. 10
( 38 .)


fully rational agent, whenever a reason does. The nature of this expla-
nation makes it tempting to say that a reason must motivate an agent,
insofar as that agent is fully rational and aware of the reason. This happens
because of the ˜transmission of motivation™ explanation of the capacity of
reasons to motivate. When Korsgaard undermines this explanation of the
motivational capacity of reasons, she removes one temptation to embrace
the internalism requirement. Of course there are other temptations. The
remainder of this chapter attempts to show how they too might be

th e de n i al of th e i nte rnal i sm re qu i re m e nt
In criticizing the fundamental nature of the internalism requirement, we
can use an argument formally identical to the one Korsgaard makes ¬rst
against Hume and then against Williams. Hume and Williams seem to
be arguing from a motivational requirement on principles of reason, to
conclusions about what to admit as rational principles. Actually, Korsgaard
shows, they are each ¬rst claiming (or assuming) something about what
rational principles there are, and then drawing the motivational principle
from their respective claims. Hume™s initial claim is that reason judges
solely of truth and falsity. He then concludes that principles of reason
cannot produce motivation on their own. Williams™s assumption is that no
rational principles exist that could form the basis of a compelling argument
showing some substantive practical principle to be valid. He concludes that
one™s acceptance of substantive practical principles will therefore always
depend on one™s antecedent motivations, and hence that the reasons for
action that stem from one™s acceptance of such principles will, like all of
one™s other reasons, also depend on those antecedent motivations.
But Korsgaard herself is assuming something nontrivial about the con-
tent of rational principles and reasons, and it is these assumptions that lead
her to a motivational conclusion about them. Her undefended assump-
tion is that reasons are prima facie rational requirements.22 That is, she
assumes that in order to act against a reason in a rationally permissible way,
one always needs some countervailing reason. This assumption, which
sounds very plausible in the abstract (where most philosophical discussion
takes place), and which is almost universally held, leads her to the motiva-
tional conclusion she calls ˜the internalism requirement.™ Now, it is almost

22 For one relatively clear statement of this assumption, see Korsgaard (1996a), pp. 225“26.

The criticism from internalism

certainly true that some reasons are prima facie rational requirements. But,
according to the view on offer in this book, not all reasons are. How
might some reasons not be prima facie rational requirements? Consider
the following possible principle of rationality:

P It is irrational to do anything that one believes will cause
one harm, unless one also believes that someone (perhaps
oneself) will thereby be spared at least as signi¬cant a harm,
or that someone (perhaps oneself) will thereby receive at least
as signi¬cant a bene¬t.23

P would be made more plausible, but more complicated than present
purposes require, if it were couched in terms of likelihoods of harms and
bene¬ts. Moreover, P stands in need of some account of what counts as
a harm or bene¬t, and how it is determined that harms or bene¬ts are to
be measured as compensating for each other. Another modi¬cation to P
might be to allow that causing harm to others also stands in need of rational
justi¬cation, though refraining from bene¬ting them does not. All these issues
will be dealt with in chapter 7. For present purposes, such modi¬cations,
clari¬cations, and additions are not required. Present purposes are served
by the recognition that a principle structurally similar to P could be a rational
Structurally, P is usefully understood as having two parts. The ¬rst part
speci¬es a class of actions that are potentially irrational: those that the
agent believes will cause him harm. The second part speci¬es the sort of
considerations that would make it rationally permissible to perform such
an action: when anyone, possibly even the agent, will thereby avoid at
least as signi¬cant a harm or gain at least as signi¬cant a bene¬t. The
unless-clause in P is to be read as involving an inclusive ˜or,™ so that it
yields a rational permission, rather than specifying conditions under which
the agent is rationally required to do something that he knows will bring
him harm. Some examples will help illustrate this.25 P makes all of the
following actions rationally permissible:

23 The corresponding principle of objective rationality, obviously, simply removes the rela-
tivization to the beliefs of the agent. In fact, the relativization in P is too simple to make P
a plausible principle of rationality, but the ways in which it oversimpli¬es are not relevant
to the present discussion. See chapter 7 for a fuller discussion.
24 B. Gert (1998), esp. chs. 2“4 defends a view with a structure similar to P.
25 All these examples assume that no other signi¬cant reasons bear on the case.

Brute Rationality

(1) Avoiding harm even though one could prevent more signi¬cant harm to someone
else. It is rationally permissible, according to P, to refrain from giving
a hundred dollars to famine relief, even though one would not miss
the money much if one did, and even if one knew that one™s donation
would probably prevent serious illness for a handful of impoverished
(2) Doing things that do not harm oneself, just because one feels like it. P claims
that it is not irrational to pick up a stone and throw it into the woods,
just because one feels like it.26
(3) Suffering harm in order to avoid equivalent or greater suffering by someone else.
That is, P claims that all the following instances of altruistic behavior
are rationally permissible: throwing oneself on a grenade to save one™s
fellows, giving away one™s last bit of food to someone who is obviously
more hungry, giving up one™s seat on a bus to someone who seems
tired, etc.
(4) Suffering harm in order to provide someone else with a bene¬t that is at
least as signi¬cant. Suppose one believes that one of one™s colleagues
would bene¬t from detailed comments on a recent draft of a paper, but
that one realizes that giving such detailed comments will be painfully
boring, will involve a lot of time and effort, and will delay one™s
own research. P claims that nevertheless it is rationally permissible to
volunteer to give detailed comments on the draft.

Of course, given the rational permissibility of actions such as those in
example 1, none of the actions speci¬ed in examples 3 or 4 are rationally
required. This is true even though the harms one would thereby prevent, or
the bene¬ts one would thereby produce, might be more signi¬cant than
the harms one would thereby suffer. But though P thus classi¬es a very
wide range of actions as rationally permissible, it by no means classi¬es
every action as rationally permissible. Unless someone, perhaps oneself,
will avoid a harm that is at least as signi¬cant, or will receive a bene¬t that
is at least as signi¬cant, P classi¬es as irrational all actions that one knows
will bring one some harm. Two examples will suf¬ce.

26 ˜Just because one feels like it™ is equivalent to ˜for no reason.™ It is not itself a reason. For,
as Derek Par¬t points out, when there is a reason to do an action, it is also a reason to
want to do it. Thus, ˜because it will give me pleasure™ is a reason both to eat chocolate,
and to want to eat chocolate. But ˜just because I feel like it™ is not a reason to want to
throw a stone into the woods “ it is just a restatement that one does want to. See Par¬t
(1997), pp. 127“28.

The criticism from internalism

(5) Suffering harm merely to avoid a less signi¬cant harm for oneself. For exam-
ple, if one has a bad toothache that promises to get worse without
treatment, then P will classify an unnecessary delay in a trip to the
dentist as irrational.
(6) Suffering harm merely to get a bene¬t that is not equally signi¬cant. For
example, if one is beginning to have respiratory problems, then P will
classify smoking as irrational, even if one gets pleasure from smoking.
Many philosophical theories of rationality cannot class all of 1 through 4 as
rationally permissible.27 And theories that can classify them as permissible
often only are able to do so at the cost of wrongly classifying 5 and 6 as
permissible also.28 But common sense classi¬es all of these cases as P does,
and this counts in favor of P. P classi¬es most of our everyday activities as
rational, whether they are sel¬sh, goodhearted, spontaneous, or carefully
deliberated. And yet P does not do this simply by placing no limits on
the object of rational desire. P classi¬es much addictive, compulsive, and
phobic behavior as irrational, as well as self-destructive actions done out
of rage, stupidity, lust, and so on. And even if an agent is acting on her
coherent, informed, and considered preferences, P is still able to classify
her action as irrational, if those preferences are self-destructive and do not
involve compensating bene¬ts for anyone else.
How does P manage to classify actions as it does? It does it by implicitly
de¬ning two logically distinct roles for normative reasons: justifying and
requiring, which were more informally introduced in the previous chapter.
We can de¬ne these roles more formally as follows:
27 For example, the Thomas Nagel of Nagel (1970) would have dif¬culties classifying 1
as rationally permissible. So would any other theorist who does not allow that there is
an essential distinction between reasons that involve one™s own interests and reasons that
involve someone else™s. ˜Essential™ here means ˜depending on more than one™s generally
knowing the content of one™s preferences better than those of other people, one™s contingent
desire to do things for oneself, the likelihood of things going wrong if one tried to satisfy
other people™s desires, etc.™
28 Bernard Williams would allow that 1 through 4 might be rationally permissible, depending
upon the agent™s subjective motivational set. But then he is also committed to allowing
the same of 5 and 6, for people with unusual motivational makeups. Hume, of course,
is in the same position. The same may appear to be true of Scanlon, for the rationality
of action for Scanlon depends only on whether or not the agent acts in accord with, or
against, that agent™s own (possibly quite drastically wrong) judgment regarding how he
ought to act. In some moods Korsgaard is very close to Williams and Hume here, allowing
that the rationality of one™s actions is totally dependent upon one™s contingent practical
identity, even if this happens to be the identity of a wanton or a Ma¬oso. See Korsgaard
(1996a), pp. 256“57. But she also holds (p. 99 n. 8) that it is impossible for humans to act
without a reason, so that 2 is not a possibility.

Brute Rationality

The requiring role: the role of making it rationally required to
do (i.e., to make it irrational to fail to do) actions to which it is
The justifying role: the role of making it rationally permissible
to do actions that would otherwise be irrational.
According to P, the only reasons that play a requiring role “ the only reasons
that are prima facie rational requirements “ are those that involve avoiding
harm to the agent. That is, it is only the presence of such a reason that
can rule out an action as irrational. But all reasons can play the justifying
role. That is, if an action would otherwise be irrational (in virtue of the
fact that it will harm the agent) it can be made rationally permissible by
the fact that it will avoid at least an equally signi¬cant harm to anyone, or
will provide at least an equally signi¬cant bene¬t to anyone.
Korsgaard, along with most other philosophers, does not even see the
possibility that some reasons may serve only to justify, and not to require.
If there are reasons that serve only to justify, and not to require, then one
is never rationally required to act on them. Such reasons are formally similar
to excuses or canceling conditions in moral theory, except that they also
provide common (but not universal) human motives. As will be discussed
in chapter 4, this link to common human motives is part of the reason why
it makes sense to offer them as reasons when one in engaging in practical
argument. But, in any given instance, one can be fully rational and have
no tendency to act on such a reason “ no motivation. The existence of
such reasons therefore is incompatible with the internalism requirement.
At this point no argument is being offered that P is indeed a principle
of practical rationality. All that is being urged is that it might be. And if it
might be, then it is logically possible that some reasons serve only to justify,
and not to require. Given the logical possibility of such reasons “ reasons
that would falsify the internalism requirement “ why does Korsgaard hold
that all reasons are prima facie rational requirements? One explanation is
that Korsgaard adheres to a view of causation as involving universal, non-
stochastic laws of force, and upon the Kantian view of the will as a rational
causality that has this same universal, law-following character.29 That is,
just as a force impressed upon an object will cause it to move in the line of
force, unless there are other forces at work, so too (on Korsgaard™s view)
will a rational agent act as any given reason directs, unless there are other

29 Korsgaard (1996a), pp. 225“28.

The criticism from internalism

reasons that also bear on the choice. Thus, the following is Korsgaard™s
view about the content of the claim that a certain type of consideration,
x, provides an agent A with a reason: such a claim always says (among
other things) ˜A is prima facie rationally required to act on considerations
of type x.™ A logical consequence of this view of the content of rational
principles is the motivational view that, in a rational agent, a reason always
supplies some motivation. And this just is the internalism requirement. It is
a logical consequence of the requirement view because if, as the require-
ment view holds, reasons are prima facie rational requirements, then a
rational agent must always act on a reason when no countervailing reasons
are present.30 When no opposing rational motivation is present, or when
such opposing motivation is removed, a rational agent cannot help but act
as the unopposed prima facie requirement requires: that is what it is for
it to be a prima facie requirement. Thus, in a rational agent there is by
de¬nition a disposition to act on any reason she has, when countervail-
ing reasons are weakened or removed. This disposition to act constitutes
the motivation that, in a rational agent, the internalism requirement states
must exist.
As Korsgaard has shown, Hume™s and Williams™s motivational view that
rational principles cannot motivate on their own is only as plausible as
their antecedent claims about the content of those principles. So too is
the internalism requirement only as plausible as the antecedent claim that
reasons always specify prima facie rational requirements. How plausible
is this view? This chapter has of course not shown that it is false, just
as Korsgaard™s argument did not show that Williams™s claims, or Hume™s,
were false. As far as Korgaard™s argument goes, Williams may still be right
that there are no substantive principles of rationality that can be shown, by
argument, to be valid. And it is equally true that, as far as the argument of
this chapter goes, Korsgaard might be right that all reasons are prima facie
rational requirements. That is, for all that has been said in this chapter, she
might be right that all reasons are ones on which we must act, in the absence
of countervailing rational considerations, on pain of acting irrationally. But
Williams cannot use the claim that it is impossible for rational principles to
provide motivation in support of his claim that there are no valid substantive
rational principles. And in the same way, neither Korsgaard, nor anyone
else, can use the internalism requirement to argue that all reasons are prima
facie rational requirements. For the internalism requirement is a consequence

30 For a similar argument in favor of the internalism requirement, see Tilley (1997), p. 113.

Brute Rationality

of the requirement view. And so neither Korsgaard, nor anyone else, can
use the internalism requirement to argue against the existence of reasons
that have only a justi¬catory role, or against principles of rationality that
have the same logical form as principle P. That is, they cannot use the
internalism requirement to argue against the conclusion of chapter 2.
The argument between those who accept the internalism requirement
and those who reject it must concern itself with the question of whether


. 10
( 38 .)