reasons that are clearly weaker than the reasons provided by saving three
or four people from serious illness and possible death. This sense is that
the latter reasons would make it rationally permissible to act against some reasons
that the former would not.
he explains this relation, however, he must be viewed as simply relying on a heteroge-
neous collection of particular normative verdicts about wholesale rational status, and the
theoretical appeal he makes to the ā˜basicā™ notion of ā˜a reasonā™ must be regarded as illusory.
4 For one self-proclaimed expression of this inability, see Kagan (1989), pp. 378ā“80.
5 The speciļ¬c details of these examples are not important. If the reader disagrees with the
assessments of the relevant reasons, any other example may be substituted in which the
reasons in favor of one action are primarily pleasure and ediļ¬cation for a few people, while
the reasons in favor of another action are primarily the prevention of signiļ¬cant suffering
and death for a few people.
A functional role analysis of reasons
But if there are clearly stronger reasons in favor of donating to charity
than in favor of donating to public radio, wouldnā™t it be irrational to choose
to donate to public radio over donating to charity?6 When one takes the
notion of a reason as basic, so that reasons can only contribute to the
rationality of action in one way (ā˜as a reasonā™), then one is almost forced
into this counterintuitive assessment of the rational status of the options.
One initially attractive escape is to move from a maximizing conception
of rationality to a satisļ¬cing one. That is, one could hold that as long as
no alternative action is a favored by reasons that are much stronger, then
one is rationally permitted to do as one pleases, and that this is why it
is rationally permissible to donate to public radio, despite the stronger
reason in favor of donating to charity. It would be hard to provide a
formal argument against this sort of claim.7 Rather, one must produce
and motivate a better overall theory, as this book is attempting to do. But
in any case, the move to satisļ¬cing seems unlikely to save our intuitions
in the present case. For, when one bothers to list them, the reasons in
favor of donating to charity seem very considerably stronger than the reasons
in favor of donating to public radio. And yet, prior to the corrupting
inļ¬‚uence of an overly simple philosophical theory, our intuitions are that
both donations are rationally permissible.
a f unc t i onal role analys i s of reas on s
In the preceding section, the reasons that favored donating to public radio
were compared in strength with the reasons that favored donating to char-
ity. The reasons that favored donating to charity were judged stronger in
the following sense: they would make it rationally permissible to act against some
reasons that the reasons in favor of donating to public radio would not. That is, it
would be rationally permissible to risk, say, imprisonment, if that were the
only way to save three or four people from the evils of severe malnutrition.
But it would be irrational to take such a risk only to fund ten minutes
6 Of course, for neo-Humeans who advocate desire-dependent views of rationality there is
a simple answer to this question: ā˜No, it isnā™t irrational; it all depends on what you care
about.ā™ This objection has already been dealt with in chapter 3, but for additional powerful
arguments against desire-dependent views of rationality, see Quinn (1995), p. 195; Dancy
(2000), pp. 35ā“38; Raz (1999b), pp. 50ā“64; Scanlon (1998), pp. 35ā“42.
7 Although chapter 5 contains a formal argument against the idea that the rational permissi-
bility of an option is the result of its being within a certain ā˜margin of practical indifference,ā™
in terms of the strengths of the reasons that favor it, of the best option.
of public radio. In light of these comparisons, we can offer the following
criterion of strength for reasons.
C1 Given two reasons, R1 and R2 , R1 is stronger than R2 iff
(= if and only if ):
(i) R1 would make it rationally permissible to do anything
that R2 would make it rationally permissible to do.
(ii) R1 would make it rationally permissible do some things
that R2 would not make it rationally permissible to do.8
ā˜To make it rationally permissibleā™ means ā˜to make it rationally permissible,
when without the reason it would not be rationally permissible.ā™ Here is
an illustrative example. If R1 would make it rationally permissible to suffer
the loss of oneā™s ļ¬nger, but R2 would only make it rationally permissible
to suffer a dayā™s mild nausea, then we can say that according to C1, R1 is
stronger than R2 .9 Given this analysis, not only can we give sense to the
notion of one reason being stronger than another, but we also have one
informative answer to Scanlonā™s question ā˜How do reasons count in favor of
actions?ā™ This answer is ā˜By being able to make it rationally permissible to
perform actions, in cases in which, without them, it would be irrational.ā™10
This is an analysis of normative reasons that picks out a functional role that
reasons play, relative to the wholesale rational status of actions. So we have
the following functional role analysis of normative reasons.
FA1 A consideration is a reason if it can make it rationally
permissible to perform actions that would be irrational
8 It is true that, if C1 is to preserve the general transitivity of ā˜stronger thanā™ among reasons,
we must deny the following possibility: that there could be two reasons, R1 and R2 ,
such that R1 could make it rationally permissible to suffer H1 but not to suffer H2 while
R2 could make it rationally permissible to suffer H2 but not to suffer H1 . This seems
a benign assumption. At least, if we deny it, then we must deny that reasons can be
generally compared in strength. Of course, a number of philosophers hold exactly this
view, claiming that reasons may sometimes be incommensurable. I argue in chapter 5 that
appeals to incommensurability are the result of a failure to distinguish the two distinct
kinds of normative strength explained here.
9 Again, we assume here that R1 would also make it rationally permissible to suffer a dayā™s
mild nausea. See note 8 above.
10 In fact, a more accurate answer would be: ā˜By counting towards their rational permissibility
by either: (1) actually making them rationally permissible, (2) reducing the number of
additional reasons it would take to make them rationally permissible, or (3) making it so
that they would continue to be rationally permissible, even if there were more reasons
against them, when, without those supporting reasons, they would become irrational.ā™
This more involved answer is required by the pro tanto or prima facie nature of reasons.
A functional role analysis of reasons
In previous chapters we have been calling the role given by FA1 ā˜justi-
fying,ā™ for when actions would be irrational without some reason, then
those actions stand in need of justiļ¬cation, and it is reasons that provide this
justiļ¬cation. Of course, if FA1 is to be useful in identifying justifying rea-
sons, it presupposes that we have some way of determining the wholesale
rational status of actions. But we should not let this worry us at this point.
The question at issue is whether it is better to take wholesale rational status
as basic, or individual reasons. Whichever choice we make, some story will
have to be told: either a story about how we come to know that an action
is irrational, or a story about how we come to know that a certain fact
provides a reason for an action.
FA1 is of course not the whole of the story. FA1 only gives the positive,
justifying role of reasons. But reasons can only play this positive role relative
to actions that would otherwise be irrational. And we have already seen
instances in which reasons rule out actions that otherwise would have been
rationally permissible. These are instances in which reasons are playing
what we have been calling the ā˜requiringā™ role. So we should add the
following to our functional role analysis.
FA2 A consideration is a reason if it can make it irrational to
do something that would, without that consideration, be
FA2 gives sense to the claim that reasons can count against actions. It
also allows a second and distinct answer to Scanlonā™s question ā˜How do
reasons count in favor of actions?ā™ This second answer is ā˜By being able to
make it irrational to fail to do those actionsā™ or, equivalently, ā˜By rationally
requiring those actions.ā™ For example, it is generally rationally permissible
to sit in oneā™s room, reading a book. But if the room is on ļ¬re, so that
one will burn to death if one continues to read, then it is irrational to fail
to get up and leave. Thus, the fact that one can only avoid a ļ¬ery death
by getting up and leaving counts as a reason in favor of getting up and
leaving. But it is a reason that favors getting up in a way that is distinct
from making it merely rationally permissible to get up. Rather, this reason
makes it rationally required to get up: required, on pain of irrationality.
As in the case of FA1, if we have some independent way of determining
the rational status of actions, we can use FA2 to determine which con-
siderations are reasons. FA2 also suggests a second criterion of strength
for reasons, although it will be strength in a second sense, and cannot be
assumed to correspond to strength in the ļ¬rst sense.
C2 Given two reasons, R1 and R2 , R1 is stronger than R2 iff:
(i) R1 would make it irrational to do anything that R2
would make it irrational to do.
(ii) R1 would make it irrational do some things that R2
would not make it irrational to do.
For example, that one will suffer some painful scrapes is a reason against
action. But this reason is not as strong as the reason provided by the prospect
of losing an arm. This follows from C2. For while it would be irrational
to reach into a lionā™s cage to pick up a fallen hundred-dollar bill, it would
not be irrational to reach into a prickly holly-bush. Again, we determine
the strengths of these two reasons based on wholesale judgments of the
rational status of actions. We do not need an antecedent notion of reasons
being stronger or weaker than one another.
FA1 and FA2 pick out two logically distinct roles for reasons. Thus the
ļ¬nal answer to Scanlonā™s question is disjunctive. That is, reasons count in
favor of actions either by being the kind of consideration that can make it
irrational to fail to perform them (which is the same as making it rationally
required to perform them), or by being the kind of consideration that can
make it rationally permissible to perform them, when otherwise it would
have been irrational.11 If one adheres to the idea of reasons as basic, it may
initially be tempting to claim that these two functional roles are simply
alternate aspects of the same underlying normative property. One might
even be tempted to make the stronger logical claim that the proposition
that one reason is stronger than another in the justifying role necessarily
implies that it is also stronger in the requiring role. But this logical claim
is demonstrably false. It is very easy to provide consistent descriptions of
cases ā“ and this is all that is required to refute the logical claim ā“ in which
two reasons have the same strength in the justifying role but very different
strengths in the requiring role. We have already seen many such examples
in the previous two chapters. Here is a further reason for holding that
those examples were completely coherent. It is plausible that, with regard
to justifying strength, it does not matter whose interests are involved in
a reason: the agentā™s, or someone elseā™s. That is, any sacriļ¬ce that would
be made rationally permissible on account of the beneļ¬ts it was likely
11 The words ā˜kindā™ and ā˜canā™ are used here, because when other reasons are involved,
a requiring reason might not actually require an action, and a justifying reason might
not actually justify an action, all-things-considered. Reasons always provide prima facie
requirements or justiļ¬cations.
A functional role analysis of reasons
to produce for the agent would also be rationally permissible if it were
undertaken in order to produce those very same beneļ¬ts for someone
else instead. But reasons involving the interests of the agent seem to have
signiļ¬cantly more requiring strength than reasons involving the interests
of others: it would be irrational to hurt oneself signiļ¬cantly for a few
hundred dollars, but hired killers are not irrational; they are immoral.12 Of
course the general normative judgments involved in these claims might be
considered controversial. But one need not agree with those judgments in
order to see that they are coherent. And their bare coherence demonstrates
the logical distinction between justifying and requiring.
Again, it is not that some reasons play a ā˜merelyā™ justifying role because
they are comparatively weak, and that if they were strengthened they might
be able to play a requiring role. To see this more clearly, it may be worth
turning our attention again to the same distinction as it appears in many
moral views. On many moral views, self-preservation can morally justify
a very great deal, but cannot morally require anything, while the avoid-
ance of some comparatively small harm to someone other than the agent
morally justiļ¬es far less, but can nevertheless morally require more than
self-preservation would. This shows that, in the moral realm, requiring
strength is not merely a higher degree of justifying strength. The same is
true for reasons as they are relevant to the rational status of actions.
reas on s and two conc e p t s of rat i onal i ty
Although the functional role account provided above allows for an infor-
mative answer to Scanlonā™s question, so far it remains too crude. FA2 seems
to classify certain things as reasons that we do not want to classify that way.
For example, John Broome has distinguished between reasons, on the one
hand, and what he calls ā˜normative requirementsā™ on the other.13 FA2
cannot account for this distinction yet. The distinction is the following.
12 Again, neo-Humeans might explain this by reference to the differences in the contingent
desires of hired killers and normal people, but the argument of this chapter is not directed
at those who hold such views. See note 6 above. Against Kantians who wish to assert
that hired killers are in fact irrational, one might urge the following. Such a view is the
result of combining a correct rejection of desire-dependent views of rationality, with a
failure to note the distinction between the justifying and requiring roles of reasons. For if
one regards reasons as a matter of desire-independent fact, and realizes that the interests
of others provide reasons, and if one can see only the requiring role of reasons, then one
is very likely to conclude that immoral behavior is irrational.
13 Broome (1999), pp. 398ā“419. It is worth noting that Broome himself sees only one role
for normative reasons: requiring. See ch. 3 n. 10.
As Broome rightly notes, it not always true that if one believes that one
ought to , then one actually ought to . For, of course, one might
wrongly believe that one ought to , without actually having any reason to
at all. Simply believing that one ought to does not, by itself, provide
a reason to . And yet, there does seem to be something irrational in
oneā™s believing that one ought to , and yet not -ing.14 The situation
is even starker if one wrongly believes that one is rationally required to .
If one fails to while one has this stronger belief, it seems appropriate
to say that one is acting irrationally. In Broomeā™s terminology, the belief
that one is rationally required to normatively requires that one . Here
is a concrete example of how FA2 will conļ¬‚ate such normative require-
ments with reasons. Suppose that without the belief that she is rationally
required to get up early and begin working on an urgent project, it would
be rationally permissible for Joanna to sleep in. It might nevertheless be
irrational for Joanna to sleep in, if we add to the description of the situa-
tion, the fact that she believes that she is rationally required to get up early.
Since this belief makes it irrational to do what, without the belief, would
be rationally permissible, FA2 will wrongly classify it as a reason against
How can we avoid labeling the belief that one is rationally required to
ā˜a reasonā™? First, note that in order to make the example more telling,
we might sensibly have stipulated that Joanna ought to sleep in, because she
is too tired to work productively, and because she is entirely unjustiļ¬ed in
her belief that the project is due soon. That we can stipulate that Joanna
ought to sleep in, even though her sleeping in would be irrational in a
certain sense, suggests that the two senses of rationality ā“ the objective
and the subjective ā“ have come apart in a way in which they have not yet
come apart in many of our earlier examples, when we were allowing the
stipulation of full information. In fact this is right. Objective rationality is
relative to the facts of the case, regardless of whether they are known to the
agent.15 It is this sense of rationality that stands behind our claim that Joanna
ought to sleep in. Subjective rationality has a more intimate connection
with freedom of the will, mental illness, moral responsibility, and so on.