. 11
( 38 .)


there are purely justi¬catory reasons or not. According to principle P,
or to any other principle of a similar formal structure, there are such
reasons. It is up to the adherents of the internalism requirement to explain
why the principles of rationality cannot take such a form. Although it is
not the primary purpose of this chapter to descend into these particular
arguments, it may be useful to ¬‚ag one objection to the possibility of
purely justi¬catory reasons at this point, if only to defer a full answer until
later. This objection claims (a) that if a putative justifying reason does
not ¬nd some corresponding motivation in an agent “ if, for example,
the agent does not care that her action will bene¬t some third party “
then it cannot rationally justify that agent in making any sacri¬ce. So
any justifying reason will in fact ¬nd a corresponding motivation in the
agent. The objection then goes on to assert (b) that if the reason does have
such a corresponding motivation “ if the agent does care “ then the reason
provides a prima facie rational requirement: after all, isn™t it irrational to act
against one™s preferences without some countervailing reason?31 Thus, the
objection seems to show that any justifying reason will be a requiring reason
as well.
It is important to ¬‚ag this objection here because it makes use of an
interesting kind of situation: a situation in which even the stipulation of
full information may not bring the objective and subjective rationality of
an action into agreement, because the agent is not acting for the reasons that
make the action objectively rational. In such situations there is a risk of error if
we take our intuitions about subjective rationality to give us direct insight
into objective rationality and, therefore, into the nature of reasons.32 Both
of the premises in the above objection fall into error in this way. First, (b)
makes use of the idea that it is irrational to act against one™s own desires.
But such action is “ at worst “ only subjectively irrational. As chapters 4,
7, and 8 will argue, this may show very little about the reasons that favor
31 For purposes of this objection, we can treat caring about, desiring, and preferring as
relevantly similar.
32 Or if, like Korsgaard, we take subjective rationality to be the end of the normative road.

The criticism from internalism

or oppose such action, or whether or not the agent ought to perform it.
It is easy to see this if one imagines a person with the irrational desire
to smash his own head open.33 And (a) is problematic for similar reasons.
For there is certainly something plausible in the claim that someone who
willingly performs an action that he knows will bring him a signi¬cant
harm would not be justi¬ed in such an action, even if it had very signi¬cant
foreseeable bene¬ts for others, if that agent regarded those bene¬ts with
indifference or antipathy. As chapter 7 will explain, such an action may
well be subjectively irrational, and it may be so because the agent lacked
a certain motive. But that is not enough to rob the altruistic reason of
justi¬catory power, or to justify calling the motive ˜a reason™ when it is
present. For, as chapter 4 will argue, and as chapter 7 will continue to
explain, it is to objective rationality that reasons are directly relevant.
Moreover, even when we are considering subjective rationality, it is
important to note that the claims made in support of the plausibility of (a)
are only appropriate for an action that has actually been performed. But
our judgments of even the subjective rationality of actions have a much
wider domain than this. We talk about what would have been rational to
have done in the past, and what would be rational to do in the future. And
when we discuss these actions, our descriptions of them generally do not
include the motives that produce them. This is why we can talk about
the same action being performed from different motives. Our judgments of
the rationality of actions that were not (or have not yet been) performed
are judgments of whether they would have been (or would be) rational if
they were chosen for the relevant reasons. This is why it is plausible to say of
essentially anyone, regardless of their current spiteful contempt for human-
ity, that it would be rational “ in the sense of rationally permissible “ for
them to do the morally obligatory thing. For it is plausible that there are
always suf¬cient agent-neutral justifying practical reasons in favor of morally
obligatory action.34 These reasons rationally justify such action, even if they
cannot rationally require it. Any agent, no matter how consistently sel¬sh
and mean he has been in the past, might, for all we can ever know, choose
to act on those rationally justifying reasons. Those who argue for desire-
satisfaction views by using examples involving altruistic or misanthropic
agents, or agents who are stipulated to have some other motivational setup,
33 It is useless to object that it is rational desires that provide reasons, since it is the presence
or absence of reasons behind a desire that determine whether or not the desire is rational.
34 At least, it is hard to see how morality could require one to make a sacri¬ce if no one at
all was going to be bene¬ted or spared a harm.

Brute Rationality

are relying on a picture of a ¬xed set of desires or preferences that deter-
mine our choices. This picture is a psychological ¬ction of the tempting
sort that Wittgenstein, for example, consistently challenges. For even if there
were such ¬xed sets of desires, we would have no way of knowing enough
about them for that knowledge to be what is behind our teaching of the
concepts ˜rational,™ ˜reason,™ and so on. Rather, the teaching of these con-
cepts, and therefore the relevant meanings of the corresponding words,
is more plausibly taken to be based on the foreseeable consequences of
Those who persist in holding that, despite a certain unavoidable igno-
rance of their precise content, we do indeed have such ¬xed sets of desires,
may be tempted to wield some version of the ˜ought-implies-can™ princi-
ple against the idea that there are desire-independent purely justi¬catory
reasons. But the existence of such reasons would not violate even the
strongest ˜ought-implies-can™ principle. For there is no ˜ought™ in play
when the question is only one of the rational permissibility to which justi-
¬catory reasons are relevant. But worse than this, for such an objector, is
the following: the necessary form of ˜ought-implies-can™ is not available in
the domain of rationality. When ˜ought to ™ means, roughly, ˜is rationally
required to ,™ it simply cannot plausibly be taken to imply ˜is psycho-
logically able to .™ Some people suffer from mental illnesses (addictions,
phobias, etc.) that essentially compel them to do things that they rationally
ought not do, or prohibit them from doing things that they rationally ought
to do: their irrational actions are precisely the result of the fact that they
are psychologically unable to act otherwise. Perhaps such mental illnesses
erase the moral ˜ought,™ but they leave the rational ˜ought™ intact.

conc lu s i on
This chapter defended the thesis that there may be such things as purely
justi¬catory reasons, and the consequent thesis that the internalism require-
ment on reasons may be false. The ¬rst part of the argument was formally
the same as Korsgaard™s own argument in “Skepticism about Practical Rea-
son.” Both Korsgaard and I conclude that when one is trying to justify
a principle of reason, whether that principle looks like the categorical
imperative or like principle P, no real work can be done merely by citing
some motivational requirement. In her own argument, the motivational
requirement shown to be insuf¬cient was motivational skepticism: the
view that principles of practical reason cannot motivate on their own. In

The criticism from internalism

this chapter, the motivational requirement shown to be insuf¬cient was the
internalism requirement on reasons. The ¬rst part of the argument thus
cleared the way for an unprejudiced examination of a putative rational
principle, P, that speci¬es some reasons that have no necessary power to
motivate even rational agents. These reasons have only justi¬catory power,
and never require action. The existence of such reasons is not to be denied
simply by pointing out that they are incompatible with the internalism
requirement. That is, one cannot simply insist that a view of rationality
must be false if it implies the existence of normative reasons that need
not motivate a rational agent to whose choice of action it applies. The
possibility of principles of rationality formally similar to principle P must
be ruled out on other grounds.
As in the previous chapter, there is room to accommodate those who do
not wish to go so far as to embrace purely justi¬catory reasons. Some may
have the intuition that extremely cruel behavior, even when it involves no
risk of harm to the agent (which could only happen under rather peculiar
circumstances, for it is very hard to be certain that one™s extremely cruel
behavior will not open one up to revenge or punishment), is irrational
in virtue of the altruistic reasons against it. An adherent to such a view can still
agree that principle P correctly classi¬es actions 1 through 6 above. For a
principle similar to P “ call it Q “ according to which altruistic reasons have
relatively little requiring power, would also classify those actions as P does. It
is true that Q would not imply the existence of purely justi¬catory reasons.
According to Q all reasons would have some power to require. Because of
this, all reasons would motivate a rational agent, if they were unopposed
by countervailing reasons. So a version of the internalism requirement
would remain true. But this will be of very little consolation to those who
wish to use the internalism requirement against the central theses offered
in this book. For even according to Q it will remain true that we will
need to distinguish the justifying and requiring roles of reasons when we
are determining the rational status of particular actions. And it will remain
true (as examples 1, 3, and 4 above illustrate) that a rational agent need
not be motivated to act on the strongest of two opposed reasons, whether
we take ˜strongest™ to mean ˜strongest in the justifying role™ or ˜strongest
in the requiring role.™


A functional role analysis of reasons

In the previous two chapters a great deal of attention was paid to the
notion of a practical reason. In chapter 2, the justifying and requiring
roles of these sorts of reasons were distinguished, and in chapter 3 this
distinction was further developed. But even in chapter 2, one might have
noticed that the roles of justifying and requiring were explained in terms of
an antecedent notion of rationality. For example, the justifying role is the
role of making an action rational, when otherwise it would be irrational.
Unless we have some idea what ˜rational™ and ˜irrational™ mean, this claim
will make little sense. What emerged more or less explicitly in chapter 3
was that the need to distinguish between these two roles for practical
reasons depends in an important way on what principles of rationality we
are willing to recognize. Principles that have the form of P imply that some
reasons can play a justifying role without being able to play a requiring
role at all. And principles similar to Q allow that some reasons might be
able play a very signi¬cant justifying role without being able to play much
of a requiring role. So, perhaps despite appearances, the basic normative
notion that has been used so far in this book has not been the notion of a
reason for action. Rather, it has been the notion of an action™s wholesale
rational status “ although, since the notions of subjective and objective
rationality are so closely related, it may not yet be entirely clear which
of these is most basic in explaining the justifying and requiring roles of
It is of course possible to begin an analysis of normativity by explic-
itly taking the notion of ˜a reason™ as basic, rather than the notion of
an action™s wholesale rational status. This is Thomas Scanlon™s strategy.1
But if one does this, then although one can say that reasons ˜count in
favor™ of things, when one asks Scanlon™s question ˜How do reasons count
in favor of actions?™ the only response available is his disappointing ˜as a

1 Scanlon (1998), p. 17.

A functional role analysis of reasons

reason.™2 This is the result of denying that there is any normative concept
more basic than that of a reason, and of de¬ning ˜reasonable,™ ˜rational,™
and so on, in terms of reasons. This chapter argues that this is a crucial
methodological mistake. Rather, for purposes of philosophical analysis,
wholesale rational status should be taken as more basic than reasons. To
make a rough analogy, the strategy of taking reasons to be the basic units
of normativity is a mistake of the same sort as taking individual words
to be the basic units of meaning in language. Just as it is more pro¬table
to take the sentence as the basic unit of meaning, it is more pro¬table to
take wholesale rational status as the basic unit of normative assessment.
And just as we can identify the various meanings of individual words with
the systematic contributions that they make to the meanings of sentences,
we can understand the various normative roles of reasons by understand-
ing how they contribute systematically to wholesale rational status. This
sort of analysis of normative reasons in terms of wholesale rational status
allows for much more interesting and informative answers to the question
˜How do reasons count in favor of actions?™ than reason-based accounts
can offer. Such an analysis also provides insight into features of reasons that
go beyond their bare normativity.

tak i ng th e not i on of ˜ a reas on ™ as bas i c
How is a normative reason relevant to an action? If one takes the notion of
˜a reason™ as basic “ inexplicable in terms of more basic normative notions “
then it is very tempting to answer ˜by counting in favor of it, or against
it, to a greater or lesser degree.™ If one does take this as an answer, the
question of importance when deciding what to do will be ˜What are the
reasons relevant to my choices, and what are their relative strengths?™ And
when one takes this as the important question, then one™s basic method for
assessing how one ought to act will almost certainly be either a maximizing
or a satis¬cing one.3 That is, one will take it as a truism that one should
always act in the way that the relevant reasons support to the highest, or
2 While Scanlon™s fundamental reliance on the notion of ˜a reason™ is by no means idiosyn-
cratic, he has the virtue of seeing, and admitting, this implication. See also Skorupski
(1999), pp. 436“37 for a similarly clear reliance on the notion of ˜a reason™ as the basis of
an analysis of the normative.
3 One way of avoiding this result is by claiming that the strengths of reasons are often incom-
mensurable. Joseph Raz embraces such a view, which I discuss and criticize in chapter 5.
Scanlon himself, despite taking reasons as basic, seems to want to resist a maximizing view
by claiming that the ways in which reasons combine are very complex. See Scanlon (1998),
p. 32. But this means that his ˜counting in favor™ relation is also extremely complex. Unless

Brute Rationality

to a suf¬ciently high, degree. Stated in the abstract, without the resistance
provided by concrete examples, this sort of account of objective rationality
(or advisability, or all-things-considered oughtness, or reasonableness, or
whatever one wants to call one™s most basic normative domain) is very
plausible-sounding and attractive. Indeed, if one does take the notion of ˜a
reason™ as normatively basic, it becomes almost inconceivable that reasons
could function in any other way.4
But there are problems with such accounts. Consider: what are the
reasons that favor a two hundred dollar donation to public radio? Let us
suppose that the primary reasons are that by doing so, roughly a hundred
people will receive some pleasure and edi¬cation that they would not
otherwise get.5 And what are the primary reasons in favor of a two hundred
dollar donation to UNICEF? Let us suppose that the primary reasons are
that such a donation would save three or four children from serious illness
and possible death. Now, according to reason-based accounts, the rational
status of an action is a function of the reasons that apply to it. And it is
to be assumed that the same reasons can reappear in different contexts,
retaining their essential normative capacities. Thus, it makes sense to ask
whether it would be irrational to risk, say, the loss of one™s arm in service
of the same reasons that favor donating two hundred dollars to public
radio (if the arm-risking method were the only available one). And the
answer to this question is ˜Yes, it would be irrational to take such a risk
for these reasons.™ But it would not be irrational to risk the loss of one™s
arm in service of the same reasons that favor donating two hundred dollars
to famine relief (again, if that were the only way to produce the same
bene¬ts). Thus there is an obvious and intuitive sense in which giving


. 11
( 38 .)