n .1 u. "L. ”Ara-:33“. w“:— . . 3&3» ..~}£..fl_: "WV... 1 I. 1 an» . . ‘ ‘ z . 1...” -' vun» o'fii- .V .l “'3', kg ‘ufifl ELM”; m: .‘1. .17... “41333: "L“ .3:— . a. - -.1A:x.l:§$.wr ‘ 2A.... 1.... \' -va.; “1.....4‘. If,“ . ) I: ‘11.}. meat-m "If: -11. a... it m‘\w ' 1......1'... . 9:“ “g J» u... .. A»: LL. 4.1. -- duh-3.: ,L L'»~x| OIL— "‘( s-«A‘ ‘. \5 1.! .-~;_ . a.“ Mile?” 3:1.“ .‘f-‘flwda 2. «a. u. .Lun-.- uuwfifi in . Mm -:~: “ W 2‘ ’ moi; . ~. ~ *4 . . ""‘ .w 1 .3; ' - “75“?" ‘3 ‘6"L3sfit '2’" "‘5 Mfifi‘w 3- "”‘~‘ 13"}; «m ‘ ' ~- . .... its ‘ ‘ ' “ ' ‘17. ‘- Wm "’5 {151“ ‘ ‘* fun“ a - J t ’ £13"? «.k-sJ-umm...“ I ‘1. . .-L ...~. A - udcvs ; . ..-:‘_.J *5 n. Jrng‘n; H: i”. .77 4._?.. .-. 4‘“. if}; {2.}.- ;' 4 v '3 ... '4‘ .u _%:_ ,w ,_£’\.L., ‘4“ “J. ‘1‘“ *. - . ."-l A --' 3 x . - . god”, , A ., r Arm“ “ u n L" ~7 ~...-_: ’ fia- run-‘7',“ ya, .: A “'4‘“ 4V LL “‘ ‘ ‘ $5.041.- w.»-. «a {(1. -\ 4”.“ . ‘ u‘ in. _:~ M 2' .. mfimgo'fimwuu 323% A . (.5: an- ~14; [Muf‘w nary “A. . .313)” ".4" ‘1. V ’3' “IA-— %,?‘:‘ . ‘ =.::~:S:% I 4' '3}? :W P ‘ ‘ C J‘ 1" 0“ . . ‘ w >‘ 1“: “‘w :31" a: . , , £3 .1. v-‘uhwtl i; «Abba-3,.“ ‘ a . M5; u: ...... ‘1 4‘ a 4i” u-fl‘” t.- .v. a. . ‘ nu‘.‘ «In .n ...m\. .4. «as. ,5: fitk an... r 5!! ”I... .;,..4 'W‘ I I. . Y- -31..» N 4 m. - :‘m-v 32441;.44“ 4.. Mulu- MI (.1 I. a»... J‘ A 4 “f1 “4 1 :- ,-~-ou-.¢‘..‘.. .. f“““_ «.424-“wsv ”8-5:, H.»- h. .. ‘3‘! Lu; :4 -n 37:71: H32" ‘5 » ‘ ' “2.. ' - . . . 'v.‘ ' , r JU§ .‘ém “Luau; J‘firfu:( ‘, - . 7 . r . “.7. . J. _ . , WW:...,},\,‘}%‘,\. “‘5" ‘ _ . — '3“ ‘ J . ‘ {iii ”3:121 r .3 :55- 7‘ N p. 4:11. 31.- t. '1’. 7'. ' : ‘uj. ‘1 .4“ 1‘1. [-1.1% as. L'fi 57‘ 5 7: . m. '1‘).- ,;;':1;.’.;.' . 5; "tr .7 {JV-1 . 1‘: , ?' :,L._.i,‘ y W... . ’ ~. 1-K. 32-}‘71.’ an”... .Y. I (‘i ‘. ~....'. . LEVI. ‘ > A. 9.: ~— .""" 1il.'~ r "4‘ u r v -.'.,Tl‘..’.‘r ”.“WJ; . . . .W ',_. 7r ”lanrlw" ' '2‘"? r:4',r,v.'.n.'.‘.lio} .. .,. .7”. .- 4 I . r..- .53». "n, 1 .v’ m 1 4n flu!“ vmro. 3‘13, ‘ Nfivr h ’ NEWS VSER ITY LB! BSRARIE IHHHHHHHHHHHHHHHHHHHHHHH H|HH HHHHH HHHH 3 1293 00909 6425 HI This is to certify that the dissertation entitled A PRETEST MODEL OF ADVERTISING EFFECTIVENESS: THE MEDIATING EFFECTS OF INVOLVEMENT, PICTURE, AND MESSAGE presented by KAK YOON has been accepted towards fulfillment of the requirements for Ph.D. degree in Mass Media can/(71¢ fl fleece, I Major professor Date December 2, 1991 MS U is an Affirmative Action/Equal Opportunity Institution 042771 “A H' _ LIBRARY ‘ Michigan State He ——~—_ w‘H University J PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or bororo date duo. EAR 3 0199 DATE DUE DATE DUE DATE DUE _-- -_ Fifi" , H __'__-_._,_s ,..__!_ , , JAN 1 3 2000 HH Qg‘H fiH‘fir‘H MSU Is An Affirmative Action/Equal Opportunity Institution amiama-p. 1 H‘" H . A PRETEST MODEL OF ADVERTISING EFFECTIVENESS: THE MEDIATING EFFECTS OF INVOLVEMENT, PICTURE, AND MESSAGE BY Kak Yoon A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Advertising 1991 ABSTRACT A PRETEST MODEL OF ADVERTISING EFFECTIVENESS: THE MEDIATING EFFECTS OF INVOLVEMENT, PICTURE, AND MESSAGE By Kak Yoon we cannot emphasize enough the importance of being able to diagnose or predict the effectiveness of an ad in a pretest setting, especially in light of the sky-rocketing media cost. This dissertation identified message processing involvement, strength of claims, and picture attractiveness as important factors that influence subjects' purchase intention of a brand after exposure to the ad of the brand. A path model is proposed to explain the relationships among subjects' cognitive responses to ad and brand, attitude toward the ad (Aad), attitude toward the brand (Ab), and purchase intention (PI), Unlike most previous studies which viewed the influence of attitude toward the ad on brand attitude as a result of peripheral persuasion, this study showed that subjects' attitude toward the ad not only represents peripheral processing of non-issue-relevant peripheral cues but it also represents central processing mechanism. This is demonstrated by decomposing Aad into the claim-based component (Aad-c) and the non-claim based component (Aad-no). The author made further efforts to improve the usefulness of the proposed model by incorporating the notion of hedonic (Ab-h) and utilitarian (Ab-u) attitudes toward the brand. A series of hypotheses were tested regarding the path strengths among the constructs across involvement conditions, claim strength, and picture attractiveness. The causal analysis found a number of relationships. Cognitions played a more important role in the persuasion process when the subjects' involvement level was high than when it was low. The major route to PI was found to be the Aad-c to Ab-u path when involvement was high. In comparison, Aad-nc was the most important component leading to brand attitude in the low involvement condition. The effect of Ab- u became relatively weaker in the low involvement condition compared with the high involvement condition, as expected. Finally, the implications of the results and direction for future research are discussed along with the limitations of the study. ACKNOWLEDGMENTS The idea of writing a dissertation was so intimidating that I was never sure if I would have an opportunity to write acknowledgments when I started my Ph.D. study. 80, it is only fair to say that this piece of work would not have been possible without the endless help of many people. First, I owe a great deal to Dr. Bonnie Reece, my dissertation chair, who was always generous with her time for me. The whole process of writing the dissertation was greatly facilitated by her prompt examination and correction of the first draft. Second, I have to give special thanks to Dr. Carrie Heeter. She has not only always provided me with insightful thoughts and new ideas but also listened to my frustrations and offered a lot of sympathy. Finally, I thank Dr. Gina Garramone and Dr. Frank Boster who helped me with statistical analysis. I feel very fortunate that several former students of mine helped me throughout the whole process. First, Edward Dobbles and Linda Yuen coded the cognitive responses for the prelim paper. Wendy Trobeck did a lot of work for stimulus material development and pretests. Lisa Hajnal created the stimulus advertisements. Brad Sharp and Cheri Rossbach coded the cognitive responses very efficiently. The dissertation iv was completed on time due to their sincere dedication. They were the best collaborators that one could have. I also thank my wife who sacrificed in many ways to support me. My father, mother, grandmother, and two aunts in Alaska and California gave me unlimited mental and financial support. They always believed that I could finish it even when I was not sure myself. To all of them I would like to dedicate this humble work with a great deal of love. TABLE OF CONTENTS Chapter Page I 0 INTRODUCTION 0 o o o o o 000000000000 o o o o o o o ccccc o o o o o o o o o o o 1 K" usage Of Terms 0 o o o o o o o o ........... o 0000000000 o o o o O o o o 2 Attitude o o o o o o o o o o o o o o o o o o o o o ..... o o o o o o o o o o o o o o o 2 _Aadgandfléb.......................................5 'IhVEIVément......................................6 Purpose and Contributions ............................9 Plan of the Dissertation ............................10 II. LITERATURE REVIEW AND HYPOTHESES ....................12 Traditional High Involvement Models ........ .........12 Low Involvement Models ..............................17 Elaboration Likelihood Model ........................19 Attitude toward the Ad...............................22 \ Empirical Literature ...........................24 A Proposed Advertising Response Model ...............49 Hypothesized relationships...........................53 III. MTHOD ...OOOOOOOOOOOOO 0000000000 ...OOOOOOOOOOOOOOOOOSS Measurement Issues....... ........... ......... ..... ...55 Experimental Design ..... ..... .................... ...61 Selection of Product..... ........ ...............61 Development of Stimulus Materials ....... . ....... 63 Pretest for Manipulation of Claim Strength......64 Pretest for Manipulation of Picture Attractiveness.................... ......... ..71 Subjects and Procedure ................. ..... ...72 Manipulation of Involvement.....................74 Measures......................... ...... ..............79 Cognitive Responses ....... . ................. ....79 Attitude toward the ad .. ...................... .80 Attitude toward the Brand . ........ .............80 Brand Beliefs ................................... 81 Purchase Intentions ........ ......... ...........81 Manipulation Checks ..... . ...................... .82 Analysis of Cognitive Responses ...................... 82 IV. RESULTS ............................................. 87 Manipulation Check .................................. 87 Additional Evidence of the Successful Involvement Manipulation .......................... 91 vi Hypothesis Testing ................................... 93 Manipulation's Effects on Dependent Variables ........ 98 Purchase Intention .............................. 98 Attitude Toward Claims in the Ad ................ 98 Attitude Toward Non-claim Elements ............. 100 Dimensionality Assessment of Brand Attitude....102 Utilitarian Attitude toward the Brand .......... 107 Hedonic Attitude toward the Brand .............. 107 Brand Beliefs .................................. 108 Estimation of the Proposed Model ................... .110 Comparison of the Path Model by Involvement Condition.. ....................... ...115 Summary ..... .. ...................................... 118 DISCUSSION AND CONCLUSIONS ......................... 120 The Formation of Aad ........................... 120 The Formation of Ab. ...................... .....122 Discussion of the Model .................. ......123 The Ab to PI Path.... ..................... .....125 Managerial Implications ..... .... ............ ...126 Limitations and Suggestions for Future Research................... ...... .......127 Conclusion......... ...................... ......129 Appendix A: Questionnaire .......................... 131 Appendix B: Coding of Cognitive Responses .......... .140 BIBLIOGRAPHY ....................................... 142 vii LIST OF TABLES Table page 1. Operationalization of Aad in Literature.........56 2. Product Involvement Level.......................62 3. Cognitive Response Categories and Frequencies...86 4. Manipulation Check for Picture Attractiveness...88 5. Comparison of the Old and New Procedures used for Manipulation of Involvement..............89 6. Involvement Manipulation Check ........ . ......... 90 7. Correlations among Constructs............ ....... 95 8. Means of Dependent Variables by Experimental conditionOOOOOOOOOOOOOOOOOOOOOOOOOOOO0.0.00.097 9a. Items and Factor Loadings for Ab-u.............104 9b. Deviation Matrix for Ab-u......... ............. 104 10a. Items and Factor Loadings for Ab-h ......... ....104 10b. Deviation Matrix for Ab-h.... ................. .105 11. Deviation Matrix Parallelism Test..............107 12. Path Analysis Results for the Proposed Model...ll3 viii LIST OF FIGURES Figure page 1. Multiattribute Attitude Model ................ ....14 2. Four Alternative Specifications of the Mediating Role of Aad.. ......... .......29 3. . Miniard et al.'s Modified Dual Mediation Model...47 4. A Proposed Advertising Response Model ............ 50 5. Interaction Effects between Claim and Involvement on Aad-c..........................99 6. Interaction Effects between Claim and Involvement on Aad-nc.... .................... 101 7. Interaction Effects between Claim and Involvement on Brand Beliefs ................. 109 8. Path Model for the High Involvement Sample......117 9. Path Model for the Low Involvement Sample ...... .119 ix CHAPTER I INTRODUCTION Throughout the 19705, consumer researchers generally viewed consumers as rational and cognitive information processors. Therefore much research has emphasized consumers' beliefs about brand attributes as determinants of attitude toward the brand. Much research on advertising effectiveness, accordingly, has focused on measuring cognitive variables such as recall of ad content, brand attribute ratings, and importance weights of the attributes (MacKenzie & Lutz, 1989). In recent years, however, researchers have begun to place more emphasis on the study of consumers' affective responses to advertising stimuli. As a specific form of affective responses to advertising, the construct of attitude toward the ad (Aad) has been extensively investigated as a mediator of advertising's effects on brand attitudes and purchase intentions. This study is based on the assumption that consumers' attitudes toward a brand (Ab) can be determined by both their beliefs about the brand attributes and their attitudes toward the advertisement. Aad in this study is viewed as partially representing consumers' affective responses. These assumptions will be discussed more in detail later. USAGE OF TERMS This study basically uses two constructs (attitude and involvement) and two extensions of one of these constructs: attitude toward the ad and attitude toward the brand. There is some inconsistency in the literature with respect to the usage of the terms "attitude" and "involvement." It is therefore important to clarify the terms before proceeding further. ATTITUDE One popular definition of attitude is "a learned predisposition to respond in a consistently favorable or unfavorable manner with respect to a given object" (Fishbein & Ajzen, 1975, p. 6). Two aspects of the definition deserve special attention for marketing and advertising purposes. First, attitudes are learned. Consumers are not born with favorable or unfavorable attitudes toward a certain brand. Rather, they develop certain attitudes toward the brand through acquisition of information about the brand, for example, from advertising, or through direct experience with the brand, or some combination of the two (Lutz, 1991). Second, attitude tends to produce consistently favorable or unfavorable responses to an attitude object. According to Petty and Cacioppo (1981, p. 7), since Allport (1935) declared that attitude is "the most distinctive and indispensable concept in contemporary social psychology," the study of attitude Change has become the focus of persuasion researchers. Attitude became the focus of research activities because of the presumption that attitudes direct behavior and thus, by knowing attitudes, we can predict behavior. For the same reasons, marketing researchers are interested in studying consumer attitudes. Despite the high interest, the concept of attitude in marketing and advertising fields has been rather loosely used, thus creating much confusion. Lutz (1991) identified two distinct theoretical orientations in the study of attitudes. The first is often known as the tripartite approach because it views attitude as consisting of three components: cognition, affect, and conation. Cognition refers to consumers' beliefs about specific attributes regarding a brand. Affect represents consumers' positive or negative emotional reactions to a brand. Finally, conation pertains to intended and actual purchase behavior regarding the attitude object. Importantly, this orientation views that there is a consistency among the three components. The second, known as the unidimensional orientation, sees attitude as consisting of a single affective component which represents the degree of favorability or unfavorability regarding an attitude object. Beliefs (cognition) and behavior (conation) are not viewed as attitude components, but rather they are viewed as antecedents and consequence of attitude respectively (Fishbein & Ajzen, 1975). The unidimensional perspective specifically defines the causal relationships among the theoretical aspects of attitudes. Beliefs are causing attitude and, in turn, attitude causes purchase intention and actual behavior. Following the most popular view (Lutz, 1991), this paper adopts the unidimensional perspective of attitude. Attitude, therefore, is used to refer to a general degree of favorability or unfavorability toward an attitude object. Further, attitude is assumed to be the result of both cognitive (beliefs) and affective (emotional) responses to the attitude object. This position is an extension of the traditional unidimensional perspective in that it also encompasses the affective or emotional responses as antecedents of attitudes. The theoretical rationale of conceptualizing affective responses as antecedents of attitude can be found in the literature on two components of brand attitudes. Holbrook and Hirschman (1982) proposed that consumers' overall attitude toward a brand is based on two distinct kinds of evaluation. The first component of brand attitudes, termed utilitarian component, is presumed to be formed as a result of logical and rational evaluation of the utility associated with the consumption of a brand. This is represented by a means-ends analysis of attributes associated with an attitude object. This component reflects the view of consumers as rational information processors. They pointed out that the information processing perspective ignored some important aspects of consumption phenomena. Those ignored consumption phenomena include sensory pleasures, aesthetic enjoyment, and emotional responses. This component of brand attitude, which is termed hedonic attitude, is presumed to be based on the consumer‘s assessment of how much pleasure the consumption will give him or her. Batra and Ahtola (1991) later empirically investigated the validity of the two components of consumer attitudes. They performed confirmatory factor analysis and found that consumer attitudes consist of utilitarian and hedonic components. Aad and Ab Following Holbrook and Hirschman (1982) and Batra and Ahtola (1991), attitude toward the brand is viewed as consisting of two components. This along with the measurement issue will be discussed more in detail later. Attitude toward the ad is a recent term and there is no consensus as to the definition of the term. “Aad has been ---__. ...mk . - measured in a variety of ways by many different researchers in the past.; The history and measurement issue of Aad ideserves a lengthy discussion because it is the key construct iin the study. This will be delegated to chapter 2, but briefly, Aad inthe study is viewed as also consisting of two components: attitude toward the claim component of the ad, and attitude toward the non-claim component of the ad. Involvement Involvement and attitude change have been important topics in social psychology for almost half a century (see Johnson & Eagly 1983 for review). Consumer researchers also have extensively investigated the relationship between involvement and consumers' reactions to advertising messages (e.g., Batra & Ray, 1983; Gardner, Mitchell, & Russo, 1985; Gill, Grossbart, & Laczniak, 1988; Park & Young, 1986; Petty, Cacioppo, & Schumann, 1983) Researchers in this field, however, have not able to establish a consensus as to the definition of involvement due to the many different conceptualizations and operationalizations (cf., Andrews, Durvasula, & Akhter, 1990). Batra and Ray (1983) suggested that the lack of consensus is due to the fact that the term involvement is used to describe two different phenomena: involvement with a product class and involvement with a message . The product class involvement refers to the degree to which a consumer makes a brand choice with care and deliberation (Batra & Ray, 1983). In this conceptualization, involvement is viewed as being determined by the inherent characteristics of a certain product class and consumers are viewed as showing relatively stable and enduring predispositions to that product class (Houston & Rothschild, 1978). One possible intricacy of using this product class involvement as an independent variable in an experimental study is that the level of involvement for a specific product (a car, for an illustration) may not be constant across subjects due to some other factors. Personal relevance and motivations can greatly influence the product involvement. For instance, subjects who plan to purchase a car would process the ad differently than those who do not plan to purchase a car for various reasons (due to financial inability or lack of need, for example). Thus, although a car is generally a high involvement product, it may not have the same level of high involvement for all the subjects. For this reason it seems difficult to manipulate involvement by using this conceptualization of involvement in an experimental study. If manipulation of involvement is a key component of an experimental study, a better strategy is using message processing or message response involvement. It refers to the way a message is processed or elaborated by individuals (Batra & Ray, 1983). Most researchers agree that the level of involvement is determined by the degree to which the message has personal relevance or importance (Petty & Cacioppo, 1979; Park & Young, 1986). Message processing involvement, thus, is not an enduring predisposition but is a transitory, situational, state variable which is influenced by many situational factors (Batra & Ray, 1983; Gardner et a1., 1985). Batra and Ray (1983), for example, identified viewers' opportunity (due to media type and/or distraction), ability (due to knowledge and/or familiarity), and motivation (due to needs and/or goals, etc) to process the advertising message as the situational factors. Gardner et a1. (1985) also viewed involvement as a state variable and defined it as having intensity and direction. By intensity they referred to the amount of attention devoted to processing or elaborating of the advertising message. Direction refers to the strategy employed by the viewers in processing the message. Individuals with a brand evaluation goal would use strategies to acquire information about the product. Those individuals will try to evaluate the brand by fully activating their schema for the product and thus linking all the relevant information stored in memory. This is referred to as the "brand processing strategy." On the other hand, individuals who do not have a recognized interest in the product at the moment of exposure to the ad would process the ad by using the "nonbrand processing strategy." Gardner et al. (1985) identified "entertainment goal" (p. 5) as an example of the nonbrand processing strategy. In this situation, individuals may not fully activate their schema for the product and thus fewer inferences and linkages will be made about the brand from the memory. This notion of brand versus nonbrand processing appears to be similar to Petty et a1.'s (1983) central and peripheral routes. Individuals with brand processing goals are more likely to activate the central route while those with nonbrand processing goals are more likely to activate the peripheral route . This paper takes the position that involvement is an individual, internal, and motivational construct having intensity and direction (Gardner et al., 1985). As noted by Andrews et a1. (1990, p. 28) "it is the individual consumer who is involved, not products, or advertising content, media, objects, or situations." Therefore, involvement levels will be manipulated by intensity of attention (high versus low) and direction of attention (brand versus nonbrand processing). It is a legitimate concern that product class involvement may confound message processing involvement. To minimize this undesirable possibility this study will use a product class which falls in the middle of a scale measuring product class involvement as will be described in Chapter HI. PURPOSE AND CONTRIBUT IONS Although many empirical studies investigated the relationships among cognitive responses, Aad, Ab, and purchase intentions, the relationship between Aad and Ab still remains equivocal, especially in light of the role of message processing involvement. As will be elaborated in chapter II, Aad has been conceptualized in the literature as representing the "peripheral cues." Elaboration Likelihood Model (Petty & Cacioppo, 1981, 1986) would predict that Aad (peripheral cues) is more likely to influence message recipients' Ab when their message processing involvement level is low. However, many studies (e.g., Homer, 1990; Yoon, 1991) found the influence of Aad on Ab to be stronger 10 in high involvement condition than in low involvement pcondition. Assuming that ELM is correct, then it follows that either the conceptualization of Aad as a peripheral cue or the measurement of Aad (or both) is wrong. This study proposes that Aad represents both "central" and "peripheral" cues that lead to Ab and presents an advertising response model. This model delineates the causal relationships among different components of Aad and Ab and among other variables. The practical implication of the study lies in the diagnostic evaluation of the effectiveness of advertising stimuli in a pretesting setting. That is, by knowing in advance viewers' attitude toward different components of the ad and their relationship to brand beliefs, marketers can predict the effectiveness of the advertisement before placing it in the media. This topic becomes even more significant considering the fact that media costs have continued to accelerate at a greater rate than the inflation rate over the past two decades (MacKenzie & Lutz, 1989). PLAN OF DISSERTATION Chapter II provides background for this study. It begins with a literature review for the traditional information processing view which emphasizes the role of cognitive responses. Then it introduces newer models which recognize that consumers are not always highly involved with advertising messages. Next, the literature on attitude toward the ad and attitude toward the brand is reviewed. A 11 new model of advertising response is then presented with hypotheses. Chapter III begins with a discussion of measurement issues and then it explains the methodology employed to test the model and hypotheses. Chapter IV reports data analysis and Chapter V follows with discussion and conclusions. CHAPTER I I LITERATURE REVIEW AND HYPOTHESES Traditional High Involvement Models As a revolt against the behaviorism which had dominated the American academic society during the first half of the twentieth century, the late 19505 and the early 19605 saw the birth of a cognitive revolution (Gardner, 1985). It was called a revolution because the basic assumption of the cognitive paradigm was vastly different from that of behaviorism. In the behaviorist's perspective of human behavior, an individual is largely a passive creature who responds to a stimulus in the way predetermined by previous conditioning. In contrast, one of the most important assumptions of cognitive psychology is that human beings are active information processors (Ashcraft, 1989). Thus, human behavior is determined as a result of thinking (cognitive) process, not as a result of conditioning. This cognitive paradigm strongly influenced most social science disciplines dealing with human behavior, including the fields of marketing and advertising. According to Smith and Swinyard (1988), the traditional models of attitude change through advertising have been based on a general hierarchy-of—effects framework proposed first by Lavidge and Steiner (1961). Lavidge and Steiner (1961) claimed that advertising is a force which moves consumers through a series of steps to the final purchase behavior. 12 13 They specified the stages as awareness —-> knowledge ——> liking --> preference ——> conviction --> purchase. In this framework, the cognitive stages (awareness and knowledge) are regarded as significant predictors of subsequent affective stages (liking and preference), which in turn determine the conative stages (conviction and purchase). This model generally emphasizes the importance of product attribute beliefs (cognition) in determining consumers' attitudes toward the product and their subsequent purchase behavior and it was soon accepted widely. Notice that affect is equated with attitude in this model; this is consistent with the paper's definition of attitude as set forth in Chapter I. One sophisticated model of human attitude change couched in the general hierarchy-of—effects framework is the multi- attribute attitude model (Fishbein, 1963; Fishbein & Ajzen, 1975). Their model is expressed in an algebraic form that specifies the relationship between the set of salient beliefs about a concept (person, issue, object) and an evaluation of the concept. It is called multi-attribute attitude model because it assumes that attitude is formed as a result of learning and elaborating on multiple attributes associated with the attitude object as shown in Figure 1. Z biei =Ao where: bi = the strength of the association between the attitude concept, 0, and the ith salient concept, ei = the evaluation of the ith salient concept, A0 = the overall evaluation of, or attitude toward, concept 0, and, n = the number of salient beliefs Figure 1: Multiattribute Attitude Model Adapted from Mitchell and Olson (1981, p. 319) In a typical advertising study, the attitude toward a concept 0 (A0) is replaced by brand attitudes and salient beliefs are replaced by consumers' beliefs about brand attributes. Again, the multi-attribute attitude model posits that consumers form a brand attitude by evaluating numerous relevant and salient attributes of the brand. Positive and negative features associated with the brand are combined together compensatorily to produce an overall evaluation of a brand. Consumers establish brand preferences after comparing the overall expected value of different alternative brands. Note that in this model, to change attitude about a brand, consumers' beliefs and/or evaluations about the attributes associated with the brand should be changed first. In other 15 words, the basic theoretical proposition of the model is that beliefs cause attitudes (Mitchell & Olson, 1981). Another popular approach to studying attitude change under the high involvement paradigm is the cognitive-response model originally proposed by Greenwald (1968) and later made popular by Wright (1973). The basic premise of this approach is that when a person is exposed to a stimulus (advertising, for example) this person will compare the incoming information to his/her existing structure of beliefs and values, thus producing cognitive responses (Wright, 1973). He postulated that it is the cognitive responses or "primary thoughts" (p. 54) rather than the message itself that mediate message acceptance and determine the message recipient's attitude toward the message. He identified counter arguments, support arguments, source derogations and curiosity statements as the four types of primary thoughts that mediate message acceptance. All the models couched within the cognitive paradigm of human behavior share several important features. First, they all assume that consumers go through a series of complicated mental stages before making a purchase decision and that a preceding stage influences the following stage (e.g., cognition—-> affect). Second, they all emphasize the importance of cognitive responses and the ensuing beliefs in determining attitude and the ultimate conative responses. Thus, these models' general assumption of consumers is that they are computer—like, rational information processors. In 16 this framework, consumers are assumed to process advertising in a problem-solving mode, implying that they are highly motivated to process brand attribute assertions in the advertisements (Batra, 1986). It is also conceivable, however, that such effortful processing of advertising does not occur under many ad reception conditions. Batra (1986) argues that such motivated information processing does not occur often particularly in television advertising, not only because of the soporific and unengaging nature of the medium but also because of the processing conditions for television advertising. The motivation to process television advertising is generally low because of the intrusive nature of the commercials in that they are not deliberately sought out by the viewer. Moreover, television commercials compete not only in a cluttered environment among other commercials but also they have to compete against the television programs for viewer attention. One of the factors that determine message recipients' processing is their level of involvement with the message as explained in Chapter I. It should be apparent now that such thoughtful processing of advertising assumed in the traditional high involvement models of advertising effectiveness is not applicable under certain circumstances. Researchers have found some evidence that the Wright-type of cognitive response explanation of advertising effects is not capable of fully accounting for low involvement advertising processing. 17 For example, Wright himself (1973) found that the explanatory power of counter argument, support argument and source derogations was significantly lower in the low involvement condition than the high involvement one. Thus, some researchers (Batra & Ray, 1985; Batra, 1986) argue that the cognitive responses usually analyzed are not the complete set of processes underlying the mechanism through which advertising works. We therefore need a new perspective on how advertising works. Low Involvement Models As early as the middle of the 19605 when the cognitive paradigm was rapidly gaining popularity, Krugman (1965) asserted that the cognition-based model of television advertising effectiveness which requires changes in attitudes prior to changes in behavior was incomplete. He proposed a drastically different view of how television advertising works. His passive learning theory offers an explanation of how viewers, without paying close attention, can be persuaded by television commercials enough to buy the advertised product after repeated exposure, especially in low involvement situations represented by low price and low risk. The passive learning theory was developed from a stream of research called hemispheral lateralization or split-brain theory (Shiffman & Kanuk, 1987). According to this theory (cf., Flemming, 1981), the left and right hemispheres of the brain specialize in different kind of information processing. 18 The left hemisphere is primarily in charge of cognitive activities such as reading and speaking. The right brain, on the other hand, is mainly responsible for nonverbal and pictorial information processing. It is possible then that, when one is engaged in right-brain processing of information, one is able to process and store such information passively without conscious effort to do so (Shiffman & Kanuk, 1987). Thus, Krugman (1965) maintained that television advertising works not as a result of viewers' cognitive processing and elaboration but as a result of repetition and overlearning. Following Krugman's theory, research (e.g., Ray, et al., 1973) found empirical evidence that television advertising produced awareness-based behavioral change that preceded, rather than followed, attitude change. The key difference of this low involvement model from the traditional high involvement model, of course, is that purchase can precede attitude formation. This pattern of consumer response seems especially plausible for low price/low risk items and it was rapidly embraced by the advertising and marketing community (Smith & Swinyard, 1988). In this low involvement paradigm, several explanations have been offered concerning the link between advertising awareness and purchase behavior. Krugman argued that familiarity is the mechanism behind the link between the two. He said: 19 "Marketers would like to think that their products were indeed better and that consumers believed them to be better. What is often the unrecognized case, however, is that their product is neither liked nor considered better, but chosen only because it is adequately good and for the pleasure of its recognition (1968, p. 223)." More recent theories, however, suggest that it is not totally awareness through which low involvement advertising is changing purchase intentions. They (Batra & Ray, 1985; Batra, 1986) suggest that awareness may induce affect and this awareness-based affect is the mechanism behind the link between awareness and purchase intentions. As a rationale for the proposition they pointed to the research on mere exposure theory (Zajonc, 1968), which demonstrated that repeated exposure to a stimulus led to conditioned liking of the stimulus. Elaboration Likelihood Model Batra (1986) also cited research on two routes to attitude change as an indication that affect is a key mechanism in the low involvement hierarchy. Petty and Cacioppo's (1981, 1986) elaboration likelihood model (ELM) incorporated the traditional high involvement persuasion model and the low involvement model and it recognized that attitude change can occur through two different routes to persuasion. The first one, which they call the central route, is activated when attitude change occurs through 20 thinking about and elaborating on the issue (brand attributes) under consideration. This central route is consistent with the traditional high involvement model of attitude change which assumes a highly involved message processor. The second route—-termed peripheral--is activated when persuasion and attitude change occur as a result of exposure to peripheral persuasion cues, rather than elaboration of the issue-relevant message arguments. Petty and Cacioppo (1981) defined persuasion cues as "factors or motives inherent in the persuasion setting that are sufficient to produce an initial attitude change without any active thinking about the attributes of the issue or the object under consideration (p. 256)." In the context of studying the effectiveness of advertising, likability of the ad due to the executional aspects would constitute an example of such persuasion cues. To understand the causal influence of affect on purchase, especially in the low involvement situation, we need to discuss two postulates of the ELM concerning which route will be activated (Petty & Cacioppo, 1986). First, motivation and ability of the message recipient to process the message are two key factors that determine the likelihood of elaboration of the message. Specifically, when the recipient does not have the motivation and ability to process the message, it is hypothesized that the likelihood of the message being elaborated is low, and thus attitude change is more likely to occur through the peripheral route. This 21 hypothesis has been supported by empirical studies. Srull (1989), for example, found that when prior knowledge about the advertised product is low, simple affective cues are more important determinant of attitudes than attribute assertions. Second, the ELM postulates that personal relevance, or level of involvement with the issue or object under consideration, determines which route will be activated. When the message is highly relevant to the audience, factors central to the issue (such as attribute assertions) are more important determinants of attitude change, while in low relevance situations peripheral cues (such as attractiveness of the commercial) are more important. Gorn (1982) studied the effect of relevance or involvement on message processing and subsequent behavior by creating high and low relevance conditions for two groups. Subjects in the high-relevance condition were told that their task was to advise an advertising agency as to whether they should purchase time on television for the commercial-and they were also told that they were to choose later one of the advertised pens a5 a gift. Subjects in the low—relevance group were not given the instruction, and they had less motivation to elaborate on the commercials. Then all the subjects viewed two different commercials for a pen. One commercial contained attribute assertions and the other one featured pleasant music rather than information about attributes of the pen. Following the experiment, all subjects were given a choice between the two pens. The majority of the subjects in the high—relevance 22 condition chose the pen advertised with attribute assertions and the majority of the subjects in the low involvement chose the pen advertised with pleasant music. As discussed earlier, some advertising processing conditions may not be highly engaging to the consumer. Then it follows that under certain situations attitude change as a result of processing of an advertisement may occur through the peripheral route rather than the central route. Petty and Cacioppo (1989) cited the Gorn (1982) and Srull (1989) studies as demonstrating that "affect can act as a simple persuasion cue and influence attitudes when people's motivation and/or ability to engage in issue-relevant thinking is low (p. 83)." In recent years more and more researchers are interested in studying consumers' affective responses to peripheral cues in ads that lead to attitude change (Aaker, Stayman, & Hagerty, 1986; Batra & Holbrook, 1990; Holbrook & Batra, 1987). Attitude toward the Ad Following the theorization in Petty and Cacioppo's ELM, researchers in recent years have paid increased attention to the notion that peripheral cues in advertising stimuli may evoke favorable emotional, affective reactions that in turn will be transferred to the advertised brand (Batra & Ray, 1986; Holbrook & Batra, 1987). As a specific form of viewers' affective response to advertising stimuli a large 23 number of researchers have shown an interest in the construct of attitude toward the advertisement (Aad). Shimp (1981) first suggested that Aad consisted of both cognitive and affective dimensions. The cognitive dimension of Aad was conceptualized as a result of consciously processing specific elements of the ad such as execution, copy, and endorser. The affective dimension, on the other hand, was seen as those emotional feelings such as happiness, love, and sorrow which occurred without any conscious processing of executional elements. He also argued these different dimensions would have differential effect on viewer's attitudes toward the advertised brand because of the different underlying mechanisms. Lutz (1985) noted that little systematic study had been conducted concerning the conceptual origins of Aad. Consistent with Petty and Cacioppo (1981, 1986) and Shimp (1981), he also viewed Aad as having different degrees of cognitive and affective antecedents such as perceived credibility of the ad (cognitive response through central route) and mood (affective response through peripheral route), for example. He offered the first definition of Aad as "a predisposition to respond in a favorable or unfavorable manner to a particular stimulus during a particular exposure occasion (p. 46)." This definition of Aad does not embrace the two-component notion and it does not distinguish between affect and evaluation (MacKenzie & Lutz, 1989). MacKenzie and Lutz (1989) proposed that the cognitive and affective 24 reactions to an advertising stimulus should be treated as antecedents to the general attitudinal response to a specific advertisement denoted as Aad. Lutz (1985, p. 46) also noted the transitory nature of Aad by saying that "Aad focuses on a particular exposure to a particular ad and not on consumers' attitudes toward advertising in general or even their attitudes toward the ad stimulus of interest at another point in time." Many researchers have empirically investigated the relationship between Aad and Ab and have documented the significant explanatory power of Aad. Presented below is a review of the empirical literature. Empirical Literature Researchers investigating the relationship between viewer's Aad and subsequent brand attitudes have focused on the advertisement itself as a possible mediating variable in determining brand attitude and subsequent purchase intention (PI). Most studies found that attitude toward advertising execution itself led to changes in brand attitude, in addition to the attitude change caused by attribute arguments. Mitchell and Olson (1981) investigated if the Fishbein type brand beliefs are the only mediators of brand attitudes by conducting a laboratory experiment. They employed a 4 X 4 Latin Square design with four groups of subjects. The first factor was ad executions. They manipulated executional 25 elements in four experimental advertisements for four hypothetical brands (the letters I, J, L, and R) of facial tissues. Of the four ads, only one provided an explicit verbal claim of the attribute "softness." Each of the other three ads was designed to convey an image. Each contained only a brand name headline and a half-page color photograph but had no explicit verbal claim of the attribute. Of the three "image" ads only one was designed to communicate the attribute "softness" by showing a picture of a fluffy kitten. The third ad contained a picture of a sunset designed to be positively evaluated and the fourth ad showed a picture of an abstract painting assumed to have a neutral evaluation. The second factor was the number of repetitions for each ad. Each of the four ads was shown on slides for ten seconds each and the ads were repeated either two, four, six, or eight times. Each group of subjects viewed a total of twenty ads for the four brands. After the experiment, a questionnaire measured typical Fishbein cognitive measures such as attribute belief strength (bi) and attribute evaluation (ei), brand attitude, attitude toward the act of purchasing and using the brand (Aact), and purchase intentions. Additionally attitudes toward the ad and attitude toward the picture (Apic) were assessed. They found that product attribute beliefs had a major mediating effect on subjects' brand attitudes. However, they also found that attribute beliefs were not the sole mediator of Ab. Aad was found to have significant effects on Ab, Aact, and purchase 26 intentions. They also found that Apic mediated Ab but only when Aad was not included in the analysis. Repetition was found to have no reliable effects on attitude formation or any other cognitive variable.. Moore and Hutchinson (1983) investigated how affective reactions to advertising (Aad) influence attitude toward the brand. Their contribution was the investigation of the effects of delay on the relationship between Aad and Ab. The experiment was conducted in two stages. In the first stage, all subjects completed a questionnaire (Brand Test 1) which asked two questions about each of the eighty brands coming from twenty product categories used for the study. The first question assessed subjects' knowledge of the brand's product category. The second question measured brand attitude indirectly by asking subjects' likelihood to purchase the brand in the future. Immediately following Brand Test 1 subjects were shown the eighty magazine ads projected on slides for the eighty brands. Then subjects' Aad was measured by a five-point Likert scale (Ad Test). Immediately following the first Ad Test, subjects' Ab was again assessed using the same scales used in Brand Test 1. The second stage of the experiment concerned the effects of delay. Subjects' Ab was measured either after two days or seven days delay. During the second session subjects' Ab were assessed following a second ad exposure and ad test. The results showed a positive linear relationship between Aad and Ab immediately following exposure to the ad. After a two-day 27 delay, the relationship persisted. After a seven—day delay, however, a J-shaped relationship developed between Aad and Ab. They attributed this phenomenon to sleeper effects or the "law of extremes" hypothesis (Silk & Vavra, 1974). According to this finding, after a sufficient delay the initial negative reactions to an ad may disappear and result in a greater probability of brand consideration. The studies reviewed so far investigated the effects of Aad on various dependent measures (such as attribute beliefs, brand attitudes, brand consideration, etc.) by using ANOVA and regression analyses. One limitation of the studies is that they examined the effects of Aad on the dependent variables separately. Our understanding of the mediating role of Aad has been vastly enhanced by a series of experiments conducted by Lutz, MacKenzie, and Belch (Lutz, MacKenzie, & Belch, 1983; MacKenzie, Lutz, & Belch, 1986). In their 1983 study they considered the relationships among the variables simultaneously and offered four alternative causal models as illustrated in Figure 2. All four alternative models are based on the general hierarchy-of— effects framework which assumes that cognition precedes affect which in turn influences behavior. Thus, in all the proposed models brand cognitions are conceived to influence attitude toward the brand, and attitude toward the brand influences purchase intention. Similarly, ad-related cognitions are posited to influence attitude toward the ad. 28 The affect transfer hypothesis is consistent with Petty and Cacioppo's ELM in that it posits a direct causal path from Aad to Ab (peripheral route) and also a path from Cb to Ab (central route). The dual mediation hypothesis (DMH) is the same as the affect transfer hypothesis except for the causal flow from Aad to Cb. The rationale of adding an indirect causal influence of Aad to Ab through Cb is that viewers' affective responses to ads can influence the way the message is processed. The reciprocal mediation hypothesis posits another path flow from Ab to Aad in addition to the same path specifications of the affect transfer hypothesis. This model finds its rationale in balance theory (Heider, 1946) and assumes that consumers' attitude toward the ad and attitude toward the brand will be consistent. Finally, the independent influence hypothesis posits no causal relationship between Aad and Ab. This model was developed from Howard's (1977) concepts of two attitudinal constructs: an "evaluative element" of the brand concept and "impersonal attitude." The evaluative element of the brand concept in this model was represented by Ab. Impersonal attitude was analyzed by Howard as representing consumers' feelings about the conditions of purchase such as price, availability, and deals. This impersonal attitude is supposed to be situation— specific and less stable as compared to the brand concept. In the independent influence model Aad was conceptualized as being situation—specific and thus corresponding to the impersonal attitude. Therefore the model assumes that Aad 29 and Ab independently influence purchase intention. (See MacKenzie, Lutz, & Belch, 1986, for a detailed explanation) 1. Affect Transfer Hypothesis 2. Dual Mediation Hypothesis Cad ____.. Aad Cad __., Aad Cb __.;A6}__...Ib Cb ___.,Ab__,1b 2. Reciprocal Mediation Hypothesis andependent Influence Hypothesis Chd ._ Ami Chd ..Amd "\ 7 ‘\ Cb____> @b’__., lb Cb ______> Ab___,, lb) Figure 2: Four Alternative Specifications of the Mediating Role of Aad Key: Cad represents ad cognitions Cb represents biand cognitions Aad represents attitude toward the ad Ab represents attitude toward the brand Ib represent intention to purchase the brand Source: MacKenzie et al. (1986), p. 131. 30 The focus of Lutz, MacKenzie, and Belch (1983) study was to examine whether the causal structure among the variables is affected by message recipients' processing involvement.1 Two specific hypotheses were developed from the notion of central and peripheral routes to persuasion (Petty & Cacioppo, 1981). They viewed Aad as representing the peripheral route. Therefore, the first hypothesis predicted that viewers low in both motivation and ability to process information would exhibit a relatively strong influence of Aad on Ab and a relatively weak influence of Cb on Ab. They viewed Cb as representing the central route, thus the second hypothesis predicting that viewers high in both motivation and ability to process information would show a relatively weak Aad -> Ab link and a relatively strong Cb -> Ab link. To test the hypotheses they conducted an experiment in which they used a television commercial for a hypothetical brand of shampoo embedded in a one-hour television program as the experimental stimulus. As anticipated, Aad was a strong mediator of Ab in low involvement condition. Also, consistent with the ELM, Cb -> Ab link was significant only for the high involvement group. Contrary to expectations, however, Aad —> Ab link was stronger for the high involvement group than the low involvement group. They offered several possible explanations for the pattern of findings. First, 1 Motivation and ability are conceptualized as the antecedents of message processing involvement. For a detailed discussion of the construct involvement refer to Chapter 1. 31 the construct involvement was not properly manipulated. Subjects all might have been limited to low levels of processing involvement. Second they suspected the quality of Cb measurement failed to capture the nature of the central processing of the commercial. They concluded that Aad is an important causal mediator of Ab in any processing situation. MacKenzie and Lutz (1983) tested the four alternative causal models using a commercial for the Ford Motor Company. Because of the strong correlation between Aad and Ab they excluded the Independent Influence Hypothesis (Model 4). They found that Model 1 (Affect Transfer Hypothesis) fit the data the best. MacKenzie, Lutz, and Belch (1986) again tested the four alternative causal models about the mediating role of Aad using data from two experiments. They pooled the two samples together because the measurement scales used in both experiments were identical. This time, unlike the MacKenzie and Lutz (1983) study, the dual mediation hypothesis (Model 2) best explained the relationship among the above mentioned variables. The difference, however, between Model 1 and Model 2 was slight in terms of fit of the data. As expected, Aad exerted a strong positive influence on Ab. Aad also was found to exert moderate positive influence on Cb. They noted that by the inclusion of the Aad -—> Cb link, this model departs from the ELM. The ELM posits that attitude change occurs either through Cb —-><ég>route (central) or through Aad -—><:? route (peripheral) depending on message 32 recipient's processing involvement. Thus, MacKenzie et a1. (1986) argued that the DMH is presenting an alternative to the ELM model in that central and peripheral processes are intertwined rather than substitutes for one another. One of the most intriguing and "disturbing" (p. 141) findings of their study, however, is the relationship between Cb and Ab. Contrary to the cognitive response literature (e.g., Wright, 1973) or expectancy value literature (e.g., Fishbein & Ajzen, 1975), this study failed to find a significant causal flow from Cb to Ab. They offered one methodological and one substantive speculation concerning this finding. First, the unknown quality of the Cb measure, generated from the open-ended verbal cognitive responses, might have been the reason for the finding. Second, because of the low involvement nature of the product used for the experiment (toothpaste), it is possible that only the peripheral route might have been invoked. They explained, "the combination of an unimportant product class, a low involvement medium, and instructions to attend to the ad may have been so powerful as to eliminate brand-attribute-based effects on Ab" (p. 141). Gresham and Shimp (1985) investigated the viability of classical conditioning explanation of the Aad -> Ab relationship. They reasoned that to satisfy the classical conditioning model, Aad, after multiple exposure, should influence Ab without changing subjects' cognitive structures (measured by Fishbein—type brand beliefs and evaluations). 33 They used fifteen actual television commercials which were found from a pretest to elicit positive, neutral, and negative reactions. Subjects were randomly assigned to five experimental groups and one control group. Each experimental group was shown a set of three stimulus ads consisting of one positive, one neutral, and one negative commercial. Their major hypothesis predicted that subjects exposed to a positively valenced commercial would exhibit more positive brand attitude than the control group. Similarly, subjects exposed to negatively valenced ad would exhibit more negative brand attitude than the control group. In other words, they predicted that affectively-valenced (positive or negative) ads would have a significant influence on subjects' attitude toward the advertised brands. However, Aad was found to be a significant predictor of Ab for only three of the ten affectively-valenced commercials. The lack of a strong support for the Aad -> Ab relationship was attributed to the use of potentially familiar commercials for mature brands as stimuli. They concluded that the lack of support for hypotheses designed to rule out alternative mechanisms leaves the classical conditioning mechanism open to question. Gardner (1985) investigated whether viewers' involvement levels affect the mediating role of Aad on Ab. High involvement was operationalized by the "brand evaluation set" and low involvement as the "nonbrand evaluation set" (cf., Gardner, Mitchell, & Russo, 1985). Two mockup print ads (one for tennis ball and one for cooking oil) for two hypothetical 34 brands were used as stimuli. The results showed that brand beliefs were a more significant mediator of Ab under a brand set than a nonbrand set condition. She also found that subjects in the brand set condition used Aad as much as those in the nonbrand set condition. These results are basically the same as the findings reported by Lutz, MacKenzie, and Belch (1983). Gardner suggested that conclusions based on previous nonbrand set experiments may be generalizable to brand set conditions and that investigation of the role of brand sets (processing involvement) need not be a top priority in future studies involving Aad. (yBatra (1985) tested a similar model called the "Override Model." This model is based on several propositions. First, likable ads generate affective responses in viewers. The frequency of the cognitive responses (support arguments and counter arguments) is the outcome of the message recipient's processing involvement. When the processing involvement is high the number of cognitive responses produced is high. Then, the production frequency of the affective response is reduced. Finally, likable message executions (Aad) influence Ab when the processing involvement is low. His concept of the affective responses consisted of subjects' written reports of moods and feelings elicited by the ad which were later coded into several categories of feelings. He tested this model by using forty thirty-second television commercials. The results were equivocal but supported the hypotheses at least partially. He argued that message 35 recipients may use the "likability cues" (p. 366), peripheral cues in an ad, to form.Ab only when cognitive responses to product attributes do not or cannot occur. Batra and Ray (1985) proposed an advertising response model called "The Percentage Contribution Model" which specified different causal relationships among Aad, Ab and purchase intentions depending on the involvement levels. They measured Aad by the affective response categories reported in Batra (1985). One distinctive feature of the model is the concept of two components of Ab. They argued that Ab is composed of an attribute-sensitive "utilitarian" component (measured by useful-useless and important- unimportant) and an execution-sensitive "hedonic" component (measured by pleasant-unpleasant, nice-awful). The model predicted that when the viewer involvement level is high the major route to purchase intention would more likely be attribute argumentation in an ad -> cognitive responses produced by the viewer -> attribute-sensitive component of Ab -> PI. On the other hand, when the involvement level is low the major route would be execution likability -> affective responses -> execution-sensitive component of Ab -> PI. They again used forty television commercials to test the conceptual model. Although the model makes much heuristic sense the data did not support the model. Most importantly, the results showed that the two Ab components turned out to be the opposite of the expectations. The feelings of social affection turned out to influence the utilitarian component, 36 while cognitive responses (counter and support arguments) influenced the hedonic component. They suspected that there might have been a "tremendous amount of concept-scale interactions" which confounded the results. This is a plausible explanation because they used forty commercials from ten product categories. Moreover, since they used existing commercials for real brands, viewers' prior attitude toward the ad and the brand may have blurred the picture. They admitted that due to the ambiguity regarding scale interpretation, the model testing was less than satisfactory and suggested that future studies be confined to single product categories. Burke and Edell (1986) noted the phenomenon of television commercial wearout and investigated whether attitude toward the ad also wears out after delay. The major contribution of this study is that, unlike most studies conducted in a laboratory setting, they studied viewers' reactions to commercials over time in a naturally occurring situation. They designed the study so that subjects' attitude toward nine actual television commercials for well known brands were measured after single exposure, normal exposure, intense exposure, short delay (one week after), and long delay (eight months). The single exposure condition was used as a measure of initial Aad. Aad declined significantly between the single exposure and the normal exposure levels and between the normal and intense exposure levels, showing that overall Aad decline over time. This finding supported 37 the notion of advertising wearout after multiple exposures. Concerning the effects of delay, they found that after one week delay Aad was not different from the baseline measurement. After eight months, however, Aad improved significantly fromche Aad measured after intense exposure. The pattern of Aad score decreasing after extensive exposure and then improving back to the initial level after long delay was prominent only for the commercials which had high initial Aad scores. Park and Young (1986) investigated the effects of two factors, involvement and background music, on viewers' television commercial processing and formation of brand attitudes. They suggested that high involvement should be differentiated into two types. An advertisement can be highly involving to a consumer because it gives information on the brand's functional performance or because it appeals emotionally to one's motive to express an actual or ideal self-image. The former situation will lead to high cognitive involvement and the latter will lead to high affective involvement. They proposed that the types and levels of involvement during exposure to a commercial and the presence of peripheral cues (music, in this study) in the commercial would influence the process of brand attitude formation. Specifically, they developed hypotheses regarding the relative importance of Aad, cognitive responses (CR Model), and Fishbein-type expectancy value (EV Model) measures in determining Ab. This was considered in relation with the 38 different types and levels of involvement and presence of music in the commercial. The hypotheses were tested by a 3 (high cognitive vs. high affective vs. low involvement) x 2 (presence of music) factorial design. They found that the effect of music on brand attitude formation depends on the type and level of involvement. The results showed that Aad, in comparison with CR and EV, was a more significant mediator of Ab in the affective and low involvement conditions, particularly the latter. In contrast, Aad was not a significant explanatory variable of Ab for the cognitively involved group. For this condition CR and EV were significant mediators of Ab. As expected, music enhanced Ab in the affective and low involvement conditions. Subjects in the cognitive condition revealed lower Ab when exposed to the commercial with music, suggesting that peripheral cues may distract subjects who are highly involved with an ad. An important finding is the similarity in results found for the affective involvement group in both music and no music conditions and the cognitive music condition group. The correlation between Aad and Ab was high for the three conditions, although the correlation coefficient for the cognitive music group (.40) was statistically not significant. Park and Young (1886) interpreted this as indicating that Aad may operate to some extent as a mediator of persuasion cues even under high cognitive involvement situations. This finding is consistent with Lutz, MacKenzie, and Belch (1983) study which showed that Aad was a 39 significant mediator of Ab even under high involvement conditions. Mitchell (1986a) examined the effects of visual and verbal elements of advertisements on brand attitudes and attitude toward the ad. To examine the effects of visual content of ads he manipulated the valence of pictures (positive, neutral, negative, and copy-only) used in the test print ads while holding the verbal content (ad copy) constant. The results showed that the positively-valenced picture created better Ab, demonstrating that visual content of an ad can create different brand attitudes. This finding supports the notion that product attribute beliefs are not the only determinant of brand attitudes (Mitchell & Olson, 1981). Thus, the results demonstrated that consumers may develop different attitudes toward a brand depending on the attractiveness of the visual content of the ad even though their brand attribute beliefs are the same. Also, the visual component in ads affected the formation of product attribute beliefs and attitude toward the advertisement. Both the visual and verbal components of ads were also found to influence Ab. Importantly, the results suggested that although Aad is strongly influenced by the attractiveness of the visual content of the ads, Aad represents viewers' evaluation of the entire advertisement. This was demonstrated by the finding that subjects' attitudes toward the brands were also influenced by their attitude toward the ads even when the ads contained only copy. Finally, the- 40 study showed that the construct of attitude toward the advertisement is empirically separate from attitude toward the brand. Moreover, multiple regression analysis showed that Aad is the most significant predictor variable of Ab compared with brand beliefs measures and elicited brand beliefs measures (similar to the cognitive response measures of Wright, 1973). , Machleit apdufiiisop,(1988) extended the Aad research by incorpSPEti;;flthe issues of brand familiarity and ad repetition. They argued that the Aad -> Ab relationship could be inaccurate if the effect of prior Ab on Aad is not controlled. They posited that if the affect-transfer- hypothesis based on classical conditioning theory is correct the Aad -> Ab relationship would increase as ad repetition increases. Their results added some new insights to the Aad research stream. The Aad -> Ab link was significant only for the two unfamiliar brands. The link was not significant for the familiar brands when the effect of prior Ab was controlled. The effect of repetition on Aad -> Ab link was unclear. One unfamiliar brand provided evidence of affect- transfer effect as the correlation between Aad and Ab increased with repetition. For the two familiar brands, again, correlation between Aad and Ab was insignificant. With repetition, interestingly, the direction of correlation became negative, thus refuting the affect-transfer hypothesis. The correlation between Aad and Ab for the other unfamiliar brand was mixed. Therefore Machleit and Wilson 41 concluded that the direct-affect-transfer hypothesis is quite context-specific. Muehling and Laczniak (1988) examined whether Aad's and brand beliefs' influence on Ab persists over time (one week delay). This was investigated also in relation with the moderating influence of message involvement. Subjects' Aad, brand beliefs, cognitive responses, and Ab were measured immediately after exposure to a mock print ad for a fictitious cassette player. Subjects' Ab was again assessed one week later. Immediately following exposure to the ad, both brand beliefs and Aad had a significant impact on Ab for the high involvement condition. For the low involvement condition, in contrast, only Aad was the significant mediator of Ab. This result is consistent with Petty and Cacioppo's (1981, 1986) ELM in that the link between Aad (peripheral cue) and Ab was stronger for the low involvement condition (.60) than the high involvement condition (.49) although the two were statistically not different. For the high involvement subjects, Ab measured a week later was still mediated by beliefs and Aad. When message involvement was low, one week delayed measure of Ab was still influenced by initial Aad. Brand beliefs, again, had no significant influence on the delayed measure of Ab. These findings suggest that Aad is an important mediator of Ab under all involvement conditions. This applies to Ab measured immediately after ad exposure and one week later. The influence of brand beliefs on Ab, on the other hand, seems to 42 diminish significantly when involvement with the message is low. Interestingly, Aad's impact remained the same even after one week's delay for both high and low involvement conditions. Burton and Lichtenstein (1988) contended that Aad consists of cognitive and affective components (Shimp, 1981) and hypothesized that the two components might have a differential impact on the dependent variables such as Ab and PI (Batra & Ray, 1985; Shimp, 1981). Following Lutz' (1985) claim that Aad has various cognitive and affective antecedents, they studied the effect of several such antecedent variables, each of which required some degree of cognitive processing, on the dependent variable. Specifically, they examined the relationships between: (1) the content of the ad (high, medium, and low discount of price offered in the ad) and.Aad; and (2) two contextual variables (the consistency with which the advertiser has advertised such an offer, and the distinctiveness of the offer compared with the competition) and Aad. The relationships between the two components of Aad and subjects' attitude toward the deal was also explored. Test print ads for a desk were used as stimulus materials in an experiment in which the three variables (price discount, consistency, and distinctiveness) were manipulated. First, the results showed that the two factor structure consisting of cognitive and affective components is a better representation of Aad than a single factor structure. Regarding the effects of the 43 antecedent variables, the results supported the contention that cognitive antecedents can affect Aad. The consistency of the deal had a significant impact on Aad. Importantly, the influence was via the cognitive dimension of Aad. The distinctiveness of the deal did not affect Aad. The level of price discount had a significant impact on both dimensions of Aad, although the size of impact was bigger on the cognitive dimension. The results also indicated that both cognitive and affective dimensions of Aad were significant mediators of attitude toward the deal. Homer (1990) extended MacKenzie et a1. (1986) study by incorporating the effects of involvement on the path specifications among Cb, Ab, Cad, Aad, and PI. She reported the results of two experiments using a television commercial and a print ad respectively. The television commercial was for a hypothetical brand of shampoo and the print ad was for a phone service. For the involvement level manipulation, Gardner et al.‘s (1985) concept of brand versus nonbrand processing was employed. As a second conceptualization of involvement, self-reported measures of high versus low product knowledge/importance were used. The results of both experiments showed a significant path coefficient between Cb and Ab. She contended that the insignificant coefficient of the Cb -> Ab link reported in MacKenzie et a1. (1986) article was due to a methodological limitation. Consistent with the MacKenzie et a1. (1986) study, the results showed that the dual mediation hypothesis model provided the best fit of the 44 data. Contrary to expectations, neither processing involvement nor product knowledge/importance produced significant differences in the path specifications. Consistent with MacKenzie et a1. (1986), Lutz et a1. (1983), and Gardner (1985), the Aad -> Ab link was significant under both brand and nonbrand processing conditions. Yoon (1991) pointed out several limitations of MacKenzie et a1. (1986) and Homer (1990) studies and conducted an experiment closely replicating the previous studies. Both studies used a commercial for a new brand of relatively low involvement products (toothpaste and shampoo, respectively) as experimental stimuli. It has been widely supported that message processing may occur through either central or peripheral route depending on the message recipient's level of involvement which in turn is determined by the motivation, ability, and opportunity to process the message (Petty, Cacioppo, & Schumann, 1983; Batra & Ray, 1983). Thus, the subjects‘ motivation to process the message in the two studies might have been extremely low due to the relatively unimportant product class. Second, Homer found that processing involvement did not produce significant differences in the path specifications among the variables, especially in the Cb -> Ab link. This result is highly unintuitive and it is inconsistent with previous research findings (Batra, 1985). ,Homer's (1990) failure to find significant differences between the high and low involvement groups in terms of the strength of the Ch -> Ab link might 45 have stemmed from the weak manipulation of the message processing conditions. Homer also acknowledged that the two processing sets might have been in fact different moderate levels as opposed to the extreme low and high levels representing the elaboration continuum end points. Yoon (1991) conducted a study and reexamined the findings of the previous studies by employing a television commercial for a relatively high involvement product class (automobile tire) as a stimulus material. Subjects' message processing involvement was manipulated by using a similar set of instructions reported in Park and Young (1986). The path analysis results showed that the Affect Transfer Hypothesis provided the best fit of the data instead of the Dual Mediation Hypothesis. Consistent with the ELM paradigm, the path coefficient between Cb and Ab was significantly stronger for the high involvement condition (.36, p<.01) than the low involvement (.29, p=.03). This finding suggests that when subjects' involvement with the product or the ad is low, the influence of brand cognitions on Ab diminishes. This also indirectly explains the insignificant Cb -> Ab link reported in the MacKenzie et a1. (1986) article. Contrary to the ELM and recent attitude toward the ad research stream which posits Aad as representing the peripheral route, the Aad -> Ab link was found to be stronger for the high involvement condition. In fact, the link was not significant for the low involvement condition at a .05 alpha level. This result is consistent with previous research (Gardner, 1985; MacKenzie 46 et al., 1986). Homer's (1990) study also revealed the same pattern concerning the link from.Aad to Ab, with the path coefficients of .91 for the brand evaluation subsample (high involvement) and .86 for the nonbrand evaluation (low involvement) subsample. Then we must suspect that either the measurement or the conceptualization of Aad as a peripheral cue is incorrect. A possible answer to the puzzling link between Aad and Ab was offered by Miniard, Bhatla, and Rose (1990). They claimed that Aad should be decomposed to represent both central and peripheral cues to persuasion. In other words, attitude toward the advertisement, representing an overall evaluation of the whole advertisement, consists of central and peripheral components (cf., Shimp, 1981; Lutz, 1985). The central component represents a cognitive evaluation of the ad based on its attribute assertions. The peripheral component, on the other hand, represents various affective responses to the peripheral cues in the ad such as attractiveness of the picture. Thus, they contended that since most advertisements contain both claim and non-claim elements Aad can influence Ab both under low and high involvement situations. 47 Non-Claim > Aad-nc COQflIthI‘IS Claim / Aad-c Cognitions Y Brand > Ab ———> Pl Cognitions Figure 3: Miniard et al.‘s Modified Dual Mediation Model Miniard et al.‘s (1990) revised dual mediation model is presented in Figure 3. By decomposing Aad into claim (Aad-c) and nonclaim (Aad-no) elements, they demonstrated that the conceptualization of Aad as a purely peripheral persuasion cue was incorrect. Consistent with the ELM predictions, the influences of all the cognitive responses (non-claim, claim, and brand cognitions) on the relevant attitudes (Aad-nc, Aad- c, and Ab) were stronger for the high involvement condition than the low involvement one. Miniard et a1. (1990) tested the influence of claim strength on subjects' attitude formation. All marketing communication, particularly advertising, is an attempt to move consumers from one mental state to another (Ray, 1982). Ray (1982) explained that one important factor which influences consumer attitude is the strength of claims made in an ad. The common assumption seems to be that the 48 stronger the claims, the greater the effect of the message ‘will be on target consumers. Quoting Maloney (1963), however, Ray (1982) argued that the effect of claim strength would be mediated by the believability of the claims. In sum, a strong claim that causes “curious nonbelief (Ray, 1982)" will be more effective than a weak claim (which is highly believable because it simply states what we already know). The results of Miniard et a1. (1990) study showed that in the high involvement condition strength of claims made in the ad affected subjects‘ Aad and Ab. In the low involvement condition, subjects' Aad and Ab were much less sensitive to the claim manipulation. Subjects in the high involvement condition based their Ab on the ad claims, while low involvement subjects' Ab was derived from peripheral cues. Aad was influenced by both the claim and picture manipulation under high involvement conditions. Aad of the subjects in the low involvement Condition, however, was only influenced by picture manipulation. The overall Aad influenced both Ab and Cb under all conditions. Decomposition of the overall Aad into claim and nonclaim components offers us new insights. For subjects reporting high involvement, only the Aad-c affected Ab.2 In contrast, for subjects reporting low involvement, their Ab was affected 2 The Aad-nc to Ab path was stronger for the ”self-reported” low involvement group than the high involvement group, thus seemingly supporting the peripheral route assumption of the ELM. This, however, is not clear because the manipulated high involvement group (as opposed to the self-reported) showed stronger path than the low involvement condition. This obviously demands further investigation. 49 by both Aad-c and Aad-nc. The authors, thus, claimed that the current conceptualization of Aad as affecting Ab only under low involvement conditions should be revised. This claim was supported by the finding that much of Aad's influence on Ab, and all of Aad's influence on brand cognitions was through Aad-c rather than Aad-nc. A Proposed Advertising Response Model Based on the above discussion of various constructs including Cad, Aad, Cb, Ab, and PI, an advertising response model is proposed as shown in Figure 4. The proposed model combines Miniard et a1.'s (1990) and Batra et al.'s (1985) models. This model hypothesizes that the claim cognition -> Aad-c -> Ab-u will be the major path to purchase intentions in the high involvement situations. On the other hand, non- claim cognitions -> Aad-nc -> Ab—h will be the major path to purchase intentions in the low involvement situations. This model clearly offers a solution to the unintuitive results regarding the stronger path from Aad to Ab reported in Homer (1990) and Yoon (1991) studies. Therefore it will contribute new insights to the existing body of literature about the role of Aad in the formation of Ab. 50 Non-claim A d Cognitions 7’ 3 -nc Claim Cogntions ’ Aad-c V Ab-h Brand Cognitions P I Ab-” / Figure 4: A proposed Advertising Response Model 51 This model views Aad as representing viewers' cognitive responses to the claim element and affective responses to the peripheral elements in the advertisement. In the past researchers conceptualized Aad as representing the peripheral route (e.g., MacKenzie et al., 1986; Gardner, 1985; Homer, 1990; Yoon, 1990) and predicted that the Aad -> Ab link would be stronger for the low involvement condition. These studies, however, found that the link was stronger for high involvement conditions. This is clearly contradictory to the predictions made by the elaboration likelihood model (Petty & Cacioppo, 1981, 1986). Perhaps the low involvement manipulation was not successful and the condition did not represent the extreme low involvement condition as envisioned in the ELM model. The heightened awareness of the test ads due to the experimental situations may have caused the unintuitively high Aad -> Ab link for the high involvement conditions reported in the studies. Park and Young's (1986) concept of "high affective involvement" may better represent the "low involvement" in the other studies. In sum, it may not possible to simulate a "low involvement" condition in an advertising pretest context. However, the argument of "failure to create low involvement condition" loses its validity when confronted by the findings of stronger Cb -> Ab link for the higher involvement conditions (Homer, 1990; Yoon, 1991). As Miniard et a1. (1990) contended and as the results of Homer and Yoon studies indicate, the conceptualization of Aad as a pure peripheral cue should be 52 revised. Therefore, the proposed model adopts the view that Aad is better measured by decomposing it into two components. A further improvement of the predictive power of the model using the constructs discussed so far may be achieved by incorporating another recent development in the measurement of Ab. Batra and Ray (1985) reviewed research'on attitude toward the brand and maintained that attitude toward the brand is composed of two components: cognitive and affective. The first component, which they called the ”utilitarian affect (Ab-u)," is formed as a result of processing the attribute arguments presented in the ad. The second component, the "hedonic affect (Ab-h)," is a general approach-avoidance feeling toward the brand as a whole. This hedonic affect, they posited, is formed by the classical conditioning of affect from executional cues in the ad and/or repeated exposure. Their percentage contribution model posits that the relative contributions of the two components of Ab to purchase intentions will be different depending on the viewers' involvement levels with the ad. ‘When viewers are highly involved, the attribute-based utilitarian Ab will be the major contributor to the purchase intentions. On the other hand, in low involvement situations, the ad execution- based hedonic Ab will be the major contributor of the purchase intentions. 53 Hypothesized Relationships As is shown in Figure 4, the model posits that cognitions precede attitudes and attitude in turn precedes conation (purchase intention in this case). From the ELM (Petty & Cacioppo, 1981, 1986), Homer (1990) and Yoon (1991) all the cognitions —-> attitude links are predicted to be stronger for the high involvement condition than for the low involvement condition. Hla: The Cb -> Ab link will be stronger for the high involvement condition than the low involvement condition. Hlb: The Non-claim.cognitions -> Aad-no link will be stronger for the high involvement condition than the low involvement condition. file: The claim cognition -> Aad-c will be stronger for the high involvement condition than the low involvement condition. It is expected that when elaboration is at a low level, persuasion occurs mostly due to the executional elements in the ad serving as peripheral cues, while, in the case of high message processing involvement, persuasion depends on viewers' responses to the claims made in the ad. Also, from research on the effects of verbal and visual components of advertisements on brand attitude (Mitchell, 1986; Miniard et a1. 1990) the following set of hypotheses were derived. H2a: In the high involvement condition, Aad-c will be more influenced by the claim strength than attractiveness of the picture. 54 H2b: In the high involvement condition, Ab-u will be more influenced by the claim strength than attractiveness of the picture. H2c: In the low involvement condition, Aad-nc will be more influenced by picture attractiveness than message strength. H2d: In the low involvement condition, Ab-h will be more influenced by picture attractiveness than message strength. Finally, the path configurations are expected as follows. Since Aad-c is conceptualized as representing the central route, the Aad-c to Ab-u link is expected to be the dominant path through which persuasion (purchase intention) occurs when viewers' processing involvement level is high. Likewise, when the elaboration level is low the Aad-no to Ab- h link is expected to be the dominant path to PI. Based on this reasoning the follwing hypotheses are proposed. H3a: In high involvement condition, the Aad-c -> Ab-u -> PI link becomes more dominant than the Aad- nc -> Ab-h -> PI link. H3b: In low involvement condition, the Aad-nc —> Ab-h -> PI link becomes more dominant than the Aad- c -> Ab-u -> PI link. H3c: The Aad-nc -> Cb link will be stronger for the low involvement condition than the high involvement condition. H3d: The Aad-c -> Cb link will be stronger for the high involvement condition than the low involvement condition. H39: The Cb -> Ab (Ab-u and Ab-h) link will be stronger for the high involvement group than the low involvement group. CHAPTER I I I METHOD Measurement Issues Aad has been operationalized in many different ways in the articles reviewed in Chapter II as illustrated in Table 1. Despite the conceptualization of Aad as a multidimensional construct (Shimp, 1981; Lutz, 1985) no systematic effort has been made to develop a measure that can capture both the cognitive and affective dimensions of Aad. As a result, researchers have used various methods to assess Aad such as a single-item measure or the mean score of a limited number of items as shown in Table 1. Most researchers in recent studies (e.g., Gresham & Shimp, 1985; Mackenzie, Lutz, & Belch, 1986; Machleit & Wilson, 1988; Muehling & Laczniak, 1988) appear to have focused on the affective dimension in measuring Aad by employing a limited number of semantic differential (SD) scales. Thus, it appears that the majority of researchers working on this topic accept the concept of Aad as a general qualitative aspect of an advertisement, or the viewer's overall affective reaCtion to the advertisement. 55 56 Table l: Operationalization of Aad in Literature Author(s) Mitchell & Olson (1981) Moore & Hutchinson (1982) Lutz, MacKenzie, & Belch (1983), MLB (1986) Gresham & Shimp (1985) Gardner (1985) Batra (1985), Batra & Ray (1985) Burke & Edell (1986) Measurement Scale 5-point SD scale good-bad, like-dislike, irritating- not irritating, interesting- uninteresting S-point Likert scale measuring subjects' emotional reactions to the ad assessed by: positive, somewhat positive, neutral, somewhat negative, negative 7-point SD scale measuring "overall reaction to the commercial" favorable-unfavorable interesting-uninteresting 5-point scale (5=extremely well, 1=not well at all) measuring how well the following adjectives describe the ad: . soothing, warmhearted, sorry, sad, affectionate, happy, elated 7-point (?) SD good-bad, like-dislike, irritating- not irritating, interesting- uninteresting Aad was measured by open-ended "cognitive" and "affective" responses which subsequently were coded into SEVA (surgency, elevation, vigor, activation) feelings, deactivation feelings, and distraction/ irrelevant thoughts 5-point Likert scale (1=not at all well, 5=extremely well) asking how well the thirty four adjectives describe the ad Also subjects' overall Aad was measured by a 7-point scale (+3=very favorable, -3=very unfavorable) 57 Table 1 (cont'd) Park a Young (1986) Mitchell (1986) Machleit & Wilson (1988) Muehling & Laczniak (1988) Burton & Lichtenstein (1988) Homer (1990) 7-point scale (1=very favorable, 7=very unfavorable) 5-point SD scale good-bad, like-dislike, irritating- not irritating, interesting- uninteresting 7-point SD good-bad, enjoyable-unenjoyable, not fond of-fond of, dislike very much- like very much, irritating-not irritating, well made-poorly made, insulting-not insulting 7-point SD Not attractive-attractive, bad-good, unpleasant-pleasant, unappealing- appealing, dull-dynamic, depressing- refreshing, not enjoyable-enjoyable 7-point SD scale Affective Aad soothing-not soothing, warmhearted- cold hearted, uplifting-depressing, pleasant-unpleasant, attractive- unattractive, affectionate-not affectionate Cognitive Aad informative—uninformative, persuasive-nonpersuasive, believable-unbelievable, convincing—not convincing 9-point SD scale positive-negative, favorable- unfavorable, interesting- uninteresting Miniard, Bhatla, & Rose (1990) 58 Table l (cont'd) 7-point SD scale Aad-c persuasive-unpersuasive, informative-uninformative, strong- weak, believable-unbelievable Aad-nc . positive-negative, good-bad, favorable-unfavorable Overall Aad good-bad, effective-ineffective, interesting-uninteresting, like very much-dislike very much, not at all irritating-very irritating 59 Given the similar scales employed by most researchers to measure Aad and Ab, one of the most important questions is whether Aad is a separate construct from Ab. Shimp (1981) pointed out that some researchers argue Aad simply represents another brand belief. Several studies reported convergent and discriminant validities of the Aad construct. Madden, Dillon, and Twible (1986) suggested that Aad and Ab are conceptually distinguishable but they may not be empirically distinguishable under certain circumstances. Specifically, they found that when the ad does not contain nonbrand- specific peripheral cues Aad may not be distinguishable from Ab. Mitchell (1986), however, found that viewers' Ab was affected by their Aad even when the ad contained only copy and he contended Aad is a separate construct from Ab. Muehling and Laczniak (1988) also investigated the construct validity of Aad through confirmatory factor analysis. Their results clearly showed that Aad and Ab were indeed tapping two different constructs. In sum, empirical research suggests that Aad can be distinguishable from Ab despite the similar scales. One unresolved empirical question in measuring the construct of Aad is whether all the usually employed Aad scales represent the affective and emotional dimension of Aad as conceptualized by Shimp (1981) and Lutz (1985). wright (1986) argued that the usual semantic differential scales may be tapping a cognitive dimension rather than the affective dimension. He contended that it is not clear whether some of 60 the most frequently utilized scales, for example "good-bad," are representing cognitive or affective reaction. Some viewers might rate the ad as ”good" because they thought the brand information in the ad was helpful. And in this case the good-bad rating would represent a cognitive response rather than an affective response to particular executional cues in the ad. Still some others might say the ad was ”good" because of some of the peripheral executional cues in the ad that made them feel happy. In this case this good-bad rating would represent the viewers' affective response. A related issue is how Aad can be represented best empirically. Burton and Lichtenstein (1988, p. 4) pointed out the importance of the multidimensionality issue by stating: "If Aad is best represented as a construct comprised of more than a single dimension, single-item global evaluation measures may underrepresent the Aad construct, and any measures that treat the construct as unidimensional may limit the scope of the diagnostic information of importance to advertisers." A few studies addressed the issue of multidimensionality and investigated whether Aad can be better represented by dividing it into cognitive and affective dimensions. Machleit and Wilson (1988) did not find evidence to support discriminant validity between the affective and cognitive dimensions. Burke and Edell (1986) reported that the thirty four adjectives used to measure Aad were better captured by three factors. Burton and Lichtenstein (1988) 61 performed confirmatory factor analysis and the results indicated that the two factor model (cognitive and affective) is a better representation of Aad than a single factor structure. More recently, Miniard et a1. (1990) contended that Aad should be decomposed into claim and nonclaim elements and the two components should be measured separately. The scales employed by them is illustrated in Table 1. This method appears to be able to resolve the question raised by Wright (1986) at least partially by measuring the claim.element (supposedly the cognitive route) and the peripheral element (supposedly the affective route). By employing this method we may not worry about the issue of multidimensionality of Aad. Experimental Design Selection of Product The selection of the product was based on several criteria. First, it had to be a product rather than a service. Second, it had to be a product class that the students who were serving as subjects would consider for purchase. Third, since the study operationalized involvement as message processing involvement, this study needed a product which would fall somewhere in the middle of a scale measuring "product involvement" to eliminate the confounding effects of product involvement. The author and two assistants identified several product categories as 62 candidates through discussion. They were hair spray, gum, tires, sunscreen lotion, and mufflers. A pretest using subjects similar to those used in the actual experiment assessed product class involvement. Product involvement was measured by asking a five-point Likert type question: “Now, think back to the last time you bought any brand of xx product (or, if that does not apply to you, imagine you were doing so). How involved were you in that purchase? Ask yourself how important is it to you that you buy exactly the brand you do. Making the right decision when I choose a particular brand of xx is, to me, something that is...(1) most important, (2) quite important, (3) somewhat important, (4) Not very important, and (5) least important." The results are presented in Table 2 below. Table 2: Product Involvement Level (N=38) Product Mean Std Dev Gum 3.78 1.01 Hair Spray 2.75 1.21 Tire 1.94 .87 Muffler 2.20 1.00 Sunscreen 3.09 .98 1=most important 5=least important 63 Sunscreen was selected for several reasons. First, the mean of responses to the product involvement question was 3.09. As you can see in Table 2, the mean of 3.09 was the closest to the scale mid-point indicating that sunscreen has the involvement neutrality appropriate for the study. Second, students represent a sizable market for this product class. Thirty-four subjects (89.5 percent) answered ”yes" to the question, "Have you purchased a sunscreen product before?" To the question ”To what extent do you think you are knowledgeable about sunscreen products?" the mean was 3.61 on a seven-point scale where one was "very much" and seven was "not at all." It was found that subjects are not very loyal to a specific brand (mean=4.55 on a seven-point scale where one represented "highly loyal to a brand"). Judging from these findings, sunscreen lotion was deemed an appropriate product for the study. Development of Stimulus Materials First, the author and the two research assistants had to come up with the names of the two brands to be portrayed in the experimental and the control ads. we decided to use ”Rayblock" and “Sun Shade" for the experimental ad and the control ad, respectively. Four test ads were developed by using two different versions of copy and two different pictures (2x2). They were produced by a student who was working for DDB/Needham in Detroit with the help of the agency's creative department. 64 As a result, the finished ads were of professional quality. The content of the control ad was constant across all conditions and contained the same copy used for the strong version of Rayblock ads. Following Miniard et a1.'s (1990) procedure, the control ad lacked any color or photography although it contained a line drawing of the bottle on which the brand name appeared. This ad was clearly inferior to the Rayblock ads in terms of overall appearance. Pretest for Manipulation of Claim Strength The manipulation for the claims involved altering the strengths of the copy presented in the test advertisements. First, the attributes to be included in the ads were determined through discussion with two undergraduate research assistants and later with the author's dissertation committee chair. These attributes were: the sun protection factor (SPF), nongreasiness, waterproofness, and hypoallergenic nature of the product. Manipulation of message strength involved altering the numbers used to describe the attributes in the two different versions of advertising copy. A series of pretests were conducted to develop the final ad copy. Presented below is the questionnaire used to verify whether the manipulation in the first draft was successful. Thank you for joining us today. we are interested in finding consumers' responses to advertising claims. 65 Please read the following advertising copy excerpted from a print ad for a suntan lotion brand and answer the questions. The Skin Cancer Feundation warned that more than 93 ,percent of all skin cancers are related to the sun. It advises that regular use of an effective sunscreen reduces the risk of developing a skin cancer by as much as 88 percent. Rayblock offers a wide range of protection to suit your needs. Rayblock with SPF 15, for example, blocks out 97 percent of the sun's burning UVB rays and deeper (penetrating UVA rays. Rayblock is a non-greasy, waterproof sunscreen that protects you far 7 hours unlike ordinary sunscreens that last only 80 minutes. All Rayblock products are also PARA-free, with virtually no chance of causing allergies. we would like your opinions about the claims made in the advertisement you have just read, NOT the product being advertised. Please describe how you feel about the advertising claims on the given scales. 1 2 3 4 5 6 7 strong __ : : : : : : weak The above copy was for the strong claim condition. As shown below, the numbers used in the weak version were altered. The Skin Cancer Foundation warned that more than 50 percent of all skin cancers are related to the sun. It 66 advises that regular use of an effective sunscreen can reduce the risk of developing a skin cancer by as much as 38 percent. Rayblock offers a wide range of protection to suit your needs. Rayblock with SPF 15, for example, blocks out 67 percent of the sun's burning UVB rays and deeper ‘penetrating UVA rays. Rayblock is a non-greasy, waterproof sunscreen that lasts 3 hours unlike ordinary sunsreens that protect you fOr only 2 hours. .All Rayblock products are also PARA-free, with a low chance of causing allergies. Fifty seven undergraduate students from three classes offered by the Department of Advertising and three Ph.D students participated in the pretest of the copy manipulation. The sequence of the two versions of the test copy was alternated and they were distributed to the student subjects so that each version was shown to thirty subjects. After seeing the copy, subjects were asked to describe how they felt about the advertising claims on a seven-point semantic differential scale where one represented "strong” and seven represented ”weak." The means were 2.97 and 3.50 for the strong claim and weak versions, respectively. The results of the t-test (t=1.51, p=.137) indicated that there was no significant difference between the two means at the alpha level of .05. It was judged that the manipulation was not successful. The intended strong claim copy was not perceived to be as 67 strong as anticipated. The weak claim version was found to be more problematic. It fell to the strong side of the scale (mean=3.5) since the scale mid-point was 4. It was speculated post has that the use of numbers (statistics) and citation of the Skin Cancer Foundation gave both versions high source credibility and caused them to be perceived as strong claims. For example, although the claim "regular use of an effective sunscreen can reduce the risk of developing a skin cancer by as much as 38 percent" as opposed to "88 percent” in the strong claim version was intended to be a weak claim, the subjects appeared to have perceived it as a strong claim. We cannot, however, rule out the possibility that the insignificant t-test results may be an artifact of the conservative research method used for the study. Since only either the strong or the weak copy was presented to the subjects, they did not have a basis of comparing the version they saw. If both versions were shown to the subjects, the t-test results might have been significant. Even so, this conservative approach was employed since it is desirable in an experimental study to be able to manipulate the construct to the extremes as discussed in Chapter 2 (see, for example, Homer, 1990). The weak version was revised by eliminating all the numbers and the "Skin Cancer Foundation" to make it weaker. It is presented below: 68 Protect your skin from dangerous sun exposure. Regular use of an effective sunscreen may reduce the risk of developing a skin cancer. Rayblock offers a wide range of protection to suit your needs. Rayblock with SPF 15, for example, blocks some of the sun's burning UVB rays. Rayblock is non-greasy and.PABA-free, with a low chance of causing allergies. Because Rayblock is waterproof, it does not need to be re-applied after brief exposure to water. Thanks to Rayblock you are finally safe. The revised weak-claim version was subjected to another pretest using thirteen student subjects enrolled in another advertising course. The mean was 4.15 to the same question asking the claim strength of the ad on the same seven-point scale where one meant "strong" and seven meant "weak." Remember, the mean for the original weak version was 3.50. A t-test result (t=2.41, p=.024) indicated that there was a significant difference between the means of the original strong version and the revised weak version at .05 alpha level. Both copies were revised again and they are shown below. The revised versions were subjected to another pretest in which thirty eight students from two other advertising courses participated. 69 Revised Strong Version Protect your skin from dangerous sun exposure. The Skin Cancer Fbundation warns that more than 93 percent of all skin cancers are related to sun exposure. It advises that regular use of an effective sunscreen reduces the risk of developing a skin cancer by as much as 91 percent. Rayblock lotions offer a wide range of protection to meet your special skin care needs. Rayblock with SPF 15, fer example, blocks out 97 percent of the sun's burning UVB rays and deeper penetrating UVA rays. Rayblock is completely hypoallergenic with its PARA- free, non-greasy formula. Thanks to Rayblock’s new waterproof fermula, it does not need to be re-applied-- even after a whole day of swimming and playing in the water. With Rayblock, you are finally safe even in the strongest summer sun. Revised Weak Version Protect your skin from dangerous sun exposure. Regular use of an effective sunscreen may reduce the risk of developing a skin cancer. Rayblock lotions offer a wide range of protection to meet your needs. Rayblock with SPF 15, for example, blocks some of the sun's burning UVB rays. Rayblock is non-greasy and PARA-free, with a low chance of causing allergies. Rayblock is also water- resistant and does not need to be re-applied after brief exposure to water. Thanks to Rayblock you are finally safe. 70 The pretest was designed so that we can compare the two versions in an absolute sense first and then in comparison to either the strong one or the weak one. To accomplish this, two sets of questionnaires were needed. One set of the questionnaires showed the strong version first and the weak version next. The other set showed the weak version first and then the strong one next. Therefore, by comparing the strong version in the first set of questionnaires and the weak version in the second set of questionnaires we could examine whether the two versions are different or not in an absolute sense. In other words, since these are the versions shown first to the subjects, the subjects had no basis of comparing the copy. After seeing either the strong or the weak version they were asked, "we would like your opinions about the claims made in the advertisement you have just read in comparison with other ads for sunscreen products. Please indicate how you feel about the claims made in the ad on a seven-point scale where one represents "strong” and seven represents "weak." The means were 2.47 and 4.05 for the strong and weak versions, respectively. A t-test (t=3.55, p=.001) indicated that there is a significant difference between the two copies. After seeing the first ad copy (either strong or weak) and answering the question, subjects saw another version (either weak or strong this time). And they were asked, ”we would like your opinions about the claims made in the advertisement you have just read in comparison with the other 71 Rayblock ad you read in the previous page. Please indicate how you feel about the claims made in this version of Rayblock ad on a seven-point scale where one represents 'strong' and seven represents 'weak'." Because the subjects were exposed to a version in the previous page of the questionnaire, they had a basis of comparison. As anticipated, the strong version was perceived as stronger (mean=1.79) than when the subjects saw the strong version first (mean=2.47). Likewise, the weak version was perceived as weaker (mean=4.68) than when it was shown first (mean=4.05). There was, of course, a significant difference between the two versions (t=6.17, p<.001). Based on the series of pretests described above it was judged that the manipulation of copy strength was successful. The next section describes the pretest for selection of the pictures to be used in the advertisements. Pretest for Manipulation of Picture Attractiveness First, the research assistants and the author collected a variety of four-color pictures of varying sizes from magazines and promotion brochures of travel agencies. Seven attractive pictures and seven unattractive pictures were selected through discussion for a pretest. These pictures contained a wide variety of stimuli ranging from portraits of animals (snakes, mouse, insects), scenery (tropical beaches, waterfall, dessert) to a human skeleton half buried in a dessert. 72 Each of the fourteen pictures was shown for about fifteen seconds to a total of thirty eight subjects in two sessions of pretesting. The order of pictures was arranged so that attractive and unattractive pictures were alternated to eliminate any potential response biases that may arise from seeing either attractive or unattractive pictures in sequence. After seeing each picture, subjects responded to a question "My overall feelings toward picture # xx is ..." anchored by four seven-point semantic differential items. The items came from.Mitchell (1986) and they were: good-bad, like-dislike, pleasant-unpleasant, and attractive- unattractive. The picture that was evaluated as the most attractive (mean=1.51 on the combined scale where 1 represented “attractive") was a portrait of a tropical beach scene. The stimulus rated as the most unattractive was a picture of a human skeleton (mean=5.79 on the same scale) half buried in a desert. These two pictures were used to manipulate the attractiveness of picture in the test ads. Subjects and Procedure A total of two hundred undergraduate students enrolled in advertising and business courses participated in the study and were given course credit. Subjects signed up for the experimental sessions all of which were conducted in groups of eight. Each session was either the high involvement or the low involvement condition. 73 To eliminate potential selection biases, the study was designed so that subjects in a single session were randomly exposed to different experimental treatments, with the one constraint that each session used either the high or the low involvement treatment to be explained later. As a result of this design, in each experimental session two subjects were given the strong/attractive version, another two were given the strong/unattractive version, another two were given the weak/attractive version, and the last two were given the weak/unattractive version. After subjects were seated at designated tables, they were instructed to read only the first page (consent form) and the second page of the questionnaire. The second page contained the instructions that manipulated involvement, which is explained below. (After reading the instructions, subjects were exposed to one of the four test ads and the control ad. After reading the stimulus materials at their own pace, the subjects answered the items in the questionnaire. When all subjects finished filling out the questionnaire, the experimenter explained that the manufacturer had not yet delivered the products. The experimenter then apologized for the deception but explained that it was necessary to increase their involvement with the product choice. They were asked not to discuss the experiment with anyone else because if those yet’ to participate knew about the experiment, the internal 74 validity of the study would be threatened. The experimenter thanked subjects for their participation and dismissed them. Manipulation of Involvement The involvement manipulation consisted of varying the personal relevance (cf., Petty & Cacioppo, 1979; Park a Young, 1986; Miniard et al., 1990) of the Rayblock ad by giving subjects different instructions regarding the task involved in the study. For the high involvement conditions the experimenter instructed the subjects that the purpose of the study was to obtain their responses to two new brands of suntan lotion being considered to be introduced in the area. Their involvement level was increased by telling them: “According to Consumer Reporth there are substantial differences in quality among major brands of suntan lotion available on the market as measured by functional performance. You will be shown a print ad for each of two new brands of suntan lotion and then asked to make a choice between the two brands. The ads may contain information on those performance characteristics as well as other features.” Further, they were told that they were going to make a choice between the two brands of suntan lotion and might receive their choice of brand as a gift. This procedure has been used quite widely to increase involvement (cf. Petty & Cacioppo, 1979; Petty et al., 1983; Miniard et al., 1990). 75 In contrast, subjects in the low involvement conditions were informed that they would make a choice between two new brands of shampoo. The Rayblock ad was expected to be more personally relevant to subjects who anticipated to make a brand choice from the sunscreen product category than to those who anticipated to make a brand choice from the shampoo product category. As briefly explained above, subjects in the low involvement were led to believe that after seeing an ad they are going to make a choice between two brands of shampoo. For this procedure to be successful, some rationale was needed as to why they would see the sunscreen ads. They were instructed: “In this study we are attempting to obtain your reactions to two new brands of shampoo which are being considered for introduction in this area. You will look at ads for the new brands of shampoo and then make a choice between the two brands. We will show you a couple of warmp up suntan lotion ads before showing the critical shampoo ads in order to familiarize you with the type of print ads used in the study." Their task, however, was over after the "warm-up." To further lower their involvement level, the following instructions were given: “According to Consumer Re orts, there are n_q substantial differences in quality among major brands of suntan lotion available on the market as measured by functional performance. You will be shown a print ad for each of two new brands of 76 suntan lotion and then asked to make a choice between the two brands. The ads may contain information on those performance characteristics as well as other features." This procedure, adapted from.Miniard et al. (1990) study, was designed to lower the test ad's personal relevance or importance to subjects. To further decrease the importance of the ad, the instruction said that there were no substantial differences in quality among major brands of suntan lotion (Park and Young, 1986). An initial manipulation check on involvement was performed after collecting data from ninety subjects. Unfortunately, the manipulation of message processing involvement was found to be unsuccessful. As explained earlier, this study replicated the procedure reported in several studies including the Miniard et a1. study (1990). A four-item semantic differential scale, adapted from.Muehling and Laczniak (1988), was used to verify the manipulation. Analysis of the four-item index (Cronbach's alpha=.74, n=90) of self-reported involvement levels with the advertising messages (assessed by the extent to which the subjects had “paid attention to the ad's claim about the performance," “concentrated on message," “put thought into evaluating messages," and “felt the information relevant for my needs") showed that the manipulation was not successful. None of the t-scores was significant at .05 level (n=90). 77 Since involvement is a crucial part of the study, it was clearly necessary to alter the procedure which was used to manipulate involvement. The author and his dissertation committee chair decided to employ the procedure used in the Park and Young study (1986). This procedure asks subjects to cooperate and assume different levels of involvement. Yoon (1991) reported that this method seemed to work well in an experiment using a television commercial as the stimulus. Presented below are the new instructions for the high and low involvement manipulations. New High Involvement Instructions Thank you for joining us today. In this study we are attempting to obtain your reactions to two new brands of suntan lotion which are being considered for introduction in this area. Please read the following instructions very carefully. According to Consumer Re rts, there are substantial differences in quality among major brands of suntan lotion available on the market as measured by functional performance. You will be shown a print ad for each of two new brands of suntan lotion and then asked to make a choice between the two brands. The ads may contain information on those performance characteristics as well as other features. 78 Please assume that you are about to take a trip to the Bahamas and you need to purchase a suntan lotion, and read the ads as though you were trying to learn about the brands' benefits and effectiveness. In other words, please focus on the product information provided in the ads. At the end of the session you may receive the brand of your choice as an additional compensation for your participation. Now read the ads right next to you. New Low Involvement Instructions Thank you for joining us today. In this study we are attempting to obtain your reactions to two new brands of suntan lotion which are being considered for introduction in this area. Please read the following instructions very carefully. According to Consumer Reports, there are pg substantial differences in quality among major brands of suntan lotion available on the market as measured by functional performance. You will be shown a print ad for each of two new brands of suntan lotion and then asked to make a choice between the two brands. The ads may contain information on those performance characteristics as well as other features. Assume that you are preoccupied by a very difficult class project due tomorrow. Furthermore, now that summer is over you do not plan to buy a suntan lotion in 79 the near future. You are asked to look at the ads as if you were sitting in your living room, worrying about the project, and knowing that you have no need to purchase a suntan lotion. Now look at the ads right next to you. Due to the importance of the involvement manipulation the author decided to collect data again using the new instructions. A total of 104 students participated in the study using the new instructions. The sample size for each cell was thirteen. Measures Cognitive Responses Subjects' cognitions were measured by the open-ended thought listing method (Wright, 1973; MacKenzie et al., 1986; Homer, 1990; Miniard et al., 1990). Immediately following the exposure to the test ad, subjects were instructed to list any thoughts, feelings, ideas, and images that occurred to them while viewing the ad. Later the thoughts were coded first into thoughts directed toward the ad and thoughts directed toward the brand. Thoughts directed to the ad were further categorized into non-claim cognitions and claim cognitions. The three groups of thoughts and feelings (brand cognitions, claim cognitions, and non-claim cognitions) finally were indexed into three measures by subtracting negative thoughts from the positive ones. 80 %(Attitude Toward the Ad For the Aad-c and Aad-no measurement this study replicated the methods used in the Miniard et al's (1990) study. The following instructions and scales were used (p. 295): In obtaining your reaction to the ad, we would like for you to distinguish between two basic components of the ad. The first component involves the claims made about the product. The second component involves the remaining elements within the ad such as the format, pictures or illustrations, colors, M type style, spacing, and so forth. Concerning the first component, how would you evaluate the claims made about Rayblock? Then subjects' Aad-c were measured using the same semantic differential scale items (persuasive/unpersuasive, informative/uninformative, strong/weak, and believable/ unbelievable) used by Miniard et a1. (1990). Subjects then were asked, "Concerning the second component, how would you evaluate the remaining elements (everything except the claims) within the Rayblock ad? Also following Miniard et al. ( 1990), this Aad-no were assessed by a three—item scale (positive/negative, good/bad, and favorable/unfavorable). Attitude Toward the Brand Ab was measured by a similar semantic differential scale developed by Batra and Ahtola (1991). Ab:u were assessed by ..l \i semantic differential items of,u§Eful/useless; \. I J 81 valuable/worthless, and beneficial/harmful. Ab-h were obtained by pleasant/unpleasant, nice/awful, and agreeable/disagreeable. These scales had been earlier subjected to a confirmatory factor analysis using pretest data (Yoon, 1991b2 ) and found to be highly stable. Brand Beliefs The next set of measures was a Fishbein-type brand belief scale adopted from.Muehling and Laczniak (1988) study. It assessed subjects' beliefs about Rayblock along the same four attributes used in the ads. These beliefs (e.g., Rayblock protects you from the harmful rays of the sun.) were assessed by a seven-point semantic differential scale (extremely likely/extremely unlikely). Purchase Intentions This construct represents the likelihood that the subjects will purchase the brand in the future. Purchase intentions, "the likelihood that you will purchase Rayblock when it becomes available in the area," were assessed via two item semantic differential scale (probable/improbable, likely/unlikely). This Operationalization has been used by MacKenzie et al. (1986), Homer (1990), Miniard et a1. (1990), and Yoon (1991). 2 Yoon and Boster performed an analysis on two components of attitude toward the brand using the confirmatory factor analysis routine in the PACKAGE (Hunter, Gerbing, Coehn, & Nicol, 1980). 82 Manipulation Checks Subjects' involvement with the ad was assessed by a four item scale used by Muehling and Laczniak (1988). Subjects were asked to indicate the degree to which they "paid attention to," "concentrated," and ”put thought into evaluating" the messages in the ad. Also the degree to which subjects thought the ad relevant for their needs was asked. Subjects' evaluation of the picture was assessed by a four- item (good/bad, like/dislike, pleasant/unpleasant, attractive/unattractive) seven-point semantic differential scale used by Miniard et a1. (1990). Finally, subjects' evaluation of the message strength was measured by a seven- point semantic differential scale (strong/weak). Analysis of Cognitive Responses The analysis of the open-ended cognitive response data was completed in three phases. In the first phase, two judges who were blind to the experimental conditions and hypotheses decomposed each protocol into separate thoughts through discussion. The judges reported that the decomposition was an easy task due to a technique used in the study to measure cognitive responses. Following a technique described in the Park and Young (1986) study, when the subjects finished the questionnaire they were asked to go back to the cognitive response page and to rate each thought they listed on a 7-point scale (1=favorable, 7=unfavorable). 83 In the second phase, the same two judges independently coded each protocol initially into thoughts directed to the brand (brand cognitions), thoughts directed to the ad (ad cognitions), and all others including random, irrelevant thoughts (refer to Appendix B for a detailed explanation of the coding of cognitive responses). Brand cognitions next were assigned either to counter argument or support argument (of. Wright, 1975). Ad cognitions were then categorized into either ad claim cognitions, nonclaim cognitions, or overall cognitions. These ad cognitions were finally assigned valence (positive or negative). Again, the judges expressed that subjects' self-evaluation of favorability on their thoughts was highly helpful in making the judgments concerning the valence of the ad cognitions. For an illustration, a cognitive response item "the ad is hilarious" would be unclear as to whether it was meant to be a positive or a negative statement about the ad. With the subjects' self-evaluation of favorability on each thought, the judges reported it was much easier to assign the thought to the proper category. Finally, random and irrelevant thoughts were categorized as “all others." As shown in Table 3, the total number of cognitions was 440. Out of the total of 440 cognitive response items the judges initially disagreed on 43 items showing an agreement rate of 90.23 percent. They could resolve all the discrepancies through discussion. The high agreement rate between the two coders without any involvement by the author 84 suggests that the coding scheme was reliable and the coder training was proper. Table 3 shows the cognitive response categories used in the study and the frequencies of each category. About 24 percent of all the responses were brand cognitions as compared to over 70 percent of ad cognitions including both the claim cognitions and non-claim.cognitions. Only 4.8 percent fell in the category of random, irrelevant thoughts. This is a striking comparison with the 26.6 percent of irrelevant thoughts reported in Yoon's (1991) study. This vast discrepancy between the two studies in the percentage of irrelevant thoughts is speculated to have been caused by the different media used in the studies. Krugman (1965) suggested that television is generally a low involvement medium. Since a print medium (a print ad in this case) is normally more engaging than a broadcast medium (a television commercial in this case) the subjects who were exposed to the television commercial produced a higher number of random and irrelevant thoughts than those exposed to the print ad. In the third phase of the analysis of the cognitive response data, cognitive response indices--brand cognitions, claim cognitions, and non-claim cognitions--were constructed from the cognitive response categories shown in Table 3. Following the procedure used by MacKenzie et al. (1986) and Homer (1990), the total number of negative statements in each category was subtracted from the total number of positive statements in the corresponding category. Each index, 85 therefore, represented the net valence of the cognitions in each of the categories (MacKenzie et al., 1986). This procedure of index construction is consistent with the Fishbein's (1975) multiattribute attitude model which assumes that individuals form attitudes toward an object in a compensatory manner by subtracting the negative attributes from the positive attributes associated with the attitude object. Following the procedure explained above, brand cognitions were combined into a single index by subtracting the counter arguments from the support arguments (SAPCA). Indices for claim cognitions and non-claim cognitions were similarly constructed by subtracting the negative statements from the positive statements in each category. Note in Table 3 that thirty two responses (7.72%) reflected either positive (24) or negative (8) overall evaluations of the ad. Because of the theoretical concerns that such responses are indicators of overall affect (as opposed to the cognitions leading to affect) they were not included in subsequent analysis (see Miniard et al., 1990; Olson, Toy, & Dover, 1982; Wright, 1980). 86 Table 3: Cognitive Response Categories and Frequencies Category Frequency % Brand Cognitions Counter argument 60 . 13.64% Support argument 44 10 Ad Claim Cognitions Positive statements 11 2.5 Negative statements 35 7.95 Non—claim Cognitions Positive statements 146 33.18 negative statements 90 20.45 Overall positive statements about the ad 24 5.45 Overall negative statements about the ad 8 1.82 Random, irrelevant thoughts 21 4.77 TOTAL 440 100 CKAPTER IV RESULTS Manipulation Check As anticipated in the pretests explained in Chapter III, it was found that the manipulation of claim strength was successful. The claim strength manipulation was verified by a seven-point semantic differential scale where “1" represented "weak" and “7" represented ”strong." The mean for the subjects in the strong claim condition was 5.02 (n=52) as compared to 3.58 (n=52) for the subjects in the weak claim condition. As shown in Table 4, verification of the picture attractiveness manipulation was achieved via a four-item (good/bad, like/dislike, pleasant/unpleasant, and attractive/unattractive) index of attitude toward the picture. Analysis of the items showed that the manipulation was successful. Scores on each item revealed significant differences (t scores ranging from 5.17 to 16.1) in the anticipated direction at .001 level. The combined overall index (Cronbach's alpha=.94, n=104) also showed a significant difference (t=9.98, p<.001), with the subjects in the attractive picture condition reporting substantially higher score than those in the unattractive condition. 87 88 Table 4: Manipulation Check for Picture Attractiveness Items“ Means for Conditions t-score, prob. Attractive Unattractive good=7, bad=1 5.85 3.85 5.38, <.001 like=7, dislikezl 5.63 3.63 5.17, <.001 pleasantz7, unpleasant=l 6.19 2.21 16.1, <.001 attractive=7, unattractivezl 6.00 2.37 12.9, <.001 Auitude to picture (index) 5.92 3.02 9.98, <.001 * Item scores were reversed for analysis so that good, for example, was recoded as 7 from 1 and bad was recoded as 1 from 7, etc. Table 5 presents the t-test results on the involvement manipulation check items for both the original and revised procedures, to make the comparison easy. Each t—score and its probability level show whether a significant difference was found between the high and the low involvement conditions when measured by the old and new procedures. As discussed in Chapter III, it shows that the original procedure did not produce significant differences between the high and low involvement conditions at the conventional .05 probability level on any of the four semantic differential items. (I) 89 Table 5: Comparison of the Old and New Procedures used for Manipulation of Involvement Items used for manipulation Original Revised check of Involvement Procedure Procedure Paid attention to the ad's t=1.12, t=3.12, claim p=.13 pp<.01 Concentrated on message t=.36 t=3.13, p=.36 p<.01 Put thought into evaluating t=.40, t=3.69, messages p=.35 .01 Felt the information relevant t=.86, t=1.79, for my needs (p=.20 p<.05 The results presented in Table 5 show that the new procedure for involvement manipulation was successful. significant difference was found in the anticipated direction between the high and low involvement conditions when measured by all the four items at .01 level except the last item which was significant at .05 level. the manipulation of involvement with the new instructions was adequate. Despite the statistically significant difference between the high and low involvement conditions, however, we are not Therefore, it is judged that sure if we successfully created a true low involvement condition. check items by involvement condition. subjects in the high involvement condition manifested higher Table 6 presents the means for the manipulation It shows that the 90 levels of involvement than those in the low involvement condition. Note, however, that the means for all four of the items for the low involvement condition were greater than the scale mid-point of 4 except for the fourth item. Miniard et al. (1991) also reported similar findings. It is inferred from these findings that creating a truly low involvement condition is highly difficult in an experimental situation, especially when the experimental stimulus is print material. Table 6: Involvement Manipulation Check Items used for manipulation High Low check of Involvement* (n=52) (n=52) Paid attention to the ad's 5.25 4.26 claim Concentrated on message 5.29 4.27 Put thought into evaluating 5.33 4.15 messages Felt the information relevant 4.31 3.62 for my needs Involvement Index** 5'04 4'08 *1=not at all, 7=very much . ** Created by adding the four items and dividing the score by four. t For significance test, see Table 5. 91 Additional Evidence of the Successful Involvement Manipulation Further evidence of the success of involvement manipulation was found in an analysis of the cognitive responses. Table 8 (page 97) contains the cell means for the open-ended cognitive measures used to evaluate the success of involvement manipulation. Subjects with high ad processing involvement are expected to utilize a brand evaluation strategy (Gardner et al., 1985). Since they will attempt to evaluate the quality of the advertised brand, they are expected to produce a higher number of brand-related thoughts than subjects with low inv01vement. Based on this reasoning, a 2x2x2 ANOVA was performed to inviestigate the effects of the three factors on the production of cognitive thoughts. Results showed that only the level of involvement had a significant effect (F=4.18, p<.05) on the number of brand related thoughts (counter arguments + support arguments). Consistent with previous research findings (Laczniak, Muehling, & Carlson, 1991; Miniard et al., 1991; Park & Young, 1986), subjects in the high involvement condition produced more brand related thoughts (M=1.21) than those in the low involvement condition (M=.79). 92 The generation of message-related thoughts is considered to be a highly involving and effortful mental activity (Wright, 1973; Laczniak et al., 1991). Therefore, it is expected that subjects with a high level of involvement would produce more message-related thoughts than subjects with a low level of involvement. It was found that subjects in the high involvement condition (M=.62, n=52) produced significantly more claim-related thoughts (F=7.07, p<.01) than those in the low involvement (M=.27) condition. This finding is consistent with the ELM which posits that the major route to persuasion will be the central processing of the claims (messages) when subjects are highly involved (Petty & Cacioppo, 1981). On the other hand, ad related thoughts (positive and negative statements about the non—claim elements in the ad) showed no significant difference (F=.02, P>.1) across the high (M=2.29) and low (M=2.25) involvement conditions. It is inferred that all subjects (regardless of the involvement condition) paid a heightened amount of attention to the picture in the ads because the subjects were highly aware of the ad processing situation. This result, consistent with the previous findings (MacKenzie et al., 1986; Miniard et al., 1990; Park a Young, 1986; Yoon, 1991), suggests that the traditional view (ELM) may not be accurate, at least not in advertising pretesting settings. The ELM posits that 93 peripheral cues such as picture will be the main route to persuasion when subjects are not highly involved. Miniard et al. (1990) suggested that level of involvement alters the amount of elaboration directed toward the product, but not the ad. Hypothesis Testing Hla, Hlb, and ch predicted that the path from cognitions to attitudes would be stronger for the high involvement condition than the low involvement condition. Three conditions should be satisfied to establish a causal relationship between two variables (Babbie, 1986). First, the cause should precede the effect in time. Second, there should be an empirical relationship between the two variables. Third, the observed empirical relationship should not be explained away by a third variable which influences both of the two variables. Since subjects were exposed to brand new test ads for hypothetical brands, it is reasonable to assume that the cognitive responses preceded the corresponding attitudes. In addition, because one cannot suspect any intervening third variable between cognition and attitude (claim cognition and Aad—c in this case, for example), it appears reasonable to assume that a significant correlation will satisfy the causality requirement in this case. Therefore, a correlation analysis was performed to test these hypotheses. 94 The correlation coefficients between the constructs used in the study are presented in Table 7 along with the associated probability levels shown in parentheses. For each correlation coefficient grouping, the top coefficient is for the pooled sample, the middle one is for the high involvement sample, and the coefficient at the bottom is for the low involvement sample. For an illustration, the correlation coefficient between brand cognitions (Cb) and hedonic attitude toward the brand (Ab-h) was .23 for the pooled data, .34 for the high involvement sample, and .05 for the low involvement sample. The corresponding probability levels were .02, .01 and .70, respectively. From Hla, it was predicted that the correlation between brand cognitions (Cb) and attitude toward the brand (operationalized as Ab-u and Ab-h in the study) would be stronger for the high involvement group than the low involvement group. As shown in Table 7, the correlation between Cb and Ab—u was stronger for the high involvement group (r=.42, p=.002) than the low involvement group (r=.24, p=.09). Similarly, the correlation between Cb and Ab-h was also stronger for the high involvement group (r=.34, p=.01) than the low involvement group (r=.05, p=.70). These findings support Hla. From Hlb, one would expect that the correlation between non-claim cognitions and Aad-no would be stronger for the high involvement condition than the low involvement condition. Table 7 shows, however, that the correlation 95 m=o=Ew8 8.8.0.5: 855 .mcoEcmoo EBB Homo ...EouEwoo 235 ”no efsoa 8.3a H... Eon—029,5 BER: 3580295 swan: GEES 3.08 5.. EooEooo cone—9:83 ”ocowoq :33. A83 5.. 63% A838. A838. 932. 23%. 2039.. 638. 23mm 83$. 83%. €38. 63 G. E 83 3. 63%. 80.5... 83%. 83 :V. 838. 839.. 8833. A832 :38. 83%. a3 :. 838. 83%. 83%. 83 a. $33. 832. :38. is. 80.5.... 839.. A833. A83 3. sin. Q38. :32. 28.39.. 83%. $38. 338. 838. 83%. $33. 632. A838. .12 63 : 83 8.. 33 mo. E; 8. 63 em. 83 as 83 3.. :3 3. 803 cm. e38. 833. $.un. 83:. 2.62 83 mm. 63 2.. 83 mm. 33 we :3 mm. 33 8. :83 a. 838. A833. 838 98... A83 cm. 83 cm. 63 cm 33 no. :3 E. 832. 63:. 056 a: 3. £3 E. 23:832. £33 mm. 35 E38. .12 .12 8-3. i! 93. .i 5392 530.1 so opossumsoo moose m=0flumaouuoo up manna 96 coefficients for both conditions were equally significant. Two post hoc speculations seem plausible for these findings. First, the high correlation coefficient for the low involvement condition can be attributed to the fact that the study could no; create a trgg low involvement condition as discussed earlier. In other words, because of the relatively high level of involvement for the low involvement group (remember that the mean for the composite manipulation check scale was greater than the scale mean of 4), subjects in the low involvement condition processed the ad cognitively just as those in the high involvement condition did. Second, as the ELM (Petty & Cacioppo, 1981) posited, subjects in the low involvement condition might have paid a high level of attention to the non-claim elements in the ad, thus naturally producing the observed strong correlation between non-claim cognitions and Aad-nc. If this is the case, the non-claim cognitions may not represent “cognitions" in its traditional sense, at least when subjects' involvement level is low. From the findings and discussions, Hlb was not supported. From ch, it was predicted that the correlation between claim cognitions and Aad-c would be stronger for the high involvement condition than the low involvement condition. Table 7 shows that the correlation coefficient for the high involvement condition (r=.40, p=.003) was stronger than for the low condition (r=.09, p=.52), thus supporting ch. 97 .osmuo 0:» chose» mospfluum oacoomn «slam venue on» cause» moouwuum omflumuflaapsuouna on one ca musmsmao Hwauo Han ounsou mosufluum nuanced on one ow moms mawmao on» ouosop convenes upload «oedema .Ho>oa woman no. no quOfluwsoflm uoztt .oowuomuca ommnousm :o Aao.vo. mucouum ucoowuflcmflm on: oowumasmwcna sflnaog no omumuouovcHt .m:0fluonmu o>wuflmom on» concommuoou ahnaam h muo:3_n on H scum ommcmm as ousuowm m>auoouvuoesumusuoaou .muouoao o>fluomuupoumusvowo+ : .omouufinu on; dado some HON onwm madame one a: we 2. Sum wmd cod _m.N mod wo.~ end ovd Sauce on :38. a: 2. 8V n fl. Tm. on. n _ . nw. No. mm. an. 3.355 850 go... a: m: no.v no. nw. mm. 8; 84 R: we; 84 35:05 .055 30,—. m: a: 50V «ed and wo.m Ed find an wmn n _ .n :52: 308029»:— m: 59v 8. nos. Q .n _m.n onn mmé end mmn non 35w 58V 5V 2. be...” 2 .n ova. finn ohm and body mun 32A? 2. noo.v m: an bod onn wmn _n.¢ mad onn Non sands. 80V m: 8. n ~ .v who and one a: "an we.»‘ won unocéa< a: 50V a: find no...» n _ .n men n m.m on .m nnn onn .8063. a: 20V tn: won nmd nod n mé wn.~ no.“ 5...». nod 382.85 08:05 220E EEO .oZ: 830E. 058ml. 230E- 2806... 2805. 2803+ 2803. «2306+ «.085 BED 3003 EEO spam EEO M83 830 spam 0382 n— _o 8:8E=w_m 2.08029»:— 33 2.080202: :3: coaufiocoo Hovcguomxm ho moannwumev ucoocomoo no name: um manna 98 Manipulation's Effects on Dependent Variables Table 8 shows the means of dependent variables by experimental condition along with the significance of F's related with independent variables. Each dependent variable was submitted to a 2 (involvement) x 2 (claim strength) x 2 (picture attractiveness) ANOVA. For significance test, the conventional probability level of .05 was employed. Purchase; Intention A purchase intention scale (Cronbach's alpha=.98, n=104) was constructed by averaging the two semantic differential items (likely/unlikely, probable/ not probable) discussed in Chapter III. Only claim strength was found to have significant effects (F=9.34, p<.01) on purchase intention. Subjects in the strong claim condition reported more favorable purchase intention (M=3.89) than those in the weak claim condition (M=2.81). Attitude toward Claims in the Ad (Aad-c) 82a predicted that in the high involvement condition, Aad-c will be more influenced by the strength of claim strength than picture attractiveness. An Aad-c scale (Cronbach's alpha=.87, n=104) was created by computing the average of the four semantic differential items (persuasive/unpersuasive, informative/uninformative, strong/ weak, and believable/unbelievable) used to measure subjects' attitude toward the claims made in the test ad. This scale 99 was then submitted to a 2x2x2 ANOVA. Only claim strength had main effects (F=37.49, p<.001) on subjects' Aad-c, supporting 82a. Subjects in the strong claims condition reported higher Aad-c score (M=5.44) than subjects in the weak claims condition (M=3.99). There were interaction effects between involvement and claim strength (F=12.29, p<.01) as illustrated in Figure 5. 6 .. o S _ é -{}- Highmfl‘flmfl1t -¢ -----O-~-~- lavhvolvemmt 4 .— 3 u . u Weak Strong Claim Strength Figure 5: Interaction effects between Claim and Involvement on Aad-c It shows that generally subjects who were exposed to strong claims reported more favorable Aad-c than those who 100 were exposed to weak claims. This phenomenon, however, was less prominent when the involvement level was low. In fact, when involvement was low, subjects exposed to the weak claims reported higher Aad-c than those who were highly involved. From these findings, it is inferred that the effects of strong claims made in an ad are reduced when consumers are not highly involved with the ad. Attitude toward Non-claim Elements (Aad-nc) 32c predicted that in the low involvement condition, Aad-no will be more influenced by picture attractiveness than message strength. To test this hypothesis, first, an Aad-no scale (Cronbach's alpha=.92, n=104) was created by calculating the average of the three semantic differential items positive/negative, good/bad, and favorable/unfavorable) and was subjected to a 2x2x2 ANOVA. Results showed that involvement and picture attractiveness had significant main effects on subjects' Aad-no. A significant difference (F=6.97, p=.01) was found between the Aad-nc scores of the subjects in the high involvement condition (M=4.32) and those who were in the low involvement condition (M=5.13). Notice that subjects in the low involvement condition reported more favorable Aad-no. This may be an indirect evidence that subjects in the low involvement condition paid more attention to the peripheral cues in the ad, thus forming a more favorable attitude to those cues than did those in the high involvement condition. 101 {As hypothesized, subjects in the attractive picture condition (M=5.85) revealed a significantly higher (F=56.69, p<.0001) Aad-nc than those in the unattractive picture condition (M:3.63). Message strength had no effects on Aad- nc. These findings support 82a. There was a moderate amount of interaction effects (F=4.12, p<.05) between involvement and claim strength on Aad-nc. As illustrated in Figure 6, strong claims produced more favorable Aad-nc than weak claims when subjects' involvement was high. When involvement was low, however, claim strength had almost no effects on Aad-no. C) ....................................... 5 _ ........... O o s 'L + High Involvement 2 4 .. "“0""- Low Involvement 3 . I Weak Strong Claim Strength Figure 6: Interaction effects between Claim and Involvement on Aad-no 102 Dimensionality Assessment of Brand Attitude Before discussing the effects of manipulations on Ab, the dimensionality of Ab should be examined. This issue is important not only for theoretical reasons (Batra & Ahtola, 1991; Holbrook & Hirschman, 1982) but also for methodological reasons because a scale (brand attitude, in this case) constitutes a valid estimate of the construct only if the scale is unidimensional (Hunter & Gerbing, 1982). For these reasons, the measurement model of Ab will be discussed before proceeding. Hunter and Gerbing (1982) explained that for the items designed to measure each construct to be meaningful, the items should represent alternative measures of the same underlying trait (factor). In other words, the cluster of items measuring a construct should be unidimensional. They proposed three criteria to evaluate the unidimensionality of items: (1) homogeneity of item content; (2) internal consistency; and (3) parallelism, or external consistency. Homogeneity of items, often referred to as face validity (Kidder & Judd, 1986), concerns whether the items for each construct appear to measure the same thing. Some researchers (e.g., Hunter & Gerbing, 1982) feel that this is the most important criterion of assessing unidimensionality of items. The determination should be made on the basis of existing theory and logical judgment (Kidder & Judd, 1982). The six items used to measure brand attitude were judged as measuring two constructs based on the face validity and existing 103 literature (Batra & Ahtola, 1991; Holbrook & Hirschman, 1982) explained above. The next criterion for scale unidimensionality is a test of the internal consistency among items in the scale. This test investigates three assumptions. First, the items measure the same underlying trait. Second, the relationship between each item and the unobservable trait (factor) is linear. Third, the underlying trait is the only significant causal factor on the items. These assumptions can be tested by the Spearman product rule which states that the correlation between any two items in a scale should be equal to the correlation between the one item and the underlying factor multiplied by the correlation between the other item and the underlying factor (Hunter & Gerbing, 1982). The internal consistency test was performed by the confirmatory factor analysis sub-routine in LIMSTAT (Hunter & Lim, year unknown). This program compares the observed correlation matrix for the items in a scale to the matrix generated by the products of their factor loadings (expected matrix). If the deviation between the observed matrix and the expected matrix is smaller than sampling error would produce, the scale is regarded as internally consistent (Hunter & Gerbing, 1982). As shown in Tables 9b and 10b, not one individual correlation coefficient deviated significantly from the expected matrix at .05 alpha level. It is therefore judged 104 Table 9a: Items and Factor Loading for Ab-u Item Content Loading=1 1 useful/useless .90 2 valuable/worthless .96 3 beneficial/harmful .79 Coefficient Alpha for Ab-u is .91. Table 9b: Deviation Matrix for Ab-u Items 1 2 3 0001 -- "'001 0000 -- No deviation was significant at .05 level. Table 10a: Items and Factor Loading for Ab-h Item Content Loading= 1 pleasant/unpleasant .93 2 nice/awful .98 3 agreeable/disagreeable .86 Coefficient Alpha for Ab-u is .94. 105 Table 10b: Deviation Matrix for Ab-h Items 4 5 6 4 -- 5 0.01 -— 6 -.01 0.00 -_ No deviation was significant at .05 level. that the items used to measure Ab-u and Ab-h passed the internal consistency test. The last criterion for scale unidimensionality is parallelism, or external consistency test, which concerns how items within a cluster correlate with items in external clusters and the external clusters themselves (Hunter & Gerbing, 1982). This criterion is based on the logic that if items within a cluster are unidimensional, they should correlate in similar ways with outside constructs. Statistically, the correlation between an item (x1) in one factor (F1) and an item (Y1) in another factor (F2) can be expressed as the product of each item's factor loading and the correlation between the two faCtors (run a rxln x run). To perform the parallelism test, first the correlations between each of the items within a cluster and an outside cluster are averaged. Then, the individual correlations for each item in the cluster with the outside cluster are 106 compared to the computed average correlation. LIMSTAT employs a similar procedure used for the internal consistency test and it computes the observed and expected matrices based on the factor loadings and the inter-factor correlations. Then it compares the two matrices to examine whether there are any significant deviations between the two matrices. If the deviations are not significant at a preset level (.05 for example), the items in each cluster are interpreted to be parallel to each other, suggesting the unidimensionality of the items in the scale. Applying the parallelism criterion, Table 11 shows that the observed matrix did not deviate significantly from the expected matrix at .05 alpha level, thus showing that the items passed the parallelism test. In sum, the three criteria for scale unidimensionality confirmed that the six items used to measure brand attitude do indeed measure two separate dimensions as expected. Therefore, the two components of Ab (utilitarian vs. hedonic) will be used in subsequent analyses. 107 Table 11: Deviation Matrix Parallelism Test Items 4 5 6 l -.03 -.01 0.07 2 -.ll ‘ -.13 0.01 3 0.04 0.02 -0.16 No deviation was significant at .05 level. Utilitarian Attitude toward the Brand (Ab-u) H2b predicted that Ab-u will be more influenced by the claim strength than attractiveness of the picture. First, an Ab-u scale (Cronbach's alpha=.91, n=104) was constructed by averaging the three semantic differential measures (useful/useless, valuable/worthless, and beneficial/harmful) explained earlier. Ab-u was submitted to a 2x2x2 analysis of variance. Only claim strength had significant effects (F=8.73, p<.005) on subjects' utilitarian attitude toward the brand, supporting H2b. Subjects in the strong claim condition reported more favorable (M=5.54) Ab-u than those in the weak claim condition (M=4.73). Hedonic Attitude toward the Brand (Ab-h) H2d predicted that in the low involvement condition, Ab- h will be more influenced by picture attractiveness than by message strength. As was done for the Ab-u scale, the Ab-h scale (Cronbach's alpha=.94, n=104) was constructed by 108 averaging the scores on the three semantic differential items (pleasant/unpleasant, nice/awful, and agreeable/disagreeable) discussed in Chapter III. ANOVA results showed that picture attractiveness (F=13.61, p<.0001) and claim strength (F=7.11, -p<.01) both had significant effects on subjects' Ab-h, thus partially supporting H2d. Subjects who saw the attractive picture reported more favorable Ab-h (M=5.10) than those who saw the unattractive picture (M=4.22). Also, the strong claims produced higher Ab-h score (M=4.97) than the weak claims (M=4.34). No interaction effects were found. Brand Beliefs Subjects' responses to the four semantic differential measures of brand attributes explained in Chapter III were summed together and then divided by four to construct their brand belief scale (Zbi in Table 8). This scale was also submitted to a 2x2x2 analysis of variance. Claim strength (F=24, p<.001) and involvement (F=6.96, p=.01) had significant main effects on subjects' brand beliefs. Subjects who were exposed to strong claims reported more favorable brand beliefs (M=S.46) than those who received weak claims (M=4.49). Subjects in low involvement condition (M=5.24) reported more favorable brand beliefs than those in high involvement condition (M=4.72). It is inferred that when consumers are not highly involved with the ad processing, their mental defense mechanism is working at a lower level, thus producing more favorable brand beliefs. In 109 contrast, when their involvement level is high, consumers are more likely to scrutinize all the claims made in an ad. This attentive processing naturally increases the probability of the claims made in the ad being discounted, thus leading to a less favorable level of brand beliefs. There were interaction effects (F=8.6, p<.01) between involvement and claim strength as shown in Figure 7. This result shows that subjects who were exposed to strong claims reported more favorable brand beliefs in general. This pattern was especially prominent when the involvement level was high. As illustrated in Figure 7, however, weak claims produced more favorable brand beliefs than strong claims when the involvement level was low. 5 .. a“! 5 . .2 8 —o— mgmwmmm 3 ---~~-O--- Lmhwohmmt 1'. O 4 _ 3 . . f Weak Strong Claim Strength Figure 7: Interaction effects between Claim and Involvement on brand beliefs 110 Estimation of the Proposed Model The proposed model was estimated by the PATH sub-routine in PACKAGE (Hunter et al., 1980) which employs the ordinary least squares method. The ordinary least squares estimation was adopted for the study because the maximum likelihood estimation method used in LISREL (J6reskog & SOrbom, 1986) tends to be unstable when the sample size is small (Fornell & Larcker, 1981). A careful examination of the correlation coefficients among the constructs (Table 7) revealed the presence of interactions by the ad processing involvement level. For an illustration, the correlation coefficient between brand cognitions (Cb) and utilitarian attitude toward brand (Ab-u) for the pooled sample was .36 and it was significant at .05 level. When broken down by the involvement levels, however, the relationship between the constructs became quite different. The high involvement sample produced a stronger correlation (r=.42, p=.002) than the pooled sample. For the low involvement sample, in contrast, there was no significant correlation (r=.24, p=.09) at the .05 alpha level. Correlation coefficients among the other constructs showed a similar pattern, clearly necessitating separate analyses of the two samples. This decision turned out to be right as will be explained below. The usually employed global test of goodness of fit for a path model is an overall Chi-square test (Bagozzi & Yi, 1988, 1989). Conceptually, the Chi-square test is a 111 likelihood ratio statistic that tests a "hypothesized model against the alternative that the covariance matrix is unconstrained" (Bagozzi & Yi, 1988). Or, the Chi square statistic tests ”the goodness of fit between the covariance matrix for the observed data and covariance matrix derived from a theoretically specified model" (Fornell & Larcker, 1981). In plain English, the null hypothesis states that the model fit the data3 and usually either .05 or .1 alpha level is used for significance test (Bagozzi & Yi, 1988, 1989; Fornell & Larcker, 1981). Therefore, when the p-value related with the chi square is greater than .05, for example, we fail to reject the null hypothesis and accordingly conclude that the hypothesized model fit the data. The proposed model was tested using the criterion explained above. Path specifications were altered systematically in an effort to find the model that fits the data best, first for the high involvement sample. Then the same effort was made for the low involvement sample. Table 12 shows the results of the path analysis. The Chi square fit statistics indicate that the model for the high involvement represents an adequate representation of the data (Chi square=16.61, p>.50). Notice the Aad-c to Aad-no path for the high involvement sample. Miniard et a1. (1990) 3The chi square test of significance for path models is conceptually distinctive from most other statistical tests in that most other tests assume that the null is correct. 112 specified reciprocal paths between the two in their study. This study did not follow their specifications for one methodological reason and one substantive reason. The methodological reason stems from the fact that the PACKAGE program does not assess nonrecursive (reciprocal pathS) models. A substantive reason is the theoretical ambiguity of the reciprocal path. Miniard et a1 maintained: The figure also indicates the possibility of causal relationships being present among the Aad components. The path from.Aad—nc to Aad—c reflects the potential for the affect generated by nonclaim elements to serve as a peripheral one that influences the acceptance of ad claims. This relationship also may occur when a nonclaim element serves as an argument relevant to the ad claims. Conversely, the claims may shape one's reactions to the nonclaim elements, thus accounting for the path from Aad-c to Aad-no (1990, p. 299). Although they discussed the possible relationship between the two Aad components, they did not clearly specify the causal path by experimental conditions. Interestingly, however, their results show that the Aad—no -> Aad-c path was significant only for the low involvement condition. In contrast, the Aad-c -> Aad-nc path was significant only for the high involvement condition. Because of these two reasons, this study specified the Aad-no to Aad-c path for 113 the low involvement sample and the Aad-c to Aad-nc path for the high involvement sample. Table 12: Path Analysis Results for the Proposed Model Coefficients by Condition Path High Invol Low Invol CRnc --> Aad-no .72 (<.01) .74 (<.01) CRc -—> Aad-c .40 (<.01) .03 (.81) Cb --> Ab-h .28 (.03) -.10 (.53) Cb --> Ab-u .11 (.47) .05 (.73) Aad-no --> Cb -.02 (.81) .22 (.13) Aad-nc --> Ab-h .37 (<.01) .40 (<.01) Aad—c --> Ch .61 (<.01) .34 (.02) Aad-c --> Ab-u .51 (<.01) .44 (<.01) Ab-h -—> PI .16 (.25) .22 (.11) Ab-u --> PI .50 (<.01) .40 (<.01) Aad-no -—> Aad-c NA .42 (<.01) Aad-c --> Aad-no .15 (p=.06) NA Fit Statistics Chi square 12.93 14.91 Degree of freedom ‘ 15 15 p-value >.50 >.50 Legend: CRnc: cognitive response about non-claim elements CRc: cognitive response about claim Cb: brand cognitions 114 Because of the lack of theory about the relationship between Ab-u and Ab-h and the methodological limitations of the PACKAGE program, this study assumed a reciprocal path between Ab-u and Ab-h without testing them. Note that path analysis results are sensitive to the number of path specifications requested in the model. A larger number of path specifications would produce a bigger Chi square statistic, hence making it easier to reject the null hypothesis which assumes that the model fits the data. When the Aad-no to Aad-c path was not requested for the low involvement sample, the model did not fit the data (Chi square=30.20, p<.05). With the addition of the Aad-no --> Aad-c path, the model fit the data very well (Chi square=14.91, p>.50). As a way of verifying whether the improvement is totally an artifact of the additional path specification, the same model was tested again with the addition of the Aad-c to Aad-nc path instead of the Aad-nc to Aad-c. The result showed that the model with the Aad-c to Aad—no path (Chi square=20.89, p>.20) is better than the original one without any path between the constructs, again suggesting the possibility of the fit improvement being a product of the statistical artifact. A comparison of the Chi square statistics for the two different specifications, however, clearly demonstrates that the Aad-no --> Aad-c specification is a better representation of the data than the other way around. Also, the Aad-no to Aad-c path seems to conform to the traditional theorization of the peripheral 115 cues influencing the claim-based attitude when the involvement level is low. Comparison of the Path Model by Involvement Condition Table 12 presents the path analysis results by involvement conditions. Based on the correlation analysis, an earlier section discussed the general hypothesis that cognition --> attitude (e.g., Cb-->Ab-h) paths would be stronger for the high involvement condition than the low involvement condition. The path analysis results seem to generally support the earlier discussion. The CRc -—> Aad-c link, for example, was significant for the high involvement sample but not for the low involvement sample. The Cb-->Ab link (especially the Ab-h component) showed the same pattern, with the Cb-->Ab-u path for the low involvement sample insignificant, thus supporting H3e. The Aad-no --> Cb link was marginally significant only for the low involvement condition (.22, p=.13) as expected in H3c. This result is consistent with Miniard et al.'s (1990) finding. It is also consistent with the Elaboration Likelihood Model (Petty & Cacioppo, 1981, 1986) which posits differentially important roles of issue-relevant central cues (claims, in this case) versus non-issue-relevant peripheral ones (the picture that led to Aad-nc, in this case) in the persuasion process. The results seem to show that the indirect mediating effects of Aad-no on Ab through brand cognitions (MacKenzie et al.'s Dual Mediation Model, 1986) is 116 present only when the message processing involvement level is low. In the same vein, H3d predicted that the Aad-c -> Cb link would be stronger for the high involvement condition than the low involvement condition. As shown in Table 9, the path coefficient for the high involvement sample (.61) is much stronger than that of the low involvement sample (.34), thus supporting the hypothesis. 33a and H3b predicted that the route through which brand attitude is formed would be different by involvement conditions. Specifically, they predicted that when involvement is high, the Aad-c -> Ab-u -> PI link would be more dominant than the Aad-nc -> Ab-h -> PI link. In contrast, in the low involvement situation, the Aad-nc -> Ab- h -> PI link would be more dominant than the Aad-c -> Ab-u -> PI link. These hypotheses were supported by the path coefficients presented in Table 12. For the high involvement condition, the path coefficient between Ab-u and PI (.50, p<.001) was found to be greater than the path coefficient between Ab-h and PI (.16, p>.10), thus supporting H3a. The Ab—u to PI path was stronger than the Ab-h to PI path also for the low involvement condition, thus seemingly negating H3b. A careful comparison of individual path coefficients, however, suggests that the hypothesis should not be rejected entirely. The Aad-no to Ab-h path was stronger for the low involvement sample (.40) than the high involvement sample (.37) although both were significant at .01 level. In addition, the Ab-h to PI path was also stronger for the low 117 involvement sample (.21, p=.10) than the high involvement sample (.16, p=.26), suggesting that the direction is consistent with the hypotheses. Non-claim .72 Cognitions " Aa‘d-nc l I .15 l Claim .40 A |d_ '37 Cogntions ’ a C .61 .51 ' Ab-h .28 Brand Cognitions 59 . y PI Ab-u Figure 8: Path Model for the High Involvement Sample 118 Summary The path models for the high and low involvement conditions are presented in Figure 8 and Figure 9, respectively. As they illustrate, when subjects' ad processing involvement was high, cognitions (CRnc, CRc, Cb) played a more important role in the persuasion process than when the subjects' involvement level was low. (As expected, the major route to purchase intention was the Aad-c to Ab-u path. And this path was initiated by cognitive response to the claims (CRc) made in the ad. The role of Cb was found to be weaker than expected. The Cb to Ab-u path (.11, p=.47) was not significant. The effects of Cb to PI was found to be indirect through Ab-h. Figure 9 presents the advertising response model for the low involvement sample. It shows that brand cognitions had no causal influence to brand attitude or to purchase intention. Similarly, claim cognitions had no effects on Aad-c when subjects' ad processing involvement was low. As anticipated, the CRnc to Aad-no was the most dominant path leading to brand attitudes. The Aad-nc to Ab-h path was strong (.40) and the Ab-h to PI path was marginally significant (.21, p=.10). The Ab-h, however, had a significant indirect effect on PI through Ab—u. There was 119 Non-claim .74 Cognitions , Aad-nc .40 Ab'h \ \.21 \ \ ‘ Brand Cognitions 36 P I ‘Ab-u",//’<2fi;”)'r Figure 9: Path Model for the Low Involvement Sample also a significant causal path (.42, p<.01) from Aad-nc to Aad-c, suggesting the possibility of peripheral cues (such as picture) exerting an important causal influence to attitude formation in low involvement situations. Remember that this was not the case for the high involvement sample. CNAPTER V DI SCUSS ION AND CONCLUS IONS The final chapter includes a discussion of the research findings followed by managerial implications of the study. It concludes with limitations of the study and suggestions for future research. The Formation of Aad Since the empirical demonstration by Michell and Olson (1981) that attitude toward the ad is an important mediator of brand attitude and purchase intention, many researchers have investigated the role of Aad and the process of its formation. Previous research (e.g., Mitchell, 1986; Park & Young, 1986) also studied the differential contribution of central processing cues (e.g., claim strength) and peripheral cues (e.g., picture attractiveness) to the formation of Aad. In a recent study, Miniard et al. (1990) enhanced our understanding of the field by decomposing Aad into the claim and non-claim components. Their basic argument was that the effects of central and peripheral cues in an ad on the formation of Aad should be examined separately by decomposing Aad in order to better understand the relationship between Aad and Ab. The ELM (Petty and Cacioppo, 1981, 1986) posited that the role of peripheral cues and central cues on Ab would be different depending on the level of ad processing involvement. Specifically, they predicted that the central 120 121 route would be the main mechanism through which persuasion (formation of Ab, in this case) occurs in a high involvement setting. In contrast, in a low involvement situation, peripheral cues would be the main route through which attitude formation would occur. Based on this theory, several studies (Homer, 1990; MacKenzie et al., 1986; Yoon, 1991) investigated the role of Aad on the formation of Ab in relation with involvement. They all conceptualized Aad as .representing the peripheral cues, hence predicting that the Aad -> Ab path would be stronger in a low involvement situation than a high involvement situation. Contrary to the expectation, however, they all found that the Aad -> Ab path was stronger for the high involvement condition. These findings clearly suggested that either the ELM theorization was wrong or the conceptualization and measurement of Aad as representing an attitude formed by peripheral processing was invalid. Decomposition of Aad into the two components seems to offer some insights as to the nature of Aad in relation with ELM theorization. _ Results of the present study indicate that manipulation of claim.and picture differentially influence Aad—c and Aad- nc. As anticipated, Aad-c was more responsive to manipulation of claim strength, whereas Aad-nc was more responsive to the picture manipulation. Only claim strength had significant effects on Aad-c. Likewise, only picture attractiveness had significant effects on Aad-no. These findings suggest the diagnostic value of decomposing Aad. 122 Regarding the differential effects of Aad-c and Aad-nc on Ah, the results indicated the validity of decomposing Aad into the central and peripheral components. The path analysis results showed that central processing (Aad-c) is a relatively more dominant route to Ab in the high involvement situation. In contrast, peripheral processing (Aad-nc) was relatively more dominant in the low involvement condition. The Formation of Ab This study posited that Ab is not a unidimensional construct and decomposed Ab into the utilitarian and hedonic components. It was hypothesized that Ab-u would be the more dominant route to PI in the high involvement condition, whereas Ab-h would be the more dominant route to PI in the low involvement condition. Also, the study expected that the two different components of Ab would have differential sensitivity to manipulations of the claim and picture in the ads. As anticipated, the results showed that only manipulation of claim strength had significant effects on Ab- u. Also, the Ab-u -> PI path was stronger for the high involvement condition than the low involvement condition. Although the coefficients for Ab-h -> PI path for both the high and low involvement samples were statistically insignificant, the low involvement sample revealed a stronger coefficient than the high involvement sample, as expected. 123 These results suggest the potential usefulness of decomposing Ab in predicting PI. Contrary to the expectation, both claim strength and picture attractiveness influenced Ab-h. This finding . suggests that consumers' hedonic attitude toward a brand is formed as a results of a general evaluation of the entire advertising stimulus, not just the peripheral elements in the ad. Discussion of the Model Claim.cognitions and brand cognitions had significant causal influence on the formation of Aad-c and Ab, respectively, only for the high involvement sample. As explained earlier, Aad-c initiated by claim cognitions was the predominant mediator to Ab-u, which in turn mediated purchase intention. One disturbing result is the insignificant path between brand cognitions and Ab-u, which was anticipated to be the major route to P1 in the high involvement condition. This unintuitively weak effect of Cb on Ab-u deserves our attention. Several post hoc speculations are possible for this result. First, this finding is essentially consistent with MacKenzie et al.'s (1986) finding. They speculated that it might have been caused by the unknown quality of the Cb measurement. Subjects' brand cognitions were measured immediately after they were exposed to the stimulus ads. The instruction to attend to the ads might have been so powerful that subjects 124 may have paid an unnaturally high amount of attention to the ads, thus producing a smaller number of brand cognitions than they would normally do in a natural setting. This unnaturally small number of brand cognitions, in turn, may be the cause of the weak Cb -> Ab-u path. Second, considering the moderately strong path coefficient (.11), it is possible that the small sample size caused the statistical insignificance. Another possibility is found in the product category used in the study. Suntan lotions may not have high enough involvement for the subjects to cause them to invoke brand cognitions at all. Another explanation is that suntan lotions may be a ”hedonic" brand category, so that the subjects only utilized hedonic cognitions (as compared to utilitarian cognitions), thus showing a significant path from Ch to Ab-h. In the low involvement condition, neither claim cognitions nor brand cognitions had significant effects on Aad-c and Ab, respectively. The path analysis results indicate that Aad-nc initiated by non-claim cognitions were the only mediator through which brand attitudes were formed. Aad-no had a strong causal influence on Ab-h. It also had a significant influence on Ab-u indirectly through Aad-c, which in turn strongly influenced Ab-u. Interestingly, Aad-nc had no significant influence on brand cognitions. This finding is consistent with Miniard et al.'s (1990) and Yoon's (1991) findings. These results suggests that MacKenzie et al.'s Dual Mediation Hypothesis (1986) may not be the best model to 125 explain the relationships among the variables used in the above mentioned studies. The Ab to PI Path By decomposing Ab into Ab-u and Ab-h, this study anticipated a differential path strength from those Ab components to PI across the involvement conditions. Specifically, it was predicted that Ab-u -> PI would be more predominant than Ab-h -> P1 in the high involvement condition. In contrast, the Ab-h -> PI path was expected to be more predominant than the Ab-u to PI path in the low involvement condition. As anticipated, the Ab—u -> PI was the predominant path in the high involvement condition. Contrary to the expectation, however, the same was true for the low involvement condition. Two speculations are offered for this finding. This study's failure to support the hypothesis may be an artifact of the failure to create a true low involvement condition. All subjects may have had such a high level of involvement that they may have employed the utilitarian brand attitudes in forming purchase intention. There is weak evidence, however, which indicates that the hypothesis of differential contribution of the Ab components to PI may be correct. The Ab-h -> PI path was stronger for the low involvement sample than the high involvement sample. Although both path coefficients were statistically insignificant, the results were in the anticipated direction, 126 suggesting that this topic is worth pursuing further in the future. The second speculation concerns the nature of the stimulus ad used in the study. The ads were fact-oriented informative ads for a hypothetical brand. Accordingly, the subjects in the low involvement conditions might not have had an opportunity to develop any affective responses to the brand, thus producing a weak path between the Ab-h to PI. This speculation is especially plausible because subjects' responses were measured after just a single exposure to the ads. Managerial Implications The most important practical implication of the study is the proposed model's diagnostic ability to predict consumers' purchase intention of a brand after seeing an ad of the brand. In other words, the model allows examination of the paths among the variables, thus providing a marketing manager with the insights of how effective the ad will be. For example, if a company is trying to market a highly hedonistic product, the company naturally should be interested in making sure that the target market has a favorable hedonic attitude toward the brand. The model can offer an indication as to whether consumers do have a good hedonic attitude toward the product or not. If consumers are found not to have favorable hedonic attitudes toward the product, the company should 127 alter either the peripheral cue used in the ad or the claims made in the ad. Limitations and Suggestions for Future Research The findings and conclusions of the study are qualified by several limitations which stem from the experimental research method. The first limitation is the external validity of the results since the research was conducted in a laboratory setting. The findings reported here may not be replicable outside the laboratory. A related problem is the subjects used in the study. Students enrolled in various advertising courses participated in the study to get extra credit. Being advertising students, their interest in advertising in general is expected to be higher than non- advertising majors. Therefore, they might have paid an unusually high level of attention to the advertising stimuli, thus producing abnormal results. Also, while suntan lotion is a product that college students would consider for purchase, the sample is not representative of the market segment for suntan lotions. To enhance the external validity of the findings, therefore, further research should be conducted under different conditions (such as natural nonlaboratory settings) with different subject populations (e.g., actual consumers, instead of students). A second limitation involves the product used in the study. A product with neutral involvement was used in this study intentionally in order to eliminate the confounding 128 effects of product involvement on the other variables such as advertising processing involvement. It is, however, not difficult to imagine that the two different types of involvement are interacting in reality. The results of the study indeed indicate the possibility of product involvement interacting with ad processing involvement in producing the paths between Cb and Ab-h, and Cb and Ab—u. Product involvement is also expected to influence the other paths. For example, if a highly involving product due to its utilitarian reasons (e.g., tires) had been used as the stimulus, the Ch -> Ab-u path (which was not significant in the study) might have turned out to be significant, especially for the high involvement sample. Future research should employ a variety of products with varying levels of product involvement and examine the proposed relationships in the study. A third limitation involves the measurement of subjects' responses immediately after just one exposure to the ads. A delayed measure after multiple exposures may produce different results. In fact, the single exposure to an ad for a hypothetical brand is speculated to be the cause of the insignificant path between Cb and Ab. Therefore, future research should investigate the proposed model by employing varying levels of exposure and delay of measurement. Another limitation is the use of human skeleton as a peripheral cue. Such a blatantly unattractive picture is very unusual, especially in suntan lotion ads, unless the 129 objective of the advertising campaign is to dramatically enhance the attention-getting ability of the ad. The skeleton picture also might have caused a confounding effect, i.e., fear appeal when the intended purpose was only ”unattractiveness.” This certainly is an important issue to be further investigated. Finally, another limitation of this research is its inability to compare the present model with Miniard et al.'s (1990) model because of the scales used to measure Ab-u and Ab-h. They employed the general measures of brand attitudes (favorable/ unfavorable, positive/negative, like/dislike). One method of comparing the power of the two models is to use both Ab-u, Ab-h scales and the general Ab scale in a study and compare the fit of the models. Conclusion The study investigated several issues related to the formation and causal role of Aad under conditions involving a new ad for a hypothetical brand. One of the contributions of the study is that it replicated Miniard et al.'s (1990) study and verified the usefulness of decomposing Aad into Aad-c and Aad-no. The decomposition of Aad enhanced our understanding of the nature of Aad's influence on Ab and Cb across involvement conditions. They explained: "The fact that much of Aad's influence on Ab, and all of Aad's influence on brand cognitions, can be tracked through Aad-c rather than Aad-no further 130 testifies to the inadequacy of attributing Aad's influence simply to peripheral persuasion mechanism." (p. 301) Thus, the consistent results of the two studies indicate that the Aad -> Ab relationship should not be regarded as a result of only peripheral persuasion. These findings then seem to indicate that the role of attitude toward the ad in the persuasion process may be even more important than believed to be until now. Another contribution of the study is the decomposition ‘of Ab into hedonic and utilitarian components. Despite several disagreements in the detail, the model generally appears to be promising in its potential as a diagnostic tool with further refinements. APPENDICES APPENDIX A APPENDIX A Questionnaire The following is the questionnaire used to collect responses after the test ads were shown to the subjects. The only difference between the high and low involvement conditions was the instructions given to the subjects as explained in Chapter III. 131 1 3 2 CONSENT FORM In turning in my questionnaire for this study, I understand that my participation has been voluntary, and that I am free to discontinue my participation at any time, without forfeiting the benefits otherwise entitled to me. I also understand that all results will be treated with strict confidentiality and that the subjects will remain anonymous in any report of research findings. Make sure you write down your name, student number and your instructor's name below to get proper credit. After you complete the questionnaire, separate this page from the rest of the questionnaire and return it to the researcher. This consent form will be given to your instructor for proper credit. This will insure that your responses to these questions will remain anonymous. Please do not talk to anyone during this session. We are interested in your opinions. We hope that you will enjoy being a participant in this study and that you find it a good learning experience. Name (pn'nt): Signature: Student Number. Instructor's Name: 9 Please turn to page 2. 133 (Instructions for the high involvement condition) Thank you for joining us today. In this study we are attempting to obtain your reactions to two new brands of suntan lotion which are being considered for introduction in this area. Please raid the followigginstructiong very carefqliy. According to Con_sumer Reports. there are substantial differences in quality among major brands of suntan lotion available on the market as measured by functional performance. You will be shown a print ad for each of two new brands of suntan lotion and then asked to make a choice between the two brands. The ads may contain information on those performance characteristics as well as other features. Please assume that you are about to take a trip to Bahamas and you need to purchase a suntan lotion, and read the ads as though you were trying to learn about the brands’ benefits and effectiveness. In other words, please focus on the product information provided in the ads. At the end of the session you may receive the brand of your choice as an additional compensation for your participation. Now read the ads right next to you. Please stop here. 134 (Instructions for the high involvement condition) Thank you for joining us today. In this study we are attempting to obtain your reactions to two new brands of suntan lotion which are being considered for introduction in this area Please read the following instructiong very rare ully. According to Consumer Repgrts, there are pp substantial differences in quality among major brands of suntan lotion available on the market as measured by functional performance. You will be shown a print ad for each of two new brands of suntan lotion and then asked to make a choice between the two brands. The ads may contain information on those performance characteristics as well as other features. Assume that you are preoccupied by a very difficult class project due tomorrow. Furthermore, now that summer is over you do not plan to buy a suntan lotion in a near future. You are asked to look at the ads as if you were sitting in your living room, worrying about the project, and knowing that you have no need to purchase a suntan lotion. Now look at the ads right next to you. Please stop here. 135 Which brand would you like to receive as a gift, Rayblock or Sun Shade? Rayblock Sun Shade In the spaces below, please write down any thoughts, feelings, ideas, and images that occurred to you while looking at the Rayblock ad. Please list the thoughts that occurred to you about this product and the ad. Also, feel free to mention any other thoughts that you had while viewing the ad. Remember, list the thoughts that occurred to you while seeing the ad. Do not worry about spelling or punctuation. 136 Next, we would like to ask you a few additional questions about the Rayblock suntan lotion advertisement In obtaining your reaction to the ad, we would like you to distinguish between two basic components of the ad. The first component involves the claims made about the product The second component involves the remaining elements within the ad such as the format, pictures or illustrations, colors, type style, spacing, and so forth. Concerning the first component, how would you evaluate the claims made about Rayblock? persuasive : : : : : : unpersuasive informative : : : : : : uni nf ormati ve strong : : : : : : weak believable : : : : : : unbelievable Concerning the second component, how would you evaluate the remaining elements (everything except the claims) within the Rayblock ad? 1 2 3 4 5 6 7 good : : : : : : bad positive : : : : : : negative favorable : : : : : : unfavorable 137 Now, turn your thoughts to the brand Rayblock, instead of the ad. Please describe how you feel about it on the given seales. useful : : : : : : useless valuable : : : : : : worthless beneficial : : : : : : harmful pleasant : : : : : : unpleasant nice : : : : : : awful agreeable : : : : : : disagreeable Please indicate the extent to which you believe the following statements to be true about the advertised brand, Rayblock. Rayblock protects you from the harmful rays of the sun. extremely : : : : : : extremely likely unlikely Rayblock is a non-greasy sunscreen lotion. extremely : : : : : : extremely likely unlikely Rayblock is a waterproof sunscreen lotion extremely : : : : : : extremely likely ~ unlikely Rayblock is a hypoallergenic sunscreen lotion that does not cause allergies. extremely : : : : : : extremely likely unlikely 138 Next, please indicate on the scales below the likelihood that you will purchase Rayblock suntan lotion when it becomes available in this area. likely : : : : : : unlikely probable : : : : : : not probable Next, indicate the extent to which you believe the following statements to be true about Rayblock advertisement: I paid attention to the ad’s claims about Rayblock‘s functional performance. very much _: : : : : : not at all I concentrated on the messages in the Rayblock ad. very much : : : : : : not at all I put thought into evaluating the messages in the Rayblock ad. very much : : : : : : not at all When I saw the advertisement for Rayblock, I felt the information might be relevant for my needs. strongly agree : : : : : : strongly disagree 139 Please indicate below your evaluation of the picture and copy of the Rayblock ad. My overall feelings toward the picture used in the Rayblock ad is... 1 2 3 4 5 6 7 Good : : : : : : Bad Like : : : : : : Dislike Pleasant : : : : : : Unpleasant Attractive : : : : : : Unattractive Please indicate how you feel about the claims made in the Rayblock ad on a seven-point scale where one represents ”strong" and seven represents "weak." I 2 3 4 5 6 7 strong : : : : : : weak Here is one last thing for you to do. Please go back to page 3 to the thoughts you listed. Indieate the extent to which each thought you listed was favorable or unfavorable on a seven-point seale where “1” means “favorable” and “7” means “unfavorable.” Example) I think RavblockLis an excellent sunscreen. (1) APPENDIX B APPENDIX B CODING OP COGNITIVE RESPONSES A. Thoughts directed to the brand 1. Counterarguments: Statements which are directed against the idea of or the use of the brand and which: a. state a specific unfavorable consequences of using the product b. state a specific undesirable attribute of the product c. suggest an alternative method for handling one of the problems cited in the advertising message d. state a specific favorable or desirable consequence or attribute of an alternative brand e. challenge the accuracy or validity of a specific argument contained in the advertising message 2. Support Arguments: Statements which are directed in favor of the idea or use of the brand and which: a. state a specific favorable consequence of using the brand b. state a specific desirable attribute of the brand c. suggest an undesirable consequence of not using the brand d. reaffirm the accuracy or validity of an argument presented in the commercial 140 B. 3. 4. 5. 7. 8. 9. 141 Thoughts directed toward the ad ' Positive statements concerning ad claims Negative statements concerning ad claims Positive statements concerning the other elements of the ad including picture, layout, typeset, etc. Negative statements concerning the other elements of the ad including picture, layout, typeset, etc. Overall positive statements about the ad Overall negative statments about the ad All others Random, irrelevant thoughts BIBLIOGRAPHY BIBLIOGRAPNY Aaker, D. A., Stayman, D. M., & Hagerty, M. R. (1986). Warmth in advertising: Measurement, impact and sequence effects. Journal of Consumer Research, 1;, 365-381. Allport, G. W. (1935). Attitudes, in Murchison, C. (Ed.). Handbook of Social Psycholggy (v01. 2). Worcester, Mass.: Clark University Press. 798-844. Andrews, J. C., Durvasula, S., & Akhter, S. H. (1990). A framework for conceptualizing and measuring the involvement construct in advertising research. Journal of Advertising, 12(4), 27-40. Areni, C. S. & Lutz R. J. (1988), The role of argument quality in the elaboration likelihood model. in Houston, M. J. (Ed.). Advances in consumer research (Vol. 15). Provo, Utah: Association for Consumer Research, 197-203. Ashcraft, M. H. (1989). Human memory and cggnition. Glenview, IL: Scot, Foresman and Company. . Babbie, E. (1986). The practice of gggial research (Fourth Edition). Belmont, CA: Wadsworth Publishing Co. Bagozzi, R. P., & Yi, Y (1988). On the evaluation of structural equation models. Journal of the Academy of Marketing Science, 16(1), 74-94. Bagozzi, R. P., & Yi, Y (1989). On the use of structural equation models in experimental designs. Journal of Marketing Research, 26 (August), 271-284. Batra, R. (1985). Understanding the likability/involvement interaction: The "Override" Model." in Hirschman, E. C., a . Holbrook M. B. (Eds.). Advances in consumer research (Vol. 142 143 12). Ann Arbor, MI: Association for Consumer Research, 362- 367. Batra, R. (1986). Affective advertising: Role, processes, and measurement. in Peterson, R. A., Hoyer, W. D., & Wilson, W. R. (Eds.). The role of affect on conggmer behavior. Lexington, MA/Toronto: Lexington and Company, 53-85. Batra, R & Ahtola, O. T. (in press). Measuring the hedonic and utilitarian sources of consumer attitudes. Marketing Letters. Batra, R. & Holbrook, M. B. (1990). Developing a typology of affective responses to advertising. Ps cholo & Marketin , 1 (1), 11-25. Batra, R., & Ray, M. (1983). Operationalizing involvement as depth and quality of cognitive response. in Bagozzi, R. P., & Tybout, A. M. (Eds.). Advances in Consumer Resgarch (Vol. 10). Ann Arbor, MI: Association for Consumer Research, 309- 313. Batra, R., & Ray, M. (1985). How advertising works at contact. in Alwitt, L. F., & Mitchell, A. A. (Eds.). Psychologicalprocesses and advertising effects. Hillsdale, NJ: Lawrence Erlbaum Associates, 13-43. Batra, R., & Ray, M. (1986). Affective responses mediating acceptance of advertising. Journal of Consumer Research, 11(September), 234-249. Burke, M. C., & Edell, J. A. (1986). Ad reactions over time: Capturing changes in the real world. Journal of Consumer Research, 11(1), 114—118. Burton, 8., & Lichtenstein, D. R. (1988). The effect of ad claims and ad context on attitude toward the advertisement. gpurnal of Advertiging, 11(1), 3-11. Cacioppo, J. T., 5 Petty, R. E. (1989). The elaboration likelihood model: The role of affect and affect-laden information processing in persuasion. in Cafferata, P., a 144 Tybout, A. M. (Eds.). Cognitive and affective respgnses to advertiging. Lexington, MA: Lexington Books, 69-89. Fishbein, M. (1963). An investigation of the relationships between beliefs about an object and the attitude toward that object. Human Relagions, 1g, 233-240. Fishbein, M., & Ajzen, I. (1975). Belief. attitude, intention and behavior: An introduction to theogy and research. Reading, MA: Addison-wesley. Flemming, H. (June, 1981). Hemispheral lateralization: Implications for understanding consumer behavior. Journal of Consumer Research, a, 23-36. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. JOurnal of Marketing Research, 18 (February), 39-50. Gardner, H. (1985). The mind's new science: A history of the cognitive revolution. New York: Basic Books, Inc., Publishers. Gardner, M. P. (1985). Does attitude toward the ad affect brand attitude under a brand evaluation set? Journal of Marketing Research, 11, 192-198. Gardner, M. P., Mitchell, A. A., & Russo, J. E. (1985). Low involvement strategies for processing advertisements. Journal of Advertising, 11(2), 4-16, 56. Gelb, B. D., & Pickett, C. M. (1983). Attitude toward the ad: Links to humor and to advertising effectiveness. Journal of Advertising, 11(2), 34-42. Gill, J. D., Grossbart, S., & Laczniak, R. N. (1988). Influence of involvement, commitment and familiarity on brand beliefs and attitudes of viewers exposed to alternative ad claim strategies. Journal of Advertisin , 11(1), 33-43. 145 Gorn, G. (1982). The effects of music in advertising on choice behavior: A classical conditioning approach. Journal of Marketing Research, 1s, 94-101. Greenwald, A. G. (1968). Cognitive learning, cognitive responses to persuasion, and attitude change. in Greenwald, A. 6., Brock, T. C., & Ostrum, T. M. (Eds.). Psychological foundations of attitudes. New York: Academic Press, 147-170. Gresham, L. G., & Shimp, T. A. (1985). Attitude toward the advertisement and brand attitudes: A classical conditioning perspective. Journal of Advertisin , 11(1), 10-17, 49. Heider, F. (1946). Attitudes and cognitive organization. Journal of Psychology, 11 (January), 107-112. Holbrook, M. B., & Batra, R. (1987). Assessing the role of emotions as mediators of consumer responses to advertising. Journal of Consumer Research, 11, 404-420. Holbrook, M. B., & Hirschman, E. C. (1982). The experiential aspects of consumption: Consumer fantasies, feelings, and fun. JOurnal of Consumer Reseach, 2 (September), 132-140. Homer, P. M. (1990). The mediating role of attitude toward the ad: Some additional evidence. Journal of Marketing Research, 11 (February), 78-86. - Houston, M. J., & Rothschild, M. L. (1978). Conceptual and methodological perspectives on involvement. in Jain, S. (Ed.). Essearch frontisps in marketing: Dialogues and directions. Chicago: American Marketing Association, 184-187. Howard, J. A. (1977). Consumer behavior: Application of theopy. NGW'YOIK: McGraw-Hill. Hunter, J. E. & Gerbing, D. W.(1982). Unidimensional measurement, second order factor analysis, and causal models. Research in Organizational Behavior, 4, 267-320. Hunter, J. E., Gerbing, D. W., Coehn, S. H., & Nicol, T. S. (1980). PACKAGE 1980: A svstepvof FORTRAN routines for the 146 gnalysis of correlational data. Waco, Texas: Baylor University, . Hunter, J. & Lim, T. S., Limstat version 2.0: Computer Program and Manua . East Lansing, MI: Michigan State University. Johnson, B. T., & Eagly, A. M. (1989). Effects of involvement on persuasion: A meta-analysis. Ps chol ical Bulletin, 106(2), 290-314. J6reskog, K. G. & Scrbom, D. (1986). LISREL VI: Analysis of linear structural relationships by the method of maximum likelihood. Mooresville, IN: Scientific Software, Inc. Kidder, L. H. & Judd, C. M. (1986). Research Methods in Social Relations. Fortworth, TX: Holt, Rinehart and Winston, Inc. Krugman, H. E. (Fall 1965). The impact of television advertising: Learning without involvement, Public Opinion Quarterly, 11, 349- 356. Krugman, H. E. (1968). The learning of consumer likes, preferences, and choices. in Bass, F. M., King, C. W. & Pessemier, E. A. (Eds.). Applicationspof the sciences in marketing managemept. New York: John Wiley, 207-225. Laczniak, R. N., Muehling, D. D., & Carlson, L. (1991). Effects of motivation and ability on ad-induced cognitive processing. in Holman, R. (Ed.). Proceeding of ths 1991 Conference of the American Acade of Advertisin , New York, 81-87. Lavidge, R. J., & Steiner, G. A. (1961). A model for predictive measurements of advertising effectiveness, Journal of Marketing, 1; (October), 59-62. Lutz, R. J. (1985). Affective and cognitive antecedents of attitude toward the ad: A conceptual framework. in Alwitt, L. F., & Mitchell, A. A. (Eds.). Psychological processes and advertising effects. Hillsdale, NJ: Lawrence Erlbaum Associates, 45-64. 147 Lutz, R. J. (1991). The role of attitude theory in marketing. in Kassarjian, H. H., & Robertson, T. S. (Eds.). Perspsctives in consumer behavior. Englewood Cliffs, NJ: Prentice Hall, 317-339. Lutz, R. J., MacKenzie, S. B., and Belch, G. E. (1983). Attitude toward the ad as a mediator of advertising effectiveness: Determinants and consequences. in Bagozzi, R. P., a Tybout, A. M. (Eds.). Advancsspin consumer research (Vol. 10). Ann Arbor, MI: Association for Consumer Research, 532-539. Machleit, K. A., & Wilson, R. D. (1988). Emotional feelings and attitudes toward advertisement: The role of brand familiarity and repetition. Journal of Advertising, 11(3), 27-35 0 MacKenzie, S. B., Lutz, R. J. (1989). An empirical examination of the structural antecedents of attitude toward the ad in an advertising pretesting context. Journal of Marketing, 11 (April), 48-65. MacKenzie, S. B., Lutz, R. J. (1983). Testing competing theories of advertisng effectiveness via structural equation models. In W. R. Darden & K. B. Monroe (Eds.), Proceedings of AMA Winter Educators' Conference (pp. 70-75). Chicago: American Marketing Association. MacKenzie, S. B., Lutz, R. J., & Belch, G. E. (1986). The role of attitude toward the ad as a mediator of advertising effectiveness: A test of competing explanations. Journal of Marksting Research, 11 (May), 130-143. Madden, T. J., Dillon, W. R., & Twible, J. L. (1986). Construct validity of attitude toward the ad: An assessment of convergent/discriminant dimensions. in Olson, J., & Sentis, K. (Eds.). Advertisingiand consumer psychology. New York: Praeger Publishing Co., 74-92. Maloney, J. C. (1963). Is advertising believability really important? Journal of Marketing 11 (October), 1-8. Miniard, P. W., Bhatla, S., a Rose, R. L. (1990). On the formation and relationship of ad and brand attitudes: An 148 experimental and causal analysis. Journal of Marketing Research, 11, 290-303. Mitchell, A. A. (1986a). The effect of verbal and visual components of advertisements and brand attitudes and attitude toward the advertisement. Journal of Consumer Research, 11 (June), 12-24. Mitchell, A. A. (1986b). Theoretical and methodological issues in developing an individual level model of advertising effects. in Olson, J. C., a Sentis, K. (Eds.). Advertising and consumer psychology. New'York: Praeger Publishing Co., 172-196. Mitchell, A. A., & Olson, J. C. (1981). Are product attribute beliefs the only mediator of advertising effects on brand attitude? JOurnal of Marketing Research, 11 (August), 318- 332. Moore, D. L., & Hutchinson, J. W. (1983). The effects of ad affect on advertising effectiveness. in Bagozzi, R. P., & Tybout, A. M. (Eds.). Advances in consumer research (V01. 10), Ann Arbor, MI: Association for Consumer Research, 526- 531. Moore, D. L.. & Hutchinson, J. W. (1985). The influence of affective reactions to advertising: Direct and indirect mechanisms of attitude change. in Alwitt, L. F., & Mitchell, A. A. (Eds.). Psychological processes and advertising effects. Hillsdale, NJ: Lawrence Erlbaum Associates, 65-87. Muehling, D. D., & Laczniak, R. N. (1988). Advertising's immediate and delayed influence on brand attitudes: Considerations across message-involvement levels. Journal of Advertisin , 11(4), 23-34. Olson, J. C., Toy, D. R., & Dover, P. A. (1982). Do cognitive responses mediate the effects of advertising content on cognitive structure? Journal of Consumer Research, 9 (December), 245-262. _ Palda, K. S. (1966). The hypothesis of a hierarchy of effects: A partial evaluation. Journal of Marketing Research, 1 (February), 13-24. 149 Park, W. C., a Young, M, S. (1986). Consumer response to television commercials: The impact of invOlvement and background music on brand attitude formation. Journal of Marketing Research, 11 (February), 11-24. Peterson, R. A., Hoyer, W. D., & Wilson, W. R. (Eds.). The role of affect on consumer behavior. Lexington, MA/Toronto: Lexington and Company. Petty, R. E., a Cacioppo, J. T. (1979). Issue involvement can increase or decrease persuasion by enhancing message-relevant cognitive responses, Journal of Personality and Social Psychology, 11(10), 1915-1926. Petty, R. E., & Cacioppo, J. T. (1981). Attitudes and psrsuasion: Classic and contempprapy approaches. Dubuque, Iowa: wm. C. Brown Company Publishers. Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. in Berkowitz, L. (Ed.). Advances in expsrimental social psychology (Vol. 19). New York: Academic Press. 123-206. Petty, R. E., & Cacioppo, J. T. (1989). The elaboration likelihood model: The role of affect and affect-laden information processing in persuasion. in Cafferata, P., & Tybout, A. M. (Eds.). Cognitive and affective respgnses to advertising. Lexington, MA/Toronto: Lexington Books, 69-89. Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement. Journal of Consumer Research, 19 (September), 135- 146. Ray, M. (1982). Advertising and Communication Management. Englewood Cliffs, NJ: Prentice Hall, Inc. Ray, M., Sawyer, A. G., Rothschild, M. L., Heeler, R. M., Strong, E. C., & Reed, J. R. (1973). Marketing communication and the hierarchy-of-effects. in Clarke, P. (Ed.). New models for communication research. Beverly Hills, CA.: Sage Publications, 147-176. 150 Shiffman, L. G., & Kanuk, L. L. (1987). Consgmer Behavior (3rd ed.). Englewood Cliffs, NJ: Prentice Hall, Inc. Shimp, T. A. (1985). Attitude toward the ad as a mediator of consumer brand choice. JOurnal of Advertisin , 19(2), 9-15. Silk, A. J., & Vavra, T. G. (1974). The influence of advertising's affective qualities on consumer response. in Hughes, G. D., 5 Ray, M. (Eds.). Buyer/consumer information processing. Chapel Hill, NC: University of North Carolina Press, 157-186. Smith, R. B., & Swinyard, W. (1988). Cognitive response to advertising and trial: Belief strength, belief confidence and product curiosity. Journal of Advertisin , 11(3), 3-14 Srull, T. K. (1989). Advertising and product evaluation: The relation between consumer memory and judgement. in Cafferata, P., & Tybout, A. M. (Eds.). Cognitive and affective respgnses to advertising. Lexington, MA/Toronto: Lexington Books, 121- 134. Wright, P. L. (1973). The Cognitive processes mediating acceptance of advertising. Journal of Marketing Research, 19, 53-62 0 Wright, P. L. (1980). Message-evoked thoughts: persuasion research using thought verbalization. Journal of Consumer Research, 1 (September), 151-175. Wright, P. L. (1986). Reactions to an ad's contents versus judgement of the ad's impact. in Olson, J., & Sentis, K. (Eds.). Advertisingpand Conspmer Psychology (Vol. 3). NY: Praeger. 108-117. Yoon, K. (1991). Involvement level and the mediating role of attitude toward the advsgtisement. Unpublished manuscript, Michigan State University, Department of Advertising, East Lansing. 151 Yoon, K. (1990). [Two components of attitude toward the brand]. Unpublished raw data. Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology Monograph, 1, 1- 27. "Illlllllllllllllllis