University of Chicago Law School

Chicago Unbound
Occasional Papers Law School Publications

2001

Law and the Emotions

Eric A. Posner

Follow this and additional works at: http://chicagounbound.uchicago.edu/occasional_papers

Part of the Law Commons

Recommended Citation
Eric Posner, "Law and the Emotions," University of Chicago Law Occasional Paper, No. 42 (2001).

This Working Paper is brought to you for free and open access by the Law School Publications at Chicago Unbound. It has been accepted for inclusion
in Occasional Papers by an authorized administrator of Chicago Unbound. For more information, please contact unbound@law.uchicago.edu.
OCCASIONAL PAPERS FROM THE
LAW SCHOOL
THE UNIVERSITY OF CHICAGO

NUMBER 42

2001
The University of Chicago Law School

Occasional Paper 42

September 2001

Law and the Emotions

By Eric A. Posner

Professor of Law

"Law and theEmotions" is a version of thejohn M. Olin Law & Economics

Working Paper No. 103 (2d Series), September 2000. A revised version
will appear in the Georgetown Law Journal, volume 89 (2001).

( 2001.
 INTRODUCTION*
The role of the emotions is much neglected in
legal theory. This should be puzzling at first sight,
because emotions play an important role in many areas
of the law. Consider the following problems.
A person who kills while angry is usually guilty of
a less serious crime than a person who kills in a calm,
unemotional state, but not if the anger is caused by
hatred rather than shame.
Judges exclude photographs of gory crime scenes,
even when the photographs have probative value,
when the photographs are likely to provoke extreme
outrage or disgust. The use of victim impact statements
in capital sentencing has been criticized for inflaming
the jury against the defendant and defended as a way of
enhancing the jury's empathy for the victim.
A common justification for workplace, environmental,
and consumer product regulation is that individuals
lack information about the risks that they face.
Unexplored is the role that fear plays in decisions to fly
on airplanes or use air bags, and how agencies should
regulate in response to panics about health risks.
Mediators have long known that anger and other
emotions interfere with bargaining, but these emotions
are overlooked in contract theory. Emotional response
to breach of contract has important implications for the
design of remedies in contract law.
Market behavior and responses to cost-benefit
surveys often reflect emotional responses that people
disclaim when they are calm. Should this information
likewise be discounted by agencies engaging in
cost-benefit analysis, or should it be treated as authentic
data on actual preferences?
These issues raise questions about the relationship
between emotion and law, but legal theory is unprepared
to answer them. One reason for the neglect of emotions
in legal theory may be that the dominant strains

*Sources for quotations, arguments, and empirical data discussed in

this paper can be found in the working paper cited on the title page.

I
of normative legal theory - economic analysis,
moral-philosophical analysis, constitutional analysis - rely
on methodologies that are not well suited to analyzing
emotion, or at least, for reasons of intellectual history,
have simply not yet focused on the emotions. Another
reason may be the primitive state of the psychology
literature on the topic. Psychologists themselves admit
that they do not have a good theory of the emotions, in
part because research in this area is relatively new. Yet a
review of that literature reveals a number of insights
that are sufficiently well-developed to be of value for
legal theory. And, indeed, in the past few year the
emotions have begun to interest a small but growing
group of legal theorists.
This paper contributes to this small literature by
providing a simple but useful framework for analyzing
the relationship between law and the emotions. The
theory draws on the economic model of consumer
choice, but it does not reduce emotional behavior to
rational self-interest, and instead relaxes the consumer
choice model in order to account for emotional behavior.
The framework assumes that people's "calm" preferences,
that is, the preferences that they have when they are
not emotionally aroused, differ from their "emotion
state" preferences, which are skewed toward the stimulus
that provokes the emotion. Further, people's abilities
and beliefs about the world often differ in emotion states
and calm states. Finally, because people can anticipate
their emotional responses to various conditions, they
will often take steps earlier on either to avoid (or to
pursue) these conditions, and to cultivate certain
beneficial emotional dispositions.
The framework does simplify, but it is not simplistic,
and it sheds light on the debates discussed above. Its
predictions are fairly straightforward; the main contribution
is clarifying the relationship between emotions and
rational action by placing them in the rational choice
framework which is now the standard approach to
analyzing private law and some aspects of public law.
After discussing the psychology literature (Part 1), and
laying out the framework (Part II), I analyze emotions
2
as excuses in criminal and tort law (Part III), regulation
of emotion-laden risks (Part IV), and bargaining in con-
tract law when parties are emotional (Part V).

I. THE EMOTIONS

Although psychology lacks a widely accepted theory

of emotion, and many fundamental issues about the
nature of emotion remain unresolved, much progress has
been made in the last thirty years, and agreement on some
important issues has been achieved. An emotion is one of
a group of psychological phenomena with the following
distinctive characteristics. Emotions are usually stimulated
by the world, either via the mediation of cognition or
through a more primitive stimulus-response-like neurological
mechanism. They have a certain feel or affect characterized
usually by a focus on a particular stimulus with the result
that the rest of the environment "fades" (a little or a lot,
depending on the strength of the emotion) though does
not disappear altogether. An angry person feels a kind of
warmth and agitation; it is directed usually at another
person, the result of a slight or offense. A person who is
disgusted feels a kind of nausea; it is directed at the
object that provokes the disgust. The rest of the world
remains, but at a remove: an angry person might restrain
himself because he does not want to be arrested for
assault; a disgusted person might overcome the urge
to withdraw because he wants to help a person with a
disgusting wound or he knows that the disgusting
substance is an important medicine. Although emotions
are usually accompanied by physiological changes, there
does not appear to be a one-to-one correspondence
between the different emotions and physiological states:
emotion has an irreducibly mental component.
One useful approach to understanding the emotions
is to think about them against the background of the
theory of consumer choice. In doing so, I follow cognitive
psychology, which got its start by criticizing expected
utility theory, and, I think, the mainstream literature on
the psychology of the emotions, which appears
to understand emotion against an implicit theory of
rational behavior.

3
as excuses in criminal and tort law (Part III), regulation
of emotion-laden risks (Part IV), and bargaining in con-
tract law when parties are emotional (Part V).

I. THE EMOTIONS

Although psychology lacks a widely accepted theory

of emotion, and many fundamental issues about the
nature of emotion remain unresolved, much progress has
been made in the last thirty years, and agreement on some
important issues has been achieved. An emotion is one of
a group of psychological phenomena with the following
distinctive characteristics. Emotions are usually stimulated
by the world, either via the mediation of cognition or
through a more primitive stimulus-response-like neurological
mechanism. They have a certain feel or affect characterized
usually by a focus on a particular stimulus with the result
that the rest of the environment "fades" (a little or a lot,
depending on the strength of the emotion) though does
not disappear altogether. An angry person feels a kind of
warmth and agitation; it is directed usually at another
person, the result of a slight or offense. A person who is
disgusted feels a kind of nausea; it is directed at the
object that provokes the disgust. The rest of the world
remains, but at a remove: an angry person might restrain
himself because he does not want to be arrested for
assault; a disgusted person might overcome the urge
to withdraw because he wants to help a person with a
disgusting wound or he knows that the disgusting
substance is an important medicine. Although emotions
are usually accompanied by physiological changes, there
does not appear to be a one-to-one correspondence
between the different emotions and physiological states:
emotion has an irreducibly mental component.
One useful approach to understanding the emotions
is to think about them against the background of the
theory of consumer choice. In doing so, I follow cognitive
psychology, which got its start by criticizing expected
utility theory, and, I think, the mainstream literature on
the psychology of the emotions, which appears
to understand emotion against an implicit theory of
rational behavior.

3
 It is immediately clear that emotions are troublesome
for the rational choice starting point that people choose
actions in a way that is consistent with their preferences.
It is common to think of emotions as "outside" forces
that compel one to act inconsistently with the interests
of the self. This conception is familiar from everyday
excuses for offensive conduct. A person might apologize
for insulting someone else by explaining that he was
"angry," or "depressed," or "irritable." The implication is
always that the speaker is not fully responsible for his
actions, because he was taken over by his emotions. But
this conception is wrong. We hold people responsible
for their emotions. If anger was justified by some earlier
wrong committed against the agent, then anger may be
an excuse; if it resulted from an innocent mistake, then
the anger is no excuse. Whether or not a person's
emotional state excuses his conduct depends on
whether the person could have avoided the emotion or
avoided the stimulus that provoked the emotion, and
whether the emotion reflects acceptable moral beliefs
about others and the world.
A better way to think about emotions is to start
by noting that a person's preferences might have different
emotional valences. X has a preference that Y not be
around. The preference could be more or less intense,
and let us fix the intensity by supposing that X is willing
to pay $100 to avoid encountering Y. Economic analysis
usually stops this point, and takes this number as given.
But it may also be relevant that X's preferences has an
emotional coloring. X's emotions toward Y are qualitatively
distinct, and this difference cannot be reduced to a dollar
amount. X might hate Y, be angry at Y, or be disgusted
with Y - while in all cases having the same monetized
preference to avoid Y.
Do these distinctions matter? In some cases the
reduction of emotionally colored desires for certain
world states to a monetized preference ordering is
unobjectionable. If we are simply concerned with deterring
certain behavior, and if the emotionality of the desire
does not result in an idiosyncratic influence on
decision-making, knowledge of an affected person's costs

4
and gains is sufficient for determining policy. For
example, the law need not make distinctions on the
basis of a tortfeasor's emotional state if its only purpose
is to ensure that the victim is compensated for his
injury. Angry and calm drivers are sufficiently deterred if
they must pay the social cost of their behavior. But in
some contexts the emotional coloring of a preference
does have instrumental and normative consequences.
From an instrumental perspective, the effectiveness
of different kinds of sanctions often depends on a person's
emotion state. An angry person discounts future sanctions
but may be quite sensitive to claims about the provocation.
If you push into a person, you are more likely to avoid
being punched in the nose if you tell him that you
tripped than if you remind him that he could go to jail
for assault. Anger is also vulnerable to delaying tactics
and cooling off periods. Hatred, by contrast, is a more
constant emotion, and also less likely to be extinguished
by a reasonable explanation. Further, we expect irascible
people to avoid places where their anger is likely to be
provoked, just as we expect epileptics to avoid driving.
Fear, disgust, and the other emotions all have their
different idiosyncrasies, and a well-designed legal system
exploits them differently.
From a normative perspective, the bare fact that
a person has acted under the influence of emotion does
not excuse his conduct. In fact, while some emotions
mitigate guilt, others enhance guilt. Anger provoked by
betrayal still mitigates guilt; but anger provoked by
unacceptable moral beliefs may increase guilt. Hate
rarely excuses murder, but real fear, even if not fully
justified, might mitigate culpability.
Both of these observations assume that people
remain rational while under the influence of emotion;
emotion is rarely, if ever, merely a reflex to some external
stimulus. An angry, disgusted, fearful, or sad person
usually can deliberate about his behavior and does not
(with the possible exception of certain kinds of fear)
engage in reflexive action. This suggests that people
continue to act rationally while in an emotion state,
even though they act differently from the way they do

5
in the calm state. One can capture this point by positing
that during the emotion state, people experience usually
temporary variations in their preferences, abilities,
and/or beliefs.
Their preferences change so that what psychologists
call the "action tendency" of an emotion becomes relatively
attractive. The action tendency of anger is to strike out;
so we can say that a person, while angry, develops a
temporary preference to strike the person who offends
him. The action tendency of disgust is to withdraw;
a person, while disgusted, develops a temporary
preference to withdraw from the disgusting object. Grief
produces withdrawal from other people and preoccupation
with the lost person or thing; fear produces flight from a
threat; pity produces aid. But before and usually after
the emotion state, the person's preferences are constant
(the "calm preferences"), so he might disapprove of
what he expects to do, or did, in the emotion state. It is
this inconsistency over time that makes emotional
behavior seem irrational; but it is important to see that
a person in an emotion state does not act irrationally
given his temporary preferences.
Abilities may also change in the grip of an emotion.
When the emotion state occurs, the agent may find
himself more alert and vigorous, even stronger, or
simply less reliant on slow-moving deliberation. The
angry person is aroused: he feels less pain, tires less
quickly, responds more rapidly to movement. The anxious
or fearful person becomes more alert to the environment,
and flees quickly from danger. A grief-stricken person
may experience a decline in abilities: everything
becomes more difficult to do. Evidence of physiological
changes - hormonal changes, increase in the heart rate,
and so forth - supports the view that abilities change
during some emotion states.
Finally, beliefs may change during emotion states.
An angry person overestimates the probability that the
offender will attack him, or that the provocation was
not an accident but the result of intent to harm or
humiliate. A fearful person overestimates the probability
of harm associated with the threat that causes his fear.
6
Joyful people underestimate risks of harm, pessimistic
people overestimate the same risks.
My claim, then, is that during the emotion state,
a person acts rationally, that is, internally consistently,
given the new and usually temporary preferences, abilities,
and beliefs that he has in that emotion state. The
actions taken during the emotion state will, of course,
affect the agent's endowments, and this may have
consequences for his behavior after the emotion state is
over; but aside from that I assume that preferences,
abilities, and emotions during the calm state are the
same before and after the emotion state.
My final point is that agents anticipate their
emotion states, and take actions in anticipation of
them. This brings us to the matter of character.
"Emotional disposition" refers to a person's tendency to
feel an emotion, which itself is no doubt connected to
genetic, cultural, and educational factors. An irascible
person is more likely to become angry; a fearful person is
more likely to become scared. People usually know their
own emotional dispositions and can take steps to modify
them or to avoid conditions that activate them.
Suppose, for example, that a person knows that if he
goes to a rowdy bar, he may be insulted, and further he
knows that he is irascible. Upon being insulted, he
might strike the person who insults him. To avoid this,
he can (1) knowing about his emotional disposition
avoid the bar, or (2) earlier on try to overcome his
irascibility through meditation (or whatever).
These actions are those of a rational person who
has self-knowledge, including knowledge about his
emotional disposition. Not all people know everything
about their emotional dispositions, but enough people
do enough of the time that we recognize that people
ought to act with attention to their emotional dispositions
and blame them if they do not. Indeed, I will argue
below that the criminal law does just this. More generally,
law cares about emotions because (1) a person responds
to incentives differently while in calm states and while
in emotion states, and across emotion states, and (2) a person
may have some power to change his or her emotional

7
disposition or to avoid stimuli that provoke certain
emotion states that increase the likelihood of harm to others.

II. A FRAMEWORK FOR ANALYZING LAW

AND THE EMOTIONS

A. Assumptions
The model of consumer choice assumes that
individuals have preferences in the sense that they are
able to rank the various states of the world that will
result from their choices. These preferences are consistent,
transitive, and complete. A utility function maps the
states of the world to a number, so that we say that an
individual's utility rises as he obtains increasingly
preferred states of the world. Individuals have certain
abilities, which enable them to engage in a particular
action at greater or less cost; and they have budgets that
limit the amount of resources that they can spend in
satisfying their preferences.
Economists usually assume that preferences
remain stable over the period of time relevant to analysis.
If individuals switch from labor to leisure after an
increase of the tax on their income, the usual explanation
is that the opportunity cost of leisure declines, not that
the tax increase happened to coincide with an exogenously
caused change in preferences, though of course the latter
is possible as a theoretical matter. The reason for this
convention is that economics has nothing to say about
how preferences change; so any economic, as opposed
to psychological, explanation for behavior must look
to other factors.
I depart from this model, though I hew as closely
as possible. I assume that under certain conditions a
"stimulus" will produce an "emotion state," during
which certain otherwise fixed attributes of the agent
change. These attributes then return to what they were
before the emotion state occurred. The attribute on
which I will focus is the preference. During the emotion
state, preferences reflect a higher ranking of world states
that are relevant to the emotion, or, as I will sometimes
say, the emotion-relevant good will become more

8
intensely preferred. In addition, abilities may change
during the emotion state: the agent may become more
or less able to convert inputs, like time and energy, into
outputs like labor or attack. And beliefs may change
during the emotion state: the emotion may cause the
agent to believe that an emotion-relevant probability is
higher than the probabilities he attaches to it during the
calm state (sometimes, the emotion-relevant probability
may be more accurate).
"Stimulus" refers to the whole set of environmental
or interpersonal conditions that, whether via cognitive
interpretation or not, set off an emotional reaction. An
insult coming from a stranger could stimulate anger, or
merely seeing someone who has offended the agent in
the past. The observation of sewage or rotting meat
could stimulate disgust. Receipt of an award might
stimulate pride or happiness. Learning about a rival's
success stimulates envy. Observation of a child's misery
might stimulate compassion or pity. In each case, some
aspect of the environment acts as a stimulus of the
emotional reaction. As should be clear from these
examples, emotional reaction to a stimulus is not necessarily,
or even usually, a reflex action but often involves
an appraisal of surrounding conditions, and all the
cognitive work that this term implies.
Our final assumption is that people can cultivate
their emotions. People are self-conscious and generally
(though not always) knowledgeable about their
emotional dispositions, recognize when these emotional
dispositions lead them astray, and take steps to modify
them. The idea of cultivating (benign) emotions or
even emotionless calm is similar to the simpler act of
avoiding stimuli of destructive emotions. A person
might seek to control his anger or envy through meditation,
yoga, religious pursuits, and so forth; or by avoidance: he
might stop going to bars where people slight him; he
might move away from cities or neighborhoods where
conspicuous consumption is the norm; he might avoid
homeless people. In these cases, the person avoids stimuli
of anger, envy, or pity that might get him trouble or
simply be unpleasant to experience. Cultivation of

9
emotional dispositions and simple avoidance are, of
course, very different behaviors, but they can be usefully
treated as the same for purpose of analysis.
Students of the emotions will notice that the
framework is better suited to some emotions (fear, disgust,
anger) than others, such as love and jealousy. The
"higher emotions" are more complex, have great cognitive
content, have less distinct physiological manifestations, are
less uniform across cultures, have fewer obvious correlates
in animals, and are - to get to the point - more poorly
understood by psychologists. Fortunately, they also seem
to play a smaller role in the law, and I will, for the most
part, avoid them.

B. Order of Actions

Our simple model of the emotions assumes a stylized

order of events, represented by the timeline in Figure 1:

Figure 1

I I I I
invest stimulus act end
- emotion state -

The first event is the agent's investment in either

cultivating an emotional disposition orvoiding condi-
tions that will stimulate the emotion. The second
event is the application of the stimulus that provokes
an emotion.
Next comes the emotion state. This refers to the
actual occurrence of the emotion. It may last for an
instant or for a very long time. For example, the agent
might be angered when someone pushes him from
behind, but the anger disappears almost immediately
when he sees that the person who pushed him had
tripped on his shoelaces. Envy might fester for months
or years; love and grief also may last for a very long time.
In any event, for purposes of analysis it is assumed that
the emotion ends at some definite point and the
non-emotional or "calm" state returns, which here may
be designated the "post-state" condition.

10
 During the emotion state the agent chooses to act,
just as he does before the state. The same set of choices
presents itself to the agent as before, except of course
that the change in environment that stimulates the
emotion might introduce new choices. Before the stimulus,
the agent has no desire to injure person P; after P insults
him, he now has such a desire, and may act on it. But he
may not. As noted earlier, the emotion state affects the
agent by changing his preferences so that the emotion-
relevant goods are more heavily weighted, and by
increasing or reducing his cost of action. Sanctions and
other costs thus continue to influence the agent's
choice; what is different is that the costs and benefits of
action differ, slightly or greatly, from what they were
during the calm state.
After the emotion state ends, the agent returns to
the calm state, which means that his preferences and
abilities return to what they were in the pre-emotion
state. But his actions during the emotion state will have
changed his endowments, so he may be better or worse
off than he was before the emotion state occurred.
The timeline highly simplifies even the simplest
emotions. One problem is that an emotional experience
can be relived, not just reimagined, and so an emotion
state will "tail off," and sometimes renew itself later in
time. I observe a disgusting slime; I withdraw and soon
forget about it; later the appearance and smell of the
syrup I am pouring on my pancakes reminds me of the
slime, and I am again disgusted. Similarly, my anger at
an insult renews itself every time I am reminded of the
insult. This complexity needs to be kept in mind.

C. Analysis
The agent's choice during the emotion state can
be represented using the standard graphical rendering of
consumer choice, as in Figure 2. The vertical axis represents
the "amount" of the emotion-relevant good (Y). If the
agent is angry, the axis represents the degree of injury
inflicted on the person at whom the agent is angry; if
the agent is disgusted, the axis represents the degree of
withdrawal from the object of disgust; if the agent is in

11
love, it represents some measure of benefit to the
beloved as experienced by the agent. The horizontal
axis represents all the other goods that the agent wants
to consume (X).

Figure 2

y
---- --- -------UE

B
.. UN

X
The graph depicts the agent's "normal" or calm-state
indifference curve (UN) and the corresponding
emotion-state indifference curve (UE). The emotion-state
indifference curve is flatter than the calm-state
indifference curve, representing the fact that in the
emotion state the agent's preferences for the emotion-
relevant good become more intense. The result is that
the agent will give up more of the ordinary goods, than
in the normal state, in return for a constant increase in
his consumption of the emotion-relevant goods. The
graphical rendering assumes that the budget line (B) is
the same in the calm and emotion state, though as
noted above it is possible that the budget lines would
diverge. The emotion-state budget line might shift out
or in: out, if the emotion makes action cheaper (anger);
in, if the emotion makes action more costly (grief). One
might assume that, say, anger makes all actions cheaper,
in which case the budget line shifts out in a uniform
manner. Alternatively, one might assume that the
emotion makes the emotion relevant good cheaper,
in which case the budget line tilts, as when it represents

12
a change in price.
When emotions make action cheaper, consumption
of the emotion-relevant good will increase, and
consumption of the other goods will either increase or
decrease. When emotions make action more costly,
consumption of the two kinds of good may rise or fall.
And even when emotions do not change the cost of
behavior, they do change preferences, so the pattern of
consumption will change. As a result, the agent's
emotional behavior will usually put him on a lower
indifference curve measured by calm state preferences
(the exception being when emotion reduces the cost of
action by a sufficient amount). Call this change the
emotion-provoked utility change (EPUC). A person
who lashes out at an enemy while angry might be better
off, with positive EPUC, after the emotion state returns,
because the enemy has removed himself, or because the
agent has developed a reputation for toughness; or
he may be worse off, with negative EPUC, because
the enemy struck back, or because the agent now finds
himself in jail.
A person in the initial calm state knows that his
emotional disposition might get him in trouble or benefit
him. This depends on the environment. A person with
an irascible disposition might do well in a shame society:
people avoid abusing him because they fear him. But he
might do poorly in a well-regulated society: a routine
altercation might spiral into a fight, landing the agent
in jail. The first person has a positive expected EPUC;
the second person has a negative expected EPUC.
(EPUC is defined only with reference to the emotion
under consideration: the same person might be easily
disgusted or not, and this might be beneficial in some
contexts but not others.) The second person can be
expected to "invest" to minimize his actual EPUC: by
cultivating a sense of calm, or by avoiding places, like
bars, where he is likely to be provoked.
It is important to emphasize that EPUC is determined
by the calm-state preferences, not the emotion-state
preferences, even though it is generated by actions
taken during the emotion state. This is necessary in

13
order to capture the idea that people in their normal
state are different from what they become in an
emotional state, and also that, anticipating this, people
will engage in certain actions in the normal state.
At the investment stage, the agent knows that he
may experience the stimulus with some probability. To
keep the analysis simple, let us suppose that during the
emotion state, preferences change but abilities and
beliefs do not. The difference between the agent's
preferences during the calm state and the emotion state
depend on the quality and intensity of the emotion,
which is reflected by the slope of the emotion state
indifference curve. As can be seen in Figure 3, the agent
will typically consume a different bundle of goods
during the emotion state and during the calm state, just
because of the change in preferences. This means that
after the emotion state has subsided and his "calm"
preferences have returned, the agent will be either
worse off (as in Figure 3) or better off than he had been
in the calm state. If he is worse off, then we can construct
a WTP measure of the difference by constructing a
hypothetical budget line in the usual manner. In Figure 3,
the agent would be willing to pay the amount depicted
in order to avoid the consumption results of the
emotion state.

Figure 3

l i

-UN

UN'

X
14
 Ex ante, the agent knows that the change in
consumption wrought by the emotional response will
occur with some probability, and so initial investment
will reflect this expected change. Investment in the current
example would be to (1) avoid the stimulus, or (2) cultivate
a less sensitive disposition. Sometimes a dollar of investment
will be relatively effective (switching from one bar to
another) and sometimes not (developing an ability to
turn the other cheek). The effectiveness of the investment
in this way will be denoted by the elasticity of the
emotional response with respect to the investment.
A high elasticity indicates greater responsiveness.
This allows us to make the following prediction:
People will invest more in cultivating dispositions
or avoiding stimuli as (1) the quality and intensity of the
emotion state increases relative to the calm state (that
is, as EPUC increases); and (2) the elasticity of the
emotion state with respect to the investment increases.
As an example, consider anger Anger can occur
in predictable and unpredictable ways. If the agent
already hates a person, then he knows that if he
encounters that person, there is a good chance that he
will become angry. By contrast, an agent cannot predict
when a stranger will offend him, and he might suppose
that his anger will on average be less from a random
incident involving a stranger than with someone he
already knows and hates. Thus, it makes more sense to
punish people who are provoked into anger by someone
they hate and could avoid, than to punish people who
are provoked into anger during a random encounter
with a stranger. The former are in a better position during
their calm state to avoid the encounter, and it is during
the calm state that they are more responsive.
Suppose, instead, that an agent knows about his
irascible disposition and has taken a number of steps to
change it. He attends anger management classes. He
avoids rowdy bars. He avoids drinking. Nonetheless, he
is provoked to anger and he injures the offender. It
again makes sense to reduce the sanction in order to
reward and encourage his behavior during the calm
state. By contrast, people who cultivate angry dispositions

15
should be punished.

III. EMOTION IN CRIMINAL LAW

It is sometimes said that a person who commits a

crime under the influence of emotion is less culpable
than a person who acts calmly and deliberately. The
standard example is of murder being reduced to
voluntary manslaughter if the defendant acted in an
uncontrollable rage. But the truth is more complex.
Certain emotion states reduce murder to voluntary
manslaughter, but others do not. For example, under
modem hate crime laws a person who kills a gay man
because he "hates" gays or is "disgusted" by gays will be
found more culpable than someone who kills the same
man just to get his wallet. This raises the question
whether the criminal law's different treatments of
emotion-influenced crime can be explained on
deterrence grounds.
There are two relevant considerations in a
deterrence approach to emotional crimes. The first
consideration is the extent to which an agent in an
emotion state is responsive to sanctions. The second
consideration is the extent to which an agent can be
deterred from being in the emotional state in situations
in which it may lead to harm.
Emotion state deterrability. A person in an emotion
state may be sensitive to sanctions. Deterrability
depends both on (1) the kind of the emotion and
(2) the intensity. In the emotion state a person experiences
a heightened preference, relative to the calm state, to
"consume" the emotion-relevant good. But the person
has other preferences as well. An angry person wants to
strike the offender, but many angry people restrain
themselves because they continue to want their freedom,
their job, their assets, and the other good things they
might lose if convicted of assault or murder. But even
putting this aside it is important to realize that most
emotions - anger is the conspicuous exception - do not
create a powerful preference to harm another person.
Disgust typically involves a preference to withdraw, not
to attack, and that is why disgust rarely excuses assault
16
or murder. Fear typically involves a preference to flee or
freeze, not to attack, though attacking may be a reaction
in some cases, and panic may lead to harm when the
agent is responsible for the well-being of an infant or
incompetent person. Although a disgusted or fearful
person's indifference curve will "tilt" dramatically, it
does not tilt in the direction of injuring another person;
indeed, that preference will become weak relative to the
desire to escape the stimulus.
In addition, a person's emotion state can be more
or less extreme relative to the calm state. A person can
be made more or less angry by a provocation. A mere
insult will, on average, anger a person less than an
unprovoked slap on the face. Graphically, an angry
person's indifference curve may shift out a lot or a little;
if a little, then his behavior will change correspondingly.
A slightly angry person will suppress the desire to lash
out if the consequences are more than trivial, and may,
if at all, indulge his preference by muttering or
complaining about the offender to a third party.
Both these observations explain why from a
deterrence perspective being in an emotion state is not
a sufficient condition for reduced culpability. A person
remains deterrable if, generally speaking, the emotion is
not anger, or the emotion is anger but it is not intense.
It makes sense to reduce the sanction for an act
committed during an emotion state if (1) the emotion is
anger (or another emotion that causes people to injure
others, like fear, possibly) and it is quite intense, so that
high expected sanctions cannot deter the behavior, and
(2) sanctions are costly, which they are when they
involve imprisonment. The justification is that expensive
sanctions should not be wasted on people who cannot
be deterred by them.
Pre-emotion (calm) state deterrability. Imagine
that a person in an emotion state is undeterrable. The
sanction cannot be made high enough to prevent him
from acting. It does not follow that he must be excused
from a crime that follows from the action tendency.
To see why, one must observe that a person in the
calm state often foresees the stimuli that will provoke

17
the emotion state and also his actions in the emotion
state. The calm state preferences might very well reflect
some things (a long-term desire to stay out of jail) and
not other things (an immediate desire to strike an
offender) that are the reverse in the emotion state.
Thus, acting on his calm state preferences the person
might take precautions against entering the emotion
state, including cultivating more peaceful emotions or
avoiding the stimulus, though he will not if he thinks
that his irascibility produces returns valued in the calm
state, including a reputation for toughness. In any
event, sanctions may be desirable for influencing
"investment" behavior during the calm state.
Reducing the sanction for emotional behavior,
relative to unemotional behavior, is justified even when
the sanction does not deter emotion state behavior
when (1) the calm state preferences and emotion state
preferences differ only a little (EPUC is low); (2) the
agent cannot avoid the stimulus or reduce the intensity
of his emotional reaction at relatively low cost (elasticity
is low); and (3) sanctions are relatively expensive (jail,
not fines). (Note that this puts aside the incapacitation
goal of imprisonment, which may justify imprisoning
highly dangerous people who engage in undeterrable
conduct.) The first factor essentially refers to the benefit
the agent obtains by avoiding the emotion state; the
second factor refers to the cost of avoiding the emotion
state. Even putting aside sanctions, we expect people
who very much dislike the preferences of their emotion-state
selves to take steps to avoid the emotion state unless
those steps are very costly. The criminal law must take
account of this first stage of action.
Table 1 summarizes the argument so far.

Table I
emotion-state deterrability
pre-emotion high low
state deterrability high punish
low do not punish

Table 1 shows that the strongest case for reducing

18
criminal punishment on deterrence grounds occurs
when the agent cannot be easily deterred in the emotion
state and cannot easily avoid the emotion state either
by avoiding the stimulus or changing his temperament.
The emotion has to be the right type (anger or fear, typically)
and it must be intense. When these conditions are not
met, the case for reducing the punishment is weaker.
Some examples. Murder is reduced to voluntary
manslaughter not just when the wrongdoer acts under
the influence of emotion. The emotion usually must be
anger, not (say) disgust. Anger causes people to attack
others; disgust causes them to withdraw. The anger must
be intense, not mild: a provocation that would excuse
murder in the heat of passion does not excuse murder
committed a while later. Distinctions regarding the
adequacy of provocation can be understood in terms of
the elasticity of investment. People who kill adulterous
spouses can hardly be expected to avoid the stimulus
(not get married in the first place?). But people who
look for trouble in bars, especially homophobes who
look for trouble in gay bars, could easily avoid provocation
that leads to murderous rage. Kahan and Nussbaum
want to put in the same category as murderous
homophobes, women who kill brutal husbands and men
who kill adulterous wives or girlfriends, so that they can
argue that the law's inconsistent treatment of
homophobes and these other wrongdoers contradicts
the deterrence argument. But a woman who kills a
drunken, brutal husband is unlikely to be a threat to the
rest of society, or even to future husbands or boyfriends
so long as they are not also drunken and brutal. A man
whose wife or girlfriend cheated on him cannot be
expected to refrain thereafter from any relationship with
another woman. These people are not generally as
dangerous as the person whose hatred causes him to kill
homosexuals, and it is less difficult for a homophobe to
avoid gay bars than for men and women to avoid having
relationships with each other. Of course, those among
the latter who are as dangerous - those who kill multiple
husbands or wives - are punished more seriously.
Similar comments apply to the case of fear. Louise

19
Woodward, who was hired as a nanny, was found guilty
of killing the infant in her care, but the trial judge
reduced a verdict of second-degree murder to
manslaughter on the ground that the evidence disclosed
"confusion, fright, a bad judgment, rather than rage or
malice." Supposing that Woodward really was panicking,
the explanation for reducing the conviction is that
sanctions do not deter people who are gripped by fear,
and although most panicking people do not attack others,
it is possible that a terrified caregiver might injure a
child. One might argue that severe sanctions would
deter fear-prone people from becoming nannies, but
fear-induced killing is so rare and bizarre that the
elasticity of investment with respect to fear must be
considered quite low. Should the convicted killer take
on a new position as nanny, however, and while panicking
kill that child as well, one anticipates a very high sanction.
She should have learned of her own unreliability, and
avoided situations in which it could harm third parties.
Not all emotions should reduce the sanction. Hate
crime laws increase, rather than, reduce sanctions, even
though the racist killer more plausibly acted under the
influence of emotion than, say, the contract killer. The
reason that hate results in greater punishment, while
anger often results in a lesser punishment, is that hate is
more susceptible to pre-emotion-state manipulation
than anger is. The person who hates members of a
particular race can often avoid interacting with them;
he can also take care not to carry weapons. The person
provoked to anger by adultery is not in a similar position.
The two emotions differ on the dimension of elasticity.
Turning now to the defense of duress, Kahan and
Nussbaum ask us to compare (1) a woman who commits
armed robbery to avoid being beaten up; (2) a woman
who fails to protect her child in order to avoid being
beaten up. Woman 1 has a duress defense; woman
2 does not. Kahan and Nussbaum say that this is
inconsistent with the narrow consequentialist view that
people who are susceptible only to extreme threats do
not pose a danger to others, so punishment would be
wasteful. But the law makes sense on deterrence
20
grounds. Sanctioning woman 1 is not likely to change
behavior because a person gripped by fear develops a
strong preference for self-preservation and ordinary
altruistic concerns diminish. But ordinary experience
suggests that altruism for one's own child remains salient
even for a scared person. When we are threatened, we run
away if alone but grab our children if they are with us. If
this is not true for our woman 2, then a heavy sanction
might offset her subjective gain, and it is no more
surprising that we punish her than that we punish anyone
who engages in child abuse. Kahan and Nussbaum
argue a consequentialist argument assumes that legal
decisionmakers "are prepared to endorse the first
woman's valuation of her own welfare but not the
second woman's." But this is not right. We want neither
woman to commit a crime: but we punish the first
woman less because a sanction is less likely to deter her.
The theory explains historical changes in the
treatment of cuckolds who murder spouses discovered in
flagrante, interracial murders, people who murder
homosexuals who proposition them, battered women
who murder husbands, and so forth. As society's evaluation
of the behavior changes, it increases or reduces
sanctions accordingly, in order to encourage people to
invest in proper emotional dispositions and to avoid
provocation otherwise, not just in the hope that the
prospect of sanctions will influence the agent while in
the emotion state. Anger and shame become less important
and desirable sources of social order as the legal system
modernizes; so over time, law will excuse less behavior
motivated by anger and shame, in the hope of encouraging
people to invest in the proper emotional dispositions.
A final point is that the criminal law should vary
from region to region if there is cross-cultural variation
in average emotional disposition. Nisbett and Cohen
provide experimental evidence that in the United
States southerners are more easily provoked to anger by
an insult than northerners are. If, as seems likely, the
elasticity of investment with respect to anger states is
low, then we would expect southern states to punish
crimes committed while angry and in response to

21
provocation, less than northern states, holding constant
the cost of imprisonment. Because southerners are less
likely to respond to sanctions while angry, already take
more care to avoid provocation, and are unlikely to be
able to modify dispositions that are mostly the result of
upbringing, scarce prison resources are better devoted to
deterring other crimes, than in then north.
Regret. A prominent feature of criminal sentencing
is the defendant's expression of regret or refusal to
express regret. If the expression is sincere, then it is
highly relevant to the purposes of criminal punishment,
for it reveals the distance between the defendant's calm
state preferences and emotion state preferences if the
crime occurred in an emotion state. If the distance is
great, then the deterrence goal requires a reduction in
the sentence, all else equal, because that suggests that
the defendant will when calm take steps to avoid the
emotion state (EPUC is high), so that criminal behavior
is unlikely to recur, and a severe sanction is not necessary
to deter further criminal behavior. Defendants who do
not express regret for criminal behavior are dangerous
because their calm state preferences are antisocial, and
they should be punished either for the purpose of deterrence
(if they are deterrable) or incapacitation (if they are not
deterrable but are sufficiently dangerous).
A note on tort. The analysis of emotion in tort law
is similar to the analysis for criminal law. Suppose that
an individual must drive to work every day but is subject
to, and knows himself subject to, road rage when other
drivers cut him off. While under the influence of road
rage, the individual drives much more carelessly than
when he is calm; for example, he might drive more
quickly and pay less attention. Formally, his emotion
state preference to drive quickly is more intense than his
calm state preference to drive quickly (or, alternatively, the
cost of driving carefully increases). For concreteness,
suppose that the individual's cost of driving carefully is
$100 when enraged, but the cost is only $10 when the
individual is calm. Further, suppose that the social cost
of driving quickly is $50 in expected accident costs
imposed on third parties. A negligence rule, technically
22
applied, would excuse the enraged driver from liability,
whereas strict liability would require him to pay $50 in
expected terms. Strict liability is the superior rule, for it
would encourage the calm state person, who is stuck
with the $50 liability imposed on him by his emotion
state self, to switch to public transportation or avoid
congested streets where road rage is more likely
provoked. Alternatively, the individual could be held
negligent for not taking public transportation when he
knows himself subject to road rage, just as an epileptic is
held liable for driving with knowledge of his epilepsy.
The latter alternative achieves the same result as strict
liability by counterfactually attributing calm state
abilities and preferences to the emotion state self that
causes the accident.

IV. CosT-BENEFIT ANALYSIS AND RISK REGULATION

In a recent article Timur Kuran and Cass Sunstein

argue that various cognitive biases can lead to panicked
government action that is socially harmful. They focus
on the availability heuristic, "a pervasive mental shortcut
whereby the perceived likelihood of any given event is
tied to the ease with which its occurrence can be
brought to mind." Because people rely on the availability
heuristic and other cognitive shortcuts, they fear risks
not solely on the basis of the expected harm, but on the
basis of strictly irrelevant factors such as the degree to
which the risks are subject to public discussion, the
harm is vivid, the technology is new, the results are
irreversible, and so forth. In the meantime, dissenting
views are ignored, dissenters are treated with opprobrium,
and so dissent is suppressed. As a consequence, initially
exaggerated beliefs about risks can feed on each other,
resulting in an "availability cascade," in which a minor
risk is blown up into a major threat. Kuran and
Sunstein's main examples are the Love Canal scare, the
Alar scare, and the response to the crash of TWA flight
800. In all these cases, a relatively trivial risk or benign
explanation is overtaken by events, resulting in a panic
in which the worst case is widely believed.
A notable aspect of Kuran and Sunstein's explanation

23
is that it is purely cognitive. People are assumed to
be rational except to the extent that cognitive biases
interfere with rational choice. Their argument predicts
that availability cascades would occur even in a
hypothetical world in which people had no emotions.
But there are some reasons for being skeptical of
this view. Start with the obvious semantic point that
Love Canal, and similar events, are often called "scares"
or "panics." Kuran and Sunstein's account of the Love
Canal affair repeatedly notes the emotional character of
the events: "frightening tales spread quickly"; "residents
feared..." various dangers; laIngered by claims in" a
newspaper, one resident organized a petition drive; a
woman at a meeting "started crying hysterically"; "[hier
anxieties and fears came to be shared by the entire
community"; one activist was "furious" ... "Now very
emotional, I said: "You can't do that! That would be
murder!"; "fears would not dissipate"; and so forth.
Although Kuran and Sunstein could argue that the
emotional responses were epiphenomenal, and that
cognitive bias fully explains the phenomenon, an
emotion-based theory cannot be discounted. Most
panics, as the term suggests, are characterized by fear.
One influential psychological theory of fear holds
that it is an evolutionarily adaptive mechanism for pro-
cessing information about potential threats to the
organism. It is evolutionarily more costly for an organism
to fail to respond to a threat than it is for the organism
to respond incorrectly. That is, you are less likely to survive
and reproduce if you sometimes neglect to flee from a
tiger than if you occasionally flee from a shadow that
looks like a tiger. Certain genetically determined or
learned stimuli (snakes, and so forth) are processed in a
rough way by a part of the brain before being subject to
cognition, a bit like the reflex action of withdrawing
one's hand from a hot object. Thus, fear (and anxiety as
well) has benefits and costs: it allows the organism to
respond quickly to a threat, but it also appears to cause
the organism to overpredict the probability of the
threat. In our framework, fear results in distorted beliefs
but also lower costs.

24
 The connection between an individual's fear and a
social panic occurs as follows. First, fear is contagious. A
person who observes that another person is fearful is
more likely to become fearful than a person who does
not make such an observation. This may, of course, be
rational in a sense; but it also seems to be a purely
psychological or even physiological response. Second,
fear feeds on itself: observing that another is fearful,
one's own fear may become heightened. Third, fear,
unlike some other emotions such as anger, can persist
for quite a long time, sometimes in the form of a general
anxiety. Fearful and anxious people continue to overestimate
the probability of a threat, and to reinforce each others'
incorrect beliefs.
The emotion perspective can be seen as a supplement
to Kuran and Sunstein's cognitive theory, which itself
indeed can be seen as a supplement to a rational choice
view based on information imperfections. It is rational
to infer that a threat exists if others believe that a threat
exists. This inference may be strengthened by the
availability heuristic, a cognitive bias. And the inference
may be strengthened further by the emotion of fear,
which leads the fearful individual to exaggerate (relative
to calm-state beliefs) not any widely publicized risk (as
suggested by the availability heuristic), but in particular
those risks that pose threats that engage the fear-centered
portions of the brain.
This last observation leads to the question of
normative implications. Kuran and Sunstein's analysis
of availability cascades lead them to call for greater
insulation of decisionmakers, greater reliance on scientists
and other experts, and greater use of analytic tools like
cost-benefit analysis. These are surely appropriate
responses when panics are the result of fear as well
as cognitive biases.
But it also would be useful to know whether different
panics have different explanations, and therefore justify
different government responses. A bank panic, for
example, is calm-state rational. Given that the bank's
liabilities exceed its liquid assets, I am rational to withdraw
my money if I believe others will withdraw their money.

25
This is a classic n-player prisoner's dilemma. Bank panics
would occur even if people never made cognitive errors
or experienced intense emotions. This may explain why
bank panics are so easily ended: the government just
needs to make a credible promise to reimburse all depositors.
By contrast, the preference for driving over flying
appears to be emotion-driven. I conjecture that people
are more anxious about flying than about driving
because when danger provokes fear, driving does not
prevent one from fleeing (one pulls over to the shoulder)
but flying does (one is stuck in the tube).
A frequent criticism of aggressive airline safety
regulation is that costly precautions cause people to
substitute to automobiles for short trips, which in fact
are more dangerous, so certain airline safety regulations
may increase fatalities rather than reduce them. A common
response is that people's subjective discomfort with airline
travel should be counted as a social cost. Although
airline safety regulation may increase aggregate fatalities
as people substitute to automobiles because of cost, it may
still increase social welfare as the travelers experience
less emotional discomfort.
The problem with this response is that fear, dread,
and anxiety are not sensitive to changes in low
probabilities. A person who feels an overpowering feeling
of dread while riding on an airplane will feel the same
amount of dread whether the probability of an accident
is 1,fin 100,000 or 1 in 1,000,000. So the safety precaution
does not increase that person's subjective well-being. If
the person rationally anticipates his dread, he will treat
it as a fixed cost regardless of the probability of accident
(within constraints, obviously), so his determination
whether to fly or drive will depend entirely on the
pecuniary costs. This observation rescues the standard
criticism of safety regulation from the economic
response: subjective (emotional) discomfort should not
play a critical role in safety regulation.
As another example of what is at stake, consider
the problem of genetically altered food, which is widely
avoided in Europe. On the rational herding view,
individuals in Europe rationally avoid this food because

26
others avoid this food, and each person, individually
rationally but collectively disastrously, infers that every
person acts on statistically independent pieces of
information. The reason that this story is probably false
is that in the herd behavior models, the suboptimal
equilibria are fragile. If genetically altered food poses
risks no greater than those that people already face, and
if people were rational, then the panic would not have
lasted as long as it has.
There is probably some truth to the cognitive view
of this panic. Concerns about genetically altered food
probably stem from the mad cow disease problem in
England, which was a highly salient event. But we know
that emotional reactions often interfere with cognition.
The argument in favor of genetic alteration of food,
which is that it is not much different from breeding that
has occurred for thousands of years, does not stick in
people's minds, and it may be that it does not for the
same reason that a person who sees a snake in a cage
jumping at him "forgets," or simply does not process, the
fact that the snake is behind a pane of glass, and that a
person will not eat a cockroach that has been sterilized.
This argument requires a jump from the primal fear and
disgust responses, to a more complex, culturally mediated
fear/disgust response, about which little is know. But it
is a reasonable hypothesis.
The cognitive and emotion views leave a lot
unexplained. If the British had all become vegetarians
in response to mad cow disease, the views could have
rationalized that response as well. But a similar objection
can be made to the rational choice argument as well.
These methodological difficulties do not alter the
fact it matters which view is correct. To see why, consider
the question over whether it is appropriate for automobile
manufacturers to install switches that allow consumers
to deactivate air bags. Reluctance about permitting
these switches stems from fears that publicity about the
dangers that air bags pose to children and small adults
may lead people to deactivate air bags even when doing
so enhances the net danger to passengers.
The proper regulatory action depends on the

27
source of people's false beliefs about air bags. If the
incorrect belief that air bags are on net dangerous results
from rational ignorance, the proper response is simply to
post labels explaining that people should activate the
air bags unless a child or small adult is in the front seat.
If the belief is the result of a cognitive bias like the
availability heuristic, the label may be an insufficient
response. Perhaps, a great deal of advertising would be
necessary in order to shift people's beliefs; or, alternatively,
nothing can shake people's beliefs and the best response
is to make it impossible to deactivate the air bags.
If the belief is the result of fear or panic, one or
more of the following responses might be appropriate.
(1) Permit installation of the switch, because emotion
states tend to weaken and then disappear as time passes.
As people become accustomed to the presence of air bags,
they will be less likely to deactivate them inappropriately.
(2) Prohibit installation of the switch, because emotions
are, even more than cognitive biases, impervious to
reasoning. When panicked people trample each other to
escape a sinking ship, no amount of reasoning will save
them. (3) Permit installation, and fight emotion with
emotion. When people are afraid, make them feel
ashamed. When they are angry, make them feel
pity. Lawyers, politicians, advertisers, and many other
professionals are skilled in the tricky art of manipulating
people's emotions. Sappy commercials have already
been used to encourage the use of seat belts by cleverly
framing seatbelt use as an obligation that people owe to
their loved ones rather than as a means of protecting
themselves. Similarly, sappy commercials could be used
to show that air bag use is necessary to protect loved
ones (unless they are small adults or children). Without
expressing any opinion on whether such commercials,
sappy or not, are effective, I will just note that Madison
Avenue as well as the government usually run vivid,
emotionally stimulating commercials, more so than dry
recitations of facts.

28
V. BARGAINING, CONTRACT LAW, AND PROPERTY
AND LIABILITY RULEs: A CRITIQUE OF SPECIFIC
PERFORMANCE

Much economic analysis of property rights

assumes that when transaction costs are low, parties can
bargain around inefficient allocations of property rights.
"Transaction cost" is an ambiguous term, but it is generally
limited to situations in which bargaining would involve
large numbers of people, or people who do not know
each other's identities. A theme of the literature is that
when transaction costs are low, it is more important for
courts to ensure that property rights are clear and so easily
tradable, than assigned to the more efficient user,
because the ability to trade ensures that the property
right will end up where it is most valued.
Ward Farnsworth informally tested this view by
interviewing lawyers who represented individuals
involved in nuisance suits. Lawyers reported that
plaintiffs who obtained judgments in nuisance cases
never sold their rights to the defendants. The most
plausible reason emerges from the descriptions of the
cases. "They [the parties] hate each other." A lawyer's
client's reaction to an offer "would have been to say 'not
only "No" but "Hell, no.""' There was "too much bad
blood." They "disliked each other intensely." No doubt
the disputes followed the same pattern. The plaintiff
was angered by the nuisance, which was considered
unneighborly. The defendant was angered by the plaintiff's
complaints and lawsuit. Irritation became enmity as the
usual indignities of litigation ensued. By the end, anger
and hatred were so intense that (in Farnsworth's view)
not even mutually beneficial trade would have
been possible.
These case studies conform to the common sense
intuition that anger interferes with bargaining. These
intuitions are consistent with theoretical and empirical
work on anger. Anger typically occurs when another
person offends or insults the agent, which usually means
violation of a recognized norm of respect. An angry person's
action tendency is to harm the offender, even at cost to

29
 oneself. When two sides of a dispute are angry with each
other, they may both do better (in subjective utility
terms) by inflicting costs on each other in the form of
continuing litigation, than by settling. If each party
gains more by injuring the other than by being injured
himself, this game is not a prisoner's dilemma: harming
the other person is both the dominant strategy for both
parties and the welfare-maximizing equilibrium. At the
same time, a person's ex ante calm preference is to avoid
fighting and to trade when there are (non-spite-related)
gains from doing so. The calm state and emotion state
preferences are in conflict, and if the former are to be
privileged (on which, see below), then the emotion
state preferences must be discounted.
These considerations suggest that the assignment
and protection of property rights should be sensitive to
the emotional valence of disputes. Consider the
well-known defense of specific performance. If the value
of performance exceeds the cost, the promisor ought to
perform, and specific performance (unlike expectation
damages) does not require an error-prone judicial
determination of the promisee's valuation. If the value
of performance is less than the cost, the promisor can
bribe the promisee to release him from the contract.
The advantage of the damages remedy is that it enables
the promisor to breach efficiently without negotiating;
but given low transaction costs this advantage has little
value, and transaction costs are likely to be low when
there are only two parties. Thus, specific performance is
superior to damages.
The problem with this argument is that the
promisor's breach may provoke the promisee's anger,
and the promisee - feeling insulted - will refuse to
release the promisee from the contract. This is very
much in the spirit of Farnsworth's interviews. This
potential refusal might not matter if the promisor tries
to buy out the promisee rather than breaching first, but
there are two further points of concern. First, it is hard to
say whether even broaching the subject of nonperformance
would not elicit an angry reaction, which would then
interfere with negotiations. Second, negotiations would

30
in any event occur in the shadow of the promisee's
angry reaction to breach.
Imagine, then, that an exogenous event causes a
seller's cost to exceed the buyer's valuation. Under specific
performance, if the seller breaches, she must either
perform or pay the buyer's entire valuation. But if the
buyer is angry, the buyer will incur a cost to spite the
seller, and this means that the buyer may insist on
performance even though he values it less than a
proposed payment equal to the entire surplus.
Anticipating this, the seller performs rather than
breaches. By contrast, under expectation damages the
seller would breach and pay damages equal to the
buyer's valuation, with the inefficient trade avoided.
Expectation damages is superior to specific performance
because the plaintiff cannot prevent an efficient trade in
order to spite the victim.
An important premise of this argument is that the
buyer's calm state preferences matter while his anger
state preferences do not matter. If the buyer's anger state
preferences matter, only specific performance can
protect them. Now, ex ante the buyer most likely would
prefer that his anger state preferences be discounted. If
that view is entitled to respect, then courts should
award expectation damages; otherwise, the seller will
pass the costs on to the buyer and the buyer will be
forced, against his calm state preferences, to purchase
the right when angry to spite the buyer! To be sure,
some buyers might want this right, but this seems
unusual. As long as calm state preferences are to be
counted, and buyers want their calm state preferences to
be counted, expectation damages is superior to specific
performance when breach would lead to anger.
Notice that emotion introduces an asymmetry
into the standard analysis of contract remedies. The person
who breaches will not feel any anger, while the victim
of the breach is likely to become angry. The reason is that
the contract sets up a background norm of performance,
and the person who breaks that norm injures the person
who complies with it. If it is not desirable for trade to
occur, which will often be the case, expectation damages
31
have the virtue of giving the property right to the calm
person, whereas specific performance has the defect of
giving the property right to the angry person. If it is
desirable for trade to occur, the choice of remedy does
not matter. Taking ex ante calm-state preferences as the
normative baseline, specific performance reduces the
value of contracts.
Further factors might bear on the choice between
remedies in specific situations. Anger subsides rapidly
when the agent realizes that the offensive act was
unintentional or in some way justified: if you push me, I
might get angry, but my anger dissolves when you
explain that you tripped and grabbed at me to prevent
yourself from falling, or that you were running away
from a mugger. Similarly, anger - and thus specific
performance - may pose fewer problems when a breach
is excusable rather than willful. It is notoriously difficult
to distinguish excusable from willful breaches in contract
cases, but one possible approach is to distinguish
breaches when the trade is ex post inefficient (the
promisor's cost exceeds the buyer's valuation) and
breaches when the trade is ex post efficient. The former
can only be an attempt to hold up the promisee for
a higher payment, and so seem willful; the latter are
at least jointly value-maximizing, and so somewhat
less objectionable.
It is worth lingering on the subtleties involved in
imposing legal sanctions against a background regime of
nonlegal sanctions. If, as Robert Frank argues, emotions
evolve to enable people to make credible their threat to
retaliate if another person breaks a promise, then this
adaptation constrains the ability of the law to serve
the same purpose. Frank argues that people restrain
themselves from breaking promises and engaging in
other opportunistic behavior in part because they fear
retaliation. Retaliation is not always individually rational
but may be compelled by emotion. Thus, emotion solves
a problem of cooperation. But if emotion causes people to
retaliate without carefully weighing the costs and benefits,
people will retaliate even when breach is jointly maximizing
and both sides would do better if the promisor compensated
32
the promisee but did not perform. Giving the enraged
promisee the tool of specific performance enables him
to exact retaliation when, given the availability of a
more modest legal remedy like expectation damages,
both sides would do better (as measured by calm state
preferences) if retaliation did not occur.
The normative implication of this discussion is
inconsistent with current law. Specific performance and
harsh remedies like punitive damages should not be
available when the promisor's breach is egregious and
thus likely to lead to an emotional response from the
promisee. Expectation damages are more appropriate.
The same is true for nuisance suits: money damages is
the superior remedy when the nuisance creates anger
and hard feelings, assuming away the cost of determining
damages. Injunctions are more suitable when emotion is
unlikely to interfere with bargaining.

CONCLUSION

Emotions matter for legal theory because much

behavior occurs under the influence of emotion. The
framework I propose is a simple alteration of the
standard rational choice approach, under which it is
assumed that an individual's preferences, abilities, and
beliefs predictably change during emotion states, and
change back afterwards. This approach allows us to
understand how individuals take steps to manipulate
their own emotional behavior (and others' as well,
though I have not emphasized this), and how the law
can influence both emotional behavior and behavior
that can be taken in anticipation of emotion.

33
Occasional Papers from
The University of Chicago Law School
1111 East 60th Street
Chicago, Illinois 60637

No. 1. "A Comment on Separation of Power"

Philip B. Kurland, November 1, 1971.
No. 2. "The Shortage of Natural Gas"
Edmund W. Kitch, February 1, 1972.
No. 3. "The Prosaic Sources of Prison Violence"
Hans W. Mattick, March 15, 1972.
No. 4. "Conflicts of Interest in Corporate Law Practice"
Stanley A. Kaplan, January 10, 1973.
No. 5. "Six Man Juries, Majority Verdicts-
What Difference Do They Make?"
Hans Zeisel, March 15, 1973.
No. 6. "On Emergency Powers of the President:
Every Inch a King?"
Gerhard Casper, May 31, 1973.
No. 7. "The Anatomy of Justice in Taxation"
Walter J. Blum and Harry Kalven Jr.,
October 1, 1973.
No. 8. "An Approach to Law"
Edward H. Levi, October 15, 1974.
No. 9. "The New Consumerism and the Law School"
Walter J. Blum, February 15, 1975.
No. 10. "Congress and the Courts"
Carl McGowan, April 17, 1975.
No. 11. "The Uneasy Case for Progressive Taxation
in 1976"
Walter J. Blum, November 19, 1976.
No. 12. "Making the Punishment Fit the Crime:
A Consumers' Guide to Sentencing Reform"
Franklin E. Zimring, January 24, 1977.
No. 13. "Talk to Entering Students"
James B. White, August 15, 1977.
No. 14. "The Death Penalty and the Insanity Defense"
Hans Zeisel, April 15, 1978.
No. 15. "Group Defamation"
Geoffrey R. Stone, August 10, 1978.
34
No. 16. "The University Law School and Practical
Education"
Carl McGowan, December 20,1978.
No. 17. "The Sovereignty of the Courts"
Edward H. Levi, July 15, 1981.
No. 18. "The Brothel Boy"
Norval Morris, March 15, 1982.
No. 19. "The Economists and the Problem of Monopoly"
George J. Stigler, July 1, 1983.
No. 20. "The Future of Gold"
Kenneth W. Dam, July 15, 1984.
No. 21. "The Limits of Antitrust"
Frank H. Easterbrook, April 15, 1985.
No. 22 "Constitutionalism"
Gerhard Casper, April 6, 1987.
No. 23. "Reconsidering Miranda"
Stephen J. Schulhofer, December 15, 1987.
No. 24. "Blackmail"
Ronald H. Coase, November 14, 1988.
No. 25. "The Twentieth-Century Revolution in
Family Wealth Transmission"
John H. Langbein, December 8,1989.
No. 26. "The State of the Modem Presidency:
Can It Meet Our Expectations?"
Stuart E. Eizenstat, March 10, 1990.
No. 27. "Flag Burning and the Constitution"
Geoffrey R. Stone, May 1, 1990.
No. 28. "The Institutional Structure of Production"
Ronald H. Coase, May 15, 1992.
No. 29. "The Bill of Rights: A Century of Progress"
John Paul Stevens, December 1, 1992.
No. 30. "Remembering 'TM'
Elena Kagan and Cass R. Sunstein, June 8, 1993.
No. 31. "Organ Transplantation: Or, Altruism
Run Amuck"
Richard A. Epstein, December 1, 1993.
No. 32. "The Constitution in Congress: The First
Congress, 1789-1791"
David P. Currie, June 15, 1994.

35
No. 33. "Law, Diplomacy, and Force: North Korea and
the Bomb"
Kenneth W. Dam, December 15, 1994.
No. 34. "Remembering Nuremberg"
Bernard D. Meltzer, December 20, 1995.
No. 35. "Racial Quotas and the Jury"
Albert W. Alschuler, February 20, 1996.
No. 36. "The Restructuring of Corporate America"
Daniel R. Fischel, June 20, 1996.
No. 37. "Constitutional Myth-Making: Lessons from the
Dred Scott Case"
Cass R. Sunstein, August 26, 1996.
No. 38. "The Role of Private Groups in Public Policy:
Cryptography and the National Research
Council"
Kenneth W. Dam, January 15, 1997.
No. 39. "Impeachment and Presidential Immunity
from Judicial Process"
Joseph Isenbergh, November 11, 1998.
No. 40. "Against Cyberanarchy"
Jack L. Goldsmith, August 13, 1999.
No. 41 "Contest and Consent:
A Legal History of Marital Rape"
Jill Elaine Hasday, May 2000.
No. 42 "Law and the Emotions"
Eric A. Posner, September 2001.

Copies of Occasional Papers from the Law School are

available from William S. Hein & Company, Inc.,
1285 Main Street, Buffalo, New York 14209, to whom
inquiries should be addressed. Current numbers are also
available on subscription from William S. Hein &
Company, Inc.

36