. 25
( 38 .)


reasons really provide them, or whether the reasons that everyone takes
as providing suf¬cient justi¬cation for a comprehensible recommenda-
tion really do provide such justi¬cation;
(b) nor, since we are trying to provide an account of the fundamental
normative principle applicable to action, is there any more fundamental
normative principle to which one might appeal.
These two points justify the heavy use of examples in this book, since those
examples might really be used by parents when teaching their children
normative concepts.12 Of course it is possible that certain isolated parts
of the actual use of terms like ˜justi¬ed™ and ˜reason™ could reasonably
be repudiated by normal language speakers. For example, we might see
that the way these terms have been taught has been in¬‚uenced by the
social or racial climate, and that we therefore have a good explanation
for the inclusion of some considerations as ˜reasons™ that really do not
¬t the broader pattern of use. But these sorts of local corrections to the
accepted meanings of such terms cannot undermine the general structure
and content of the concepts as revealed in actual considered usage.13
It should not be thought that appeal is being made here to some principle
of the following sort: if the overwhelming majority of people regard an
action as irrational, that makes it irrational. That would be quite a dubious
normative principle. Moreover, it would leave the account open to a
standard objection that is raised against overly simple response-dependent
accounts of notions of all sorts. In this case the objection would be that if
we all suddenly went crazy, perhaps as the result of a psychoactive gas from
outer space, this would be completely irrelevant to the rational status of
actions. One standard response to this objection is to rigidify the account
by the inclusion of the word ˜actually.™ In this case, the word ˜actually™
could be inserted, for example, before the word ˜would™ in the right-hand
side of A2, yielding the following:
12 Again, it does not matter if the examples go against popular theoretical views that imply,
for example, that immoral behavior is really irrational. Such views cannot be right if the
method advocated here is correct.
13 See Raz (1999b), pp. 161“81 for arguments in a similar spirit, that allow only limited
improvements in our understandings of basic normative notions.

Two concepts of rationality

AA An action is objectively irrational iff virtually everyone act-
ually would regard the action as irrational, if they were fully
informed about all nontrivial consequences of the action.
With this modi¬cation, ˜irrational™ would continue to refer to its actual
referent even if the responses of the overwhelming majority of people
should change from what they actually are at the moment.14 This move,
however, though popular, and though it does save response-dependent
accounts from the objection, seems motivated entirely by the need to
avoid the force of the objection. Otherwise it seems ad hoc. Why, that
is, should the concept of ˜irrational action™ (or other response-dependent
concepts, such as ˜yellow™) privilege the actual responses of human beings
over those of the future, especially since we will be using those same
words in the future, when our then-actual responses may have changed?
The current account does not provide an answer to precisely this question,
since it denies that any reference to the responses of human beings, actual
or otherwise, is part of the meaning of ˜irrational action™ (or of ˜yellow™).
Nevertheless, the current account does explain why there is a connection
between the overwhelming responses of actual human beings, and the
referent of ˜irrational action™ (and of ˜yellow™): it is actual human beings
who have participated in the processes whereby people have come to learn
the meanings of these words.
Thus, on the current account A2 is not a meaning claim, and is not
asserted to be analytic. If it were, then we should have to suspend our judg-
ment as to the rational status of certain actions in all cases in which we
were unsure whether our views were shared by the overwhelming major-
ity of other people.15 But the overwhelming agreement of other people,
though relevant to the rational status of actions, is not relevant in this
way. Rather, the fact that there happens to be overwhelming agreement
in what people regard as irrational (again, in the sense given by A1) is what
allows for the ostensive teaching of the concept of irrational action, and
the related concept of a normative practical reason. And once the meaning
is taught, people can rely on their spontaneous judgments of rationality
just as reliably as they can rely on their spontaneous judgments of color.
The meaning of ˜irrational™ contains no more reference to the existence

14 For the canonical explanation of this use of ˜actually,™ see Davies and Humberstone (1980),
pp. 22“25.
15 This seems to be the sort of view David Lewis has in mind with respect to the notion of
value. See Lewis (1989).

Brute Rationality

of overwhelming agreement than does the meaning of ˜yellow.™ Rather,
the meaning of this word can be explained by reference to the class of
actions that members of the linguistic community would use in teaching
the concept. A3 speci¬es this class. That means that the meaning of ˜irra-
tional™ is more adequately given by A3 than by A2. Rather than giving the
meaning of ˜irrational,™ A2 is an important part of what, following Philip
Pettit, we might call the ˜genealogy™ of the concept of rational action. But
it is a substantive description, such as A3, that comes much closer to giving
the meaning.
˜But I don™t want to know what comes close to giving the meaning, I
want to know the meaning.™ This seems to me a misguided request. One
explains a concept when one explains how to use a word in a certain
way. This is why verbal de¬nitions often serve to explain concepts and
give the meanings of words. But often such purely verbal de¬nitions are
not suf¬cient “ as in the case of ˜yellow,™ for example, and other concepts
taught primarily by ostension “ and then other explanations are required.
I am claiming that A3 is much more useful in such an explanation than A2
would be. But by itself A3 might well be insuf¬cient. It will then be useful,
in explaining the concept of objectively irrational action, to mention that
such actions are the actions to which one typically will have the attitude
given in A1. The possible usefulness, in explaining the concept of an
objectively irrational action, of mentioning the attitude of ˜regarding as
irrational,™ may be what explains the attraction of expressivist views such
as those of Allan Gibbard and Simon Blackburn. These philosophers seem
content with claims such as A1: claims that limit themselves to describing
the attitude one typically (they might use a stronger word) expresses when
one uses a normative word. But such an account, unsupplemented by
claims like A2 and A3, cannot differentiate between response-dependent
words such as ˜funny,™ for which the criterion of correct use involves the
possession of the right response by the speaker, and response-dependent
words such as ˜yellow,™ for which the criterion involves picking out the
correct items. There is a fact of the matter with regard to yellowness, in
virtue of the overwhelming agreement of people in their visual responses,
so that anyone who does not share this response is called ˜color-blind.™16

16 This fact of the matter is consistent with there being no fact of the matter, with regard
to certain yellows, as to whether or not they are slightly greenish or slightly reddish. For
an opposed view, see Hardin (1993), p. 91. It seems to me that Hardin makes too much
of the fact that very ¬ne-grained distinctions can be made differently by people who are
equally ˜normal.™ It is possible that there is no fact of the matter as to whether the color

Two concepts of rationality

A similar claim is not true for ˜funny,™ since you and I can ¬nd different
things funny, without either of us having to regard the other as wrong.
But the overwhelming agreement in response to irrational actions (and as
will be seen presently) to harms and bene¬ts means that people who fail
to have the typical response can be identi¬ed, and their responses labeled
defective. When the degree of agreement allows for this to happen, this
permits the development of an objective notion for which an expressivist
analysis will be wrong.17
To return to the description: when we call an action ˜irrational™ in the
fundamental sense described above, we take it that the action involves a
risk of harm to the agent without a compensating bene¬t to anyone else.18
Put this way, it should seem plausible that we have indeed reached the end
of the normative road. For there does not appear to be any answer to
the question ˜Why shouldn™t I act irrationally, in that sense?™ One could
of course answer ˜Because such action will harm you without bene¬ting
anyone else.™ But that is no further reason. It is simply a repetition of the
de¬nition. It only sounds like an answer because it would have been an
answer if the question had been ˜Why shouldn™t I do this particular action?™
in a case in which the particular action turned out to be irrational.
As has already been mentioned, the above response-dependent account
of objectively irrational action is similar to response-dependent accounts
that might be given of other notions that are objective despite being rel-
ative to human nature, such as accounts of what is yellow, poisonous,
comfortable, sweet, and so on. There will always be a small minority of
people who are not harmed by certain substances, who dislike a certain
room-temperature, who think oranges are the same color as lemons, or
who cannot taste the difference between sugar and salt. The existence of
such people does nothing to falsify the claims that arsenic is a poison,

of a ¬re engine perfectly matches a particular tristimulus value, while there is a fact of the
matter as to whether it is red.
17 I explain this in more detail in J. Gert (2002a), in which I use the ˜possession of the atti-
tude by the language learner™ criterion to argue against an expressivist analysis of ˜morally
wrong.™ Michael Ridge has presented interesting counterarguments to this particular illus-
tration of the criterion, but his arguments do nothing to diminish the force of the points
here, as applied to the concepts of irrational action, reason, harm, or bene¬t. Compare
B. Gert (1998), pp. 90“91, in which the objectivity of yellowness is explained in terms
of ˜standard conditions™ and so on. This common method of ensuring the objectivity of
response-dependent concepts waits on an analysis of ˜standard™ in a way that my statistical
method does not.
18 I use the normative words ˜harm™ and ˜bene¬t™ here as a shorthand for the sort of conse-
quences mentioned in A3. This use is justi¬ed below.

Brute Rationality

that seventy degrees Fahrenheit is a comfortable room temperature, that
oranges are not the same color as lemons, and that sugar is sweet. This is
true despite the fact that the tiny minority may sometimes be the only
group to hold some true belief. Let us suppose, for example, that there was
a time when the overwhelming majority of people believed that the Earth
was ¬‚at. This did nothing to the shape of the Earth, or to the falsity of the
claim ˜The Earth is ¬‚at.™ But when there is an overwhelming agreement
in judgments that have no independent method of veri¬cation, or about
which the only arguments will be those initiated by the skeptic, then the
judgments or reactions of the tiny minority are not rightly regarded as
legitimate alternate views, but are, rather, to be regarded as wrong. It can-
not be, for example, that the vast majority of people are wrong in thinking
that ˜hello™ is a greeting, that grass is green, that sugar is sweet, or that it
is objectively irrational to risk the loss of one™s arm if no one is going to
A2 and A3 can provide an account of harm and bene¬t in the follow-
ing way. The class of actions that are objectively irrational, according to
A2 and A3, is pretty well ¬xed. Therefore, even if we did not have the
normative concepts of harm and bene¬t, we could notice, as A3 claims,
that objectively irrational actions are those that involve risk to the agent of
pain, premature death, disability, and so on, without a certain chance of
bringing someone pleasure, freedom, and so on, or of helping someone
to avoid pain, premature death, disability, and so on. If we did notice this
descriptive structure, we could invent the notions of harm and bene¬t, and
of a bene¬t compensating for a harm, in the following ways.
Harms are the types of consequences of an action, to an agent,
that can make the action objectively irrational: i.e., death, pain,
Bene¬ts are the types of consequences of an action, to an agent,
that can prevent the action from being objectively irrational,

19 It is no objection to this sort of response-dependent analysis of rationality that, whereas
blueness and sweetness are descriptive properties, rationality is a normative one. For the
point here is that rationality, in both senses at issue in this chapter, is a descriptive property,
as well as a normative one. The normativity comes from the nature of the subjective
response mentioned in A1, which is linked to human motivation in a way that having a
subjective experience of blueness or sweetness is not.
20 The reference to the agent here does not prevent us from talking about harms that my
actions might do to you. For this de¬nition of harm serves to pick out certain substantive
consequence types. The same is true for de¬nitions of bene¬t and of compensating.

Two concepts of rationality

in those cases where the action would be objectively irrational
without those consequences: i.e., pleasure, ability, reduction of
pain, etc.
Compensating bene¬ts, relative to certain harms, are bene¬ts that
actually would make it so that it was not objectively irrational for
an agent to suffer the harms: i.e., the unpleasantness of running
¬ve miles is compensated for by the pleasure one feels afterwards,
and by the contribution the running makes towards increasing
various of one™s abilities.
Unsurprisingly, given the extreme usefulness of the concepts of harm,
bene¬t, and compensating bene¬t, we did not have to wait to invent them.
De¬ning ˜harm™ and ˜bene¬t™ in terms of objective irrationality allows us
to understand why avoiding a harm always counts as a bene¬t, but why
not getting a bene¬t does not always count as a harm. It also explains why
it may be a harm to lose a certain bene¬t, when it would not be a harm
to be prevented from getting it.
It may seem implausible to some readers that notions such as rationality,
harm, and bene¬t are as objective as the notions of sweetness or yellow-
ness.21 For it may seem that there simply will not be any actions of which
it is true that ˜virtually everyone would regard the action as irrational, if
they were fully informed,™ as A2 claims. But the strength of this objection
should be signi¬cantly reduced by attention to the following three points.
First, there is vagueness in the notions of sweetness and yellowness, just as
there is vagueness in the notions of objective and subjective rationality, and
harm. Everyone admits that there are irresolvable disputes as to whether a
certain substance is sweet, or a certain object yellow. Therefore, similarly
irresolvable disputes as to whether a certain type of action is objectively
irrational, or whether a certain consequence counts as a harm, are not in
themselves any argument against the objectivity of those notions. It is in
fact a virtue of the above account that it does not contentiously impose
an unrealistically rigid precision on the relevant notions. As long as the
disputes are suf¬ciently marginal, they can be attributed to the vagueness
inherent in almost all objective notions.
Second, the disputes are marginal, especially with regard to the question
of what counts as a harm or a bene¬t. There is overwhelming agreement
21 Precisely how objective these notions are is a matter of philosophical interest, but is not
relevant for present purposes. I will be satis¬ed to convince readers that it is just as much
a matter of fact that pain is a harm as that sugar is sweet.

Brute Rationality

that pain and death, for example, are harms, even though there may not
be such overwhelming agreement on the question of precisely how much
pain it is worth risking in order to avoid a certain chance of death. One
need not agree that a very painful treatment is worth a 5 percent chance
of avoiding a premature death, in order to agree that the choice involves
a balance of harms. That is, virtually everyone agrees that one absolutely
ought not take the treatment if it has no chance of increasing one™s life-span
(or providing some other bene¬t). And virtually everyone agrees that one
absolutely ought not refuse the treatment if it were completely painless
and cost-free (unless living the extra time would bring with it some other
Third, the disputes are marginal even with regard to the question of
the relative strengths of reasons, and thus with regard to the question of


. 25
( 38 .)