Naturalism in Moral Philosophy
Gilbert Harman
Princeton University
November 25, 2008
1 Introduction
1.1 Narrow and Wide Conceptions of Philosophy and Philo-
sophical Method
How you conceive of philosophy and philosophical method may depend on
where you first studied philosophy. While some people may have learned
philosophy as a distinctive subject with a special subject matter and special
methods, I and many other living philosophers learned philosophy as a sub-
ject whose methods and subject matter are continuous with those in other
subjects. This wider conception of philosophy is the more traditional one.
Relatively recent American philosophers who accept the wider concep-
tion of the subject and its method include John Dewey, W.V. Quine, Morton
White, Richard Brandt, Monroe Beardsley, John Ladd, Thomas Kuhn, John
Rawls, Arthur Danto, Noam Chomsky, Donald Davidson, Michael Scriven,
Richard Rorty, Thomas Nagel, Robert Nozick, and many others.1
1
E.g., Richard Brandt, Hopi Ethics: A Theoretical Analysis (Chicago: University of
1
I myself was brought up in the wider methodological tradition. I stud-
ied philosophy with many of the people just mentioned: Brandt, Beardsley,
Scriven, Rawls, Quine, White, Chomsky, and Ladd. When I started teaching
at Princeton, my cohort included Nozick, Nagel and Rorty, among others.
Other similarly inclined philosophers who taught at Princeton in the early
part of my career there included Stuart Hampshire, Walter Kaufmann, Alas-
dair McIntyre, Charles Taylor, and Donald Davidson.
Davidson and I organized conferences bringing linguists and philosophers
together. He was also quite involved with psychologists like Tversky and
Kahneman, trying to get them to accept that there were limits on how
irrational people could be interpreted to be.
I went into philosophy because it allowed me to pursue my own inter-
ests in issues in many fields, especially linguistics, computer science, and
psychology. My earliest publication was in linguistics.2 Later George Miller
and I started the Princeton University Cognitive Science Laboratory and
an undergraduate program in Cognitive Studies. More recently, I regularly
teach courses with faculty in linguistics, psychology, computer science, and
engineering.
I don t think I am special. Most of my colleagues at Princeton take a
wider view of philosophy in one or another respect.
Chicago Press, 1954); John Ladd, The Structure of a Moral Code: A Philosophical Anal-
ysis of Ethical Discourse Applied to the Ethics of the Navaho Indians (Cambridge, MA:
Harvard University Press, 1957); Thomas Nagel,  Brain Bisection and the Unity of Con-
sciousness, Synthese 22 (1971); Robert Nozick,  Newcomb s Problem and Two principles
of Choice, in Essays in Honor of Carl G. Hempel, ed. Nicholas Rescher, Synthese Library
(Dordrecht, the Netherlands: D. Reidel, 1969).
2
 Generative grammars without transformation rules: a defense of phrase structure,
Language 39 (1963).
2
1.2 Naturalism
Philosophical naturalism is a special instance of the wider conception of
philosophy, taking the subject matter and methods of philosophy to be con-
tinuous with the subject matters and methods of other disciplines, especially
including the natural sciences. From a naturalistic perspective, productive
philosophers are those who (among other things) produce fruitful more or
less speculative theoretical ideas, with no sharp distinction between such
theorizing by members of philosophy departments and such theorizing by
members of other departments. (In my view, department boundaries are of
interest only to administrators.)
Naturalism also often has an ontological or metaphysical aspect in sup-
posing that the world is the natural world, the world that is studied by the
the natural sciences, the world that is available to methodological natural-
ism. But the main naturalistic theme is methodological.
On this occasion I will discuss certain prospects for naturalism in moral
philosophy. I begin with metaphysical issues of the sort just mentioned, hav-
ing to do with naturalistic reduction in ethics, and then say something about
a few recent naturalistic methodological approaches in moral psychology,
2 Naturalistic Reduction
Naturalistic reduction in ethics attempts to locate the place of value in a
world of (naturalistically conceived) facts.
In one view, goodness and evil and rightness and wrongness are not
features that have a place in the naturalistic world as described by science.
3
Naturalists who take this view either abandon ethics altogether or try to
provide a nonfactual account of it.
Alternatively, naturalists might try to identify an act s being morally
right and wrong, good and bad, just and unjust, etc., with certain natural
properties of the act.
The most straightforward naturalistic reductive strategy appeals to the
supervenience of the moral on the natural facts. Any change in what the
agent ought morally to do requires a change in the (natural) facts of the case.
This appears to imply that there is a more or less complex natural relation
between an agent, a possible act, and the agent s situation (conceived as
a whole possible world) that holds when and only when the agent in that
situation is morally permitted to do that act. The idea then is to identify
the property of being what an agent is morally permitted to do in a given
situation with the property of being a possible act for which this natural
relation holds.
For example, suppose that act utilitarianism provided the correct ac-
count of what an agent is morally permitted to do. Given that supposition,
the supervenience strategy identifies a possible act s being what an agent
is morally permitted to do in a given situation with its being an act that
maximizes utility in that situation.
More generally, the strategy identifies a possible act s being what an
agent is morally permitted to do in a given situation with the holding of the
relevant natural relation, whatever it is, which exists between agent, act,
and situation if and only if that the agent is morally permitted to do that
act in that situation.
4
It is not a good objection that such an identification fails to capture the
meaning of  morally permitted. To suppose that water can be identified
with H2O is not to say what the word  water means as used by ordinary
people.
It is true that the moral case raises a methodological issue for naturalism,
since different moral theories disagree with each other. There are competing
versions of utilitarianism, social contract theory, virtue theory, Kantianism,
and many others. Is there a naturalistically acceptable way to resolve these
disputes by testing them against the world as competing scientific theories
can be tested?
Instead of trying to answer this question directly, let us consider three
kinds of naturalistic reduction, associated with theories of normative func-
tionalism, response dependent theories, and social convention theories.
2.1 Normative Functionalism and Virtue Ethics
One kind of virtue ethics3 appeals to a normative functionalism that seeks to
derive normative results from assumptions about functions about designed
or natural functions, purposes, roles, etc. For example, the most important
function of a clock is to keep time. Whether something is a clock depends
on its function, not on what it is made of or what it looks like, as long as it
can serve to indicate to an observer what the time is.
Furthermore, a clock can be evaluated in terms of its function. So, a
good clock is one that keeps time accurately. That s what a clock ought to
3
Philippa Foot, Natural Goodness (Oxford: Clarendon, 2001); Rosalind Hursthouse,
On Virtue Ethics (Oxford: Oxford University Press, 1999).
5
do. If it does not do so, something is wrong with it. The features of a good
clock that contribute to its accurate functioning are virtues of the clock.
Bodily organs are also defined by their proper functions. A heart is an
organ whose nature or function is to pump blood steadily. Lungs are organs
that function in breathing. Whether something counts as a heart or lung
is not a matter of its shape or what it is made of, but whether it has the
relevant function. One that actually does so is to that extent a good heart
or lung. A heart that fulfills its function poorly, by irregular pumping, or by
leaking blood, is a bad heart and something is wrong with it. The virtues
of a heart include steady pumping and not leaking.
People who have social roles have associated functions or purposes. A
good teacher is one who teaches well, who enables students to learn. Some-
thing is wrong with a teacher whose students do not learn. Virtues in a
teacher are those characteristics that enable the teacher s students to learn
as well also they can. A teacher who cannot get students to learn is not a
good example of a teacher, not a real teacher.
It is in the nature of human beings and certain other animate beings
(bees and chimpanzees, for example) that they are social beings. A good
human being has various virtues, like courage and compassion. A man
lacking courage is not a good example of a man, not a real man.
Various issues arise for views that attempt to derive moral assessments
from functionalism. Do human beings have functions or purposes as part of
their nature as human beings? Is the relevant function or purpose to lead a
good life, or even the best life? Can this function or purpose be characterized
naturalistically? Given competing views of the best life, is there a way of
6
testing these views against the world in the way that scientific hypotheses
can be tested?
2.2 Response Dependent Theories and Social Convention
Theories
Another rather different naturalistic approach identifies moral categories in
terms of something about human responses to the consideration of pos-
sibilities, in the way in which colors are sometimes identified in terms of
something about the responses of normal human perceivers.
In this approach, an act s being wrong might be identified with the dis-
positions of impartial unbiased sympathetic people to feel moral disapproval
of the act on being made vividly aware of the facts of the situation.
David Hume and Adam Smith defend different versions based on different
interpretations of sympathy. Hume has a tuning fork account of sympathy:
Humean sympathy leads someone to vibrate in tune with others and feel
similarly (if less intensely) what others are perceived to be feeling. This
yields a utilitarian result. Since people would rather be happy than unhappy,
they will favor situations in which there is more net happiness.
Smith objects that Hume s conception of sympathy cannot account for
the fact that unhappy people crave sympathy and feel better when they
receive it. Humean sympathetic vibrations would make an acquaintance of
an unhappy person sympathetically unhappy and then the unhappy person
would vibrate with the acquaintance s unhappiness, making the originally
unhappy person even more unhappy. Since the sympathy of an acquain-
tance makes an unhappy person less unhappy, Hume is wrong about what
7
sympathy is.
Smith observes that ordinary sympathy involves approval. If someone
gets a minor bump and moans and complains, observers who are aware of
the minor pains involved will not sympathize, because they will not approve
of the complainer s reactions. According to Smith, people want sympathy
because they want approval. Furthermore, in Smith s view, the relevant sort
of approval tends to be an internalized reflection of community standards.
My desire for the approval of others leads me to imagine how they will
react to me. I imagine being one of them to consider how I would react,
in this way internalizing their standards. This yields a different view of
morality from Hume s one in which what counts as right or wrong is more
heavily influenced by the conventional practices of one s society. Smith s
theory, while response dependent, sees morality as more of a matter of social
convention than Hume s does.
It is true that Hume takes social convention to be important for those
aspects of morality having to do with justice: People are disposed to approve
of those conventions that promote the general welfare. But for Smith social
conventions affect approval and disapproval more directly.4
2.3 Worries about Relativism
I think that the most promising naturalistic reductions have relativistic im-
plications. Adam Smith s is explicitly relativistic, because what captures
one s sympathy is directly affected by local customs. The point general-
4
I say more about this difference between Hume and Smith in  Moral Agent and
Impartial Spectator, in Explaining Value, Oxford, Clarendon Press, 2000.
8
izes to other response dependent theories to the extent that the relevant
response, usually approval, is directly influenced by varying customs or per-
sonal values. And functionalist theories may have to suppose that moral
conclusions are relative to one or another competing conception of the best
life, the purpose of life, etc.
Any absolutist (non-relativist) reduction of morality faces the episte-
mological problem of showing how that conception of morality is better
supported than its competitors. The problem is that there are competing
moral frameworks and no obvious way to test them against the world. Com-
pare the dispute between Cardinal Belarmine and Galileo about whether the
earth is at rest. Their dispute assumed that there is such a thing as being
absolutely at rest. The correct resolution of their dispute is that motion
is always relative. This conclusion is grounded in the fact that there is no
empirical difference between competing views about what is at absolute rest.
Similarly, there appears to be no empirical difference that would resolve
fundamental moral differences, which suggests that, from a naturalistic per-
spective, there may be no reason to believe in absolute right and wrong.5
Some might respond that to believe in moral relativism would be to
accept moral nihilism, at least if one s initial conception of moral values is
absolutist and not relativistic. But that would be like saying that to believe
in the relativity of motion would be to give up on the idea that things move
or are at rest!
Consider atheists who were brought up to believe that, when they said
5
Gilbert Harman,  Moral Relativism, in Gilbert Harman and Judith Jarvis Thomson,
Moral Relativism and Moral Objectivity (Cambridge, MA: Blackwell, 1996).
9
that something was wrong, they intended to be saying that it violated God s
commands. They may have at one time firmly believed that, if God had not
existed,, nothing would have been morally prohibited. But now, having
come to doubt that God exists, they continue to accept the same moral
principles as before (at least as regards nonreligious matters) and instead
have stopped believing that morality is the expression of God s will. In
the same way, someone who is initially committed to moral absolutism who
later decides that moral relativism is true can still accept (most of) the same
moral principles as before.
2.4 Naturalism as a response to evolutionary debunking
Recent attempts to debunk nonutilitarian moral intuitions appeal to possible
explanations of the intuitions in terms of evolution by natural selection. The
claim is that the relevant moral intuitions result from factors having nothing
to do with their truth, namely tendencies to develop whatever intuitions
might help to get one s genes into following generations.
While some theorists have argued that such evolutionary explanations
debunk intuitions that conflict with utilitarianism, leaving utilitarianism
unchallenged as the correct normative view,6 others say that, if the expla-
nations debunk nonutilitarian intuitions, they also debunk intuitions that
appear to support of utilitarianism.7
Naturalism offers a response to these debunking arguments. Compare a
6
Peter Singer,  Ethics and Intuitions, The Journal of Ethics (2005); Joshua Greene,
 The Secret Joke of Kant s Soul in W. Sinnott-Armstrong, ed., Moral Psychology 3: The
Neuroscience of Morality, Cambridge, MA: MIT Press, 2008
7
Guy Kahane,  Evolution and Debunking Arguments, Nous, forthcoming.
10
naturalistic response to a corresponding attempt at evolutionary debunking
of color perception. The response is to identify colors as response depen-
dent properties, determined by how we perceive them. If that s right, an
evolutionary account of how we happen to have the color experience is not
a debunking account. Similarly if what is right or wrong is response de-
pendent in one or another of the ways already considered, an evolutionary
account of moral intuitions is not a debunking account8.
This is all I am going to say about a possible naturalistic reduction of
morality.
3 Moral Psychology
I now want to consider how certain issues in moral psychology look from a
naturalistic point of view.
Philosophers have at times treated philosophical moral psychology as an
armchair subject based on a priori reflection on meanings and concepts.9
A naturalistic approach rejects that methodology and takes moral psy-
chology to be an empirical subject that is continuous with scientific psychol-
ogy, requiring a more intersubjective approach.10
I am going to say something (briefly) about three rather different topics
in moral psychology: linguistics as a guide to moral theory, guilt feelings,
8
Sharon Street,  A Darwinian Dilemma for Realist Theories of Value, Philosophical
Studies 127 (2006).
9
E.g., David K. Chan, editor, Moral Psychology Today: Essays on Values, Rational
Choice, and the Will (Berlin: Springer, 2008).
10
John Doris and Stephen Stich,  Moral Psychology: Empirical Approaches, The
Stanford Encyclopedia of Philosophy (Fall 2008 Edition), Edward N. Zalta (ed.), URL
=http://plato.stanford.edu/archives/fall2008/entries/moral-psych-emp/.
11
and character traits.
3.1 Linguistic Analogy
Despite great superficial differences between languages, linguistics aims at
a theory of universal grammar. Faced with apparently great differences in
moralities, perhaps moral theory can similarly develop a universal moral
grammar.
3.1.1 Generative grammars
The primary object of study of contemporary linguistics is not language
in the ordinary sense in which English, German, Mohawk, and Chinese
are languages. Any ordinary language in this sense has different dialects
that blend into one another in ways that do not correspond to national or
geographical boundaries. There is a well known saying that a language (in
the ordinary sense) is a dialect with an army and a navy. What counts
as a dialect of French rather than Flemish is a social or political issue,
not an issue in the science of language. It may be that any two speakers
have at least somewhat different dialects, with at least somewhat different
vocabularies. Chomsky and other linguists are concerned with language as
a property of a particular person, assumed to be abstractly specifiable by an
internal grammar, or I-grammar. Indeed, individuals are assumed to have
more than one I-grammar if they are bilingual.
The analogy between language and morality suggests that the primary
object of study for moral theorists, then, might not be the morality of a
group but rather a property of a particular person, abstractly specified by
12
something like an internal  moral grammar or I-morality.
A  generative grammar of a particular person s language would specify
the linguistic structure of expressions of that language, indicating the nouns,
verbs, adjectives, prepositions, etc., as well as the phrases of which these are
the  heads, as when a noun is the head of a noun phrase and a preposition
the head of a prepositional phrase. A generative grammar might specify
important aspects of the meanings of expressions, depending on how the
expressions are structured and possibly indicating the scope of quantifiers
and other such operators. The grammar would also specify aspects of the
pronunciation of expressions and would relate pronunciations of expressions
to possible interpretations, in this way indicating certain sound-meaning
(phonetic-semantic) relationships.
An analogous generative moral grammar would attempt explicitly to
characterize an individual s moral standards. Just as an I-grammar speci-
fies the structure of a well-formed linguistic sentence by using a specialized
linguistic vocabulary, an I-morality might specify the structure of permissi-
ble and impermissible actions, using a specialized vocabulary. A generative
moral grammar might be an action grammar (or a situation grammar, etc.,
depending on what sorts of things can be morally assessed).
The rules of a generative grammar have a recursive character. These
rules imply that a larger sentence can contain a smaller sentence within it,
which can contain another even smaller sentence, and so on without limit.
Analogously, an act may contain a smaller act within it. A generative I-
morality might imply that it is wrong to encourage someone to do something
that is wrong, a principle that is recursive in the sense that it implies that
13
it is wrong to encourage someone to encourage someone to do something
that is wrong, etc. Similarly, the I-morality might imply that it is wrong to
promise to do something that it would be wrong to do.
There is evidence that whether people accept certain seemingly descrip-
tive claims about whether an agent does something intentionally can depend
on whether they accept certain normative claims about side effects of ac-
tions. And whether people judge that a particular person caused a certain
result can depend on whether they judge that the person is morally at
fault for doing what he or she did. For example, whether one judges that
Tom s omitting to water the plants caused their death can depend in part
on whether one thinks Tom was obligated to water the plants.
That suggests that relevant structural descriptions of the situations or
actions may themselves sometimes involve normative assessments.
In developing a theory of linguistic grammar, Chomsky famously dis-
tinguishes what he calls  competence from what he calls  performance .11
By  competence Chomsky means an individual s internalized grammar. He
uses the term  performance to refer to all other factors that may affect the
use of language. In Chomsky s framework an individual s linguistic intu-
itions may or may not reflect the individual s competence, in Chomsky s
sense.
In moral theory as in linguistics, we can consider whether it makes sense
to postulate a distinction between an individual s internal morality or moral
 competence in contrast with various other factors that determine the per-
11
Noam Chomsky, Language and Mind, Enlarged Edition (New York: Harcourt Brace
Jovanovich, 1972).
14
son s intuitions and actions. For linguistics, the corresponding assumption
has been theoretically fruitful and illuminating. With respect to moral the-
ory, the assumption amounts to assuming that there is a distinctive moral
competence or I-morality.
3.1.2 Universals
A child picks up language from its interactions with others; there is no need
for explicit teaching, even though basic linguistic principles have only been
formulated recently and cannot have been explicitly taught to the child, who
therefore seems innately predisposed to acquire an I-grammar satisfying such
principles.
Similarly, a number of moral philosophers have suggested that ordinary
moral intuitions obey certain non-obvious rules. For instance, it may be
that moral intuitions reflect some version of a Principle of Double Effect.
Double Effect: It is worse knowingly to harm one person X in
saving another Y if (a) the harm to X is intended as part of the
means to saving Y than (b) if the harm to X is merely a foreseen
unintended side-effect of the attempt to save Y .
For another example, Thomson12 suggests that some ordinary moral intu-
itions might reflect a principle of the following sort:
Deflection: It is better to save a person X by deflecting a harmful
process onto another person Y than by originating a process that
harms Y.
12
Judith Jarvis Thomson,  Killing, Letting Die, and the Trolley Problem, Monist 59
15
These principles are not generally recognized and it is unlikely that they are
explicitly taught to children. If children do acquire moralities containing
such principles, it would seem that they must be somehow predisposed to
do so. If so, we might expect such principles to be found in all natural
moralities that children naturally acquire.
On the other hand, moralities differ about abortion and infanticide,
about euthanasia, about slavery, about the moral status of women, about
the importance of chastity in women and in men, about caste systems, about
cannibalism, about eating meat, about how many wives a man can have at
the same time, about the relative importance of equality versus liberty, the
individual versus the group, about the extent to which one is morally re-
quired to help others, about duties to those outside one or another protected
group, about the importance of religion (and which religion is important),
about the importance of etiquette (and which standards of etiquette), about
the relative importance of personal virtues, and so on.
Linguists have evidence that the child uses its limited experience to set
a relatively small number of parameters and acquire the  core syntax of
a language. Other  non-core aspects of syntax, perhaps involving stylistic
variation, are learned as exceptions.13
The linguistic analogy suggests considering whether it is possible to de-
velop a theory of morality involving principles and parameters. Perhaps, all
moralities accept principles of the form: avoid harm to members of group
G, share your resources with members of group F , etc., where G and F are
parameters that vary from one morality to another.
13
Mark Baker, Atoms of Language (New York: Basic Books, 2001.
16
However, we will need to have more explicit generative accounts of a
variety of moralities before being able to think about moral principles and
parameters.
I have described various issues suggested by the analogy between moral-
ity and linguistics. I discussed what might be involved in a generative  moral
grammar and have discussed what might be involved in a universal moral
grammar that parallels universal linguistic grammar. I believe that the is-
sues suggested by such analogies are worth pursuing.14
3.2 Guilt
I believe that moral theorists have made a number of related mistakes about
guilt defining morality in terms of nontrivial guilt, identifying having a
moral conscience with susceptibility to nontrivial guilt, and holding that
one cannot be a moral person if one lacks a susceptibility to nontrivial
guilt. (I suspect that these ideas arise from theorists over-generalizing from
their own experience, failing to realize that many quite moral people do not
experience the sort of non-trivial guilt required for these claims.)
What is the required (nontrivial) emotion of guilt? It must at least
include the thought or impression that one is guilty of having done something
wrong. Feelings of guilt trivially involve this. But a cold hearted psychopath
might have such thoughts without experiencing nontrivial feelings of guilt.
More is needed. At least (agent) regret. But a feeling of agent regret
is not enough either. One can regret doing things without feeling guilty in
14
For more discussion see Erica Roedder and Gilbert Harman,  Grammar (forthcoming
in a volume devoted to Moral Psychology, edited by John Doris) and references given there.
17
any nontrivial sense. In particular, a cold hearted psychopath might regret
having done something morally wrong without suffering from non-trivial
feelings of guilt.
Typically, the theorists I am talking about suppose that nontrivial guilt
feelings involve more or less agonized self-punishment or deep shame. (Deep
shame, of course, is more general than guilt in that one can feel shame for
one s appearance, accent, parents, or president.)
Some nontrivial feelings of guilt and/or shame may derive from internal-
ized embarrassment at being seen by others to have done something wrong or
for something shameful. Nontrivial guilt feelings might also be or instead be
internalizations of external resentment at others wrongdoing, internalized
blame and outrage.
It seems to me that we should reject all of the following proposed con-
nections between morality and such nontrivial guilt feelings: (1) Wrong acts
warrant such nontrivial guilt feelings. (2) Morality can be defined in part in
terms of such nontrivial guilt feelings. (3) It is abnormal not to be suscepti-
ble to such nontrivial guilt feelings for wrongdoing. (4) Susceptibility to such
nontrivial guilt feelings for wrongdoing is necessary for moral motivation.
Here are two reasons to doubt that such nontrivial guilt feelings are
central to morality. First, and most obviously, it is easy to imagine an oth-
erwise moral agent not susceptible to such nontrivial guilt feelings. Second,
some more or less famous moral paragons are not obviously susceptible, for
example, Nietzsche s overman and Aristotle s fully virtuous person.
In addition, I believe that I am not susceptible to nontrivial guilt feelings.
I think I am reasonably moral and many others have told me that they also
18
do not seem to be susceptible to nontrivial guilt feelings.
It is sometimes thought that there are good empirical psychological rea-
sons to think nontrivial guilt feelings are central to morality, based on studies
of psychopaths, narcissists, and children with certain brain deficits. Hare s
standard criteria for psychopathology include lack of guilt for wrongdoing.15
But, to conclude from these studies that nontrivial guilt feelings are central
to morality is to confuse  All As are Bs with  All Bs are As. There
are many moral people who are not susceptible to guilt and are not psy-
chopathic, narcissistic, or brain damaged. And psychopaths (for example)
not only lack nontrivial guilt feelings but also (and more importantly) lack
empathy and other sources of moral motivation.
It is also sometimes thought that the anticipation of nontrivial guilt is
needed as a way of indicating that a given act would be wrong. But it is
enough to have the belief or impression that the act would be wrong with
no need for any anticipation of nontrivial guilt feelings.
It may also be that some people are motivated to do the right thing in
order to avoid nontrivial guilt feelings. But not everyone needs this moti-
vation. It can be enough that one cares about others and about doing the
right thing.
It is true that the expression of nontrivial guilt feelings can have the
social function of appeasing people who one has wronged. But a moral
person not susceptible to nontrivial feelings of guilt who does something
wrong can at least honestly express regret and sympathy.
15
R. D. Hare, Without Conscience: The Disturbing World of the Psychopaths Among
Us (New York: Pocket Books, 1993).
19
These points imply that it is a mistake to define morality in terms of
nontrivial guilt feelings. It is a mistake to define moral standards as those
standards violation of which warrants nontrivial guilt feelings or to define
one s individual moral standards as those for which one is susceptible to guilt
for violating. My diagnosis of these common mistakes is that they arise from
introspective moral psychology treated as apriori insight by people subject
to guilt.
To summarize my own views about guilt. First, morality does not require
a susceptibility to nontrivial feelings of guilt. Second, I am inclined to think
it is a defect in an otherwise moral person to be susceptible to nontrivial
guilt feelings. Third, I think is possible to be brought up so as not to
be susceptible to nontrivial guilt feelings. And fourth, I think if one is
susceptible to such feelings, it is possible to get oneself not to be susceptible.
Where does the philosophical emphasis on guilt come from? I suggest
it comes from the method of introspection. Some philosophers find a con-
nection with their moral views and guilt feelings, a connection they imagine
holds for everyone. But they do not actually undertake an empirical study
of the extent to which that speculation is true.16
3.3 Character
Finally, let me say something about character traits. People are very hasty
in attributing broad and robust character traits to others, but it seems that
such traits are very rare, if they exist at all. This may cast doubt on versions
16
For further discussion see Gilbert Harman,  Guilt-Free Morality, to appear in Russ
Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume 4 (2009).
20
of virtue ethics that treat virtues as robust character traits.
Social psychologists have discovered that ordinary people have a bias
toward explaining behavior by appeal to robust character or personality
traits while overlooking important situational factors. This bias seems to
be associated with a perceptual tendency to pay more attention to a figure
than to its ground, and there appear to be significant cultural differences in
the extent to which people are subject to this tendency.
Having once attributed a trait to a given person, an observer has a strong
tendency to continue to attribute that trait to the person even in the face
of considerable disconfirming evidence, a tendency psychologists sometimes
call  confirmation bias, a bias toward noting evidence that is in accord
with one s hypothesis and disregarding evidence against it.17
Even in a world with no individual differences in robust character traits
or personality traits, people would still strongly believe that there were such
differences. This means that the apparent obviousness of the claim that
people differ in such traits is less evidential than one might think. True, it
may seem  obvious that some people have different character and person-
ality traits than others. But our finding this fact so obvious is predicted by
our tendency to a  fundamental attribution error whether or not there are
such differences.
Sometimes a person acts well or badly in a seemingly unusual way. Con-
cerning any such case, there is an issue as to what makes the difference
that leads to such seemingly unusual behavior. When you perceive or learn
17
Lee Ross and Richard Nisbett, The Person and the Situation (Philadelphia: Temple
University Press, 1991).
21
about someone you do not know doing such an unusual thing, you have a
strong tendency to attribute the behavior to some good or bad trait of the
person in question. For example, when you learn that a certain seminary
student walked right past someone who seemed to be having a heart attack,
actually stepping right over the person, you tend to think of the student as
incredibly callous, but the student may have just been in a hurry.
In the Milgram experiment, subjects were led by gradual steps to ad-
minister what they thought were very severe electrical shocks to another
person. The gradualness of the process with no obvious place to stop seems
an important part of the explanation why they obeyed a command to shock
the other person in that experiment although they would not have done so
if directly ordered to give the severe shock at the very beginning.18
What these and many many related experiments show is that aspects
of a particular situation can be important to how a person acts in ways
that ordinary people do not normally appreciate, leading them to attribute
distinctive actions to an agent s distinctive character rather than to subtle
aspects of the situation. Observers of participants in the Milgram experi-
ments are strongly inclined to blame those participants who did not stop to
help or who provided intense shocks, thinking that the explanation of these
agent s immoral actions lies in their terrible character. But the observers
are wrong: that cannot be the explanation.
The evidence indicates that people may differ in certain relatively narrow
traits but do not seem to have broad and stable dispositions corresponding
to the sorts of character and personality traits we normally suppose that
18
Stanley Milgram, Obedience to Authority (New York: Harper and Row, 1974).
22
people have.19
We must distinguish individual acts of honesty or dishonesty, courage or
cowardice, compassion or coldness from the corresponding character traits.
The ordinary conception of a character or personality trait is of a relatively
broad-based disposition to respond in the relevant way with acts of the
corresponding sort.
Aristotelian style virtue ethics shares with folk psychology a commitment
to broad-based character traits of a sort that people do not seem to possess.
Why have so many philosophers taken character based virtue ethics se-
riously? Relying on the method of introspective intuition, it seems obvious
that people have the relevant sorts of character traits. But social psychology
casts doubt on the existence of such traits and explains why philosophers
think they do.
4 Final remarks
To sum up, I have tried to say something about contemporary naturalistic
approaches to morality. I started by describing more traditional issues of
naturalistic reduction. I then discussed three current issues in naturalistic
moral psychology, having to do with the possibility of a linguistic analogy
for moral theory, with the unimportance to moral theory of guilt feelings,
and with the nonexistence of character traits.
19
Ziva Kunda, Social Cognition: Making Sense of People (Cambridge, MA: MIT Press,
1999).
23