Measuring wanting and liking from animals to humans: A systematic review

Article

Reference

Measuring wanting and liking from animals to humans: A systematic review

POOL, Eva, et al.

Abstract

Animal research has shown it is possible to want a reward that is not liked once obtained.

Although these findings have elicited interest, human experiments have produced contradictory results, raising doubts about the existence of separate wanting and liking influences in human reward processing. This discrepancy could be due to inconsistences in the operationalization of these concepts. We systematically reviewed the methodologies used to assess human wanting and/or liking and found that most studies operationalized these concepts in congruency with the animal literature. Nonetheless, numerous studies operationalized wanting in similar ways to those that operationalized liking. These contradictions might be driven by a major source of confound: expected pleasantness.

Expected pleasantness underlies cognitive desires and does not correspond to animal liking, a hedonic experience, or to animal wanting, which relies on affective relevance, consisting of the perception of a cue associated with a relevant reward for the organism's current physiological state. Extending the concept of affective relevance and differentiating [...]

POOL, Eva, et al . Measuring wanting and liking from animals to humans: A systematic review.

Neuroscience and Biobehavioral Reviews , 2016, vol. 63, p. 124-142

DOI : 10.1016/j.neubiorev.2016.01.006

Available at:

http://archive-ouverte.unige.ch/unige:90359

Disclaimer: layout of this document may differ from the published version.

1 / 1

Contents lists available atScienceDirect

Neuroscience and Biobehavioral Reviews

j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / n e u b i o r e v

Review

Measuring wanting and liking from animals to humans: A systematic review

Eva Pool

, Vanessa Sennwald

, Sylvain Delplanque

, Tobias Brosch

, David Sander

aSwissCenterforAffectiveSciences,UniversityofGeneva,Switzerland

bLaboratoryfortheStudyofEmotionElicitationandExpression,DepartmentofPsychology,FPSE,UniversityofGeneva,Switzerland

cConsumerDecisionandSustainableBehaviorLaboratory,DepartmentofPsychology,FPSE,UniversityofGeneva,Switzerland

a r t i c l e i n f o

Articlehistory:

Received2October2015

Receivedinrevisedform5January2016 Accepted21January2016

Availableonline3February2016

Keywords:

Incentivesalience Wanting Liking

Affectiverelevance Pleasure

Expectedpleasantness

a b s t r a c t

Animalresearchhasshownitispossibletowantarewardthatisnotlikedonceobtained.Although thesefindingshaveelicitedinterest,humanexperimentshaveproducedcontradictoryresults,raising doubtsabouttheexistenceofseparatewantingandlikinginfluencesinhumanrewardprocessing.This discrepancycouldbeduetoinconsistencesintheoperationalizationoftheseconcepts.Wesystematically reviewedthemethodologiesusedtoassesshumanwantingand/orlikingandfoundthatmoststudies operationalizedtheseconceptsincongruencywiththeanimalliterature.Nonetheless,numerousstud- iesoperationalizedwantinginsimilarwaystothosethatoperationalizedliking.Thesecontradictions mightbedrivenbyamajorsourceofconfound:expectedpleasantness.Expectedpleasantnessunderlies cognitivedesiresanddoesnotcorrespondtoanimalliking,ahedonicexperience,ortoanimalwant- ing,whichreliesonaffectiverelevance,consistingoftheperceptionofacueassociatedwitharelevant rewardfortheorganism’scurrentphysiologicalstate.Extendingtheconceptofaffectiverelevanceand differentiatingitfromexpectedpleasantnessmightimprovemeasuresofhumanwantingandliking.

©2016ElsevierLtd.Allrightsreserved.

Contents

1. Introduction...125

1.1. Theincentivesaliencehypothesis...125

1.2. Keyelementsofthewantingandlikingmeasuresinanimals...126

1.3. Wantingandlikinginhumans:successandcontroversy...126

1.4. Thepresentreview ... 127

2. Method...127

2.1. Inclusioncriteria...127

2.2. Literaturesearchstrategy ... 127

2.3. Dataextraction...127

3. Results...128

3.1. Population ... 129

3.2. Typesofstudies...129

3.3. Objectofthemeasure ... 130

3.4. Measures...130

3.5. Reward,cueandtiming...131

4. Discussion...132

4.1. Expectedpleasantnessasamajorconfound...133

∗Correspondingauthorat:CaliforniaInstituteofTechnology,1200EastCaliforniaBoulevard,MC228-77,Pasadena,CA91125,UnitedStates.

E-mailaddress:[email protected](E.Pool).

http://dx.doi.org/10.1016/j.neubiorev.2016.01.006 0149-7634/©2016ElsevierLtd.Allrightsreserved.

4.2. Differentiatingexpectedpleasantnessfromaffectiverelevanceasasolution...133

5. Conclusion...135

Acknowledgments...135

AppendixA. :Summaryofthestudiesincludedinthepresentsystematicreview...135

References...140

1. Introduction

Psychologistsandneuroscientistshavelongtriedtounderstand howindividualsdecidetoinvesttheirlimitedresourcestopursue aparticularrewardingoutcome(e.g.,Delgado,2007;Rescorlaand Solomon,1967;Spence,1956).Commonsensesuggeststhatpeople decidetoinvesttheirresourcestopursuetheoutcometheylikethe most.However,inmanysituations,individualsinvestaconsider- ableamountofefforttopursueanoutcomeeventhoughafterthey obtainit,theydonotexperienceitaspleasurable.Aclearexam- pleoccursinthecaseofdrugaddiction,inwhichindividualsare willingtogotoextraordinarylengthstoobtainasubstancethat willeventuallyelicitnopleasurablefeelingsduringitsconsump- tion(RobinsonandBerridge,2003).Inthenineties,Berridgeand Robinson,1998proposedtheincentivesaliencehypothesis that challengedthehedonicperspective.Proponentsofthishypothesis suggestedthatthepursuitofanoutcomeisnotalwaysdirectlypro- portionaltothepleasureexperiencedduringconsumption,because rewardprocessingisaprocessinvolvingmultipledistinctparal- lelcomponents,includingthemotivationtoobtainareward(i.e., wanting)andthehedonicpleasurefeltduringitsconsumption(i.e., liking;seealsoBerridge,2009b).Thesecomponentsaretypically positivelycorrelatedbutcanalsobedissociated,therebymaking organismsworkforarewardthattheywillnotappreciateonce obtained.

Thisproposal,basedonananimal model,hasgarneredgreat interestamongresearchersinvestigating motivationalprocesses inhumans(e.g.,Finlaysonetal.,2007b;Kringelbachetal.,2012;

Mela,2006;Nawijnetal.,2015).Severalscholarshaveconsidered theindependenceofwantingandlikingasapotentialmechanism underlyingavarietyofhumanbehaviorsthatnegativelyimpact well-beingsuchasovereating,pathologicalgamblingandthecon- sumptionof addictivesubstances (Finlaysonet al.,2007b; Pool etal.,2015c;Tibboeletal.,2011;Wölflingetal.,2011).However, humanexperimentshaveledtocontradictoryresults,openinga debateontheexistenceoftwodissociablecomponentsinhuman reward processing (Havermans, 2011, 2012). It has even been claimedthatacorrectoperationaldefinitionofwantingandliking asconceivedbyBerridgeandRobinson,1998islackinginhuman research(Havermans,2012).Clearoperationaldefinitionsarepar- ticularlyimportantforstudiesconductedonhumansbecausesuch studiesoffermuchlargervariabilityinoperationalizingpsychologi- calconstructs–whichcanbeanimportantsourceofconfound–than animal studies.In thepresentarticle, wesystematically review theliteraturethatinvestigateswantingandlikingamonghuman populations, aswell assystematically describe how these con- ceptswereoperationalizedregardingtheimportanttenetsofthe incentivesaliencehypothesis.Wetherebyaimedto(1)systematize andquantify,acrossallkindsofhumanrewards,thecontradic- toryoperationalizationsofwantingand/orlikingthathavebeen previouslyhighlighted(Havermans,2011,2012)and (2) identify possiblesourcesofconfoundsthat mightberesponsibleforthe contradictoryresults.

1.1. Theincentivesaliencehypothesis

The incentive saliencehypothesis has beenconceived asan extensionofearlymodelsofincentivemotivation(Bindra,1974;

Bolles,1972;Spence,1956;Toates,1998).Thesemodelschallenged thedrivereductiontheorythataccountedformotivatedbehaviors exclusivelyintermsoftheneedtoreduceaparticularimbalanced physiologicalstate,suchashunger(i.e.,drives),inordertoreestab- lishhomeostasis.Spence(1956)wasthefirsttoproposethatthe amountof energyinvestedin anaction(e.g.,walkingtoward a restaurant)canbeinfluencedbytheperceptionofexternalstimuli (e.g.,therestaurantlogo)thathavebeenassociatedwithareward throughtheorganism’sexperiences.Subsequently,otherincentive motivationtheorists(Bindra,1974;Bolles,1972;Toates,1998)sug- gestedthatthemotivationalincreasesubsequenttotheperception ofthereward-associatedcueis proportionaltotheexperienced hedonicpleasure,which istriggeredbytheconsumption ofthe reward:themorepleasurablethereward,thebiggertheincrease inmotivationtriggeredbytherewardcue.Therefore,accordingto thissuggestion,incentivesshouldinfluencetheorganism’smoti- vationinalogicalway:theamountofeffortmobilizedtoobtainthe rewardisalwaysjustifiedbythehedonicexperienceduringreward consumption.Forseveralyears,thisintrinsicrelationshipbetween motivationand hedonicpleasurehasbeensodeeplyintegrated in affective neuroscience that the amountof hedonic pleasure foraparticularrewardhasbeenmeasuredinmultiplestudiesas theamountofeffortmobilizedtoobtainit.Thisoperationaliza- tionhasbeenusedinresearchconductedonanimalsinparticular because theycannot verbally report the hedonicpleasure that theyexperienced(seeBindra,1974;Bolles,1972;Toates,1998).

Mostresearchersassumedthatifanorganismworkstoobtaina reward,itmustmeanthatitlikesit.Inthenineties,Berridgeand co-workerschallengedthishedonicperspectiveofincentivemoti- vationaltheoriesthroughacorpusofexperimentsconductedon rodents(BerridgeandRobinson,1998;MahlerandBerridge,2012;

Pecinaetal.,2003;WyvellandBerridge,2000,2001).Theydemon- stratedthatitispossibletomakearodentworktoobtainareward thatitdoesnotlike.Themostinnovativeaspectofthisseriesof experimentswastheuseoftwodifferentmeasuresforincentive motivationandhedonicpleasure:theformerwasmeasuredina classicway(e.g.,theincreaseinmobilizedeffortafterthepercep- tionofarewardingcue),and,critically,thelatterwasmeasuredby adistinctdependentvariableconsistingofprototypicalorofacial expressionsduringrewardconsumption.Theseorofacialexpres- sionsareelicitedbytheconsumptionofpleasant(e.g.,sweettaste) orunpleasant(e.g.,bittertaste)foodandseemtobereliableindexes ofhedonicexperiencesinseveralorganisms(e.g.,rats,apes,mon- keys,humanbabies;seeBerridge,2000).Throughthesemeasures, Berridgeand co-workers showedthat two different dissociable neuronalnetworksunderliehedonicpleasureandincentivemoti- vationinrodents(Berridge,2000;PecinaandBerridge,2000,2005;

Pecinaetal.,2003;WyvellandBerridge,2000,2001).Animpor- tant demonstration in this work is that increasing thelevel of dopamineinthemesolimbicregionincreasestheamountofeffort mobilizedtoobtainarewardwithoutsimultaneouslymodifying themeasureofhedonicpleasureexperiencedduringitsconsump- tion.Fromtheseempiricalfindings,theinvestigatorsformulated

the incentive salience hypothesis, which postulates that reward processinginvolvesmultiple components,includingone that is motivational(wanting)andanotherthatishedonic(liking),which relyonseparateneuralnetworksthatcanbedissociatedunderpar- ticularcircumstances(Berridgeand Kringelbach,2015;Berridge andRobinson,2003).Theinteractionbetweentheorganism’sbrain state(e.g.,increasedlevelofmesolimbicdopamine)orphysiolog- icalstate(e.g.,hunger/satiety) and theelementspresent inthe environment(e.g.,reward-associatedcue)isanimportanttenetof theincentivesaliencehypothesis.Indeed,computationsofwanting dynamicallyincorporatethecurrentphysiologicalstate,reflecting therealinternalstateoftheorganismataparticulartimewith respecttoanidealsetpointthatregulateshomeostasis(alsocalled kfactor;Zhangetal.,2009).Examplesofsuchphysiologicalstates couldbesatiation,hunger,andthirst,aswellasdrugeffectsorstress (BerridgeandO’Doherty,2014).Theabilityofareward-associated stimulustotriggeramotivationalstateisstronglymodulatedby therelevanceoftherewardforthephysiologicalstateoftheindi- vidual(RobinsonandBerridge,2013;Zhangetal.,2009).Insome cases,theorganism’sstatecanincreasebothcue-triggeredwanting andthelikingexperienceduringrewardconsumption;forinstance, hungerincreasestherelevanceofafoodrewardthatbecomesboth morewantedandliked(Havermansetal.,2009).Inothercases,the organism’sstatecanselectivelyincreasewantingwithoutmodify- ingliking;forinstance,stressprioritizesrewardrelevance(Leyton, 2010),increasingcue-triggeredwantingforaparticularrewardbut notlikingduringrewardconsumption(seePooletal.,2015cfora review).

BerridgeandRobinson(2003)proposedthatwantingandliking canbefurtherclassifieddependingonwhethertheyareprocessed atanimplicit orexplicitlevel (seealsoAnselmeand Robinson, 2015).Explicitandimplicitlikingbothrefertothehedonicimpact oftherewardduringitsconsumptionandsimplydifferintermsof explicitness/implicitness;however,explicitandimplicitwanting relyondifferentpsychologicalmechanisms.Implicitwanting,also calledincentivesalience,reliesonaPavloviansystemandrefersto cuetriggeredmotivationalreactionsthatcanoccurwithoutacon- sciousexperience.Explicitwanting,alsocalledcognitivedesires, reliesona goal-directed systemand ofteninvolvesthesubjec- tivefeelingofbeingattractedtowardadesiredobject.Moreover, cognitivedesiresrelyonexpectationsindividualshaveaboutthe pleasantnessofthereward,whicharebuiltbasedonpastliking experiences.Therefore,cognitivedesiresarenotcompletelyinde- pendentfromliking,whereasimplicitwantingorincentivesalience ispotentiallyindependentfromanyhedonicaspectofthereward includingexpected pleasantness (Berridge and Aldridge, 2008).

Pleasenotethatinthecontextoftheincentivesaliencehypothesis andtheaforementionedanimalliterature,thetermwantingrefers toimplicitwantingorincentivesalience.Therefore,inthepresent articlethetermwantingisusedtorefertoincentivesalienceor implicitwanting.

1.2. Keyelementsofthewantingandlikingmeasuresinanimals

Asanextensionofincentivemotivationtheories,theincentive salience hypothesis considers three keys elements when mea- suring wanting and liking. The first is the rewarding outcome (alsoreferredtoastheunconditionedstimulus),thesecondisthe reward-associatedcue(alsoreferredtoastheconditionedstim- ulus),and thethird is thephysiological stateof theindividual.

Wantingand likingdependondifferentinteractions ofsomeof theseelementsatspecificmomentsintime.Wantingistriggered bytheinteractionbetweenanindividualinaparticularstateand theperceptionofarewardcueandcanbemeasuredbytheeffort mobilizedin the instrumental action. Notice that theincentive (i.e.,thereward-associatedcueorthereward)isthuspresented

beforetheinstrumentalaction.Timingisparticularlyimportant;

indeed,iftheincentiveispresentedaftertheinstrumentalaction, theprocessnolongerrelatestoincentivemotivationbutratherto reinforcementlearning.Inaddition,thespecificinfluenceofwant- ingis strongerbeforereward consumption,since duringreward consumption,thehedonicexperienceisdominant.Similarly,liking istriggeredbytheinteractionbetweenanindividualinaparticular stateandtheconsumptionofarewardthatismeasuredthroughthe hedonicreactionduringorimmediatelyafterrewardconsumption.

Hereagain,timingiscriticalbecauselikingisconceivedasahedo- nicexperience;ifthemeasurementisnottakenclosetoreward consumption,itwillreflecttheencodedmemoryofthehedonic experienceratherthanthehedonicexperienceitself.Thismightbe particularlyproblematicbecausememoriesofpasthedonicexpe- riencesare usedtobuildexpectedpleasantness (Balleine,2005).

Expectedpleasantness,whichconsistsofpredictionandexpecta- tionsabouthowpleasantorunpleasantsomethingisgoingtobe, representsthemechanismunderlyingcognitivedesiresthatdonot correspondtoeitheranimallikingoranimalwanting,butrathera distinctmotivationalcontrolsystem(i.e.,goal-directedsystem)of reward-seekingbehaviors(BerridgeandAldridge,2008;Berridge andO’Doherty,2014;DickinsonandBalleine,1994;Wassumetal., 2011b).

Given the importance of the rewards and the reward cues formeasuresoftheincentivesaliencehypothesisinanimals,we decidedtosystematicallydescribedifferentaspectsoftherewards or reward cues (e.g.,kind of reward, format) presented in the methodological proceduresthat measure wanting and liking in humans.

1.3. Wantingandlikinginhumans:successandcontroversy

The incentive salience hypothesis, which has been formu- latedonthebasis of an animal model(Berridgeand Robinson, 1998,2003), hasgarneredgreat interestamongresearcherswho areinvestigatingmotivationalprocessesinhumans.Severallines ofresearchhavesubsequentlybeenlaunchedtoinvestigatethe effectofdopaminederegulationonmotivationandhedonicplea- sureforaparticularreward(e.g.,Braueretal.,2001;Evansetal., 2006;Volkowetal.,1997),theroleofwantingandlikinginaddic- tivebehaviors(e.g., Goldsteinet al.,2010; Tibboel et al.,2011;

Wachtel et al., 2002), or the role of these two components in thenormalprocessingofrewardsrelatedtodifferentneeds,such asoffspringcaretaking(withbabies)ornourishment(withfood) (e.g.,Finlaysonetal.,2007a;Parsonsetal.,2011).Thiscorpusof experimentsprovidedevidencesupportingtheideathatthesame processesfoundinrodentscouldpotentiallyexistinhumans.More particularly,inclinicaldisordersinvolvingdopaminederegulation (Evansetal.,2006;Volkowetal.,1997),ithasbeenarguedthat thelevelofmesolimbicdopamineinfluencesthemotivationalpro- cesseswithoutnecessarily modifyingthehedonicexperienceof rewardconsumption.Moreover,ithasbeensuggestedthatdiffer- entbrainregionsareactivatedbyamotivationalstatesuchasthe expectationof areward (e.g.,amygdala;O’Doherty etal.,2002;

Smalletal.,2008)andbyahedonicstatesuchastheconsump- tionofareward(e.g.,orbitofrontalcortex;O’Dohertyetal.,2002;

Smalletal.,2008).Althoughtheaforementionedfindingssupport theexistenceoftwodistinctcomponentsinhumanrewardpro- cessingthatreflect wantingand liking, otherexperimentshave provided contradictoryevidence againsta dissociation between wantingandliking inhumans(Havermans,2011,2012;Tibboel etal.,2011).Specifically,Havermans(2011,2012)highlightedthat in studies investigating wanting and liking for food reward in humans,construct operationalizationsare farfrom theoriginal incentivesaliencehypothesisandoftencontradicteachother:in somecases,asimilaroperationalizationisusedtomeasurewanting

inonestudyandlikinginanother(Finlaysonetal.,2007a;Lemmens etal.,2009).Hearguedthat,inresearchconductedonhumans, measuresofwantingandlikingstillneedtobevalidatedbydisso- ciatingthemunderprecisecircumstancesthatareclearlypredicted bytheincentivesaliencehypothesis.Intheabsenceofsuchaval- idation,thedifferentialcontributionsofwantingandlikingfound in studiesconductedonhumans arelikely toreflect poor con- structvalidityratherthanrealeffects.Healsoproposedabandoning thedistinctionbetweenwantingandlikingintheinvestigationof foodrewardinhumans,claimingthatwantingandlikingareso intrinsicallyrelatedthattheycannotbeconsideredastwodistinct componentshavingseparateinfluences.

1.4. Thepresentreview

Severalresearchersagreethattheincentivesaliencehypoth- esis has important explanatory powerin theunderstanding of varioushumanbehaviors,inparticularproblematicbehaviorsuch asovereating,addictiveconsumptionofsubstancesorpathologi- calgambling(Finlaysonetal.,2007b;Goldsteinetal.,2010;Pool etal.,2015c;Wölflingetal.,2011).Nonetheless,resultsofstudies investigatingtheincentivesaliencehypothesiswithfoodreward inhumansledtoskepticalconclusions beingdrawn concerning theexistenceofwanting andlikingastwodistinctcomponents withseparateinfluences(Havermans,2011,2012).Thesecriticisms raisedtwoimportantaspectsthatseemtobeproblematicforthe hypothesis:(1)operationalizationsofwantingandlikingareoften farfromtheoriginalincentivesaliencehypothesisand(2) mea- suresofwantingandlikingareinconsistentacrossstudiesandoften contradictoneanother.

Here,wesystematicallyreviewstudiesonwantingandliking forallrewardsinhumanstoestimatetheextenttowhichthese aspectsrepresentaproblem.Wesystematicallydescribe(1)how wantingandlikingweremeasuredacrossthesestudiesand(2)how themethodologicalproceduresintegratedthekeyelements(i.e., rewardcue,rewardconsumptionandtheirrespectivetiming)ofthe incentivesaliencehypothesis.Fromtheresultsofthissystematic descriptionof theexisting literature,we argue that themajor- ityofstudiesseemtocorrectlyintegratethemaintenets ofthe incentivesaliencehypothesis;however,numerousstudiesopera- tionalizetheconceptsofwantingandlikingincontradictoryways.

We claimthat thesecontradictoryoperationalizationsareoften derivedfromconfusionovertheconceptofexpectedpleasantness thatis sometimesconsideredliking, but atothertimes consid- eredwanting.Finally,wesuggestthatclarifyingthedistinctions between(1)expectedpleasantnessandaffectiverelevanceand(2) experienceandmemorycouldimprovetheconceptualclarityof themechanismsinvolvedinwantingandliking,therebyreducing sourcesofconfusionwhenoperationalizingtheseconstructs.

2. Method

2.1. Inclusioncriteria

Toselectthestudiesincludedinthissystematicreview,weused thefollowingcriteria:

1.Thearticlehadtobepublishedinapeer-reviewedjournaland writteninEnglish

2.Thearticlehadtoreportoriginaldatacollectedfromahuman populationbetweenJanuary1990andApril2015.

3.Thestudyhadtohavemeasuredatleastoneoftheconstructsof interest(i.e.,“incentivesalience”,“wanting”,“incentivemotiva- tion”,“liking”,“hedonicpleasure”)withanexplicitreferenceto

theincentivesaliencetheoreticalframework(e.g.,Berridgeand Robinson,1998,2003;RobinsonandBerridge,2003).

2.2. Literaturesearchstrategy

The first potential studies were identified by searching the electronicProQuestandPubMed databases.Wesearchedfor all availablerecordsstartingfromJanuary1990untilApril2015,using thefollowingcombinationofkeywordsinthetitleorabstractof thearticle:(wantingOR“incentivemotivation”)AND(likingORplea- sure)OR“incentivesalience”.Thissearchyielded545hits.Afterthe removalofduplicatesandclearlyoff-topic articles(e.g.,geogra- phyorcityplanning);weobtainedaninitialpoolof378articles.

Toensurethattheymettheinclusioncriteria,theinitialpoolwas winnowedthroughafive-stepprocess(seeFig.1).Duringthefirst foursteps,onlytheabstractsofthearticleswereread.Atthelast step,thearticleswerereadinfull.Iftherewasadoubtatanystep, thearticlewaskeptforfurtherinspection.Step1wasdesignedto includeonlythosearticlesreportingoriginalexperimentaldata;at thisstage,allreviewsandmeta-analyseswereexcluded.Intotal, 268articlessurvivedStep1.AtStep2,articleswereincludedonly iftheywereconductedonahumanpopulation;143articlessur- vived.Step3wasdesignedtoexcludeallarticlesthatexclusively measuredperceptualprocessingorattentionalorientingtoward therewardingstimulus.Thisparticularprediction oftheincen- tivesaliencehypothesishasbeenfullyreviewedelsewhere(Pool et al.,2015a).Here, weaimedtoreview studiesmeasuringthe motivationalandhedonic,ratherthanattentional,correlatesofthe incentivesaliencepredictions.Intotal,126articlessurvivedStep3.

AtStep4,articlesthatdidnotuserewardingstimuliwereexcluded and125articlessurvived.Atthelaststep,theremainingarticles werereadinfullbyonerater(graduatelevelandauthorofthisarti- cle).Thisstepwasdesignedtoincludeonlythosearticlesthataimed tomeasureatleastoneoftheconstructsofinterest(i.e.,“incen- tivesalience”,“wanting”,“incentivemotivation”,“liking”,“hedonic pleasure”),withanexplicitreferencetotheincentivesaliencethe- oreticalframework(e.g.,BerridgeandRobinson,1998).Tworaters (bothgraduatelevelandauthorsofthisarticle)firstreadthesame 20%oftheabstractsofthearticles.Theoverallagreementwasvery high(Cohen’sk=.92),disagreementswerediscussedanda con- sensualsolutionwasused.Onlyonerater(graduatelevelandfirst authorofthisarticle)readtheremainingarticlesinfull.Foreachof the51articlesthatsurvivedStep5wedidanelectronicsearchin theGoogleScholardatabasetofindoutwhetherthefirstauthors publishedotherrelevantarticles.Weobtained33otherrelevant hits,leadingtoafinaldatabaseof84articles.

2.3. Dataextraction

For each of the selected articles, we summarized different aspectsofthestudy(seeAppendixAandTable1foranoverview).

First,wecharacterizedthetypeofstudybythemeasureused (e.g., behavioral, questionnaires, electroencephalography [EEG], functionalmagneticresonanceimaging[fMRI],positronemission tomography[PET]).Studiesthatusedphysiologicalmeasures(e.g., acousticstartle)ormanipulationofphysiologicalfactors(e.g.,food, alcoholordrugadministration)weredescribedasphysiological.

Second,wespecifiedthetypeofpopulationthatthestudyinves- tigated:forinstance,thestudyauthorsmayhavebeeninterested inapopulationofhealthyindividuals,orofindividualswithprob- lematicfoodconsumptionorofthosewhoreportedaproblematic useofalcohol.

Third,wecharacterizedwhichparticularrewardwastheobject ofthemeasureofwantingandliking;forinstance,wedescribed whetherthemeasureofwantingquantifiedwantingforfood,for alcohol,forapleasantphotographorforapleasantodor.

Table1

Summaryofthemethodologicalaspectsdescribedinthepresentreview.

Aspect Variable Examples

Population Populationtargetedinthestudy Healthy,problematicuseofsubstances,problematicfoodconsumption

Type Typeofstudy Behavioral,fMRI,physiological

Method Objectofthemeasure Food,nicotine,alcohol

Rewardformat Photo,odor,sample

Cueformat Photo,symbol,sightofasample

Measure Self-reports,implicitassociations,effortmobilized

Timingofthemeasure Aftercueexposure,duringcueexposure,afterrewardconsumption Note.fMRI:functionalmagneticresonanceimaging.

Fig.1.Flowchartillustratingthesearchandwinnowingprocesses.

Fourth,wespecifiedtheformat(e.g.,asample,anodor,apho- tograph)inwhichtherewardwaspresentedtotheparticipants.

Thesamewasdonefortheformatofthereward-associatedcue whenthemethodologicalprocedureinvolvedanexposuretoany cuesassociatedwiththerewardthatwastheobjectofthewanting orlikingmeasures.Thecue-rewardassociationscouldhavebeen learnedinthelaboratory(e.g.,throughassociativelearningproce- dures)oroutsidethelaboratory(e.g.,picturesoffoodassociated withfoodthrougheverydaylifeexperiences).

Fifth, we described how wanting and liking weremeasured intheexperimentalprocedure.Forinstance,somestudiesasked participantstoreporttheirlevelofwantingorlikingbyusingquan- titativescales(e.g.,visualanaloguescales,Likertscales),whereas othersmeasuredtheparticipants’brainactivityduringamotiva- tionalorahedonicstate.

Finally,wedescribedwhenthismeasurewastakenduringthe procedure. This description was based on the reward and cue

presentationandcouldbecodedasbefore,duringorafterthecue orrewardconsumption.Incasesinwhichtherewardwasadmin- isteredduringthemotivationaltask(e.g.,Epsteinetal.,2011),we consideredthemeasuretobetakenduringconsumption.Incases wherewantingand/orlikingforaparticularrewardwasmeasured whileperceivingtherewardcue(e.g.,aphotographinMcNeiletal., 2015a,b),weconsideredthemeasuretobetakenduringthecue perception.

Weassessedinterratervariabilitybycomparingthedescriptions oftworaters(bothgraduatelevelandauthorsofthisarticle)for20%

ofthejournalarticlesincludedinthesystematicreview.Cohen’sk variedbetween.76and1acrossthedifferentvariables,withamean of.93.Thedisagreementswerediscussedandaconsensualsolution wasusedforthefinaldescription.

Fig.2. Frequency(inpercentage)ofthedifferenttypesofmeasuresusedtoassess(a)wantingand(b)liking.EMA:evaluativemovementtask;exp.:expected;facialexpr.:

facialexpression;IA:implicitassociation;PAR:postauricularreflex;QUEST:questionnaire;QQ:quantitativequestion;will.:willingness.

3. Results 3.1. Population

Themajorityoftheselectedstudies(55.55%)investigatedwant- ing and/or liking in healthy humans. However, the interest in these constructs as potentialmechanisms underlying problem- aticbehaviorswasevident:alargeproportion(25.55%)ofhuman studiesinvestigatedwantingand/orlikinginpopulationsreport- ingproblematicconsumptionofsubstancessuchasdrugs,alcohol andnicotine;animportantproportion(11.11%)targetedapopula- tionreportingproblematicconsumptionoffood,mostlyrelatedto excessivefoodconsumption(e.g.,overeating,bulimia,bingeeat- ing);andasmallerproportionofrecentstudies(3.33%)extended thisinvestigationtobehavioraladdictionsuchasexcessivevideo gameplayingorgambling.Finally,a smallsetofstudies(4.44%) triedtomeasurewanting and/orlikinginpopulationsreporting otherdisorderssuchasschizophreniaanddepression(seeTable2).

3.2. Typesofstudies

Physiologicalstudies(e.g.,mobilizedeffort,electromyography, foodor drugadministration)representedthelargestproportion (53.57%)ofstudiesinvestigatinghumanwantingand/orliking.The interestinphysiologicalmanipulationiscongruentwiththeincen- tivesaliencehypothesis,accordingtowhichthephysiologicalstate oftheindividualrepresentsacriticalfactorindeterminingboth

Table2

Frequency(inpercentage)ofthepopulation,typeofstudyandobjectofstudies investigatinghumanwantingand/orliking.

Variable Descriptor Percentage

Population Healthy 55.55

Problematicuseofsubstance 25.55 Problematicfoodconsumption 11.11

Otherdisorders 4.44

Behavioraladdiction 3.33

Typeofstudy Physiological 53.57

Neurobiological 30.95

Behavioral 10.71

Questionnaire 4.76

Object Food 52.79

Dependencesubstances 17.25

Erotic/attractive 8.12

Money 7.61

Multiple 6.59

Odor 4.06

Activity/behavior 2.53

Touch 1.01

wanting and liking (Berridge andRobinson, 1998).Because the incentivesaliencehypothesiswasconceivedinneuroscience,itis notsurprisingthat neurobiologicalstudies(e.g.,fMRI, PET,EEG, brainlesions)alsorepresentedalargeproportionoftheselected studies (30.95%). Behavioral (10.71%) and survey/questionnaire (4.76%)studieswerelessfrequent(seeTable2).

Fig.3.Frequency(inpercentage)ofwanting(a)andliking(b)measurestakeninmethodologicalproceduresthatpresentedparticipantswithacue,arewardorneither ofthesetwoelements.Frequency(inpercentage)ofwanting(c)andliking(d)measurestakenduringorafterthecuepresentation,orbefore,duringorafterthereward consumptionorreceipt.cons.:consumption/receipt.

3.3. Objectofthemeasure

Mostof themethodological proceduresintheselectedstud- iesmeasuredwantingand/orlikingforfoodreward(52.79%).This mightberelatedtothehistoryoftheincentivesaliencehypoth- esis,whichwasinitiallydevelopedforanimalstudiesusingfood rewards(Berridge,2000;BerridgeandRobinson,1998).Moreover, foodhastwomainadvantageswithrespecttotheincentivesalience hypothesis:first, beinga primary reward, it canbe consumed, thereby triggering a hedonic experience that can bemeasured andreported;second,therelevantphysiologicalstate(i.e.,hunger) playsacriticalroleintheincentivesaliencehypothesis,whichcan easilybemanipulated.Studiesmeasuringwanting and/orliking forpotentiallyaddictivesubstances(e.g.,cocaine,alcohol,nicotine) wererelativelyfrequent(17.25%),againhighlightingtheinterestin usingtheseconceptstoexplaindysfunctionalbehaviorsinhumans suchassubstanceaddiction.Lessfrequentwerestudiesmeasuring wantingand/orlikingformoney(7.61%),erotic/attractivestimuli (8.12%),othertypesofprimaryrewardsuchaspleasanttouches (1.01%)orpleasantactivities(e.g.,videogaming,physicalactivity;

2.53%),ormultipletypesofrewards(6.59%;seeTable2).

3.4. Measures

The present systematic review has highlighted how human wantingandlikinghavebeenstudiedthroughalargevarietyof measures,whichcanbedescribedashavingadoptedfourdifferent strategies.

Themostwidespreadstrategyreliedontheparticipants’knowl- edgeof motivational andhedonicterms (e.g.,wanting, craving, liking,appreciating). Mostofthemeasures(37.03%for wanting

studiesand49.43%forlikingstudies;seeFig.2)consistedofasingle quantitativequestionaskingtheparticipantsabouttheirfeelings.

Theseratingscaleshavebeenlargelyusedtomeasureaffective experiencessuchaspleasureandpain,andhaveshownhighvalid- ity(e.g.,Bartoshuk,2014).Severalauthors(seeAppendixA)have adaptedthismeasurebyaskingparticipantstoreport,througha varietyofdifferentterms,theirmotivationalfeelingsforareward (i.e.,wanting,craving,desire,urgetoconsume,desiretoconsume more)andtheirhedonicfeelingfora reward(i.e.,liking,pleas- antness,appreciation,positivefeelings).Aparticularcaseofthese ratingscalemeasuresisaquestionon“expectedpleasantness”that hasbeenusedtomeasurewantinginsomemethodologicalpro- cedures(1%;seeFig.2a).Othermethodologicalprocedureshave useditasameasureofliking(12.35%;seeFig.2b).Othermeasures (8.33%forwanting;4.49%forliking)consistedofquestionnaires orquestionnairesubscalesthattargetedmotivational(e.g.,crav- ing)orhedonicfeelings(e.g.,rememberedorimaginedliking).In somecases,ratingscalesandquestionnairesusedtomeasureliking targetedprocessesthatareoftenconsideredmotivational,suchas excitement(1.12%)orarousalandattractiveness(3.37%).Asmall proportionofmeasures(2.27%forwanting;3.37%forliking)aimed todevelop animplicit indexbyadapting a tasklargely usedin psychology:theimplicitassociationtask(Greenwaldetal.,1998).

Intheclassicversionofthetask, participantsareaskedtoclas- sifywordsintofourcategories:tworepresentingtargetconcepts (e.g.,peaceandwar)andtworepresentingattributes(e.g.,positive andnegative).Intheassociationcompatibleblocks, participants areaskedtopressonabutton(e.g.,therightarrowkey)forone conceptanditscongruentattribute(e.g.,peace/positive)and on adifferentbutton(e.g.,theleftarrowkey)fortheotherconcept anditscongruentattribute(e.g.,war/negative).Intheassociation

incompatibleblocks,participantsareaskedtopressonabutton foraconceptanditsincongruentattribute(e.g.,peace/negative) andonadifferentbuttonfortheotherconceptanditsincongru- entattribute(e.g.,war/positive).Participantsrespondfasterinthe associationcompatibleblocksthanintheassociationincompatible blocks.Thedifferenceinreactiontimesbetweenthetwoblocksis thoughttoreflectthestrengthoftheassociationbetweenatarget categoryanditscompatibleattribute.Thistaskhasbeenadapted tomeasurethestrengthofanimplicitassociationbetweentherep- resentationofaparticularrewardandtheconceptsofwantingand likingbyusingtheattributes“Ilike”,“Idonotlike”and“Iwant”,

“Idonotwant”(seeTibboeletal.,2015bforadetailedreview).

Althoughthistaskprovidesanimplicitmeasure,intermsofunder- lyingmechanismsitstillrequireshigh-levelprocessingsuchasthe semanticrepresentationoftheconceptsofwantingandliking.

Thesecondstrategyconsistedoftryingtoadaptmeasuresfrom theanimalliterature.Thesestudiesmeasuredwantingbytheeffort mobilized (11.11%), thewillingness to work or to pay (4.62%), otherindexesmixingperformancesandwillingnesstoconsume thereward(5.55%),ortheamountofrewardconsumed(1%;see Fig.2a).Researcherswhoadoptedthisstrategyusedmethodologi- calproceduresthatmeasuredlikingthroughelectromyographyof thefacialmuscles.Eventhoughthismeasurecorrespondsmostto theorofacialexpressionsusedtoassessanimalliking,ithasbeen relativelylittleused(3.37%;seeFig.2b).Thismightbeduetothe difficultyinfindingaclearindicatorofhedonicpleasureinhuman facialexpressions,whichseemsmoresuitableformeasuringaver- siveexperiencessuchasdisliking;inparticular,theactivityofthe corrugatorsuperciliiseemstoreflectdislikingexperiences(Booth etal.,2010;Horio,2003;Huetal.,1999).Otherexperimentalwork triedtodevelop anotherindexofimplicit liking:theevaluative movementtask(1.12%seeFig.2b).Inthistask,participantswere askedtopressonakeyboardtomoveaphotographofareward towardor awayfromtheirfirst name,whichwasdisplayed on acomputerscreen.Pressingthekeytomovetherewardtoward theirfirstnameistakenasanimplicitindexofliking.However, thismeasurehasastrongmotivationalcomponent(i.e.,approach, avoidance)thatraisesdoubtsofitsvalidityasapurehedonicindex.

Thethirdstrategyusedtotesttheincentivesaliencehypothesis onhumansistoinduceamotivationaland/orahedonicstateand tomeasurethecorrespondingneuralcorrelates(15.5%forwanting;

6.74%forliking;seeFig.2).Beyondstudiesselectedinthepresent systematicreview,thisstrategyhassuccessfullybeenusedtoinves- tigatebraincorrelatesofhumanhedonicpleasure(DeAraujoetal., 2003;Kringelbachetal.,2003)andhumanincentivemotivation (Prevostetal.,2012;Talmietal.,2008).Inthesestudies,behav- ioralperformancesandsubjectiveratingsoflikingweremeasured throughout theneuroimaging experiments and correlated with changesinbrainactivity.

The final strategy consisted of assessing participants’ pref- erencesfor a particular reward over other rewards.Preference measureshavebeenequallyusedtoreflectwanting(12.03%;see Fig.2a)insomestudiesandliking(12.35%;seeFig.2b)inothers.

Severalpreferenceindexeswereusedacrossstudies:somestudies explicitlyaskedparticipantstoreporthowmuchtheypreferreda particularrewardingeneral(e.g.,Bornetal.,2011),whereasothers measuredrelativepreferencesbypresentingparticipantswithpos- siblecombinationsofdifferenttypesofrewardsandaskingthemto rapidlyindicatewhichonetheylikedthemost(e.g.,Lemmensetal., 2009).Finlaysonetal.(2007a)alsodevelopedimplicitpreference indexes:participantsarepresentedwithpairsofdifferentrewards andtheyhadtoselecttherewardtheywantedthemost.Thereac- tiontimeofeachdecisionisthoughttoreflectthedegreetowhich arewardiswantedoveritsalternative.Thisimplicitindexhasbeen widelyusedintheliteratureasameasureofwantingorincentive salience(seeAppendixA).Notethatalthoughpreferenceindexes

Table3

Frequency(inpercentage)oftheformatinwhichthecueandrewardwerepresented inthemethodologicalproceduresassessinghumanwantingand/orliking.

Variable Descriptor Percentage

Cueformat Photo 76.04

Symbol 9.37

Samplesight 5.21

Video 5.21

Odor 4.16

Rewardformat Sample 50.72

Photo 27.53

Odor 13.04

Other 4.34

Caress 2.89

Symbol 1.44

computedthroughparticipants’choicesareequallytakentoreflect humanwantingandliking,theyaremorelikelytoreflectwanting.

Differenttheoreticaldescriptions(BerridgeandO’Doherty,2014;

BerridgeandRobinson,1998)havestatedthat wantingroughly correspondstotheconceptof“decisionutility”(Kahnemanetal., 1997)that referstothedegreetowhich anoutcomeischosen.

Indeed,BerridgeandAldridge,2008proposedthatirrationalwant- ingcanbeobservedincompulsiverewardseekingbehaviors(e.g., drugaddiction,bingeeating)andcanbeinterpretedasacaseof

“decisionutility”,wherethe“decisionutility”isdisconnectedfrom the“predictedutility”(i.e.,theexpectationofhowmuchafuture rewardwillbeliked)andthe“experiencedutility”(i.e.,thehedonic pleasureexperienceduringtherewardconsumption).Suchapro- cesscouldtherebyresultinthedecisiontopursuearewardthatis notexpectedtobelikedandthatisnotactuallylikedonceobtained.

3.5. Reward,cueandtiming

Rewardandreward-associatedcuesarecriticalelementsforthe incentivesaliencehypothesis:wantingistriggeredbythepercep- tionofacue,whilelikingistriggeredbyrewardconsumptionor receipt.Thepresentsystematicreviewhighlightsthatmostofthe methodologicalproceduresthathavebeenusedininvestigating humanwantingand/orlikingincludedtheseelements(seeFig.3a andb).Moreimportant,incongruencewiththeincentivesalience hypothesis,thelargestproportion(62.96%)ofmethodsassessing humanwantingincludedthepresentationofacue,whereasthe largest proportion(52.80%) of methods assessing humanliking includedthepresentationofanactualreward.However,arelatively highproportion ofstudies (31.46%;see Fig.3b)presented cues whilemeasuringliking. Ifincentivemotivationalelementssuch asrewardcuesarepresentedduringthehedonicmeasure,then themeasuredoesnotpurelyreflectthehedonicexperience,butis likelytoreflectbothmotivationalandhedonicinfluences.Inpar- ticular,takingthehedonicmeasureduringcuepresentationrather thanduringrewardconsumptionmightbeproblematicbecause themeasurereflectstheencodedmemoryofthehedonicexperi- enceratherthanthehedonicexperienceitself.Suchmemoriesof pasthedonicexperiencesareusedtobuildexpectedpleasantness, whichdoesnotcorrespondtoanimalliking.

Thecuewasvisualinmostmethodologicalprocedures(95.83%):

participants werepresented with photographs(76.04%),videos (5.21%)andsymbols(9.37%)associatedwiththereward,orpartici- pantsweresimplypresentedwitharewardsamplethattheycould notyetconsume(5.21%;seeTable3).Asmallerproportionofmeth- odsinvolvedolfactorycues(4.16%;seeTable3)suchasfoododors predictingafoodtaste.Inmostofthestudies,therewardwasasam- ple(e.g.,foodordrug)thatcouldbeconsumed(50.72%;seeTable3).

Rewardspresentedinthisformatareadvantageousformeasuring likingreactions,sincetheycantriggerastronghedonicexperience

ofsensorypleasurethatcaneasilybereported.Otherresearchers haveadaptedasimilarstrategybypresentingtherewardaspleas- antodors(13.04%)orcaresses(2.89%;seeTable3).Inanimportant proportionofmethods,therewardwaspresentedvisuallywhen therewardobjectwas,e.g.,apleasantphotograph(27.53%),ora symbolindicatingthereceiptofamonetaryreward(1.44%).

For themethodological procedures in which thecue and/or therewardwerepresented,wecodedwhenwantingand/orlik- ingmeasureswereadministeredwithrespecttotheseelements.

Timingisparticularlyimportantfortenetsoftheincentivesalience hypothesis:wantingisamotivationalcomponent;thus,itsspe- cificinfluenceispresentbeforerewardconsumption.Wantingis triggeredbytheperceptionofacue;thus,itshouldbemeasured duringorafterthecueperception,whereaslikingisanexperience triggeredbyrewardconsumptionandthusitshouldbemeasured duringorimmediatelyafterrewardconsumption.

Thevast majorityofmethods(80.00%)investigating wanting integratedthetimeaspectaccordingly,i.e.,measuredwantingdur- ing(60.00%;seeFig.3c)orimmediatelyafter(20.00%;seeFig.3c)a cuepresentation.Asmallerproportionofmethods(14.44%)mea- suredwanting during (1.11%; seeFig. 3c)or immediatelyafter (13.33%;seeFig.3c)rewardconsumption.Inthesecases,thewant- ingmeasure islikely toreflect hedonicinfluencesand learning processes,astheadministrationofa rewardafterastimulusor aninstrumentalactiontriggersPavlovianorinstrumentallearning processes.

Themajorityof studiesin whichmethodological procedures (59.21%)wereusedtoinvestigatelikinginvolvedtimingsthatwere similar tothose used in animal studies, measuringliking dur- ing(18.42%) or immediatelyafter (40.78%; seeFig. 3d) reward consumption. However,in a substantial proportion of methods (39.47%),likingwasmeasuredduring(35.52%;seeFig.3d)orimme- diatelyafter(3.94%;seeFig.3d)thepresentationofacue.Thismight beproblematicforseveralreasons.First,aspreviouslymentioned, accordingtotheincentivesaliencehypothesis,thepresentationofa cuetriggerswanting;thus,thesemeasuresarelikelytoreflectmoti- vationalinfluences.Second,inanimalstudies,likingisdefinedas anexperience;thus,ifthemeasurementistakenwhenthereward hasnotbeenconsumed,itisunlikelytoreflectthehedonicexperi- enceitself.Finally,studiesoftenmeasuredlikingduringoraftercue perceptioninpreferenceindexesandexpectedpleasantnessques- tions,and,asillustratedintheprevioussection(Section3.4),these twomeasurescouldpotentiallybeproblematicsincetheyarealso usedtoreflectwantinginotherstudies.

4. Discussion

Theaimofthepresentreviewwastodescribeassystematically aspossiblehow wantingandliking havebeenmeasuredacross studiesinvestigatinghumanrewardwithrespecttothekeyele- mentsoftheincentivesaliencehypothesis(i.e.,cue,rewardand theirrespectivetiming). Through this systematic review of the humanliterature,weaimedtoquantifythecontradictoryopera- tionalizationsofthewantingandlikingconstructsthathavebeen previously highlighted (Havermans, 2011,2012)and toidentify potentialconfoundsthatmighthaveledtothesecontradictions.

Wewereabletoinclude84publicationsinthepresentreview byusingstringentcriteria:we includedonlythose studiesthat explicitlyaimedtomeasurewantingand/orlikingwithspecificref- erencetotheincentivesaliencehypothesis(BerridgeandRobinson, 1998,2003;Robinson andBerridge,2003).Thisnumber ofstud- iesconfirmsthatamongresearchersinvestigatinghumanreward, thereisagreatdealofinterestintestingpredictionsoftheincen- tivesaliencehypothesis.Basicresearchtestedwhetherresultsfrom animalstudiescouldbereplicatedinhumanstudiesbyusingbrain

imaginingtechniques(e.g.,fMRI, PET;Born et al.,2011;Leyton etal.,2002),dopaminergicmanipulations(e.g.,dopaminiergicdrug administration;Leytonetal.,2002,2005)ormethodsthatareas similaraspossibletotheoriginalanimalstudies(e.g.,Pooletal., 2015b).More applied researchexplored whether thepotential independenceofwantingandlikingmightrepresentamechanism underlyinga varietyofproblematicbehaviorssuchasexcessive foodconsumption(e.g.,Lemmensetal.,2011c),substanceaddic- tion (e.g.,Goldstein et al., 2010) or behavioral addictions (e.g., gambling,excessivevideogameplaying;Thalemannetal.,2007;

Wölflingetal.,2011).

Overall, this systematic review showed that themajority of studiesonhumanwantingandlikinghaveintegratedkeyelements oftheincentivesaliencehypothesisintheirmethodologicalpro- cedures. Nonetheless,an importantnumber ofstudiesincluded measuresthatdonotreflectwantingandlikingasdefinedinthe animal literature. These studies generated confusion aboutthe wantingand/orlikingconstructsandmightrepresentthesource ofthecontradictoryfindingsproducedbythehumanexperimental literature.

Moreprecisely,mostofthestudiesmeasuredhumanwanting afteror duringthe perceptionof a reward-associatedcue.This measureiscongruentwiththeideathatwantingisproducedby asynergeticinteractionbetweenthecurrentphysiologicalstate ofanindividual(e.g.,hunger)andtheencounterofacue(realor vividlyimagined;e.g.,afoodphotograph)associatedwithareward (e.g.,food)thatisrelevanttotheindividual’scurrentphysiological state(BerridgeandO’Doherty,2014;BerridgeandRobinson,1998;

Zhangetal.,2009).Neitherthecuenorthephysiologicalstateisby itselfsufficienttotriggerwanting:thesynergeticcombinationof thesetwoelementsiscritical.Anindividualinaparticularphysio- logicalstatewillnotshowanywantingbehaviorifheorshedoes notencounteracue,andacuewillnotelicitwantingbehaviorif theassociatedrewardisnotrelevantforthephysiologicalstateof theindividual(RobinsonandBerridge,2013;Tindelletal.,2009;

WyvellandBerridge,2000;Zhangetal.,2009).Therefore,allthe proceduresthatdidnotmeasurewantingduringoraftertheper- ceptionofarealorvividlyimaginedcueareunlikelytotrulyreflect thespecificinfluenceofwanting.Eventhoughthesestudiesarenot themajority,theystillrepresentmorethanathirdofthestudies selectedinthepresentsystematicreview.

Similarly,mostofthestudiesmeasuredlikingduringorimme- diatelyaftertheconsumptionofthereward,whichisinlinewith theincentivesaliencehypothesisthatdefineslikingasthehedo- nicexperienceoftheconsumptionorthereceiptofanimmediate reward(Berridge,2000;BerridgeandKringelbach,2015;Berridge andO’Doherty,2014;BerridgeandRobinson,1998,2003).There- fore,measurementsoflikingshouldbetakenascloseaspossible toreward consumption in order toreflect thehedonic experi- ence. Nonetheless, the present review revealed that in almost halfofthemethodologicalproceduresassessinghumanliking,the reward itself wasnot presented to theparticipants, but rather onlyrewardcues,orquestionswereaskedonexpected,remem- beredorimaginedlikeability.Allthesemeasuresarebasedonthe encodedepisodicmemoryofthepasthedonicpleasureexperienced.

KahnemanandRiis(2005)illustratedseveralcasesinwhichthe memoryofanexperiencedidnotcorrespondtotheexperienceitself, becauseofvariablesbiasingtheencodingprocessoftheexperience.

Forinstance,thesameexperiencewillberememberedinaremark- ablydifferentwayifthemostintenseemotionalmomentissituated atthebeginningortheendoftheexperience.Thus,remembered likingusuallydivergesfromexperiencedliking.Rememberedlik- inginhumansdoesnotrefertothesameconceptbeingmeasured inanimalstudies,inwhichlikingisclearlyconceivedasahedonic experience.

Amajorproblematic aspectofhumanwanting and/orliking investigationsisrepresentedinthemeasuresofpreferencesand expectedpleasantness.Differentimplicitandexplicitindexesof preferencesandexpectedpleasantnesstakenatthesametimein themethodologicalprocedure(i.e.,duringcueperception)were usedtoreflectwantingin13%ofthestudies,whereastheseindexes wereusedtoreflectlikingin25%ofthestudies.Thisfindingdescrip- tively quantifies theobservation of Havermans (2011,2012) on thedifficultyofcongruentlyoperationalizingwantingandliking amonghumanresearchers.Butwhatarethereasonsunderlying thisdifficulty?

4.1. Expectedpleasantnessasamajorconfound

Wearguethatexpectedpleasantnessrepresentsamajorcon- ceptualconfoundunderlyingtheproblematicoperationalization ofwantingandlikinginhumans.Morethan12%ofthemeasures specificallyaskedparticipantstoreporttheirexpectanciesofplea- sure,mostly to measure liking but sometimesalso tomeasure wanting.Inmorethan47%ofthestudiesthataimedtomeasure liking,theresearchersdidnotpresenttherewardintheproce- duresthattheyused,butsimplyaskedparticipantstoremember orimaginehowmuchtheylikedorwouldlikeaparticularreward.

Thesequestionsreflectmemoriesorexpectationsoflikingrather thantheexperienceitself.

Cognitive representationssuchas expectedpleasantness are easiertoaccessinhumansthaninanimals;therefore,theyaremore widelypresentinthehumanliterature.Expectedpleasantnessisan evaluationofhowgoodorhowbadaparticularrewardisgoingto be.Thispredictioninvolvesactivereconstructionofpastepisodic memoriesoflikingexperienceswiththecurrentrewardandtheuse oftheseepisodicmemoriestoanticipateorpredictafutureexperi- ence(DickinsonandBalleine,1994;KahnemanandTversky,1984).

Ithasbeenwidelydemonstratedthatwhilerewardmemoriesand rewardanticipationarebasedonpastlikingexperiences,mostof thetimetheydo notcorrespondperfectly(Kahnemanand Riis, 2005).Intheliteratureonhumanwantingandliking,somescholars considerexpectedpleasantnesstobepartofthelikingcomponent ofrewardprocessing,sinceitsrepresentationismainlybasedon pastlikingexperiences (e.g.,Finlaysonetal.,2007a;Soussignan etal.,2012).Otherscholars,however,considerittobethemecha- nismunderlyingwanting:expectationsofpleasurearepartofthe anticipatoryrewardcomponentanddeterminethemotivationto obtainthereward(e.g.,Dawkinsetal.,2006;Gardetal.,2007).

Onewaytomovetowardresolvingthiscontroversyistoanalyze theroleofexpectedpleasantnessintheoriginalincentivesalience hypothesis.Inoneofitsfirstformulations(BerridgeandRobinson, 1998),thehypothesisdidnotclearlyspecifythedifferencebetween themechanismunderlyingwanting andthatofexpectedpleas- antness.Later,however,theexactroleofexpectedpleasantness withrespecttowantingwasfurtherformulated(Berridge,2007;

BerridgeandAldridge,2008;Zhangetal.,2009).Moreprecisely, theincentivesaliencehypothesisdistinguishesbetween(a)cogni- tivedesiresorexplicitwantingbasedonahigh-levelgoal-directed system,and(b)incentivesalienceorimplicitwanting,basedona moreprimaryPavloviansystem(seeFig.4).Itisimportanttonote thateventhoughcognitivedesiresaresometimesalsoreferredto asexplicitwanting,theydonotcorrespondtothewantingcom- ponentthat canbe dissociated from thehedonic properties of therewardintheframeworkoftheincentivesaliencehypothe- sis.Cognitivedesiresdonotsimplydifferfromwantinginterms ofexplicitness/implicitness,buttheyrelyondifferentunderlying mechanisms.Cognitivedesiresaredrivenbytheexpectedpleas- antnessofthereward. Expectedpleasantness isbuildbased on memoriesofpastlikingexperiences.Sincecognitivedesiresrely ona mechanism that depends onpast liking experiences, they

Reward-associated cue

Relevance to current concerns

Expected Pleasantness

Reward-seeking behaviors Perceptual salience

Liking experience Wanting

Pavlovian System

Cognitive desires Goal-directed System

Fig.4.Illustrationofthemechanismsproposedtobeinvolvedinwantingandliking.

Wantingisunderlainbytheinteractionbetweentheperceptionofacue(condi- tionedstimulus;CS)associatedwithareward(unconditionedstimulus;UCS)and therelevanceofthisrewardforthecurrentconcernsoftheindividual.Itisdis- tinctandpotentiallyindependentfromtheexpectedpleasantnessaswellasthe likingexperienceduringtherewardconsumptionorreceipt.Therewardrelevance alsoincreasestheperceptualsalienceoftherewardandthereward-associatedcue.

Liking,whichconsistsinthehedonicexperienceduringtherewardconsumption orreceipt,influencesexpectedpleasantnessofarewardbasedonthememories ofpastlikingexperiences,inturndeterminingcognitivedesires.Whilecognitive desiresrelyonthegoal-directedsystem,wantingreliesonthePavloviansystem, thustheyrepresenttwodistinctmotivationalcontrolsystemsofrewardseeking behaviors.

arenotcompletelyindependentfromliking.On theotherhand, wantingispotentiallyindependentfromanyhedonicaspectofthe reward,includingexpectedpleasantness.Thispotentialindepen- denceofwantingfromanyhedonicaspectoftherewardimplies thatitistheoreticallypossiblethatindividualscouldmobilizeeffort toobtainarewardthattheydonotexpecttolikeandthatthey arenotgoingtolikewhentheyobtainit(BerridgeandAldridge, 2008;BerridgeandO’Doherty,2014).Sincesuchaconditionhas mostlybeenobservedthroughspecificmanipulationsofmesolim- bicdopamineinrodents,scholarsworkingwithhumanparticipants havearguedthatsuchdissociationdoesnothaveecologicalvalidity (Havermans,2011,2012).Theyarguethatsuchbrainmanipulations donotprovideanalternativeexplanationfortheincreasedwant- ingforarewardthatisnotliked.Onthebasisoftheabsenceof analternativemechanismunderlyingwanting,someauthorshave suggestedthattheconceptofwantingstillimpliesatheoreticalgap thatneedstobefilled(Frijda,2010b).

4.2. Differentiatingexpectedpleasantnessfromaffective relevanceasasolution

Themajorcriticismoftheconstructofwantingconcernsthe lack of an alternative explanation to expected pleasantness in termsofunderlyingmechanisms.However,theincentivesalience hypothesisdoesproposeaclearcomputationalmechanismunder- lyingwanting(Zhangetal.,2009).Thismechanismconsistsofa synergeticinteractionbetweenthephysiologicalstateoftheindi- vidualandtheperceptionofacueassociatedwithareward:after thecueperception,wanting isnotdeterminedby theexpected

pleasantness,butisdirectlymodulatedbytherelevantphysiolog- icalstate.

Onemightobjectthatphysiologicalstatessuchashungerand thirst might also engender a parallel modification of expected pleasantnessofrewardssuchasfoodandwater.However,alarge corpusofanimal studiessuggeststhat thefunctioning ofthese twomechanismsisdifferentandthattheyrelyondissociableneu- ralnetworks(e.g.,Cardinaletal.,2002;Wassumetal.,2011b),as developedhereafter.

First,therepresentationofexpectedpleasantness,determining goal-directedactionsininstrumentallearning,criticallydepends uponepisodicmemoryofthepastpleasantexperiences(Balleine, 2005;DickinsonandBalleine,1994).Onlyifthereward(e.g.,apar- ticularfood)isconsumedinanewshiftedphysiologicalstate(e.g., hunger),andtheindividualexperiencestheincreasedpleasantness ofconsumingtherewardintheshiftedphysiological state(e.g., eatingthatparticularfoodwhenhungry),istheexpectedpleas- antnessconsequentlymodified (Balleine,1992,2005;Dickinson andBalleine,1994).Studiesshowedthatiftheencoding ofthe increasedvalenceoftherewardunderashiftedphysiologicalstate ispharmacologicallyblocked,individualsdonotadapttheeffort theymobilizetoobtainthefoodrewardaccordingtothenewphysi- ologicalstate(Wassumetal.,2011a).Motivationalbehaviorsdriven byexpectedpleasantness thusdependupontheepisodicmem- oriesofpriorexperienceswiththerewards.Researchconducted onanimalsdemonstratedthatthismechanismreliesonanetwork thatincludesthebasolateralnucleusoftheamygdala,whichplays acriticalroleinencodingandupdatingexperiencedpleasantness (Johnsonetal.,2009;Wassumetal.,2011a,b;Wellmanetal.,2005), aswellasprelimbicregionsoftheprefrontalcortex(Balleineand Dickinson,1998;KillcrossandCoutureau,2003)anddorsomedial regionsofthestriatum(Yinetal.,2005),whichretrieveandcom- puteexpectedpleasantnesstomodulatethebehavioraloutput.

Second,themechanismunderlyingwantingdoesnotnecessar- ilydependonepisodicmemories;itdynamicallyvariesaccording tothephysiologicalstate,withoutrequiringthere-experienceof rewardpleasantnessintheshiftedphysiologicalstate(Robinson andBerridge,2013;Tindelletal.,2009;Zhangetal.,2009).This findingimpliesthatwantingofaparticularrewardcandramati- callychangeaccordingtoashiftinthephysiologicalstate,even thoughtheexpectedpleasantnessforthatrewardhasnotchanged.

A growing corpus of experiments has demonstrated that cues thathave beenassociated withnon-attractive outcomesduring learningtriggerwantingiftheyarepresentedunderashiftedphys- iologicalstateinwhichthepreviouslynon-attractiveoutcomeis nowrelevant(DayanandBerridge,2014;DickinsonandBalleine, 1990;DickinsonandDawson,1987;RobinsonandBerridge,2013;

Tindelletal.,2009).Aclearexampleofsuchaphenomenonhas beenprovidedbyRobinson andBerridge(2013).In theirstudy, rodentslearnedtoassociateaPavloviancuewithasaltoutcome, whichwasexperiencedasunpleasantduringthelearningphase.

Afterthelearningphase,rodentswereputinasodium-depleted state(whichrodentshadneverexperiencedbefore)thatinduceda strongappetiteforsalt.Theserodentshadneverbeforeexperienced orconsumedsaltinthisnewphysiologicalstate;therefore,they couldnotupdatetheirexpectanciesaboutthepleasantnessofthe saltoutcome.Subsequently,thecuepreviouslyassociatedwiththe saltwaspresented.Eventhoughrodentshadneverconsumedsalt inthenewphysiologicalstateofsodiumdepletion,thepresentation ofthesalt-associatedcueinducedastrongwanting:thePavlovian cuethatduringthelearningphasewasavoidedhadnowbecome stronglyattractive,resultinginrodentsshowingseveralapproach behaviorstowardit(e.g.,sniffing,grasping,nibbling).Theseexper- imentssuggestthatthemechanismunderlyingwantingisnota simplePavlovianreflex,butthatthePavloviancueactivatesthe identityoftheassociatedreward,whichisrelevantforthecurrent

physiologicalstate,thusdeterminingwanting.Fromaneuralpoint ofview,thismechanismseemstorelyonadifferentnetworkthan expectedpleasantness,whichincludes,amongothers,thecentral nucleioftheamygdala(MahlerandBerridge,2009;Robinsonetal., 2014),theventralpallidum(Tindelletal.,2009),theventralstria- tumandtheventraltegmentalarea(Wassumetal.,2013;Wyvell andBerridge,2000).

Insummary,theoriesinvestigatinganimalaffectiveprocesses suggesttheexistenceoftwodifferentmechanisms:(a)expected pleasantness, which criticallyrelies ontheepisodic memoryof past liking experiences and (b) synergetic interaction between aphysiological state’sneedanda cueassociatedwithareward thatisrelevantforthecurrentphysiologicalstate’sneed.Whereas expected pleasantness drives cognitive desires, the interaction betweentheindividual’sphysiologicalstateandtheperceptionof therelevantreward-associatedcuedeterminesincentivesalience orwanting(Berridgeand O’Doherty,2014).Therefore, cognitive desiresbeingdrivenbymemoriesofpastlikingexperiencesarenot completelyindependentfromliking,whereaswantingisunderlain byamechanismthatiscompletelyindependentfromtheliking component(seeFig.4).

Likethetheoriesbasedonanimalresearch,theoriesinterested intheelicitationofaffectiveprocessesinhumans(Moorsetal., 2013;Sanderetal.,2005;Schereretal.,2001;SmithandEllsworth, 1985)statedacleardistinctionbetweenpleasantnessevaluation, consistingofexpectationsabouthowpleasantorpainfulastimu- luseventwillbe,andaffectiverelevanceevaluation,consistingof theinteractionbetweenthestimuluseventandthecurrentcon- cernsoftheindividualperceivingit.Currentconcernsareaffective representationsofpsychologicaland physiologicalmotives(e.g., self-achievement),needs(e.g.,hunger)andvalues(e.g.,security) that are of major importance for the individual (Frijda, 1988).

Therefore, affective relevance represents a mechanism that is similartothatproposedtounderliewantingintheanimalliter- ature:botharecomposedoftheinteractionbetweentheoutcome attributesandtheorganism’smotivationalstate(e.g.,Cunningham andBrosch,2012;RobinsonandBerridge,2013;Sanderetal.,2005;

Zhang et al., 2009).Although animal experiments manipulated affectiverelevancebyinducingphysiologicalmotivationalstates suchashungerorthirst(Balleine,1994;DickinsonandDawson, 1987;RobinsonandBerridge,2013),scholarsinvestigatingaffec- tive processing highlightedthat in humans, a larger variety of motives(i.e.,concernsbasedonsocialization,personalsensitivity ormomentarygoals)iseasilyaccessibleinanexperimentalsetting (Frijda,2010a;Sanderetal.,2005).Whileseveralconcernsexistand areaccessibleinhumans,theydonothavethesameimportance fortheindividual.Concernsareorganizedinadynamichierarchy ofprioritiesthatcanvarydependingonthesituation.Thedegree ofaffectiverelevanceisdeterminedbythenumberandtheimpor- tanceofconcernsforwhichtheoutcomepropertiesarerelevant (Sanderetal.,2005).Theaffectiverelevanceofanoutcomeisthus notbasedonpleasure,eventhoughtheyoftencorrelateinthecase ofrewardprocessing.

Considerationofthedistinctionbetweenexpectedpleasantness andaffectiverelevanceastwodifferentmechanismsunderlying cognitivedesiresandwantingorincentivesaliencemightsignifi- cantlyimprovethequalityofthemethodologicalproceduresthat areused toassessthespecificinfluencesof wantingand liking andmoregenerallycontributetoabetterunderstandingofhuman rewardseekingbehaviors.

First,expectedpleasantnessthatdeterminescognitivedesiresis builtontheepisodicmemoryofthishedoniclikingduringreward consumption.Thisfunctioningimpliesthatself-reportedmeasures ofwanting,likelytoreflectcognitivedesires,areunderlainbya mechanism thatrelies onpastliking experiencesand therefore arenotrecommendedinstudiesthataimtomeasuretheselective