The Moderating Effect of Attitude Toward an Ad on Ad Effectiveness Under Different Processing Conditions

Julie A. Edell, Duke University
Marian C. Burke, Duke University
ABSTRACT - This paper is based on the proposition that we gain leverage in understanding attitude toward the brand (Ag) if we focus on attitude toward the ad (AAd) as well. Six competing hypotheses regarding the process by which the AAd affects AB are reviewed. We then argue that in order to determine which process is operating we must study the AAdAb relationship over time. Hypotheses concerning the nature of that relationship under each processing condition are proposed.
[ to cite ]:
Julie A. Edell and Marian C. Burke (1984) ,"The Moderating Effect of Attitude Toward an Ad on Ad Effectiveness Under Different Processing Conditions", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 644-649.

Advances in Consumer Research Volume 11, 1984      Pages 644-649

THE MODERATING EFFECT OF ATTITUDE TOWARD AN AD ON AD EFFECTIVENESS UNDER DIFFERENT PROCESSING CONDITIONS

Julie A. Edell, Duke University

Marian C. Burke, Duke University

ABSTRACT -

This paper is based on the proposition that we gain leverage in understanding attitude toward the brand (Ag) if we focus on attitude toward the ad (AAd) as well. Six competing hypotheses regarding the process by which the AAd affects AB are reviewed. We then argue that in order to determine which process is operating we must study the AAdAb relationship over time. Hypotheses concerning the nature of that relationship under each processing condition are proposed.

INTRODUCTION

The effects of persuasive communications on attitude formation and change have intrigued social scientists for years. Marketers have paid special attention to advertising because of its effects on brand attitude, purchase intention, and sales. The latest outcome measure to intrigue marKeters, because of its potential role in mediating the effects of advertising, is attitude toward the ad (AAd). while the notion that consumers' subjective reactions to an advertisement may influence their receptivity to the ad is not new, interest in more carefully specifying the nature or the role of AAd in the communications process is re Cent

The increased interest in attitude toward the ad (considered here to be consumers' evaluations of the advertisement as opposed to evaluations of the product or brand) is due, in part, to the recent evidence that AAd acts as a mediator of advertising effectiveness. This evidence comes from studies which focused on understanding the effects of advertising on consumers' attitude toward the brand (Ag). The research results support the conclusion that consumers' responses to advertisements consist of more than cognitions about the brand. For instance, the multi-attribute research suggests that an important dimension of cognitive structure may be attitude toward the ad (Mitchell and Olson 1981). In addition, as the cognitive response literature evolves, it is increasingly evident that advertising messages evoke thoughts about the ad itself as well as cognitive responses about the brand being advertised (Wright 1973, Belch 1981). For example the categories of cognitive responses used by MacKenzie and Lutz (1982) included ad idea cognitions and ad execution cognitions as well as the traditional brand cognitions. Thus, the interest in AB has led us to an interest in AAd, as a construct which is conceptually distinct from AB (e.g., Shimp 1981, Mitchell and olson 1981) yet is strongly linked to AB. We have identified that the AAd-Ag relationship exists; however our understanding of the nature of the relationship and the underlying process mechanisms is embryonic.

This paper is based on the proposition that we gain leverage in understanding AB if we focus on AAd as well. MacKenzie and Lutz (1982) presented four alternative specifications of the mediating role of AAd (which are reviewed later in this paper). These four models result in six competing hypotheses regarding the process by which AAd and AB influence advertising effectiveness. We review evidence that supports all six hypotheses and then suggest a method by which we should be able to differentiate more clearly among the six. Specifically, we propose that studying AAd and AB over time, preferably with exposures to the ad occurring in a natural environment, should provide evidence for which of the six underlying processes are operating. The six hypotheses we consider here are not meant to constitute an exhaustive list of the processes by which AAd and AB might operate. They do, however, represent the explanations which are offered most often in the literature.

MODELS, HYPOTHESES, AND EVIDENCE

MacKenzie and Lutz (1982) identified four alternative specifications of the manner in which AAd mediates the effectiveness of advertising. Their four models are presented in Figure 1 and will be only briefly summarized here. (Refer to MacKenzie and Lutz 1982 and Lutz, MacKenzie, and Belch 1983 for a full description. )

FIGURE 1

FOUR ALTERNATIVE SPECIFICATIONS OF THE MEDIATING ROLE OF AAD

Model 1

Model 1 postulates a direct, one-way causal influence of AAd on AB and not of AB on AAd or AAd on beliefs. This model has received the most empirical support (Mitchell and Olson 1981, MacKenzie and Lutz 1982, Lutz, MacKenzie and Belch 1983, Moore and Hutchinson 1983). Several different hypotheses have been offered concerning the mechanism by which AAd directly influences AB. Three potential competing hypotheses are Presented here.

Classical Conditioning Hypothesis. The most widely discussed mechanism by which AAd may directly influence AB is classical conditioning. The simple transfer of affect

was suggested by Shimp (1981), and the researchers cited above proposed classical conditioning as one explanation of their results. Under the classical conditioning process, AAd's direct effect on AB occurs because affective responses generated by the ad get transferred to the brand. Thus, the ad (or some element of the ad) acts as the unconditioned stimulus and the brand becomes the conditioned stimulus, eventually eliciting the same affective response as the ad.

The previous research provides only the suggestion that classical conditioning may be operating; the evidence is not conclusive. For instance. classical conditioning requires that the unconditioned and conditioned stimulus be repeatedly paired. Further, because the conditioning is not permanent, the relationship must be reinforced, at least intermittently or extinction will set in. Thus, a true test of the classical conditioning explanation would require that consumers be repeatedly exposed to an ad over some time period. We feel that this time period is probably longer than the typical one-hour laboratory session. In some of the research cited above, consumers did see the at more than once; however, the multiple exposures occurred in a single setting. Further, AAd and AB measures should be taken after a period of no exposure in order to test for the extinction phenomenon of classical conditioning. (Moore and hutchinson (1983) incorporated delayed measures of AAd and Ag, implicitly supporting the extinction aspect of classical conditioning.)

Salient Attribute Hypothesis. A competing hypothesis for the direct effect of AAd on AB is that the ad is a salient attribute of the brand. Following the Fishbein (1967) attitude formation framework, consumers develop the belief (bi) that the ad is associated with the brand. That belief is combined with the evaluation of the ad (bi) to influence AB in the same manner as any other brand attribute. Conceptually this mechanism assumes that consumers consciously associate the ad and the brand. The salient attribute explanation makes a great deal of sense given the experimental method and the nature of the stimuli used in the research cited above. In all cases the research was conducted in a lab, which may increase the salience of the ad (Mitchell and Olson 1981, MacKenzie and Lutz 1982). Often the stimulus was a novel ad for a hypothetical product. Mitchell and Olson (1981) note that in such cases the ad may be the subjects' only source of information about the brand, resulting in a high correlation between AAd and AB. Again, this explanation is merely a hypothesis at this stage of our knowledge of the AAd-Ag relationship, especially if we want to generalize to other than an advertising pretest situation.

Measurement Artifact Hypothesis. The final hypothesis stemming from the finding that AAd affects AB directly is the proposition that AAd actually has no effect. Rather, the effects which have been attributed to AAd may be artifacts of the way AAd and AB have been measured. That is, often both constructs are measured using similar semantic differential scales. Hence, any relationship between the two may be due solely to method variance. Or, it is possible that the AAd measure actually captured evaluations of salient but unmeasured attributes of the brand.

Model 2

The second model proposed by MacKenzie and Lutz is considered to be a cognitive response model of the role of AAd. is model proposes that, in addition to the direct effects on AB, the evaluation of the ad itself (AAd) impacts brand cognition directly, and, thereby, AB indirectly. Holbrook (1978) found support for both paths. Lutz, MacKenzie, and belch (1983) report some support for this model. Their data, however, more strongly support Model 1. Thus, while the evidence for the indirect effect of AAd is not strong, it is not clear that this model should be ruled out. a e conceptual basis for the indirect effect comes, in part, from traditional communications theory concerning the impact of the source of a persuasive message. MacKenzie and Lutz (1982), following Lutz and Swasy (1977), argue that an advertising stimulus can be considered a source. This conceptual framework is too appealing to dismiss at this point. Therefore, the Cognitive Response Hypothesis remains a viable alternative explanation of the mediating effects of AAd.

Model 3

Balance theory (Heider 1946) provides the conceptual basis for the third model of the AAd-Ag relationship. The Balance Theory hypothesis states that there is a reciprocal relationship between AAd and AB, i.e., that there is mutual two-way causation. The basic premise of balance theory is that a person strives to maintain balance among the components of any cognitive unit. Heider's example involves a person-object-event triad. The cognitive unit is balanced if all three pairs of relationships are positive or if two are negative and one is positive (p. 110). The balance theory triad in an advertising situation would be the consumer, an ad, and the brand which is advocated in the ad. Since the ad-brand relationship is always positive a balanced state will exist only if the consumer dislikes both the ad and the brand or likes them both. Therefore, the prediction would be that AAd and AB are highly positively correlated.

MacKenzie and Lutz (1982) note that the relative strengths of the causal flows from AAd to AB or from AB to AAd may vary across consumers and situations. For instance, they speculate that the AB to AAd flow would dominate in the case of a mature brand while the AAd flow AB would dominate for a new brand. Although their stimulus was a new ad for a mature brand, MacKenzie and Lutz did not find strong support for this model. Their measure of Ag, however, was taken after the subjects viewed the test commercial and, therefore, cannot be considered a measure of prior AB, i.e., the brand attitude that might causally affect AAd. Messmer (1979) did find that prior AB significantly mediated the effect of exposure to the ad. He used previously unbroadcast TV ads for well-known products (as did MacKenzie and Lutz) but measured subjects' AB prior to exposure to the commercial. He found that positive prior levels of AB favorably influenced AAd. Thus, the balance theory hypothesis has received partial support.

Model 4

The final model postulates that there is no relationship between AAd and AB but both AAd and AB have a direct effect on Purchase Intention. This hypothesis is based on Howard's (1977) model of routinized response behavior. Howard proposed that there are two attitudinal elements which determine a consumer's purchase intention - the brand concept and impersonal attitude. The brand concept represents the 'subjective meaning' which the consumer gives to the brand. It is composed of perceptions about the brand's attributes and is quite stable. Impersonal attitude is less stable and more situation specific as it represents aspects of the purchase situation which are important but are not enduring aspects of the brand. MacKenzie and Lutz (1982) extend the idea of impersonal attitude to include AAd. The Routine Response Hypothesis states that AAd and AB are independent and both affect purchase intention.

MacKenzie and Lutz (1982) and Lutz, MacKenzie, and Belch (1983) rule out this model. Shimp and Yokum (1980), however, found that AAd directly influenced repeat purchase behavior. They did not report the Ag-choice relationship or the AAd-Ag correlation so it is difficult to use their findings as strong support for the Routine Response Hypothesis. If, in fact, AAd was correlated with purchase and uncorrelated with AB then their findings are consistent with this hypothesis.

Summary

The four models of the causal role of AAd as a mediator of advertising effects yield six hypotheses regarding the underlying processes which the models could represent. Some evidence from past research on the AAd-AB relationship is compatible with each of the hypotheses; some studies, in fact, provide support for more than one. Further, a direct causal flow from AAd to AB (Model l) could result from two different processes or be merely a measurement artifact. Thus, it is not yet possible to ascertain the underlying mechanism by which AAd affects AB. One reason may be that most studies have been conducted at one setting in a lab, which enhances the salience of the at and of any repetition of the ad that may have been part of the design. We feel that in order to test the competing hypotheses, the AAd-Ag relationship must be studied over time with multiple exposures to the ad occurring at a natural rate (rather than, say, ten exposures in a one-hour show). The processes that were Presented above need time to work. For instance, classical conditioning requires repeated pairing of the conditioned and unconditioned stimuli - probably over a longer period than one hour. Therefore, we feel that studying AAd and AB over time provides a way of sorting out which alternative proceSS hypotheses are most plausible.

Hypotheses Concerning the AAd-AB Relationship Over Time

In this section we will explore what each of the theories presented above predicts about how the AAd-Ag relationship changes as the number of exposures to the ad increases and as the period of no exposure to the ad (following repeated exposures) increases. We are interested in being able to distinguish among the hypothesized mechanisms by which AAd impacts AB. Making explicit the way each of these theories predicts the AAd-Ag relationship to change over time provides researchers a method for determining which process is taking place. It is also important to note that the changes we propose below are those we would expect to observe over time across people. The hypotheses do not necessarily apply to an individual.

We will not examine any factors which mediate the effect of AAd. For example, Moore and Hutchinson (1983) examined the ways in which likability of the ad mediated its effect on changes in brand consideration. Rather, this paper assumes that the various mechanisms by which AAd has its effect do not vary systematically with the value of AAd. Further, stating the assumptions with which each process as well as he implications of the theory for changes in the AAd - AB relationship over rime may, in fact, cause us to question whether these often proposed explanations are truly viable.

Classical Conditioning Hypothesis

The essence of the classical conditioning mechanism is that over repeated exposures to the ad the affective response to the ad becomes associated with the brand. When this affective response is elicited by the brand in the absence of the ad, the brand has become the conditioned stimulus. However, if the conditioned response is not reinforced by repeatedly associating the ad and the brand, the response will weaken and be extinguished. Thus, the classical conditioning hypothesis of how AAd impacts AB requires repeated exposures to the ad.

After a single exposure to an ad, there should be little or no correlation between AAd and AB. With repeated exposures to the ad, the brand should begin to elicit the same response as the ad. Classical conditioning research has identified several factors which affect the rate at which the conditioning takes place and its resistance to extinction across a number of situations. Those which are applicable to the transfer of affect from the ad to the brand are the number of exposures to the ad, the time interval between exposures, and the percentage of the ad exposures for the brand which elicit the same affective response. For instance, as the length of time from the last exposure to the ad (reinforced trial) increases, the likelihood of the brand eliciting the affective response declines. The rate at which this decline takes place depends on how well the response was learned.

If classical conditioning is the underlying process by which AAd affects AB, we should expect AAd and AB to be uncorrelated initially, but become increasingly positively correlated as the number of ad exposures increases. Then, as the period of no exposure to the ad increases, the correlation between AAd and AB should decrease due to extinction. Additionally, classical conditioning theory requires that the valence of the unconditioned response (AAd) remain constant regardless of the number of times the ad is seen. Thus, the AAd "score" should not change systematically.

Salient Attribute Hypothesis

The salient attribute notion of how AAd impacts AB is based on Fishbein's (1967) multi-attribute formulation, where AB = n/i=1 biei and postulates that the ad is a salient attribute of the brand. [This hypothesis should be distinguished from an alternative hypothesis which suggests that the probability of the ad being a salient attribute of the brand is a function of the number of exposures to the ad.] Thus, initially the ad is associated with the brand at a positive level. The strength of this initial association between the brand and the ad is not specified. Rather it would depend on the mechanism by which the ad becomes associated with brand. Two scenarios are compatible with the theory. First, the belief that a particular ad is for a given brand may be held with maximal certainty even after just one exposure to the ad. In this situation repeated exposures would not increase the strength of the association but would delay the forgetting which might otherwise reduce the belief strength. With this process operating the bi measure associated with the ad (bAd) would be at the maximum level after a single exposure to the ad and remain maximal over repeated exposures and then begin to decline when the ad is no longer seen.

Alternatively, the salient attribute hypothesis is compatible with a learning, model. Initially, the ad would be weakly associated with the brand. As the number and frequency of the exposures to the ad increase, the ad would become more strongly associated with the brand. This association would weaken as the period of no exposure to the ad increases. The bAd measure would get larger then smaller. The theory sheds no light on how the evaluation of the ad (AAd) would change over time. Rather it gives us a precise formula for how AAd influences AB. This formula is AB = n/i-1 biei + bAd AAd. Since AAd can change in various ways and bAd can take either of two patterns just described, as well as the fact that the belief strength and evaluation of the other salient attributes will change over time, it is not possible to specify the exact nature of the correlation between AB and AAd over time. Therefore, in order to test this hypothesis, one would examine how well the data fits the specified formula. This requires that the belief strengths and the evaluations of all the salient attributes, as well as the overall brand evaluation, be measured over time.

Cognitive Response Hypothesis

The third hypothesis indicates that one's evaluation of the ad will impact the number and type of cognitive responses generated in response to the ad. These cognitive responses then mediate the formation of or changes in AB. The cognitive response theory of brand attitude change assumes a highly involved consumer seeking to process the information contained in the ad. As mentioned above, this model of AAd's impact is based upon Holbrook's (1978) findings that the brand's cognitive structure index was more positive when the ad was evaluated as more convincing, believable, and persuasive.

Holbrook did not collect cognitive response measures. It is not possible to specify from his study the exact nature of the mediation. A more positive cognitive structure could be the result of more support arguments, fewer counter-arguments or some combination of the two.

Holbrook and others (e.g., Mitchell and Olson 1981, Calder and Sternthal 1980) factor analyzed subjects' responses to a set of adjectives which described the ad and, following Osgood, Suci, and Tannenbaum (1957), treated the first factor as the evaluative factor, i.e., AAd. Therefore, in discussing changes in the relationship between AAd and brand cognitions over time we will limit our attention to the traditional evaluative aspect. Additionally, since the generation of brand cognitions occurs primarily in those who are highly involved, the following discussion pertains mainly to them. our specifications of how the cognitive responses mediate AAd's effect on AB is quite speculative. Others (Lutz and Swasy 1977) have proposed that non message cognitions will affect AB directly. However that model has not received empirical support (Swasy 1980).

If an ad's score on evaluation is initially positive, then it is more likely that the brand message will be support argued than counterargued. Upon subsequent exposures to the ad, the ad evaluation score is likely to become more positive because the ad's message is now even more congruent with information stored in memory (due to the earlier support arguing). Likewise, if the ad initially scores low on evaluation, consumers are more likely not to believe the message and to counterargue against it. The process of counterarguing the message makes the beliefs about the brand even more negative. Therefore, upon successive exposures to the ad, the ad's content will be more discrepant resulting in an even lower evaluation of the ad.

The cognitive response hypothesis of AAd's impact over repeated exposures postulates polarization of AAd brand cognitions, and AB. Thus subjects who initially evaluated an ad negatively would race it more negative after repeated exposures. The same holds rue for brand cognitions and Ag. however, the correlations between AAd, brand cognitions, and AS should not charge over multiple exposures if this hypothesis of the process by which AAd impacts AB is supported. The cognitive response theory makes no predictions concerning the AAd-AB relationship after a period of no exposure.

This model emphasizes the importance of measuring subjects' overall ad evaluation as well as the various aspects of the ad Which may be evaluated (e.g., whether the ad is believable, convincing, irritating, and so forth). The results or research by Holbrook (1978), Calder and Sternthal (1980), and Wells, Leavitt, and McConville, (1971) suggest that AAd is a multi-dimensional construct, although typically the first factor to emerge from a factor analysis has been treated as the evaluative factor and, therefore, AAd. None of the earlier research has used an overall measure of AAd. It is quite possible that for those consumers who are highly involved the first factor captures those aspects of the ad which are the basis of the overall evaluation. However, for those consumers who are only casually viewing the ad, the overall evaluation of the ad is likely to be based on other aspects of the ad (e.g., humor, tenderness, excitement) which may impact AB quite differently than does the first factor. Therefore, we need an overall measure of AAd. Without measuring both how the components of AAd change over time and how the overall evaluation of the ad changes we have incomplete information and cannot verify the cognitive response explanation. a is hypothesis is the only one that requires an overall AAd measure and measures of a sufficient number of aspects of the ad to capture the multidimensional nature of AAd. (We suggest, however, that multiple measures be taken in examining an) of the hypotheses.)

Balance Theory Hypothesis

In its simplest form, balance theory implies that the initial attitude formed (either AAd or Ag) will determine al] subsequent attitudes formed. Thus, if one's first encounter with a new brand is via a well-liked ad, then the brand and the ad will be liked until some external event upsets the balance. For instance, if the consumer purchases the brand and finds it to be an unsatisfactory product then imbalance is introduced into the system and will be resolved by either disliking both the ad and the brand or deciding it wasn't such a bad product after all and continuing to like both the brand and the product. Or, if the consumer's evaluation of the ad decreases as exposure to the ad increases, (due to irritation, tedium, etc.), the evaluation of the brand may suffer as well.

Congruity theory (Osgood and Tannenbaum 1955) provides some insight regarding the AAd-Ag relationship. The theory postulates that when a state of incongruity exists (i.e., the relationship between the a person and the object is not as equally polarized as that between the person and the event) the less polarized relationship will change more than the more polarized one.

In order to predict when AAd will affect AB and vice versa, we must know which attitude is more strongly held. When consumers have considerable purchase and usage experience with a brand, their AB would likely be composed of strongly held beliefs about the brand. It is much more likely in this situation that AB influences the evaluation of the ads for that brand than that AAd affects AB. For a brand that a consumer has never evaluated, however, the AB may not be strongly held; and, therefore, an ad which is evaluated very negatively (or positively) could have a large impact on AB. This suggests that, at least for well known products, one must control for PriOr AB in studies of affective responses to ads.

To test the balance theory explanation of the impact of AAd on AB and of AB on AAd We must have measures of AAd and AB over time. Further, we need some indication of the basis for AAd and AB- If AB is strongly held or based on usage experience, AB will influence AAd and AAd will have little influence on subsequent AB. The difference between AB after exposure and AB prior to exposure will not be correlated with AAd but prior AB and AAd will be positively correlated. When AB is formed from only weakly held beliefs or from no beliefs, then the potential for AAd to affect AB is much greater. In this case the AB after the initial exposure to the ad should be more correlated with AAd than the AB taken prior to exposure. Neither balance theory nor congruity theory specify how quickly one's cognitive structure reaches a balanced state. If balance is obtained immediately then the correlation between AAd and AB should be at a maximum level after the initial exposure to the ad. If, however, balance is obtained gradually in a more iterative fashion then the correlation between AAd and AB would become more positive over repeated exposures to the ad. The correlation should remain very high even over the period of no exposure as long as no extraneous events affect the cognitive structure.

Routine Response Hypothesis

In the Howard Model of routinized response behavior, AB and AAd are each hypothesized to impact the purchase intention with no influence on one another. This model is applicable only when the consumer has a well-formed and stable brand concept (Ag). In this model AAd's impact is through the impersonal attitude construct's effect on purchase intentions.

If this theory is correct in explaining AAd's impact we should see several effects over time. First, because AB is stable it should not vary over multiple exposures to the ad or as the length of time from the last exposure increases. Second, because AB is independent of changes in the purchase environment and thus AAd, AAd and AB should be uncorrelated and remain uncorrelated over time. Third, although this theory does not predict how AAd will change over time, any changes that do occur should be reflected in changes in the purchase intention and not in AB.

Measurement Artifact Hypothesis

The final hypothesis proposes that AAd has no impact on AB. Rather, the effects which have been attributed to AAd are only artifacts of the way AAd and AB have been measured in many studies. If this is the case, completing one scale is influencing the completion of the second, so AAd and AB should be equally correlated over time. This would be unrelated to the extent to which the consumer has seen the ad. In order to eliminate this potential explanation for the AAd-Ag relationship, care must be taken in the measurement process. At a minimum, the AAd and AB measures should be separated in space and time. Multiple methods of measuring each construct are also advisable.

It has also been suggested that AAd is not a distinct concept but that it is capturing the evaluations of salient but unmeasured attributes of the brand. Since both the evaluation of these unmeasured attributes and AB could change in various ways over multiple exposures to the ad we are unable to generalize about changes in the pattern of these measures over time. This potential explanation can be eliminated, however, by eliciting salient attributes from each consumer rather than using a common set of attributes for all consumers. This procedure will reduce the probability that attributes which are salient remain unmeasured, reducing the feasibility of this explanation.

TABLE 1

SUMMARY OF HYPOTHESES

DISCUSSION

The previous section described the patterns of the AAd-AB relationship that should hold given that a particular process is driving that relationship. Cross-sectional studies do not provide sufficient information to determine which of several mechanisms is operating. Therefore, the patterns of the AAd-AB relationship which are examined are those which should occur over time. Table 1 provides a summary of the models, assumptions, required measures, and expected effects for each hypothesis. Often the differences are very subtle. As can be seen from examining the last several columns of Table 1, it is difficult to distinguish the various models using any single measure. Looking at the pattern of expected results over several measures can, however, separate the models. For instance, a positive correlation between AAd and AB after initial exposure to the ad is consistent with the salient attribute, cognitive response, balance theory, and measurement artifact hypotheses. Other measures such as AAd over time or bAD over time help to distinguish one process from others.

The importance of Table 1 is threefold: 1) the phenomena one needs to examine over time are clearly specified, thus providing a guide for future research; 2) it points out how to distinguish among the potential moderating mechanisms; and 3) the Table suggests additional analyses that might be done on data from existing research. Thus, a great deal of care must be taken in designing experiments and measuring constructs.

An issue that is critical to studying the AAd-Ag relationship is that the hypothesized processes take place over time. "Over time" involves multiple exposures to the ad (i.e., repetition) at a naturally occurring race, and the passage of real time, including periods of time with no exposure to the ad. Most past studies of AAd (and of repetition), however, have taken place in a lab. This creates very artificiaL exposure conditions and, specificalLy, very unnatural repeated exposure conditions. Thus, makes it difficult to generalize the empirical results beyond an advertising pretest situation. Advertising practitioners have used time as a surrogate for exposure (Grass and Wallace 1970, Appel 1971, Greenberg and Suttoni 1973), and we may need to consider that option in this program of research. Thus, in order for our understanding of the relationship among AAd, AB, and advertising effectiveness to advance we need to begin to triangulate with respect to research methods. Lab studies have provided us with some insight, yet the experimental control provided by the lab studies has been at the expense of external validity. Field studies will allow us to examine some or the time dependent phenomena we have presented here.

Finally, we feel that there is probably not one exclusive mechanism by which AAd affects AB. In fact, the appropriate mechanism may not have been identified yet. The six hypotheses presented in this paper are, at first, appealing intuitively. Yet rigorous consideration of the assumptions that are critical to a particular explanation tend to reduce the appeal of many of the theories. For instance, the classical conditioning hypothesis requires that the valence of AAd remain constant over time. To the extent that this assumption does not fit with our experience or past empirical results, the viability of classical conditioning as the mechanism at work is reduced. Further, we to not expect one process to explain the role of AAd in every situation. We feel, however, that by systematically varying important situational variables (such as the level of involvement, the nature of the ad, or the viewing environment) and extending our inquiry to include field studies we will begin to develop a knowledge base that indicates which processes operate under which conditions.

We hope that the hypotheses presented in this paper will aid in structuring thinking regarding the AAd-Ag relationship and encourage a systematic research program.

REFERENCES

Appel, Valentine (1971), "On Advertising Wearout, Journal of Advertising Research, 11 (February), 11-13.

Belch, George E. (1981), 'An Examination of Comparative and Noncomparative Television Commercials: The Effect of Claim Variation and Repetition on Cognitive Response and Message Acceptance,' Journal of Marketing Research, 18 (August), 333-49.

Calder, Bobby J. and Brian Sternthal (1980), "Television Commercial Wearout: An Information Processing View,' Journal of Marketing Research, 17 (May), 173-186.

Fishbein, Martin (1967), Readings in Attitude Theory and Measurement. New York: Wiley.

Grass, Robert and Wallace H. Wallace (1969), 'Station Effects of Television Commercials," Journal of Advertising Research, 9 (September), 3-8.

Greenberg, Allan and Charles Suttoni (1973), 'Television Commercial Wearout,' Journal of Advertising Research, 13 (October), 47-54.

Heider, Fritz (1946), Attitudes and Cognitive Organization,Journal of Psychology, 21, 107-112.

Holbrook, Morris B. (1978), "Beyond Attitude Structure: Toward the Informational Determinants of Attitude," Journal of Marketing Research, 15 (November), 546-556.

Howard, John A. (1977), Consumer Behavior: Application of Theory. New York: McGraw Hill.

Lutz, Richard J., Scott B. MacKenzie, and George E. Belch (1983), "Attitude toward the Ad as a Mediator of Advertising Effectiveness: Determinants and Consequences, in Advances in Consumer Research, Vol. 10, eds.

Lutz, Richard J. and John Swasy (1977), "Integrating Cognitive Structure and Cognitive Response Approaches to Monitoring Communication Effects," in Advances in Consumer Research, Vol. 4, William Perreault, Jr., ed. Atlanta: Association for Consumer Research, 363-371.

MacKenzie, Scott B. and Richard J. Lutz (1982), "Monitoring Advertising Effectiveness: A Structural Equation Analysis of the Mediating Role of Attitude toward the Ad.working paper # 117, Center for Marketing Studies. UCLA.

Messmer, Donald J. (1979), "Repetition and Attitudinal Discrepancy Effects on the Affective Response to Television Advertising," Journal of Business Research, 7, 75-93.

Mitchell, Andrew A. and Jerry C. Olson, (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitudes?" Journal of Marketing Re- search, 18 (August), 318-322.

Moore, Danny L. and J. Wesley Hutchinson, (1983), "The Effects of Ad Affect on Advertising Effectiveness,-' in Advances in Consumer Research, Vol. 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Association for Consumer Research, 526-531.

Osgood, C. E., G. J. Suci and P. H. Tannenbaum (1957), The Measurement of Meaning, Urbana: University of Illinois Press.

Osgood, C. E. and P. H. Tannenbaum (1955), "The Principle or Congruity in the Prediction of Attitude Change," Psychological Review, 62, 42-55.

Shimp, Terence A. (1981), "Attitude Toward the Ad as a Mediator or Consumer Brand Choice, Journal of Advertising, 10 (2), 9-15. and J. Thomas Yokum (1980), "The Influence of Deceptive Advertising on Repeat Purchase Behavior, working paper #86, University of South Carolina, Columbia, SC.

Shimp, Terence A. and J. Thomas Yokum (1980), "The Influence of Deceptive Advertising on Repeat Purchase Behavior," working paper #86, University of South Carolina, Columbia, SC.

Swasy, John L. (1980), "Conceptual and Modeling Issues in a Cognitive Response Perspective on Persuasion Processes," unpublished doctoral dissertation, University of California, Los Angeles, CA.

Wells, William D., Clark Leavitt, and Maureen McConville, (1971), "A Reaction Profile for Commercials," Journal of Advertising Research. 11 (December), 11-17.

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

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