ANRV314-NE30-02 ARI 7 May 2007 17:6

Orbitofrontal Cortexand Its Contributionto Decision-MakingJonathan D. WallisHelen Wills Neuroscience Institute and the Department of Psychology,University of California, Berkeley, California 94720-3190; email: wallis@berkeley.edu

Annu. Rev. Neurosci. 2007. 30:31–56

First published online as a Review in Advance onApril 6, 2007

The Annual Review of Neuroscience is online atneuro.annualreviews.org

This article’s doi:10.1146/annurev.neuro.30.051606.094334

Copyright c© 2007 by Annual Reviews.All rights reserved

0147-006X/07/0721-0031$20.00

Key Words

prefrontal cortex, reward, neurophysiology, neuroeconomics,choice behavior

AbstractDamage to orbitofrontal cortex (OFC) produces an unusual patternof deficits. Patients have intact cognitive abilities but are impaired inmaking everyday decisions. Here we review anatomical, neuropsy-chological, and neurophysiological evidence to determine the neu-ronal mechanisms that might underlie these impairments. We sug-gest that OFC plays a key role in processing reward: It integratesmultiple sources of information regarding the reward outcome toderive a value signal. In effect, OFC calculates how rewarding areward is. This value signal can then be held in working memorywhere it can be used by lateral prefrontal cortex to plan and organizebehavior toward obtaining the outcome, and by medial prefrontalcortex to evaluate the overall action in terms of its success and theeffort that was required. Thus, acting together, these prefrontal ar-eas can ensure that our behavior is most efficiently directed towardssatisfying our needs.

31

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

OFC: orbitofrontalcortex

Contents

INTRODUCTION. . . . . . . . . . . . . . . . . 32ANATOMICAL

ORGANIZATION. . . . . . . . . . . . . . . 33Structural Anatomy. . . . . . . . . . . . . . . 33Overview of Connections . . . . . . . . . 34Other PFC Areas . . . . . . . . . . . . . . . . . 34

NEUROPSYCHOLOGY OF OFC. . 34Stimulus-Reward Learning and

Flexible Behavior . . . . . . . . . . . . . . 35Somatic Marker Hypothesis and

Decision-Making . . . . . . . . . . . . . . 35Summary . . . . . . . . . . . . . . . . . . . . . . . . 37

NEURONAL MECHANISMSWITHIN OFC . . . . . . . . . . . . . . . . . . 37Specialization of OFC for Reward

Processing . . . . . . . . . . . . . . . . . . . . 37Which Aspect of Reward is OFC

Encoding? . . . . . . . . . . . . . . . . . . . . 38Complexity of Reward Processing:

Calculating a Reward’s Value . . . 41Neuroeconomics . . . . . . . . . . . . . . . . . 43Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Probability of Success. . . . . . . . . . . . . 45Integrating Multiple Decision

Parameters . . . . . . . . . . . . . . . . . . . . 45A MODEL OF

DECISION-MAKING WITHINPFC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Nature of the OFC Reward

Representation: WorkingMemory for Value . . . . . . . . . . . . . 46

Interaction of OFC with OtherPFC Regions . . . . . . . . . . . . . . . . . . 47

Application of the Model to theCurrent Empirical Evidence . . . 48

CONCLUSION . . . . . . . . . . . . . . . . . . . . 49

INTRODUCTION

The orbitofrontal cortex (OFC) is the regionof the brain directly behind our forehead rest-ing on top of our eye orbits. Its position inthe skull, on top of the ridges created by thesphenoid bone, makes it particularly suscep-

tible to damage from head trauma. Yet dam-age to OFC often appears to have remark-ably little effect. Consider the case of Elliott,a happily married young man in his thirties(Damasio 1994, Eslinger & Damasio 1985).Elliott excelled in college and rose rapidlythrough the ranks of a building firm to becomecomptroller at the age of 32. People describedhim as a role model and a natural leader. Un-fortunately, at the age of 35 doctors diag-nosed Elliot with a brain tumor. The oper-ation to remove the tumor was successful, butthe surgery left Elliot with bilateral damage tohis OFC. However, neuropsychological testscould find no evidence of brain damage. Testsof his intelligence, memory, reading and writ-ing comprehension, verbal fluency, visuospa-tial abilities, and facial recognition revealedaverage to superior performance. He couldtalk intelligently and knowledgeably aboutcurrent issues. Even tests designed specificallyto tax frontal lobe processes, such as workingmemory, rule switching, and cognitive estima-tion, failed to reveal any deficits.

So was Elliot unaffected by the damage?Sadly, the answer is no. Within months of theoperation, he had quit his job, lost a largesum of money to a scam artist, divorced hiswife, lost contact with family and friends, andremarried a prostitute he had known for amonth. He had trouble holding down a job;employers complained about his tardiness anddisorganization. His second marriage endedin divorce six months later, and he moved inwith his parents. In short, prior to his tumorElliot had made a series of excellent life deci-sions, but within months of the operation hemade a series of catastrophic ones. Even sim-ple decisions were difficult because he wouldagonize over every possible consideration. Forexample, deciding where to dine would takehours as he considered the menu, the seatingarrangement, and the atmosphere. He wouldeven drive to each restaurant to see how busyit was.

Here then is the paradox of OFC: Howcan damage to this area leave so many ofour cognitive abilities intact, yet devastate our

32 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

ability to make the decisions that enable us tonavigate everyday life? This chapter focuseson understanding the neuronal mechanismsthat contribute to decision-making to help usmake sense of the deficits seen in OFC pa-tients. We begin with a brief overview of OFCanatomy and explore the deficits that occur af-ter OFC damage. We then examine the func-tional properties of OFC neurons, first con-centrating on their role in processing rewardsand then exploring more specifically the hy-pothesis that OFC is responsible for calculat-ing the value of a reward. We finish by propos-

Working memory:a memory systemenabling thetemporarymaintenance andmanipulation of alimited amount ofinformation overshort delays

PFC: prefrontalcortex

ing a model demonstrating how other brainareas might use the information from OFC tocontrol decision-making.

ANATOMICAL ORGANIZATION

Structural Anatomy

OFC is part of prefrontal cortex (PFC) andoccupies the ventral part of the frontal lobe(Figure 1). Discrepancies in the early maps ofhuman and monkey OFC organization havebeen resolved, and researchers now agree that

Figure 1Ventral view of the macaque (left) and human (right) brains illustrating the major cytoarchitectonicallydistinct regions of OFC (Petrides & Pandya 1994) and the main sulci. Olf = olfactory sulcus,M = medial orbital sulcus, T = transverse orbital sulcus, L = lateral orbital sulcus. In the macaquebrain preparation, the olfactory tubercle obscures the olfactory sulcus.

www.annualreviews.org • OFC and Decision-Making 33

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

DLPFC:dorsolateralprefrontal cortex

VLPFC:ventrolateralprefrontal cortex

MPFC: medialprefrontal cortex

its basic structure and organization are sim-ilar across primates. It consists of five cy-toarchitectonic subregions: frontal polar area10, area 11 anteriorly, area 13 posteriorly,area 14 medially, and area 47/12 laterally(Carmichael & Price 1994, Petrides & Pandya1994). Four sulci divide the surface into fivegyri (Chiavaras & Petrides 2000). Runningalong the anterior-posterior axis are three par-allel sulci. Most medially is the olfactory sul-cus, followed by the medial orbital sulcus andlateral orbital sulcus. The transverse orbitalsulcus connects the medial and lateral or-bital sulci approximately halfway along theirlength.

Overview of Connections

The connections of OFC exhibit three promi-nent features:

1. Within frontal cortex it is unique inreceiving information from all sensorymodalities (Carmichael & Price 1995b,Cavada et al. 2000, Romanski et al.1999). Area 47/12 receives highly pro-cessed visual information from areassuch as inferior temporal cortex, audi-tory information from secondary andtertiary auditory areas, somatosensoryinput from secondary somatosensorycortex and parietal cortex, and inputsfrom polysensory areas such as the su-perior temporal cortex. Primary olfac-tory and gustatory cortex both projectto posterior area 13.

2. OFC has only weak motor connections.Some connections exist between area47/12 and the supplementary eye fields,and between area 13 and the ventralpremotor cortex (Carmichael & Price1995b). In comparison, the medial wallof PFC densely connects with cingu-late motor areas, whereas dorsal andlateral PFC densely connect with pre-motor cortex (Chiba et al. 2001, Luet al. 1994). OFC may influence behav-ior through a subcortical route because

it strongly connects with the nucleus ac-cumbens (Haber et al. 1995).

3. OFC extensively connects with the lim-bic system, including the amygdala,cingulate gyrus, and the hippocampus(Carmichael & Price 1995a). It can alsoinfluence the autonomic nervous systemthrough its connections with the hy-pothalamus and other brainstem struc-tures, such as the periaqueductal graymatter (Ongur et al. 1998).

In summary, the connections of OFC arecompatible with a structure that integratessensory and reward information.

Other PFC Areas

Throughout this review we frequently com-pare and contrast the functions of OFCwith other major PFC subregions. Using thenomenclature of Petrides & Pandya (1994),dorsolateral PFC (DLPFC) consists of areas9, 46, and 9/46. Ventrolateral PFC (VLPFC)consists of areas 47/12 and 45. Medial PFC(MPFC) consists of area 32 and the anterior-most portions of area 24. MPFC is often con-sidered part of anterior cingulate cortex. Theanterior cingulate cortex is a large area that ex-tends posteriorly to about midway along thecingulate gyrus. The region we discuss is theanterior-most portion of the anterior cingu-late cortex. Although the nomenclature of thisregion is controversial, for brevity we simplyrefer to it as MPFC. Clear differences exist inthe pattern of connections of these differentPFC regions. The lateral regions of DLPFCand VLPFC connect predominately with sen-sory and motor areas (Carmichael & Price1995b, Lu et al. 1994), whereas MPFC con-nects predominately with motor and rewardareas (Carmichael & Price 1995a, Chiba et al.2001).

NEUROPSYCHOLOGY OF OFC

Having familiarized ourselves with theanatomical organization of OFC, let us nowreturn to patients with OFC damage. Recall

34 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

that such patients often show intact perfor-mance on a wide range of neuropsychologi-cal tests, including those designed to detectfrontal lobe dysfunction. Indeed their braindamage was so difficult to detect that they of-ten had trouble obtaining insurance paymentsor disability benefits (Damasio 1994). In theearly 1990s two tests were developed that weresensitive to OFC damage in humans. The firstof these derived from research into the ef-fects of lesions of OFC in monkeys and testedpatients’ ability to switch stimulus-reward as-sociations. The second led to an influentialtheory called the somatic marker hypothesis,which described how OFC might use auto-nomic signals to guide decision-making.

Stimulus-Reward Learningand Flexible Behavior

One of the earliest deficits associated withOFC lesions in monkeys was a failure toperform stimulus-reward reversals (Mishkin1964). A monkey learns that choosing one oftwo objects will lead to a reward. Then thecontingencies reverse and the monkey mustlearn that to get a reward he now has to choosethe previously unrewarded object. Monkeyswith lesions of OFC were impaired at the task.Following the reversal, they were unable to in-hibit responding to the previously rewardedobject, a behavior called perseveration. Themonkeys seemed to have difficulty modify-ing their behavior even when it was no longersuccessful in obtaining reward. Later studiesrevealed that the deficit was specific to OFC(monkeys with lesions of lateral PFC wereunimpaired) and depended on serotonergicinnervation (Clarke et al. 2004, 2006; Diaset al. 1996).

The task was adapted to test humans, andinvestigators found that frontal lobe dam-age impaired performance (Rolls et al. 1994).In addition, the extent of the patient’s im-pairment on the task correlated with the ex-tent to which his/her day-to-day behavior hadchanged. Thus, the same deficit that under-lies the inability to reverse stimulus-reward

Somatic marker: abodily state orcentralrepresentation of thebodily statecorresponding to theconsequences ofchoosing a particularcourse of action

Perseveration: thetendency to continueor repeat apreviously rewardedact or activity evenwhen no longerappropriate

associations might also underlie the patient’spoor decisions. One possibility is that the pa-tient is unable to modify his/her behavior inresponse to negative feedback. For example,scam artists might initially work to gain ourtrust, but we realize their intentions before weare taken advantage of and modify our behav-ior accordingly. In contrast, Elliott may havebeen unable to modify his initial trust and sowas swindled by the fraudster. Only patientswith OFC damage have problems with rever-sal learning: Patients with damage to DLPFCare unimpaired (Fellows & Farah 2003), con-sistent with the previous findings in monkeys.

Somatic Marker Hypothesisand Decision-Making

Another neuropsychological test that aims tomimic patients’ day-to-day impairments is theIowa gambling task (Bechara et al. 1994). Inthis task, there are four decks of cards, and thesubject must choose from these decks one cardat a time. Each card wins the subject a smallamount of money, but some of the choices alsolose money. The aim of the game is to win asmuch money as possible. Unbeknownst to thesubject, two of the decks are risky: They are as-sociated with large gains but also large and fre-quent losses. In the long term, choosing fromthese decks is a losing strategy. In contrast, theother two decks are associated with small gainsbut small and infrequent losses, so choosingfrom these decks will win money overall. Con-trol subjects initially favor the decks associatedwith the largest gains but, after encounteringlosses, gradually realize that this choice willlose them money, and they alter their choicesaccordingly. Patients with OFC damage like-wise initially favor the decks associated withlarger gains, but unlike control subjects, theycontinue to favor these decks until they havelost all their money. The deficit was specificto patients with damage to OFC: Lesions ofDLPFC did not affect performance (Becharaet al. 1998).

OFC patients also had unusual autonomicresponses during the performance of the task

www.annualreviews.org • OFC and Decision-Making 35

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

SCR: skinconductanceresponse

Outcome: theconsequence of anaction; can be eitherpositive (reward) ornegative(punishment)

(Bechara et al. 1997). During learning, con-trol subjects showed a marked increase in theirskin conductance responses (SCRs) immedi-ately before making a selection from one ofthe risky decks. This anticipatory SCR wasmissing in OFC patients. However, there wasnot a general disruption of autonomic pro-cessing. OFC patients continued to show anSCR to the delivery of the monetary reward orpunishment. Nor was there a general failure oflearning to affect the autonomic nervous sys-tem. OFC patients developed normal SCRsduring conditioning to a loud noise (Becharaet al. 1999).

From these results, Damasio (1994) devel-oped a theory of how autonomic responsesmight facilitate decision-making called thesomatic marker hypothesis. He argued thatbodily states corresponding to the emotionsproduced while evaluating different coursesof action (so called somatic markers) help tofacilitate normal decision-making. The role ofOFC is to store associations between patternsof environmental inputs and the somatic statesthat those inputs produce. When making adecision, OFC activates the somatic states,which can then bias decision-making. Dam-age to OFC destroys patients’ ability to ac-tivate the somatic states, and so all choiceoutcomes become emotionally equivalent. Inthis state, the patient must rely on a cog-nitive appraisal of a decision. Consequently,the myriad of variables needed to assess achoice can easily overwhelm the decision-making process. In effect, the patient loses theability to make a decision by gut feeling.

These ideas embody many of the conceptsfrom the James-Lange theory of emotions,which argued that changes in our autonomicstate gave experiences an emotional quality( James 1884, Lange 1922). Accordingly,many of the criticisms leveled at the James-Lange theory (Cannon 1927) might apply tothe somatic marker hypothesis. For example,patients with autonomic failure due to periph-eral denervation of autonomic neurons do nothave deficits in everyday decision-making and

perform normally on the Iowa gambling task(Heims et al. 2004). To circumvent such criti-cisms, Damasio proposed that somatic mark-ers did not necessarily have to operate throughthe peripheral autonomic nervous systembut might operate through an “as-if ” centralrepresentation. In effect, OFC might use so-matosensory cortex to simulate the emotionthat a particular course of action would pro-duce. Unfortunately no direct evidence yetdemonstrates that somatosensory cortex is in-volved in decision-making. Although it is in-volved in recognizing emotion (Adolphs et al.2000), we do not know how patients with dam-age to somatosensory cortex would performon the gambling task. Furthermore, centrallymediated responses to emotional stimuli,such as the P300 orienting response, showenhancement, not reduction, in OFC patients(Rule et al. 2002). It is difficult to reconcile thisenhanced orienting response with an inabilityto activate a central representation of somaticstates.

Another difficulty with interpreting the re-sults from the gambling task is that it is un-clear precisely which psychological mecha-nisms the task taxes. For example, the originalformulation of the task requires a reversalof a stimulus-reward contingency. The re-ward contingencies cause subjects to respondinitially to the high-reward decks, but theymust then switch to the lower-reward decksto win money in the long term. To addressthis confound, Fellows & Farah (2005) de-veloped a “shuffled” version of the gamblingtask, which used reward contingencies thatdid not initially bias the subject toward anyof the decks. Patients with OFC damage werenot impaired on this version of the task. Fur-thermore, there was a positive correlation be-tween the size of impairment that the patientsdisplayed on a stimulus-reward reversal taskand the size of their impairment on the orig-inal Iowa gambling task. The results suggestthat the deficits on the gambling task mighthave arisen from the problems that OFCpatients have in reversing stimulus-reward

36 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

associations. However, despite these con-cerns, the somatic marker hypothesis remainsthe only theory to attempt to explain both theautonomic changes observed in OFC patientsas well as their decision-making impairments.

Summary

The 1990s saw the development of two theo-ries to account for deficits seen in humans withOFC damage: stimulus-reward inflexibilityand the somatic marker hypothesis. However,one difficulty with interpreting the neuropsy-chological literature is determining whichmechanisms underlie the deficits. There is notnecessarily a straightforward link between theprocess performed by a region and the behav-ioral deficit that damage to that region pro-duces (if one removes the capacitor from aradio, the radio will howl, but that does notmean the function of the capacitor is to in-hibit howling). To understand what the un-derlying mechanisms might be, we turn to theneurophysiological literature. A caveat, how-ever, is that any properties we observe in OFCneurons must be capable of explaining theneuropsychological impairments that followOFC damage.

NEURONAL MECHANISMSWITHIN OFC

Specialization of OFC for RewardProcessing

The first neurophysiological studies of OFCnoted the frequency of neurons that showedselective responses to the delivery of foodand liquid rewards (Rosenkilde et al. 1981).They also supported the notion that OFCwas important for stimulus-reward reversals.In a stimulus-reward reversal task, neuronsshowed differential activity to two visual stim-uli, one of which predicted the delivery offruit juice and the other of which predictedthe delivery of saline (Thorpe et al. 1983).Such neurons were not simply encoding the

visual properties of the stimulus; when the re-ward contingencies reversed, the neuronal se-lectivity would also reverse. Thus, the neu-rons appeared to be encoding the rewardpredicted by the stimulus and expected bythe monkey.

Later studies began to challenge the notionthat these properties were unique to OFC.Reward-selective neurons, that is, neuronsthat show different firing rates depending onthe expected reward outcome, were found inmany different brain areas. For example, manystudies demonstrated that reward-selectiveneurons were also in DLPFC (Amemori& Sawaguchi 2006, Hikosaka & Watanabe2000, Kobayashi et al. 2002, Leon & Shadlen1999, Watanabe 1996), and DLPFC neuronsshowed similar responses to OFC neuronsduring the performance of stimulus-rewardreversal tasks (Wallis & Miller 2003). Partic-ularly challenging was a study by Roesch &Olson (2003), which examined the influenceof reward expectation on neurons throughoutthe frontal lobe. The prevalence and strengthof reward selectivity were weakest in PFC andstrongest in motor areas such as premotorcortex. Reward-selective neurons were alsofound in posterior cortex, such as perirhinalcortex (Liu & Richmond 2000), parietal cor-tex (Musallam et al. 2004, Platt & Glimcher1999), and even primary visual cortex (Shuler& Bear 2006). However, we must be carefulin interpreting these results. A neuron is notnecessarily encoding a reward just because itsfiring rate correlates with some parameters ofthat reward. Many behavioral and cognitivemeasures also correlate with expected rewardand may equally be driving the neuron’s re-sponse. For example, an animal’s muscles willoften tense when it expects a large reward,and its behavior will be quicker and more ac-curate (Roesch & Olson 2003). Another ar-gument is that animals pay more attention tocues that predict reward (Maunsell 2004) andenter a state of higher autonomic arousal. Anyof these processes may be driving neuronalfiring rates.

www.annualreviews.org • OFC and Decision-Making 37

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Hedonic value: theamount of pleasureor pain associatedwith an outcome(how much we likesomething)

Incentive value: thedegree of desirabilityof an outcome (howmuch we wantsomething)

Which Aspect of Reward is OFCEncoding?

Given that reward has multiple behavioral andcognitive correlates, how are we to determinewhich aspect of reward OFC neurons encode?Because we know that OFC damage impairsdecision-making, a sensible place to begin isby concentrating on those aspects of rewardthat drive our choices and decisions. How-ever, multiple aspects of a reward can driveour behavior. For example, rewards have ahedonic value (how much we like something)and an incentive value (how much we wantsomething) (Robinson & Berridge 1993). Fornow, we refer to these multiple aspects simplyas value, but we return later to look at whichfactors make one reward more valuable thananother. The first question we must address,however, is whether OFC neurons are encod-ing value or are, in fact, encoding one of thecorrelates of value.

One approach is to compare the neuronalresponse to rewards and punishment. The ra-tionale is that punishers should have manyof the same behavioral and cognitive seque-lae that rewards do. For example, punish-ment motivates behavior, focuses attention,and produces arousal. However, in terms ofits value punishment is clearly different fromreward. We try to obtain reward and avoidpunishment. Thus, neurons encoding valueshould show a difference in activity betweenrewards and punishments. In contrast, neu-rons encoding some sequelae of the rewardshould show a similar response to both re-wards and punishers. Using this rationale,Roesch & Olson (2004) compared neuronalresponses in OFC and premotor cortex whenan animal made choices based on the size ofeither a reward or a punishment. OFC neu-rons tended to fire more strongly to choicespredicting larger rewards and showed a de-creased firing rate when choices predictedlarger punishments. In contrast, neurons inpremotor cortex showed stronger responsesto choices indicating larger punishments orrewards. From these results, they concluded

that OFC was encoding the value of a choice,whereas the reward-selective responses inpremotor cortex were actually indicative ofthe increased motor readiness an animal ex-hibits when it is making an important choice(that is, one associated with large amounts ofreward or punishment).

A second approach has been to comparethe neuronal latency at which reward selectiv-ity appears across various brain regions. Therationale in this case is that brain areas extract-ing the value of a choice should display re-ward selectivity before those areas responsiblefor using the value information to control be-havior and cognition. To investigate this idea,we trained monkeys to choose between dif-ferent pictures associated with delivery of dif-ferent amounts of fruit juice (Wallis & Miller2003). Pictures appeared on the left and rightof a screen, and monkeys were required tomake a saccade to the picture they wanted tochoose. Monkeys soon learned to maximizetheir reward by selecting pictures associatedwith larger rewards. We recorded simultane-ously from DLPFC and OFC and found neu-rons in both areas that encoded the size ofreward. Figure 2 illustrates the time-courseof this encoding across the DLPFC and OFCpopulations of neurons. The measure of selec-tivity is derived from the receiver operatingcharacteristic (ROC) of each neuron’s firingrate. The ROC is the probability that an in-dependent observer could correctly identifythe payoff given the firing rate of the neu-ron. No selectivity equates to an ROC valueof 0.5 (in practice it is slightly higher thanthis because we rectify the ROC value dur-ing its calculation. Small fluctuations due tonoise push the value to about 0.52). Maximalselectivity equates to a value of 1.0. Both pop-ulations begin to encode the expected pay-off at about the same time, but selectivityreaches its peak value in OFC ∼60 ms be-fore it does in DLPFC. In addition, OFCneurons tended to encode the reward alone,whereas DLPFC neurons encoded a combi-nation of the reward and the upcoming motorresponse (Figure 3). We recently replicated

38 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

0

Time from onset of reward-predictive cue (ms)

500 1000

0

5

10

15

0

5

10

0.52

0.54

0.56

0.58

0.60

0.62

OFCDLPFC

Sele

cti

vit

y f

or

rew

ard

pre

dic

tio

n

(RO

C v

alu

e)

Nu

mb

er

of

neu

ron

s

Time from onset of reward-predictive cue (ms)

OFC

DLPFC median 570 ms

median 510 ms

a

b

250 500 750 1000 1250

250 500 750 1000 1250

Figure 2(a) Time-course ofmean selectivity forencoding anexpected rewardacross the DLPFC(blue) and OFC (red )population ofneurons. Error barsindicate the standarderror of the mean.(b) Distribution ofpeak selectivityacross the populationof DLPFC and OFCneurons. The OFCpopulation reachesits peak selectivity∼60 ms before theDLPFC population(Wilcoxon’srank-sum test,P < 0.05).

these findings, using a very different behav-ioral paradigm designed to investigate the in-teraction of reward information with spatialworking memory (Kennerley & Wallis 2006).In this paradigm, OFC neurons encoded thevalue of a reward-predictive cue 110 ms beforeDLPFC neurons did. From these results, we

have concluded that OFC encodes the valueof a choice outcome and then passes this in-formation to DLPFC, which uses the infor-mation to control behavior. Our results areconsistent with previous neurophysiologicalstudies because reward-selective neuronswere in both DLPFC and OFC. However,

www.annualreviews.org • OFC and Decision-Making 39

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

08 4 2

10

20

0

10

20

0 500

Left saccade Right saccade

0

20

40

0 500

0

20

40

0

5

10

15

20

25

0

10

20

30

40

50

0

5

10

15

20

25

0

10

20

30

40

50

Time from onset of reward-predictive cue (ms)

Neu

ron

al fi

rin

g r

ate

(H

z)

a b

c d

8 4 2

8 4 2 8 4 2

Figure 3Spike histograms from two single neurons encoding the expected reward and/or the monkey’s response (aleft or right saccade). Inset bar graphs indicate the mean neuronal firing rate ( ± standard error) duringthe presentation of the reward-predictive cue (the first 500 ms). Gray indicates that the cue predicted thedelivery of eight drops of juice, blue four drops, and red two drops. (a, b) OFC neuron encoding thepredicted reward in a parametric fashion irrespective of saccade direction. This neuron showed adepression in its firing rate that was greatest for eight drops of juice, less for four drops, and least of allfor two drops. Its firing rate, however, was the same irrespective of whether the monkey would make aleft or right saccade to earn the reward. Significantly more OFC neurons (28%) showed this pattern ofselectivity compared with DLPFC neurons (13%, chi-squared = 9.8, P < 0.005). (c, d ) A DLPFCneuron that showed a complex pattern of selectivity that encoded a combination of the reward and theupcoming saccade. During the cue epoch, the neuron discriminated between the different expectedreward amounts only when the monkey would make a rightward saccade (showing a high firing rate wheneight drops of juice were expected). In contrast, during the subsequent period the same neuron wasreward-selective only when the monkey would make a leftward saccade. Significantly more DLPFCneurons (43%) encoded a combination of the reward and response compared with OFC neurons (19%,chi-squared = 19, P < 0.00,005).

40 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

they are also consistent with the neuropsy-chological literature, suggesting a preeminentrole for the OFC in reward processing, be-cause our results suggest that OFC is thesource of reward signals to DLPFC.

In summary, neurons sensitive to rewardparameters are widespread in the brain, whichis not surprising. Obtaining rewards is criti-cal to our survival, and so rewards drive manyof our behavioral and cognitive processes.Thus we must be careful to differentiate be-tween neurons that are directly responsiblefor encoding the value of a reward and thosethat encode a cognitive process that rewardshappen to affect. Two different approachesboth point to the importance of OFC withinthe frontal lobe for encoding a reward’svalue. OFC neurons respond quickly to re-wards and differentiate between rewards andpunishments.

However, OFC is not critical for muchsimple reward processing. For example, ananimal with bilateral OFC removal is stillmotivated to work for reward (Izquierdoet al. 2004, Pears et al. 2003), can learnthat a neutral stimulus predicts food (Pickenset al. 2003), and can make choices be-tween rewarded and unrewarded alternatives(Izquierdo et al. 2004, Rudebeck et al. 2007).In real life, however, most of our choices arenot so simple. We often have to consider mul-tiple parameters of an outcome and weigh itspros and cons. We may or may not experiencea physically identical outcome as rewarding,depending on our needs and the context inwhich it occurs. In all, considerable processingmay be required to determine that a reward isactually rewarding. In the following section,we explore some of these processes and ex-amine the evidence for their dependency onOFC.

Complexity of Reward Processing:Calculating a Reward’s Value

Rewards involve integration and trade-off.Many of us could live in a larger house if wewere prepared to accept a longer commute

to the workplace. Our decisions about resi-dence location depend on a trade-off betweenthese two factors. Many of our everyday de-cisions are similarly complex often requiringus to weigh the pros and cons of several vari-ables. A recent study by Padoa-Schioppa andAssad (2006) shows that OFC neurons inte-grate multiple sensory features of a reward todetermine its value. Monkeys made choicesbetween different volumes of different typesof juice reward. To make its choice effectively,the monkey needed to consider both variables.For example, a thirsty monkey might preferthe taste of fruit juice to water. If so, then ifthe choice is between equal volumes of both,he will obviously choose the juice. However,increasing the volume of water available cancompensate for its less desirable taste. If thevolume of water is sufficiently large, relativeto the juice volume, then the monkey will pickthe water. At some point, the volume of wa-ter will compensate for its less desirable tasteexactly, and the monkey will be indifferentbetween the two choices. This measures themonkey’s value of one reward’s taste relativeto the other. For example, if the monkey isequally likely to choose four drops of wateror one drop of fruit juice, we know that themonkey considers the taste of juice four timesmore valuable than water.

The firing rates of OFC neurons weremore likely to vary systematically withthe value of the drinks, rather than withthe drinks’ physical properties, such as theirtaste or volume. To see how the authors de-termined that the neurons were encoding thevalue of the drinks, we return to the juiceand water example. A neuron that was en-coding the value of the chosen reward mightshow a higher firing rate when the monkeywas choosing one drop of juice comparedwith when he was choosing one drop of wa-ter. However, the neuron’s firing rate wouldbe the same when the monkey was choos-ing one drop of juice compared with whenhe was choosing four drops of water. We can-not explain this pattern of neuronal activity onthe basis of the drinks’ volume because equal

www.annualreviews.org • OFC and Decision-Making 41

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

volumes of the drinks produce different neu-ronal firing rates. Nor can we explain it solelyby the drinks’ taste because certain volumesof the drinks produce equal levels of neuronalfiring. However, we can explain it in termsof the monkey’s valuation: When his valua-tion of the two drinks is the same (such aswhen there are four drops of water or onedrop of juice), the neuronal firing rate is alsoequivalent.

Although the authors focused on gustatorystimuli, these processes could easily apply tohigh-level decision-making. Indeed, patientswith OFC damage have difficulty integratingmultiple attributes pertaining to a decision(Fellows & Farah 2005). For example, patientswere poor at integrating the factors that mightgo into choosing an apartment (size, location,etc.). Anatomically, OFC is ideal for the mul-timodal integration of the parameters neces-sary to evaluate an outcome because it receivesinputs from all sensory modalities. OFC neu-rons in the monkey respond to visual, olfac-tory, and gustatory aspects of rewards (Rolls& Baylis 1994). Neuroimaging reveals thathuman OFC is activated by pleasant and un-pleasant smells, sights, sounds, and touches(Rolls et al. 2003, Royet et al. 2000), as well asmore abstract rewards and punishments, suchas receiving or losing money (Breiter et al.2001, O’Doherty et al. 2001).

Reward is relative. Unlike physical proper-ties, such as luminance or pitch, reward is dif-ficult to measure in absolute terms. The valueof a reward depends on other potential re-wards. For example, you might be delightedto receive a $1000 pay raise until you findout that all your coworkers received $5000.In an analogous laboratory situation, investi-gators explored the ability of subjects to expe-rience regret by having them rate their emo-tional experience during a task where theyhad to choose between two spinners (Camilleet al. 2004). Depending on where the spinnerlanded, the subject might win or lose money.Critically, however, in some conditions thesubject saw what would have happened had

they chosen the alternate spinner. For controlsubjects, this simple manipulation could evokeregret by turning an otherwise positive expe-rience (winning money) into a negative expe-rience (if the subject would have won moremoney by choosing the alternative spinner).Patients with OFC damage showed a differ-ent pattern of emotional experience duringthis task. They still reacted positively or neg-atively to winning or losing money, but theoutcome of the alternative spinner did not af-fect their emotional experience and they didnot experience regret. Their deficit seemed tobe in representing or simulating what wouldhave happened had they chosen the alterna-tive. Some evidence also indicates that OFCneurons encode rewards in a relative manner(Tremblay & Schultz 1999). For example, if amonkey prefers raisins over cabbage and or-anges over both raisins and cabbage, then anOFC neuron might respond to raisins if themonkey’s choice is limited to raisins and cab-bage, but to oranges if the choice is betweenoranges and raisins.

Rewards must satisfy a need. Something isvaluable only in the sense that it meets someneed. We might pay $40 for a good steak, butif we were extremely thirsty, we would pre-fer to spend the money on water. Our needsare often complex, encompassing physiolog-ical, cognitive, emotional, and social factors,but in all cases, these needs affect how valu-able a reward is. It also means that the exactsame physical stimulus might be rewarding (acold beer on a hot summer night) or aversive(that same beer the morning after) depend-ing on our motivational state. Indeed, someevidence has shown that neuronal activity inOFC reflects our physiological needs. For ex-ample, some OFC neurons initially respondwhen a thirsty monkey tastes fruit juice, butthe neuronal response declines as the animaldrinks more juice and becomes sated (Rollset al. 1989). This contrasts with gustatorycortex, where neuronal responses to gusta-tory stimuli remain constant irrespective ofthe animal’s motivational state (Yaxley et al.

42 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

1988). Neuroimaging studies have shownsimilar results in human OFC (Gottfried et al.2003).

Associations can make neutral eventsrewarding. Primary reinforcement, i.e.,things we find intrinsically pleasant such assweet tastes or orgasms, does not drive allour behavior. Instead, we direct much of ourbehavior toward secondary reinforcers, whichwould be otherwise neutral if it were not fortheir association with primary reinforcers. Agood example in humans is money. Moneyis simply paper, with no intrinsic rewardingproperties. Yet by its association with primaryreinforcement (the good things in life thatmoney can buy), the sight of a large sum ofmoney has the same effect on people as doesprimary reinforcement. It is a positive andarousing emotional experience, and peoplewill work for money just as they might workfor food.

Turning to the neuronal mechanisms thatunderlie this process, OFC neurons do re-spond to previously neutral stimuli that pre-dict the delivery of food (Schoenbaum et al.1998, Thorpe et al. 1983, Wallis & Miller2003). However, this does not necessarilymean that such neurons are encoding sec-ondary reinforcement. We must be carefulto distinguish between cues that simply pre-dict primary reinforcement and cues that aregenuine secondary reinforcers. The differ-ence is that secondary reinforcers, throughtheir association with primary reinforcement,are now rewarding in their own right. Wecan demonstrate this by teaching animals newresponses using solely secondary reinforcersas the reward (Mackintosh 1974). Lesions ofOFC in monkeys disrupt the ability of sec-ondary reinforcers to support new learning(Pears et al. 2003), which suggests that OFCis critical in mediating the rewarding effectsof secondary reinforcement.

Summary. To summarize, OFC seems to beparticularly involved in complex situationswhere significant processing is required to

Primaryreinforcement:outcomes that haveinnate reinforcingqualities, such as thepleasure from a sweettaste or an orgasm

Secondaryreinforcement:outcomes where thereinforcing qualitiesare learned throughtheir association withprimary reinforcers,such as money

Neuronal currency:an abstract signalencoded by neuronsto indicate the valueof a behavior

Neuroeconomics:the relation ofneuronal activity tomodels derived fromeconomics andbehavioral ecology

determine the value of the outcome. Howfar can we extend this idea? One proposalsuggests that OFC encodes a “neuronal cur-rency” by integrating all the relative param-eters pertinent to a decision (Montague &Berns 2002). We explore this idea in the nextsection.

Neuroeconomics

To make sense of neuronal data, neurophys-iologists must compare neuronal responsesagainst a model of the behavioral or cogni-tive process that the neuron is putatively en-coding. Traditionally neurophysiologists haveused models derived from sensorimotor psy-chophysics or animal learning theory. Overthe past decade, however, scholars have real-ized that to understand the neuronal mech-anisms underlying decision-making, it mighthelp to widen the fields from which we con-struct our behavioral models. Evolutionarybiologists and economists have constructeddetailed models of the parameters that ani-mals and humans use to make everyday deci-sions. Neuroeconomics refers to the nascentfield that attempts to relate these models topatterns of neuronal firing (Glimcher 2003,Sanfey et al. 2006, Schultz 2004).

These models emphasize the considera-tion of three basic parameters that one needsto consider when making a decision: theexpected reward or payoff, the cost in termsof time and energy, and the probability of suc-cess (Kahneman & Tversky 2000, Stephens& Krebs 1986). Determining the value ofa choice involves calculating the differencebetween the payoff and the cost and dis-counting it by the probability of success. Onesuggestion is that OFC integrates all theseparameters to derive an abstract measure ofthe value of a choice (Montague & Berns2002). This encoding scheme offers distinctcomputational advantages. When faced withtwo choices, A and B, one might imagine itwould be simpler to compare them directlyrather than going through an additionalstep of assigning them an abstract value.

www.annualreviews.org • OFC and Decision-Making 43

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

The problem with this process is that as thenumber of available choices increases, thenumber of direct comparisons increases expo-nentially. Thus, choosing between A, B, andC would require three comparisons (AB, AC,and BC), and choosing between A, B, C, andD requires six comparisons (AB, AC, AD, BC,BD, and CD). The solution quickly suffersfrom combinatorial explosion as the numberof choices increases. In contrast, valuingeach choice along a common reference scaleprovides a linear solution to the problem.

An abstract representation provides im-portant additional behavioral advantages,such as flexibility and a capacity to deal withnovelty, both of which are hallmarks of pre-frontal function. For example, suppose an an-imal encounters a new food type. To deter-mine whether it is worth choosing relative toother potential food sources, the animal mustdetermine the value of that food. If the an-imal relies on making direct comparisons, itcan determine this only by iteratively compar-ing the new food with all previously encoun-tered foods. If the animal calculates an abstractvalue, however, it has to perform only a sin-gle calculation. By assigning the new food avalue on the common reference scale, it knowsthe value of this foodstuff relative to all otherfoods. Second, it is often unclear how to com-pare directly very different outcomes. Howdoes a monkey decide between grooming aconspecific and eating a banana? Valuing thealternatives along a common reference scalecan help. For example, although I have neverneeded to value my car in terms of bananas,I can readily do so by assigning each item anabstract, monetary value.

Thus, there are good theoretical groundsfor expecting a neuronal system to encode thevalue of behavioral choices in an abstract man-ner, but is there empirical evidence for sucha system and does it reside in OFC? As wehave seen, some evidence demonstrates thatOFC encodes payoff information, and it doesso by calculating a value signal. This signal isabstract because a single sensory feature can-not explain it (Padoa-Schioppa & Assad 2006).

But does OFC also encode the other factorsrelevant to a decision, cost, and probability ofsuccess?

Cost

Although an outcome may be highly desir-able, we may not pursue it if the cost to obtainit is too great. For example, behavioral ecolo-gists have specified several costs in obtainingfood, such as search costs to find the food andhandling costs to render the food consumable(Stephens & Krebs 1986). If these costs ex-ceed the energy that the animal will derivefrom the food, then the animal does not at-tempt to obtain that food. Although there aremany different types of cost, we can describethem largely in terms of either energy (effort)or time (delay).

A recent study by Roesch & Olson (2005)shows that OFC neurons encode time costs.Monkeys performed a cognitive task, and thefinal reward for correct performance was ei-ther large or small or occurred after eithera short or a long delay. Confirming previ-ous results, OFC neurons tended to fire morestrongly when the monkey anticipated a largereward as opposed to a small one. However,they also tended to fire more strongly whenthe monkey anticipated reward after a shortdelay as opposed to after a long delay. Fur-thermore, the strength of an individual neu-ron’s response to the delay manipulations cor-related with its response to the manipulationsof reward size. Thus, the value signal encodedby OFC neurons incorporates not only mul-tiple sensory parameters of a reward but alsotemporal information.

A different picture emerges when consid-ering effort. Studies by Rushworth and col-leagues have implicated MPFC as responsi-ble for effort-based decisions (Walton et al.2002, 2003). Most recently, they have demon-strated a double dissociation between OFCand MPFC in the types of cost the two areasuse to guide decisions (Rudebeck et al. 2007).Rats learned to make decisions on a T-maze.The two arms contained different amounts of

44 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

reward as well as either barriers of differentheights that the rats had to clamber over(effort manipulation) or gates that would openonly after the rat had been in the arm for aspecific amount of time (delay manipulation).Lesions of MPFC biased rats toward choos-ing less effortful alternatives but did not affectdecisions involving delays. In contrast, lesionsof OFC biased rats toward choices with moreimmediate access to the reward but did notaffect decisions involving effort. The impair-ments did not relate to changes in sensitiv-ity to reward or motor abilities. If the armswere equal in terms of effort or delay, then therats consistently chose the arm associated withlargest reward. Thus, whereas ecological andeconomic models often lump together costsin terms of time and effort, the brain has notadopted this solution. OFC factors time costsinto decisions, whereas MPFC factors effortcosts. This functional dissociation is consis-tent with the stronger connections betweenMPFC and motor regions relative to OFC,which places it in a better position to evalu-ate effort (Carmichael & Price 1995b, Cavadaet al. 2000, Chiba et al. 2001).

Probability of Success

A second parameter that we need to considerwhen assessing a choice alternative is the un-certainty in obtaining the outcome. However,this uncertainty could arise owing to sourcesof variation anywhere along our cognitiveprocessing pathways. For example, we mightnot get the behavioral outcome we expectedbecause we misinterpreted sensory informa-tion owing to perceptual ambiguity. A batterhas a fraction of a second to determine whatpitch has been thrown; failure to do so couldresult in a strike rather than a home run. Al-ternatively, perhaps the relationship betweenthe sensory stimulus and the response is un-certain; a goalkeeper does his best to predictwhich direction a striker intends to shoot apenalty, but even so, he will often dive in thewrong direction. Sometimes our interpreta-tion of the sensory situation is correct, and we

make the correct response, but the outcomeis inherently uncertain. The poker player ishappy to put his money into a pot when dealtpocket aces, even though he may be outdrawn,because in the long run this is the most prof-itable course of action. Even our estimationof risk can itself vary. For example, some risksare certain (the probability that the rouletteball will land on black), whereas other risks areambiguous (the probability that it will rain to-morrow). Finally, risk interacts with emotion.Uncertainty in a negative situation makes thesituation even more unpleasant, whereas un-certainty in positive situations can be excitingand fun (as a visit to the casino can attest).

Given the different sources of variance thatcan affect the probability of a given outcome,it is perhaps not surprising that such ma-nipulations activate diverse brain areas. Neu-roimaging studies have revealed that multiplebrain areas activate during decisions involv-ing uncertainty, including OFC and MPFC(Hsu et al. 2005, Knutson et al. 2005), butalso lateral PFC, parietal, cingulate, and in-sular cortex (Critchley et al. 2001, Kuhnen& Knutson 2005, Yoshida & Ishii 2006). Fur-thermore, some studies have begun to dissoci-ate how different brain regions are involved inprocessing different types of risk. For exam-ple, different networks process certain versusambiguous risks (Huettel et al. 2006). Neu-rophysiology studies show that neurons in avariety of regions, including MPFC (Amiezet al. 2006), parietal cortex (Platt & Glimcher1999), and dopamine neurons (Fiorillo et al.2003), combine information about the size ofa payoff and its probability of occurrence toderive an expected value for a given action.

Integrating Multiple DecisionParameters

Thus, several different brain regions appear toencode the parameters underlying decisions.Is there any evidence that a single region in-tegrates these parameters to derive an overallvalue for a decision? A recent study in ourlaboratory tested this question by recording

www.annualreviews.org • OFC and Decision-Making 45

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

simultaneously from OFC, MPFC, DLPFC,and VLPFC while monkeys made decisionsguided by the amount of reward, the probabil-ity of reward delivery, and the effort requiredto obtain the reward (Kennerley et al. 2005).Within OFC, DLPFC, and VLPFC, few neu-rons encoded these parameters (<10%). InMPFC, however, more than half the neuronsencoded at least one of the parameters, and aquarter of the neurons encoded two or more ofthe parameters. Thus, MPFC neurons seemto encode a variety of information pertinentto decision-making. Lesion studies also sug-gest a role for MPFC in using behavioral out-comes to guide actions (Kennerley et al. 2006).In sum, MPFC may be a better candidate forencoding an abstract value signal than is OFCbecause it accounts for not just the value of theoutcome, but also the effort involved in ob-taining that outcome and the likelihood of theaction being successful. It can then integratethis information to derive an overall value ofthe behavior.

A MODEL OF DECISION-MAKING WITHIN PFC

Nature of the OFC RewardRepresentation: Working Memoryfor Value

Having considered the type of informationthat OFC encodes, we now consider the na-ture of this encoding. Although OFC is a sub-region of PFC, many of the current mod-els of PFC function pay less attention toOFC compared with lateral PFC (Duncan2001, Goldman-Rakic 1987, Koechlin et al.2003, Miller & Cohen 2001, Petrides 1996,Shimamura 2000). In addition, there is a dis-connection in the types of tasks used to testfunctions of different PFC subregions. Forexample, tests of lateral PFC function typ-ically focus on some type of sensory work-ing memory (Funahashi et al. 1989, Rao et al.1997, Romo et al. 1999, Wilson et al. 1993).In contrast, tasks used to examine OFC func-

tions test stimulus-reward associations held inlong-term memory (Roesch & Olson 2004;Schoenbaum et al. 1998, 1999; Thorpe et al.1983; Tremblay & Schultz 1999; Wallis &Miller 2003).

Yet when one records the activity of OFCneurons during the performance of workingmemory tasks, their activity can closely re-semble that of lateral PFC neurons (Walliset al. 2001). Thus, we suggest that OFC neu-rons operate in much the same way as doneurons in the rest of PFC: by holding in-formation in working memory across shortdelays and using that information to bias ac-tivity in other areas in a behaviorally relevantmanner (Miller & Cohen 2001, Shimamura2000). The difference between these areas liesin the nature of the information encoded, anidea originally espoused by Goldman-Rakic(1987). Lateral PFC encodes sensory infor-mation, behavioral responses, and the contextin which these occur. In contrast, OFC en-codes the potential goals and outcomes to-ward which we can direct our behavior. Itencodes this information as a value signal, en-suring that the outcomes that satisfy our needsreceive behavioral priority. This value signalrelates only to the outcome itself and not tothe means to achieve that outcome.

What advantages does such a system con-fer over simpler mechanisms such as asso-ciative learning? Associative learning dependson trial-and-error, which has inherent draw-backs. First, to modify our behavior, we mustexperience the outcome. This is problem-atic particularly for learning about aversiveoutcomes, which may be physically harmful.Second, obtaining a particular outcome mayrequire considerable planning and effort. Asystem that can encode the value of the out-come beforehand can avoid wasting time andenergy on outcomes that are not sufficientlyvaluable. Thus, the OFC system may be par-ticularly important for planning behavior to-ward distant rewards, compared with morelow-level subcortical systems, which might beadequate for obtaining immediate rewards.

46 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Third, in the real world we often do not havethe opportunity to engage in trial-and-errorlearning: Choices are often one-time deals.

As an example, consider an experimentby Murray and colleagues (Izquierdo et al.2004). In one task, they allowed monkeys tochoose between different food items. The ex-perimenters could manipulate the monkeys’choices: If the animals were fed one of thefoods prior to testing, they were less likelyto choose this food during testing, a phe-nomenon called “sensory-specific satiety.” Ina second version, they trained monkeys suchthat selecting particular objects would leadto the delivery of specific food items. Dur-ing testing, the monkey chose between pairsof objects, but each object pair appeared onlyonce. Again, the experimenters could manip-ulate the monkeys’ choices by varying whatthe monkeys ate prior to testing.

Monkeys with OFC lesions showed nor-mal patterns of sensory-specific satiety on thefirst version of the task, but not on the sec-ond. A key difference between the two tasksis that the monkeys can engage in trial-and-error learning in the first version of the taskbut not in the second, where they face eachchoice only once. Instead, they must rely on aninternal representation of value to guide theirchoice. As an analogy, compare the processesthat might take place if you were selecting ameal from an all-you-can-eat buffet or froma menu. In the first case, you are free to se-lect foods, try them, and decide whether youwant more. In the second, this is not an option.You can guide your choice only by using infor-mation about your current motivational stateand the sensory properties of the food item togenerate a representation of the expected con-sequences of choosing that food. Indeed a re-cent study showed OFC activation when sub-jects considered food items on a menu (Aranaet al. 2003). This mechanism might also ex-plain why OFC patients fail to experience re-gret (Camille et al. 2004): To experience re-gret we must generate a representation of theconsequences of the alternative outcome.

Thus, we speculate that a major functionof OFC neurons is to encode a value repre-sentation in working memory, which can beused to anticipate the future consequences ofour behavior. However, this value represen-tation might take one of two forms: either ahedonic representation (liking), which wouldindicate that a choice outcome would pro-duce a pleasurable emotional experience, oran incentive representation (wanting), whichwould indicate that a choice outcome was de-sirable. To date, research dissociating thesetwo systems has focused on subcortical andneurotransmitter systems rather than on cor-tical areas (Pecina & Berridge 2005, Wyvell& Berridge 2000). In the healthy individual,however, these two systems probably operatein a tight coupling. Only in unhealthy indi-viduals, such as drug addicts, do we see peo-ple expending considerable effort to obtain anoutcome that provides them with no pleasure.We do not know which system OFC uses, butit is also possible that the signals are integratedby the time they reach cortical areas so thatrewards are valued by both their hedonic andincentive properties.

Interaction of OFC with Other PFCRegions

Although OFC encodes the value of an out-come, it encodes little information aboutthe means to achieve the outcome. We sug-gest that lateral PFC and MPFC are cru-cial in this regard. Lateral PFC is responsiblefor the top-down control of cognitive pro-cesses that enable the construction of plansand organization of behavior necessary toobtain goals and outcomes. These processesare beyond the scope of this chapter, butseveral recent reviews have described themin detail (Duncan 2001, Miller & Cohen2001, Shimamura 2000). In contrast, MPFCcan use information from OFC as to thevalue of the outcome and evaluate the plansgenerated in lateral PFC (for example, interms of probability of success or effort) to

www.annualreviews.org • OFC and Decision-Making 47

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Sensoryinformation

e.g., gustatory cortex,temporal cortex

Affectiveinformation

e.g., amygdala

Motivationalinformation

e.g., hypothalamus

OFC

W O R K I N G

M E M O R Y

Integration of information

to derive the value of

potential reward outcomes

DLPFC

Construction of

plan to obtain

reward outcome

MPFC

Evaluation of

effort involved

in plan

Behavioral

response

Figure 4Model of the neuronal mechanisms underlying decision-making in PFC.

perform a cost-benefit analysis and generatean overall value for an action. Operating to-gether these major PFC divisions can controlbehavior.

Figure 4 summarizes these ideas in theform of a speculative model. Sensory, affec-tive, and motivational information about anoutcome enter OFC. For example, deliveryof a juice reward might involve informationabout the juice’s taste arriving from gusta-tory cortex, information that the taste is pleas-ant arriving from the amygdala, and informa-tion that the juice is thirst-quenching arrivingfrom hypothalamus. OFC then uses this infor-

mation to calculate the value of the outcomeand determine how rewarding it is. This in-formation passes to lateral PFC, which canuse it to construct behavioral plans, to priori-tize goals, and to generate expectancies aboutfuture events. In turn, MPFC can use infor-mation about the value of the outcome and thebehavioral plan to determine whether an ac-tion is worth performing. These calculationstake place in working memory, so the ongoingpattern of neuronal activity in each of theseareas reflects these processes.

Application of the Model to theCurrent Empirical Evidence

How well does this model stand up to the cur-rent empirical evidence? Some studies suggestthat OFC is not involved in working memory(Bechara et al. 1998, Hikosaka & Watanabe2000), but they have focused on spatial work-ing memory. Therefore, if a task was used thatrequired subjects to hold the value of a rewardor outcome in working memory, an involve-ment of OFC may be apparent.

How does the model relate to the deficitsseen in OFC patients? Concerning the gam-bling task, Damasio originally argued thatsubjects performed the task with little knowl-edge of the contingencies underlying theirsuccessful performance (Bechara et al. 1997).The contents of working memory are, by def-inition, consciously accessible, so this findingwould seem to preclude a role for workingmemory in the task. However, recent findingsshow that subjects are trying to track explic-itly the experimental contingencies (Maia &McClelland 2004), which suggests that hold-ing outcome information in working memorywould be useful for performance. The modelmay also help explain some of the neuropsy-chiatric illnesses in which OFC dysfunction isimplicated (see Clinical Implications).

Working memory for rewards might alsobe useful to solve stimulus-reward rever-sal tasks. Successfully performing these tasksinvolves gradually learning and unlearningstimulus-reward associations in long-term

48 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

memory. However, although these tasks canbe learned using stimulus-reward associa-tions, working memory might also contributeto the learning. Explicitly keeping track ofa stimulus and its associated outcome mightenable faster reversal. A role for workingmemory is most noticeable in trained mon-keys, who are capable of performing rever-sals following a single error, a phenomenonthought to depend on the development of areversal “learning set.” Increasing the lengthof the intertrial interval impairs the devel-opment of a learning set, suggesting an in-volvement for working memory (Deets et al.1970). Indeed, recent computational modelsof stimulus-reward reversals have incorpo-rated neurons that encode the reward contin-gencies of the task in their ongoing pattern ofneuronal activity (Daglish et al. 2001).

There is one class of deficits relating toOFC damage that does not fit with our model.OFC lesions in monkeys impair the learn-ing of stimulus-response associations (Busseyet al. 2001, Parker & Gaffan 1998). Howdo we reconcile these results with the ob-servations of a lack of neuronal activity inOFC relating to behavioral responses (Wallis& Miller 2003)? Closer examination of thedeficits of OFC animals reveals that theirproblems stem not from learning stimulus-response associations per se, but rather fromimplementing strategies that speed learn-ing (Bussey et al. 2001). Specifically, con-trol monkeys adopt a “win-stay, lose-shift”strategy, which is absent in monkeys withOFC damage. Such a strategy conceivablyrequires the monkey to hold in workingmemory the outcome of its choice on theprevious trial across the intertrial interval,which is compatible with our account of OFCfunction.

CONCLUSION

Damage to OFC produces a unique deficit,impairing everyday decision-making whileleaving other cognitive capabilities intact. Anextensive literature implicates the OFC in

CLINICAL IMPLICATIONS

OFC dysfunction is associated with disorders involving com-pulsive behavior (Volkow & Fowler 2000) such as obsessive-compulsive disorder, substance abuse, eating disorders, obe-sity and pathological gambling. Subjects report feeling out ofcontrol of their behavior, an immense desire to engage in thecompulsive behavior and a feeling of release once they do. Thisbehavior is thought to depend on the nucleus accumbens, aregion with which OFC heavily connects (Haber et al. 1995).For example, drugs of abuse are thought to sensitize the nu-cleus accumbens, which misdirects behavior towards drug ac-quisition (Everitt & Robbins 2005, Hyman & Malenka 2001,Robinson & Berridge 2003).

How might OFC influence this process? We have empha-sized how OFC is important for planning and obtaining dis-tant rewards and goals. Thus, one of its functions may be toprovide top-down control to the nucleus accumbens, biasingbehavior away from immediate rewards in the environment.This might be important when trying to quit a drug. For exam-ple, the recovering alcoholic must bring to working memorythe long-term goal to remain sober in order to inhibit thehabitual response to enter the liquor store. Disorders such assubstance abuse and obesity might result from lack of top-down control, while disorders such as anorexia might arisefrom too much control.

processing reward information, but in the lab-oratory situation, the experimenter usuallyendeavors to ensure that it is obvious whethera choice was rewarding. In real life, it is of-ten not so clear cut. The ultimate outcome ofour choice may not be apparent until sometime distant from when we make the choice,or it may occur unpredictably. We may needto consider multiple variables, some of whichwill be more or less important to us depend-ing on our present needs. Some choices couldhave negative consequences that we will needto offset against the positive. In short, in thereal world, considerable processing is oftenrequired to determine just how rewarding areward actually is. We suggest that this isthe role of the OFC. Furthermore, we haveproposed a mechanism by which this mighttake place. We suggest that the value of the

www.annualreviews.org • OFC and Decision-Making 49

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

expected outcome is held in working memory,in much the same way that sensorimotor andcontextual information are held in workingmemory in the lateral PFC. MPFC can thenuse this information to determine whether a

given action path is worthwhile, while lateralPFC can use the information to plan and co-ordinate behavior. In this way, these distinctPFC areas can act in concert to ensure that weget what we need.

SUMMARY POINTS

1. The anatomy of OFC suggests a region that integrates multiple sensory propertieswith affective information.

2. The functional properties of OFC are consistent with a direct role in reward process-ing rather than the performance of a function that merely correlates with reward.

3. OFC is responsible for calculating the value of a reward outcome, which includesassessing trade-offs, determining how well the outcome satisfies current needs, andcomparing the outcome with other potential reward outcomes.

4. OFC conceivably operates in a fashion analogous to the rest of PFC, holding infor-mation about the value of reward outcomes in working memory. This would be usefulfor formulating action plans, as well as predicting and monitoring expected outcomes.

5. Lateral PFC could use the value signal to plan the most efficient behavior, whereasmedial PFC may use the signal to perform a cost-benefit analysis of the plan.

FUTURE ISSUES TO BE RESOLVED

1. We need to specify more precisely the differences in function of OFC and MPFC.This will require experiments designed to detect double dissociations in the functionsof the two areas, including direct comparison of lesions of the two structures, as wellas the properties of neurons in both areas.

2. We need to understand how value information controls behavior: How does infor-mation pass between the different PFC areas, and how do subcortical structures, suchas the nucleus accumbens, use the information?

3. We need to determine the nature of the value signal in OFC. Does it relate more tohedonic value, incentive value, or a combination of the two?

4. We need to understand the neuromodulation of PFC. For example, does rewardinformation carried by dopamine neurons drive reward encoding in OFC or viceversa? How does the involvement of dopamine in spatial working memory relate tothe capacity of dopamine neurons to encode rewards?

ACKNOWLEDGMENTS

Grants from NIDA R01-DA019028 and the Hellman Family Faculty Fund support ourwork. I thank Elisabeth Murray and Arthur Shimamura for their thoughtful comments onthe manuscript. I also thank Steven Kennerley for valuable conversations that went into thedevelopment of many of the ideas in this review.

50 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

LITERATURE CITED

Adolphs R, Damasio H, Tranel D, Cooper G, Damasio AR. 2000. A role for somatosensory cor-tices in the visual recognition of emotion as revealed by three-dimensional lesion mapping.J. Neurosci. 20:2683–90

Amemori K, Sawaguchi T. 2006. Contrasting effects of reward expectation on sensory andmotor memories in primate prefrontal neurons. Cereb. Cortex 16:1002–15

Amiez C, Joseph JP, Procyk E. 2006. Reward encoding in the monkey anterior cingulate cortex.Cereb. Cortex 16:1040–55

Arana FS, Parkinson JA, Hinton E, Holland AJ, Owen AM, Roberts AC. 2003. Dissociablecontributions of the human amygdala and orbitofrontal cortex to incentive motivation andgoal selection. J. Neurosci. 23:9632–38

Bechara A, Damasio AR, Damasio H, Anderson SW. 1994. Insensitivity to future consequencesfollowing damage to human prefrontal cortex. Cognition 50:7–15

Bechara A, Damasio H, Damasio AR, Lee GP. 1999. Different contributions of the humanamygdala and ventromedial prefrontal cortex to decision-making. J. Neurosci. 19:5473–81

Bechara A, Damasio H, Tranel D, Anderson SW. 1998. Dissociation of working memory fromdecision making within the human prefrontal cortex. J. Neurosci. 18:428–37

Bechara A, Damasio H, Tranel D, Damasio AR. 1997. Deciding advantageously before knowingthe advantageous strategy. Science 275:1293–95

Breiter HC, Aharon I, Kahneman D, Dale A, Shizgal P. 2001. Functional imaging of neuralresponses to expectancy and experience of monetary gains and losses. Neuron 30:619–39

Bussey TJ, Wise SP, Murray EA. 2001. The role of ventral and orbital prefrontal cortex inconditional visuomotor learning and strategy use in rhesus monkeys (Macaca mulatta).Behav. Neurosci. 115:971–82

Camille N, Coricelli G, Sallet J, Pradat-Diehl P, Duhamel JR, Sirigu A. 2004. The involvementof the orbitofrontal cortex in the experience of regret. Science 304:1167–70

Cannon WB. 1927. The James-Lange theory of emotions. Am. J. Psychol. 39:115–24Carmichael ST, Price JL. 1994. Architectonic subdivision of the orbital and medial prefrontal

cortex in the macaque monkey. J. Comp. Neurol. 346:366–402Carmichael ST, Price JL. 1995a. Limbic connections of the orbital and medial prefrontal cortex

in macaque monkeys. J. Comp. Neurol. 363:615–41Carmichael ST, Price JL. 1995b. Sensory and premotor connections of the orbital and medial

prefrontal cortex of macaque monkeys. J. Comp. Neurol. 363:642–64Cavada C, Company T, Tejedor J, Cruz-Rizzolo RJ, Reinoso-Suarez F. 2000. The anatomical

connections of the macaque monkey orbitofrontal cortex. A review. Cereb. Cortex 10:220–42

Chiavaras MM, Petrides M. 2000. Orbitofrontal sulci of the human and macaque monkeybrain. J. Comp. Neurol. 422:35–54

Chiba T, Kayahara T, Nakano K. 2001. Efferent projections of infralimbic and prelimbicareas of the medial prefrontal cortex in the Japanese monkey, Macaca fuscata. Brain Res.888:83–101

Clarke HF, Dalley JW, Crofts HS, Robbins TW, Roberts AC. 2004. Cognitive inflexibilityafter prefrontal serotonin depletion. Science 304:878–80

Clarke HF, Walker SC, Dalley JW, Robbins TW, Roberts AC. 2007. Cognitive inflexibilityafter prefrontal serotonin depletion is behaviorally and neurochemically specific. Cereb.Cortex 17:18–27

Critchley HD, Mathias CJ, Dolan RJ. 2001. Neural activity in the human brain relating touncertainty and arousal during anticipation. Neuron 29:537–45

www.annualreviews.org • OFC and Decision-Making 51

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Daglish MR, Weinstein A, Malizia AL, Wilson S, Melichar JK, et al. 2001. Changes in regionalcerebral blood flow elicited by craving memories in abstinent opiate-dependent subjects.Am. J. Psychiatry 158:1680–86

Damasio AR. 1994. Descartes’ Error: Emotion, Reason, and the Human Brain. New York: Putnam.336 pp.

Deets AC, Harlow HF, Blomquist AJ. 1970. Effects of intertrial interval and Trial 1 rewardduring acquisition of an object-discrimination learning set in monkeys. J. Comp. Physiol.Psychol. 73:501–5

Dias R, Robbins TW, Roberts AC. 1996. Dissociation in prefrontal cortex of affective andattentional shifts. Nature 380:69–72

Duncan J. 2001. An adaptive coding model of neural function in prefrontal cortex. Nat. Rev.Neurosci. 2:820–29

Eslinger PJ, Damasio AR. 1985. Severe disturbance of higher cognition after bilateral frontallobe ablation: patient EVR. Neurology 35:1731–41

Everitt BJ, Robbins TW. 2005. Neural systems of reinforcement for drug addiction: fromactions to habits to compulsion. Nat. Neurosci. 8:1481–89

Fellows LK, Farah MJ. 2003. Ventromedial frontal cortex mediates affective shifting in humans:evidence from a reversal learning paradigm. Brain 126:1830–37

Fellows LK, Farah MJ. 2005. Different underlying impairments in decision-making followingventromedial and dorsolateral frontal lobe damage in humans. Cereb. Cortex 15:58–63

Fiorillo CD, Tobler PN, Schultz W. 2003. Discrete coding of reward probability and uncer-tainty by dopamine neurons. Science 299:1898–902

Funahashi S, Bruce CJ, Goldman-Rakic PS. 1989. Mnemonic coding of visual space in themonkey’s dorsolateral prefrontal cortex. J. Neurophysiol. 61:331–49

Glimcher PW. 2003. Decisions, Uncertainty, and the Brain: The Science of Neuroeconomics. Cam-bridge, MA: MIT Press. 400 pp.

Goldman-Rakic PS. 1987. Circuitry of primate prefrontal cortex and regulation of behavior byrepresentational memory. In Handbook of Physiology, The Nervous System, Higher Functionsof the Brain, ed. F Plum, pp. 373–417. Bethesda, MD: Am. Physiol. Soc.

Gottfried JA, O’Doherty J, Dolan RJ. 2003. Encoding predictive reward value in human amyg-dala and orbitofrontal cortex. Science 301:1104–7

Haber SN, Kunishio K, Mizobuchi M, Lynd-Balta E. 1995. The orbital and medial prefrontalcircuit through the primate basal ganglia. J. Neurosci. 15:4851–67

Heims HC, Critchley HD, Dolan R, Mathias CJ, Cipolotti L. 2004. Social and motivationalfunctioning is not critically dependent on feedback of autonomic responses: neuropsycho-logical evidence from patients with pure autonomic failure. Neuropsychologia 42:1979–88

Hikosaka K, Watanabe M. 2000. Delay activity of orbital and lateral prefrontal neurons of themonkey varying with different rewards. Cereb. Cortex 10:263–71

Hsu M, Bhatt M, Adolphs R, Tranel D, Camerer CF. 2005. Neural systems responding todegrees of uncertainty in human decision-making. Science 310:1680–83

Huettel SA, Stowe CJ, Gordon EM, Warner BT, Platt ML. 2006. Neural signatures of eco-nomic preferences for risk and ambiguity. Neuron 49:765–75

Hyman SE, Malenka RC. 2001. Addiction and the brain: the neurobiology of compulsion andits persistence. Nat. Rev. Neurosci. 2:695–703

Izquierdo A, Suda RK, Murray EA. 2004. Bilateral orbital prefrontal cortex lesions in rhesusmonkeys disrupt choices guided by both reward value and reward contingency. J. Neurosci.24:7540–48

James W. 1884. What is an emotion? Mind 9:188–205

52 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Kahneman D, Tversky A. 2000. Choices, Values and Frames. New York: Cambridge Univ. PressKennerley SW, Lara AH, Wallis JD. 2005. Prefrontal neurons encode an abstract representation of

value. Presented at Annu. Meet. Soc. Neurosci., Washington, DCKennerley SW, Wallis JD. 2006. Interaction of spatial working memory and reward across prefrontal

cortex. Presented at Annu. Meet. Soc. Neurosci., AtlantaKennerley SW, Walton ME, Behrens TE, Buckley MJ, Rushworth MF. 2006. Optimal decision

making and the anterior cingulate cortex. Nat. Neurosci. 9:940–47Knutson B, Taylor J, Kaufman M, Peterson R, Glover G. 2005. Distributed neural represen-

tation of expected value. J. Neurosci. 25:4806–12Kobayashi S, Lauwereyns J, Koizumi M, Sakagami M, Hikosaka O. 2002. Influence of reward

expectation on visuospatial processing in macaque lateral prefrontal cortex. J. Neurophysiol.87:1488–98

Koechlin E, Ody C, Kouneiher F. 2003. The architecture of cognitive control in the humanprefrontal cortex. Science 302:1181–85

Kuhnen CM, Knutson B. 2005. The neural basis of financial risk taking. Neuron 47:763–70Lange C. 1922. The Emotions. Baltimore: Williams & WilkinsLeon MI, Shadlen MN. 1999. Effect of expected reward magnitude on the response of neurons

in the dorsolateral prefrontal cortex of the macaque. Neuron 24:415–25Liu Z, Richmond BJ. 2000. Response differences in monkey TE and perirhinal cortex: stimulus

association related to reward schedules. J. Neurophysiol. 83:1677–92Lu MT, Preston JB, Strick PL. 1994. Interconnections between the prefrontal cortex and the

premotor areas in the frontal lobe. J. Comp. Neurol. 341:375–92Mackintosh NJ. 1974. The Psychology of Animal Learning. London: AcademicMaia TV, McClelland JL. 2004. A reexamination of the evidence for the somatic marker

hypothesis: what participants really know in the Iowa gambling task. Proc. Natl. Acad. Sci.USA 101:16075–80

Maunsell JH. 2004. Neuronal representations of cognitive state: reward or attention? TrendsCogn. Sci. 8:261–65

Miller EK, Cohen JD. 2001. An integrative theory of prefrontal cortex function. Annu. Rev.Neurosci. 24:167–202

Mishkin M. 1964. Perseveration of central sets after frontal lesions in monkeys. In The FrontalGranular Cortex and Behavior, ed. JM Warren, K Akert, pp. 219–41. New York: McGraw-Hill

Montague PR, Berns GS. 2002. Neural economics and the biological substrates of valuation.Neuron 36:265–84

Musallam S, Corneil BD, Greger B, Scherberger H, Andersen RA. 2004. Cognitive controlsignals for neural prosthetics. Science 305:258–62

O’Doherty J, Kringelbach ML, Rolls ET, Hornak J, Andrews C. 2001. Abstract reward andpunishment representations in the human orbitofrontal cortex. Nat. Neurosci. 4:95–102

Ongur D, An X, Price JL. 1998. Prefrontal cortical projections to the hypothalamus in macaquemonkeys. J. Comp. Neurol. 401:480–505

Padoa-Schioppa C, Assad JA. 2006. Neurons in the orbitofrontal cortex encode economicvalue. Nature 441:223–26

Parker A, Gaffan D. 1998. Memory after frontal/temporal disconnection in monkeys: condi-tional and nonconditional tasks, unilateral and bilateral frontal lesions. Neuropsychologia36:259–71

Pears A, Parkinson JA, Hopewell L, Everitt BJ, Roberts AC. 2003. Lesions of the orbitofrontalbut not medial prefrontal cortex disrupt conditioned reinforcement in primates. J. Neurosci.23:11189–201

www.annualreviews.org • OFC and Decision-Making 53

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Pecina S, Berridge KC. 2005. Hedonic hot spot in nucleus accumbens shell: Where do mu-opioids cause increased hedonic impact of sweetness? J. Neurosci. 25:11777–86

Petrides M. 1996. Specialized systems for the processing of mnemonic information within theprimate frontal cortex. Philos. Trans. R Soc. London B Biol. Sci. 351:1455–61

Petrides M, Pandya DN. 1994. Comparative architectonic analysis of the human and macaquefrontal cortex. In Handbook of Neuropsychology, ed. F Boller, J Grafman, pp. 17–57. NewYork: Elsevier

Pickens CL, Saddoris MP, Setlow B, Gallagher M, Holland PC, Schoenbaum G. 2003. Dif-ferent roles for orbitofrontal cortex and basolateral amygdala in a reinforcer devaluationtask. J. Neurosci. 23:11078–84

Platt ML, Glimcher PW. 1999. Neural correlates of decision variables in parietal cortex. Nature400:233–38

Rao SC, Rainer G, Miller EK. 1997. Integration of what and where in the primate prefrontalcortex. Science 276:821–24

Robinson TE, Berridge KC. 1993. The neural basis of drug craving: an incentive-sensitizationtheory of addiction. Brain Res. Brain Res. Rev. 18:247–91

Robinson TE, Berridge KC. 2003. Addiction. Annu. Rev. Psychol. 54:25–53Roesch MR, Olson CR. 2003. Impact of expected reward on neuronal activity in prefrontal

cortex, frontal and supplementary eye fields and premotor cortex. J. Neurophysiol. 90:1766–89

Roesch MR, Olson CR. 2004. Neuronal activity related to reward value and motivation inprimate frontal cortex. Science 304:307–10

Roesch MR, Olson CR. 2005. Neuronal activity in primate orbitofrontal cortex reflects thevalue of time. J. Neurophysiol. 94:2457–71

Rolls ET, Baylis LL. 1994. Gustatory, olfactory, and visual convergence within the primateorbitofrontal cortex. J. Neurosci. 14:5437–52

Rolls ET, Hornak J, Wade D, McGrath J. 1994. Emotion-related learning in patients withsocial and emotional changes associated with frontal lobe damage. J. Neurol. Neurosurg.Psychiatry 57:1518–24

Rolls ET, O’Doherty J, Kringelbach ML, Francis S, Bowtell R, McGlone F. 2003. Represen-tations of pleasant and painful touch in the human orbitofrontal and cingulate cortices.Cereb. Cortex 13:308–17

Rolls ET, Sienkiewicz ZJ, Yaxley S. 1989. Hunger modulates the responses to gustatory stimuliof single neurons in the caudolateral orbitofrontal cortex of the macaque monkey. Eur. J.Neurosci. 1:53–60

Romanski LM, Bates JF, Goldman-Rakic PS. 1999. Auditory belt and parabelt projections tothe prefrontal cortex in the rhesus monkey. J. Comp. Neurol. 403:141–57

Romo R, Brody CD, Hernandez A, Lemus L. 1999. Neuronal correlates of parametric workingmemory in the prefrontal cortex. Nature 399:470–73

Rosenkilde CE, Bauer RH, Fuster JM. 1981. Single cell activity in ventral prefrontal cortex ofbehaving monkeys. Brain Res. 209:375–94

Royet JP, Zald D, Versace R, Costes N, Lavenne F, et al. 2000. Emotional responses to pleasantand unpleasant olfactory, visual, and auditory stimuli: a positron emission tomographystudy. J. Neurosci. 20:7752–59

Rudebeck PH, Walton ME, Smyth AN, Bannerman DM, Rushworth MF. 2006. Separateneural pathways process different decision costs. Nat. Neurosci. 9:1161–68

Rule RR, Shimamura AP, Knight RT. 2002. Orbitofrontal cortex and dynamic filtering ofemotional stimuli. Cogn. Affect. Behav. Neurosci. 2:264–70

54 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Sanfey AG, Loewenstein G, McClure SM, Cohen JD. 2006. Neuroeconomics: cross-currentsin research on decision-making. Trends Cogn. Sci. 10:108–16

Schoenbaum G, Chiba AA, Gallagher M. 1998. Orbitofrontal cortex and basolateral amygdalaencode expected outcomes during learning. Nat. Neurosci. 1:155–59

Schoenbaum G, Chiba AA, Gallagher M. 1999. Neural encoding in orbitofrontal cortex andbasolateral amygdala during olfactory discrimination learning. J. Neurosci. 19:1876–84

Schultz W. 2004. Neural coding of basic reward terms of animal learning theory, game theory,microeconomics and behavioural ecology. Curr. Opin. Neurobiol. 14:139–47

Shimamura AP. 2000. The role of the prefrontal cortex in dynamic filtering. Psychobiology28:156–67

Shuler MG, Bear MF. 2006. Reward timing in the primary visual cortex. Science 311:1606–9Stephens DW, Krebs JR. 1986. Foraging Theory. Princeton, NJ: Princeton Univ. PressThorpe SJ, Rolls ET, Maddison S. 1983. The orbitofrontal cortex: neuronal activity in the

behaving monkey. Exp. Brain Res. 49:93–115Tremblay L, Schultz W. 1999. Relative reward preference in primate orbitofrontal cortex.

Nature 398:704–8Volkow ND, Fowler JS. 2000. Addiction, a disease of compulsion and drive: involvement of

the orbitofrontal cortex. Cereb. Cortex 10:318–25Wallis JD, Anderson KC, Miller EK. 2001. Single neurons in prefrontal cortex encode abstract

rules. Nature 411:953–56Wallis JD, Miller EK. 2003. Neuronal activity in primate dorsolateral and orbital prefrontal

cortex during performance of a reward preference task. Eur. J. Neurosci. 18:2069–81Walton ME, Bannerman DM, Alterescu K, Rushworth MF. 2003. Functional specialization

within medial frontal cortex of the anterior cingulate for evaluating effort-related deci-sions. J. Neurosci. 23:6475–79

Walton ME, Bannerman DM, Rushworth MF. 2002. The role of rat medial frontal cortex ineffort-based decision making. J. Neurosci. 22:10996–1003

Watanabe M. 1996. Reward expectancy in primate prefrontal neurons. Nature 382:629–32Wilson FA, Scalaidhe SP, Goldman-Rakic PS. 1993. Dissociation of object and spatial pro-

cessing domains in primate prefrontal cortex. Science 260:1955–58Wyvell CL, Berridge KC. 2000. Intra-accumbens amphetamine increases the conditioned

incentive salience of sucrose reward: enhancement of reward “wanting” without enhanced“liking” or response reinforcement. J. Neurosci. 20:8122–30

Yaxley S, Rolls ET, Sienkiewicz ZJ. 1988. The responsiveness of neurons in the insular gustatorycortex of the macaque monkey is independent of hunger. Physiol. Behav. 42:223–29

Yoshida W, Ishii S. 2006. Resolution of uncertainty in prefrontal cortex. Neuron 50:781–89

RELATED RESOURCES

Schultz W. 2007. Midbrain dopamine systems: multiple behavioral functions at different timecourses. Annu. Rev. Neurosci. 30:259–88

Shadlen MN, Gold JI. 2007. The neural basis of decision making. Annu. Rev. Neurosci. 30:535–74

Glimcher PW, Rustichini A. 2004. Neuroeconomics: the consilience of brain and decision.Science 306:447–52

Miller EK, Cohen JD. 2001. An integrative theory of prefrontal cortex function. Annu. Rev.Neurosci. 24:167–202

www.annualreviews.org • OFC and Decision-Making 55

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

ANRV314-NE30-02 ARI 7 May 2007 17:6

Schoenbaum G, Roesch MR, Stalnaker TA. 2006. Orbitofrontal cortex, decision-making anddrug addiction. Trends Neurosci. 29:116–24

Walton ME, Kennerley SW, Bannerman DM, Phillips PEM, Rushworth MFS. 2006. Weighingup the benefits of work: behavioral and neural analyses of effort-related decision making.Neural Netw. 19:1302–14

56 Wallis

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

AR314-FM ARI 20 May 2007 16:26

Annual Review ofNeuroscience

Volume 30, 2007Contents

Information Processing in the Primate Retina: Circuitry and CodingG.D. Field and E.J. Chichilnisky � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �1

Orbitofrontal Cortex and Its Contribution to Decision-MakingJonathan D. Wallis � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 31

Fundamental Components of AttentionEric I. Knudsen � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 57

Anatomical and Physiological Plasticity of Dendritic SpinesVeronica A. Alvarez and Bernardo L. Sabatini � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 79

Visual Perception and Memory: A New View of Medial TemporalLobe Function in Primates and RodentsElisabeth A. Murray, Timothy J. Bussey, and Lisa M. Saksida � � � � � � � � � � � � � � � � � � � � � � � � 99

The Medial Temporal Lobe and Recognition MemoryH. Eichenbaum, A.P. Yonelinas, and C. Ranganath � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �123

Why Is Wallerian Degeneration in the CNS So Slow?Mauricio E. Vargas and Ben A. Barres � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �153

The Head Direction Signal: Origins and Sensory-Motor IntegrationJeffrey S. Taube � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �181

Peripheral RegenerationZu-Lin Chen, Wei-Ming Yu, and Sidney Strickland � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �209

Neuron-Glial Interactions in Blood-Brain Barrier FormationSwati Banerjee and Manzoor A. Bhat � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �235

Multiple Dopamine Functions at Different Time CoursesWolfram Schultz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �259

Ventral Tegmental Area Neurons in Learned Appetitive Behavior andPositive ReinforcementHoward L. Fields, Gregory O. Hjelmstad, Elyssa B. Margolis,

and Saleem M. Nicola � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �289

v

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.

AR314-FM ARI 20 May 2007 16:26

Copper and Iron Disorders of the BrainErik Madsen and Jonathan D. Gitlin � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �317

The Micromachinery of Mechanotransduction in Hair CellsMelissa A. Vollrath, Kelvin Y. Kwan, and David P. Corey � � � � � � � � � � � � � � � � � � � � � � � � � � � � �339

Neurobiology of Feeding and Energy ExpenditureQian Gao and Tamas L. Horvath � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �367

Mechanisms that Regulate Establishment, Maintenance, andRemodeling of Dendritic FieldsJay Z. Parrish, Kazuo Emoto, Michael D. Kim, and Yuh Nung Jan � � � � � � � � � � � � � � � � �399

Dynamic Aspects of CNS Synapse FormationA. Kimberley McAllister � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �425

Adhesion Molecules in the Nervous System: Structural Insights intoFunction and DiversityLawrence Shapiro, James Love, and David R. Colman � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �451

Development of Neural Systems for ReadingBradley L. Schlaggar and Bruce D. McCandliss � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �475

Molecular Architecture of Smell and Taste in DrosophilaLeslie B. Vosshall and Reinhard F. Stocker � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �505

The Neural Basis of Decision MakingJoshua I. Gold and Michael N. Shadlen � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �535

Trinucleotide Repeat DisordersHarry T. Orr and Huda Y. Zoghbi � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �575

Indexes

Cumulative Index of Contributing Authors, Volumes 21–30 � � � � � � � � � � � � � � � � � � � � � � � �623

Cumulative Index of Chapter Titles, Volumes 21–30 � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �627

Errata

An online log of corrections to Annual Review of Neuroscience chapters (if any, 1997to the present) may be found at http://neuro.annualreviews.org/

vi Contents

Ann

u. R

ev. N

euro

sci.

2007

.30:

31-5

6. D

ownl

oade

d fr

om w

ww

.ann

ualr

evie

ws.

org

by N

ew M

exic

o St

ate

Uni

vers

ity o

n 10

/06/

11. F

or p

erso

nal u

se o

nly.