Brand Cognitions As Determinants of Brand Attitudes: the Influence of Measurement and Processing Involvement

Randall L. Rose, University of South Carolina
Paul W. Miniard, Ohio State University
Sunil Bhatla, Case Western University
ABSTRACT - We examined the influence of brand cognitions on the formation of brand attitudes after exposure to an advertisement under conditions of higher or lower involvement. Brand cognitions were measured using both a cognitive response index and an expectancy-value measure. The strength of the brand cognitions - attitude relationship proved to be stable across involvement conditions and relatively insensitive to measurement. Further analyses using groups based on extremely high or low levels of self-reported involvement indicated the expected involvement moderation of the impact of brand cognitions on brand attitudes.
[ to cite ]:
Randall L. Rose, Paul W. Miniard, and Sunil Bhatla (1990) ,"Brand Cognitions As Determinants of Brand Attitudes: the Influence of Measurement and Processing Involvement", 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: 128-134.

Advances in Consumer Research Volume 17, 1990      Pages 128-134

BRAND COGNITIONS AS DETERMINANTS OF BRAND ATTITUDES: THE INFLUENCE OF MEASUREMENT AND PROCESSING INVOLVEMENT

Randall L. Rose, University of South Carolina

Paul W. Miniard, Ohio State University

Sunil Bhatla, Case Western University

ABSTRACT -

We examined the influence of brand cognitions on the formation of brand attitudes after exposure to an advertisement under conditions of higher or lower involvement. Brand cognitions were measured using both a cognitive response index and an expectancy-value measure. The strength of the brand cognitions - attitude relationship proved to be stable across involvement conditions and relatively insensitive to measurement. Further analyses using groups based on extremely high or low levels of self-reported involvement indicated the expected involvement moderation of the impact of brand cognitions on brand attitudes.

INTRODUCTION

The role of brand cognitions in the formation of post-communication brand attitudes has been well documented in research applying expectancy-value attitude theory (e.g., Fishbein and Azjen 1975; Lutz 1975; Mitchell and Olson 1981) and cognitive response models of persuasion (e.g., Greenwald 1968; Wright 1973). More recently, researchers have begun intensive study of affective reactions to advertisements with a focus on the influence of ad attitudes (Aad) on the formation of brand attitudes (Ab) (e.g., Mitchell and Olson 1981; Shimp 1981). One interesting, and somewhat surprising, observation drawn from this stream of research has been that brand cognitions, when operationalized in the form of cognitive responses, often do not affect brand attitudes in models including ad attitude as a second predictor of Ab. Both conceptual and methodological explanations have been offered for the relatively weak causal role played by brand-oriented cognitive responses. This research attempts to shed some further light on these issues.

BACKGROUND

The surprising weakness of the link between brand-oriented cognitive responses (CRb) and brand attitudes in models including Aad was first noted by Lutz, MacKenzie, and Belch (1983) and MacKenzie and Lutz (1983). Similarly, Mackenzie, Lutz, and Belch (1986) reported a non-significant relationship between CRb and Ab in two experiments which supported their model of brand attitude formation. In a study which included affective responses as well as cognitive responses to advertisements, Batra and Ray (1986) also found that CRb did not influence Ab when Aad was included as a copredictor of brand attitudes.

At this point, the only supportive evidence of the CRb - Ab relationship when Aad also serves as a determinant of Ab has been reported by Park and Young (1986). In their study, CRb was related to Ab when the commercial lacked music and was processed under cognitive involvement conditions. However, a change in involvement (i.e., the affective and low involvement conditions) or the presence of music under cognitive involvement reduced the relationship to non-significance.

Mackenzie et al. (1986) proposed a measurement artifact as one explanation for the lack of a significant CRb - Ab relationship. They pointed out that the cognitive response index used in their studies is inferior psychometrically to measures based on expectancy-value models. Moreover, cognitive response measures may be vulnerable to restrictions in range, especially when motivation to process message arguments is low. This line of reasoning suggests that causal relationships involving brand cognitions and attitudes should be stronger when the former is operationalized using an expectancy-value formulation instead of cognitive responses. Consistent with this, research using such measures rather than cognitive responses has reported stronger support for the causal role of brand cognitions (Gardner 1985; Mitchell 1986; Mitchell and Olson 1981). Given, however, the possibility that other inter-study differences could account for the more favorable findings involving expectancy-value measures, a more precise test of potential measurement effects would entail comparisons within the same investigation.

Such a comparison was provided by Park and Young (1986). They observed relatively little difference between cognitive response and expectancy-value measures. Both measures produced largely equivalent estimates of the relationship between brand cognitions and attitudes under all of their experimental conditions but one: affective involvement and a musical commercial. In this condition, the relationship was significant only for the expectancy-value measure. This particular finding is questionable on several grounds. First, although the regression coefficient between Ab and the expectancy-value measure was significant, the simple correlation between the two was not. Such differences typically indicate that the predictor is simply serving as a suppressor variable. It is also interesting to note that the expectancy-value measure was unrelated to Ab when subjects processed the same musical commercial but under cognitive involvement conditions. Why the measure should relate to Ab under low involvement but not cognitive involvement is not at all clear, and runs counter to predictions about the role of involvement discussed below.

One objective of the present study was to provide a further examination of this measurement issue. In particular, we examine how changes in the operationalization of brand cognitions (i.e., expectancy-value versus cognitive response measures) influence its relationship with brand attitude.

A second explanation for the weak relationship between brand-oriented cognitive responses and brand attitude is based on the level of involvement that occurs during message processing. As suggested earlier, the ELM predicts that brand cognitive responses will play an increasingly important role in the formation of brand attitude when the person becomes more motivated and/or able to engage in a careful evaluation of a message's arguments. Consistent with this, several researchers (Batra and Ray 1986; MacKenzie et al. 1986) have argued that brand cognitions are less important contributors to the formation of Ab in typical advertising exposure settings since viewers are unlikely to be motivated to elaborate on the brand-relevant arguments presented in the ad.

Thus, the weak relationships observed in prior research may simply reflect the undermining effect of limited processing of the message claims due to weak involvement. This basic reasoning is supported by the previously discussed finding from Park and Young (1986) that CRb was related to Ab when a non-musical ad was processed under cognitive involvement conditions, but not under low involvement conditions. Additional confirmation of involvement's moderating role is furnished by Gardner (1985). Gardner (1985) found that an expectancy-value measure was more strongly associated with Ab under brand than under non-brand processing sets. Recently, Hastak and Olson (1989) reported a similar effect of brand versus non-brand processing set on the CRb - Ab relationship.

In the present study, involvement is used to compare further the potential impact of different approaches for assessing brand cognitions. Research participants examined a print ad for a hypothetical new brand of soft drink under conditions designed to alter their involvement during processing. Consequently, it was possible to make comparisons between cognitive response and expectancy-value measures within each involvement condition. Both measurement approaches were expected to reveal the same basic pattern of stronger relationships among brand cognitions and attitude as involvement increases.

METHOD

Subjects

A total of 170 male and female students enrolled in an undergraduate business course at a midwestern university participated in the study to earn extra credit. Subjects were randomly assigned to experimental conditions and processed in groups of five to ten.

Procedure

After subjects were seated at partitioned tables, the experimenter stated that the purpose of the study was to obtain their reactions to two new products being considered for introduction in their area. It was explained that they would look at ads for the new products and then make a choice between the two products. Subjects were told that they would receive the product of their choice as additional compensation for their participation.

The experimenter then drew subjects' attention to a set of folders in front of each subject. Starting with the top folder, subjects worked at their own pace through the set and were not allowed to return to a particular folder after they had finished with it. The first folder contained an ad for a fictitious soft drink called Sunburst. The ad heading read "Introducing the next generation of soft drinks for today's generation." Beneath the heading appeared two color pictures side-by-side: one was of the soft drink can, the other varied as explained below. The bottom half of the ad contained copy making a number of claims about the product.

The next folder held a questionnaire that assessed subjects' reactions to the ad and product. The third folder contained an ad for a fictitious soft drink called Sparkle. The fourth folder contained a form asking the subject to indicate whether they would like to receive the Sparkle or the Sunburst brand. As explained below, subjects in the low involvement condition also received a fifth folder.

Involvement Manipulation

The involvement manipulation consisted of varying the personal relevance of the Sunburst ad by telling subjects they would eventually make a choice among either soft drinks or an unrelated product category. Subjects in the higher involvement condition were told that they would be shown a print ad for each of two new soft drink brands and then make a choice between the two brands. They would then receive a six pack of the brand of their choice. This basic procedure has been used previously for enhancing involvement (cf. Petty, et al. 1983; Wright 1973). In contrast, subjects in the lower involvement condition were informed that they would be asked to choose between two disposable razor brands. It was expected that subjects would view the Sunburst ad as more personally relevant when they anticipated making a choice involving the product relative to those expecting to make a brand choice in another product category.

This manipulation necessitated some changes in the procedure and materials for low involvement subjects. First, some rationale was needed as to why they would encounter the soft drink ads. Consequently, the experimenter explained that subjects would be exposed to a couple of "warm-up" soft drink ads before seeing the critical razor ads in order to familiarize themselves with the type of print ads used in the study. It was also necessary to make an alteration in the brand choice form within the fourth folder since subjects were not anticipating making a soft drink choice. The low involvement brand choice form stated that some of the subjects would receive a six pack of soda rather than a disposable razor. Subjects were therefore asked to indicate which brand they would prefer in the event that they received the soda. Finally, low involvement subjects received a fifth folder containing two ads for fictitious disposable razors and a razor choice form.

Claims Manipulation

This manipulation involved varying the strength of the arguments within the Sunburst ad. Common to both versions were the following claims: great taste, low in calories, thirst quenching, natural ingredients, real fruit juices, comparable price, and just the right amount of carbonation. The versions differed only in the presence and form of additional copy that attempted to substantiate these claims. One version contained information that was intended to provide stronger support for the claims (e.g., Sunburst has 50% fewer calories than the average non-diet soft drink) while a second version presented weaker support (e.g., Sunburst has 5% fewer calories than the average non-diet soft drink.). The claims and substantiation copy were placed in the bottom half of each ad used in the final experiment. Each claim was printed in bold block letters (e.g., GREAT TASTE!) which was followed by substantiation copy that comprised the claim manipulation.

Picture Manipulation

This manipulation involved altering the color picture that appeared at the top of the ad next to the picture of the Sunburst soft drink can. The favorably evaluated stimulus was a tropical beach scene with a rising sun. The disliked stimulus was a side view of five ugly iguanas facing the same direction.

Measures

Only those measures germane to the current research issues are described below. The key dependent variables consisted of thoughts, beliefs, attitudes, and purchase intentions. The thought elicitation measure, which was completed immediately after exposure to the Sunburst ad, asked subjects lo list all of the thoughts, ideas, and images they had while looking at the ad. These thoughts were listed in the blank space, roughly three-fourths of a page, beneath the measure. Although a time constraint was not imposed, almost all of the subjects completed the measure in a couple of minutes.

The expectancy-value measure was operationalized in accordance with a Fishbein multiattribute model (Fishbein and Azjen 1975). Subjects' evaluations (ei) of the same seven soft drink attributes used in the ad (e.g., "For me, a soft drink with more real fruit juices than other brands is:") were measured on two scales (good - bad, desirable - undesirable). Beliefs (bi) about Sunburst (e.g., "how likely is it that Sunburst has more real fruit juices than other brands?") were assessed on two scales (likely - unlikely, probable improbable). Responses to the two scales for each attribute evaluation and belief were summed since the correlations between them were very high (average r = .87 for ei, average r = .94 for bi).

Ad attitude was assessed on five scales (good - bad, effective - ineffective, interesting uninteresting, like very much - dislike very much, not at all irritating - very irritating) which were summed (alpha = .92). Attitude toward the Sunburst soft drink was represented by the sum (alpha = .97) of three scales (favorable - unfavorable, positive negative, like very much - dislike very much). Intention to purchase Sunburst when it became available was captured by two scales (likely unlikely, probable - improbable) which were added together (r= .98).

The following measures were used to examine the success of the involvement manipulation. Subjects' involvement while processing the ad was assessed by asking subjects to respond to the statement "While reading the Sunburst ad, I was:" using three scales (very involved - very uninvolved, concentrating very hard - concentrating very little, paying a lot of attention - paying very little attention). Subjects also reported their agreement on an agree - disagree scale with the statement "I carefully considered the claims about Sunburst in the ad." These measures were combined to form an indicator of processing involvement (alpha = .91).

Cognitive Response Coding Procedures

The coding of the cognitive response data consisted of two phases. First, each protocol was decomposed into separate thoughts by two judges blind to the experimental condition. A total of 90 disagreements occurred between the two judges which were then resolved by discussion. The total number of thoughts after resolving the disagreements was 689 or an average of four thoughts per subject. Judges therefore agreed on nearly 87 percent of the thoughts expressed by subjects.

In the second phase, the same two judges assigned each thought to categories reflecting the valence (positive, neutral, and negative) and content. The basic distinction drawn in thought content was whether it pertained to the ad versus to the product. For ad thoughts, a total of 13 categories were developed which reflected whether the thought was directed at the picture, copy, or some other ad element. One of these categories represented expressions reflecting an overall evaluation of the ad. Thirteen categories were also developed for brand thoughts. These categories represented differences in terms of whether the thought focused on product attributes, brand name, or the can/packaging. Thoughts indicating overall brand evaluation were classified separately.

Twenty-four of the 689 total responses were excluded on grounds of being either unclassifiable or irrelevant. Favorable thoughts accounted for 204 of the usable responses, 65 of the thoughts were neutral, and the remaining 396 thoughts were coded as unfavorable. In addition, 103 responses reflected an overall evaluation of the ad (n=36) or product (n=67). Because of concerns about such responses being indicators or overall affect (see Olson, Toy, and Dover 1982; Wright 1980), they were excluded, thus leaving 562 thoughts about various aspects of the ad (n=217) and product (n=345) for use in the analyses involving cognitive responses.

FIGURE

THE DUAL MEDIATION MODEL

In terms of thought valence, the judges were very consistent in their ratings as reflected by an agreement rate of nearly 96 percent. For thought content, the judges agreed 81 percent of the time. While this agreement rate is lower, we believe it represents an acceptable level considering the large number of categories used to classify thought content. Disagreements for both valence and content were resolved through discussion.

RESULTS

Manipulation Checks

As expected, subjects' responses to the processing involvement measures significantly differed between high (M=4.47) and low (M=.038) involvement conditions (F=26.3, p<.001). Note that the mean for the low involvement condition was at the scale midpoint (scale ranges from +12 to -12). Consequently, this involvement condition is more appropriately viewed as inducing a moderate level rather than a truly low level of involvement.

Similar support was derived from two cognitive response indices representing (I) the total number of brand thoughts and (2) the ratio of brand thoughts to total thoughts. Higher involvement subjects reported more total brand thoughts ( M=2.03 vs. 1.41; F=32 9, p<.001) as well as a greater proportion of brand thoughts (M=.57 vs. .45; F=25.2, p<.001) than those less involved.

Models

The following analyses focus on causal configurations of the relationships among brand and ad attitudes and antecedents. The dual mediation model (Figure A) supported by MacKenzie et al. (1986) was tested using the maximum likelihood procedure for path analysis in LISREL VI (Joreskog and Sorbom 1984). We report path analysis results for models in which brand cognitions are represented by either a cognitive response index or a multiattribute model score in three cuts of the data -- (1) pooled across treatment conditions, (2) for the higher involvement group (n=86), and (3) for the lower involvement group (n=84).

Results for CRb

Table 1 summarizes the results of the path analyses for the dual mediation model using the cognitive response measure (cf. MacKenzie et al. 1986). An examination of the goodness-of-fit statistics indicates that all three models provide a very good description of data relationships. Two of the three chi-square statistics are non-significant and the root mean square residuals (RMSR) are small. Joreskog and Sorbom's goodness of fit estimates range from .960 to .977. The variance explained in the endogenous constructs ranges from a low of 15.1% for CRb to a high of 58.6% for intentions. Taken together, these results indicate that the dual mediation model provides an acceptable description of data relationships in all three samples. Therefore, interpretation of the parameter estimates is appropriate. Given the focus of the present paper, discussion will center on the CRb - Ab path estimates.

Results based on the pooled data revealed a significant (b32 = .216, p<.01) path estimate between CRb and Ab. The path estimate maintained significance (b32 = .275, p<.01) under higher involvement conditions, but becomes somewhat attenuated under lower involvement conditions (b32 = .151, p<.l). Even so, the results of a chi-square difference test did not reject the null hypothesis of equivalence in the CRb - Ab path between the higher and lower involvement conditions (x2(1) = 1.19, p >.1).

TABLE 1

PATH ANALYSIS RESULTS FOR MODELS USING BRAND COGNITIVE RESPONSE MEASURE

This unexpected stability of the CRb- Ab relationship prompted us to adopt another approach for testing involvement's moderating role. Based on the responses to the involvement measures, subjects were divided into three groups of roughly equal sizes. Path analyses were then performed on the two extreme groups representing either higher or lower levels of self-reported involvement for models including either the brand CR or expectancy-value indices. While this procedure sacrifices strict experimental control and runs the risk of confounding involvement with other individual difference variables, the potential for differences to emerge is also enhanced by using the two extreme groups which represent greater polarization in their level of involvement (M=7.92 versus -3.84; n=61 for each group) than captured by the involvement conditions (M=4.47 versus 0.38). Additional analyses were performed on the higher and lower self-reported involvement groups to validate presumed differences in their degree of elaboration and their responsiveness to the ad content manipulations. As expected, the higher SRI group (M=2.38) reported more brand-oriented cognitions than the lower SRI group (M=1.57, p<.01). A 2 (Claim) X 2 (Picture) ANOVA was also performed on Ab for each SRI group. The higher SRI group was affected only by the claim manipulation (F=21.8, p<.001), while the lower SRI group was affected only by the picture manipulation (F=7.5, p<.01).

The results of the path analyses are reported in the last two columns of Table 1. As can be seen, the difference in the strength of the CRb - Ab path was intensified by using the extreme self-reported involvement (SRI). Indeed, the CRb - Ab path becomes non-significant (p>.1) for the lower SRI group. A chi-square difference test indicated that the influence of CRh on Ah was significantly stronger in the high SRI group (X2(1) = 5.59, p<.05).

Results for the Expectancy-Value Models

The goodness-of-fit statistics and parameter estimates for the models using the multi-attribute measure of brand cognitions are reported in Table 2. The path estimate for Sbiei - Ab is significant (532 = .330 p<.01) in both the pooled data analyses and as well as for the lower (b32 = .353, p<.01) and higher (b32 = .328, p<.01) involvement conditions. These estimates did not differ between involvement conditions (X2(1) = 0.05, p>.1).

TABLE 2

PATH ANALYSIS RESULTS FOR MODELS USING MULTI-ATTRIBUTE MEASURE

In contrast, the results based on self-reported involvement revealed the anticipated difference in the strength of the Sbiei - Ab relationships. Whereas this link is marginally significant for the lower SRI group (b32 = .181, p<.l), it achieves strong significance for the higher SRI group (b32 = .594, p<.01). A chi-square difference test confirmed the Sbiei - Ab relationship to be stronger in the higher SRI condition (X2(1) = 7.20, p<.05).

These results are essentially the same as those reported for the CRb measure. The only substantive difference between the two measurement approaches is that the strength of the relationship between brand cognitions and attitudes is somewhat enhanced for the models using the expectancy-value measure. For example, when the analysis is based on the pooled data, the path estimate (b32) based on the expectancy-value measure increases by nearly fifty percent in size relative to the same estimate based on cognitive responses.

DISCUSSION

Mackenzie et al. (1986) were disturbed by the surprising weakness of the relationship between brand cognitions as measured by a cognitive response index and brand attitudes. They provided two explanations for this finding. First, they suggested that the crudeness of the cognitive response index as an indicator of brand cognitions may have contributed to the lack of relationship with brand attitudes. Second, based on the ELM, they built an argument for the primacy of peripheral processing over central processing in brand attitude formation in a typical advertising exposure setting.

Our results indicate that the use of CRb did reduce the apparent strength of the relationships between brand cognitions and attitude in comparison with an expectancy-value measure. However, it should be noted that this attenuation may also be attributable to the greater availability of subjects' responses to the expectancy-value measure at the time of their response to the brand attitude items (Feldman and Lynch 1988). This is possible because Ab was measured immediately after the expectancy-value measures were taken, while a number of measures separated Ab and the thought elicitation task. Nonetheless, conclusions about the causal role of brand cognitions are essentially the same across the models. Both operationalizations yielded significant relationships between brand cognitions and attitude in the pooled data analyses. These relationships were also found to be invariant across the involvement conditions for both measurement approaches. Finally, the alternative measures produced the same pattern of differences in the significance of the brand cognitions - attitude relationship across the self-reported involvement analyses. Consequently, it would appear that prior research observing null relationships may be more indicative of the type of processing that occurred during message exposure than of any major weakness in cognitive response measures.

The results based on self-reported involvement also support the presumed moderating influence of involvement on the relationship between brand cognitions and attitude. The brand attitudes of subjects reporting lower involvement during processing were unrelated to their brand cognitions. Strong relationships, on the other hand, were observed for subjects reporting higher involvement during processing. These findings provide additional evidence for the moderating role of involvement in persuasion.

REFERENCES

Batra, Rajeev and Michael L. Ray (1986), "Affective Responses Mediating Acceptance of Advertising," Journal of Consumer Research, 13(September), 234-249.

Feldman, Jack M. and John G. Lynch, Jr. (1988), "Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention, and Behavior," Journal of Applied Psychology, 73(3), 421-435.

Fishbein, Martin and Icek Ajzen (1975), Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research, Reading, MA: Addison-Wesley.

Gardner, Meryl P. (1985), "Does Attitude Toward the Ad Affect Brand Attitude Under a Brand Evaluation Set?", Journal of Marketing Research, 22(May), 192- 198.

Greenwald, Anthony (1968), "Cognitive Learning, Cognitive Response to Persuasion, and Attitude Change," in Psychological Foundations of Attitudes, eds. A.G. Greenwald, T.C. Brock, and T.C. Ostrom, New York: Academic Press.

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.

Joreskog, Karl G. and Dag Sorbom (1984), LISREL VI: Analysis of Linear Structural Relationships by Maximum Likelihood and Least Squares Methods, Chicago: National Education Resources, Inc.

Lutz, Richard J. (1975), "Changing Brand Attitudes Through Modification of Cognitive Structure," Journal of Consumer Research, l (March), 49-59.

Lutz, Richard J., Scott B. MacKenzie, and George Belch (1983), "Attitude Toward the Ad as a Mediator of Advertising Effectiveness: Determinants and Consequences," in Advances in Consumer Research, Vol. 10, Richard Bagozzi and Alice Tybout, eds. Ann Arbor, MI: Association for Consumer Research, 532-539.

Mackenzie, Scott B. and Richard J. Lutz (1983), 'Testing Competing Models of Advertising Effectiveness Via Structural Equations Models," in Proceedings, AMA Winter Educators Conference, Chicago: American Marketing Association, 70-75.

Mackenzie, Scott B., Richard J. Lutz, and George E. Belch (1986), 'The Role of Attitude Toward the Ad as a Mediator of Advertising Effectiveness: A Test of Competing Explanations," Journal of Marketing Research, 23(May), 130-143.

Mitchell, Andrew A. (1986), "The Effect of Verbal and Visual Components of Advertisements on Brand Attitudes and Attitude Toward the Advertisement," Journal of Consumer Research, 13(June), 12-24.

Mitchell, Andrew A. and Jerry C. Olson (1981), "Are Product Attributes Beliefs the Only Mediator of Advertising Effect on Brand Attitudes?", Journal of Marketing Research, 18(Aug.), 318-331.

Olson, Jerry C., Daniel R. Toy, and Philip A. Dover (1982), "Do Cognitive Responses Mediate the Effects of Advertising Content on Cognitive Structure?", Journal of Consumer Research, 9(December), 245-262.

Park, C. Whan and S. Mark Young (1986), "Consumer Response to Television Commercials: The Impact of Involvement and Background Music on Brand Attitude Formation," Journal of Marketing Research, 23(Feb.), 11-24.

Petty, Richard E. and John T. Cacioppo (1986), Communication and Persuasion: Central and Peripheral Routes to Attitude Change, New York: Springer-Verlag.

Petty, Richard E., John T. Cacioppo, and David Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10(Sep.), 135-146.

Shimp, Terence A. (1981), "Attitude Toward the Ad as a Mediator of Consumer Brand Choice," Journal of Advertising, 10(2), 9-15.

Wright, Peter (1973), "The Cognitive Processes Mediating Acceptance of Advertising," Journal of Marketing Research, 10(Feb.), 53-62.

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

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