You are on page 1of 57

mr05_Article 1 3/28/05 3:25 PM Page 9

Journal of Economic Literature

Vol. XLIII (March 2005), pp. 9–64

Neuroeconomics: How Neuroscience

Can Inform Economics

Who knows what I want to do? Who knows what anyone wants to do? How can you
be sure about something like that? Isn’t it all a question of brain chemistry, signals
going back and forth, electrical energy in the cortex? How do you know whether
something is really what you want to do or just some kind of nerve impulse in the
brain. Some minor little activity takes place somewhere in this unimportant place in
one of the brain hemispheres and suddenly I want to go to Montana or I don’t want
to go to Montana. (White Noise, Don DeLillo)

1. Introduction such as finance, game theory, labor econom-

ics, public finance, law, and macroeconomics
In the last two decades, following almost a
(see Colin Camerer and George Loewenstein
century of separation, economics has begun
2004). Behavioral economics has mostly been
to import insights from psychology.
informed by a branch of psychology called
“Behavioral economics” is now a prominent
“behavioral decision research,” but other
fixture on the intellectual landscape and has
cognitive sciences are ripe for harvest. Some
spawned applications to topics in economics,
important insights will surely come from neu-
roscience, either directly or because neuro-

Camerer: California Institute of Technology. science will reshape what is believed about
Loewenstein: Carnegie Mellon University. Prelec: psychology which in turn informs economics.
Massachusetts Institute of Technology. We thank partici- Neuroscience uses imaging of brain activity
pants at the Russell Sage Foundation-sponsored confer-
ence on Neurobehavioral Economics (May 1997) at and other techniques to infer details about
Carnegie Mellon, the Princeton workshop on Neural how the brain works. The brain is the ultimate
Economics December 8–9, 2000, and the Arizona confer- “black box.” The foundations of economic
ence in March 2001. This research was supported by NSF
grant SBR-9601236 and by the Center for Advanced Study theory were constructed assuming that details
in Behavioral Sciences, where the authors visited during about the functioning of the brain’s black box
1997–98. David Laibson’s presentations have been partic- would not be known. This pessimism was
ularly helpful, as were comments and suggestions from the
editor and referees, and conversations and comments from expressed by William Jevons in 1871:
Ralph Adolphs, John Allman, Warren Bickel, Greg Berns,
Meghana Bhatt, Jonathan Cohen, Angus Deaton, John
Dickhaut, Paul Glimcher, Dave Grether, Ming Hsu, David I hesitate to say that men will ever have the means
Laibson, Danica Mijovic-Prelec, Read Montague, Charlie of measuring directly the feelings of the human
Plott, Matthew Rabin, Antonio Rangel, Peter Shizgal, heart. It is from the quantitative effects of the
Steve Quartz, and Paul Zak. Albert Bollard, Esther feelings that we must estimate their comparative
Hwang, and Karen Kerbs provided editorial assistance. amounts.

mr05_Article 1 3/28/05 3:25 PM Page 10

10 Journal of Economic Literature, Vol. XLIII (March 2005)

Since feelings were meant to predict behav- drug consumption limits pleasure from
ior but could only be assessed from behavior, future consumption of other goods (dynam-
economists realized that, without direct ic cross-partial effects in utility for commod-
measurement, feelings were useless inter- ity bundles) and how environmental cues
vening constructs. In the 1940s, the con- trigger unpleasant craving and increase
cepts of ordinal utility and revealed demand. These effects can be approximated
preference eliminated the superfluous inter- by extending standard theory and then
mediate step of positing immeasurable feel- applying conventional tools (see Douglas
ings. Revealed preference theory simply Bernheim and Antonio Rangel 2004; David
equates unobserved preferences with Laibson 2001; Ted O’Donoghue and
observed choices. Circularity is avoided by Matthew Rabin 1997).
assuming that people behave consistently, The radical approach involves turning
which makes the theory falsifiable; once back the hands of time and asking how eco-
they have revealed that they prefer A to B, nomics might have evolved differently if it
people should not subsequently choose B had been informed from the start by insights
over A. Later extensions—discounted, and findings now available from neuro-
expected, and subjective expected utility, science. Neuroscience, we will argue, points
and Bayesian updating—provided similar to an entirely new set of constructs to under-
“as if ” tools which sidestepped psychological lie economic decision making. The standard
detail. The “as if ” approach made good economic theory of constrained utility maxi-
sense as long as the brain remained substan- mization is most naturally interpreted either
tially a black box. The development of eco- as the result of learning based on consump-
nomics could not be held hostage to tion experiences (which is of little help when
progress in other human sciences. prices, income, and opportunity sets
But now neuroscience has proved change), or careful deliberation—a balanc-
Jevons’s pessimistic prediction wrong; the ing of the costs and benefits of different
study of the brain and nervous system is options—as might characterize complex
beginning to allow direct measurement of decisions like planning for retirement, buy-
thoughts and feelings. These measurements ing a house, or hammering out a contract.
are, in turn, challenging our understanding Although economists may privately acknowl-
of the relation between mind and action, edge that actual flesh-and-blood human
leading to new theoretical constructs and beings often choose without much delibera-
calling old ones into question. How can the tion, the economic models as written invari-
new findings of neuroscience, and the theo- ably represent decisions in a “deliberative
ries they have spawned, inform an econom- equilibrium,” i.e., that are at a stage where
ic theory that developed so impressively in further deliberation, computation, reflec-
their absence? tion, etc. would not by itself alter the agent’s
In thinking about the ways that neuro- choice. The variables that enter into the for-
science can inform economics, it is useful to mulation of the decision problem—the pref-
distinguish two types of contributions, which erences, information, and constraints—are
we term incremental and radical approach- precisely the variables that should affect the
es. In the incremental approach, neuro- decision, if the person had unlimited time
science adds variables to conventional and computing ability.
accounts of decision making or suggests spe- While not denying that deliberation is part
cific functional forms to replace “as if ” of human decision making, neuroscience
assumptions that have never been well sup- points out two generic inadequacies of this
ported empirically. For example, research on approach—its inability to handle the crucial
the neurobiology of addiction suggests how roles of automatic and emotional processing.
mr05_Article 1 3/28/05 3:25 PM Page 11

Camerer, Loewenstein, and Prelec: Neuroeconomics 11

First, much of the brain implements role in economics and other social sciences
“automatic” processes, which are faster than (e.g., law, Terrence Chorvat, Kevin McCabe,
conscious deliberations and which occur and Vernon Smith 2004). Indeed, a new area
with little or no awareness or feeling of effort of economics that has been branded “neu-
(John Bargh et al. 1996; Bargh and Tanya roeconomics” has already formed the basis
Chartrand 1999; Walter Schneider and of numerous academic gatherings that have
Richard Shiffrin 1977; Shiffrin and brought neuroscientists and economists
Schneider 1977). Because people have little together (see also Paul Zak, in press).1
or no introspective access to these processes, Participating in the development of a shared
or volitional control over them, and these intellectual enterprise will help us ensure
processes were evolved to solve problems of that the neuroscience informs economic
evolutionary importance rather than respect questions we care about. Stimulated by the
logical dicta, the behavior these processes authors’ own participation in a number of
generate need not follow normative axioms such meetings, our goal in this paper is to
of inference and choice. describe what neuroscientists do and how
Second, our behavior is strongly influ- their discoveries and views of human behav-
enced by finely tuned affective (emotion) ior might inform economic analysis. In the
systems whose basic design is common to next section (2), we describe the diversity of
humans and many animals (Joseph LeDoux tools that neuroscientists use. Section 3
1996; Jaak Panksepp 1998; Edmund Rolls introduces a simplified account of how cog-
1999). These systems are essential for daily nition and affect on the one hand, and auto-
functioning, and when they are damaged or matic and controlled processes on the other,
perturbed, by brain injury, stress, imbal- work separately, and interact. Section 4 dis-
ances in neurotransmitters, or the “heat of cusses general implications for economics.
the moment,” the logical-deliberative sys- Section 5 goes into greater detail about
tem—even if completely intact—cannot implications of neuroeconomics for four top-
regulate behavior appropriately. ics in economics: intertemporal choice, deci-
Human behavior thus requires a fluid sion making under risk, game theory, and
interaction between controlled and automat- labor-market discrimination. Through most
ic processes, and between cognitive and of the paper, our focus is largely on how neu-
affective systems. However, many behaviors roscience can inform models of microfoun-
that emerge from this interplay are routine- dations of individual decision making.
ly and falsely interpreted as being the prod- Section 6 discusses some broader macro
uct of cognitive deliberation alone (George implications and concludes.
Wolford, Michael Miller, and Michael
Gazzaniga 2000). These results (some of 2. Neuroscience Methods
which are described below) suggest that Scientific technologies are not just tools
introspective accounts of the basis for choice scientists use to explore areas of interest.
should be taken with a grain of salt. Because New tools also define new scientific fields
automatic processes are designed to keep and erase old boundaries. The telescope cre-
behavior “off-line” and below consciousness, ated astronomy by elevating the science from
we have far more introspective access to
controlled than to automatic processes. 1
The first meeting was held at Carnegie Mellon in
Since we see only the top of the automatic 1997. Later meetings were held in Arizona and
iceberg, we naturally tend to exaggerate the Princeton, in 2001, in Minnesota in 2002, and on
importance of control. Martha’s Vineyard in 2003 and Kiawah Island in 2004.
Occasional sessions devoted to this rapidly growing topic
Neuroscience findings and methods will are now common at large annual meetings in both
undoubtedly play an increasingly prominent economics and neuroscience.
mr05_Article 1 3/28/05 3:25 PM Page 12

12 Journal of Economic Literature, Vol. XLIII (March 2005)

pure cosmological speculation. The micro- resolution (on the order of one millisecond)
scope made possible similar advances in biol- and is the only method used with humans
ogy. The same is true of economics. Its that directly monitors neural activity, as
boundaries have been constantly reshaped by opposed to, e.g., blood flow. But spatial reso-
tools such as mathematical, econometric, and lution is poor and it can only measure activi-
simulation methods. Likewise, the current ty in the outer part of the brain. EEG
surge of interest in neuroscience by psychol- resolution has, however, been improving
ogists emerged largely from new methods, through the use of ever-increasing numbers
and the methods may productively blur the of electrodes. Interpolation methods, and the
boundaries of economics and psychology. combined use of EEG and fMRI for measur-
This section reviews some of these methods. ing outer-brain signals and inner-brain sig-
nals at the same time, promise to create
2.1 Brain Imaging
statistical methods which make reasonable
Brain imaging is currently the most popu- inferences about activity throughout the
lar neuroscientific tool. Most brain imaging brain from EEG signals. For economics, a
involves a comparison of people performing major advantage of EEG is its relative unob-
different tasks—an “experimental” task and trusiveness and portability. Portability will
a “control” task. The difference between eventually reach the point where it will be
images taken while subject is performing the possible to take unobtrusive measurements
two tasks provides a picture of regions of the from people as they go about their daily
brain that are differentially activated by the affairs. PET and fMRI provide better spatial
experimental task. resolution than EEG, but poorer temporal
There are three basic imaging methods. resolution because blood-flow to neurally
The oldest, electro-encephalogram (or EEG) active areas occurs with a stochastic lag from
uses electrodes attached to the scalp to meas- a few seconds (fMRI) to a minute (PET).
ure electrical activity synchronized to stimu- Brain imaging still provides only a crude
lus events or behavioral responses (known as snapshot of brain activity. Neural processes
Event Related Potentials or ERPs). Like are thought to occur on a 0.1 millimeter
EEG, positron emission topography (PET) scale in 100 milliseconds (msec), but the
scanning is an old technique in the rapidly spatial and temporal resolution of a typical
changing time-frame of neuroscience, but is scanner is only 3 millimeters and several sec-
still a useful technique. PET measures blood onds. Multiple trials per subject can be aver-
flow in the brain, which is a reasonable proxy aged to form composite images, but doing so
for neural activity, since neural activity in a constrains experimental design. However,
region leads to increased blood flow to that the technology has improved rapidly and will
region. The newest, and currently most pop- continue to improve. Hybrid techniques that
ular, imaging method is functional magnetic combine the strengths of different methods
resonance imaging (fMRI), which tracks are particularly promising. Techniques for
blood flow in the brain using changes in mag- simultaneously scanning multiple brains
netic properties due to blood oxygenation (“hyperscanning”) have also been developed,
(the “BOLD signal”). Simultaneous direct which can be used to study multiple-brain-
recording of neural processing and fMRI level differences in activity in games and
responses confirms that the BOLD signal markets (Read Montague et al. 2002).
reflects input to neurons and their processing
2.2 Single-Neuron Measurement
(Nikos Logothetis et al. 2001).
Although fMRI is increasingly becoming Even the finest-grained brain imaging
the method of choice, each method has techniques only measure activity of “cir-
pros and cons. EEG has excellent temporal cuits” consisting of thousands of neurons. In
mr05_Article 1 3/28/05 3:25 PM Page 13

Camerer, Loewenstein, and Prelec: Neuroeconomics 13

single neuron measurement, tiny electrodes For a big series of EBS pulses, rats will leap
are inserted into the brain, each measuring a over hurdles, cross electrified grids, and
single neuron’s firing. As we discuss below, forego their only daily opportunities to eat,
single neuron measurement studies have drink, or mate. Animals also trade EBS off
produced some striking findings that, we against smaller rewards in a sensible fash-
believe, are relevant to economics. A limita- ion—e.g., they demand more EBS to forego
tion of single neuron measurement is that, food when they are hungry. Unlike more
because insertion of the wires damages naturalistic rewards, EBS does not satiate.
neurons, it is largely restricted to animals. And electrical brain stimulation at specific
Studying animals is informative about sites often elicits behaviors such as eating,
humans because many brain structures and drinking (Joseph Mendelson 1967), or copu-
functions of non-human mammals are simi- lation (Anthony Caggiula and Bartley
lar to those of humans (e.g., we are more Hoebel 1966). Many abused drugs, such as
genetically similar to many species of mon- cocaine, amphetamine, heroin, cannabis,
keys than those species are to other species). and nicotine, lower the threshold at which
Neuroscientists commonly divide the brain animals will lever-press for EBS (Roy Wise
into crude regions that reflect a combination 1996). Despite its obvious applications to
of evolutionary development, functions, and economics, only a few studies have explored
physiology. The most common, triune divi- substitutability of EBS and other reinforcers
sion draws a distinction between the “reptil- (Steven Hursh and B. H. Natelson; Leonard
ian brain,” which is responsible for basic Green and Howard Rachlin 1991; Peter
survival functions, such as breathing, sleep- Shizgal 1999).
ing, eating, the “mammalian brain,” which
2.4 Psychopathology and Brain Damage in
encompasses neural units associated with
social emotions, and the “hominid” brain,
which is unique to humans and includes Chronic mental illnesses (e.g., schizophre-
much of our oversized cortex—the thin, fold- nia), developmental disorders (e.g., autism),
ed, layer covering the brain that is responsi- degenerative diseases of the nervous system,
ble for such “higher” functions as language, and accidents and strokes that damage local-
consciousness and long-term planning (Paul ized brain regions help us understand how
MacLean 1990). Because single neuron the brain works (e.g., Antonio Damasio
measurement is largely restricted to nonhu- 1994). When patients with known damage to
man animals, it has so far shed far more light an area X perform a special task more poor-
on the basic emotional and motivational ly than “normal” patients, and do other tasks
processes that humans share with other equally well, one can infer that area X is used
mammals than on higher-level processes to do the special task. Patients who have
such as language and consciousness. undergone neurosurgical procedures such as
lobotomy (used in the past to treat depres-
2.3 Electrical Brain Stimulation (EBS)
sion) or radical bisection of the brain (an
Electrical brain stimulation is another extreme remedy for epilepsy, now rarely
method that is largely restricted to animals. used) have also provided valuable data (see
In 1954, two psychologists (James Olds and Walter Freeman and James Watts 1942;
Peter Milner 1954) discovered that rats Gazzaniga and LeDoux 1978).
would learn and execute novel behaviors if Finally, a relatively new method called
rewarded by brief pulses of electrical brain transcranial magnetic stimulation (TMS)
stimulation (EBS) to certain sites in the uses pulsed magnetic fields to temporarily
brain. Rats (and many other vertebrates, disrupt brain function in specific regions.
including humans) will work hard for EBS. The difference in cognitive and behavioral
mr05_Article 1 3/28/05 3:25 PM Page 14

14 Journal of Economic Literature, Vol. XLIII (March 2005)

functioning that results from such disrup- the trajectories which project from one neu-
tions provides clues about which regions ral region to others (like watching patterns
control which neural functions. The theoret- of car traffic from a helicopter can tell you
ical advantage of TMS over brain imaging is roughly the ebb and flow of economic and
that TMS directly leads to causal inferences social activity). Knowing where a region’s
about brain functioning rather than the neurons project is extremely useful in
purely associational evidence provided by understanding neural circuitry and is an
imaging techniques. Unfortunately, the use important complement to simply imaging
of TMS is currently limited to the cortex (it activity in multiple regions (using fMRI)
is particularly useful for studying visual with little ability to pin down which activity
processes in the occipital lobe, in the back of occurs earliest. Furthermore, the technique
the head). It is also controversial because it can be used after autopsies, which is an
can cause seizures and may have other bad obvious advantage.
long-run effects.
2.7 Is Neuroscience Just About Where
2.5 Psychophysical Measurement Things Happen in the Brain?
An old and simple technique is measure- Neuroscience is sometimes criticized as
ment of psychophysiological indicators like providing little more than a picture of
heart rate, blood pressure, galvanic skin “where things happen in the brain” or, more
response (GSR, sweating in the palms), and cynically, as simply showing that behavior is
pupil dilation (pupils dilate in response to caused by action in the nervous system
arousal, including monetary reward). These (which was never in doubt). Indeed, some
measurements are easy, not too obtrusive, neuroscientists who are purely interested in
portable, and very rapid in time. The draw- the functionality of different brain regions
back is that measurements can fluctuate for would endorse such a characterization
many reasons (e.g., body movement) and unapologetically.
many different combinations of emotions However, the long-run goal of neuro-
lead to similar psychophysiological output, science is to provide more than a map of the
just as being pulled over by a cop and meet- mind. By tracking what parts of the brain
ing a blind date may produce very similar are activated by different tasks, and espe-
emotional anxiety responses. Often these cially by looking for overlap between
measurements are useful in combination diverse tasks, neuroscientists are gaining an
with other techniques or in lesion patients understanding of what different parts of the
who are likely to have very different physio- brain do, how the parts interact in “circuit-
logical reactions (e.g., sociopaths do not ry,” and, hence, how the brain solves differ-
show normal GSR fear reactions before a ent types of problems. For example,
possible monetary loss). Facial musculature because different parts of the brain are
can also be measured by attaching small more or less associated with affective or
electrodes to smiling muscles (on the cognitive processing (a distinction we
cheekbones) and frowning muscles (between define more precisely below), imaging peo-
the eyebrows). ple while they are doing different types of
economic tasks provides important clues
2.6 Diffusion Tensor Imaging (DTI)
about the mix of affective and cognitive
Diffusion tensor imaging (DTI) is a new processes in those tasks. For example,
technique which exploits the fact that water Sanfey et al. (2003) find that the insula cor-
flows rapidly though myleinated (sheathed) tex—a region in the temporal lobe that
neural axons (Dennis Bihan et al. 2001). encodes bodily sensations like pain and
Imaging the water flow can therefore reveal odor disgust—is active when people receive
mr05_Article 1 3/28/05 3:25 PM Page 15

Camerer, Loewenstein, and Prelec: Neuroeconomics 15

low offers in an ultimatum bargaining humans, provides a level of experimental

game. This means that even if rejecting a control that economists conducting field
low offer is done because of an adapted research should envy.
instinct to build up a reputation for tough- As always, “the proof is in the pudding.”
ness (in order to get more in the future), Here, our goal is not to review the many
the circuitry that encodes this instinct clear- ways in which neuroscience will rapidly
ly has an affective component, which is not change economic theory, because that is not
purely cognitive. where we are yet. Our goal is to showcase
For neuroeconomists, knowing more some key findings in neuroscience, and stim-
about functional specialization, and how ulate the reader’s curiosity about what these
regions collaborate in different tasks, could findings might mean for economics.
substitute familiar distinctions between cat-
egories of economic behavior (sometimes
3. Basic Lessons from Neuroscience
established arbitrarily by suggestions which
become modeling conventions) with new Because most of these techniques involve
ones grounded in neural detail. For exam- localization of brain activity, this can easily
ple, the insula activity noted by Sanfey et al. foster a misperception that neuroscience is
in bargaining is also present when subjects merely developing a “geography of the
playing matrix games are asked to guess brain,” a map of which brain bits do what
what other subjects think they will do (sec- part of the job. If that were indeed so, then
ond-order beliefs) (see Meghana Bhatt and there would be little reason for economists
Camerer, in press). This suggests that maybe to pay attention. In reality, however, neuro-
the subjects who receive low offers in the science is beginning to use regional activity
ultimatum study are not disgusted, they are differences and other clues to elucidate the
simply evaluating a second-order belief principles of brain organization and func-
about what proposers expect them to do, as tioning, which in turn, is radically changing
an input into an emotional evaluation. This is our understanding of how the brain works.
just a speculation, of course, but it shows Our goal in this section is to highlight some
how direct understanding of neural circuitry of the findings from neuroscience that may
can inspire theorizing and the search for prove relevant to economics.
new data.
3.1 A Two-Dimensional Theoretical
In light of the long list of methods
reviewed earlier in this section, it is also
worth emphasizing that neuroscience isn’t Our organizing theme, depicted in Table 1,
only about brain imaging. Brain lesion stud- emphasizes the two distinctions mentioned
ies (as well as TMS) allow one to examine in the introduction, between controlled and
the impact of disabling specific parts of the automatic processes, and between cognition
brain. They have often provided clearer evi- and affect.
dence on functionality of specific brain
3.1.1 Automatic and Controlled Processes
regions—even with tiny sample sizes from
rare types of damage—than imaging meth- The distinction between automatic and
ods have. Single neuron measurement pro- controlled processes was first proposed by
vides information not just about what parts Schneider and Shiffrin (1977). Many others
of the brain “light up,” but about the specif- have developed similar two-system models
ic conditions that cause specific neurons to since then, with different labels: rule-based
fire at accelerated or decelerated rates. And and associative (Steven Sloman 1996);
electrical brain stimulation, though it is rational and experiential systems (Lee
largely off limits to studies involving Kirkpatrick and Seymour Epstein 1992);
mr05_Article 1 3/28/05 3:25 PM Page 16

16 Journal of Economic Literature, Vol. XLIII (March 2005)


Cognitive Affective
Controlled Processes
■ serial
■ effortful
■ evoked deliberately I II
■ good introspective
Automatic Processes
■ parallel
■ effortless
■ reflexive III IV
■ no introspective

reflective and reflexive (Matthew Automatic processes are the opposite of

Lieberman et al. 2002); deliberative and controlled processes on each of these
implementive systems (Peter Gollwitzer, dimensions— they operate in parallel, are
Kentaro Fujita, and Gabriele Oettingen not accessible to consciousness, and are
2004); assessment and locomotion (Arie relatively effortless. Parallelism facilitates
Kruglanski et al. 2000), and type I and type rapid response, allows for massive multi-
II processes (Daniel Kahneman and Shane tasking, and gives the brain remarkable
Frederick 2002). power when it comes to certain types of
Controlled processes, as described by the tasks, such as visual identification.
two rows of Table 1, are serial (they use Parallelism also provides redundancy that
step-by-step logic or computations), tend to decreases the brain’s vulnerability to injury.
be invoked deliberately by the agent when When neurons are progressively destroyed
her or she encounters a challenge or sur- in a region, the consequences are typically
prise (Reid Hastie 1984), and are often gradual rather than sudden (“graceful
associated with a subjective feeling of degradation”).3 “Connectionist” neural net-
effort. People can typically provide a good work models formulated by cognitive psy-
introspective account of controlled process- chologists (David Rumelhart and James
es. Thus, if asked how they solved a math McClelland 1986) capture these features
problem or choose a new car, they can and have been applied to many domains,
often recall the considerations and the
steps leading up to the choice.2 Standard 3
The brain’s ability to recover from environmental
tools of economics, such as decision trees damage is also facilitated by a property called plasticity. In
and dynamic programming, can be viewed one study that illustrates the power of plasticity, the optic
nerves of ferrets (which are born when they are still at a
as stylized representations of controlled relatively immature state of development when the brain is
processes. still highly plastic) were disconnected at birth and recon-
nected to the auditory cortex (the portion of the brain that
processes sound). The ferrets learned to “see” using audi-
tory cortex, and some neurons in their auditory cortex
Elaborate methods have been developed that maxi- actually took on the physical characteristics of neurons in
mize the validity of such “verbal protocols” (see, e.g., the visual cortex (Lauire von Melchner, Sarah Pallas, and
Herbert Simon). Mriganka Sur 2000).
mr05_Article 1 3/28/05 3:25 PM Page 17

Camerer, Loewenstein, and Prelec: Neuroeconomics 17


Brodmann”10 Visual

L insula R insula

Figure 1. The human brain with some economically relevant areas marked.

including commercial ones. Models of this Automatic and controlled processes can
type have a very different structure from be roughly distinguished by where they
the systems of equations that economists occur in the brain (Lieberman et al. 2002).
typically work with. Unlike systems of Regions that support cognitive automatic
equations, they are “black-box”; it is hard to activity are concentrated in the back (occip-
intuit what they are doing by looking at ital), top (parietal), and side (temporal)
individual parameters. parts of the brain (see Figure 1). The amyg-
Because automatic processes are not dala, buried below the cortex, is responsible
accessible to consciousness, people often for many important automatic affective
have surprisingly little introspective insight responses, especially fear. Controlled
into why automatic choices or judgments processes occur mainly in the front (orbital
were made. A face is perceived as “attrac- and prefrontal) parts of the brain. The pre-
tive” or a verbal remark as “sarcastic” auto- frontal cortex (pFC) is sometimes called the
matically and effortlessly. It is only later “executive” region, because it draws inputs
that the controlled system may reflect on from almost all other regions, integrates
the judgment and attempt to substantiate it them to form near and long-term goals, and
logically, and when it does, it often does so plans actions that take these goals into
spuriously (e.g., Timothy Wilson, Samuel account (Timothy Shallice and Paul
Lindsey, and Tonya Schooler 2000). Burgess 1996). The prefrontal area is the
mr05_Article 1 3/28/05 3:25 PM Page 18

18 Journal of Economic Literature, Vol. XLIII (March 2005)

region that has grown the most in the (LeDoux 1996; Piotr Winkielman and Kent
course of human evolution and which, Berridge 2004). As Rita Carter (1999) com-
therefore, most sharply differentiates us ments, “the conscious appreciation of emo-
from our closest primate relatives (Stephen tion is looking more and more like one
Manuck et al. 2003). quite small, and sometimes inessential, ele-
Automatic processes—whether cognitive ment of a system of survival mechanisms
or affective—are the default mode of brain that mainly operate—even in adults—at an
operation. They whir along all the time, unconscious level.”
even when we dream, constituting most of For most affect researchers, the central
the electro-chemical activity in the brain. feature of affect is not the feeling states asso-
Controlled processes occur at special ciated with it, but its role in human motiva-
moments when automatic processes become tion. All affects have “valence”—they are
“interrupted,” which happens when a per- either positive or negative (though some
son encounters unexpected events, experi- complex emotions, such as “bittersweet,”
ences strong visceral states, or is presented can combine more basic emotions that have
with some kind of explicit challenge in the opposing valences). Many also carry “action
form of a novel decision or other type of tendencies” (Nico Frijda 1986; Leonard
problem. To the degree that controlled Berkowitz 1999)—e.g., anger motivates us to
processes are well described by economic aggress, pain to take steps to ease the pain,
calculation but parallel processes are not, and fear to escape or in some cases to
one could say that economics is about the freeze—as well as diverse other effects on
“interrupt” or “override.” 4 sensory perception, memory, preferences,
and so on (see, e.g., Leda Cosmides and
3.1.2 Affective and Cognitive Processes
John Tooby 2004). Affective processes,
The second distinction, represented by according to Zajonc’s (1998) definition, are
the two columns of table 1, is between those that address “go/no-go” questions—
affective and cognitive processes. Such a that motivate approach or avoidance behav-
distinction is pervasive in contemporary ior. Cognitive processes, in contrast, are
psychology (e.g., Robert Zajonc 1980, 1984, those that answer true/false questions.5
1998; Zajonc and Daniel McIntosh 1992) Though it is not essential to our overall argu-
and neuroscience (Damasio 1994; LeDoux ment, our view is that cognition by itself can-
1996; Panksepp 1998), and has an historical not produce action; to influence behavior,
lineage going back to the ancient Greeks the cognitive system must operate via the
and earlier (Plato described people as driv- affective system.
ing a chariot drawn by two horses, reason Affect, as we use the term, embodies not
and passions). only emotions such as anger, fear, and jeal-
The distinguishing features of affective ousy, but also drive states such as hunger,
processing are somewhat counterintuitive. thirst and sexual desire, and motivational
Most people undoubtedly associate affect states such as physical pain, discomfort
with feeling states, and indeed most affect (e.g., nausea) and drug craving. Ross Buck
states do produce feeling states when they (1999) refers to these latter influences as
reach a threshold level of intensity. “biological affects,” which he distinguishes
However, most affect probably operates
below the threshold of conscious awareness
By this definition, neural processes that don’t have
valence are not affects. There are also neural processes
that produce actions that are not best defined as affects –
As David Laibson and Andrew Caplin, respectively, e.g., the reflexes that cause you to draw away from a hot
have aptly expressed it. object or an electric shock.
mr05_Article 1 3/28/05 3:25 PM Page 19

Camerer, Loewenstein, and Prelec: Neuroeconomics 19

from the more traditional “social affects.” processing. Research on “emotional regu-
Affect, thus, coincides closely with the his- lation” shows many ways that cognition
torical concept of the passions. Although influences emotion, which implies the
emotions such as anger and fear might capacity for controlling emotion.7
seem qualitatively different than the bio- The four-quadrant model is just a way to
logical affects, they have more in common remind readers that the cognitive–affective
that might be supposed. Thus, a recent and controlled–automatic dimensions are
study showed that hurt feelings activated not perfectly correlated, and to provide a
the same brain regions activated by broken broad view to guide exploratory research.
bones or other physical injuries (Naomi For some purposes, reducing the two
Eisenberger et al. 2003).6 dimensions to one, or the four quadrants to
two, will certainly be useful. Furthermore,
3.1.3 The Quadrants in Action: An
noting all four cells is not a claim that all are
equally important. It is just a suggestion that
These two dimensions, in combination, leaving out one of the combinations would
define four quadrants (labeled I to IV in lead to a model which is incomplete for
Table 1). Quadrant I is in charge when you some purposes.
deliberate about whether to refinance your Consider what happens when a party host
house, poring over present-value calcula- approaches you with a plate of sushi.
tions; quadrant II is undoubtedly the rarest Quadrant III: Your first task is to figure
in pure form. It is used by “method actors” out what is on the plate. The occipital cor-
who imagine previous emotional experi- tex in the back of the brain is the first on the
ences so as to actually experience those scene, drawing in signals from your eyes via
emotions during a performance; quadrant your optic nerves. It decodes the sushi into
III governs the movement of your hand as primitive patterns such as lines and corners
you return serve; and quadrant IV makes then uses a “cascading process” to discern
you jump when somebody says “Boo!” larger shapes (Stephen Kosslyn 1994).
Most behavior results from the interac- Further downstream, in the inferior tempo-
tion of all four quadrants. A natural instinct ral visual cortex (ITVC), this information
of economists trained in parsimonious becomes integrated with stored representa-
modeling is to think that cognition is typi- tions of objects, which permits you to rec-
cally controlled, and affect is automatic, so ognize the objects on the plate as sushi. This
there are really only two dimensions latter process is extraordinarily complicated
(quadrants I and IV) rather than four. But (and has proved difficult for artificial intelli-
a lot of cognitive processing is automatic as gence researchers to recreate in computers)
well—e.g., visual perception or language because objects can take so many forms,
orientations, and sizes.
Quadrant IV: This is where affect enters
The researchers scanned the brains of subjects using the picture. Neurons in the inferior tempo-
fMRI (functional magnetic resonance imaging) as they ral visual cortex are sensitive only to the
played a video game designed to produce a feeling of social
rejection. Subjects thought they were playing a game that
involved throwing a ball back and forth with two other
people, but in fact the computer controlled the two ani-
mated figures that they saw on the screen. After a period For example, when people are shown unpleasant pho-
of three-way play, the two other “players” began to exclude tographs and told to interpret the photos so that they don’t
the subjects by throwing the ball back and forth between experience negative emotions (e.g., imagine a picture of
themselves. The social snub triggered neural activity in a women crying as having been taken at a wedding), there is
part of the brain called the anterior cingulate cortex, which less activity in the insula and amygdala, emotional areas,
also processes physical pain, and the insula, which is active and in medial orbitofrontal cortex, and area which the
during physical and social discomfort. insula projects to (Kevin Ochsner et al., in press).
mr05_Article 1 3/28/05 3:25 PM Page 20

20 Journal of Economic Literature, Vol. XLIII (March 2005)

identity of an object; they don’t tell you when she turns to serve other guests).8
whether it will taste good. Outputs of the These explicit thoughts involve anticipated
inferior temporal visual cortex as well as out- feelings (your own and the host’s) and
puts from other sensory systems feed into draw on explicit memories from a part of
the orbitofrontal cortex to determine the the brain called the hippocampus (see fig-
“reward value of the recognized object.” ure 1), inputs from the affective system
This is a highly particular representation. In (sometimes referred to as the “limbic sys-
economic terms, what is represented is nei- tem”), and anticipation (planning) from
ther pure information (i.e., that this is sushi) the prefrontal cortex.
nor pure utility (i.e., that it is something I Because standard economics is best
like) but rather a fusion of information and described by the controlled, cognitive,
utility. It is as if certain neurons in the processes of quadrant 1, in the remainder of
orbitofrontal cortex are saying “this is sushi this section we focus on the other half of
and I want it.” each dichotomy—automatic and affective
The reward value of sushi depends in processes—providing further details of their
turn on many factors. First, there is your functioning.
personal history with sushi. If you got sick
on sushi in the past, you will have an uncon- 3.2 Automatic Processes
scious and automatic aversion to it (“taste Here, we review some key principles of
aversion conditioning”). The amygdala neural functioning that characterize auto-
seems to play a critical role in this kind of matic processes. Our short-list includes:
long-term learning (LeDoux 1996). parallelism, specialization, and coordina-
Second, the reward value of the sushi will tion. Unpacking this a bit, we would say that:
depend on your current level of hunger; (1) much of the brain’s processing involves
people can eat almost anything—grass, processes that unfold in parallel and are not
bugs, human flesh—if they are hungry
enough. The orbitofrontal cortex and a sub-
cortical region called the hypothalamus are 8
Paul Romer (2000) uses the example of peanut tastes
sensitive to your level of hunger (Rolls as an illustration of how understanding the cause of
1999). Neurons in these regions fire more revealed preference matters. One person loves the taste of
rapidly at the sight or taste of food when peanuts, but is allergic to them and knows that the conse-
quences of eating would be disastrous. When she is hun-
you are hungry, and fire less rapidly when gry, her visceral system motivates her to eat peanuts, but
you are not hungry. her deliberative system, with its ability to consider delayed
Quadrants I and II: Processing often consequences, inhibits her from eating them. The other
person developed a “taste aversion” to peanuts many years
ends before quadrants I and II go to work. ago, as a result of having gotten sick right after eating
If you are hungry, and like sushi, your them. She knows at a cognitive level that the peanuts were
motor cortex will guide your arm to reach not the cause of her sickness, but her visceral system over-
rules cognitive awareness. Revealed preference theory
for the sushi and eat it, drawing on auto- would stop at the conclusion that both women have disu-
matic quadrant III (reaching) and IV tility for peanuts. But the fact that the mechanisms under-
(taste and enjoyment) processes. Under lying their preferences are different leads to predictable
differences in other kinds of behavior. For example, the
some circumstances, however, higher level taste-averse woman will have a higher price elasticity
processing may enter the picture. If you (she’ll eat peanuts if paid enough) and she will learn to
saw a recent documentary on the risks of enjoy peanuts after eating them a few times (her taste-
aversion can be “extinguished”). The allergic woman will
eating raw fish, you may recoil; or if you also like the smell of peanuts, which the taste-averter
dislike sushi but anticipate disappointment won’t. Treatments also differ: cognitive therapies for treat-
in the eyes of your proud host who made ing harmless phobias and taste-aversion train people to use
their conscious quadrant I processing to overrule visceral
the sushi herself, you’ll eat it anyway (or quadrant IV impulses, whereas treatment of the allergic
pick it up and discreetly hide it in a napkin peanut-avoider will concentrate on a medical cure.
mr05_Article 1 3/30/05 3:19 PM Page 21

Camerer, Loewenstein, and Prelec: Neuroeconomics 21

accessible to consciousness; (2) the brain the automatic activity produced by quadrant
utilizes multiple systems specialized to per- III and IV processes when they are wrong?
form specific functions; and (3) it figures out Indeed sometimes it does: pilots learn to
how to use existing specialized systems to trust their instrument panels, even when
accomplish new tasks efficiently, whatever they conflict with strong sensory intuitions
functions they originally evolved to perform. about where their plane is headed, but such
cognitive override is more the exception
3.2.1 Parallelism
than the rule. To override automatic
The brain performs a huge number of dif- processes, quadrant I has to (a) recognize
ferent computations in parallel. Because of the that an initial impression is wrong (which
massively interconnected “network” architec- requires self-awareness about behavior in
ture of neural systems, computations done in the other quadrants), and then (b) deliber-
one part of the brain have the potential to ately correct that impression. But when
influence any other computation, even when sense making works outside of consciousness
there is no logical connection between the two. it will not generate alarm bells to trigger the
Recent psychological research on automat- recognition required in (a). This is surely the
ic processing provides many striking examples case in the Epley and Gilovich studies. Even
of such spurious interactions. In one particu- when the external influence is obvious and
larly clever study, Gary Wells and Richard inappropriate, or the subject is warned
Petty (1980) had subjects listen to an editori- ahead of time, the deliberation required to
al delivered over headphones while shaking correct the first impression is hard work,
their head either up–down, or right–left (sub- competing for mental resources and atten-
jects were led to believe that shaking was a tion with all the other work that needs to be
legitimate part of the “product test”). Those done at the same moment (Daniel Gilbert
who were instructed to shake up–down 2002). The struggle between rapid uncon-
reported agreeing with the editorial more scious pattern-detection processes and their
than those instructed to shake left–right, pre- slow, effortful modulation by deliberation is
sumably because in our culture up–down not a fair contest; so automatic impressions
head movement is associated with an attitude will influence behavior much of the time.
of acceptance, and left–right with an attitude
3.2.2 Specialization
of rejection. A similar impact on preference
was also observed in a study in which subjects Neurons in different parts of the brain have
were asked to evaluate cartoons while holding different shapes and structures, different func-
a pen either clamped horizontally between tional properties, and operate in coordination
their teeth or grasping it with pursed lips as if as systems that are functionally specialized.
puffing on a cigarette (Fritz Strack, Leonard Progress in neuroscience often involves trac-
Martin, and Sabine Stepper 1988). The hori- ing well-known psychological functions to cir-
zontal clamp forces the mouth into a smile, cumscribed brain areas. For example, Broca’s
which enhances ratings, while “puffing” and Wernicke’s areas are involved, respective-
forces the mouth into a pursed expression, ly, in the production and comprehension of
which lowers ratings. What the brain seems to language. Patients with Wernicke damage can-
be doing, in these examples, is seeking a not understand spoken words. While they can
“global equilibrium” that would reconcile the produce words themselves, they can’t monitor
forced action (e.g., forced smile) with the their own speech, which results in sentences
response and the perceived attributes of the of correctly articulated words strung together
evaluated object. into unintelligible gibberish. People with dam-
Given that people are capable of delibera- age to Broca’s area, in contrast, can understand
tion, why doesn’t quadrant I thinking correct what is said to them and typically know what
mr05_Article 1 3/28/05 3:25 PM Page 22

22 Journal of Economic Literature, Vol. XLIII (March 2005)

they want to say, but have difficulty articulat- in his hand that pricked a patient when they
ing it, sometimes to the point of not being able shook hands. Later the doctor returned and,
to generate any words at all. although he did not remember the doctor, he
Beyond uncovering the nature of these nevertheless responded negatively to the
specialized systems, neuroscience has led to doctor’s arrival and refused to shake his hand.
the discovery of new functional systems, An example of the latter involves the expe-
some of which are quite surprising. For rience of fear, and recognition of fear in oth-
example, surgeons conducting brain surgery ers. Experiencing an emotion and
on an epileptic patient discovered a small recognizing the emotion when displayed
region of her brain which, when stimulated, intuitively seems to involve entirely different
caused her to laugh (Itzhak Fried 1998), processes. However, recent findings suggest
hinting at the existence of a “humor system.” that this is not the case. Numerous studies
Neuroscientists have also located an area in have implicated the amygdala—a small
the temporal lobe that, when stimulated “organ” in the brain that is also intimately
electrically, produces intense religious feel- linked to the sense of smell—to fear pro-
ings—e.g., the sense of a holy presence or cessing. Lesions to the amygdala of rats and
even explicit visions of God or Christ, even other animals disrupt or even eliminate the
in otherwise unreligious people (Michael animals’ fear-responses. Humans with stroke
Persinger and Faye Healey 2002). damage in the amygdala show similar
More generally, neuroscience has begun deficits in reacting to threatening stimuli.
to change our classification of functional The same damage that disrupts fear
processes in the brain, in some cases identi- responses in humans also cripples a person’s
fying distinct brain processes that serve the ability to recognize facial expressions of fear
same function, and in others drawing con- in others, and to represent such expressions
nections between processes that have been in pictures. Figure 2 shows how a patient
commonly viewed as distinct. with amygdala damage expressed different
An example of the former is memory. emotions when asked to draw them in the
Studies of patients with localized brain dam- form of facial expressions. The patient was
age have confirmed the existence of distinct able to render most emotions with remark-
memory systems, which can be selectively able artistic talent. But when it came to
“knocked out.” Anterograde amnesiacs, for drawing an expression of fear she felt clue-
example, are commonly able to recall infor- less. She didn’t even try to draw an adult
mation acquired before injury, and they are face; instead she drew a picture of a crawling
able to acquire implicit information, includ- infant looking apprehensive.
ing perceptual-motor skills (like the ability to The finding that people who don’t experi-
read text in a mirror), but they are unable to ence fear also can’t recognize it or represent it
recall new explicit information for more than pictorially, suggests that two phenomena that
a minute.9 Anterograde amnesiacs can also had been viewed as distinct—experiencing
acquire new emotional associations without and representing fear—in fact have important
the explicit memories necessary to make commonalities. Beyond this, they point to the
sense of them. In one famous example, a doc- intriguing notion that to recognize an emotion
tor introduced himself with a tack concealed in others, one needs to be able to experience
it oneself (see Alvin Goldman 2003).
Retrograde amnesia is the more familiar kind, in The idea that people have specialized sys-
which people forget their past but can form new memo- tems that are invoked in specific situations
ries. The 2000 movie “Memento” drew an accurate por- could have dramatic consequences for eco-
trait of anterograde amnesia and—interestingly for
economists— its vulnerability to exploitation by people nomics. The standard model of economic
who know you are forgetting what they did to you. behavior assumes that people have a unitary
mr05_Article 1 3/28/05 3:25 PM Page 23

Camerer, Loewenstein, and Prelec: Neuroeconomics 23

Figure 2. Representations of different emotions drawn by a patient with amygdala damage

(Adolphs et al. 1995).

set of preferences which they seek to satisfy, brains have different systems that they use to
and economists often criticize psychology for solve different problems.”
lacking such a unified perspective. The exis-
tence of such selectively—invoked, special- 3.2.3 Coordination
ized, systems, however, raises the question
of whether a unified account of behavior In a process that is not well understood,
is likely to do a very good job of capturing the brain figures out how to do the tasks it is
the complexities of human behavior. As assigned efficiently, using the specialized sys-
Jonathan Cohen (personal communication) tems it has at its disposal. When the brain is
aptly expressed it at a recent conference on confronted with a new problem it initially
neuroeconomics, “Economics has one theo- draws heavily on diverse regions, including,
ry, psychologists many—perhaps because often, the prefrontal cortex (where controlled
mr05_Article 1 3/28/05 3:25 PM Page 24

24 Journal of Economic Literature, Vol. XLIII (March 2005)

Figure 3. Regions of brain activation when first playing Tetris (left) and after several weeks of
practice (right) (Haier et al. 1992).

processes are concentrated). But over time, Andrew Lo and Dmitry Repin (2002)
activity becomes more streamlined, concen- observe a similar result in a remarkable
trating in regions that specialized in process- study of professional foreign-exchange and
ing relevant to the task. In one study derivatives traders who were wired for psy-
(Richard Haier et al. 1992), subjects’ brains chophysiological measurements while they
were imaged at different points in time as traded. Less-experienced traders showed
they gained experience with the computer significant physiological reactions to about
game Tetris, which requires rapid hand–eye half of the market events (e.g., trend rever-
coordination and spatial reasoning. When sals). More experienced traders reacted
subjects began playing, they were highly much less to the same events. Years of
aroused and many parts of the brain were automatizing apparently enabled seasonal
active (Figure 3, left panel). However, as they traders to react calmly to dramatic events
got better at the game, overall blood flow to that send a novice trader on an emotional
the brain decreased markedly, and activity roller coaster.
became localized in only a few brain regions Given the severe limitations of controlled
(Figure 3, right panel). processes, the brain is constantly in the
Much as an economy ideally adjusts to the process of automating the processing of
introduction of a new product by gradually tasks—i.e., executing them using automatic
shifting production to the firms that can pro- rather than controlled processes. Indeed,
duce the best goods most cheaply, with expe- one of the hallmarks of expertise in an area is
rience at a task or problem, the brain seems the use of automatic processes such as visual
to gradually shift processing toward brain imagery and categorization. In one now-
regions and specialized systems that can famous study, Fernand Gobet and Simon
solve problems automatically and efficiently (1996) tested memory for configurations of
with low effort. chess pieces positioned on a chessboard.
mr05_Article 1 3/28/05 3:25 PM Page 25

Camerer, Loewenstein, and Prelec: Neuroeconomics 25

They found that expert chess players were neural activity is distilled into a categorical
able to store the positions of players almost percept, or a decision, we do know that the
instantly—but only if they were in positions brain doesn’t invariably integrate (i.e., aver-
corresponding to a plausible game. For ran- age) over the signals carried by individual
domly arranged chess pieces, the experts neurons. In particular, when two distinct
were not much better than novices. Further groups of neurons convey different informa-
investigations have found that chess grand- tion about the external world, the resulting
masters store roughly 10,000 different possi- perceptual judgment often adopts the infor-
ble board setups in memory which they can mation of one neuronal group and entirely
recognize almost instantly and respond to. suppresses the information carried by the
More recent research in decision making other. This is referred to as a “winner-take-
suggests that this is a more general phenom- all” principle of neural signal-extraction. As a
enon, because decision making often takes result, many brain processes are fundamen-
the form of pattern matching rather than of tally categorical, yielding well-defined per-
an explicit weighing of costs and benefits ceptions and thoughts even if the incoming
(e.g., Robyn Leboeuf 2002; Doug Medin information is highly ambiguous.10 The
and Max Bazerman 1999). advantages of this principle are clear if the
In some cases, repeated use of particular ultimate job of the brain is to initiate a dis-
specialized systems can produce physically crete action, or categorize an object as one of
recognizable changes. Studies have found several discrete types. The disadvantage is
that, for example, violinists who finger violin that updating of beliefs in response to new
strings with their left hand show enlarged information can procede in fits and starts,
development of cortical regions which corre- with beliefs remaining static as long as the
spond to fingers on the left hand (Thomas new information does not lead to recatego-
Elbert et al. 1995), and the brain regions rization, but changing abruptly and dramati-
responsible for navigation and spatial mem- cally when the accumulation of evidence
ory (the hippocampus) of London taxi driv- results in a change of categorization (see
ers are larger than comparable areas in Sendhil Mullainathan 2002).
non-taxi drivers (Eleanor Maguire et al.
3.3 Affective Processes
2000). While the exact causality is difficult to
determine (perhaps preexisting differences The way the brain evolved is critical to
in size of brain region affect people’s talents understanding human behavior. In many
and occupational choices), a causal connec- domains, such as eating, drinking, sex, and
tion in the posited direction has been even drug use, human behavior resembles
demonstrated in songbirds, whose brains that of our close mammalian relatives, which
show observable differences as a function of is not surprising because we share many of
whether or not they have been exposed to the neural mechanisms that are largely
the song that is characteristic of their species responsible for these behaviors. Many of the
(Carol Whaling et al. 1997).
3.2.4 The Winner Take All Nature of 10
Alternatively, one could imagine that perceptions or
Neural Processing beliefs aggregate all relevant neural information (which
would also be more in tune with Bayesian updating). So far
Another feature that is reminiscent of the as we can tell, the brain can use both principles of aggre-
gation: when different neural populations carry “similar”
operation of an economy is the “winner-take- information, then the overall perceptual judgment is an
all” nature of neural information processing average of all information; when they carry very different
(M. James Nichols and Bill Newsome 2002). information, then the overall judgment follows the “win-
ner-take-all” rule (Nichols and Newsome 2002). For a sim-
While there is a lot that we do not know ple model of a neural network that exhibits these
about the way in which the huge amount of properties, see Richard Hahnloser et al. 2000.
mr05_Article 1 3/28/05 3:25 PM Page 26

26 Journal of Economic Literature, Vol. XLIII (March 2005)

processes that occur in these systems are LeDoux and his colleagues (summarized in
affective rather than cognitive; they are LeDoux 1996) have arrived at similar conclu-
directly concerned with motivation. This sions about the primacy of affect using very
might not matter for economics were it not different research methods and subjects.
for the fact that the principles that guide the Based on studies conducted on rats, they dis-
affective system—the way that it operates— covered that there are direct neural projec-
is so much at variance with the standard tions from the sensory thalamus (which
economic account of behavior. performs crude signal-processing) to the
amygdala (which is widely believed to play a
3.3.1 Primacy of Affect
critical role in the processing of affective
In contrast to the intuitive view of human stimuli) that are not channeled through the
behavior as driven by deliberations about neocortex. As a result, animals can have an
costs and benefits, it does not do a terrible affective reaction to stimuli before their cor-
injustice to the field of psychology to say that tex has had the chance to perform more
a growing consensus has developed around refined processing of the stimuli—they are lit-
the view that affect is primary in the sense erally afraid before knowing whether they
that it is “first on the scene” and plays a dom- should be. Such immediate affective respons-
inant role in behavior. Indeed, as we discuss es provide organisms with a fast but crude
below (section 3.4.2), the conscious brain assessment of the behavioral options they face
often erroneously interprets behavior that which makes it possible to take rapid action.
emerges from automatic, affective, processes They also provide a mechanism for interrupt-
as the outcome of cognitive deliberations. ing and refocusing attention (Simon 1967),
In a series of seminal papers, Zajonc (1980, typically shifting processing from automatic to
1984, 1998) presented the results of studies controlled processing. As Jorge Armony et al.
which showed, first, that people can often (1995) note, “A threatening signal, such as
identify their affective reaction to some- one indicating the presence of a predator,
thing—whether they like it or not—more arising from outside of the focus of attention
rapidly than they can even say what it is, and, will have a reduced representation in the cor-
second, that affective reactions to things can tex. Thus, if the amygdala relied solely on the
be dissociated from memory for details of cortical pathway to receive sensory informa-
those things, with the former often being bet- tion, it would not be capable of processing,
ter. For example, we often remember and responding to, those danger signals that
whether we liked or disliked a particular per- are not within the focus of attention” (see also,
son, book or movie, without being able to Armony et al. 1997; Gavin DeBecker 1997).11
remember any other details (Bargh 1984). A similar pattern to the one LeDoux
Subsequent research in social psychology observed in rats can also be seen in humans.
takes Zajonc’s initial research a step further Gilbert and Michael Gill (2000) propose that
by showing that the human brain affectively people are “momentary realists” who trust
tags virtually all objects and concepts, and their immediate emotional reactions and
that these affective tags are brought to mind only correct them through a comparatively
effortlessly and automatically when those laborious cognitive process. If the car behind
objects and concepts are evoked (e.g., Russell
Fazio et al. 1986; Anthony Greenwald, Mark The amygdala has excellent properties for perform-
ing such a function. Recent research in which amygdalae
Klinger, and Thomas Liu 1989; Greenwald of subjects were scanned while threatening stimuli were
1992; Bargh, Shelly Chaiken, Paula flashed at various positions in the visual field found that
Raymond, and Charles Hymes 1996; Jan De amygdala activation is just as rapid and just as pronounced
when such stimuli are presented outside as when they are
Houwer, Dirk Hermans, and Paul Eelen presented inside the region of conscious awareness (Adam
1998; David Houston and Fazio 1989) Anderson et al. 2003).
mr05_Article 1 3/28/05 3:25 PM Page 27

Camerer, Loewenstein, and Prelec: Neuroeconomics 27

you honks after your light turns green, you cold or turning on the air conditioner when
are likely to respond with immediate anger, hot. The brain motivates one to take such
followed, perhaps, by a sheepish acknowl- actions using both a “carrot” and a “stick.”
edgment that maybe the honking person The stick reflects the fact that departing
behind you had a point, as you were dis- from a set-point usually feels bad—e.g., it
tracted when the light turned green. feels bad to be either excessively hot or
It is important to note that while emotions cold—and this negative feeling motivates
may be fleeting, they can have a large eco- one to take actions that move one back
nomic impact if they create irreversible rash toward the set-point. The carrot is a process
decisions (as in “crimes of passion”). Even called “alliesthesia” (Michael Cabanac
emotions which could be transitory, such as 1979) whereby actions that move one
embarrassment, can have long-run effects if toward the set-point tend to feel pleasura-
they are kept alive by memory and social ble. When the body temperature falls below
reminders. For example, Dora Costa and 98.6°, for example, almost anything that
Matthew Kahn (2004) found that deserters raises body temperature (such as placing
in the Union Army during the Civil War, who one’s hand in warm water) feels good, and
were allowed to return to their hometowns conversely when the body temperature is
without explicit penalty, often moved away elevated almost anything that lowers body
because of the shame and social ostracism of temperature feels good.
being known to their neighbors as a deserter. As economists, we are used to thinking of
preferences as the starting point for human
3.3.2 Homeostasis
behavior and behavior as the ending point. A
To understand how the affective system neuroscience perspective, in contrast, views
operates, one needs to recall that humans did explicit behavior as only one of many mech-
not evolve to be happy, but to survive and anisms that the brain uses to maintain home-
reproduce. An important process by which ostasis, and preferences as transient state
the body attempts to achieve these goals is variables that ensure survival and reproduc-
called homeostasis. Homeostasis involves tion. The traditional economic account of
detectors that monitor when a system departs behavior, which assumes that humans act so
from a “set-point,”12 and mechanisms that as to maximally satisfy their preferences,
restore equilibrium when such departures starts in the middle (or perhaps even toward
are detected. Some—indeed most—of these the end) of the neuroscience account.
mechanisms do not involve deliberate action. Rather than viewing pleasure as the goal of
Thus, when the core body temperature falls human behavior, a more realistic account
below the 98.6° F set-point, blood tends to would view pleasure as a homeostatic cue—
be withdrawn from extremities, and when it an informational signal.13,14
rises above the set-point one begins to sweat.
But other processes do involve deliberate 13
Even deliberative behavior generated by quadrant I
action—e.g., putting on one’s jacket when is typically organized in a fashion that resembles home-
ostasis (Miller, Eugene Gallanter, and Karl Pribram 1960;
Loewenstein 1999). Rather than simply maximizing pref-
12 erences in a limitless fashion, people set goals for them-
The mechanisms that monitor the body’s state can
be exquisitely complex, and can include both internal and selves, monitor their progress toward those goals, and
external cues. In the case of nutritional regulation, for adjust behavior when they fall short of their goals.
example, internal cues include gastric distension (James Of course, even the neuroscience account begins, in
Gibbs et al. 1981), receptors sensitive to the chemical some sense, in the middle. A more complete understand-
composition of the food draining from the stomach D. ing of behavior would also ask how these different mecha-
Greenberg, Smith and Gibbs (1990); Arthur L. nisms evolved over time. Since evolution selects for genes
Campfield and Smith 1990). External cues include time that survive and reproduce, the result of evolution is
of day, estimated time till the next feeding, and the sight unlikely to maximize pleasure or minimize pain. (See Gary
or smell of food. Becker and Luis Rayo 2004.)
mr05_Article 1 3/28/05 3:25 PM Page 28

28 Journal of Economic Literature, Vol. XLIII (March 2005)

An important feature of many homeosta- encodes whether an outcome is a gain or a

tic systems is that they are highly attuned to loss (see section 5), why self-reported hap-
changes in stimuli rather than their levels. A piness (and behavioral indicators like sui-
dramatic demonstration of such sensitivity cide) depend on changes in income and
to change came from single-neuron studies wealth, rather than levels (Andrew Oswald
of monkeys responding to juice rewards 1997), and why violations of expectations
(see Wolfram Schultz and Anthony trigger powerful emotional responses
Dickinson 2000). These studies measured (George Mandler 1982).
the firing of dopamine neurons in the ani-
3.4 Interactions between the Systems
mal’s ventral striatum, which is known to
play a powerful role in motivation and Behavior emerges from a continuous
action. In their paradigm, a tone was sound- interplay between neural systems support-
ed, and two seconds later a juice reward was ing activity within each of the four quad-
squirted into the monkey’s mouth. Initially, rants. Three aspects of this interaction bear
the neurons did not fire until the juice was special emphasis, which we labeled “collab-
delivered. Once the animal learned that the oration,” “competition,” and “sense-mak-
tone forecasted the arrival of juice two sec- ing.” Collaboration captures the insight that
onds later, however, the same neurons fired decision making, which is to say “rationality”
at the sound of the tone, but did not fire in the broad, nontechnical sense of the
when the juice reward arrived. These neu- word, is not a matter of shifting decision-
rons were not responding to reward, or its making authority from quadrants II, III, and
absence . . . they were responding to devia- IV toward the deliberative, nonaffective
tions from expectations. (They are some- quadrant I, but more a matter of maintain-
times called “prediction neurons.”) When ing proper collaboration in activity across all
the juice was expected from the tone, but four quadrants. If quadrant I tries to do the
was not delivered, the neurons fired at a job alone, it will often fail.16 Competition
very low rate, as if expressing disappoint- reflects the fact that different processes—
ment. The same pattern can be observed at most notably affective and cognitive—often
a behavioral level in animals, who will work drive behavior in conflicting directions and
harder (temporarily) when a reinforcement compete for control of behavior. Sense-
is suddenly increased and go “on strike” making refers to how we make sense of such
when reinforcement falls.15 Neural sensitiv- collaboration and competition—how we
ity to change is probably important in make sense of our behavior. While behavior
explaining why the evaluation of risky gam- is, in fact, determined by the interaction of
bles depends on a reference point which all four quadrants, conscious reflection on
our behavior, and articulating reasons for it,
is basically quadrant I trying to make sense
As Rolls (1999) writes, “We are sensitive to some of that interaction. And, not surprisingly,
extent not just to the absolute level of reinforcement being
received, but also to the change in the rate or magnitude
of reinforcers being received. This is well shown by the
phenomena of positive and negative contrast effects with Roy Baumeister, Todd Heatherton, and Dianne Tice
rewards. Positive contrast occurs when the magnitude of a (1994) review studies which suggest that excessively high
reward is increased. An animal will work very much hard- incentives can result in supra-optimal levels of motivation
er for a period (perhaps lasting for minutes or longer) in that have perverse effects on performance (known as the
this situation, before gradually reverting to a rate close to Yerkes–Dodson law). Beyond documenting such “choking
that at which the animal was working for the small rein- under pressure,” the review research showing that it often
forcement. A comparable contrast effect is seen when the occurs because it causes people to utilize controlled
reward magnitude (or rate at which rewards are obtained) processes for performing functions, such as swinging a golf
is reduced—there is a negative overshoot in the rate of club, that are best accomplished with automatic processes.
working for a time.” See Dan Ariely et al. (2004) for the latest evidence.
mr05_Article 1 3/28/05 3:25 PM Page 29

Camerer, Loewenstein, and Prelec: Neuroeconomics 29

quadrant I has a tendency to explain behav- There is also interesting experimental evi-
ior in terms it can understand – in terms of dence that deliberative thinking blocks
quadrant I processes. access to one’s emotional reactions to objects
and so reduces the quality of decisions (e.g.,
3.4.1 Collaboration and Competition
Wilson and Schooler 1991). In one study
Although it is heuristically useful to distin- (Wilson et al. 1993) college students select-
guish between cognitive and affective ed their favorite poster from a set of posters.
processes, and between controlled and auto- Those who were instructed to think of rea-
matic processes, most judgments and behav- sons why they liked or disliked the posters
iors result from interactions between them. before making their selection ended up less
Collaboration, delegation of activity, and happy on average with their choice of
proper balance across the quadrants are posters (and less likely to keep them up on
essential to normal decision making. Many their dorm room wall) than subjects who
decision-making disorders may originate in were not asked to provide reasons.
an improper division of labor between the Needless to say, affect can also distort cog-
quadrants. For example, psychiatry recog- nitive judgments. For example, emotions
nizes a decision-making continuum defined have powerful effects on memory—e.g.,
by the impulsive, “light” decision-making when people become sad, they tend to recall
style at one end and the compulsive, “heavy” sad memories (which often increases their
style at the other. The decisions of an impul- sadness). Emotions also affect perceptions of
sive individual are excessively influenced by risks—anger makes people less threatened
external stimuli, pressures, and demands. by risks, and sadness makes them more
Such a person may not be able to give a more threatened (Jennifer Lerner and Dacher
satisfying explanation of an action except Keltner 2001). Emotions also create “moti-
that “he felt like it” (David Shapiro 1965). By vated cognition”—people are good at per-
contrast, an obsessive–compulsive person suading themselves that what they would
will subject even the most trivial decisions to like to happen is what will happen. Quack
extensive deliberation and calculation. In sit- remedies for desperate sick people, and get-
uations in which it is entirely appropriate to rich-quick scams are undoubtedly aided by
make a quick decision based “on impulse”— the human propensity for wishful thinking.
e.g., when choosing which video to rent Wishful thinking may also explain high rates
for the evening or what to order in a of new business failure (Camerer and Dan
restaurant—the obsessive-compulsive will Lovallo 1999), trading in financial markets,
get stuck. undersaving, and low rates of investment in
We are only now beginning to appreciate education (foregoing large economic
the importance of affect for normal decision returns). As LeDoux (1996) writes, “While
making. The affective system provides inputs conscious control over emotions is weak,
in the form of affective evaluations of behav- emotions can flood consciousness. This is
ioral options—what Damasio (1994) refers so because the wiring of the brain at this
to as “somatic markers.” Damasio and his point in our evolutionary history is such that
colleagues show that individuals with mini- connections from the emotional systems to
mal cognitive, but major affective deficits the cognitive systems are stronger than con-
have difficulty making decisions, and often nections from the cognitive systems to the
make poor decisions when they do (Antoine emotional systems.”
Bechara et al. 1994; Bechara, Damasio, When it comes to spending money or
Damasio, and Lee 1999; Damasio 1994). It is delaying gratification, taking or avoiding
not enough to “know” what should be done; risks, and behaving kindly or nastily toward
it is also necessary to “feel” it. other people, people often find themselves
mr05_Article 1 3/28/05 3:25 PM Page 30

30 Journal of Economic Literature, Vol. XLIII (March 2005)

of “two minds”; our affective systems drive us well understood. At a neurological level, it
in one direction and cognitive deliberations appears that an organ in the reptilian brain
in another. We find ourselves almost compul- called the striatum (part of a larger system
sively eating our children’s left-over called the basal ganglia) plays a critical role.
Halloween candy, while obsessing about how The striatum receives inputs from all parts of
to lose the extra ten pounds; gambling reck- the cerebral cortex, including the motor cor-
lessly at the casino, even as a small voice in tex, as well as from affective systems such as
our head tell us to rein it in; trying to build up the amygdala. Lesions of pathways that sup-
courage to step up to the podium; or tempt- ply dopamine to the striatum leads, in ani-
ed to give a little to the pathetic street-corner mals, to a failure to orient to stimuli, a failure
beggar though we know our crumpled dollar to initiate movements, and a failure to eat or
will go further if donated to United Way. drink (J. F. Marshall et al. 1974). In humans,
All of these deviations occur because our depletion of dopamine in the striatum is
affective system responds to different cues, found in Parkinson’s disease, the most dra-
and differently to the same cues, as our matic symptom of which is a lack of volun-
cognitive system does. As Rolls (1999) tary movement. The striatum seems to be
writes, involved in the selection of behaviors from
competition between different cognitive and
emotions often seem very intense in humans, affective systems—in producing one coher-
indeed sometimes so intense that they produce ent stream of behavioral output, which can
behaviour which does not seem to be adaptive,
such as fainting instead of producing an active
be interrupted if a signal of higher priority is
escape response, or freezing instead of avoiding, received, or a surprising stimulus appears
or vacillating endlessly about emotional situa- (Carolyn Zink et al. 2003).
tions and decisions, or falling hopelessly in love The extent of collaboration and competi-
even when it can be predicted to be without tion between cognitive and affective sys-
hope or to bring ruin. The puzzle is not only that
the emotion is so intense, but also that even with
tems, and the outcome of conflict when it
our rational, reasoning, capacities, humans still occurs, depends critically on the intensity of
find themselves in these situations, and may find affect (Loewenstein 1996; Loewenstein and
it difficult to produce reasonable and effective Lerner 2003). At low levels of intensity,
behaviour for resolving the situation (p. 282). affect appears to play a largely “advisory”
role. A number of theories posit that emo-
Such divergences between emotional reac- tions carry information that people use as an
tions and cognitive evaluations arise, Rolls input into the decisions they face (e.g.,
(1999) argues, because Damasio 1994; Ellen Peters and Paul Slovic
2000). The best-developed of these
in humans the reward and punishment systems approaches is affect-as-information theory
may operate implicitly in comparable ways to
those in other animals. But in addition to this,
(Norbert Schwarz and Gerald Clore 1983;
humans have the explicit system (closely related Schwarz 1990; Clore 1992).
to consciousness) which enables us consciously At intermediate levels of intensity, peo-
to look and predict many steps ahead. (p. 282). ple begin to become conscious of conflicts
between cognitive and affective inputs. It
Thus, for example, the sight or smell of a is at such intermediate levels of intensity
cookie might initiate motivation to consume that one observes the types of efforts at
on the part of the affective system, but might self-control that have received so much
also remind the cognitive system that one is attention in the literature (Jon Elster 1977;
on a diet. Walter Mischel, Ebbe Ebbesen, and
Exactly how cognitive and affective sys- Antonette Zeiss 1972; Thomas Schelling
tems interact in the control of behavior is not 1978, 1984).
mr05_Article 1 3/28/05 3:25 PM Page 31

Camerer, Loewenstein, and Prelec: Neuroeconomics 31

Finally, at even greater levels of intensity, both the sensation of intention and the overt
affect can be so powerful as to virtually pre- action are caused by prior neural events
clude decision making. No one “decides” to which are inaccessible to consciousness (see
fall asleep at the wheel, but many people do. Daniel Wegner and Thalia Wheatley 1999).17
Under the influence of intense affective Quadrant I tends to explain behavior ego-
motivation, people often report themselves as centrically—to attribute it to the types of
being “out of control” or “acting against their deliberative processes that it is responsible
own self-interest” (Baumeister, Heatherton, for (see Richard Nisbett and Wilson 1977). A
and Tice 1994; Stephen Hoch and dramatic study demonstrating this phenom-
Loewenstein 1991; Loewenstein 1996). As enon was conducted with a “split-brain”
Rita Carter writes, “where thought conflicts patient (who had an operation separating the
with emotion, the latter is designed by the connection between the two hemispheres of
neural circuitry in our brains to win” (1999). his brain). The patient’s right hemisphere
could interpret language but not speak, and
3.4.2 Spurious Sense-Making
the left hemisphere could speak (LeDoux
Sense-making is an important form of 1996). The patient’s right hemisphere was
interaction between quadrant I and the instructed to wave his hand (by showing the
other quadrants. The brain’s powerful drive word “wave” on the left part of a visual
toward sense making leads us to strive to screen, which only the right hemisphere
interpret our own behavior. Since quadrant I processed). The left hemisphere saw the
often does not have conscious access to right hand waving but was unaware of the
activity in the other quadrants, it is perhaps instructions that had been given to the right
not surprising that it tends to over-attribute hemisphere (because the cross-hemisphere
behavior to itself—i.e., to deliberative deci- connections were severed). When the
sion processes. Even though much of the patient was asked why he waved, the left
brain’s activity is “cognitively inaccessible,” hemisphere (acting as spokesperson for the
we have the illusion that we are able to make entire body) invariably came up with a plau-
sense of it, and we tend to make sense of it sible explanation, like “I saw somebody I
in terms of quadrant I processes. knew and waved at them.”
Research with EEG recordings has shown
(Benjamin Libet 1985) that the precise
4. General Implications of
moment at which we become aware of an
Neuroscience for Economics
intention to perform an action trails the ini-
tial wave of brain activity associated with that To add value to economics, neuroscience
action (the EEG “readiness potential”) by needs to suggest new insights and useful
about 300 msec. The overt behavioral perspectives on old problems. This section
response itself then follows the sensation of discusses some broad implications for eco-
intention by another 200 msec. Hence, what nomics of the ideas and findings reviewed in
is registered in consciousness is a regular the previous section. First, we show that
pairing of the sensation of intention followed neuroscience findings raise questions about
by the overt behavior. Because the neural the usefulness of some of the most common
activity antecedent to the intention is inac- constructs that economists commonly use,
cessible to consciousness, we experience such as risk aversion, time preference, and
“free will” (i.e., we cannot identify anything
that is causing the feeling of intention).
Because the behavior reliably follows the “… the brain contains a specific cognitive system that
binds intentional actions to their effects to construct a
intention, we feel that this “freely willed” coherent experience of our own agency” (Patrick Haggard,
intention is causing the action—but in fact, Sam Clark, and Jeri Kalogeras 2002).
mr05_Article 1 3/28/05 3:25 PM Page 32

32 Journal of Economic Literature, Vol. XLIII (March 2005)

altruism. Second, we show how the exis- economic analyses typically assume that the
tence of specialized systems challenges same degree of time preference is present
standard assumptions about human infor- for all intertemporal tradeoffs— saving for
mation processing and suggests that intelli- the future, flossing your teeth, dieting, and
gence and its opposite—bounded getting a tattoo. For example, whether a per-
rationality—are likely to be highly domain- son smokes is sometimes taken as a crude
specific. Third, brain-scans conducted while proxy for low rates of time discounting, in
people win or lose money suggest that studies of educational investment or savings.
money activates similar reward areas as do Thinking about the modularity of the brain
other “primary reinforcers” like food and suggests that while different intertemporal
drugs, which implies that money confers tradeoffs may have some element of plan-
direct utility, rather than simply being val- ning in common (e.g., activity in prefrontal
ued only for what it can buy. Fourth, we cortex), different types of intertemporal
show that research on the motivational and choices are likely to invoke qualitatively dif-
pleasure systems of the brain human chal- ferent mixtures of neural systems and hence
lenges the assumed connection between to produce entirely different patterns of
motivation and pleasure. Finally, we behavior. Then measured discount rates will
describe some of the important implications not be perfectly correlated across domains,
of cognitive inaccessibility for economics. and might hardly be correlated at all.
Much as the study of memory has been
4.1 Economic Constructs
refined by the identification of distinct mem-
Knowing how the brain solves problems, ory systems, each with its own properties of
and what specialized systems it has at its dis- learning, forgetting, and retrieval; the study of
posal to do so, challenges some of our fun- intertemporal choice might be enhanced by a
damental assumptions about how people similar decomposition, most likely informed
differ from one-another when it comes to by neuroscience research. For example,
economic behavior. Economists currently unpublished research by Loewenstein and
classify individuals on such dimensions as Roberto Weber suggests that, in normal sam-
“time preference,” “risk preference,” and ples, future-oriented behaviors tend to clus-
“altruism.” These are seen as characteristics ter in tasks which tap different dimensions of
that are stable within an individual over time self-control. For example, flossing your teeth
and consistent across activities; someone is statistically associated with a number of
who is risk-seeking in one domain is expect- other minor, repetitive, helpful behaviors
ed to be risk-seeking in other domains as such as feeding parking meters religiously
well. But empirical evidence shows that risk- and being on time for appointments. These
taking, time discounting, and altruism are behaviors seem to involve “conscientious-
very weakly correlated or uncorrelated ness,” an important measure in personality
across situations. This inconsistency results theory.18 John Ameriks, Andrew Caplin, and
in part from the fact that preferences are John Leahy (in press) measured the propen-
state-contingent (and that people may not sity to plan by asking TIAA–CREF enrollees
recognize the state-contingency, which—if to disagree or agree with statements like “I
they did—would trigger overrides that have spent a great deal of time developing a
impose more consistency than observed). financial plan.” These measures correlate
But it also may point to fundamental prob- with actual savings rates.
lems with the constructs that we use to Dieting and use of addictive drugs, in
define how people differ from each other. contrast to punctuality and flossing, might
As an illustration, take the concept of be involve very different circuitry and
time-preference. In empirical applications, hence reveal fundamentally different dis-
mr05_Article 1 3/28/05 3:25 PM Page 33

Camerer, Loewenstein, and Prelec: Neuroeconomics 33

counting behavior. Both of these activities processed by specialized systems, but rela-
involve visceral motives that seem to differ tively obtuse when they are forced to rely on
reliably across persons and may be only controlled processes.
weakly linked to conscientiousness. If A neat illustration of this is provided by the
you’re the kind of person who loves to eat, “Wason four-card problem” in logic. Subjects
or drink alcohol, then resisting indulging in are shown four cards, each with a letter on
those activities requires difficult exertion of one side and a number on the other. The
self-control. exposed card faces read “X,” “Y,” “1,” and “2.”
Subjects are asked which cards would need to
4.2 Domain-Specific Expertise
be turned over to test the rule: “If there is an
Economics implicitly assumes that people X on one side there is a 2 on the other.” Few
have general cognitive capabilities that can subjects give the right answers, which are X
be applied to any type of problem and, and 1 (if there’s an X on the opposite side of
hence, that people will perform equivalently the “1” the rule is broken). However, most
on problems that have similar structure. The subjects give the right answer when the logi-
existence of systems that evolved to perform cally equivalent problem is put in a cheating-
specific functions, in contrast, suggests that detection frame. For example, if there are
performance will depend critically on four children from two different towns and
whether a problem that one confronts can two school districts and the rule is “If a child
be, and is in fact, processed by a specialized lives in Concord he or she must go to
system that is well adapted to that form of Concord High,” most subjects realize that the
processing. When a specialized system exists home address of the student who does not go
and is applied to a particular task, processing to Concord High must be checked to see if
is rapid and the task feels relatively effort- she is cheating and not going to Concord
less. Automatic processes involved in vision, High when she should (Cosmides 1989).
for example, are lightning fast and occur Of particular interest to economics, many
with no feeling of mental effort, so people neuroscientists believe there is a specialized
are not aware of the power and sophistica- “mentalizing” (or “theory of mind”) module,
tion of the processes that allow it to happen. which controls a person’s inferences about
Even the most powerful computers don’t what other people believe, feel, or might
hold a candle to humans when it comes to do. The hint at the existence of such a ded-
visual perception or voice recognition. icated module came from tests conducted
When we lack such tailored systems, how- by developmental psychologists in which
ever, we are likely to seem extraordinarily two children are shown an object in the
flat-footed because we will be forced to process of being concealed (see Uta Frith
“muscle it out” with quadrant 1 processing, 2001a). One child then leaves, and the
much as autistic individuals seem to solve other child observes as the object is moved
theory of mind problems by building up a to a new location. The child who remains in
statistical understanding of appropriate the room is then asked to predict where the
social behavior. As a general rule, we should child who left will look for the object when
expect people to be geniuses when present- she returns. Normal children are typically
ed with problems that can be, and are able to solve this problem around age three
or four. Autistic children as a rule master
Based on the observation that drugs affecting sero- this distinction much later (8–12 years), and
tonin receptors are used to treat compulsion disorders, with great difficulty, although some (espe-
compulsivity appears to have some connection to the neu- cially those with “Asperger syndrome”)
rotransmitter serotonin, In addition, compulsion disorders
are thought to be related to hyperactivity in the caudate have normal or superior intelligence. At the
nucleus, a “reminder” region of the limbic system. same time, the autistic child will have no
mr05_Article 1 3/28/05 3:25 PM Page 34

34 Journal of Economic Literature, Vol. XLIII (March 2005)

difficulty with general inferences having a specifically turned on by mentalizing activi-

similar logical form (e.g., if a photo is taken ty. Neuroscience is not there yet, but recent
of the object’s location, and the object is single-cell recordings in monkeys have iden-
then moved, they will correctly infer that tified an intriguing class of “mirror neurons”
the photo will register the object in the old in the prefrontal cortex, which fire either
place, before it was moved). when an experimenter performs a physical
Autistic adult individuals may compensate action (e.g., grasping a peanut) or when the
in many ways and eventually pass such basic monkey performs (“mirrors”) the same
tests of mentalizing. However, they have dif- action. Having such neurons makes learning
ficulty appreciating more subtle social by imitation easy and supports mind reading
meanings (e.g., irony), and will sometimes by, for example, internally simulating the
wonder at the “uncanny ability” of facial expressions of others.
non–autistic persons to “read minds” (Frith Mentalizing is relevant for economics
2001b). People with Asperger’s syndrome, because many judgments require agents to
who are intelligent but have trouble under- make guesses about how other people feel or
standing emotions in others, show lower what they will do. The concept of equilibri-
activation in the medial prefrontal regions as um requires that agents correctly anticipate
compared with normals when presented what others will do; presumably this arises
with problems that involve mentalizing, but because of accurate mentalizing, or through
also show greater activation in more ventral some kind of specific learning about behav-
(lower) region of the prefrontal cortex, ior which may not transfer well to new
which is normally responsible for general domains or when variables change.
reasoning (Francesca Happe et al. 1996). A Furthermore, the kind of learning about
natural interpretation is that Asperger indi- players’ “types” from observation in
viduals eventually deduce the answer Bayesian–Nash equilibrium is modeled as
through a complex process of reasoning simply Bayesian updating of the likelihoods
instead of grasping it directly by a special- of chance events in the face of new informa-
ized module, as if they create a neural tion. Since mentalizing is a special ability,
“workaround” or “long-cut.” In the medical and logical-deductive reasoning can only
literature, one can also find patients with partially compensate for its absence, treating
brain lesions who exhibit difficulties with inference about behavior of other agents and
mentalizing tasks but not with other types of “nature” is a simplification which may be
cognition (Andrea Rowe et al. 2001; James wrong.19
Blair and Lisa Cipolotti 2000). This, too, is
consistent with the hypothesis of a separable
mentalizing module. Our point here is simple: Circuitry that controls
The possibility of a mentalizing module updating of frequencies of events in the world—whether it
snows in Chicago in January—may be dissociated from cir-
has gained credibility and substance through cuitry that controls updating of personal “types” from
converging neuroscientific evidence. fMRI behavior. The circuitry which controls attributions of pay-
studies have shown that when normal adults off types to people, from their behavior, requires a concept
of how types link to behavior (i.e., what are types?), as well
are given pairs of closely matching judgment as updating of types (i.e., reputations) from behavior. An
problems, differing only in whether they do autist with theory of mind deficits, for example, might
or do not require mentalizing, the mentaliz- keep very accurate counts of relative frequencies (in fact,
many autists are obsessed with counting objects in special
ing problems lead to greater activation in the categories) but be unable to tell whether a person behaved
left medial prefrontal cortex (Fletcher et al. badly in a game because of bad intentions or situational
1995; Rebecca Saxe and Nancy Kanwisher influences (such as a bureaucrat who is pressured to stick
to rules). So there is no reason to think that updating about
2003). As ultimate proof, one would like to event frequencies is necessarily the same as updating
identify neuronal populations that are about personal types.
mr05_Article 1 3/28/05 3:25 PM Page 35

Camerer, Loewenstein, and Prelec: Neuroeconomics 35

The existence of domain-specific expertise regions (at the top of Figure 4) both receive
suggests that people will appear to be input from “lower” systems (dopaminergic
geniuses at some tasks but will seem neurons and the amygdala), and also feed
remarkably flat-footed when dealing with back processed information to the striatum.
other tasks that may be only superficially dif- The idea that many rewards are processed
ferent. Domain-specific processing has similarly in the brain has important implica-
important implications for economics, and tions for economics, which assumes that the
specifically for the organization of labor. marginal utility of money depends on what
Bundling tasks together into jobs, for exam- money buys. Of course, it is possible that
ple, requires an understanding of which kind money rewards activate the same circuitry as
of skills are general (useful for many tasks) sports cars, cocaine, and jokes because the
and which are neurally separate. circuitry is being used to evaluate money in
terms of the goods that it buys (which
4.3 Utility for Money
requires a cortical process that “simulates”
As discussed earlier, neuroscience can the value of those goods internally to the
point out commonalities between categories brain). But it is also possible that money
that had been viewed as distinct. An example becomes what psychologists call a “primary
of this with important implications for eco- reinforcer,” which means that people value
nomics is the utility for money. The canoni- money without carefully computing what
cal economic model assumes that the utility they plan to buy with it. Neuroeconomics is
for money is indirect—i.e., that money is a not nearly advanced enough to separate these
mere counter, only valued for the goods and two roles for midbrain responses to money,
services it can procure. Thus, standard eco- but suggests the possibility that the brain
nomics would view, say, the pleasure from value of money is only loosely linked to con-
food or cocaine and the “pleasure” from sumption utility, which in turn suggests a
obtaining money as two totally different phe- wide range of potential experiments to
nomena. Neural evidence suggests, however, explore implications.
that the same dopaminergic reward circuitry An example of how consumption-driven
of the brain in the midbrain (mesolimbic sys- and primary reinforcement from money can
tem) is activated for a wide variety of differ- matter is asset pricing. Since Robert Lucas
ent reinforcers (Montague and Berns 2002), (1978), many models of stock prices have
including attractive faces (Itzhak Aharon et assumed that investors care about the utili-
al. 2001), funny cartoons (Dean Mobbs et al. ty of consumption they can finance with
2003), cultural objects like sports cars stock market returns, rather than with
(Susanne Erk et al. 2002), drugs (Schultz returns per se. Simple models of this sort
2002), and money (e.g., Hans Breiter et al. make many counterfactual predictions—
2001; Brian Knutson and Richard Peterson, most famously that the “equity premium,”
in press; Delgado et al. 2000). This suggests or marginal return to stocks over bonds,
that money provides direct reinforcement. would be much lower than it has actually
Figure 4 is a rough guess about the neural been if investors only disliked risk because
circuitry of reward (from Schultz 2002). It is of its impact on consumption, as conven-
useful as a pictorial reminder that, while neu- tional models assume. Shlomo Benartzi and
roscience (and our review) often emphasize Richard Thaler (1995) and Nicholas
specific regions which are central in different Barberis, Tano Santos, and Ming Huang
kinds of processing, the focus in thinking (2001) have had more success explaining
about economic decisions should be on cir- returns patterns using a model in which
cuitry or systems of regions and how they investors care directly about stock returns.
interact. The diagram also shows how frontal This alternative assumption fits a brain that
mr05_Article 1 3/28/05 3:25 PM Page 36

36 Journal of Economic Literature, Vol. XLIII (March 2005)

Figure 4. Hypothesized neural circuitry of reward processing (Schultz 2000).

gets a kick out of earning high returns for plans are often sold as packages, making it
their own sake. impossible to compute the cost of the indi-
If gaining money provides direct pleasure, vidual components (hotel, food, transporta-
then the experience of parting with it is tion). Often components of the package are
probably painful. While there is no direct presented as “free” (like Microsoft’s internet
evidence that paying is painful, the assump- browser) even though the claim is meaning-
tion that paying hurts can explain many mar- less from an economic standpoint, given that
ket phenomena which are otherwise the package is presented on a take-it-or-
puzzling (Drazen Prelec and Loewenstein leave-it basis. One can interpret the appeal
1998). An example is the effect of payment- of ad-hoc currencies, such as frequent-flyer
neutral pricing schemes on choices. miles, chips in casinos, or the beads used for
Companies often go to great lengths to dis- incidental expenses at all-inclusive resorts
guise payments, or reduce their pain. like Club Med, as an attempt to reduce the
Consumers appear to oversubscribe to flat- pain-of-payment. The ad-hoc currency,
rate payment plans for utilities and tele- whether miles or beads, feels like “play
phone service and health clubs (Kenneth money,” and spending it does not seem to
Train 1991; Train, Daniel McFadden, and exact the same psychic cost.
Moshe Ben-Akiva 1987; Della Vigna and In experiments, we have observed a prefer-
Ulrike Malmendier 2003, Anja Lambrecht ence for prepayment for certain items, even
and Bernd Skiera 2004). A flat-rate plan where prepayment is financially irrational
eliminates marginal costs and allows con- because it incurs an opportunity cost of lost
sumers to enjoy the service without thinking interest payments (Prelec and Loewenstein
about the marginal cost. Similarly, travel 1998). When questioned, respondents claim
mr05_Article 1 3/28/05 3:25 PM Page 37

Camerer, Loewenstein, and Prelec: Neuroeconomics 37

to prefer prepayment for, e.g., a vacation trip experiments create a situation where the
because then they could relax and enjoy the utility of food and disutility of work remain
vacation more knowing it was paid up. It is an the same, but the amount of work-for-
interesting question whether part of the reward goes up. This implies that it is possi-
motive for owning rather than renting prod- ble to be motivated to take actions that bring
ucts is precisely to create prepayment, and so no pleasure.
enjoy consumption without being reminded of Berridge believes that the later stages of
the cost. The pain-of-paying may also explain many drug addictions presents prototypical
why we are willing to pay less for a product if examples of situations of what he terms
paying in cash than by credit card. Although “wanting” without “liking”; drug addicts
there are financial reasons to prefer paying by often report an absence of pleasure from
credit-card, the size of credit card relative to taking the drugs they are addicted to, cou-
cash premium that people reveal, in an exper- pled with an irresistible motivation to do so.
iment, is much too high (up to 100 percent) to Other examples of situations in which there
be rationalized by liquidity preference and often seems to be a disconnect between
other economic considerations (Prelec and one’s motivation to obtain something and
Duncan Simester 2001). the pleasure one is likely to derive from it
are sex and curiosity (Loewenstein 1994).
4.4 Wanting and Liking
Thus, for example, you can be powerfully
Economists usually view behavior as a motivated to seek out information, even
search for pleasure (or, equivalently, escape when you are quite certain that it will make
from pain). The subfield of welfare econom- you miserable, and can feel quite unmotivat-
ics, and the entire ability of economists to ed to engage in activities that, at a purely
make normative statements, is premised on cognitive level, you are quite sure you would
the idea that giving people what they want find deeply pleasurable.
makes them better off. But, there is consid- Economics proceeds on the assumption
erable evidence from neuroscience and that satisfying people’s wants is a good thing.
other areas of psychology that the motivation This assumption depends on knowing that
to take an action is not always closely tied to people will like what they want. If likes and
hedonic consequences. wants diverge, this would pose a fundamen-
Berridge (1996) argues that decision mak- tal challenge to standard welfare economics.
ing involves the interaction of two separate, Presumably welfare should be based on
though overlapping systems, one responsi- “liking.” But if we cannot infer what people
ble for pleasure and pain (the “liking” sys- like from what they want and choose, then
tem), and the other for motivation (the an alternative method for measuring liking
“wanting” system). This challenges the fun- is needed, while avoiding an oppressive
damental supposition in economics that one paternalism.
only strives to obtain what one likes.
4.5 Cognitive Inaccessibility
Berridge finds that certain lesions and phar-
macological interventions can selectively The fact that people lack introspective
enhance a rat’s willingness to work for a access to the sources of their own judg-
food, without changing the pleasure of eat- ments of behavior, and tend to overattribute
ing the food, as measured, admittedly some- both to controlled processes, has many
what questionably, by the animal’s facial important implications for economics.
expression (Animal facial expressions, like When it comes to discriminatory biases, for
those of humans, provide at least a clue example, because people lack introspective
about whether something tastes good, bad access to the processes that produce such
or indifferent. In economic language, the biases, they are unable to correct for them
mr05_Article 1 3/28/05 3:25 PM Page 38

38 Journal of Economic Literature, Vol. XLIII (March 2005)

even when they are motivated to make choose one way or another. Paradoxically, this
impartial judgments and decisions. Indeed, can be very beneficial. In many situations, an
they are very likely to deny that they are action may be diagnostic of a good outcome,
biased and, hence, likely to not even per- without having any significant ability to cause
ceive the need for such correction. Such that outcome. For instance, participating in
unconscious discrimination may explain some socially desirable activity (e.g., voting or
why otherwise identical job application not littering) may be quite diagnostic of the
resumes for candidates with statistically desired collective outcome (your candidate
“white” rather than “African–American” winning, clean streets) without being able to
names have a 50 percent higher chance of cause that outcome, because the impact of
generating callbacks from job application your action is negligible. This gives rise to the
letters, as shown recently by Marianne notorious “voter’s paradox” in rational choice
Bertrand and Mullainathan (2004). It can theory, which seems to imply that no one
also help explain why physicians are so con- should ever vote. Fortunately, the distinction
vinced that gifts from pharmaceutical com- between diagnosticity and causality, which is
panies do not bias their prescription absolute in the rational-choice model, is quite
practices even though research (and the fuzzy psychologically. There is experimental
continuing practice of gift giving by the evidence that people will cheat on their own
companies) suggests that it does (Jason medical tests so as to “manufacture” a good
Dana and Loewenstein 2003). diagnosis (George Quattrone and Amos
A second class of implications is related to Tversky 1984), and their own personality
the phenomena of apparent self-deception inventories, to yield a personality assessment
and self-manipulation, where, for instance, diagnostic of success. This can only be possi-
economic agents (investors, consumers, ble if the true motive for action — the desire
entrepreneurs) are overly optimistic about to get “good news” — is hidden from the
their chances for success. These phenomena agent at the moment of choice; if it were not
have been richly catalogued by social psy- hidden, then awareness that the action was
chologists, beginning with the research on taken precisely to get the good news would
motivated cognition and cognitive disso- instantly void the diagnostic value of the
nance in the 1950s. Neuroscience suggests action (see Ronit Bodner and Prelec 2003 for
that they are all related to chronic cognitive an economic self-signaling model that allows
inaccessibility of automatic brain processes. for noncausal motives on action). Cognitive
Attention, for example, is largely controlled inaccessibility prevents this logical short-cir-
by automatic processes, and attention in cuit and dramatically expands the range of
turn determines what information we motives that can influence behavior. Because
absorb. If attention is chronically drawn to the hedonic system (quadrant IV) is not con-
information that is favorable to us, we will strained by logical considerations, a person
emerge with an overoptimistic sense of our who takes a small step toward a larger virtu-
abilities and prospects. It is a case of quad- ous objective, such as joining a gym to get fit,
rants III and IV collaborating, without the or buying a copy of Stephen Hawking’s A
“adult supervision” provided by quadrant I. Brief History of Time to feel scientifically lit-
A third class of implications arises from the erate, may experience a feeling of pleasure
cognitive inaccessibility of our own motives from the small step “as if” they have in fact
for action. The fact that we are not con- secured their objective (which they most like-
sciously aware of the moment of decision (as ly will not). At the same time, cognitive inac-
shown by Libet’s research, mentioned in the cessibility means that quadrant I doesn’t have
previous section) strongly suggests that we to acknowledge that the feeling of pleasure
may also not understand the reasons why we was in fact the cause of his action.
mr05_Article 1 3/28/05 3:25 PM Page 39

Camerer, Loewenstein, and Prelec: Neuroeconomics 39

5. Specific Economic Applications achieves this function in part by motivating

individuals to take certain actions. In most ani-
We now go into greater detail about rami- mals, emotions and drives motivate behaviors
fications of neuroscience findings for four that have short-term goals, such as eating,
specific topics in economics: intertemporal drinking, and copulating. As a result, the
choice, decision making under risk and affective systems that we share with wide
uncertainty, game theory, and labor-market range of other animals are inherently myopic.
discrimination. Though some animals display far-sighted
5.1 Intertemporal Choice and Self-Control behaviors, such as storing food for winter,
these are specialized, preprogrammed, behav-
The standard perspective in economics iors that are distinctly different from the type
views intertemporal choice as a trade-off of of spontaneous delay of gratification observed
utility at different points in time. Individual in humans. Humans appear to be unique
differences in the way that people make this among animals in terms of caring about, mak-
tradeoff are captured by the notion of a dis- ing immediate sacrifices for, and flexibly
count rate—a rate at which people discount responding to, desired future consequences.
future utilities as a function of when they This capacity to take long-term conse-
occur. The notion of discounting, however, quences of our behavior into account seems
gained currency not because of any support- to be the product of our prefrontal cortex,
ive evidence, but based only on its conven- which, tellingly, is the part of the brain that
ient similarity to financial net present value is uniquely human (see, e.g., Manuck et al.
calculations (Loewenstein 1992). Indeed, in 2003). Patients with damage to prefrontal
the article that first proposed DU in detail, regions—most famously Phineas Gage,
Samuelson (1937) explicitly questioned its whose injury led to a reformulation of our
descriptive validity, saying “It is completely understanding of the function of the pre-
arbitrary to assume that the individual frontal cortex—tend to behave myopically,
behaves so as to maximize an integral of the paying little heed to the delayed conse-
form envisaged in [the DU model].” quences of their behavior. As Hersh Shefrin
In fact, more recent empirical research on and Thaler (1988) suggested many years ago,
time discounting challenges the idea that intertemporal choice can be viewed as a
people discount all future utilities at a con- splice of two processes—an impulsive, affec-
stant rate (see Frederick, Loewenstein, and tive, process and a more far-sighted process
O’Donoghue 2002). The notion of time dis- guided by the prefrontal cortex.
counting, it seems, neither describes the Recent brain imaging research provides
behavior of individuals nor helps us to classi- support for such an account. Samuel
fy individuals in a useful fashion. How can an McClure et al. (2004) scanned subjects using
understanding of the brain help us to estab- fMRI while they made a series of preference
lish a better understanding of intertemporal judgments between monetary reward options
choice behavior? Our two central distinc- that varied by amount and delay to delivery.
tions, between affect and cognition, and For some pairs of choices, the earlier reward
between automatic and controlled processes, would be received immediately; for others,
both have important ramifications. both rewards were delayed (though one by
5.1.1 Affect and Cognition in Intertemporal more than the other). Consistent with the
Choice idea that intertemporal choice is driven by
two systems, one more cognitive and one
As discussed above, the affective system is more deliberative, they found that parts of
designed to ensure that certain survival and the limbic—i.e., affective—system associated
reproduction functions are met and it with the midbrain dopamine system were
mr05_Article 1 3/28/05 3:25 PM Page 40

40 Journal of Economic Literature, Vol. XLIII (March 2005)

preferentially activated by options involving (low-load) were given a two-digit number that
immediately available rewards. In contrast, they were instructed to memorize, and the
regions of the lateral prefrontal cortex and other half (high load) were given a seven-digit
posterior parietal cortex—typically viewed as number to memorize. Subjects were then
more cognitive regions—were engaged uni- instructed to walk to another room in the
formly by intertemporal choices irrespective building. On the way they passed by a table at
of delay. Furthermore, the relative activity of which they were presented with a choice
the two systems actually predicted the choic- between a caloric slice of cake or a bowl of
es that subjects made: Greater relative activ- fruit salad. More than half (59 percent) chose
ity in affective systems was associated with the cake in the high load (seven-digit) condi-
choosing earlier rewards more often. tion, but only 37 percent chose the cake in the
The notion of quasi-hyperbolic time dis- low load (two-digit number) condition. This
counting, of course, provides a mathematical finding is consistent with the idea that the
representation of precisely such a splicing of effort required to memorize seven-digit num-
two processes, and has been shown to use- bers drew deliberative resources away from
fully describe behavior in a wide range of self-control, leading those subjects who had
domains. However, an understanding of the to remember more to eat more cake.
neural underpinnings of these dual process- Second, prior exercise of self-control
es allows for more nuanced predictions and seems to diminish the capacity and, hence,
formulations. The notion of hyperbolic time- propensity to exert self-control in the present.
discounting predicts that people will always Recall that the prefrontal cortex is the part of
behave impulsively when faced with the the brain that is associated with a subjective
right combination of incentives (typically feeling of effort. It is tempting to attribute
those involving some immediate cost and this to the fact that self-control involves the
benefit), but this does not seem to be the same part of the brain—the executive pre-
case. Understanding that hyperbolic time frontal cortex—that is itself associated with
discounting stems, in part, from competition feelings of mental effort. Perhaps this is why
between the affective and cognitive systems, exercising willpower feels so difficult, and
leads to the prediction that factors that why exercising self-control in one domain can
strengthen or weaken one or the other of undermine its exercise in another, as demon-
these influences will cause people to behave strated by a series of clever experiments con-
more or less impulsively. ducted by Baumeister and colleagues (see,
e.g., Baumeister and Kathleen Vohs 2003). In
5.1.2 Determinants of the Relative Strength
a typical study, subjects on diets who resisted
of Affect and Cognition
temptation (by foregoing the chance to grab
There are a variety of factors that affect the snacks from a nearby basket) later ate more
relative strength of affective and cognitive ice cream in an ice-cream taste test and also
influences on intertemporal choice which quit earlier when confronted with an intellec-
can help to explain what could be called tual problem they couldn’t solve. They acted
“intraindividual” variability in impatience. as if their ability to resist temptation was tem-
First, any factor that increases the demands porarily “used up” by resisting the snacks (or,
on the prefrontal cortex—on the controlled, alternatively, that they had “earned” a reward
cognitive, system, should decrease the influ- of ice cream by skipping the tempting
ence of this system and, hence, decrease indi- snacks). Other factors that seem to under-
viduals’ control over their own behavior. This mine this self-control resource are alcohol,
possibility was demonstrated by Baba Shiv stress, and sleep deprivation.
and Alexander Fedorikhin (1999). To manip- Turning to the other side of the equation,
ulate “cognitive load,” half of their subjects activation of affective states should, by the
mr05_Article 1 3/28/05 3:25 PM Page 41

Camerer, Loewenstein, and Prelec: Neuroeconomics 41

same token, tend to accentuate temporal Tranel, and Damasio 1997; Damasio 1994)
myopia. Indeed, there is considerable evi- lends further credence to this perspective,
dence of such effects (Janet Metcalfe and as does research on psychopaths, who are
Mischel 1999). Research has shown, for characterized by both emotional deficits
example, that addicts display higher discount when it comes to imagining the future and
rates, not only for drugs but also for money, by insensitivity to the future consequences
when they are currently craving drugs than (as well as consequences to others) of their
when they are not (Giordano et al. 2002). behavior (Hervey Cleckley 1941; Hare
Other research has shown that sexual arousal 1965, 1966; David Lykken 1957). Whether
produces greater time-discounting of money deliberately, as when one conjures up an
rewards (Ariely and Loewenstein 2003). image of a “fat self” exposed on the beach,
The main exceptions to the rule that affec- or without conscious intention, self-control
tive states tend to engender short-sighted often involves an interaction of affective and
behavior involve interactions between the cognitive mechanisms.
cognitive and affective systems. In fact, deci- How might one model intertemporal
sions to delay gratification often involve a choice differently as a result of the insights
mixture of affect and cognition. They from neuroscience? First, the neuroscience
require a cognitive awareness of the delayed research points to ways to “unpack” the con-
benefits in delaying gratification—e.g., that cept of time preference. Clearly, ability to
desisting from eating cake today will mean a think about future consequences is impor-
more pleasing body type in the future. But, tant, which is probably why time preference
as many researchers have observed, cogni- is correlated with measured intelligence
tive awareness alone is insufficient to moti- (Mischel and Robert Metzner 1962).
vate delay of gratification; emotions play a Second, because people are likely to make
critical role in forward-looking decision myopic choices when under the influence of
making. As Barlow (1988) notes, “The capac- powerful drives or emotions (Loewenstein
ity to experience anxiety and the capacity to 1996), this suggests that a key to under-
plan, are two sides of the same coin.” standing impulsivity in individuals might be
Thomas Cottle and Stephen Klineberg to understand what types of situations get
(1974), similarly argue that people care them “hot.” Third, we might be tempted to
about the delayed consequences of their look for individual differences in what could
decisions only to the degree that contem- be called “willpower”—i.e., the availability
plating such consequences evokes immedi- of the scarce internal resource that allows
ate affect. In support of this view, they cite people to inhibit viscerally driven behaviors
the effects of frontal lobotomies which harm (see Loewenstein and O’Donoghue 2004 for
areas of the brain that underlie the capacity a recent two-system model of intertemporal
for images of absent events to generate choice and other economic behaviors).
experiences of pleasure or discomfort. The A model of intertemporal choice that took
neurosurgeons who performed these opera- account of interactions between affect and
tions wrote of their frontal-lobotomy cognition can help to explain not only impul-
patients that: “the capacity for imagination is sivity, but also why many people have self-
still present, and certainly not sufficiently control problems of the opposite type of
reduced to render the patients helpless, and those typically examined in the literature—
affective responses are often quite lively, e.g., tightwads who can’t get themselves to
[but there is] a separation of one from the spend enough; workaholics who can’t take a
other” (Freeman and Watts 1942). break, and people who, far from losing con-
The work of Damasio and colleagues dis- trol in the bedroom, find themselves frus-
cussed earlier (Bechara, Damasio, Daniel tratingly unable to do so. All of these patterns
mr05_Article 1 3/28/05 3:25 PM Page 42

42 Journal of Economic Literature, Vol. XLIII (March 2005)

of behavior can easily be explained by the subjects to choose whether to eat a fancy
possibly uniquely human propensity to expe- French Restaurant dinner in one month or
rience emotions, such as fear, as a result of two, and others to choose between the
thinking about the future. Indeed, it is likely sequence [fancy French this month, eat at
that one of the main tools that the prefrontal home next month] or the same choice in
cortex uses to impose self-control when reverse order. A majority of subjects preferred
affective forces would otherwise favor short- the isolated French dinner earlier but
sighted self-destructive behavior is to create deferred the French dinner when it was
“deliberative affect” via directed imagery and embedded in a sequence with eating at home.
thought (Roger Giner-Sorolla 2001). Frederick and Loewenstein (2000) conducted
Such a framework might also help to a study in which subjects were presented with
explain why people appear so inconsistent a series of intertemporal choices that were
when their behavior is viewed through the framed in ways intended to evoke different
lens of discounted utility. The ability to think considerations. For example, in one version
about future consequences may not be they asked respondents to allocate pleasurable
strongly correlated with the degree to which outcomes, such as massages, over time. They
different experiences produce visceral reac- expected, and found, that the allocation for-
tions, and these in turn might not be corre- mat evoked a choice heuristic that caused peo-
lated with an individual’s level of willpower. ple to spread consumption relatively evenly
Indeed, Loewenstein et al. (2001) found over time, implying a preference for flat
close to zero correlations between numerous sequences. In another version, they asked
behaviors that all had an important intertem- respondents to state a maximum buying price
poral component, but much higher correla- for the Greek-then-French and French-then-
tions between behaviors that seemed to Greek sequences, rather than choose between
draw on the same dimension of intertempo- them, anticipating that the mention of money
ral choice—e.g., which required suppression would evoke considerations of the time value
of specific emotions such as anger. of money and, hence, cause subjects to place
higher value on the option that provided
5.1.3 Automatic Processes in Intertemporal
greater value earlier—i.e., the declining
sequences—which is what they found.20
To the extent that intertemporal choice is,
in fact, driven by cognitive considerations,
much of this cognition does not take the usu- Another phenomenon that may be driven by such
ally assumed form of a weighing of costs and automatic processes is the “diversification bias” (Itamar
Simonson 1989). When people choose several alternatives
benefits discounted according to when they from a set, they choose more variety when they choose
occur in time. Rather, consistent with the ear- them all simultaneously than when they choose them
lier claim that people often make decisions via sequentially. This phenomenon has been demonstrated
with snack food, audio pieces, gambles, and lottery tick-
a two-part process that takes the form of first ets. Daniel Read and Loewenstein (1995) tested various
asking “what situation am I in?” then continu- explanations for overdiversification (e.g., they diversify in
ing “how does one behave in such a situation,” simultaneous choice to gather information), but conclud-
ed that it results from a rule of thumb that that they apply
much intertemporal choice is driven by auto- whenever choices are expressed in a fashion that high-
matic processes involving pattern-matching, lights diversification (see also Thomas Langer and Craig
recognition, and categorization. Fox 2004). This viewpoint is supported by a study (Read,
Loewenstein, and S. Kalyanaraman 1999) in which sub-
A pattern of choice that seems to be driven jects make successive choices between groups of objects
by such a process is the preference for which are easily categorized or not. When categorization
sequences of outcomes that improve over was easy (e.g., virtues and vices), subjects diversified
more in simultaneous than in sequential choice. When
time. In one demonstration of this effect, there were multiple competing categorizations, the
Loewenstein and Prelec (1993) asked some overdiversification bias disappeared.
mr05_Article 1 3/28/05 3:25 PM Page 43

Camerer, Loewenstein, and Prelec: Neuroeconomics 43

In sum, neuroscience points to some defi- it comes to risks; we drive (or wish we were
ciencies in the way that economists currently driving as we sit white-knuckled in our air-
model intertemporal choice and also suggests plane seat) when we know at a cognitive level
directions for future modeling. A somewhat that it is safer to fly. We fear terrorism, when
stylized interterpretation of the results just red meat poses a much greater risk of mor-
reviewed would be that some intertemporal tality. And, when it comes to asking someone
decisions are, in fact, well represented by the out on a date, getting up to speak at the podi-
discounted utility model—specifically those um, or taking an important exam, our delib-
involving detailed deliberation but minimal erative self uses diverse tactics to get us to
affect. However, a wide range of other take risks, or to perform in the face of risks,
intertemporal choices are influenced by that our visceral self would much prefer to
affectively “hot” processes such as drives and avoid. Perhaps the most dramatic illustra-
emotions, or result from processes that auto- tions of the separation of visceral reactions
matically evoke a response which depends on and cognitive evaluations, however, comes
the situation. Models which focus on how from the phobias that so many people suffer
these discrepant processes interact are prom- from; the very hallmark of a phobia is to be
ising (e.g., Bernheim and Rangel 2004; unable to face a risk that one recognizes,
Loewenstein and O’Donoghue 2004). objectively, to be harmless. Moreover, fear
unleashes preprogrammed sequences of
5.2 Decision-Making under Risk and
behavior that aren’t always beneficial. Thus,
when fear becomes too intense it can pro-
Both collaboration and competition duce counterproductive responses such as
between affect and cognition, and between freezing, panicking, or “dry-mouth” when
controlled and automatic processes, can also speaking in public. The fact that people pay
be seen in the domain of decision making for therapy to deal with their fears, and take
under risk and uncertainty. drugs (including alcohol) to overcome them,
can be viewed as further “evidence” that peo-
5.2.1 Affect versus Cognition
ple, or more accurately, people’s deliberative
The expected utility model views decision selves, are not at peace with their visceral
making under uncertainty as a tradeoff of reactions to risks.
utility under different states of nature—i.e.,
5.2.2 Affective Reactions to Uncertainty
different possible scenarios. But, much as
they do toward delayed outcomes, people A lot is known about the neural processes
react to risks at two different levels. On the underlying affective responses to risks.
one hand, as posited by traditional econom- Much risk averse behavior is driven by
ic theories and consistent with quadrant I of immediate fear responses to risks, and fear,
table 1, people do attempt to evaluate the in turn, seems to be largely traceable to the
objective level of risk that different hazards amygdala. The amygdala constantly scans
could pose. On the other hand, and consis- incoming stimuli for indications of potential
tent with quadrant IV, people also react to threat and responds to inputs both from
risks at an emotional level, and these emo- automatic and controlled processes in the
tional reactions can powerfully influence brain. Patrik Vuilleumier et al. (2001)
their behavior (Loewenstein, Weber, observed equivalent amygdala activation in
Christopher Hsee, and Ned Welch 2001). response to fearful faces that were visually
The existence of separate affective and attended to or in the peripheral region which
cognitive systems that respond differently to falls outside of conscious perception (cf. de
risks is most salient when the two systems Beatrice de Gelder, Jean Vroomen, Gilles
clash. People are often “of two minds” when Pourtois, and Lawrence Weiskrantz 1999;
mr05_Article 1 3/28/05 3:25 PM Page 44

44 Journal of Economic Literature, Vol. XLIII (March 2005)

Nouchine Hadjikhani and de Gelder 2003; and affective systems) and normal subjects
LeDoux 1996; John Morris, C. Buchel, and chose a sequence of cards from four decks
Raymond Dolan 2001; Whalen et al. 1998). whose payoffs the subjects only learned from
But the amygdala also receives cortical experience (a “multiarmed bandit” prob-
inputs, which can moderate or even override lem). Two decks had more cards with
its automatic quadrant IV response. extreme wins and losses (and negative
In a paradigmatic experiment that illus- expected value); two decks had less extreme
trates cortical overriding of amygdala activa- outcomes but positive expected value. Both
tion (LeDoux 1996), an animal such as a rat groups exhibited similar skin conductance
is “fear-conditioned”—by repeatedly admin- (sweating—an indication of fear) after large-
istering a signal such as a tone followed by loss cards were encountered, but, compared
administration of a painful electric shock. to normals, prefrontal subjects rapidly
Once the tone becomes associated in the returned to the high-paying risky decks after
animal’s mind with the shock, the animal suffering a loss and, as a result, went “bank-
responds to the tone by jumping or showing rupt” more often. Although the immediate
other over signs of fear. In the next phase of emotional reaction of the prefrontal patients
the experiment, the tone is played repeated- to losses was the same as the reaction of nor-
ly without administering the shock, until the mals (measured by skin conductance), the
fear response becomes gradually “extin- damaged patients apparently do not store
guished.” At this point, one might think that the pain of remembered losses as well as nor-
the animal has “unlearned” the connection mals, so their skin conductance rose much
between the tone and the shock, but the less than normals when they resampled the
reality is more complicated and interesting. high risk decks. Subsequent research found a
If the neural connections between the cortex similar difference between normal subjects
and the amygdala are then severed, the orig- who were either high or low in terms of emo-
inal fear response to the tone reappears, tional reactivity to negative events. Those
which shows that fear conditioning is not who were more reactive were more prone to
erased in “extinction” but is suppressed by sample from the lower-paying, safer decks of
the cortex and remains latent in the amyg- cards (Peters and Slovic 2000).
dala. This suggests that fear learning may be Damasio et al.’s research shows that
permanent, which could be an evolutionarily insufficient fear can produce nonmaximiz-
useful adaptation because it permits a rapid ing behavior when risky options have nega-
relearning if the original cause of the fear tive value. But, it is well established that
reappears. Indeed, other studies (reported fear can also discourage people from taking
in the same book) show that fear-condition- advantageous gambles (see, e.g., Uri
ing can be reinstated by the administration Gneezy and Jan Potters 1997). Indeed, Shiv
of a single shock. et al. (2005) found that frontal patients
Decision making under risk and uncer- actually make more money on a task in
tainty, like intertemporal choice, nicely illus- which negative emotions cause normal sub-
trates both collaboration and competition jects to be extremely risk averse: a series of
between systems. When it comes to collabo- take-it-or-leave-it choices to play a gamble
ration, risk taking (or avoiding) behavior with a 50 percent chance of losing $1.00 or
involves an exquisite interplay of cognitive gaining $1.50. Normal subjects and frontal
and affective processes. In a well-known subjects were about equally likely to play
study that illustrates such collaboration the gamble on the first round, but normals
(Bechara et al. 1997), patients suffering pre- stopped playing when they experienced
frontal damage (which, as discussed above, losses, while frontal patients kept playing.
produces a disconnect between cognitive Clearly, having frontal damage undermines
mr05_Article 1 3/28/05 3:25 PM Page 45

Camerer, Loewenstein, and Prelec: Neuroeconomics 45

the overall quality of decision making; but et al. 2002) which corroborates the subjects’
there are situations in which frontal damage self-reports. Using fMRI, Ming Hsu et al.
can result in superior decisions. (2005) found frontal insula and amygdala
At a more macro level, emotional reac- activation when subjects faced ambiguous
tions to risk can help to explain risk-seeking choices, compared to risky ones. They also
as well as risk-aversion (Caplin and Leahy found that patients with orbitofrontal corti-
2001). Thus, when gambling is pleasurable, a cal (OFC) lesions are ambiguity-neutral,
model that incorporates affect naturally pre- compared to brain-damaged controls. Since
dicts that people will be risk-seeking and the OFC receives input from the limbic sys-
that self-control will be required to rein in tem (including the insula and amygdala), the
risk-taking. Indeed, about 1 percent of the fMRI and lesion evidence together imply
people who gamble are diagnosed as “patho- that in normal subjects, ambiguous gambles
logical”—they report losing control, “chasing often create discomfort or fear which is
losses,” and harming their personal and work transmitted to the OFC. Ironically, patients
relationships by gambling (National with OFC brain damage therefore behave
Academy of Sciences 1999). The standard more “rationally” (treating ambiguous and
economic explanations for gambling—con- risky gambles similarly) than normals, a
vex utility for money or a special taste for the reminder that logical principles of rationality
act of gambling—don’t help explain why and biological adaptations can be different
some gamblers binge and don’t usefully (cf. Shiv et al. 2005).
inform policies to regulate availability of
5.2.3 Automatic Versus Controlled
gambling. Neuroscience may help.
Pathological gamblers tend to be over-
whelmingly male and tend to also drink, The divergence between different sys-
smoke, and use drugs much more frequent- tems’ evaluations of risk can also be seen
ly than average. Genetic evidence shows that when it comes to judgments of probability.
a certain gene allele (D2A1), which causes Numerous studies by psychologists have
gamblers to seek larger and larger thrills to observed systematic divergences between
get modest jolts of pleasure, is more likely to explicit judgments of probability in different
be present in pathological gamblers than in settings (presumably the product of con-
normal people (David Comings 1998). One trolled processing) and implicit judgments
study shows tentatively that treatment with or judgments derived from choice (which
naltrexone, a drug that blocks the operation are more closely associated with automatic
of opiate receptors in the brain, reduces the processing and/or emotion). For example,
urge to gamble (e.g., Paula Moreyra et al. Kirkpatrick and Epstein (1992) found that
2000). The same drug has been used to suc- people prefer to draw a bean from a bowl
cessfully treat “compulsive shopping” (Susan containing ten winning beans and ninety
McElroy et al. 1991). losing beans than from a bowl containing
Neural evidence also substantiates the dis- one winning bean and nine losing beans (see
tinction between risk (known probability) also Veronika Denes-Raj, Epstein, and
and “Knightian” uncertainty, or ambiguity. Jonathon Cole 1995; Paul Windschitl and
Subjects facing ambiguous gambles—know- Gary Wells 1998). Subjects say that they
ing they lack information they would like to know the probabilities of winning are the
have about the odds—often report a feeling same, but they still have an automatic quad-
of discomfort or mild fear. Brain imaging rant III preference for the bowl with more
shows that different degrees of risk and winning beans.
uncertainty activate different areas of the An important feature of good probability
brain (McCabe et al. 2001; Aldo Rustichini judgment is logical coherence: probabilities
mr05_Article 1 3/28/05 3:25 PM Page 46

46 Journal of Economic Literature, Vol. XLIII (March 2005)

of mutually exclusive and exhaustive events whether any theory that fails to incorporate
should add to one, and conditional probabil- the affective dimensions of risk will be capa-
ities should be linked to joint and marginal ble of shedding much light on such important
probability according to Bayes rule phenomena as stock market booms and busts,
(P(A|B)=P(A and B)/P(B)). Logical coher- the ubiquity of gambling (e.g., slot machines
ence is violated in at least two neurally revenues dwarf revenues from movies), and
interesting ways. One is “conjunction falla- the vicissitudes of public responses to threats
cy”—the tendency to judge events with two as diverse as terrorism and global warming.
components A and B as more likely than A
5.3 Game Theory
or B alone. While most subjects (even statis-
tically sophisticated ones) make conjunction Neuroscientific data are well-suited to
errors on some problems, when those errors exploring the central assumptions on which
are pointed out, quadrant I wakes up, and game theory predictions rest. These assump-
the subjects sheepishly recognize the error tions are that players: (1) have accurate
and correct it (Kahneman and Frederick beliefs about what others will do (i.e., play-
2002). For example, the famous “Linda” ers are in equilibrium); (2) have no emotions
problem describes an earnest young politi- or concern about how much others earn (a
cally minded student. In one condition, sub- useful auxiliary assumption); (3) plan ahead;
jects are asked to rank statements about and (4) learn from experience.
Linda—Is she a bank teller? A feminist bank
5.3.1 Theory of Mind and Autism
teller? A large majority of subjects, even
highly educated ones, say Linda is more In strategic interactions (games), knowing
likely to be a feminist bank teller than she is how another person thinks, and how anoth-
to be a bank teller, a violation of the con- er person thinks you think, etc., is critical to
junction principle. But when subjects are predicting the other person’s behavior (and
asked, “Out of one hundred people like for inferring the other player’s intentions,
Linda, how many are bank tellers? Feminist which underlie emotional judgments of fair-
bank tellers?” conjunction errors disappear ness and obliged reciprocity in more mod-
(Tversky and Kahneman 1983). ern theories (e.g., Rabin 1993). From a
Another violation is that subjects often neural view, iterated strategic thinking con-
report probabilities which are logically inco- sumes scarce working memory and also
herent. fMRI evidence suggests an explana- requires a player to put herself in another
tion for why probability judgments are player’s “mind.” There may be no generic
incoherent, but can be corrected upon human capacity to iterate this kind of think-
reflection: when guessing probabilities, the ing beyond a couple of steps.21 Studies that
left hemisphere of the brain is more active; examine either subject’s choices, or that
but when answering logic questions, the monitor what type of information subjects
right hemisphere is more active (Lawrence look up or pay attention to in experimental
Parsons and Daniel Osherson 2001). Since games, suggest only one–two steps of strate-
enforcing logical coherence requires the gic thinking are typical in most populations
right hemisphere to “check the work” of the (e.g., Eric Johnson et al. 2002; Miguel
left hemisphere, there is room for slippage. Costa-Gomes et al. 2001; Camerer, Teck
Once again, it can be seen that neuro- Ho, and Kuan Chong 2004), though up to
science, and specifically, a consideration of
affective and automatic processes that have
been largely neglected by economists, could Of course, equilibration might occur through a
process other than introspection—e.g., adaptive learning,
potentially inform an important line of imitation, communication, or evolution. But these processes
research and theory. We are skeptical of must have a neural basis too.
mr05_Article 1 3/28/05 3:25 PM Page 47

Camerer, Loewenstein, and Prelec: Neuroeconomics 47

three–four steps are observed in analytically Ironically, while the subject’s reasoning
skilled and specially trained populations. matches exactly how conventional game
As discussed in section 4.2, many neuro- theory approaches the game, it also sounds
scientists believe there is a specialized autistic, because this subject is surprised
“mind-reading” (or “theory of mind”) area, and perplexed by how normal people
perhaps in prefrontal area Brodmann 10, behave.
which generates reasoning about what oth- This anecdote is complemented by
ers believe and might do (e.g., Simon Baron- Elizabeth Hill and David Sally’s (2003)
Cohen 2000).22 Autism is thought to be a extensive comparison of normal and autistic
deficit in this area (and related circuitry). children and adults playing ultimatum
People with autism often have trouble figur- games. Nearly half of the autistic children
ing out what other people think and believe, offer zero or one unit (out of ten) in the ulti-
and are consequently puzzled by behavior matum game, and relatively few offer half.
that most people would consider normal. Many autistic adults also offered nothing,
One tool used in behavioral game theory is but a large number of autistic adults offered
the “ultimatum game.” In this game, a “pro- half—as if they had developed a reasoning
poser” offers a division of a sum of money, or experiential workaround which tells
generically $10.00, to another “responder” them what other people think is fair in
who can accept or reject it, ending the game. games that involve sharing, even though
If the responder has no emotional reaction they cannot guess what others will do using
to the fact that the proposer is earning more normal circuitry.
than she is (“envy” or “disgust”), then the McCabe et al. (2001) used fMRI to meas-
responder should accept the smallest offer. ure brain activity when subjects played
If the proposer also has no emotional reac- games involving trust, cooperation, and pun-
tion to earning more (“guilt”) and anticipates ishment. They found that players who coop-
correctly what the responder will do, then erated more often with others showed
the proposer should offer the lowest increased activation in Brodmann area 10
amount. This pattern is rarely observed: (thought to be one part of the mind-reading
Instead, in most populations the proposer circuitry) and in the thalamus (part of the
offers 40–50 percent and about half the emotional “limbic” system). Players who
responders reject offers less than 20 percent. cooperated less often showed no systematic
When players do follow the dictates of activation.
game theory, the result can be a low payoff Bhatt and Camerer (in press) used fMRI
and confusion. Consider this quote from an to compare brain activity when subjects
upset subject, an Israeli college student, make choices and express beliefs in matrix
whose low offer in a $10.00 ultimatum game games. They found that when players were
was rejected (from Shmuel Zamir 2000): in equilibrium (choices were best-responses
and beliefs were accurate), the same circuit-
I did not earn any money because all the other ry was being used when making a choice
players are stupid! How can you reject a positive and expressing a belief. This means equilib-
amount of money and prefer to get zero? They
just did not understand the game! You should rium is a “state of mind,” which can be iden-
have stopped the experiment and explained it to tified by a tight overlap in activity in the two
them . . . tasks, as well as a mathematical restriction
on best response and belief accuracy. They
also found evidence suggesting that, when
22 subjects guessed what beliefs other subjects
Grether et al. (2004) also find BA 10 activity, along
with interesting activity in anterior cingulate and the basal had about their own behavior, they tended
forebrain, as subjects bid in second-price auctions. to anchor on their own choices. This means
mr05_Article 1 3/28/05 3:25 PM Page 48

48 Journal of Economic Literature, Vol. XLIII (March 2005)

beliefs about beliefs are not just iterations Ernst Fehr and Simon Gachter 2000;
of some belief processing mechanism; Camerer 2003). The fact that unfair offers
instead, self-referential beliefs use compo- activate insula means that a verbal statement
nents of circuitry for forming beliefs and for like “I am so disgusted about being treated
making choices. that way” is literal, not metaphorical—they
really do feel disgusted.24
5.3.2 Emotions and Visceral Effects
Zak et al. (2003) explored the role of hor-
One of the most striking neuroscientific mones in trust games. In a canonical trust
findings about game theory comes from game, one player can invest up to $10.00,
Sanfey et al.’s (2003) fMRI study of ultima- which is tripled. A second “trustee” player
tum bargaining. By comparing the brains of can keep or repay as much of the tripled
subjects responding to unfair ($1.00–$2.00 investment as they want. Zak et al. measured
out of $10.00) and fair ($4.00–5.00) offers, eight hormones at different points in the
they found that very unfair offers differen- trust game. The hormone with the largest
tially activated three regions: Dorsolateral effect was oxytocin—a hormone that rises
prefrontal cortex (DLPFC), anterior cingu- during social bonding (such as breast-feed-
late (ACC), and insula cortex (see Figure 4). ing and casual touching). They found that
DLPFC is an area involved in planning. The oxytocin rose in the trustee if the first player
insula cortex is known to be activated during “trusts” her by investing a lot. (They also
the experience of negative emotions like found that ovulating women were particular-
pain and disgust. ACC is an “executive func- ly untrustworthy—they did not repay as
tion” area which often receives inputs from much of the investment.)
many areas and resolves conflicts among Roxanna Gonzalez and Loewenstein
them.23 Therefore, it appears that, after an (2004) examined the impact of circadian
unfair offer, the brain (ACC) struggles to rhythms in a repeated trust (centipede)
resolve the conflict between wanting to game. They sorted people into “morning”
accept the money because of its planned and “night” people (which can be done
reward value (DLPFC) and disliking the rather reliably) and had them play a cen-
“disgust” of being treated unfairly (insula). tipede game when they were on- or off-peak
In fact, whether players reject unfair (e.g., the morning people were off-peak
offers or not can be predicted rather reliably when playing in the evening). Based on prior
(a correlation of 0.45) by the level of their research showing that sleep-cycle affects
insula activity. It is irresistible to speculate emotional regulation—i.e., people’s ability to
that the insula is a neural locus of the distaste suppress or avoid acting on unwanted feel-
for inequality or unfair treatment posited by ings—they predicted and found much lower
models of social utility, which have been suc- levels of cooperative behavior when people
cessfully used to explain many varying pat- played at off-peak times.
terns in experiments — robust ultimatum Tania Singer et al. (2004) report an impor-
rejections, public goods contributions, and tant link between reward and behavior in
trust and gift-exchange (e.g., Bazerman, games. They played repeated prisoners’
Loewenstein, and Leigh Thompson 1989; dilemma games in which one player, in the
fMRI scanner, faced a series of opponents.
The ACC also contains a large concentration of
“spindle cells”—large neurons shaped like spindles, It also suggests an intriguing follow-up experiment
which are almost unique to human brains (Allman et al. that no previous theory would have predicted—patients
2002). Loosely speaking, these cells are probably impor- with damage to their insula regions should feel no disgust
tant for many of the activities which distinguish humans and accept low offers, unless those patients have devel-
from our primate cousins, particularly language, and oped a “workaround” or alternative method to “feel”
complex decision making. unfairness in nonvisceral terms.
mr05_Article 1 3/28/05 3:25 PM Page 49

Camerer, Loewenstein, and Prelec: Neuroeconomics 49

Figure 5. Coronal slices showing regions which are differentially active after an unfair offer
($1–2 out of $10), relative to activity after a fair off ($4–5). Regions are anterior cingulate (ACC),
right and left insula, and dorsolateral prefrontal cortex (DLPFC). See Sanfey et al. (2003).

The scanned subject was told that some and reasoning ability, might legitimately vary
opponents cooperated intentionally (i.e., with these biological factors. The theory is
they could choose freely) while others coop- easily patched by inserting variables, like “an
erated, but unintentionally. Afterwards, sub- envy/disgust coefficient,” which depends on
jects were shown different faces of those some biological state. But game theory also
they had played against. The faces of the assumes that players will recognize the state-
intentional cooperators activated insula, dependence in others and adjust their guess-
amygdala, and ventral striatal areas (among es about how other people will play. We have
others). Since the striatum is an all-purpose no idea if they can, and it is likely that people
reward area, activation in that region means are generally limited at simulating emotional
that simply seeing the face of a person who states of others (see Leaf van Boven,
intentionally cooperated with you is reward- Loewenstein, and David Dunning 2003).
ing. In game theory terms, a person’s “repu- Cognitive inaccessibility also implies that
tation” in a repeated game is a perception by people may not fully understand the influ-
other players of their “type” or likely behav- ence of exogenous changes in visceral
ior based on past play. Singer et al.’s results states on their own behavior. For example,
mean that a good reputation may be neural- if being trusted produces oxytocin, then
ly encoded in a way similar to beautiful or when oxytocin surges for exogenous rea-
other rewarding stimuli. sons—from a relaxing massage, or when
The facts that insula activity, oxytocin lev- synthetic oxytocin is administered—the
els, and sleep cycles affect behavior in games, brain might misread this surge in oxytocin
and that cooperators’ faces “feel beautiful,” as a sign of being trusted and react accord-
does not “disprove” game theory, per se, ingly (e.g., by acting in a more reciprocal
because preferences for different outcomes, trustworthy way).
mr05_Article 1 3/28/05 3:25 PM Page 50

50 Journal of Economic Literature, Vol. XLIII (March 2005)

5.3.3 Backward Induction Levine 1998; George Mailath 1998). Since

many types of learning rules have been pro-
A central principle in game theory is posed, fitting rules to data from experiment
“backward induction” in extensive-form has proved useful in showing when intuitive-
(“tree”) games that are played over time. ly appealing rules might fit badly, and sug-
Backward induction means figuring out what gesting improvements to those rules (e.g.,
to do today by and reasoning how others will Camerer 2003). Camerer and Ho (1999)
behave at all possible future points and showed that both simple reinforcement of
working backward. Behavioral evidence, and chosen strategies, and learning by updating
direct evidence from measuring where play- beliefs about other players, are really two
ers look on a computer screen, shows that polar opposite types of generalized rein-
people have trouble doing more than a cou- forcement learning in which strategies have
ple of steps of backward induction (e.g., numerical propensities or attractions which
Johnson et al. 2002). However, Johnson et al. are adjusted over time by experience.
also found that, when players were briefly In neural terms, the Camerer–Ho theory
instructed about how to do backward induc- can be interpreted as a splice of two process-
tion, they could learn to do it rapidly and es—a rapid, emotional process in which a
with little effort (total response times were chosen strategy is quickly reinforced by the
similar to in pre-instruction trials). This is a gain or loss that resulted and a slower delib-
reminder, in the game theory context, of the erative process that requires players to cre-
important distinction between controlled ate counterfactuals about how much they
and automatic behavior stressed in section 3 would have earned from other strategies that
above. Initially, subjects look automatically were not chosen. Conventional reinforce-
at early periods and hardly notice future ment learning neglects the second process.
periods they realize (correctly, in a statistical “Fictitious play” belief learning assumes the
sense) are unlikely to occur. However, with second process completely overrides the
instruction and practice, backward induction first. A parameter in the theory which repre-
quickly becames automated, yielding fast sents the relative strength of the second
responses and few errors. process, compared to the first, is usually
Economists are naturally inclined to estimated to be between zero and one. This
model cognition in terms of costs and bene- implies that reinforcement is stronger, but
fits. If forced into this framework, backward both processes are at work.
induction would probably be characterized Michael Platt and Paul Glimcher (1999)
as cognitively costly. But the fact that it is found corroborating evidence for reinforce-
easily learned and automated suggests the ment learning of the first, rapid-process sort
costs of backward induction has a special in single-neuron recording in monkey pari-
structure— at first, it is unnatural (not spon- etal cortex. They measured neuron firing
taneously intuited by subjects), like a piece rates in advance of choices in a game
of software not yet installed; but once between a monkey and a computerized
installed, it is cheap to run. opponent. They found that firing rates are
5.3.4 Learning closely related to the average reinforcement
received for that choice in the last ten trials.
The idea that a game-theoretic equilibri- Dominic Barraclough, Michelle Conroy, and
um resulted from learning, imitation, or evo- Daeyeol Lee (2004) found similar evidence
lution, rather than simple introspection, has of “learning neurons” in dorsolateral pre-
led to a large literature on what results in the frontal cortex of rhesus monkeys; their
long-run from different types of learning parameter estimates of learning models also
models (e.g., Drew Fudenberg and David supports the two-process Camerer–Ho
mr05_Article 1 3/28/05 3:25 PM Page 51

Camerer, Loewenstein, and Prelec: Neuroeconomics 51

Before EV Cue After Cue

Action Potentials
per second

Relative Expected Value

Figure 6. ( from Glimcher et al., in press). The left graph (“before EV cue”) shows the LIP neuron firing
rate (y axis) plotted against the relative expected value of one of two possible reward-producing locations.
The right graph (“after cue”) shows the firing rate after the monkey learns which movement is rewarding.

theory.25 Gathering neural evidence from theory. Strategic thinking is likely to require
humans (to see whether a second, delibera- specialized circuitry and iterated strategic
tive process is being used as the parametric reasoning is likely to be bounded (as indicat-
estimates suggest) is feasible and could ed by ample behavioral data). Early studies
prove insightful. suggest hormones and biological factors
Figure 6 illustrates the behavior of (such as circadian rhythyms) play a role in
Glimcher’s parietal neurons in a task with social preferences like trust. Thinking about
two targets of different expected value. The the brain—and some experimental facts—
left graph shows the actual firing rates (on also suggest that logical principles like back-
the y axis) plotted against the expected ward induction may be neurally unnatural
reward value of the target before the “best” and that learning processes are likely to be a
target is revealed. The correspondence is splice of cognitive and affective processes. At
remarkably close—these neurons are encod- the same time, single-neuron monkey meas-
ing expected value. The righthand graph urement suggests that expected-value com-
shows that, after the winning target is putation and reinforcement learning may
revealed (so that the expected value is no control behavior in highly adapted tasks. The
longer relevant), the firing rates increase to emerging picture is one in which the simplest
close to their maximum capacity. elements of game theory, like keeping track
In sum, neuroscience provides some new of what has worked in a mixed-equilibrium
ways to think about central elements of game game with limited scope for outguessing one’s
opponent, may be alive and well in the brain,
while higher-order cognition and affective
They estimate two reinforcements: When the mon- influence on social preferences depart from
keys choose and win (reinforcement by 1), and when they
choose and lose (2). In their two-strategy games, the the standard ideas we teach our students.
model is equivalent to one in which monkeys are not rein-
forced for losing, but the unchosen strategy is reinforced 5.4 Labor-Market Discrimination
by 2. The fact that 2 is usually less than 1 in magnitude
(see also Lee et al., in press) is equivalent to   1 in the Our last specific application is labor-mar-
Camerer–Ho theory. ket discrimination. Economic models assume
mr05_Article 1 3/28/05 3:25 PM Page 52

52 Journal of Economic Literature, Vol. XLIII (March 2005)

that labor-market discrimination against As the name suggest, the implicit associa-
minorities is either a taste (a distaste for tion test taps into “implicit,” as opposed to
working with minorities, or a distaste passed “explicit,” attitudes. One can think of
on from customers), or a belief that minority implicit attitudes, roughly, as those associat-
workers are less productive (for example, a ed with automatic processing, whereas
belief that minority status is a proxy for unob- explicit attitudes are associated with con-
servable differences in skill, also known as scious, controlled, processing. New meth-
“statistical discrimination”). ods developed by psychologists, such as the
Neuroscience suggests a different answer. IAT, have begun to reveal that implicit and
Automaticity contributes to discrimination explicit attitudes can sometimes diverge
because neural networks rapidly spread acti- from one-another, with implicit attitudes, in
vation through associated concepts and some cases, exerting a more reliable influ-
stereotypes. Affect contributes to discrimina- ence on behavior. Thus, in one recent study,
tion because automatic affective reactions Allen McConnell and Jill Leibold (2001)
have such a powerful effect on cognitive judg- administered the IAT to subjects, had them
ments. Discrimination in this view involves complete measures of explicit attitudes
rapid, automatic, associations between social toward blacks and whites, and also had them
categories, stereotypes, and affect. interact with two experimenters, one black
Such an account receives support from and the other white. Coders blind to either
remarkable experiments that demonstrated the implicit or explicit measures then coded
subtle “implicit associations” between demo- the subjects’ interactions with the experi-
graphic categories and good or bad adjectives menters, including such objective measures
(try it out on yourself at http:// as how closely the subject moved their chair Subjects taking to the experimenter, and the experimenter
this computer-administered implicit associa- also rated their perception of the prejudice
tion test (IAT) are shown a mixed up series of of the subject. Although the experimenters
stereotypically black or white names (Tyrone ratings of the subjects’ degree of prejudice
or Chip) and positive or negative adjectives correlated with both explicit (r = 0.33, p 
(mother or devil). They are asked to tap one 0.05) and implicit (r = 0.39, p  0.05) meas-
key when they see one type of name or adjec- ures, the other behavioral measures of bias
tive, and a different key if they see the other that correlated with either the IAT or self-
type of name or adjective, at which point the reports of prejudice (these were: coders’
computer moves on to the next name or adjec- overall ratings, speaking time, smiling,
tive. The dependent variable is how long it speech errors, speech hesitation, and extent
takes the subject to work their way through of extemporaneous social comments) all
the complete list of names and adjectives. correlated with the IAT but not with the
White subjects work their way through the list explicit measures of prejudice.
much more quickly when one key is linked to Implicit attitudes have also been linked to
the pair (black or negative) and the other to neural processing. In one study (Elizabeth
(white or positive) than they are when one key Phelps et al. 2000), researchers adminis-
is linked to [black or positive] and the other to tered the IAT to Caucasian subjects and also
[white or negative]. What’s going on? The asked them explicit questions about their
brain encodes associations in neural networks, attitudes toward African–Americans. Then
which spread activation to related concepts. they scanned the subjects’ brains with fMRI
For white students, black names are instantly while exposing them to photographs of unfa-
associated with negative concepts, whether miliar black and white males, and focusing
they realize it or not, because the association specifically on a subcortical structure in the
is automatic (rapid and unconscious). brain called the amygdala, which numerous
mr05_Article 1 3/28/05 3:25 PM Page 53

Camerer, Loewenstein, and Prelec: Neuroeconomics 53

studies have linked to processing of fear. other than prices and choice sets so it is a
They found that the relative strength of stretch to think of them of as conventional
amygdala activation to black as compared tastes or beliefs.
with white faces was correlated with the IAT
measure of implicit attitudes, but not with
6. Conclusions
direct, conscious, expression of race atti-
tudes. (A study by Hart et al. (2000) suggests Economics parted company from psychol-
the effect is roughly symmetric for black ogy in the early twentieth century.
subjects seeing white faces.) Furthermore, Economists became skeptical that basic psy-
the same pattern was not observed when the chological forces could be measured without
black and white faces were well known, pos- inferring them from behavior (the same posi-
itively regarded, celebrities (such as Michael tion “behaviorist” psychologists of the 1920s
Jordan and Denzel Washington). People reached), which led to adoption of the useful
hold implicit attitudes involving not only tautology between unobserved utilities and
race, but also a wide range of other individ- observed (revealed) preferences. But
ual characteristics, such as height, weight, remarkable advances in neuroscience now
attractiveness, religion, and national origin, make direct measurement of thoughts and
and, like race, many of these attitudes feelings possible for the first time, opening
undoubtedly have real economic conse- the “black box” which is the building block of
quences. Thus, although most people reject any economic interaction and system—the
height and attractiveness as indicators of human mind.
marginal productivity, taller and more attrac- Most economists are curious about neuro-
tive people are more likely to earn higher science, but instinctively skeptical that it can
wages (e.g., Nicola Persico, Andrew tell us how to do better economics. The tra-
Postlewaite, and Dan Silverman 2002) and dition of ignoring psychological regularity in
other rewards (e.g., the U.S. Presidency). making assumptions in economic theory is
Does the implicit association view imply so deeply ingrained—and has proved rela-
that labor-market discrimination is due to a tively successful—that knowing more about
taste, a statistical shortcut, or something the brain seems unnecessary. Economic the-
else? One piece of evidence suggests the sta- ory will chug along successfully for the next
tistical interpretation is on the right track: few years paying no attention at all to neuro-
the amygdala that is active when people see science (just as it paid little attention to psy-
other-race faces seems to be sensitive to chology until recently). But it is hard to
familiarity of faces, not race per se (Dubois believe that some neuroscientific regularities
et al. 1999). This is consistent with an inter- will not help explain some extant anomalies,
pretation of statistical discrimination in particularly those that have been debated
which employers do not fear minority work- for decades.
ers, they are simply less sure of their abilities. Indeed, in many areas of economics there
But other evidence suggests discrimination are basic constructs or variables at the heart
is “something else.” Amygdala activity can be of current debates which can be usefully
dampened if subjects are shown pictures of thought of as neural processes, and studied
black and white faces and asked to judge using fMRI and other tools. For example,
how much the pictured people like vegeta- finance is a field awash in literally millions of
bles (Wheeler and Fiske 2005). This suggests observations of daily price movements.
that automatic reactions to race can be Despite having widespread access to terrific
erased (or substituted for) depending on the data, after decades of careful research there
question being asked when the faces are per- is no agreed-upon theory of why stock prices
ceived. These reactions respond to variables fluctuate, why people trade, and why there
mr05_Article 1 3/28/05 3:25 PM Page 54

54 Journal of Economic Literature, Vol. XLIII (March 2005)

are so many actively managed mutual funds product’s quality or, for “network” or “sta-
despite poor fund performance. Perhaps tus” goods, a product’s likely popularity.
knowing more about basic neural mecha- Many of these models seem like strained
nisms that underlie conformity, attention attempts to explain effects of advertising
paid to large price changes, wishful thinking, without incorporating the obvious intuition
sense-making of random series, and percep- that advertising taps neural circuitry of
tions of expertise can help explain these reward and desire.
puzzles (e.g., Lo and Repin 2002). Finally, economic models do not provide a
Furthermore, some scholars have argued satisfying theory of how individuals differ. As
that large fluctuations in stock prices are laymen, we characterize other people as
due to reasonable time-variation of risk pre- impulsive or deliberate, stable or neurotic,
mia. But there is no theoretical basis in decisive or indecisive, mature or immature,
finance for why attitudes toward risk would foolish or wise, depressed or optimistic, scat-
vary over time. Maybe neuroscience can terbrained or compulsively organized. The
supply one. consumers who spend countless dollars on
In labor markets, a major puzzle is why self-help, “organize your life” manuals, and
wages are so downward sticky. Firms say they who sustain the huge, and infinitely varied
are afraid to cut wages because they want to psychological counseling industry, are typi-
maintain worker morale (e.g., Truman cally unhappy where they stand on some of
Bewley 2002); and experiments show that these dimensions and are looking for ways to
when worker productivity is valuable, paying change. Comparative economic develop-
a high wage induces effort, even when work- ment, entrepreneurial initiative and innova-
ers are free to shirk (Fehr and Gachter tion, business cycle sensitivity, and other
2000). Presumably morale is some combina- important macroeconomic behaviors are
tion of workers’ emotional feelings toward probably sensitive to the distribution of these
their employer and may be very sensitive to and other psychological “assets.” Yet there is
recent experience, to what other workers no complete way to discuss them with the
think, to whether wage cuts are “procedural- language of beliefs and desires, which is the
ly just,” and so forth. There is no reason only language operating in quadrant I.
these processes could not be described as
6.1 Can Neuroscience Save Rational
neural processes and studied that way.
Choice Economics?
There are many anomalies in intertempo-
ral choice. In the United States, credit card Many neuroscientists are now using the
debt is substantial ($5,000.00 per household) most basic elements of rational choice theo-
and a million personal bankruptcies have ry to explain what they see. Ironically, they
been declared in each of several years are taking up rational choice theory at the
(Laibson, Andrea Repetto, and Jeremy same time as more and more economists are
Tobacman 1998). Healthier food is cheaper moving away from rational choice toward a
and more widely available than ever before, behavioral view anchored in limits on ration-
but spending on dieting and obesity are both ality, willpower, and greed (which we expect
on the rise. Surely understanding how brain to be informed by neural detail). For exam-
mechanisms process reward, and curb or ple, neuroscientific studies of simple reward
produce compulsion, and their evolutionary circuitry in rats and primates vindicate some
origins (e.g., Trent Smith 2003) might help of the simplest ideas of economics—namely,
explain these facts and shape sensible policy a “common brain currency” which permits
and regulation. substitution (Shizgal 1999), existence of neu-
Prevailing models of advertising assume rons which encode expected reward (see
that ads convey information or signal a Glimcher 2002 and Figure 3), and revealed
mr05_Article 1 3/28/05 3:25 PM Page 55

Camerer, Loewenstein, and Prelec: Neuroeconomics 55

preference (macaques viewing socially val- Repetto and Tobacman 1998; O’Donoghue
ued images, see Deaner, Khera, and Platt and Rabin 1999). Laibson’s (2001) model of
2005). Other groups are using Bayesian homeostatic response to environmental cues
models. One study concluded that “the cen- in addiction is an incremental model well
tral nervous system therefore employs grounded in recent neuroscience.
[Bayesian] probabilistic models during sen- However, we believe that in the long run a
sorimotor learning” (Konrad Körding and more “radical” departure from current theo-
Daniel Wolpert 2004). ry will become necessary, in the sense that
Our view is that establishing a neural basis the basic building blocks will not just consist
for some rational choice principles will not of preferences, constrained optimization and
necessarily vindicate the approach as widely (market or game-theoretic) equilibrium.
applied to humans. The reason is that the After all, the point of constrained optimiza-
studies which have most clearly established tion is to model behavior precisely and pre-
expected-value and Bayesian encoding use dict how behavior changes in response to
simple tasks which monkeys and humans are changes in budget constraints and prices.
well-evolved to perform (e.g., reaching to There is no reason other models starting
earn a juice reward). It is quite possible that from a very different basis could not be con-
simple rational mechanisms are neurally structed, while also predicting behavioral
instantiated to do these tasks well. But the responses to constraints and prices, and pre-
most important kinds of economic behavior dicting responses to other variables as well.
involve manipulating abstract symbols, think- Furthermore, thinking about the brain does
ing about groups of people and complex not so much “falsify” rational choice theories
institutions, trading off very different types of as suggest entirely new distinctions and
rewarding objects across time, and weighing questions.
them by probabilities which are not always For example, is learning that a particular
learnable by experience. Ironically, rational person cooperated with you in a game
choice models might therefore be most use- encoded as a cognitive reputational statistic
ful in thinking about the simplest kinds of about that person, like a numerical test score,
decisions humans and other species make— or a “warm glow” which produces a surge of
involving perceptual tradeoffs, motor move- dopamine when you see the person’s face
ments, foraging for food, and so forth—and (Singer et al. 2004; cf. James Rilling et al.
prove least useful in thinking about abstract, 2002). Standard game theory has no answer
complex, long-term tradeoffs which are the for this question. But the distinction matters.
traditional province of economic theory. If reputations are encoded dopaminergically,
then they may spill over across groups of
6.2 Incremental Versus Radical
people who look the same, for example, or
who are perceived as being part of a group
How should neuroeconomics contribute who behave similarly (Bill McEvily et al.
to advances in economic analysis? In the 2003; Paul Healy 2004). Or the mechanism
short-run, an “incremental” approach in can work in reverse—since attractive faces
which psychological evidence suggests func- are known to produce dopaminergic surges
tional forms will help enhance the realism of (Aharon et al. 2001), the cortex may mistake
existing models. For example, the two- these positive signals of cooperativeness
parameter “ß–δ“ hyperbolic discounting (which also produces pleasurable sensations)
approach (where ß expresses preference for and automatically judge attractive faces as
immediacy, and is equal to one in the stan- likely to be cooperative. This simple neural
dard model) is an example that has proven question about reward spillover might form
productive theoretically (e.g., Laibson, the foundation of group affiliation and social
mr05_Article 1 3/28/05 3:25 PM Page 56

56 Journal of Economic Literature, Vol. XLIII (March 2005)

capital and have something to do with why trade could also flow in the opposite direc-
some countries are rich and others poor. tion. Neuroscience is shot through with
Our main theme in this paper is that radi- familiar economic language — delegation,
cal models should respect the fact that brain division of labor, constraint, coordination,
mechanisms combine controlled and auto- executive function — but these concepts are
matic processes, operating using cognition not formalized in neuroscience as they are in
and affect. The Platonic metaphor of reason economics. There is no overall theory of how
as a charioteer, driving twin horses of passion the brain allocates resources that are essen-
and appetite, is on the right track—except tially fixed (e.g., blood flow and attention).
reason has its hands full with headstrong An “economic model of the brain” could help
passions and appetites.26 Of course, the chal- here. Simple economic concepts, like mech-
lenge in radical-style theorizing is to develop anisms for rationing under scarcity, and gen-
models of how multiple mechanisms interact eral versus partial equilibrium responses to
which are precise. Can this be done? The shocks, could help neuroscientists under-
answer is Yes. Bernheim and Rangel (2004), stand how the entire brain interacts. At a
Loewenstein and O’Donoghue (2004), and technical level, neuroscientists are using
Jess Benhabib and Alberto Bisin (2002) have tools imported from econometrics, such as
all proposed recent models with interacting Granger causality (e.g., Wolfram Hesse et al.
mechanisms much like those in our Table 1. 2003), to draw better inferences from neural
Furthermore, while interactions of multi- time series. Finally, as the center of gravity in
ple brain mechanisms might appear to be neuroscience research shifts from elemen-
too radical a change from equilibrium with tary cognitive processes to the study of so-
utility maximization, we think many familiar called higher functions—reasoning, social
tools can be used to do radical neuroeco- inference, and decision making—neurosci-
nomics. Interactions of cognition and affect entists will increasingly reference, and draw
might resemble systems like supply and inspiration from, the conceptual apparatus of
demand, or feedback loops which exhibit economics, a unique distillate of our century-
multiple equilibria. The interaction of con- long reflection on individual and strategic
trolled and automatic processes might be behavior.
like an inventory policy or agency model in
which a controller only steps in when an
Adolphs, Ralph, Daniel Tranel, Hanna Damasio, and
extreme state of the system (or unusual Antonio R. Damasio. 1995. “Fear and the Human
event) requires controlled processes to Amygdala.” Journal of Neuroscience, 15(9): 5879–91.
override automatic ones. The influence of Aharon, Itzhak, Nancy Etcoff, Dan Ariely, Chris F.
Chabris, Ethan O’Connor, and Hans C. Breiter. 2001.
affect on choices is a very general type of “Beautiful Faces Have Variable Reward Value: fMRI
state-dependence (where the “state” is and Behavioral Evidence.” Neuron, 32(3): 537–51.
affective, and is influenced by external cues Allman, John, Atiya Hakeem, and Karli Watson. 2002.
“Two Phylogenetic Specializations in the Human
and also by internal deliberation and Brain.” Neuroscientist, 8(4): 335–46.
restraint). Instead of solving for equilibria in Ameriks, John, Andrew Caplin, and John Leahy. 2003.
these interacting-mechanism models, solve “Wealth Accumulation and the Propensity to Plan.”
The Quarterly Journal of Economics, 118(3):
for steady states or cyclic fluctuations. 1007–47.
Instead of summarizing responses to Anderson, Adam K., Kalina Christoff, David Panitz,
changes by comparative statics, study Eve De Rosa, and John D. Gabrieli. 2003. “Neural
Correlates of the Automatic Processing of Threat
impulse-response functions. Facial Signals.” Journal of Neuroscience, 23(13):
Although we focused solely on applications 5627–33.
of neuroscience to economics, intellectual Ariely, Dan and George Loewenstein. 2003. “The Heat
of the Moment: The Effect of Sexual Arousal on
Decision Making,” CMU Working Paper.
Thanks to Nava Ashraf for clarifying this point. Ariely, Dan, Uri Gneezy, George Loewenstein, and
mr05_Article 1 3/28/05 3:25 PM Page 57

Camerer, Loewenstein, and Prelec: Neuroeconomics 57

Nina Mazar. 2004. “Large Stakes and Big Mistakes,” Bechara, Antoine, Hanna Damasio, Daniel Tranel,
CMU Working Paper. and Antonio R. Damasio. 1997. “Deciding
Armony, Jorge L., David Servan-Schreiber, Jonathan Advantageously Before Knowing the Advantageous
D. Cohen, and Joseph E. LeDoux. 1995. “An Strategy.” Science, 275(5304): 1293–95.
Anatomically Constrained Neural Network Model of Becker, Gary and Luis Rayo. 2004. “Evolutionary
Fear Conditioning.” Behavioral Neuroscience, Efficiency and Mean Reversion in Happiness.”
109(2): 246–57. University of Chicago Graduate School of Business.
Armony, Jorge L., David Servan-Schreiber, Jonathan D. Benartzi, Shlomo and Richard H. Thaler. 1995. “Myopic
Cohen, and Joseph E. LeDoux. 1997. “Computational Loss Aversion and the Equity Premium Puzzle.” The
Modeling of Emotion: Explorations through the Quarterly Journal of Economics, 110(1): 73–92.
Anatomy and Physiology of Fear Conditioning.” Benhabib, Jess and Alberto Bisin. 2002. “Self-Control
Trends in Cognitive Sciences, 1(1): 28–34. and Consumption-Savings Decisions: Cognitive
Barberis, Nicholas, Ming Huang, and Tano Santos. Perspectives.” New York University.
2001. “Prospect Theory and Asset Prices.” The Berkowitz, Leonard. 1999. “Anger,” in Handbook of
Quarterly Journal of Economics, 116(1): 1–53. Cognition and Emotion. Tim Dalgleish and Mick J.
Bargh, John A. 1984. “Automatic and Conscious Power, eds. New York: Wiley, 411–28.
Processing of Social Information,” in Handbook of Bernheim, B. Douglas and Antonio Rangel. 2004.
Social Cognition. R. S. Wyer Jr. and T. K. Srull, eds. “Addiction and Cue-Conditioned Cognitive
Hillsdale, NJ: Erlbaum, 1–43. Processes.” American Economic Review, 94(5):
Bargh, John A., Shelly Chaiken, Paula Raymond, and 1558–90.
Charles Hymes. 1996. “The Automatic Evaluation Berridge, Kent C. 1996. “Food Reward: Brain
Effect: Unconditional Automatic Attitude Activation Substrates of Wanting and Liking.” Neuroscience
with a Pronunciation Task.” Journal of Experimental and Biobehavioral Reviews, 20(1): 1–25.
Social Psychology, 32(1): 104–28. Bertrand, Marianne and Sendhil Mullaninathan. 2004.
Bargh, John A. and Tanya L. Chartrand. 1999. “The “Are Emily and Greg More Employable than Lakisha
Unbearable Automaticity of Being.” American and Jamal? A Field Experiment on Labor Market
Psychologist, 54(7): 462–79. Discrimination,” NBER Working Paper 9873.
Barlow, David H. 1988. Anxiety and its Disorders: The Bewley, Truman F. 2002. “Fairness, Reciprocity, and
Nature and Treatment of Anxiety and Panic. New Wage Rigidity,” Cowles Foundation Working Paper.
York: Guilford Press.
Baron-Cohen, Simon. 2000. “Theory of Mind and Bhatt, Meghana and Colin F. Camerer. In Press. “Self-
Autism: A Fifteen Year Review,” in Understanding Referential Strategic Thinking and Equilibrium as
Other Minds: Perspectives from Developmental States of Mind in Games: Evidence from fMRI.”
Cognitive Neuroscience. Simon Baron-Cohen, Helen Games and Economic Behavior.
Tager-Flusberg and Donald J. Cohen, eds. Oxford: Blair, R. James and Lisa Cipolotti. 2000. “Impaired
Oxford University Press, 1–20. Social Response Reversal: A Case of ‘Acquired
Barraclough, Dominic J., Michelle L. Conroy, and Sociopathy.’” Brain, 123(6): 1122–41.
Daeyeol Lee. 2004. “Prefrontal Cortex and Decision Bodner, Ronit and Drazen Prelec. 2003. “The
Making in a Mixed-Strategy Game.” Nature Diagnostic Value of Actions in a Self-Signaling
Neuroscience, 7(4): 404–10. Model,” in The Psychology of Economic Decisions
Baumeister, Roy F., Todd F. Heatherton, and Dianne Vol. 1. Isabelle Brocas and Juan D. Carillo, eds.
M. Tice. 1994. Losing Control: How and Why People Oxford: Oxford University Press.
Fail at Self-Regulation. San Diego: Academic Press. Breiter, Hans C., I. Aharon, Daniel Kahneman, A. Dale,
Baumeister, Roy F. and Kathleen D. Vohs. 2003. and Peter Shizgal. 2001. “Functional Imaging of
“Willpower, Choice, and Self-control,” in Time and Neural Responses to Expectancy and Experience of
Decision: Economic and Psychological Perspectives Monetary Gains and Losses.” Neuron, 30(2): 619–39.
on Intertemporal Choice. George F. Loewenstein, Buck, Ross. 1999. “The Biological Affects: A Typology.”
Daniel Read and Roy F. Baumeister, eds. New York: Psychological Review, 106(2): 301–36.
Russell Sage Foundation, 13–86. Cabanac, Michael. 1979. “Sensory Pleasure.”
Bazerman, Max, George Loewenstein, and Leigh Quarterly Review of Biology, 54(1): 1–29.
Thompson. 1989. “Social Utility and Decision Caggiula, Anthony R. and Bartley G. Hoebel. 1966.
Making in Interpersonal Contexts.” Journal of “‘Copulation–Reward Site’ in the Posterior
Personality and Social Psychology, 57(3): 426–41. Hypothalamus.” Science, 153(3741): 1284–85.
Bechara, Antoine, Antonio R. Damasio, Hanna Camerer, Colin F. 2003. Behavioral Game Theory:
Damasio, and Steven W. Anderson. 1994. Experiments in Strategic Interaction. Roundtable
“Insensitivity to Future Consequences Following Series in Behavioral Economics. Princeton: Princeton
Damage to Human Prefrontal Cortex.” Cognition, University Press; New York: Russell Sage Foundation.
50(1-3): 7–15. Camerer, Colin F. and George Loewenstein. 2004.
Bechara, Antoine, Hanna Damasio, Antonio R. “Behavioral Economics: Past, Present, and Future,”
Damasio, and G. P. Lee. 1999. “Different in Advances in Behavioral Economics. Colin F.
Contributions of the Human Amygdala and Camerer, George Loewenstein and Matthew Rabin,
Ventromedial Prefrontal Cortex to Decision- eds. Princeton: Princeton University Press.
Making.” Journal of Neuroscience, 19(13): 5473–81. Camerer, Colin F. and Teck Hua Ho. 1999.
mr05_Article 1 3/28/05 3:25 PM Page 58

58 Journal of Economic Literature, Vol. XLIII (March 2005)

“Experience-Weighted Attraction Learning in American Medical Association, 290(2): 252–55.

Normal Form Games.” Econometrica, 67(4): 827–74. de Becker, Gavin. 1997. The Gift of Fear: Survival
Camerer, Colin F., Teck Hua Ho, and Kuan Chong. Signals That Protect Us from Violence. Boston:
2004. “A Cognitive Hierarchy Model of One-Shot Little, Brown and Co.
Games.” Quarterly Journal of Economics, 119(3): Deaner, Robert O., Amit V. Khera, and Michael Platt.
861–98. 2005. “Monkeys Pay Per View: Adaptive Valuation of
Camerer, Colin F., George Loewenstein, and Martin Social Images by Rhesus Macaques,” Current
Weber. 1989. “The Curse of Knowledge in Economic Biology.
Settings: An Experimental Analysis.” Journal of 0960-9822/PIIS960982205001041.pdf
Political Economy, 97(5): 1232–54. de Gelder, Beatrice, Jean Vroomen, Gilles Pourtois,
Camerer, Colin F. and Dan Lovallo. 1999. and Lawrence Weiskrantz. 1999. “Non-Conscious
“Overconfidence and Excess Entry: An Recognition of Affect in the Absence of Striate
Experimental Approach.” American Economic Cortex.” Neuroreport, 10(18): 3759–63.
Review, 89(1): 306–18. De Houwer, Jan, Dirk Hermans, and Paul Eelen. 1998.
Campfield, L. Arthur and Francoise J. Smith. 1990. “Affective and Identity Priming with Episodically
“Systemic Factors in the Control of Food Intake,” in Associated Stimuli.” Cognition and Emotion, 12(2):
Handbook of Behavioural Neurobiology, Vol. 10: 145–69.
Neurobiology of Food and Fluid Intake. E. M. Delgado, Mauricio R., Leigh E. Nystrom, C. Fissell, D.
Stricker, ed. New York: Plenum, 183–206. C. Noll, and Julie A. Fiez. 2000. “Tracking the
Caplin, Andrew and John Leahy. 2001. “Psychological Hemodynamic Responses to Reward and
Expected Utility Theory and Anticipatory Feelings.” Punishment in the Striatum.” Journal of
The Quarterly Journal of Economics, 116(1): 55–79. Neurophysiology, 84(6): 3072–77.
Carter, Rita. 1999. Mapping the Mind. Berkeley: DeLillo, Don. White Noise. East Rutherford, NJ:
University of California Press. Penguin Books.
Chorvat, Terrence, Kevin McCabe and Vernon L. della Vigna, Stefano and Ulrike Malmendier. 2003.
Smith. 2004. “Law and Neuroeconomics.” George “Overestimating Self-Control: Evidence from the
Mason Working Paper. Health Club Industry,” Berkeley Working Paper.
Cleckley, Hervey. 1941. The Mask of Sanity. St. Louis: trol_nov03.pdf.
C. V. Mosby. Denes-Raj, Veronika, Seymour Epstein, and Jonathan
Clore, Gerald L. 1992. “Cognitive Phenomenology: Cole. 1995. “The Generality of the Ratio–Bias
Feelings and the Construction of Judgment,” in The Phenomenon.” Personality and Social Psychology
Construction of Social Judgments. Leonard L. Bulletin, 21(10): 1083–92.
Martin and Abraham Tesser, eds. Hillsdale: Dubois, Samuel, Bruno Rossion, Christine Schiltz,
Erlbaum, 133–63. Jean-Michel Bodart, Christian Michel, Raymond
Comings, David E. 1998. “The Molecular Genetics of Bruyer, and Marc Crommelinck. 1999. “Effect of
Pathological Gambling.” CNS Spectrums, 3(6): Familiarity on the Processing of Human Faces.”
20–37. Neuroimage, 9: 278–89.
Cosmides, Leda. 1989. “The Logic of Social Exchange: Eisenberger, Naomi I., Matthew D. Lieberman, and
Has Natural Selection Shaped How Humans Kipling D. Williams. 2003. “Does Rejection Hurt?
Reason? Studies with the Wason Selection Task.” An fMRI Study of Social Exclusion.” Science,
Cognition, 31(3): 187–276. 302(5643): 290–92.
Cosmides, Leda and John Tooby. 2004. “Evolutionary Elbert, Thomas, Christo Pantev, Christian Wienbruch,
Psychology and the Emotions,” in Handbook of Brigitte Rockstroh, and Edward Taub. 1995.
Emotions. Michael Lewis and Jeannette M. “Increased Cortical Representation of the Fingers of
Haviland-Jones, eds. New York: Guilford. the Left Hand in String Players.” Science, 270(5234):
Costa-Gomes, Miguel, Vincent P. Crawford, and Bruno 305–07.
Broseta. 2001. “Cognition and Behavior in Normal- Elster, Jon. 1977. Ulysses and the Sirens. Cambridge:
Form Games: An Experimental Study.” Cambridge University Press.
Econometrica, 69(5): 1193–235. Erk, Suzanne, Manfred Spitzer, Arthur P. Wunderlich,
Costa, Dora L. and Matthew E. Kahn. “Shame and Lars Galley, and Henrik Walter. 2002. “Cultural
Ostracism: Union Army Deserters Leave Home.” Objects Modulate Reward Circuitry.” Neuroreport,
MIT Working Paper. 13(18): 2499–503.
shame11b.html. Farrer, Chlöé and Chris D. Frith. 2001. “Experiencing
Cottle, Thomas. 1974. The Present of Things Future: Oneself versus Another Person as Being the Cause of
Explorations of Time in Human Experience. New an Action: The Neural Correlates of the Experience
York: Free Press. of Agency.” Neuroimage, 15(3): 596–603.
Damasio, Antonio R. 1994. Descartes’ Error: Emotion, Fazio, Russell, David M. Sanbonmatsu, Martha C.
Reason, and the Human Brain. New York: G. P. Powell, and Frank Kardes. 1986. “On the Automatic
Putnam. Activation of Attitudes.” Journal of Personality and
Dana, Jason and George Lowenstein. 2003. “A Social Psychology, 50(2): 229–38.
Psychological Perspective on the Influence of Gifts Fehr, Ernst and Simon Gachter. 2000. “Cooperation
to Physicians from Industrry.” Journal of the and Punishment in Public Goods Experiments.”
mr05_Article 1 3/28/05 3:25 PM Page 59

Camerer, Loewenstein, and Prelec: Neuroeconomics 59

American Economic Review, 90(4): 980–94. Journal of Economics, 112(2): 631–45.

Fletcher, Paul C., Francesca Happe, Uta Frith, S. C. Gneezy, Uri and Aldo Rustichini. 2004. “Incentives,
Baker, Ray J. Dolan, Richard S. Frackowiak, and Punishment and Behavior,” in Advances in
Chris D. Frith. 1995. “Other Minds in the Brain: A Behavioral Economics. Colin F. Camerer, George F.
Functional Imaging Study of “Theory of Mind” in Loewenstein and Matthew Rabin, eds. New York:
Story Comprehension.” Cognition, 57(2): 109–28. Russell Sage Foundation, 572–89.
Frederick, Shane and George Loewenstein. 2000. “The Gobet, Ferhand and Herbert A. Simon. 1996. “Recall
Psychology of Sequence Preferences.” CMU of Random and Distorted Chess Positions:
Working Paper. Implications for the Theory of Expertise.” Memory
Frederick, Shane, George Loewenstein, and Ted and Cognition, 24(4): 493–503.
O’Donoghue. 2002. “Time Discounting and Time Goldman, Alvin. 2003. “Emotion, Mindreading,
Preference: A Critical Review.” Journal of Economic Simulation, and Modularity.” Other Minds: An
Literature, 40(2): 351–401. Interdisciplinary Conference. Institute of Cognitive
Freeman, Walter J. and James W. Watts. 1942. and Decision Sciences: University of Oregon,
Psychosurgery in the Treatment of Mental Disorders September 27–28, 2003.
and Intractable Pain. Springfield: Thomas. Gollwitzer, Peter M., Kentaro Fujita, and Gabriele
Fried, Itzhak. 1998. “Technical Comment: The Oettingen. 2004. “Planning and Implementation of
Hippocampus and Human Navigation.” Science, Goals,” in Handbook of Self-Regulation: Research,
282(5397): 2151. Theory, and Applications. Roy F. Baumeister and
Frijda, Nico. 1986. The Emotions. Cambridge: Kathleen D. Vohs, eds. New York: Guilford Press,
Cambridge University Press. 211–28.
Frith, Uta. 2001a. “Mind Blindness and the Brain in Gonzalez, Roxanna and George Loewenstein. 2004.
Autism.” Neuron, 32(6): 969–79. “Effects of Circadian Rhythm on Cooperation in
Frith, Uta. 2001b. “What Framework Should We Use an Experimental Game.” Working Paper. http://
for Understanding Developmental Disorders?”
Developmental Neuropsychology, 20(2): 555–63. Green, Leonard J. and Howard Rachlin. 1991.
Fudenberg, Drew and David K. Levine. 1998. The “Economic Substitutability of Electrical Brain
Theory of Learning in Games. Cambridge and Stimulation, Food, and Water.” Journal of the
London: MIT Press. Experimental Analysis of Behavior, 55(2): 133–43.
Gazzaniga, Michael S. and Joseph E. LeDoux. 1978. Greenberg, D., G. P. Smith, and J. Gibbs. 1990.
The Integrated Mind. New York: Plenum. “Intraduodenal Infusions of Fats Elicit Satiety in
Gibbs, James, S. P. Madison, and Edmund T. Rolls. Sham-Feeding Rats.” American Journal of
1981. “Satiety Role of the Small Intestine Examined Physiology: Regulatory, Integrative and
in Sham-Feeding Rhesus Monkeys.” Journal of Comparative Physiology, 259(1 Pt. 2): R110–18.
Comparative Physiology and Psychology, 95(6): Greenwald, Anthony G. 1992. “New Look 3:
1003–15. Unconscious Cognition Reclaimed.” American
Gilbert, Daniel T. 2002. “Inferential Correction,” in Psychologist, 47(6): 766–79.
Heuristics and Biases: The Psychology of Intuitive Greenwald, Anthony G., M. R. Klinger, and T. J. Liu.
Judgment. Thomas Gilovich, D. Griffin and Daniel 1989. “Unconscious Processing of Dichoptically
Kahneman, eds. Cambridge: Cambridge University Masked Words.” Memory and Cognition, 17(1):
Press, 167–84. 35–47.
Gilbert, Daniel T. and Michael J. Gill. 2000. “The Grether, David, Charles Plott, Daniel Rowe, Martin
Momentary Realist.” Psychological Science, 11(5): Sereno, and John Allman. 2004. “Mental Processes
394–98. and Strategic Equilibration: An fMRI Study of
Giner-Sorolla, Roger. 2001. “Guilty Pleasures and Selling Strategy in Second Price Auctions.” Caltech
Grim Necessities: Affective Attitudes in Dilemmas Working Paper No. 1189. http://www.
of Self-Control.” Journal of Personality and Social
Psychology, 80(2): 206–21. Hadjikhani, Nouchine and Beatrice de Gelder. 2003.
Giordano, Louis A., Warren K. Bickel, George “Seeing Fearful Body Expressions Activates the
Loewenstein, Eric A. Jacobs, Lisa Marsch, and Gary J. Fusiform Cortex and Amygdala.” Current Biology,
Badger. 2002. “Opioid Deprivation Affects How 13(24): 2201–05.
Opioid-Dependent Outpatients Discount the Value of Haggard, Patrick, Sam Clark, and Jeri Kalogeras. 2002.
Delayed Heroin and Money.” Psychopharmacology, “Voluntary Action and Conscious Awareness.”
163(2): 174–82. Nature Neuroscience, 5(4): 382–85.
Glimcher, Paul W. 2003. Decisions, Uncertainty, and Hahnloser, Richard H., Rahul Sarpeshkar, Misha A.
the Brain: The Science of Neuroeconomics. Mahowald, Rodney J. Douglas, and H. Sebastian
Cambridge and London: MIT Press. Seung. 2000. “Digital Selection and Analogue
Glimcher, Paul W., Michael Dorris, and Hannah Bayer. Amplification Coexist in a Cortex-Inspired Silicon
In Press. “Physiological Utility Theory and the Circuit.” Nature, 405(6789): 947–51.
Neuroeconomics of Choice.” Games and Economic Haier, Richard J., Benjamin V. Siegel, Jr., Andrew
Behavior. MacLachlan, Eric Soderling, Stephen Lottenberg,
Gneezy, Uri and Jan Potters. 1997. “An Experiment on and Monte S. Buchsbaum. 1992. “Regional Glucose
Risk Taking and Evaluation Periods.” The Quarterly Metabolic Changes After Learning a Complex
mr05_Article 1 3/28/05 3:25 PM Page 60

60 Journal of Economic Literature, Vol. XLIII (March 2005)

Visuospatial/Motor Task: a PET Study.” Brain Dissociated by Demand Elasticity.” Physiological

Research, 570(1–2): 134–43. Behavior, 26(3): 509–15.
Happe, Francesca, Stefan Ehlers, Paul Fletcher, Uta Jevons, William S. 1871. The Theory of Political
Frith, Maria Johansson, Christopher Gillberg, Ray Economy. London: Macmillan and Co.
Dolan, Richard Frackowiak, and Chris Frith. 1996. Johnson, Eric J., Colin F. Camerer, Sen Sankar, and
“‘Theory of Mind’ in the Brain: Evidence from a Talia Tymon. 2002. “Detecting Failures of Backward
PET Scan Study of Asperger Syndrome.” Induction: Monitoring Information Search in
Neuroreport, 8(1): 197–201. Sequential Bargaining.” Journal of Economic
Hare, Robert D. 1965. “Psychopathy, Fear Arousal And Theory, 104(1): 16–47.
Anticipated Pain.” Psychological Reports, 16: Kahneman, Daniel. 2003. “Maps of Bounded
499–502. Rationality: Psychology for Behavioral Economics.”
Hare, Robert D. 1966. “Temporal Gradient of Fear American Economic Review, 93(5): 1449–75.
Arousal in Psychopaths.” Journal of Abnormal Kahneman, Daniel and Shane Frederick. 2002.
Psychology, 70(6): 442–45. “Representativeness Revisited: Attribution
Hart, Allen J., Paul Whalen, Lisa Shin, Sean Substitution in Intuitive Judgment,” in Heuristics of
McInerney, Hakan Fischer, and Scott Rauch. 2000. Intuitive Judgment: Extensions and Applications.
“Differential Response in the Human Amygdala to Thomas Gilovich, Dale Griffin and Daniel
Racial Outgroup versus Ingroup Face Stimuli.” Kahneman, eds. New York: Cambridge University
Neuroreport, 11(11): 2351–55. Press, 49–81.
Hastie, Reid. 1984. “Causes and Effects of Causal Kirkpatrick, Lee A. and Seymour Epstein. 1992.
Attribution.” Journal of Personality and Social “Cognitive–Experiential Self-Theory and Subjective
Psychology, 46(1): 44–56. Probability: Further Evidence for Two Conceptual
Healy, Paul J. 2004. “Group Reputations and Systems.” Journal of Personality and Social
Stereotypes as a Contract Enforcement Device.” Psychology, 63(4): 534–44.
Caltech Working Paper. Knutson, Brian and Richard Peterson. In Press.
/~pj/papers/Healy-StereotypeContracts.pdf. “Neurally Reconstructing Expected Utility.” Games
Hebb, Donald. 1949. The Organization of Behavior: A and Economic Behavior.
Neuropsychological Theory. New York: Wiley. Körding, Konrad P. and Daniel M. Wolpert. 2004.
Hershey, John C., Howard C. Kunreuther, and Paul J. “Bayesian Integration in Sensorimotor Learning.”
H. Schoemaker. 1982. “Sources of Bias in Nature, 427(6971): 244–47.
Assessment Procedures for Utility Functions.” Kosslyn, Stephen M. 1994. Image and Brain.
Management Science, 28(8): 936–54. Cambridge: MIT Press.
Hesse, Wolfram, Eva Möller, Matthias Arnold, and Kruglanski, Arie W., Erik P. Thompson, E. Tory
Bärbel Schack. 2003. “The Use of Time-Variant Higgins, M. Nadir Atash, Antonio Pierro, James Y.
EEG Granger Causality for Inspecting Directed Shah, and Scott Spiegel. 2000. “To ‘Do the Right
Interdependencies of Neural Assemblies.” Journal of Thing’ or to ‘Just Do It’: Locomotion and Assessment
Neuroscience Methods, 124(1): 27–44. as Distinct Self-Regulatory Imperatives.” Journal of
Hill, Elisabeth and David Sally. 2003. “Dilemmas and Personality and Social Psychology, 79(5): 793–815.
Bargains: Autism, Theory-of-Mind, Cooperation and Laibson, David. 2001. “A Cue-Theory of Consumption.”
Fairness.” University College London Working The Quarterly Journal of Economics, 116(1): 81–119.
Paper. Laibson, David I., Andrea Repetto, and Jeremy
Hirst, William. 1993. “On the Nature of Systems,” in Tobacman. 1998. “Self-Control and Saving for
Conceptions of the Human Mind: Essays in Honor of Retirement.” Brookings Papers on Economic
George A. Miller. Gilbert Harman and George A. Activity. 1: 91–196.
Miller eds. Hillsdale: Erlbaum, 1–11. Lambrecht, Anja and Bernd Skiera. 2004. “Paying Too
Hoch, Stephen J. and George F. Loewenstein. 1991. Much and Being Happy About It: Causes and
“Time-Inconsistent Preferences and Consumer Self- Consequences of Tariff Choice–Bias.” Johann
Control.” Journal of Consumer Research, 17(4): Wolfgang Goethe-University Frankfurt am Main
492–507. Working Paper.
Hoffman, Elizabeth, Kevin McCabe, Keith Shachat, Langer, Thomas and Craig R. Fox. “Partition
and Vernon L. Smith. 1994. “Preferences, Property Dependence and Naïve Diversification in
Rights, and Anonymity in Bargaining Games.” Investment Decision Making.” University of
Games and Economic Behavior, 7(3): 346–80. Mannheim Working Paper.
Houston, David and Russell H. Fazio. 1989. “Biased Le Bihan, Denis, Jean-François Mangin, Cyril Poupon,
Processing as a Function of Attitude Accessibility: Chris A. Clark, Sabina Pappata, Nicolas Molko, and
Making Objective Judgments Subjectively.” Social Hughes Chabriat. 2001. “Diffusion Tensor Imaging:
Cognition, 7(1): 51-66. Concepts and Applications.” Journal of Magnetic
Hsu, Ming, Meghana Bhatt, Ralph Adolphs, Daniel Resonance Imaging, 13(4): 534–46.
Tranel, and Colin Camerer. 2005. “Ambiguity- Leboeuf, Robyn Aimee. 2002. “Alternating Selves and
Aversion in the Brain: fMRI and Lesion-Patient Conflicting Choices: Identity Salience and
Evidence.” Caltech Working Paper. Preference Inconsistency.” Dissertation Abstracts
Hursh, Steven R. and B. H. Natelson. 1981. “Electrical International, 63(2–B): 1088.
Brain Stimulation and Food Reinforcement LeDoux, Joseph E. 1996. The Emotional Brain: The
mr05_Article 1 3/28/05 3:25 PM Page 61

Camerer, Loewenstein, and Prelec: Neuroeconomics 61

Mysterious Underpinnings of Emotional Life. New Stephen Manuck and Matthew Muldoon. 2001.
York: Simon and Schuster. “Dimensions of Time Discounting.” Conference on
Lee, Daeyeol, Michelle L. Conroy, Benjamin P. Survey Research on Household Expectations and
McGreevy and Dominic J. Barraclough. 2004. Preferences. Ann Arbor, MI: November 2–3, 2001.
“Reinforcement Learning and Decision Making in Logothetis, Nikos K., Jon Pauls, Mark Augath, Torsten
Monkeys During a Competitive Game.” Cognitive Trinath, and Axel Oeltermann. 2001.
Brain Research, 22(1): 45–58. “Neurophysiological Investigation of the Basis of the
Lerner, Jennifer S. and Dacher Keltner. 2001. “Fear, fMRI Signal.” Nature, 412(6843): 150–57.
Anger, and Risk.” Journal of Personality and Social Lucas, Robert E., Jr. 1978. “Asset Prices in an
Psychology, 81(1): 146–59. Exchange Economy.” Econometrica, 46(6): 1429–45.
Libet, Benjamin. 1985. “Unconscious Cerebral Lykken, David T. 1957. “A Study of Anxiety in the
Initiative and the Role of Conscious Will in Voluntary Sociopathic Personality.” Journal of Abnormal
Action.” Behavior and Brain Sciences, 8(4): 529–66. Psychology, 55(1): 6–10.
Lieberman, Matthew D. In Press. “Reflective and MacLean, Paul D. 1990. The Triune Brain in Evolution:
Reflexive Judgment Processes: A Social Cognitive Role in Paleocerebral Function. New York: Plenum.
Neuroscience Approach,” in Responding to the Maguire, Eleanor A., David G. Gadian, Ingrid S.
Social World: Implicit and Explicit Processes in Johnsrude, Catriona D. Good, John Ashburner,
Social Judgments and Decisions. Joseph P. Forgas, Richard S. Frackowiak, and Christopher D. Frith.
Kipling D. Williams and William von Hippel, eds. 2000. “Navigation-Related Structural Change in the
Philadelphia: Psychology Press. Hippocampi of Taxi Drivers.” Proceedings of the
Lieberman, Matthew D., Ruth Gaunt, Daniel T. National Academy of Sciences of the United States of
Gilbert, and Yaacov Trope. 2002. “Reflection and America, 97(8): 4398–403.
Reflexion: A Social Cognitive Neuroscience Mailath, George. 1998. “Do People Play Nash
Approach to Attributional Interference,” in Equilibrium? Lessons from Evolutionary Game
Advances in Experimental Social Psychology. Mark Theory.” Journal of Economic Literature, 36(3):
P. Zanna, ed. New York: Academic Press, 199–249. 1347–74.
Lo, Andrew W. and Dmitry V. Repin. 2002. “The Mailath, George J., Michihiro Kandori, and Rafael
Psychophysiology of Real-Time Financial Risk Rob. 1993. “Learning, Mutation, and Long Run
Processing.” Journal of Cognitive Neuroscience, Equilibria in Games.” Econometrica, 61(1): 29–56.
14(3): 323–39. Mandler, George. 1982. “The Structure of Value:
Loewenstein, George F. 1992. “The Fall and Rise of Accounting for Taste,” in Affect and Cognition: The
Psychological Explanations in the Economics of Seventeenth Annual Carnegie Symposium on
Intertemporal Choice,” in Choice Over Time. Cognition. M. S. Clark and S. T. Fiske, eds. Hillsdale:
George Loewenstein and Jon Elster, eds. New York: Erlbaum, 3–36.
Russell Sage Foundation, 3–34. Manuck, Stephen B., Janine Flory, Matthew Muldoon,
Loewenstein, George F. 1994. “The Psychology of and Robert E. Ferrell. 2003. “A Neurobiology of
Curiosity: A Review and Reinterpretation.” Intertemporal Choice,” in Time and Decision:
Psychological Bulletin, 116(1): 75–98. Economic and Psychological Perspectives on
Loewenstein, George F. 1996. “Out of Control: Visceral Intertemporal Choice. George F. Loewenstein,
Influences on Behavior.” Organizational Behavior Daniel Read and Roy F. Baumeister, eds. New York:
and Human Decision Processes, 65(3): 272–92. Russell Sage Foundation, 139–72.
Loewenstein, George F. 1999. “Because It Is There: Marshall, J. F., J. S. Richardson, and P. Teitelbaum.
The Challenge of Mountaineering…for Utility 1974. “Nigrostriatal Bundle Damage and the Lateral
Theory.” Kyklos, 52(3): 315–43. Hypothalamic Syndrome.” Journal of Comparative
Loewenstein, George F. and Jennifer S. Lerner. 2003. and Physiological Psychology, 87(5): 808–30.
“The Role of Affect in Decision Making,” in McCabe, Kevin, Daniel Houser, Lee Ryan, Vernon L.
Handbook of Affective Sciences. Richard J. Davidson, Smith, and Theodore Trouard. 2001. “A Functional
Klaus R. Scherer and H. Hill Goldsmith, eds. Imaging Study of Cooperation in Two-Person
Oxford: Oxford University Press, 619–42. Reciprocal Exchange.” Proceedings of the National
Loewenstein, George F. and Ted O’Donoghue. 2004. Academy of Sciences of the United States of America,
“Animal Spirits: Affective and Deliberative Influences 98(20): 11832–35.
on Economic Behavior,” CMU Working Paper. McClure, Samuel M., David I. Laibson, George F.
Loewenstein, George F. and Drazen Prelec. 1993. Loewenstein and Jonathan D. Cohen. 2004.
“Preferences for Sequences of Outcomes.” “Separate Neural Systems Value Immediate and
Psychological Review, 100(1): 91–108. Delayed Monetary Rewards.” Science, 306(5695):
Loewenstein, George F., Leigh Thompson, and Max H. 503–07.
Bazerman. 1989. “Social Utility and Decision McConnell, Allen and Jill Leibold. 2001. “Relations
Making in Interpersonal Contexts.” Journal of Among the Implicit Association Test, Discriminatory
Personality and Social Psychology, 57(3): 426–41. Behavior, and Explicit Measures of Racial Attitudes.”
Loewenstein, George F., Elke U. Weber, Christopher Journal of Experimental Social Psychology, 37(5):
K. Hsee, and Ned Welch. 2001. “Risk as Feelings.” 435–42.
Psychological Bulletin, 127(2): 267–86. McElroy, Susan L., Andrew Satlin, Harrison Pope,
Loewenstein, George F., Robert Weber, Janine Flory, Paul E. Keck, and James I. Hudson. 1991.
mr05_Article 1 3/28/05 3:25 PM Page 62

62 Journal of Economic Literature, Vol. XLIII (March 2005)

“Treatment of Compulsive Shopping with Influences Veridical Judgments of Motion

Antidepressants: A Report of Three Cases.” Annals Direction.” Journal of Neuroscience, 22(21): 9530–40.
of Clinical Psychiatry, 3(3): 199–204. Nisbett, Richard E. and Timothy D. Wilson. 1977.
McEvily, Bill, Roberto Weber, Cristina Bicchieri and “Telling More Than We Can Know: Verbal Reports
Violet Ho. 2003. “Can Groups Be Trusted? An on Mental Processes.” Psychological Bulletin, 84(3):
Experimental Study of Collective Trust,” CMU 231–59.
Working Paper. http://www.andrew.cmu/user/rwe- O’Donoghue, Ted and Matthew Rabin. 1997.
ber/CollectiveTrust10.pdf. “Addiction and Self-Control,” in Addiction: Entries
McNeil, Barbara J., Stephen G. Pauker, Harold C. Sox, and Exits. Jon Elster, ed. New York: Russell Sage
Jr., and Amos Tversky. 1982. “On the Elicitation of Foundation, 1999, 169–206.
Preferences for Alternative Therapies.” New O’Donoghue, Ted and Matthew Rabin. 1999. “Doing It
England Journal of Medicine, 306(21): 1259–62. Now or Later.” American Economic Review, 89(1):
Medin, Douglas L. and Max H. Bazerman. 1999. 103–24.
“Broadening Behavioral Decision Research: Ochsner, Kevin N., Silvia A. Bunge, James J. Gross, and
Multiple Levels of Cognitive Processing.” James D.E. Gabrieli. 2002. “Rethinking Feelings: An
Psychonomic Bulletin and Review, 6(4): 533–47. fMRI Study of the Cognitive Regulation of Emotion.”
Mendelson, Joseph. 1967. “Lateral Hypothalamic Journal of Cognitive Neuroscience, 14(8): 1215–29.
Stimulation in Satiated Rats: The Rewarding Effects Ochsner, Kevin N. and Matthew D. Lieberman. 2001.
of Self-Induced Drinking.” Science, 157(3792): “The Emergence of Social Cognitive Neuroscience.”
1077–79. American Psychologist, 56(9): 717–34.
Metcalfe, Janet and Walter Mischel. 1999. “A Hot/Cool- Olds, James and Peter Milner. 1954. “Positive
System Analysis of Delay of Gratification: Dynamics Reinforcement Produced by Electrical Stimulation
of Willpower.” Psychological Review, 106(1): 3–19. of Septal Area and Other Regions of Rat Brain.”
Miller, George A., Eugene Galanter, and Karl H. Journal of Comparative and Physiological
Pribram. 1960. Plans and the Structure of Behavior. Psychology, 47(6): 419–27.
New York: Holt. Oswald, Andrew J. 1997. “Happiness and Economic
Mischel, Walter, Ebbe B. Ebbesen, and Antonette R. Performance.” Economic Journal, 107(445): 1815–31.
Zeiss. 1972. “Cognitive and Attentional Mechanisms Panksepp, Jaak. 1998. Affective Neuroscience. Oxford:
in Delay of Gratification.” Journal of Personality and Oxford University Press.
Social Psychology, 21(2): 204–18. Parsons, Lawrence M. and Daniel Osherson. 2001.
Mischel, Walter and Robert Metzner. 1962. “Preference “New Evidence for Distinct Right and Left Brain
for Delayed Reward as a Function of Age, Systems for Deductive versus Probabilistic
Intelligence, and Length of Delay Interval.” Journal Reasoning.” Cerebral Cortex, 11(10): 954–65.
of Abnormal Social Psychology, 64(6): 425–31. Persico, Nicola, Andrew Postlewaite, and Dan
Mobbs, Dean, Michael D. Greicius, Eiman Abdel- Silverman. “The Effect of Adolescent Experience on
Azim, Vinod Menon, and Allan L. Reiss. 2003. Labor Market Outcomes: The Case of Height,”
“Humor Modulates the Mesolimbic Reward University of Pennsylvania Working Paper.
Centers.” Neuron, 40(5): 1041–48. Persinger, Michael A. and Faye Healey. 2002.
Montague, P. Read and Gregory S. Berns. 2002. “Experimental Facilitation of the Sensed Presence:
“Neural Economics and the Biological Substrates of Possible Intercalation Between the Hemispheres
Valuation.” Neuron, 36(2): 265–84. Induced by Complex Magnetic Fields.” Journal of
Montague, P. Read, Gregory S. Berns, Jonathan D. Nervous and Mental Disease, 190(8): 533–41.
Cohen, Samuel M. McClure, Giuseppe Pagnoni, Peters, Ellen and Paul Slovic. 2000. “The Springs of
Mukesh Dhamala, Michael C. Wiest, Igor Karpov, Action: Affective and Analytical Information
Richard D. King, Nathan Apple, and Ronald E. Processing in Choice.” Personality and Social
Fisher. 2002. “Hyperscanning: Simultaneous fMRI Psychology Bulletin, 26(12): 1465–75.
During Linked Social Interactions.” Neuroimage, Phelps, Elizabeth, Kevin J. O’Connor, William A.
16(4): 1159–64. Cunningham, E. Sumie Funayama, J. Christopher
Moreyra, Paula, Angela A. Ibáñez, Jerónimo Saiz-Ruiz, Gatenby, John C. Gore, and Mahzarin R. Banaji.
Kore Nissenson, and Carlos Blanco. 2000. “Review 2000. “Performance on Indirect Measures of Race
of the Phenomenology, Etiology and Treatment of Evaluation Predicts Amygdala Activity.” Journal of
Pathological Gambling.” German Journal of Cognitive Neuroscience, 12(5): 729–38.
Psychiatry, 3(2): 37–52. Pillutla, Madan M. and Xiao-Ping Chen. 1999. “Social
Morris, John S., C. Buchel, and Raymond J. Dolan. Norms and Cooperation in Social Dilemmas: The
2001. “Parallel Neural Responses in Amygdala Effects of Context and Feedback.” Organizational
Subregions and Sensory Cortex During Implicit Fear Behavior and Human Decision Processes, 78(2):
Conditioning.” Neuroimage, 13(6 Pt. 1): 1044–52. 81–103.
Mullainathan, Sendhil. 2002. “Thinking Through Platt, Michael L. and Paul W. Glimcher. 1999. “Neural
Categories,” MIT Working Paper. Correlates of Decision Variables in Parietal Cortex.”
National Academy of Sciences. 1999. Pathological Nature, 400(6741): 233–38.
Gambling. Washington, D.C.: National Academy Press. Prelec, Drazen and George F. Loewenstein. 1998.
Nichols, M. James and William T. Newsome. 2002. “The Red and the Black: Mental Accounting of
“Middle Temporal Visual Area Microstimulation Savings and Debt.” Marketing Science, 17(1): 4–28.
mr05_Article 1 3/28/05 3:25 PM Page 63

Camerer, Loewenstein, and Prelec: Neuroeconomics 63

Prelec, Drazen and Duncan Simester. 2001. “Always “Controlled and Automatic Human Information
Leave Home Without It.” Marketing Letters, 12(1): Processing: I. Detection, Search and Attention.”
5–12. Psychological Review, 84(1): 1–66.
Quattrone, George A. and Amos Tversky. 1984. Schultz, Wolfram. 2002. “Getting Formal with
“Causal Versus Diagnostic Contingencies: On Self- Dopamine and Reward.” Neuron, 36(2): 241–63.
Deception and on the Voter’s Illusion.” Journal of Schultz, Wolfram and Anthony Dickinson. 2000.
Personality and Social Psychology, 46(2): 237–48. “Neuronal Coding of Prediction Errors.” Annual
Rabin, Matthew. 1993. “Incorporating Fairness into Review of Neuroscience, 23: 473–500.
Game Theory and Economics.” American Economic Schwarz, Norbert. 1990. “Feelings as Information:
Review, 83(5): 1281–302. Informational and Motivational Functions of
Read, Daniel and George Loewenstein. 1995. “The Affective States,” in Handbook of Motivation and
Diversification Bias: Explaining the Difference Cognition: Foundations of Social Behavior. E. T.
Between Prospective and Real-Time Taste for Higgins and R. M. Sorrentino, eds. New York:
Variety.” Journal of Experimental Psychology: Guilford Press, 527–61.
Applied, 1(1): 34–49. Schwarz, Norbert and Gerald L. Clore. 1983. “Mood,
Read, Daniel, George Loewenstein, and Shobana Misattribution and Judgments of Well-Being:
Kalyanaraman. 1999. “Mixing Virtue and Vice: The Informative and Directive Functions of Affective
Combined Effects of Hyperbolic Discounting and States.” Journal of Personality and Social
Diversification.” Journal of Behavioral Decision Psychology, 45(3): 513–23.
Making, 12(4): 257–73. Shallice, Timothy and Paul W. Burgess. 1996. “The
Rilling, James K., D. A. Gutman, T. R. Zeh, Giuseppe Domain of Supervisory Processes and Temporal
Pagnoni, Greg S. Berns, and Clint D. Kilts. 2002. “A Organization of Behaviour.” Philosophical
Neural Basis for Social Cooperation.” Neuron, 35(2): Transactions of the Royal Society B: Biological
395–405. Sciences, 351(1346): 1405–12.
Rolls, Edmund T. 1999. The Brain and Emotion. New Shapiro, David. 1965. Neurotic Styles. New York: Basic
York: Oxford University Press. Books.
Romer, Paul M. 2000. “Thinking and Feeling.” Shefrin, Hersh and Richard Thaler. 1988. “The
American Economic Review, 90(2): 439–43. Behavioral Life Cycle Hypothesis.” Economic
Ross, Lee, Mark R. Lepper, and Michael Hubbard. Inquiry, 26(4): 609–43.
1975. “Perseverance in Self-Perception and Social Sherry, David F. and Daniel L. Schacter. 1987. “The
Perception: Biased Attributional Processes in the Evolution of Multiple Memory Systems.”
Debriefing Paradigm.” Journal of Personality and Psychological Review, 84(2): 127–90.
Social Psychology, 32(5): 880–92. Shiffrin, Richard M. and Walter Schneider. 1977.
Rowe, Andrea D., Peter R. Bullock, Charles E. Polkey, “Controlled and Automatic Human Information
and Robin G. Morris. 2001. “‘Theory of Mind’ Processing: II. Perceptual Learning, Automatic
Impairments and their Relationship to Executive Attending and a General Theory.” Psychological
Functioning Following Frontal Lobe Excisions.” Review, 84(2): 127–90.
Brain, 124(3): 600–16. Shiv, Baba and Alexander Fedorikhin. 1999. “Heart
Rumelhart, David E. and James L. McClelland. 1986. and Mind in Conflict: The Interplay of Affect and
Parallel Distributed Processing: Explorations in the Cognition in Consumer Decision Making.” Journal
Microstructure of Cognition, Vol. 1: Foundations. of Consumer Research, 26(3): 278–92.
Cambridge: MIT Press. Shiv, Baba, George Loewenstein, Antoine Bechara,
Rustichini, Aldo, John Dickhaut, Paolo Ghirardato, Kip Antonio Damasio and Hanna Damasio. 2005.
Smith and Jose V. Pardo. 2002. “A Brain Imaging “Investment Behavior and the Dark Side of
Study of Procedural Choice,” University of Emotion.” Psychological Science. Forthcoming.
Minnesota Working Paper. Shizgal, Peter. 1999. “On the Neural Computation of
Samuelson, Paul. 1937. “A Note on Measurement of Utility: Implications from Studies of Brain
Utility.” Review of Economic Studies, 4: 155–61. Reward,” in Foundations of Hedonic Psychology:
Sanfey, Alan G., James K. Rilling, Jessica A. Aaronson, Scientific Perspectives on Enjoyment and Suffering.
Leigh E. Nystrom, and Jonathan D. Cohen. 2003. Daniel Kahneman, Edward Diener and Norbert
“The Neural Basis of Economic Decision-Making in Schwarz eds. New York: Russell Sage Foundation,
the Ultimatum Game.” Science, 300(5626): 1755–58. 502–26.
Saxe, Rebecca and Nancy Kanwisher. 2003. “People Simon, Herbert A. 1967. “Motivational and Emotional
Thinking about Thinking People: The Role of the Controls of Cognition.” Psychological Review, 74(1):
Temporo–Parietal Junction in ‘Theory of Mind.’” 29–39.
Neuroimage, 19(4): 1835–42. Simonson, Itamar. 1990. “The Effect of Purchase
Schelling, Thomas C. 1978. “Egonomics, or the Art of Quantity and Timing on Variety-Seeking Behavior.”
Self-Management.” American Economic Review, Journal of Marketing Research, 27(2): 150–62.
68(2): 290–94. Singer, Tania, Stefan J. Kiebel, Joel S. Winston,
Schelling, Thomas C. 1984. “The Intimate Contest for Raymond J. Dolan, and Chris D. Frith. 2004. “Brain
Self-Command,” in Choice and Consequence. Responses to the Acquired Moral Status of Faces.”
Cambridge: Harvard University Press, 57–81. Neuron, 41(4): 653–62.
Schneider, Walter and Richard M. Shiffrin. 1977. Sloman, Steven A. 1996. “The Empirical Case for Two
mr05_Article 1 3/28/05 3:25 PM Page 64

64 Journal of Economic Literature, Vol. XLIII (March 2005)

Systems of Reasoning.” Psychological Bulletin, Bases for Innate Recognition of Song.” Proceedings
119(1): 3–22. of the National Academy of Sciences of the United
Smith, Trenton. “The McDonald’s Equilibrium: States of America, 94(23): 12694–98.
Advertising, Empty Calories, and the Endogenous Wheeler, Mary E. and Susan T. Fiske. 2005.
Determination of Dietary Preferences,” University “Controlling Racial Prejudice: Social–Cognitive
of Bonn Working Paper. Goals Affect Amygdala and Stereotype Activation.” Psychological Science, 16(1): 56–63.
Strack, Fritz, Leonard L. Martin, and Sabine Stepper. Wilson, Timothy D., Samuel Lindsey, and Tonya Y.
1988. “Inhibiting and Facilitating Conditions of the Schooler. 2000. “A Model of Dual Attitudes.”
Human Smile: A Nonobtrusive Test of the Facial Psychological Review, 107(1): 101–26.
Feedback Hypothesis.” Journal of Personality and Wilson, Timothy D., Douglas J. Lisle, Jonathan W.
Social Psychology, 54(5): 768–77. Schooler, Sara D. Hodges, Kristen J. Klaaren, and
Titmuss, Richard M. 1970. The Gift Relationship. Suzanne J. LaFleur. 1993. “Introspecting about
London: Allen and Unwin. Reasons Can Reduce Post-Choice Satisfaction.”
Train, Kenneth E. 1991. Optimal Regulation: The Personality and Social Psychology Bulletin, 19(3):
Economic Theory of Natural Monopoly. Cambridge: 331–39.
MIT Press. Wilson, Timothy D. and Jonathan W. Schooler. 1991.
Train, Kenneth E., Daniel L. McFadden, and Moshe “Thinking Too Much: Introspection Can Reduce the
Ben-Akiva. 1987. “The Demand for Local Telephone Quality of Preferences and Decisions.” Journal of
Service: A Fully Discrete Model of Residential Personality and Social Psychology, 60(2): 181–92.
Calling Patterns and Service Choices.” Rand Journal Windschitl, Paul D. and Gary L. Wells. 1998. “The
of Economics, 18(1): 109–23. Alternative–Outcomes Effect.” Journal of
Tversky, Amos and Daniel Kahneman. 1983. Personality and Social Psychology, 75(6): 1411–23.
“Extensional Versus Intuitive Reasoning: The Winkielman, Piotr and Kent C. Berridge. 2004.
Conjunction Fallacy in Probability Judgment.” “Unconscious Emotion.” Current Directions in
Psychological Review, 90(4): 293–315. Psychological Science, 13(3): 120–23.
Van Boven, Leaf, George Loewenstein, and David Wise, Roy A. 1996. “Addictive Drugs and Brain
Dunning. 2003. “Mispredicting the Endowment Stimulation Reward.” Annual Review of
Effect: Underestimation of Owners’ Selling Prices by Neuroscience, 19: 319–40.
Buyer’s Agents.” Journal of Economic Behavior and Wolford, George, Michael B. Miller, and Michael
Organization, 51(3): 351–65. Gazzaniga. 2000. “The Left Hemisphere’s Role in
Viner, Jacob. 1925. “The Utility Concept in Value Hypothesis Formation.” Journal of Neuroscience,
Theory and its Critics.” Journal of Political Economy, 20(6): 1–4.
33(4): 369–87. Zajonc, Robert B. 1980. “Feeling and Thinking:
von Melchner, Laurie, Sarah L. Pallas, and Mriganka Preferences Need No Inferences.” American
Sur. 2000. “Visual Behaviour Mediated by Retinal Psychologist, 35(2): 151–75.
Projections Directed to the Auditory Pathway.” Zajonc, Robert B. 1984. “On the Primacy of Affect.”
Nature, 404(6780): 871–76. American Psychologist, 39(2): 117–23.
Vuilleumier, Patrik, Jorge L. Armony, Jon Driver, and Zajonc, Robert B. 1998. “Emotions,” in Handbook of
Ray J. Dolan. 2001. “Effects of Attention and Social Psychology. Daniel T. Gilbert, Susan T. Fiske
Emotion on Face Processing in the Human Brain: An and Gardner Lindzey, eds. New York: Oxford
Event-Related fMRI Study.” Neuron, 30(3): 829–41. University Press, 591–632.
Wegner, Daniel M. and Thalia Wheatley. 1999. Zajonc, Robert B. and Daniel N. McIntosh. 1992.
“Apparent Mental Causation. Sources of the “Emotions Research: Some Promising Questions
Experience of Will.” American Psychologist, 54(7): and Some Questionable Promises.” Psychological
480–92. Science, 3(1): 70–74.
Wells, Gary L. and Richard E. Petty. 1980. “The Zak, Paul. In Press. “Neuroeconomics.” Transactions of
Effects of Overt Head Movements on Persuasion: the Royal Philosophical Society B.
Compatibility and Incompatibility of Resources.” Zak, Paul, Robert Kurzban and William Matzner. 2003.
Basic and Applied Social Psychology. 1(3): 219–30. “Oxytocin is Associated with Interpersonal Trust in
Whalen, Paul J., S. L. Rauch, Nancy Etcoff, S. C. Humans.” Under Review.
McInerney, M. Lee, and M. A. Jenike. 1998. Zamir, Shmuel. 2000. “Rationality and Emotions in
“Masked Presentations of Emotional Facial Ultimatum Bargaining,” Discussion Paper #222.
Expressions Modulate Amygdala Activity without
Explicit Knowledge.” Journal of Neuroscience, 18(1): Zink, Caroline F., Giuseppe Pagnoni, M. E. Martin, M.
411–18. Dhamala, and Greg S. Berns. 2003. “Human Striatal
Whaling, Carol S., M. M. Solis, Allison J. Doupe, Jill A. Response to Salient Nonrewarding Stimuli.” Journal
Soha, and Peter Marler. 1997. “Acoustic and Neural of Neuroscience, 23(22): 8092–97.