Advertising=S Longitudinal Effects on Brand Attitudes: the Moderating Roles of Evaluation Goals and Attitude Confidence

ABSTRACT - In this study, we proposed our research model based on DMH model in a longitudinal perspective and explored how some variables moderate the relationships among constructs in this model. Subjects answered questionnaires immediately after ad exposure and ten days later. First, it was found that DMH model can apply at a later time as well as at an immediate time after ad exposure. Second, DMH model appeared to work better when consumers are in the brand evaluation set relative to the ad evaluation set. Third, we found that the effects of antecedent variables on the formation of brand attitudes at a later time vary with evaluation goal and brand attitude confidence. At the end of the paper, theoretical contributions, limitations, and future research directions are discussed.

Citation:

Haksik Lee, Jaerock Lee, and Gilbert D. Harrell (2002) ,"Advertising=S Longitudinal Effects on Brand Attitudes: the Moderating Roles of Evaluation Goals and Attitude Confidence", in NA - Advances in Consumer Research Volume 29, eds. Susan M. Broniarczyk and Kent Nakamoto, Valdosta, GA : Association for Consumer Research, Pages: 173-179.

Advances in Consumer Research Volume 29, 2002     Pages 173-179

ADVERTISING=S LONGITUDINAL EFFECTS ON BRAND ATTITUDES: THE MODERATING ROLES OF EVALUATION GOALS AND ATTITUDE CONFIDENCE

Haksik Lee, Hong Ik University

Jaerock Lee, Chongju University

Gilbert D. Harrell, Michigan State University

ABSTRACT -

In this study, we proposed our research model based on DMH model in a longitudinal perspective and explored how some variables moderate the relationships among constructs in this model. Subjects answered questionnaires immediately after ad exposure and ten days later. First, it was found that DMH model can apply at a later time as well as at an immediate time after ad exposure. Second, DMH model appeared to work better when consumers are in the brand evaluation set relative to the ad evaluation set. Third, we found that the effects of antecedent variables on the formation of brand attitudes at a later time vary with evaluation goal and brand attitude confidence. At the end of the paper, theoretical contributions, limitations, and future research directions are discussed.

Only a few researchers have studied what attitude is reported at a later time (e.g., Chattopadhyay and Nedungadi 1992; Mitchell 1993; Muehling and Laczniak 1988). In these studies, some variables have been identified as antecedents of brand attitudes at a later time. For example, Mitchell (1983a, 1983b) found that the ad attitudes persist over time and have a significant influence on brand attitudes two weeks later. The effect of immediate ad attitudes also was found in other studies (Chattopadhyay and Nedungadi 1992; Muehling and Laczniak 1988). Other researchers have reported the effects of immediate brand attribute beliefs (Mitchell 1993; Muehling and Lacznick 1988) and brand attitudes on brand attitudes at a later time (Mitchell 1993).

MacKenzie, Lutz, and Belch (1986) compared four alternative models in relation to the effect of ad attitudes, and reported that DMH (Dual Mediation Hypothesis) model appeared to be superior to the other three alternative models. Although their study made a significant contribution to the understanding of advertising effects, DMH model has never been studied in a longituinal context. It is generally accepted that the causal relationships among related constructs for advertising effects vary with consumer’s involvement or evaluation goal (Gardner 1985; Gardner, Mitchell and Russo 1985). However, we do not know whether DMH model can apply regardless of involvement or evaluation goal.

In this study, we propose our research model based on DMH model in a longitudinal perspective. Then, we examine the applicability of DMH model in different contexts and explore how some variables moderate the relationships among constructs in this model. More specifically, our purpose is fourfold. First, we investigate whether DMH model works at both immediate and later times. Second, we study whether DMH model can apply regardless of evaluation goal. Third, we examine how evaluation goal moderates the influences of immediate brand attitudes and ad attitudes at a later time on brand attitudes at the later time. Finally, we investigate how brand attitude confidence moderates the influences of immediate brand attitudes and brand-related attribute beliefs at a later time on brand attitudes at the later time.

RESEARCH MODEL AND HYPOTHESES

Research Model

If consumers are asked to express their brand attitudes just after ad exposure, those attitudes are likely to be very salient and easily accessible, but they rarely influence purchase decisions immediately. Rather, the attitudes may be stored in long-term memory and retrieved later in a purchase situation. The network model asserts that long-term memory can be represented as a network of nodes (information) and connecting links. When consumers are exposed to a stimulus, the stimulus may activate a particular node. Consequently, activation may spread from that node to other linked nodes (Keller 1987).

Consumer brand attitudes (A_b2) at a later time after ad exposure (t₂) can be influenced by brand attitudes (A_b1) formed immediately after ad exposure (t₁) to the extent that A_b1 can be retrieved. A_b2 can also be influenced by brand attribute beliefs at the later time (C_b2), if present at that time. Especially if consumers can’t retrieve or did not form A_b1, A_b2 is more likely to be influenced by C_b2. C_b2 can be influenced by the initial brand beliefs (C_b1) to the extent that C_b1 can be retrieved from memory (Lichtenstein and Srull 1985) and by the information provided at that time (Mitchell 1993). Sometimes C _b2 can be inferred by other cues (Ford and Smith 1987). In addition, as A_b1 can be influenced by immediate ad attitudes (A_ad1) (Mitchell and Olson 1981), A_b2 can be influenced by ad attitudes at the later time (A_ad2). To summarize, A_b2 can be influenced by A_b1, C_b2, and A_ad2, and C_b2 and A_ad2 can be influenced by C_b1 and A_ad1, respectively, to the extent that they can be retrieved.

Based on DMH model, we propose the research model of our study in a longitudinal perspective, as shown in Figure 1. The research model consists of two parts. One involves brand attitudes formation mechanisms immediately after ad exposure, and the other concerns a later time.

Hypotheses

DMH Model in a longitudinal context and unde different evaluation goals. As reviewed, MacKenzie, Lutz and Belch’s (1986) DMH model appeared to be superior to the other three models. This finding was supported also by Homer’s (1990) study. Then, how does A_ad influence A_b at a later point in time? We propose that as long as A_ad and C_b exist in consumers’ long-term memory, they would influence A_b in the same way as they do immediately after exposure. In other words, if consumers are asked about A_b at time 2, they may retrieve their A_ad and C_b stored at time 1, and these retrieved A_ad and C_b would influence A_b.

The next question that can be raised is how consumers’ evaluation goal moderates the causal relationships among A_ad, C_b, and A_b at both time 1 and time 2. It is generally accepted that A_ad influences A_b more in the case of the low rather than high involvement whereas C_b influences A_b more under the high rather than low involvement (Gardner 1985). In addition, the strength of causal paths among the variables in DMH model was examined by Miniard, Bhatla and Rose (1990). In this study, the influence of C_b on A_b appeared to be stronger in the high rather than low involvement condition (see p.298, Table 2; .275 vs. .151 .458 vs. .097). However, this result is inconsistent with Homer’s (1990) finding in that the coefficients for the same path appeared to be not significantly different between the two groups (see p.82, Table 1)

FIGURE 1

RESEARCH MODEL

Therefore, there have been mixed findings for the moderating role of involvement in the causal relationships between C_b and A_b. We think, however, consumers would elaborate the features of the brand more deeply in the brand evaluation condition rather than the ad evaluation condition. Therefore, C_b would influence A_b more significantly when consumers are in the brand rather than ad evaluation condition. These discussions lead to the following hypotheses concerning DMH model.

H1-a: DMH model will work at a later time (paths 5, 6, 7) as well as at an immediate time after ad exposure (paths 1, 2, 3).

H1-b: DMH model will work better under the brand evaluation strategy relative to the ad evaluation strategy (paths 1, 2, 3 and 5, 6, 7).

The Moderating Role of Evaluation Goals in the Effects of A_ad1, A_ad2, and A_b1. Numerous studies in the persuasion literature suggest that an individual’s information processing, evaluation process, and knowledge structure vary with evaluation goals and strategies (brand evaluation and nonbrand evaluation). For instance, based on Anderson’s (1983) network model of long-term memory structure, Gardner, Mitchell, and Russo (1985) identified two different types of processing strategies (see also Gardner 1985; Mitchell 1983b). When exposed to an ad, individuals with a brand evaluation goal may fully activate their brand-relevant schema and generate elaborations about the brand. They are expected to engage in deep, elaborate, and systematic processing of brand-related information in the ad. Thus, argument-related cognitive responses, such as counterarguments and support arguments, elicited during ad exposure would influence brand-related beliefs and brand attitudes (Hastak and Olson 1989).

In contrast, individuals with a nonbrand evaluation goal would not activate their relevant schema and generate brand elaborations (Gardner, Mitchell, and Russo 1985). They would engage in relatively shallow and limited processing of brand-related information and generate few, if any, argument-related cognitive responses. Therefore, in this case, individuals are not likely to form brand attitudes based on well-organized brand-related beliefs (Beattie and Mitchell 1985). Even when they form certain attitudes, these ae likely to be influenced by ad attitudes (MacKenzie, Lutz, and Belch 1986; Mitchell and Olson 1981). This reasoning leads to the following hypotheses.

H2-a: The effect of A_ad1 on A_b1 (path 2) will be stronger when consumers are in the ad evaluation set relative to the brand evaluation set.

H2-b: The effect of A_ad2 on A_b2 (path 6) will be stronger when consumers are in the ad evaluation set relative to the brand evaluation set.

Since information processing goals can play an important role in affecting the nature and amount of elaboration, they would influence the encoding and retrieving of ad effects. If consumers have a brand evaluation goal, they are more likely to process the message as it is received and form brand attitudes based on the brand information contained in the message. These attitudes would be relatively easily retrieved at some later time. However, consumers who do not have a brand evaluation goal are less likely to process the message and form brand attitudes based on the brand information contained in the message. If the consumers must express brand attitudes at a later point in time, then the attitudes should be dependent mostly on the information that can be recalled (Petty, Unnava, and Strathman 1991). Therefore, the relationship between the evaluative implications of the information that are recalled and the attitudes would be stronger in the latter condition than in the former condition (Lichtenstein and Srull 1985). This reasoning leads to the following hypothesis.

H3: The effect of A_b1 on A_b2 (path 4) will be stronger when consumers are in the brand evaluation set relative to the ad evaluation set.

The Moderating Role of Brand Attitude Confidence in the Effects of A_b1 and C_b2. Zanna and Fazio (1982) argued that attitudes have not only an evaluative dimension (attitude valence) but also a nonevaluative dimension, such as attitude accessibility, attitude confidence, and attitude strength. For example, Bennett and Harrell (1975) found that high confidence in attitudes strengthens the attitude-intentions link. Fazio, Powell, and Herr (1983) argued that attitudes with high confidence do not change easily. Furthermore, Fazio (1986) asserted that strongly held attitudes are more likely than weakly held attitudes to be spontaneously retrieved when people are confronted with a decision situation. Laczniak and Muehling (1993) reported that individuals are more confident with high involvement in brand belief and A_ad confidence. It is argued, therefore, that the attitude-behavior relation varies with attitude confidence.

Therefore, if consumers are very confident in their brand attitudes (A_b1), then those attitudes will not change easily at a later time. If consumers are not confident in their A_b1, those attitudes may change easily at a later time. Instead, their brand attitudes (A_b2) are more likely to be influenced by brand attribute beliefs (C_b2) at that time (t₂). Therefore, the following hypotheses are established.

H4-a: The effect of A_b1 on A_b2 will be stronger when brand attitude confidence is high (path 4).

H4-b: The effect of C_b2 on A_b2 will be stronger when brand attitude confidence is low (path 5).

METHOD

Subjects and Experimental Ads

There were two waves of experimentation with an interval of ten days. In the first wave, 424 undergraduates in a business school participated. Just before conducting the experiment in their classroom, they were asked to participate. No students refused. Ten days later, in the second wave, 353 students from among the first wave participated. In the brand evaluation group, 177 of 211 first-wave subjects participated in the second experiment; in the ad evaluation group, 176 of 213 first-wave subjects participated in the second experiment. Among the 302 usable questionnaires, exactly half were from each group.

A pretest was conducted with 30 undergraduates to select a proper experimental product. The choice was a personal computer, which was very relevant to the students, required cognitive effort in the purchase decision, and was likely to be purchased in the near future by the subjects.

The ad for the product was composed of a picture of a real computer, headline, body copy, and company name. The headline was "The Ultimate Computer at an Affordable Price." The body copy consisted of CPU, RAM size, CD-ROM speed, MODEM speed, basic O/S, warranty, and price, which emerged as salient attributes in the pretest. The filler ad was about a toothbrush. Print ads of fictitious brands (Zecca for the personal computer) were made by an advertising agency.

Procedure

In the first wave of experimentation, twelve sessions were conducted. Each time, 25-50 students participated. Subjects were randomly divided into two groups, one manipulated as the brand evaluation goal group and the other manipulated as the ad evaluation goal group. They were asked to read statements for the manipulation.

The brand evaluation group was asked to pay attention to information in the ads during exposure and to determine whether they liked the brands and had an intention to buy them. The ad evaluation group was asked to assume they were ad planners and to evaluate whether the ads were well organized and would appeal to consumers, as well as whether they liked the ads.

Subjects then viewed one target ad and one filler ad for one minute, respectively. Half the subjects viewed the filler ad first and then the target ad, and the other half viewed the target ad first and then the filler ad. After viewing both the ads, they were requested to fill out the questionnaire. The instrument used in the first wave consisted of measures for brand beliefs structure (C_b1), ad attitudes (A_ad1), brand attitudes (A_b1), attitude confidence, and a manipulation check.

Before the experimentation began in the first wave, subjects were informed that there would be a second meeting ten days later. They were asked to write their student number on the cover page for identification for both the first and second waves of experimentation. A small gift was given to each subject. The questionnaire used in the second wave consisted of measures for brand beliefs structure (C_b2), ad attitudes (A_ad2), and brand attitudes (A_b2).

Measures

Brand attribute beliefs structure. Based on Fishbein’s (1967) attitude model (A_o=Sb_ie_i), brand attribute belief (b_i) and evaluation (e_i) for each attribute were measured. For the six salient attributes found in the pretest (CPU speed, RAM size, CD-ROM speed, MODEM speed, price, warranty), a seven-point scale was used for each attribute belief and each attribute evaluation.

Ad attitudes. Following the measures used in other research (e.g., Gardner 1985), ad attitudes were measured by five seven-point semantic differential scales (-3 to +3): good/bad, dislike/like, not irritating/irritating, uninteresting/inteesting, well made/poorly made.

Brand attitudes. Following the measures used in other studies (e.g., Hastak and Olson 1989; MacKenzie, Lutz and Belch 1986), brand attitudes were measured by five seven-point semantic differential scales (-3 to +3): good/bad, dislike/like, feel negatively/feel positively, unpleasant/pleasant, favorable/unfavorable.

Brand attitude confidence. Following the measures used in other studies (e.g., Berger and Mitchell, 1989), brand attitude confidence was measured by three seven-point semantic differential scales (1 to 7): completely confident/not at all confident, very certain/not at all certain, sure/not at all sure.

RESULTS

Manipulation Check

Two kinds of questions were asked to check manipulation of evaluation goals. One was to determine whether subjects viewed the ads in the way they were asked. Two seven-point scales were used (ranging from disagree to agree): "All the time during exposure to the ads, I thought if those brands were relevant to me," and "All the time during exposure to the ads, I thought how well those ads were made." The second kind of question was to determine how correctly they remembered the information given in the ad. Four multiple-choice questions were asked (for example, "What is the CPU speed of the computer advertised?").

The means of the brand evaluation group and the ad evaluation group for the first question were 5.1 and 3.4, respectively. The means of the brand evaluation group and the ad evaluation group for the second question were 3.0 and 4.6, respectively. The differences were statistically significant for both questions (p<.01). In the manipulation check asking brand information, 25 points were assigned for each question, totaling to 100. The scores of the brand evaluation group and the ad evaluation group were 95.1 and 61.3, respectively. The difference was also statistically significant (p<.01). To summarize, the manipulation appeared to be successful.

TABLE 1

ANALYSIS RESULTS FOR THE RESEARCH MODEL

Scale Refinement

To refine the measures of each construct (brand attitudes at t₁ and t₂, ad attitudes at t₁ and t₂, attitude confidence), we followed the scale refinement procedure suggested by Singh and Rhoads (1991).

First, an exploratory factor analysis was performed for each construct. All items for each construct loaded on a single factor, and the variance extracted by each factor exceeded .50 (see Bagozzi and Yi 1988; Challagalla and Shervani 1996; Singh and Rhoads 1991). To check reliability, Cronbach’s alpha was computed for each construct. All five alphas were greater than .88. Therefore, no items were discarded due to factor analyses or reliability checks.

Next, confirmatory factor analyses (CFA) were performed for the measures of each construct. One item was discarded respectively for brand attitudes (t₁, good/bad; t₂, good/bad,) and ad attitudes (t₁, good/bad; t₂, good/bad) since modification index was greater than 3.84 respectively (see Hayduk 1987, pp. 177-178).

Another CFA was performed for all remaining items, except for the confidence items, to assess the measurement model. The standardized residuals for all the items were smaller than 3.00. All AVE for the brand attitudes and ad attitudes at t₁ and t₂ exceeded the minimum of .50 suggested by Fornell and Larcker (1981). These results indicated that the measures of all constructs had acceptable levels of validity. In this regard, it can be said that the scale refinement procedure warrants unidimensionality of the scales of each construct.

Research Model and Hypotheses Testing

To test relationships among constructs, structural equation modeling was used. The covariance matrix among indicators was used for input data for LISREL 8 (J÷reskog and S÷rbom 1993). The goodness-of-fit evaluation for the overall model was as follows: c² = 213.621, d.f. = 126, p-value = .000017, GFI = .924, AGFI = .897, Std. RMSR = .043, and NFI = .928 (see table 1). These values satisfied or were close to the criteria for goodness-of-fit assessment.

H1-a concerns whether DMH model works at both immediate and later times. The standardized coefficients for A_ad¦A_b (.342 and .451), A_ad¦C_b (.339 and .497), and C_b¦A_b (.188 and .203) were all statistically significant at time 1 and time 2 (see pooled data of Table 1). This is consistent with DMH model and supports H1-a. The explained variance for each endogenous construct is as follows: A_b2, 44.7%; C_b2, 22.6%; A_ad2, 5.6%; A_b1, 10.7%; and C_b1, 22.0%.

In addition, the standardized coefficients for A_ad¦C_b were tended to be highest (.339, .497), the standardized coefficients for A_ad¦A_b the next (.342, .451), and the standardized coefficients for C_b¦A_b lowest (.188, .203). This order of coefficients is consistent with those reported by MacKenzie, Lutz and Belch (1986; Table 3, DMH .79, .21, -.04; .71, .12, -.03), Homer (1990; Table 1, total sample .89, .24, .17), and Miniard, Bhatla and Rose (1990; Table 2, pooled data .631, .389, .216).

H1-b concerns whether the causal relationships among A_ad, C_b, and A_b are different between the brand evaluation group and the ad evaluation group. As shown in Table 1, all the causal paths were statistically significant (p < .05) in the brand evaluation group. In contrast, for the ad evaluation group, the paths for C_b1¦A_b1, "_b1¦A_b2 , and C_b2¦A_b2 were statistically insignificant (p > .1). Therefore, it can be said that DMH model can apply to the brand evaluation group better than the ad evaluation group. This is consistent with H 1-b.

H2 concerns whether the effect of A_ad on A_b (at immediate and later times) varies with evaluation goal, and H3 concerns whether the effect of A_b1 on A_b2 varies with evaluation goal. To test these hypotheses, the data for each evaluation group were analyzed respectively, and the standardized path coefficients for the two groups were compared.

Table 2 shows the standardized coefficients for A_ad1¦A_b1 (.295 vs. .503) and for A_ad2¦A_b2 (.272 vs. .735) for the brand and ad evaluation groups. As expected, the standardized coefficients for the ad evaluation group appeared to be greater than for the brand evaluation group. To see the statistical significance of any difference between the two groups, a chi-square difference test was performed. This test compares path coefficients between two groups, with all other paths constrained to be equal across groups (see MacKenzie and Spreng 1992). In this test, the difference was significant for A_ad1¦ A_b1 (c²=4.717, p<.05) and not significant for A_ad2¦ A_b2 (c²=1.514, p>.10). Therefore, H2-a is supported, and H2-b is not supported.

TABLE 2

ANALYSES RESULT ACROSS EVALUATION GOALS

TABLE 3

ANALYSIS RESULT ACROSS CONFIDENCE LEVEL GROUPS

H3 was tested in a similar way. Table 2 shows the standardized coefficients for A_b1¦A_b2 for the brand and ad evaluation group (.415 vs. .117). As hypothesized, the standardized coefficient for the brand evaluation group appeared to be greater than for the ad evaluation group. In addition, a chi-square test showed statistical significance in the difference (c²=9.872, p<.005). Therefore, H3 is supported. This result supports Keller’s (1987) argument that information processing goals influence the encoding and retrieving of ad effects. It also empirically supports the LM argument (Petty and Cacioppo 1981) that brand attitudes formed through central route processing are more enduring.

H4 concerns whether the effects of A_b1 and C_b2 on A_b2 vary with attitude confidence level. We performed moderated regression analysis to test the moderating effect of attitude confidence. This approach is contrasted with subgroup analysis where sample is split into two groups (i.e., high vs. low confidence groups). Since moderated regression analysis maintains original scores on a moderator variable, the loss of information resulting from the artificial transformation of a continuous variable into a categorical variable can be avoided (Bagozzi, Baumgartner, and Yi, 1992; Baron and Kenny 1986).

The following regression equation was used by treating confidence level as a moderator:

"_b2 = A_b1 + C_b2 + A_ad2 + Cf + A_b1-Cf + C_b2-Cf + A_ad2-Cf.

Table 3 shows the results. The hypotheses predict a positive parameter for A_b1 X Cf term and a negative parameter for C_b2 X Cf term. Results support this prediction: The parameter estimate was positive and significant for A_b1 X Cf term (b_Ab1xCf=.387, t=3.286, p < .01) and negative and significant for C_b2 X Cf term (b_Cb2xCf= -.325, t= -3.154, p < .01). As expected, the parameter estimate was insignificant for attitude confidence (b_Cf= -.017, t= -.280). These results support H4-a and H4-b.

Although our major concern was the effect of ad exposure on brand attitudes at a later time, we additionally investigated the effects of A_ad1 and A_b1 on A_ad2 in each evaluation group. In the ad evaluation group, A_ad2 appeared to be influenced by A_ad1 more than by A_b1 (.357 vs. -.122), whereas in the brand evaluation group A_ad2 appeared to be influenced by A_b1 more than by A_ad1 (.605 vs. .046). The reason is thought to be that in the ad evaluation group A_ad1 is more salient than A_b1, whereas in the brand evaluation group A_b1 is more salient than A_ad1.

CONCLUSION

Summary and Theoretical Contributions

This study investigated in a comprehensive framework the relationships among antecedent variables and brand attitude formation at a later time after ad exposure. In this study, it was found that DMH model can apply at a later time as well as at an immediate time after ad exposure. We also found that DMH model works better when consumers are in the brand evaluation set rather than the ad evaluation set. This study also found that the effect of A_b1 on A_b2 is stronger when consumers have a brand evaluation goal rather than an ad evaluation goal. It also appeared that the effect of A_b1 on A_b2 is stronger when brand attitude confidence is high rather than low, while the effect of C_b1 is stronger when brand attitude confidence is low rather than high.

By explaining advertising effects based on DMH model in a longitudinal perspective and by identifying moderating variables that influence the relationships of the variables in the model, this study contributes to an understanding how ad exposure influences consumer’ brand attitudes formation after a while.

Managerial Implications

In this study, DMH model appeared to be effective even in a longitudinal context and to be more effective when consumers have a brand processing strategy. A striking finding is that A_ad greatly influences C_b in both immediate and later times. These results imply that practitioners have to be concerned about A_ad more than they have been so far. It is because A_ad may directly influence A when consumers have an ad processing strategy, whereas it may directly and indirectly through C_b influence A_b when consumers have a brand processing strategy.

In addition, when consumers are in the brand evaluation set, their A_b2 tend to be influenced by A_b1 (.415) much more than by A_ad2 (.272) and C_b2 (.212). In contrast, when consumers are in the ad evaluation set, their A_b2 is likely to be influenced by A_ad2 (.735) much more than by A_b1 (.117) and by C_b2 (.075). In this regard, the importance of concern on A_ad2 is more emphasized when consumers are not in the brand evaluation condition. The importance of A_b1, however, is more emphasized when consumers are in the brand evaluation condition. In reality, most consumers are less likely to be in the brand evaluation set. Rather, they are more likely to passively see and/or hear the ads. Therefore, in a longitudinal perspective, it is extremely important to keep (or form) consumers’ ad attitudes in a more favorable way.

Limitations and Future Research Directions

Since this study focused on one product and used student subjects, there are some limitations in generalizing the results. Furthermore, in order for the experimental condition to be closest to reality, the print ad should be presented along with articles and several other clutter ads. Finally, we measured A_ad, C_b, and A_b immediately after exposure and again after ten days, but we could not control subjects’ behaviors during the interval. In the future, replication studies are needed in different settings to address these limitations.

At least two areas can be considered for further research. First, in our experimental setting, new information was not given after ad exposure. Ten days after ad exposure, subjects were asked to express their A_ad2, C_b2, and A_b2. However, in reality consumers could unintentionally be exposed to additional information or even intentionally seek to make a purchase decision. Therefore, for A_b2 to have a more realistic meaning, new information may be provided to subjects some time between time 1 and time 2 or at time 2. The studies addressing this point may explain the longitudinal advertising effects in a more realistic way.

Second, in this study, A_b2 appeared to be influenced by A_b1, C_b2, and A_ad2. If the interval is short, A_b1 may remain salient in long-term memory and influence A_b2 significantly. After longer intervals, however, A_b1 may be less salient due to a forgetting effect and may not influence significantly the formation of A_b2. In this situation, consumers may form brand attitudes based on their brand-related attribute beliefs, which can be retrieved from memory, obtained from information sources at that point, and/or inferred from other cues. Future research may address the moderating role of a given delay interval in the relationships of all the related constructs.

REFERENCES

Anderson, John. R. (1983), "A Spreading Activation Theory of Memory," Journal of Verbal Learning and Verbal Behavior, 22, 261-95.

Bagozzi, Richard P., Hans Baumgartner, and Youjae Yi (1992), "State Versus Action Orientation and the Theory of Rasoned Action: An Application to Coupon Usage," Journal of Consumer Research, 18 (March), 505-518.

Bagozzi, Richard P. and Youjae Yi (1988), "On the Evaluation of Structural Equation Models," Journal of the Academy of Marketing Science, 16 (Spring), 74-94.

Baron, Reuben M. and David. A. Kenny (1986), "The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations," Journal of Personality and Social Psychology, 5 (6), 1173-1182.

Beattie, Ann. E. and Andrew A. Mitchell (1985), "The Relationship Between Advertising Recall and Persuasion: An Empirical Investigation," in Psychological Processes and Advertising Effects, ed. Linda F. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: Lawrence Erlbaum Associates, Inc., 129-156.

Bennett, Peter D. and Gilbert D. Harrell (1975), "The Role of Confidence in Understanding and Predicting Buyers’ Attitudes and Purchase Intentions," Journal of Consumer Research, 2 (September), 110-117.

Berger, Ida E. and Andrew A. Mitchell (1989), "The Effect of Advertising on Attitude Accessibility, Attitude Confidence, and Attitude-Behavior Relationship," Journal of Consumer Research, 16 (December), 269-279.

Challagalla, Goutam N. and Tasaddug A. Shervani (1996), "Dimensions and Types of Supervisory Control: Effects on Salesperson Performance and Satisfaction," Journal of Marketing, 60 (January), 89-105.

Chattopadhyay, Amitava and Prakash Nedungadi (1992), "Does Attitude toward the Ad Endure? The Moderating Effects of Attention and Delay," Journal of Consumer Research, 19 (June), 26-33.

Fazio, Russel H.(1986), "How Do Attitudes Guide Behavior?" in Handbook of Motivation and Cognition: Foundations of Social Behavior, ed. R. M. Sorrentino and E.T. Higgins, New York: Guilford, 204-243.

Fazio, Russel H., Martha C. Powell, and Paul M. Herr (1983), "Toward a Process Model of the Attitude-Behavior Relation: Accessing One’s Attitude upon Mere Observation of the Attitude Object," Journal of Personality and Social Psychology, 44 (April), 723-735.

Fishbein, Martin A. (1967), "Attitude and the Prediction of Behavior," in Readings in Attitude Theory and Measurement, ed. Martin Fishbein, New York: Wiley, 477-92.

Ford, Gary T. and Ruth A. Smith (1987), "Inferential Beliefs in Consumer Evaluations: An Assessment of Alternative Processing Strategies," Journal of Consumer Research, 14 (December), 363-371.

Fornell, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error," Journal of Marketing Research, 18 (February, 39-50.

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.

Gardner, Meryl P., Andrew A. Mitchell, and J. Edward Russo (1985), "Low Involvement Strategies for Processing Advertisements," Journal of Advertising, 14 (2), 4-12.

Hastak, Mmnoj 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.

Hayduk, Leslie A. (1987). Structural Equation Modeling with LISREL: Essentials and Advances, The Johns Hopkins University Press.

Homer, Pamela M. (1990), "The Mediating Role of Attitude toward the Informational Determinants of Attitude," Journal of Marketing Research, 15 (November), 78-86.

J÷reskog, Karl G. and Dag S÷rbom (1993), LISREL 8: A Guide to the Program and Applications, Chicago: Scientific Software International.

Keller, Kevin L. (1987), "Memory Factors in Advertising: The Effect of Advertising Retrieval Cues on Brand Evaluations," Journal of Consumer Research, 14 (December), 316-333.

Laczniak, Russel N. and Darrel D. Muehling (1993), "Toward a Better Understanding of the Role of Advertising Message Involvement in Ad Processing," Psychology & Marketing, 10, 301-319.

Lichtenstein, Meryl and Thomas K. Srull (1985), "Conceptual and Methodological Issues in Examining the Relationship between Consumer Memory and Judgment," in Psychological Processes and Advertising Effects, ed. Linda F. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: Lawrence Erlbaum Associates, 113-128.

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), 130-43.

MacKenzie, Scott B. and Richard A. Spreng (1992), "How Does Motivation Moderate the Impact of Central and Peripheral Processing on Brand Attitudes and Intentions?," Journal of Consumer Research, 18 (March), 519-529.

Miniard, Paul W., Sunil Bhatla, and Randall L. Rose (1990), "On the Formation and Relationship of Ad and Brand Attitudes: An Experimental and Causal Analysis," Journal of Marketing Research, 27 (August), 290-303.

Mitchell, Andrew A.(1983a), "The Effects of Visual and Emotional Advertising: An Information-Processing Approach," in Advertising and Consumer Psychology, ed. Larry Percy and Arch G. Woodside, Lexington, MA: D.C. Heath and Company, 197-217.

Mitchell, Andrew A.(1983b), "Cognitive Processes Limited by Exposure to Advertising," in Information Processing Research in Advertising, ed. Richard J. Harris, Hillsdale, NJ: Lawrence Erlbaum Associates, Inc., 13-42.

Mitchell, Andrew A. (1993), "Attitude toward the Advertisement Effects over Time and in Attitude Change Situations," in Advertising Exposure, Memory, and Choice, ed. Andrew A. Mitchell, Hillsdale, NJ: Lawrence Erlbaum Associates, Inc., 209-238.

Mitchell, Andrew A. and Jerry C. Olson (1981), "Are Product Attribute Beliefs the Only Meiator of Advertising Effects on Brand Attitude?" Journal of Marketing Research, 18 (August), 318-332.

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

Petty, Richard E. and John T. Cacioppo (1981), Attitudes and Persuasion: Classic and Contemporary Approaches, Dubuque, Iowa: Wm. C. Brown Company.

Petty, Richard E., Rao H. Unnava, and Alan J. Strathman (1991), "Theories of Attitude Change," in Handbook of Consumer Behavior, ed. Thomas S. Robertson and Harold H. Kassarjian, Englewood Cliffs, NJ: Prentice-Hall, 241-280.

Singh, Jagdip and Garry K. Rhoads (1991), "Boundary Role Ambiguity in Marketing-Oriented Positions: A Multidimensional, Multifaceted Operationalization," Journal of Marketing Research, 23 (August), 328-338.

Zanna, Mark P. and Russell H. Fazio (1982), "The Attitude Behavior Relation: Moving toward a Third Generation of Research," in Consistency in Social Behavior: The Ontario Symposium, Vol. 2, ed. Mark P. Zanna et al., Hillsdale, NJ: Lawrence Erlbaum Associates, 283-301.

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