ELSEVIER

The Nature of Satisfaction:

An Updated Examination and Analysis
Barry J. Babin
THE UNIVERSITY OF SOUTHERN MISSISSIPPI

Mitch Griffin
BRADLEY UNIVERSITY

The preponderance of recent satisfaction research focuses on things that retention (Jones and Sasser, 1995), while in theoretical re-
cause satisfaction. Measurement articles addressing the precise conceptual- search, satisfaction indicators often provide insufficient dis-
ization of consumer satisfaction are more than a decade old. In that criminant validity (Taylor and Baker, 1994).
time, many advances in analytical approaches used to delineate and Few articles focusing specifically on the construct validity
operationalize latent constructs have been made. The research presented and measurement of satisfaction have appeared in top market-
here uses advances in both satisfaction research and measurement theory ing journals. The early works of Oliver (1980, 1981) and
to provide a more precise view of consumer satisfaction and dissatisfaction Westbrook ( 1980, ] 981; Westbrook and Oliver, 1981) remain
than has been previously offered. While some researchers have used the most commonly cited sources for satisfaction scales. Table
indicators that, arguably, lack face validity due to contamination of other 1 provides an overview of satisfaction operationalizations in
closely related constt-ucts, the conceptualization offered here maximizes major marketingjournals and attests to the merits of this work.
face validity and is more true to the nature of reflective indicators of However, given recent criticisms of satisfaction measurement
latent constructs. Additionally, the possibility of distinct satisfaction and mentioned above and elsewhere (Fornell, 1992), conceptual
dissatisfaction constructs is investigated, j BUSN RES 1998. 41.127-- and analytical advances in theory and measurement (Anderson
136. © 1998 Elsevier Science Inc. and Gerbing, 1992) may suggest improvements in the manner
with which satisfaction is assessed and validated. Specifically,
the use of structural equations modeling and confirmatory
" hat more can be said about consumer satisfaction?

W Surprisingly, marketing theory formally recog-

nized that postpurchase evaluations and reactions
are as important as those occurring before a purchase only
factor analysis has been suggested as overcoming many short-
comings of previous measurement validation approaches (Bag-
ozzi, 1993).
As a further step toward "better theoretical specification of
about two decades ago. Since 1971, hundreds of academic
articles have been written trying to explain satisfaction and the satisfaction and postpurchase process" (Oliver, 1993,
its related nomological network (Perkins, 1993). In practice, p. 428) and the development of better satisfaction measures
the era of relationship marketing has emphasized customer (Westbrook and Oliver, 1991), this research provides a de-
satisfaction as a key measuring stick indicating overall market- tailed analysis of the construct validity of satisfaction/dissatis-
ing performance. Clearly, satisfaction is one of only a few key faction. One goal of the research is to sort through potential
building blocks in marketing philosophy, theory, and practice. measurement errors that can contaminate traditional satisfac-
The importance of postpurchase reactions to marketing tion items and lead to interpretational confounding with other
practice and theory makes clear the need for precise measure- constructs with which it may relate (i.e., performance, disso-
ment. Yet, satisfaction's diagnostic and explanatory power is nance, disconfirmation, etc.). Further, the possibility of dis-
often restricted by problems with measurement (Peterson and tinct but related satisfaction and dissatisfaction constructs is
Wilson, 1992). Although prevalent in practice, the common also examined. While such an examination should be part
4-item check box indicator often fails to predict customer of any assessment of dimensionality, it may be particularly
important with the satisfaction construct due to reports of
consumers' differing reactions to like levels of satisfaction and
Address correspondence to Barry J. Babin, The University of Southern Missis- dissatisfaction (Jones and Sasser, 1995; Peterson and Wilson,
sippi Department of Marketing & Finance, College of Business, Marketing-
Box 5091, Hattiesburg, MI 39406 (e-mail: babinb@cba.usm.edu). 1992). In short, this research examines items commonly used

Journal of Business Research 41, 127-136 (1998)

© 1998 Elsevier Science inc. All rights reserved. ISSN 0148-2963/98/$19.00
655 Avenue of the Americas, New York, NY 10010 PII SO148-2963(97)00001-5
128 J Busn Res B.J. Babin and M. Griffin
1998:41 :I 27-I 36

in assessing satisfaction in an attempt to identify a reliable, 1987, p. 260). These definitions leave the researcher with ques-
valid, and parsimonious scale for measuring satisfaction. tions concerning what satisfaction is and what it is not.
While totally new items could be developed, previous research Given the wide variance in conceptual underpinnings,
has clearly produced some suitable satisfaction indicators. common measurement approaches may provide indicators
This article focuses more on eliminating the use of items that better represent closely related--but potentially dis-
potentially contaminating the construct, and by this contami- tinct--constructs such as perceived performance, disconfir-
nation, potentially distorting empirical results. mation, happiness, or decision regret. For example, phrases
such as "my decision was right" and "I believe I would be
Conceptual Development satisfied with XYZ services" sometimes appear as indicators of
"satisfaction." Although these generally produce highly reliable
The Nature of "Satisfaction" scales (Oliver, 1993), reliability is no guarantee of veracity or
The variance in definitions of consumer satisfaction/dissatis- unidimensionality (Gerbing and Anderson, 1988).
faction suggest it is a rather illusive concept. An often cited In some ways, the job satisfaction literature exhibits similar
definition is provided by Hunt (1977, p. 49): "[c]onsumer problems. For example, consider the item "I am paid enough
satisfaction with a product refers to the favorableness of the for what I do" (see Johnston et al., 1990). In terms of face
individual's subjective evaluation of the various outcomes and validity, this item seemingly reflects more clearly a cognitive
experiences associated with buying it or using it." Alterna- belief about pay, or perhaps even a disconfirmation or compar-
tively, "satisfaction may best be understood as an evaluation ison with expectancies, than it does a direct expression of the
of the surprise inherent in a product acquisition and/or con- feeling of satisfaction. Another scale includes items better
sumption experience" (Oliver, 1981, p. 27). More recently, reflecting specific beliefs about the work itself ("I feel that my
some have suggested a clearer separation between satisfaction job is no more interesting [emphasis added] than others I could
and knowledge, noting that satisfaction is distinct from "cogni- get") (Brayfield and Rothe, 1951). Perhaps in recognition of
tive beliefs about product/consumption outcomes" (Westbrook, these potential shortcomings, some organizational researchers

Table 1. Review of Satisfaction Operationalizations Since 1981

Article t Source (Description) Primary Validation Reported

Anderson, Fornell, and Lehmann (1994) Fornell (1992) Variance extracted

Andreasen (1984) Single 3-point item NA
Andreasen (1985) Binary (yes/no) NA
Bearden and Teel (1983) Oliver (1980) Reliabilities. Loadings given from a structural model.
Bolton and Drew (1991) Binary (yes/no) NA
Churchill and Surprenant (1982) 4 bipolar items assessing "beliefs Reliabilities, item-total correlation, loadings from a
and affect" (see p. 495) structural model.
Crosby and Taylor (1982) Single 5-point item NA
Fornell and Robinson (1983) Secondary data used NA
Fomell (1992) Ten 10-point items Variance extracted
Herbertein, Linz, and Ortiz (1982) Single 5-point item NA
Johnson, Anderson, and Fomell (1995) SCSB (Fomell, 1992) NA
Mano and Oliver (1993) Oliver (1980) Coefficient or; exploratory factor analysis
Oliver and Swan (1989) Westbrook and Oliver (1981) Coefficient or; CFA performed for discrimination (no
details given)
Oliver (1993) Oliver (1980) Coefficient c~; individual factor analyses (not on all
measures simultaneously)
Oliver and DeSarbo (1988) Oliver (1981)
Richins (1983) Inferred
Rust and Zahorik (1993) Oliver (1980) 2 Exploratory factor analysis
Swan and Oliver (1989) Westbrook and Oliver (1981) Reliabilities
Taylor and Baker (1994) Four 7-point items Coefficient
Tse and Wilton (1988) Single 5-point bipolar item NA
Westbrook (1987) Andrews and Withey (1976) Reliability
(Average of circles and a
delighted-terrible scale)
Westbrook and Oliver (1991) Oliver (1980) Coefficient c~3
Woodruff et al. (1983) Not empirical

This table is meant to be illustrative rather than comprehensive.

2 The actual scale was used as a basis for forming attribute specific ratings.
Separate 11-point "satisfied" and "dissatisfied" scales were used but treated as distinct scales rather than as part of the Oliver (1980) Likert scale.
Nature of Satisfaction J Busn Res 129
1998:41:127-136

have shunned "accepted" job satisfaction scales in favor of ad cult to discriminate the definition of satisfaction with this view
hoc measures (e.g., Brown and Peterson, 1994). of emotion. However, compared to other emotions, instances
Thus, attempts to assess "satisfaction" are often contami- of satisfaction can be categorized as positive and fairly passive
nated by items indicating cognitions or knowledge equally or in motivating further action. Its core relational theme is, argua-
better than they do an affective reaction resulting from, but bly, success, and it is motivationally congruent in that it helps
distinct from, assimilation, accommodation, or contrast with a consumer obtain desired states that may contribute to overall
an active cognitive structure comprising a consumer's refer- well-being (Lazarus, 1991; Smith et al., 1993). Satisfaction's
ence point (Stayman, Alden, and Smith, 1992). Alternatively, closest synonyms are contentment and pleasure (Mehrabian
measures may represent thoughts that may cause satisfaction and Russell, 1974). More specific definitions become difficult
more than they indicate satisfaction itself, and thus, they given the "fuzziness" of emotional constructs, thus it has been
would be more accurately modeled as formative rather than argued that emotions' meanings are best conveyed by describ-
reflective indicators. However, a cursory examination of the ing typical instantiations (Fehr and Russell, 1984; Russell,
literature shows an absence of this approach. Therefore, as 1991). For example, consumer satisfaction is the positive feel-
expressed previously (Westbrook, 1987), to measure satisfac- ing resulting from thoughts comparing the expected gas mile-
tion accurately it seems a measurement device more free from age of a Honda Civic (rated 34 mpg) with the gas mileage
"cognitive beliefs" is needed. actually obtained after years of use (e.g., 41 mpg). The emotion
Part of the problem may be that the marketing lexicon literature suggests that if consumer "satisfaction" is the thing
treats consumer satisfaction as an omnibus noun representing to be measured, it should be done so in a manner that is free
most events occurring after a customer makes a purchase. of potential contamination from beliefs and other cognitions
While many postpurchase phenomena may covary signifi- that comprise the appraisal process.
cantly with satisfaction, the question of whether they are iden-
tical to it remains unanswered. Some authors appropriately Satisfaction~Dissatisfaction
discriminate between the satisfaction appraisal (involving cog- The distinctiveness of positive and negative affect is a recurring
nitive beliefs) and satisfaction itself. Operationally, distinct issue in the emotion measurement literature. Although much
disconfirmation items, representing the assimilation/contrast empirical and conceptual research is based on bipolar emotion
process comparing perceived outcomes with salient expecta- constructs (e.g., Russell, 1978; Smith and Ellsworth, 1985;
tions, have been proposed and are used (e.g., Oliver, 1980; Plutchik, 1980; Havlena and Holbrook, 1986), numerous
Tse and Wilton, 1988). Consistent with this view, some au- studies report separate (not very highly correlated) positive
thors are clear in defining satisfaction as these processes' emo- and negative emotions (Babin and Darden, 1996; Darden and
tional result (Woodruff, Cadotte, and Jenkins, 1983; Woodruff Babin, 1994; Bagozzi and Moore, 1994; Edell and Burke,
et al., 1991). Confirmatory factor analyses of common post- 1987; Mano and Oliver, 1993; Oliver, 1993; Lacher and Miz-
purchase evaluations and reactions supports the affective na- erski, 1994). Support for this view (i.e., happy is not the
ture of the satisfaction experience over a cognitive interpreta- opposite of sad) is found in the work of Bradburn (1969),
tion (Babin, Griffin, and Darden, 1994). Izard (1977), and Watson and Tellegen (1985). Under this
Consumer satisfaction, therefore, can be described as an "parallel" view, the presence of a positive emotion does not
emotion resulting from appraisals (including disconfirmation, preclude the presence of a negative emotion. In fact, in some
perceived performance, etc.) of a set of experiences (Locke, cases, positive correlations between linguistic opposites might
1969; Westbrook, 1980; Woodruff et al., 1983). The apprais- be expected. For example, Diener and Emmons (1984) report
als consist of various categorizational processes that trigger positive correlations between positive and negative mood
affective responses (Smith, Haynes, Lazarus, and Pope, 1993). among subjects asked for their affective reaction to an evalua-
For instance, a busy marketing professor may get a Big Mac tion of their day-to-day activities over the previous weeks. As
for lunch that is exactly what he/she expected and experience a result, commonly used emotion measurement scales like the
dissatisfaction because the appraisal results in contrast with PANAS (Positive and Negative Affect Scale) are based on the
his/her schema for "a good lunch." Yet, a high school junior notion that pleasant feelings are not the opposite of unpleasant
may consume an identical Big Mac, again just as expected, feelings (Watson, Clark, and Tellegen, 1988). Bagozzi (1993)
and experience satisfaction because of differing schematic ref- supports this measurement theory in one of the few confirma-
erence points. tory investigations of emotion indicators. Despite considerable
The emotion literature provides additional insight into the empirical evidence, this issue continues to spark debate
definition of consumer "satisfaction." First, cognitive apprais- (Green, Goldman, and Salovey, 1993).
als elicit emotions (Stein and Oatley, 1992; Arnold, 1945), Logically, if pleasant and unpleasant feelings can comprise
which are defined broadly as subjective feeling states that two separate constructs (one negative and one positive), then
result from cognitive appraisals of one's environment (Plut- it seems reasonable to raise the same question for satisfaction/
chik, 1980; Izard, 1977; Lazarus, 1982). These emotions then dissatisfaction. Surprisingly, little research has addressed this
influence reactions to events. Conceptually, it becomes diffi- issue directly (cf., Leavitt, 1977; Maddox, 1981), but a few
 130 J Busn Res B.J. Babin and M. Griffin
1998:41:127-136

results exist that suggest some degree of discrimination. For center adjacent to the university. The three services include
example, a common consumer satisfaction scale (Oliver, 1980) a hair-styling salon, a supermarket, and a restaurant. These
showed only a modest level of intercorrelation (Irl = .44) services were selected based on their overall familiarity among
with a unipolar dissatisfaction item (Mano and Oliver, 1993), the respondent population and because they represent rather
and unipolar measures displayed better predictive validity typical consumption experiences. The data reflect actual reac-
than did conventional bipolar satisfaction measures (West- tions to consumption in a real setting in which consumers are
brook and Oliver, 1991). Peterson and Wilson (1992) report undertaking processes leading to future consumption choices.
variance in consumer responses based upon queries to either Respondents were randomly assigned to a business and
a "satisfaction" item or a "dissatisfaction" item. Further, the screened for familiarity. The sample contained similar num-
observed differences in behavioral reactions of satisfied and bers of men and women overall and within each of the three
dissatisfied customers (Jones and Sasser, 1995; Zeithaml, product categories.
Berry, and Parasuraman, 1996) may be due to the fact that
these are distinct feeling states.
Many satisfaction measurement devices, particularly items
Measures
using a semantic differential format or those that are anchored SATISFACTION. The intent was to develop an overall satis-
by satisfaction and dissatisfaction, imply that satisfaction and faction device with high face validity that could be readily
dissatisfaction are opposite ends of the same construct (corre- applied in most consumer situations. Thus, we sought to
lated - 1). Further, academic studies involving multiple-item develop a multi-item scale with a reasonable number of items.
satisfaction measures operationalize them as a single construct Potential satisfaction items were taken or adapted from avail-
with little consideration that consumers may feel some dissat- able sources (e.g., Oliver, 1981; Westbrook, 1987). Items
isfaction and some satisfaction, and that these feelings may judged by the authors to lack face validity were not included.
exist, at least in part, independently. Distinct (correlated sig- Also, scale devices were selected in an effort to lessen potential
nificantly less than - 1 ) satisfaction and dissatisfaction con- problems that might occur because of the common left skew-
structs would mean that greater information can be extracted ness associated with satisfaction assessments (Peterson and
by considering two constructs than by placing these items Wilson, 1992). Where possible, different rating scales were
into a single measure. Thus, if satisfaction and dissatisfaction used across the items in an effort to reduce skewness. Four
are distinct, a two-factor confirmatory factor model positing satisfaction and four dissatisfaction items were included
separate satisfaction and dissatisfaction constructs should pro- and are shown in Table 2. Additionally, given that the D-T
vide a significantly better fit than would a one-factor model (delighted-terrible) scale is a commonly applied satisfaction
comprised of all the dissatisfaction and satisfaction items. device, separate items indicating how delighted consumers
felt and how terrible they felt were included.
Summary
OTHER MEASURES. Separate measures of conceptually corre-
The preceding section has attempted to identify some key
lated constructs were also collected. These include the cogni-
issues surrounding the measurement of satisfaction and illus-
tively oriented disconfirmation of expectations scale assessed
trate the importance of identifying a succinct, reliable, and
in the "worse than expected" and "better than expected"
valid measure of the satisfaction construct. The following sec-
method (Oliver, 1980; Tse and Wilton, 1988), decision regret,
tion describes a series of analyses investigating these issues.
life satisfaction (Westbrook, 1987), attitude, product category
First, it builds on previous satisfaction measurement theory
involvement, expectations, and future purchase intentions.
(e.g., Babin, Darden, and Griffin, 1994; Oliver, 1981; Peterson
These items were interspersed with the satisfaction/dissatisfac-
and Wilson, 1992; Westbrook, 1981) and tests a measurement
tion items to minimize potential order effects.
device representing satisfaction/dissatisfaction for construct
and face validity. Second, the issue of distinct positive and
negative satisfaction factors is investigated. Finally, the impli-
cations of these analyses are discussed, particularly in terms
Results
of implications for practice and for the direction of future Construct Validity
research trying to better understand and capture consumers' Confirmatory factor analysis was used to investigate the pri-
postpurchase reactions. mary research questions concerning the validity of the satisfac-
tion items and the distinctiveness of positive and negative
satisfaction. Initially, the 10 items discussed above were en-
Research Methods tered into a unidimensional confirmatory factor model. These
Sample results suggested that neither the "delighted" nor "terrible"
Data were gathered from a convenience sample of 250 respon- item achieved sufficient convergence relative to the remaining
dents living in a mid-sized university community. The partici- items to be included on the scale (Anderson and Gerbing,
pants were asked to rate their satisfaction/dissatisfaction with 1988). Briefly, the loading estimate from satisfaction to "de-
one of three services available in a heavily frequented shopping lighted" was .44 and the loading estimate from dissatisfaction
Nature of Satisfaction J Busn Res 131
1998:41:127-136

Table 2. Original Satisfaction/Dissatisfaction Items Drawn from Previously Cited Sources

Abbreviation Item Scale Type

S1 Which of the following choices best descnbes the level of satisfaction you experi- Extreme satisfaction
enced from [name of the relevant business] Very much satisfaction
Satisfaction
Some satisfaction
No satisfaction
$2 Use the following percentage scale to indicate your level of satisfaction. Please Not at all Completely
circle the percentage best describing your level of satisfaction experienced satisfied satisfied
from [name of business] 0% 10% ... 90% 100%
$3 1 feel satisfied with my experience at [name of business] 5-point Likert
$4 Please respond to the following based on how you feel about your overall Not at all Very much
experience with [name of business]. The scale ranges from 1 = "Not at all," 1 2 3 4 5 6 7
meaning you did not feel that emotion at all, to 7 = "Very much felt,"
meaning you felt that emotion very much: Satisfaction

D1 Which of the following choices best describes the level of dissatisfaction you Extreme dissatisfaction
experienced from [name of the relevant business] Very much dissatisfaction
Dissatisfaction
Some dissatisfaction
No dissatisfaction
D2 Use the following percentage scale to indicate your level of dissatisfaction. Please Not at all Completely
circle the percentage best describing your level of dissatisfaction experienced dissatisfied dissatisfied
from [name of businessl 0% 10% ... 90% 100%
D3 I felt some dissatisfaction based on my experience with [name of the business] 5-point Likert
D4 Same as $4 except for the term "Dissatisfaction" Same as $4
Del. Same as $4 except for the term "Delighted" Same as $4
Ter. Same as $4 except for the term "Terrible" Same as $4

tO "terrible" was .58.: Thus, the additional analyses described affects, the eights items were rearranged into a two-factor
below were conducted using the remaining eight items. model and fit reassessed.
The two-factor model produced a X2 of 24.6 with 19 de-
ONE-FACTOR MODEL. The eight items were modeled as indi-
grees of freedom (p < . 17) suggesting that the model explained
cating a single construct. The proposed one-factor model
the original correlations within sampling variation. The corre-
yielded a X2 goodness of fit residual of 65.8 with 20 degrees
sponding GFI and CFI are .97 and .997, respectively, also
of freedom (p < .001). The resulting GFI (goodness of fit
supporting the model's validity. The PNFI is .68. The loading
index) and CFI (comparative fit index) indices were encourag-
estimates support each dimension's convergent validity, rang-
ing at .92 and .97, respectively. The parsimony normed fit
ing between .75 and .90. The variance extracted estimates are
index was .68 (Mulaik et al., 1989). 2 Table 3 shows the stan-
also supportive at .77 and .68 for satisfaction and dissatisfac-
dardized scale loading estimates. All loading estimates ex-
tion, respectively. Further, the X2 difference statistic between
ceeded a magnitude of. 73 and were highly significant. Coeffi-
this and the above model is 41.2 with 1 degree of freedom
cient ot for this scale was .91 and the variance extracted
suggesting a significant improvement in fit over a one-factor
estimate was .68. Each of these measures indicates sufficient
model (p < .001) and evidence of discriminant validity (An-
convergent validity (Anderson and Gerbing, 1988).
derson and Gerbing, 1988). The ~ coefficient representing
TWO-FACTOR MODEL. Although coefficient ot is the most com- the interfactor correlation between the positive and negative
monly used indicator of reliability in the satisfaction literature satisfaction dimensions is strong ( - . 9 1 ) but significantly less
and elsewhere, it does not assess a measure's unidimensional- than - 1 in magnitude (SE = .02; p < .05). Thus, the model
ity (Gerbing and Anderson, 1988). Considering this, and the
theoretical case for distinct positive and negative satisfaction
2The PNFI does not have an absolute interpretation or rule of thumb as
other indices might. Rather it provides a useful relative assessmentof models,
The X 2 overall fit for the 10-itemmodel was 92.9 with 34 degrees of freedom parlicularlywhen other indices are high. All things equal, more parsimonious
with a GH and CFI of .92 and .96, respectively. Overall, these suggest a representations are preferred and are represented by relatively high values
relativelypoorer fit compared to the 8-item model. of the PNH (Mulaik et al., 1989).
132 J Busn Res B.J. Babin and M. Griffin
1998:41:127-136

Table 3. Satisfaction and Dissatisfaction Item Loading Estimates between the two 7-point "degree felt" scales was - . 0 5 (t =
Model - 2 . 1 , p < .05). Second, the ¢p estimate ( - . 9 0 ) was virtually
Item 1-factor 2-factor Model w/4 8s's Overall Model unaffected by this "correction." Thus, the measurement analy-
ses provide considerable support for separate but highly corre-
$1 .88 .90 .90 ° .89 lated satisfaction and dissatisfaction scales.
$2 .86 .86 .86 b .87
$3 .84 .86 .86~ .86
$4 .87 .88 .87 d .85 Nomological Validity
D1 -.77 .81 .8P .78 A final analysis was conducted to investigate discriminant
D2 -.81 .84 .84 b .84 and nomological validity. A confirmatory factor analysis was
D3 -.73 .75 .75~ .76
D4 -.83 .88 .88d .87 conducted including satisfaction, dissatisfaction, and other
.91 .92 .90 .92 .90 .92 .90 measures related to consumer evaluations and reactions. The
Var. Exp. .68 .77 .68 .76 .68 .75 .67 overall fit of this model was adequate considering the large
number of observed variables and sample size considerations
" Errorterm betweenS1 and D1 (0~5~)= .018 (ns). (X2 df = 296 = 660.0; CFI = .90), with items generally
hErrorterm between$2 and D2 (0~<2) = -.001 (ns).
' Errorterm between$3 and D3 (0870= -.011 (ns). displaying high factor loadings, and factors exhibiting high
J Errorterm between$4 and D4 (08~4)= -.047 (t = -2.16, p < .05).
reliability and variance extracted estimates. In all cases, the
correlations were more than two standard errors below one
passes conventional tests of discriminant validity described and in no case did the combination of two factors into one
by Anderson and Gerbing (1988). lead to an insignificant X2 difference statistic (Oliver and Swan,
1989). In other words, pair-wise, two-factor solutions pro-
Satisfaction and Dissatisfaction vided better fits than one-factor solutions, providing evidence
COMMON METHOD BIAS. Recently, Green et al. (1993) have of adequate discrimination. For example, combining satisfac-
argued against a bipolar view of affect ratings by suggesting tion and attitude into a single factor produced a significantly
that bipolarity is "masked" by measurement error arising from worse fit than did a model treating those as separate factors (X2
common methods bias. In three studies using commonly ap- difference = 78.4; df = 1; p < .001). The corresponding
plied mood scales including PANAS and the mood adjective combination of dissatisfaction and attitude also produced a
checklist, 3 they provide evidence that the relatively modest significantly worse fit (X2 difference = 136.4; df = 1; p <
observed correlations suggesting separate "happiness" and •001). This is particularly noteworthy since many of the items
"sadness" dimensions were misleading. Rather, after correcting indicating other constructs are sometimes included as "satis-
for common method covariance by allowing for additional faction" items as discussed earlier. For example, the cognitive
covariation between items of the scale format, the correlations dissonance measure consists of items (for example, "My choice
(ranging from - . 9 7 to - 1 ) did not differ significantly from to go to [name of service provider] was a wise one," and "My
- 1, or at the least, the correlations substantially increased in decision to go to [name of service provider] was the right
magnitude calling into question their discriminant validity. one") drawn from a frequently employed satisfaction scale
The close link between positive and negative affect ratings (Oliver, 1980). This analysis suggests that these items are not
considered in Green et al. (1993) and satisfaction and dissatis- part of the satisfaction construct, but as their content suggests,
faction suggest a similar investigation. Thus, a model was are part of a distinct decision regret factor.
tested that allowed correlated error terms between items shar- Table 4 displays the correlations between satisfaction and
ing a common scale format. For example, an estimate was dissatisfaction and these other measures. These correlations
freed between the 100-point satisfaction and 100-point dissat- are generally consistent with theoretical expectations. For ex-
isfaction scale. The three parallel coefficients representing the ample, both satisfaction and dissatisfaction are highly corre-
similar scale formats were also freed (see Bagozzi, 1993). lated with consumers' future patronage intentions. Note that
The resulting model X2 with 15 degrees of freedom was previous researchers had questioned previous measures ability
19.08 (p -< .21), a GFI of .98, a CFI of .999, and a PNFI of to predict intentions (Jones and Sasser, 1995). Likewise, as
.60. Thus, the overall fit is good but less parsimonious than expected, attitude toward the service provider and satisfaction
the congeneric two-factor model. The loading estimates were and dissatisfaction are highly related. In contrast, consumers'
virtually unchanged as were the variance extracted estimates mood, which can sometimes bias satisfaction measures in low
(see Table 3). Two results contrast with those reported in involvement situations (Babin, Griffin, and Babin, 1994) at
Green et al. (1993). First, only one of the four common the time of responding showed significant but relatively low
method estimates was significant. Specifically, the estimate levels Of correlation. Therefore, the first eight items displayed
in Table 2 exhibit acceptable convergent, discriminant, and
3Scales contain common adjectives such as "pleased," "happy," "satisfied," nomological validity in representing consumer satisfaction and
and "unhappy." dissatisfaction.
Nature of Satisfaction J Busn Res 133
1998:41:127-136

Table 4. Correlations of Other Constructs with Satisfaction and correlation between satisfaction and dissatisfaction. Although
Dissatisfaction one might argue that the size of relationship between satisfac-
Correlation With tion and dissatisfaction is such that the results are equivocal,
Construct Satisfaction Dissatisfaction Reliability at the very least, the scale containing equal numbers of posi-
tively and negatively valenced items allows researchers a fair
Mood .15 -.12 examination of discriminant validity from context to context.
(.07)* COS) .83 Cumulating evidence may provide more insight into when,
Cognitive -.57 .58
dissonance (.03) (.03) .73 if, and how much the two constructs are distinct.
Future patronage .87 -.79 l~he scale's parsimony also makes it advantageous for use
intentions (.04) (.05) .83 in practice. Respondent fatigue is clearly not an issue, yet the
Expectations .59 -.46 advantages of multi-item (more than two) scales are gained
(.06) (.07) .77 (Costner, 1969). Therefore, the scale can be applied widely
Disconfirmation -.32 .19
(.08) (.09) .73 both in theoretical and practical applications and no problems
Attitude .77 - .62 are presented in assessing both internal and external consis-
(.03) (.04) .85 tency.
Some items previously used as indicators of satisfaction
*Standard errors are shown in parentheses beneath the corresponding correlation esti-
mate.
did not provide results consistent with those of the eight
items mentioned above. For example, neither "delighted" nor
"terrible" provided the same level of convergent validity as
did the other items. This should not be taken as diminishing
Discussion the importance of customer delight, rather, it only indicates
The growing recognition of relationship marketing makes the that delight and satisfaction may be distinct at this relatively
importance of understanding postpurchase consumer evalua- basic level of abstraction. Also, items previously used to indi-
tions and reactions precisely all the more important. Among cate satisfaction (e.g., "My decision to . . . was a wise one")
all postpurchase phenomena, consumer satisfaction/dissatis- were shown more representative of distinct constructs (e.g.,
faction has received the most attention. Yet, compared to other cognitive dissonance) with which satisfaction and dissatisfac-
constructs (e.g., value, attitude, self-esteem, quality, etc.), few tion relate. Analyses investigating discriminant validity sup-
articles address specifically the conceptualization and opera- ported this conceptualization.
tionalization of consumer satisfaction/dissatisfaction. Research Hopefully, this research stimulates other research investi-
presented here is directed toward understanding postpurchase gating the precise nature of postpurchase phenomena closely.
phenomenon better by directing attention toward the episte- Results presented here clearly suggest the emotional nature
mology of consumer satisfaction/dissatisfaction. of "satisfaction." Left for further research is the role of other
Eight items that maximize the face validity of satisfaction/ emotions commonly evoked by evaluations of product use
dissatisfaction were used to capture consumers' reactions to outcomes. Work in social psychology suggests numerous ad-
common service encounters. The items were selected to max- ditional emotional outcomes of cognitive appraisals, all of
imize face validity and to minimize potential contamination which have important implications for predicting future be-
from items that might better represent closely related postpur- havior. It is argued here, that "satisfaction" and "dissatisfac-
chase constructs such as cognitive dissonance, beliefs about tion," as in the psychological studies of emotions arising as
disconfirmation and expectations, and others. The model the result of cognitive appraisals (Smith et al., 1993), are two
achieved high overall fit as indicated by confirmatory factor of a potential number of human emotions that result from
analysis and provides previously called for and needed im- these cognitive processes (Van Overwalle, Mervielde, and De
provement to previous measures by more accurately reflecting Schuyter, 1995). Thus, rather than being caused by positive
satisfaction's affective nature (Westbrook and Oliver, 1991). and negative affect occurring after consumption (cf., Oliver,
Further analysis suggested the appropriateness of measur- 1993), satisfaction is one of many potential emotional out-
ing dissatisfaction distinct from satisfaction. Although these comes. Future research should be directed at a better delinea-
measures are correlated highly, a two-factor model positing tion of these emotions. Perhaps appraisals resulting in high
both satisfaction and dissatisfaction as latent constructs, pro- levels of emotions, such as "anger," "disgust," or "elation" are
vides a better fit than does a single satisfaction/dissatisfaction equally or more important than "satisfaction and dissatisfac-
construct. This work can be compared to work in social psy- tion" given the stronger motivational implications. If so, rigor-
chology on state dependent mood that suggests the magnitude ous investigation of the conceptualization and distinctiveness
of correlation between good and bad mood is inflated due to of satisfaction and dissatisfaction from these other feelings is
error arising through use of common measurement devices needed (see Babin, Darden, and Griffin, 199+). Further re-
(Green et al., 1993). The results presented here suggest that search should be directed at the potential moderation of the
correction for common methods variance did not influence the correlation between satisfaction and dissatisfaction. Similar
134 J Busn Res B.J. Babin and M. Griffin
1998:41:127-136

work has shown how the distinctiveness of positive and nega- Consumer Satisfaction and Complaint Reports. Journal of Market-
tive moods are moderated by other factors (Diener and Em- ing Research 20 (March 1983): 21-28.
mons, 1984), so it seems reasonable that similar processes Bolton, Ruth N., and Drew, James H.: A Multistage Model of Cus-
tomer Assessment of Service Quality and Value. Journal of Con-
could occur for postconsumption affects. However, given the
sumer Research 17 (March 1991): 375-384.
past broad-sweeping connotation of satisfaction, this research
Bradburn, N. M.: The Structure of Psychological Well-Being, Adeline
provides a start in the better measurement and operationaliza- Press, Chicago, IL. 1969.
tion of postpurchase constructs by providing parsimonious
Brayfield, Arthur H., and Rothe, Harold F.: An Index of Job Satisfac-
and valid devices for assessing consumer "satisfaction" and tion. Journal of Applied Psychology 35 (October 1951): 307-311.
"dissatisfaction." Our hope is that by providing a succinct,
Brown, Steven P., and Peterson, Robert A.: The Effect of Effort on
reliable, and valid satisfaction and dissatisfaction scale, re- Sales Performance and Job Satisfaction. The Journal of Marketing
searchers can take a common approach to measuring this 58 (April 1994): 70-80.
important, yet elusive, construct. Churchill, Gilbert A., and Surprenant, Carol: An Investigation into
the Determinants of Consumer Satisfaction. Journal of Marketing
An earlier version of this paper was selected for the 1996 Steven J. Shaw Research 19 (November 1982): 491-504.
Award as the outstanding paper presented at the 1996 Southern Marketing Costner, Herbert L.: Theory, Deduction and Rules of Correspon-
Association Annual CoMerence. dence. American Journal of Sociology 75 (September 1969): 245-
263.
Crosby, Lawrence A., and Taylor, James R.: Consumer Satisfaction
References with Michigan's Container Deposit Law An Ecological Perspec-
Anderson, James C., and Gerbing, David W.: Structural Equation tive. Journal of Marketing (Winter 1982): 47-60.
Modeling in Practice: A Review and Recommended Two-Step Darden, William R., and Babin, Barry J.: Exploring the Concept of
Approach. Psycholo~cal Bulletin 103 (March 1988): 411--423. Affective Quality: Expanding the Concept of Retail Personality.
Anderson, James C., and Gerbing, David W.: Assumptions of the Journal of Business Research 29 (February 1994): 101-110.
Two-Step Approach to Latent Variable Modeling. Sociological Diener, Ed, and Emmons, Robert A.: The Independence of Positive
Methods and Research 20 (February 1992): 321-333. and Negative Affect. Journal of Personality and Social Psychology
Anderson, Eugene W., Fornell, Claes, and Lehmann, Donald R.: 47 (November 1984): 1253-1265.
Consumer Satisfaction, Market Share, and Profitability: Findings Edell, Julie, and Burke, Marian Chapman: The Power of Feelings in
from Sweden. Journal of Marketing 58 (July 1994): 53-66. Understanding Advertising Effects. Journal of Consumer Research
Andreasen, Alan R.: Consumer Responses to Dissatisfaction in Loose 14 (December 1987): 421-433.
Monopolies. Journal of Consumer Research 12 (September 1985): Fehr, Beverley, and Russell, James A.: Emotion Viewed from a Proto-
135-141. type Perspective. Journal of Personality and Social Psychology 113
Andreasen, Alan R.: Life Status Changes and Changes in Consumer (October 1984): 464-486.
Preferences and Satisfaction. Journal of Consumer Research 11 (De- Fornell, Claes: A National Customer Satisfaction Barometer: The
cember 1984): 784-794. Swedish Experience. Journal of Marketing 56 (January 1992):
6-21.
Andrews, Frank M., and Withey, Stephen B.: Social Indicators of Well-
Being, Plenum Press, NY. 1976. Fornell, Claes, and Robinson, William T.: Industrial Organization and
Consumer Satisfaction and Dissatisfaction. Journal of Consumer
Arnold, M. B.: Physiological Differentiation of Emotional States. Psy-
Research 9 (March 1983): 403-412.
chological Review 52 (1945): 35-48.
Gerbing, David A., and Anderson, James C.: An Updated Paradigm
Babin, Barry J., and Darden, William R.: Good and Bad Shopping for Scale Development Incorporating Unidimensionality and Its
Vibes: Spending and Patronage Satisfaction. Journal of Business Assessment. Journal of Marketing Research 25 (May 1988): 186-
Research 35 (March 1996): 201-207. 192.
Babin, Barry J., Darden, William R., and Griffin, Mitch: An Empirical Green, Donald Phillip, Goldman, Susan Lee, and Salovey, Peter:
Comparison of Alternative Conceptualizations of Postconsump- Measurement Error Masks Bipolarity in Affect Ratings. Journal of
tion Reactions. Journal of Consumer Satisfaction~Dissatisfaction and Personality and Social Psychology 64 (June 1993): 1029-1041.
Complaining Behavior 7 (1994): 172-182.
Havlena, William J., and Holbrook, Morris B.: The Varieties of Con-
Babin, Barry J., Griffin, Mitch, and Babin, Laurie A.: The Effect of sumption Experience: Comparing Two Typologies of Emotion in
Motivation to Process on Consumers' Satisfaction Reactions, in Consumer Behavior. Journal of Consumer Research 13 (December
Advances in Consumer Research, vol. 21, Chris Allen and Deborah 1986): 394-404.
Roedder-John, eds., Assoc. of Consumer Research, Provo, UT.
1994, pp. 406-411. Herberlein, Thomas A., Linz, Daniel, and Ortiz, Bonnie P.: Satisfac-
tion, Commitment and Knowledge of Customers on a Mandatory
Bagozzi, Richard P., and Moore, David J.: Public Service Announce- Participation Time-of-Day Electricity Pricing Experiment. Journal
ments: Emotions and Empathy Guide Prosocial Behavior. Journal of Consumer Research 9 (June 1982): 106-114.
of Marketing 58 (January 1994): 56-70.
Hunt, Keith H.: CS/D-Overview and Future Directions, in Conceptual-
Bagozzi, Richard P.: An Examination of the Psychometric Properties ization and Measurement of Consumer Satisfaction and Dissatisfac-
of Measures of Negative Affect in the PANAS-X Scales. Journal of tion, H. Keith Hunt, ed., Marketing Science Institute, Cambridge,
Personality and Social Psycholow 65 (October 1993): 836-851. MA. 1977.
Bear&n, William O., and Teel, Jesse E.: Selected Determinants of lzard, Carrol E.: Human Emotions, Plenum Press, NY. 1977.
Nature of Satisfaction J Busn Res 135
1998:41:127-136

Johnson, Michael D., Anderson, Eugene W., and Fornell, Claes: Plutchik, Robert: A General Psychoevolutionary Theory of Emotion,
Rational and Adaptive Performance Expectations in a Customer in Emotion: Theory, Research, and Experience, R. Plutchik and H.
Satisfaction Framework. Journal of Consumer Research 21 (March Kellerman, eds., Academic Press, New York. 1980, pp. 3-31.
1995): 491-504. Richins, Marsha L.: Negative Word-of-Mouth by Dissatisfied Custom-
Johnston, Mark W., Parasuraman, A., Futrell, Charles M., and Black, ers. Journal of Marketing 47 (Winter 1983): 68-78.
William C.: A Longitudinal Assessment of the Impact of Selected Russell, James A.: In Defense of a Prototype Approach to Emotion
Organizational Influences on Salespeople's Organizational Com- Concepts. Journal of Personality and Social Psychology 60 (January
mitment During Early Employment. Journal of Marketing Research 1991): 37-47.
27 (August 1990): 333-334.
Russell, James A.: Evidence of Convergent Validity on the Dimensions
Jones, Thomas O., and Sasser, Earl W. Jr.: Why Dissatisfied Custom- of Affect. Journal of Personality and Social Psychology 36 (October
ers Defect. Harvard Business Review (November/December 1995): 1978): 1152-1168.
88-99.
Rust, Roland T., and Zahorik, Anthony J.: Customer Satisfaction,
Lacher, Kathleen T., and Mizerski, Richard: An Exploratory Study Customer Retention, and Market Share. Journal of Retailing 69
of the Responses and Relationships Involved in the Evaluation (Summer 1993): 193-215.
of, and in the Intention to Purchase, New Rock Music. Journal
of Consumer Research 21 (September 1994): 366-380. Smith, Craig A., Haynes, Kelly N., Lazarus, Richard S., and Pope,
Lois K.: In Search of the "Hot" Cognitions: Attributions, Apprais-
Lazarus, Richard S.: Emotion and Adaption, Harper, New York. 1991. als, and Their Relation to Emotion. Journal of Personalityand Social
Lazarus, Richard S.: Thoughts on the Relations Between Emotion Psychology 65 (November 1993): 916-929.
and Cognition. American Psychologist37 (September 1982): 1019- Smith, Craig A., and Ellsworth, Phoebe C.: Patterns of Cognitive
1024. .Appraisal in Emotion. Journal of Personality and Social Psychology
Leavitt, Clark: Consumer Satisfaction and Dissatisfaction: Bipolar or 48 (April 1985): 813-838.
Independent, in Conceptualization and Measurement of Consumer Stayman, Douglas M., Alden, Dana L., and Smith, Karen H.: Some
Satisfaction and Dissatisfaction, H. Keith Hunt, ed., Marketing Sci- Effects of Schematic Processing on Consumer Expectations and
ence Institute, Cambridge, MA. 1977, pp. 137-149. DisconfirmationJudgments. Journal of Consumer Research 19 (Sep-
Locke, Edwin A.: What is Job Satisfaction? Organizational Behavior tember 1992): 240-255.
and Human Perlbrmance 4 (November 1969): 309-336. Stein, Nancy L., and Oatley, Keith: Basic Emotions: Theory and
Maddox, R. Neff: Two-Factor Theory and Consumer Satisfaction: Measurement. Cognition and Emotion 6 (March/April 1992): 161-
Replication and Extension. Journal qf Consumer Research (june 168.
1981): 97-103. Swan, John E., and Oliver, Richard L.: Postpurchase Communications
Mano, Haim, and Oliver, Richard L.: Assessing the Dimensionality by Consumer. Journal of Retailing 65 (Winter 1989): 516-533.
and Structure of the Consumption Experience: Evaluation Reel- Taylor, Steven A., and Baker, Thomas L.: An Assessment of the
ing, and Satisfaction. Journal of Consumer Research 20 (December Relationship Between Service Quality and Customer Satisfaction
1993): 451-466. in the Formation of Consumers' Purchase Intentions. Journal of
Mehrabian, Albert, and Russell, James A.: An Approach to Environmen- Retailing 70 (Summer 1994): 163-178.
tal Psychology, MIT Press, Cambridge, MA. 1974. Tse, David K., and Wilton, Peter C.: Models of Consumer Satisfaction
Mulaik, Stanley A., James, Larry R., Van Alstine, Judith, Bennett, Formation: An Extension. Journal of Marketing Research 25 (May
Nathan, Lind, Sherri, and Stilwell, C. Dean: Evaluation of Good- 1988): 204-212.
ness-of-Fit Indices for Structural Equations Models. Psychological Van Overwalle, Frank, Mervielde, Ivan, and De Schuyter, Joris: Struc-
Bulletin 105 (March 1989): 430-445. tural Modeling of the Relationships between Atributional Dimen-
Oliver, Richard L.: A Cognitive Model of the Antecedents and Conse- sions, Emotions, and Performance of College Freshmen. Cognition
quences of Satisfaction Decisions. Journal of Marketing Research and Emotion 9 (January 1995): 59-86.
17 (November 1980): 460-469. Watson, David, Clark, L. A., and Tellegen, Auke: Development and
Oliver, Richard L.: Cognitive, Affective, and Attribute Bases of the Validation of a Brief Measure of Positive and Negative Affect: The
Satisfaction Response. Journal of Consumer Research 20 (December PANAS Scales. Journal of Personality and Social Psychology 54 (June
1993): 418-430. 1988): 1063-1070.
Oliver, Richard L.: Measurement and Evaluation of Satisfaction Pro- Watson, David, and Tellegen, Auke: Toward a Consensual Structure
cesses in Retail Settings.Journal of Retailing 57 (Fall 1981): 25-48. of Mood. Psychological Bulletin 98 (1985): 219-235.
Oliver, Richard L., and DeSarbo, Wayne S.: Response Determinants Westbrook, Robert A: Intrapersonal Affective Influences on Con-
in Satisfaction Judgments. Journal of Consumer Research 14 (March sumer Satisfaction with Products. Journal of Marketing Research 7
1988): 495-507. (June 1980): 49-54.
Oliver, Richard L., and Swan, John E.: Equity and Disconfirmation Westbrook, Robert A.: Sources of Consumer Satisfaction with Retail
Perceptions as Influences on Merchant and Product Satisfaction. Outlets. Journal of Retailing 57 (Fall 1981): 68-85.
Journal of Consumer Research 16 (December 1989): 372-383. Westbrook, Robert A.: Product/Consumption-Based Affective Re-
Perkins, Deborah S.: An Update of the Consumer Satisfaction, Dissat- sponses and Postpurchase Processes. Journal of Marketing Research
isfaction and Complaining Behavior Bibliography: Revolution and 24 (August 1987): 258-270.
Evolution. Journal of Consumer Satisfaction/Dissatisfactionand Com- Westbrook, Robert A., and Oliver, Richard L.: Developing Better
plaining Behavior 6 (1993): 217-279. Measures of Consumer Satisfaction: Some Preliminary Results,
Peterson, Robert, and Wilson, William R.: Measuring Customer Satis- in Advances in Consumer Research, vol. 8, Kent B. Monroe, ed.,
faction: Fact and Artifact. Journal of the Academy of Marketing Association for Consumer Research, Ann Arbor, MI. 1981, pp.
Science 20 (Winter 1992): 61-72. 94-99.
136 J Busn Res B.J. Babin and M. Griffin
1998:41:127-136

Westbrook, Robert A., and Oliver, Richard L.: The Dimensionality Woodruff, Robert B., Clemons, Scott, Schumann, David W., Gardial,
of Consumption Emotion Patterns and Consumer Satisfaction. Sarah F., and Burns, Mary Jane: The Standards Issue in CS/D
Journal of Consumer Research 18 (June 1991): 84-91. Research: A Historical Perspective. Journal of Consumer 5atisfac-
Woodruff, Robert B., Cadotte, Ernest R., and Jenkins, Roger L.: tionIDissatisfaction and Complaining Behavior 4 (1991): 103-109.
Modeling Consumer Satisfaction Processes Using Experience- Zeithaml, Valarie A., Berry, Leonard L., and Parasuraman, A.: The
Based Norms. Journal of Marketing Research 20 (August 1983): Behavioral Consequences of Service Quality. Journal of Marketing
296--304. 60 (April 1996): 31-46.