Normative Strength and the Balance of Reasons

Normative Strength and the Balance of Reasons*

Joshua Gert

“Normative Strength and the Balance of Reasons,” Philosophical Review, Vol. 116, No. 4

(2007), pp. 533-562. Please cite published version.

1. Introduction

The concept of normative practical reasons has become one of the most basic and central

topics in contemporary ethical theory.1 Such reasons are normative in that they either favor, or

disfavor, the various options to which they are relevant, in the sense of being directly relevant to

what the agent rationally ought or ought not do. Since many actions can be assessed as rational

or irrational without being morally significant in any way, it should be clear that assessments of

practical rationality, and claims about normative practical reasons, are distinct from claims about

moral status. Indeed, one extremely popular approach to moral theory takes the notions of

reasons and rationality as basic and then goes on to try to explain and justify morality in terms of

these notions.2 This paper is concerned with these prior notions, and not directly with morality.

Normative reasons need to be distinguished from motivating or explanatory reasons, for it

is possible for there to be normative reasons that do not move an agent in any way, even though

they are relevant to his choice of action. This might happen because the agent is unaware of

those reasons, or because the agent is irrational, or because some normative reasons are in some

sense optional, so that it would be equally rational to act on them, or to ignore them altogether.

And it is possible for there to be “reasons why” an agent acted as he did (for revenge, say, or

jealousy) that do not correspond to normative reasons: not all motives are rational motives. This

paper has to do only with normative reasons.

My aim in this paper is to call attention to a key feature of normative reasons: namely,

that they can play two quite different roles in determining overall rational status. I will later call

these roles ‘justifying’ and ‘requiring’. Given my definitions of these terms, the thesis that

2

normative reasons play these two roles should be uncontroversial. What philosophers have not

appreciated, however, is that as a consequence of the existence of these two roles, any account of

the normative strength of reasons will need to distinguish two dimensions of normative strength,

corresponding to each of these roles. For two reasons with the same justifying strength might

have very different requiring strengths. I will also argue that because reasons have these two

distinct dimensions of normative strength, the common view that the (most) rational action is the

one favored by the strongest reasons is completely ill conceived.

Most contemporary ethical theorists who make much use of the notion of a practical

reason seem to rely on the fact that some reasons are stronger than others, for phrases such as ‘the

stronger reason’, ‘most reason’ and ‘the balance of reasons’ are ubiquitous. It is certainly hard to

deny the plausibility of the idea that normative reasons have relative strengths. The fact that an

action will cause me to die a horrible fiery death surely provides a stronger reason against it than

does the fact that it will prevent me from seeing a movie that I have been looking forward to

seeing. But despite the apparently straightforward truth of this particular claim, if reasons have

the two dimensions of normative strength this paper claims that they do, then we ought to give up

using phrases such as ‘the stronger reasons’, at least in an unqualified way. For this phrase, and

others, conceal the assumption that any given reason has a single strength value, and they also

encourage the false view that rationality is best understood in a maximizing or satisficing way.

But when one appreciates the nature of the two kinds of normative strength, it will become clear

that maximizing is not really a coherent goal, that the general advice ‘Act on the stronger

reasons’ is typically quite useless and confused, and that phrases such as ‘the balance of reasons’

or ‘what there is most reason to do’, even taken as metaphorical, are so misleading that they

ought never to be used.

A recognition of the existence of two dimensions of normative strength is almost an

immediate consequence of the recognition of the two roles that such reasons play. How is it that

these roles have not generally been recognized? Philosophers have certainly asked themselves

about the role that normative reasons play in determining rational status. The problem is that

3

they generally have given only the most abstract and formal of answers: ‘They count in favor of

actions’, ‘They support ought-claims’.3 These answers may be true as far as they go, but they do

nothing to indicate what ‘counting in favor of’ or ‘supporting’ amount to, and they certainly

make it extremely unlikely that different ways of counting in favor or supporting will be

distinguished. Consequently, they do not allow for a clear view of the nature of normative

strength. Because of this most philosophers who make use of the notion of ‘the strength of a

reason’ have only been able to use that notion in a metaphorical way. The natural metaphor here

is with the strength of a physical force such as weight. As John Broome puts it, “Weighing is

just what reasons are made for.”4 As a result, these philosophers have thought that in contests

between two opposed reasons, they could discern the stronger reason as the one that rationally

compels action and thereby “wins,” and have supposed that they could define rational action as

‘the one supported by a preponderance of reasons’.5 The idea that practical reasons have

univocal strength values may also partly explain the temptation of views that associate such

reasons with idealized desires, since desires at least plausibly have univocal strengths.6 But the

strength of a reason is not a physical or a psychological matter. Therefore we should not think of

normative reasons as pushing and pulling an object (here, a rational agent) in one direction or

another. Rather, we should think of normative reasons as considerations that play certain

systematic roles in determining the rational status of actions. And since these roles are neither

physical nor psychological, they are limited only by the form that principles of rationality can

take. That is, if the fundamental principles of rationality allow reasons to contribute to the

rational status of actions in two, three, or four distinct ways, then there may well be two, three, or

four distinct dimensions along which we can specify a given reason’s various normative strength

values. For example, a principle of rationality may entail that a certain class of reasons serves

only to mitigate (sometimes completely) the irrationality of certain actions, in the same sort of

way that some principles of morality entail that a certain class of considerations (say,

considerations of self-defense) can serve to mitigate (sometimes completely) the immorality of

harming others. In that case, some reasons will be stronger than others as mitigators of

4

irrationality. But this will not entail anything about their capacity to render actions rationally

required. And yet some reasons surely serve to require actions, and can be compared along that

dimension, yielding a notion of strength in generating rational requirements.

To repeat: if reasons can play more than one distinct role in determining the rational

status of an action, then more than one dimension of strength may well be required to

characterize the normative capacities of any given reason. As will be explained in the next

section, there are – clearly and uncontroversially – two such roles. This would not matter a great

deal for those who wish to associate a single strength value with any given reason, if strength in

one role were always correlated with strength in the other. But unless we are given some reason

to believe in such a universal correlation, it seems completely unjustified to theorize as if it were

a given. This, however, is what virtually all of those who theorize about practical reasons do,

since they virtually all write as if one strength value were sufficient to characterize the normative

capacities of any given practical reason. The view is so ubiquitous that it is typically not

explicitly expressed or endorsed, but Joseph Raz put it fairly clearly when he wrote that “[a]ll

reasons are comparable with regard to strength […and…] this is their only feature relevant to the

outcome of practical inferences.”7 And of course the view that rationality is a matter of

maximizing the satisfaction of our considered preferences also suggests this view, as long as one

associates a reason in a natural way with each considered preference. But even when the view is

not articulated as clearly as this, it remains implicit in the use of maximizing phrases such as ‘the

strongest reasons’ or ‘what we have most reason to do’. Anyone who is persuaded by this paper

should see that these phrases, which one encounters everywhere one looks in the literature on

practical reasons and rationality, conceal a very significant assumption that is almost certainly

false.

2. Two Roles

What do normative reasons for action do? One thing they do is explain why actions that

would otherwise be rationally permissible are in fact irrational. For example, in normal

5

circumstances it is rationally permissible for me to have my usual cup of coffee in the morning.

But if I am suffering from a stomach ailment and the doctor has informed me that coffee in the

morning will make me very uncomfortable in the afternoon, then I have a reason not to have the

coffee. The reason against having the coffee – that it will make me quite uncomfortable in a few

hours – explains why having the coffee is irrational in this case, rather than rationally

permissible.8 This is an illustration of a reason playing what we might call ‘the irrational-making

role’, or ‘the requiring role’, since to make an action irrational is to require, rationally, that we

not perform it. It is uncontroversial that reasons play this role. Corresponding to this role is one

kind of normative strength, which we can call ‘requiring strength’. One reason, A, has more

requiring strength than another reason, B, if, in playing the requiring role in actual and

counterfactual circumstances, A can overcome any reason or set of reasons that B can overcome,

and there are some reasons or sets of reasons that A can overcome but B cannot. For example,

the fact that an action will save my life has more requiring strength than the fact that an action

will save me from a mild hangover. For despite there being the normal social and hedonic

reasons in favor of (say) having another drink, and despite the fact that it would otherwise be

rationally permissible to have that drink, I am rationally required to refuse the drink it if having it

would cost me my life. But I am not rationally required to refuse the drink if having it would

only end up giving me a mild hangover. On the other hand, if I am rationally required to act

against a certain set of reasons simply because otherwise I will suffer a mild hangover, I would

certainly be required to act against that same set of reasons if that were what I needed to do to

save my life. Let us call this criterion of relative requiring strength ‘the counterfactual criterion’.

Later, in section 5, I will argue that anyone willing to speak about the strength of reasons is

committed to some such criterion.

Another thing that normative reasons do is explain why actions that would otherwise be

irrational are in fact rationally permissible.9 For example, in normal circumstances it would be

irrational to rush out into a busy street while traffic was still flowing. But if one saw a small

child in the road who was in immediate danger, then it would be rationally permissible to rush

6

into the street to rescue the child. This is an illustration of a reason playing what we can call ‘the

rational-making role’, or ‘the justifying role’, since to justify an action (in a common sense, and

in the sense relevant to this paper) is to show that it would not be irrational to perform it, despite

the existence of considerations that speak against it. It is uncontroversial that reasons play this

role. Corresponding to this role is another kind of normative strength, which we can call

‘justifying strength’. One reason, A, has more justifying strength than another reason, B, if, in

playing the justifying role in actual and counterfactual circumstances, A can overcome any

reason or set of reasons that B can overcome, and there are some reasons or sets of reasons that A

can overcome but B cannot. For example, the fact that my action will save an innocent person

from going to prison has more justifying strength than that my action will save me from two

nights of insomnia. For despite there being the obvious reasons against (say) making myself the

potential target of some dangerous criminals – reasons which would otherwise make the action

irrational – it would be rationally permissible make myself such a target in order to save an

innocent stranger from going to prison. But it would not be rationally permissible to expose

myself to the same danger merely to spare myself two nights of insomnia. On the other hand,

any set of reasons that I would be justified in acting against, simply in order to avoid two nights

of insomnia, I would also be justified in acting against in order save an innocent person from

going to prison.

One useful way of thinking about the justifying role of practical reasons, and

distinguishing it from the requiring role, is by analogy with justifying considerations in the moral

domain. Considerations of self-defense, for example, can be compared in terms of strength; that

is, they can be compared in terms of how much they can morally justify. Such considerations are

directly relevant to the moral status of actions. And yet it is clear that their role is logically

distinct from the role of another sort of moral consideration: the sort of consideration that

morally requires one to act, or to refrain from action. Indeed, it is plausible to regard

considerations of self-defense as completely lacking in any power to generate moral

requirements. And this is true despite the fact that considerations of self-defense can be, in an

7

obvious and intuitive sense, extremely strong: such considerations might sometimes justify

killing someone, or hurting them in a way that would be grievously immoral in the absence of

such a justification. It is one thesis of this paper that, within the domain of practical rationality

too, it is perfectly possible – indeed quite likely – that some reasons might have only, or

primarily, justifying strength.10 If this is right, then in order to know the normative capacities of

some particular reason – say, that an action will save an innocent stranger from going to prison –

we must know both how much it can justify, and also how much it can require: its justifying

strength and its requiring strength. This means that we cannot simply talk about “the” strength of

a reason. Nor, as will be made clear in what follows, can we talk about the action, or class of

actions, supported by the best, most, or strongest reasons.11

Some remarks on the terminology might be in order here. Requirements are an all-or-

nothing affair, so it may seem strange to talk about requiring strength. And if the requiring

strength of a reason could only be understood as indicating the strength of the requirement it

yielded, then the terminology might not make a great deal of sense. But it does make sense to

ask, of a given reason, how much it can require of one: can it require me to make this very great

sacrifice, or only this rather smaller one?12 And if it makes sense to ask how much a reason can

require, then it makes sense to talk about its requiring strength. Similarly, it makes sense to ask

how much a given reason can justify: this much, or only that much? So it also makes sense to

talk about justifying strength, even though the question of whether an action is justified is also

plausibly regarded as an all-or-nothing affair.

It may seem that the justifying role is merely the mirror image of the requiring role.

Certainly many reasons (say, the reason provided by the fact that the building is on fire) can be

described in either positive or negative terms: as favoring actions in one class (getting out of the

building), or as disfavoring actions in the complementary class (remaining in the building).

Since the requiring role was initially defined in negative terms (making actions irrational despite

positive reasons), and the justifying in positive terms (making actions rationally permissible

despite negative reasons), one might suspect that the distinction between them marks nothing but

8

a difference in terminology, and that the two roles are nothing but grammatically distinct

formulations of the very same underlying role. But this is not correct.

In fact, it is possible to draw the justifying/requiring distinction in terms of a certain kind

of positive/negative distinction, but not the philosophically uninteresting one just described.

Patricia Greenspan, in recent work, has done something very much like this, classifying certain

reasons as negative in virtue of their providing a basis for criticism, and other reasons as positive

in virtue of their providing an answer to such criticism.13 That is, negative reasons function to

rule out actions – or at least to form the basis for an argument that those actions are ruled out –

while positive reasons serve to answer such arguments. To take one of Greenspan’s examples,

one positive reason for serving on university committees is that it gives one some degree of

power. But Greenspan holds that the normative function of this reason is best seen as that of

providing an answer to potential criticism: for example, the criticism that serving on committees

is a boring waste of time. That it serves this function does not imply that one need be moved in

any degree by this reason. One can be indifferent to the potential for this sort of power, even

when one can obtain it at little or no cost, and one’s indifference need not make one liable to any

sort of criticism. It should be clear that if we restrict our view of the relevant form of criticism to

criticism given in terms of rationality, then Greenspan’s distinction between negative and

positive reasons corresponds to the distinction between the two roles given above. But this will

be of little use to those who wish to think of requiring strength and justifying strength simply as

alternate descriptions of the same underlying capacity. For that thesis to be true, it must be the

case that that any particular positive reason can also be described as a negative reason, although

of course for a distinct class of actions. But this is not true on Greenspan’s understanding of the

positive/negative distinction. That a consideration provides a positive reason for -ing, on

Greenspan’s understanding of that notion, does not imply that it can be thought of as providing a

negative reason against failure to . For a consideration can provide an answer to criticism for

-ing without being able to form the basis of any criticism of its own.

9

That the distinction between the justifying and requiring roles cannot be understood

merely as a superficial grammatical distinction will become even clearer below, when it is shown

that reasons with the same justifying strength might have distinct requiring strengths. Of course,

and as Greenspan’s position shows, it is not necessary to think of the positive/negative

distinction as a distinction that is merely grammatical. Jonathan Dancy, too, makes a distinction

between favoring and disfavoring, and explicitly allows that to favor a certain action may not be

to disfavor performing the only alternative action. So for Dancy, also, the positive/negative

distinction cannot be conceived in a way that would make positive and negative strength mere

terminological variants.14 Both Greenspan and Dancy, then, allow practical reasons to play two

logically distinct roles.

That there are two roles for reasons should not be taken to imply that there are two

distinct types of reasons: those that require, and those that justify. Indeed, such a suggestion is

false, for any reason that can require me not to act on contrary reasons also justifies me in acting

against those reasons, and therefore has both requiring and justifying strength. For example, let

us say that it would be irrational to refuse to suffer a certain amount of pain in order to prevent

the loss of one’s arm. If this is right, then the prospect of saving one’s arm is a reason with some

requiring strength: it can rationally require one to suffer the pain. But since this very same

reason also explains why an action that would otherwise be irrational – voluntarily suffering pain

– is not irrational, it is also true that the prospect of saving one’s arm is a reason with some

justifying strength. But although in many cases the very same reason that justifies an action may

also require it, and though a reason that requires an action also justifies acting against any

reasons that may oppose it, it still makes sense to separate the requiring and justifying roles, and

to distinguish requiring from justifying strength. For even in the case in which the same reason

that justifies an action also requires it, it may well be true that it could justify acting against still

more reasons, but could not require doing so.

10

3. Two Distinct Strength Values

We have just seen that there are two ways in which one might understand the normative

strength of a reason: as requiring strength, or as justifying strength. Which understanding should

one choose? It may seem a matter of indifference. For, one might object, even granting that the

justifying and requiring roles are somehow distinct, surely if a reason can justify more, it can

require more, and if it can require more, it can justify more. The answer to this objection is that

while this might be true, there is no reason to suppose that it must be true, or indeed that it even is

true. It is plausible, for example, that the fact that one’s action will save a handful of strangers

from malnutrition can rationally justify much more than can the fact that one’s action will save

one a hundred dollars. And yet we do not seem to be rationally required to do much to help

prevent malnutrition, while we do seem to need some non-trivial reason to part with a hundred

dollars if our action is not to be considered irrational. To put the point in Greenspan’s terms,

why should we think that a consideration that provides a very good answer to criticism must also

be a consideration that would form the basis for a very strong criticism, should one fail to act on

it?15 Certainly this is not the case in the moral domain, as the example of self-defense shows.

Why then should it be true in the domain of rationality? Similarly, Dancy takes the favoring and

disfavoring relations to be distinct. And because he is a particularist, he also holds that the

degrees to which reasons bear these two relations to various actions are matters of brute fact, not

amenable to systematic explanation: certainly strength in one role is not to be explained in terms

of strength in the other. In his framework, then, a constant correlation between strength in the

two roles would be an incredible cosmic coincidence. In this way, Dancy’s particularism actually

provides the basis for an argument in favor of the significance of the distinction between

justifying and requiring strength, for the distinction is significant if we cannot expect justifying

and requiring strength always to covary. Of course, there is another way in which particularism

might seem to argue against the attempt to assign two stable strength values. After all,

particularists deny even single stable strength values to practical reasons. I will address this

objection in the final substantive section of the paper, which also offers what I believe to be a

11

novel criticism of particularism: that without the notion of stable strength values for practical

reasons – values they keep from context to context – there does not seem to be any coherent

notion of the strength of a reason even in a particular context.16 But let us put this issue aside for

the moment.

Actual examples very strongly suggest that the strengths of reasons in the justifying role

and the requiring role do not always covary. I have argued elsewhere that what seems to unify

our judgments of irrationality is something like the following principle:

P: An action is irrational iff it will bring some harm to the agent and will not bring

any compensating benefit to anyone (including, but not limited to, the agent).17

This principle allows varying degrees of altruism to be equally rationally permissible, all the way

from an almost selfless devotion to others, to a selfish devotion to one’s own pleasure. Principle

P allows us to say that those who commit harmful crimes for profit are as rational as the next

person, although of course quite immoral, and this allows us to avoid the typical problems with

moral responsibility that infect views that imply that all immoral action is irrational, and that

severely immoral action is severely irrational. But P also allows us to say that the self-

destructive vengeance of a jealous husband, or the compulsive action of the addicted gambler, or

the shortsighted behavior of someone who does not follow an unpleasant but necessary course of

medical treatment, are all irrational. And it also allows us to say that self-destructive behavior

for no reason at all, as one finds in certain psychotic people, is irrational even if such action is in

line with their strongest preferences.

Admittedly, there are problems with P as it is presented here, in a very crude form.

Moreover, there is an ambiguity in the term ‘rationality’ which may make P seem completely

inadequate. This is the ambiguity between a sense of ‘rationality’ that has a very close

connection to proper mental functioning – let us call it ‘subjective rationality’ – and a distinct

sense according to which the rationality of an action is a matter of the reasons that favor and

12

disfavor it, whether or not an agent is aware of those reasons. This latter sense, which we might

call ‘objective rationality’, is one in which an agent would do research in order to find out what

his rational options might be in a given situation. P should clearly be taken as an account of

objective rationality. Moreover, it is objective rationality that is the more central notion in

contemporary ethical theory, subjective rationality typically being defined in terms of it. Despite

the existence of these two senses of ‘rationality’, it is agreed on all sides that there is a close

conceptual connection between the two, and that people who are informed of all the relevant

reasons are acting in a subjectively irrational way if they perform actions that are objectively

irrational. This explains why P can yield intuitively plausible results about the status of self-

destructive actions performed by psychotic people, even though it is not, directly, an account of

subjective rationality.

Despite all that can be said in defense of the adequacy of a principle of rationality very

close to principle P, it is important to stress that the actual adequacy of P is of no significance for

current purposes. In particular, the rather consequentialist (though non-maximizing) flavor of P

is incidental to the point that it illustrates: in the statement of P, the term ‘harms’ could be

replaced with the less euphonious ‘irrational-making features’ and the term ‘benefits’ could be

replaced with ‘rationally eligible-making features’. What is important is that the existence of P

shows that there can be coherent principles that yield a consistent set of judgments regarding the

rationality of action, but according to which justifying and requiring strength do not covary. For

according to P some reasons (altruistic ones) can have significant justifying strength without

having any requiring strength at all. That is, suppose we have an action that is rationally

permissible. If we add to the description of the action the fact that there is an additional reason –

an altruistic reason – against it, this by itself can never, according to P, change the status of the

action from rationally permissible to irrational.18 Equivalently, if an action is not rationally

required, then, according to P, the addition of altruistic reasons that favor it can never make it

rationally required. On the other hand, P claims that altruistic reasons can change the status of an

13

action from irrational to rationally permissible, even when the reasons against the action are very

significant.

P does not entail that altruistic reasons are weaker than self-interested ones. For

according to P the following two reasons have the same strength in the justifying role:

A: that the action will spare the agent a certain harm, H

B: that the action will spare someone other than the agent the same harm, H.

This is sufficient to show that two dimensions of normative strength are required to describe the

normative capacities of reasons, if something formally (but not necessarily substantively) similar

P is the right account of practical rationality. For according to P, A and B have the same strength

in the justifying role, yet B has no strength in the requiring role. I am aware that many practical

reasons theorists would not want to draw such a stark distinction between self-interested and

altruistic reasons. They might therefore want to defend a principle according to which altruistic

reasons had some significant requiring strength. But as long as we can reasonably hold that two

reasons with the same justifying strength might have different requiring strengths, or vice versa,

we will need to distinguish the strength of a reason in the requiring role from its strength in the

justifying role. Consider again reasons A and B above. It is extremely plausible that they have

the same justifying strength: that is, that any sacrifice that would be justified by the fact that it

would spare the agent a certain harm, H, would also be justified if it would spare someone else

that same harm. But even if we allow B to have quite significant requiring strength, we will still

need to distinguish requiring and justifying strength as long as we hold that one cannot be

rationally required to take as much trouble to spare other people from harm as we are to spare

ourselves. And this seems extremely plausible as well. After all, without the introduction of

additional reasons into the story, it seems irrational to refuse to suffer a mildly serious harm now

if that would prevent one’s suffering twice the harm a short while later. But we do not seem

14

rationally required to prevent other people’s suffering whenever we can do so by suffering half

the amount ourselves, even if we would be rationally justified in doing so.

The mere possibility of a principle such as P, or a version of P that gives altruistic reasons

some non-negligible requiring strength, shows why it is important to ask what roles normative

reasons can play. For it is by understanding what roles reasons play that we can understand what

it means to say that one reason is stronger than another. If something like P should turn out to be

correct, then there really are two quite distinct roles, and therefore two quite distinct kinds of

normative strength that any given reason will have. It should not surprise us that normative

reasons have more than one dimension of strength, since they clearly and uncontroversially

perform more than one role in determining the rational status of actions: requiring and justifying.

Since reasons play both the requiring role and the justifying role, philosophers who are willing to

talk about the strength of reasons at all should be willing to talk about the justifying strength and

the requiring strength. And there is no reason to assume that two reasons with the same strength

in one role must necessarily have the same strength in the other. The mere existence of a

principle such as P, quite independently of its correctness as an account of practical rationality,

shows that justifying strength and requiring strength are conceptually distinct, so that it makes

sense to ask if they might come apart in actual cases. And consideration of actual cases strongly

suggests that they do.

Again, it is worth stressing that the argument of this paper is not dependent in any way on

the existence of the sort of basic distinction between self-interested and altruistic reasons that

principle P happened, for illustrative purposes, to make use of.19 An account of practical

rationality need not make any distinction between self-interested and altruistic reasons in order to

require two dimensions of normative strength. For example, one might very plausibly hold that

reasons involving harms and reasons involving benefits are usefully differentiated in terms of the

two dimensions of normative strength. On such a view, reasons involving the getting of benefits

may have a great deal of strength in the justifying role (making it rational to risk one’s life for a

very great pleasure), while not having much strength in the requiring role (so that it would not be

15

irrational to forego that very same pleasure because of some small inconveniences required to get

it). Or one might hold that reasons lose strength in the requiring role as they concern more and

more remote future times, even though they retain an undiminished strength in the justifying role.

Thus it would be rationally permissible to suffer a great deal at the present in order to gain

security for one’s retirement, but it would not be irrational to spend one’s money now. I have

argued at length elsewhere that the distinction between self-interested and altruistic

considerations is relevant to the justifying and requiring strengths of practical reasons.20 But the

point of this paper is emphatically not to offer further arguments in support of this point: it is

only to show that philosophers ought to recognize that there are at least two dimensions of

normative strength.

4. Why There is No Balance of Reasons

If the above is right then we should regard it as a misleading metaphor to talk about the

balance of reasons. For on a balance, if one side has sufficient weight to make the other side go

up, it also makes its own side go down. This is a consequence of the physical nature of balances

and weights. In the normative domain, this metaphor suggests that if reason R is sufficiently

weighty to change an irrational action into a rationally optional one (that is, into an action that is

rationally permissible but not required), this is because it has just managed to balance out the

countervailing reasons: the reasons in favor of the action and against it are roughly equal. If this

were the correct explanation, then if R were weightier to any non-trivial degree, it would make it

rationally required to perform the action it favored. But this sort of claim is often false. Indeed,

the falsity of such claims is one of the central phenomena exploited by those who argue for a

widespread incommensurability of reasons.21 Philosophers who offer this kind of argument for

incommensurability clearly are assuming that the balance metaphor would be apt if reasons could

indeed be assigned univocal strength values, and as far as I know no opponent of

incommensurability has challenged this assumption. But principles of rationality that have the

form of P can explain the phenomena that seem to support incommensurability without leading

16

to the obscurities that surround that notion. According to P, it is rationally permissible to make

certain painful sacrifices in order to alleviate a certain amount of the suffering of others. And it

is the alleviation of the suffering of others that makes it rationally permissible to make those

painful sacrifices: that justifies making them. But this does not mean that one would be

rationally required to make such sacrifices, even if the benefits for others were significantly

increased.

Since reasons have both justifying and requiring strength, it also makes little sense to use

maximizing phrases such as ‘the action that is supported by the best or strongest reasons’, or

‘what we have most reason to do’. Current writing about the rationality of action is filled with

such language. Indeed, rational action is often simply defined as action that is supported by the

best, strongest, or most reasons.22 But given the distinction between requiring and justifying

strength, it should be recognized, as it generally is not, that such maximizing phrases, even taken

metaphorically, are extremely problematic.23 An example will help illustrate this. Assume for

the moment that P is correct, and consider a situation in which an agent must choose between

avoiding some small harm for himself, and preventing some more significant harm for some

strangers. Suppose, to make things more concrete, that on the way to the pharmacists to buy

some medicine for a very bad headache that he is currently suffering, he is asked by a stranger,

whom he has no reason to suspect of fraud, to donate some money to a worthy cause. Suppose

further that the agent knows that the money will prevent suffering that is clearly more significant

than his headache and that he knows his own psychology well enough to know that if he does not

give the money in response to the personal request, he is not going to make an unsolicited

donation later. The reason in favor of spending the money on headache medicine is the

following: that it will spare the agent a very bad headache. Call this reason A. The reason in

favor of donating the money to charity is the following: that it will prevent more significant

harms for others. Call this reason B. Reason A has more requiring strength than does reason B

for the simple reason that, according to P, self-interested reasons have requiring strength, but

altruistic ones do not. Although we are merely stipulating this to be true in order to illustrate a

17

formal point, it should nevertheless seem plausible, since it does not seem irrational to refuse to

give money to charity “just because one doesn’t feel like doing it.”24 Now, despite the fact that

reason B has less requiring strength than does reason A, it has more justifying strength than does

reason A. For the example stipulates that the harms prevented by the donation would be more

significant than even a very bad headache, and, according to P, the justifying strength of a reason

is a matter only of the severity of the harms avoided, and not of whether those harms would

befall the agent or other people. This also should seem at least plausible, for it amounts to the

claim that if it would be rationally permissible to suffer a lesser harm now to avoid a greater

harm later, it would also be rationally permissible to suffer the same lesser harm in order to help

some other person avoid the greater. So reason A has more requiring strength than reason B, but

reason B has more justifying strength than reason A. Which action, then, should we say is

“supported by the best, strongest, or most reasons?” The question is ill-posed. Rather, we

should say that according to P either option is rationally permissible, since the altruistic reason

can justify acting against the self-interested one, but cannot require one to do so, and vice versa.

And since this can remain true even if we increase (up to a point) the severity of the headache, or

the benefits that the charity provides, we cannot consider the original case to be one in which the

reasons are simply “balanced,” or in which each option was supported by reasons of equal

strength. Although it is implicit in the example, it may be worth making it explicit that we

cannot even say that one is rationally required to act on the reason with the greatest requiring

strength. For in the example, this would mean that one would be rationally required to act on

reason A, and continue to the pharmacist’s.

The problem with phrases like ‘what we have most reason to do’ is similar to the

historical problem with the utilitarian phrase ‘the greatest good for the greatest number’. Those

who used this phrase wrongly assumed that these two criteria – quantity and number – combined

smoothly to yield unambiguous moral advice. But in many cases the greatest good, taken as a

sum of utility, might concentrate it in the hands of a smaller number than some wider distribution

of a lesser sum of utility. This sort of possibility reveals the false assumption, or confusion,

18

hidden in the elegant utilitarian phrase. We might call this ‘the problem of multiple criteria’.

Utilitarians, at least, have some ready answers to this problem. One is simply to say that they

really meant only ‘the greatest good’, however distributed. And another is to say that, as a matter

of empirical fact, given the strife caused by inequalities and also the fact of diminishing marginal

utility, achieving the greatest good typically involves distributing it amongst the greatest number,

so that in most cases we won’t have a conflict of criteria. But those who use phrases such as

‘what we have most reason to do’ to indicate what we rationally ought to do or what we would

do if we were ideally rational have no such answers available.25 For example, it will not do to

say that ‘what we have most reason to do’ is meant to refer only to what we have most requiring

reason to do. For it can often be rationally permissible to do what one has less requiring reason

to do. This was the case in the headache/charity example offered above. For as long as the

altruistic reasons in that case are sufficient to rationally justify acting against the reasons to spend

the money on headache medicine – reasons which have greater requiring strength – one is not

irrational in acting on those altruistic reasons: that is, it is false that one rationally ought not act

on those altruistic reasons, even though they have less requiring strength. This means that it is

false that one rationally ought to act on the reasons with greater requiring strength. It is false

even if we accept principle P in its starkest form, and conclude that buying the headache

medicine is favored by a reason with nontrivial requiring strength, while giving the money to

charity is favored only by a reason that has no requiring strength at all. The parallel with

morality may again be useful here. That an action will hurt someone else is a consideration that

can morally require me not to perform the action. We can therefore say it is a consideration with

morally requiring strength. And considerations of self-defense, for familiar reasons, are

plausibly held to be considerations that have only morally justifying strength. Still, when I act in

a way that hurts someone else, in self-defense, I may well be acting in a way that is completely

morally permissible. The fact that a “morally requiring consideration” speaks against my action

does not mean that I morally ought not defend myself.

19

5. The Particularist Objection

At this point in the paper a reader with particularist tendencies might well concede that

the notion of “the strength” of a reason, naïvely based on a metaphor of physical or psychological

strength, is inadequate to capture the actual way in which reasons contribute to the rational status

of actions. But such a reader might think that the paper wrongly attempts to remedy the defects

of such a position by the equally naïve expedient of simply adding another dimension of strength

that remains just as wrongly conceived. That is, the postulation of two distinct dimensions of

normative strength, in each of which dimensions a given reason has its own constant strength

value – one that it keeps as it reappears in distinct contexts – might be viewed as the unhappy but

understandable result of an inability to give up completely on the idea that we can assign stable

strength values to practical reasons. The correct view, the particularist might urge, simply gives

up on this idea altogether, and holds that reasons cannot be assigned any stable strength values at

all: neither univocal ones nor the pairs of such values that this paper suggests. According to this

objection, a specific reason may count a great deal in favor of an action in one context, while

counting much less in its favor in a second, while in a third context it might even count against

the action to which it is relevant. What I would like to argue in this section is that this sort of

particularism, represented paradigmatically in the work of Jonathan Dancy, has consequences

that those who are initially drawn to it are unlikely to be willing to embrace. In particular, this

sort of particularism makes it impossible to speak of the strength of a reason at all, even when we

restrict ourselves to a particular maximally specific context. That is, the particularist who

mounts this objection cannot merely deny that it makes sense to ascribe stable strength values to

reasons: values that they keep from context to context. Rather, the particularist must make the

blanket claim that talk of the strength of a reason makes no real sense even in a restricted

context.26 At least this is true if – as I will argue in what follows – the particularist cannot

present any view to rival the understanding of strength offered in this paper.

The notion of strength being used in this paper to explain justifying and requiring strength

is one that unavoidably makes use of or implies conditionals. On this understanding, when we

20

talk about something’s strength, even in a particular case, we are talking about a kind of power

that it has in that case, and it is not possible to explain power without committing oneself to

certain conditionals. That is, on the view of strength values as indicators of power, one cannot

use the strength of anything – a weightlifter, an electromagnetic force, or a reason – to predict or

explain what is going on even in a particular instance if one is unwilling to make conditional

claims about what that lifter, force, or reason would be able to do in other circumstances. For it

is the regularities that the set of relevant conditionals express that are doing the explaining.

Consider, for example, three different fishing lines and their corresponding strength values: 10

pound, 15 pound, and 20 pound. If we tie a 17 pound weight to each of the lines, and then lift

them, and the first two break, while the 20 pound line does not, the explanation for this last fact

is that the 20 pound line has sufficient strength to lift 20 pounds, while the others do not. And

this is a way of saying something very much like the following: in standard circumstances, if you

put any weight up to 20 pounds on this line, it doesn’t break, but if you put more, it does. If this

were not true – if the line quite often broke even with weights less than 17 pounds – then no

appeal to the strength of the line could explain why it did not break in this instance.27 A strength

value is a way of representing regularities of a certain kind: it is an index that tells us where, in

an ordered series of conditional claims, truth leaves off and falsity begins.

When we assign a strength value to a reason we are providing a concise representation of

the way it affects rational status across a range of cases, in comparison with other reasons. For

example, if one assigns a greater justifying strength to the fact that an action will save a

stranger’s life than to the fact that it will save the agent’s own little finger, this is a way of

endorsing the following claim, which makes essential use of conditionals: any sacrifice that

would be rationally permissible for the sake of saving one’s little finger would also be rationally

permissible for the sake of saving a stranger’s life, although the converse is not true. And if one

assigns a greater requiring strength to the fact that an action will save the agent’s own little finger

than to the fact that it will save a stranger’s life, this is a way of endorsing the following claim:

21

any action that would rationally required in order to save a stranger’s life would also be rationally

required in order to save one’s own little finger, although the converse is not true.28

It may be worth remarking here, to avoid misunderstanding, that I am in broad sympathy

with many of the intuitions that lead to particularism in domains other than that of practical

rationality, such as the domain of theoretical rationality, and perhaps the moral domain also,

which I take to be distinct from the domain of practical rationality.29 For example, I agree with

the particularist that it is not at all plausible that one could assign constant weights to reasons for

belief that would yield the correct rational status of any given belief. But this does not lead me to

particularism about theoretical reasons because I simply do not believe that the rational status of

belief is a matter of the strengths of reasons at all. If one is antecedently committed to the idea

that reasons for belief bear weights, then it is hard to avoid the conclusion that these weights

must vary a great deal from context to context.30 But another option is to give up on the idea that

reasons for belief have weights at all, and that they contribute to the rational status of belief in the

same way that reasons for action contribute to the rationality of action. It is surprising to me that

particularists, given the emphasis they place on the ubiquity of contextual dependence, should

assume that there is a univocally correct account of ‘reason’ as it figures in epistemic contexts,

moral contexts, and the context of judgments of practical rationality.31

But let us return to the domain of practical rationality, and let us assume – since the

particularists we are addressing share this assumption – that the rational status of an action is a

function of the strengths of the reasons that favor and disfavor it.32 The problem for

particularists then is that until the counterfactual understanding of strength is supplanted, they

will not be able to define rational action by means of such phrases as ‘action favored by the

balance of reasons’, or ‘action based on the best or most reasons’. But all of the particularists of

whom I am aware do wish to talk about rational action in this way.33 Note that the current

problem for the particularist is not the same as the one described above as ‘the problem of

multiple criteria’. That problem had its roots in the fact that requiring and justifying strength

need to be distinguished. The current problem for the particularists would arise even on the

22

assumption that reasons have only one strength value. For the problem is in explaining what

such a strength value indicates, if it does not indicate a stable pattern of the sort described above

in terms of conditionals.

Even those particularists who do not define rational action in term of the “best” or “most”

reasons clearly still want to be able to talk about the strengths of reasons.34 But unless one is

willing to ascribe stable, non-particularistic strength values, such talk is problematic. Why can’t

a particularist claim that, in a particular context, a reason has a specific strength? The answer is

that it is completely unclear what such a claim could mean, unless it means that it can be

assigned a value that could be used to figure out the rational status of other actions in other

contexts. But that is just the non-particularist, conditional notion of a stable strength value.

Particularists cannot criticize the notion of a stable strength value used in non-particularist

accounts of reasons – a notion that gives sense to the idea of rational status as a function of the

strengths of the reasons that favor and oppose action – and then go on to help themselves to the

same notion, just “subtracting the stability.” For it is the stability – the regular contribution that a

reason makes across a range of cases, so that certain conditional claims are true and others false –

that gives sense to the notion of strength, even in a particular instance.

It may be useful to explain this stability requirement in a context less conceptually

confusing than the domain of the normative. Imagine therefore an object being pushed around by

various rockets that are attached to it, pushing it in different directions: say a red rocket, a blue

one, and a yellow one. Suppose that the mixture of fuel in each rocket is very uneven, so that the

rockets push with different forces at different times. Someone might ask “What is the force of

the yellow rocket?” In this case, one might hold a position superficially similar to the

particularist position. That is, it might be correct to answer “There is no constant force

associated with the rockets; one can only ask what the force is at a given moment.” But the

reason it makes sense to talk about the force at a given moment, and to say that the resultant force

on the object at that moment is a function of such forces, is that we know that there is, at any

given moment, something going on in each rocket – the fuel mixture is of such-and-such density,

23

etc. – and we know that if that same thing were going on in a different context, and we knew the

other relevant forces, we could calculate the resultant force and predict the motion of the object.

So it makes sense to refuse to talk about “the force that a yellow rocket exerts,” even thought it

makes sense to talk about “the force that the yellow rocket exerts at time t,” because talk about

the force it exerts at time t is parasitic on talk of the stable force exerted by, say, the combustion

of a mixture of so-and-so chemicals in a tube of so-and-so dimensions. If we were convinced

that there was absolutely no underlying regularity of this sort – if, despite the rockets being

completely uniform in their chemical constitution, the objects to which they were attached moved

in an erratic and unpredictable manner – then I do not think we could understand the motion of

the object as a function of forces exerted by the rockets. Rather, we would see the whole

composite – object and attached rockets – as subject to invisible and inexplicable influences. For

what conceivable experiment could ever convince us that, even in a particular context, the yellow

rocket was exerting a particular force?35 Indeed, what would it even mean to say that it was

exerting a particular force? All we would see was the resultant motion of the object to which the

rocket was attached. Perhaps for objects to which only one rocket was attached, we might be

able to conclude that the rocket was exerting some force, if the object moved. But in fact, even

this is a little suspicious, unless we knew that objects to which no rockets were attached never

moved on their own. And in any case, this would be of very little use to anyone who ever wanted

to explain the motion of an object to which more than one rocket was attached.36

If one does not see that strength values presuppose the above sort of stability, then the

following particularist rejoinder may appear tempting. Surely it makes sense to imagine various

forces or weights – indeed, any quantifiable magnitudes associated with particular objects –

varying over time in such a way that they have a value in a context, but no stable value. Why

could this not be the particularist picture of the strength of a reason? True, the weights of actual

objects do not change in such ways, but there is no contradiction in imagining that they might.

This rejoinder may seem plausible initially. But the idea that a given reason, or a given object,

could have different strengths, or different weights, in different contexts makes sense only on a

24

picture according to which the reason, or the object – which does not have any strength, or

weight “of its own” – is associated, in each different context, with some other item or items to

which it does make sense to ascribe stable values: in the weight case, this might be the mass of

the object and the local gravitational field. In the rocket case, it was the force exerted by the

combustion of a mixture of so-and-so chemicals in a tube of so-and-so dimensions. These other

items serve to explain why the fluctuating strength, weight or force varies as it does. This claim

is quite general, and applies to any items to which one can ascribe something like a strength

value (i.e., a value that describes the item’s strength, weight, force, capacity, power, etc.) in a

particular context, but which do not keep those strength values from context to context. I call

this ‘the underlying stability claim’. If the underlying stability claim is correct, then our

particularist faces a dilemma. Either there is something available to play the role of the

associated item with stable strength, or there is not. If there is such an associated item, then the

natural response of a non-particularist would be to call such items ‘the real reasons’ or ‘the basic

reasons’.37 But if there is no such associated item, then the particularist’s claim that a reason has

a strength value in a context needs to be given some sense completely independently of any

conditional claims that rely on the idea of something having a stable strength value: that is, the

analogy with a world in which objects change their weights will not be available.

Now, despite the above considerations, it is certainly possible that some sense could be

given to the particularist’s claim that a reason has a strength value in a particular context. But

until such a sense has been given, the particularist objection to the view put forward in this paper

lacks force. This seems to me to be the position in which Jonathan Dancy finds himself. He

associates the strength or weight of a contributory reason with the degree to which it bears what

he calls ‘the favoring relation’ to the action it favors. But he is skeptical about giving any

content to claims about the relative degrees with which reasons bear this relation to particular

actions: in particular, he is skeptical about giving that content in terms of counterfactual

conditionals.38 But some sense needs to be given to the comparative claim that, of two reasons

favoring the same action, one bears the favoring relation to that action to twice the degree that the

25

other does. One cannot simply say “You know what ‘twice the degree’ means: just apply it to the

case of bearing the favoring relation.” For ‘twice as much’, and other comparative terms,

sensibly apply to length, age, weight, and so on, in virtue of the content of length, age, and

weight claims. But Dancy refuses to give any content to the claim that a consideration bears the

favoring relation to an action. Sense could be given to Dancy’s claims, if he were willing to talk

of stable strength values, cashed out in terms of counterfactual conditionals. But Dancy rejects

this strategy.

Another way of putting the point is that if two different assignments of strength values to

the various reasons relevant to an action both yield the same overall status, we need to know

what it means to say of one of the assignments that it is correct, while the other is not. One sort

of answer, unavailable to the particularist, is that various counterfactual conditionals are made

true by the one assignment, and not by the other. An artificial example may help here, although

it is of course very implausible that actual normative reasons can usefully be assigned strength

values that have kind of numerical precision the discussion will presume. Moreover, in order to

simplify matters still further, I will assume, falsely, that practical reasons have only one

dimension of strength, and not two. Suppose, then, that I am a smoker who already has some

symptoms of lung disease. I am feeling very anxious for a cigarette, and smoking one would rid

me of this unpleasant feeling. But my spouse will get angry with me for smoking, and that too

would be unpleasant. On due consideration, I conclude that it would be irrational to smoke the

cigarette. Suppose that this rational status is the result of the strengths that the relevant reasons

have in this context. One assignment of values might give the health-related reason (H) a

strength of 4, the spouse-related reason (S) a strength of 4, and the anxiety-related reason (A) a

strength of 5. This would yield a judgment against smoking. But so too would the following

assignment: H: 6, S: 2, and A: 5. What then favors one assignment over the other? Well, on the

first assignment, if we remove S counterfactually (imagining that my spouse does not object),

then it becomes rationally permissible (indeed, on this artificial single-strength view, rationally

required) to smoke the cigarette; on the second assignment it remains irrational.39

26

There are of course other sorts of particularists: ones who admit that the notion of the

strength of a reason in a context has counterfactual implications. What these particularists assert

is that we can only have insight into counterfactuals that are quite local. The view of these

particularists is similar to the view one might have about the weight of an object, if one were to

inhabit a world with an extremely complex and wildly varying gravitational field. For even if

one knew the weight of an object at a particular location, this might not serve as the basis for

accurate predictions about how that object would behave ten miles away. But if one knew the

“local gravitational topology” fairly well, then one could presumably take the weights of various

objects as approximately constant when one was doing things within the confines of a small

room, or in one’s own house. There is something about such a view that merits the label

‘particularism’. It may be worth noting, however, that in the “local normative domain,” reasons

will perform the two functions outlined at the beginning of this paper. Thus, this sort of

particularist does not escape the force of the argument. Such a particularist will have to admit

that reasons play the two roles of justifying and requiring, and ought therefore to abandon talk of

the action that is favored by the most, best, or strongest reasons. In this case the reason for

abandoning such talk will not be that the particularist cannot make sense of strength claims at all.

Rather, it will be the problem of multiple criteria.

6. Conclusion

The requiring strength of a reason is an index of its capacity to require us to act against

other reasons. And its justifying strength is an index of its capacity to justify us in acting against

other reasons. If one is willing to talk about the strengths of reasons at all, one should be willing

to talk about both of these kinds of strength, for it is uncontroversial that reasons play the two

associated roles. The only question is whether two reasons with the same strength in one role

must have the same strength in the other, in which case we could, for all practical purposes, think

of reasons as having a single univocal strength value. The plausibility of principles formally

similar to P, and of claims regarding the distinction between avoiding harms and getting benefits,

27

and between more proximate and more distant future considerations, all support a negative

answer. If all this is right, then philosophers should abandon talk of the balance of reasons, and

such phrases as ‘what we have most reason to do’. They should abandon the idea that we can

always make sense of the phrase ‘the stronger of these two reasons’, and the associated idea that

it is a fundamental principle of rationality that one should always act on the stronger of two

opposed reasons. For it is not even reliably the case that, in conflicts of two reasons, one ought

to act on the reason that has the greatest requiring strength. For the opposing reason may well

have sufficient justifying strength to make it rationally permissible to act on it instead.

One very general response to all of this is to deny that one ever meant to endorse the

simplistic view of reasons and rationality that is being criticized here, and only intended such

phrases to be taken in a metaphorical way. One general form this response takes is particularism.

But we have seen that if one is willing to talk about the strengths of reasons, one needs to explain

what strength claims amount to, and why one is willing to say that a given reason has a certain

strength in a certain context. The particularist’s refusal to say anything informative here at all

robs him of the right to speak of the strengths of reasons. But the same criticism applies to non-

particularists who simply say that they never meant such phrases as ‘the strongest reasons’ to be

taken literally.

* Thanks to Victoria Costa, Jonathan Dancy, Patricia Greenspan, Alfred Mele, G. F. Schueler, to

audiences at the Universidad Torcuato Di Tella in Buenos Aires and Bowling Green State

University.

1 The notion is central to Michael Smith’s The Moral Problem (Cambridge: Blackwell, 1994),

and to Thomas Scanlon’s in What We Owe to Each Other (Cambridge: Harvard University Press,

1998). A huge literature has also grown up around Bernard Williams’s controversial claim that

practical reasons depend in a constitutive way on the antecedent motivations of the agent. See

Bernard Williams “Internal and External Reasons,” in his Moral Luck (Cambridge: Cambridge

28

University Press, 1981), 101-113. For a presentation of some of the problems with Williams’

view, and proposed solutions, see Robert Johnson, “Internal Reasons and the Conditional

Fallacy,” Philosophical Quarterly 49 (1999): 53-71 and Christine Korsgaard, “Skepticism about

Practical Reason,” in her Creating the Kingdom of Ends (Cambridge: Cambridge University

Press, 1996), 311-334. See also the essays collected in Ethics and Practical Reason, eds. G.

Cullity and B. Gaut (New York: Clarendon Press, 1997).

2 For example, Stephen Darwall writes that “It is part of the very idea of the [rationally normative

system] that its norms are finally authoritative in settling questions of what to do.” Allan

Gibbard’s notion of rationality “settles what to do [...] what to believe, and [...] how to feel.”

And Michael Smith claims that it is all and only reasons which spring from the norms of

rationality that make actions desirable. See Stephen Darwall, Impartial Reason (Ithaca: Cornell

University Press, 1983), 215-6; Allan Gibbard, Wise Choices, Apt Feelings (Cambridge, Harvard

University Press, 1990), 49; Smith, The Moral Problem, 150ff.

3 The first of these formal answers is Thomas Scanlon’s, from What We Owe to Each Other, 17,

50. The second is John Broome’s, from “Normative Requirements,” Ratio 12 (1999): 398-419.

See also Darwall, Impartial Reason, 19.

4 See Broome “Normative Requirements,” 412.

5 See Allan Gibbard, Wise Choices, 160 and Derek Parfit, “Reasons and Motivation,” Aristotelian

Society Supplement Vol. 71 (1997): 99-131, at 99.

6 But see Joshua Gert, “Breaking the Law of Desire,” Erkenntnis 62 (2005): 295-319 and Brute

Rationality (Cambridge: Cambridge University Press, 2004), chapter 6.

7 Joseph Raz, Practical Reason and Norms (New York: Oxford University Press, 1999), 43. This

is not Raz’s current position: he now advocates widespread reasons incommensurability. See his

Engaging Reason (New York: Oxford University Press, 1999), chapter 5. Elsewhere I argue that

the phenomena that Raz takes to support incommensurability is better accounted for by the view

29

advocated in this paper. See Joshua Gert, “Requiring and Justifying: Two Dimensions of

Normative Strength,” Erkenntnis 59 (2003): 5-36. Even when Raz originally expressed the

universal comparability view, he knew it yielded strange consequences. To eliminate these

consequences he introduced second-order exclusionary reasons and permissions. But a

recognition of the two dimensions of normative strength renders these expedients unnecessary

and unmotivated.

8 Of course this depends to some degree on how much I will enjoy the coffee. The reader should

therefore assume that the pleasure I get from the coffee is within the normal range. Also, the

term ‘irrational’, as it is being used in this paper, is meant to include actions that we would

ordinarily not describe with this word, since it connotes a certain degree of severity. As the term

is being used here, the silly, the stupid, the wrongheaded, and the insane are all subtypes of the

irrational, as long as they are taken to indicate a status conferred by the relevant reasons.

9 ‘Rationally permissible’ of course includes ‘rationally required’, but is wider, since it also

includes ‘rationally optional’. Also note the importance of the phrase ‘that would otherwise be

irrational’. Justification is always relative to some opposed reason or reasons that put an action

in need of justification in the first place. Some actions may not be irrational to perform, not

because they are justified by reasons, but because there is no reason why one should not perform

them: whistling while one walks, for example.

10 Moral justification is therefore to be understood as formally identical to rational justification:

the difference is only with regard to the normative domain within which a negative assessment is

mitigated. Moral excuses should be distinguished from moral justifications for acting against

moral rules. Moral excuses seem best understood as considerations that undermine the normal

presuppositions that are in the background when we make normative assessments: that the agent

knew what he was doing, was rational, and so on. As a result of considerations that go beyond

30

the scope of this paper, I do not find a formal parallel to moral excuses in the domain of practical

rationality.

11 It may well be that many or even most philosophers who make use of such maximizing phrases

would deny that they are using them literally. But such philosophers must then tell us what they

think the strength of a reason amounts to, if not a value that it contributes to a total. For they do

not seem to be speaking metaphorically when they talk about reasons coming with strength

values and, as will be argued in section 5, to assign a strength value, or even to claim that one

reason is stronger than another, is to commit oneself to a certain kind of structure within which

reasons contribute systematically to overall rational status. If one is unwilling to commit oneself

to such a structure, then one ought to leave off talking about the strengths of reasons at all.

12 I do not mean to suggest that there will always be determinate answers to such questions.

Surely any account of normative strength will make room for a realistic degree of vagueness.

13 Patricia Greenspan, “Asymmetrical Practical Reasons,” in Experience and Analysis:

Proceedings of the 27th International Wittgenstein Symposium, eds. J. C. Marek and M. E.

Reicher (Vienna: Öbv & Hpt, 2005), 387-394.

14 Jonathan Dancy, Ethics Without Principles (Oxford: Oxford University Press, 2004), 93.

15 Greenspan actually expresses doubts that a failure to act on adequate positive reasons need be

the basis for any criticism. See her “Adequate Reasons,” available on her web page. In holding

this view she differs from Dancy, for whom a failure to act on one’s greatest positive reasons will

count at least as ‘silly’. See his “Enticing Reasons” in Reason and Value: Themes from the Moral

Philosophy of Joseph Raz, eds. R. Wallace, P. Pettit, S. Scheffler and M. Smith (Oxford:

Clarendon Press, 2004), 91-118.

16 Of course there are many criticisms of particularism that hold that there must be a stable core of

invariant reasons in order for variant ones to exist. See, e.g., David McNaughton, “An

Unconnected Heap of Duties?”, Philosophical Quarterly 46 (1996): 433-47 and David

31

McNaughton and Piers Rawling, “Unprincipled Ethics,” in Moral Particularism, eds. B. Hooker

and M. Little (Oxford: Oxford University press, 2000), 256-75. But in these arguments variance

and invariance are taken merely to be matters of valence, and not of strength values. A different

and, it seems to me, much stronger argument is available once one realizes that invariable

valence alone is completely inadequate for those who wish to think of rational status as a

function of reasons.

17 See Brute Rationality, especially chapters 2 through 4. Principle P is based on a definition of

rational action that plays a central role in Bernard Gert’s moral theory. It was discussions of this

principle that first suggested to me the possibility of distinguishing the justifying and requiring

dimensions of normative strength. One should read P as having the following logical form

(without details about the question of the adequacy of the benefits):

~Rational (Harm to agent ~Benefit to someone).

This is equivalent to

Rational (~Harm to agent Benefit to someone).

The first of these forms makes it easy to see why it is only harm to the agent that plays the

requiring role, while the second form makes it easy to see why benefits for others can

nevertheless play the justifying role.

18 I say ‘by itself’, since adding an altruistic reason, in real-life circumstances, will typically add

other self-interested reasons. For example, if an action will kill someone or hurt them very

badly, it will also probably bring personal dangers to the agent who performs it. Also, the phrase

‘changing the status of an action’ should be understood as a convenient shorthand for ‘explaining

why an action to which the same reasons were relevant – with the exception of the reason at issue

– would have a different status’.

19 Although I hope it will be conceded that the idea that the self/other distinction is relevant to

practical rationality is prima facie quite plausible. It is, after all, basic to most moral theories.

32

20 See Brute Rationality.

21 See Ruth Chang, Incommensurability, Incomparability, and Practical Reason (Cambridge:

Harvard University Press, 1997); Joseph Raz, Engaging Reason, chapter 5.

22 See Darwall, Impartial Reason, 208; Gibbard, Wise Choices, 160; Parfit, “Reasons and

Motivation,” 99. Scanlon, What We Owe, 30, defines ideal rationality in these terms. See also

Barbara Herman, The Practice of Moral Judgment (Cambridge: Harvard University Press, 1993),

166-8. There Herman indicates the common nature of this view by explaining an interesting way

in which it might be argued that Kant does not hold it. Some philosophers do distinguish

rationality from what there is most reason to do, but nevertheless define the former in terms of

the latter. See David Sobel, “Subjective Accounts of Reasons for Action,” Ethics 111 (2001):

461-492.

23 Satisficing accounts do not avoid essentially the same criticism, since such accounts also make

essential use of the notion of ‘the action supported by the best reasons’, even though they do not

define ‘rational action’ by simply equating it with such action. A similar point goes for Thomas

Scanlon’s views.

24 This latter response, as Anscombe has pointed out, does not offer a reason at all, but is a denial

that there was a reason, or that one was called for. See G.E.M. Anscombe, Intention (Oxford:

Basil Blackwell, 1958), §17.

25 See Scanlon, What We Owe, p. 30.

26 Compare Margaret Little’s “Moral Generalities Revisited,” in Moral Particularism, eds. B.

Hooker and M. Little, 276-304, at 299-300. There Little’s point is that the explanation of the

moral status of a particular action – something that particularists are committed to saying that

reasons can provide – requires that the considerations doing the explaining fall into some sort of

pattern. Little plausibly takes her point to be compatible with particularism. My point in what

33

follows, however, is that when the explanation is of a certain sort – one that appeals to strength

values – then the required patterns are no longer consistent with particularism.

27 In this case, one might also want to know why the line is stronger: what it is that makes it

stronger. And the answer here might have something to do with electron bonding, or with the

area of the cross section of the line. But despite the fact that one might seek such an explanation,

it remains true that one can answer the question ‘Why did this line hold, while the others did

not?’ by saying ‘Because it is a 20 pound line, and the others are 15 and 10 pound lines’.

28 Although this last claim is at odds with kind of a moralistic outlook that is often found in

rationalist moral philosophers, it seems true. That is, although it seems irrational not to

immediately agree to spend quite a bit of money to save one’s little finger, it is not plausible to

regard it as irrational to refuse to donate a similar quantity to charity, even if one is quite sure that

it will save a life. There is a difference between being irrational and being selfish.

29 One takes the domains of morality and practical rationality to be distinct if one holds, as I do,

that people who act immorally are not necessarily acting irrationally. Although this position

seems to me commonsensical, it is admittedly controversial. I provide some considerations in

favor of it in “Korsgaard’s Private-Reasons Argument,” Philosophy and Phenomenological

Research 64 (2002): 303-324, and in Brute Rationality, chapters 1, 4, and 7.

30 Dancy gives a very nice example. The blue appearance of an object is typically a reason to

believe it is blue. But if I know I am experiencing a temporary color inversion, then its blue

appearance may be a reason to believe it is not blue. See Dancy, Ethics Without Principles, 74.

31 See Dancy, Ethics Without Principles, 74, 132, 191. G. F. Schueler defends the idea that the

notion of a reason for belief is very closely linked to notion of a reason for action in his Reasons

and Purposes (Oxford: Oxford University Press, 2003), 97-105. On his view, reasons for belief

are a species of practical reason, since believing is a thing we do. I would want to distinguish

reasons for the belief that p more sharply from reasons for believing that p than Schueler does.

34

32 At least, once various relevant considerations that are not themselves reasons have done their

work of enabling or disabling, intensifying or weakening the considerations that are in fact

reasons. See note 34 below.

33 See, for example, McNaughton and Rawling, “Unprincipled Ethics,” 260. McNaughton and

Rawling acknowledge the complexity of the relation between reasons and ought-judgments, but

the primary source of this complexity comes from an epistemic filter. And in any case, they

certainly think that the phrase ‘the most reasons’ makes univocal sense. Jonathan Dancy, in his

Ethics Without Principles, also makes liberal use of phrases such as ‘more reason to do this or

that’ and ‘most reason to do some third thing’. He also seems committed to the idea that “the

value of the whole is identical to the sum of the values of the contributing parts” (181). Given

the very strong link he maintains between values and reasons this suggests a fairly literal

interpretation ought to be given to ‘most reason’.

34 See Dancy, Ethics Without Principles, chapter 2. Dancy (105) opposes what he calls the

‘kitchen scales’ model of balancing reasons, which might suggest that he is not committed to the

picture I am here ascribing to him. But the only motivated deviation from the kitchen scales

model that Dancy’s particularism makes is that it allows many considerations that are not

themselves reasons to disable, intensify, or weaken a reason in a given context. After eliminating

the disabled reasons and modifying the enabled ones in the ways dictated by the relevant

intensifiers and weakeners, it seems that Dancy’s picture resembles the kitchen scales model in

its essentials.

35 Appeal to the ‘felt pressure’ at the surface of the object does not help, unless we have

independent reasons for associating this felt pressure with the truth of various counterfactuals.

36 Interestingly, Dancy – a defender of one of the strongest formulations of particularism – makes

a similar criticism of views that focus on only one relevant “force.” See Dancy, Moral Reasons,

(Oxford: Blackwell, 1993), 98.

35

37 Ross would call them ‘underived reasons’. See McNaughton, “An Unconnected Heap of

Duties?”. It seems to me very likely that Dancy’s denial of the plausibility of conditional

analyses of the notion of the strength of a reason is the result of a conflation of (a) real-life

reason-giving considerations, which are derived and cannot be altered counterfactually without

complex effects on the normative landscape, and (b) the reasons that those considerations give,

which are underived and can be used to specify the relevant counterfactuals much less

problematically. But to argue for this claim would take us too far afield.

38 See Dancy, Ethics Without Principles, 29, where he admits that his appeal to an unanalyzed

notion of favoring is “terribly uninformative,” and where he associates himself with Scanlon as a

partner in the crime of stopping the analysis of the normative at this level. I criticize Scanlon’s

failure to provide any analysis of the basic notion of a normative reason in “A Functional Role

Analysis of Reasons,” Philosophical Studies 124 (2005): 353-378, in which I also provide a

conditional analysis of the strength of a reason that (I think) avoids the criticisms that Dancy

levels at similar accounts.

39 This formulation emerged from a series of pressing questions from Janice Dowell.