Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitude? Author(s

Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitude?Author(s): Andrew A. Mitchell and Jerry C. OlsonSource: Journal of Marketing Research, Vol. 18, No. 3 (Aug., 1981), pp. 318-332Published by: American Marketing AssociationStable URL: http://www.jstor.org/stable/3150973 .

Accessed: 06/11/2014 07:06

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access toJournal of Marketing Research.

http://www.jstor.org

This content downloaded from 200.89.68.193 on Thu, 6 Nov 2014 07:06:56 AMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=ama

http://www.jstor.org/stable/3150973?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


ANDREW A. MITCHELL and JERRY C. OLSON*

Fishbein's attitude theory posits that beliefs are the only mediators of attitude formation and change. The validity of this proposition for consumers' beliefs about product attributes and brand attitudes was exmained in the context of an advertising effects study. To manipulate product attribute beliefs and to create settings in which other mediation processes might occur, the authors exposed subjects to simple advertisements that contained either a verbal claim or visual information. Level of repetition also was varied. As expected, product attribute beliefs mediated attitude formation. However, another variable, termed attitude toward the advertisement, also mediated brand attitudes and purchase intentions. The authors discuss alternative explanations for the results

and offer suggestions for future research.

Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand

Attitude?

Attitude, defined here as an individual's internal evaluation of an object such as a branded product, has been an important concept in marketing research for the past 20 years. There are at least two major reasons for this long-term interest. First, attitudes often are considered relatively stable and enduring predisposi- tions to behave. Consequently, they should be useful predictors of consumers' behavior toward a product or service. Second, social psychology has provided several theoretical models of the attitude construct (e.g., Fishbein 1963; McGuire 1968; Rosenberg 1956; Triandis 1971; Wyer 1974). These conceptual frame- works, especially Fishbein's, have stimulated much of the attitude research in marketing. Most of this work, however, has been descriptive and pragmatic in orientation; theoretical issues have been relatively neglected.

To increase the usefulness of the attitude construct,

*Andrew A. Mitchell is Associate Professor of Marketing, Carne- gie-Mellon University. Jerry C. Olson is Associate Professor of Marketing, The Pennsylvania State University. The research was funded by a Research Initiation grant to both authors from The Pennsylvania State University and an NSF grant to the first author. The authors thank Meryl Gardner and S. P. Raj for their assistance in data analysis and three reviewers for their constructive comments.

marketers must develop a clearer understanding of the causal determinants of attitude formation and change (Olson and Mitchell 1975; also see Lutz 1977; Lutz and Bettman 1977). These causal influences are of critical importance because they mediate the effects of marketing decision variables such as advertising or price on consumer attitudes. Thus, a better understanding of the causal dynamics of attitude formation would not only aid marketing researchers in measuring the attitudinal impact of marketing variables, but also help managers develop more effective marketing strategy.

ATTITUDE THEORY

Fishbein (1963, 1967; Fishbein and Ajzen 1975) presented perhaps the clearest theoretical exposition of the causal basis of attitudes. According to Fishbein and Ajzen (1975, p. 222), "A person's attitude is a function of his salient beliefs at a given point in time." Beliefs are the subjective associations between any two discriminable concepts. Salient beliefs are those activated from memory and "considered" by the person in a given situation (Fishbein and Ajzen 1975; Olson, Kanwar, and Muderrisoglu 1979). As these ideas are operationalized in the typical marketing research study, the attitude concept of interest is a brand and the related concepts are product attributes.

318

Journal of Marketing Research Vol. XVIII (August 1981), 318-32

I ?

I



PRODUCT ATTRIBUTE BELIEFS AND BRAND ATTITUDE

That is, marketing researchers have been mainly concerned with consumers' beliefs about attributes of a brand.

Fishbein's now-familiar attitude model specifies the relationship between the set of salient beliefs about a concept (often termed cognitive structure) and an overall evaluation of, or attitude toward, the concept.

n

( b,e, = A, i=l

where:

b, = the strength of the association between the attitude concept, o, and the ith salient concept,

e, = the evaluation of the ith salient concept, Ao = the overall evaluation of, or attitude toward, con-

cept o, and n = the number of salient beliefs.

Fishbein clearly intended the attitude model in equation 1 to describe only the predicted relationship between measures of the theoretical constructs (e.g., see Fishbein and Ajzen 1975, p. 222-3). That is, the algebraic model provides a means of estimating the belief-attitude relationship that was formed as a result of causal mechanisms. The model itself does not "state" the causal proposition nor does it specify the causal mechanisms.

Fishbein also proposed that the attitude-belief relationship specified in equation 1 holds for attitudes toward a specific behavior such as buying a product, Aact. However, the set of salient beliefs may not be the same as for Ao (Fishbein and Ajzen 1975). In turn, attitude, especially Aact, is presumed to have a causal influence on behavioral intentions (BI).' In sum, Fishbein proposed a causal flow among three cognitive variables-beliefs, evaluations or attitudes, and intentions (see Lutz 1977). According to this view, a marketing stimulus such as an advertisement affects consumers' beliefs first. Then the influenced salient beliefs mediate the marketing variable's effect on attitude, and attitude in turn mediates subsequent effects on behavioral intention.

The basic theoretical proposition of Fishbein's attitude theory is that beliefs cause attitude. Because attitude is determined by a set of salient beliefs, changes in attitude must be mediated by changes in those beliefs. Therefore, to change a person's attitude toward a concept, one must modify the salient beliefs about that concept (Fishbein and Ajzen 1975, Ch.

'Actually, Fishbein's extended attitude model assumes that behavioral intentions are a function of both A^,c and a social norm construct (Fishbein and Ajzen 1975). Recent research findings, however, raise the question of whether the social norm element is really a separate construct (Miniard and Cohen 1979). Because of these operational and conceptual problems, the social norm construct is not included in our research.

6). Beliefs can be modified by changing the strength of a salient belief (be), changing the evaluation of a belief (e,), creating a new salient belief, or making a salient belief nonsalient (Lutz 1975, 1977). Fishbein's algebraic model (equation 1) then can be used to estimate the subsequent effect on overall attitude. Our key concern is with the proposition that beliefs cause attitudes. Stated in its strongest sense, the presumption is that beliefs mediate all effects on attitude formation or attitude change.

Developing an empirical test of this causal relationship is straightforward, at least logically. Assume that a persuasive message has a significant effect on attitudes. Assume further that we can identify and measure all salient beliefs. If Fishbein's theory that beliefs mediate effects on attitude is correct, we should find that statistically removing the effect of the message on beliefs (e.g., by analysis of covariance procedures) also removes the significant message effect on attitude. To the extent that the message still has significant effects on attitude, other causal processes must have occurred (or else measurement errors, etc., are present). In reality, problems arise because empirical data cannot always be interpreted as rigorous tests of the theory (cf. Carnegie-Mellon University Marketing Seminar 1978; Dickson and Miniard 1978; Lutz 1978a, b). Perhaps the most difficult problem is demonstrating that all of the salient beliefs, and only the salient beliefs, were measured. Another problem is unambiguously demonstrating the temporal ordering of the effects. Perhaps because of these difficulties, very few studies in either psychology or marketing have addressed the validity of the basic proposition that beliefs are the sole determinant of attitudes. That issue motivated the research reported here.

PAST A TTITUDE RESEARCH IN MARKETING

Much of the marketing research based on the Fish- bein model has focused on demonstrating statistically significant correlations between a "direct" measure of brand attitude (e.g., A,) and the cognitive structure index of attitude (Ebiei) based on product attribute beliefs (see Wilkie and Pessemier 1973). Most of the studies obtained statistically significant correlations. Moreover, when the model constructs were measured correctly (Cohen, Fishbein, and Ahtola 1972), the resulting correlations were reasonably high, ranging from .40 to .70 (e.g., Lutz 1977). Considerable research effort has been directed at determining whether various modifications of the basic model are empirically "better" than the Fishbein formulation, even though the appropriate criteria for selecting one model over another are somewhat ambiguous. In general, this research has emphasized measurement issues and predictive validity (Lutz and Bettman 1977).

Recently, other interests related to Fishbein's attitude model have begun to emerge. Several researchers

319

(1)



JOURNAL OF MARKETING RESEARCH, AUGUST 1981

have suggested that this model provides a theoretically integrated set of measures of the cognitive effects of marketing variables, particularly advertising (Lutz 1975; Mazis and Adkinson 1976; Olson and Mitchell 1975). In applying this approach, beliefs about attributes of the advertised brand, brand attitudes, and purchase intentions are measured to indicate the multiple effects of a particular communication message on cognitive structure variables (e.g., Olson and Dover 1978b). The measures of brand attribute beliefs seem particularly useful in providing diagnostic information about the effectiveness of a message strategy.

A second emerging interest of a few researchers is the theoretical basis for attitude (cf. Calder 1975, 1978; Holbrook 1978; Lutz 1975; Lutz and Swasy 1977; Mitchell, in press; Olson and Dover 1978a; Olson and Mitchell 1975). Although massive correlational evidence supports the static relationship between beliefs and attitudes (cf. Fishbein and Ajzen 1975; Wilkie and Pessemier 1973), experimental evidence support- ing the causal, mediating effect of beliefs in attitude formation and change is meager (Lutz and Bettman 1977). Lutz (1975), for instance, demonstrated that manipulations of product attribute belief strength were mirrored by corresponding shifts in Ao. Olson and Dover (1976) created changes in the strength of product attribute beliefs that were consistent with changes in brand attitudes. However, few other studies have examined the causal influence of product attribute beliefs on brand attitudes. In fact, relatively few marketing researchers have even included beliefs, attitudes, and intentions as multiple dependent variables in the same study.

RESEA R CH PURPOSE

The purpose of our study was to examine the validity of the basic theoretical proposition of the Fishbein attitude theory in a marketing research context. Spe- cifically, we were interested in whether beliefs about product attributes are the only mediator of brand attitude. The experiment involved exposing subjects to different advertisements intended to create different brand attitudes. Because the brands were fictitious, consumers had no prior brand beliefs or attitudes. Thus, the research concerned only belief and attitude formation, not change. If Fishbein's theory is correct, the advertisements should also create differences in brand attribute beliefs. Moreover, the variation in these beliefs should account for (i.e., mediate) all the experimentally produced variation in brand attitudes.

To enhance the inferential logic of the study (Platt 1964), we also examined two alternatives to the beliefs-cause-attitude model. Both imply that an attitude toward a concept may be formed (or changed) without the corresponding formation of salient beliefs about the concept (or changes in those beliefs). One alternative proposition suggests that sheer repetition of a stimulus may cause changes in an individual's attitude toward that stimulus. Since Zajonc's (1968) original

research suggested that repeated exposure enhances attitudes, there has been considerable controversy about the necessary conditions for these effects (e.g., Matlin 1970; Stang 1975) and their theoretical basis (e.g., Berlyne 1970; Harrison 1968; Sawyer 1977). However, substantial evidence supports their exis- tence (Harrison 1977). Recently, Wilson (1979) demonstrated that repetition affects attitude even when subjects are unaware of being exposed to the stimulus. On the basis of such evidence, Zajonc (1980) suggested that attitudes may precede beliefs in certain situations; that is, the causal flow may be reversed. Briefly stated, the notion is that attitude formation apparently can occur without belief formation under certain conditions of repeated exposure.

The other alternative proposition investigated is derived from the classical conditioning approach to attitude formation (e.g., Staats and Staats 1967). Ac- cording to this perspective, attitudes may be formed by repeatedly pairing a neutral (unconditioned) stimulus with a positively or negatively evaluated (conditioned) stimulus. For instance, an unknown brand name (UCS) could be paired in an advertisement with a picture (CS) that evokes positive feelings. The association might cause the positive feelings evoked by the picture to become conditioned to the brand name.

Both alternative propositions identify conditions under which attitudes may be formed or changed without forming or changing beliefs. Our study was designed to create the conditions necessary to test whether brand attribute beliefs or the causal factors suggested by the alternative propositions mediate attitude formation. We manipulated two variables: (1) repetition-the number of times a particular advertisement was repeated, and (2) advertising content- whether the information presented in each ad was a verbal claim about a product attribute or merely a pairing of a brand name with one of three visual stimuli. By repeatedly exposing subjects to different types of information, we created conditions in which attitudes could have been formed through mechanisms other than belief formation. That is, we experimentally created opportunities for a disconfirmation of the proposition that beliefs are the sole mediator of attitudes. For instance, if advertising content creates brand attitudes without parallel effects on product beliefs, such results would constitute a "strong" disconfirmation of the beliefs-cause-attitudes proposition.

In another article based on this research (Mitchell and Olson 1977), we presented a theoretical discussion of repetition effects and the results of a preliminary data analysis.2 The analysis reported here produced

In that analysis only the cell means were used as the dependent variable. Because we used all the observations in each cell as the dependent variable, our analysis is more complete.

320




the same results for repetition as reported previously; that is, there were no effects of repetition on any of the cognitive variables. Thus, the data cannot be used to address the question of whether sheer repetition might "directly" affect attitudes but not beliefs, because repetition had no effect on brand attitudes or beliefs. Therefore, in this article we are concerned only with the effects of the visual-verbal information manipulation of advertising content.

METHOD

Subjects

Subjects were 71 junior and senior undergraduate students of both sexes who were recruited from an introductory marketing course. Subjects volunteered to attend one of four 45-minute experimental sessions in return for a $2.00 payment. As there is no reason to suspect that subjects' choices of an experimental session were biased, this procedure is assumed to have produced a random assignment of subjects to groups.

Experimental Design To avoid the demand characteristics that might be

created by requiring subjects to rate the same stimulus repeatedly, we used an experimental design in which reactions were measured only once, after all exposures to the four different advertisements. The resulting design was a 4 x 4 Latin square with four groups of subjects and repeated measures over both treatment factors (Table 1). Advertising content was manipulated by four different advertisements, each for a different "brand" of facial tissue. Three advertisements were nonverbal "image ads," each containing a different picture, and the fourth contained a verbal product claim but no picture. Repetition was manipulated at four levels-2, 4, 6, and 8 exposures. As illustrated in Table 1, subjects in the four groups were exposed to all four advertisements, each advertisement under a different level of repetition. In total, each subject received 20 exposures divided among four different advertisements.

This design does not allow an ideal examination of the belief and attitude formation processes as they occur for each individual subject, over time, with successive repetitions. However, it does eliminate the potential reactivity of within-subject designs that require repeated responses to the same measures. The design provides complete information about the main effects of group, advertising content, and repetition level, but only partial information on the interaction between advertising content and repetition. Finally, this design requires the assumption that interactions involving the group factor are negligible. There was no a priori reason to expect treatment x group interactions because subjects were presumed to have been assigned randomly to groups.

Table 1 EXPERIMENTAL DESIGN SHOWING NUMBER OF

REPETITIONS

Brand name/advertising content

Brand I, Brand R, Experi- explicit Brand J, Brand L, picture of mental claim of picture of picture of abstract group softness kitten sunset painting

A (n = 21) 4 8 6 2 B (n = 19) 2 6 8 4 C (n = 17) 6 2 4 8 D(n= 14) 8 4 2 6

Product

For several reasons, facial tissue was selected as the product to be advertised. First, the salient product attributes of the product category were well known to the subjects and thus could be determined relatively easily. Second, subjects' knowledge structures for facial tissues seemed rather simple, involving a reasonably small number of salient product attributes. Unstructured interviews with several secretaries and students indicated six salient attributes for the facial tissue product class: softness, price, absorbency, tearing ease, scent, and color. Finally, it appeared likely that the softness attribute could be communicated in a pictorial, nonverbal manner as well as through an explicit verbal claim.

Brand Advertisements

For each of the four "brands" of facial tissue, advertisements were designed to look like mockup, partially completed print advertisements. To eliminate the influence of prior brand learning on the measured effects of repetition and advertising content, we used unfamiliar brand names. The letters I, J, L, and R were selected as the "brand names" on the basis of pilot research indicating that individuals made relatively few associations with these letters. Each advertisement contained a headline and about the same amount of simulated copy in the form of drawn lines representing rows of words arranged in paragraph form below the headline. The headlines were set in the same type face and size (e.g., "Brand J Facial Tis- sues"). The artwork, simulated copy, and reproduction of the finished advertisements on 35-mm slides were of professional quality.

Of the four advertisements, only one contained an explicit product claim but no picture. Each of the three nonverbal "image" advertisements contained only a brand name headline and a half-page color photograph but no product claim. The verbal claim advertisement and one of the nonverbal advertisements were intentionally designed to communicate that the brand had the specific attribute of softness. That nonverbal advertisement contained a picture of a fluffy kitten, assumed to be a positively evaluated stimulus

321




and to connote "softness," photographed against a dark background. Below the picture was the headline, "Brand J Facial Tissues," and the simulated copy. The verbal claim ad contained only the headline, "Brand I Facial Tissues are Soft," below a half-page blank space where a picture might have been. The other two nonverbal advertisements contained pictures considered to be essentially irrelevant to the product. The advertisement for Brand L included a picture of a spectacular sunset over an ocean, which was assumed to be a positively evaluated stimulus. The advertisement for Brand R contained a presumably neutral picture of an abstract painting.

Experimental Procedures

After being seated at tables in a large room, subjects in each group were told that they would see partially completed advertisements for four different brands of facial tissues. Then, to meet university regulations,

subjects signed an informed consent form. As a rationale for the experiment, subjects were told that the basic purpose of the study was to investigate whether different advertisements affected their ability to remember brand names. Subjects were told to concentrate on remembering the brand names and the advertisement for each brand. This was done to ensure that subjects could associate each "brand name" with the appropriate advertising content, because all post- exposure measures of cognitive structure were brand specific.

Then subjects were shown 20 advertisements (4 brand advertisements with either 2, 4, 6 or 8 repetitions), timed so each remained on the screen for 10 seconds. The ordering of the advertisements was randomly determined for each group, except for the stipulation that the same ad not appear consecutively. After the 20 exposures, subjects responded to the dependent variable measures contained in a question-

Table 2 MAIN EFFECTS OF ADVERTISING CONTENT ON BELIEF STRENGTH, ATTITUDES, AND INTENTIONS

Brand name/advertising content

marginal means

F-ratio' for Brand I, Brand R,

advertising explicit Brand J, Brand L, picture of content claim of picture of picture of abstract

Dependent variables main effect softness kitten sunset painting

Belief levels (main e,) Not at all soft (-2.82) 27.30' 2.49 2.32 2.59 [3.96]

Slightly soft (-0.74) 3.82 4.83 4.51 4.25 4.46

Fairly soft (1.11) 16.58' 5.42 5.31 4.99 [4.45]

Very soft (2.86) 25.83' (4.80) 5.45 4.89 [3.59]

Tears easierb (-1.88) 4.30d 3.99 4.08 3.68 3.46

Tears about as easilyb (0.07) 2.80 5.18 5.06 4.92 4.73

Does not tear as easily (1.58) 5.42d 3.34 3.30 3.68 (3.90)

More absorbentb (2.21) 3.98d 3.87 4.17 3.90 (3.48) About as absorbentb (0.47) 4.99d 5.21 5.17 4.96 (4.70) Less absorbentb (-1.56) 1.95 3.49 3.25 3.39 3.68

More economicalb (1.83) 9.35' 3.82 3.28 3.31 (4.18) About as economicalb (0.56) 2.34 4.79 4.76 4.45 4.82

Less economicalb (-1.25) 4.86d 3.70 4.04 4.25 (3.51)

Comes in more attractive colorsb (0.90) 30.92c [3.04] 4.51 5.11 4.79

Comes in the same colorsb (0.35) 6.46c (5.04) 4.93 4.37 4.34

Attitudes and intentions Attitude toward brand (A,) 27.43c [3.03] 3.65 3.68 [2.82] Attitude toward act of purchasing (Aa,,) 14.54c 3.36 3.61 3.61 [2.95] Behavioral intention to purchase (BI) 15.62c 4.08 4.66 4.81 [3.20]

"d.f. = 3 and 204. bTo complete description read "than (or as) most other brands."

< .001. p < .01.

Notes: Interpretation of the Newman-Keuls pairwise comparisons of advertising content means is: (a) means enclosed by [ ] are

significantly different (p < .01) from each of the other three means, (b) means enclosed by ( ) are significantly different

(p < .05) from underlined means, (c) underlined means are not significantly different (p > .05) from nonunderlined means or other underlined means.

322




naire booklet, then they were debriefed, paid, and dismissed.

Dependent Variables

The major dependent variables were the various cognitive elements defined by Fishbein and Ajzen (1975) that are commonly measured in consumer research (e.g., Gardner, Mitchell, and Staelin 1977; Lutz 1977; Mazis and Adkinson 1976; Olson and Dover 1978b). These included belief strength for the salient brand attributes (b,), evaluations of each attribute belief (e,), attitude toward each brand Ao), attitude toward the act of purchasing and using each brand Aact), and behavioral intention to purchase each brand (BI).

Because of the importance of precisely measuring subjects' attribute beliefs about facial tissues, Ahtola's (1975) vector model modification of Fishbein's basic attitude model was adapted as the belief measurement procedure. In terms of attitude formation, Ahtola's vector model is theoretically identical to Fishbein's. However, the Ahtola approach seems to capture more subtle aspects of consumers' belief structures (Ahtola 1975; Olson and Dover 1978b). To operationalize Ahtola's model one must know not only the salient attributes for a product, but also the levels, categories, or amounts of each attribute dimension that consumers use to discriminate objects in terms of that attribute. In the Ahtola model beliefs are measured for each discriminable attribute level, and the set of beliefs about the various levels of an attribute constitutes a belief vector. The discriminable belief levels for each salient attribute (listed in Table 2) were derived from subjects' comments in the pilot interviews and the authors' intuition.

Following Ahtola, we measured belief strength for each attribute level, although our measurement procedure differed from the constant sum method suggested by Ahtola (1975). For each of the four brands, respondents rated each of the 15 attribute levels as to how likely that level was to be associated with the brand (1-7 scale, very unlikely-very likely). The evaluation (ei) of each attribute level was measured on a bipolar scale labeled good-bad (+3 to -3). Because e, values for particular attribute levels are generally assumed to be the same for all brands within a product category, these measures were not brand specific.

Attitude toward the brand (A ) was measured by the mean of four five-point evaluative scales (good- bad, dislike very much-like very much, pleasant- unpleasant, poor quality-high quality). The mean of three five-point evaluative scales (good-bad, foolish- wise, beneficial-harmful) was used as a measure of attitude toward the act of purchasing and using the brand (A.,t). Multiple scales were used to enhance the reliability of the important attitude measures. Coefficient alpha scores for the Ao and Aat measures

were 0.88 and 0.85, respectively, indicating acceptable levels of internal consistency. Behavioral (purchase) intention toward each brand (BI) was measured on a single seven-point bipolar scale anchored by the phrases "not at all likely to buy" and "very likely to buy."

Consumer evaluations of each advertisement and the two pictorial symbols were also measured. Near the beginning of the post-exposure questionnaire and before the cognitive structure measures, subjects rated each advertisement on seven five-point bipolar scales. A factor analysis of these measures yielded two factors with eigenvalues greater than 1, which together accounted for 68% of the variance. A varimax rotation of these factors identified four variables (good-bad, like-dislike, irritating-not irritating, interesting-unin- teresting) that loaded highly on the first, clearly evaluative, factor. The mean of the four evaluative measures was interpreted as attitude toward the ad (Aad, coefficient alpha = .87). Later in the questionnaire, after the belief and attitude measures, subjects rated the concepts of kittens and sunsets on four five-point scales (good-bad, like-dislike, pleasant- unpleasant, and nice-awful; coefficient alpha = .85). The mean of these evaluative measures was interpreted as indicating attitude toward the pictorial symbols (Apic) portrayed in the nonverbal advertisements.

RESULTS

Demand Characteristics

Immediately after completing the questionnaire, subjects wrote a brief paragraph describing their perceptions of the purpose of the experiment. These comments indicated that virtually all of the subjects accepted the explanation given to them. No subjects mentioned that the experiment might pertain to the effects of conditioning or of different types of advertising content on the formation of beliefs, attitudes, or intentions. Thus, although demand characteristic influences cannot be completely ruled out by this rough check, at least the basic purposes of the experiment were not apparent.

Brand Name Learning To ensure that the measures of beliefs, attitudes,

and intentions for each brand were interpretable, we first had to determine that all subjects had accurately learned the correct association between the advertising content and the four brand names. Therefore, the first question in the post-experimental questionnaire asked subjects to match each brand name with the correct written description of its advertisement. Four subjects were unable to answer this question correctly and were dropped from the analysis along with two other subjects who failed to complete substantial parts of the questionnaire. The resultant sample sizes were 21, 19, 17, and 14 for groups A, B, C, and D, respectively.

323




Analyses An ANOVA for a 4 x 4 Latin square design with

repeated measures on both the advertising content and repetition factors was used to analyze each of the cognitive variables. Following Winer (1971) and Kirk (1968), we used an estimate of the mean square due to subjects within groups to test the between-subjects group factor. Advertising content, repetition, and their interaction are within-subject factors which were tested against the residual error after the sum of squares due to subjects within groups was removed.

Because the cell sizes were unequal, the advertising content and repetition factors are not orthogonal. However, the group factor is orthogonal to these two. As Perreault and Darden (1975) point out, nonorthog- onality creates problems in testing the significance of the nonorthogonal effects, because a portion of the total sum of squares will be common to the nonorthogonal factors. Perreault and Darden suggest a procedure for hypothesis testing that sequentially adds the common variance to each nonorthogonal factor. However, this procedure may produce a significant effect for a particular factor when the common variance is included and a nonsignificant effect when it is not included. A more conservative procedure is to test the significance of each nonorthogonal factor using the sum of squares unique only to that factor (Timm and Carlson 1975). A preliminary analysis of the mean scores of each cell had indicated that only the advertising content factor was significant (Mitchell and Olson 1977). Therefore, only the unique sum of squares due to advertising content was used to test that effect, and the sum of squares common to both nonorthogonal effects was included with the sum of squares unique to repetition to test that effect. In essence, this procedure "stacked the deck" against finding an advertising content effect and for finding a repetition effect.

Effects on Belief Strength

Using these procedures, we ran ANOVAs on the belief strength scores for each of the salient beliefs about facial tissue (i.e., the 15 attribute levels for the five belief vectors, as shown in Table 2). Neither the group nor repetition factors had significant main effects on any of these beliefs (p < .05). However, significant main effects due to advertising content were obtained for 11 (73%) of the 15 belief strength scores (p < .01). Only one (7%) of the 15 possible interaction effects between repetition and advertising content was significant (p < .05), an event possibly due to chance. Thus, only the main effects of advertising content on belief strength are of interest.

As can be seen in Table 2, the advertising content effects were most pronounced for the beliefs about softness. Significant effects of advertising content were obtained for three of the four levels of the

softness attribute (not at all soft, fairly soft, and very soft, all p < 0.001). Pairwise tests indicated that for all three of these attribute levels the beliefs created by the abstract painting advertisement for Brand R were significantly different from the beliefs created by the other three advertisements (Newman-Keuls tests, p < .01). In addition, the advertisement pairing the J brand name with the kitten picture created a stronger belief that the tissues were very soft (the most positively evaluated attribute level) than did the explicit verbal claim advertisement (5.45 vs. 4.80; Newman-Keuls test, p < .01), perhaps because the verbal claim stated that Brand I tissues were soft, not very soft.

The advertising content factor also had strong main effects on both belief levels for the color attribute (p < 0.001). For Brand I in the explicit verbal claim advertisement, the strength of the positively evaluated belief, comes in more attractive colors than other brands, was significantly lower than that for the three visually advertised brands (Newman-Keuls test, p < .01). For the essentially neutral comes-in-the-same- color attribute level, stronger beliefs were created for the verbal-claim brand than for the brands paired with the sunset and abstract paintings.

Finally, scattered, somewhat weaker effects of advertising content were found for six of the remaining nine brand attribute beliefs (p < .01). These advertising content effects were due primarily to differences between the advertisement containing the abstract painting and one or more of the other brand advertisements. Thus, Brand R paired with the abstract painting was perceived to be less absorbent, to not tear as easily, and to be more economical than the other brands. In summary, the advertising content manipulation had substantial multiple effects on product attribute beliefs.

Effects on Attribute Evaluations

Fishbein's attitude theory proposes that attitudes are a function of belief strength and evaluation. Thus, whether or not the advertising content manipulation had any effects on the 15 attribute evaluation (ei) scores is of interest. Our design, however, is not capable of detecting specific effects of advertising content (or repetition) on attribute evaluations. Be- cause advertising content was a within-subject manipulation and the ei's were stated in terms of the generic product category (i.e., they were not brand specific), the ei's were measured only once for each subject, after multiple exposures to the varied contents of the four advertisements. Therefore, the best that can be done is to examine the ei scores as a function of the four groups to determine whether different confoundings of the advertising content and repetition factors (or a nonrandom assignment of subjects to experimental groups) influenced the e,'s (see Table 1). One-way ANOVAs for each of the 15 ei scores

324




indicated a significant effect of groups for only one belief, fairly soft, which is close to the chance level. On the basis of this evidence, and because the advertising content of the four test advertisements does not seem relevant to. attribute evaluations, we have assumed that any observed mediation effect of beliefs on attitudes is probably due to the impact of advertising content on belief strengths, not belief evaluations.

Effects on Attitudes and Intentions

The ANOVA procedure used for the belief strength measures was also used to examine the effects of advertising content on the more global cognitive variables, Ao, Aa,, and BI. Again, neither the main effects due to groups or repetition nor any of the interactions were statistically significant (all p > .05). However, as is consistent with the belief strength results, significant main effects due to advertising content were obtained for the Ao, Aact, and BI variables (p < .01). The marginal means and main effect F-ratios for these effects are reported in Table 2. Specifically, the two brands associated with the positively evaluated visual stimuli (Brands J kitten and L sunset) had significantly more positive brand attitudes (3.65 and 3.68) than did the brand associated with the neutral abstract painting (2.82) or the brand in the explicit verbal claim advertisement (3.03; Newman-Keuls test, p < .01). The Aact and BI scores followed the same pattern, except that the two brands associated with the kitten and sunset pictures were not significantly different from the verbal-claim brand. However, all three of the latter brands were significantly more positive than the brand associated with the presumably neutral abstract painting (Newman-Keuls tests, p < .01).

The preceding results show that the manipulation of advertising content had significant effects on the strength of beliefs about several product attributes, attitudes toward both the brand and the act of buying the brand, and purchase intentions. Now let us consider the mediation processes by which the attitude and intention effects may have occurred.

Alternative Mediators of Attitude Formation

Fishbein proposed that attitudes toward either the act or the object are a function only of the strength of the respective salient beliefs and the evaluations of those beliefs. Alternatively, classical conditioning principles suggest that if a brand is repeatedly paired with a positively evaluated stimulus a "direct transfer" of that evaluation to the brand might occur (Olson and Mitchell 1975). In terms our study, an individual's attitude toward an advertised brand might be affected by that person's evaluation of the advertisement as a whole (Aad), or of a visual stimulus in the ad (Ap,). Presumably, such attitudinal impact would be independent of the semantic meanings that might be acquired from an advertisement, as represented here by the product attribute beliefs. The relative validity of these two alternative explanations for the mediation of attitude formation was examined by using regression analysis and analysis of covariance.

Regression analysis. First, AO and A act were treated as separate criterion variables in four multiple regression models. The independent variables included the evaluation index for the softness belief vector (. b,e,),3 Aad, and Api, as well as dummy variables for three of the four brands when A ,, was included in the regression. The results of these alternative prediction models, summarized in Table 3, indicate that either Aad or Aic, but not both, contributes significantly

3The predicted attitude toward the brand was computed by using the Ahtola/Fishbein formula for all the attribute vectors (X:b,ei) and for only the softness vector (bi ei). Correlations were then computed between the respondents' rated brand attitude (Ao) and these two variables. The softness vector score alone predicted A, better (r = .42) than did the evaluation score based on all the attribute vectors (r = .28). One possible reason for these results is that because softness was the only attribute emphasized in the advertisements, it represented the most stable and salient component of cognitive structure for these brands. Olson and Dover (1978a) suggested that the predictive validity of a Fishbein-type model may depend on the stability of the salient beliefs in cognitive structure.

Table 3 ALTERNATIVE MODELS FOR PREDICTING BRAND ATTITUDES

(beta coefficients and t-statistics)

Criterion Predictor variables

variable Model bie, A A pic R2

Ao 1 0.40 (7.33)b -.16 2 0.31 (6.10)" 0.31 (3.37)b .35 3 0.22 (5.37)" 0.65 (15.74) - .56 4 0.20 (4.71)b 0.61 (11.73)" 0.11 (1.35) .57

A^t 1 0.39 (7.12)b - .16 2 0.31 (5.65)b 0.14 (1.40) .23 3 0.28 (5.45)" 0.41 (7.99) .31 4 0.24 (4.47)" 0.42 (6.54)b -0.001 (0.01) .34

aEvaluation based on softness belief vector. bp < 0.005.

325




Table 4 EFFECT OF COVARIATES ON THE F-RATIO FOR THE MAIN EFFECT OF ADVERTISING CONTENT

Covariate(s)

Dependent No variable covariate Ib, eia ?bie,a & Aad Aac, Aact & .bie,I Aac, & Aad

A. 27.43b 16.03b 0.13 -

Aact 14.54b 3.18' 0.86

BI 15.62- 3.61b 2.52 2.47

aEvaluation based on softness belief vector. bp < 0.001. cp < 0.05.

to the prediction of Ao beyond that provided by the evaluation (5 b,ei) associated with the softness beliefs (R = .74 or .59 vs. r = .40, respectively). Because the correlation between Aad and A pi was only 0.34, multicollinearity probably did not account for this result. Much the same pattern of results was obtained for the prediction of A at, except that only Aad (not A ic) contributed significantly to the prediction of Aact in addition to lbiei. Empirically, Aad is a stronger predictor of brand attitudes than Apic. Conceptually, Aad seems to be the more general construct because it may incorporate the evaluation associated with the picture symbol.

Covariance analysis. Next, using analysis of covariance, we examined the mediating roles of product attribute beliefs and Aad in explaining the obtained differences in Ao and Aact (e.g., Osterhouse and Brock 1970). If product attribute beliefs are the only mediator of brand attitudes, including the presumed mediator (represented by the evaluative belief index, lbie,) as a covariate should reduce the main effect of advertising content on Ao to a nonsignificant level. However, if a significant Ao effect remains after statistical removal of the presumed mediating effect of brand attribute beliefs, these beliefs may not be the only mediator of attitude formation. The ANCOV results are summarized in Table 4 in terms of the changing F-ratios for the main effects of advertising content given different mediators/covariates. The results indicate that including the evaluation associated with the softness belief vector (bi,ei) as a covariate did not completely remove the effects of advertising content on either Ao or A at. However, using both .biei and Aad as covariates eliminated any reliable

effects on attitude due to advertising content. When these analyses were repeated using the combined evaluations associated with all the attribute belief vectors (5zbiei) as a covariate, the results were identical.

In summary, the results of both the regression and ANCOV analyses suggest that brand attitudes are not solely a function of the attribute beliefs that are formed about the brand, but may also be influenced by consumers' general liking for the ad itself or the visual stimulus presented in the advertisement.

Mediators of Behavioral Intentions

For completeness, the BI scores were analyzed similarly (see Table 4). As specified by the Fishbein extended model (cf. Lutz 1977), Aact is the critical mediator of BI. Including Aact as a covariate reduced the magnitude of the advertising content main effect on BI considerably, but not quite to nonsignificant levels. When either Xbie, or Aad was included as a second covariate, the advertising content effect on BI became marginally significant (F = 2.52 and 2.47, both p = .06). These results suggest that Aad had its major mediating effect on attitudes, either Ao or Aac,, and had only a weak mediating influence on BI. Thus, as predicted by the Fishbein theory, Aact was the major mediator of BI.

DISCUSSION

We tested the basic theoretical proposition of Fish- bein's attitude theory in a marketing research context. Specifically, we examined whether product attribute beliefs were the only mediator of advertising effects on brand attitudes. The beliefs-cause-attitudes proposition was contrasted with two alternative models of attitude formation, sheer repetition and classical conditioning, neither of which treats beliefs as a causal mediator. Thus the study was designed to provide conditions in which either beliefs or the alternative causal processes could have mediated the effects of advertising on attitude formation.

The experiment involved two manipulations: (1) the number of times a subject saw a particular advertisement (repetition) and (2) the type of visual or verbal information contained in the advertisement (advertising content). Repetition had no reliable effects on attitude formation or any other cognitive variable. Therefore, we could not test the proposition that multiple exposures to a stimulus can directly affect attitude without influencing beliefs.4 In contrast, the

4However, the lack of repetition effects on both beliefs and attitudes is consistent with, although only weak evidence for, the beliefs-cause-attitude proposition. The Fishbein proposition would have been disconfirmed only if repetition had affected either beliefs or attitudes, but not the other.

326




advertising content factor produced significant effects on three cognitive variables-product attribute beliefs, attitudes, and purchase intentions.

Given these main effects of advertising content, the critical question was whether product attribute beliefs were the only mediator of the variation in brand attitude produced by the advertisements, or whether another process such as classical conditioning was operating. The obtained mediation relationships are summarized in Figure 1. They indicate that beliefs about product attributes were a major mediator of the advertising content effect on attitudes, but not behavioral intentions, as predicted by the causal flow of the Fishbein extended model. These results are consistent with those obtained by Lutz (1977). However, the mediation analyses also suggest that product attribute beliefs were not the sole mediator of attitudes. The Aad measure also seemed to mediate attitude formation. In fact, together, the Aad scores and the evaluative belief indices (X bie,) totally mediated the effects of advertising content on brand attitudes. These results are similar to the mediation effects of product attribute beliefs and advertising evaluations reported by Holbrook (1978).

Do the results obtained confirm or disconfirm Fish- bein's basic proposition that beliefs cause attitude? They seem to do both. Although product attribute beliefs did mediate message effects on attitude as specified by Fishbein's attitude theory, such beliefs do not appear to be the only mediator. Another potential mediator, represented by the Ad measure, may be operating. To understand better the meaning of these results, we must examine the Aad concept more closely.

Attitude Toward the Advertisement A straightforward interpretation of the Aad measure

is that it accurately reflects subjects' evaluations of the overall advertising stimulus. Accordingly, Aad

should be treated as a construct that is conceptually distinct from brand attribute beliefs and brand attitude. Thus the mediation effect of Aad can be interpreted as capturing the classical conditioning effect of pairing an unknown brand name (unconditioned stimulus) with

Figure 1 OBSERVED MEDIATORS OF ADVERTISING CONTENT ON

COGNITIVE VARIABLES

a highly valenced visual stimulus (conditioned stimulus). In this interpretation, the conditioning process due to contiguous presentation caused the evaluation associated with the advertisement in general (Aad), or with a prominent part of the advertisement such as a picture (Api), to become associated with the brand name (see Staats and Staats 1967). Presumably, this "direct" influence on attitude is independent of the message's effect on the formation of or change in product attribute beliefs. If this classical conditioning interpretation is accepted, our results are partially inconsistent with the information processing perspective of Fishbein.

Alternatively, one might argue that A ad is a surrogate indicator of unmeasured salient beliefs about product attributes not included in the experimental questionnaire. Hence, processing the advertised information, impoverished as it was, may have created beliefs about unmeasured product attributes that somehow are re- flected by the Aad measure. The idea is that if we had measured all the salient attribute beliefs for each subject, the Aad mediation effect would not have been obtained. However, as it is not obvious how the four-item Aad score could represent beliefs about unmeasured product attributes, this explanation seems somewhat strained. In any case, this explanation cannot be addressed by our data. Because we used a structured questionnaire to measure b, and e,, all subjects rated a fixed set of presumably salient attributes. To examine this issue we must identify and measure the set of beliefs that are salient for each subject. Lutz and Swazy (1977) proposed such a procedure, and Olson and Muderrisoglu (1979) de- scribed free-elicitation procedures by which it could be accomplished. To date, however, no marketing- oriented study has investigated how subjects' idio- syncratic attribute beliefs are influenced by advertising messages.

Along similar lines, the advertising content manipulation could have created beliefs about concepts other than product attributes (Calder 1978). For example, a consumer could have a belief that X Facial Tissue is manufactured by a particular company, or a person may associate a visual image with Brand Z. Because in a general sense beliefs are the subjective associations between cognitive representations, consumers presumably can have beliefs about any concept that they have represented cognitively. However, beliefs about the nonattribute concepts that may be associated with a brand are not measured by the typical marketing-oriented multiattribute questionnaire. The point is that we know relatively little about the types of cognitive representations consumers may associate with a brand. Possibly at least some of these salient beliefs do not concern traditional product attributes.

If semantic memory is conceptualized as a network of associations between cognitive representations (cf. Anderson 1976), a set of product beliefs is a semantic

327




memory structure. A semantic memory concept provides the basis for another explanation of the Aad mediation effect. Assume that a visual image of each advertisement is associated with the brand name representation in a consumer's memory. Although marketing researchers usually would not treat the "connection" between a brand name and the remem- bered advertisement as a belief, such an association is conceptually equivalent to a belief. Thus, if the consumer is asked to evaluate the brand, the related visual image might be activated from memory along with other beliefs about the brand, perhaps including product attribute beliefs. If so, the evaluation of the visual image, indicated by Aad in our study, should influence the overall brand attitude, Ao. By this interpretation the Aad concept is absorbed into the semantic memory structure of beliefs that are associated with the attitude concept. Thus, Aad is not a separate construct, bttt rather is a surrogate measure of biei; that is, Aad is the evaluation of the advertisement "attribute" of the brand. If this explanation is accepted, our results are totally consistent with Fishbein's proposition that beliefs cause attitude, because Aad is considered to be another belief mediator of Ao.

In sum, the interpretation of our results hinges on the conceptual meaning of the Aad construct. At minimum, Fishbein's theoretical proposition is partially supported. However, other processes may be operating that are not captured by the belief structure model, at least as typically operationalized by marketing researchers in terms of product attribute beliefs. Although empirical data will probably be valuable in reaching an accepted interpretation, the decision about how to treat Aad ultimately may hinge on more pragmatic considerations. In advertising effects research, for example, it may be useful to consider Aad as a separate construct that can provide separate diagnostic information about an advertisement's attitudinal impact on consumers. In contrast, if the researcher is interested in memory structure, it may be more useful to consider A ad as the evaluation of an image-like cognitive representation that has become associated with the brand name representation. Fishbein's attitude theory can incorporate the latter approach, but not the former (see Zajonc 1980).

Future research. Our research design is not capable of distinguishing between these alternative explanations for the Aad effect. In fact, designing an empirical test of these causal explanations will be difficult. Researchers might begin by reviewing past research on how consumers' reactions to the advertisement per se are related to other responses (e.g., March and Swinbourne 1974; Schlinger 1979; Silk and Vavra 1974; Wells, Leavitt, and McConville 1971). As our results and those of Holbrook (1978) indicate, such evaluations of the advertisement seem to have an independent impact on brand-related responses. More research is needed, however, to explicate consumers'

reactions to advertisements and to differentiate those effects from advertising impacts on brand attribute beliefs, attitudes, and intentions.

How can these ideas be explored in future research? If Aad is a direct mediator of attitude formation, the Aad score might be expected to have a greater mediation effect on brand attitudes for advertisements that arouse more intense evaluative or affective reactions, whether positive or negative. Such responses might be due to vivid visual content or strong symbols used in the advertisement. Also, the verbal (written or spoken) content of the advertisement might evoke intense affective feelings, perhaps by stimulating visual imagery (cf. Kosslyn 1975; Pavio 1977). Thus, we might "test" the Aad mediation hypothesis by examin- ing the magnitude of its effect for advertisements that differ greatly in evaluative appeal but contain about the same verbal content.

Another way in which the evaluation associated with an advertisement may become conditioned to a brand occurs when an individual pays general attention to an advertisement, but does not actively process the brand-relevant information in the advertisement (Krugman 1965; Mitchell, Russo, and Gardner 1980)- for instance, when a consumer attends to an advertisement only for its entertainment value. Under such conditions, a consumer may achieve only limited comprehension of the brand-related information in the message. However, when considering the advertised brand at some later time (say, during the purchase decision process), the individual might recall characteristics of the advertisement from memory. Then the Aad that is also retrieved might influence the formation of a brand attitude (Mitchell, in press).

Future research could examine the relative mediation effects of beliefs (Eb,ie) and advertising evaluations (e.g., Aad) for advertisements of widely differing semantic and affective content, in conditions involving different purchase deliberation goals, and for varied time intervals after exposure. Such research would help to clarify the construct of Aad by determining the specific conditions under which Aad seems to mediate attitude formation and change.

Visual Versus Verbal Information in Advertising

Relatively few studies of advertising effects have examined the influence of visual versus verbal information content. Exceptions include Wright's (1979) study of the efficacy of different visual content for inducing certain desired behaviors, Rossiter and Percy's (1978, in press) examinations of the attitudinal impact of visual information, and Holbrook's (1978) research on the effects of more and less affective verbal advertising copy on brand attitudes. In our study we examined four ads, only one of which contained verbal information about a product characteristic. The other three advertisements contained only visual information (except for the brand and product names). Despite the relative lack of specific informa-

328




tion, the belief strength measures indicate that subjects formed rather different perceptions of the four brands (see Table 2).

These results are of interest for several reasons. At a descriptive level they suggest that subjects, apparently by some inferential process, develop beliefs about brand attributes based on minimal brand-specific information. In addition, they show that the measures used in our study are sensitive enough to detect these effects. The results also suggest that subjects convert- ed visual information that was not directly related to the product into meaningful semantic information. This process is indicated by the obtained differences between brands in the strength of beliefs about brand attributes, particularly the softness beliefs. For instance, the picture of a fluffy kitten produced the strongest beliefs that the facial tissue was very soft.

A possible explanation for the softness belief strength effects is that seeing the advertisement containing the verbal claim about softness primed subjects to think about softness during exposure to the other three visual advertisements. This alternative hypothesis was examined by taking written cognitive responses (thought protocols) from two additional groups of student subjects (n = 18 and 16) immediately after exposure to each of the four advertisements. We found that when the verbal claim preceded the kitten advertisement, about as many subjects mentioned softness in their protocols for the kitten advertisement (72.2%) as when the kitten advertisement preceded the verbal claim (75.0%). Very few mentions of softness were made in response to the advertisements portraying the sunset and the abstract painting. Thus, these data provide additional support for the idea that the kitten picture independently connoted the semantic concept of softness.

For the other product attribute beliefs, the connec- tions between the pictorial stimuli and the inferentially derived beliefs about facial tissue are somewhat less obvious. For instance, subjects seem to have interpreted the picture of the sunset as meaning that the advertised brand comes in more attractive colors (than other brands). Subjects also seemed to make inferences about other characteristics of the four brands (e.g., absorbency) even though no relevant information was provided. Such inferences were especially evident for the brand paired with the abstract painting, which was perceived as more negative on nearly every attribute. Inferences are clearly important in attitude formation (cf. Fishbein and Ajzen 1975), and more research on consumer inference processes is needed (Lutz 1976; Olson 1978). To stimulate that research, we offer the following comments on inferential belief formation.

Inferential Belief Formation

Inferences may be formed in at least three ways. First, as suggested by Wyer (1974), individuals may believe a product having a particular level of one

attribute is likely to have a certain level of another attribute. For instance, if consumers believe a brand of facial tissue is very soft, they may also consider it likely the brand will tear easily or be more absorbent. Such inferences could be based on previously learned associations about the generic product category which enable a person to form numerous beliefs based on only one piece of information (cf. Olson 1978). Second, memory structures for generic product categories may provide default values which enable a person to make inferences about specific attributes when information is missing (e.g., Cantor and Mischel 1977; Rosch 1978). For instance, individuals may have generic knowledge structures for different categories of facial tissue which provide typical values for the absorbency and tearing ease of brands in that category. Basically, both of these explanations derive from semantic memory theory. That is, people have structures of knowledge or memory schemata which may be activated and used as a basis for interpreting new information and for making inferences (cf. Olson 1978, 1980).

A third explanation for the observed "inferences" was suggested by Mitchell (1979). The overall brand attitude may influence specific ratings of belief strength through a "halo effect" process. The process seems especially likely to occur when subjects are forced to rate beliefs about nonsalient attributes. This explanation cannot be ruled out for some of our belief strength results. Our procedures for establishing the saliency of the attribute dimensions were admittedly intuitive, and all subjects had to rate every attribute. Thus, some subjects probably rated some nonsalient beliefs. The ideal situation would be for each subject to rate only those beliefs that are salient for him or her. Although difficult, a procedure of eliciting the salient beliefs of each individual is not impossible (cf. Lutz and Swazy 1977; Olson and Muderrisoglu 1979). In fact, such an approach is probably necessary in future research if we are to develop clearer answers to these theoretical questions about the basis for inferential belief formation.

Finally, like most studies of attitude formation, our design cannot rigorously establish when the measured attribute beliefs were formed. Possibly some of the brand beliefs were formed during completion of the questionnaire, rather than during or immediately after exposure to the advertisements. Cognitive response protocols taken during exposure could be used in future research to provide evidence of the extent of belief and attitude formation at exposure.

Limitations

Certain characteristics of our research were included to enhance internal validity, but obviously limit exter- nal validity-that is, the generalizability of the results to natural advertising settings. Certainly the test advertisements contained much less information than actual print advertisements. Unraveling the relative effects of verbal and visual content in real, complex

329




advertisements will be difficult because most contain substantial amounts of both verbal and visual content designed to be complimentary. Another limitation of our research is the artifical exposure environment. For instance, the advertisements were not embedded in editorial material, exposure was forced and timed, and the repetitions occurred within a short time period. The last feature, in fact, may account for the nonsignificant effect of repetition. Multiple exposures to an advertisement over more natural intervals (e.g., weeks) might produce measurable changes in beliefs, attitudes, and intentions. Also, effects measured in our study were taken almost immediately after exposure. Whether similar mediation effects would be obtained after post-exposure delays of days or weeks is not known. Finally, by using fictitious "brand names," we avoided the effects of previously acquired brand knowledge. In natural settings, however, peo- ples' stored knowledge will interact with the advertised information. In sum, moving to more natural research settings and advertising stimuli will complicate the research design problems and will require rigorous controls and measurements.

SUMMARY

Our results indicate that product attribute beliefs had a major mediating effect on brand attitudes, whereas attitudes substantially mediated behavioral intentions. These findings are consistent with Fish- bein's attitude theory. However, contrary to Fish- bein's major assumption as it is operationally interpreted in most marketing research, the product attribute belief index, lbiei, was not the sole mediator of attitude formation. Rather, a measure of attitude toward the advertisement, Aad, partially mediated advertising effects on brand attitudes. Because these results can be accounted for within the beliefs-cause- attitudes theoretical framework, however, our data cannot be interpreted as strongly disconfirming the basic proposition of Fishbein's attitude model. Addi- tional research is needed to develop a better understanding of the Aad construct, its measures, and its causal dynamics, as well as to identify the specific conditions under which it may have separate mediation effects on attitudes and intentions relative to belief measures.

Our results also indicate that individuals can develop different perceptions of brands based only on visual information that provides no explicit brand information. That is, consumers seem able to convert visual information into semantic knowledge or beliefs about attributes of the advertised brand. The process seems to occur even for visual stimuli that are seemingly irrelevant to the product. These phenomena may be considered in terms of people's memory structures for the generic product categories and the inference processes that are guided by these cognitive structures. As these theoretical issues also have implications for

advertising and marketing practice, they warrant further research attention.

REFERENCES

Ahtola, Olli T. (1975), "The Vector Model of Preferences. An Alternative to the Fishbein Model," Journal of Mar- keting Research, 12 (February), 52-9.

Anderson, John R. (1976), Language, Memory and Thought. Hillsdale, New Jersey: Lawrence Erlbaum.

Berlyne, D. B. (1970), "Novelty, Complexity and Hedonic Value," Perception and Psychophysics, 8, 279-86.

Calder, Bobby J. (1975), "The Cognitive Foundations of Attitudes: Some Implications for Multi-Attribute Models," in Advances in Consumer Research, Vol. 2, Mary J. Schlinger, ed. Chicago: Association for Consumer Research, 241-7.

(1978), "Cognitive Response, Imagery, and Scripts: What Is the Cognitive Basis of Attitude?" in Advances in Consumer Research, Vol. 5, H. Keith Hunt, ed. Ann Arbor, Michigan: Association for Consumer Research, 630-4.

Cantor, Nancy and Walter Mischel (1977), "Traits as Proto- types: Effects on Recognition Memory," Journal of Per- sonality and Social Psychology, 33 (January), 38-48.

Carnegie-Mellon University Marketing Seminar (1978), "Attitude Change or Attitude Formation? An Unanswered Question," Journal of Consumer Research, 4 (March), 271-6.

Cohen, Joel B., Martin Fishbein, and Olli T. Ahtola (1972), "The Nature and Uses of Expectancy-Value Models in Consumer Attitude Research," Journal of Marketing Re- search, 9 (November), 456-60.

Dickson, Peter R. and Paul W. Miniard (1978), "A Further Examination of Two Laboratory Tests of the Extended Fishbein Attitude Model," Journal of Consumer Research, 4 (March), 261-6.

Dover, Philip A. and Jerry C. Olson (1977), "Dynamic Changes in the Expectancy-Value Attitude Model as a Function of Multiple Exposures to Product Information," in Contemporary Marketing Thought, Barnett A. Green- berg and Danny A. Bellenger, eds. Chicago: American Marketing Association, 455-60.

Fishbein, Martin (1963), "An Investigation of the Relation- ships Between Beliefs About an Object and the Attitude Toward the Object," Human Relations, 16, 233-40.

(1967), "A Consideration of Beliefs and Their Role in Attitude Measurement," in Readings in Attitude Theory and Measurement, Martin Fishbein, ed. New York: John Wiley & Sons, Inc., 257-66.

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

Gardner, Meryl, Andrew A. Mitchell, and Richard Staelin (1977), "The Effects of Attacks and Inoculations in a Public Policy Context: A Cognitive Structure Approach," in Contemporary Marketing Thought, Barnett A. Green- berg and Danny A. Bellenger, eds. Chicago: American Marketing Association, 292-7.

Harrison, A. A. (1968), "Response Competition, Frequency, Exploratory Behavior and Liking," Journal of Personality and Social Psychology, 10 (August), 363-8.

(1977), "More Exposure," in Advances in Experi-

330




mental Social Psychology, Vol. 10, Leonard Berkowitz, ed. New York: Academic Press, 39-83.

Holbrook, Morris B. (1978), "Beyond Attitude Structure: Toward the Informational Determinants of Attitude," Journal of Marketing Research, 15 (November), 545-56.

Kirk, R. E. (1968), Experimental Design: Procedures for the Behavioral Sciences. Belmont, California: Brooks/ Cole.

Kosslyn, Stephen M. (1975), "Information Representation in Visual Images," Cognitive Psychology, 7 (July), 341-70.

Krugman, Herbert E. (1965), "The Impact of Television Advertising: Learning Without Involvement," Public Opinion Quarterly, 29 (Fall), 349-56.

Lutz, Richard J. (1975), "Changing Brand Attitudes Through Modification of Cognitive Structure," Journal of Con- sumer Research, 1 (September), 49-59.

(1976), "First-Order and Second-Order Cognitive Effects in Attitude Change," in Communicating with Consumers, Michael L. Ray and Scott Ward, eds. Beverly Hills, California: Sage Publications, 101-12.

(1977), "An Experimental Investigation of Causal Relations Among Cognitions, Affect, and Behavioral In- tention," Journal of Consumer Research, 3 (March), 197-208.

(1978a), "Rejoinder," Journal of Consumer Behav- ior, 4 (March), 266-71.

(1978b), "Rejoinder," Journal of Consumer Re- search, 4 (March), 276-8.

and James R. Bettman (1977), "Multiattribute Models in Marketing: A Bicentennial Review," in Consumer and Industrial Buying Behavior, Arch G. Woodside, Jagdish N. Sheth, and Peter D. Bennett, eds. New York: North- Holland, 137-50.

and John L. Swazy (1977), "Integrating Cognitive Structure and Cognitive Response Approaches to Moni- toring Communications Effects," in Advances in Con- sumer Research, Vol. 4, William D. Perreault, Jr., ed. Atlanta: Association for Consumer Research, 363-71.

March, Robert M. and Donald W. Swinbourne (1974), "What is 'Interest' in TV Commercials?," Journal of Advertising Research, 14 (August), 17-22.

Matlin, M. W. (1970), "Response Competition as a Mediating Factor in the Frequency-Affect Relationship," Journal of Personality and Social Psychology, 16 (November), 536-52.

Mazis, Michael B. and Janice E. Adkinson (1976), "An Experimental Evaluation of a Proposed Corrective Adver- tising Remedy," Journal of Marketing Research, 13 (May), 178-83.

McGuire, William J. (1968), "The Nature of Attitudes and Attitude Change," in Handbook of Social Psychology, 2nd ed., Vol. 3, Gardner Lindzey and Elliot Aronson, eds. Reading, Massachusetts: Addison-Wesley, 136-314.

Miniard, Paul W. and Joel B. Cohen (1979), "Isolating Attitudinal and Normative Influences in Behavioral Inten- tions Models," Journal of Marketing Research, 16 (Febru- ary), 102-10.

Mitchell, Andrew A. (1979), "Predicting Choice with Prefer- ence Functions Estimated from Perceptions of Brands Within a Product Category," in Analytic Approaches to Product and Marketing Planning, Allan Shocker, ed. Cambridge, Massachusetts: Marketing Science Institute, 368-87.

(in press), "Cognitive Processes Initiated by Expo-

sure to Advertising," in Information Processing Research in Advertising, Richard J. Harris, ed. Hillsdale, New Jersey: Lawrence Erlbaum.

and Jerry C. Olson (1977), "Cognitive Effects of Advertising Repetition," in Advances in Consumer Re- search, Vol. 4, William D. Perreault, Jr., ed. Atlanta: Association for Consumer Research, 213-110.

, J. Edward Russo, and Meryl Gardner (1980), "Strategy Induced Low Involvement Processing of Ad- vertising Messages," working paper, Graduate School of Industrial Administration, Carnegie-Mellon University.

Olson, Jerry C. (1978), "Inferential Belief Formation in the Cue Utilization Process," in Advances in Consumer Research, Vol. 5, H. Keith Hunt, ed. Ann Arbor, Michi- gan: Association for Consumer Research, 706-13.

(1980), "Encoding Processes: Levels of Processing and Existing Knowledge Structures," in Advances in Consumer Research, Vol. 7, Jerry C. Olson, ed. Ann Arbor, Michigan: Association for Consumer Research, 154-60.

and Philip A. Dover (1976), "Effects of Expectation Creation and Disconfirmation on Belief Elements of Cognitive Structure," in Advances in Consumer Research, Vol. 3, Beverly B. Anderson, ed. Cincinnati: Association for Consumer Research, 168-75.

and (1978a), "Attitude Maturation: Changes in Related Belief Structures Over Time," in Advances in Consumer Research, Vol. 5, H. Keith Hunt, ed. Ann Arbor, Michigan: Association for Consumer Research, 333-42.

and (1978b), "Cognitive Effects of Deceptive Advertising," Journal of Marketing Research, 15 (Fe- bruary), 20-38.

, Rajesh Kanwar, and Aydin Muderrisoglu (1979), "Clarifying the Confusion Regarding Salience, Impor- tance, and Determinance Concepts in Multi-Attribute Attitude Research," in AMA Educators' Conference Pro- ceedings, Neil Beckwith, et al., eds. Chicago: American Marketing Association, 286-90.

and Andrew A. Mitchell (1975), "The Process of Attitude Acquisition: The Value of a Developmental Approach to Consumer Attitude Research," in Advances in Consumer Research, Vol. 2, Mary J. Schlinger, ed. Chicago: Association for Consumer Research, 240-64.

and Aydin Muderrisoglu (1979), "The Stability of Responses Obtained by Free Elicitation: Implications for Measuring Attribute Salience and Memory Structure," in Advances in Consumer Research, Vol. 6, William L. Wilkie, ed. Ann Arbor, Michigan: Association for Con- sumer Research, 269-75.

Osterhouse, R. A. and Timothy C. Brock (1970), "Distrac- tion Increases Yielding to Propaganda by Inhibiting Coun- terarguing," Journal of Personality and Social Psychology, 15 (August), 344-58.

Pavio, Allan (1977), Imagery and Verbal Processes. New York: Holt, Rinehart and Winston.

Perreault, William D. and William R. Darden (1975), "Un- equal Cell Sizes in Marketing Experiments: Use of General Linear Hypotheses," Journal of Marketing Research, 7 (August), 333-42.

Platt, John R. (1964), "Strong Inference," Science, 146 (October), 347-53.

Rosch, Eleanor (1978), "Principles of Categorization," in Cognition and Categorization, Eleanor Rosch and Barbara

331




B. Lloyd, eds. Hillsdale, New Jersey: Lawrence Elbaum, 27-48.

Rosenberg, Milton J. (1956), "Cognitive Structure and At- titudinal Affect," Journal of Abnormal and Social Psy- chology, 53, 367-72.

Rossiter, John R. and Larry Percy (1978), "Visual Imaging Ability as a Mediator of Advertising Response," in Ad- vances in Consumer Research, Vol. 5, H. Keith Hunt, ed. Ann Arbor, Michigan: Association for Consumer Research, 621-28.

and - (in press), "Visual Communication in Ad- vertising," in Information Processing Research in Adver- tising, Richard J. Harris, ed. Hillsdale, New Jersey: Lawrence Erlbaum.

Sawyer, Alan G. (1977), "Repetition and Affect: Recent Empirical and Theoretical Developments," in Consumer and Industrial Buying Behavior, Arch G. Woodside, Jagdish N. Sheth, and Peter D. Bennett, eds. New York: North Holland, 229-42.

Schlinger, Mary Jane (1979), "A Profile of Responses to Commercials," Journal of Advertising Research, 19 (April), 37-46.

Silk, Alvin J. and Terry G. Vavra (1974), "The Influences of Advertising's Affective Qualities on Consumer Re- sponse," in Buyer/ Consumer Information Processing, G. David Hughes and Michael L. Ray, eds. Chapel Hill: University of North Carolina Press.

Staats, Arthur W. and Carolyn Staats (1967), "Attitudes Established by Classical Conditioning," in Readings in Attitude Theory and Measurement, Martin Fishbein, ed. New York: John Wiley & Sons, Inc., 337-81.

Stang, David J. (1975), "Effects of 'Mere Exposure' on

Learning and Affect," Journal of Personality and Social Psychology, 31 (January), 7-12.

Timm, N. H. and J. E. Carlson (1975), Analysis of Variance Through Full Rank Models, Multivariate Behavioral Re- search Monograph No. 75-1, Texas Christian University, Forth Worth, Texas.

Triandis, Harry C. (1971), Attitude and Attitude Change. New York: John Wiley & Sons, Inc.

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

Wilkie, William L. and Edgar A. Pessemier (1973), "Issues in Marketing's Use of Multi-Attribute Attitude Models," Journal of Marketing Research, 10 (November), 428-41.

Wilson, William R. (1979), "Feeling More Than We Can Know: Exposure Effects Without Learning," Journal of Personality and Social Psychology, 37 (June), 811-21.

Winer, B. J. (1971), Statistical Principles in Experimental Design, 2nd ed. New York: McGraw-Hill Book Company.

Wright, Peter L. (1979), "Concrete Action Plans in TV Messages to Increase Readings of Drug Warnings," Jour- nal of Consumer Research, 6 (December), 256-69.

Wyer, Robert S. (1974), Cognitive Organization and Change: An Information Processing Approach. Hillsdale, New Jersey: Lawrence Erlbaum.

Zajonc, Robert B. (1968), "Attitudinal Effects of Mere Exposure," Journal of Personality and Social Psychology Monograph Supplement, 9 (February), 1-27.

(1980), "Feeling and Thinking: Preferences Need No Inferences," American Psychologist, 35 (February), 151- 75.

AG-CHEM MARKETING CONFERENCE

October 25-27, 1981 Stouffer's National Center Hotel Arlington, Virginia

FEES: AMA Members Non-Members

Early Late $105 $120 $120 $135

How-to conference for professional agricultural researcher covering all aspects of marketing research including theory and application. General sessions on technique and discussion sessions on problem solving. Sponsored by the Agricultural-Chemical Marketing Research Section of the Industrial Marketing Division of AMA.

EDUCATION DEPARTMENT American Marketing Association 250 S. Wacker Drive Chicago, Illinois 60606

332



Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitude? Author(s

Documents