The Relation Between Valence and Arousal in Subjective Experience Varies With Personality and ...

13 Pages·2017·0.25 MB·English

JournalofPersonality85:4,August2017
The Relation Between Valence and VC 2016WileyPeriodicals,Inc.
DOI:10.1111/jopy.12258
Arousal in Subjective Experience
Varies With Personality and Culture
Peter Kuppens,1 Francis Tuerlinckx,1 Michelle Yik,2
Peter Koval,1,3 Joachim Coosemans,1 Kevin J. Zeng,2 and
James A. Russell4
1KULeuven–UniversityofLeuven
2HongKongUniversityofScienceandTechnology
3AustralianCatholicUniversity
4BostonCollege
Abstract
Objective: Whileingeneralarousalincreaseswithpositiveornegativevalence(aso-calledV-shapedrelation),therearelarge
differences among individuals in how these two fundamental dimensions of affect are related in people’s experience. In two
studies,weexaminedtwopossiblesourcesofthisvariation:personalityandculture.
Method: In Study 1, participants (Belgian university students) recalled a recent event that was characterized by high or low
valence or arousal and reported on their feelings and their personality in terms of the Five-Factor Model. InStudy 2, partici-
pants from Canada, China/Hong Kong, Japan, Korea, and Spain reported on their feelings in a thin slice of time and on their
personality.
Results: InStudy1,wereplicatedtheV-shapeascharacterizingtherelationbetweenvalenceandarousal,andidentifiedper-
sonality correlates of experiencing particular valence–arousal combinations. In Study 2, we documented how the V-shaped
relationvariedasafunctionofWesternversusEasternculturalbackgroundandpersonality.
Conclusions: The results showed that the steepness of the V-shaped relation between valence and arousal increases with
Extraversionwithincultures,andwithaWest-Eastdistinctionbetweencultures.Implicationsforthepersonality–emotionlink
andresearchonculturaldifferencesinaffectarediscussed.
Keywords: Valence, arousal, pleasure, activation, structure of affect, structure of emotion, personality, culture
Subjectively experienced affect—how we feel—is a central encedaffect(Yik,Russell,& Barrett,1999).Thecurrentstudy
aspectofthemind,playingafundamentalroleindiversepsycho- focusesonthesetwodimensionswithoutassumingvalenceand
logical phenomena, including perception (Vuilleumier, 2005), arousalaretheonlydimensionsofaffect.
categorization (Niedenthal, Halberstadt, & Innes-Ker, 1999), The second basic question then becomes how valence and
decision making (Forgas, 1995), well-being, and psychopathol- arousal are related to each other. Answering this question is
ogy(Watson,2000).Consequently,understandinghowaffectis importantfortheoretical(whatisthestructureofaffectiveexpe-
rience?),empirical(whatistherelationofaffecttootherpsycho-
structuredinsubjectiveexperienceisoneofthemostfundamental
logical phenomena, from attention to well-being?), and
and long-standing challenges to the science of psychology
psychometricreasons(canvalenceandarousalbemeasuredor
(Diener, 1999; Wundt, 1924). To address this challenge, two
manipulated independently?). Several relationships have been
basic questions need to be answered: What are the fundamental
suggested,or,moreoften,assumed,including(a)independence
dimensionsunderlyingaffect,andhowaretheyinterrelated?
betweenthetwo(howpleasantorunpleasantoneisfeelingcon-
Regardingthefirstquestion,differentanswershavebeenfor-
veysnoinformationabouthowarousedoneisfeelingandvice
mulated(Fontaine,Scherer,Roesch,&Ellsworth,2007;Larsen
versa; e.g., Barrett & Russell, 1999; Carver & Scheier, 1990),
&Diener,1992;Osgood,May,&Mirron,1975;Russell,1980;
Thayer,1989;Watson&Tellegen,1985),andyettheyallcon-
CorrespondenceconcerningthisarticleshouldbeaddressedtoPeter
verge on identifying valence (i.e., pleasure-displeasure or
Kuppens,FacultyofPsychologyandEducationalSciences,KULeuven–
hedonictone)andarousal(i.e.,senseofenergyoractivation)as
UniversityofLeuven,Tiensestraat102,3000Leuven,Belgium.Email:peter.
amongthe mostfundamental properties of subjectivelyexperi- [email protected].
Relation of Valence to Arousal 531
Figure1 Overviewofpossiblerelationsandassociatedmodelequationsbetweenvalenceandarousal:(a)independence,(b)(positive)linearrelation,(d)
symmetricV-shapedrelation,and(e)asymmetricV-shapedrelationincludingbothpositivityoffsetandnegativitybias.
(b) a strictly linear relation between valence and arousal (how aresult,wedonotknowwhetheranyrelationfoundcharacter-
pleasant one feels increases [or decreases] how aroused one izes the structure of affect universally. Second, Kuppens,
feels), (c) the equation of arousal with the intensity of positive Tuerlinckx,Russell,andBarrett(2013)observedthatthesepro-
andofnegativevalence(themorepositiveornegativeonefeels, posalsareallaimedatanomotheticlevel.Kuppensetal.(2013)
themorearousedonefeels—graphicallyseenasaV-shapedpat- demonstrated that (a) any relationship between valence and
ternofarousalasa function of valence;e.g.,Clore,Ortony, & arousalatthenomotheticlevelaccountsforonlyasmallpropor-
Foss,1987;Lang1994),and(d)anasymmetricV-shapedrela- tionofthetotalvariance,and(b)thereissufficientvariationat
tionshipwithapositivityoffset(feelingsofarousalarehigherat theidiographic(orwithin-person)levelthatvalenceandarousal
lowlevelsofpositivevalenceascomparedtonegativevalence) caninprincipleshowanyrelationshipdependingontheperson
and a negativity bias (negative feelings increase more strongly and the situation (e.g., the nomothetic relation accounted for
thandopositivefeelingswithincreasingarousal;e.g.,Cacioppo only 1–18% of the total observed variance across studies). In
&Gardner,1999).Figure1providesavisualdepictionofeach otherwords,howvalenceandarousalarerelatedcanbemark-
oftheseproposedrelations. edly different from one individual toanother, ranging from V-
Inadditiontotheexistenceofcompetingproposals,twofur- shaped, to linear, to inverted V-shaped relationships, or even
ther limitations preclude us from accepting any such proposal. independence. This malleability means that for some individu-
First,thefindingsandconclusionsfrompreviousresearchwere als, high pleasant and/or unpleasant affect is habitually experi-
basedexclusivelyonWesternsamples.Moreover,toourknowl- encedincombinationwithhigherlevelsofarousal,whereasfor
edge, no cross-cultural research has addressed these four com- others it may be experienced with lower levels of arousal, and
peting modelsor indeed the general question ofthe relationof for some, valence and arousal are experienced relatively inde-
valencetoarousal(e.g.,Perunovic,Heller,&Rafaeli,2007).As pendently. Large variation at the idiographic level underscores
532 Kuppens, Tuerlinckx, Yik, et al.
the large malleability of the structure of affect. This variability predictthe habitualexperience of specific affective states(e.g.,
reflectsanimportantyetunderstudiedsourceofdiversityinsub- DeNeve & Cooper, 1998; Kuppens et al., 2007; Larsen &
jective emotional experience and thus has important implica- Ketelaar, 1991; Watson & Clark, 1984; Yik, Russell, Ahn,
tions for understanding the antecedents and consequences of Fern(cid:2)andezDols,&Suzuki,2002).
affectivestates.Forexample,idiographicvariationinthestruc- Extraversion has consistently been found to be related to
ture of affect would cast doubt on the existence of a universal increased experience of positive emotions (see Lucas, Le, &
neuralarchitectureasthedeterminantoftherelationofvalence Dyrenforth,2008,forameta-analysis).
to arousal. Moreover, idiographic variation in the valence– Perhaps because of its specific responsivity to appetitive
arousalrelationmayimplythatvalenceandarousalmaynotuni- stimuli that involve the possibility of reward, Extraversion is
formlyaffectperception,memory,well-being,andsoon(fora particularly related to the experience of high arousal positive
more elaborate discussion, see Kuppens et al., 2013). Thus, affect(Smillie,Cooper,Wilt,&Revelle,2012).Inotherwords,
explainingthisvariationisofcriticalimportanceforunderstand- when aroused or when feeling pleasant, it is more likely for
ingindividualdifferencesinthesubjectiveexperienceofaffect extravertscomparedtointrovertstoalsoexperiencehighlevels
andhowaffectimpactsotherphenomena.Furthermore,theexis- ofpositivefeelingsandarousal,respectively.
tence of reliable and meaningful individual differences in the Neuroticism,inturn,isconsideredtoreflectthedispositional
valence–arousal relationship would be further established by tendencytoexperiencenegativeemotions(Larsen&Buss,2015).
showing that these differences are related to known sources of Unlike Extraversion, which seems to be particularly
individual differences in order to explain part of the observed geared to the experience of high arousal states, Neuroticism
idiographicvariation. has equally been related to the habitual experience of both
Thepresentarticleaddressestwoissues.First,weprovidefur- low arousal negative feelings, such as sadness and depres-
ther evidence on the relation between valence and arousal. Our sion, as to the experience of high arousal negative feelings,
hypothesis is that the relationship is characterized by a weak, suchasstressandanxiety(Matthews&Deary,1998;Watson,
asymmetric V-shape atthe nomothetic level but bylarge varia- Gamez,&Simms,2005).
tionattheidiographiclevel(Kuppensetal.,2013).Second,we Based on this literature, we hypothesized that Extraversion
aimtocontributetounderstandingthenatureoftheidiographic wouldbepositivelyrelatedtoindividualdifferencesinvalence
variation by looking into how it correlates with personality when people experience high arousal (e.g., extraverts gravitate
(Study 1) and culture (Studies 1 and 2). Indeed, personality is toward feeling pleasant when feeling active) and to higher
assumed toreflect the mostbasic individual differences dimen- arousal when feeling positive (e.g., extraverts gravitate toward
sions in how people behave, think, and feel (Larsen & Buss,
feeling active when feeling pleasant). In other words, people
2014),andcultureisassumedtohaveaprofoundinfluenceon,
who are high on Extraversion would experience more positive
orisevenconstitutedof,thewayspeopleexperiencetheirworld
feelings accompanied by higher levels of arousal. In addition,
emotionally(e.g.,Mesquita,DeLeersnyder,&Boiger,inpress).
we hypothesized that Neuroticism would be negatively related
tovalencewhetherarousalisloworhigh(i.e.,neuroticsgravi-
tatetowardfeelingunpleasant).
STUDY 1:PERSONALITYCORRELATES
Totestthesehypotheses,weconductedastudyinwhichpar-
OF VARIATIONIN THE RELATION
ticipants completed a personality questionnaire assessing the
BETWEEN VALENCE ANDAROUSAL
FFM dimensions, recalled four recent events (one highly acti-
Thelargeindividualdifferencesinthevalence–arousalrelation vated,onelowinactivation,onehighlypleasant,andonehighly
raise the important question of the extent to which this idio- unpleasant), and reported their experienced levels of valence
graphicvariationismeaningful: Isthe variation inthe habitual andarousalinthoseevents.Thisevent-recallstudyallowedfor
occurrence of particular valence–arousal combinations simply idiographic variability in experiencing valence in relation to
random samplingerror,orcanitbeexplainedintermsofrela- highversuslowlevelsofarousal(Events1and2),andvariabili-
tionships with well-established, independently defined sources tyinexperiencingarousalinrelationtopositiveversusnegative
ofindividualdifferences?Ananswertothisquestionwouldcon- valence(Events3and4),anditenabledustoexaminehowthis
tributetobuildinganomologicalnetworkaroundtheobserved variabilitywasrelatedtopersonality.
idiographic variation in valence–arousal relations and would
highlight the emotional concomitants of important established
Method
dimensionsofindividualdifferences.
In this study, we examined the associations between idio- Participants. A total of 482 first-year Belgian, Dutch-
graphic variability in experiencing valence–arousal combina- speakinguniversitystudents(398females;M 519years)par-
age
tions and individual differences in the personality dimensions ticipatedinreturnforcoursecredit.
definedbytheFive-FactorModelofpersonality(FFM;McCrae
&Costa,1987).Alargeliteraturehasdocumentedhowperson- Materials and Procedure. In a Dutch-language computer-
ality dimensions, particularly Extraversion and Neuroticism, izedtask,participantsfirstreadshortdescriptionsofthepleasure
Relation of Valence to Arousal 533
Figure2 Relationship between valence and arousal in the (a) high arousal (based on all data), (b) low arousal (based on all data), (c) positive valence
(basedonpositivevalencedataonly),and(d)negativevalence(basedonnegativevalencedataonly)conditions.Forillustrativepurposes,thedashedline
reflectsthenon-parametricrelationship.Thepointsaretheplottedrawdata(Study1).
andarousaldimensions.Next,participantswereaskedtorecall .44 to –.58 per condition). The final score for valence was the
four different events: the last time they had felt echt actief meanofthreescores:thebipolarSAMvalenceratings,thebipo-
(“really active”; high arousal condition), echt inactief (“really lar verbal scale valence score, and the valence score based on
inactive”;lowarousalcondition),echtgoed(“verygood”;posi- theunipolaritems.Thefinalarousalscorewassimilarlydefined.
tive valence condition), and echt slecht (“very bad”; negative Thereliabilityofthesemultisourcecompoundscoreswasgood
valencecondition). (Cronbach’sasforvalencerangedfrom.68to.95acrosscondi-
Thefoureventdescriptionswerepresentedinaseparateran- tions;thoseforarousalrangedfrom.63to.79).
domorderforeachparticipant.Whentryingtorecalleachevent, Finally,participantscompletedacomputerizedversionofthe
participantswereinstructedtoclosetheireyes,relivetheexperi- Dutch translation of the Ten-Item Personality Inventory (TIPI;
ence,andrememberexactlyhowtheyfeltatthetime.Theynext Hofmans,Kuppens,&Allik,2008),whichmeasureseachofthe
typedashortdescriptionoftheeventintoatextbox.1 fivepersonalitydimensionsbymeansoftwo(opposite)itemsto
Next, their feelings of arousal and pleasure were assessed bescoredona7-pointscalerangingfrom1(describesmenotat
using multiple measures (while their description of the event all)to7(describesmeverywell).
remainedatthetopofthescreen).Allratingsweremadeon9-
point scales. The first measure was the Self-Assessment-
Results
Manikin(SAM;Bradley&Lang,2007),whichconsistsoftwo
bipolarratingsscaleswithpictorialmanikinsrepresentingvary- Nomothetic Relationship. While the design of this study
ingvaluesofpleasure(rangingfromunpleasanttopleasant)and was primarily intended to explain idiographic variation in the
arousal(rangingfromlowtohigharousal),whichwererecoded experienceofparticularvalence–arousalcombinations,thedata
torangefrom24to14.Thesecondmeasureconsistedoftwo canneverthelessbeusedtoverifywhethertheyconformtopre-
semanticdifferentialitemsrangingfromverybadtoverygood dictions that follow from the hypothesis that a nomothetic V-
(valence) and from very passive to very active (arousal), again shapedrelationbetweenvalenceandarousalfitsthedatabetter
recodedtorangefrom24to14.Thethirdsetofmeasurescon- than other models suggested in the literature. To this end, we
sistedof16unipolaradjectivescales,twoitemsforeachofeight examinedthe nomotheticrelationbetweenvalence andarousal
combinations of arousal and valence (high arousal-positive: ineachcondition.Wefitdifferent statistical regression models
enthusiastic, peppy; neutral arousal-positive: pleasant, good; inwhicharousalwaspredictedbyvalenceinwaysreflectingthe
low arousal-positive: calm, relaxed; low arousal-neutral differenttheoreticalmodels.Forthehighandlowarousalcondi-
valence:passive,sleepy;lowarousal-negative:bored,sad;neu- tions, we tested models involving (a) independence (Model 1:
tral arousal-negative: bad, unpleasant; high arousal-negative: predictingarousalbyaninterceptonly;seeequationwithFigure
nervous, stressed; high arousal-neutral valence: active, acti- 1a),(b)alinearrelation(Model2:predictingarousalbyvalence;
vated).Onthe basis ofthese items,a valence scorewas calcu- seeequationwithFigure1b),(c)asymmetricalV-shapedrela-
latedbysubtractingthemeannegativevalencecompositescore tion (Model 3: predicting arousal by the absolute value of
from the mean positive valence composite score (intercorrela- valence;seeequationwithFigure1c),and(d)variousinstances
tionbetweenboth5–.90,p<.001acrossconditions,andrang- of asymmetrical V-shaped relations (Models 4–6: predicting
ing between –.48 to –.84 per condition); an arousal score was arousalbytheabsolutevalueofvalenceandallowingfordiffer-
calculatedbysubtractingthemeanlowarousalcompositescore ences in intercept and/or slope for positive and negative
from the mean high arousal score (intercorrelation between valence—models that reflect positivity offset and/or negativity
both5–.77,p<.001acrossconditions,andrangingbetween– bias; see equation with Figure 1d). Finally, Model 7 tests a
534 Kuppens, Tuerlinckx, Yik, et al.
2aR .02 .05 .00 .03 .05.00 .04.02.03 aredis-
el
d
c 00 00 01 00 0000 000001 mo
Model7parametri PostP . . . . .. ... best-fitting
n
No BIC 459 611 867 022 870359 463111853 the
1 1 1 2 11 11 of
es
u
Model6AsymmetricVPositivity1OffsetNegativ-ityBias BICPostP 456.00 608.02 —— —— 864.03358.00 462.01108.01855.00 enmodels.Theval
1 1 11 11 ev
s
of
Model5mmetricVativityBias PostP .05 .07 — — .05.00 .05.09.06 mongtheset
AsyNeg BIC 1451 1605 — — 18621358 14571102850 dataa
chStudy Vset P .04 .33 — — .04.02 .09.10.07 giventhe
ValenceforEa Model4AsymmetricPositivityOff BICPost 1452 1602 — — 18631355 14561102850 ofeachmodel
y
asaFunctionof Model3mmetricV PostP .78 .57 — — .88.05 .84.61.49 osteriorprobabilit
odeled Sy BIC 1446 1601 — — 18571353 14521098846 5ostPp
M P
WhenArousalIs Model2LinearRelation BICPostP 453.02 623.00 866.08 001.99 881.00353.04 469.00106.01852.02 esreflectbetterfit);
cs 1 1 1 2 11 11 alu
atisti 1 0 1 1 07 084 werv
SelectionSt Model1ependence PostP .1 .0 .9 .0 .0.8 .0.1.3 criterion(lo
ofModel Ind BIC 1449 1618 1861 2010 18781347 14641101847 nformation model.
Table1Summary Study Study1HigharousalconditionLowarousalconditionPositivevalenceconditionNegativevalenceconditionStudy2CanadaChina/HongKongJapanKoreaSpain 5Note.BICBayesianiplayedinboldface.a2Rofthebest-fitting
Relation of Valence to Arousal 535
nonparametric relation between valence and arousal, where conditions),orinonelegoftheV-shape(thepositiveleginthe
arousalismodeledasafreefunctionofvalenceinordertodetect positive valencecondition andthe negativeleg inthe negative
possibleshapesnotseeninthepreviousmodels. valence condition) was moderated by personality. To this end,
However, the positive and negative valence conditions we performed 20 separate regression analyses across partici-
allowed us to test only part of the nomothetic V-shape predic- pantsinwhicharousalwaspredictedbyaV-shapedfunctionof
tion, namely, the increase of arousal with increasing positive valence(inthecaseofthehighandlowarousalconditions)ora
and negative valence, respectively. Indeed, due to the event- linearfunction(inthecaseofthepositiveandnegativevalence
recall manipulation, valence scores were negatively skewed in conditions) in interaction with each of the five personality
thepositivevalenceconditionandpositivelyskewedintheneg- dimensions(separatelyforeachpersonalitydimension).Innone
ativevalencecondition.Asaresult,itdoesnotmakesensetofit of these analyses, however, was the V-shape or one leg of it
V-shaped models in these conditions. Therefore, only models moderatedbypersonality(allps>.05).
involving independence (Model 1), a linear relation (Model 2; Second, we examined the extent to which personality was
whichwoulddetectonelegofaV-shape),andanonparametric related to the experience of each particular valence–arousal
relation(Model7)fittedtothedatainvolvedpositiveandnega- combination. Specifically, we regressed the valence scores on
tivevalencescoresundertheseinstructions,respectively. allfive personality dimensions separatelyfor the high and low
Subsequently, we applied model-selection procedures to
arousal conditions, and the arousal scores on the same five
select the best-fitting model, relying on the Bayesian informa-
dimensionsseparatelyforthehighandlowvalenceconditions.
tion criterion (BIC), a model selection criterion that balances
Theseanalysesinformushowindividualdifferencesintheper-
modelfitwithparsimony,andposteriorprobabilityestimatesof
sonality dimensions uniquely relate to idiographic variation in
model fit, which provide a quantification of uncertainty as a
valence when feeling highly active or passive, and to arousal
probability statement (i.e., determining the probability that the
whenfeelingverypositiveornegative.Byprovidingthisinfor-
modelistrueafterhavingseenthedata;Raftery,1995).
mationaboutthevalence–arousalcombinationspeoplewithcer-
The results can be found in Table 1. In the high and low
tain personality traits tend to experience, we can infer how
arousal conditions, both the BIC and the posterior probability
personalityshapestherelationbetweenvalenceandarousal.
statisticsfavoredaV-shapedrelationship(Model3).Intheposi-
In the high arousal condition, Extraversion predicted more
tive valence condition, independence was favored, and in the
positivevalence(b5.19,SE5.06,p<.01;theotherfourper-
negative valence condition, a linear relationship was favored.
sonality dimensions failed to make a significant contribution,
Theselectedmodelsforthedifferentconditionsaregraphically
ps>.05). In the low arousal condition, only Neuroticism pre-
illustrated in Figure 2, along with the nonparametric model to
dicted more negative valence (b5–.16, SE5.08, p<.05; all
illustratetherelationbetweenvalenceandarousalwhennolin-
otherps>.05).2Inthepositivevalencecondition,Extraversion
earrestrictionsareimposed.Thesefiguresclearlyshowthatfeel-
(b5.20, SE5.05, p<.001) and Agreeableness (b5.30,
ing more positive or negative is on average accompanied by
SE5.08, p<.001) each predicted higher arousal (all other
feeling more arousal in almost all conditions, essentially con-
ps>.05). In the negative valence condition, Neuroticism
firmingtheV-shapedrelationbetweenvalenceandarousal.Fig-
(b5.15, SE5.06, p<.05) and Extraversion (b5.15,
ure 2c, reflecting the positive valence condition, in which an
SE5.06,p<.05)predictedhigherarousal,whereasAgreeable-
independence model was favored, is an exception to this rela-
ness (b5–.21, SE5.09, p<.05) predicted lower arousal (all
tion, although the nonparametric function again goes in the
otherps>.05).
directionofa(one-armed)V-shape.
Idiographic Variation. Thewidescatteringoftheplotteddata
Discussion
in the Figure 2 panels clearly illustrates, however, that partici-
pantsstronglydifferedinhowarousedtheyfeltwhentheyfelt As expected, valence and arousal were found to be related to
verygoodorverybad,andinhowpositiveornegativetheyfelt oneanotheratthenomotheticlevel.Therewerereliabletenden-
when they felt very high or low arousal, showcasing the large cies for certain combinations of valence and arousal to occur
idiographicvariationinvalence–arousalcombinations.Thesize morefrequentlythanothercombinations.Thisvalence–arousal
of the idiographic deviation from the overall relation is also relationship was not simply linear. On average, as people felt
illustrated bythe very low R-squared values (ranging from .00 more intensely positive or more intensely negative, they felt
to.05)reportedinTable1(seealsoKuppensetal.,2013,report- morearoused: Intensityineithervalence directionwas accom-
ingvaluesbetween.01and.34,withamedianof.08)thatindi- paniedbymorearousal.
cate howmuchofthe varianceinthe data ofeachconditionis However,ascanbeobservedfromthedatainFigure2,this
explainedbytheselectednomotheticmodel. relationshipwasbynomeanstrueforallindividuals.Themag-
Toexaminehowmuchofthevariationcouldbeexplainedby nitudeoftherelationwassmallatthenomotheticlevel,andany
personality, we took two approaches. First, we examined to combinationofvalenceandarousalispossible,andindeedfre-
what extent variation in the slope of the entire V-shape or the quent. Contrary to the overall V-shaped relation at the nomo-
positiveornegativelegoftheV-shapeseparately(inthearousal theticlevel,someindividualsreportedfeelinglowarousallevels
536 Kuppens, Tuerlinckx, Yik, et al.
incombinationwithhighlypositiveorhighlynegativevalence. (e.g., relaxed and peaceful feelings). Such differences in ideal
Similarly, when participants were instructed to remember affectarethoughttooriginatefromculturaldifferencesininter-
instances when they felt high or low arousal, they reported a personalgoals.InWesternsocieties,theemphasisliesoninde-
widerangeofvalencescores.However,thiswaslessthecasein pendence, influencing the environment, and personal
thehigharousalcondition,inwhichparticipantspredominantly achievement, whereas in Eastern societies, the focus lies more
reportedpositiveexperiences(seeFigure2a).Thus,theinstruc- on interdependence, adjustment to the environment, and inter-
tion to remember a highly active experience seemed to prime personalharmony(Markus&Kitayama,1991;Tsai,2007).
primarilypositivememories,atleastinaWesternEuropeancul- Althoughidealaffectisnotalwaysachievedinactualexperi-
turalcontext. encedaffect,culturaldifferencesinthetwocertainlyalign(Tsai,
The observed idiographic variation in scores was related to 2007).Asa consequence,people fromdifferent cultures might
established dimensions of personality. Extraverts tended to differintheirnomotheticexperienceofvalence–arousalcombi-
experiencehigherarousalcomparedtointrovertsincombination nations:Peoplemightsucceedinachievingtheirideal.InWest-
withpositiveandnegativevalence,andhigherpositivevalence ern societies, the preference for high arousal positive affect
in combination with high arousal. Individuals scoring high on wouldcreateahigherprevalenceofhigh(vs.low)arousalposi-
Neuroticism tended to experience high arousal in combination tive states; in Eastern societies, the preference for low arousal
withnegativevalenceandnegativevalenceincombinationwith positive affect would create a higher prevalence for low (vs.
lowarousal.Inaddition,agreeableindividualstendedtoexperi- high)arousalpositivestates.Combinedwithareasonablyantici-
encehigharousalincombinationwithpositivevalence,butlow pated preponderance of average arousal levels experienced at
arousal in combination with negative valence. These results lowpositiveandnegativevalence,thesepreferencescouldtrans-
showed that individual differences in experiencing particular lateintomoreorlesssteep,orevennegative,positivevalence–
valence–arousalcombinationsaremeaningfullyrelatedtobroad arousalrelationsacrosscultures.
personality dimensions in ways consistent with previous find- Predictionsregardingnegativeaffectarelessstraightforward,
ings on these personality traits. Regarding Extraversion, early as ideal affect is understandably situated on the positive rather
theoriesalreadylinkedthispersonalitydimensionwiththeseek- than negative side, and less research on ideal negative affect
ing of high arousal states (Eysenck, 1967); Neuroticism is a exists.Still,peopledoonoccasionseekoutandachievenegative
strong predictor of negative emotionality both in the high and affect(Tamir,2009).Anintriguingquestion,therefore,iswhether
lowarousalrealm(Barlow,Sauer-Zavala,Carl,Bullis,&Ellard, a cultural proneness for experiencing high or low arousal affect
2014); the low anger-proneness associated with Agreeableness on the positive side also generalizes to the negatively valenced
(e.g.,Kuppens,2005)maypredisposeindividualstoexperience realmoffeelingstates.Thereareseveraltheoreticalreasonswhy
this may be so. High arousal negative states such as anger and
lowratherthanhigharousalnegativestates.
anxietyarise,inpart,tomobilizeenergyforinfluencingtheenvi-
Thecurrentstudydesignfocusedonparticularregionsofthe
ronment,andthemotivationtoexertinfluenceoneventsiscon-
affectivespace.Hence,itwasnotdesignedtoestablishpersonal-
sideredtobemoredominantinWesternthaninEasterncultures,
itymoderationofglobalpatternsofvalenceandarousalacross
which are more concerned with adjustment (Tsai, 2007). The
theentireaffectivespace(suchasaV-shape).Forthispurpose,
oppositeholdsforlowarousalnegativestatessuchassadnessand
dataareneededthatpotentiallyrangeacrosstheentireaffective
depression, which reflect passive compliance rather than active
realm (such as those based on uninstructed self-reported affect
interactionwithongoingevents.Empirically,culturaldifferences
inathinsliceoftime).
inthevalence–arousalrelationremainunexplored.
InStudy2,we examinedpossible culturaldifferences inthe
STUDY 2:CULTURALCORRELATESOF nomothetic relation between valence and arousal by analyzing
VARIATIONINTHE RELATION existingdatacollectedbyYiketal.(2002)infivedistinctcultural
contexts:Canada,China/HongKong,Japan,Korea,andSpain.3
BETWEEN VALENCE ANDAROUSAL
Thesecountrieswereselectedbecausetheyrepresentfivedif-
Most data reported in the literature (including our Study 1) on ferentlanguagegroups(seeYiketal.,2002)andshowsignifi-
therelationbetweenvalenceandarousaloriginatedfromWest- cant variation on basic cross-cultural dimensions (e.g., with
ernsamples.Asaresult,thequestionremainswhetherthecon- Spain being more collectivistic thanmostother Western coun-
clusions drawn could be generalized to non-Western samples. tries,butmoreindividualisticthanEasterncountries;Hofstede,
Perhapstherelationsfound,suchastheshallowV-shape,reflect 2001). Participants recalled a moment from the preceding day,
a culture-specific feature of the structure of affect stemming basedonwhichtheyprovidedaffectratingsthatwerethenused
fromaWesternlikingofhigharousalaffectivestates. tocalculatecompoundscoresofvalenceandarousal.Thesedata
Therearegoodreasonstocontemplatethislatterpossibility. also contained scores on the FFM personality dimensions,
Tsai,Knutson,andFung(2006)convincinglydemonstratedthat therebyallowingustoexaminefurtherthepersonalitycorrelates
members from Western cultures prefer high arousal positive ofthevalence–arousalrelationshipbasedondataforwhichpar-
affect(e.g.,feelingexcitedandenthusiastic),whereasmembers ticipants were not instructed to report on particular restricted
of Eastern cultures instead prefer low arousal positive states regionsoftheaffectivespace.
Relation of Valence to Arousal 537
Figure4 Relationship between valence and arousal for people scoring
Figure3 Relationshipbetweenvalenceandarousalpercountryaccording highversuslow(basedonmediansplit)onExtraversion(E)(Study2).
toamodelthatassumesasymmetricV-shaperelation(Study2).
aroused, hyperactivated, activated), and deactivation (adjec-
Method tives: sleepy,still,quiet).Next,the study authorsidentifiedthe
Participants. Participants were undergraduate student volun- items thatwereculturally and theoreticallyappropriatetoindi-
teers.TheCanadian(Vancouver)datacamefrom535participants catepositiveandnegativevalenceandactivatedanddeactivated
(294 females; M 520 years); the Chinese/Hong Kong data arousalineachculture.Measurementinvarianceoftheresulting
age
from487participants(323females;M 520years);theJapa- item groups was tested using multigroup confirmatory factor
age
nese (Kyoto) data from 450 participants (222 females; analyses (Brown, 2015), with corresponding item parcels for
M 520 years); the Korean (Busan) data from 365 partici- eachresponseformatservingasindicatorsforpositivevalence,
age
pants (189 females; M 521 years); and the Spanish (Mad- negative valence, activation, and deactivation. Model selection
age
rid)data from233participants (37females;M 520years). (based on BIC) indicated weak invariance (CFI5.946,
age
Theyparticipatedinthestudyinreturnforcoursecredit. RMSEA5.094),suggestingthatthefactorloadingsofthepar-
celscanbeconsideredtobeequalacrossgroups.Becauseposi-
Materials and Procedure. To measure momentary affect, tive and negative valence scores were strongly negatively
wereliedonthe“rememberedmomentsmethod”inwhichpar- correlated (intercorrelation between both5–.81, p<.001
ticipantswereaskedtodescribetheiraffectinamomentduring acrosscountries,andrangingbetween–.73to–.87percountry),
theprecedingdaythattheyrememberedwell.4Theywereasked wecreatedavalencescorebysubtractingnegativefrompositive
toreporttheiraffectusingfourquestionnaires,eachinadifferent valence scores for each of the three formats. Similarly, an
format.The rationale for the inclusion of multipleformatswas arousalscorewascalculatedbysubtractingdeactivedfromacti-
toobtainscoresthatarenotdependentonaparticularscalefor- vated scores (intercorrelation between both5–.51, p<.001
mat,andtocaptureemotionalresponsesindependentofassess- acrosscountries,andrangingbetween–.35to–.63percountry).
ment method (see also Yik et al., 2002). A first format used Next,valencearousalscoreswereaveragedacrossthethreefor-
visual analogue scales with opposing emotion terms at either matstoobtainafinalvalenceandarousalscoreforeachsubject.
end, a second format involved indicating to what extent they Cronbach’salphasfromthefivesamplesrangedfrom.95to.97
experienced a number of emotions termed as adjectives on a for valence,and from.75 to.91 for arousal,indicatingreliable
scalerangingfrom1(notatall)to5(extremely),athirdformat scales.
involved indicating their agreement on a scale ranging from 1 In addition, personality was measured with the NEO Five-
(stronglydisagree)to5(stronglyagree)withanumberofstate- FactorInventory(NEO-FFI;Costa&McCrae,1992),a60-item
ments regarding their feeling state, and a fourth format asked questionnairedesignedtomeasuretheFFMdimensionsofper-
how well a number of statements describe their feeling at the sonality. Each of the five factors is represented by 12 items.
timeonascalerangingfrom1(notatall)to4(verywell).Each Responses are made on a 5-point rating scale ranging from
formatexceptforthefirstonecontaineditemsthatcouldbeused stronglydisagreethroughneutraltostronglyagree.Cronbach’s
tocomputescalescoresforpositivevalence(adjectives:happy, alphasfromthefivesamplesrangedfrom.62to.83,indicating
pleasant,content,satisfied),negativevalence(adjectives:miser- thatthescaleswereinternallyconsistent,similartotheoriginal
able, troubled, unhappy, dissatisfied), activation (adjectives: Englishscales(McCraeetal.,1998).Modelselection(basedon
538 Kuppens, Tuerlinckx, Yik, et al.
BIC) testing different types of measurement invariance using p5.038),butnotofthenegativevalence–arousalslope(B5–
multigroupconfirmatoryfactoranalysespointedtoweakinvari- .07,SE5.08,p5.398).TovisualizethemoderationbyExtra-
ance, indicating that factor loadings can be considered to be version, Figure 4 displays the V-shape relationship between
equal across groups (both for models involving 12 individual valenceandarousalforparticipantswithhighandlowlevelsof
itemsperscale,CFI5.602,RMSEA5.063,orparcelsconsist- Extraversion. While in the model underlying this figure Extra-
ingofthreeorfouritemsperscale,CFI5.832,RMSEA5.094, versionisallowedtomoderatetheinterceptandslopeofthepos-
andCFI5.825,RMSEA5.080,respectively). itive and negative leg of the V-shape separately (based on
median split), it is important to point out that again only the
moderationofthepositiveslopereachesstatisticalsignificance.6
Results
Second, although no clear-cut predictions could be formu-
First,weexaminedthenomotheticrelationshipbetweenvalence lated based on the results from Study 1, we examined in two
andarousalpercountry.AsinStudy1,thebest-fittingofseven similar models how Neuroticism (again country-centered)
modelswasselectedbasedontheassociatedBICandposterior related to the intercept and slope(s) of the V-shape. The first
probabilityestimates,separatelyforeachcountry.Theresultsof model showed that Neuroticism predicted a higher level of
thismodelselectioncanbefoundinTable1.Inallculturesbut arousal at neutral valence (intercept: B5.20, SE5.09,
one (China/Hong Kong), a symmetric V-shape (Model 3) was p5.027) and a slightly less steep slope overall (B5–.08,
the most dominant pattern in the data. The order of countries, SE5.04,p5.046).However,the secondmodel didnotshow
fromlargesttosmallest,basedonthesizeoftheslope(steepness that Neuroticism moderated the steepness of the separate posi-
oftheV-shape)inthosemodelswasCanada(B5.32,SE5.06,
tive and negative valence–arousal slopes (B5–.06, SE5.05,
p<.001),Spain(B5.27,SE5.11,p5.013),Japan(B5.25,
p5.283;B5.10,SE5.07,p5.189).
SE5.06, p<.001), and Korea (B5.18, SE5.06, p 5 .004).
Agreeableness,OpennesstoExperience,andConscientious-
For the data from China/Hong Kong, independence between
ness (country-centered) were not related to the intercept, V-
valenceandarousal(Model1)fitthedatabest.
shape,orseparateslopesinsimilarmodels(allps>.10).
Wenextfitamultipleregressionmodelassumingasymmet-
Inacombinedmodelinwhichallfivepersonalitydimensions
ricV-shapeonalldatacombined,allowingforseparateintercept
wereincludedininteractionwiththeinterceptandslopeofthe
and slope values per country. As seen in Figure 3, Canada,
V-shape,onlyNeuroticismsignificantlypredictedtheintercept,
Spain,andJapanwerecharacterizedbyasteeperV-shapedrela-
with higher Neuroticism being related to a higher level of
tionship,whereastheV-shapewasweakerinKoreaandalmost
arousal at neutral valence (B5.21, SE5.10, p5.033), and
absentinChina/HongKong.Theinteractionbetweenslopeand
Extraversion remained the only predictor of the slope, with
country was significant, F(4)53.58, p5.006, indicating that
higher Extraversion being related to a steeper slope of the V-
the country differences in steepness of the V-shape were reli-
shape(B5.12,SE5.06,p5.036;allotherps>.05).Finally,
able. A comparison of the BIC of a joint model with
inacombinedmodelinwhichallpersonalitydimensionswere
(BIC56710)andwithout(BIC56694)thisinteractionfavored
entered as predictors of the intercept and separate positive and
amodelwithouttheinteraction,however.
negative slopes, the only significant moderation that emerged
Finally,weexaminedhowtheshapeoftheV-shapeisafunc-
was Extraversion being positively related to the positive
tionofpersonalityacrossdatacombinedfromallcultures.The
valence–arousalslope(B5.15,SE5.07,p5.027).
results from Study 1 provide the basisfor a clear predictionin
thisrespect.AsExtraversionpredictedhigherarousalinthecase
of both positive and negative valence, we expected that the
Discussion
steepness of the V-shape would be predicted by Extraversion.
Totestthishypothesis,weestimatedaregressionmodelacross The findings from this study again replicated the previously
all data predicting arousal from valence in which country- documentedrelationbetweenvalenceandarousal,whileatthe
centered Extraversion interacted with the intercept (level of sametimedemonstratingthatculturalandindividualvariability
arousalatneutralvalence)andabsolutevalueoftheslope(steep- in that relationship are predictable and therefore meaningful.
ness of the V-shape).5 The results indeed showed that higher Specifically, all but one country showed a V-shaped relation
scores for Extraversion were related to a more steep V-shape between valence and arousal, with the exception being China/
(slope;(B5.12,SE5.05,p5.016;Extraversiondidnotrelate Hong Kong, in which valence and arousal were independent.
totheintercept,however,B5.01,SE5.10,p5.943).Inaddi- Moreover,thesteepnessoftheV-shapevariedbycountry.The
tion, we also examined how country-centered Extraversion V-shapewassteepestinCanada,thenSpain,thenJapan,moder-
moderatedthesteepnessoftheslopeseparatelyforpositiveand ateinKorea,andalmostabsent(i.e.,flat)inChina/HongKong.
negative valence (to examine whether Extraversion indeed Thesefindingsattesttothemalleabilityofthestructureofaffect
relatestoasteeperV-shapeoverall,orperhapsparticularlytoa (seealso,e.g.,Feldman,1995;Kuppens,2008).
steeper slope for positive or negative valence). The result Inaddition,theystronglyresonatewiththedocumentedpref-
showed that Extraversion only positively moderated the steep- erence for high arousal positive affect in Western cultures and
ness of the positive valence–arousal slope (B5.13, SE5.06, lower arousal positive affect in Eastern cultures (Tsai et al.,
Relation of Valence to Arousal 539
2006). Indeed, Figure 3 and the accompanying reported V- Extraversionhasbeenassociatedwiththehabitualexperienceof
shape slopes illustrate that the five samples can be ordered in positivehigharousalaffectivestates,butnotmuchwithpositive
termsoftheirarousallevelsatbothhighpositiveandhighnega- low arousal or negative states (Yik, 2009; Yik, Russell, &
tivevalenceinlinewithanEast-Westculturaldistinction.This Steiger,2011).Assuch,Extraversioncanberegardedasagen-
indicates that differences in ideal affect extend to experienced eral arousal-infusing factor for positive feelings. The approach
affect(see also Tsai et al.,2006),and italsodemonstrates that motivationandrewardsensitivitythatisconsideredtounderlie
cultural differences in preference for higher or lower levels of Extraversion (e.g., Smillie, 2013) may be the driving force
arousal extend to negative affect. The largest exception to this behindthelargersurgeinarousalinreactiontopositiveevents.
wasJapan,however,whichwascharacterizedbylowerlevelsof Otherpersonalitydimensionswerealsorelatedtomorespe-
arousalatneutralvalence,butdidshowcasearelativelystronger cific aspects of idiographic variation in the structure of affect.
V-shapecompared to the other Eastern countries (although we Neuroticism was generally related to higher levels of arousal,
note thattheaverage arousal level atneutralvalencewas quite particularly of negative affect. Neuroticism has indeed been
lowinJapan).Apossibleexplanationforthismaybedifferences associated with negative emotionality in numerous studies,
in scale use when reporting emotions between cultures. Also, although it must be said that this dimension is considered to
the data from Hong Kong, while displaying a flat relation makeindividualspronetoexperiencingbothhigharousalnega-
between valence and arousal, were characterized by relatively tivestates(e.g.,stress,anxiety)aswellaslowarousalnegative
states (e.g., sadness, depression). This duality is perhaps also
high arousal levels overall. Especially at neutral valence, the
reflectedinthefindingsfromStudy1that,ontheonehand,indi-
arousallevelofHongKongparticipantsexceededthatofpartici-
viduals high in Neuroticism tend to feel more negative when
pants from other nations. More research is needed to pinpoint
experiencinglowarousal(correspondingtosadnessanddepres-
theexactinterpretationofthesefindings,however.
sion), while at the same time also reporting increased arousal
Inaddition,thecurrentresultsallowedustofine-tuneinsights
when feeling negative (corresponding to stress and anxiety).
from the first study in that Extraversion was found to be the
Thecombinationofthesefindingsmayhelponetounderstand
most consistent independent personality predictor of the
theroleofNeuroticismin,forinstance,thereportedhighcomor-
arousal–valence relationship. Specifically, the findings showed
bidity between depression and anxiety (Barlow et al., 2014;
that Extraversionisrelated to a steeper V-shapeinthe relation
Mineka,Watson,&Clark,1998).
between valence and arousal, which is particularly driven by
In terms of culture, the first important aspect to highlight is
Extraversion being related to increasing positive valence with
thatdespitetheobservedculturalvariation,theV-shapebetween
increasingarousal.
valence and arousal still remains the dominant pattern across
Anotablelimitationofthisstudy,however,isthatitisdiffi-
cultures, albeit to varying degrees, ranging from positive to
culttoruleoutthepossibilitythatpersonalityand/orculturemay
totally flat.Our data didnot indicatea culture inwhich the V-
partlyhavedriventhespecificeventspeoplerecallwhenasked
shape was reversed, or made thoroughly asymmetric, as could
to report on a particularly memorable event from the previous
perhaps be expected on the basis of research on ideal affect
day (cf. Robinson & Clore, 2002), and that this may partly
(which would predict decreasing arousal with positive valence
accountforthefindings.Whilethiswouldreflectadifferentway
inEasterncultures).The observeddifferences moreor lessdid
personalityorculturemaybiaspeopleaffectively,experimental
align,however,withthedistinctionbetweenWestern/independ-
methods are needed to dissociate this possibility from the
entversusEastern/interdependentculturesthatareconsideredto
hypothesis that personality and culture predispose people to
lieatthebaseofdifferencesinidealaffect.Also,whileexperi-
respondaffectivelytoeventsinparticularways.
encedandidealaffectfolloweachother’scourse,theyarecer-
tainly not identical. One way to interpret thisset of findingsis
that although experienced affect does not coincide with ideal
GENERALDISCUSSION
affect,culturalpreferencesforhighorlowarousalstatesseemto
The present studiesreplicatedKuppens andcolleagues’(2013) pull at the overall nomothetic V-shape in one direction or
findings, in six new samples in both Western and Eastern cul- another,creatingculturaldifferencesinhowvalenceandarousal
tures, regarding an overall nomothetic V-shaped association relatetooneanother.Anadditionalfindingwasthatthecultural
betweenvalenceandarousal,witharousalincreasingaspositive ordering in experienced arousal levels at higher levels of posi-
ornegativevalenceincreases.Yet,atthesametime,thecurrent tive and negative valence was similar, suggesting that habitual
studies highlight significant and meaningful variation in this experience of high versus low arousal feeling states generalize
association across individuals and cultures. One of the main acrossvalence.
objectivesofthetworeportedstudieswastoidentifycorrelates Combining the findings from both studies, it is striking to
ofthesedifferences. note that cultural differences in Extraversion (e.g., McCrae,
Intermsofpersonality,themostconsistentfindingwasthat 2004; McCrae & Terracciano, 2005) seem to align with the
Extraversion was associated with how much arousal increases observed cultural differences in the V-shaped relation between
with increasing valence, particularly positive valence, under- valence and arousal, which showed a steeper slope in Western
scoringthecrucialroleofExtraversioninemotionalexperience. comparedtoEasternsamples.Thisobservationtiesthefindings