ple, consider the case discussed in Par´¬üt.40 Some thieves are threatening
38 It is probably worth making it clear that the account of objective rationality offered in
this chapter does not do so: it only results in the extension being speci´¬üable in terms of a list.
39 See especially Nozick (1993) and Brandt (1979).
40 Par´¬üt (1986), pp. 12ÔÇ“13.
Two concepts of rationality
oneÔÇ™s family with death unless one opens a safe containing valuables within
the next ´¬üfteen minutes. Under the circumstances, if one takes an ÔÇ˜irra-
tionality pill,ÔÇ™ this will prevent the thieves from being able to press one
to open the safe, since one will not respond rationally to threats to one-
self or the things one cares about. Taking the pill is both objectively and
subjectively rational, according to R5 and A3. What about the subse-
quent ÔÇ˜crazyÔÇ™ actions? Since, by stipulation, the pill puts one in a state in
which one does not try to avoid harms to oneself or to the things one
cares about, the actions that are the result of having taken the pill are
irrational.41 And yet the actions are also, in a sense, objectively rational:
like the original action of taking the pill, these actions do involve, albeit
somewhat indirectly, an adequate compensating bene´¬üt. Given that the
actions are objectively rational in this sense, it might seem reasonable to
suggest that they must also be rational. After all, they are the result of a
state that is, in the circumstances, producing objectively rational actions. But
the relevant state of the agent here is something like the following: not
being disposed to avoid harms to oneself, or to the things one cares deeply
about. It is just false that the state the pill induces is the following: not
being disposed to avoid harms to oneself, or to the things one cares deeply
about, when this disposition is useful.42 After all, if the thieves suddenly run
away, one will continue to act in crazy ways. Thus the actions that result
from having taken the pill count as objectively rational in a sense, but as
paradigmatically subjectively irrational.
conc lu s i on
The purely justi´¬ücatory role of altruistic reasons is of great importance to
an understanding of rationality in the ÔÇ˜proper mental functioningÔÇ™ sense.
Because of their exclusively justi´¬ücatory role, altruistic reasons need not
motivate even rational agents to any degree. That is, even fully informed
agents can in general be completely unmotivated by such reasons, and yet
always act in an objectively rational way.43 For if they always avoid harms
41 The pill does not simply make one no longer care about these things.
42 For related points, see Par´¬üt (2001), pp. 85ÔÇ“86.
43 A slightly moderated claim would be that altruistic reasons are primarily justifying, so that
fully informed rational agents could be almost completely unmotivated by them. This
moderation of the view does not have any substantial effect on the position for which this
book is arguing, or on its criticisms of other views of rationality. For it still requires
a sharp logical distinction between the justifying and requiring strengths of reasons.
See pp. 89ÔÇ“92.
to themselves, they are at no increased risk of acting in an objectively
irrational way. Now consider the case in which an agent performs an action
that he knows will harm him considerably, but where he also knows that
the action will bene´¬üt someone else to a comparable extent. In such a
case there are two possible explanations for the action. One is that the
agent was motivated by the altruistic reason. The other is that the agent
was insuf´¬üciently averse to his own harm. In the former case, the action
does not involve any failure of mental functioning. That is, the agent who
acts from such motives is not at increased risk of performing objectively
irrational actions. But in the latter case, in which the agent is insuf´¬üciently
averse to his own harm, the action does involve such a failure. That is,
the agent is at increased risk of performing objectively irrational actions.
It is primarily in distinguishing these two cases that the motivations of the
agent become relevant to the rational status of an action. And the relevant
question is: ÔÇ˜Is the agent motivated by the prospect of bene´¬üting someone
Thus there is an important role for the contingent desires of the agent
in determining which reasons are relevant to the (subjective) rational status
of her actions, and how important those reasons are. But this role is limited
to reasons that play an exclusively or primarily justi´¬ücatory role: reasons
involving the interests of others, and reasons that involve bene´¬üts to the
agent, as opposed to harms. This accounts for much of what is plausible in
Humean accounts of rationality. But the account offered in this chapter also
allows for the Kantian intuition that there are some categorical require-
ments of reason ÔÇ“ that there are some actions that are irrational regardless of
the contingent desires of the agent. Of course, these requirements are not,
as the Kantian typically argues, moral requirements.44 However, much of
the appeal of Kantian views stems from dissatisfaction with accounts that
invest brute desire with too much normative signi´¬ücance, rather than from
a prior sense that morality is rationally required. Kantians deny that our
desires provide rational justi´¬ücation. Instead, they hold that justi´¬ücation is
provided by the reasons that stand behind our desires. The view offered
in this book is in wholehearted agreement with this intuition.
44 Philippa Foot (1978b), pp. 148ÔÇ“56 offers a view that also responds to this combination
of Humean and Kantian intuitions, suggesting that reasons stem from the interests and
desires of the agent, but that only the former provide categorical rational requirements.
David Copp (1995) offers a similar ÔÇ˜needs and valuesÔÇ™ theory of rationality. The account
offered in this chapter explains the plausibility of these other views.
Internalism and different kinds of reasons
The purpose of the present chapter is to bring the requiring/justifying
distinction to bear on a central controversy in contemporary ethical theory:
the internalism/externalism debate.1 This debate concerns the nature of
the relation between practical reasons and the desires of the agents who
have those reasons. Crudely put, internalists hold that there is a very strong
relation between the desires of a (rational) agent, and the reasons that such
an agent has, while externalists hold that the reasons that an agent has are
given by features of her situation in the world, and are independent of
her attitudes towards those features. The nature of the relation between
practical reasons and desire is of obvious relevance to a large number of
central philosophical and practical questions, including the rational status
of morally required behavior, and the reasonableness of punishing people
who act in signi´¬ücantly immoral ways.
Parties to the internalism/externalism debate have typically assumed
that, with regard to practical reasons, either internalism or externalism is
correct. And they have assumed that if, for example, internalism should
turn out to be the correct account, then there will be a single correct
interpretation of internalism that holds for all practical reasons. In bringing
the requiring/justifying distinction to bear on the internalism/externalism
debate, one major point of this chapter is that these assumptions are almost
certainly false, and that a failure to see this has hamstrung the discussion
1 The terms ÔÇ˜internalismÔÇ™ and ÔÇ˜externalismÔÇ™ are of course used to describe a wide range of
views. As understood in this chapter, internalism and externalism are views about whether
or not it is a necessary condition on the existence of a practical reason, that the agent have
a related motivation. Other versions of internalism concern a putative necessary condition
on judgments or beliefs about reasons, rather than on the existence of reasons. For the
distinction between judgment and existence internalism, see Darwall (1983), pp. 54ÔÇ“56.
There are also debates between internalists and externalists about moral reasons, moral
obligations, and moral judgments, although such debates often presuppose some view
about practical reasons internalism more generally.
from the beginning.2 Practical reasons have, as we have seen, two logi-
cally distinct normative roles in determining the rational status of action:
requiring and justifying. The relation of desire to the requiring role of a
practical reason will almost certainly be very different from the relation of
desire to the justifying role of the very same reason.
i nte rnal i sm and e xte rnal i sm
All internalists could endorse the following general claim about the link
between practical reasons and the motivations of the agents who have those
reasons: any practical reason must ´¬ünd a corresponding motivation in the
agent. But there are two signi´¬ücantly different ways of reading the ÔÇ˜mustÔÇ™ in
ÔÇ˜must ´¬ünd a corresponding motivation.ÔÇ™ Some internalists take it as making
the existence of reasons depend on the existence of some corresponding
contingent antecedent motivation in the agent. These internalists hold
that if an agent has a reason, then it follows as a matter of conceptual
necessity that the agent actually has a relevant antecedent motivation. For
example, an agent would be held to have a reason to take a walk, or to help
a stranger, only if that agent had some relevant desire, or adhered to some
relevant principle, or was (in the case of helping the stranger) benevolently
disposed, or if some other relevant motivational claim were antecedently
true of the agent. Of course it need not be the case that the agent actually
has a motivation to do the very action for which he has a reason: internalists
of the sort under discussion here do not deny the possibility of irrational
action, or of ignorance or mistake about the reasons one has. But they do
hold that if an agent has a reason to perform some action, then the agent
does indeed actually have some kind of motivation from which rational
processes could produce a motivation to perform the action. Moreover,
according to this form of internalism, the relevant rational processes can
only produce such motivation from antecedent motivation, so that the
actual reasons an agent has are always contingent on his antecedent moti-
vational setup. I will call philosophers who hold this form of internalism
ÔÇ˜Humeans,ÔÇ™3 and I will call their interpretation of internalism ÔÇ˜the Humean
2 The most recent instantiation of the debate might be dated as beginning in 1981, with the
republication of Williams (1981).
3 Bernard Williams is such a philosopher. See Williams (1981).
Internalism and different kinds of reasons
Other internalists take the ÔÇ˜mustÔÇ™ in ÔÇ˜must ´¬ünd a corresponding moti-
vationÔÇ™ as a normative requirement on agents, rather than as a necessary
condition for the existence of reasons. These internalists hold that prac-
tical reasons are given by the situation in which the agent ´¬ünds herself,
and these reasons must ´¬ünd a corresponding motivation if the agent is to
avoid the charge of irrationality. What is distinctive about the internalists
of this second type is that they hold that reasons for action are conceptually
independent of any contingencies in the motivational setup of the agent
who has those reasons.4 I will call philosophers who hold an internalism of
this sort ÔÇ˜Kantians,ÔÇ™ and I will call their interpretation of internalism ÔÇ˜the
Kantian interpretation.ÔÇ™5 There are two ways of being a Kantian inter-
nalist. The ´¬ürst way is to hold that there are certain rational processes
that would produce speci´¬üc motivations in any agent, completely inde-
pendently of that agentÔÇ™s contingent antecedent motivation. This way of
being a Kantian internalist involves denying the Humean internalistÔÇ™s claim
that rational processes cannot produce motivation except from antecedent
motivation.6 A second way of being a Kantian internalist is to hold that
there are simply some motivations that one is rationally required to have,
even though there may be no rational processes that would necessarily
bring one to have them.7 This sort of Kantian might hold that the desires
to avoid death and pain, for example, are simply rationally required, in
virtue of death and pain being bad things. For such a Kantian, failure to
have one of these desires would by itself be enough to convict one of irra-
tionality. Sometimes this sort of Kantian internalist is called an externalist,
because the status of a consideration as a reason seems so far removed from
the nature of the agent whose reason it is.8 But this chapter will reserve
4 Of course our desires are likely to have a causal impact on substantive consequences such
as our likelihood of feeling frustration or satisfaction, or our likelihood of success in the
relevant action. And these consequences may constitute reasons.
5 The labels ÔÇ˜HumeanÔÇ™ and ÔÇ˜KantianÔÇ™ are not intended to smuggle in covert interpretations of
Hume and Kant. In fact, neither Hume nor Kant talk explicitly about normative reasons in
the typical senses in which such reasons are understood in current debates about internalism.
But there are modern Humeans and modern Kantians for whom the labels are appropriate.
6 Christine Korsgaard is such a philosopher. See Korsgaard (1996b). She does not actually
deny the HumeanÔÇ™s claim in this article, but only makes room for the denial, which she
makes explicit in, for example, Korsgaard (1996a), ch. 4.
7 This type of Kantian internalist would be classi´¬üed as a ÔÇťnon-constitutive existence inter-
nalistÔÇŁ by Darwall. The former Kantian internalist, and the Humean, would both be
classi´¬üed as ÔÇťconstitutive existence internalists.ÔÇŁ See Darwall (1992), pp. 158ÔÇ“59, 165.
8 See, for example, Par´¬üt (1997), p. 101 and Brink (1986), p. 36.
the term ÔÇ˜externalistÔÇ™ for an easily distinguishable and much more extreme
Externalists about practical reasons, as here understood, deny the claims
of both Humean and Kantian internalists. They deny the Humean claim
by holding that our reasons, in a given situation, do not depend in any
conceptual way upon our contingent desires. Thus, a characteristic exter-
nalist claim would be that the prospect of restored health always provides
an agent with a reason, whether or not that agent is interested in being
restored to health. Of course such an externalist would also grant that if
restored health will bring a host of signi´¬ücant problems, then there might
be countervailing reasons that make it rationally permissible not to want,
overall, to be restored to health. But the reasons at issue in the internal-
ism/externalism debate are pro tanto ones: the sort that can be opposed or
augmented by other pro tanto reasons.
The externalist also denies the Kantian internalistÔÇ™s claim, and holds,
against the Kantian, that some reasons need not motivate us even if we are
fully rational. Thus, the externalist might claim that while it would not
be irrational in any way to be completely unmotivated to do anything to
entertain a small child left in oneÔÇ™s charge for the afternoon, the prospect
of providing pleasure to that child still provides a reason that would justify
taking some trouble to do so. Such a position is not very popular, since
it is almost impossible to see what such a reason claim would amount to
without ´¬ürst recognizing the requiring/justifying distinction.9 This near
impossibility is perhaps the result of the philosopherÔÇ™s tendency to simplify
matters in order to see more clearly. In this case, such a tendency leads to
the consideration of cases in which only one reason is relevant. But when
one considers a reason in isolation in this way, only the requiring role is
likely to be apparent. In order for justifying strength to be relevant, there
must be something to justify: that is, there must already be opposing reasons
with some requiring strength. If one simpli´¬ües by removing countervailing
reasons, and thus only sees the requiring role, then the externalist position is
going to appear conceptually confused. This explanation also helps explain
why Joseph Raz, who holds a position very similar to the externalist in
most cases, nevertheless seems to hold that if there is only one reason
relevant to a choice, then the agent is rationally required to be moved
See, e.g., Kagan (1989), pp. 378ÔÇ“80. See Raz (1999b), pp. 90ÔÇ“117.