Research
Note
Calibrating
30
Years
of
Experimental
Research:
A
Meta-Analysis
of
the
Atmospheric
Effects
of
Music,
Scent,
and
Color
Holger
Roschk
a,∗,
Sandra
Maria
Correia
Loureiro
b,
Jan
Breitsohl
caDepartmentofServiceManagement,Alpen-Adria-UniversitätKlagenfurt,Universitätsstraße65-67,9020Klagenfurt,Austria
bMarketing,OperationsandGeneralManagementDepartment,InstitutoUniversitáriodeLisboa(ISCTE-IUL),BusinessResearchUnit(BRU/UNIDE),Av.For¸cas
Armadas,1649-026Lisbon,Portugal
cSchoolofManagement&Business,AberystwythUniversity,CeredigionSY233AL,UnitedKingdom
Availableonline27October2016
Abstract
Atmosphericin-storestimulihavebeenthesubjectofconsiderableempiricalinvestigationforover30years.Thisresearchpresentsa meta-analysisof66studiesand135effects(N=15,621)calibratingtheatmosphericeffectsofmusic,scent,andcoloronshoppingoutcomes.Atan aggregatelevel,theresultsrevealthatenvironmentsinwhichmusicorscentarepresentyieldhigherpleasure,satisfaction,andbehavioralintention ratingswhencomparedwithenvironmentsinwhichsuchconditionsareabsent.Warmcolorsproducehigherlevelsofarousalthancoolcolors, whilecoolcolorsproducehigherlevelsofsatisfactionthanwarmcolors.Theestimatedaveragestrengthoftheserelationshipsrangedfromsmall tomedium.Effectsizesexhibitedsignificantbetween-studyvariance,whichcanbepartlyexplainedbythemoderatorsinvestigated.Forinstance, largereffectsizeswereobservedfortherelationshipbetweenscentandpleasureinthosesampleswithahigher(vs.lower)proportionoffemales. Dataalsoindicatedatendencytowardstrongermusicandscenteffectsinservicesettingsascomparedtoretailsettings.Theresultsofthisanalysis, basedondataaggregatedacrosstheresearchstream,offerretailersaguidetoenhancecustomers’shoppingexperiencethroughjudicioususeof in-storeatmosphericstimuli.
©2016TheAuthor(s).PublishedbyElsevierInc.onbehalfofNewYorkUniversity.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Music;Scent;Color;Meta-analysis;In-storeatmospherics;Retailcustomerbehavior
One of the key success factorsfor any retailer or service providerispresentingcustomerswithapleasurable
consump-tion environment (Pan and Zinkhan 2006). A well-designed
storeenvironmentmaypositivelystimulatecustomers’senses, enhance their shopping experience, and ultimately translate intolargersalesrevenues(DoucéandJanssens2013;Sullivan 2002).Thesubtletyofatmosphericeffectsoftenresultsin cus-tomersbeingunawareoftheir exposuretothem,eventhough theirbehaviorisaffected(MorrinandRatneshwar2000).
Aca-demic researchers have explored how environmental stimuli
affect customers’ shopping behavior for more than 30 years
(Bellizzi, Crowley,and Hasty1983; Ludvigson andRottman 1989;Milliman1982).Inparticular,scholarshaveinvestigated
how music, scent, and color influence shopping outcomes,
∗Correspondingauthor.Fax:+434632700994091.
E-mailaddresses:holger.roschk@aau.at(H.Roschk),
sandra.loureiro@iscte.pt(S.M.C.Loureiro),jab99@aber.ac.uk(J.Breitsohl).
affectingemotionalreactions,satisfactionandpurchase inten-tion,andhaveproducedavoluminousliteraturewithsubstantial variation in sample composition, industry context,and study design(BellizziandHite1992;MattilaandWirtz2001;Sayin etal.2015).
This body of work hasproduced mixed results, including
significant and non-significant findings, as well as effects in opposingdirections,evenforthesamerelationship(Andersson et al. 2012; Cyr, Head, and Larios 2010; Michon, Chebat, and Turley 2005; Morrin and Ratneshwar 2000; Yalch and Spangenberg1988).Furthermore,estimatesofthestrength of therelationshipshaverangedfromsmalltolarge(Jacob,Stefan, and Guéguen2014; Morrisonet al.2011),rendering conclu-sionsabouttheelasticityof atmosphericeffects,animportant questionforretailexecutives,uncertain.Generalizableestimates of effect sizes are therefore badly needed. Previous reviews of atmospheric effects have, however, been limited to narra-tive or vote-counting methods(Bone andEllen1999; Turley and Milliman 2000), and generalized estimates among the
http://dx.doi.org/10.1016/j.jretai.2016.10.001
0022-4359/©2016TheAuthor(s).PublishedbyElsevierInc.onbehalfofNewYorkUniversity.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).
relationshipshereinvestigatedhavebeenreportedonlyforthe effectofmusiconpleasure(GarlinandOwen2006).In addi-tiontothelackofaggregatedeffectsizeestimates,theseearly summariesdatebackmorethan10years.
Thefirstcontributionofthisstudyisthereforetopresenta meta-analysis attemptingto calibrate the size of atmospheric effects onshopping outcomes.The second contribution isan attempttoaccountforbetween-studyvarianceineffectsizesand toinvestigateanumberofmoderatorsthatreflectstudydesign choicesmadebytheresearchers.Themoderatorsinclude samp-lingframe(studentsversuscustomers),gendersplit(lowversus highproportionoffemalesinasample),industrysetting(retail versusserviceversusonlinesettings),andexperimentaldesign (fictitiousversus actualenvironments).Through thesemeans, weaimtopresentretailerswithareliableguidetotheeffectsof atmosphericstimulionshoppingoutcomesbasedonananalysis ofdataaggregatedacrosstheresearchstream.
TheoreticalBackground
Theframeworkfor thismeta-analysis isdepictedinFig.1
togetherwith the investigated variables. It follows the extant literature in using the so-called stimulus-organism-response
paradigm (Mattila and Wirtz 2001; Mehrabian and Russell
1974).
ShoppingOutcomes
Frequently studied shopping outcomes at the organism
levelincludecustomers’emotionalreactionsandjudgmentsof satisfaction.Emotionalreactionsareconceptualizedasa combi-nationofarousalandpleasure.Arousalrepresentstheactivation dimensionandcanbedefinedastheperceiveddegreeof stimula-tion,whilepleasurerepresentsthevalencedimensionandrefers tothe perceived degreeof enjoyment(Donovan andRossiter 1982).Satisfactionreflectsanoverallevaluativejudgmentabout theshoppingexperience(MattilaandWirtz2001).Satisfaction isdistinct from pleasurein that it relatesto outward-looking judgmentsaboutexternalentitiessuchasastore’satmosphere (“Shoppinginthisstoreis apositiveexperience”,as adapted fromSayinetal.2015,p.5);whilepleasurereflectsasubjective, inwardfocus(“I’mexperiencingpleasantfeelings”).
Attheresponselevel,themostcommonlystudiedvariables includepurchase(Fiore,Yah,andYoh2000),visiting(Doucé and Janssens 2013), shopping (Broekemier, Marquardt, and Gentry2008),orpatronageintention(Grewaletal.2003). Stud-iesalsocaptureactualexpenditures(Sullivan2002).Together, thesevariablesreflecttheunderlyingobjectiveofcustomersto dobusinesswithanorganization,andareheresubsumedunder
behavioralintentions. AtmosphericStimuli
Theintegrationofpriorfindingsintoacommonframework necessitates a concentration on the most frequently studied variables.Amongthewidevarietyofinvestigatedatmospheric stimuli,music,scent,andcolorhavereceivedthemostresearch
attentionandarethereforethefocusofthisanalysis(Bellizzi, Crowley, andHasty 1983; Bone andEllen 1999; Garlin and Owen2006).
Music
Asanatmosphericstimulus,musicreferstohuman compo-sitionsfunctioningasanambient elementinthe consumption environment(GarlinandOwen2006).Atthemostbasiclevel,
musichasbeenstudiedbycomparingtheeffectsofthepresence andabsenceofmusic;that iscustomeremotions,satisfaction, andbehavioralintentionsarecomparedacrossconditionswhere musicispresentandwheremusicisabsent(e.g.,Grewaletal. 2003).
Authorsarguethat musiccanbeseenasacomplementary
productorservicefeaturethatisconsumedduringthepurchase processandisthereforelikelytoinfluenceshoppingoutcomes (Hui,Dubé,andChebat1997).Anotherexplanationfortheeffect ofmusiccomesfromoptimalarousaltheory,whichpositsthat peopleseektoaligntheircurrentlevelofarousaltoalevelthey findpersonallyoptimal(Berlyne1971).Customerswhoarein an“understimulated”statewillbeseekingheightenedarousal, whichtheymayrealizethrough the presence ofmusicin the shoppingenvironment(MattilaandWirtz2001).Sincearousal operatesasanamplifierofpositivein-storeexperiences(Oliver, Rust, and Varki 1997),downstream positive effects on plea-sure,satisfaction,andbehavioralintentionscanbeanticipated. Therefore,wehypothesize:
H1. The presence (versus absence) of music has apositive
effecton(a)arousal,(b)pleasure,(c)satisfactionand(d) behav-ioralintentions.
Scent
Ambientscentreferstoascentpresentintheenvironment thatdoesnotemanatefromaparticularobject(BoneandEllen 1999).Scenthasbeenemployedasanaturallyoccurring stimu-lus(e.g.,inbakeries)aswellasanartificiallyinducedstimulus toenhancestoreambience (Spangenbergetal.2006).Similar tomusic,scenteffectsareusuallymeasuredbycomparing
cus-tomers’ shoppingexperiences ina scentedenvironment with
thoseinascent-freeone(e.g.,DoucéandJanssens2013). Researchsuggeststhat,relativetoothersensorycues,scent isprocessedinamoreprimitiveportionofthebrain(Herzand Engen1996),andscentthereforerequireslittleornocognitive effort to enhance alertness, improve in-store experience, and promote positiveshopping outcomes (Bone andEllen1999). Studies havealso founda privilegedneural link between the olfactorynerveandthearearesponsibleforemotionalmemory (Herz2004).Thisisunderstoodasthephysiologicalexplanation for whysmell evokessignificantlystrongeremotional memo-riescomparedtothosetriggeredbyauditoryandvisualstimuli (Herz 2004).Therefore, when evoked in-store, such memory associationsmaystimulatepositiveemotionsandleadtoamore enjoyableshoppingexperience(BoneandEllen1999).Hence,
Fig.1.Meta-analyticframework.
H2. Thepresence(versusabsence)ofscenthasapositiveeffect on(a)arousal,(b)pleasure,(c)satisfactionand(d)behavioral intentions.
Color
Asanatmosphericvariable,colordescribesthevisual appear-anceof theconsumption environment(Bellizzi,Crowley,and Hasty1983).Scholarsusuallycomparenon-whitecolorsasthese canbeorderedbywavelength,i.e.,fromlongtoshort(Crowley 1993).Researchdefinesthosewithalongerwavelength—such asred,orange,andyellow—aswarmcolors,whilethosewith
a shorter wavelength—such as green, blue, and violet—are
describedascool(Crowley1993).Althoughwhiteisgenerally regardedasneutral,itissometimesascribedtothecoolcolor cat-egory(ChebatandMorrin2007).Accordingly,researchershave predominantlystudiedcolorbycomparingcustomers’internal dispositionsandbehavioralintentionsinresponse towarmor cool conditions in amanner analogous tothe present versus absentcomparisonusedinstudiesonmusicandscent(e.g.,van Rompayetal.2012).
Forconceptualizing coloreffects inretail settings, studies drawonphysiologicalandpsychologicalfindingsreportedfor human behaviorin general inresponse tocolor (Bagchi and Cheema2013;BellizziandHite1992).Physiologically,red (rel-ativetoblue)hasbeenfoundtobemoreactivatingintermsof bloodpressure,respiratoryrate,andskin-conductance. Psycho-logically,warmcolors(especiallyred)areseenasemotionally arousing,exciting,anddistracting,whilecoolcolors(especially blue)arelinkedtofeelingsofrelaxation,peacefulness,calmness, and pleasantness(for moredetailed summaries, see Bellizzi, Crowley,andHasty 1983 andLabrecque,Patrick, andMilne 2013).
Findings in the atmosphericsdomain mirror these results. Red(comparedtoblue)hasbeenlinkedtogreaterarousal,and
blue (compared to red) is reported to be more relaxing and
pleasantandtofacilitatepurchaseincidencewithinashopping
environment (Bagchi and Cheema 2013; Bellizzi and Hite
1992;Crowley1993).However,theoverallempiricalevidence remainsinconclusive,withsomestudiesbeingunableto iden-tifydifferencesinresponsetowarmandcoolcolors,andothers revealingeffectsoppositetothedirectionhypothesized(Babin, Hardesty,andSuter2003;BagchiandCheema2013;Bellizzi andHite1992;Cyr,Head,andLarios2010).Forourhypothesis, wefollowthemaintheoreticalperspectiveandpropose: H3. Warm (versuscool)colorshaveapositiveeffecton(a) arousal andanegativeeffecton(b) pleasure,(c)satisfaction, and(d)behavioralintentions.
Moderators
SubstantiveInfluences
Substantive influences relate to boundary conditions that describehowastimulusshouldbedesignedinordertomaximize theimpactofitspresence.Variationsinstimuliinvestigatedin thecontextofmusic,scent,andcolorcanbecategorizedas struc-turalandcongruitycharacteristics(BoneandEllen1999;Mattila andWirtz2001).Althoughthenumberofempirical operational-izationsofthesecharacteristicswasinsufficientforthemtobe includedandcomparedinthemeta-analysis,wedescribethem herebrieflyastheydofigureintheresearchwearereviewing.
Structural characteristics describe the ‘components’ of a stimulusandincludevalence,intensity,andcomplexity.Valence reflectswhetherastimuluscarriesapositiveornegative conno-tationandreflectsdistinctionssuchashappyversussadmusic, the bright-to-darkgradationof color,andthehedonicor utili-tariantoneofascent(Cheng,Wu,andYen2009;Hui,Dubé, andChebat 1997;Knasko1995).Stimulus intensity captures, forexample,thetempoandvolumeofmusic,theconcentration ofscentintheair,andthedegreeofsaturationofagivencolor (BoneandEllen1999;Cheng,Wu,andYen2009;Herrington
andCapella1996).Complexitydescribesthesimplicityversus elaborationofastimulus,including,forinstance,howeasyitis foracustomertofollowapieceofmusic(NorthandHargreaves 1996),toidentifyascent(plainversusblendedodors;Herrmann etal.2013),ortounderstandacolorscheme(plainversus gra-dientcolorcompositions).
Congruityisdefined as thefit of astimulusto theoverall storeambience.Sinceaperceivedincongruenceinterfereswith customers’informationprocessing,studiesprovide consistent evidencethatthefitofastimulustoeitherapart(e.g.,a prod-uctassortment)or toanentirestoreenvironmentenhancesits impactonshoppingoutcomes(MattilaandWirtz2001;Mitchell, Kahn,andKnasko1995;Spangenbergetal.2006).The same also applies tothe fit of a stimulustoindividual preferences (Broekemier,Marquardt, and Gentry 2008). In consequence, researcherstypicallyemploystimuliwithstructural character-isticsthat suitthe shoppingenvironment,as for instancewas implementedbyGrewaletal.(2003,p.262),whousedclassical music“becauseit‘fits’thecontextofluxurygoods”.
MethodologicalInfluences
Study designs used in research on atmospheric effects
reflect the differences in methodological decisions made by researchers. One objective of this research was to consider
the impact of these methodological decisions on the effect
sizes observedin the primary studiesand touse coded vari-ablesreflectingthesemethodologicaldecisionsinanattemptto accountforbetween-studyvarianceineffectsizes.Four poten-tialmoderatorswereincluded:samplingframe(studentversus customer),gendersplit(lowversushighproportionoffemalesin asample),industrysetting(retailversusserviceversusonline), andexperimentaldesign(actualversusfictitiousenvironments). Variationsbetweenstudentandcustomer(i.e.,non-student) samples maybe ascribed todifferences inpersonality
devel-opment. Specifically, students are regarded as “unfinished
personalities”(Carlson1971,p.212)withlessdefined prefer-ences.Peterson’s(2001)secondordermeta-analysisshowsthat effectsizesfromstudentsamplesfrequentlydifferfromthose derivedfromnon-studentsubjects.However,nosystematic pat-ternwithrespecttoeffectsizesordirectionofeffectsbetween students and customers could be firmly established. Expect-ingasimilarunsystematicvariation,we useanon-directional hypothesisandpropose:
H4. Theeffectsizesofatmosphericstimulionshopping out-comesvarybetweenstudent-andcustomer-basedsamples.
Although some authors have made suggestions regarding
musicandcolor(Anderssonetal.2012;vanRompayetal.2012), empiricalfindingsongendereffectshavebeenreportedmainly inrelationtoscent(BoneandEllen1999).Physiological evi-dencepointstowomenasbeingemotionallymoresensitiveand responsivetoscent(HerzandEngen1996).Yousemetal.(1999)
forinstancefindthatscentactivatesalargerareainawoman’s brainthaninaman’s.Intheatmosphericsdomain,Lehrneretal. (2000)indicatethatthepresenceofanambientorangescentin thewaitingroomof adentalsurgeryevokedmorepleasurein
womenthaninmen.Consequently,wepropose:
H5. The effectsizeof scenton(a)arousaland(b) on plea-sureislargerinsampleswithahigherproportionoffemalesas comparedtosampleswithalowerproportionoffemales.
Industrysetting(retail,service,oronline)mayaccountfor between-studyvarianceintheeffectsizeofatmosphericstimuli. Among investigations into settings, the differences between servicecontextsandproductretailsettingshavebeen particu-larlydiscussed(Bitner1990).Relativetoproducts,servicesare definedbytheirinherentlylargerdegreeofintangibility(Bitner 1990).Itissuggestedthat,intheabsenceof tangibleproduct features,environmentalcuesregardingservicesserveassubtle messagesthathelpcustomerstoknowwhattoexpectfroma cer-tainoffering(BoomsandBitner1982).Accordingly,customers maydependtoalargerextentonatmosphericstimuliwhenin contactwithaserviceenvironmentthaninaretailsetting.This leadstothefollowinghypothesis:
H6. Theeffectsizesof music,scent, andcolorarelargerin servicethaninretailsettings.
Finally, research designs inthe literature on experimental
atmospherics can be grouped into those based on fictitious
versusactual(i.e.,in-store)environments.Extantresearch pro-videsmixedfindings.Ontheonehand,BatesonandHui(1992)
find evidencefor the ecologicalvalidity of fictitious environ-ments. Onthe otherhand,fictitiousenvironmentscanleadto inflatedeffectsizesbecausetheyallowresearchersmore con-trolinisolatingextraneousfactorsthatmayhavebearingonthe hypothesizedrelationshipsandindirectingtheattentionofthe studyparticipants(Shadish,Cook,andCampbell2010).Thus, wehypothesize:
H7. Theeffectsizesofatmosphericstimulionshopping out-comesarelargerinfictitiousthaninactualenvironments.
Method
DatabaseDevelopment
Abibliographickeywordsearchwasconducted toidentify studies reporting on customers’ reactions tothe atmospheric stimulusvariablesasspecifiedinFig.1.Databasesusedincluded
EBSCO Business Source Complete, Science Direct, Emerald ManagementXtra,ABI/Inform,PsycINFO,GoogleScholarand theSocialScienceCitationIndex.Inaddition,areference analy-siswasconductedonprevioussummaries(BoneandEllen1999; GarlinandOwen2006;TurleyandMilliman2000),anda cita-tionanalysisofpertinentarticleswasmade(Bellizzi,Crowley, andHasty 1983; Milliman1982; Spangenberg, Crowley,and Henderson 1996). Unpublished work was also obtained by searchingtheSSRNdatabaseandGoogleScholar.
Astudywasincludedinthefinaldatasetif(1)music,scent, andcolorweremanipulatedexperimentally,(2) thestudywas independent (i.e., if the results of two different studies were derivedfromthesamesample,thestudywhichprovidedmore detail wasused), (3) themeasurementitem(s) of anoutcome variableaccuratelyreflectedourconstructspecification,and(4) aneffectsizeorsufficientstatisticalinformationwasprovided
foratleastoneoftherelationshipsspecifiedinFig.1.Astudy wasexcludediftheexperimentaldesignpreventedtheisolation ofeffectsforasinglestimulus(i.e.,multiplemanipulationswere notfullycrossed).
Thesearchidentified66studies(64articlespublishedin
aca-demic journals, 2 unpublished working papers) that met the
criteriaforinclusion.Theincludedstudiesreferredto74
inde-pendent samplesfrom thetimeperiod 1982to2016 (March)
witha combined totalN of 15,621 respondents. Overall, the averagesamplehadamedianageof33.2years(student sam-ples=22.0years,generalcustomersamples=39.1years)anda mediangendersplitof 61.7%infavorof femalerespondents. Followingtheremovaloftwooutliers,1weobtainedafinaldata setof135studyeffects.
EffectSize Computation
Common meta-analytic guidelines were followed for the
experimentalstudiesreviewed inthemeta-analysis. First,the standardizedmeandifferences(Cohen’sd)werecomputed, fol-lowed by conversion of these to correlation coefficients (De Matos, Henrique, andRossi 2007). The choiceof using cor-relationcoefficientsastheeffect-sizemetricwasbasedonthe relative ease with which these can be interpreted, and their generalutility as astandardmeta-analytic metric inthe mar-keting literature(Gelbrich and Roschk2011; Palmatier etal. 2006).
When calculating r, a positive (negative) value indicates
that—inthecaseof musicandscent—thepresencecompared
totheabsenceconditionincreases(decreases)thevalueofthe outcome variable. Likewise, in the case of color, a positive (negative)valueindicatesthatthewarm(i.e.,red,orangeand yel-low)comparedtothecool(i.e.,green,blue,violet,andwhite) color schemeincreases (decreases) the value of the outcome variable.
Integration
Correlationcoefficientsweredirectlyderivedfromthe stud-iesathand orwerecalculatedthroughstatistical datasuchas Student’st,η2,andF-ratios(Cohen1988;Glass,McGaw,and Smith1981).In somecases,the studyauthorsprovidedmore thanonemeasure of correlationfor the samerelationship by analyzingdifferentresponsemeasuresormultipleexperimental comparisonsbetweenastimulus-absentgroup(e.g.,nomusic) andastimulus-presentgroup(e.g.,differentstylesofmusic).If thiswasthecase,weaveragedtheeffectsizesinordertoavoid bias from the overrepresentationof samples (Palmatieret al. 2006).
Theeffectsizeswerenextadjustedforreliabilitytocorrect
for attenuation from randommeasurement error (Hunter and
Schmidt2004).Forstudiesthatdidnotprovidereliabilityindices orused single-itemmeasures,themeansample size-weighted
1 Theoutlierswereexcludedbecausetheylayoutsideoftherangeofthree
standarddeviationstotherespectivesample-sizeweightedmeaneffectsize.
reliability across allstudieswasused. Wethen computedthe sample-sizeweightedmeansofallavailableeffectsizeestimates foreachrelationship,whichwerefertoasr(HunterandSchmidt 2004).
CalculationofAssociatedStatistics
Wecalculatedthe‘fail-safeN’ statistictofind theaverage numberofdiscardednullresultsnecessarytorenderthe relation-shipsnon-significant.Forameaningful,robustresultRosenthal (1979)suggeststhattheobtainedfail-safeNshouldbegreater thanorequaltofivetimesthenumberofobservationsplus10 (referredtoas‘requiredfail-safeN’).Heterogeneityofthe inte-grated effectsizes(i.e.,between-studyvariance)wasassessed via theQ-Statistic(HedgesandOlkin1985).Moderator anal-yses are justifiedwhen between-study variance issignificant, indicating that resultsof extant studiesdo not convergeona
commonpopulationvalue.
Coding
Two judges independentlycoded the dependent and
mod-erating variables based on the theoretical definitions of the constructs.Thetwojudgesconcurredonmorethan95percent of independentlydetermined codingdecisions;disagreements wereresolvedbydiscussion.Atableofthecodingsisprovided intheAppendix.
Results
AtmosphericEffects
Theresultsoftheimpactofatmosphericstimulionarousal, pleasure, satisfaction,andbehavioral intentionsare shownin
Table1.FollowingCohen’s(1988,p.82)criteria,aneffectsize of .10canbeconsideredas small,.30asmedium,and.50as large.
Ourfindingsindicatedthatthepresenceofmusic(compared toitsabsence)wassignificantlyandpositivelyrelatedtopleasure (r=.098),tosatisfaction(r=.226),andtobehavioralintentions (r=.130).Fail-safeNvaluesexceededtherequiredfail-safeN
(see Table 1). Presence of musicdid not significantly affect
arousal.Hence,H1b,H1c,andH1dweresupportedwhileH1a
wasnot.
Scentwassignificantlyrelatedtoalloutcomevariables. Pres-enceofscentledtohigherarousal(r=.079),pleasure(r=.093), satisfaction (r=.183), and behavioral intentions (r=.054) as comparedtoscent-absentconditions,supportingH2a,H2b,H2c, andH2d.Theeffectofscentonarousalshouldbetreatedwith cautionbecausetheobtainedfail-safeNwasbelowthenormative value.
Color schemessignificantly affectedbotharousal and
sat-isfaction. As hypothesized, these effects were in opposing
directions.Warmascomparedtocoolcolorschemesproduced higherarousal(r=.157)butlowersatisfaction(r=−.254).Both effectshadrobustfail-safeNs.Colorwasnotsignificantlyrelated toeitherpleasureorbehavioralintention.Thus,H3aandH3c
Table1
Meta-analyticresultsfortheinfluenceoftheatmosphericstimuliontheshoppingoutcomes.
95%CI Fail-safeN
k N r Lowerbound Upperbound SD p Obtained Required Min Max Q-Statistica
Music(present=1,absent=0)
Arousal 11 2441 .042 −.021 .106 .108 .193 – – −.272 .265 28.87**
Pleasure 12 2489 .098 .043 .153 .097 <.001 77 70 .000 .334 24.47*
Satisfaction 5 877 .226 .170 .282 .064 <.001 63 35 .151 .354 4.03ns
Behavioralintentions 19 3116 .130 .078 .181 .114 <.001 314 105 −.059 .395 43.03***
Scent(present=1,absent=0)
Arousal 14 2763 .079 .033 .124 .087 <.001 49 80 −.071 .277 20.96ns
Pleasure 18 3793 .093 .039 .148 .117 <.001 184 100 −.193 .331 53.36***
Satisfaction 4 938 .183 .099 .268 .086 <.001 30 30 .031 .293 7.40ns
Behavioralintentions 21 5280 .054 .013 .094 .094 .009 129 115 −.154 .458 49.32***
Color(warm=1,cool=0)
Arousal 9 1724 .157 .020 .294 .210 .025 156 55 −.198 .522 85.69***
Pleasure 8 1499 .031 −.129 .192 .232 .704 – – −.380 .446 88.46***
Satisfaction 7 947 −.254 −.299 −.210 .060 <.001 79 45 −.320 −.105 3.52ns
Behavioralintentions 7 1225 −.057 −.190 .076 .179 .402 – – −.292 .301 40.47***
Note:k,numberofeffects;N,totalsamplesize;r,samplesize-weightedmeancorrelationcoefficientadjustedforreliability;SD,standarddeviation.Obtained/required fail-safeNarenotcalculatedforinsignificantrelationships(indicatedbyadashedline).Figuresinboldindicatesignificanceand,withregardstothefail-safeN,
robustcorrelations.ns,notsignificant.
a df=k−1.
* p<.05.
** p<.01.
***p<.001.
ModeratingEffects
ThelastcolumninTable 1depicts the Q-statistic.Results
indicated that between study variance was non-significant
for four relationships (music→satisfaction, scent→arousal, scent→satisfaction, color→satisfaction). Forthe remaining eightrelationships,theresults weresignificant,indicatingthe appropriatenessofmoderatoranalysesinattemptingtoexplain between-studyvariance.Moderatoranalyseswereconductedby partitioningstudiesaccordingtocodedlevelsofstudy character-isticsandcalculatingrwithinsubgroups.Forgender,weusedthe medianproportionoffemalestocreatetwosub-groups(≤61.7% versus >61.7%). Subsequently, a t-test for small samples, as describedbyWiner(1971),wasused(Palmatieretal.2006).2 Itshouldbenotedthatthesmallnumberofstudyeffectsrender thesetestslowinstatisticalpower,andthereforeTypeIIerror ismorelikelythanTypeIerror(Brown1996).Theresultsof thesemoderatoranalysesaredepictedinTables2and3,andare reportedbelow.
For most relationships, effect sizes did not differ
signifi-cantlybetweenstudentandnon-studentsamples.Only music
hadstrongereffectsonbehavioralintentionsinnon-student sam-ples(r=.169)thaninstudentsamples(r=.069).Overall,H4was notconfirmed.
Withregardtogendereffects,significantlylargereffectsizes
were found for the relationship between scent and pleasure
2 Thesmallsample t-statisticisbased on thesinglereliability corrected
correlationcoefficientsandiscalculatedbyt= (¯xa−¯xb)
(s2 a/na)+(s2b/nb) ,withdf= (V+W)2 [V2/(n a−1)]+[W2/(nb−1)],whereV=s 2 a/naandW=s2b/nb.
amongsampleswithahigherproportionoffemales(r=.176) comparedtothosewithalowerproportion(r=.069).Asimilar between-groupdifferencewasevidentforarousal,butwasnot statisticallysignificant.Hence,H5bwassupportedwhileH5a
wasnot.Additionally,thepresenceofmusichadasignificantly strongereffectonbehavioralintentionamongsampleswitha lowerproportionoffemales(r=.165)ascomparedtothosewith ahigherproportion(r=.006).
The industry settinganalyses indicatedatendency toward
stronger music and scent effects in service as compared to
retailsettings.However,theresultsweresignificantonlyforthe effectofmusiconbehavioralintentions(servicer=.200,retail
r=.112),andtheyweremarginallysignificantfortheeffectof musiconpleasure(servicer=.263,retailr=.073)andfor the effectofscentonbehavioralintentions(servicer=.342,retail
r=.041).Thus,H6waspartiallysupportedformusicandscent. Theresultsalsoshowedsignificantlylargereffectsizesinonline comparedtoretailsettingsfortherelationshipsbetweencolor andarousal(onliner=.401,retailr=.040)andbetweencolor andpleasure(onliner=.208,retailr=−.059).
Finally,sevenrelationshipshadasufficientnumberof avail-ableeffectstomakeitpossibletocompareeffectsizesbetween experimentscarriedoutinactualversusfictitiousenvironments. Resultsshowedthat noneoftheanalyzeddifferenceswereof statisticalsignificanceandH7wasnotsupported.
GeneralDiscussion
Inthelight ofthevolumeoftheliteraturethatisuncertain abouttheimpactofatmosphericstimulionshoppingoutcomes, thefirst objectiveofthismeta-analysiswastocalibrateeffect sizes. On an aggregate level, the results revealed predictable
Table2
Meta-analyticmoderationresultsfortheinfluenceofthemoderatorsontheeffectsizesbetweenatmosphericstimuliandshoppingoutcomes.
Samplingframe Gendersplit(%female)
Level r k LB UB t df p Level r k LB UB t df p
Music(present=1,absent=0)
Arousal Customer .045 6 −.063 .152 0.53 8 .305 ≤61.7 .081 4 −.037 .199 1.29 4 .133
Student .041 5 −.032 .114 >61.7 .033 5 −.028 .094
Pleasure Customer .099 6 −.011 .208 0.01 7 .496 ≤61.7 .065 4 −.043 .173 0.05 6 .481
Student .098 6 .055 .140 >61.7 .115 5 .037 .194
Satisfaction
Behavioralintentions Customer .169 10 .102 .235 1.76 17 .048 ≤61.7 .165 8 .089 .240 3.78 9 .002
Student .069 9 .006 .132 >61.7 .006 3 −.044 .056
Scent(present=1,absent=0)
Arousal Customer .084 9 .030 .138 0.25 8 .404 ≤61.7 .059 7 .004 .115 1.25 8 .124 Student .058 5 −.028 .144 >61.7 .132 4 .061 .203 Pleasure Customer .087 13 .025 .149 0.69 8 .255 ≤61.7 .069 8 .000 .138 3.00 11 .006 Student .130 5 .016 .245 >61.7 .176 5 .124 .227 Satisfaction Customer .158 2 .148 .168 0.04 1 .489 ≤61.7 .213 2 .046 .381 0.04 1 .489 Student .213 2 .046 .381 >61.7 .158 2 .148 .168
Behavioralintentions Customer .042 14 −.011 .094 0.67 19 .254 ≤61.7 .091 5 .018 .165 1.47 7 .093
Student .096 7 .054 .138 >61.7 −.002 6 −.048 .044
Color(warm=1,cool=0)
Arousal Customer .130 5 −.041 .301 0.37 7 .362 ≤61.7 .286 3 .073 .499 0.90 4 .209 Student .237 4 .013 .462 >61.7 .050 4 −.104 .204 Pleasure Customer .087 3 −.077 .250 1.45 6 .098 ≤61.7 .239 2 −.069 .547 1.80 2 .107 Student −.038 5 −.294 .219 >61.7 −.035 4 −.146 .077 Satisfaction Customer −.260 3 −.317 −.203 0.05 5 .483 Student −.233 4 −.315 −.151
Behavioralintentions Customer −.075 4 −.210 .060 0.49 3 .330 ≤61.7 .087 2 −.102 .276 2.29 2 .074
Student −.025 3 −.286 .235 >61.7 −.150 4 −.280 −.020
Note:r,samplesize-weightedmeancorrelationcoefficientadjustedforreliability;k,numberofeffects;LB/UB,lower/upperboundofthe95%CI.Duetomissing information,thecombinednumberofobservationacrosssubgroupscanbelowerthanthetotalfromTable1.Subgroupcomparisonsareconductedifatleasttwo observationsineachsubgroupcouldbeidentified.Significantdifferencesarehighlightedinbold.
patternsfor theeffectsof music(presencevs.absence),scent (presencevs.absence)andcolors(warmvs.cool)onshopping outcomes.Theeffectsizesforthesignificantrelationshipswere estimated atan average rthat spanned from.054 to.254 (in absoluteterms). Thesevalues indicatedthat the relationships weresmall-to-mediuminstrength,whichcanbeseenas reflect-ing the subtlenature of the atmosphericstimuli. Overall, the resultsprovideresearcherswithaquantitativesummarythat fig-uresthepatternoftheeffectsthathavebeenincludedwithinthe researchstream.Specificinsightsformusic,scent,andcolorare asfollows:
Formusic,theresultsrevealedsignificantandpositive aggre-gateeffectsonpleasure,satisfaction,andbehavioralintentions. A link witharousal could not be established. It appearsthat musicis,againstabackgroundnoisesuchascustomerchatter, toosubtletotriggercustomers’arousal.Hence,musical stimula-tionmaybeseenasapleasure-inducingsubstitutefordistracting in-storesounds,whichwillenhancetheshoppingexperience. Furthermore,thegeneralizedestimatesreportedhereaddtoand updatethoseprovidedby GarlinandOwen(2006),thus
pro-vidingresearcherswithtwocomplementarysummarieswhich
maybeusedasreferencesforfurtherresearchintotheeffectsof atmosphericmusic.
At a meta-analytic level, scent positively influences cus-tomers’ pleasure, satisfaction, and behavioral intentions. A positive effectof scenton arousalwas also evident, withthe limitationthatitsrobustnessagainstdiscardednull-resultscould notbeverified.Thesefindingssuggestaneedtoreconsiderthe conclusioninBoneandEllen’s(1999)review,whichstatesthat “countingonsuch [scent]effects isanunwise strategyatthis point in time” (p. 259). Instead, scentmay be regarded as a reliablemeansforenhancingtheshoppingexperience.
The integratedeffects showcolortobebipartiteinnature. Whilewarmcolorscausehigherlevelsofarousalthancool col-ors,theoppositeistrueforsatisfaction.Researcherscandrawon theseresultswhenconceptualizingcoloreffectsforspecific in-storebehavioralphenomena.Forinstance,Labrecque,Patrick, andMilne(2013)discussthearousing yetdissatisfyingeffect of warm(vs.cool)colorsontheperceptionofcustomerswho arewaitinginthecheck-outline.Ourdatadidnotconfirma sig-nificanteffectofcoloronpleasureoronbehavioralintentions. Opposingcoloreffectsmay,duetointerrelationshipsamongthe
Table3
Meta-analyticmoderationresultsfortheinfluenceofthemoderatorsontheeffectsizesbetweenatmosphericstimuliandshoppingoutcomes.
Industrysetting Experimentaldesign
Level r k LB UB ta df p Level r k LB UB t df p
Music(present=1,absent=0)
Arousal Retail .022 5 −.084 .129 1.31 2 .160 Actual .041 7 −.055 .138 0.82 8 .219
Service .142 2 −.032 .316 .37 2 .374 Fictitious .043 4 −.040 .126
Online .043 4 −.040 .126 1.26 7 .124
Pleasure Retail .073 6 −.020 .166 2.79 2 .054 Actual .104 7 .005 .203 0.41 8 .346
Service .263 2 .152 .374 2.45 2 .067 Fictitious .093 5 .049 .138
Online .094 4 .044 .145 .76 8 .234
Satisfaction Retail .236 2 .200 .272 .51 1 .349
Service .273 2 .153 .393
Behavioralintentions Retail .112 8 .031 .194 2.03 13 .032 Actual .169 10 .102 .235 1.53 16 .073
Service .200 7 .142 .258 4.40 7 .002 Fictitious .076 8 .007 .144
Online .022 3 −.036 .081 1.64 9 .067
Scent(present=1,absent=0)
Arousal Retail .107 9 .057 .158 .14 6 .446 Actual .089 7 .029 .149 1.17 9 .135
Service .037 4 −.042 .116 Fictitious .051 6 −.024 .126
Pleasure Retail .078 12 .008 .148 .11 6 .459 Actual .090 11 .023 .157 0.09 12 .466
Service .120 5 .031 .208 Fictitious .115 6 .011 .219
Satisfaction Actual .158 2 .148 .168 0.04 1 .489
Fictitious .213 2 .046 .381
Behavioralintentions Retail .041 17 .006 .076 3.88 1 .080 Actual .051 13 .000 .101 0.27 15 .395
Service .342 2 .191 .493 Fictitious .064 7 −.012 .140
Color(warm=1,cool=0)
Arousal Retail .040 4 −.096 .177 .78 2 .260 Service .088 2 −.122 .297 1.89 1 .155 Online .401 3 .283 .518 3.63 5 .008 Pleasure Retail −.059 4 −.219 .101 Service Online .208 3 −.095 .511 2.33 4 .040 Satisfaction Retail −.237 2 −.279 −.196 .67 1 .311 Service −.277 2 −.346 −.207 .27 2 .405 Online −.179 3 −.292 −.066 1.11 3 .174
Behavioralintentions Retail −.094 5 −.216 .027 Service
Online .037 2 −.280 .354 1.14 1 .229
Note:r,samplesize-weightedmeancorrelationcoefficientadjustedforreliability;k,numberofeffects;LB/UB,lower/upperboundofthe95%CI.Duetomissing information,thecombinednumberofobservationacrosssubgroupscanbelowerthanthetotalfromTable1.Subgroupcomparisonsareconductedifatleasttwo observationsineachsubgroupcouldbeidentified.Significantdifferencesarehighlightedinbold.
a Subgroupcomparisonsinthefollowingsequence:retailvs.service,servicevs.online,onlinevs.retail.
shoppingoutcomes,indirectlyexhibitajointeffectonpleasure andonbehavioralintentions,whichthuscanceleachotherout. Thesecond objective of thismeta-analysis wastoattempt toexplainbetween-studyvarianceineffectsizeestimates.The resultsofprimarystudiesshowsignificanteffectsizevariations forthe majorityof relationships,indicatingthat the effective-nessofatmosphericstimuliislikelytodependonthespecific contextinwhichtheyareemployed.Partofthisvariationcanbe accountedforbytheresearchers’studydesignchoices. Specif-ically,gendersplitandindustrysettingprovide the following insights.
Forgender,themoderationresultsindicatethatwomenderive morepleasurefromscentthanmen.Thisconfirms physiolog-ical research on scent (Yousem et al. 1999) and encourages researcherstopayattentiontogenderdifferencesinscenteffects.
Itfurthersuggeststhatpriorresearchusingfemale-onlysamples (Morrison etal.2011) may haveexperienced inflatedeffects for this particular relationship. Additionally, our data shows that musichas a weakerimpact on the behavioral intentions ofwomenthanonthoseofmen,confirmingrecentfindingsin theatmosphericsliterature(Anderssonetal.2012).
With regard to industry setting,the effect sizes for music andscenttended tobestronger inservice comparedtoretail settings. Thisfindingisinlinewithpriorresearchsuggesting thatatmosphericstimuli,as subtlesignalsforcustomers,gain greatersaliencewhentangibleproductcuesareabsent(Bitner 1990). Asimilarlogic mayalso explain the largereffects of coloron arousalandpleasureinonline as comparedtoretail settings.Additionally,alackofdataregardingscenteffectsin online settingsreflects theneed for abetterunderstandingof
how virtual environmentscan leveragescent effects (e.g., by scentingproductpackages).
Between-studyvarianceofeffectsizescouldnotbeexplained byeithersamplingframeorbyexperimentaldesign.Thus,the effectsizesappearnottovarysystematicallyinrelationtothese methodologicalspecifications.Oneexceptionwasaweakerlink betweenmusicandbehavioralintentionsinstudentthanin cus-tomersamples,whichmaybeseenasapreliminaryindication thatthereislessexplicablevarianceinstudent(versuscustomer) behavioralintentions.
In addition to the heterogeneity of effect size estimates, moderatoranalysisalsoindicatedthatbetween-studyvariance wasnotsignificantforfourrelationships:music→satisfaction, scent→arousal,scent→satisfaction,andcolor→satisfaction. Fortheserelationships,extant primarystudiesappearto con-verge of what is tentatively, and pending further research, consideredtorepresentapopulationestimateofthe elasticity oftheatmosphericeffects.
Managerially, establishing reliable effect patterns has the potential to provide retail executives withthe necessary
pre-dictability over the uncertainty currently conveyed by the
literatureand thustoenable themto purposefullyimplement music, scent,andcolortoenhancetheir customers’shopping
experience. In doing this, two aspects need consideration.
First, congruence of the stimuli with the environment needs tobeensuredotherwisethestimulimaycauseadverseeffects (Mitchell, Kahn, and Knasko 1995). Second, the small-to-medium size of the effects suggests that the stimuli are of a subtlenatureandthusshouldbeconsideredaslong-term strate-gies.Specifically,musicandscentfostertheshoppingoutcomes withtheexceptionofarousal.Musicoffersthegreatest poten-tialforbeingtailoredtothepurchasesetting,whilescentbears the advantage that pleasant scentsmay occur naturally (e.g., roastedcoffeebeans)andcanbeventedtoaligningstoreareas. Further,warmcolorscausehigher levelsof arousalthancool colors, while the opposite is true for satisfaction. To benefit fromtheseeffects,warmcolorsmaybefavoredfornewproduct aisles,leveraging ontheir arousal property,while coolcolors maybepreferredforcomplainthandlingdesks,leveraging on theirsatisfyingproperty,forinstance.
Finally,two mainlimitationsthatofferpotentialfor future researcharediscussed.First,considerablevariationin concep-tualframeworksandinconsistentevidencepreclude definitive conclusionsonthecausalpriorityamongarousal,pleasure, sat-isfaction,andbehavioral intentions(BabinandDarden 1996; ChebatandMichon2003;DonovanandRossiter1982;Morrison et al. 2011). Therefore, our analysis is based on bivariate
correlations. This choice was due to the main study focus,
i.e., calibrating effect sizes to understandtheir direction and
strength, and accounting for moderators to understand
spe-cificboundaryconditions.Therefore,futureworkisencouraged to test for causal effects using larger meta-analytic datasets
as downstream sequential effects relate to many additional
types of exogenous influences besides those of atmospheric
stimuli.
Second, the meta-analytic design relies on secondary
data and thus only those relationships for which there was
a sufficient quantity of empirical evidence were included.
In particular, substantive moderators (valence, intensity, and complexity) could not be integrated and therefore represent an under-researched field worthy of future investigation and meta-analytic integration. Further, examining the extent to whichatmosphericstimulivariablesaffectnon-purchaserelated behavior, like in-store product trial participation, would be worthexploring.
Executivesummary
Retailers use subtle atmospheric stimuli to purposefully
enhance customers’shopping experience. Academicresearch
hasstudiedatmosphericstimuliwithaparticularemphasison music, scent, and color in a large variety of study designs. Thefindingsofthisvoluminousbodyofwork,spanningmore than 30 years, are however inconclusive, rendering reliable predictions about the stimuli’s effects on shopping outcomes uncertain. Wetherefore synthesizedthe quantitativeevidence from66experimentalstudiesandexaminedtheeffectsofmusic (presence vs.absence),scent(presencevs.absence),and col-ors(warm:red,orange,andyellowvs.cool:green,blue,violet, andwhite)onshoppingoutcomes.Shoppingoutcomescomprise customers’arousal(emotionalstimulation),pleasure(emotional enjoyment), satisfaction (evaluation of the shopping experi-ence), and behavioral intentions (e.g., purchase- or visiting intentions).
Resultsoffertwo keyinsightsrelevant toretailexecutives. First,inthelightofthecurrentuncertaintyaboutthestimuli’s atmosphericeffects,thefindingsshowpredictablepatternsfor
how music, scent, and color impact on shopping outcomes.
Morespecifically,environmentsinwhicheithermusicorscent is present, yield higher pleasure, satisfaction, andbehavioral
intention ratings compared withenvironments inwhich such
conditions are absent. Warm colors produce higher levels of arousalthancoolcolors,whiletheoppositeistruewithrespect tolevelsofsatisfaction.Itisimportanttonotethatthestimuli typically shouldfit theconsumptioncontext.Retailmanagers can use these findings as a guide,based as theyare on data aggregated acrossthe researchstream,toenhancecustomers’ shoppingexperience.Asonesuchstimulus,musicoffersa flex-ible tool that easily can be tailored tothe shopping context, pleasantscentsmayoccurnaturallyandcanbeventedto align-ingstoreareas,andcolorscanbeusedspecificallyinthoseareas whereeither arousal (e.g.,newproduct aisles)or satisfaction (e.g.,complainthandlingareas)shouldbetriggered.
Second,thesizeoftheeffectofstimulionshoppingoutcomes variedacrossstudies.Weattemptedtoexplainthesevariations
by conducting subgroupanalyses using coded moderators to
reflectthestudydesigndecisions.Resultsshowed,forinstance, atendencytowardlargermusicandscenteffectsinserviceas comparedtoretailenvironments.Wealsofoundevidence indi-catingthatwomenderivemorepleasurefromscentthanmen. Effect sizevariationspointtowardthe contextdependenceof atmospheric effects, justifying the efforts of marketing intel-ligence todesignandimplement thestimuliaccording tothe
individual consumption environment. Overall, the average strength of the relationships across contexts was small-to-medium,whichcanbeinterpretedasareflectionofthesubtle nature of stimuli. Therefore, their implementation should be consideredasalong-termstrategy.
FinancialSupport
Theauthorsreceivednofinancialsupportortechnical assis-tancefortheresearchand/orauthorshipofthismanuscript.
Acknowledgments
Theauthorsareverygratefultothreeanonymousreviewers andtheeditorfortheirhelpfulandconstructivecomments.The firstauthoralsolikestothankWendyTraboldforherconstant supportduringthedevelopmentofthisproject.
Appendix. Codedcharacteristicsofincludedsamples
Sample(inalphabeticalorder) Independentvariables Dependentvariables Moderatingvariables Music Scent Color Arousal Pleasure Satisfaction Behavioral
intentions Sampling frame Gender split(% female) Industry setting Experimental design
1 AlpertandAlpert(1990) × × × Student NA Retail Fictitious
2 Anderssonetal.(2012–Study1) × × × × Customer 43.3 Retail Actual
3 Anderssonetal.(2012–Study2) × × × × Customer 56.5 Retail Actual
4 Babin,Hardesty,andSuter(2003) × × × × Customer 100.0 Retail Fictitious
5 BagchiandCheema(2013– Study3)
× × Customer 59.0 Online Fictitious
6 Barlietal.(2012) × × Customer 52.0 Retail Actual
7 Baron(1997) × × Customer NA Retail Actual
8 BellizziandHite(1992–Study 1)
× × Customer 100.0 Retail Fictitious
9 BellizziandHite(1992–Study 2)
× × × × × Student NA Retail Fictitious
10 Bouzaabia(2014) × × × × Customer NA Retail Actual
11 Bramley,Dibben,andRowe (2016)
× × Student 64.2 Online Fictitious
12 ChebatandMichon(2003) × × × Customer 57.0 Retail Actual
13 ChebatandMorrin(2007) × × × Customer 64.6 Retail Actual
14 Chebat,Morrin,andChebat (2009)
× × Customer NA Retail Actual
15 Cheng,Wu,andYen(2009) × × × Customer NA Online Fictitious
16 Cyr,Head,andLarios(2009– Canadiansample)
× × Student 76.7 Online Fictitious
17 Cyr,Head,andLarios(2009– Germansample)
× × Student 66.7 Online Fictitious
18 Cyr,Head,andLarios(2009– Japanesesample)
× × Student 86.7 Online Fictitious
19 deWijkandZijlstra(2012) × × × Customer 59.1 NA Fictitious
20 Dijkstra,Pieterse,andPruyn (2008–Study2)
× × Student 63.6 Service Fictitious
21 DoucéandJanssens(2013) × × × × Customer 89.7 Retail Actual
22 Fiore,Yah,andYoh(2000) × × × × Student 100.0 Retail Fictitious
23 Grewaletal.(2003) × × × Student 47.0 Retail Fictitious
24 GuégenandPetr(2006) × × Customer NA Service Actual
25 Harrington,Ottenbacher,and Treuter(2015)
× × Customer NA Service Actual
26 HerringtonandCapella(1996) × × Customer 20.0 Retail Actual
27 Herrmannetal.(2013–Study1) × × Customer NA Retail Actual
28 Herrmannetal.(2013–Study3) × × Student NA Retail Fictitious
29 Hui,Dubé,andChebat(1997) × × × Student 49.1 Service Fictitious
30 Jacob,Stefan,andGuéguen (2014)
× × Customer NA Service Actual
31 KimandLennon(2012) × × × Student 100.0 Online Fictitious
32 Kim,Kim,andLennon(2009) × × × Student 100.0 Online Fictitious
33 Knasko(1995) × × × Student NA Retail Fictitious
34 Lehrneretal.(2000–Female sample)
Appendix(Continued)
Sample(inalphabeticalorder) Independentvariables Dependentvariables Moderatingvariables Music Scent Color Arousal Pleasure Satisfaction Behavioral
intentions Sampling frame Gender split(% female) Industry setting Experimental design
35 Lehrneretal.(2000–Male sample)
× × × Customer 0.0 Service Actual
36 Lehrneretal.(2005) × × × × Customer 49.5 Service Actual
37 Lorenzo-Romero,Gómez-Borja, andMollá-Descals(2011)
× × × × × Student NA Online Fictitious
38 LudvigsonandRottman(1989) × × Student 70.8 NA Fictitious
39 Madzharov,Block,andMorrin (2015–Study3)
× × Customer NA Retail Actual
40 MattilaandWirtz(2001) × × × × × × Customer 75.0 Retail Actual
41 McDonnell(2007) × × Customer 45.0 Service Actual
42 McGrath,Aronow,andShotwell (2015)
× × Customer NA Retail Actual
43 MichonandChebat(2007) × × Customer 62.0 Retail Actual
44 Michon,Chebat,andTurley (2005)
× × Customer NA Retail Actual
45 Middlestadt(1990) × × Student 100.0 Retail Fictitious
46 Milliman(1982) × × Customer NA Retail Actual
47 Mitchell,Kahn,andKnasko (1995–Study2)
× × × Customer NA Retail Fictitious
48 Moore(2014) × × Student 54.3 Retail Fictitious
49 MorrinandChebat(2005) × × Customer 61.7 Retail Actual
50 MorrinandRatneshwar(2000) × × × Student NA Retail Fictitious
51 Morrisonetal.(2011) × × × × × Customer 100.0 Retail Actual
52 NorthandHargreaves(1996) × × Student 62.0 NA Fictitious
53 North,Shilcock,andHargreaves (2003)
× × Customer 50.0 Service Actual
54 Novak,LaLopa,andNovak (2010)
× × × Student 62.8 Service Actual
55 Parsons(2009) × × Customer NA Retail Fictitious
56 Poon(2014) × × × Student 54.3 Retail Fictitious
57 Price(2010) × × × Student 71.5 Online Fictitious
58 Sayinetal.(2015–Study2) × × Student NA Service Fictitious
59 Sayinetal.(2015–Study3) × × Student NA Service Fictitious
60 Sayinetal.(2015–Study4) × × Student NA Service Fictitious
61 SchiffersteinandBlok(2002) × × Customer NA Retail Actual
62 Schifferstein,Talke,and Oudshoorn(2011)
× × × Customer 50.1 Service Actual
63 Spangenberg,Crowley,and Henderson(1996)
× × × Student 46.0 Retail Fictitious
64 Spangenbergetal.(2005) × × × × × Student 50.7 Retail Fictitious
65 Sullivan(2002) × × Customer NA Service Actual
66 vanHagenetal.(2008) × × × Student 50.0 Service Fictitious
67 vanRompayetal.(2012) × × × × Customer 61.8 Retail Fictitious
68 VinitzkyandMazursky(2011) × × Student 29.8 NA Fictitious
69 Wilson(2003) × × Customer 54.6 Service Actual
70 Wu,Cheng,andYen(2008) × × × × × Student 49.0 Online Fictitious
71 YalchandSpangenberg(1988) × × × × Customer NA Retail Actual
72 YalchandSpangenberg(1993) × × × Customer 68.6 Retail Actual
73 Yildirimetal.(2012) × × Customer 100.0 Service Fictitious
74 Yildirim,Akalin-Baskayab,and Hidayetoglu(2007)
× × Customer 51.0 Service Actual
Note:NA,notavailable(i.e.,characteristicsthatwereeitherabsentfromornotcodeableinthestudies).Individualstudiesmaycontaindependentvariableswhich arenotconsideredinthisanalysissincethestatisticaldatawasnotsufficienttocalculateaneffectsize.Areferencelistoftheincludedstudiesisavailablefromthe authorsuponrequest.
References
Andersson,PernilleK.,PerKristensson,ErikWästlundandAndersGustafsson (2012),“LettheMusicPlayorNot:TheInfluenceofBackgroundMusicon ConsumerBehavior,”JournalofRetailingandConsumerServices,19(6), 553–60.
Babin,BarryJ.andWilliamR.Darden(1996),“GoodandBadShoppingVibes: SpendingandPatronageSatisfaction,”JournalofBusinessResearch,35(3), 201–6.
Babin,BarryJ.,DavidM.HardestyandTracyA.Suter(2003),“Colorand ShoppingIntentions:TheInterveningEffectofPriceFairnessandPerceived Affect,”JournalofBusinessResearch,56(7),541–51.
Bagchi,RajeshandAmarCheema(2013),“TheEffectofRedBackground ColoronWillingness-to-Pay:TheModeratingRoleofSellingMechanism,” JournalofConsumerResearch,39(5),947–60.
Bateson,JohnE.G.andMichaelK.Hui(1992),“TheEcologicalValidityof Pho-tographicSlidesandVideotapesinSimulatingtheServiceSetting,”Journal ofConsumerResearch,19(2),271–81.
Bellizzi,JosephA.,AynE.CrowleyandRonaldW.Hasty(1983),“TheEffects ofColorinStoreDesign,”JournalofRetailing,59(1),21–45.
Bellizzi,JosephA.andRobertE.Hite(1992),“EnvironmentalColor, Con-sumerFeelings,andPurchaseLikelihood,”Psychology&Marketing,9(5), 347–63.
Berlyne,DanielE.(1971),AestheticsandPsychology,NewYork,NY:Meredith.
Bitner,MaryJ.(1990),“EvaluatingServiceEncounters:TheEffectsof Physi-calSurroundingsandEmployeeResponses,”JournalofMarketing,54(2), 69–82.
Bone,PaulaF.andPamS.Ellen(1999),“ScentsintheMarketplace:Explaining aFractionofOlfaction,”JournalofRetailing,75(2),243–62.
Booms,BernardH.andMaryJ.Bitner(1982),“MarketingServicesby Man-agingthe Environment,” CornellHotel and Restaurant Administration Quarterly,23(1),35–40.
Broekemier,Greg,RayMarquardtandJamesW.Gentry(2008),“AnExploration ofHappy/SadandLiked/DislikedMusicEffectsonShoppingIntentionsin aWomen’sClothingStoreServiceSetting,”JournalofServicesMarketing, 22(1),59–67.
Brown,StephenP.(1996),“AMeta-AnalysisandReviewofOrganizational ResearchonJobInvolvement,”PsychologicalBulletin,120(2),235–55.
Carlson,Rae(1971),“WhereIsthePersoninPersonalityResearch?,” Psycho-logicalBulletin,75(3),203–19.
Chebat,Jean-CharlesandRichardMichon(2003),“ImpactofAmbientOdorson MallShoppers’Emotions,Cognition,andSpending:ATestofCompetitive CausalTheories,”JournalofBusinessResearch,56(7),529–39.
Chebat, Jean-Charles and Maureen Morrin (2007), “Colors and Cultures: ExploringtheEffectsofMallDécoronConsumerPerceptions,”Journal ofBusinessResearch,60(3),189–96.
Cheng,Fei-Fei,Chin-ShanWuandDavidC.Yen(2009),“TheEffectofOnline StoreAtmosphereonConsumer’sEmotionalResponses:AnExperimental StudyofMusicandColour,”Behaviour&InformationTechnology,28(4), 323–34.
Cohen,Jacob(1988),StatisticalPowerAnalysisfortheBehavioralSciences, 2nded.Hillsdale,NJ:LawrenceErlbaumAssociates.
Crowley,AynE.(1993),“TheTwo-DimensionalImpactofColoronShopping,” MarketingLetters,4(1),59–69.
Cyr,Dianne,MilenaHeadandHectorLarios(2010),“ColourAppealinWebsite DesignWithinandAcrossCultures:AMulti-MethodEvaluation,” Interna-tionalJournalofHuman-ComputerStudies,68(1–2),1–21.
DeMatos,CelsoAugusto,JorgeL.HenriqueandCarlosA.V.Rossi(2007),
“ServiceRecoveryParadox:AMeta-Analysis,”JournalofServiceResearch, 10(1),60–77.
Donovan,RobertJ.andJohnR.Rossiter(1982),“StoreAtmosphere:An Envi-ronmentalPsychologyApproach,”JournalofRetailing,58(1),34–57.
Doucé,LieveandWimJanssens(2013),“ThePresenceofaPleasantAmbient ScentinaFashionStore:TheModeratingRoleofShoppingMotivationand AffectIntensity,”EnvironmentandBehavior,45(2),215–38.
Fiore,AnnM.,XinluYahandEunahYoh(2000),“EffectsofaProductDisplay andEnvironmentalFragrancingonApproachResponsesandPleasurable Experiences,”Psychology&Marketing,17(1),27–54.
Garlin,FrancineV.andKatherineOwen(2006),“SettingtheTonewiththe Tune:AMeta-AnalyticReviewoftheEffectsofBackgroundMusicinRetail Settings,”JournalofBusinessResearch,59(6),755–64.
Gelbrich,KatjaandHolgerRoschk(2011),“AMeta-AnalysisofOrganizational ComplaintHandlingandCustomerResponses,”JournalofServiceResearch, 14(1),24–43.
Glass,GeneV.,BerryMcGawandMaryL.Smith(1981),Meta-Analysisin SocialResearch,NewburyPark,CA:SagePublications.
Grewal,Dhruv,JulieBaker,MichaelLevyandGlennB.Voss(2003),“The EffectsofWaitExpectationsandStoreAtmosphereEvaluationson Patron-ageIntentionsinService-IntensiveRetailStores,”JournalofRetailing,79 (4),259–68.
Hedges, LarryV.and IngramOlkin (1985), StatisticalMethods for Meta-Analysis,Orlando,FL:AcademicPress.
Herrington,J.D.andLouisM.Capella(1996),“EffectsofMusicinService Environments:AFieldStudy,”JournalofServicesMarketing,10(2),26–41.
Herrmann,Andreas,ManjaZidansek,DavidE.SprottandEricR.Spangenberg (2013),“ThePowerofSimplicity:ProcessingFluencyandtheEffectsof OlfactoryCuesonRetailSales,”JournalofRetailing,89(1),30–43.
Herz,RachelS.(2004),“ANaturalisticAnalysisofAutobiographicalMemories TriggeredbyOlfactoryVisualandAuditoryStimuli,”ChemicalSenses,29 (3),217–24.
Herz,RachelS.andTryggEngen(1996),“OdorMemory:ReviewandAnalysis,” PsychonomicBulletin&Review,3(3),300–13.
Hui,MichaelK.,LauretteDubéandJean-CharlesChebat(1997),“TheImpact ofMusiconConsumers’ReactionstoWaitingforServices,”Journalof Retailing,73(1),87–104.
Hunter,JohnE.andFrankL.Schmidt(2004),MethodsofMeta-Analysis: Cor-rectingErrorandBiasinResearchFindings,2nded.ThousandOaks,CA: SagePublications.
Jacob,Céline,JordyStefanandNicolasGuéguen(2014),“AmbientScentand ConsumerBehavior:AFieldStudyinaFlorist’sRetailShop,”The Inter-nationalReviewofRetail,DistributionandConsumerResearch,24(1), 116–20.
Knasko,SusanC.(1995),“PleasantOdorsandCongruency:EffectsonApproach Behavior,”ChemicalSenses,20(5),479–87.
Labrecque,LaurenI.,VanessaM.PatrickandGeorgeR.Milne(2013),“The Marketers’PrismaticPalette:AReviewofColorResearchandFuture Direc-tions,”Psychology&Marketing,30(2),187–202.
Lehrner, Johann,ChristineEckersberger,PeterWalla,G.PötschandLüder Deecke(2000),“AmbientOdorofOrangeinaDentalOfficeReduces Anx-ietyandImprovesMoodinFemalePatients,”Physiology&Behavior,71 (1–2),83–6.
Ludvigson,WayneH.andTheresaR.Rottman(1989),“EffectsofAmbient OdorsofLavenderandClovesonCognition,Memory,AffectandMood,” ChemicalSenses,14(4),525–36.
Mattila,AnnaS.andJochenWirtz(2001),“CongruencyofScentandMusicas aDriverofIn-storeEvaluationsandBehavior,”JournalofRetailing,77(2), 273.
Mehrabian,AlbertandJamesA.Russell(1974),AnApproachtoEnvironmental Psychology,Cambridge,MA:TheMITPress.
Michon,Richard,Jean-CharlesChebatandL.W.Turley(2005),“Mall Atmo-spherics:TheInteractionEffectsoftheMallEnvironmentonShopping Behavior,”JournalofBusinessResearch,58(5),576–83.
Milliman,RonaldE.(1982),“UsingBackgroundMusictoAffecttheBehavior ofSupermarketShoppers,”JournalofMarketing,46(3),86–91.
Mitchell,DeborahJ.,BarbaraE.KahnandSusanC.Knasko(1995),“There’s SomethingintheAir:EffectsofCongruentorIncongruentAmbientOdor onConsumerDecisionMaking,”JournalofConsumerResearch,22(2), 229–38.
Morrin,MaureenandS.Ratneshwar(2000),“TheImpactofAmbientScenton Evaluation,Attention,andMemoryforFamiliarandUnfamiliarBrands,” JournalofBusinessResearch,49(2),157–65.
Morrison,Michael,SarahGan,ChrisDubelaarandHarmenOppewal(2011),
“In-StoreMusicandAromaInfluencesonShopperBehaviorand Satisfac-tion,”JournalofBusinessResearch,64(6),558–64.
North,AdrianC.andDavidJ.Hargreaves(1996),“ResponsestoMusicina DiningArea,”JournalofAppliedSocialPsychology,26(6),491–501.
Oliver,RichardL.,RolandT.RustandSajeevVarki(1997),“CustomerDelight: Foundations,Findings,andManagerialInsight,”JournalofRetailing,73 (3),311–36.
Palmatier,RobertW.,RajivP.Dant,DhruvGrewalandKennethR.Evans(2006),
“FactorsInfluencingtheEffectivenessofRelationshipMarketing:A Meta-Analysis,”JournalofMarketing,70(4),136–53.
Pan,Yueand GeorgeM.Zinkhan(2006),“Determinantsof Retail Patron-age: A Meta-Analytical Perspective,” Journal of Retailing, 82 (3), 229–43.
Peterson,RobertA.(2001),“OntheuseofCollegeStudentsinSocial Sci-enceResearch:InsightsfromaSecond-OrderMetaanalysis,”Journalof ConsumerResearch,28(3),450–61.
Rosenthal,Robert(1979),“The‘File-DrawerProblem’andToleranceforNull Results,”PsychologicalBulletin,86(3),638–41.
Sayin,Eda,AradhnaKrishna,CarolineArdelet,GwenaëlleB.DecréandAlain Goudey(2015),“‘SoundandSafe’:TheEffectofAmbientSoundonthe PerceivedSafetyofPublicSpaces,”InternationalJournalofResearchin Marketing,32(4),343–53.
Shadish,WilliamR.,ThomasD.CookandDonaldT.Campbell(2010), Experi-mentalandQuasi-ExperimentalDesignsforGeneralizedCausalInference, Belmont,CA:WadsworthCengageLearning.
Spangenberg, EricR., AynE. Crowleyand Pamela W.Henderson (1996),
“ImprovingtheStoreEnvironment:DoOlfactoryCuesAffectEvaluations andBehaviors?,”JournalofMarketing,60(2),67–80.
Spangenberg,EricR.,DavidE.Sprott,BiancaGrohmannandDanielL.Tracy (2006),“Gender-CongruentAmbientScentInfluences onApproachand AvoidanceBehaviorsinaRetailStore,”JournalofBusinessResearch,59 (12),1281–7.
Sullivan,Malcolm(2002),“TheImpactofPitch,VolumeandTempoonthe AtmosphericEffectsofMusic,”InternationalJournalofRetail& Distribu-tionManagement,30(6),323–30.
Turley,L.W.andRonaldE.Milliman(2000),“AtmosphericEffectsonShopping Behavior:AReviewoftheExperimentalEvidence,”JournalofBusiness Research,49(2),193–211.
vanRompay,ThomasJ.L.,KarinTanja-Dijkstra,JoostW.M.Verhoevenand Annemiek F.van Es(2012),“OnStore Designand Consumer Motiva-tion:SpatialControlandArousalintheRetailContext,”Environmentand Behavior,44(6),800–20.
Winer,BenJ.(1971),StatisticalPrinciplesinExperimentalDesign,2nded. NewYork,NY:McGraw-Hill.
Yalch,RichardF.andEricR.Spangenberg(1988),“AnEnvironmental Psycho-logicalStudyofForegroundandBackgroundMusicasRetailAtmospheric Factors,” inEfficiencyandEffectiveness inMarketing,FrazierGaryed. Chicago,IL:AmericanMarketingAssociation,106–10.
Yousem,DavidM.,JosephA.Maldjian,FaezSiddiqi,ThomasHummel,David C.Alsop,RenaJ.Geckle,WarrenB.BilkerandRichardL.Doty(1999),
“GenderEffectsonOdor-StimulatedFunctionalMagneticResonance Imag-ing,”BrainResearch,818(2),480–7.