The Impact of Measurement Context on the Relationship Between Attitude Toward the Ad and Brand Attitude For Familiar Brands

Karen A. Machleit, University of Cincinnati
Arti Sahni, University of Cincinnati
ABSTRACT - This study investigates the possibility that measurement context (the material which precedes a question on a questionnaire) can affect the observed relationships between Aad and brand attitude. The data illustrate that for familiar brands in particular, the magnitude of the Aad/brand attitude relationship can be affected by measurement context. This study demonstrates the need for careful questionnaire construction to reduce concerns regarding the internal validity of experiments measuring both Aad and brand attitude.
[ to cite ]:
Karen A. Machleit and Arti Sahni (1992) ,"The Impact of Measurement Context on the Relationship Between Attitude Toward the Ad and Brand Attitude For Familiar Brands", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 279-283.

Advances in Consumer Research Volume 19, 1992      Pages 279-283

THE IMPACT OF MEASUREMENT CONTEXT ON THE RELATIONSHIP BETWEEN ATTITUDE TOWARD THE AD AND BRAND ATTITUDE FOR FAMILIAR BRANDS

Karen A. Machleit, University of Cincinnati

Arti Sahni, University of Cincinnati

ABSTRACT -

This study investigates the possibility that measurement context (the material which precedes a question on a questionnaire) can affect the observed relationships between Aad and brand attitude. The data illustrate that for familiar brands in particular, the magnitude of the Aad/brand attitude relationship can be affected by measurement context. This study demonstrates the need for careful questionnaire construction to reduce concerns regarding the internal validity of experiments measuring both Aad and brand attitude.

INTRODUCTION

The primary focus of the literature addressing attitude toward the ad (Aad) has been on determining the effects of Aad on brand attitude (Ab); the empirical evidence strongly supports a positive relationship between the constructs (e.g., Batra and Ray 1985; Gardner 1985; MacKenzie, Lutz and Belch 1986; Mitchell and Olson 1981). Recently, there has been some interest in studying the moderating impact of brand familiarity on this relationship. The findings regarding this issue, however, have been somewhat inconsistent. For example, Gresham and Shimp (1985) found a significant effect of Aad on Ab for only six of fifteen familiar brands. Edell and Burke (1986) report that for highly familiar brands, the effect of prior Ab on Ab was greater than the effect of Aad on Ab. In addition, the effect of Aad on Ab was weaker for the highly familiar brands than for brands with low familiarity. Phelps and Thorson (1991) report a similar finding; that the effect of Aad on Ab was stronger for unfamiliar brands than for familiar brands when prior brand attitude was controlled. Yet, Machleit and Wilson (1988) found no relationship between Aad and Ab for two familiar brands after controlling for prior Ab. Similarly, Machleit, Madden, and Allen (1990) report nonsignificant effects of Aad and positive affect on brand attitude for two familiar brands. This study is designed to address the inconsistent Aad/Ab findings for familiar brands by investigating the degree to which measurement procedures might impact the relationship. It has been demonstrated that item context (the material which precedes an item in the questionnaire) can have an impact on the responses subjects provide to that item (cf. Bradburn 1982; Schuman and Presser 1981; Tourangeau and Rasinski 1988). More specifically, correlations between constructs can be increased when they are measured contiguously (Feldman and Lynch 1988). Indeed, this is a problem which researchers have recognized in the Aad context (e.g., Edell and Burke 1986; Mitchell 1986). Edell and Burke (1984) caution that "(a)t a minimum, the Aad and Ab measures should be separated in space and time" (p. 648) to insure that method variance can be eliminated as a possible explanation for the Aad/Ab relationship. Unfortunately, by reviewing the method section of many Aad studies, one can find that such precautions are rarely undertaken. Thus, the purpose of this study is to demonstrate that context effects in measurement can increase the Aad/Ab association for familiar brands and may offer a potential explanation for the inconsistent findings regarding Aad's impact on Ab. We begin with a brief review of the relevant literature before proceeding with the hypotheses.

BACKGROUND

Moderating Effects of Brand Familiarity

The literature offers many studies demonstrating Aad's impact on brand attitude for unfamiliar and fictitious/hypothetical brands. The theoretical justification typically offered for this relationship involves a classical conditioning and/or direct affect transfer explanation (e.g., Gardner 1985; Gresham and Shimp 1985, Machleit and Wilson 1988; MacKenzie, Lutz, and Belch 1986; Mitchell and Olson 1981). However, this theory does not support predictions of an Aad/Ab effect for familiar brands.

McSweeney and Bierley (1984) review classical conditioning procedures and their implications for consumer behavior. They note that "it will be easier to classically condition behaviors to new products (CSs) than to products with which people have had previous experience" (p. 624). Their conclusion is based on a conditioning principle called "latent inhibition" which contends that little conditioning will occur when the CS (in this case, the brand) is presented several times without the US (in this case, the ad and resulting affective responses). Thus, when consumers already have some sort of knowledge and/or experience with a familiar brand and they are subsequently exposed to an ad for that brand, little conditioning/affect transfer should occur. This principle has been demonstrated empirically by Shimp, Stuart, and Engle (1991) who, in a series of experiments, found that for highly familiar brands of cola, little conditioning effects were observed in brand attitude. Indeed, this finding is not surprising given that attitudes for familiar, experienced individuals are difficult to change (Fazio and Zanna 1981; Wu and Shaffer 1987). One would expect that it would take more than just one or two exposures to affect-producing ads to change such a well-formed attitude.

The Impact of Context Effects

The process of measurement and the context of questioning have been demonstrated to impact the correlations among constructs by shaping the computational and retrieval processes that individuals use to generate answers to survey questions. Feldman and Lynch (1988) present an "Accessibility-Diagnosticity" model that can contribute to our understanding of the effects of measurement context. According to their model, the likelihood that a person's response to a question will be used as a basis for a response to a second question depends upon the accessibility of the first input in memory, the accessibility of alternative inputs in memory, and the perceived diagnosticity or relevance of the input and alternative inputs (Lynch, Marmorstein, Weigold 1988).

It is not uncommon for respondents in a data collection situation to have a relatively low interest in the survey task, as well as to feel some time pressure to complete the questionnaire. As a result, the response process performed by the individual is likely to be carried out superficially, and respondents will often base their opinions on easily accessed situational cues from the immediate context rather than retrieving the appropriate information from memory (Tourangeau and Rasinski 1988). Even without time pressure or low involvement of the part of the respondent, the measurement process may direct the respondent's attention to subsets of their prior knowledge structure. Through output interference, "cognitions activated in the process of making the first judgment may suppress the retrieval of cognitions that would have influenced the second judgment or behavior in the absence of prior measurement" (Feldman and Lynch 1988, p. 422). Further, multiple measures of a construct (necessary for increased measurement reliability) will magnify this problem since they provide increased rehearsal and thus increased memory accessibility (Feldman and Lynch 1988). Tourangeau and Rasinski (1988) note that the retrieval process will reflect what is most accessible to the respondent rather than what is most important. This problem "is likely to be heightened in attitude measurement settings, in which few respondents have either the motive or the opportunity to reflect carefully on their answers. . . what is most readily retrieved from memory does not necessarily reflect either reality or the contents of memory" (p. 301).

HYPOTHESES

Expanding on the "accessibility" portion of their model, Feldman and Lynch (1988) note that separating two mildly similar responses (e.g., responses about Aad and brand attitude) on a questionnaire can reduce the likelihood that the respondent will retrieve the first response as direct input to the second. When making a judgment, subjects begin by searching working memory; the search concludes when a sufficient basis for making the judgment is retrieved. By including "filler" or intervening items between the two measures, it is less likely that the response to the first question will still be in working memory when the second question is encountered. The respondent must then rely on long-term memory (Feldman and Lynch 1988).

However, for the unfamiliar brand, the only information that the respondent has about the brand comes from the ad to which s/he was exposed. This is the only information which the respondent has available for use in forming a judgment or attitude about the brand. Thus, this limited information should be relevant in forming brand attitude regardless of whether it was made salient and in working memory through contiguous measurement or whether the brand attitude was taken after some "filler" questions. Given this limited information, we anticipate that measurement context effects induced by measuring Aad and Ab contiguously in a questionnaire will have minimal effects for the unfamiliar brand. Thus, we hypothesize:

Hypothesis 1: For unfamiliar brands, Aad will significantly impact brand attitude regardless of measurement context.

For the familiar brand, respondents should already have an attitude in long-term memory that will serve as the basis for attitude questioning. However, Tourangeau and Rasinski (1988) report that "(e)ven when the underlying attitude structure is stable, the response process need not be very reliable" (p. 301). This unreliability in the retrieval process will be reflective of the ease in accessing information for generating a response. When Aad and Ab questions are presented contiguously, working memory provides convenient Aad cues for basing responses of brand attitude; when the questions are presented noncontiguously, subjects will be more likely to engage in the more involved and effortful search of long-term memory (Tourangeau and Rasinski 1988). As a result of processing ease, we hypothesize that, for familiar brands, measurement context effects will produce a significant correlation between Aad and Ab when the measurements are taken contiguously; however, with noncontiguous measures, we hypothesize that the Aad/Ab relationship will be, consistent with theory, nonsignificant.

Hypothesis 2: For familiar brands, Aad will significantly impact brand attitude only when the measures are taken contiguously.

METHODOLOGY

Experimental Design

A 2x2 between subjects design (familiar vs. unfamiliar brands x contiguous vs. noncontiguous measures) was conducted. Data were collected within this design for two different product categories relevant to student subjects (soft drinks and beer) to enable a replication of the findings. Product category was not a factor in the design since we simply wanted to insure and demonstrate that the results would be replicable in different product category conditions.

Stimuli

Four commercials were selected as the test stimuli. For soft drinks, ads for Coca-Cola Classic and Mug root beer were selected; for beer, the ads were for Michelob and Kronenbourg brands. The Coke commercial was a 1950s style song and dance production, the Michelob ad included singing artist Steve Winwood, the Mug ad was a humorous animated sketch, and the Kronenbourg ad was a humorous discussion of the European brand. These ads were selected for their ability to generate some type of affective response from the subjects, and they fit nicely into the familiar (Coke and Michelob) and unfamiliar (Mug and Kronenbourg) factor necessary in the experimental design. Recognizing the limitations inherent in using "real" ads, we concluded that they would produce responses more natural than what would have been elicited through mock commercials (Mitchell 1986).

The test ads were embedded in a ten minute segment of a game show to mimic a more natural viewing environment. The first commercial break included three non-affective "filler" ads. During the second commercial break, subjects were exposed to either the two ads for the familiar brands, or the two ads for the unfamiliar brands. These test ads were then followed by a filler ad.

Subjects and Procedure

Student subjects were processed during undergraduate marketing courses. One of the two videotapes was shown to the students and they were told that they would be asked some questions when the tape concluded. Unknown to the students, questionnaires including either the contiguous or noncontiguous measures were randomly ordered and distributed. A total of 88 subjects participated: 29 in the familiar-contiguous condition, 21 in the familiar-noncontiguous condition, 20 in the unfamiliar-contiguous condition, and 18 in the unfamiliar-noncontiguous condition. A manipulation check indicated the success of the brand familiarity manipulation; on a 1-7 semantic differential scale, Coke and Michelob were rated as familiar (mean values of 6.7 and 5.3, respectively) and Mug and Kronenbourg were rated as unfamiliar (mean values of 1.8 and 2.1, respectively).

Questionnaire Design and Measures

The questionnaire began with the measures of positive affect and Aad validated by Madden, Allen, and Twible (1988). Positive affect was measured by asking subjects reflect how they felt during the ad and respond to five adjectives such as "cheerful" and "pleased" on a six point scale (with endpoints "very much so" to "not at all"). Aad was then measured with a seven-point semantic differential format. The scale items included: unpleasant/pleasant, likeable/unlikeable, interesting/boring, good/bad, tasteful/tasteless, artless/artful.

In the contiguous condition, the Aad measure was followed by the brand attitude measure. Brand attitude was measured with the items good/bad, unfavorable/favorable, and like/dislike with a seven-point semantic differential format. Three pages of "filler" questions then followed. In the noncontiguous condition, the "filler" questions followed the Aad measure, then the brand attitude measure was taken. The "filler" questions (asked with several different scale formats) included a variety of questions regarding the subjects' television viewing preferences and habits.

Coefficient alpha values were high for the positive affect, Aad, and brand attitude measures for each of the four brands. They ranged from .84 to .97 and demonstrate the reliability of the measures.

RESULTS AND DISCUSSION

Table 1 presents the correlations between Aad and brand attitude for the familiar and unfamiliar brands in both the contiguous and noncontiguous conditions. Hypothesis 1 predicted significant Aad/Ab correlations regardless of measurement context for the unfamiliar brands. This hypothesis is supported; the correlations for Mug and Kronenbourg are highly significant in both the contiguous and noncontiguous conditions.

These data were further examined to determine whether the correlations in the contiguous condition were significantly higher than those in the noncontiguous condition (Snedecor and Cochran 1980). For Mug, the correlations are not significantly different (.71 vs. .63). For Kronenbourg, however, the difference (.89 vs. .65) approaches significance (p=.0524). We argued earlier that Aad should affect Ab regardless of measurement context since these were affective ads and the ad provided the only basis from which to form a brand attitude. However, in the Kronenbourg ad, the brand was compared to Heineken beer. Thus, in the noncontiguous measure condition, the respondent may have accessed information regarding Heineken from long-term memory, using this information as input into brand attitude and thereby reducing the Aad effect.

Hypothesis 2 predicted significant Aad/Ab correlations for the familiar brands only when the measures are taken contiguously. The data for Michelob support this hypothesis. The high correlation (r=.65) in the contiguous condition does not occur in the noncontiguous condition (r=.07); these correlations are significantly different at p < .05.

It has been suggested that the significant effects of Aad on Ab for familiar brands could simply be spurious and due to the fact that prior brand attitude may act as a common antecedent to both Aad and Ab (Machleit and Wilson 1988). To determine that the effect for Michelob in the contiguous condition was likely due to measurement context rather than a lack of control for prior brand attitude, brand attitude was regressed on both Aad and prior brand attitude (this measure was included with questions for an unrelated experiment one week earlier). The results indicate that while prior brand attitude significantly impacts brand attitude (beta=.52, p<.0001), the impact of Aad on brand attitude (beta=.62, p<.0001) was not mediated by prior Ab. We therefore conclude that measuring Aad and brand attitude contiguously can inappropriately result in a significant correlation between the constructs.

For Coke, however, the results were not as anticipated. The Aad/Ab correlations were nonsignificant regardless of the measurement context (see Table 1). In an effort to post hoc explain this finding, we began by looking for differences between the two familiar brands, Coke and Michelob. Brand familiarity was measured on a semantic differential scale (7=highly familiar) for both brands. ANOVA results indicate that the respondents were significantly less familiar (p<.0001) with Michelob than with Coke (mean values: Michelob=5.3, Coke=6.7). Tourangeau and Rasinski (1988) note that "with highly familiar issues, an attitude structure may be activated automatically when the issue is confronted" (p. 309, emphasis ours), and context effects may be eliminated. Further, issue expertise may reduce susceptibility to context effects (Fiske and Kinder 1981); such "experts" are likely to have thorough retrieval processes and, as a result, are less likely to be affected by measurement context. Given that the mean familiarity value for Coke was virtually at the extreme end of the scale, we conclude that this very high level of familiarity and expertise beyond that of Michelob is the reason that different effects were observed for the two familiar brands.

TABLE 1

AAD/AB CORRELATIONS

TABLE 2

POSITIVE AFFECT/AB CORRELATIONS

Since the Aad and Ab measures contained similar scale formats, we wished to consider whether method variance played a role in the study. Feldman and Lynch (1988) indicate that shared method variance, like context effects, can be minimized by increasing the amount of unrelated material separating the questions. However, we hoped to discount the method variance explanation in favor of the context effect explanation by evaluating whether context effects were also present for a different type of scale format. The positive affect measure was taken along with the Aad measure although it was not measured with the similar semantic differential scale format. Table 2 presents the correlations between positive affect and brand attitude for the four brands by question context. Note that while these correlations as a whole are lower in magnitude (as we would anticipate since positive affect may precede Aad in a causal chain leading to brand attitude (e.g., Batra and Ray 1986; Edell and Burke 1987; Machleit and Wilson 1988)), the pattern of results is similar to the Aad results. In particular, for the moderately familiar brand Michelob, the correlation in the contiguous condition is highly significant while the correlation in the noncontiguous condition is not. Thus, the data in Table 2 also demonstrate the impact of measurement context.

CONCLUSION

This study underscores the importance of questionnaire design and contextual factors when researching associations among constructs. This issue should be an important concern for those who attempt to understand how affect-producing ads can impact brand attitude. In particular, we demonstrate that when researching familiar brands, measurement context has the potential to seriously inflate the magnitude of the relationship between Aad (or positive affect) and brand attitude. Researchers measuring the Aad/Ab relation for familiar brands should take care that these measures are separated by filler questions to reduce concerns regarding the internal validity of their experiment.

REFERENCES

Batra, Rejeev and Michael L. Ray (1985), "How Advertising Works at Contact," in Psychological Processes and Advertising Effects: Theory, Research and Application, Linda F. Alwitt and Andrew A. Mitchell, (eds.), Hillsdale, NJ: Lawrence Erlbaum Associates, 13-43.

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

Bradburn, N. (1982), "Question-Wording Effects in Surveys," in Question Framing and Response Consistency, R. Hogarth (ed.), San Francisco: Jossey-Bass, 65-76.

Edell, Julie A. and Marian C. Burke (1984), "The Moderating Effect of Attitude Toward an Ad on Ad Effectiveness under Different Processing Conditions," in Advances in Consumer Research, 11, T.C. Kinnear (ed.), 644-649.

Edell, Julie A. and Marian C. Burke (1986), "The Relative Impact of Prior Brand Attitude and Attitude Toward the Ad on Brand Attitude After Ad Exposure," in Advertising and Consumer Psychology, 3, Jerry Olson and Keith Sentis, (eds.), New York: Praeger Publishers, 93-107.

Edell, Julie A. and Marian Chapman Burke (1987), "The Power of Feelings in Understanding Advertising Effects," Journal of Consumer Research, 14 (December), 421-433.

Fazio, Russel M. and Mark P. Zanna (1981), "Direct Experience and Attitude-Behavior Consistency," Advances in Experimental Social Psychology, (14), 161-202.

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.

Fiske, S. T., and D. R. Kinder (1981), "Involvement, Expertise, and Schema Use: Evidence from Political Cognition" in Personality and Social Interaction, N. Cantor and J. Kihlstrom (eds.), Hillsdale, NJ: Erlbaum, 171-192.

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

Gresham, Larry G. and Terence A. Shimp (1985), "Attitude Toward the Advertisement and Brand Attitudes: A Classical Conditioning Perspective," Journal of Advertising, 14 (1), 10-17, 49.

Lynch, John G. Jr., Howard Marmorstein, and Michael F. Weigold (1988), "Choices from Sets Including Remembered Brands: Use of Recalled Attributes and Prior Overall Evaluations," Journal of Consumer Research, 15 (September), 169-184.

Machleit, Karen A., Thomas J. Madden, and Chris T. Allen (1990), Measuring and Modeling Brand Interest as an Alternative Aad Affect With Familiar Brands," in Advances in Consumer Research, 17, Marvin E. Goldberg, Gerald Gorn and Richard W. Pollay (eds.), 223-229.

Machleit, Karen A. and R. Dale Wilson (1988), "Emotional Feelings and Attitude Toward the Advertisement: The Roles of Brand Familiarity and Repetition," Journal of Advertising, 17 (3), 27-35.

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.

Madden, Thomas J., Chris T. Allen, and Jacqueline L. Twible (1988), "Attitude Toward the Ad: An Assessment of Diverse Measurement Indices Under Different Processing 'Sets'," Journal of Marketing Research, 25 (August), 242-252.

McSweeny, Frances K. and Calvin Bierley (1984), "Recent Developments in Classical Conditioning," Journal of Consumer Research, 11 (September), 619-631.

Mitchell, Andrew A. (1986), "Theoretical and Methodological Issues in Developing an Individual Level Model of Advertising Effects" in Advertising and Consumer Psychology, Jerry C. Olson and Keith Sentis (eds.), New York: Praeger Publishing Co., 172-196.

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

Phelps, Joseph and Esther Thorson (1991), "Brand Familiarity and Product Involvement Effects on the Attitude Toward an Ad-Brand Attitude Relationship," in Advances in Consumer Research, 18, Rebecca Holman and Michael R. Solomon (eds.), 202-209.

Schuman, H. and Presser, S. (1981), Questions and Answers in Attitude Surveys: Experiments in Question Form, Wording and Context, New York: Academic Press.

Shimp, Terence A., Elnora W. Stuart and Randall W. Engle (1991), "A Program of Classical Conditioning Experiments Testing Variations in the Conditioned Stimulus and Context," Journal of Consumer Research, 18 (June), 1-12.

Snedecor, George W. and William G. Cochran (1980), Statistical Methods, Seventh Edition, Ames, IA: Iowa State University Press.

Tourangeau, Roger and Kenneth A. Rasinski (1988), "Cognitive Processes Underlying Context Effects in Attitude Measurement," Psychological Bulletin, 103 (3), 299-314.

Wu, Chenghuan and David R. Shaffer (1987), "Susceptibility to Persuasive Appeals as a Function of Source Credibility and Prior Experience With the Attitude Object," Journal of Personality and Social Psychology, 52 (4), 677-688.

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