8/10/2019 art%3A10.1177%2F0092070399272005

1/23

The Role of motions in Marketing

Richard P Bagozzi

University o f Michigan

Mahesh Gopinath

Tulane University

Prashanth U Nyer

Chapman University

Emotions are mental states of readiness that arise rom ap-

praisal s of events or one s own thoughts. In this article, the

authors discuss the differentiation o f emotions fro m affect,

moods, and attitudes, and outline an appraisal theory of

emotions. Next, various measurement issues are consid-

ered. This is follo wed by an analysis o f the role of arousal

in emotions. Emotions as markers, mediators, and mod-

erators of consumer responses are then analyzed. The

authors tu rn next to the influence o f emotions on cognitive

processes, which is ollo wed by a study of the implications

of emotions fo r volitions, goal-dire cted behavior, and de-

cisions to help. Emotions and customer satisfaction are

briefly explored, too. The article closes with a number of

questions for fu ture research.

This a r t icle addresses emotiona l behavior in marketing.

In comparison to information processing and behavioral

decision research, we know much less about the role of

emotions in marketing behavior . Much of what we do

know is conf in ed to consum er behavior , as opposed to the

behavior of salespeople or ma rketing managers.

Never theless, emotions are central to the act ions of

consum ers and manag ers al ike. Our goal in this ar ticle will

be to present a f ram ework for thinking about emotions; to

discuss the m easurem ent of emotions; to review how emo-

tions function as causes, effects, mediators, and moderato rs

Journal of the Acade my of Marketing Science.

Volume 27 No. 2 pages 184-206.

Copyright 9 1999 by Academ y of Marketing Science.

in marketing behavior ; and to provide suggestions for

future research.

THEORY ND FUNCTION OF EMOTIONS

Litt le consistency can be found in the use of terminol-

ogy related to emotions. For purposes of organization and

discussion, we begin with a def ini t ion of emotions and

then tu rn to a f r amework fo r in terpre t ing emot ional

behavior.

Definitions

The term affect will be conceived herein as an umbrella

for a set of more specif ic mental processes including em o-

tions, moods, and possibly) att i tudes. Thus, af fect migh t

be cons idered a genera l ca tegory fo r menta l f ee l ing

processes, rather than a par t icular psychologica l process,

per se.

By emotion, we mean a mental s tate of readiness that

ar ises f rom c ognit ive appraisals o f events or thoughts; has

a phenomenolog ical tone; is accom panied by physiologi-

cal processes; is of ten expressed physically e.g. , in ges-

tures, posture, facial features) ; and m ay result in specif ic

actions to aff irm or cope with the em otion, depend ing on

its nature and m eaning for the person having i t . For a s imi-

lar perspective, see Lazarus 1991) and Oatley 1992) .

The l ine between an emot ion and m ood i s f r equen t ly

dif f icult to draw but of ten by convention involves conce iv-

ing of a mood as being longer last ing f rom a few hours up

to days) and lowe r in intensi ty than an emo tion. Yet, excep-

tions to this construal can be found . Still anoth er distinction

8/10/2019 art%3A10.1177%2F0092070399272005

2/23

Bagozzi et al. / THE ROLEOF EMOTIONS 185

between emo tions and m oods is that the former typically is

intentional (i.e. , it has an object or referent), whereas

moods are generally nonintentional and global or diffused

(Frijda 1993). Also, moods are not as directly coupled with

ac t ion tendenc ies and expl ic i t ac t ions as a re many

emotions.

Finally, attitudes, too, are often conside red instances of

affect, with the sam e measures used on oc casion to indicate

emotions and attitudes (e.g., pleasant-unpleasant, happy-

sad, or interested-b ored semantic differential items). How-

ever, some authors take a narrower view of attitudes and

define them as evaluative judgmen ts (measured, e .g. , by

good-bad reactions) rather than em otional states. Cohen

and A reni (1991), for instance, reserve the term affect for

valenced feeling states, with emotions and moods as

specif ic examples. Atti tudes are evaluative judgments in

their view. Nevertheless, other researchers do not make a

distinction betw een affect and evaluative judgments. For

example, Eagly and Chaiken (1993) point out that Fish-

bein and Ajzen (1975) and other social psychologists have

regarded affect as isomorphic with evaluation i tself and

used the terms interch angea bly (p. 12). Still others pro-

pose that attitudes have two distinct, but generally highly

cor re la ted , components : a f fec t ive and cogni t ive (or

evaluative) dimensions. So me empirical support exists for

this interpretation (Bagozzi and Burnkrant 1979; Batra

and Ahtola 1990; Breckler and Wiggins 1989; Crites, Fab-

rigar, and Petty 1994; Eagly, Mladin ic, and Otto 1994).

I t should be recogn ized that the terms

affect emotions

moods and attitudes have frequently been used inconsis-

tently in the literature. We will revisit this issue when we

consider both me asuremen t issues and custom er satisfac-

tion research below. For now, we stress that when reading

the literature, it is important to pay attention to how

authors define affective (and related) terminologies and

how they measure the variables to which the terminologies

refer . One's definit ion of terms permits an interpretation of

their meaning, but equ ally important is how the variables

to which the terms refer are operationalized. Some authors

have defined key variables as emotions, moods, or att i-

tudes but have used operationalizations corresponding to

different concepts. Oth er authors have u sed operationali-

zations for a single variable that cut across two or more

instances of affect. To make clear our definit ion of emo-

tions and ho w it differs from definit ions of mood and att i-

tudes, we present the follow ing point of view.

Organizing ramework

Above we noted that emotions are mental states of

readiness. Bu t so, too, are moods and att itudes. How then

might we distinguish between these affective states? For

one thing, the state of readiness character ized by an em o-

tion tends to be more intense than that character ized by

moods or att i tudes. I t is more intense in the sense of

strength of felt subjective experience, plus magn itude of

physiological response (e .g. , autonomic nervous system

activity) and extent of bodily expression (e.g., facial dis-

plays) , when these latter reactions accomp any an em otion.

Probably the most important factor differentiating

emotions from moods and att i tudes is the way emotions

arise. Emotio ns are said to have a s pecific referen t (e.g., a

consum er becomes pleased when a new detergent removes

grass stains from clothing; he or she is angered by p oor ser-

vice in a restaurant) . Specif ically, emotions ar ise in

response to appraisals one makes for som ething o f rele-

vance to one 's well-being. By appraisal , we mean an

evaluative judgm ent and interpretation thereof . By some-

thing of relevance, we mean an incident or episode that

happens to on eself (e .g., an unp lanned event) ; a behavior

one performs or a result one produces (e .g. , engaging in an

activity or receiving or fail ing to receive a planned out-

come); or a change in an object, person, or thought that has

personal meaning.

I t is important to stress that a lthough categories of

events or physical c ircumstances are frequen tly associated

with particular emotional responses, i t is not the specif ic

events or physical c ircumstances that produce the emo-

tions but rather the unique psycho logical appraisal mad e

by the person evaluating and interpreting the events and

circumstances. Different people can have different emo-

tional reactions (or no emotional reactions at a ll) to the

same event or happening. Note, too, that appraisals can be

deliberative, purposive, and conscious, but also unrefiec-

f ive, automatic, and unconscious, depending on the person

and elicit ing conditions fo r emotional arousal. Th e central

role of appraisals in the formation o f emotions has co me to

define what are aptly called appraisal theories in psychol-

ogy (e.g., Frijda 1986; Lazarus 1991; Ortony, Clore, and

Collins 1988; Roseman 1991; Smith and Ellsworth 1985).

Appraisal theorists maintain that the critical determi-

nant of any em otion is the resultant evaluation and inter-

pretation that ar ise af ter comparing an actual state with a

desired state. Two appraisals are particu larly crucial at this

stage of emotion formation: goal relevance and goal con-

gruence (Lazarus 1991). That is , a necessary condition for

an emotional response to an even t or happening is that a

person has a personal stake in it and at the same time

judges the event or happening to facili ta te or thwart this

stake. Again, the appraisal can occur consciously or

unconsciously.

A distinctive feature of appraisal theories is their speci-

f ication of the conditions leading to discrete emotional

responses. Forced to be brief , we focus on Roseman's

(1991) version of appraisal theories, whic h differs in rela-

tively minor ways from other leading theories. Roseman

hypothesized that particular combinations of f ive apprais-

als determine which o f 16 unique emotions will be experi-

enced in any given situation. Figure 1 summarizes his the-

ory , where the f ive appra isa ls a re labe led mot ive

8/10/2019 art%3A10.1177%2F0092070399272005

3/23

186 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

FIGURE 1

Roseman s (1991) Appraisal

Theory of Emotions

PosiuveEmotions

Motive-Consistent

Appetitive [

verslve

ircumstance

caused

unexpected

11riceraila

certain

uncert in

eertam

Other-caused

uncertain

ceruun

u n ~

Self-caused

uncertain

certain

uncertain

c~l tain

NegativeEmotions

Motive-lnconslstent

Appetitive I

versive

Sur~se

Hope

Joy [ Relief Sadnes s

Hope

Joy I Relief

Ldong

Pride

Fear

Distress Disgust

Frustratton

Weak

Strong

Dishke Weak

nger Strong

Shame Guilt

Regret

Weak

Strong

SOURCE: Roseman 1991:193). Reprintedwith permission.

consistent/motive inco nsistent (i .e., positive em otions ver-

sus nega tive emotions), appetitive/aversive (i.e., presence

of a reward vs. absence of a punishmen t), agency (i.e., out-

come is perceived caused by impersonal circumstances,

some other person, or the self), probability (i .e., an out-

come is certain or uncertain), and pow er (i .e., strong ver-

sus weak co ping potential).

For examp le, pride occurs w hen one evaluates his or her

own performan ce of an act ion or achievement of an out-

come in a positive light (e.g., a feeling of having done

well). Here the positive emotion is motive consistent,

either appetitive (e.g., having attained a positive goal) or

aversive (e.g. , having avoided a punishment), self-

produced under weak or low coping potent ial , and ei ther

certain or uncertain, depending on the circumstances. Sad-

ness happens when one experiences a loss for which one

recognizes tha t nothing can b e done to restore it . The loss,

which is of some thing or someo ne valued, is experienced

negat ively and with high certainty under condi t ions of low

coping power. It is perceived to be caused by impersonal

circumstances.

One value of appraisal theories is that it is possible to

account for m ost emotions. Indeed, subt le combinat ions of

appraisals yield discrete emotional responses. Anger and

regret, for example, differ primarily in only one type of

appraisal and sh are in the other four, namely, anger occurs

when a person sees an other person as the source of injury

to onese lf or to another person v iewed as a victim of injus-

tice, whereas regret results when one's negative outcome

is attributed to actions or inactions of the self.

Not every emotion is accounted for by Roseman's

framework (or by any other framework for that matter).

For instance, pride is regarded as a positive emotion in

Rosema n's framework, yet excessive or exaggerated pride

(sometimes termed

hubris

can invite retribution. Like-

wise, shame and gui l t are thought by Ro seman to be pro-

duced by similar appraisals, but other researchers have

found important distinctions between shame, guilt , and

embarra ssment (e.g., Lew is 1993). Like wise, disgust has

been studied extensively and found to differ from distress

(e.g., R ozin, Haidt, and McC auley 1993). Nevertheless, in

contrast to other theories of emotion that conc eive of it in

b ipo lar t erms (e .g . , p leasure-d i sp leasure and h igh

arousal-low arousal [Russell 1980] or high negative

affect-low negative affect and high positive affect-low

positive affect [Watson and Tellegen 1985]), Roseman's

framew ork and other appraisal theories not only allow for

many discrete emotions but specify condi t ions for their

occurrence.

An elaboration of appraisal theories that is especially

relevant for marketing is the treatment of goals, which m ay

be defined as internal representations of desired states,

where states are broadly defined as outcomes, events, or

processes (Austin and Vancouver 1996:338). Oatley and

Johnson-L aird (1987) proposed w hat they termed a com-

municative theory o f emotions wherein events are evalu-

ated in relation to a person's goals. Emotion s are though t

to function to coordinate parts of one 's cognitive syste m so

as to manage responses to events and in so doing change

from ongo ing to new activities or to maintain desired states

or activities. The self-regulation of goa ls is believed to be

the main function of emotions:

Each goal and plan has a m onitoring mechanism that

evaluates events relevant to it . When a substantial

change in probability occurs of achieving an impor-

tant goal or subgoal , the monitoring mechanism

broadcasts to the whole cognitive system a signal

that can set it into readiness to respond to this

change. Humans experience these signals and the

states of readiness they induce as emotions. (Oatley

1992:50)

Accor ding to Oatley and Johnson-Laird (1987), emotions

are evoked at a s ignificant juncture of a p la n . . , typi-

c al ly . . , when the evaluat ion (conscious or unconscious)

of the likely success of a plan cha nges (p. 35). Positive

emotions (e.g., happiness, elation, joy ) are associated w ith

the attainment of a (sub)goal, wh ich usually leads to a de-

cision to continue with the plan, whereas negative emo-

tions (e.g., frustration, disappointment, anxiety) result

from problems w ith ongoing plans and failures to achieve

desired goals (see also Stein, Liwag, and W ade 1996).

Emo tions have implications for action and goal attain-

ment . Lazarus (1991) ident i f ies coping responses as

8/10/2019 art%3A10.1177%2F0092070399272005

4/23

Bagozzi et al. / THE ROLEOF EMOTIONS 187

importan t mec hanism s in this regard. When w e experience

a negative emotion (e.g., anger, sadness, fear), we are in

disequilibrium and wish to return to our normal state.

Either one or both of two coping processes are typically

used: problem -focuse d coping, where we attempt to alle-

viate the sources o f distress, or emotion -focuse d coping,

where we ei ther change the meaning o f the source of dis-

tress (e.g., deny that a threat exists, distance oneself from

the source of distress) or avoid thinking about a problem.

By contrast , coping with posi t ive emotions often

involves sharing one's good fo rtune, savoring the experi-

ence, working to continue or increase the rewards, and

increasing phy sical activity. Positive emotions are som e-

times acc ompa nied as well by higher levels of physiologi-

cal arousal , expanded at tent ion, increased opt imism,

enhanced recall, and a shift from self- to other-centered

orientations (e.g., becoming friendlier, caring about oth-

ers), when compared, say, to sadness. Indeed, positive

emotions, part icularly happiness , frequent ly s t imulate

helping or altruistic actions. Why? Schaller and Cialdini

(1990) offe r two explanations: First, we may propose that

positive mood leads to enhanced helping via the more

positive outlook and enhanced activity that appear to

spring automatically from the experience of happiness,

and second, we argue that happiness is associated with a

motivat ion toward disequi l ibrium--toward the possible

attainment of additional personal rewards that transcend

the basic conce rn over one's mo od (pp. 284-285). The

personal rewards referred to here concern such self-

enrichment motives as affiliation, achievement, compe-

tence, and esteem.

Closely related to coping responses are action tenden-

cies. An ac tion tendency is a readiness to engage in or dis-

engage from interact ion with some goal object and

includes (i)mpulses of 'mov ing towards, ' 'mov ing away,'

and 'movin g against ' (Frijda, Kuipers, and ter Schure

1989:213). Som e theorists maintain that emotions are not

merely reactions to appraisals of events but also include

action tendenc ies as part of their mean ing (Frijda 1986).

Others go fur ther and maintain that action tendencies are

automatic, prew ired responses connec ted to emotions

(LeDoux 1996). And in Frijda's (1986) treatment, emo-

tions are conceived as the entire process from stimulus

event to action an d arousal:

Stimulus eve nt~ eve nt coding~appraisal (evaluation of

relevance, context, and urgency/difficulty/seriousnessof

action

event)~ action readiness

arousal

Mu ch as e motion s arise in response to patterns of apprais-

als, Frijda (1986; F rijda et al. 1989) has shown that pat-

terns of action readiness correspond to distinct emotion

categories (e.g., avoidance with fear, helping with caring,

helplessness with sadness, assault with anger, withdraw al

with shame).

Finally, it has been argued that man y coping responses

to emotio ns are volitional (B agoz zi 1992:186-189). The

process begins with outcome-desire units and appraisals

of changes or ant icipated changes in goal a t tainment or

goal progress. Four appraisal classes can be identified.

Outcome-des i re conf l i c t s happen when one fa i l s to

achieve a goal or when one experiences an unpleasant

ev en t . O n e o r m o re em o t i o n a l r eac t i o n s o ccu r t o

outcome-desire confl icts (e.g. , dissat isfact ion, anger,

shame, guilt , sadness, disappointment, disgust, regret),

depending on attributions of the source of goal failure or

the unpleasa nt event (i .e., self, other person, or external

cause). The coping respon se(s) to these emotion s, in turn,

is selected from the following: intent to remove or undo

harm, obtain help or support, decrease outco me, reevalu-

ate goal, or redouble effort, if appropriate, depe nding on

the specific emotion involved.

Outcom e-desire fulfi l lment takes place when one

achieves a goal, experiences a pleasant event, or avoids an

unpleasant event. One or more emotional react ions com e

about when o utcome-d esire fulfillment happen s (e.g., sat-

isfaction, joy, elation, pleasure, pride, relief, caring, love),

again depending on attributions of the source of good for-

tune. The coping responses to these emotions include an

intention to maintain, to increase, to share, or to enjoy the

outcome.

Outcome-desire conflicts and fulfillment refer to out-

comes in the past or present. The following two appraisal

classes go on with regard to planned outcomes. Outcome-

desire avoidances transpire in anticipation of unpleasant

outcome s or goals. Fear o r its variants (e.g., worry, anxi-

ety, distress) are the emotiona l reactions to this appraisal.

The coping responses to these emotions entail either an

intention to avoid undesirable outcomes or to reinterpret

the threat.

The final class of appraisals, outcom e-desire pursuits,

happen in anticipation of pleasant goals or outcomes.

Hope is the emotional reaction to such appraisals. The

coping response(s) to hope include s intentions to realize or

facilitate outcom e attainment and to sustain one's co mmit-

men t and vigilance.

The theory of self-regulation suggests that unique voli-

tional responses underly coping for each particular emo-

tion or class of emotions (Ba gozzi 1992). In addition, the

specific intent ion enacted depends on one's degree of

self-efficacy n executing the coping responses. Some what

analogous (appraisal---~emotional reactions---~coping)

responses occur for outcome-identity conflicts, fulfill-

ments, avoidances, and pursuits in social situations related

to norm ative expe ctation s (B agoz zi 1992:191-194).

Returning to the dis t inct ion between emotions and

mood s and attitudes, we migh t say that in addition to the

8/10/2019 art%3A10.1177%2F0092070399272005

5/23

188 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

things mentioned earl ier , emotions differ from moods and

attitudes in the manner in w hich they arise and in their rep-

resentation in memory. Emotions occur in response to

changes in specific plans or goal-relevant events. As O at-

ley (1992) points out, emotions are manife st as transi-

t ions from one sequence of act ion and another, but moods

occur when the cogni t ive system is maintained in an emo-

tion mod e for a period (pp. 64, 91-92). Ind eed, moods are

often resistant to changes in events surrounding them. On e

reason for this is that moo ds depen d on the dissociability

of control emotion s ignals from semantic information

about causation (Oatley 1992:64). In general, moo ds are

elicited by (a) after effects of emotions; (b) organismic

conditions such as illness, fatigue, previous exercise, and

good health, or pha rmaco logical agents; (c) general envi-

ronme ntal cond itions and side-effects of activities: heat,

noise, environm ental variety, stressful conditions (Frijda

1986:289).

Like emotions, attitudes can arise from changes in

events, but attitudes also occur in response to mundane

objects. In addition, arousal is a nec essary part of emotions

but not nec essarily attitudes. Moreover, attitudes seem to

have the capac ity to be stored during long periods of time

and retrieved, whereas emotions are not experienced in

this way (i.e., emotions are ongoing states of readiness;

they are n ot stored a nd retrieved, per se, although it is pos-

sible to recreate the conditions originally producing them

in our mem ory a nd react em otional ly to the thoughts so

generated, at least up to a point; emotio ns can, however, be

classically conditioned, but it is uncle ar whether attitudes

can). Finally, the connection of emotions to volition and

action is stronger an d more direct than it is for attitudes.

Emotions directly stimulate volitions and initiate action,

but attitudes may require an additional motivation impe-

tus, such as desire (Bag ozzi 1992).

ME SUREMENT OF EMOTIONS

The m easuremen t of emotions could focus on a ful l set

of signs or evidence, inc luding evaluative appraisals, sub-

jective feelings, bod y posture a nd gestures, facial expres-

sions, physiological responses, action tendencies, and

overt act ions. Whatever mea surements one uses should, of

course, be t ied to an underlying theory of emotions.

Som e authors (e.g., most appraisal theorists) construe

emo tions as menta l states or processes, and thus it would

be prudent to directly measure the cognitive activities

compris ing the emotional content of these s tates or

processes, fr om this perspective. Self-reports of one's sub-

ject ive experiences const i tute the m ost frequent ly used

procedures in this regard, altho ugh other methods for indi-

cat ing emotional memor y processes might be used as w el l

(e.g., response time, sublimina l priming). From the po int

of v iew of menta l conceptual i za t ions of emot ions ,

physiological, motor, or biological indicators would be at

bes t cons idered corre la tes o r ind i rec t measures o f

emotions.

Other researchers who interpret emotions in broader

terms, as ei ther the whole process from the coding of

events to action responses (e.g., Frijda 1986) or as com -

plex patterns of physiological responses (e.g., Caciopp o,

Bem tson, and Klein 1992 maintain that emotion al experi-

ence is a fun ction of somatoviscerai activation, afferentia-

tion, and cognitive operations; LeDou x 1996 emph asizes

brain processes, especially the role of the amyg dala) stress

the need for measurement processes going be yond self-

reports. Depending on the theorist, overt behaviors or

physiological reactions ma y be considere d either a part of

what i t means to have an em otion or antecedents , concomi-

tants , or possibly even effects of an em otion. More behav-

iorally or physiolog ically oriented researchers obviously

employ measures o f emot ions cons i s ten t wi th these

interpretations.

Marketers have tended to take an empirical approach to

the measurement of emotions and to rely on self-reports

(i.e., either unipolar or bipolar items on questionnaires). In

the typical application, many items cutting across num er-

ous positive and negative emotions are administered to

measu re reactions to a stimulus, and such meth ods as fac-

tor analysis, multidime nsional scaling, or cluster analysis

are used to ident ify the underlying emotional dimensions

for the sample at hand. The number of i tems invest igated in

this regard has been as large as 180 (Aaker, Staym an, and

Vezina 1988), while a paper-and-pe ncil tech nique (basi-

cal ly a s ingle i tem measured cont inuously while viewing

an ad) has even been suggested to regis ter warm th

toward a stimulus ad (Aaker, Stayman, and H agerty 1986;

see also Russell, Weiss, and Me ndelso hn 1989).

Two influential studies in the measurement of emo-

tional responses toward adv ertisements are those by E dell

and Burke (1987) and Holbrook and Batra (1987). Edell

and Burke (see also Burke and Edell 1989) developed a

52-item scale for measu ring emotions towards ads, while

Holbrook and Batra worked with a 94-i tem scale, which

was later reduced to 34 items (Batra and H olbro ok 1990).

With so many i tems measuring emotions, a quest ion

arises whether a small number of basic dimen sions under-

lie people's responses. Edell and Burke (1987) analyzed

the items in their scale and found three factors: upbeat feel-

ings, negative feelings, and warm feelings. Likewise, Hol-

brook and Batra (1987) used factor analysis, but in a

slightly different way. Their 94 items w ere first generated

a priori to measure 29 emotional indices. For example,

joyful , happy, delighted, and pleased w ere hypo thesized to

indicate a joy index, and ashamed, embarrassed, and

humiliated were hypothesized to indicate a shame index.

Then, based on fac tor analyses of the 29 indices, a three-

factor solution for emotions was fou nd: pleasure, arousal,

and domination. Similarly, Batra and Holbrook (1990)

8/10/2019 art%3A10.1177%2F0092070399272005

6/23

Bagozzi et al. / THE ROLE OF EMOTIONS 189

factor analyzed 12 emotional indices (largely a subset of

those revealed in Holbrook and Batra, with a few excep-

tions) and discovered three factors corresponding closely

to those discovered by Ede ll and Burke. O ther researchers

(e.g., Oliver 1994; Westbrook 1987), also using factor

analysis, have found em otional items to load on two fac-

tors: positive affect and negative affect.

Richins (1997) recent ly argued that consumption-

related emotions are more complex than the two- and

three-factor solutions observed in studies of reactions to

ads or custom er satisfaction. Moreover, because explora-

tory factor analyse s often yield a small numbe r of factors,

she used a multidimensional scaling procedure, in con-

junct ion w ith examinat ion of clusters based on locat ion

and semantic similarity of emotional descriptors in two-

dimensional space. Sixteen clusters of emotions were

identified, each measu red by 2 to 8 indicators (in Study 4):

anger, discontent, w orry, sadness, fear, shame, envy, lone-

liness, romantic love, love, peacefulness, contentment,

optimism, joy, exciteme nt, and surprise.

Although the approaches used by marketers to date

have been largely empirically driven (e.g., Edell and Burke

1987; Holbrook and Batra 1987; Oliver 1994; Richins,

1997; Westbr ook 1987), the y are consistent in certain senses

with leading perspectives on emotions in psychology. Fo r

example , H olbrook and Bat ra s (1987) th ree-fac tor

pleasure-a rousal-dom ination findings are somew hat simi-

lar to R ussel l and Meh rab ian s (1977) th ree-fac tor

pleasure-arousal-dominance model. However, some dif-

ferences can be pointed out. The most important are the

high loadings of sadness and fear on the dominat ion factor

in Holbroo k and Batra s study, in contrast to the more com-

mon outcome o f sadness loading on or near the negat ive

pole of a pleasure-displeasure factor, and fear loading

about 45 degrees away from displeasure and toward

greater arousal (e.g., Russell, 1997).

Similarly, the three factors fou nd in Edell and Burk e s

(1987) s tudy (i.e., upbeat feelings, negative feelings, and

warm feelings), correspond roughly to the high positive

affect-low positive affect, high negative affect-low nega-

tive affect, and pleasantness-unpleasantness dimensions,

respectively, of Watson and T ellegen s (1985) circumplex

mode l ; the posi t ive-affect and negat ive-affect factors

observed by Oliver (1994) and Westbrook (1987) also

align to a great extent with factors on the circumplex

mod el (see also Mano a nd Oliver 1993).

The circumplex structure of emotions is shown in Fig-

ure 2. This representation o f emotions has also been called

the two-factor model, because, based on the techniques

used to generate it (e.g., factor analysis or multidimen-

sional scaling), emotions can be arranged around two

orthogona l axes. Russell (1997) terms the axes pleasure-

displeasure an d arousal-sleepiness, wh ile Watson and Tel-

legen (1985) label them high positive affect-low positive

FIGURE 2

Watson and Tellegen s Two-Factor

Structure of Affect

CONSENSUALMOOD STRUCTURE

~ G t4 pOSItIVE AFF~CCT

kindly

I s t ~ d

warmhearted

lit rlist

rl i~xed

i

o ~ n l

O~ qu~ol

~ 9 111111

a~ve

elaled

e n t h u ~ b c

excited

pepp

1rofl

J

j l

s

J

dull

klepy

r

LOwPOSITIVEAFF~.C~

J

dlstrUSed

hos~ie

8/10/2019 art%3A10.1177%2F0092070399272005

7/23

190 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

feeling hostile (see high negative affect in Figure 2).

Then, too, the circumple x contains categories that may not

correspond to em otions. For instance, surprised, drowsy,

and sleepy do not seem to reflect emotions. On the other

hand, the circumplex fails to represent well instances of

emotion im portant in every day life and marketing. It does

not accom mod ate love, disgust, pride, hope, guilt, shame,

or embarrassment very well, to name a few. The various

two- and three-fa ctor summaries o f emotions disclosed in

marketing a lso include variables not reflective of emotions

and exclude instances ge nerally recognized as emotions.

Richins 's (1997) Consumption Emotions Set (CES)

with its 16 descriptors is appealing because it covers most

emotional reactions one encounters in consumption, and

its measures achieved satisfactory reliability (except for

measure of envy, loneliness, peacefulness, and content-

ment). The CES would be best used within the context of a

particular theory of emotions to operationalize specific

categories of emotions hypothesized to serve as antece-

dents, consequences, or moderating variables. The CES

could also be used to operationalize emotions in more

empirically oriented studies, but whether measures of each

dimension would achieve discriminant validity is prob-

lematic. Most studies inco rporating multiple instances of

both positive emotion s and negative emotions find that the

measures load on two factors corresponding to positive

and negative emotions (e.g., Bagozzi, Baumgartner, and

Pieters, 1998; Oliver, 1994).

This raises the question of wh en one can expect discrete

emotional reactions versus amalgamated groupings of,

say, positive emotions and negative emotions (e.g., highly

correlated feelings o f anger, sadness, and fear). The advan-

tage of a theory-based approach to emotions is that spe-

cific conditions can be specified for the occurrence of dis-

tinct emotions, and the se hypothese s can be tested. It is for

these reasons tha t we used appraisal theories as our organ-

izing framework fo r looking at emotions. Discrete emo-

tional reactions are likely to happen wh en one manipulates

conditions produc ing specific appraisals or when natura lly

occurring events correspond to unique appraisal condi-

tions. Gopina th and Bagozz i (1999), for example, were

able to induce independent emotional reactions toward

three targets in a moviegoing context. On the basis of

Roseman's (1991) theory, distinct emotional reactions

were produced as a function of three-way interactions

between motive consis tency-inconsis tency, appet i t ive-

aversive, and self-other agency conditions. For example,

admirat ion, affect ion, dis l ike, and contempt resul ted

toward a group membe r in decis ion making w ith regard to

movie choice, and pride, shame /guilt , and regret resulted

toward the self in group decis ion making, depending on

the three appraisal conditions. Likewise, happiness, satis-

faction, annoyance, or frustration occurred toward the

movie, and pleasure, contentment, irritation, or anger

occurred toward the theater, depending on the three

appraisal conditions. Scenarios were used to create the

appraisal conditions.

By contrast, research not based on manipulations of

appraisal conditions or based on reactions to a single

stimulus frequently finds that emotions cluster in two and,

on occasion, three factors (e.g., Edell and Burke 1987;

Holbroo k and Batra 1987; Oliver 1994; Westbrook 1987).

Furthermore, research exam ining the construct validity of

measures of discrete emotions obtained in nonexperimen-

tal survey settings shows that discrim inant validity is often

lacking among measures of different posi tive or different

negat ive emotions (e.g. , Bagozzi 1993) or between mea-

sures within a particular subcategory of positive or negative

emotions, such as among measures of elation, gladness,

and joy (B agozzi 1991a).

What accounts for the differences in findings between

experimental research based on appraisal theories a nd sur-

vey research or research based on reactions to a single

st imulus measured by inventories of emotional i tems? O ne

possibility may be that discrete emotional reactions are

short-lived or, once activated, stimulate other emotional

reactions closely related to them. C onsider, for example,

sadness: [W] hen we experience loss, we rarely feel a sin-

gle emotion such as sadness. We grieve, are angry, anx-

ious, guilty, envious, even hopeful, and defe nsive (Lazar us

1991:250). A reason why these emo tions ma y go in tan-

dem is that coping processes for sadness may involve

active struggle or even protest against loss, w hich results in

other emotions. Alternatively, the absence of discrete

emotions may simply reflect how difficult i t is to create

them. As Izard (1972) noted, pure emotions are virtually

impossible to obtain in the laboratory or in any research

setting (p. 103). However, as noted above, researchers

have recently found that the use of scena rios in an experi-

mental context can generate discrete emotional responses

(e.g., Gopinath and Bag ozzi 1999; Ro sema n 1991).

Another factor that might account for a coalescence of

multiple positive emotions and m ultiple negative emotions

in two corresponding groupings is the nature of the stimu-

lus under study. Most stimulus ads, products, or brands are

complex, and the appraisals engendered are typically

variegated, but related. Also the way in which items are

presented on some questionnaires makes it difficult to

uncover discrete emotional components . W hen mult iple

measures of a single discrete emotional response (e.g.,

happy, pleased, and joy ful for joy ) are interspersed

throughout a questionnaire, this tends to reduce correla-

tions among items purported to indicate the same respon se

and to increase correlations of these items with me asures

of other responses. The result is predictably a redu ction in

discriminant validity and high correlations among items

8/10/2019 art%3A10.1177%2F0092070399272005

8/23

Bagozzi et al. / THE ROLEOF EMOTIONS 191

measuring posi t ive emotions an d am ong i tems measuring

negative emotions. Th e alternative is to group items by the

emotion al response they are intend ed to tap, which tends to

increase correlations among measures of the same thing

and decrease correlat ions among measures of different

emotional responses. Thus, a tra de-off is entailed by use of

either practice.

An issue that has received little attention in marketing is

whether to use unipolar or bipolar items to measure em o-

tions. The choice ca n influen ce findings and their interpre-

tation in fund ame ntal ways. Some lea ding scholars claim

that emotions are in the final analysis bipolar states or

processes. We are either happy or sad, for example, and

any other pattern (e.g., independe nce or concom itance) is

though t to be an artifact of measu reme nt error. Altho ugh a

numb er of studies have shown that pleasant and unpleasant

emotions are independent (e.g., Bradburn 1969; Diener

and Emmons 1985; Zevon and Tellegen 1982), Green,

Goldman, and Salovey (1993) and Barrett and Russell

(1998) argue and prese nt findings demonstrating that emo-

tions are bipolar, once random or both random and system-

atic errors are taken into account.

Bagozzi, Wong, and Yi (1998) challenge the conclu-

sions made by G reen et al. (1993) and Barrett and Russell

(1998). They hypoth esize that bipolarity, independence,

and conc omitan ce depen d on gender, culture, and the tar-

get of one's emotions. Briefly, Bagozzi, Wong, and Yi

found that positive and negative emotions were highly

negat ively correlated for American women but highly

positively correlated for Chine se wome n. The form er pat-

tern is evidenc e for b ipolarity (i .e., either positive or nega-

tive emotion s occur bu t not both), the latter for concomi-

tance (i .e., both positive and ne gative emotions occur at the

same time). For m en, the correlations between positive and

negat ive emotions were m uch smaller in magnitude but in

the sam e direc tion across ethnicity, that is, slight negative

correlations were fo und for Ame rican men, slight positive

correlat ions for Chinese men. Thus, the relat ionship

between positive and negative emotions for men was

nearly independent. Bagozzi, Wong, and Yi argued that

differences in culture (i .e., a tendency for Americans to

view things in d ichotom ies or discrete categories, i .e., in

opposition; and a te ndenc y for Chinese to view things dia-

lectically, i.e., in balance or ha rmony ) interact with gender

differences (i .e., a tendency for wom en to be more knowl-

edgeable and skilled in the use of emotions than men) to

produce the divergent pat terns. The above findings

resul ted when people were asked to express how they fel t

at the moment according to their idiosyncratic reasons,

which is the standard procedure use d in the literature. That

is, the stimulu s for each person could be considere d het-

erogen eous across individuals. However, when America ns

and Chinese were asked to give their emotiona l reactions

to eating in fast-food restaurants (a common, singular

stimulus), positive and negative emotions were indepen-

dent for men and w omen al ike.

It is perhaps too early to give definitive recom mend a-

tions on which emotional scales to employ in empirical

work, but for now, we think that it is advisable to recom-

mend use of unipolar scales that ask respondents to

express to what extent each emotion describes their own

subjective feelings, rather than bipolar scales that can

obscure differences in emotional responses across the

various dimensions. Also, at least five, preferably sev en to

nine, scale steps should be used for each item to enhance

the chances that optimal distributional properties o f mea-

sures will be achieved. In addition, at least three, prefera-

bly more, i tems should be used for each emotional

subcategory.

A final measure ment issue we wish to raise is the fol-

lowing. To what extent are emotions blends o f categories?

For ease of discussion, we co nsider the categories of emo-

t i o n s p re s en t ed o n t h e c i r cu m p l ex . P l eas an t n es s -

unpleasantness migh t combine, for instance, with arousal

to produce different kinds or intensities of emotion. To take

a particular example, consider happiness, an instant of

pleasantness on the circumplex. Intense forms o f happiness

occur when pleasantness combines with high arousal:

elated, excited, enthusiastic, euphoric, gleeful, joyo us, ec-

static, and exultant are examples. Mild f orms o f happiness

occur when pleasantness combines with low arousal :

peaceful, calm, serene, or quietude of mind are examples.

Happiness, itself, might be at an intermediate level of

arousal. Other perspectives on blended emotions, based

not on a blend of arousal with emotion categories but

rather on combinat ions of basic emotions, can be found

in Izard (1991, 1992) and Plutc hik (1980).

What more general role, if any, does arousal play in

emotions (Bago zzi 1991b)? We turn to this issue next.

ROUS L

An early, influential point of view on em otions was pro-

fessed by James ([1890] 1950) who claim ed that bodily

changes foll ow directly the perception of the exciting fact,

an d.., our eeling o f the same changes as they occur lS the

emotion (p. 449, emphasis in original). For James, differ-

ent stimuli lead to different bodily responses (e.g., sweaty

palms, racing heart, etc.), these physiological responses

are then detected as bodily sensations in our mind, an d the

result is interpreted by us as emo tional experiences. B ut it

is important to note that James reserv ed this interpretation

for what he termed the coarser emotions (e.g., grief, fear,

rage, love ), which involve strong bodily perturbations; he

was less clear about what he termed the subtler emotions

8/10/2019 art%3A10.1177%2F0092070399272005

9/23

192 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

(e.g., mora l, intellectual, and aesthetic feelings (James

[1890] 1950:468).

An equal ly influent ial theory of emotions was proposed

by Schach ter and Singer (1962), wh o argued that emotion

is essentially bodily arousal plus a cognitive label one pro-

vides to diagno se his or h er felt arousal, a perspective con-

sistent with James's point of view. The idea is that we first

experience physio logical arousal, and, especially when w e

are unaware or uncertain of the origin of the arousal, we

look for evidence in the physical and social situation

accompanying the arousal to label our emotional state.

Although this theory had considerable impact in psychol-

ogy for nearly three d ecades, it has largely been discred-

i ted and has not received much support ing evidence

beyo nd Schach ter and Sin ger's original experiment (e.g.,

Manstead and Wagner 1981; Reisenzein 1983). One

except ion to the above observat ion is the frequent ly

repeated find ing that arousal misattributed to an extrane-

ous source intensifies emotions (e.g., Cantor, Bryant, and

Zillman 1974; Dutton a nd Aron 1974; Zillman 1971).

An im portant contribution of James ([1890] 1950) and

Schachter and Singer (1962) was the recognition that

arousal plays an essent ial role in emotion. Before we

address arousal mo re fully, we shou ld mention research in

marke ting that addresses the acquisition of affect, without

necessarily involving concepts of arousal or information

processing. Both in the practice of marketing and market-

ing research, considerable emphasis has been placed on

the effects of various stimuli on c onsum er behavior. Retail

store environment cues, advertising, background music,

brand names, packages, celebrity endorsers, and other

stimuli are frequently administered to produce emotional

reactions in consumers. The premise is that emotions or

moo ds trigger buying responses (e.g., Gardn er 1985; Hill

and Gard ner 1987).

If not by appraisal processes or direct arousal, per se,

how doe s presentation o f a stimulus under repetitive con-

ditions induce affect? A seductively simple explanation

was provided by Zajonc, who argued that when objects

are presented to the individual on repeated occasions, the

mere ex posure is capable of makin g the individual 's atti-

tude toward these objects more positive (Zajonc and Mar-

kus 1982:125). This mer e exposure effect has been

found primarily when the stimulus is simple and previ-

ously unknow n or else has l i t tle or no semantic content

(e.g., nonsens e syllables, foreign words o f Chinese char-

acters). One mechanism that has been offered to explain

the mere exposure effect is familiarity: we come to like

things that are familiar to us, perhaps because o f feelings

of security (Zajonc 1968). However, a full theoretical

explanat ion for the mere exposure effect has not been

developed. When a person is exposed to m eaningful s tim-

uli, i t has been m ore difficult o produce the mere exposure

effect (e.g., Obermiller 1985). This is, in part, a conse-

quence o f the cognitive processing that occurs in response

to awareness of the meaningful s t imuli . Repeated expo-

sure to a meaningful stimulus can lead to increased or

decreased positive or negative feelings, dep end ing on its

meaning. Mere exposure, thus, loses its utility in such

cases, both as a theory and a practical tactic.

Based loosely on the not ion that emotions or moods

induced by one s t imulus become at tached to another, some

researchers have investigated the effects of (a) music on

length of stay and money spent in supermarkets and res-

taurants (e.g., Mil lima n 1982, 1986) and (b) affectiv e tone

of stores on purchase intentions (e.g., Don ovan and Rossi-

ter 1982) and evaluations (e.g., Gardner and Simokos

1986). How can the observed at tachment of affect from

one st imulus to another be explained? Shimp (1991)

reviews seven studies in consum er research th at test vari-

ous facets of classical conditioning explanations. The id ea

behind classical conditioning is that the repeated pairing

of a conditioned stimulus (e.g., a new brand na me) w ith an

unconditione d stimulus (e.g., an attractive spokesperson)

will eventually lead to the new brand name, on its own,

stimulating the unconditioned response (e.g., positive

affect) originally induced by the u ncond itioned stimulus.

Very few studies have been performed in m arket ing that

confo rm to the conditions required to test classical condi-

tioning. It is unclear w hether classical condition ing studies

can be designed to rule out such rival explanations as

dema nd characteristics or cognitive interpretations of the

results. Allen and Janiszewski (1989) provide some evi-

dence that at least one type of cognitive med iation is neces-

sary for classical conditioning to occur: namely, subject

awareness of the cont ingency between the condi t ioned

stimulus and unconditioned stimulus. On the other han d,

classical conditioning, particularly for fear responses, has

been shown to involve unconsciou s arousal processes con-

nected with the amygd ala (LeDoux 1996).

Another way to explain the observed at tachment of

affect from one stimulus to another is by Z illman 's (1971)

excitation-transfer model. Briefly, Zillman propo sed tha t

exposure to one stimulus ma y produce arousal. If a second

stimulus is presented close on that also is capa ble o f pro-

ducing arousal on i ts own, the two sources of arousal may

comb ine to produce intensely experienced arousal. Und er

certain conditions (e.g., unawareness of the source of

arousal from the first stimulus, recency o f the second), a

person may attribute the arousal to the second stimulus.

Zillman (1983) interpreted arousal as undifferentiated

sympathe tic activation.

An issue in need of resolut ion is whether em otions can

occur without arousal. Cognitive theories of emotions

(e.g., appraisal theories) seem to allow tha t emotions ca n

be produce d by cognition alone, without arousal (e.g., Par-

rott 1988). But does arousal alway s accom pany the experi-

ence of emotions?

Recent research suggests that arousal is an essential

component of emotion and is manifest in neural systems in

8/10/2019 art%3A10.1177%2F0092070399272005

10/23

Bagozzi et al. / THE ROLEOF EMOTIONS 193

the brain. LeDoux (1996) reviews evidence suggesting

that there are at least five arousal system s in the brain con-

tributing com plexly to emotiona l experience. Four of these

are in regions of the brain stem and rely respectively on

acetylcholine, noradrenaline, dopamine, and serotonin for

activation. A fifth (the nucleus basalis) is in the forebrain

and also relies on acetylcholine to arouse cortical cells.

LeDoux notes that the arousal systems act in nonspecific

way s throughout the fo rebrain to make cells more sensitive

to incoming signals. In a sense, the nonspecific arousal

interacts w ith the information proc essing o f a particular

stimulus. T he am ygdala acts as a kind o f central pro cessor

and interacts with the prefrontal cortex (working memo ry

and attention), hippocampus (long-term explicit memory),

and sensory cortex (perception and short-term storage) to

influence emotional responses. The amygdala not only

influences cortical a reas of the brain but also receives input

from arousal networks (which themselves also influence

the forebrain) and feedback from bodily expression of

em otion s. In addition, signals from th e amygdala are sent to

muscles and internal organs and glands.

Mo st of the research to date into the role o f arousat sys-

tems and the amygdala in emotional behavior has been

limited to a small numbe r of emotions (e.g., fear). B ut it is

believed that each emotional response is mediated by sepa-

rate neural sy stems, although each m ay overlap or resem-

ble each other in many respects. In sum, LeDoux (1996)

and other brain researchers (e.g., Dama sio 1994) construe

emotions as biological functions of the nervous system

(see also Zajonc 1998).

To our know ledge, there has been little work to either

integrate or reconcile cognitive theories o f emotions with

neural and b iological theories. M uch remains to be done in

psychological research before we can make definitive

statements abou t the precise role of arousal in emotional

experience and behavior.

At least three studies have examined limited asp ects of

arousal in marketing-related contexts. Sanbonmatsu and

Kardes (1988 ) found that arousal ma y govern attitude for-

mation in persuasive message settings. Attitudes were

based on peripheral cues wh en respondents w ere highly

aroused but on argument strength when they were moder-

ately aroused. It is unclear whether arousal functioned

here to reduce inform ation-processing capacity of external

arguments or focused attention on internal reactions.

Bagoz zi (1994) found that consisten t with predictions

by knowledge-assembly theory (Hayes-Roth 1977),

arousal t rans formed a two-dimens ional , a f fec t ive-

cognitive representation of evaluations of giving blo od

into a one-d imen siona l, unitized representation. L ikew ise,

arousal increased the association between attitudes and

positive beliefs about the consequences of giving blood

and decrea sed the association betw een attitudes and nega-

tive beliefs. These pre dictions on the associations betwee n

attitudes and beliefs were explained by the implications of

spreading activation effects of arousal and coping

responses, wherein individuals attemp t to avoid negative

and facilitate positive associations of beliefs with atti-

tudes. In another study, Bagoz zi (1996) found that for atti-

tudes toward giving blood, high arousal tend s to enhance a

halo effect from attitudes to positive beliefs and reduce the

halo from attitudes to negative beliefs about the conse-

quences of giving b lood.

Clearly, arousal is a fundamental aspec t of behavior

related to emotions. We must a cknow ledge that appraisal

theories have not done a good, or at least complete, job o f

incorporating arousal into their frameworks. In their

defense, however, we should mention the following.

Appraisal theorists recognize that the intensity of emo -

tional experience consists of two componen ts: arousal and

self-control (e.g., Frijda 1994:120). Lik ew ise, resea rche rs

accept that autonomic nervous system and other physio-

logical processes at least accom pany subjectively felt

emotion s (e.g., Oatley 1992: 21) and that [i]f the criterion

of physiological activity w as eliminated fro m the defini-

tion, the concept of emotion w ould be left without one o f

the important response boundaries with which to distin-

guish it from nonem otion (Lazarus 1991:58-59). Bu t is

there more to arousal in emotion than this?

Some psycho logists and marketers have been quick to

dismiss Zajo nc's (1980) claim that preferen ces need no

inferences (see Cohen and Areni 1991:215-216; Lazarus

1982; Marcel 1983; Zajon c 1984). Howe ver, we be lieve it

is important, at the present, to recognize that emotional

meanings can be processed subconsciously, emo tions can

be activated automatically, and responses to em otions

(e.g., coping, action tendencies, actions) also can occur

automatically. W e leave open the possibility that emotio n

and cognition are best thought o f as separate bu t interact-

ing mental functions mediate d by separate but interacting

brain system s (LeD oux 1996:69; see also Oatley 1992,

chap. I). It appears that arousal is a key part of emotional

functions in the brain that underlies muc h o f its automatic-

ity. Cognitive appraisals and arousal ne ed to be better

incorporated into our theories of emotion.

EMOTIONS AS MARKERS MEDIATORS

AND MODERATORS OF CONSUMER

RESPONSES

Based on content, mo st advertisements can be divided

into two categories: (a) thinking ads, where f ocus is placed

on either factual information (e.g., prod uct attributes) or

utilitarian consequ ences of product/service use (e.g., sav-

ings in time or money) or (b) feeling ads, where conc entra-

tion is placed o n the e motio ns one wiI1 experien ce through

use or ownership o f a product (see Puto and W ells's [ 1984]

s i m i l a r d i s t i n c t i o n b e t w e e n i n f o r m a t i o n a l a n d

8/10/2019 art%3A10.1177%2F0092070399272005

11/23

194 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

transformationa l advertising). Rather than focu sing on the

stimulus, per se, it is more important to emphasize the

processes and experiences comprising a person's response

to ads, to better understa nd the emotional meaning o f ads

(e.g., Friestad and Th orson 1986).

Paralleling the above differentiation between types of

ads and the emot ional -cogni t ive d iv i s ion in menta l

processes me ntione d earlier, Batra and R ay (1986) devel-

oped a framework and coding scheme for classifying

affective responses to ads, as a com plemen t to cognitive

responses. Specifically, Batra and Ray (1986) identified

three positive affective response categories: surgency-

elation-vigor/activation (SEVA), deactivation, and social

affection feelings. The SEVA feelings refer to upbeat,

happy mood reactions (e.g., the ad's music was 'catchy,'

the ad was 'fun to watch or breezy, ' o r . . . a l ikable use of

humo r ); deactivation includes soothing, relaxing, quiet,

or pleasing reactions; and social affection encompasses

feel ings of warmth, tenderness , and caring (Batra and R ay

1986:241). The three positive affective responses w ere

used along with six cognitive responses (i .e., support argu-

ments, counterarguments, execution discounting, execu-

tion bolstering, neutral distracters, and other reactions) in

a s tudy o f the impa ct of televis ion commercials on con-

sumers. A total of 12 percent o f reactions to ads were clas-

s i fied as posi t ive affect : SEVA (3.7%), deact ivat ion

(2.5%), and social affect (6.1%).

One use of emotional react ions in the above sense

might be as markers or indicators of the effectiveness of

advertising copy, particularly with respect to the overall

persuasiveness of the ad, the appeal of spokespersons,

evaluation of particular product claims, and appraisals of

other aspects of the ex ecution (Wiles and Cornw ell 1990).

Also, the program surrounding an ad (e.g., happy versus

sad content) has been found to have main effects on one's

evaluation of an ad and recall (e.g., Goldberg and Gorn

1987). Future rese arch is neede d to identify how progra m

conten t and advertising appeals interact to influence con-

sumer emo tional responses.

A research question that has received quite a bit of

attention in recent years is how and to what extent emo-

tional reactions to ads influence consume r decision mak-

ing. Most ofte n these reactions have been mea sured as atti-

tudes (e.g., l iking) toward the ad (e.g., Brown and Sta yma n

1992; Mitchell and Olson 1981; Shimp 1981).

Attitude tow ard the ad (A d) is thought to be a function

of feelings (and thoughts) about the ad itself (e.g., Batra

and Ray 1986; Mac Kenz ie, Lutz, and Belch 1986). In fact,

Batra and Ray (1986) found that the three affect ive

responses discu ssed above significan tly predicted A deven

after controlling fo r the effec ts of cognitive responses. A

number of researchers have examined the condi t ions

under w hich em otions influence A~. For instance, under

low-involvement viewing conditions, feelings about the

ad have been found to be more important determinants of

A~ than thoughts (e.g., how informative or useful the ad

is), but under higher involvement viewing, both feelings

and thoughts may be importan t (e.g., Miniard, Bh atla, and

Rose 1990). Presumably, high involvem ent promote s cog-

nitive processing of the usefulness of the ad and its

content.

A majori ty of research has addressed the effects of A~,

especially on attitudes toward the bra nd (Ab). Classic atti-

tude theory maintains that A b s a func tion of beliefs about

brand attributes or consequence s of product use. Mitchell

and Olson (1981) and Shimp (1981) were the first to find

that A~ provided additional explana tory pow er for A b over

and above brand beliefs (see also Edell and Burke 1987).

Batra and Ray (1986) found, however, that affect ive

responses toward the ad in fluenc ed A b only indirectly

throu gh Am.

Some research has addressed the condi t ions under

which Aad nfluenc es A~. For examp le, Brow n and Staym an

(1992) revealed in their meta- analysis that the effects of

Aad on A bare greater for novel than well-know n brands and

for durable and other goods versus nondurables. So me evi-

dence also exists showing that Aad influences A b ndirectly

through its effect on beliefs about the brand (M acKe nzie et

al. 1986). Finally, Staym an and Aaker (1988) showe d that

repetition governs the feelings to Aa~ relationship. U nder

levels of low versus high repetition, feelings have a

stronger effect on Aad. This m ay b e a c onsequ ence of

greater information processing under high versus low

repetition.

In addition to the transfer of affec t from ad emotion s to

A d, research shows that brand names and feelings toward

ads can become l inked in memory. Stayman and Batra

(1991) found that respondents exposed to an affective, as

opposed to an argument, ad w ere able to retrieve brand atti-

tudes faster, when prime d with the brand nam e. In addi-

tion, the positive retrieved affect had a strong er influenc e

on choice in low- than in high -involvem ent contexts. In a

second study, Stayman and Batra dem onstrated that view-

ers of an ad who were in a positive affective state more

strongly evoked the affect when given the brand name as a

subsequent retrieval cue than viewers exposed to the ad

while n ot in a positive affect state.

Olney, Holbr ook, a nd B atra (1991 ) investiga ted a hier-

archical model explaining advertising viewing time. The

effects of ad emotions (i .e., pleasure and arousal) were

mediate d by A d and reactions to a d content. Interestingly,

arousal had both indirect and direct effects on viewing

time, even after co ntrolling for ad conten t and Aad. Arousal

was measu red by self-reports.

Emotions have been found to serve as moderators in

their impact on A b. Batra and Stay man (1990), in one o f the

few studies to examine mood and print ads, found that

positive mood s enhanc e A b hrough the ir interaction with

two cognitive processes: (1) a bias against the generation

of negat ive thoughts (such as evoked by w eak arguments),

8/10/2019 art%3A10.1177%2F0092070399272005

12/23

Bagozzi et al. f THE ROLEOF EMOTIONS 195

leading to a mo re favorab le evaluation of message argu-

ments, and (2) a reduction in total cognitive elaboration,

making processing more heuristic than systematic (pp.

212-213). An interesting finding was that positive moods

seem to reduce counterargumentation when we ak argu-

ments are use d in ads (see also Worth and Mackie 1987).

Batra and Steph ens (1994) also investigated the moderat-

ing effec ts of m ood on A~. Specifically, they show ed that

mood and motivation (the latter conceived as degree of

relevance of the product category for consumers) inter-

acted to go vern A b when consum ers watche d television

ads. The greatest impact on A~ occurred when positive

moods en sued under conditions of low motivation. The

rationale is that positive moods and lo w motivation sup-

press counterargumentation (and therefore lead to more

favorable Ab) in comparison with high-motivation

conditions.

Following the elaboration likelihood model (ELM)

(e.g., Petty and Cacioppo 1986), we might expect emo-

tions to have one or both of two effects. Emotions might

operate centrally to influence cognitive processes (e.g.,

with regard to argumentation), or emotions might function

peripherally (perhaps through associative or affect transfer

mechanisms). Early predictions under the ELM took an

either-or perspective and stressed that when the processing

of information in a comm unication is low (e.g., due to low

motivation, distraction, low need for cognition, w eak argu-

ments), emotiona l content in the comm unication (e.g., an

attractive spokesperson) is processed directly and trans-

fers to, or influences, attitude toward the produc t or mes-

sage. When issue-relevant thinking is high, attitude

change is thought to be a function of the balance of pros

and cons in the commu nication, a largely rational proc ess,

and emoti on ma y not be a factor.

Over time, the role of em otion in persuasive communi-

cation has bee n found to be more com plex than the simple

central versus peripheral pro cessi ng alternatives spe cified

in the ELM (e.g., W egener and Pe tty 1996). For example,

when the likelih ood of information processing is moder-

ate, emotions have b een found to affect the extent to which

arguments in a comm unication b ecom e elaborated. Posi-

tive (versus neutral) moo ds tend to lead to less processing

of arguments (e.g., Bless, Bohner, Schwarz, and Strack

1990; see also discussion below on the effects of emotions

on cognitive processes). By contrast, when peop le process

the arguments in a messag e closely, moo d might bias

information proc essin g or even function as an argument

itself (Forgas 1995). Another explanation for mood effec ts

is that sad or neutral (versus happy) moo ds lead to more

effortful processing, w hich is believed to be do ne sponta-

neously (e.g., Bohner, Chaiken, and Hunyadi 1994). Note

that this prediction seem s to conflict with the observations

made by Schaller and Cialdini (1990), who analyzed

mostly nonpersuasive com munica tion studies. So me clari-

fication for the discrepan cy in interpretations can be seen

in a study by Wegener, Petty, and S mith (1995). W egener

et al. (1995) discovered that happy (versus sad) moods

lead to the processing of more arguments in a message

wh en a proattitudinal/uplifting posi tion wa s taken, but

happy (versus sad) moods lead to less processing of argu-

ments when a

counterattitudinal/depression

position

was taken. Wegener et al. (1995) explain the findings, in

part, by suggesting that when in a happy mo od, peop le try

to maintain their moo d and thus process less of the coun-

terattitudinal/depression content.

In addition, interactions of emotions sometim es occur

with other variables, such as motivational or ability fac-

tors. For instance, W egener, Petty, and Klein (19 94) found

that for people high in need for cognition, messages

framed positively (i.e., good things will happen i f peop le

adopt the advocacy) were more persuasive when the audi-

ence was happy (versus sad), but messages framed nega-

tively (i.e., ba d things w ill happen if peopl e do not adop t

the advocacy) were more persuasive when the audience

was sad (versus happy). Wegener et al. (1994) explained

these findings by claiming that good outco mes seem bette r

and more likely to occur while in a good (versus sad)

mood, and bad outcomes seem wo rse and more likely to

occur while in a sad (versus happy) mood.

Finally, even when the likelihood of information pro-

cessing is low (due, e.g., to low need fo r cognition or low

relevance of a product for a consumer), the moo d o f the

audience can have a direct effect on A b, but w hen th e likeli-

hood of information processing is high, the m ood o f the

audience affected message-generated thoughts consistent

wi th the mood (Pet ty , Schumann, Richman, and

Strathman 1993). Mood w as induced by a television pro-

gram o r music.

THE EFFECTS OF EMOTION 1:

INFLUENCES ON COGNITIVE

PROCESSES

A person's emotional state can influence various

aspects of information processing including encodin g and

retrieval of information, different strategies used to

process information, evaluations and judgm ents, and c rea-

tive thinking. In this section we examine the influence o f

affective states o n various aspe cts of cognitive processe s.

Emotion Mood Effects on Memory

The influence of mood states on memory can be

broadly classified into three categories: retrieval effects,

encoding effects, and state-dependent learning effects. In

the following section, we review s ome o f the studies that

have investigated these effects and discuss the mechanism

by w hich affect influences m emory.

8/10/2019 art%3A10.1177%2F0092070399272005

13/23

196 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

Retrieval effects. Affec t has been shown to influenc e re-

trieval of informa tion, whereby persons in a positive mood

state at the time o f retrieval have been f ound to show supe-

rior recall of positive material learned during encoding,

relative to neutral or negative material (Isen, Shalker,

Clark, and Karp 1978; Laird, Wagener, Halal, and Szegda

1982; Nasb y and Yando 1982; Teasdale and Russell 1983).

For exam ple, Isen et al. (1978) had respond ents study posi-

tive, negative, and neu tral words. E ither positive, neutral,

or negative mood states were induced in these respon-

dents. Respondents in the positive-mood condition re-

trieved more positive words compared with neutral or

negative words. Isen et al. (1978) suggest that thinking

about mood-incongruent material involves shifting one s

focus, which is cognitively taxing, and therefore people

are more likely to focus on mood-c ongrue nt material. An-

other mec han ism proposed to explain the retrieval effects

of positive affect suggests that positive mood at the time of

retrieval functions as a cue that primes the positive mate-

rial in memory, making these material more accessible

(Isen 1989; Isen et al. 1978; also see Tulving and Pearl-

stone 1966 for a discussion on the effe cts of priming on ac-

cessibility). The easier accessibility of positive material

may then influence other cognitive processes such as

evaluations and decision mak ing, and also subsequent be-

haviors. While retrieval effects have been replicated by

many researchers using different mood induct ion and test-

ing techniques, a few prom inent studies have failed to de-

tect retrieval effects (Bower, Monteiro, and G illigan 1978;

Bower, Gilligan, and M onteiro 1981), leading Isen (1984)

to speculate that this failure may have been c aused by the

specific material and induction methods (such as hypno-

sis) used in these studies.

Encoding effects. Mo od states have also been shown to

exhibit encod ing effects wh ereby the a ffective state at the

t ime of learning is associated with superior memory for

similarly valenced material (Bowe r and Cohen 1982; For-

gas and Bowe r 1987). Nas by and Yando (1982) foun d that

positive mo od a t the time of learning led to an improved re-

call of positive material at a later point in time regardless of

the mood state at the time of recall. Bower et al. (1981)

found evidence for the encoding effect of both posi tive and

negat ive affect. Respondents were made to feel happy or

sad and then read descriptions o f various psychiatric inter-

views. Happy respondents learned many m ore happy facts

than sad facts, while sad respondents learned many more

sad facts than happy facts.

How can the encoding effects of mood be explained?

Bow er and colleagues (e.g., Bow er and Cohen 1982) have

suggested that mood-congruent material is likely to be

m o r e s e m a n t i c a l l y e l a b o r a t e d r e l a t i v e t o m o o d -

incon gruen t material. Forgas and Bow er (1987) found that

in impression formation situations, sad individuals spent

more t ime examining negat ive ra ther than pos i t ive

information, and they subsequently recalled the negative

information better. Conversely, happy individuals spent

more time on the positive information and recalled that

better at a later time. The greater levels of associations

evoked by mood-congruent material may have caused a

more extensive elaboration, which, in turn, requires m ore

time. However, Isen et al. (1978) and Srull (1983) failed to

find any encodin g effects of affect.

State dependent learning effects. A third mem ory ef-

fect of mood is the s tate-dependent learning effect of af-

fect, where any material regardless o f its affective valence

learned under a particular mood state is recalled better

when the p erson is again in that affective state (Bartlett ,

Burleson, and Santrock 1982; Ba rtlett and Santrock 1979;

Bower, Monteiro, and Gilligan 1978; Bower, Gilligan, and

Monteiro 1981). Bow er et al. (1978) had resp ondents learn

two sets of words, one while they we re in a positive affec-

tive state and the other in a negative mood . Wh en respon-

dents who learne d the two lists in different moods re called

the words in the wrong mood (e.g., when words learne d in

a positive mood were recalled w hile respondents were in a

negative mood), they experienced interferenc e and the av-

erage recall rate was less than 50 percent. When respon-

dents who learne d two lists in different moods rec alled the

words in the correct mood, the average recall rate was

more than 70 percent. Control respondents who learned

and recalled both lists while in the same m ood showe d an

average recall rate between 50 an d 60 percent. Bow er and

Cohen (1982) suggest that the respondents mood at the

t ime of learning becomes associated with the learned m a-

terial and that these associations facilitate the recall of

learned material when the mood state at recall matches the

moo d state at encodin g (also see Bower 1981 for a descrip-

tion of his semantic-network theory).

Evidence for mood state-dependent learning has been

ambiguous. Many studies have failed to find any state-

dependent effects of mood state (Bow er and May er 1985;

Isen et al. 1978; Laird et al. 1982; Nasb y and Yando 1982).

Eich and Birnbaum (1982) and Isen (1984, 1989) have

suggested that when the material to be learned has sem an-

tic meaning, the stimulus will be enc oded ac cording to this

meaning, and the influence of the mood state in the encod-

ing and subsequent retrieval processes will be minimal.

However, when the stimulus lacks meaning, contextual

cues such as affective states at the time of learning m ay be

more s t rongly encoded with the learned material . At the

t ime of retrieval , these memory i tems, which have few

semantic associations, are more primed by the matching

affectiv e state at recall.

Asymmetric effects of positive a nd negative moods.

While positive affective states have been shown to have

significant influences on recall, nega tive affect has some-

t imes been found to have ei ther no effect or a much smaller

8/10/2019 art%3A10.1177%2F0092070399272005

14/23

Bagozzi et al. / THE ROLEOF EMOTIONS 197

effect on the recal l of negat ive material from m emory.

Asymm etric effects of posi tive and negative moods have

been fou nd for retrieval effects (Isen et al. 1978; Nasby a nd

Yando 1982; Teasdale and Foga rty 1979), enc oding effects

(Nasby an d Yando 1982), and state-depend ent learning ef-

fects (Bartlett and Sa ntrock 1979; Bartlett et al. 1982). Isen

(1984) speculated that positive affect is structured in a

broad and extensive manner (i .e., highly interconnected

with other mem ories), w hile negative affect is more nar-

rowly and less w ell connected with other material, and that

specific negative affective states such as an ger and sadness

may be organized separately in memory. This, in turn,

would ma ke i t difficult for any given negative mood to act

as an effective retrieval cue. It is not readily apparent why

negat ive affect would be less wel l connected in mem ory

and positive affect more widely interconnected as Isen

suggests, conside ring that negative affect usually signals

problematic environmental conditions that may require

problem so lving (Schwarz an d Clore 1983; Wegene r et al.

1995). One of the side effects of this hypothesized in-

terconnectedness of positive mem orie s is the greater crea-

tiveness and cognitive flexibility demo nstrated by people

in positive mood s, a topic we revie w briefly later in this

article.

Mood maintenance and repair have also been put for-

ward as exp lanations for the asym metric effects of positive

and negative moods. Ise n (1984) has suggested that happy

individuals attem pt to prolon g their positive affective state

by foc using on the positive aspects of their stimulus (mood

maintena nce), wh ile individuals in a negative mood try to

improve their situation by not focusing on negative mem o-

ries (mo od repair). Isen (1989) noted that in some of the

studies showing sym metrical effects of positive and nega-

tive mood (Bower et al. 1978, 1981), respondents were

instructed to maintain their induced moods, and this may

have discouraged them from engaging in mood repair

strategies. However, the mood repair explanation is not

without problems. If sad respondents engage in mood

repair, why is there no evidence for mood-incongruent

recall effec ts? After all , an effective strategy to improve a

depressed m ood state is to engage in pleasant thoughts and

memories. Yet, evidence seems to show that negative

moo d inhibits the recall of positive mem ories (e.g., Isen et

al. 1978). The comp eting explanations of memo ry struc-

ture differences and m ood maintenance/repair have also

b een u s ed t o ex p l a i n d i f f e r en ces i n i n fo rm a t i o n -

processing strategies of happy and sad individuals, a topic

we d iscuss later in this article.

Mood C ategorization and Creativity

Various studies investigating the influence of mo od on

categorization have found that people in positive mood

states, compared with those in neutral or negative mood

states, te nd to be better at integrating inform ation, finding

relationships am ong stimuli, and at find ing creative solu-

tions (Isen and Daubman 1984; Isen, Daubman, and

Nowicki, 1987; Isen, Johnson, Mertz, and Rob inson 1985;

Isen, Niedenthal, and Cantor 1992; Murray, Sujan, Hirt,

and Sujan 1990). For example, respondents in positive

mood condi t ions tended to group a wider range of neutral

stimuli together (Isen and Daubman 1984). They also

rated words such as

cane ring

and

purse

as being better

exemplars of the category clothing, than did the neutral-

mood respondents . Murray et al . (1990) found that

positive-mood respondents, compared with respondents

in other mood states, formed broader categories when

focusing on similarities among exemplars and narrower

categories when focusing on differences, prompting them

to suggest that what posi tive mood promotes is not broader

categorization but rather cognitive flexibility. Isen et al.

(1985) found that positive-mood respondents tended to

give more unusual responses to neutral words in word

association tests. For example, in response to the word

house

positive-mood respondents were more likely than

were neutral-mood respondents to mention unusual first

associates such as

security residence

and

apartment

sug-

gesting that positive-mood states may influence cognitive

organization, resulting in more flexible interpretation of

relationships among stimuli. This cognitive flexibility also

results in enhanced creativity. Isen et al. (1987) foun d that

respondents in positive-mood states outperf orme d those in

neutral- and negative-mo od states on tests requiring crea-

tive solutions. These researchers suggest that positive-

moo d respondents were better at creative problem solving

since such tasks required the ability to see relatedness

among seemingly unrelated stimuli, and as we have seen

earlier, positive affect results in mo re fle xible cognitive

organization.

Mood Effects on Evaluation

One of the best-recognized and most robust effects of

mood is i ts influence on evaluat ion. Individuals in

positive-mood states have been shown to evalua te stimuli

more positively than individuals in neutral- or negative-

mood states, whether the stimuli being studied are other

people (Clore and Byrn e 1974; Forgas and Bow er 1987),

consumer goods (Isen et al. 1978; Srull 1983), life satis-

faction (Schwarz and Clore 1983), or past life events

(Clark and Teasdale 1982). The reliability of moo d effects

on evaluation is perhaps best illustrated by the fact that

evaluat ions of the pleasantness of neutral /ambiguous

stimuli are used as checks for mood manipulation (e.g.,

Isen et al. 1985, 1987). Isen et al. (1978) fo un d that respon-

dents in whom posi t ive mood was induced were more

likely to rate their cars and televisions more favorably

8/10/2019 art%3A10.1177%2F0092070399272005

15/23

198 JOURNALOF THE ACADEMYOF MARKETINGSCIENCE SPRING 1999

compared with respondents in neutral-mood states. In the

case of memory-based evaluations, if the recalled infor-

mation is biased by the mood (as discussed in the follow-

ing section), then evaluations that follow will be biased

too. On-line processing of evaluations can also be influ-

enced by mood states through the retrieval of information

congruent with the mood (Clore, Schwarz, and Conway

1994).

A competing explanation based on the feelings-as-

information model (see Schwarz 1990; Schwarz and Bless

1991; Schwarz and Clore 1983) suggests that individuals

may assume that their mood states are affective reactions

to the object being evaluated and thus base their evalua-

tions on their affective states. For example, a happy indi-

vidual when asked to evaluate a painting may ask the ques-

tion, How do I feel? and infer that his or her positive

mood is a reaction to the painting and therefore come to the

conclusion that he or she likes the painting. The feelings-

as-information hypothesis suggests that when individuals

attribute their mood state to something else other than the

object being evaluated, the effect of mood on evaluation

should disappear. Schwarz and Clore (1983) found sup-

port for this hypothesis when they showed that although

people called on sunny days reported more life satisfaction

than people called on cloudy days, the differences disap-

peared when the interviewer casually mentioned the

weather to the individuals.