Cognitive, Affective, And Attribute Bases of the Satisfaction Response

Cognitive, Affective, and Attributeof the Satisfaction Response

RICHARD L. OLIVER*

An attempt to extend current thinking on postpurchase response to include attributesatisfaction and dissatisfaction as separate determinants not fully reflected in eithercognitive (i.e.. expectancy disconfirmation) or affective paradigms is presented. Inseparate studies of automobile satisfaction and satisfaction with course instruction,respondents provided the nature of emotional experience, disconfirmation percep-tions, and separate attribute satisfaction and dissatisfaction judgments. Analysisconfirmed the disconfirmation effect and tbe effects of separate dimensions of positiveand negative affect and also suggested a multidimensional structure to the affectdimensions. Additionally, attribute satisfaction and dissatisfaction were significantlyrelated to positive and negative affect, respectively, and to overall satisfaction. It issuggested that all dimensions tested are needed for a full accounting of postpurchaseresponses in usage.

T he study of postpurchase processes as central tothe consumer's decision to repeat-buy products

has attracted increasing attention from consumer re-searchers (e.g.. Bearden and Teei 1983; Bolton andDrew 1991: Cadotte, Woodruff, and Jenkins 1987;Churchill and Surprenant 1982; Oliver 1980; Oliver andDeSarbo 1988: Tse and Wilton 1988: Westbrook 1987).Interest in this topic results from the high likelihoodthat many purchases in any one consumer's routine aresecond or n-time purchases and not initial decisions.

A satisfactory purchase experience would appear tobe one requirement for the type of continued interestin a product that might lead to repeat purchasing. Tothis end, researchers have turned increasing interest tothe study of consumer satisfaction/dissatisfaction (CS/D). Until lately, however, the literature has focused ondissatisfaction, especially complaining behavior, thedemographic correlates of satisfied/dissatisfied con-sumers, or the attribute performance correlates ofproduct satisfaction judgments.

The present article attempts to extend this literaturewith the following additions to existing work. Drawingon previous conceptualizations of satisfaction in con-sumption, focus is dravi'n to the role of attributes ascausal agents for affect in this process. Specifically, sep-arate dimensions of positive and negative affect are hy-

* Richard L. Oliver is professor, Owen Graduate School of Man-agement. Vanderbilt University, Nashville. TN 37203, He thanks thereviewers for providing constructive comments on previous versionsof the manuscript. Appreciation is also expressed to the Center forServices Marketing and the Dean's Fund for Faculty Research oftheOwen Graduate School of Management of Vanderbilt University forproviding partial support for this project.

418

pothesized, separated into meaningful subcomponents,and related to summary occurrences of satisfying ordissatisfying attribute experiences, with the goal of ex-panding perspectives of the satisfaction process to itsattribute-determined origins.

To achieve this end, the article first describes the CS/D literature and outlines present knowledge. Next, thiscurrent perspective is viewed in light of new develop-ments in affect researeh. A conceptual attempt is thenmade to integrate attribute experience with existingframeworks, as represented by the expectancy discon-firmation and performance model with affect augmen-tation. The proposed framework is then tested in twosamples of consumers with a causal modeling approach.Results are then analyzed, and suggestions for futureresearch drawn.

STRUCTURE OF THE SATISFACTIONRESPONSE

Cognition in SatisfactionWhile personality and attribute correlates of sum-

mary satisfaction/dissatisfaction judgments are valuablefrom a segmentation or diagnostic perspeetive, they dolittle to reveal the psychological processes that motivatea satisfaction response. That is, how do consumers pro-cess attribute experiences into summary form that laterinfluences satisfaction? Early strides toward answeringthis question were made when the expectancy discon-firmation model was adopted. In this process (see Oliver1980, 1981; Tse and Wilton 1988: Yi 1990), consumersare posited to form preconsumption expectancies, ob-serve product (attribute) performance, compare per-formance with expectations, form disconfirmation per-

1993 by JOURNAL OF CONSUMER RESEARCH, Inc. • Vol. 20 • December 1993All rights reserved. 0093-5301/94/2003-0005S2.00

ATTRIBUTE BASES OF SATISFACTION 419

ceptions. combine these perceptions with expectationlevels, and form satisfaction judgments. According tothis view, disconfirmation is the most immediate influ-ence on satisfaction.

Other works show that the process above is morecomplicated than originally proposed. Churchill andSurprenant (1982). for example, demonstrated that, forthe durable product tested in their study (a VCR), per-formance appears to impact CS/D directly rather thanthrough disconfirmation. This finding has been repli-cated by Tse and Wilton (1988) for compact disc playersand by Bolton and Drew (1991) in the category of con-sumer telephone services. Moreover. Oliver and De-Sarbo (1988) conducted an experiment using securitytransactions and showed that a significant direct per-formance effect can operate in tandem with the discon-firmation effect. Apparently, performance influencesCS/D through two mechanisms, directly via consumerobservation of good and bad performance and indirectlyas an input to the disconfirmation comparison (Boltonand Drew 1991).

Additional variance in CS/D has been explained byinclusion of a direct effect of expectations or norms.Oliver (1981) suggests that this results from a framingor assimilation effect whereby the expectation level es-tablishes the (high or low) baseline around which sat-isfaction judgments are made. Other standards of com-parison, including ideals and brand-based norms (see,e.g., Cadotte et al. 1987; Tse and Wilton 1988), havealso been suggested.

Additional cognitive themes such as that of equity(see, e.g., Oliver and Swan 1989) and attribution (Folkes1990) have also been proposed as satisfaction deter-minants but have not generated the same level of in-terest in CS/D researeh. Generally, consumers arethought to be more satisfied when they perceive fair(i.e.. equitable) treatment and when they attribute fa-vorable outcomes to themselves and unfavorable onesto others. While these perspectives are promising, theyare not studied here, although attribution is addressedindirectly. These concepts have been investigated,however, as integral to a multifactor satisfaction design(see Oliver and DeSarbo 1988).

As generally expressed, the processes outlined in thepreceding discussion are thought to be primarily cog-nitive. The expectation formation process, the com-parison of perf"ormanee to expectations, equity judg-ments, and causal attributions are mostly conscious,overt activities that consumers may or may not perform.Nevertheless, evidence for each of these paradigms isgrowing in the CS/D literature (see, e.g., Yi 1990).

Affect Augmentation in SatisfactionNot well represented in the cognitive perspectives is

consideration of the affective basis of satisfaction.Westbrook (1987; Westbrook and Oliver 1991) hasmade progress in this area. Westbrook (1987) posited

FIGURE 1

COMBINED COGNITIVE AND AFFECT-AUGMENTEDCS/D MODEL

Expectation!

Attrlbut*Performanc* Satisfaction

Positlve/NegallvaAftecl

Attribution

Equity/Inequity

NOTE.—Constructs and relations in bold represent the CS/D submodel testedin the present study.

that consumers form two summary affect states, onebased on the positive affects in consumption and theother on the negative affects. Using Izard's (1977) Dif-ferential Emotions Scale (DES), he showed that the af-fects of joy and interest load on a factor separate fromthat of anger, disgust, and contempt, and that these fac-tors are relatively uncorrelated. Moreover, both weresignificantly related to satisfaction in the expected di-rection. These relationships held up to the introductionof cognition (i.e.. disconfirmation) in the satisfactionequations. Similar results obtained in Westbrook andOliver (1991) except that interest appeared separatefrom joy in that analysis.

A theoretical perspective on affect in consumptionwas presented in Oliver (1989), based on postpurehasecausal attributions. Using Weiner's (1986) frameworkfrom social psychology, Oliver described how consumeraffect can arise as "attribution dependent." Specifically,attributions defined in the context of Weiner's locus-stability-controllability matrix are known to evoke spe-cific affects depending on whether the (purchase) out-come is defined as a success or failure (Weiner. Russell,and Lerman 1979). An example in marketing where anexternal and controllable failure on the part of a firmresults in consumer anger appears in Folkes, Koletsky,and Graham (1987).

On the basis ofthe CS/D literature reviewed in thepreceding discussions of cognitive and affect influences,the satisfaction response is thought to be representedby the constructs shown in Figure I. The cognitive an-tecedents include expectations, performance, diseon-firmation, attribution, and equity/inequity. Further, the

420 JOURNAL OF CONSUMER RESEARCH

model shows affect as augmenting these variables in theprediction of satisfaction and, in addition, shows affectas mediating an indirect attribution influence in accor-dance with Oliver (1989).

The emergence of affect as a proposed component ofpostpurchase expression is not unusual in light of otheremerging work in the consumer area (e.g., Batra andHolbrook 1990: Havlena and Holbrook 1986). Earlierwritings on consumer affect, however, can be traced toBradburn (1969) in the quality-of-life area. Bradburnwas, perhaps, the first to propose an "affect balance the-ory." which recognized that the positive experiences oflife (e.g., excitement, pleasure) are not necessarily in-versely correlated with negative experiences (e.g., lone-liness, boredom). Generally, positive and negative affecthave been found to make independent contributions tolife satisfaction/dissatisfaction judgments in the lifequality literature (e.g., Horley and Little 1985). Thisindependence argument has taken two forms. In thefirst, researchers have raised the issue of how the con-figural structure of affect could accommodate indepen-dent positive and negative affect dimensions. A secondquestion relates to the conceptual mechanism that couldexplain their independent operation.

Conftgural Structure. With regard to the dimen-sionality question, theoretical perspectives on affect'sstructure have been presented by Russell (1980) andWatson and Tellegen (1985). Russell's "circumplexmodel of affect" posits two dimensions of affectivestructure, pleasure/displeasure and arousal/boredom.If one narrowly defines affect as pleasure, then pleasure(positive affect) and displeasure (negative affect) will bepolar opposites on the same continuum and, thus, neg-atively correlated. In addition, the Russell model as-sumes that (pure) pleasure and displeasure are inde-pendent of (orthogonal to) arousal.

Watson and Tellegen (1985) have provided a secondperspective by noting that pleasure is only one com-ponent of affect. By rotating the axes ofthe pleasure-arousal space by 45° (without changing the locationsofthe affects in the scaling solution), they are able todefend axes of independent positive and negative affect.The resulting axis descriptions are (1) aroused positiveaffect (e.g.. delight) and (2) aroused negative affect (e.g.,anger). One conceptual attraction of this interpretationis that affect appears more complex than a simple plea-sure explanation would suggest; pleasure is viewed onlyas moderately aroused positive affect (cf. Diener andEmmons 1984).

Conceptual Mechanism. This dimensional expla-nation is consistent with theory and data relating to themechanism by which affect operates in everyday life.Most references on this subject rely on the aforemen-tioned Bradburn (1969) affect-balance theory or the"independence hypothesis." which maintains thatevents in life alternate between the positive and nega-tive, and that instances of one do not preclude occur-

rences of the other. This argument would seem es-pecially relevant to product and service consumption,where numerous and varied attributes exist at differentlevels of abstraction. Because each attribute is a poten-tial source of pleasure or frustration, the likelihood ofpositive and negative experience is enhanced, a premiseunderlying the use of multiattribute models of attitude.This perspective is developed further below.

An Extension ofthe Affect PerspectiveThe role of events (e.g,, attribute performance ex-

periences) as causal agents for positive and negative af-fective states has not been well eonceptualized in theCS/D literature. Via a mechanism similar to the emer-gence of attribution-dependent affect, research in theaffect literature (e.g., Scherer and Tannenbaum 1986)suggests that various affects are also event specific. Thatis, different types of everyday experiences may directlytrigger different types of affect. For example, a loss (e.g.,financial, emotional) would cause sadness while pleas-antly surprising product performance would cause de-lighL This analogy is extended to the summary attributelevel, where the sum of positive product experiences(i.e., satisfactory attribute performances) should relateto positive affect, and negative experiences (i.e., dissat-isfactory attribute performances) to negative affect. Thefollowing discussion elaborates this additional perspec-tive, comparing it to the present literature, and expandsthe attributional basis ofthe affect response suggestedin Oliver (1989) and Westbrook (1987),

CONSUMER SATISFACTION/DISSATISFACTION MODEL

EXTENSION

The "Attribute Basis" of SatisfactionPrior research on affect in consumption (e.g., West-

brook 1987: Westbrook and Oliver 1991) has not elab-orated on how consumers could entertain both positiveand negative affect toward the same product simulta-neously. Yet regressions of satisfaction on positive andnegative affect in both of these studies clearly showconcurrent effects. The explanation draws on the Brad-burn (1969) affect-balance theory, as discussed, whichwould propose the simultaneous operation of multifac-eted product or service attributes, thereby providing dif-ferentially valenced product experience. A restaurantdining eneounter, for example, evokes both negativeand positive affective reactions because of the com-plexity of this type of service (Derbaix and Pham 1991).It is proposed, therefore, that dual positive and negativeaffect influences in consumption can arise from dualconcurrent positive and negative experienees at the at-tribute level. That is, a range of multifaceted productfeatures allows for varying experiences with the productor service, some negative, some positive. These expe-


riences become the source of summary positive andnegative affect.

Attribute Satisfaction versus Attribute Performance.Up to this point, the phrase "attribute experience" hasbeen used to refer to the consumer's perception of aproduct or service feature. I now address the issue ofequating attribute experiences of a widely disparate na-ture. It is proposed that these experiences can be com-pared, not at the level of "performance units" (since,for example, the noise level of a car's interior is notcomparable in decibels to the fuel economy in milesper gallon), but at the level of "satisfaction units." Thus,positive (negative) affect should be a function of theamount of positive (negative) attribute-level units ofsatisfaction. Attribute satisfaction, then, is the consum-er's subjective satisfaction judgment resulting from ob-servations of attribute performance and can be consid-ered to be the psychological fulfillment responseconsumers make when assessing performance.

Thus, a distinction is made between attribute per-formanee and attribute satisfaction. As originally ex-pressed by Bettman (1974). attribute satisfaction is afunction of a decision heuristic applied to attribute per-formance and, as such, is a higher-order concept in theconsumer's product judgment. According to Bettman.performance itself is not sufficient to cause attributesatisfaction; rather, consumers are thought to assessperformance in the context of its variance and thestrength oftheir beliefs in the performance judgment.Although Bettman did not carry his analysis beyondthe formation of attribute satisfactions, he was the firstto describe a process by which attribute performance istranslated into attribute satisfaction and to distinguishbetween these concepts. This article extends Bettman'swork by showing how attribute satisfaction may affectstill higher-order affective judgments in product or ser-vice evaluation.

Crossover Itifluences. The present framework alsoallows for the likelihood that "crossover" effects operatebetween the attribute states and the affect states. In thiscontext, crossover effects would be represented by in-verse infiuences between attribute satisfaction (dissat-isfaction) and negative (positive) affect. A crossover ef-fect from attribute dissatisfaction to positive affect islikely if negative product experience at the attributelevel has the effect of decreasing positive affect in ad-dition /(increasing negative affect.' Work in psychology(e.g., Taylor 1991) suggests that negative events detractfrom one's ability to experience positive affect. Simi-

'This proposition may at first appear to contradict the affect-balanceassumption, as it suggests that the affects may have common butinverse causes. However, it is noted that (a) crossover effects assumethe prior existence ol' simultaneous positive and negative affect, {b)the proposed effects are at the level of a modification to the affects(e.g.. a raising or lowering) and not at the level ofdetermining theirpresence or absence, and (r) the existence of these effects remainssubject to empirical verification.

larly. positive events, most notably successes, seem tomitigate the occurrence of negative affect, although thesymmetry of this effect with that of negative events onpositive affect is not presently known. Nonetheless, bothcrossover effects from attribute experience to affect arespecifically hypothesized in the model.

This perspective augments that of Westbrook (1987;Westbrook and Oliver 1991) by explicitly linking pos-itive and negative attribute experienee to affective states.Whereas these works did not entertain sources of affec-tive responses, positive and negative aflect are explicitlymodeled as emerging from consumer reactions to prod-uct feature performance. Moreover, the possibility thatperformance affects satisfaction directly as shown byChurchill and Surprenant (1982). Oliver and DeSarbo(1988), and Tse and Wilton (1988) is tested simulta-neously in the model. Thus, the current study modelssatisfaction as a function of cognition, affect, and directexperience, a three-determinant perspective that has notappeared in the literature to date. In addition, the studysuggests the mechanism by which attribute experiencecan operate through affect to influence satisfaction in-directly. Taken together, these direct and indirect at-tribute influences comprise the "attribute basis" of sat-isfaction.

Specification of Affect DimensionsAttributional Stmrces of Negative Affect. As noted,

Westbrook (1987) examined the contribution of positiveand negative affect on satisfaction and showed that bothwere significantly related to satisfaction when testedagainst other satisfaetion antecedents. The present studyextends this work by expanding the components of bothpositive and negative affect. Specifically, negative affectis expanded to include different attributional sourcesof causality for dissatisfaction. This draws on the locus-of-causality literature in attribution theory, where thelocus dimension is "concerned with the source of cau-sality; that is. either the cause resides in you, in someother people, or in the situation" (Wong and Weiner1981. p. 655). Generally, the mechanisms by whichconsumers may assign causality to any of these dimen-sions are straightforward. In essence, marketer-causeddissatisfaction is thought to be externally attributed,consumer mistakes are apparently internally attributed,while unfortunate environmental effects are expectedto be situationally attributed (e.g., Bitner 1990; Folkes1990). Studies of affeet resulting from subject attribu-tions frequently find these distinctions. For example.Smith and Ellsworth (1985) found that emotions couldbe described as falling on a self-other responsibilitycontinuum and a situational-human-eontrol contin-uum. Anger, disgust, and contempt were clearly otheroriented; shame and guilt were self oriented; and fearand sadness were situationally oriented.

While Westbrook's (1987) work distinguished be-tween the three different responses among the negative


affects, only the externally attributed affects (anger, dis-gust, and contempt) were used under the assumptionthat marketer-caused problems are primary reasons fordissatisfaction. The present model is extended to boththe internally and situationally attributed negative af-fects because it is suggested that dissatisfaction can arisefrom alternative sources of product frustration. For ex-ample, if a buyer understands that s/he is solely re-sponsible for the purchase of an unsuitable or badproduct, it is proposed that dissatisfaction would be ag-gravated by the consumer's guilt. Alternatively, if theproduct is susceptible to situationally caused failings,dissatisfaction could be worsened by sadness. Thus, allofthe negative affects may be necessary for a fuller ac-counting of consumer responses.

Study ofa greater variety of negative affect is poten-tially important because of an innate positive hedoniebias on the part ofthe consumer. Generally, individualsare thought to attempt to maximize positive affectivestates and minimize negative affective states. Negativeexperiences are thought to be more salient, are perceivedwith greater intensity, and are expressed with greatervariety (Derbaix and Pham 1991). This perhaps explainswhy emotional typologies typically have a greater num-ber of negative affects. The Izard (1977) DES scale, forexample, has seven negative and two positive affects,reflecting Izard's view of "fundamental" emotions.

Interest and Joy as Separate Positive Ajjects. Thepositive affect dimension tested by Westbrook (1987)is decomposed to separate interest from joy, the twopositive affects in the Izard (1977) scheme. Recent tax-onomic research in the affect literature now distin-guishes the dimension of pleasure (including joy) fromthat of arousal, attention, or interest (Daly, Lancee, andPolivy 1983: Watson and Tellegen 1985). Moreover,discussions of the content of positive affect either donot include interest (de Rivera etal. 1989) or specificallyseparate states of arousal and joy (Storm and Storm1987). Apparently interest is a low-intensity but directedstate of arousal or activation while joy is a somewhatintense form of pleasure. Although Izard's typology in-cludes interest in the categories of discrete emotions,others do not (e.g., Plutchik 1980). This argues fortreating interest and joy in the Izard scheme as separateaflects in the same way that the categories of attribution-specific negative affeet are analyzed separately,^

The Model to Be TestedThe model to be tested here is the subsystem shown

in boldface in Figure 1 with the attribute extension aug-

^Attribution-based categorizations of positive affect have not beenforthcoming. Weiner et al. (1979) do describe different discrete pos-itive affects resulting from differeni attributions. However, these dis-tinctions are not applicable to the two positive affects suggested byIzard (1977) and used here. As the list of consumption-related positiveaffect IS studied further, attribution categorizations of positive affectwill likely appear.

FIGURE 2

EXPANDED ATTRIBUTE-BASED SATISFACTION MODELWITH PROPOSED CAUSAL PATHS

OiKonllrTnUlon

menting the affect sequence. The subsystem includes(overall) disconfirmation to provide a partial replicationof Westbrook and Oliver (1991). Note that the effectsof expectation and equity are not tested. Prior expec-tation measures were not available and equity consid-erations were not thought to operate in the researchcontexts investigated here.^ In addition, and as dis-cussed, attribution influences are manifested within thecategories of (negative) affect.

Figure 2 shows the expanded model, as tested, withthe following modifications to Figure 1. First, attributeinfluences are expressed as attribute satisfaction units,which are posited to arise from summary episodes ofnet attribute satisfaction and net attribute dissatisfac-tion. Seeond, attribute satisfaction and dissatisfactionare shown as antecedent to both positive and negativeaffect. Third, positive affect is now represented by thetwo affect states of interest and joy while negative aflectis portrayed by the three attribution-dependent statesof external, internal, and situational affect. Finally,overall disconfirmation is shown as direetly influencingsatisfaetion to represent a cognitive source of CS/D re-sponse.

HypothesesThe preeeding discussion suggests the study hy-

potheses as follows:

HI: Affective response in usage processes can bedescribed by a positive affective dimension

omission of expectations may constitute a limitation if biasis introduced in the estimates ofthe remaining variables. However,other research suggests that expectations may not operate for certaindurables (Churchill and Surprenant 1982) and continuing services(Bolton and Drew 1991).


eonsisting of interest and joy and. separately,by three negative dimensions defined on thebasis ofthe causal agent.

H2: Attribute satisfaction experiences will influ-ence the degree of positive affect and will haveinverse effects on negative affeet. Conversely,attribute dissatisfaction experiences will in-fluence negative affeet and also inversely affectpositive affect.

113: Attribute satisfaction and dissatisfaction willhave direct effects on overall product or ser-vice satisfaction judgments.

H4; Positive and negative affect will directly influ-ence overall product or service satisfactionjudgments.

H5: Disconfirmation will be significantly relatedto satisfaction.

METHOD

Sample and ProcedureTwo field studies were performed to sample affective

response in ongoing consumption settings to insure thatconsumers would experience actual usage situations(see. e.g.. Scherer and Tannenbaum 1986). Also, toprovide a "bridge" from Westbrook's (1987) study, oneproduct category was used that was common to hisstudy (autos) and one was used that was not. WhereasWestbrook used cable television for his second product,student evaluations of a required course in principlesof marketing are used here. Automobiles were selectedbecause the high level of variation in involvement inthis product category (Bloch 1981) aflbrds similar vari-ation in affect. The course evaluation was chosen be-cause of similar variation in involvement and becausemuch ofthe work on affect, particularly with regard toattribution-generated affect, has been done in academiccontexts (e.g.. Smith and Fllsworth 1987; Weiner et al.1979).

As an additional consideration, inclusion of a servicecategory is important in light of emerging research onsatisfaction in the delivery of service quality. A numberof studies specifically address the role of satisfaction inservice contexts. These include Botton and Drew (1991),Cadotte et al. (1987), Oliver (1980). Oliver and DeSarbo(1988). Oliver and Swan (1989), and Swan and Trawick(1981). In these studies, those parts ofthe expectancydisconfirmation model that were included were gen-erally supported, as shown in Figure 1. Thus, thisframework provides an encouraging perspective for ap-plication to service contexts.

Separate samples were surveyed for the two usagesituations. As in Westbrook (1987). the auto samplewas drawn according to an area-based judgmental sam-pling plan. This resulted in 125 usable subject surveys.

The student sample was drawn from nine sections of arequired marketing principles class. Section instructorswere asked to solicit 20 volunteers from each sectionto provide a representative cross section ofthe entireclass. Responding anonymously, the students provided178 usable surveys.**

MeasuresAffect. In accord with Westbrook (1987). Izard's

(1977) DES was chosen for the affect scale because ex-tensive developmental work has produced validity andreliability data on this instrument (e.g., Boyle 1984).Moreover, Westbrook has shown this seale to be validin consumption settings and useful in explaining therole of affect in consumption; its use here will providefor comparisons across the Westbrook and currentstudies.'

The DES contains 10 subscales representing the in-tensity with which subjects experienee the 10 funda-mental emotions in Izard's (1977) theory. These includethe positive aflects of interest and joy. the negative af-fects of anger, disgust, contempt, shame, guilt, sadness,and fear, and the additional affect of surprise. Note thatsurprise is not used in the present framework. Surpriseis a bivalent affect, having positive (e.g., delight), neg-ative (shock), and neutral (amazement) interpretations(Charlesv^-orth 1969). Because the surprise scale in theIzard scheme does not reveal the subject's intended va-lence, it was not used here as a measure of affect. SeeWestbrook (1987) for elaboration of the content of theIzard scales.

Descriptive statistics and reliability estimates for thenine DFS subscales used for the auto and course samplesare reported in Table 1. Generally, the respondents ex-perienced greater levels ofthe positive affects. The samepattern was also observed in Westbrook (1987). wherethe auto results are very similar to those found here.

Identical sets ofthe summary measures were used inboth studies (with the exception ofthe satisfaction scaleisee below) as follows:

Affects. The negative affect subscales were formedon the basis of the attribution agency proposed in

*Note that, although students are able to select specific sections ofa marketing principles course, they are not able to make a choiceabout whether to take it or not. Thus, while differences in the causalrelationships in the suggested framework over the two sample domainsare not predicted, certain indicators may not emerge as anticipated.One likely example is that of internally attributed affect. Because the"consumption experience" was mandatory for the course sample.little self-blame would be expected for this group.

'It should be noted that other emotion typologies are available toconsumer researchers. In fact, Havlena and Holbrook (1986) haveperformed various tests on paradigms suggested by Mehrabian andRussell (1974) and Plutchik (1980). However, it appears that neitherof these have been used in the CS/D literature, which motivates theuse of Izard's (1977) typology. It is hoped that validation work forthe Plutchik and Mehrabian-Russell frameworks in a CS/D contextwill be forthcoming.


TABLE 1

DESCRIPTIVE STATISTICS FOR STUDY MEASURES

Measure

Affects:InterestJoyAngerDisgustContemptShameGuiltSadnessFear

Model variables:Attribute satisfactionAttribute dissatisfactionPositive affect (see text)Negative affect:

ExternalInternalSituation

DisconfirmationSatisfaction

Length

55555555S

66

55575

X

3.313.411.581.551.491,381.491.791.37

3.911.47

1,541,441,584.873,96

Autos (n = 125)

a

.72

.991.031,01

.89

.81

.83

.99

.75

1.15.61

,94.76.81

1.23.90

«

,82,70.94.92.8887.86.90.89

. . . '

. . , '

.96

.85

.82

.89

.98

X

3.072.661.581.661,561.731.342,091,45

3,661.55

1.571,531.773.893.49

Course (n = 178)

a

.87

.96

.89

.85

.86

.86

.65

.94

.82

1.18.74

.71,63.76,95.91

a

.77

.84

.86

.80

.85

.81

.80

.81,88

. . . *

,88.60,61.65.92

'Not intended to be an internally consistent scale.

Smith and Ellsworth (1985). Externally attributednegative affect was composed ofa summated indexof the anger, disgust, and contempt DES subscales.Internally attributed negative affect was based on theshame and guilt DES subscales, while situationailyattributed affect was based on the fear and sadnessDES subscales. The positive affect subscales wererepresented by the interest and enjoyment scales ofthe DES.

Attribute Satisfaction. For the auto data, 19 attri-butes were selected as representative of automobile fea-tures. These were determined from informal focus groupdiscussions, salient features typically reviewed by con-sumer product magazines (e.g.. Consumer Reports), andfeatures prominently displayed in automobile advertis-ing. Subjeets were asked to state their level of satisfac-tion and dissatisfaction with each feature on separatesatisfaction and dissatisfaction six-point scales rangingfrom "not at all" to "a great deal." Respondents couldindicate levels of both satisfaction and dissatisfactionfor the same attribute, as would occur for '"mixed ex-perience," although many did not. For the course study,the literature on course evaluation (e.g.. Marsh 1984)was studied for generally agreed upon dimensions ofeducational experience. Selected items in each of themajor categories (e.g., instructor, work load, evaluation)were culled from this literature and used to constructan i8-item list. The attribute (dis)satisfaction indices

were created by summing over the 19 (auto) or 18(course) attributes.''

Disconftrmation. Disconfirmation was measured ona disconfirmation scale reported in Oliver (1980). Itcontains three "better-worse than expected" itemssampling benefits, problems, and overall performance.

Sati.sfaction. A 12-item Likert-type scale expandedfrom one originally reported in Oliver (1980) and testedin Oliver and Westbrook (1982) was used in its entiretyfor the automobile sample. Reliabilities in the 0.9 rangeare reported in these tests as well as in other studieswhere this scale or abbreviations of it have been used(Bearden and Teel 1983; Oliver and Bearden 1985;Oliver and Swan 1989). This scale differs from the at-tribute satisfaction scales in that it measures domainsof satisfaction (e.g., right decision, enjoyment, good ex-perience) and not product features per se.

Modification of this scale was required for the coursesample. Because the class was a required marketingprinciples course and students did not have the optionof selecting it (see n. 4), items in the global satisfactionscale regarding course selection (e.g., "My decision to

••Dimensional (e.g.. factor) analysis was not used because ofthenature of attribute experience, which is cumulative over discreteevents. Two highly correlated attributes, which would load on a singledimension, could have singular, additive impacts on (dislsatisfaction.This additivity would be obscured if dimensions were used.


take this course was a wise one") were eliminated, whichresulted in a six-item scale for the course sample.

Alpha reliability coefficients for all scales are shownin Table 1. For Izard's (1977) affects, all values equaledor exceeded 0.7, a level considered acceptable for basicresearch. Generally the results were comparable acrosssamples. Results from the other study variables showedthat all coefficients were in excess of 0.8 in the autosample. In the course evaluation study, coefficients inthe 0.6 range appeared for internally and situationallyattributed negative affect and for disconfirmation.'While these values are lower than one might prefer,they were thought to be sufficient for model comparisonpurposes. Finally, no reliabilities are reported for thepositive affect eonstruct as interest and joy are usedhere as separate indicators and the reliabilities for thesescales appear in the upper part ofthe table.

AnalysisThe system of equations represented by the model

in Figure 2 is as follows:

satisfaction = ./(positive affect, negative

affect, attribute satisfaction,

attribute dissatisfaction,

disconfirmation),

positive affect ^ /(attribute satisfaction,

attribute dissatisfaction),

negative affect = /(^attribute dissatisfaction,

attribute satisfaction),

where the following items are used to represent each ofthe constructs above: satisfaction: the 12 (auto) or six(course) satisfaction scale items; positive affeet: the in-terest and joy scales; negative affect: the negative inter-nal, external, and situational scales; attribute satisfac-tion: the net attribute satisfaction index; attributedissatisfaction: the net attribute dissatisfaction index;disconfirmation: the three disconfirmation scale items.

The model in Figure 2 is recursive and includes se-quential relations where some variables (e.g., the affects)are both dependent and independent. Thus, ordinaryleast squares analysis will not give consistent or un-biased estimates because ofthe fact that the nondepen-dent endogenous variables in an estimated equation willbe correlated with the disturbance term. In this situa-tion, a procedure such as two-stage least squares (TSES)is required. Two-stage least squares analysis is often

' A likely explanation for these low attribution dimension reliabil-ities in the course sample is the possibility that little self-blame andfear or sadness would be expected for this group because ofthe lackof choice addressed in n. 4. The item contributing to the low discon-firmation reliability is that of problem disconfirmation. No expla-nation for this is immediately apparent.

preferred over other instrumental variable techniquesfor its simplicity and applicability to a variety of models(Johnston 1984), Bolton and Drew (1991) have previ-ously used TSLS in a satisfaction investigation wherethey modeled satisfaction as a function of performanceand quality as a function of satisfaction.

The specific TSES estimation procedure used is fromJoreskog and Sorbom (1987). who employ a measure-ment model for latent variables with multiple indicators(i.e., positive affeet, negative affect, satisfaction, anddisconfirmation as modeled here). The procedure usesthe TSES instrumental variable technique for both themeasurement and structural models and producesstandard error estimates for the model parameters.Model estimates were calculated for each data set sep-arately.

RESULTS

Measurement Model and the Structure ofAffect

The top part of Table 2 shows the measurementmodel for the two latent affect constructs. (Significant/-statistics, all /? < .01. were obtained for all indicatorsfor disconfirmation and satisfaction over both samplesand are not shown here for brevity.) The summary af-fects were represented by interest and joy for positiveaffect, and external, internal, and situational affect fornegative affect. The significance of the standardizedloadings for these five affect categories is used to testHypothesis 1, which proposes the affect structure ofthemodified Izard (1977) scheme tested here.

The findings in Table 2 show that the negative affectdimensions of external, internal, and situational affectare supported. All nonfixed loadings over both sampleswere highly significant (the fixed parameter having thehighest loading). This same pattern was also true forpositive affect in the course sample, where both interestand joy attained high loadings. In the auto sample,however, joy did not load significantly on the positiveaffect dimension, which was represented primarily byinterest. Thus. Hypothesis 1 was not supported for pos-itive affect in the auto sample, which suggests that in-terest and joy may be separate constructs for this prod-uct category.

To test for the dependence or independence of pos-itive and negative affect, the estimated correlation be-tween the positive and negative affect latent constructswas determined. For the course sample, this value was0.092 (not significant), which supports the indepen-dence argument. However, for the auto sample, thiscorrelation was 0.264 {p < .01), suggesting a positivecorrelation. Analysis showed that this finding was dueto the positive correlation between interest and negativeaffect {r = 0.288. p < .01) in the auto sample. The cor-relation between joy and negative affect was —0.328 {p< .01). This latter set of findings in the auto sample


TABLE 2

PATH ESTIMATES FOR THE PROPOSED SATISFACTION FRAMEWORK

Causal paths

Autos' Course*"

Standardized loading f-Value Standardized loading f-Value

Positive affect indicators:InterestJoy

Negative affect indicators:ExtematIntemalSituationai

Positive affect relationships:Attribute satisfactionAttribute dissatisfaction

Negative affect relationships:Attribute satisfactionAttribute dissatisfaction

Satisfaction relationships:Attribute satisfactionAttribute dissatisfactionPositive affectNegative affectDisconfirmation

.753

.178

.976

.878

.951

.510

.150

.257

.288

,135.175.142.110,634

1,47

17.2123.69

5.041.48

-2.622.93

2.24-3.87

3.07-2.2912.86

.771

.882

.894

.613

.843

.603

.031

.065

.235

.130

.059

.385

.219

.426

10.65

e.0512.46

9.59-.049

-.0843.06

2.98-1.85

7.73-5.33

8,48

•Forpositiveaffectrelatlonshjps,R* = 0.190. F= 14.31,df = 2,122. For negative affect relationships. fl* = 0,238.F= t9.05.d/ = 2,122. For satisfaction relationships.f = 0.851, F - 135.9, d / - 5,119,

Tor positive affect relationships, fl^ = 0.376. F = 53.63. d/= 2,175, Fof negative affect relationships, fl'= 0.069, F = 6.60, df = 2,175, For satisfaction relationships.* = 0.815, F - 151.6, d/ = 5.172.•"Fixed parameter.

adds further to the argument that interest and joy maybe separate and distinct dimensions for the durablegoods category investigated here.

Model RelationshipsTable 2 also tests for the effect of attribute satisfac-

tion/dissatisfaction influences on the affects (Hypothesis2) over both samples separately. Generally, positive af-fect was strongly related to attribute satisfaction, as hy-pothesized. Contrary to Hypothesis 2, however, attri-bute dissatisfaction did not reduce the expression ofpositive affeet. Both coefficients were nonsignificant. Inlike manner, negative affect was related to attribute dis-satisfaction in the expected direction in both samples.In addition, negative affect was inversely related to at-tribute satisfaction in the auto sample, a hypothesizedcrossover effect.

The satisfaction relationships are also shown in Table2, where the results for both samples are found to besimilar. Satisfaction was a positive function of attributesatisfaction and a negative function of attribute dissat-isfaction, although only the auto sample generated asignificant coefficient in this latter case. Thus, Hypoth-esis 3 was fully supported for the auto sample and par-tially supported for the course sample. The affects, inturn, contributed further to the satisfaction equation,with positive affect and negative affect displaying sig-nificant coefficients ofthe correct sign in both samples.

Thus, Hypothesis 4 was supported. Finally, the discon-firmation coefficient is highly significant in both sam-ples, indicating that cognition remains as a necessarycomponent of satisfaction. Thus, Hypothesis 5 is alsosupported.

Generally, the coefficients of determination for thesatisfaction relationships were large and highly signifi-cant ip < .001). Those for the affect relationships werevery adequate (;7 < .01), the exception being negativeaffect in the course sample, which, nonetheless, was sig-nificant at the .01 level.

For comparative purposes,, various subsets ofthe fullmodel in Figure 2 were tested. These included satis-faction variance estimates as eaused by (a) attributesatisfaction/dissatisfaction only, (b) affect only, (c) dis-confirmation only, id) affect and attribute (dis)-satisfaction, and (e) affect and disconfirmation; model(/) is the full mode! variance estimate. These results areshown in Table 3; all are significant atp < .01.

Note that attribute satisfaction correlates more highlywith satisfaction in the auto sample while affect is morehighly correlated in the course sample and that discon-firmation is the best predictor in the auto sample whileaffect remains a better predictor in the course sample.Moreover, the combination of affect and disconfirma-tion is more predictive of satisfaction than is the com-bination of affect and attribute satisfaction/dissatisfac-tion. The full model integrates the variance explainedover all subsets in both samples.


TABLE 3

VARIANCE ESTIMATES FOR SATISFACTION MODEL SUBSETS

Auto Course

a] Attribute (dls)satisfactionb) Affectc) Disconfirmationd) Affect and attribute (dis)satisfactione) Affect and disconfirmation0 Full model

.609

.418

.750

.651

.816

.851

F

95.043.8

369.056.2

178.9135.9

df

2.1222,1221,1234,1203.1215.119

R"

.449

.705

.675

.738

.800

.815

F

71.3209.1365.5121.8232.0151.6

df

2.1752.1751.1764,1733.1745.172

Because interest and joy did not appear to be equallyrepresentative of positive affect in the auto sample, thepossibility that these discrete emotions contribute tosatisfaction separately was entertained in a separateanalysis ofthe auto sample, shown in Table 4. Thesefindings show that attribute satisfaction has a greatereffect on joy than on interest, although both relationsare significant. Attribute dissatisfaction was unrelatedto either emotion, as was true in the combined anaiysisin Table 2. Whereas the coefficient of determination inthe joy equation was highly significant, the R~ value forthe interest equation attained significance at only the.05 level. The negative affect relations were not affectedby the separation of interest and joy in the analysis.

The prediction of satisfaction for this new analysisshowed some divergence from that in Table 2. Attributesatisfaction and dissatisfaction had direct effects in thehypothesized direction as before. However, joy ap-peared as the only affect influence on satisfaction. Nei-ther the interest nor negative affect influences were sig-nificant. Disconfirmation emerged as a majorsatisfaction influence as before. Apparently, interestdoes not appear as a component of̂ positive affect forautomobile satisfaction, which explains the divergenceof positive affect loadings in Table 2.

DISCUSSIONThe discussion addresses four major issues. First, the

structure of affect in the consumption process is dis-cussed as it relates to the framework tested here. Seeond,the role of attribute satisfaction/dissatisfaction is intro-duced as possibly instrumental in generating these feel-ings. Third, the degree to which attribute satisfactionand affective states complement one another and addto disconfirmation in explaining satisfaction is dis-cussed. Fourth, the theoretical implications of this ex-panded theory are explored with reference to existingviews of satisfaction.

Structure of AffectThis study suggests that the structure of affect in con-

sumption may have two dimensions relating to a pos-

itive affect factor and three negative factors based onthe locus-of-causality attributional dimension (Smithand Ellsworth 1985;Weiner 1986). The first of the neg-ative causal ascriptions appears to be externally targetedat the provider of the product or service, the secondapparently is directed inward at the user him/herself,while the third may be situational in nature.

The positive dimension findings did not appear con-sistently across the two product and service categoriesinvestigated here. For the course instruction sample,both interest and joy were representative of positiveaffect with similar loadings. However, the automobilesample findings suggested that these two affects are pro-cessed separately and cannot be assumed to be corre-lated. When analyzed separately, only joy was found tobe related to satisfaction. Apparently interest is a stateof arousal or activation that, in the case of automobiles,may pertain to positive or negative events, while joy isa somewhat intense form of pleasure (Daly et al. 1983;Storm and Storm 1987; Watson and Tellegen 1985).Further research is needed to determine when, and forwhat product and service categories, interest and joyare correlated and when they are not.

Role of Attribute ExperienceThe possibility that satisfaction with the functional

elements of the product could have some relation tothe affects experienced by the consumer was investi-gated. This phenomenon was referred to as the attributebasis of satisfaction and provides an analogue to mostcommercial surveys where attribute ratings of one sortor another are correlated with overall satisfaction judg-ments. The role that the affects play in this process wasalso studied, and a likely mechanism (i.e.. whether thisis a mediating role in a true causal sequence) was alsotested.

The results from the two samples suggest quitestrongly that (a) attribute satisfaction affects overallsatisfaction directly, (b) attribute satisfaction influencespositive affect, (c) attribute dissatisfaction affects neg-ative affect, and id) positive and negative affect are pos-itive and negative influences, respectively, on satisfac-tion. In addition, for the auto sample only, attribute


TABLE 4

PATH ESTIMATES FOR THE PROPOSED SATISFACTIONFRAMEWORK WITH SEPARATE INTEREST AND JOY: AUTO

DATA

Causal paths

Interest relationships:'Attribute satisfactionAttribute dissatisfaction

Joy relationships:''Attribute satisfactionAttribute dissatisfaction

Negative affect relationships:'^Attribute satisfactionAttribute dissatisfaction

Satisfaction relationships:*^Attribute satisfactionAttribute dissatisfactionInterestJoyNegative affectDisconfirmation

Standardized loading

.278

.119

.540-

.002

-.257.288

.148-.194-,001

.247-.035

,528

f-Value

5.041.09

5.70.02

-2.622,93

3.00-4.76-.026.27-.86

11.34

= 0.051, F = 3,28, cff = 2,122-= 0,290, F = 24,92, dt = 2.122.= 0.238. F = 19.05, df - 2.122.= 0.879, F - 142.9, df = 6,118.

satisfaction decreases negative affect while attributedissatisfaction decreases satisfaction.

Thus, ofthe five hypothesized predictors of satisfac-tion in Figure 2, all were supported for the automobilesample with interest and joy as joint indicators (al-though theinfluencesof negative affect and interest werenot supported with joy and interest separated). In thecourse sample, all hypothesized predictors except at-tribute dissatisfaction were supported. In addition, pos-itive affect was a function of attribute satisfaction andnegative affect was a function of attribute dissatisfactionin both samples. However, only one ofthe two hypoth-esized crossover effects was observed in the auto samplewhere attribute satisfaction was found to decrease neg-ative affect. Finally, attribute dissatisfaction had no ef-fect on positive affect in either sample.

Relation to Disconfirmation InfluencesAs suggested in Oliver (1980), the expectancy dis-

confirmation paradigm is primarily cognitive in naturebecause the comparison process in disconfirmationjudgments requires the deliberate processing of infor-mation. In contrast, the affective proeess is thought tobe, at least in part, not under conscious control. Thus,the cognitive and affective responses in postpurchasejudgments may be seen as distinct components in re-sponse to environmental events, and each would appearto introduce its own influence on the consumption pro-cess.

The relative effects of affect and cognition pose an-other issue. It appears from the findings that disconfir-

mation is the more influential of the variables testedhere. It had the greatest effect, as measured by the struc-tural coefficient, in both the auto and course satisfactionequations. Whether this suggests that disconfirmationis the most important antecedent in a group of inde-pendent effects or is an artifact of the present data isunknown. It is noted, however, that the dominatingeffect of disconfirmation over affective response was alsofound in Westbrook (1987), and the present study nowcorroborates this result.

Theoretical ImplicationsThe theoretical implications exist primarily in the

domain of discovery ofthe underlying structure ofthesatisfaction response. As noted, an attribute-only sat-isfaction analysis provides little insight as to the mech-anism by which consumers process attribute outcomesinto satisfaction. The expectancy disconfirmation par-adigm filled some of this void, but did not seem capableof explaining more than 35 percent ofthe variance insatisfaction (e.g., Oliver 1980). By merging affect, dis-confirmation, and attribute-based satisfaetion judg-ments, explained variance values of 85 percent (autos)and 81 percent (course) were obtained here. And, with-out the benefit of attribute judgments, Westbrook(1987) reported R^ values of 59 percent (autos) and 37percent (cable television). Thus, the three sets of vari-ables studied here appear to capture more variance insatisfaction than has been obtained elsewhere.

Awaiting further development is a better theoreticalspecification ofthe satisfaction and postpurehase pro-cess. Insights available from this study revolve aroundthe separate roles played by the positive and negativeaffects and their relation to satisfying and dissatisfyingattribute experiences. The data show that positive affectis mostly influenced by attribute satisfaction and notdissatisfaction. In contrast, negative affect appears torespond to both positive and negative attribute expe-riences in the auto sample. This implies that, for durablegoods products, negative affect may respond to attributedissatisfaction as one would expect and may be reducedby generally positive attribute experiences. Thus, attri-bute satisfaction may have dual beneficial effects, whileattribute dissatisfaction may work primarily to causeincreased levels of negative affect.

Another theoretical implication concerns the directeffects of attribute satisfaction/dissatisfaction on overallsatisfaction. This is one mechanism contained in theattribute basis of satisfaetion whereby satisfaction is seenpartly as a function of direct attribute experiences.However, the data show clear mediating linkagesthrough the affects, which then affect satisfaction. Thisimplies that attribute experiences may have both directand indirect effects on product satisfaction.

LimitationsAs with cross-sectional studies generally, the findings

here may have been affected by common method vari-


ance. This is a particularly troublesome issue in satis-faction research because the consumption experienceitself may generate a primarily positive or negative feel-ing state that will obviously correlate with other re-sponses in consumption. This same difficulty can alsopose problems in obtaining accurate estimates of themodel coefficients.

A second concern is that the affect process in thesatisfaction response varies over product and servicecategories and populations. To some extent, this issuewas addressed with the use of two consumption settingsand two samples of respondents. However, the datashow that, despite differences in the magnitude of somerelations, the structural equation results were fairlyconsistent across both sets of data tested here. Differ-ences were most apparent in the auto sample, whereinterest and joy appeared to be separable affects andwhere more ofthe hypothesized relations were signifi-cant. Nevertheless, further testing on more variedproduct categories is needed.

SummaryThis study both extends and departs from prior work

on the role of affect in the consumption experience ina number of ways. It confirms the existence of positiveand negative affect in consumption, corroborates thedegree to which affect appears to be a response separatefrom disconfirmation. and supports the indirect mech-anism ofthe attribute basis of satisfaction, which sug-gests a mechanism by which affective response mediatesthe effects of the attributes on satisfaction. Thus, theattribute influences, the influence of positive and neg-ative affect, and the disconfirmation coefficients pointto complexity in the satisfaction formation process notyet fully understood; additional work in the area isneeded to address this complexity.

[Received March 1992. Revised March 1993.]

REFERENCESBatra. Rajeev and Morris B. Holbrook (1990). "Developing

a Typology of AfTective Responses to Advertising," Psy-chology and Marketing, 1 (Spring). 1 f-25.

Bearden, William O. and Jesse E. Teel (1983). "Selected De-terminants of Consumer Satisfaction and Complaint Re-ports." Journal of Marketing Research, 20 (February).21-28.

Bettman. James R. (1974), "A Threshold Model of AttributeSatisfaction Decisions," Journal of Consumer Research,1 (September). 30-35.

Bitner. Mary Jo (1990). "Evaluating Service Encounters: TheEffects of Physical Surroundings and Employee Re-sponses." Journal of Marketing. 54 (April). 69-82.

Bloch. Peter H. (1981), "An Exploration into the Scaling ofConsumers' Involvement with a Product Class," in Ad-vances in Consumer Research. Vol. 8, ed. Kent B. Mon-roe. Ann Arbor, MI: Association for Consumer Researcfi,61-65.

Bolton, Ruth N. and James H. Drew (1991). "A MultistageModel of Customers' Assessments of Service Quality andVa]ue,^' Journal of Consumer Research, 17 (March), 375-384.

Boyle, Gregory J. (1984), "Reliability and Validity of Izard'sDifferential Emotions Scale." Personality and IndividtmlDifferences. 5 (6). 747-750.

Bradburn, Norman M. (1969), The Str tat ure of PsychologicalWell-Being. Chicago: Aldine.

Cadotte. Ernest R., Robert B. Woodruff, and Roger L. Jenkins(1987), "Expectations and Norms in Models ofConsumerSatisfaction." Journal of Marketing Research, 24 (Au-gust). 305-314.

Charleswortfi, William R. (1969), "The Role of Surprise inCognitive Development," in Studies in Cognitive Devel-opment, ed. David Elkind and John H. Flavell, New York:Oxford University Press, 257-314.

Churefiilf, Gilbert A., Jr. and Carol Surprenant (1982), "AnInvestigation into the Determinants of Customer Satis-faction." Journat of Marketing Re.search, 19 (November),491-504.

Daly. Eleanor M,, William J. Laneee, and Janet Polivy (1983),"A Conical Model forifie Taxonomy of Emotional Ex-perience," Journat of Personality and Social Psychology,45 (August). 443-457.

Derbaix. Christian, and Michel T. Pfiam (1991), "AffectiveReactions to Consumption Situations: A Pilot Investi-gation," Journal of Economic Psychology. 12 (June), 325-355.

de Rivera, Joseph, Lois Posseil, Julie A. Verette, and BernardWeiner (1989), "Distinguishing Elation, Gladness, andJoy," Journal of Personality and Social Psychology, 57(December). 1015-1023.

Diener, Ed and Robert A. Emmons (1984). "The Indepen-dence of Positive and Negative Affect," Journal of Per-sonality and Social Psychology, 47 (November), 1105-1117.

Folkes. Valerie S. (1990). "Conflict in the Marketplace: Ex-plaining Why Produets Fail." in Attribution Theory: .Ap-plications to Achievement. Mental Health, and Interper-sonal Conflict." ed. Sandra Graham and Valerie S. Folkes,Hillsdale, NJ: Eribaum, 143-160.

, Su^an Koletsky, and John L. Graham (1987), "AField Study of Causal Inferences and Consumer Reaction:The View from the Airport," Journal of Consumer Re-search, 13 (March). 534-539,

Havlena, William J. and Morris B. Holbrook (1986). "TheVarieties of Consumption Experience: Comparing TwoTypologies of Emotion in Consumer Behavior." Journalof Consumer Research, f 3 (December), 394-404.

Horley. James and Brian R. Little (1985), "Affective andCognitive Components of Global Subjective Well-BeingMeasures," Social Indicators Research. 17 (August), 189-197.

Izard. Carroll E. (1977), Human Emotions. New York:Plenum.

Johnston, J. (1984), Econometric Methods. New York:McGraw-Hill.

Joreskog. Karl G. and Dag Sorbom (198,7).SIMPLIS. Moores-ville. IN: Scientific Software.

Marsh. Herbert W. (1984). "Students' Evaluations of Uni-versity Teaching: Dimensionality, Reliability, Validity,


Potential Biases, and Utility," Journal of EducationalPsyehology. 76 (5). 707-754.

Mehrabian, Albert and James A. Russell (1974), An Approachto Environmental Psychology. Cambridge, MA: MITPress.

Oliver, Richard L. (1980), "A Cognitive Model ofthe Ante-cedents and Consequences of Satisfaction Decisions,"Journal of Marketing Research. 17 (November), 460-469.

(1981), "Measurement and Evaluation of SatisfactionProcesses in Retail Settings, Journal of Retailing. 57(Fall), 25-48.

(!989), "Processing ofthe Satisfaction Response inConsumption: A Suggested Framework and ResearchPropositions," Journal of Consumer Satisfaction. Dis-satisfaction and Complaining Behavior. 2, i-16.

and William O. Bearden (1985), "DisconfirmationProcesses and Consumer Evaluations in Product Usage,"Journal of Business Research. 13 (June). 235-246.

and Wayne S. DeSarbo (1988), "Response Determi-nants in Satisfaction Judgments," Journal of ConsumerResearch. 14 (March), 495-507.

and John E. Swan (1989), "Equity and Disconfir-mation Perceptions as Influences on Merchant andProduct Satisfaction," Journal of Consumer Research,16 (December), 372-383.

and Robert A. Westbrook (1982), "The Factor Struc-ture of Satisfaction and Related Postpurchase Measures,"in New Findings on Consumer Satisfaction and Com-plaining, ed, Ralph L. Day and H, Keith Hunt, Bloom-ington: Indiana University School of Business, 11-14.

Plutchik, Robert (1980), Emotion: A PsychoevolutionarySynthesis, New York: Harper & Row.

Russell, James A. (1980), "A Circumplex Model of Affect,"Journal of Personality and Social Psychology. 39 (De-cember), II6I-I178.

Scherer, Klaus R. and Percy H, Tannenbaum (1986), "Emo-tional Experiences in Everyday Life: A Survey Ap-proach," Motivation and Emotion, 10 (December), 295-314,

Smith. Craig A. and Phoebe C, Ellsworth (1985), "Patternsof Cognitive Appraisal in Emotion," Journal of Person-ality and Social P.sychology, 48 (April). 813-838.

and Phoebe C. Ellsworth (1987), "Patterns of Ap-praisal and Emotion Related to Taking an Exam," Jour-nal of Personality and Social Psychology. 52 (March).475-488.

Storm, Christine and Tom Storm (1987), "A TaxonomicStudy ofthe Vocabulary of Emotions," Journal of Per-sonality and Social Psychology. 53 (October), 805-816.

Swan, John E. and 1. Frederick Trawick (1981), "Disconfir-mation of Expectations and Satisfaction with a RetailService," Journal of Retailing. 57 (Fall), 49-67.

Taylor, Shelley E. (1991), "Asymmetrical Eflects of Positiveand Negative Events: The Mobilization-MinimizationHypothesis," Psychological Bulletin, I 10 (July). 67-85.

Tse, David K. and Peter C. Wilton (1988), "Models of Con-sumer Satisfaction Formation: An Extension," Journalof Marketing Research. 25 (May), 204-212.

Watson, David and Auke Tellegen (1985), "Toward a Con-sensual Structure of Mood," Psychological Bulletin. 98(September), 219-235.

Weiner, Bernard (1986), An Attributional Theory of Moti-vation and Emotion, New York: Springer.

, Dan Russell, and David Lerman (1979), "The Cog-nition-Emotion Process in Achievement-related Con-texts," Journal of Personality and Social Psychology. 37(July), 1211-1220.

Westbrook, Robert A. (1987). "Product/Consumption-basedAffective Responses and Postpurchase Processes," Jour-nal of Marketing Re.search, 24 (August), 258-270.

and Richard L, Oliver (1991), "The Dimensionalityof Consumption Emotion Patterns and Consumer Sat-isfaction," Journal of Consumer Research. 18 (June), 84-91.

Wong, Paul T. P. and Bernard Weiner (1981), "When PeopleAsk 'Why' Questions, and the Heuristics of AttributionaiSearch," Journal of Personality and Social Psychology.40 (April), 650-663.

Yi, Youjae (1990), "A Critical Review of Consumer Satis-faction," in Review of Marketing 1990. ed. Valarie A.Zeithaml, Chicago: American Marketing Association,68-123.

Cognitive, Affective, And Attribute Bases of the Satisfaction Response

Documents

Transcript of Cognitive, Affective, And Attribute Bases of the Satisfaction Response