This article was downloaded by: [University of Oxford]
On: 21 January 2011
Access details: Access Details: [subscription number 773573297]
Publisher Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 3741 Mortimer Street, London W1T 3JH, UK
Australasian Journal of Philosophy
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713659165
The Standard-Relational Theory of 'Ought' and the Oughtistic Theory of
Reasons
Daan Eversa
a
University of Oxford,
First published on: 22 February 2010
To cite this Article Evers, Daan(2011) 'The Standard-Relational Theory of 'Ought' and the Oughtistic Theory of Reasons',
Australasian Journal of Philosophy, 89: 1, 131 — 147, First published on: 22 February 2010 (iFirst)
To link to this Article: DOI: 10.1080/00048401003610890
URL: http://dx.doi.org/10.1080/00048401003610890
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Australasian Journal of Philosophy
Vol. 89, No. 1, pp. 131–147; March 2011
THE STANDARD-RELATIONAL THEORY
OF ‘OUGHT’ AND THE OUGHTISTIC
THEORY OF REASONS
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
Daan Evers
The idea that normative statements implicitly refer to standards has been
around for quite some time. It is usually defended by normative antirealists,
who tend to be attracted to Humean theories of reasons. But this is an
awkward combination: ‘A ought to X’ entails that there are reasons for A to
X, and ‘A ought to X all things considered’ entails that the balance of reasons
favours X-ing. If the standards implicitly referred to are not those of the agent,
then why would these entailments hold? After all, Humeanism says that ‘A has
a reason to X’ is true if and only if A has some desire which is furthered by
X-ing.
In this paper, I develop a standard-relational theory of ‘ought’ and a nonHumean theory of reasons (oughtism). Together, they explain why ‘A ought to
X’ entails not only that there are reasons for A to X, but also that the balance
of reasons favours X-ing. The latter explanation depends on a theory of
weight, in which the weight of a reason depends on the position of a rule
(standard) in an order of priorities. The theories are truth-conditional, but do
not require objective normative facts for the truth of ‘ought’ judgments and
judgments about reasons.
1. Introduction
Antirealists or subjectivists about normativity deny the existence of
objective normative facts (in some sense of ‘objective’ which I will not
attempt to delineate here). Concerning the meaning of normative statements
(statements involving words like ‘ought’, ‘good’, ‘wrong’, etc.), antirealists
roughly have three options: (1) they can say that normative statements
purport to describe objective normative facts which do not exist. This entails
that all positive and non-analytic normative statements are false. Or (2) they
can deny that normative statements purport to describe anything at all.
Instead, they might be expressions of approval or disapproval, much like
‘Hurray!’ is an expression (but not a description) of excitement or
enthusiasm. This means that the surface grammar of normative language
is deceptive. Finally, (3) they can say that normative statements do purport
to describe normative facts, but not objective ones (in the sense of ‘objective’
I have not bothered to explicate). Or at least, they can say that the meaning
of normative statements is not such as to require the existence of objective
Australasian Journal of Philosophy
ISSN 0004-8402 print/ISSN 1471-6828 online Ó 2011 Australasian Association of Philosophy
http://www.tandf.co.uk/journals DOI: 10.1080/00048401003610890
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
132
Daan Evers
normative facts in order for these statements to be true. I want to develop a
version of the third option. It is a form of contextualism about normative
language.
Here is one reason why the third option is not easy to develop. It is not
particularly plausible that ‘ought’ statements are sensitive to the desires of
agents. Take the statement: ‘Hitler ought not to have ordered the
extermination of the Jews’. Few think that its truth is affected by whether
Hitler had any desires that were furthered by that action. So it is, at least
prima facie, plausible that its truth conditions do not involve Hitler’s desires.
However, ‘A ought to X’ arguably entails that there are reasons for A to X.
But many antirealists are Humeans about reasons: they believe that the
truth of ‘A has a reason to X’ requires that A has some desire which is
furthered by X-ing. But if the truth conditions of ‘A ought to X’ have
nothing to do with A’s desires, then why should it entail anything about A’s
reasons, if the reasons that A has do depend on A’s desires? This challenge is
not often met. In this paper, I provide a theory of ‘ought’ and reasons which
explains why ‘A ought to X’ entails not only that A has reasons to X, but
also that the balance of reasons favours X-ing. My attempt to meet the
challenge, however, involves abandoning Humeanism about practical
reasons.
2. The Standard-Relational Theory of ‘Ought’
In this section, I outline a standard-relational theory of ‘ought’. The idea
behind it is that all statements about what we ought to do are made relative
to some contextually-supplied standards. In other words: there are no
categorical ‘ought’s, if by that we mean a truth about what we ought to do
which does not implicitly refer to a standard. This type of theory has been
mentioned or developed (more or less explicitly) by authors like Mackie
[1977], Wong [1984], Harman [1975], and Harman and Thomson [1996]. But
so far it lacks enough semantic detail to allow a fair assessment. And it
hasn’t been combined with a non-Humean theory of reasons.
The standard-relational theory of ‘ought’ says this: ‘ought’ refers to a
three-place relation between an agent, an act, and a system of rules. It is the
relation of being required by. This is not always true, but as a first
approximation it will do.1 This proposal comes to the following: if I say ‘A
ought morally to X’, the meaning of my statement can be represented as: X
is required of A by [the relevant system of] moral rules. If I say ‘A ought
1
It isn’t always true, because we can judge that things ought to be the case. There needn’t be anyone who
‘owns’ this requirement in the sense that s/he ought to bring it about that the state of affairs obtains. It is not
obvious what to say about this, but I suggest that in such cases ‘ought’ refers to a two-place relation between
a state of affairs and a system of rules. As will become clear, I analyse the notion of an act being required by a
system in terms of rules in the system entailing that an act is (not) done by the agent. Rules entail various
things in virtue of their propositional contents. Normally, these contents involve agents. E.g. the content of
‘Don’t kill!’ involves a quantifier which ranges over agents. But this needn’t be the case. There could also be
‘rules’ for states of affairs, which reflect certain ideals about the world. These rules can be represented by
means of constructions like: ‘Let there be no suffering!’, whose propositional content is that there is no
suffering. I suggest that ‘ought’ in statements like ‘It ought to be the case that p’ refers to a relation between a
state of affairs and a system of non-agent-involving rules.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
The Standard-Relational Theory of ‘Ought’
133
prudentially to X’, the meaning of my statement can be represented as: X is
required of A by [the relevant system of] prudential rules.2
Several things need specification. First, what are rules? Second, what is it
to be required by a system of rules? Third, what determines the relevant
rules? Fourth, in what way do the rules figure in the truth conditions of an
‘ought’ statement? I will answer these questions in turn.
First: what are rules? The kind of rules relevant here can be represented
by imperatives: ‘Do/don’t do X!’.3 We can think of them as propositional
contents which are not asserted but put forward with imperative force. The
propositional content of ‘Don’t steal (in circumstances C)!’ is that no one
steals (in C). However, it is not essential to think of rules as propositions
put forward with some kind of non-assertoric force. We can also think of
rules as propositions describing (aspects of) ideal worlds.4 What matters is
that rules do not involve normative concepts like ought or permissible. For I
want to analyse such concepts in terms of rules.
Second: what is it for an act to be required by a system of rules? For ease
of exposition, I will stipulate that ‘X’ is a variable that takes both acts and
omissions as values. So ‘X’ can both stand for the act of killing and the
omission of refraining from killing. I will sometimes simply talk about acts,
which are supposed to comprise both doing something and refraining from
doing something. Now I can say that X is required by a system if and only if
conformity to the system requires that one X-es. This can mean either one of
two things:
(a) In a simple case, where the rules have no inconsistent implications with
respect to X in C, conforming to the system means doing what is required by
at least one rule in the system. An act (omission) is required by a rule just in
case the act (omission) has some feature such that the rule entails that you
do acts (refrain from doing acts) with that feature. Say that pushing a
button in C will kill many people. In that case, the rule ‘Don’t kill in C!’
requires you not to push the button, because the act has a feature (being
such as to kill people) such that the rule entails that you refrain from acts
with that feature in C. (I am assuming that the propositional content of the
rule is something like that no one kills in C, which entails that you don’t kill
in C and thus not push the button either.)
(b) In a complex case, where the system contains rules with inconsistent
implications with respect to X in C, conforming to the system means doing
2
The standard-relational theory of ‘ought’ theoretically allows for objective standards: standards that are
objectively correct. In this respect, it is neutral with respect to realism or antirealism.
3
This is of course exactly what David Wong says in [1984: 37]. Not all rules can be represented by means of
imperatives. For example, it is a rule of chess that one wins by checkmating one’s opponent. The rules that
matter for my theory can all be represented by imperatives. (Thanks to an anonymous referee for this
journal.)
4
(Part of) what makes an item with this propositional content a rule rather than, say, a report about the
world, is, in the case of the speaker, the attitude which s/he takes towards the proposition (i.e. some
attitude other than belief). Or, in the case of the hearer, it is his/her understanding that the proposition is
not to be considered as a claim about the world. This may itself be a belief about the role of the content in
the context. Because rules (and ideal worlds) are individuated with reference to the attitude taken towards
propositional contents, I rely on a substantive assumption about mental states like desires, namely that
they can be understood in non-normative terms. This assumption is often made by antirealists about
normativity, but is not uncontroversial. I do not have the space to argue for it here. If this assumption is
false, I may not without circularity be able to analyse ‘ought’ in terms of rules. (Thanks to an anonymous
referee for this journal.)
134
Daan Evers
what is required by the rule with the highest priority. For example: a system
may contain both a rule of morality (which requires X in C), and a rule of
prudence (which requires not-X in C). In this case, X is required of you in C
by the system if and only if (1) the moral rule requires that you X in C (as
explained in (a)) and (2) the moral rule has higher priority than the rule of
prudence.
We can now state a general definition which covers both cases of
conforming to a system of rules:
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
Conformity: an agent A conforms to a system of rules S by X-ing in C if and
only if X-ing has a feature such that a rule in S entails that A performs acts
with that feature in C and X-ing does not have a feature such that another rule
in S with the same or higher priority entails that A does not perform acts with
that feature in C.
What does it mean to say that a rule has higher priority than another rule? It
means that there is a relevant system of higher-order rules conformity to
which requires that the first rule (or the system of which it is a part) prevail.
These rules can be represented by means of imperatives like: ‘In C, act in
accordance with the moral rule!’ (but, again, that is not essential).
This sheds light on ‘ought’ all things considered (or rather some such
judgments). Suppose that I ought morally to X, but prudentially to not-X.
After some reflection, I judge that I ought to X, all things considered
(henceforth I will call such judgments ‘all-in ‘‘ought’’s’). What is the content
of this judgment?
If we want a unified semantics, we should say that this judgment is also
about the requirements of rules. But it is implausible that the judgment is
relative to the same moral system that is in conflict with the prudential one.
If ‘A ought morally to X’ means that X is required of A by moral system M,
then it would be surprising if (in this situation) ‘A ought all-in to X’ means
that, all-in, X is required of A by moral system M. That would be to repeat
the original moral judgment, this time with the redundant add-on that we
have considered everything. (It may be possible for pragmatics to explain
away the oddness. But there is at least some pressure to avoid the need for
such an explanation.)
In this case, it is more plausible that the rules referred to are higher-order
ones. When the speaker says that A ought all-in to X, s/he is not just
indicating that s/he has considered all relevant things, but also that there is a
system of rules conformity to which requires acting in accordance with the
moral requirement.5
One may wonder whether this system was already in place prior to the
speaker’s reflection on the conflict at hand. And one may also wonder
whether s/he was conscious of being committed to the system prior to this
reflection. But I don’t think that matters, so long as it is true that the speaker
5
It is not ruled out in advance that the system contains rules of higher orders than the second. But this will be
rare in practice.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
The Standard-Relational Theory of ‘Ought’
135
is, at the time of speaking, committed to a system conformity to which
requires that the moral ‘ought’ prevail.
I don’t think this is ad hoc. In the situation described, the all-in ‘ought’
reflects a certain ordering of rules or ends which the speaker either arrives at
on considering all things, or which was already in place before, and which
reflection reveals to require that the moral rule prevail.6
Third: what determines the relevant standards? Stephen Finlay claims
that ‘A ought to X’ means that A ought to X in order that E, where ‘E’ is an
end which is determined by context [Finlay 2009]. In some contexts, the
speaker’s desires play a role in determining the end. But even in these cases,
Finlay does not build those desires into the truth conditions of normative
statements. The identity of the end (which is part of the truth conditions) is
determined by the speaker’s desires, but normative statements are not about
anyone’s desires. Rather, they are about acts and the ends to which the acts
are instrumental.
I think this idea can be retained in the current theory of ‘ought’. My
answer to the third question, then, is this: the relevant standards are
determined by context. Exactly how this works varies from case to case. In
sincere moral ‘ought’s, the speaker’s own commitments are involved in the
identification of the standards. In inverted-commas uses, the speaker refers
to standards s/he does not (necessarily) endorse. The same happens if the
relevant standards are the legal requirements of a country. In some cases, we
identify with people in order to advise them (as it were from their
perspective). Here, the standards are (at least in part) determined by the
advisee’s commitments, or more precisely, by our beliefs about what those
standards are. But although what the standards are determined by varies,
attitudes or beliefs themselves never occur in the truth conditions of
normative statements. Only the standards determined by them do.
Fourth: in what way do the standards determined by context figure in the
truth conditions of an ‘ought’ statement? The standards can figure in the
truth conditions in different ways: for example, their content can be
described, or they can simply be referred to. I don’t think it’s plausible that
the content of the rules is part of the truth conditions of ‘ought’ statements.
If I say: ‘You ought not to lie’, I indicate that lying is forbidden by a system
of rules, but I don’t indicate that lying is forbidden by a system of rules part
of the content of which is ‘Don’t lie!’.
Another question about the way in which the standards figure in the truth
conditions of ‘ought’ statements is to do with quantification over systems
versus reference to a system. Take the following explication of ‘ought’ by
David Wong, who claims that ‘A ought morally to X’ means that
By not doing X under actual conditions C, A will be breaking a rule of an
adequate moral system applying to him or her.
[Wong 1984: 40]
6
The above considerations apply to all-in ‘ought’s made in the light of conflicting requirements. But all-in
‘oughts’ are not always relative to higher-order rules. One can also consider what is all-in morally required.
The ‘all-in’ indicates that we have considered all relevant features of the situation. But the ‘ought’ itself may
be relative to a system of first-order moral rules (unless these rules themselves conflict). This shows that if any
special interest accrues to all-in judgments, they are judgments made in the light of conflicting requirements.
136
Daan Evers
In Wong’s case, it is enough for an ‘ought’ judgment to be true that there is
some adequate system or other which contains a rule which would be broken
by A if s/he refrained from X-ing. So according to Wong, ‘ought’ judgments
quantify over systems of rules. There is not a unique system of rules to
which the speaker is referring. I disagree with Wong in this respect. It is
more plausible that ‘ought’ judgments refer to a particular system of rules.
My standard-relational analysis of ‘ought’, then, is:
Ought: ‘A ought to X in C’ is true iff conformity to the contextually salient
system of rules requires that A X-es in C.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
Fully spelt out, this means:
Oughtfull: ‘A ought to X in C’ is true iff X has a feature such that the
contextually salient system of rules contains at least one rule which entails that
A does acts with that feature in C and X has no feature such that a rule with
the same or higher priority entails that A does not do acts with that feature
in C.
On this theory, relativized ‘ought’ statements are treated as follows: ‘A
ought morally to X’ means that conformity to the contextually salient
system of moral rules requires X-ing. ‘A ought prudentially to X’ means that
conformity to the contextually salient system of prudential rules requires
X-ing.
To sum up: the standard-relational theory of ‘ought’ says that ‘ought’
(often) refers to a three-place relation between an agent, an act, and a system
of rules. It is the relation of being required by. The system is referred to in an
‘ought’ statement, but its content is not described (although the system may,
but needn’t, be explicitly identified as moral or prudential). The kind of
system referred to varies with the type of ‘ought’ in question. By ‘type’ of
‘ought’, I don’t mean that the word ‘ought’ itself has variable meaning. The
semantic value of ‘ought’ is always the relation of being required by. But
the proposition expressed by ‘A ought morally to X’ is not the same as the
proposition expressed by ‘A ought prudentially to X’ (which is what one
would expect). Similarly, the proposition expressed by ‘A ought to X all
things considered’ is (in our example) different from the proposition
expressed by ‘A ought morally to X’ or ‘A ought prudentially to X’.
The standard-relational theory of ‘ought’ has a number of virtues: first of
all, it is unified for all non-epistemic uses of ‘ought’ (and may well allow
extension to epistemic uses). Second, it has intuitive appeal, at least for
lower-level ‘ought’s: it is plausible that ‘A ought morally to X’ is about the
requirements of moral rules or standards. But since it is also plausible that
the meaning of ‘ought’ is similar in moral, prudential, and other contexts, it
is a virtue that the semantics assigns the same value to all such ‘ought’s (i.e.
the relation of being required by). Third, it does not build the desires of
either speakers or agents into the truth conditions of normative statements.
Moral statements are about the requirements of rules, not about anyone’s
desires. Fourth: it allows us to explain why speakers are often motivated to
act in accordance with their ‘ought’ judgments: these judgments often
The Standard-Relational Theory of ‘Ought’
137
concern rules to which the speaker is him/herself committed. And (in
ordinary cases), such commitment will in part be a matter of motivational
states. Fifth: as we will see, it is compatible with a theory of reasons which
explains why ‘A ought to X’ entails that there are reasons for A to X, and
also why ‘A ought all-in to X’ entails that the balance of reasons favours
X-ing.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
3. The Oughtistic Theory of Reasons
Recently, several authors have begun to work on theories of reasons in
which (a) reasons are analysed in terms of ‘ought’ and (b) the strength of a
reason is analysed in terms of ‘ought’ as well.7 I want to sketch a view of
reasons with similar properties.
The standard-relational theory of ‘ought’ says that ‘ought’ statements are
about the requirements of systems of rules. A system of rules requires X-ing
if and only if conformity to the system requires X-ing. The proposal about
reasons is that:
Reasons are features of acts indispensable in an explanation of why a relevant
rule entails that the act is (not) done in C.
Reasons comprise both reasons for and reasons against acts. We can spell this
out as follows:
Reason For: A reason for an agent A to X in C is a feature of X indispensable
in an explanation of why a relevant rule entails that A X-es in C.
Reason Against: A reason for an agent A not to X in C is a feature of X
indispensable in an explanation of why a relevant rule entails that A not-X-es
in C.
The definition is restricted to relevant rules, which are rules that are
contextually salient in the conversation or thought process at hand. This is
necessary in order to forestall overgeneration: there is an infinite number of
rules in logical space, but that does not mean that there are reasons for
anything whatsoever.8
I have said that reasons are those features of acts indispensable in an
explanation of why a relevant rule entails that the act is (not) done. I have
not said that reasons are those features indispensable in an explanation of
why the system requires the act. This is to allow for the possibility that one
has reasons for acts which one ought not, on the whole, to do.
7
Examples include Gert [2005] and Horty [2007] (or at least his view can be read along these lines). Theories
of this type are intended as theories of normative as opposed to motivating reasons. Normative reasons count
in favour of or against actions, whereas motivating reasons play a causal-explanatory role in the genesis of an
agent’s action.
8
Or at least, there aren’t reasons to do anything whatsoever relative to the system salient in the conversation.
There is a sense in which there are reasons to do anything whatsoever. Like all reasons, these are relative
reasons in the sense that they are reasons relative to certain possible rules.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
138
Daan Evers
The definition involves the notion of indispensability, which should be
understood as follows. It may be thought that if a rule entails that A X-es,
then X’s being such that A is capable of it is part of an explanation of why
the rule entails this. But it is not, intuitively, a reason for an act that one is
capable of it (or at least not in most contexts).9 However, explanations are
answers to why-questions and what counts as an adequate answer depends
on what one is interested in. By (my) hypothesis, when one asks ‘What
reasons do I (or some agent) have to X?’, one is interested in learning
whether there are rules that apply to X (i.e. entail something about it). But
when ‘X’ is an act, and one is interested in the reasons in favour of it, one is
already presupposing that one (or some relevant agent) is capable of it. For
reasons count in favour of things that can be done. This makes the answer
‘You’re capable of X’ irrelevant and thus dispensable. Similarly for such
answers as ‘X has a spatio-temporal boundary’. Hence my definition of a
reason as a feature of X indispensable in an explanation of why some
relevant rule entails that the act is (not) done in C. (But I will sometimes
abbreviate ‘being indispensable in an explanation of why’ to ‘explaining
why’.)
To illustrate the definition: if a salient rule is ‘Don’t lie in C!’, then those
features of an act which make it into a case of lying explain why the rule
entails that the agent does not do it in C. And thus they are reasons to
refrain from doing it. If the rule is ‘Preserve life in C!’, then those features of
an act which make it life-preserving explain why the rule entails that the
agent does it. And thus they are reasons for the action.
So my theory of reasons says that reasons for (against) an act are features
of the act which explain why a relevant rule entails that the act is (not) done
(in the situation). But what about the strength of reasons? I will say that a
feature of an act triggers a rule (terminology I take from Horty [2007]) just
in case the rule entails that we (not) do that act at least in part in virtue of
that feature and the feature is a reason. I suggest that a feature is a stronger
reason than some other feature just in case the rule which is triggered by the
first has higher priority than the rule triggered by the second. For example:
the fact that A will die unless I X is a stronger reason to X than the fact that
B will be disappointed is a reason not to X, because the rule which tells us to
preserve life is higher up in the ordering of rules than the rule which tells us
to avoid disappointment.
So we have arrived at the following view of the strength of reasons:
Stronger Than: A feature (or features) F1 is stronger reason for A to X in C
than a feature (features) F2 iff F1 triggers a rule (or rules) higher in the order of
priorities in C than the rule (rules) triggered by F2.10,11
9
This problem was raised by an anonymous referee for this journal.
This formulation is compatible with a situation in which two (or more) features are individually weaker
reasons than a third, though collectively stronger. They will be collectively stronger if and only if the
contextually salient system of rules contains a higher-order rule which requires giving priority to both rules
triggered by these features, and no rules which require giving either one of them priority.
11
This theory of the weight of reasons has a lot in common with the theory developed by Horty [2007]. But
there are some differences. First, Horty is primarily interested in formal models of reasoning (about beliefs or
action). He is not in the business of doing semantics. Second, Horty does not explain what makes it the case
that a rule is higher up the order of priorities than another rule. Third, Horty only explains what makes an
10
The Standard-Relational Theory of ‘Ought’
139
This by itself does not tell us what makes it the case that a set of reasons is
collectively stronger than another set of reasons. I propose to deal with that
as follows:
Collective Weight: If, in C, a feature F1 triggers a rule higher in the order of
priorities than any other rule triggered in C, then the features that trigger rules
prescribing the same act as the one triggered by F1 are, in C, collectively
stronger reasons than the features triggering rules prescribing incompatible
acts lower in the order.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
4. Explaining the Entailments
This brings me to the challenges I promised to discharge at the start of this
paper. The first is to explain why ‘A ought to X in C’ entails that there are
reasons for A to X in C (at least one reason). The second is to explain why
‘A ought all-in to X in C’ entails that the balance of reasons favours X in C.
The first challenge is met as follows: I have said that the truth of ‘A ought
to X in C’ requires that the contextually salient system of rules is such that
conformity to it requires that A X-es in C. In order for this to be true, there
has to be at least one rule in the system which entails that A X-es in C. I
have also said that a reason is a feature of an act which explains why a
relevant rule entails that it is (not) done in C. But if a rule entails that A X-es
in C, then X must have some feature in virtue of which that is entailed. So in
all possible worlds in which A ought to X in C, there is a reason for A to X
in C. Therefore: the standard-relational theory of ‘ought’ paired with
oughtism about reasons explains why ‘A ought to X in C’ entails that there
is at least one reason for A to X in C.12
We also want to show that ‘A ought all-in to X in C’ entails that the
balance of reasons favours X in C (provided that the judgment is made in
the light of conflicting requirements). I have argued that such all-in ‘ought’s
are relative to a system of higher-order rules which determine the relative
priorities among rules. If conformity to this system requires acting in
accordance with the rule entailing that A X-es in C, then there must be a
first-order rule which entails that A X-es in C. And so there is a feature of X
which explains why this first-order rule entails that A X-es in C (there is a
reason for A to X in C). But since we know (from Oughtfull) that this rule is
highest in the order of priorities (at least higher up than any other rule that
bears on the situation), it follows that there can be no feature such that it
explains why a rule higher or at the same level in the order of priorities
individual reason stronger than another individual reason. He does not explain what makes several reasons
collectively weightier than other reasons.
12
One worry: if explanation is interest-relative, can it be necessary that X has some feature which satisfies the
interest? Can it be ruled out that there will only be dispensable features that explain why the rule entails that
A ought to X? Yes it can, because if otherwise dispensable features are the only ones that play a role in the
explanation, the rules themselves must be concerned with those features. But when that is so, and if one is
interested in learning whether any rules apply to the action, then it is in the relevant sense explanatory to
show that the act has those features.
140
Daan Evers
entails that A not-X-es in C. Given Collective Weight, this means that the
collective weight of the reasons in favour of X is greater than the collective
weight of the reasons in favour of not-X. Therefore the standard-relational
theory of ‘ought’ paired with oughtism about reasons explains why ‘A ought
all-in to X in C’ entails that the balance of reasons favours that A X-es in C.
5. Some Questions and Objections
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
Many questions and objections can be raised about the theory of ‘ought’
and reasons described above. I will attempt to answer some of them in the
rest of this paper.
5.1 Gert on the Balance of Reasons
Joshua Gert [2007] believes that reasons (can) play two distinct roles in
determining the rational status of an action. The first is the requiring role:
reasons can make an act rationally required (i.e. such that one ought to do
it). The second is the justifying role: reasons can make otherwise irrational
acts permissible (i.e. not irrational), without requiring one to do them. For
example, harms to oneself ordinarily have requiring strength (they require
one to prevent them). Altruistic reasons (preventing harms to others)
needn’t have requiring strength, although they can justify incurring harm
oneself. Gert believes two types of strength correspond to both roles:
requiring strength and justifying strength [2007: 537–8].
If Gert is right, then a theory of reasons should be able to make sense of
these two roles. My theory can do this as follows:13 the requiring strength of
a feature (reason) in C is measured by the number of rules that the rule
triggered by it trumps in C. If, in C, a rule is at the same position in the
hierarchy as another rule, I will say that these rules are equalized in C. The
justifying strength of a feature in C can then be thought of in terms of the
number of rules that the rule triggered by the feature equalizes in C.
Gert also claims that requiring strength and justifying strength can come
apart. For example, a reason can justify acts without requiring one to do
them. This is also possible on my theory, since the number of rules trumped
by a rule triggered by a feature may be higher or lower than the number of
rules equalized by that rule.
Gert is sceptical about metaphors like ‘the balance of reasons’ and ‘the
collective weight of reasons’ in so far as they suggest that there is only one
kind of strength. But he does not dispute that reasons with different kinds of
strength still collectively determine what one ought to do. In other words:
what one ought to do is a function of the reasons pro and con. If so, we
should be able to make sense of their collective force or what they require on
balance.
13
The following is an indication of how my theory would deal with these distinctions. It is not meant to
capture all the details. Thanks to an anonymous referee for this journal for forcing me to consider this.
The Standard-Relational Theory of ‘Ought’
141
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
If, in a context, a reason with justifying strength cancels a requirement,
there is a sense in which the reason corresponding to the latter is not
stronger than the former. It is this relation between reasons that Stronger
Than intends to capture. Some such relation should exist if what one ought
to do is a function of the reasons pro and con.
My theory, then, not only makes sense of the notions of requiring
strength and justifying strength, but also tells us how to determine what one
ought to do in the light of different reasons (which may exemplify both types
of strength): one has to check whether any of the reasons trigger rules higher
in the order of priorities than any other rules. If so, then one ought to do the
act which is uniquely favoured by those reasons (if any). This is intended to
capture the idea that if one ought (all-in) to X, then the reasons favour X on
balance.14
5.2 Reasons for Permissible Acts
Someone might object to oughtism as follows: some acts are permissible, but
neither forbidden nor required. Some of those are the acts which the rules
are silent about. But can’t we have reasons for such acts? So perhaps we
should say that reasons are those features of acts which either explain why a
rule entails that the act is (not) done, or why it is neither the case that it is
done nor not done. This is essentially John Broome’s [2004] definition of
reasons.
I think Broome’s definition overgenerates, however.15 Suppose that a
trivial act is an act which has no important consequences whatsoever
(such as rubbing one’s chin). Further suppose that the rules say that
trivial acts are neither required nor forbidden. In that case, the fact that
an act is trivial plays a role in an explanation of why the act is neither
required nor forbidden. But it is not intuitively a reason for or against
the act.
So I prefer my definition of a reason. It’s true that oughtism denies that
one could have reasons for acts which are permissible in virtue of the rules’
silence about them (although it allows for reasons for acts which are
permissible in virtue of a tie between the rules). Is this a problem?
Philosophers who think it is believe that reasons do not always ground
requirements. In other words, they deny the Reason–Ought Link:
14
Here is another way of putting it: Gert does not dispute that acts which are rationally required are such that
one ought to do them. Neither does he dispute that if an act is required then it is impermissible not to do it.
So if X is such that one ought to do it, we can infer at least three things: (1) that there is a set of reasons with
requiring strength favouring X, (2) that there is no set of reasons with equal or more requiring strength
disfavouring X and (3) that there is no set of reasons with justifying strength which makes not-X permissible.
If conditions (1) to (3) are satisfied, then one ought to X, and vice versa: if one ought to X, then conditions (1)
to (3) are satisfied. When I say that the balance of reasons favours X, I mean that these three conditions are
satisfied. A theory of ‘ought’ and reasons should explain why this is entailed by ‘A ought all-in to X’ and vice
versa. The standard-relational theory of ‘ought’ paired with oughtism about reasons enables us to do so,
although my presentation in this paper does not put it in these terms.
15
I owe the example in this paragraph to Ralph Wedgwood.
142
Daan Evers
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
ROL: if one has a reason to X and no reason to do anything else, then one
ought to X.
Oughtism says that reasons for A to X in C are features of X which explain
why a relevant rule entails that A X-es in C. It follows from this view (and
the standard-relational theory of ‘ought’) that if one has a reason to X and
no reasons to do anything else, then one ought to X (this roughly follows
from the fact that in such a situation, there is some rule requiring X and no
rule requiring not-X). So my theories entail ROL.16
One argument against ROL is by Jonathan Dancy [2004]. He claims that I
may have a reason to buy an ice cream (and no reason to do anything else),
but that it’s odd to say that I ought to buy the ice cream. This, I take it, is a
semantic observation. However, as argued by Broome [2004], the impression
that there is something odd about saying that one ought to buy the ice cream
is plausibly due to the impression that ‘ought’ is a very serious word, a word
with moral connotations. But ‘ought’ isn’t intrinsically moral (or even
serious). It may be that you ought to buy a pair of shoes. Some things may
be required to do yourself a favour, even if no one will blame you if you
don’t.
Another argument against ROL derives from Gert [2007]. As we’ve seen,
Gert distinguishes between justifying and requiring strength (of reasons).
Although my theory makes sense of the distinction conceived in terms of
making permissible (justifying) versus grounding ‘ought’s (requiring), it does
not cater for reasons that merely justify (i.e. don’t require in any context,
including ones where only one reason applies).
I don’t think that is clearly wrong, however. The kind of cases that lend
plausibility to Gert’s distinction are cases where an act involves some cost to
onself. Such acts are normally prohibited, but the prohibition can be lifted
by benefits to others. Such benefits make it permissible to incur the cost
without requiring us to do so. But now imagine there to be no cost at all
(which is actually impossible given our limited natures). Is it clear that we
are not required to prevent a harm to others even then? I don’t find that
obvious (in fact I find it wrong). So we cannot be confident that there are
any reasons with only justifying strength.
5.3 Wrong Order of Explanation
It seems unlikely that one could explain the greater weight of a feature in
terms of a higher-order rule which arbitrates between (potentially)
conflicting requirements. Doesn’t this higher-order rule itself depend on
prior insight into the relative significance of the conflicting features?
16
They share this entailment with a popular view which says that if the balance of reasons favours X, then
one ought to X. If the balance of reasons favours X, then X is required and not-X impermissible. This kind of
view can allow for permissible acts for which one has reasons, but only if the balance of reasons does not
favour one particular act. So this view also entails that if one has a reason to X and no reason to do anything
else, then one ought to X. In other words, this view also entails that what reasons do is, most fundamentally,
to require.
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
The Standard-Relational Theory of ‘Ought’
143
But, if so, then I haven’t explained what the weight of a reason consists
in.17
The answer to this objection emerges when we remind ourselves that I am
engaged in a particular project. It is the project of doing normative
semantics on the assumption that there are no objective normative facts.
Given normative antirealism, the normative significance of a feature derives
most fundamentally from the stance taken towards it by a subject (at least in
cases where the identification of the relevant standards proceeds via the
commitments of the subject18). For example, the fact that A cares about
harm to others is what gives such harm normative significance for A. This is
a metaphysical claim: a claim about the generation of normative
significance. It entails that reasons are always (in a sense) derivative:
features of the world matter (count as reasons) only because we care about
some things. But if caring to some degree is metaphysically rock-bottom in
the generation of normative significance, caring to different degrees is
metaphysically rock-bottom in an explanation of relative significance. These
are metaphysical claims. My suggestion is that these phenomena find
semantic expression in claims about higher-order rules and their requirements.
The idea is this: the fact that F1 is weightier than F2 is a reflection of one’s
preference for F1 over F2 in case of conflict (which preference is rockbottom19). This preference explains one’s commitment to a rule which
requires that standards furthering F1 take priority. Statements about
relative weight can then be understood as statements about the priorities
determined by such rules.
5.4 Complicated Rules
The next set of questions I want to discuss concerns the nature of the rules to
which I claim speakers are referring with ‘ought’ statements. The questions I
want to focus on are relevant mainly when the relevant rules are ones to
which the speaker is him/herself committed.
(1) Most plausible rules are rules that hold only ceteris paribus. For
example, it is implausible that it is always bad to lie, and most people
know that. So hardly anyone is committed to the rule ‘Don’t lie!’. Most
people are committed to something like: ‘Don’t lie, other things being
equal!’. But that rule does not entail that one does (not) lie if an act is an
instance of lying. It only does so if other things are equal. However,
spelling out what that amounts to makes the rule impossibly complicated.
And it is implausible that anyone is committed to such a complicated
rule. So a view where ‘ought’ is related to what such a rule entails is
implausible.
17
This objection was pressed by an anonymous referee for this journal.
In cases where this is not so, the higher-order standards are determined by the subject’s beliefs about the
standards in force. These beliefs may themselves be based on information about what tends to take priority
(e.g. in court rulings or etiquette).
19
Or at least some such preferences will be rock-bottom for antirealists. Others may be derivative (depend on
preferences which are themselves rock-bottom).
18
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
144
Daan Evers
This type of objection is based on the observation that most people
allow exceptions to rules. However, in order to allow for that, it is not
necessary to build the exceptions into the content of the rules themselves.
It suffices if we have a structure which explains why certain rules
(which can themselves be simple) sometimes are to be heeded and
sometimes not. And I have offered just such a structure: a rule wins out in
a context just in case it is (in that context) higher in the order of priorities
than any other (relevant) rule with inconsistent implications. In other
contexts it may be lower than another rule, in which case we allow for an
exception.
(2) My answer to the first question naturally raises the second: does my
view not commit me to the claim that people tend to be committed to
impossibly complicated structures of first-order and higher-order rules?
Even if the first-order rules are simple, the second-order ones are not. There
are indefinitely many situations in which different rules are higher up the
order of priorities. But is anyone committed to higher-order rules which
legislate for all such situations in advance? It seems that such a rule would
have to be very complicated: ‘Let the moral system prevail over the
prudential system, except if a, or b, or c, or . . .’.
I agree that if the higher-order rules had to legislate for all possible
situations in advance, then they would have to be very complicated. But
there is no reason why this should be so. Take (again) a case where a moral
rule (or system of rules) which requires A to X conflicts with a prudential
rule which requires A to not-X. Now imagine that a speaker makes the
judgment that A ought, all things considered, to X (i.e. respect the moral
requirement). It suffices for this judgment to be true that the speaker is, at
the time of speaking, committed to a rule (subscribes to an imperative)
which entails that the moral rule prevail in this context. But the rule needn’t
say anything more general than that.
5.5 Normativity
Another problem for the theory is that of normativity [Boghossian 2006:
24]: saying that X is required by some system does not seem normative. But
‘A ought to X’ does.
In order to respond to this problem adequately, we need to know what is
meant by ‘normative’. What makes a statement normative? A plausible and
popular answer is that a statement is normative if and only if it entails
something about reasons (in this case, ‘A ought to X’ is normative if and
only if it entails that there are (sufficient) reasons for A to X). But on my
theory, ‘ought’ statements are normative in this sense. For I have argued
that the standard-relational theory and oughtism about reasons explain why
‘A ought to X’ entails not only that there are reasons for A to X, but also
that the balance of reasons favours X-ing.
If there is a problem of normativity for my theory, it must stem from
something else. Some philosophers believe that normative statements
cannot be reduced to non-normative statements. But what is meant by
The Standard-Relational Theory of ‘Ought’
145
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
‘non-normative statement’? Presumably, a successful analysis of ‘ought’
statements in terms of different vocabulary makes that vocabulary itself
normative, at least as used in the analysis. This must be so if the ‘ought’
statement is itself normative and the analysis successful.20
However, anti-reductionists think that no analysis of claims involving
words like ‘ought’ and/or ‘reasons’ can work precisely because the
meaning of these claims has a feature, normativity, which is lacked by the
meaning of claims couched in other vocabulary. However, such claims are
question-begging unless the feature can be identified in some way other
than by saying that the analysis does not contain words like ‘ought’ and
‘reason’. So I believe that, in this context, the burden of proof is on the
opposition.
5.6 Disagreement
The last problem I will discuss is that of disagreement. The standardrelational theory does not always allow inconsistent propositions to be
expressed by speakers who respectively say ‘A ought to X’ and ‘A ought not
to X’. If the first statement refers to a system of rules accepted by one
speaker, and the second to a different system accepted by the other, they are
not inconsistent. The speakers merely state facts about the requirements of
different systems. This problem is sometimes called the problem of
disagreement: on the face of it, ‘A ought to X’ and ‘A ought not to X’
are incompatible. But the kind of view I have sketched does not always
make it so.
Notice that my view does allow some genuine disagreement. Because I
don’t use possessive pronouns in representing the meaning of ‘ought’
statements, two speakers referring to the same system genuinely disagree if
the one says that A ought to X and the other denies it.21 But even in cases
where speakers do not express inconsistent propositions, it is not hard to see
(1) how the appearance of inconsistency may arise or (2) why the (verbal)
disagreement does not cease.
The appearance of inconsistency may arise from the supposition that
one’s interlocutor is referring to similar rules as oneself. It is hard to be
certain that people do not share our values because normative disagreement
is often explained by differences in non-normative beliefs. There is much
overlap in values [Finlay 2008: 356]. But even if we are certain of a difference
in rules, we can still explain why we might insist that our interlocutor is
wrong. Björnsson and Finlay [unpublished ms] plausibly claim that in
(some) normative matters our interest in truth is secondary to our interest in
the promotion or protection of values. This explains why, in normative
debates, the proposition assessed for truth or falsity is not always the one
expressed by my interlocutor. It is sometimes one closely related to it. When
20
I am assuming that the reduction in question is analytic. Thanks to an anonymous referee for this journal.
The same is true if speakers subscribe to many rules different to those accepted by the other, but (1) only a
subset of the rules apply to the case and (2) these rules have the same content. In such cases, it makes sense to
construe the speakers’ statements as involving reference to the same subsystem of rules.
21
146
Daan Evers
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
John says ‘A ought morally to X’ I don’t check whether X-ing is required by
John’s standards (even if the truth conditions of the statement do involve
them). Rather, I check whether it is required by mine. The explanation for
this can be found in the point of moral discourse, which is not entirely
theoretical:
Part of what it is to engage in a practice of moral discourse is to subscribe to a
particular standard, to the exclusion of any rival, as determinative of what to
do—both for one’s own conduct and others’. Here too there is a context which
for us is privileged, consisting in our own standards. The truth of oughtpropositions relativized to other standards is irrelevant to this conversational
interest. To address the question of whether [John] was right to think that
[X-ing] is required [of A] by [John’s] standard . . . would therefore be a
perverse fixation on truth for truth’s sake, in neglect of what is relevant to the
purpose of moral thought and discourse.
[Björnsson and Finlay, unpublished ms]
6. Conclusion
Many more objections ought to be addressed, but space does not permit. I
have outlined a standard-relational theory of ‘ought’ and an oughtistic
theory of reasons. The combination of these theories meets the explanatory
challenges stated at the start of this paper. The theory of ‘ought’ allows
many normative judgments to be true even if there are no objective
normative facts. Oughtism about reasons explains why ‘A ought to X’
judgments entail that there are reasons for A to X. Furthermore, it explains
why ‘A ought all-in to X’ entails that the balance of reasons favours X-ing. I
take the ability of the theories to explain these (often neglected) phenomena
to be a major point in their favour. It lends further plausibility to the ancient
idea that normative judgments are relative to standards.22
Received: June 2009
Revised: December 2009
University of Oxford
REFERENCES
Björnsson, G. and S. Finlay unpublished ms. Metaethical Contextualism Defended, URL¼5http://wwwrcf.usc.edu/*finlay/MCD.pdf4.
Boghossian, P. 2006. What is Relativism?, in Truth and Realism, ed. P. Greenough and M. Lynch, Oxford:
Clarendon Press: 13–37.
Broome, J. 2004. Reasons, in Reason and Value: Essays on the Moral Philosophy of Joseph Raz, ed. R. Jay
Wallace, M. Smith, S. Scheffler and P. Pettit, Oxford: Oxford University Press: 28–55.
Dancy, J. 2004. Enticing Reasons, in Reason and Value: Essays on the Moral Philosophy of Joseph Raz, ed.
R. Jay Wallace, M. Smith, S. Scheffler and P. Pettit, Oxford: Oxford University Press: 91–118.
22
I would especially like to thank Ralph Wedgwood for extremely helpful comments and suggestions. I have
also benefited from commentary by Natalja Deng, Barbara Vetter, Stephen Finlay, Herman Philipse and
anonymous referees for this journal. Thank you all.
The Standard-Relational Theory of ‘Ought’
147
Downloaded By: [University of Oxford] At: 14:04 21 January 2011
Finlay, S. 2008. The Error in the Error Theory, Australasian Journal of Philosophy 86/3: 347–69.
Finlay, S. 2009. Oughts and Ends, Philosophical Studies 143/3: 315–40.
Gert, J. 2005. A Functional Role Analysis of Reasons, Philosophical Studies 124/3: 353–78.
Gert, J. 2007. Normative Strength and the Balance of Reasons, The Philosophical Review 116/4: 533–62.
Harman, G. 1975. Moral Relativism Defended, The Philosophical Review 84/1: 3–22.
Harman, G. and J. Thomson 1996. Moral Relativism and Moral Objectivity, Oxford: Blackwell.
Horty, J. 2007. Reasons as Defaults, Philosophers’ Imprint 7/3: 1–28.
Mackie, J. 1977. Ethics: Inventing Right and Wrong, Harmondsworth: Penguin.
Wong, D. 1984. Moral Relativity, Berkeley: University of California Press.