. 26
( 38 .)


the objective rationality of particular types of actions. To see this, it is
important to keep in mind that such agreements are not the same as, and
do not entail, agreement in what one actually does, recommends, or even
allows. It is possible to recommend one option while recognizing that
many other options “ even options one would recommend that a friend
not do “ are not objectively irrational. You and I might agree, for example,
that there is nothing objectively irrational in giving up a high-paying job
in order to do less pleasant but more socially bene¬cial work. And yet you
might choose to make this sacri¬ce and urge me to do the same, while I
might continue to draw my six-¬gure income as a computer consultant,
and press you to act similarly. That there is overwhelming agreement in
what kinds of actions are objectively rational is shown by the fact that
almost all of us regard as rational (in the sense given by A1) almost all of
the actions that almost all people actually perform, whether we approve of
them or not. That is, almost all of the different career and lifestyle choices,
almost all of the choices that medical patients make, either to undergo or
to refrain from treatment, almost all of the mundane choices of what to
eat, wear, watch, say, and so on, are ones for which we can see the reasons,
and which do not mystify us or call for the special explanations that actions
typically require when we regard them as irrational.22

22 Thomas Scanlon describes something very similar to the attitude of seeing the reasons
that other people have, without being moved by them oneself. See the Appendix to
Scanlon (1998). But because Scanlon takes the notion of a reason as primitive (p. 17), he
is forced to take the idea of ˜counting in favor of an action™ as univocal, and he cannot
distinguish justifying power from requiring power. As a result, his univocal interpretation
is essentially ˜requiring™ (p. 61). Thus he cannot explicitly say that, without some sort of

Two concepts of rationality

rat i onal i ty as re late d to p rope r m e ntal f unc t i on i ng
The above account of objectively irrational action makes it objectively
rational to make great sacri¬ces for others, and it also makes it objectively
rational to ignore the interests of others to a very great degree. But accord-
ing to A3, an action that is likely to hurt the agent a great deal, but which
is also likely to save someone else™s life, would be classi¬ed as objectively
rational, even if the agent were completely unaware of the potential bene¬ts
of the action. This seems wrong in an account of rational action, if such
an account is meant to capture something about proper practical mental
functioning. But this is ¬ne. A1 through A3 do not provide an account
of rationality in this sense at all. Rather, the class of actions A3 describes
includes mistaken action that is harmful to the agent. Since A2 and A3
emerged at the end of an argument that started with the idea that there is a
class of actions of which it should make no sense to ask the question ˜Why
shouldn™t I ever perform actions that belong to that class?™ it should be clear
that objective irrationality is the fundamental normative notion applying to
There is, however, a notion that is more properly termed ˜rationality,™
and which has an intimate connection with proper practical mental func-
tioning. Of course the use of the phrase ˜proper mental functioning™ here
has nothing to do with the contingent ˜purposes™ served by human reason:
purposes that might ¬gure in an evolutionary or theological explanation
of why our reasoning abilities have their current shape. Rather, the cur-
rent account is an analysis of the concept of ˜irrational action™ as it is used
by philosophers, doctors, lawyers, and others, in a way that is intimately
related to our assessment of the agents who perform them, and that is
therefore related to the notions of moral responsibility, freedom of the
will, competence to give consent, and so on. The link with these other
important notions should always be kept in mind when rival accounts of
rationality are offered, especially Kantian accounts according to which it
is irrational to be immoral. Perhaps there is some sense of ˜irrational™ that
these accounts capture. But if there is, it is a sense that has very little to do
with moral responsibility, competence to give consent, self-control, and so
on. One main point of this chapter is to distinguish the notion of objective

de¬ciency, one can see that someone else has reasons, and that one is in the same relevant
circumstances, and yet remain unmoved. Instead, the most he can say is that one can see
that one has, oneself, reason “not to scorn” the ideals that someone else takes to provide
him with reasons, and reason “not to mock those who take it seriously” (p. 370).

Brute Rationality

rationality from the notion of rationality in this latter sense, and to provide
an account of the relation between the two. In general, philosophers have
used the term ˜rationality™ to refer to both of these notions. When I use the
term without quali¬cation, I should always be understood to be talking
about the subjective notion.
One might think that there is a simple de¬nition of subjective rationality
in terms of objective rationality, which runs as follows:

R1 An action is subjectively irrational if, relative to the beliefs
of the agent, it is objectively irrational.23

Such an account would correctly make it subjectively irrational to do
something that one knew would be harmful to oneself, even if, unbe-
knownst to one, it were very likely to save someone else™s life, and therefore
even if the action were objectively rational. Thus R1 accounts for the type
of case that shows that A3 does not capture the notion of rationality that is
related to proper mental functioning. But such a simple de¬nition will not
work. One problem is that an agent may conspicuously lack a belief: for
example, the belief that her action will be extremely harmful to her. If the
agent should believe that an action will be harmful in this way, but does not,
then that agent™s action may well be subjectively irrational.24 But R1 will
not classify it as subjectively irrational, since R1 relativizes to actual beliefs.
For example, suppose that I believe that I can ¬‚y, and therefore that I will
not fall to my death when I jump off of the roof of my apartment building.
Despite the fact that I do not believe I will be harmed by jumping, it is
still subjectively irrational to jump.25 Why? Because I should believe that
I will be harmed. In response to such cases one might try to patch up the
de¬nition in the following way:

23 See Brandt (1979), pp. 72“73; Gibbard (1990), pp. 18“19; Harman (1982), p. 127; Raz
(1999a), p. 22.
24 For present purposes, ˜should believe™ can be taken to mean something like ˜could be
faulted for not believing.™ It does not mean ˜would be delusional not to believe™ “ although
of course delusional agents will also often fail to believe what they should, in the relevant
sense, believe. Worries about the sense of ˜should™ at issue here will not carry over into
the account of rationality offered below, since that account does not refer to what the
agent should believe. In fact, the account can be used in a straightforward way to give a
relatively clear content to this important sense of ˜should.™
25 This example was chosen for vividness, but one might also be acting irrationally, although
less drastically so, in eating another oyster. Eating the oyster would be irrational if
one should (but does not) believe that it will cause the same distress that it typically

Two concepts of rationality

R2 An action is subjectively irrational if, relative to the beliefs
that the agent should have, it is objectively irrational.26
But suppose now that an agent should believe that her action will save
someone™s life, but that the agent does not actually believe this. If the action
is one that she should and does believe will be very painful, then it would
clearly be irrational for her to do it, for she does not believe it will have
any compensating bene¬t. Yet de¬nition R2 will not classify it as irra-
tional. What is going on here? Our strategy, in de¬nitions R1 and R2,
for determining the subjective rationality of an action has been ¬rst to
relativize the action to a set of beliefs, and then to determine its objective
rationality. But it seems as though we cannot relativize to the actual beliefs
of the agent, because of the relevance of beliefs that the agent should have.
But we cannot relativize to the beliefs that the agent should have either,
because of the relevance of the beliefs that the agent actually does have.
And of course one cannot simply relativize to both, since there may then
be no univocal verdict in cases in which an agent has one belief about the
consequences, but should have a different one.
At this point one might make the following suggestion, which, since it
seems to follow from some remarks of Hume™s, we can call ˜the Humean
suggestion.™ This suggestion admits that, speaking in a rough and inaccurate
way, we often call an action irrational if it is based on an irrational belief.
But strictly speaking, the suggestion continues, irrational action is captured
perfectly well by R1. That is, this suggestion holds that someone™s jumping
out of a window, irrationally believing that he can ¬‚y, may nevertheless
be a perfectly rational action. It is just the belief that it is based upon that
is irrational. There are a number of responses to the Humean suggestion.
The ¬rst response is that one should not lose sight of the ultimate goal
of the account of subjective irrationality here. We are trying to capture a
sense of ˜irrational™ that already exists, and that bears a close relation to
freedom, moral responsibility, disabilities of the will, competence to give
consent, and so on. It is easy to become diverted from this purpose by the
attractions of conceptual simplicity. One possible result of such a diversion

26 This de¬nition follows a pattern used by Rawls in de¬ning what he calls ˜subjective
rationality™ in relation to what he calls ˜objective rationality.™ See Rawls (1971), p. 417.
See also Cullity and Gaut (1997), p. 2. Cullity and Gaut relativize practical rationality not
to what we actually believe, but to what we are rationally justi¬ed in believing. This last phrase
is ambiguous as between ˜what, given the evidence, we would be rationally justi¬ed in
believing (whether or not we actually believe it)™ and ˜those of our actual beliefs that are
also rational.™ Either choice will encounter problems.

Brute Rationality

is that one ends up de¬ning two ˜cleaner™ notions of irrational action, the
¬rst of which is R1, and the second of which is ˜action based on irrational
belief or culpable ignorance.™ Call the ¬rst ˜acting on ¬‚awed desires™ and
the second ˜acting on ¬‚awed beliefs.™ Neither of these notions does what
we want. For example, when considering questions of moral responsibility,
it is often a matter of indifference whether someone™s irrational behavior
is the result of an affective disorder, or of delusions.27 Moreover, not all
actions that are based on irrational beliefs are viewed in the same way, with
reference to moral responsibility, free will, etc. If I irrationally believe that
wearing green brings me small pieces of good luck, and for that reason
I wear green, this action would not typically be regarded as irrational at
all. And if it were, it would not be regarded as being nearly as irrational
as my jumping out of the window, thinking I could ¬‚y. This difference
in the rational status of the action has nothing to do with the degree of
irrationality of the belief.
The second response to the Humean suggestion is directed against one of
the worries that inspires it. The worry is that unless we take the suggestion,
our account of irrational action is going to be drastically inelegant. That
is, one might worry that unless we divide the actions we call ˜irrational™
into the two suggested classes, the resulting account will include reference
to an unwholesome mix of actual beliefs and beliefs that the agent should
have. Readers who are sympathetic to the Humean suggestion for this
reason are advised to wait until the of¬cial account is offered below. That
account captures the differential relevance of actual beliefs and beliefs that
the agent should have. But it does so in a way that is neither ad hoc nor
disjunctive, and it also does so in a way that ¬nds a parallel in theoretical
Let us return therefore to the failures of R1 and R2. The problem,
again, is that we can relativize neither to the actual beliefs of the agent,
nor to the beliefs that the agent should have. In seeing why this is so, it is
useful to note that it is a matter of importance who is going to be bene¬ted
or harmed as a result of one™s action (or failure to act). Failing to note
the more or less obvious evidence that one™s action will result in one™s
own signi¬cant harm can, by itself, be suf¬cient to convict one of acting
irrationally. Failing to note evidence of the same sort that one™s action
will result in the same harm for someone else is not, by itself, suf¬cient.
27 Consider a case in which one™s agoraphobia provides a valid moral excuse. In such a case,
it does not really matter if the phobia is described as an irrational fear of the outdoors, or
the irrational belief that something terribly bad will happen if one does go outdoors.

Two concepts of rationality

Of course such a failure still counts as failure of theoretical rationality.
Depending on the circumstances it may also count as a moral failure. And
if the person is, for example, one™s beloved child, spouse, or friend, it may
also count as a failure of practical rationality. But the claim here is about
what counts, by itself, as a suf¬cient condition for acting irrationally. The
loss of a beloved child, spouse, or friend normally entails substantial harms
for oneself, and it is these harms that are most plausibly taken to ground
the charge of irrationality in such cases. Of course this does not mean that
when one acts to save one™s beloved child, one is motivated by the prospect
of avoiding harms to oneself. In fact, love is as much a matter of acting on
the reasons that stem from the interests of one™s beloved as it is a matter of
suffering when one™s beloved is harmed.
The relevance of who is going to be harmed by one™s action, when
one should be but is not aware of the likelihood of such harm, not only
supports the preceding arguments against the Humean suggestion, but also
points towards the following relativization:
R3 An action is irrational if, relative to the beliefs that the agent
should have about harms the agent may suffer, and to the
beliefs that the agent actually has about bene¬ts to anyone
at all, it is objectively irrational.
This relativization is simply a response to the counterexamples to R1
and R2. It is therefore, and admittedly, ad hoc. But it comes very close
to working. Nevertheless, it does not work. Consider the following case.
Suppose someone decides to kill herself in a painful way in order to produce
terrible guilt feelings in her parents. In the circumstances she can see that
a side-effect will be that her death will save the lives of a handful of other
people. But let us suppose these others are people whom the agent hates
and wishes dead. Thus the side-effect is completely undesired. But, let us
further suppose, the agent™s wish for the death of these other people is
not very great. Though she feels mild regret that her own death will save
them, she still believes that, overall, it is worth killing herself to punish her
parents. Now, if she does kill herself, her action is not irrational according
to de¬nition R3. For that de¬nition makes no reference to the motives
of the agent. It mentions only the agent™s beliefs, and the beliefs the agent
should have. Because it involves only beliefs, it cannot distinguish cases
in which a heroic person dies in order to save other people, from cases in
which a spiteful person dies with the mere distasteful knowledge that she will
thereby save other people. What seems important in this latter case is that

Brute Rationality

the reason in favor of suicide, namely that suicide will save several other
people from death, does not motivate the agent at all. In fact, she would
kill herself even more readily in the absence of this consideration.
Note, however, that though the motivations of the agent are relevant
to the question of whether saving other people™s lives rationally justi¬es
suicide, the motivations of the agent are not relevant to the question of
whether saving the agent™s own life is (in the absence of strong counter-
vailing reasons) rationally required. That is, when we are considering the
subjective rationality of an action, the motivations of the agent generally
impact the relevance of possible bene¬ts to others in a logically distinct and
more signi¬cant way than such motivations impact the relevance of harm
to the agent. For if the agent is not motivated by bene¬ts to others, those
bene¬ts generally play little or no role in determining the rationality of
the action; roughly speaking, it is as if those bene¬ts were not possible
consequences. But even if the agent is not motivated by possible harm
to herself, those harms continue to provide requiring reasons against the
Derek Par¬t™s notions of what we have most reason to do and what is most
rational for us to do parallel my notions of objective and subjective rationality.
Trying to clarify the relation between the former and the latter, Par¬t notes
that “[w]hile reasons are provided by the facts, the rationality of our desires


. 26
( 38 .)