Mediators of Message Sidedness Effects on Cognitive Structure For Involved and Uninvolved Audiences

Manoj Hastak, The American University
Jong-Won Park, University of Illinois
ABSTRACT - Persuasive effects due to one versus two-sided messages were examined within the framework of the Elaboration Likelihood Model (Petty and Cacioppo 1986). It was proposed that message sidedness effects on brand attitude are mediated by brand-related cognitive responses for involved audiences, and by perceptions of message/source believability for uninvolved audiences. These propositions were tested in a laboratory experiment involving a 3 (high involvement, low involvement, control) by 2 (one versus two-sided message) design. Results suggest, as-expected, that involved subjects are more sensitive to manipulation of message structure. However, no effects due to message sidedness on brand beliefs or attitude were obtained for either involved or uninvolved subjects. The paper concludes with a discussion of the implications of these results for future research on message sidedness effects, and on involvement.
[ to cite ]:
Manoj Hastak and Jong-Won Park (1990) ,"Mediators of Message Sidedness Effects on Cognitive Structure For Involved and Uninvolved Audiences", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 329-336.

Advances in Consumer Research Volume 17, 1990      Pages 329-336

MEDIATORS OF MESSAGE SIDEDNESS EFFECTS ON COGNITIVE STRUCTURE FOR INVOLVED AND UNINVOLVED AUDIENCES

Manoj Hastak, The American University

Jong-Won Park, University of Illinois

ABSTRACT -

Persuasive effects due to one versus two-sided messages were examined within the framework of the Elaboration Likelihood Model (Petty and Cacioppo 1986). It was proposed that message sidedness effects on brand attitude are mediated by brand-related cognitive responses for involved audiences, and by perceptions of message/source believability for uninvolved audiences. These propositions were tested in a laboratory experiment involving a 3 (high involvement, low involvement, control) by 2 (one versus two-sided message) design. Results suggest, as-expected, that involved subjects are more sensitive to manipulation of message structure. However, no effects due to message sidedness on brand beliefs or attitude were obtained for either involved or uninvolved subjects. The paper concludes with a discussion of the implications of these results for future research on message sidedness effects, and on involvement.

INTRODUCTION

Consumer and advertising researchers have for long been interested in the attitudinal effects of one versus two-sided ad message appeals. A one-sided appeal is characterized by favorable claims on all brand attributes mentioned in the message, whereas a two-sided appeal contains disclaimers or admissions of inferior performance on one or more relatively unimportant attributes. Empirical research in this area has been driven by two theoretical frameworks that focus on different mediating mechanisms in predicting more favorable attitudinal effects due to a two-sided message. Inoculation theory (McGuire 1961) suggests that two-sided appeals are likely to be more effective because presenting negative information about a brand serves to inhibit counter-argumentation. Attribution theory (e.g., Jones and Davis 1965) suggests that disclaimers used in a two-sided message serve to enhance the credibility of the message and/or of the message endorser, and thus induce a more favorable attitude. Detailed reviews of these frameworks, and there implications for the study of message sidedness effects are available elsewhere (see Kamins and Assael 1987 for a recent review).

In this paper, we focus on the role of perceived message/source believability, and spontaneous cognitive responses generated in response to the ad message as alternative mediators of message sidedness effects on attitude. Although several empirical studies in the consumer behavior literature have examined message sidedness effects (Belch 1981; Settle and Golden 1974; Golden and Alpert 1978; Hunt, Domzal, and Kernan 1982; Swinyard 1981; Smith and Hunt 1978; Edgar and Goodwin 1982, Kamins and Assael 1987), evidence for mediation due to either cognitive responses or message/source believability is very limited. For instance, two studies have shown that two-sided appeals inhibit counter-argumentation and/or enhance support argumentation (Kamins and Assael 1987; Swinyard 1981). Unfortunately, measures of brand attitude were not obtained in these studies, so no tests of mediation could be conducted. Similarly, three of five studies that obtained significant effects due to message sidedness on source/message believability did not measure brand attitude (Kamins and Assael 1987; Swinyard 1981; Smith and Hunt 1978), while the two remaining studies obtained n effects on attitude (Golden and Alpert 1978; Kamins and Marks 1988).

One goal of the present study was simply to create the conditions under which mediation effects due to both mediators (cognitive responses and message/source believability) on attitude could be examined. A second, more important goal, was to investigate the conditions under which one of these mediators would likely dominate the other. The extant literature on message sidedness provides virtually no guidelines for understanding this issue We believe that the Elaboration Likelihood Model persuasion processes proposed by Petty and Cacioppo (1979, 1986) provides a suitable conceptual framework within which multiple mediators of message sidedness effects (such as perceived message believability, and message induced cognitive responses) can be examined. In the following sections, we first present a brief overview of the ELM model. We then derive testable implications of the model for mediation processes for message sidedness effects for involved and uninvolved audiences. Finally, we report on a laboratory experiment designed to test these implications.

THEORETICAL FRAMEWORK

The Elaboration Likelihood Model

Extensive reviews of the ELM are available elsewhere (Petty and Cacioppo 1986). Briefly, this framework suggests that the way in which recipients process a persuasive message depends on their motivation and ability to process the informational claims in the message. Individuals who are high on motivation and ability (e.g., involved and knowledgeable subjects) adopt a "central" route to persuasion. Specifically, they engage in detailed, elaborate processing of message claims en route to forming a carefully reasoned opinion about the advocacy issue. By contrast, unmotivated or unable individuals follow a "peripheral" route to persuasion. These individuals judge the advocacy issue based on a superficial assessment of readily available cues in the message. Attitudes may be formed or changed via peripheral processing either because the communication object is associated with positive or negative cues (e.g., an attractive source), or because the individual can rely on a simple heuristic to make a quick evaluative inference about the object based on cues in the persuasive message (e.g., a product endorsed by an expert must be good). These ELM predictions have been tested in several psychology and marketing studies, and have received good empirical support (Petty and Cacioppo 1979; Petty, Cacioppo, and Goldman 1981; Petty Cacioppo and Schumann 1983). Also, several recently proposed models of advertising effects have suggested processing differences between high and low involvement audiences similar to the ELM (Batra and Ray 1985; Burnkrant and Sawyer 1983; Greenwald and Leavitt 1984: Mitchell 1986).

Message sidedness effects on involved audiences.

ELM implicates processing of message arguments, and hence the cognitive responses generated during such processing as the key mediators of message effects on attitude for involved audiences.- Thus, a two-sided appeal will induce more favorable attitudinal effects to the extent it is successful in biasing the evaluative tone of message processing by either suppressing counter-arguments and/or enhancing support arguments.

It is important to note here that the ELM does not predict that two-sided ad messages will result in a more favorable brand attitude than one-sided messages. All that this model asserts is that attitudinal advantage for a two-sided message (if any) will be mediated by message-related cognitive responses. However, there are several reasons why two-sided messages may produce favorable effects on attitude and cognitive responses. First, since a two-sided message explicitly admits to inferior performance on disclaimed attributes, it is likely to inhibit the recipient from generating counter-arguments on those attributes. Also, these disclaimers may enhance the believability of other claims made in the message, and hence suppress counter-argumentation and enhance support argumentation targeted at these claims. This logic suggests that effects of message sidedness on attitude and on beliefs for attributes not disclaimed in the message will be mediated by parallel effects on the evaluative focus of message processing as indicated by degree of counter and/or support argumentation.

Message sidedness effects on uninvolved audiences

For uninvolved audiences to be influenced by a two-sided message, they must first recognize the "two-sided" nature of the message, and then perceive this cue as being relevant to generating an overall evaluation of the advertised brand.

Since the disclaimed attributes are embedded in the verbal message of a two-sided communication, a distinct possibility is that uninvolved audiences may not even recognize the two-sided nature of the communication. This suggests that a two-sided message may frequently-have no advantage over a one-sided message in influencing brand attitude. If the two-sided nature of the message is recognized, then a two-sided message will induce a more favorable attitude to the extent that subjects rely on message sidedness as a cue to inferring source/message believability.

Note, again, that the ELM does not directly predict an attitudinal advantage for two-sided messages. It simply suggests that attitudinal effects due to message sidedness (if any) for uninvolved audiences will be mediated by a peripheral cue such as perceived message/source believability. The logic for expecting sidedness effects on believability has been clearly explicated in the attribution theory literature as applied to message sidedness effects. (see Kamins and Assael 1987 for a review). This reasoning suggests that effects of message sidedness on attitude will be mediated by perceptions of source credibility (if a source is explicitly identified) and/or message believability. Importantly, ELM predicts no effects due to message sidedness on cognitive responses (counter and support arguments). Also effects on beliefs about attributes not disclaimed (if any) should be due to halo effects, and should not be mediated by cognitive responses.

Summary

The ELM framework suggests that audience involvement with the advertised brand should influence the degree to which they recognize the message sidedness manipulation as well as the mediators of sidedness effects on attitude. Since involved audiences are expected to carefully process brand message claims in the ad, they should clearly recognize that some attributes are disclaimed in the two-sided message, and hence form distinctly inferior beliefs about these attributes. By contrast, uninvolved audiences should show much weaker effects of message sidedness on beliefs about disclaimed attributes since they are expected to process the ad in a-cursory and superficial manner. Also, significant effects due to message sidedness on attitude (if any) should be mediated by brand-related cognitive responses (counter and/or support arguments) for involved audiences, and by perceptions of message/source believability for uninvolved audiences. Finally, message sidedness should produce no effects on brand-related cognitive responses for uninvolved audiences.

METHOD

Subjects and Design

130 undergraduate student subjects participated in the study in exchange for course credit. Of these, 124 provided usable responses. The design was a 3 (high involvement, low involvement, control) by 2 (one vs. two sided message) between subjects factorial. Subjects were run in groups of 12 to 15. Each session contained subjects from all six treatment conditions.

Procedure

Upon arrival, subjects were asked to read a one page description of the study. Subjects were told that they would be asked to look through a booklet containing black and while copies of several print ads. Subjects were also told that other participants would be exposed to full color versions of these ads, and that the purpose of the study was to examine the effects of color on consumer reactions to ads. This statement purpose was given for two reasons. First, it was necessary to give subjects some justification for why they were being shown black and white copies of ads. Second, we wanted to hold constant subjects' processing objectives so that natural variations in these would not weaken our involvement manipulation or mediation tests (see Hastak and Olson 1989; Wright 1980).

All subjects were exposed to a booklet containing seven black and white ads. The fifth ad in the booklet was the target ad (for a fictitious brand of ball point pen). Each ad in the booklet was preceded by a brief written description, which subjects were instructed to read carefully. After all subjects had finished examining the booklet, they were given a series of questionnaires to complete. Finally, subjects were debriefed and thanked for their participation.

Independent Variables

Involvement. Our manipulation of brand involvement closely paralleled the manipulation used by Petty, Cacioppo, and Schumann (1983). This manipulation was embedded in two places in the ad booklet. First, the one page study description for high (low) involvement subjects stated that at the end of the study they would be given an opportunity to choose one among several brands of ball point pen (pain reliever). A real ad for an unfamiliar pain reliever brand that was not locally available was included in the third position in the ad booklet. Subjects in the control condition were told nothing.

Second, the brief description that preceded the ball point pen ad for high involvement subjects stated that the pen was to be test marketed in the Champaign-Urbana area, and hence would soon be locally available for purchase. Subjects assigned to the low involvement condition were told that the pen was to be test marketed exclusively on the west coast. Also, the description for the pain reliever for low involvement subjects stated that the pain reliever brand would soon be locally introduced, while high involvement subject were told that there were no plans to introduce this brand in the local area. Control subjects were told nothing about the availability of either the pen or the pain reliever brands.

Message sidedness. The pen ad showed a picture of the pen, a photograph of a male college student, and a message that was attributed to the student. In the two-sided version, two relatively unimportant attribute (styling and color variety) were disclaimed in the opening paragraph of the message:

Quite frankly, the Spree pen doesn't make a particularly good first impression--a rather ordinary plastic barrel that is not uniquely styled, and a limited assortment of colors to choose from. This impression changes, however, when one starts writing with the Spree.

In contrast, the one-sided message opened with the following paragraph:

The Spree pen makes a great impression right from the start--a distinctive metal barrel that reflects a unique styling in its timeless simplicity of line, and a wide assortment of colors to choose from. And this impression is further strengthened when one starts writing with the Spree.

Both messages then presented the following positive information about the pen:

This is truly one of the most consistent, smooth writing pens I have experienced in its price class. They tell me it is the tough tungsten carbide ball and the patented free flowing ink that does the trick. Whatever the reasons, the results are truly remarkable. And virtually smudge proof.

If smooth, skip free, virtually effortless writing is what you are looking for, you owe it to yourself to try the new Spree pen. You will be glad you did. I certainly am.

Dependent Measures

Cognitive responses. After subjects had viewed the ad booklet, they were asked to proceed through a series of booklets which contained the dependent measures. The first booklet asked subjects to list all the thoughts, reactions, and feelings they had while looking at the pen ad. Subjects were given three minutes to list their thoughts, but were told to stop writing whenever they had no more thoughts to report. These thoughts were subsequently classified in terms of their target (brand message, source, ad execution, other) and polarity (favorable, unfavorable, neutral).

Brand, ad, and source evaluation. These variables were all measured on 9 point rating scales. Brand attitude was measured on three scales (bad-good, unsatisfactory-satisfactory, dislike-like), purchase intention on a single scale (definitely not-definitely yes) attitude towards the ad on three scales (bad-good, dislike-like, unattractive-attractive), message believability on a single scale (unbelievable-believable), message quality on two scales (unpersuasive-persuasive, weak-strong), source credibility on three scales (unbelievable-believable, untrustworthy-trustworthy, unconvincing-convincing), and source attractiveness on two scales (unlikable -likable, unattractive- attractive). Constructs that were measured on multiple scales were operationalized as the average of those scales.

Attribute beliefs. Beliefs about 12 attributes of the Spree pen were measured on 9 point rating scales (strongly disagree- strongly agree). Six of these attributes were claimed in the ad (variety of colors, unique styled barrel, writes smoothly, smudge proof, comfortable grip, skip free), while the other six were not (even thickness in writing, does not blotch, inexpensive, durable, does not dry out, leak proof). These latter belief measures were included to explore differences (if any) between involved and uninvolved subjects in terms of the effects of message sidedness on inferential belief formation.

RESULTS

Manipulation Checks

Involvement. Three measures intended to assess the success of this manipulation were obtained at the end of the study. Two of these paralleled measures used in the ELM literature to measure brand involvement, or personal relevance of the advertised brand (Petty, Cacioppo, and Goldman 1981; Petty, Cacioppo, and Schumann 1983). First, subjects were asked to recall the product category in which they were to be asked to choose one of several brands. As expected, 40 of 44 subjects in the high involvement condition correctly identified the ball point pen category, while 34 of 42 subjects in the low involvement condition correctly identified the pain reliever category. Importantly, only one subject in the low involvement condition incorrectly identified the pen category.

Second, subjects were asked to indicate how likely they personally thought it was that the Spree pen would be locally available in the next few months on a 9 point (very unlikely- very likely) scale. An involvement (3) by message sidedness (2) ANOVA revealed a significant main effect on this measure (F(2,119)=10 2, p<.01). As expected, an a priori contrast revealed that involved subjects believed that it was much more likely that the pen would be available locally (mean=5.82) than did uninvolved subjects(mean=3.62, F(1,82)=17.15, p<.01). The belief for control subjects on this measure (5.26) was marginally bellow the mean for involved subjects.

These results suggest that our manipulation successfully influenced perceived personal relevance of the pen brand. However, it could be argued that these results simply represent instructional checks (i.e, checks to confirm that subjects remembered orienting instructions) rather than evidence for successful manipulation of brand involvement. Furthermore, since the intent of the brand involvement manipulation in the ELM framework is to create differences in message response involvement (i.e., degree of brand message elaboration, see Batra and Ray 1985), we also examined the effects of this manipulation on reported levels of message response involvement, or MRI. Message response involvement (MRI) was operationalized as the average of three 9 point scales designed to indicate how carefully subjects actually examined and processed the pen message (very uninvolved- very involved, concentrating very little- very hard, paying little- a lot of attention). A 3 by 2 (involvement by message sidedness) ANOVA on this measure revealed no main effect due to the involvement factor. The mean MRI levels for the high and low involvement groups were virtually identical (mean MRI=S.32 for high involvement, 5.34 for low involvement, 4.80 for control subjects).

In sum, these results suggest that variations in perceived personal relevance induced by our manipulation did not produce the anticipated parallel effects on brand message processing. To deal with this problem, the analyses were conducted in three ways. First, the manipulated levels of involvement (high, low, and control) were maintained in all analyses. Next, a median split on the MRI measure was used to create two groups that differed significantly on stated levels of MRI, and all analyses were repeated using this blocking factor. Finally, a median split on the number of reported brand-related thoughts (i.e., counter + support arguments) was used to create two group of that differed significantly on amount of processing related to the brand message. This final approach is discussed in Batra and Ray (1985) and has been employed in past research (e.g., see Muehling and Laczniak 1988).

The three sets of analyses produced similar results and identical conclusions. To conserve space, we report only the analyses involving the MRI blocking variable. As a result of the median split, 64 subjects were assigned to the high MRI condition (mean MRI=6.50) and 60 subjects were assigned to the low MRI condition (mean MRI=3.74). These groups differed significantly on reported MRI (F(1, 123)=204.73, p<.01).

Message sidedness. Effectiveness of this manipulation was checked by examining beliefs on the two attributes (variety of colors, and uniquely styled barrel) that were disclaimed in the two-sided message. Means for these beliefs are displayed in Table 1. As expected, subjects exposed to the two-sided message reported a significantly weaker belief in these attributes than did subjects exposed to the one-sided message. Thus, the message sidedness manipulation was successful. Also, the MRI by message sidedness interaction was significant or marginally significant for both beliefs. A-priori contrasts revealed that message sidedness had stronger effects on both beliefs for involved subjects (F(1, 62)= 37.1 for variety of colors, 11.59 for uniquely styled barrel, both p's <.01) compared to uninvolved subjects (F(1, 58)= 6.69, p<.05 for variety of colors, 0.53, p>.4 for uniquely styled barrel). This indicates, as expected, that high MRI subjects processed the ad message more carefully and deeply, and hence were more sensitive to a manipulation embedded in the message claims. Also note that the interaction obtained here parallels the (message quality by involvement) interaction obtained in past ELM research (e.g., Petty, Cacioppo and Goldman 1981; Petty, Cacioppo, and Schumann 1983). Thus The median split on MRI used in this study likely created differences in brand message involvement similar to those obtained in past research.

TABLE 1

CELL MEANS FOR F-RATIOS FOR KEY MEASURES

Effects on Mediators

The ELM suggests that effects due to message sidedness on attitude are mediated by counter and support arguments for involved subjects, and by message/source believability for uninvolved subjects. Table 1 presents results of 2 (high versus low MRI) by 2 (one versus two-sided message) ANOVAs on measures of counter and support argumentation, and message quality, message believability, and source credibility. As is evident, message sidedness had no effects on any of these variables, and the two-way (MRI by message sidedness) interaction was also not significant in each case. Also, no effects were found on measures of the proportion of counter and support arguments, or on the number/proportion of positive and negative ad and source-related thoughts produced by subjects -- these are not reported to conserve space. MRI did produce a main effect on perceptions of message quality and source credibility suggesting that involved subjects judged the ad message to be stronger, and the endorser to be more credible than uninvolved subjects. These results suggest that the message sidedness manipulation had no effects on any of the potential mediators of effects on attitude.

Effects on Belief and Attitude Measures

Results for these measures paralleled the results obtained for the mediators (see Table 1). The MRI variable produced main effects on the belief and attitude measures such that these were more positive for involved subjects. Message sidedness produced no significant (main or interactive) effects on any of the belief measures or on brand attitude. A similar pattern of results was obtained for the six beliefs on attributes not claimed in the message (i.e., inferential beliefs) -- these are not reported to conserve space.

DISCUSSION

Summary of Results

Our results suggest that a two sided message does not produce more favorable effects on brand-related cognitive structure than does a one sided message. Although all subjects clearly recognized the message sidedness cue, and formed distinctly inferior beliefs on attributes disclaimed in the two-sided message, message sidedness had no effects on any of the brand belief and attitude measures, nor did it influence measures of message processing and message/source believability. Note that this failure to find significant effects cannot be attributed to potential problems with our manipulation of involvement, since control subjects showed an identical pattern of results.

Although disappointing, these results are not surprising in light of past research in this area. In the recent marketing literature, only one study showed clear effects due to message sidedness on brand evaluation (Etgar and Goodwin 1982), while several studies failed to show a significant attitudinal advantage due to a two-sided message (Belch 1981; Golden and Alpert 1978; Kamins and Marks 1988). These results (in conjunction with ours) seem to suggest that predicted advantages for a two-sided message are relatively weak and practically inconsequential. This conclusion seems particularly robust in light of the fact that studies in this area differ quite a bit in terms of contexts in which subjects are exposed to the ad messages, nature of the two-sided message, message modality, etc.

Our results also suggest that variations in involvement with the brand or the brand message do not alter the effects due to message sidedness. Clearly, no analyses for differential mediators of message sidedness effects under high versus low involvement could be conducted because there were no effects to be mediated. It is conceivable that weak recognition of the message sidedness cue among uninvolved subjects produced the pattern of results obtained for this group. What is surprising is that involved subjects, who strongly responded to the message sidedness manipulation, showed no effects on cognitive response or attitude measures. A possible explanation for this result may reside in the mix of counter and support arguments produced by our subjects in response to the pen ad. Past marketing research indicates that significant effects due to message sidedness on cognitive responses are usually obtained when counter-argumentation is the dominant mode of message processing (e.g., Kamins and Assael 1987). Also, Inoculation theory (McGuire 1961) suggests that two-sided messages are most effective when audiences are initially negatively predisposed towards the advocacy issue, and hence likely to predominantly counter-argue with the message. Our subjects generated more support than counter arguments in response to the ad, and reported very favorable brand attitudes. Thus, a two-sided message could not have influenced attitudes through suppression of counter-argumentation. Note, however, that message sidedness could still have produced attitudinal effects through enhancement of support argumentation, although this did not happen in the present study.

Manipulation and Measurement of Involvement

We manipulated brand involvement using procedures similar to those that have been successfully used in past ELM research. Also, two of the manipulation check measures we employed were identical to measures used by Petty, Cacioppo, and Goldman (1981,; Petty, Cacioppo, and Schumann 1983) in their research. Interestingly, the manipulation produced significant effects on these measures, but not on a different measure of MRI. Also, attempts to examine involvement effect by using different theoretically justifiable- blocking variables yielded similar results. Finally, the significant interaction effect due to MRI and message sidedness on beliefs about disclaimed attributes seems to suggest that The MRI blocking variable adjusted levels of involvement in a manner similar to past ELM studies. Nevertheless, the failure to unambiguously validate the involvement manipulation suggests that our results concerning involvement effects should be cautiously interpreted.

Although not specifically designed to do so, we believe the present study provides some interesting data on the manipulation and measurement of involvement. For instance, three sets of measures obtained in this study can be justified as measuring or partly measuring brand and/or brand message involvement. First, subject perceptions of the likelihood that the advertised brand would be locally available have been used in past ELM research as a measure of personal relevance of the brand. Second, self report measures similar to the ones we used have been used by other researchers (e.g., Laczniak, Muehling and Grossbart 1989) to measure MRI, or degree of brand message elaboration. Finally, the number of message CRs (i.e., CA + SA) has been suggested as an alternative measure of MRI (Batra and Ray 1985; Muehling and Laczniak 1988).

Table 2 shows the intercorrelations we obtained among these three sets of "involvement" measures. We find the lack of association between these measures surprising. The lack of association between the self report and cognitive response measures of message response involvement is particularly difficult to explain. Overall, our results suggest that a manipulation of involvement based on personal relevance may not produce parallel effects on self report measures of MRI, or that blocking on MRI may not create differences in degree of brand message processing as indicated by cognitive responses. Thus, future researchers should consider employing multiple maximally different measures of "involvement" (as we did) in order to provide compelling support for the successful manipulation of this variable. This approach seems particularly useful given the fuzzy distinctions between concepts such as product class involvement, brand involvement, personal relevance of brand, and brand message response involvement.

TABLE 2

CORRELATIONS AMONG VARIOUS INVOLVEMENT MEASURES

Finally, we believe that implications derived from the ELM for message sidedness effects are useful because they reconcile the role of two competing mediators that have heretofore been independently theorized about and examined in the literature. Future research incorporating a stronger and more convincing manipulation of involvement should prove useful in developing more unambiguous tests of ELM based predictions than were achieved in the present study.

REFERENCES

Batra, R. and M. L. Ray (1985), "How Advertising Works at Contact," In L. Alwitt and A. Mitchell (Eds.), Psychological processes and Advertising Effects: Theory, Research, and Applications, Hillsdale, NJ: Lawrence Erlbaum, 13-43.

Belch, George E (1981), "An Examination of Comparative and Noncomparative Television Commercials: The Effects of Claim Variation and Repetition on Cognitive Response and Message Acceptance," Journal of Marketing Research, 18 (August), 333-349.

Burnkrant, R. E. and A. G. Sawyer (1983), "Effects of Involvement and Message Content on Information-Processing Intensity," in Information Processing Research in Advertising, R. J. Harris (ed.), Hillsdale, NJ: Lawrence Erlbaum Associates, 43-64.

Etgar, Michael and Stephen A. Goodwin (1982), "One-Sided Versus Two-Sided Message Appeals for New Brand Introductions," Journal of Consumer Research, 8 (March), 460-465.

Golden, Linda and Mark Alpert (1978), "The Relative Effectiveness of One-Sided and Two Sided Communication for Mass Transit Advertising," in Advances in Consumer Research, 5, H. Keith Hunt ed., Ann Arbor, MI: Association for Consumer Research, 12-18.

Greenwald, A. and C. Leavitt (1984), "Audience Involvement in Advertising: Four Levels," Journal of Consumer Research, 11, 581-592.

Hastak, Manoj and Jerry C. Olson (1989), "Assessing the Role of Brand-Related Cognitive Responses as Mediators of Communication Effects on Cognitive Structure," Journal of Consumer Research, 15 (March), 444-456.

Hunt, James M., Teresa J. Domzal, and Jerome B. Kernan (1982)," Causal Attributions and Persuasion: The Case of Disconfirmed Expectancies," in Advances in Consumer Research, Vol. 9, Andrew Mitchell, ed., 287-292.

Jones, Edward E. and Keith E. Davis (1965), "FromActs to Dispositions," in Advances in Experimental Social Psychology, 2, L. Berkowitz, ed., New York: Academic Press, Inc., 219-266.

Kamins, Michael A. and Henry Assael (1987), 'Two Sided Versus One-Sided Appeals: A Cognitive Perspective on Argumentation, Source Derogation, and the Effects of Disconfirming Trial on Belief Change," Journal of Marketing Research, 23 (March), 29-39.

Laczniak, Russel N., Darrel D. Muehling, and Sanford Grossbart (1989), "Manipulating Message Involvement in Advertising Research," Journal of Advertising, 18 (2), 28-38.

McGuire, William J. (1961), "The Effectiveness of Supportive and Refutational Defenses in Immunizing and Restoring Beliefs Against Persuasion," Sociometry, 24 (June), 184-197.

Mitchell, Andrew A. (1986), "Theoretical and Methodological Issues in Developing an Individual-Level Model of Advertising Effects," in Advertising and Consumer Psychology, J.C. Olson and K. Sentis, eds., 3, 172-196.

Muehling, Darrel D. and Russel N. Lacznick (1988), "Advertising's Immediate and Delayed Influence on Brand Attitudes: Considerations Across Message-Involvement Levels," Journal of Advertising, 17 (4), 23-34.

Petty, R. and J. Cacioppo (1979), "Issue-Involvement Can Increase or Decrease Persuasion by Enhancing Message-Relevant Cognitive Responses," Journal of Personality and Social Psychology, v.37, 1915-1926.

Petty, R. E. and John T. Cacioppo (1986), 'The Elaboration Likelihood Model of Persuasion," in Advances in Experimental Social Psychology, L. Berkowitz, ed., 19, Orlando, FL: Academic Press, 123 -205.

Petty, R. E., John T. Cacioppo, and R. Goldman (1981), "Personal Involvement as a Determinant of Argument-Based Persuasion," Journal of Personality and Social Psychology, 41, 847-855.

Petty, R. E., John T. Cacioppo, and D. Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10, 134-148.

Settle, Robert B. and Linda L. Golden (1974), "Attribution Theory and Advertising Credibility," Journal of Marketing Research, 11 (May), 181185.

Smith, Robert E and Shelby D. Hunt (1978), "Attributional Processes and Effects in Promotional Situations," Journal of Consumer Research, 5 (December), 149-158.

Swinyard, William R. (1981), 'The Interaction Between comparative Advertising and Copy Claim Variation," Journal of Marketing Research, 18 (May), 175-186.

Wright, Peter L. (1980), "Message-Evoked Thoughts: Persuasion Research Using Thought Verbalization," Journal of Consumer Research, 7, 151 -175.

----------------------------------------