven't seen any|that do seem to establish the intransitivity
of reasons, we could even accommodate that, or something like it, on this picture. We would just
let some ends provide reasons of no weight, relative to some sets of alternatives. So it would still
true to say that the fact is a reason, and we could retain the entailment relations. It's just that
the reasons have no weight.
92
reasons may vary depending on the comparison. We might have some very weighty
reasons forA rather thanB and forB rather thanC which are much weaker reasons
forA rather thanC, because the relative importance of the ends that provide these
reasons varies with specic comparisons being made. Then we can fail to have more
reason for A than for C without giving up on the transitivity of reasons.
This is a way of spelling out the intuition, often appealed to in the anti-
transitivity literature, that some values or ends are \essentially comparative".
19
Dierences with respect to some value like equality, total utility, or average utility
might make a big dierence when comparing some alternatives, but matter much
less when comparing other alternatives. We can allow that having more total utility,
for example, is always a reason to choose an outcome rather than choosing one with
less total utility. But depending on other features of the alternatives, this reason
can be stronger or weaker. And that's because considerations of total utility might
be more or less important depending on the comparison.
Here's an illustrative and particularly relevant example. When comparing two
outcomes with relatively close levels of average utility, like any two adjacent out-
comes in the sequence I used in the argument for the Repugnant Conclusion, total
utility is more important than average utility. That's why, when comparing any
two adjacent outcomes, there's always more reason to choose the one later in the
sequence|more reason to choose G than to choose F out offchoose F, choose
Gg, for example. But when we have vast dierences in average utility, total utility
is much less important than average utility. If that's right, it would explain why
there's more reason to choose A than to choose Z, even though there's so much
more total utility in Z. Even though considerations of total utility do provide a
reason to choose Z out offchoose A, choose Zg, this reason is much weaker than
the reason to choose A provided by average utility. That's why there's more reason
to choose A than to choose Z out offchoose A, choose Zg. On the other hand,
the reason provided by average utility to choose A out offchoose A, choose Bg
is apparently weaker than the reason provided by total utility to choose B out of
this set|that's why we think there's more reason to choose B. The importance of
total utility depends (probably among other things) on the relative levels of average
utility, and vice versa.
19
See especially Temkin (1987, 1996, 2012). See also Rachels (2001), p. 218: \[A] factor deter-
mining how A&Z compare diers dramatically in signicance from how it gures in comparing
A&B, B&C, C&D, and so on".
93
5.4.3 Contrastivism about reasons and contrastivism about
importance
Contrastivism about reasons is the view that whether or not some consideration
is a reason for an action or attitude depends on the alternatives. The strategy for
accommodating the intransitivity of `more reason' that I've just outlined involves
making the importance of reason-providing ends, and thus the strength of reasons,
depend on the alternatives. What is the relationship between these two ideas?
First, we can be contrastivists about reasons without adopting a contrast-
sensitive view of the importance of ends. That is, the strategy I've outlined in
this section is detachable from the rest of the contrastivist theory I've developed in
previous chapters. So if you do not nd the arguments for intransitivity compelling,
or just nd the diculties that come along with intransitivity (some of which I'll
discuss in the next section) too troubling, that's no reason not to be a contrastivist.
You can just hold the importance of ends xed across sets of alternatives to rule
out intransitivity.
Second, we can adopt a contrast-sensitive view of the importance of ends without
being contrastivists about reasons. On this kind of view, whether or not some fact is
a reason forA does not depend on the alternatives, though the weight of that reason
does depend on the alternatives. For example, the total utility in Z always provides
a reason to choose Z (simpliciter, of course). But when we compare choosing Z
to choosing A, that reason is simply not very weighty, though it is quite weighty
when we compare choosing Z with choosing Y. Note that this is precisely what
the contrastivist says about these particular comparisons. The dierence between
the two views will be the following. The non-contrastivist about reasons says that
the total utility in A provides a reason to choose A even when we compare it to
choosing Z. But the contrastivist denies this: total utility doesn't provide any reason
to choose A when we compare it with choosing Z (or, as the contrastivist would
rather say, no reason to choose A rather than choose Z). But the important point
for now is that the non-contrastivist about reasons can still be a contrastivist about
importance, and accommodate intransitivity in the way I've suggested here.
It's worth pointing out, though, that accepting contrast-sensitive importance
is a particularly natural thing for a contrastivist to do. The contrastivist about
reasons is impressed by the role that relevant alternatives seem to have in our
94
judgments about what is a reason for what. So thinking that alternatives can also
matter for how strong those reasons are is right up the contrastivist's alley.
5.5 Remaining Questions
In this section I'll discuss some remaining issues for the view I've developed here.
Throughout the discussion, I'll
ag issues that would need to be resolved before we
should feel comfortable adopting the account in this chapter. Thus, the upshot of
this discussion isn't that this account is the correct one. It's rather that it oers a
plausible start, if we want to accommodate intransitivity.
5.5.1 What should I do?
An important question for any theory that allows for intransitivity of `more reason'
concerns how we should decide what to do. A natural view is that you should do
what you have most reason to do. But consider the imagined choice you face in the
Repugnant Conclusion case:fchoose A, choose B, ..., choose Zg. What out of this
set do you have most reason to do? In Chapter 3, I adopted CRO, which says that
we have to look at the two-member subsets: the option you should take out of the
larger set is the one that wins in all of its pairwise comparisons.
So suppose we know what to say for any two member subset of this larger set.
20
Well, part of the point of this case was that nothing wins in all of the pairwise
comparisons. And this is the result that the account I developed here gives. So the
account I've oered so far doesn't tell us what to do in this case. And a theory of
reasons, we might hope, will tell us what to do, once we know the facts about the
reasons and their weights. After all, what we should do is intuitively determined
by our reasons.
I think that in fact a theory should not tell us that we should perform some
particular alternative out of the larger set in this case. That's just because any
answer the theory gives will seem intuitively wrong. If the theory gave us some
answer here, there are two possibilities. First, it might say that we should choose
Z. Well, that seems wrong, since there's more reason to choose A. Second, it might
20
This doesn't actually seem true to me|what should we say aboutfchoose A, choose Ng?
More on this shortly. The important point, though, is that we do know what to say about any
adjacent pair, and aboutfchoose A, choose Zg.
95
say that we should choose some outcome besides Z. That also seems wrong, though,
since there's intuitively more reason to perform the next outcome in the sequence.
So I take it that it's actually a virtue of the account that it doesn't deliver an
(inevitably unsatisfying) answer as to what we should do in a case of intransitivity.
Contrastivism (plus contrast-sensitive importance) gets all of the pairwise cases
for adjacent alternatives, which we do have strong intuitions about, correct. But
we don't have strong intuitions about pairwise comparisons between non-adjacent
alternatives. For example, what is there most reason to do out offchoose D, choose
Pg? It's hard to say. This is going to depend on whether we think the additional
total utility in P provides a strong enough reason to choose P to outweigh the
reason provided by the higher average utility in D. These are hard questions, but I
won't address them here.
So the view developed in this chapter gives the correct results in cases that we
have strong intuitions about, and tells us exactly how to decide in pairwise cases
we don't have strong intuitions about|we just have to compare the weights of the
reasons provided by the various factors. A disadvantage is that, since the weights
of these reasons can vary widely depending on the comparison being made, there
doesn't seem to be a neat, systematic way to do that. A theory that did not allow
the weight of the reasons to vary|or at least made them vary in a systematic way|
would be neater, but I don't see how to give such a theory while still capturing the
intuitively clear cases. This is an important open issue. If there's no good way to
make progress on it, then we should be wary of accepting this account.
5.5.2 Resolution of pairwise intransitivity?
I've just argued that, if the two-member subsets of a set of alternatives exhibit
intransitivity, as in the Repugnant Conclusion case, we want a theory that does not
say that we ought to do some particular thing out of the larger set. This leaves us
in a precarious practical position|there's just no answer about what we should do,
out of the larger set|but any answer would seem incorrect.
But this raises a problem for the theory I've suggested in this chapter. The
strategy for accommodating cases of intransitivity was to allow the importance of
the reason-providing ends to vary with sets of alternatives, which generates variation
in the weight of the reasons. Suppose the weights of the reasons to A, B, and C
96
vary in such a way that we have a case of pairwise intransitivity: more reason for
A than B out offA, Bg, more reason for B than C out offB, Cg, but not more
reason for A than C out offA, Cg. I've argued that we just can't say what there
is most reason to do out of the larger set,fA, B, Cg. And again, this seems to
be the right thing to say about the Repugnant Conclusion case, since no answer
is satisfying. But|and here's the problem|for all the theory I've developed so
far says, pairwise intransitivities might be resolved when we consider a larger set.
The weights of reasons for the dierent alternatives relative to the pairwise sets
do nothing to constrain the weights of the reasons relative to the larger set. So
it might be that, even though no single alternative wins out in all of its pairwise
comparisons, one does win out relative to the larger set. And I've just argued that
this is exactly what we don't want to happen, at least in many cases.
In a way, it would be comforting if intransitivity was always resolved in larger
sets of alternatives. Out of any pairwise set, we'd know what to do (except in cases
of ties or incommensurability). And we would know what to do out of larger sets, as
well|there'd be a right answer to the question of what there is most reason to do.
Moreover, we'd still be able to admit the possibility of pairwise intransitivities. But
the problem is that this simply doesn't seem to be what happens in paradigm cases
of intransitivity, like the case in the argument for the Repugnant Conclusion. Still,
the framework seems to formally allow for it. And if it does, given the argument in
the last subsection, that's a problem for the theory. In the rest of this subsection,
I'll show how to block this result: we can guarantee that pairwise intransitivity will
not be resolved in larger sets.
First, I need to say more precisely what it is for pairwise intransitivity to be
resolved in a larger set. Roughly, it's just for there to be some thing that we should
do out of the larger set, even though there is pairwise intransitivity between its
members. But this is just rough. We might have pairwise intransitivity between
three members, B, C, and D, of the larger four-member setfA, B, C, Dg. But
suppose that A wins in all of its pairwise comparisons with these three. Then
intuitively there is something we ought to do out of the larger set, namely A, even
though there is pairwise intransitivity. But this isn't really the interesting sort of
resolution of pairwise intransitivity. The kind of case we're interested in, then, is
one in which no option wins in all of the pairwise comparisons, but one does win
out of the larger set.
97
We can guarantee the result we want by adopting CRO, which I argued for in
Chapter 3:
CRO: s ought to A out of Q i s has most reason to A out offA, Bg for all of
the other alternatives B in Q.
I argued that this principle is preferable to a somewhat more natural principle,
called CRO*, that says simply that what you ought to do out of a set is what
you have most reason to do out of that set. This is because CRO lets us explain
more easily why it is that you ought to A out of Q when A is, roughly, best on
balance, even though it's not best with respect to any particular factor. CRO tells
us to look at the reasons relative to the two-member subsets of Q; whenA wins in
all of those pairwise comparisons, it's what you ought to do out of the larger set.
CRO*, on the other hand, tells us to look only at the reasons relative to Q. If
we adopt CRO, and assume that intransitivity is resolved in the larger set only if
there's something you ought to do out of that larger set, then CRO guarantees that
pairwise intransitivity will never be resolved in a larger set, in the sense I described
in the last paragraph. If no option wins in all of its pairwise comparisons, then
there's nothing you ought to do out of the larger set.
One strange result that the view, even with the addition of CRO seems to allow
for, is the following. No alternative wins in all of the pairwise comparisons, and so
according to CRO, there is no alternative that we ought to do out of the larger
set. But the weights of the reasons may still shift in such a way that there is most
reason to do one of them out of the larger set. But since it doesn't win out in all
of its pairwise comparisons, again, it's not the case that we ought to do it.
This highlights the fact that, once we adopt contrast-sensitive importance, we
leave room for the possibility of a case in which what you ought to do out of a set is
not what you have most reason to do out of that set. In Chapter 3, I argued against
the principle, CRO*, which I mentioned above. I rejected this principle in favor of
CRO because there are cases in which you intuitively ought to do something even
though it's hard to nd any reason to do it out of the relevant larger set. But I
claimed that this wasn't so counterintuitive, since in these cases, the thing that you
ought to do, according to CRO, is what's best on balance. But the kind of case
we're considering now isn't like that.
Here's a schematic description of a way to rule out this last result. This sort
of strategy would actually make appealing to CRO unnecessary. To accommodate
98
intransitivity, I appealed to contrast-sensitive importance of reason-providing ends,
and thus to contrast-sensitive weight of reasons. The issue is then that there seems
to be nothing to predict that the weight won't always shift in such a way that
pairwise intransitivity is resolved in a larger set. So we could let the importance
of ends relative to subsets of a larger set in
uence the importance of those ends
relative to the larger set. For example, we might say that when there is signicant
variation in how important an end X is relative to the dierent subsets of a set Q,
there's simply no answer to the question of how importantX is relative toQ. Then
there would be no answer to the question of how weighty the X-provided reasons
are relative to Q. And nally, there would be no answer to the question of what
there is most reason to do relative to Q|we can't decide, since we can't say how
weighty theX-provided reasons are. Obviously this view would require much more

eshing out to properly evaluate it. There might even be ways of developing the
idea so that pairwise intransitivity was sometimes resolved in larger sets, though I
don't see exactly how that would go. I just present it here as an alternative to the
strategy of appealing to CRO, which can avoid the potentially troubling results.
5.5.3 Money pumps
A famous worry about intransitivity concerns cycling, sometimes presented as a
\money pump". The idea is that if we allow intransitivity, there might be situations
in which doing what you have most reason to do would predictably lead you to
a worse outcome. Suppose you have more reason to choose a chocolate ice cream
cone and pay a dime than to choose vanilla, more reason to choose vanilla and pay
a dime than to choose strawberry, but more reason to choose strawberry and pay
a dime than to choose chocolate. If we allow intransitivity of `more reason', this
seems like a possible case. But it looks like you could be pumped for arbitrarily
large amounts of money, simply by doing what you have most reason to do.
Suppose you have a vanilla ice cream cone. Then I might oer you a trade: your
vanilla cone plus a dime for my chocolate cone. By hypothesis you have more reason
to choose a chocolate cone plus paying a dime than to choose a vanilla cone. So
if you do what you have most reason to do|as presumably you ought to do|you
will make the trade. Now suppose I oer you a second trade: your chocolate cone
plus another dime for my strawberry cone. Again, by hypothesis, you have more
99
reason to choose a strawberry cone plus paying a dime than to choose a chocolate
cone. So you'll make the trade. Now I oer you a third trade: my vanilla cone
(which, recall, was originally your vanilla cone) for your strawberry cone plus a
dime. Again, if you do what, by hypothesis, you have most reason to do, you'll
make the trade. So after this cycle, you'll be right back where you started, with a
vanilla cone, only you'll have paid thirty cents to get there. So you're certainly in a
worse position than you were to begin with. And we could obviously make things
even worse by going around the loop again: I could oer you the chocolate cone
for your vanilla cone plus a dime, then the strawberry cone for your chocolate cone
plus a dime, and so on, at least until the ice cream melts. So you could end up
paying an arbitrarily large amount of money to end up with the same vanilla cone
you started with. But all you did was what you had most reason to do in each case.
This looks like a big problem. We might have hoped that doing what you
have most reason to do could not|at least not predictably|land you in a worse
position overall. And all we needed, it seems, to get this going is the assumption
that intransitivity of `more reason' is possible. Thus, according to this objection,
we should conclude that intransitivity of `more reason' is impossible, despite the
powerful arguments for its possibility.
I think this would be hasty. There are two popular ways to respond to this
type of concern, which correspond to two dierent ways to conceive of the money
pump case. First, suppose you're minding your own business, getting ready to
enjoy a vanilla ice cream cone, when I approach you with a chocolate cone. You
don't suspect any bad intentions on my part. Then it's hard to deny that you
should make the trade. Now suppose a dierent person comes up to you with a
strawberry cone, and again you don't suspect any bad intentions. Again, it's hard
to deny that you should make the trade. At this point you're down twenty cents
and have a strawberry cone. But there's nothing irrational about ending up in this
situation. Intuitively, it's only when you get back to where you started, but have
paid thirty cents to get there, that you're in an obviously worse situation than you
were to begin with. So if another person comes up to you with a vanilla cone,
according to this line of response, you should refuse to make the trade, even though
by hypothesis you have more reason to make the trade, at least synchronically. You
just recognize that the facts about what reasons you have (plausibly due to your
preferences) make you vulnerable to being money pumped, if you aren't careful.
100
Once you recognize this, you will not do what you recognize you have most reason
to do, in the particular situation. That is, at least in cases of intransitivity, it's
not true that you should do what you have most reason to do. Rachels (1998,
2001) seems to think of the money pump case in this way, and oers this sort
of response. He says, \The money-pump objection assumes that a rational agent
would always prefer what is better and act on those preferences, no matter what.
But that assumption would be rejected along with Transitivity".
21
It's straightforward to put this kind of response in a contrastivist framework.
The relevant sets are just the obvious ones: rst,fkeep the vanilla cone, trade for
the chocolate coneg; second,fkeep the chocolate cone, trade for the strawberry
coneg, etc. But we just deny the initially attractive idea that what you ought to
do is what you have most reason to do. This is an unattractive consequence; but
perhaps Rachels is right that once we deny that transitivity holds, some intuitive
principles of rationality simply have to go, as well.
Here's the second way to think of the money pump objection, and the second
way to respond to it. You're getting ready to enjoy your vanilla cone, when you
see me approach with a chocolate cone and a strawberry cone, looking mischievous.
If you think clearly about the situation, it seems that you are plausibly rationally
required not to be pushed around the cycle|not to make the trades, at least not
more than the rst two. The idea is that rationality requires more than simply
looking at the synchronic choices you have to make. If we think that intransitivity
is possible, we should adopt a more global perspective. Darwall (1983) discusses
this kind of response to intransitivity. Sometimes we have to make what looks
synchronically like the choice we have less reason to make to avoid ending up in
a worse situation overall|\take one step back to take two forward", or perhaps
better, \stay put to avoid taking a big step back". If you see what I have in mind
with my ice cream cones, you should rationally not make the trades.
One way to make this response more precise, in my contrastivist framework,
is to say that if you see what I have in mind, the relevant set of alternatives for
the rst trade isn't actuallyfchoose the chocolate cone and pay a dime, choose
the vanilla coneg, but rather something likefchoose the chocolate cone and pay
a dime and then stay put, keep the vanilla cone, choose the chocolate cone and
pay a dime and then accept the trade for the strawberry cone minus another dime
21
Rachels (2001), p. 218.
101
and then stay put, choose the chocolate cone and pay a dime and then accept the
trade for the strawberry cone minus another dime and then accept the trade for the
vanilla cone minus another dimeg. Spelling out exactly what the set is might get
a bit complicated, but the idea is simply to include descriptions of what you'll do
later|to look at more ne-grained alternatives. When we do this, staying put, or
at least only trading once or twice, will very plausibly be what you actually have
most reason to do, out of these ne-grained sets of alternatives. When you have a
vanilla cone, an alternative that involves paying thirty cents to end up with that
same vanilla cone will certainly be worse than the others. This is a very natural
thing for a contrastivist to do, since (i) she thinks that reasons are always relativized
to sets of alternatives, and (ii) one important way in which sets of alternatives can
vary is by being more or less ne-grained. So this strategy appeals to what I've
been calling, following Yalcin (2011) and Cariani (fc), the resolution-sensitivity of
reasons.
I suspect that this strategy can be extended to stop the money pump, even
when it's conceived of in the rst way. The basic idea is that after you make the
rst two trades, you should recognize that you are vulnerable to the money pump,
if you aren't careful. At that point, you should be considering a more ne-grained
set of alternatives like the one I described in the last paragraph. And again, once
you consider this ne-grained set, alternatives in which you get pushed around the
loop will very plausibly be worse. So if you don't anticipate getting money pumped
to begin with, you might end up losing some money. But you won't be forced
around the loop twice. I think this response is more attractive than the idea that
you should simply refuse to make the trades, and knowingly do what you have less
reason to do, because it lets us retain the idea that you ought to do what you have
most reason to do.
So there are ways to respond to the money pump objection to intransitivity.
Moreover, an attractive way to respond to the problem can be made precise by
appealing to the important contrastivist idea of resolution-sensitivity. However,
what I've said here likely isn't completely satisfying. Given the importance of this
objection, then, this is another place where more work would need to be done before
we could be comfortable accepting an account like the one developed here.
102
5.5.4 Why so uncommon?
The last issue I want to consider is the following. Even if we think intransitivity is
theoretically possible, and so want a theory that can accommodate it, it's plausibly
a desideratum on any theory that allows for intransitivity to give some explanation
for why intransitivity is so rare. After all, the cases that are most often appealed
to as alleged examples of intransitivity are Part's cases I described above. And
these are extraordinary cases, to say the least. In any ordinary case, transitivity
seems to hold. But if there's no formal obstacle to intransitivity, it looks like every
single situation any real person will ever face is just a special case. On the theory
I've suggested here, in particular, cases in which transitivity holds look like very
special cases. The importance of the reason-providing ends across dierent sets of
alternatives just happen to work out conveniently. We need some explanation of
this. If we can't oer such an explanation, then we should be very skeptical that
the theory I've suggested here is correct.
22
To satisfy this desideratum, I need to give an explanation of why intransitivity
will be rare, even though there's no formal obstacle. I don't see how to adjust the
structural features of the account to give such an explanation. So the strategy has to
be to appeal to substantive facts about particular cases. As long as the substantive
facts that generate intransitivity are rare enough, we can get the explanation we
want.
The crucial feature of the account is what I called contrast-sensitive importance
of ends. That's the feature that lets the weight of reasons vary with sets of alterna-
tives. So if the importance of ends only varies|or only varies signicantly|when
extraordinary sets of alternatives are involved, we could explain why intransitivity
is so rare. If relative to all the sets of alternatives that are ordinarily relevant,
the importance of a given end is relatively constant, we won't get intransitivity in
ordinary cases. But we can still get intransitivity when we compare the strength of
reasons between ordinary and extraordinary cases, or among extraordinary cases.
Consider the sequence of outcomes from the argument for the Repugnant Con-
clusion. Now, any given set of alternatives consisting of choosing members of this
22
It's worth reiterating here that the ocial view of this dissertation is the one developed in
Chapter 4. So if you nd intransitivity too problematic, you can simply reject the amendment
I proposed in this chapter, namely tacking on contrast-sensitive importance of reason-providing
ends.
103
sequence aren't really ordinary. But what is ordinary about sets likefchoose A,
choose Bg orfchoose Y, choose Zg is that we're choosing between outcomes in
which the average levels of utility are relatively close. Usually when we choose be-
tween some alternatives, our choice won't make a drastic change in the average level
of utility, at least not while leading to an increase in the total amount of utility.
Of course, some people may face choices between one outcome and another which
would lead to a much lower average level of utility for some part of the population.
But the second choice will generally lead to a much lower total level of utility, as
well.
The importance of the ends of total utility and average utility, I suggest, remains
relatively stable between sets likefchoose A, choose Bg andfchoose A, choose Cg.
It's only when we consider a set likefchoose A, choose Zg that the importance
of the ends can shift signicantly. So the intransitivity of `more reason' is only
going to arise when some of the sets we're considering are extraordinary. That
explains why intransitivity is so rare: we don't generally consider extraordinary
sets of alternatives.
Some explanation along these lines has to work, if the theory I developed in
this chapter is going to be acceptable. I think the prospects of developing this
explanation into something plausible are promising. But it would obviously take
more work than what I've done here. All I want to suggest is that the package
I've developed in this chapter is one very promising way to begin accommodating
intransitivity.
5.6 Conclusion
In Chapter 4, I developed a contrastivist view of reasons, and showed that it rules
out a certain kind of troubling intransitivity. In this chapter, I considered an
objection from the other direction: intransitivity actually is possible, and moreover,
an initial attraction of contrastivism|ignoring the framework from Chapter 4|is
that it seems especially well-placed to accommodate it. I argued that the sort of
intransitivity that this objector has in mind actually is compatible with, though
not required by, the framework from Chapter 4, and showed how to accommodate
it. I want to remain neutral on whether intransitivity is possible; all I meant to do
in this chapter was demonstrate that contrastivism can accommodate it.
104
Another more general goal of this chapter was to show that the view I developed
in the rst four chapters of the dissertation provides an interesting way to think
about important debates in ethics and practical reasoning. We can reframe these
debates in contrastivist terms and make progress by exploiting the extra resources
that contrastivism gives us. In this chapter, I hope to have shown that appealing to
natural contrastivist ideas|in particular, contrast-sensitive importance of reason-
providing ends|lets us make some progress in the debate over the possibility of
intransitivity. In the next chapter, I'll apply the theory to an interesting issue
from epistemology, and an analogous issue in practical reasoning: the rationality of
withholding belief, on the one hand, and intention, on the other.
105
Chapter 6
Withholding
My goal for both the previous chapter and this one is to illustrate some interest-
ing applications of contrastivism about reasons. In the last chapter I showed how
contrastivism lets us make progress in the debate over the possibility of the in-
transitivity of `more reason than'. In this chapter, I turn to an application from
theoretical and practical rationality. An important, but often neglected, issue from
these domains is the rationality of withholding an attitude|belief, in the theoretical
case, and intention, in the practical case. Sometimes the most rational thing to do
is to neither believe thatp nor believe that:p, but to not make up your mind with
respect top, or withhold belief. Similarly, sometimes it's more rational to withhold
intention with respect to , rather than intend to or intend not to .
In this chapter I will develop an account of when an agent ought to withhold,
which applies to both belief and intention, that follows very naturally from the con-
trastive account of reasons I have developed in this dissertation. This account has
several attractive explanatory features, and avoids problems that face an initially
attractive non-contrastive account of when agents ought to withhold.
6.1 Withholding Belief and Contrastive Reasons
I will develop the account for the epistemic case of withholding belief rst, and then
extend it to the practical case of withholding intention.
Since we are trying to provide an account of what doxastic attitude you ought to
have, and what you ought to do depends on your reasons, we need to gure out what
your reasons are in this kind of case. Given the contrastive theory of reasons I've
developed in this dissertation, we need to look at your reasons relative to some set
106
of alternatives or other. If you are trying to decide what doxastic attitude to take
towards p, the most natural set of alternatives isfbelieve p, believe:p, withhold
belief with respect to pg.
1
That is, the relevant alternatives will be believing that
p is true, believing that p is false (or disbelieving p), or not making up your mind
one way or the other.
Recall the principle relating what an agent ought to do with her contrastive
reasons that I have adopted:
CRO: s ought to A out of Q i s has most reason to A out offA, Bg, for all of
the other alternatives B in Q.
WhenA-ing wins in all the pairwise comparisons, you ought toA. Now we can just
straightforwardly apply this to belief:
O(Bel)*: s ought to believe p out offbelieve p, believe:p, withholdg i s has
most reason to believep out offbelievep, believe:pg ands has most reason
to believe p out offbelieve p, withholdg
This principle is very plausible: you ought to believe p just in case you have better
reason to do so than to either disbelieve p or withhold belief.
2
More importantly
for the purposes of this chapter, CRO gives us the following account of when an
agent ought to withhold belief:
O(Wh-B)*: s ought to withhold belief inp out offbelievep, believe:p, withholdg
i s has most reason to withhold belief in p out offbelieve p, withholdg and
s has most reason to withhold out offbelieve:p, withholdg.
This principle, like O(Bel)*, is just the one that follows straightforwardly from the
theory I've developed in this dissertation.
6.1.1 Contrastive epistemic reasons
So in providing an account of when an agent ought to withhold belief, we need to
look at four kinds of reasons: (i) reasons to withhold rather than believe p, (ii)
1
I will ignore more ne-grained attitudes, like credences, here.
2
Here is Chisholm (1976), p. 27: \h is certain forS att =
d
f (i) Acceptingh is more reasonable
forS att than withholdingh (i.e., not acceptingh and not accepting:h)". And here is Conee and
Feldman (2004), p. 3: \a person is justied in believing a proposition when the person's evidence
better supports believing that proposition than it supports disbelieving it or suspending judgment
about it".
107
reasons to believe p rather than withhold, (iii) reasons to withhold rather than
believe:p, and (iv) reasons to believe:p rather than withhold. Also important,
if we want to gure out what doxastic attitude an agent ought to take towards p
will be (v) reasons to believe p rather than believe:p and (vi) reasons to believe
:p rather than believe p. What we need next, then, is some account of what these
reasons are, so that we know what to compare.
Before providing that account, though, I need to make one important clari-
cation. I said above that I am giving a view of what doxastic attitude an agent
ought to take toward a proposition, and more specically, when an agent ought to
withhold belief in a proposition. The `ought' here needs to be read in a particular
way, as the `ought' of epistemic rationality. So, I will provide an account of when
you ought, epistemically, to withhold belief in a proposition. This epistemic `ought'
is the one in play in intuitive claims like, `You ought to believe what your evidence
supports', or `You ought not believe contradictions'. Similarly, the reasons I will be
talking about are epistemic reasons|reasons that bear on what you epistemically
ought to do.
3
The most useful way to characterize epistemic reasons for my purposes here is by
distinguishing them from non-epistemic reasons to believe a proposition or withhold
belief in a proposition. If a powerful demon threatens to kill you if you believe that
there is a glass on the table, that is a (very strong!) reason not to believe that there
is a glass on the table.
4
But this reason is importantly dierent than the reasons
provided by evidence that there is a glass on the table, like the fact that you seem
to see a glass on the table. The fact that you seem to see a glass on the table is a
reason to believe that there is a glass on the table.
5
This is a paradigm example
of what I mean by an epistemic reason|assuming that you don't have any other
evidence that there is not a glass on the table, or any reasons to disregard your
visual evidence, the epistemically rational thing for you to do is believe that there
3
Two clarications: (i) I do not mean to claim that `ought' is ambiguous, in any deep sense,
between this use and a practical use; for all I have said, the dierence can be traced to contextual
elements, like Kratzer (1981)'s ordering source and modal base. (ii) There is another use of
`epistemic `ought" that diers from my use in this paragraph; this is sometimes also called the
predictive use, and it is in play in the most natural interpretation of `The roast ought to be done
by now'.
4
For discussion of this \wrong kind of reasons" problem, see D'Arms and Jacobson (2000a,b);
Rabinowicz and Rnnow-Rasmussen (2004); Schroeder (2012b). Pascal's Wager provides another
classic example of a practical reason for belief.
5
I am glossing over issues about perceptual justication and evidence here.
108
is a glass on the table. But this may not be the practically rational thing to do in
this case. In this practical sense, plausibly, you ought not believe that there is a
glass on the table, since the demon's threat gives you very strong practical|but
not epistemic|reason not to believe this. The demon's reason does not bear on
what it is epistemically rational for you to do, though it does bear on what it is
practically rational for you to do. There is a huge literature on these issues. But
in my discussion of belief and withholding belief in this chapter, I will ignore them,
and stipulate that I am only talking about epistemic reasons.
One nal point before I move on to give a contrastivist account of epistemic
reasons. This last example may tempt you to think that epistemic reasons are all
evidence for or against the proposition, while non-epistemic reasons involve good or
bad features of being in the doxastic state (belief, disbelief, or withholding) itself. In
fact lots of people have been tempted to exactly this kind of view.
6
But as Schroeder
(2012b) argues, there are epistemic reasons besides evidence. For example, the fact
that you have very little information and expect to get more soon is intuitively an
epistemic reason to withhold belief. It bears on what doxastic attitude it is rational
for you to have|in this case, withholding belief. This is, not surprisingly, a very
important point for this chapter, since I will be talking mostly about reasons to
withhold belief, and paradigm cases of reasons to withhold belief are not evidence.
Now I am ready to give a contrastivist view of the six types of epistemic reasons
listed above. Begin with reasons to believe p rather than believe:p. This is
probably the easiest case: a reason to believe p rather than believe:p is simply
evidence for p. Similarly, of course, a reason to believe:p rather than believe
p is evidence for:p (or evidence against p). Evidence for p is also a reason to
believe p rather than withhold|the fact that Tom's ngerprints are on the book
is a reason to believe that Tom stole the book, rather than withhold belief in this
proposition. Of course, this reason might not be weighty enough to justify belief in
this proposition, but that is compatible with saying that it is a reason to believe it.
Similarly, evidence for:p is a reason to believe:p rather than withhold. Let `Ev
p
'
stand for the set of evidence for p and `Ev
:p
' stand for the set of evidence for:p.
This leaves reasons to withhold rather than believe p and reasons to withhold
rather than believe:p. Begin with considerations like the fact that you will receive
6
For views along these lines (ignoring important dierences between them), see Part (2001);
Piller (2001, 2006); Olson (2004); Stratton-Lake (2005); Hieronymi (2005).
109
more evidence regarding p tomorrow. This is both a reason to withhold rather
than believe p and a reason to withhold rather than believe:p. Similarly, the fact
thatp is completely irrelevant is plausibly a reason to withhold rather than believe
that p and a reason to withhold rather than believe that:p. After all, making
up your mind about p may take some cognitive eort and distract you from more
pressing matters, and given the unimportance of p, it would be a waste to do so.
7
These types of reasons are both reasons to withhold rather than believe that p and
reasons to withhold rather than believe that:p. Call the set of such reasons for a
proposition p `W
p
'.
Sometimes, though, the reasons to withhold rather than believe that p and the
reasons to withhold rather than believe that:p can come apart. An advantage of
contrastivism is that it gives us a very straightforward way to capture this. Stanley
(2005) presents cases in which believing p is very risky, while believing:p is not,
or vice versa.
8
In a case like this, the potential costs of believing p provide reasons
to withhold rather than believe p, but not reasons to withhold rather than believe
:p. Similarly, the potential costs of falsely believing:p provide reasons to withhold
rather than believe:p, but not reasons to withhold rather than believep.
9
Let `C
p
'
stand for the set of reasons provided by potential costs of falsely believing p and
`C
:p
' stand for the reasons provided by potential costs of falsely believing:p.
Finally, I claim that evidence can also be reason to withhold belief. Initially,
this claim seems completely wrong.
10
I said above that the fact that Tom's n-
gerprints are on the book is a reason to believe that Tom stole the book rather
than withhold|it is a reason not to withhold, and to believe that Tom stole the
book, instead. But remember that I am providing a contrastive account of epis-
temic reasons. So evidence thatp, I claim, is reason to withhold rather than believe
that:p. Similarly, evidence that:p is reason to withhold rather than believe that
p. Though the fact that Tom's ngerprints are on the book is a reason to believe
that Tom stole the book rather than withhold, it is also a reason to withhold belief
rather than believe that Tom did not steal the book. As we will see below, this
7
I am not claiming that making up your mind about propositions is always, or even usually,
cognitively taxing. But sometimes, it is, even for unimportant propositions.
8
The riskiness depends on the belief thatp playing its characteristic role in guiding action|to
suer the potential consequences of believing that p, you have to actually act as if p is true.
9
See Schroeder (2012a) for more discussion of these kinds of cases, which focuses on costs of
believing and withholding.
10
See Schroeder (2012a) for an argument that evidence cannot be reason to withhold.
110
is the feature of this contrastive account that gives it explanatory advantages over
the rival non-contrastive picture.
So to sum up, I have oered the following account of epistemic reasons. Let `'
stand for the combining reasons operation (which is not simply additive):
reasons to believe thatp rather than believe that:p = reasons to believe that
p rather than withhold = Ev
p
reasons to believe that:p rather than believe thatp = reasons to believe that
:p rather than withhold = Ev
:p
reasons to withhold rather than believe that p = W
p
C
p
Ev
:p
reasons to withhold rather than believe that:p = W
p
C
:p
Ev
p
6.1.2 A contrastive account of rational withholding
Now we can plug this account of epistemic reasons into the principles about when
an agent ought to believe a proposition, and when an agent ought to withhold belief
in a proposition. Let `A>B' mean that the set of reasons A is weightier than the
set of reasons B.
O(Bel): s ought to believe p i (i) Ev
p
>Ev
:p
, and (ii) Ev
p
>W
p
C
p
Ev
:p
O(Wh-B): s ought to withhold belief with respect top i (i) W
p
C
p
Ev
:p
>Ev
p
,
and (ii) W
p
C
:p
Ev
p
>Ev
:p
Note that condition (i) in O(Bel) is actually redundant, assuming W
p
and C
p
cannot have negative weight. That is, whenever an agent has more reason to believe
that p than to withhold belief (out of these two options), she will also have more
reason to believe that p than to believe that:p (out of these two options). This
is a direct result, of course, of the fact that this account treats evidence that:p
as reason to withhold rather than believe that p, along with the assumption that
there are not epistemic reasons not to withhold belief.
11
As we will see, though, the
analogous condition governing when agents ought to intend will not be redundant.
In the next subsection, I will discuss a particularly nice feature of this account
that an otherwise promising existing account, developed by Schroeder (2012a), does
11
This assumption is not uncontroversial, but I will not defend it here.
111
not share. As we will see, contrastivism has this feature because of the claim that
evidence for a proposition p is not only reason to believe p rather than either
believe:p or withhold, but also reason to withhold rather than believe:p. As I
mentioned above, and will further argue below, this move seems proprietary to the
contrastivist.
6.1.3 Ties
When your evidence forp and your evidence for:p are balanced, or tied, you ought
to withhold belief. This is a very plausible requirement on any theory of rational
withholding. If we accept the appealing idea that what you ought to do is what
you have most reason to do, this means that, whenever your evidence is tied, you
must have more reason to withhold belief than to believe p or believe:p. So when
your evidence provides very little reason to believe p and equally little reason to
believe:p, you have to have more reason to withhold. And|crucially|when your
evidence provides very strong reason to believe p, and equally strong reason to
believe:p, you still have to have more reason to withhold.
What we need, then, is some explanation for why your reasons to withhold
belief will always be weightier than your reasons to believe p and your reasons to
believe:p, no matter how weighty they are, when your evidence is tied. If we only
recognize as reasons to withhold (i) costs of falsely believing and (ii) considerations
like the fact that you will get more evidence regardingp tomorrow, it is hard to see
why this should be so. The weights of these kinds of reasons seem to have nothing
to do with whether or not your evidence is tied. So we lack an explanation for why
these reasons to withhold will always outweigh both the reasons to believe p and
the reasons to believe:p, in cases of ties.
An account of rational withholding developed by Schroeder (2012a) faces exactly
this problem. According to his account, you ought to withhold when the costs of
falsely believing p outweigh the evidence for p and the costs of falsely believing
:p outweigh the evidence for:p. We can add to his picture reasons like the fact
that you will get more evidence tomorrow. But as I said above, the weights of
these reasons seem to have nothing whatsoever to do with the weight of the reasons
provided by the evidence for and againstp. So we have no reason to expect them to
112
be weighty enough to make it the case that you ought to withhold belief whenever
your evidence for and against p is tied.
One response here is to claim that in cases of ties, there is a special kind of
reason to withhold belief:
Meta-Reason: The fact that your evidence is tied is a reason to withhold belief.
Thus, in cases of ties, you get a new reason to withhold belief. But to predict that
you always ought to withhold belief in the case of ties, we need to add the following
stipulation:
Stipulation: This Meta-Reason is always suciently weighty that your reasons to
withhold belief outweigh both your reasons to believe thatp and your reasons
to believe that:p.
With these two assumptions, a non-contrastive account like Schroeder's can predict
that you ought to withhold belief whenever your evidence is tied.
But I think this strategy is problematic for two reasons. First, it is simply not
very explanatory. While it is plausible that the fact that your evidence is tied is a
reason to withhold, I do not think it is very explanatory to simply stipulate that
it is suciently weighty to do the job. This comes very close to simply stipulating
that in cases of ties, you ought to withhold belief. But it would be better if we
could explain this fact, instead.
Second, these assumptions do not go far enough. In most cases, if your evidence
is almost tied, you still ought to withhold belief. But if your evidence is merely
almost tied, we cannot appeal to Meta-Reason, since you only have this reason
when your evidence is in fact tied. We could try modifying Meta-Reason so that
you had this sort of reason even when your evidence is merely almost tied. But this
is starting to look very unexplanatory. Further, we need some account of how close
your evidence must be, in order for this sort of meta-reason to apply. Complicating
things even further, it will likely be a variable matter how close your evidence must
be, for withholding to be what you ought to do.
The contrastivist account I've given here, on the other hand, explains this ob-
servation nicely. To see this, assume that your evidence for p and your evidence
for:p are tied; so set Ev
p
= Ev
:p
in O(Wh-B) above. It is easy to see, then,
that in any case like this, you can never have more reason to believe p than to
113
withhold or more reason to believe:p than withhold. So as long as either (i) there
is always some cost of falsely believing, e.g. C
p
and C
:p
are not completely empty,
or (ii) some reason to withhold rather than either believe p or believe:p, e.g. W
p
is not empty, it will turn out that in every case in which your evidence is tied, you
ought to withhold out offbelieve p, believe:p, withholdg. It is overwhelmingly
plausible that at least one of (i) and (ii) is true. I acknowledge, though, that for
my account to make the prediction that you ought to withhold in cases of ties, I do
need one of these to be true.
12
If there are cases of ties in which neither (i) nor (ii)
is true, my account can at least explain why it is never the case that you ought to
believep and not the case that you ought to believe:p when your evidence is tied.
Non-contrastive accounts like Schroeder's do not explain this.
Moreover, the contrastive account can explain why you ought to withhold belief,
even in many cases in which your evidence is not quite tied. As long as there are
either (i) potential costs of falsely believing or (ii) reasons to withhold rather than
either believep or believe:p that are suciently weighty to outweigh the dierence
in the weight of the evidence forp and for:p, you ought to withhold. So this account
tells us how close to balanced your evidence must be, in order for withholding to
be what you ought to do. The dierence must be small enough for the potential
costs of falsely believing and your reasons to withhold in W
p
to outweigh it. This
explains why we get the sort of variation I mentioned above: in cases in which
you have very weighty reasons to withhold, in W
p
, or in which the stakes are very
high (as in Stanley (2005)'s bank cases), you ought to withhold belief even if your
evidence is quite unbalanced. In lower stakes cases in which you don't have very
weighty reasons to withhold in W
p
, your evidence must be much closer to balanced,
for withholding to be what you ought to do.
The key to this contrastivist account is that it allows us to treat evidence as
reason to withhold. This is what allows us to explain why you should always
withhold belief in cases of ties, and why you should often withhold in the case
of near-ties. This move seems proprietary to the contrastivist. A non-contrastivist
12
One way we could guarantee this is to accept a contrastivist version of Meta-Reason above:
the fact that your evidence is tied is always a reason to withhold rather than believe p and a
reason to withhold rather than believe:p. Then W
p
will not be completely empty in cases of
ties. Notice, that this strategy does not require accepting the unexplanatory Stipulation that
this meta-reason is always suciently weighty. Rather, it would explain why it was suciently
weighty: in cases of ties, all you need, according to O(Wh-B) is some reason in W
p
, so the
meta-reason will be weighty enough.
114
could not let evidence thatp and evidence that:p provide (non-contrastive) reasons
to withhold, since in that case, there would always be at least as much reason to
withhold belief as to believe p or believe:p, and ordinarily there would be most
reason to withhold. Thus, if the non-contrastivist were to attempt to co-opt this
strategy of treating evidence as reason to withhold, she would commit herself to
the view that in almost every case, what you ought to do is withhold belief. This
would be a strange route to skepticism.
6.2 Withholding Intention
Just as agents often ought to withhold belief in a given proposition, they often
ought to withhold intention, with respect to a given action. For example, suppose
I've been invited to a party next weekend. I'd like to go, but there's a chance an
old friend will be in town the evening of the party, and if she is, then I should see
her. I will know in the next couple of days if she will in fact make the trip. In this
case, the thing for me to do is withhold intention about whether to go to the party.
I shouldn't intend to go, since my friend may be in town, and I shouldn't intend not
to go to the party, since she may not be. Rather, I should withhold intention until I
get more information. Now that I have provided the contrastivist account of when
agents ought to withhold belief, I want to generalize to an account of when agents
ought to withhold attitudes that applies not only to belief but also to intention.
6.2.1 A unied account?
My goal is to give a unied account of rational withholding that applies to both the
practical and theoretical cases. Besides the theoretical virtues, there are also strik-
ing structural similarities between between withholding intention and withholding
belief that make it reasonable to expect such a unied account. In both cases,
there is what we might call the `positive' attitude|believing p or intending to|
the `negative' attitude|believing:p or intending not to |and the `undecided'
attitude|withholding belief or withholding intention.
13
13
By calling believing:p and intending not to `negative', I don't mean to suggest that there
are any deep dierences between these and what I called the `positive' attitudes. This is especially
obvious in the case of belief: believing that:p is simply bearing the same attitude|believing|to
115
But Harman (2004) makes an observation that seems to raise a problem for this
idea. He points out that there is, at least in many cases, an important dierence
between the theoretical and practical cases with respect to what you ought to do
when your reasons for the positive and negative attitude are tied. In the practical
case, it is generally permissible to simply decide. In fact, it's often impermissible
to not decide, even if your reasons to intend to and reasons to intend not to
are tied. Think of the case of Buridan's Ass, who starved to death trying to
decide whether to take the bale of hay on the right or identical bale on the left.
Since the reasons to intend to take the bale on the right were perfectly matched by
reasons to intend to take the bale on the left, it withheld intention about whether
to take the bale of hay on the right (as well as about whether to take the bale
of hay on the left), and ended up starving to death. But things are dierent in
the theoretical case. When your reasons to believe p and your reasons to believe
:p are tied, it is in general not rationally permissible to just pick one to believe.
In this case, you ought to withhold belief. Since cases of ties are cases in which
you ought to withhold belief but not, or not always, cases in which you ought to
withhold intention, it looks like we can't give a unied account of when you ought
to withhold that applies to both.
Harman's observation is correct|it is often (practically) permissible to simply
form an intention to or an intention not to when your reasons for each are tied,
and it is never (epistemically) permissible to simply form a belief inp or a belief in
:p when your reasons for each are tied. But it is a mistake to conclude that this
must be due to any structural dierences between the correct accounts of when you
ought to withhold intention, on the one hand, and belief, on the other. Rather,
we can explain the asymmetry Harman observes by noticing that though there are
frequently weighty reasons not to withhold intention (e.g. you'll starve to death,
if you do) there are never, or at least very rarely, reasons not to withhold belief
(besides, of course, evidence). Thus, we should actually expect the asymmetry
Harman observes, even if we have a unied account of rational withholding. It
is just that in the practical case, it will often be much harder for the reasons to
withhold intention to outweigh the reasons not to withhold intention, but this will
not be true for withholding belief.
14
a dierent proposition. Perhaps there's some more important dierence between intending to ,
where is incompatible with , and intending not to . But I'll ignore this issue here.
14
Schroeder (2012a) brings out essentially the same point, but in a dierent way.
116
6.2.2 Contrastive practical reasons
Now I will provide an account of what I'll call, for the purposes of this chap-
ter, practical reasons. By this I simply mean reasons that bear on what practical
attitude|intending to , intending not to , or withholding intention|it is ratio-
nal for you to take toward a given action, . This is to use the term `practical
reasons' in a dierent way than it is often used. This term is often used to refer to
reasons for action rather than reasons for intention. But again, since I am talking
about intentions, I am just talking about reasons for intention.
Just as I restricted my discussion in the epistemic case to epistemic reasons,
ignoring reasons like the fact that a demon has threatened to kill you if you believe
that p, I will here restrict my discussion to practical reasons that bear on whether
it is rational to intend to , intend not to , or withhold intention. So I will set
aside reasons like those present in Kavka (1983)'s Toxin Puzzle. In this case, you
are oered a lot of money to intend to drink a mild toxin, with the catch that you
know you will receive the money before the time comes to drink the toxin. In this
case, you have a reason to intend to drink the toxin that is importantly dierent
than a reason like the fact that someone else will give you a lot of money if you
actually drink the toxin, or the fact that the toxin is tasty. I do not have a good
way to distinguish between these \right kind" and \wrong kind" of reasons in the
practical case, just as I had no good way to distinguish between the right kind
and wrong kind of reasons in the epistemic case. But rather than entering into
this important but dicult debate, I will rely on an intuitive understanding of the
dierence, and only talk about reasons that, I hope, fall clearly on the \right kind"
side.
I am concerned here with the question of what attitude an agent ought to take
toward an action, . So the relevant alternatives will be intending to , intending
not to , or withholding intention. As before, we can simply appeal to CRO.
Applying CRO to the case of intention, we get the following principles:
O(Int)*: s ought to intend to i (i) s has most reason to intend to out of
fintend to , intend not to g and (ii) s has most reason to intend to out
offintend to , withhold intentiong
117
O(Wh-I)*: s ought to withhold intention with respect to i (i)s has most reason
to withhold intention out offintend to , withhold intentiong and (ii) s has
most reason to withhold intention out offintend not to, withhold intentiong
So what we need is some account of what these reasons are: (i) reasons to intend
to rather than intend not to , (ii) reasons to intend to rather than withhold,
(iii) reasons to intend not to rather than intend to , (iv) reasons to intend not
to rather than withhold, (v) reasons to withhold rather than intend to , and
(vi) reasons to withhold rather than intend not to .
Perhaps the easiest cases are reasons to intend to rather than intend not to,
and reasons to intend not to rather than intend to . These are simply reasons
for and against -ing, respectively. Similarly, reasons for and against -ing will
be among the reasons to intend to rather than withhold, and to intend not to
, rather than withhold, respectively. Call these sets of reasons `R

' and `R
ag
'.
As a contrastivist, though, I need to say what sets these reasons for and against
-ing are relativized to. I conceive of things as follows. Deliberating agents face
two subtly dierent kinds of questions. First, there is the kind of question I have
been dealing with in most of the dissertation, the question of what to do. This
question will ordinarily provide a set of alternative actions, likef, ,g. Second,
the agent will face the kind of question I have been considering in this chapter, the
question of whether to . This question will provide a set of alternatives like the
one I am concerned with here,fintend to , intend not to , withhold intentiong.
So the reasons for and against-ing will be relativized to the contextually relevant
set, likef, ,g. The reasons to relative to this set will then also be reasons to
intend to rather than intend not to , and to intend to rather than withhold
intention. The reasons not to , R
ag
, will also be reasons to intend not to
rather than intend to , and to intend not to rather than withhold intention.
But in addition, we need to account for some other reasons not to withhold|
or, on this framework, reasons to intend to rather than withhold and to intend
not to rather than withhold. These will include considerations like the fact that
you will starve to death if you withhold intention. For a more realistic example of
reasons not to withhold intention, notice that agents often simply have to decide
because the time to decide has come. If you are trying to decide whether to go see
the guest speaker today, at some point|say, shortly before the talk is scheduled to
start|you simply have to make up your mind. Another related source of reasons
118
not to withhold intention come from the fact that intentions aid in coordination.
15
Forming intentions can aid in your own planning and in making joint plans with
others. You miss out on these benets if you withhold intention. Thus, these
considerations about coordination can provide reasons to intend to rather than
withhold, and reasons to intend not to rather than withhold. Notice how common
these kinds of reasons are. This reinforces the point above that there are frequently
reasons not to withhold intention, and this lets us explain Harman's observation.
Call the set of such reasons `:W

'.
That just leaves reasons to withhold rather than intend to and reasons to
withhold rather than intend not to . In the epistemic case, we had reasons in
the set W
p
, like the fact that you'll get more information regarding p tomorrow,
or the fact that you simply don't have very much evidence one way or the other.
These are reasons to withhold belief rather than believe p and to withhold belief
rather than believe:p. Analogously, in the case of intention, there are reasons to
withhold intention rather than intend to and to withhold intention rather than
intend not to . For example, the fact that you'll get more relevant information
tomorrow is such a reason. For another example, the fact that there's no pressing
need to make up your mind yet (e.g., there's no one to coordinate with, not much
advance planning required, and the time to , if you do it at all, isn't for several
months) is also a reason to withhold intention rather than either intend to or
intend not to . Call the set of reasons like this `W

'.
Just as the reasons to withhold belief rather than believe p could come apart
from the reasons to withhold belief rather than believe:p, the reasons to withhold
intention rather than intend to may come apart from the reasons to withhold
intention rather than intend not to. First, there may be particular costs associated
with intending to that are not associated with intending not to , or vice versa.
Suppose that intending to cuts o some other options that are desirable in certain
ways, while intending not to does not. This is plausible, given the coordinating
role of intention, especially in the interpersonal case. Forming intentions to do or
not to do certain things will lead others to form intentions that may then disqualify
certain options that would otherwise be open to you. These kinds of costs can
provide reasons to withhold rather than intend to that are not reasons to withhold
15
As emphasized by Bratman (1987).
119
rather than intend not to (or vice versa). Call the set of such reasons `C

' and
`C
not
', respectively.
16
Finally, just as I held that evidence for p provides reasons to withhold rather
than believe that:p, I hold that reasons to , R

, are also reasons to withhold
intention rather than intend not to , and that reasons not to , R
ag
, are also
reasons to withhold rather than intend to .
6.2.3 A contrastive account of rational withholding of intention
Now we can plug these reasons into the principles O(Int)* and O(Wh-I)* to
get the account of when agents ought to intend and when they ought to withhold
intention:
O(Int): s ought to intend to i (i) R

>R
ag
and (ii) R

:W

>R
ag
C

W

O(Wh-I): s ought to withhold intention with respect to i (i) W

C

R
ag
>R

:W

and (ii) W

C
not
R

>R
ag
:W

As I said above, condition (i) in O(Int) is not redundant, as it was in the epistemic
case, O(Bel). In the epistemic case, as long as an agent has more reason to believe
that p than to withhold (out of those two options), she was guaranteed to have
more reason to believe thatp rather than believe that:p (out of those two options).
This relied on the assumption that there are not epistemic reasons not to withhold
belief. But as I said in defending the possibility of a unied account of rational
withholding from Harman (2004)'s observation, there frequently are reasons not to
withhold intention. On the contrastivist view I'm developing here, these are the
reasons to intend to rather than withhold and to intend not to rather than
withhold|:W

. The upshot here is that an agent may have more reason to intend
to than to withhold intention (out of those two options) without having more
reason to intend to rather than intend not to . In fact, this kind of thing
seems to happen frequently. I have to settle on an intention to go to the talk or
an intention not to go to the talk|I can't just keep deliberating forever|even if
neither of these options is more attractive than the other.
In the epistemic case, I claimed that whenever your evidence for and againstp is
tied, you ought to withhold belief. The analogous claim is not true in the practical
16
I use `C
not
' simply for notational convenience, for the set of reasons associated with costs
of intending not to .
120
case, as Harman observed. Sometimes you just have to decide. But it is still true
that in some cases in which your reasons for and against -ing are tied, you ought
to withhold intention. And, crucially, there are both cases in which you ought to
withhold and have very little reason either way, and cases in which you ought to
withhold and have strong reasons each way. So we still need an explanation for
why your reasons to withhold intention are weightier than your reasons to intend
to and your reasons to intend not to in these cases.
A non-contrastivist account, for example, a practical analog of Schroeder (2012a)'s
view, will not have such an explanation. That is because the weights of the kinds of
reasons to withhold that this kind of account could recognize|the fact that you'll
get more information before the time for action, costs of closing deliberation, and
so on|seem to have nothing to do with the weight of your reasons to and the
weight of your reasons not to .
But the account I've developed here, captured in O(Wh-I) can provide such an
explanation of how your reasons to withhold intention could vary in the necessary
ways with the weights of your reasons to and your reasons not to . And just
as in the epistemic case, this feature comes from the move, proprietary to the
contrastivist, that reasons to are also reasons to withhold intention rather than
intend not to , and reasons not to are also reasons to withhold intention rather
than intend to .
Finally, just as in the epistemic case, this account can explain why sometimes
you ought to withhold intention even if your reasons to and reasons not to
are merely almost tied. As long as they are close enough to be outweighed by (i)
the reasons to withhold intention in W

and (ii) the costs of settling deliberation,
and there aren't weighty reasons not to withhold intention, you ought to withhold
intention.
6.2.4 A unied account
So far I have shown how the following two principles make attractive predictions
about when an agent ought to withhold belief and intention, respectively:
O(Wh-B): s ought to withhold belief with respect top i (i) W
p
C
p
Ev
:p
>Ev
p
,
and (ii) W
p
C
:p
Ev
p
>Ev
:p
121
O(Wh-I): s ought to withhold intention with respect to i (i) W

C

R
ag
>R

:W

and (ii) W

C
not
R

>R
ag
:W

These two principles are structurally identical, other than the addition of `:W

'|
reasons not to withhold intention|in O(Wh-I). Given the (somewhat controver-
sial) assumption that there are not reasons not to withhold belief, we can simply add
to O(Wh-B an empty set, `:W
p
', without changing anything substantive about
the account. Then the two principles are structurally identical, giving us a unied
account of when agents ought to withhold that applies to both belief and intention.
O(Wh): s ought to withhold attitude with respect to content i (i) W

C

R
:
> R

:W

, and (ii) W

C
:
R

>R
:
:W

Moreover, this account is just a special case of the much more general principle re-
lating an agent's reasons with what she ought to do, CRO. Thus, the contrastivist
view I have developed in this dissertation provides an attractive account of when
agents ought to withhold that can solve problems facing otherwise attractive non-
contrastive views. Crucially, as I've argued, the advantages of the contrastive ac-
count actually come from contrastivism: only the contrastivist, it seems, can allow
(i) evidence to be reasons to withhold belief, rather than the appropriate alterna-
tive and (ii) reasons for action to be reasons to withhold intention, rather than the
appropriate alternative.
6.3 A Non-Contrastive Alternative
So far I have presented a theory of rational withholding that falls out of the more
general contrastivist theory of reasons and `ought' that I have developed in the
dissertation, along with a particular conception of epistemic and practical (in my
restricted sense) reasons. A central attraction of this theory is that it can explain
why agents ought to withhold in cases of ties|always, in the case of belief, and
often, in the case of intention. The theory is able to explain this because reasons
for the \positive" attitude (believingp, intending to) are also reasons to withhold
rather than have the \negative" attitude, and reasons for the \negative" attitude are
also reasons to withhold rather than have the \positive" attitude. I argued above
that this kind of move is proprietary to the contrastivist. If a non-contrastive
theory, like the one developed by Schroeder (2012a), allowed reasons to believe p
122
and reasons to believe:p, or reasons to intend to and reasons to intend not to
, to also be reasons to withhold, then withholding would nearly always be what
agents have most reason to do|what they ought to do.
But in this section I want to consider a non-contrastive alternative that seems
to have this same advantage, without resorting to contrastive reasons. One way to
think about the key to the contrastivist account's explanation of why you ought
to withhold in the case of ties is the following. Non-contrastive accounts only let
evidence thatp, or reasons to intend to, count in favor of believingp, or intending
to . When we're trying to determine whether your reasons to withhold outweigh
your reasons to believe:p, or to intend not to, these reasons to believep or intend
to do not have any role to play. The contrastivist, on the other hand, gives these
reasons for the positive attitude a role to play when comparing withholding with
the negative attitude. The account I want to consider now takes this insight from
the contrastivist, and builds it into a non-contrastive theory.
17
6.3.1 The account
As before, I will begin with the epistemic case. Rather than conceiving of the
alternatives as believing p, believing:p, and withholding belief, we think of them
as believing p, not believing p, believing:p, and not believing:p. Withholding
seems to have dropped out of the picture. But we can simply say that withholding
belief with respect to p is neither believing p nor believing:p.
18
Just like we have reason to and reasons not to for a given action, we have
reasons to believe p and reasons not to believe p, for propositions p. Reasons to
believep are again just evidence that p. Reasons not to believe p include potential
costs of falsely believingp, as before, but also evidence that:p. Similarly, of course,
reasons to believe:p are evidence for:p, and reasons not to believe:p include
both potential costs of falsely believing:p and evidence that p.
We then adopt the following principles about when you ought to believe (listing
both is actually redundant, but helps make things clear):
17
Mark Schroeder has suggested this account in conversation.
18
This is one place where the question of whether withholding is best thought of as a special
kind of attitude or as simply the absence of belief is important. If withholding is a distinct kind
of attitude, this account would need to be modied.
123
Believe p: You ought to believe that p i your reasons to believe that p outweigh
your reasons not to believe that p.
Believe:p: You ought to believe that:p i your reasons to believe that:p out-
weigh your reasons not to believe that:p.
Here is how this account explains why you ought to withhold belief in cases of
ties. In cases in which your evidence is tied, these principles will say that it's not the
case that you ought to believep and that it's not the case that you ought to believe
:p. Again, this is because we let evidence for p provide reasons against believing
:p, and vice versa. But to get the result that you ought to withhold whenever
your evidence is tied, we need some principle about when you ought to withhold.
A natural choice, given the two principles about belief above, is the following:
Withhold: You ought to withhold belief with respect to p i (i) your reasons not
to believe p outweigh your reasons to believe p, and (ii) your reasons not to
believe:p outweigh your reasons to believe:p.
When your evidence is tied, your reasons to believe p will never outweigh your
reasons not to believep (similarly for:p). If we adopt a version of Meta-Reason,
and say that the fact that your evidence is tied is always some reason not to believe
p and some reason not to believe:p, then we'll get the result that you ought to
withhold in every case in which your evidence is tied.
This appeal to Meta-Reason raises an important question about this approach.
We have reasons to believe p and reasons not to believe p, as well as reasons to
believe:p and reasons not to believe:p. What about reasons to withhold? Since
withholding isn't a distinct alternative, on this approach, it's not surprising that
we don't have a distinct category of reasons to withhold. But some considerations,
like the fact that you'll get more evidence soon, and the fact that your evidence is
tied, seem best thought of as reasons to withhold. On this picture, we can think
of these kinds of considerations as both reasons not to believe p and reasons not to
believe:p. So while we don't have a distinct category of reasons to withhold, we
can just think of the set of reasons to withhold as the intersection of the sets of
reasons not to believe p and reasons not to believe:p.
This alternative view gains the advantages of the contrastive view by letting
evidence that p count against believing:p, in addition to counting in favor of
124
believing p. The contrastivist implementation of this idea is to treat evidence that
p as providing not only reasons to believe that p rather than believe that:p and
to believe that p rather than withhold belief, but also reasons to withhold rather
than believe that:p. This account, on the other hand, lets evidence thatp provide
both reasons to believe that p and reasons not to believe that:p.
Generalizing this account to the practical case is straightforward. The main
dierence is that we need to say what reasons not to withhold intention are. Again,
withholding is not a separate attitude on this account, but instead must be reduced
to not intending to and not intending not to. So just as we can think of reasons
to withhold as both reasons not to intend to and not to intend not to , we can
think of reasons not to withhold as both reasons to intend to and to intend not
to .
6.3.2 Comparing
Which of these two accounts|the contrastivist view I developed in the last section,
or the non-contrastivist alternative I've been developing in this section|should we
prefer? Of course, the main dierence between the two accounts is that one appeals
to contrastive reasons and one does not. So if the arguments I've given in favor
of contrastivism in the rest of the dissertation are compelling, we should prefer
the contrastivist account. If not, perhaps we should prefer the non-contrastivist
alternative. But besides this obvious dierence, the two accounts are quite similar.
In particular, both treat evidence forp as counting against believing:p in addition
to counting in favor of believing p, though they do so in dierent ways.
One striking dierence between the two accounts is that the contrastivist ac-
count, but not the non-contrastive account, posits withholding belief and with-
holding intention as a third alternative, not necessarily reducible to believing and
disbelieving. On the non-contrastive account, we reduce withholding belief with
respect to p as not believing p and not believing:p, and withholding intention to
not intending to and not intending not to .
Corresponding to this dierence is a dierence in what kinds of reasons each ac-
count recognizes. While the contrastivist account recognizes reasons for and against
withholding belief or intention as a distinct class of reasons, the non-contrastive ac-
count reduces them to reasons for and against believing and reasons for and against
125
intending. What we would naturally call a reason to withhold belief with respect to
p, like the fact that you'll get more information soon, the non-contrastive account
understands as a reason both against believing p and against believing:p; simi-
larly for reasons to withhold intention, like the fact that you'll get more relevant
information soon. Finally, reasons against withholding intention, like the fact that
you need to coordinate with a friend, are reasons to intend to and to intend not
to . This is where I think the contrastivist account gains an advantage over the
non-contrastive account.
In Chapter 4, I argued that reasons for one option are reasons against incompat-
ible options. A reasonr for an actionA highlights some good feature ofA|on the
picture I've been working with, it explains whyA-ing would promote some valuable
objective. So doing something incompatible with A-ing would cause you to miss
out on this good feature of A-ing, and r would be part of the explanation for why.
Thus, it's very plausible to think that reasons for one option are reasons against
incompatible options. This is just a way to capture the very plausible idea that the
opportunity costs of an option|the things you miss out on by performing it|are
reasons against it.
19
So, while not totally uncontroversial, this thesis is attractive.
But the advocate of the non-contrastive account I'm discussing cannot accept it.
I will focus on the case of withholding intention, since it should be uncontroversial
that there are reasons not to withhold intention (beyond just reasons for and against
-ing). Consider the fact that I need to coordinate my plans for tomorrow with
my friend so that we can meet for lunch. This fact is a reason against withholding
intention about whether to go in to campus tomorrow; call it r. On the non-
contrastive account, we understand reasons against withholding like r as reasons
both to intend to and to intend not to. But now add the attractive thesis that
a reason for one option is a reason against incompatible options. Sincer is a reason
to intend to, given this thesis, it is a reason against intending not to. Similarly,
sincer is a reason to intend not to, given this thesis, it is a reason against intending
to. Sor is both a reason against intending to and against intending not to. But
according to this account, that is just what it is forr to be a reason for withholding
intention. So given (i) the non-contrastive account and (ii) the attractive thesis that
reasons for one option are reasons against incompatible options, we get the absurd
19
I cited Greenspan (2005) as someone who rejects this view, and Sinnott-Armstrong (2006);
Broome (ms); Ruben (2009) as supporters.
126
result that all reasons against withholding are also reasons for withholding.
20
Thus,
the advocate of this non-contrastive account must reject the attractive thesis that
reasons for one option are reasons against incompatible options. This is a theoretical
cost.
The contrastivist, on the other hand, can accept this thesis, since she does
not reduce reasons for and against withholding to reasons for and against be-
lieving/intending. Moreover, the contrastivist, unlike the advocate of the non-
contrastive account, does not need to claim that a reason against withholding in-
tention with respect to is always a reason to intend to . This is an unintuitive
claim, even when we add that it's also a reason to intend not to (similar remarks
go in the belief case, of course). Some reasons against withholding intention, on
the contrastivist account, will not be reasons for intending to or for intending
not to , relative tofwithhold, intend to , intend not to g, since it could be
that the objectiveX which providesr as a reason against withholding is such that
P (Xjyou intend to ) = P (Xjyou intend not to ), even though both are greater
than P (Xjyou withhold intention).
21
I conclude that the contrastivist account I
developed has theoretical advantages over the non-contrastive account, even be-
yond the advantage of tting into an independently attractive contrastivist view of
reasons in general.
6.4 Wrap Up
In this chapter I've argued that the contrastivist can give an attractive account of
when one ought to withhold both belief and intention. Given the importance of
20
This absurd result does not follow if we treat reasons against withholding as reasons to intend
to or reasons to intend not to, instead of as both reasons to intend to and reasons to intend
not to . And this treatment may actually seem more plausible. The problem, though, is that it
is not clear how we could incorporate reasons not to withhold, thought of disjunctively, into this
picture. In order to count against withholding intention, on this picture, the reasons must either
count in favor of intending to or count in favor of intending not to , or both|counting in
favor of doing one or the other will not work. Given this, the only reasonable choice is to let these
reasons not to withhold both count in favor of intending to and count in favor of intending not
to , since just picking one at the exclusion of the other would be arbitrary.
21
This is one place where we can see the importance of being able to distinguish between reasons
against an option which imply a signicant criticism of the option (in Greenspan (2005)'s sense)
and those which do not. I argued in Chapter 4 that the account I developed there can make that
distinction.
127
withholding and the problems facing other accounts, if I'm right, then this is a com-
pelling argument in favor of contrastivism about reasons. Moreover, this chapter
and the last show that contrastivism has interesting applications in areas in which
reasons are important, including ethics, practical reasoning, and epistemology.
I've argued in this dissertation for contrastivism about reasons. After introduc-
ing the project in Chapter 1, I began, in Chapter 2, by arguing that our talk about
reasons supports a contrastive semantics for `reason'. I then argued, in Chapter 3,
that we shouldn't stop there: normative favoring itself is contrastive. And nally,
in Chapter 4, I developed the theory in more detail by analyzing reasons in terms of
promotion|a popular strategy which itself provides independent support for con-
trastivism. Moreover, the theory I developed in Chapter 4 provides some needed
constraints on the independence of reasons relative to dierent sets of alternatives,
giving the theory structure. In Chapters 5 and 6, I illustrated some applications of
the theory I developed in Chapters 2 through 4. I think that, taken together, all of
this gives us strong reason to accept contrastivism, rather than either reject it or
withhold.
128
References
Anderson, E. (1993). Value in Ethics and Economics. Harvard University Press.
Baron, M., Pettit, P., and Slote, M. (1997). Three Methods of Ethics: A Debate.
Blackwell Publishers.
Bedke, M. (ms). Ends to means: Transmitted reasons and their weights. Unpub-
lished manuscript, University of British Columbia.
Behrends, J. and DiPaolo, J. (2011). Finlay and Schroeder on promoting a desire.
Journal of Ethics and Social Philosophy.
Blaauw, M. (2012). Contrastive belief. In Blaauw, M., editor, Contrastivism in
Philosophy. Routledge.
Bratman, M. (1987). Intention, Plans, and Practical Reason. Harvard University
Press.
Broome, J. (1991). Weighing Goods. Blackwell Publishers.
Broome, J. (1999). Incommensurable value. In Crisp, R. and Hooker, B., edi-
tors, Well-Being and Morality: Essays in Honour of James Grin, pages 21{38.
Oxford University Press.
Broome, J. (2004). Reasons. In Wallace, Pettit, Scheer, and Smith, editors,
Reason and Value: Themes from the Moral Philosophy of Joseph Raz, pages
28{55. Oxford University Press.
Broome, J. (ms). Rationality Through Reasoning. Manuscript.
Cariani, F. (2009). The Semantics of `Ought' and the Unity of Modal Discourse.
PhD thesis, University of California, Berkeley.
Cariani, F. (fc). `Ought' and resolution semantics. No^ us.
Chandler, J. (2007). Solving the tacking problem with contrast classes. British
Journal for the Philosophy of Science, 58(3):489{502.
129
Chandler, J. (2013). Contrastive conrmation: Some competing accounts. Synthese,
190(1):129{138.
Chisholm, R. (1976). Person and Object. Open Court.
Conee, E. and Feldman, R. (2004). Evidentialism: Essays in epistemology. Oxford
University Press.
Crisp, R. (2000). Particularizing particularism. In Moral Particularism, pages
23{47. Oxford University Press.
Dancy, J. (1993). Moral Reasons. Blackwell Publishers.
Dancy, J. (2004). Ethics Without Principles. Oxford University Press.
Danielsson, S. and Olson, J. (2007). Brentano and the buck-passers. Mind, 116:511{
522.
D'Arms, J. and Jacobson, D. (2000a). The moralistic fallacy: On the `appropriate-
ness' of emotions. Philosophy and Phenomenological Research, 61(1):65{90.
D'Arms, J. and Jacobson, D. (2000b). Sentiment and value. Ethics, 110(4):722{748.
Darwall, S. (1983). Impartial Reason. Cornell University Press.
Dretske, F. (1970). Epistemic operators. Journal of Philosophy, 67:1007{1023.
Evers, D. (2009). Humean agent-neutral reasons? Philosophical Explorations,
12(1):55{67.
Finlay, S. (2001). What Does Value Matter? The Interest-Relational Theory of the
Semantics and Metaphysics of Value. PhD thesis, University of Illinois, Urbana-
Champaign.
Finlay, S. (2006). The reasons that matter. Australasian Journal of Philosophy,
84(1):1{20.
Finlay, S. (2009). Oughts and ends. Philosophical Studies, (143):315{340.
Finlay, S. (ms). Confusion of Tongues. Unpublished book manuscript.
Finlay, S. and Snedegar, J. (fc). One `ought' too many. Philosophy and Phenomeno-
logical Research.
Fitelson, B. (2012). Contrastive Bayesianism. In Blaauw, M., editor, Contrastivism
in Philosophy. Routledge.
Friedman, A. (2009). Intransitive ethics. Journal of Moral Philosophy, 6:277{297.
130
Greenspan, P. (2005). Asymmetrical practical reasons. In Reicher, M. and
Marek, J., editors, Experience and Analysis: Proceedings of the 27th Interna-
tional Wittgenstein Symposium, pages 387{394. oebv&hpt.
Greenspan, P. (2007). Practical reasons and moral `ought'. In Shafer-Landau, R.,
editor, Oxford Studies in Metaethics, volume 2, pages 172{194. Oxford University
Press.
Grice, H. (1989). Logic and conversation. In Studies in the Way of Words. Harvard
University Press.
Groenendijk, J. and Stokhof, M. (1997). Questions. In van Benthem, J. and ter
Meulen, A., editors, Handbook of Logic and Language, pages 1055{1124. Elsevier
Science Publishers.
Hamblin, C. L. (1958). Questions. Australasian Journal of Philosophy, 36:159{168.
Hampton, J. (1998). The Authority of Reason. Cambridge University of Press.
Harman, G. (2004). Practical aspects of theoretical reasoning. In Mele, A. and
Rawling, P., editors, The Oxford Handbook of Rationality. Oxford University
Press.
Hieronymi, P. (2005). The wrong kind of reason. The Journal of Philosophy,
102(9):437{457.
Higginbotham, J. (1993). Interrogatives. In Hale, K. and Keyser, S. J., editors, The
View from Building 20: Essays in honor of Sylvain Bromberger, pages 195{228.
The MIT Press.
Higginbotham, J. (1996). The semantics of questions. In Lappin, S., editor, The
Handbook of Contemporary Semantic Theory, pages 361{383. Oxford University
Press.
Hitchcock, C. (1996). The role of contrast in causal and explanatory claims. Syn-
these, 107(3):395{419.
Horty, J. F. (2007). Reasons as defaults. Philosophers' Imprint, 7(3).
Horty, J. F. (2012). Reasons as Defaults. Oxford University Press.
Jackson, F. (1985). On the semantics and logic of obligation. Mind, 94(374):177{
195.
Jackson, F. and Pargetter, R. (1986). Oughts, options, and actualism. Philosophical
Review, 95(2):233{255.
131
Kavka, G. (1983). The toxin puzzle. Analysis, 43(1):33{36.
Kierland, B. (ms). A new approach to the future dependence issue. Unpublished
manuscript, Boise State University.
Kolodny, N. (ms). Instrumental reasons. Unpublished manuscript.
Kratzer, A. (1981). The notional category of modality. In Eikmeyer, H. J. and
Rieser, H., editors, Words, Worlds, and Contexts, pages 38{74. de Gruyter,
Berlin.
Lehrer, K. and Paxson, T. (1969). Knowledge: Undefeated justied true belief.
Journal of Philosophy, 66(4):225{237.
Lewis, D. (1996). Elusive knowledge. Australasian Journal of Philosophy, 74:549{
567.
Lipton, P. (1990). Contrastive explanation. Royal Institute for Philosophy Supple-
ment, 27:247{266.
Moore, G. (1903). Principia Ethica. Cambridge University Press.
Moore, G. (1912). Ethics. Williams & Norgate.
Morton, A. and Karjalainen, A. (2003). Contrastive knowledge. Philosophical
Explorations, 6(2):74{89.
Nagel, T. (1970). The Possibility of Altruism. Princeton University Press.
Norcross, A. (1997). Good and bad actions. Philosophical Review, 106(1):1{34.
Norcross, A. (2005a). Contextualism for consequentialists. Acta Analytica, 20(2).
Norcross, A. (2005b). Harming in context. Philosophical Studies, 123(1-2).
Olson, J. (2004). Buck-passing and the wrong-kind of reasons. The Philosophical
Quarterly, 54(3):295{300.
Part, D. (1984). Reasons and Persons. Oxford University Press.
Part, D. (2001). Rationality and reasons. In Egonsson, Josefsson, Petterson,
and Rnnow-Rasmussen, editors, Exploring Practical Philosophy: From action
to values, pages 17{39. Ashgate.
Part, D. (2011). On What Matters. Oxford University Press.
Pettit, P. (1991). Consequentialism. In Singer, P., editor, A Companion to Ethics,
pages 230{237. Blackwell Publishers.
132
Piller, C. (2001). Normative practical reasoning. Proceedings of the Aristotelian
Society, 25:195{216.
Piller, C. (2006). Content-related and attitude-related reasons for preferences. Phi-
losophy, 59:155{181.
Rabinowicz, W. and Rnnow-Rasmussen, T. (2004). The strike of the demon: On
tting pro-attitudes and value. Ethics, 114(3):391{423.
Rachels, S. (1998). Counterexamples to the transitivity of `better than'. Aus-
tralasian Journal of Philosophy, 76(1):71{83.
Rachels, S. (2001). A set of solutions to Part's problems. No^ us, 35:214{238.
Raz, J. (1999). Engaging Reason: On the theory of value and action. Oxford
University Press.
Reisner, A. (2009). Abandoning the buck-passing analysis of nal value. Ethical
Theory and Moral Practice, 12(4):379{395.
Rickless, S. (fc). The contrast-insensitivity of knowledge ascriptions. Philosophy
and Phenomenological Research.
Rooth, M. (1992). A theory of focus interpretation. Natural Language Semantics,
1:75{116.
Ross, J. (2006). Acceptance and Practical Reason. PhD thesis, Rutgers University.
Ruben, D.-H. (1987). Explaining contrastive facts. Analysis, 47(1):35{37.
Ruben, D.-H. (2009). Con-reasons as causes. In Sandis, C., editor, New Essays on
the Explanation of Action, pages 62{74. Palgrave Macmillan.
Scanlon, T. (1998). What We Owe to Each Other. Harvard University Press.
Schaer, J. (2004). From contextualism to contrastivism. Philosophical Studies,
119(1-2):73{104.
Schaer, J. (2005a). Contrastive causation. Philosophical Review, 114(3):327{358.
Schaer, J. (2005b). Contrastive knowledge. In Gendler and Hawthorne, editors,
Oxford Studies in Epistemology, volume 1, pages 235{271. Oxford University
Press.
Schaer, J. (2007). Knowing the answer. Philosophy and Phenomenological Re-
search, 75(2):383{403.
133
Schaer, J. (2008). The contrast-sensitivity of knowledge ascriptions. Social Epis-
temology, 22(3):235{245.
Schaer, J. (2012). Causal contextualisms. In Blaauw, M., editor, Contrastivism
in Philosophy. Routledge.
Schroeder, M. (2007). Slaves of the Passions. Oxford University Press.
Schroeder, M. (2010). Value and the right kind of reason. In Shafer-Landau, R.,
editor, Oxford Studies in Metaethics, volume 5, pages 25{55. Oxford University
Press.
Schroeder, M. (2012a). Stakes, withholding, and pragmatic encroachment on knowl-
edge. Philosophical Studies, 160(2):265{285.
Schroeder, M. (2012b). The ubiquity of state-given reasons. Ethics, 122(3):457{488.
Searle, J. (2001). Rationality in Action. MIT Press.
Sinnott-Armstrong, W. (1992). An argument for consequentialism. Philosophical
Perspectives, 6:399{421.
Sinnott-Armstrong, W. (2004). Classy Pyrrhonism. In Sinnott-Armstrong, W.,
editor, Pyrrhonian Skepticism. Oxford University Press.
Sinnott-Armstrong, W. (2006). Moral Skepticisms. Oxford University Press.
Sinnott-Armstrong, W. (2008). A contrastivist manifesto. Social Epistemology,
22(3):257{270.
Skorupski, J. (2010). The Domain of Reasons. Oxford University Press.
Sloman, A. (1970). `Ought' and `better'. Mind, 79(315):385{394.
Snedegar, J. (2012). Contrastive semantics for deontic modals. In Blaauw, M.,
editor, Contrastivism in Philosophy. Routledge.
Stanley, J. (2005). Knowledge and Practical Interests. Oxford University Press.
Stanley, J. (2011). Know How. Oxford University Press.
Stratton-Lake, P. (2005). How to deal with evil demons: Comment on rabinowicz
and rnnow-rasmussen. Ethics, 115(4):778{798.
Suikkanen, J. (2005). Reasons and value { A defence of the buck-passing account.
Ethical Theory and Moral Practice, 7(5):513{535.
Suikkanen, J. (2009). Buck-passing accounts of value. Philosophy Compass,
4(5):768{779.
134
Temkin, L. (1987). Intransitivity and the Mere Addition Paradox. Philosophy and
Public Aairs, 16(2):138{187.
Temkin, L. (1996). A continuum argument for intransitivity. Philosophy and Public
Aairs, 25:175{210.
Temkin, L. (2012). Rethinking the Good: Moral Ideals and the Nature of Practical
Reasoning. Oxford University Press.
Temple, D. (1998). The contrast theory of why-questions. Philosophy of Science,
55(1):141{151.
Toulmin, S. (1950). Reason in Ethics. Cambridge University Press.
van Fraassen, B. (1980). The Scientic Image. Oxford University Press.
V ayrynen, P. (2010). A wrong turn to reasons? In Brady, M. S., editor, New Waves
in Metaethics. Palgrave Macmillan.
Way, J. (fc). Value and reasons to favor. In Shafer-Landau, R., editor, Oxford
Studies in Metaethics, volume 8. Oxford University Press.
Wedgwood, R. (2009). Intrinsic values and reasons for action. Philosophical Issues,
19:342{363.
Yalcin, S. (2011). Nonfactualism about epistemic modality. In Egan, A. and Weath-
erson, B., editors, Epistemic Modality. Oxford University Press.
135

Abstract (if available)

Abstract This dissertation develops and defends a contrastive view of normative reasons, according to which they are fundamentally reasons for one thing rather than an- other, or for one thing out of some range of alternatives. Reasons for something relative to one set of alternatives may not be reasons for that same thing relative to some other alternatives. This is in opposition to nearly all existing work on reasons, which assumes without argument—even without mention—that reasons are reasons for actions simpliciter, independently of what we’re comparing the action to. ❧ I argue that this contrastive view is supported by (i) our talk about reasons, (ii) the idea that reasons favor the things they are reasons for, and (iii) the idea that reasons involve the promotion of things like desires, goals, or values. These arguments build, step by step, to a detailed version of contrastivism that can both claim advantages over the traditional non-contrastive picture and avoid problems that face less well-developed versions of contrastivism. ❧ One thread running through the dissertation is an exploration of the depth of contrastivism. Chapter 2 argues that contrastivism goes at least as deep as our use of the word ‘reason’—this word (as used to ascribe normative reasons) should be given a contrastive treatment. Chapter 3 argues that contrastivism actually goes deeper, to the normative reason, or favoring, relation that normative philosophers are concerned with. Chapter 4, though, argues for a somewhat moderate contrastive theory. The normative reason relation is indeed contrastive, but we need not—in fact, we’d better not—adopt the deepest possible form of contrastivism. ❧ Given the central place of reasons in normative philosophy, we should expect that contrastivism will have upshots throughout areas like ethics, practical rational- ity, and epistemology. In the final two chapters of this dissertation, I start to explore this idea. This is just a start, but it suggests that exploring further applications for contrastivism is likely to be a fruitful area for future research.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Rationality and the primacy of the occurrent

PDF

Reasons, obligations, and the structure of good reasoning

PDF

Positivist realism

PDF

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

PDF

Process-oriented rationality

PDF

Reasoning with uncertainty and epistemic modals

PDF

The case for moral skepticism

PDF

A deontological explanation of accessibilism

PDF

Reasoning with degrees of belief

PDF

Meaningfulness, rules, and use-conditional semantics

PDF

Units of agency in ethics

PDF

A perceptual model of evaluative knowledge

PDF

Rethinking reductive realism in ethics

PDF

Belief as credal plan

PDF

Beliefs that wrong

PDF

Public justification beyond legitimacy

PDF

Constraining assertion: an account of context-sensitivity

PDF

Telling each other what to do: on imperative language

PDF

The mental states first theory of promising

PDF

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

Asset Metadata

Creator Snedegar, Justin L. (author)

Core Title Contrastive reasons

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Philosophy

Publication Date 03/22/2013

Defense Date 03/04/2013

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

Tag contrastivism,metaethics,OAI-PMH Harvest,philosophy,practical rationality,reasons

Language English

Contributor Electronically uploaded by the author (provenance)

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

Creator Email jsned12@gmail.com,snedegar@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-226299

Unique identifier UC11293380

Identifier usctheses-c3-226299 (legacy record id)

Legacy Identifier etd-SnedegarJu-1483.pdf

Dmrecord 226299

Document Type Dissertation

Rights Snedegar, Justin L.

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