The Effects of Expectancy Disconfirmation and Argument Strength on Message Processing Level: an Application to Personal Selling

James M. Hunt, Temple University
Michael F. Smith, Temple University
Jerome B. Kernan, University of Cincinnati
ABSTRACT - Message processing levels (central versus peripheral) were assessed as a function of expectancy confirmation/disconfirmation (all salient product features superior/one salient product feature not superior) and argument strength (strong = arguments stated in terms of tangible product features; weak = intangible feature arguments ) in an experimental setting couched in a personal selling context involving a home computer. Results--measured by message beliefs, overaLl affect toward product, behavioral intention, causal attributions, and assessments of the spokesperson--indicate strong effects (but opposite from chose hypothesized) for expectancy confirmation and virtually no effects for argument strength. Several explanations for these findings are offered.
[ to cite ]:
James M. Hunt, Michael F. Smith, and Jerome B. Kernan (1985) ,"The Effects of Expectancy Disconfirmation and Argument Strength on Message Processing Level: an Application to Personal Selling", in NA - Advances in Consumer Research Volume 12, eds. Elizabeth C. Hirschman and Moris B. Holbrook, Provo, UT : Association for Consumer Research, Pages: 450-454.

Advances in Consumer Research Volume 12, 1985      Pages 450-454


James M. Hunt, Temple University

Michael F. Smith, Temple University

Jerome B. Kernan, University of Cincinnati

[This study was supported by a grant from the Temple University Faculty Senate Research Program.]


Message processing levels (central versus peripheral) were assessed as a function of expectancy confirmation/disconfirmation (all salient product features superior/one salient product feature not superior) and argument strength (strong = arguments stated in terms of tangible product features; weak = intangible feature arguments ) in an experimental setting couched in a personal selling context involving a home computer. Results--measured by message beliefs, overaLl affect toward product, behavioral intention, causal attributions, and assessments of the spokesperson--indicate strong effects (but opposite from chose hypothesized) for expectancy confirmation and virtually no effects for argument strength. Several explanations for these findings are offered.


The ways in which individuals process information from persuasive communications and assess its veridicality hold considerable interest for those in the field of consumer research. In attempting to analyze these processes, researchers typically have followed one of two rather distinct perspectives (Chaiken 1980); Petty and Cacioppo 1981, 1983). The first of these treats message recipients as active processors, who systematically scrutinize the validity of arguments contained in persuasive messages. When acceptance occurs as a result of such cognitive effort, persuasion is said to have taken place through the central route. As noted by Petty and Cacioppo, persuasion induced in this manner is likely to be relatively enduring and predictive of overt behavior. In contrast, a second, core passive, view of persuasion emphasizes what can be called the peripheral route. According to this perspective, message acceptance occurs not so much because recipients have considered the central arguments of a message, but because they make simple inferences about the merits of the advocated position based on various peripheral cues--factors that are external to the message arguments, such as the speaker's attractiveness, background music, etc.

Both perspectives have spawned a considerable amount of research. However, as Petty, Cacioppo and Schumann (1983) note, neither perspective by itself is capable of ordering the diverse results obtained in the field of persuasion research. Thus, they argue for a more general framework--one that is capable of explaining when recipients will instigate active processing of message arguments and when they will become more cursory in their analysis of such communications. The research reported here is an attempt to address this issue through attribution theory (Heider 1458). In particular, this study draws on one concept thought to underlie the instigation of attributional analysis-expectancy disconfirmationCas a means of explaining when message recipients will instigate active cognitive processing of message arguments.

Expectancy disconfirmation refers to the degree to which an event conforms to or deviates from an individual observer's expectations (see Pyszczynski and Greenberg 1981). Events that conform to expectations are thought to require little cognitive effort to interpret, because people can rely on a heuristic form of analysis, using pre-existing knowledge structures, or the causal theories that gave rise to the expectancies originally. In contrast, when events disconfirm expectations, there are no pre existing theories on which to rely. Accordingly, people are more likely to undertake active, or systematic, forms of cognitive analysis in order to interpret or explain disconfirming events.

This line of reasoning meshes nicely with the foregoing assumptions about the processing of persuasive communications. If expectancy disconfirmation leads to a relatively active form of cognitive processing, then persuasive appeals that disconfirm persuasion-related expectancies should receive a more active and thorough scrutiny and produce persuasion through the central route. On the other hand, when persuasive communication conforms to typical persuasion-related expectancies, message recipients should exert less cognitive effort in analyzing, the message. Accordingly, persuasion through the peripheral route would he more likely.

The effects of expectancy disconfirmation on persuasion proceed largely from studies dealing with communicators who take a stand against their own best interest (e.g. Eagly, Wood and Chaiken 1978; Koeske and Crano 1968; McPeek and Edwards 1975; Walster, Aronson and Abrahams 1966). The proposition underpinning these studies is that speakers who willingly, and unexpectedly, take a position against what seems to be their own best interest appear to be more sincere and, thus, more credible to an audience than do those who argue solely in favor of their own interest. The latter speakers generally are thought to appear motivated," or biased, and as having something to Rain. As such, their credibility is discounted and their persuasiveness inhibited.

This line of reasoning also seems to order the results of studies dealing with the effects of one-sided and two- sided communication. In general, these studies show that communicators tend to be more persuasive when they are willing to present arguments for and against the advocated position (e.g. Chu 1967; Jones and Brehm 1970; Lumsdaine and Janis 1953). Presumably, the presentation of arguments on both sides of the issue, or advocated position, conveys an impression of fairness. Thus, speakers who employ two-sided messages are thought to appear less biased and more sincere than speakers who "limit" their arguments solely to those that support their point of view.

In a recent attempt to deal with the persuasive effects of messages that seemingly depart from a speaker's best interest, EaRly, Chaiken and Wood (1981) have proposed an expectancy-disconfirmation hypothesis based on attribution theory. According to this formulation, message recipients form pre-message expectancies regarding the likely position a communicator will advocate. Typically, these expectancies are biased in that a persuasive communication is anticipated to be skewed toward the communicator ' s position or interests. The origin of such expectancies is thought to be the discernible background traits of the communicator and existing pressures of the communication setting. Message acceptance hinges on whether these bias-related expectancies are confirmed by the actual message. If the actual message confirms (agrees with) the biased expectancies, recipients are thought to "explain" the message in terms of the biasing elements. Accordingly, the validity (veridicality) of the message is discounted and the credibilitY of the source questioned. when a message disconfirms bias-related expectations, however, recipients purportedly invoke an augmentation principle they discount the biasing elements and place greater emphasis on the communicator's willingness to report his/her own judgment in explaining the message). As a result, the communicator is judged to be more sincere and the likelihood of message acceptance is enhanced. At least one study (Hunt, Domzal and Kernan 1982) found partial support for the Eagly et al. formulation.

Since the Eagly, Chaiken and Wood paradigm does not speak directly to the issue of message processing level (see Hunt and Kernan 1994) it likely captures only a portion of the process by which expectancy disconfirmation influences persuasion. Not only does expectancy disconfirmation influence persuasion through attributional processes related to source credibility, but it also might affect message acceptance as a result of more detailed processing of message content. Thus, a message that deviates from persuasion-related expectations might be hypothesized to have a two-fold effect on persuasion: (I) it influences ho-- credible the source is judged to be; and (2) it affects the degree to which message argumentation is processed actively.

Such a two-fold effect is similar to the workings of two-sided messages. As in this latter case, however, one can expect disconfirming messages to be effective only under certain conditions -- when the unfavorable/disconfirming information is: (l) placed at the beginning of a message; (2) already familiar to recipients; and (3) subsequently refuted in a compelling manner (Hass and Linder 1972; Sternthal, Phillips and Dholakia 1978). The first condition serves to activate the receiver's cognitive processes immediately; the second precludes the introduction of surprise elements into the recipient's counter-arguments repertoire; and the third is necessary lest (under the increased message scrutiny induced by expectancy disconfirmation) the unfavorable information remain unchallenged.

The criticality of this last condition cannot be overemphasized. If the message arguments that are intended to refute the disconfirming information do not do so, the overall persuasiveness of the presentation is likely to suffer. Thus, "weak" arguments are far more damaging to the efficacy of a presentation when a receiver's expectations are disconfirmed (because then they play a pivotal role) than when they are confirmed (since then they are no more important than other arguments).

Evidence of this effect is reported by Hunt and Kernan (1984), who found that subjects exposed to two-sided advertising appeals formed credibility judgments largely in terms of unverifiable, or intangible, message arguments--something not found when subjects were exposed to one-sided claims. It was reasoned that, since unverifiable claims and intangible product features tend to he judged idiosyncratically (see Hirschman 1981), positive claims regarding such issues are inferred as disingenuous and presumptuous; thus, they constitute relatively weak arguments compared with claims on more tangible or palpable features or issues. If this is so, then a source's credibility will suffer to the extend that s/he makes positive claims about a product's performance on intangible attributes. Moreover, this effect should be heightened in the case of expectancy disconfirmation (or two-sided messages) since induced cognitive effort is greater and message processing deeper. In the case of one-sided (or expectancy-confirmed) messages, product claims of an intangible nature should have a minimal effect on source credibility since message processing is more likely to occur in a heuristic or automatic fashion. In essence, the disingenuous product claims go undetected.

In summary, expectancy disconfirmation is hypothesized to be a moderating variable influencing when message recipients will engage in active and elaborate message processing. Further, this moderating effect is thought to either facilitate or hinder persuasion, depending on the strength of the arguments used in behalf of the advocated position or product. The goals of the present study were two: (1) to investigate the operation of the Eagly, Chaiken and Wood paradigm in a promotional context; and (2) to assess the effects of expectancy disconfirmation on message recipients' cognitive workload. Manipulation of expectancy disconfirmation and argument strength was carried out by exposing subjects to differing versions of a sales presentation for a home computer. Expectancy confirmation/disconfirmation was operationalized by exposing subjects to a presentation that contained "superior" product performance claims only or to one that disclaimed superiority on one of the performance attributes. Argument strength was manipulated through the tangibility of product claims; the strong-argument treatment contained only tangible (verifiable) claims, while the weak-argument treatment contained one intangible performance claim. Based on these manipulations, it was hypothesized that argument strength would have a greater influence on subjects' message acceptance when the sales presentation disconfirmed, rather than confirmed, subjects' bias-related expectancies.



A total of 174 (female and male) undergraduate students was selected from the student body of a large eastern university to participate in a study designed to evaluate various sales presentations. These volunteer subjects were paid $3 for their participation. Subjects' ages ranged from 18 to 55, with a mean of 22 years. Most were either business-administration majors (88%) or students majoring in communication (9%). Each subject was assigned at random to one of five groups: an expectancy group, or one of four treatment groups that formed a 2 (expectancy confirmed or disconfirmed) x 2 (tangible or intangible arguments) factorial design. The expectancy group was used to establish subjects' pre-message expectancies regarding the sales presentation.

Procedure and Stimulus Materials

[Owing to space limitations, this section is condensed. A complete packet of experimental materials is available from the authors.]

Upon reporting for the experiment, subjects were informed by an experimenter that they would be participating in the evaluation of a sales presentation for a home computer designed for student use. Their participation would entail: (1) reading a transcript of an actual sales presentation; and (2) responding to questions designed to ascertain their evaluations of both the presentation and the person who made it. Nominally, the presentation was one among many made by college applicants in the course of interviewing for sales positions with a local computer firm. Supposedly, this presentation--required of all applicants--had been made to a group of students interested in home computers. The experimenter informed subjects that the researchers conducting the study were currently working with the computer firm to identify qualified sales applicants and that subjects' evaluations would aid the researchers in this process.

After receiving initial instructions, each treatment subject was given a booklet containing the experimental materials--the presentation transcript and a questionnaire designed to measure the dependent variables. Expectancy subjects also received pre-message instructions but they did not read an experimental transcript. Instead, they were asked to respond to items designed to assess their pre-message expectancies.

The presentation revolved around four product features previously elicited through a focus group interview: (1) the computer's hardware--described as being extremely powerful and reliable, backed by many years of experience in the computer industry, and easily expanded and very versatile, with many built-in features that, unlike other computers, were available at no extra cost; (2) the cost of the computer--represented as exceptional, relative to its performance; (3) the computer's software--used to manipulate expectancy confirmation/disconfirmation; and (4) the computer's ease of use--used to manipulate argument strength.

Independent Variables

Manipulation of the expectancy confirmation/disconfirmation factor was carried out by varying the information subjects received about the computer's software. Half of the subjects read a transcript in which the software was described as being superior, while the other half received a message containing less-than-superior information regarding this feature. Argument strength was manipulated by exposing subjects to different information concerning the computer's ease of use. Some subjects read a description that presented specific, or tangible, arguments on behalf of the computer's easy handling characteristics. Others were exposed to less specific (more intangible) argumentation.

Dependent Measures

Overall, subjects responded to 18 dependent measures, grouped in five categories. There were four message-acceptance beliefs about the computer (that it was a Rood dollar value, was easy to use, had superior hardware, and superior software), each measured on a point aided-recall basis (where 7 = "strongly agree" and 1 = strongly disagree"); an overall affect measure how superior the computer was), assessed on a 7-point scale; an intention measure (whether subjects would "consider" the computer), also on a 7-point scale; the applicant-presenter's true feelings, hope of getting a sales-rep job, and the real facts about the computer in influencing him/her to say what s/he said in the presentation), measured on an extremely important

(7)/"extremely unimportant" (1) scale; and nine source evaluations (honesty, usefulness .s an information source, sincerity, knowledgeability, believability, expertise, trustworthiness, like-ability, and overall impression as a good sales rep), each measured on a point scale where 7 = "strongly agree" and 1 = "strongly disagree."


Certain design requirements must be satisfied before the experimental results can be interpreted properly. First, it is necessary to establish that superior product claims indeed are the normal expectation in a sales presentation. This was verified by the expectancy group, whose 23 members were asked to compare the likelihood that an applicant would describe the computer as superior (a) in a sales presentation and (b) in a private (non-selling) conversation with a friend. These subjects (who did not actually read a sales-presentation transcript, but rather assessed what they expected one would contain) averaged 5.17 (where 7 = "extremely likely" and 1 = "extremely unlikely") for the sales-presentation condition but only 4.52 for the private-conversation one. A repeated-measures t-test (df = 21) yields a value of 41.25, significant at p < .001. Thus, it is reasonable to assume that superior claims were expected in the sales presentations read by treatment subjects.

The other design requirements have to do with the manipulation checks: whether treatment subjects actually perceived the treatment variations that were intended. As noted above, expectancy confirmation/disconfirmation was manipulated by varying the description of the computer's software. In the confirmed condition, it was described as superior; in the disconfirmed condition, it was described as less than-superior. Subjects in the confirmation condition had a mean rating for the computer's software of 5.66, while the comparable statistic for the disconfirmed condition was only 2.68. The difference in these means suggests that this manipulation was effective--F (1,148) = 104.77; p < .001. A similarly effective manipuLation occurred in the case of argument strength. The strong-argument (tangible) condition described the computer's easy-to-use function keys that would draw graphs or construct tables, while the weak-argument (intangible) condition suggested merely that one could master the computer's operation in a short time. Mean recall of the function-key description was 6.23 in the tangible condition and 2.68 in the intangible one--F (1,148) = 142.50; p < .001.

Thus, there is little question that these design requirements were met--subjects expected superior product claims and had these expectations either confirmed or disconfirmed. Moreover, subjects had no difficulty in distinguishing the tangible arguments from the intangible ones. Indeed, based on the experimental results about to be described, one might speculate that some of the foregoing requirements were met too well. As we shall see, the results were remarkably impervious to the hypotheses.

The results are described according to the five sets of dependent measures listed above. Each of the 18 dependent variables was treated as a criterion in a two-way ANOVA, where expectancy confirmation/disconfirmation and message strength (tangible/intangible) served as treatments. Since the resulting data are quite voluminous, only selected findings are reported here. [Detailed results, (e.g. cell means) are available from the authors.]

Message Beliefs

The clearest indication of how centrally, actively, or deeply receivers process information should be manifested in their post-message beliefs. Compared with heuristic or peripheral processers, central processers should attend to these elements more thoroughly and, as a consequence, recall them more veridically. In the present case, since expectancy disconfirmation was hypothesized to induce central processing, subjects in the disconfirmed conditions should exhibit "higher message-belief scores than those in the confirmed conditions. For each of the four beliefs measured-, however, the opposite result occurred: confirmed subjects rated the computer significantly higher on dollar value, hardware, and software (as well as directionally higher on ease-of-use) than did disconfirmed subjects. ["Significant" differences refer to p < .10; i.e., .10, .05, .01, and .001.] The main effect of argument strength was insignificant and there were no significant interactions. Thus, it would appear that the one-sided (confirmation) version of the sales presentation induced more central processing than lid the two-sided (disconfirmation) version.

Affect and Intention

A similar result occurred with the overall affect and intention measures. Confirmation subjects evinced significantly higher agreement that the computer was a superior one and were significantly more likely to consider looking at it than subjects exposed to the disconfirmation treatments. Again, neither argument strength nor interactions were significant.


Subjects' attributions of the influences that "caused" the applicant-presenter to make product claims (as a function of expectancy confirmation/disconfirmation and message strength) are mixed. Contrary to our hypotheses, confirmation of expectations induced significantly greater internal (spokesperson's true feelings) and entity (real facts about the computer) attributions; but consistent with our hypotheses, weak (intangible) arguments induced significantly greater external (hope of getting a sales-rep job) attributions. (There were no significant interactions.) These admittedly perplexing results seem to suggest that the presenter was interpreted as representing reality when s/he made all superior claims about the computer except in the case of intangible arguments, whose disingenuity is explained by the presenter's motivation to get a job Overall, however, these results are not consistent with the Eagly, Chaiken and Wood paradigm.

Source Evaluations

The results that are most inconsistent with the Eagly et al. hypothesis, however, are receivers' evaluations of the spokesperson. Eagly et al. would predict more favorable evaluations under conditions of expectancy disconfirmation yet, in general, our results are in the opposite direction. Confirmation produced significantly higher source ratings on usefulness as an information source, knowledgeability, believability, expertise, liking, and overall impression as a good sales rep. There were no significant differences on the honesty, sincerity, and trustworthiness dimensions. Furthermore, neither argument strength nor interactions were significant. Thus, it appears that disconfirming receivers' expectations did not induce them to evaluate the spokesperson more favorably (and thereby become more susceptible to his/her persuasive arguments--see message belief results. above).


The results of this study are substantially different from what was hypothesized. Given that the Eagly, Chaiken and Wood paradigm is the linchpin of the prediction framework proffered, one conclusion might be that their model has no predictive efficacy. We are inclined to reject that explanation, however; there is too much verification of the paradigm elsewhere--e.g., Hunt, Domzal and Kernan 1982; Hunt and Kernan 1984. Another explanation for these "contrary" findings might be that this study represents an ineffectual operationalization of the Eagly et al. model. Based on the strength of the expectancy-group results and manipulation checks, however, we doubt that conclusion as well. Rather, we suspect that the principal reason for our results is to be found not in the Eagly et al. model per se, but in some limiting condition; that apparently attend it. At least three such conditions warrant comment.

First, the Eagly et al. paradigm supposes that increased persuasion is effected through heightened source evaluation which, in turn, is produced by expectancy disconfirmation. But this assumes that there is no ceiling effect on source evaluation. In the present case, because students evaluated students it is possible that the disconfirming of expectations induced attributions that the spokesperson was inept ("what a stupid thing to do when you' re trying to sell something!") rather than of his/her being a more fair, hence believable person. Support for this possibility can be seen in subjects' significantly different (p < .01) ratings of how good a sales rep the spokesperson vas. In the confirmed conditions, the mean rating was 5.13 (out of a possible 7) while in the disconfirmed conditions, it was only 4.06. Thus, disconfirmation might indeed have induced more central processing, but since it was regarded as so stupid: (a) the spokesperson was rendered less, rather than more credible; and (b) message acceptance was less, rather than more under disconfirmation conditions. This interpretation would explain the empirical results achieved for all five kinds of dependent variables measured in this study.

Second, the potential effects of involvement cannot be ignored. Although no specific measures of involvement were made (the focus grouP analysis suggested that student-oriented home computers were at least moderately involving to subjects), it is altogether plausible that this product (coupled with the "evaluation task") produced a highly involving situation--one so high that, even in the expectancy confirmation treatments, central processing occurred. If so, disconfirmation did not have the opportunity to trigger message (cf. source) processing; the former was preordained as a result of the high involvement. Such an explanation also would account for our findings.

A third limiting condition stems from the work of Hass and Linder (1972), Jones and Brehm (1970), and Sternthal et al. (1978), referred to above. Their work indicates that negative (disconfirming) information is effectual only when it: (1) appears at the beginning of a message; (2) is already familiar to recipients; and (3) is refuted in a compelling manner. The present study satisfied the first of these criteria, but probably neither of the other two. The second criterion obviously was not satisfied because, rather than admit to familiar deficiencies in the computer, the disconfirmation treatments surprised recipients and likely provided them with grist for counterargumentation. and the third criterion is suspect because we have no way of knowing whether recipients interpreted the disconfirming message as containing a compelling refutation of the less-than-superior-software description. Thus the kind of expectancy disconfirmation contemplated by the Eagly et al. paradigm might not have been present in this study. If that is the case, one can hardlY expect the model's predictions to hold.

In summary, there seem to be at least three conditions in the present study that subverted a fair test of the Eagly, Chaiken and Wood paradigm. A ceiling effect on source evaluations, excessively high issue/task involvement, and an incongruous kind of disconfirmation separately or in combination might account for the "failure" of their model to order this study's results--to predict when central or peripheral processing would occur.

Given this overwhelming perturbation, little explanation is required to account for the apparent lack of effect produced by the argument-strength manipulations. As hypothesized, argument strength should emerge under disconfirmation conditions because they produce central, more scrutinous processing. Since this necessary condition did not occur, whatever effect strong/weak (tangible/intangible) arguments had was obscured. The (nonsignificant) directional effects they produced were correct. Hence we are satisfied that Holbrook's (1978) contention regarding the persuasive superiority of factual claims over intangible ones is a sound one.


Chaiken, S. (1980), "Heuristic Versus Systematic Information Processing and the Use of Source Versus Message Cues in Persuasion," Journal of Personality and Social Psychology, 39, 752-66.

Chu, G. C. (1967), "Prior Familiarity, Perceived Bias, and One-Sided Versus Two-Sided Communications," Journal of Experimental Social Psychology, 3, 243-54.

Eagly, A. H., S. Chaiken and W. Wood (1981), "An Attributional Analysis of Persuasion," in New Directions in Attributional Research, J. H. Harey et al. (eds.). Hillsdale, NJ: Lawrence A. Erlbaum Associates.

Eagly, A. H., W. Wood and S. Chaiken (1978), "Causal Inferences About Communicators and Their Effect on Opinion Change," Journal of Personality and Social Psychology, 36, 424-35.

Hass, R. and D. Linder (1972), "Counterargument Availability and the Effects of Message Structure on Persuasion," Journal of Personality and Social Psychology, 23, 219-33.

Heider, F. (1958), The Psychology of Interpersonal Relations, New York: Wiley.

Hirschman, E. C. (1981), "Cognitive Complexity and the Perception of Intangible Attributes," in Educators' Conference Proceedings, K. C. Bernhardt et al. (eds.). Chicago. American Marketing Association.

Holbrook, M. B. (1978), "Beyond Attitude Structure: Toward the Informational Determinants of Attitude," Journal of Marketing Research, 15, 515-56.

Hunt, J. M., T. J. Domzal and J. B. Kernan (1982), "Causal Attributions and Persuasion: The Case of Disconfirmed Expectancies," in Advances in Consumer Research, Vol. 9, A. Mitchell (ed.). Ann Arbor, MI: Association for Consumer Research.

Hunt, J. M. and J. B. Kernan (1986), "The Role of Disconfirmed Expectancies in the Processing of Advertising Messages," Journal of Social Psychology, in press.

Jones, R. A. and J. W. Brehm (1970), 'Persuasiveness of One- and Two-Sided Communications as a Function of Awareness There are Two Sides," Journal of Experimental Social Psychology, 6, 47-56.

Koeske, G. F. and W. D. Crano (1968), "The Effect of Congruous and Incongruous Source Statement Combinations Upon the Judged Credibility of a Communication," Journal of Experimental Social Psychology, 4, 384-99.

Lumsdaine, A. A. and I. L. Janis (1953), Resistance to 'Counterpropaganda' Produced by One-Sided and Two-Sided 'Propaganda' Presentations, Public Opinion Quarterly 17, 311-18.

McPeek, R. W. and J. D. Edwards (1975), Expectancy Disconfirmation and Attitude Change, Journal of Social Psychology, 96, 193-208.

Petty, R. E. and J. T. Cacioppo (1981), Attitudes and Persuasion- Classic and Contemporary Approaches Dubuque, IA: William C. Brown.

Petty, R. E. and J. T. Cacioppo (1983), "Central and Peripheral Routes to Persuasion: Application to Advertising," in Advertising and Consumer Psychology L. Percy and A. Woodside, (eds.). Lexington, MA: Lexington Books.

Petty, R. E., J. T. Cacioppo and D. Schumann (1983), "Central and Peripheral Routes to Advertising effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10, 135-46.

Pyszczynski, T. A. and J. Greenberg (1981), "Role of Disconfirmed Expectancies in the Instigation of Attributional Processing," Journal of Personality and Social Psychology, 40, 31-38.

Sternthal, B., L. W. Phillips and R. Dholakia (1978), "The Persuasive Effect of Source Credibility: A Situational Analysis," Public Opinion Quarterly, 42, 285-314.

Walster, E. E., E. Aronson and D. Abrahams (1966), On Increasing the Persuasiveness of a Low Prestige Communicator," Journal of Experimental Social Psychology, 2, 325-42.