Advances in Consumer Research Volume 10, 1983 Pages 250-255
THE ROLE OF INVOLVEMENT IN SATISFACTION PROCESSES
Richard L. Oliver, Washington University
William O. Bearden, University of South Carolina
A two-wave panel study of users of a new encapsulated appetite suppressant was performed to determine if an expectancy disconfirmation model of consumer satisfaction operated in identical fashion for high and low involvement users. The results showed that mean pre-exposure affect and intention levels were greater for high involvement users. This effect generalized to the post-usage time period where satisfaction and other cognitions were also higher in the high involvement group. A path-analytic regression showed that the satisfaction model was equally valid for both involvement groups although a small number of differences was found when coefficient parameters were compared.
Two streams of research have received some attention in the consumer behavior literature of late. One involves the meaning, measurement, and proper positioning of satisfaction in behavioral process (LaTour and Peat 1979; Oliver 1980, 1981; Westbrook and Oliver 1981), while the second addresses the degree to which product involvement or importance influences or moderates behavioral process (Bloch 1981, 1989; Kassarjian 1981; Ray 1979; Robertson 1976). Both research areas have now been developed to the point where investigators can draw on some theory, some well designed and executed empirical works, and some meaningful findings to structure future studies. This paper adopts an emerging framework of satisfaction processes in the context of a product category which is involving to consumers to varying degrees.
A Satisfaction Framework
line satisfaction framework employed here essentially replicates Oliver's (1980, p. 462) two-stage expectancy disconfirmation model of satisfaction effects on attitude change and intention. Oliver proposed that pre-usage expectations are instrumental for pre-trial attitude development and that attitude influences intention to buy in the manner suggested by Fishbein and Ajzen (1975). The post-usage experience begins with disconfirmation (fulfillment) perceptions which, together with expectations, affect satisfaction. Satisfaction, in turn, influences one's post-usage attitude which subsequently affects new purchase intentions.
The theoretical underpinnings of the model derive from Howard and Sheth's (1969, p. 147) discussion of postpurchase attitude which was thought to be a function of prepurchase attitude and satisfaction relative to prepurchase attitude. Based on work by Helson (1964) and Weaver and Brickman (1974), the Howard and Sheth satisfaction concept was described as having two antecedents, a separate disconfirmation variable and one's expectations. When this framework was integrated with Fishbein's (197;) attitude model, the expectancy disconfirmation model tested in Oliver (1980) resulted.
A number of studies have shown support for this scheme. These include Bearden, Crockett, and Teel (1981), Oliver (197,, 1980), Oliver and Linda (1931), Swan (1977), and Swan and Trawick (1981). While no study shows unequivocal support, the model coefficients tend to be consistently significant. Moreover, when the model is tested in path analytic modes (e.g., Oliver, 1980), only the immediate (direct) coefficients tend to be significant. Variables having less immediate (indirect) effects decline in significance as later stages in the model are tested. Given this background, the time would appear to be ripe for extensions or modifications of the model.
The Role of Involvement
One impetus for testing the effects of product involvement on satisfaction was Day's (1977) observation that satisfaction decisions are contingent on the incidence of postpurchase evaluation. He argued that, for products of low importance, evaluations ma not be triggered. Swan and Trawick (1979) tested this notion and found everyday products resulted in fewer evaluations than more involving products. Bloch (1982) drew a similar conclusion in his theoretical discussion of product importance. He argued that important products are more likely to affect attitudes and behaviors and that importance may be situation-specific.
Rothschild (1978, 1979) presents a theoretical basis for the effect of involvement on cognitive process. He argued that high involvement may increase one's arousal a: early stages of consumer decision making (e.g., awareness), but should have reduced effects on later stages because the latter stages are more directly antecedent to behavior. Moreover, he suggested that low involvement consumers should have less consistent attitude structures before purchase due to a lower level of information processing motivation, but should have more consistent structures after usage. Rothschild (1978) reported data supporting the arousal hypothesis. Higher levels of awareness and attitude (but not intention) were found for more involved political contests.
Bowen and Chaffee (1974) also present empirical evidence that highly involved consumers make different prepurchase judgments than do low involved consumers. They found that highly involved individuals appear to display higher advertisement evaluations and intentions to buy in the relevant product class. It is not clear, however, whether involvement universally causes higher evaluations. Rather, Houston and Rothschild (1978) argue that involvement enhances the intensity of judgment; the direction is influenced by the affective tone attached to the object. Thus, the Bowen and Chaffee findings may simply reflect an overall generally favorable reflection of the communications used in their study.
Objectives and Hypotheses
The preceding observations advanced by involvement theorists and by Day (1977) in the context of satisfaction would suggest that low involved consumers should be less likely to make strong evaluations, if they make them at all. This suggests that the variables tested in the proposed satisfaction framework may differ across involvement categories on two levels. First, significant differences are expected on pre- and post-usage mean ratings when highly involved consumers are compared to those less involved. Second, the variable relationships in the model (i.e., the model coefficients) may be expected to differ when separate tests are conducted for high and low involvement subsamples.
The directions of effects are suggested by the preceding discussion. Because all subjects were users by choice, one would expect higher pre-usage evaluations from more involved subjects (the intensity effect). Post-usage ratings should also be higher for involved subjects due to either a dissonance effect or to an adaptation level effect (Helson 1964). The effect of involvement on post-usage rating levels has not been addressed to date. With regard to the model parameters, pre-usage model coefficients should be lower for less involved consumers as they would be expected to have lower levels of cognition in their attitude structures than more highly involved consumers. No differences would be expected for the post-usage coefficients as both high and low involved consumers would have equal amounts of cognitive information.
In broader terms, two overriding objectives guided the study. The first was to test the validity of the Oliver (1980) satisfaction model in a new context while the second was to determine if involvement levels affect the variable levels and the performance characteristics of the model.
Data were collected via three mailings to 800 members of a regional consumer panel. In the first wave, expectations, attitudes, and intentions regarding a new heavily marketed diet suppressant were collected. One-month after the initial survey, a four-week supply of 24-hour capsules was mailed to each respondent. Following the four-week consumption period, a second mailing collected data regarding product usage, disconfirmation of expectations, feelings of satisfaction, and follow-up attitudes and intentions. A third mailing five months after the first survey also assessed satisfaction, attitude, and intention data. This study is based on the data from the first two periods of the longitudinal design. Ninety-one users and 269 non-users provided complete responses to all three mailings.
Panel members were selected to be representative of urban and suburban households with family incomes above $10,000 in the two states covered by the survey. As such, the demographic profile of the panel is somewhat older and upscale economically. Of the respondents in this study, 56% of the respondents were female; 89% were white. Seventy percent had at least some education beyond high school and the median income category was $20-29,999.
Expectations were viewed as the sum of the Fishbein ab products (Fishbein and Ajzen 1975) where all beliefs (bs) were worded in terms of expected future outcomes. These beliefs were obtained by asking subjects to scale the possibility of receiving 14 outcomes of taking the appetite suppressant on 7-point probability scales ranging from "no chance" to "certain." The evaluation (a) component was measured by asking respondents to evaluate each outcome on a 7-point good-bad scale.
Oliver's (1980),9-item semantic differential attitude scale was used for both the pre- and post-usage measures of affect toward the new product. The pre- and post-intention measure consisted of a 3-item semantic differential recommended by Ajzen and Fishbein (1980). The three bi-polar adjectives were likely-unlikely, probable-improbable, and possible-impossible.
The disconfirmation measure was an extension of Oliver's two-item "greater-than-expected, worse-than-expected" scale. The one used in the present study contained four items covering an overall better-worse than expected evaluation plus one relating to benefits only, one relating to problems only, and one evaluating the inherent positive and negative surprise over the product's performance. Satisfaction was measured with the Westbrook and Oliver (1981) Likert scale.
Finally, involvement was measured in the pre-usage wave of the study on a nine-item Likert scale suggested by Bloch's (1981) procedure. He recommended that such scales contain items reflecting (a) interest in the product area, (b) readiness to talk about the product, (c) the relation between the product and important needs or values, and (d) use of the product to express one's self-concept. Items reflecting each of these dimensions were constructed and embedded in a larger section on "general opinions" about weight loss.
Coefficient alpha reliabilities, calculated for all scales used here, are shown in Table 1. All reliabilities exceeded .8, indicating that the measures were sufficiently stable for basic research (Nunnally 1978). Of special note are the satisfaction and pre- and post-attitude reliabilities which exceeded .9 (.92, 94, and .95 respectively).
The initial analysis examined mean variable differences between high and low involvement users. User involvement categories were defined by a median split. Because of tied scores at the median, the final subgroup samples consisted of 47 high and 44 low involvement users.
The second phase of the study examined the performance of the proposed satisfaction model for the entire group of users and for the two involvement subsets. This was accomplished in the manner proposed by Oliver (1980). He used a just-identified, fully recursive path analysis (Duncan 1975) whereby all temporally defined subsequent variables in the model are regressed on all antecedents. This is considered to be a rather stringent test of causal effects because variable interrelationships (laths) hypothesized to be zero are required to not attain significance while paths hypothesized to be significant must compete with extraneous relationships, multicollinearity, and other system "noise." The model predicts that only the most immediate antecedent coefficient should be significant, plus any adaptation level effects (see Oliver 1980).
When t-tests were performed on all variables in the model across the high and low involvement groups, results consistent with the hypotheses were found. With the exception of the expectation measure, the high involvement group scored significantly higher (p < .01) on all pre- and post-usage variables. The expectation measure, although not significant, yielded findings consistent with the direction of effect. In accord with the bowen and Chaffee (1974) results, it appears that involvement served to inflate judgments in some consistent fashion.
REGRESSION RESULTS FOR ALL USERS AND FOR THE HIGH AND LOW INVOLVEMENT SUBGROUPS
The path analysis results, shown in Table 2, compare the model coefficients for the entire sample and the high and low involvement groups respectively. The model performed fairly well when tested on the entire sample in that the results generally conformed to those in Oliver (1980). Post-usage intention was clearly a function of post-usage attitude and, to a lesser but significant degree, satisfaction. Pre-usage intention appeared to provide the adaptation level function discussed by Oliver. Post-usage attitude was a function of satisfaction as hypothesized and, to a lesser extent, disconfirmation. Pre-usage attitude coefficients were in the expected direction, but were not significant.
Satisfaction was a very significant function of disconfirmation only. No expectation effects were evident in contrast to the findings reported in Oliver (1980). Disconfirmation, in turn, was exogenous to the system, having no antecedents. This is consistent with prior findings and with the basic expectancy disconfirmation paradigm. Finally, pre-usage intention was a function or pre-usage attitude which, in turn, was a function of expectations, as hypothesized. This latter relationship was rather modest, however, and raises questions about the attribute list used in the expectation measure.
Adjusted coefficients of determination showed that all post-exposure phenomena were adequately predicted. In fact, the proportion of future intention variance explained approached 75%. Among the pre-exposure variables, intention was predicted well, but attitude was not because of the low correlation between expectation and attitude.
Comparison of the high and low involvement findings showed a great deal of similarity and some interesting differences. Generally, the basic satisfaction model operated in the hypothesized manner for both subgroups. Differences existed largely in the size of certain coefficients. summarized below.
First, expectations were related to attitudes, but only in the high involvement group. Second, disconfirmation was a positive function of pre-attitude in the low involvement group. Third, satisfaction was modestly (but not significantly) positively related to prior attitude in the low involvement group. The same coefficient for the high involvement group was negative. Finally, all post-exposure criteria yielded higher coefficients or determination for the low involvement group, indicating that the model may be a more accurate representation or satisfaction processes for that subsample.
The results would appear to have implications for both satisfaction theory and the role of involvement in satisfaction processes. The involvement issue can be further divided into one of level, process, and effect. Level refers to involvement's influence on variable states in a descriptive sense; process refers to its effect on the operation of constructs in the satisfaction model under levels of involvement; while effect refers to the strength and magnitude or the coefficients. Each will be discussed in turn.
The Satisfaction Process
The basic satisfaction model proposed here received support in some areas and only mixed support in others. Of the features receiving tentative support was the postexposure causal chain: disconfirmation 9 satisfaction P post-attitude - post-intention. All path coefficients in the regressions testing this scheme were above .5 (.52 - .65). This is consistent with Oliver's (1980) findings where coefficients ranged between .32 and .55. The basic Fishbein pre-exposure model (Eab > attitude w intention) also received support except that the ab products and attitude were not related for the low involvement group. Surprisingly, the correlation between the ab products and attitude, even in the high involvement group, did not exceed .4. The authors have no explanation for this except that examination of the individual ab products showed that some of the attributes appeared to be extraneous.
Unfortunately, the adaptation-level hypothesis advanced by Oliver (1980) was not supported at conventional significance levels in this study. Specifically, satisfaction was a function only of disconfirmation, except in the low involvement group where attitude generated a moderate (.28) but non-significant path coefficient. Similar results were obtained for attitude and intention except that the pre-intent on coefficient was just significant for the post-intention criterion.
A number of explanations are possible for the lack of adaptation level findings. It is suggested here that the product category is or the "strong disconfirmation" variety where disconfirmation (positive or negative) is so apparent that the disconfirmation effect overwhelms the expectation effect. This would require that both expectations and performance be objective (e.g., pounds or weight lost in the present case) so that comparisons can be easily made.
Alternatively, history effects may have intervened between the time the pre-measure was taken and the second wave or the study, thereby obscuring the effect of expectation levels. While nothing was done to monitor history over the wave interval, it is possible that competitive promotion or the sponsoring company's own advertising may have altered expectation levels. Still another possibility is the intervention of alternative dietary factors which subjects were employing at the same time. Other activities including smoking, caloric intake reduction, exercise, etc., could also have caused weight change which may have been attributed incorrectly to the appetite suppressant. This phenomenon would serve to enhance the disconfirmation effect relative to expectation.
The possibility remains, however, that the basic model is specified incorrectly. Perhaps satisfaction is a function or disconfirmation and other unmeasured factors, a criticism which is relevant to all variables in the causal chain. In light of prior research showing significant expectation effects in similar product usage situations, the authors are inclined to view this study as a special case which requires further study to determine why expectation effects were not more prominent.
The Role of Involvement
Level effects. The results showed rather dramatically that involvement tends to raise evaluations prior to product use and that these elevated expectations carry over to post-usage evaluations as well. The literature is not clear on why this should be. In one study where this was also empirically observed (Bowen and Chaffee 1974), the authors did not address the phenomenon from a deterministic standpoint.
Two explanations may pertain. The first is that involvement intrinsically elevates ratings in the relevant product class. The second is that involvement interacts with actual states of nature in the manner proposed by Sherif and Sherif (1967). Namely, involvement is thought to affect the acceptance region, in effect generating more extreme evaluations of disconfirmed expectations. The authors are inclined to subscribe to the latter interpretation because of the pattern of findings generally. For example, it was found that the high involved consumers were more pleased with the product's performance. In essence,-their positive discrepancies were higher, resulting in more extreme positive ratings in the post-usage stage, In the pre-usage stage, they may have been more receptive (accepting?) of product claims, although this interpretation is highly speculative at this time.
Differences in model specification. The results provide some insights to determine if the same satisfaction process operates for high and low involved consumers. As was noted, three differences were found. In contrast to the high involvement group, low involved users did not tie their prior attitudes to expectations, but did use their prior attitude to respond to the disconfirmation and satisfaction items. The first of these observations has implications for attitude theory in that rational cognitive process may not be relevant for low involved consumers (Krugman 1965).
The second observation would appear to have implications for satisfaction management as it suggests that low involved consumers use their prior attitudes to interpret disconfirmation, i.e., the more favorable one's attitude, the more likely one is to perceive positive disconfirmation. Thus, low involvement disconfirmation may follow the same psychological process used in forming satisfaction judgments. That is, prior attitudes provide a psychological bias either for or against the object.
Influences in coefficient magnitudes. Where similar coefficients across the involvement groups were in identical directions, differing only in magnitude, no meaningful differences were found. It was apparent that the low involvement group generated larger coefficients of determination. Because the path coefficients were approximately the same size across groups, the larger coefficients may be attributed to less multicollinearity in the low involvement group or to smaller standard errors. In any event, these differences appear to have little meaning for either satisfaction or involvement theory.
Summary and Limitations
This study sought to determine how involvement affects satisfaction judgments among users or a non-prescription appetite suppressant. The results tend to raise more questions than they answer. Involvement was related to the absolute level of pre and post evaluations for reasons which are not entirely clear. It appeared to affect certain of the model parameters, reducing them for low involved consumers in one case (expectations S attitude) and increasing them for low involved users in two other cases (attitude s disconfirmation and attitude s satisfaction). In contrast, high involved users seemed not to follow an adaptation level process for making satisfaction judgments. However, in spite of the differences noted here, the basic post-usage satisfaction process (disconfirmation S satisfaction - post-attitudes post-intention) operated for both involvement groups in almost identical fashion.
In summary, one might propose the following regarding the role of involvement in satisfaction: High involvement decreases one's sensitivity to pre-usage phenomena while low involvement causes the general tone of pre-usage affect (attitude) to influence post-usage evaluations. In Oliver's (1980) terms, low- involvement enhances adaptation level effects while high involvement tends to negate these effects.
The reason for this is conjecture at this point. However, it may be said that high involvement increases sensitivity to outcome phenomena (e.g., performance, disconfirmation) irrespective of input cognition. Low involvement, in contrast, is more subject to halo across the buying process. Low involvement consumers should be less motivated to process performance distinct from prior evaluations. Perhaps motivation is the central concept underlying the phenomena observed here. It would be an understatement to say that more work is needed.
Limitations. A number of limitations are apparent from the design itself. One potentially serious problem is the complexity of self-selection biases. First, panel members represent a volunteer sample; second, users were allowed to declare themselves; and third, high and low involvement subgroups were defined relative to their distribution in the user sample. It may be that all users were highly involved to varying degrees.
Another problem, alluded to previously, was the inability to control for history and for the degree to which usage regimens were adhered to. Another similar concern involves the effect of other weight loss activities on weight reduction and attributions of causality.
Common method bias is also a possible explanation for certain of the findings.. Although the pre- and post-usage surveys were temporally separated, all pre-measures and all post-measures were obtained on two respective instruments . This problem poses particular concern for the validate of the causal chain. It may be that the high coefficients of determination for some of the criteria may reflect the juxtaposition of measures.
In defense, it must be recognized that the problems noted above are direct outcomes of the panel design, the survey approach, and the nature of the product category. In an effort to obtain realism and external validity, the benefits of laboratory control and manipulation were sacrificed. However, it can be said that potential members of the appetite suppressant market were surveyed without inducements or demand artifacts, that satisfaction was based on actual voluntary usage of the product in the normal usage environment, and that, although methods variance remains a problem, the scales used did demonstrate high reliabilities.
As a final note, readers are cautioned that the regression approach used here presents further shortcomings. It assumes that the variables are measured without error, that they are pure measures of their constructs, that the regression weights mirror construct relationships, and that the structural equations in Table 2 are specified correctly. In contrast, Bagozzi (1977) has suggested that maximum likelihood structural equation modeling be used for validating causal frameworks such as that suggested here. In addition to providing estimates of measurement error in the constructs and errors in the equations, it also generates diagnostic data for testing rival hypotheses and provides one goodness-of-fit test for the system of equations. This technique is highly recommended and should be used for future refinements in satisfaction and involvement models.
With this recommendation in mind, others are encouraged to replicate this study in the field or in the laboratory, perhaps using a different operationalization of involvement. The Likert scale is widely accepted (Bloch 1981) but other forms of involvement measurement have been applied (e.g., Clarke and Belk 1979). Resolution of some of the issues raised here may be dependent on improvements in the measurement of involvement.
Ajzen, Icek and Fishbein, Martin (1980), Understanding Attitudes-and Predicting Social Behavior, Englewood Cliffs, N.J.: Prentice-Hall. Inc.
Bagozzi, Richard P. (1977), "Structural Equation Models in Experimental Research," Journal of Marketing Research, 14, 209-226.
Bearden, William O., Crockett, Melissa and Teel, Jesse E. (1981), "Disconfirmation of Purchase Expectations and Subsequent Consumer Attitudes and Patronage Intentions," paper delivered at the American Marketing Association Retail Patronage Theory Conference. May 1981.
Bloch, Peter H. (1981), "An Exploration into the Scaling of Consumers' Involvement with a Product Class," Advances in Consumer Research, Vol. 8, R. Monroe, ed., Ann Arbor: Association for Consumer Research, 61-65.
Bloch, Peter H. (1982), "Involvement Beyond the Purchase Process: Conceptual Issues and Empirical Investigation," in Advances in Consumer Research, Vol. 9, Andrew Mitchel ed., Ann Arbor: Association for Consumer Research.
Bowen, Lawrence and Chaffee, Steven H. (1974), "Product Involvement and Pertinent Advertising Appeals," Journalism Quarterly, 51, 613-691.
Clarke, Keith and Belk, Russell (1979), "The Effects of Product Involvement and Task Definition on Anticipated Consumer Effort," In Advances in Consumer Research, Vol. 6, W. L. Wilkie, ed., Ann Arbor: Association for Consumer Research, 313-318.
Day, Ralph L. (1977), "Toward a Process Model of Consumer Satisfaction," in Conceptualization and Measurement of Consumer Satisfaction and Dissatisfaction, H. Keith Hunt, ed., Cambridge, Mass.: Marketing Science Institute, 153-183.
Duncan, Otis Dudley (1975), Introduction to Structural Equation Models. New York: Academic Press.
Fishbein, Martin, and Ajzen, Icek (1975), Belief, Attitude, Intention, and Behavior, Reading, Mass.: Addison-Wesley.
Helson, Harry (1964), Adaptation-Level Theory, New York: HarDer & Row.
Houston, Michael J. and Rothschild, Michael L. (1978), "Conceptual and Methodological Perspectives on Involvement," in 1978 Educators' Proceedings, S.C. Jain, ed., Chicago: American Marketing Association, 184-187.
Howard, John A. and Sheth, Jagdish N. (1969), The Theory of Buyer Behavior, New York: John Wiley & Sons.
Kassarjian, Harold H. (1981), "Low Involvement: A Second Look," in Advances in Consumer Research, Vol. 8, K. B. Monroe, ed., Ann Arbor: Association for Consumer Research, 31-34.
Krugman, Herbert E. (1965), "The Impact of Television Advertising: Learning Without Involvement," Public Opinion Quarterly, 99, 349.-356.
LaTour, Stephen A. and Peat, Nancy C. (1979), "Conceptual and .Methodological Issues in Satisfaction Research," in Advances in Consumer Research, Vol. 6, W. L. Wilkie, ed., Ann Arbor: Association for Consumer Research, 431-37.
Nunnally, Jum C. (1978). Psychometric Theory, Second Ed, New York: McGraw-Hill.
Oliver, Richard L. (1977), "Effect of Expectation and Disconfirmation on Post-exposure Product Evaluations: An Alternative Interpretation," Journal of Applied Psychology, 62, 480-486.
Oliver, Richard L. (1980), "A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions," Journal or Marketing Research, 17, 460-469.
Oliver, Richard L. (1981), "Measurement and Evaluation of Satisfaction Processes in Retail Settings," Journal of Retailing, 57, 95-48.
Oliver, Richard L. and Linda, Gerald (1981), "Effect of Satisfaction and Its Antecedents on Consumer Preference and Intention," in Advances in Consumer Research, Vol. 8, Kent B. Monroe, ed. Ann Arbor, Mich.: Association for Consumer Research, 88-93.
Ray, Michael L. (1979), "Involvement and Other Variables Mediating Communication Effects as Opposed to Explaining All Consumer Behavior," in Advances in Consumer Research, Vol. 6, W.L. Wilkie, ed., Ann Arbor: Association for Consumer Research, 197-199.
Robertson, Thomas S. (1976), "Low Commitment Consumer Behavior," Journal or Advertising Research, 16, 19-94.
Rothschild, Michael L. (1978), "Political Advertising: A Neglected Policy Issue in Marketing," Journal of Marketing Research, 15, 58-71.
Rothschild, Michael L. (1979), Marketing Communications in Nonbusiness Situations," Journal of Marketing, 43, 11-90.
Sherif, Muzafer, and Cantril, Hadley, (1947), The Psychology of Ego-Involvement, New York: Wiley.
Swan, John E. (1977), "Consumer Satisfaction with a Retail Score Related to the Fulfillment of Expectations on an Initial Shopping Trip," in Consumer Satisfaction, Dissatisfaction, and Complaining Behavior, Ralph L. Day, ed., Bloomington: Indiana University, School of Business, 10-17.
Swan, John E., and Trawick, I. Fredrick (1979), "Testing an Extended Concept or Consumer Satisfaction," in New Dimensions or Consumer Satisfaction and Complaining Behavior, Ralph L. Day and Keith Hunt, eds. Bloomington: Indiana University, School of Business, Division of Research, 56-61.
Swan, John E. and Trawick, I. Fredrick (1981), "Disconfirmation of Expectations and Satisfaction with a Retail Service," Journal of Retailing, 57, 49-67.
Weaver, Donald and Brickman, Philip (1974), "Expectancy, Feedback and Disconfirmation as Independent Factors in Outcome Satisfaction," Journal of Personality and Social Psychology, 30, 420-428.
Westbrook, Robert A., and Oliver, Richard L. (1981), "Developing Better Measures of Consumer Satisfaction: Some Preliminary Results," in Advances in Consumer Research, Vol. 8, Kent B. Monroe, ed ., Ann Arbor, Mich.: Association for Consumer Research, 94-99.