.. ‘ J. L.» . .. ‘ : .«uwrsm‘xxxm, ; ‘1 (r. . A lira I ‘ a .1), :14... 5.. V I: s - 1:...5. II (6.7.: 19:. \ not}: .1,¥ ii. I... :(«triEIes .1. |P74Lutz, MacKenzie, and Belch (1983) also postulated that the effect of Aad on A.b would be stronger for peripheral processing because under that condition the consumer does not process the message actively. [ sf //, MacKenzie and Lutz (1983) evaluated four possible 6 causal relationships among Amr Cb, Ab, and purchase intentions. They concluded that a one—way causal flow from Aad to A5 is superior to the other models although all four of the proposed models are significant explanations of the data. The strongest support was for causal flow from Aad to Cb and from Aad to Ab. Weaker support was found for dual causal flow between Aad and Ab.. A replication of the study by MacKenzie, Lutz, and Belch (1986), however, shows different results. That study found that Dual Mediation Hypothesis (DMH) model (causal flow from Aad to Cb as well as from Aad to Ab) appeared to fit the data significantly better than the other models. Park and Young (1983, 1986) suggested that type and level of involvement (cognitive, affective, and low involvement) appear to moderate the effect of Aad and Cb on brand attitude. Their research suggested that cognitive involvement would lead to a stronger effect of brand cognition than the effect of Aad on Ab. They also indicated that affective involvement and low involvement conditions would lead to stronger effect of Aad on Ab than the effect of brand cognition on Ab. ’/Gardner (1985a) reviewed studies on the effects of Aac on Ab. She found that Aad affects Ab to an approximately equal extent under both consumer information processing sets -— brand evaluation set (evaluation of the advertised brand) and nonbrand set conditions (enjoyment or evaluation of the 7 advertisement for its own sake). She also noted that brand related beliefs are more significant mediators for brand attitudes under a brand evaluation set condition than under a nonbrand set condition. Regarding affective responses (ARs), Batra and Ray (1985, 1986) confirmed that affective responses represent strong mediating influences on brand attitudes and affective responses appear to be antecedents of the Aad to have impact on Am- They, therefore, suggested the relevant chain of effect is affective responses (ARs) ——> Aad -—> Ab ——> PI. A study by Lee (1987) indicated that, in addition to thoughts (cognitive responses), feelings (affective responses) are useful for the study of consumer's response to advertisements. He manipulated the number of advertisement repetitions and concluded that high involvement consumers tend to process information more cognitively and low involvement consumers tend to process information more affectively at any level of ad repetitions. Building on the notion that the source of affect toward the brand may/be different in high and low involvement situationsp/Batra and Ray (1985) proposed the percentage contribution model. Their model posits that ad execution— based on a hedonic component should be the major contributor to purchase intentions in low involvement situations; the attribute—based utilitarian attitude component is important in high involvement situations. Additionally, the model 8 indicates that brand familiarity should be significantly greater in low involvement situations than in high involvement situations Batra and Ray (1985) suggested that brand attitudes result from two sources of affect and are, in turn, mediated by different responses. Using LISREL— V, they found no support for the contribution model. In high involvement situations, the attribute sensitive component did have a significant causal effect on purchase intentions while the ad execution sensitive component had no effect. In low involvement situations the ad execution sensitive component had a stronger causal influence on purchase intentions than the attribute sensitive component. The result, however, was not statistically significant. They also found that familiarity is the major determinant of purchase intentions in low involvement situations and that it works more through the ad—execution—sensitive component than the attribute—sensitive component. Finally, they proposed a model that suggests (a) in high message response involvement, cognitive processing evoked by message argumentation leads to changes in attribute argumentation sensitive attitudes; (b) in low message response conditions, the execution sensitive attitudes are led by execution likability and exposure frequency.“ In addition, brand familiarity influences purchase intentions directly in low message response conditions. Later, Batra (1986) made a note that this hypothetical model requires empirical 9 verification. Additionally, he said that the model ignores the Aad construct and that if included Aad would be affected by both attribute and affective processing, which in turn would influence both attitude components. Regarding familiarity, Batra argued that ”the liking phase of the mere exposure effect is not due to conscious (recall—evidenced) learning but to some kind of preattentive recognition.” Overview of Major Theoretical Perspectives on the Static Aspects of the Relationship between Aad and Ab Several studies have attempted to explain the influence of Aad and Ab. Most support a positive relationship between Aad and Ab. There is, however, a seemingly contradictory belief about this positive relationship. Both marketing practice and academic literature suggest that disliked advertisements can be effective and the negative Aad is, in fact, more effective than neutral Aw° Do these two arguments contradict each other? If so, how can the relationship between Aad and Ab be explained? When a negative Aad is formed, should one expect a negative or positive Ab response? In order to answer these questions, several possible theoretical perspectives need to be synthesized. One of the early researchers in the area, Shimp (1981) categorized nine theoretical combinations between Aad and Ab. 10 He divided both Aad and At) into three categories (positive, neutral, and negative) and produced Figure 1, which presents a good overview of possible relationships. It seems that the argument, which supports positive relationship, explains the cells 1 (++) and 9 (-—) in Figure 1. The other argument, which indicates a negativeAad can be effective, explains cell 7 (-+). In a sense, both arguments are too myopic to represent a general explanation. However, with the inclusion of time, the two approaches can be integrated into a general explanation. Because most consumer purchasings occur some time after exposure to advertisements, time is an important variable. Figure 1 Nine theoretical combinations between Aad and Ab Ab + 0 - 1 2 3 + (+1 +) (+1 0) (+1 ') 4 5 6 A3, 0 (o. +) (o, 0) (o. -) 11 In this section, for an overview, theoretical explanations on the static aspects of the relationship between Aad and Ab are reviewed. The static explanations will be extended into dynamic views in the following section. The theoretical perspectives, (a) Classical Conditioning theory, (b) Balance Theory, (c) Mood Research, and (d) Salient Attribute Hypothesis, offer convergent interpretations of how Aad and Ab are related over time. Classical Conditioning Theory Shimp (1981) posited that an attitude toward ad is transferred to an attitude toward brand via a conditioning effect. He expected that the pairing of a connotative response (e.g. a feeling of joy) with a denotative response (e.g., a new brand name) would lead to a conditioned emotional affect toward the advertised brand. Classical conditioning theory (Pavlov, 1927) indicates that a conditioned stimulus (CS) followed by unconditional stimulus (US) will automatically evoked a response, called the unconditioned response (UR). As the result of the pairing of CS and US, one could remove the US and the CS would produce the CR. The classical conditioning process is shown in Figure 2. In an advertising context, when an advertisement (US) creates positive (negative) attitude toward the ad (UR), the advertised brand (CS) can precipitate the same positive (negative) attitude toward the 12 brand (CR) due to classical conditioning. Figure 2 Classical Conditioning Process in Advertising unconditioned conditioned stimulus response ——> advertisement Aad conditioned unconditioned stimulus response brand A b Some researchers are critical on the generalizability of Pavlov's classical conditioning on advertising. McSweeney and Bierley (1984) claimed that the animal laboratory principles may not hold in the more complicated consumer behavior context. Allen and Madden (1985) agree with McSweeney and Bierley's critique and suggest that affective conditioning is a better focus in consumer behavior context. Empirically, however, they found little support for the affective-conditioning hypothesis. On the other hand, several researchers have supported the conditioning effect in advertising conceptually and 13 empirically. Nord and Peter's (1980) behavioral modification theory explains the effects of advertising based on classical/operant conditioning. Gorn (1982) posited that background music could be conditioned with products, colored pens. When consumers were exposed to music they liked 0r disliked while being exposed to a product, their product preferences were influenced. Seventy-nine percent of the subjects selected the product (a pen of a given color) associated with the liked music, while only thirty percent picked the product (a pen of another color) associated with the disliked music. Yet, when the subjects were asked why they chose the specific color pen, no one mentioned music as the reason. Clearly, one might conclude that behaviors might be influenced by factor about which consumers unaware. Kroeber-Riel's (1983) experiment also supports the association between brand and emotion by classical conditioning. He paired a model brand name with emotional pictures (eroticism, social happiness, and exotic reactions) in slide advertisements according to the classical conditioning paradigm. The brand name alone aroused significant emotional reactions and the stronger the emotional scenes the more effective the conditioning that resulted. The direct association between affect and brand attitude was also supported by Bierley, McSweeney, and Vannienwkerk (1985). They manipulated four sets of three colored arbitrary geometric stimuli and the presence/absence 14 of well—liked music. The result revealed that the preference for a stimulus with well-liked music was higher than without the music. Recently, Machleit and Wilson (1988) also provided evidence that conditioning based direct—affect-transfer hypothesis may be an adequate explanation for the effects of emotional feelings andAad on A.b in some situations. They hypothesized that if direct-affect-transfer parallels a conditioning process, correlations between emotional feeling and 'Ab and between Aald and Ab would be significant and would increase with increased exposure to the advertisement. The result showed that only one correlation out of four different brand manipulations exhibited the significance and the antiCipated pattern. They reasoned that for unfamiliar brands, direct—affect—transfer should be expected, but when the brand is familiar it should not be expected. They finally suggested that future research should evaluate brand familiarity as a moderator variable. Shimp (1981) and Lutz (1985) added involvement to the picture. They suggested that direct-affect transfer may occur only in low involvement conditions. Additionally, MacKenzie and Lutz (1989) posited that pure affect transfer would occur in the situation of low ad message involvement and low ad execution involvement. Using four ad exposure situations (high/low ad message involvement x high/low ad execution involvement) they hypothesized the relationship ——i—— ' '5" m.” 57“!" "‘".' ’."t~‘-‘v“.$ - 15 and called for empirical support. ‘qTo summarize, classical conditioning theory can give a good explanation of theAad and.A$ association. Based on the theory, a positive relationship is expected between Aad and Ab. Therefore, it is expected that a positive (negative) Am would be formed when a positive (negative) Aad is developed from advertisements. fiéBalance Theory V Balance theory offers another conceptual perspective to explain the positive relationship between Aad and Ab. Borrowed from Heider (1946), Edell and Burke (1984) attempted to explain the balance among the consumer, the advertisement, and the brand. Balance theory assumes that a person strives to maintain balance among the components of any cognitive unit. Balance theory states that a balance would occur when the relationship between all three pairs is positive or when the relationship between one pair is positive and the other two are negative. Figure 3 represents balance theory in advertising context. Since the relationship between the advertisement and the brand is always positive, the consumer must either like both the advertisement and the brand or dislike them both (Edell and Burke, 1984). In the Same vein, if Aad is positive then Ab must be positive and if Aad is negative then Ab must be negative to maintain the balance. The expected relationship 16 is a reciprocal between Aad and Ab. That is, Ab also can influence A“. The theory has received partial support by Messmer (1979) in the study of unbroadcasted TV ads for well-known products. After measuring A.b before measuring AM, he found that prior Ab favorably influenced A“. In summary, balance theory predicts a positive relationship between Aad and Ab. Based on the theory, the expected relationship is that positive (negative) Aad will cause positive (negative) Ab and vice versa. Figure 3 Balance theory in advertising context advertisement Mood Research Research on the effects of mood states on affective reactions and judgments centers around two different links 17 (Gardner, 1985b). A direct linkage may be explained by associations in memory between mood states and affective reactions. According to Gardner (1985b), the association is due to a conditioned response based on Grifitt and Guay‘s (1969) perspective. They posited that stimuli which possess positive and negative reinforcement properties act as unconditioned stimuli which, in turn, evoke implicit affective responses. Gardner (1985b) views the stimuli as inducers of positive or negative mood states, also an indirect linkage may exist between mood inducers and evaluations. Isen, Shalker, Clark, and Karp (1978) postulated that the effects of positive mood may be mediated by cognitive activity such as information retrieval. According to this theory, advertising effects can be affected by moods evoked by programs. For example, a positive mood created by a television program can increase the accessibility of the positive material stored in memory. Consequently, judgement made on any part of the advertisement or brand evaluations would be biased congruent with a mood state. Consequently, mood research also supports only a positive relationship between Aad and Ab. As such, that research does not provide a rationale as to why some negative Aad can be related to positive Ab. Salient Attribute Hypothesis From a different point of view, Edell and Burke (1984) developed the salient attribute hypothesis. This hypothesis states that an advertisement is a salient attribute of the brand. The logic is based on Fishbein's (1967) attitude formation model. It is reasoned that because most ad research is done in labs and advertisements are an attribute for the brand, there is the high correlation between the attribute (Am) and Ab (Mitchell and Olson, 1981). This hypothesis, however, has not received much attention or empirical support. As Edell and Burke (1984) pointed out, this hypothesis is merely a hypothesis because there is no empirical support. In summary, the four theoretical perspectives described in the section indicate that a general understanding of the relationship between Aad and Ab is not close at hand. Although the static explanations support positive relationship between Aad and Ab, the theories do not provide explanations on how negative Aad can be related to positive Ab. Available research, nevertheless, offers useful insights into what aspects of the Aad and Ab relationship have been studied and supported. The research covers the common dimensions across theories and what aspects of the Aad and A.b relationships have been overlooked. 19 Dynamic Views of the Relationship between Aad and Ab This section reviews what is known about Aad and Ab over time. The longitudinal studies focus on how and why the association between Aad and Ab changes dynamically. Some of the researchers attempted to explain why negative Aad can be related to positive A.b after some time. The theoretical perspectives that will be reviewed are (a) superiority-of—the-pleasant Hypothesis, (b) balance theory, (c) law of extremes hypothesis, (d) mere exposure effect hypothesis, and (d) sleeper effect theory. Superiority—of-the-Pleasant Hypothesis (Generalization Hypothesis) Silk and Vavra's (1974) superiority—of—pleasant hypothesis is based on the classical conditioning mechanism. They postulated that reactions to an advertisement are generalized to the brand by the viewer. Later Shimp (1981) posited that Amiis linearly related to A.b so that the more positive consumers' reactions are to the advertisement, the more positive their reactions to the brand. In order to acquire conditioning effects several implications are suggested by McSweeney and Bierley (1984). 20‘ First, the advertisement must be presented several times to insure that a change in preference occurs. Second, the advertisement which uses a salient conditioned stimulus (CS) and a strong unconditioned stimulus (US) will be the most effective. Third, advertisements which are well—spaced in time should be more effective than ones which are not. Last, although there are no fixed rules about how quickly responses reach and asymptote, there will be some optimal number of presentations. I Conditioned responses will weaken over time due to extinction and/or forgetting (Domjan and Burkhart, 1986, p.64). Extinction and forgetting need to be distinguished. According to Domjan and Burkhart (1986, p.64), extinction is a gradual decline of the conditioned response because of repeated presentations of the CS without US. Forgetting, in contrast, is a decline in the strength of the conditioned response due to the passage of time. Extinction occurs when the CS and US relation is broken, but forgetting takes place with prolonged absence of exposure to the conditioned stimulus. The implications of extinction for consumer research is that conditioning will be most effective for products that are rarely seen outside of the commercials. Extinction, on the other hand, will occur rapidly for frequently seen product when the commercial is discontinued (McSweeney and Bierley, 1984).. I To summarize, conditioning based superiority—of—the- 21 pleasant hypothesis predicts positive relation between Aad and Ab and the association becomes stronger as the number of exposure increases until it reaches ceiling. The relationship weakens due to extinction and/or forgetting over time. Although the intensity of the relationship can be strengthened by increasing the number of ad exposures (reenforcement) or weakened by discontinuing the advertisement (extinction/forgetting), the valence of the relation remains the same. Balance Theory Balance theory predicts that the initial relationship between Aad and Ab will remain over time. For example, once a positive Ab is formed due to a positive A“, then it will be sustained until other sources disturb or break the balance. If too many exposures create tedium, then the evaluation of Aad will be worse and in turn it will decrease Ab to maintain the balance. Edell and Burke (1984) expect the correlation between Aad and A.b will be maximum if the balance is obtained immediately. They also anticipate that if balance is obtained gradually, then the correlation would become more positive over repeated exposure to the advertisement. Over time, the correlation should remain very high even without more exposures unless no events break the balance. 22 Therefore, balance theory predicts a positive relationship between Aad and A.b over time. Furthermore, it is expected that the intensity of the affect or the valence will not be changed unless the balanced state between Aad and Ab is affected by an external event. This prediction, however, does not seem to be acceptable intuitively and conceptually. If the expectation is true, continuous repetition of commercials would be a waste of resources. Also the theory does not specify how quickly one's cognitive structure reaches a balanced state. Law of Extremes H othesis Distinctiveness H othesis Considering a distinction between immediate and delayed effects, Silk and Vavra (1974) have posited a J-shaped relationship between Aad and advertising effectiveness. The assumptions behind the hypothesis are (a) advertisements eliciting strong affective reactions are more involving, (b) brands associated with positive or negative ads are distinctive in memory and consequently increased familiarity with the brand will lead to more favorable attitudes toward the brand, and (c) reactions to the ad and reactions to the brand are separated in memory. Silk and Vavra (1974) predicted that both liked and disliked ads should develop more brand awareness and form more favorable Ab than neutral ads. Liked ads are expected to produce more favorable Ab 23 than that of disliked ads. Three possible explanations tell why a disliked ad can be effective (Aaker and Bruzzone, 1985). First, attention and processing could be stimulated without the negative reaction being transferred directly to the brand. This process assumes that ads eliciting strong affective reactions are more involving (Moore and Hutchinson, 1983). Second, irritation could distract, inhibiting counterarguing and, therefore, enhance persuasion (Aaker and Bruzzone, 1985). Finally, and perhaps the most importantly, brand familiarity created by exposures over time may increase positive brand attitude. Silk and Vavra (1974) argue that Aad can not generalize or condition to Ab if the law of extreme explanation is true. Therefore, they postulate that dissociation between Aa‘d and Ab occurs over time and that leads to a "sleeper” effect. The law-of—extremes hypothesis expects that intensity of affect, regardless of valence has the major role of developing a positive Ab if the reactions improve memory for the brand or facilitate processing of persuasive advertising copy and they are dissociated with Ab. .Although the hypothesis is conceptually appealing, a question arises when the dissociation occurs: Why does the dissociation take place at an arbitrary time all of sudden? Even if the dissociation happens sometime, it is of no use to managers because no one can predict when it will occur. 24 Mere Exposure Effect Hypothesis Zajonc (1968, 1980, 1986) has proposed mere exposure effect that extreme affective reactions facilitate memory for the advertised brand. He posited that the increased brand familiarity, in turn, may increase liking for the brand. Zajonc (1986) attempted to explain the mere exposure effect based on Waynbaum's (1907) vascular theory that expression of emotion may cause an emotional subjective state. Zajonc (1986) argued that if familiar objects induce a facial expression then it is possible that the facial expression can cause a subjective state which leads the person to say ”I like it." Obermiller (1985) also suggested the "exposure without cognition model” and explained that familiarity may create positive feelings. Sleeper Effect Theory In social psychology the sleeper effect has been investigated extensively in understanding persuasion over time. According to Hovland, Lumsdaine, and Sheffield (1949), the sleeper effect describes a phenomenon whereby the impact of a persuasive message increases with the passage of time. According to the hypothesis, with introduction of a discounting cue an immediate attitude change is suppressed. With passage of time, however, the message and the discounting cue become dissociated from one 25 another so that the impact of message—related knowledge increases and attitude change. That is, the association between message and the discounting cue weakens over time and so the discounting cue becomes irrelevant for the evaluation of the message. Many studies demonstrated the sleeper effects (Hovland et al. 1949; Weiss, 1953; Kelman and Hovland, 1953; Johnson and Watkins, 1971). The recent perspectives have developed in two stems. First, Gillig and Greenwald (1974) questioned the validity and replicability of the sleeper effect in their study ”Is It Time To Lay The Sleeper Effect To Rest?" Followed by their study, other researchers (Gruder, Cook, and Hennigan, 1978; Cook and Fray,1978; Cook, Gruder, Hennigan, and Fray, 1979), focused on refining the definition and the operationalization of constructs. Despite criticisms, in the past several years, there has been empirical evidence which supports the sleeper effect. Therefore, the main research issue is no longer whether a sleeper effect exists, but when it is likely to occur (Hannah and Sternthal, 1984; Mazursky and Schul, 1988). According to Cook et al. (1979), four criteria are important to achieve the sleeper effect: (a) a message with a strong initial impact on attitude, (b) a discounting cue that inhibits immediate attitude change, (c) the dissociation of the discounting cue and message over time, and (d) the rapid dissociation of the discounting cue and message so that the message has some residual impact. 26 The other perspective concerns the process underlying the sleeper effect. Hannah and Sternthal (1984) applied the availability valence hypothesis to explain the sleeper effect in contrast to the traditional approach of dissociation of a discounting cue from the message. The availability valence hypothesis states that individuals' attitudinal judgments in response to a persuasive message are determined by the favorableness (valence) of the issue and relevant information available in memory at the time of judgement. Their test results were consistent with the availability—valence predictions. Attitude toward the message advocacy became more favorable with passage of time in the relevance cue conditions, but in the irrelevant cue conditions the passage of time had no effect of on attitudes. Mazursky and Schul (1988) supported the notion that the magnitude of the sleeper effect is influenced by the relative availability if the product information and the discounting cue appeal. The investigation of the sleeper effect has important implications for marketing. The existence of the sleeper effect encourages the consideration of timing of attitude measurement as a relevant factor in predicting purchase decisions. Studies on the persuasion mechanism of advertisement can be enhanced by the sleeper effect explanation. Also, as a separate route from cognitive message processing, the relationship between Aad and Ab can 27 be understood with the sleeper effect over time. In this context, Moore and Hutchinson (1983) have postulated the familiarity-based sleeper hypothesis and the affect—based sleeper hypothesis. Their work is of importance to review in more detail because it is a notable attempt to explain the relationship between Aad and Abihl conjunction with the sleeper effect. The familiarity-based sleeper hypothesis states that if Aacl generalizes to Ab, and if the influence ofAad on Ab is initially strong, but weakens over time, then it is possible that increased brand familiarity resulting from ads that elicit strong affective reactions will eventually lead to increased favorability toward the advertised brand. The hypothesis predicts that the generalization hypothesis will be hold immediately after exposure to an ad, but with passage of time a J-shaped relationship will appear between Aad and Ab. It also predicts that either a negative or a positive ad affect will result in better rememberingof the brand. According to the hypothesis, over time the direct influence of ad affect on Ab should decay and the impact of brand familiarity should increase. On the other hand, Moore and Hutchinson's (1983) affect—based—sleeper hypothesis predicts that if ads eliciting strong positive or negative reactions are attended to more than neutral ads, then one might expect a J-shaped relation between Aad and Ab immediately following ad exposure. The hypothesis also posits that after some delay 28 period, the relation between Aad and A.b will be positive and linear because reactions to the ad and to the brand may become confused in memory. It is assumed that the influence of product attributes and brand familiarity on Ab is initially more important than ad affect in contrast to the familiarity—based sleeper hypothesis. In order to test the hypotheses, Moore and Hutchinson (1983) manipulated forty—six subjects with eighty print ads for existing products in two-stage experiments. Because the products existed in the market, the major dependent variable was measured in terms of change in brand knowledge and brand consideration from base line measured in the Experiment 1. From an analysis of variance, they found: (a) immediately following exposure to the ads, brand consideration and-ad affect were linearly related, (b) after a delay period, a J—shaped relationship between ad affect and change in brand consideration developed, and (c)there is marginal support for the familiarity—based sleeper hypothesis because of the fact that only the seven—day delay group showed the expected U—shaped relation. The study supports the familiarity—based sleeper hypothesis only marginally, and it is subject to a few caveats as well. First, the hypothesis states that "over time the direct influence of ad affect on Ab should decay and the impact of brand familiarity should increase (p 527)." This statement appears to be contradictory to their 29 critical assumption that ”ad affect is forgotten more readily than brand familiarity (p 530)." If brand familiarity is forgotten as the assumption implies, the impact of brand familiarity cannot increase but will decrease with the passage of time. If the assumption is accepted, it would be a fair statement that over time both the direct influence of ad affect on Ab and the impact of brand familiarity should decrease, however, after a period of time the familiarity impact will outweigh the ad affect influence because brand familiarity is forgotten more slowly than ad affect. Second, the above assumption lacks theoretical support. Intuitively the assumption makes sense but more rigorous theoretical and empirical supports are needed. Third, they have used ads for existing products as stimuli. It is possible that initial brand familiarity might have a critical role whether ad affect would have a significant impact on Ab. According to Machleit and Wilson (1988), the level of brand familiarity may moderate the Aad and Apzrelationship. That is, for unfamiliar brands the Aad effect on A1) is significant but for familiar brand it is not. Therefore, if ads for unfamiliar brands were used as stimuli, Moore and Hutchinson's (1983) experiment results would be clearer. Fourth, as they admitted, ad affect was measured rather than manipulated. Consequently, the findings are not causal but correlational. A research with ad manipulation is needed to clarify the causal 30 relationship. Fifth, they assumed that brand consideration reflects attitude toward brand and that brand knowledge is a proxy to brand familiarity. As they indicated (p. 530), brand consideration might be weakly related to real purchase behavior and brand knowledge may not be a good measure of brand familiarity. Rather than brand knowledge, brand name recognition or brand name recall would be a better measure of the brand familiarity because in the sleeper effect, the brand familiarity is more related to the initial memorial response to the brand name and not to knowledge about the brand, per se. Finally, the study underscored the role of memory in determining Ab (Moore and Hutchinson, 1983, p. 530). They suggested that future research needs to include memory effect, particularly the relative forgetting curves for various types of information that influence Ab. In the context, the different durations of affect influence and familiarity effect need to be investigated. To summarize, the sleeper effect theory suggests another valid perspective in attempting to understand the relation between Aad and Ab. Particularly, the modified sleeper effect hypothesis (i.e. the familiarity-based— sleeper hypothesis) gives an insightful approach to explain the Aad and Ab relationship in separation of brand cognition (Cb). It can be derived from the above discussion that the main dimensions in the hypothesis consist of the valence and intensity of affect and familiarity of brand name. 31 Other Relevant Theories on the Role of Affect In discussing why affective advertising may be more effective, Ray and Batra (1983) posited the following four reasons: (a) people may pay more attention to affective advertising, (b) affect may enhance the degree of processing, (c) affective executions may lead to more positive judgments of the advertised message, and (d) affective executions may be remembered better. Some of the reasons are reviewed extensively in this section. Attention and Affect For an ad to have an effect, it must draw attention. If an ad gets no attention, it may not have any effect in this overflow of information era. Affective advertising should be more effective in getting such attention (Ray, 1977). Ittelson (1973) also argues that the first level of response to the environment is affective. Emotions are perceived more readily than others (Broadbent, 1977). Bower and Cohen (1982) also assert a similar opinion that a person's feelings act like a selective filter. It, therefore, appears that both positive and negative affects get more attention than neutral and boring advertisements. For example, a humorous ad such as Bud Light beer gets more 32 attention than other dull ads. Also, ads with fear appeal (i.e. ad against drunken drive) create more attention. Perhaps an ad with a negative affect may create more attention than that with positive affect, because negative attributes tend to remain highly distinctive (Srull, 1983) and are relatively scarce whereas either positive or neutral ads are generally abundant. Degree of Processing and Affect Affect may enhance the degree of processing. Ray (1973, 1982) asserts that affective executions facilitate learning. He, therefore, suggests affective executions should be used in the situations where the natural drive is low. Kroeber—Riel (1979) also argues that the degree of information processing for a message is function of the degree of the message evokes arousal or phasic arousal. Memory and Affect Affect may facilitate memory. According to Dutta and Kanungo (1975), intensity of affect, regardless of valence, results in better memory. Particularly, affective executions can be remembered better over time than immediately after the exposure (Silk and Vavra, 1974). Srull (1983) examined the intensity hypothesis in his 33 experiment. The result reasonably strongly supported the finding that more intense moods, regardless of valence, facilitate better overall levels of recall for each of the manipulated ads. The aroused mood state by the intensity of affect leads to better overall levels of recall. He also found that initial evaluations are strongly influenced by subjective mood states at the time of information acquisition. The results reveal that positive (negative) mood states lead to more (less) favorable evaluations than neutral mood states. The research also found interaction between the subject's mood state at the time of retrieval and the type of information provided. The evidence shows the contrast effect that people remember better when the mood state at time of retrieval is in consistent with the valence of attributes. He tried to find the reason in the cue overload theory (Watkins and Watkins, 1975, 1979; Watkins, 1979). When a person is in positive mood, the retrieval of positive attribute is possible because the mood serves as a cue. Since the cue is associated with nearly all other positive events in the person's life, however, the cue becomes overloaded and less effective. It is known that the more things are associated with a cue, the less effective the cue becomes ( Watkins, 1979; Mueller and Watkins, 1977; Watkins and Watkins, 1975). Based on the same logic, negative attributes can be recalled better in the neutral or positive mood states, since 34 negative attributes will remain highly distinctive (Srull, 1983). In the same vein, if a brand name is encoded in association with negative (positive) Aad then the brand name can be recalled when the retrieval mood is positive (negative). CHAPTER THREE HYPOTHESES AND METHODOLOGY This chapter discusses hypotheses and methodology. It begins with the identification of primary dimensions in the relationship between and Aad and Ab to integrate the theories. Then, derived hypotheses are discussed. The following sections provide definitions of constructs, a description of research design and subjects, development of stimuli and pretesting, and measurements of constructs. Primary Dimensions in the Relationship Between Aad and Ab: Integration of the Current Theories The theories and hypotheses thus far reviewed in Chapter Two contribute to understand the relationship between Aad and Ab over time. They have, however, limited power for explaining the relationship because each theory or hypothesis has been developed in different bodies of literature and in different domains, so that the explanation is more sectional than comprehensive. Therefore, an integration of the various limited but appealing theories is 35 36 desired to give a comprehensive explanation of the relation over time. In this section, the primary dimensions are identified and formal hypotheses are generated in the following section. Given the review of the literature in the previous chapter, three general dimensions seem to be important to explain and predict A.b based on A“. First, valence of affect seems to have a great influence on Ab. The valence can be either positive, negative, or neutral. All of the reviewed theories consider the valence of Aad as an important cause on the direction of Ab. Second, the intensity of affect appears to be another important factor. The theories imply that more intense affect results in stronger Ab. Finally, familiarity is considered as another important factor in addition to the above two factors that are related to Aw~ Familiarity has been considered a primary factor in the mere exposure effect hypothesis (Zajonc, 1968; Harrison, 1977; Sawyer, 1981; Zajonc and Markus, 1982) and in the familiarity based sleeper effect (Moore and Hutchinson, 1983). Here familiarity refers to brand familiarity and how much the brand name can be recognized (or recalled). With the identified primary dimensions, a hypothetical affective information processing model is developed based the integration of the theories. Figure 4 represents the model. 37 Figure 4 Hypothesized Affective Information Processing Model Hypotheses When people watch an advertisement, they develop an attitude toward the advertisement. In classical conditioning context, when an unconditioned stimulus is given, a conditioned response is developed. Hypothesis 1 focuses on the unconditioned response in the context of advertising. In advertising, the level of an advertisement is an unconditioned stimulus and attitude toward the advertisement is a unconditioned response. The relationship between these two constructs is expected to be positive. That is, as the level of advertisement increases, attitude toward the advertisement increases. Figure 5 38 represents the relationship. H1: There is a positive relationship between the level of the advertisement (Lu) and the attitude toward the advertisement (Am). Figure 5 Hypothesis 1 / ad Hypothesis set 2 posits a relationship between attitude toward the advertisement (Am) and attitude toward the brand (Ab), the effect of brand familiarity on Ab, and joint effect of the Aad and BF on Ab. As reviewed in Chapter Two, classical conditioning theory, balance theory, and mood research expect a positive relation between Aad and Ab. In classical conditioning theory an Aad is considered to be an unconditioned response and Ab is a conditioned response. Many studies support applications of the theory in advertising (Shimp, 1981; Nord and Peter, 1980; Gorn, 1982; Kroeber—Riel, 1983; 39 Bierley et al., 1985; Machleit and Wilson, 1988). Balance theory also leads one to expect a positive relationship between Aad and Ab. According to balance theory, since the relationship between advertisement and the brand is always positive, the consumer must either like both the ad and brand or dislike them both. Edell and Burke (1984) and Messmer (1979) support the idea. Mood research generally supports the positive relationship (Gardner, 1985; Isen et al., 1978). Therefore, Hypothesis 2a is derived from the theories and is shown in figure 6. H2a: There is a positive relationship between the attitude toward the advertisement (Am) and the attitude toward the brand (Ab). Figure 6 Hypothesis 2a / ad Hypothesis 2b was established based on mere exposure effect (Zajonc, 1968, 1980, 1986; Obermiller, 1985), 40 familiarity based sleeper effect (Moore and Hutchinson, 1983), and the law of extremes hypothesis (Silk and Vavra, 1974). The theories predict that increased familiarity will lead to more favorable attitude toward the brand. Figure 7 shows the relationship. H2b: There is a positive relationship between brand familiarity (BF) and the attitude toward the brand (Ab). Figure 7 Hypothesis 2b / Hypothesis Zc is the combination of hypothesis 2a and BF 2b. It is based on familiarity-based sleeper effect (Moore and Hutchinson, 1983) H2c: Brand familiarity (BF) and the attitude toward the advertisement (Am) are jointly positively related to the attitude toward the brand (Ab). 41 Either a positive or negative advertisement draws more attention than a neutral advertisement (Ray, 1977; Ittelson, 1973; Broadbent, 1977; Bower and Cohen, 1982). Affect also may enhance the degree of information processing (Ray, 1973, 1982; Kroeber—Riel, 1979). Dutta and Kanungo (1975), furthermore, posited the intensity hypothesis; intensity of affect, regardless of valence, results in better memory. Other researchers (Silk and Vavra, 1974; Srull, 1983), in addition, suggested that intensity of affect facilitates memory. Hence, hypothesis 3a and 3b are derived. Figure 8 represent Hypothesis 3a and 3b. H3a: There is a positive relationship between brand familiarity (BF) and the intensity of an advertisement (Iad) ° H3b: There is a positive relationship between the intensity of the attitude toward the advertisement (IAN) and the intensity of the advertisement (1“): 42 Figure 8 Hypothesis 3a & 3b BF , IAad / ad Hypothesis 3c and 3d are established based on cue overload theory (Watkins, 1975, 1979; Mueller and Watkins, 1977; Wakins and Wakins, 1975) and distinctiveness of a negative attribute in advertisements (Srull, 1983). The theories suggest that a negative affect may be more effective in memory and perceived more intensely than a neutral or positive affect. Figure 9 shows Hypothesis 3c and 3d. H3c: At a given level of intensity of an advertisement, brand familiarity (BF) is higher for advertisements with a negative valence than for those with a positive valence. 43 H3d: At a given level of intensity of an advertisement (Lu), the intensity of the attitude toward the advertisement (IAw) is higher for advertisements with a negative valence than for those with a positive valence. Figure 9 Hypothesis 3c & 3d BF, IAd a negative positive / In Perceived intensity of the attitude toward the advertisement (Am) and brand familiarity (BF) are expected to be weaken over time due to forgetting and extinction. The following hypotheses H4a and H4b attempt to examine the decrease for the given period of time (10 days). Figure 10 shows the possible decrease. H4a: Intensity of the attitude toward the advertisement (IAN) decreases over time. 44 H4b: Brand familiarity (BF) decreases over time. Figure 10 Hypothesis 4a and 4b IA“ , BF \\\\‘\\\\‘F“~l time The following sections provide definitions of constructs, a description of research design and subjects, development of stimuli and pretesting, and measurement of constructs. Definitions of Constructs The definitions of constructs in this study are as follows: Attitude: a favorable or unfavorable predisposition toward an object, including an evaluative component and an affective component 45 Attitude toward the Advertisement (Am): an attitude toward a particular advertising stimulus during a particular exposure occasion Level of Advertisement (Lu): perceived levels of advertisement categorized as five levels (strongly positive, weakly positive, neutral, weakly negative, and strongly negative) Valence of Advertisement (Va): The direction (positive, neutral, or negative) of the level of advertisement Intensity of Advertisement (In): the strength of the level of advertisement Valence of Attitude toward the Advertisement (VAN): the direction (positive, neutral, or negative) of attitude toward the advertisement (Am) Intensity of Attitude toward Advertisement (IAw): the strength of attitude toward the advertisement (Am) Attitude toward the brand (Ab): an attitude toward a particular brand Brand Familiarity (BF): the degree of recognizing or 46 recalling a particular brand name Purchase Intention (PI): a person's likelihood of purchasing a particular brand Research Design and Subjects In this research two factors were employed. The first manipulation factor was the levels of advertisement. The levels were manipulated with five combinations of the valence and the intensity of the advertisement (strongly positive (++), weakly positive (+), neutral (0), weakly negative (—), and strongly negative (~—)). The five different levels were manipulated by the background music and the pictures used in the advertisements. The experimental design included three different times of measurement (immediately after advertisement exposure, four days later, and ten days later). All of the dependent variables (BF, AMI and Ab) were measured three different times. In sum, the design of the experiment was a 5 x 3 between—subject design with five levels of advertisement (++, +, 0, —, ——) and three different times of measurement (immediately after advertisement exposure, four days later, and ten days later). A within—subject design was initially considered it 47 because, generally, it is more powerful and requires fewer subjects than an equivalent between—subject design. The within—subject design, however, was eliminated due to possible carryover effect. Carryover effects occur when a previous treatment alters the observed behavior in a subsequent treatment and pose a serious threat to the internal validity of the experiment (Bordens and Abbott, 1988, p.228). In the current experiment, the carryover effect could be a serious problem because Aad immediately after ad exposure may affect Aad measured four days later and Aad ten days later in turn. Development of stimuli and pretesting The product advertised in the advertisement was desired to be low involvement consumer product with an unfamiliar brand name. A chewing gum was selected as the product category because, generally, it can be considered a not important, less risky, low priced product. The brand name had to be unfamiliar and so a creation of a fictitious brand name was desirable. The new name should not have any meanings or connotations because the attitude toward the brand (Ab) can be affected by the meaning of the brand name. An unknown brand name was necessary because brand name familiarity would be measured. In the selection of the most desirable brand name, 20 alternatives were initially 48 considered and compared. Out of the 20 alternatives, the name Chowa was selected after discussion with several native English—speaking Americans. The consensus was that Chowa does not have any special meaning and sounds fine as a chewing gum brand name. The advertisements were made of three elements-brand name, picture, and background music. There was no additional message in order to control Fishbeininan cognitive effects. The types of picture and background music were expected to generate different levels of advertisement. In selecting stimuli, pretesting was necessary to get a successful manipulation. In order to decide background music, 60 junior or senior university students were selected and asked to write one or two songs and/or any type of music that they found the best, the most likeable, non—irritating, interesting, favorable, enjoyable, and pleasant. Since the answers were varied, a professional music reviewer was consulted to reduce the number of alternative kinds of music that would be used in the music pretest. The reviewer also made lists of music considered to be moderately good, neutral, and music that would be perceived by college junior and seniors as the worst (the least likeable, irritating, boring, unfavorable, unenjoyable, and unpleasant). Based on the lists, 23 musical pieces were selected for pretesting. The most representative part of each musical piece was 49 recorded for 30 seconds. Thirty—five students listened to and rated 23 musical pieces right after they listen to each selection. The scale used had a 7—point semantic differential between 7 adjective pairs (bad—good, like— dislike, irritating—not irritating, interesting—not interesting, favorable—unfavorable, enjoyable—unenjoyable, pleasant-unpleasant). Based on the average scores of the scale, the musical selections were categorized from strongly negative to strongly positive. For picture selection, the researcher collected 25 possible pictures for chewing gum advertisements with consideration given to five different levels of advertisement (from very negative to very positive). After consulting a professional artist and other experts, three pictures were selected for each of five categories. The Chowa chewing gum's logo and package were designed by a professional artist and approved by the researcher. The profeSsional artist added the package design to each of the 15 pictures, that is, the picture part of each advertisement consisted of a Chowa chewing gum package and a background picture. The final advertisement pictures were transformed into slides by a professional photographer. The 15 slides were matched with the music by considering the types of pictures and music. Therefore, three final alternative advertisements for each of the five levels of advertisement were identified for use in 50 advertisement pretest. In the pretest of the advertisements, a total of 76 junior or senior university students participated. The pretest was conducted in three different sessions. In the first session, 22 students participated; 25 students participated in the second session, and 29 students in the third session. In each session the research let subjects watch five advertisement consecutively without rating them in order for them to develop consistent judgmental level. In each session, the order of the advertisements were randomly altered to prevent any order biases. In each session, after watching each advertisement slide with a background music, the students answered a question to identify their perceptions of the advertisements: strongly negative (—2), weakly negative (— 1), neutral (0), weakly positive (+1), or strongly positive (+2). Then, the mean values of the fifteen advertisements were compared. Finally five advertisements were identified and selected to represent the five levels of advertisements in the experiment. The five advertisements selected are summarized in Table 1. Next, the mean difference between adjacent categories were compared and the means were found to be different. The mean of strongly positive advertisement is significantly higher than that of weakly positive advertisement (p < .01); the mean of weakly positive advertisement is significantly 51 Table 1 Selected Advertisements Level Mean Score Description (Std. Dev.) Strongly positive 1.63 Picture: Three young people ( .67) are sledding with big smiles Background music: ”In The Air Tonight” by Phil Collins Weakly positive .64 Picture: A young woman is (1.22) blowing bubble gum Background music: ”If Only Love" by Liona Boyd (Guitar Music) Neutral —.03 Picture: A package of Chowa (1.17) in a briefcase with a Wall Street Journal, a baby toy, etc. Background music: "Tara's Theme” by Liona Boyd (Guitar Music) Weakly negative -.30 Picture: Eight men are (1.39) swimming in a pool Background music: ”Metal Storm Face the Slayer” by Slayer Strongly negative —.71 Picture: A fat woman (made (1.16) of play—dough) is holding a package of Chowa Background music: Traditional Indian music higher than that of neutral advertisement (p < .01); the mean of neutral advertisement is marginally higher than that of weakly negative advertisement (p < .10); the mean of 52 weakly negative is significantly higher than that of strongly negative advertisement (p < .05). Consequently, the five selected advertisements were concluded to be reasonably good stimuli for further experiment. Experimental Procedure The experiment was carried out in three different stages in ten days. In the first stage, the subjects were divided into five different groups based on five levels of advertisement (strongly positive (++), weakly positive (+), neutral (0), weakly negative (—), and strongly negative (-—)). Each group watched a slide show "Getting Ready To Ride" in which three times of an advertisement were embedded. The slide show "Getting Ready To Ride” is an instructive program that tells viewers how to find the right bicycle, how to ride, and how to maneuver in emergency. The slide program is considered to be a nonemotional and neutral. Total running time of the slide show was 11 minutes and 12 seconds. The placement of the advertisements in the slide show is shown in the Table 2. Placement of Advertisements 53 Table 2 Time Sequence 00:00 Getting Ready To Ride 02:35 Advertisement 03:05 Getting Ready To Ride 06:22 Advertisement 06:52 Getting Ready To Ride 09:22 Advertisement 09:52 Getting Ready To Ride 11:12 End No explanation about the purpose of the experiment was given to the subjects before they watched the slide show in order to ensure the watching environment as natural as possible. After watching the show, only one third of subjects in each group remained to answer the questionnaire. In the second stage (four days later), each group answered the same questionnaire. stage (ten days later), group answered the questionnaire. another one third of In the third the other one third of the each 54 Subjects The subjects were students in a large midwestern university who received extra credit in a marketing class for answering the questionnaires. After all data were collected, the subjects were debriefed. Using relatively homogeneous student subjects gave the benefit of easy control. The subjects had to sign up in advance for the best option out of the possible three dates for answering the questionnaire. Reminder calls were made to each student to ensure the attendance in the second and third stages. As the result, a total of 254 students participated in the experiment. Due to the difficulty of controlling many class sections and absences, however, the cell sizes were turned out to be unequal as shown in the Table 3. If cell sizes are different, unequal weight will be assigned to the experimental conditions in the statistical analysis and the effect of violating the distribution assumptions of the analysis of variance (i.e., normality and homogeneity of variance) would be increased. Each cell size was reduced to 11 because that was the smallest obtained cell size in the experiment. Even though the cell size was reduced, the degrees of freedom in the error term was as large as 150. The power in the analysis of variance would not decrease much from that with the degrees of freedom of 241, because 55 there is no big improvement of the power after the degrees of freedom exceeds 100 in the F distribution. In fact, the power of the analysis variance was well above 90 percent, assuming the effect size was between medium and large at the a level of .05 (Cohen, 1977). In 13 cells, the desired numbers of subjects, except one subject, were eliminated randomly. One subject was initially eliminated from the sample because he or she answered that he or she has heard the brand name Chowa once before the experiment. Presumably the subject had participated in the pretest. The result was that 11 subjects in each of the 15 cells made a total of 156 subjects. The mean age of the subjects was 21.4 and ranged from 20 to 30 years of age. Most, or 90.9 percent, of the subjects were aged between 20 and 22. Most of the subjects (71.5 percent) were seniors and 28.5 percent were juniors. Among the subjects, 52.1 percent were male and 47.9 percent were female. 56 Table 3 Cell sizes Time Immediately 4 Days 10 Days after later later ++ 11 (22) 11 (16) 11 (18) Level of + 11 (21) 11 (14) 11 (19) Ad 0 11 (15) 11 (18) 11 (11) — 11 (21) 11 (16) 11 (12) -- 11 (16) 11 (24) 11 (11) * Initial cell sizes are in the parenthesis Measurement The constructs in the experiment and their operationalizations are in the following section. Attitude toward the Advertisement (Aw) The attitude toward the advertisement (Aw) was measured by the mean scores obtained from seven-point semantic differential items of bad-good, dislike—like, irritating—not irritating, interesting-boring, favorable- 57 unfavorable, enjoyable—unenjoyable, unpleasant—pleasant. These scales are similar and consistent with the literature (Mitchell and Olson, 1981; Mitchell, 1983; Gardner, 1985; MacKenzie, Lutz and Belch, 1986; Machleit and Wilson, 1988). The items randomly altered the direction of the adjective continua following a common practice. This convention is intended to overcome response sets that could adversely affect the validity of the instrument (Mueller, 1986, p.52). Valence of Attitude toward the Advertisement (VAN) The valence of attitude toward the advertisement (VAw) was measured by the sign of the Aad measurement. Since the adjectives in the semantic differential in each item were opposite, the middle point was considered to be neutral, and the other two parts were considered as either positive or negative according to the direction of the adjectives. Intensity of Attitude toward the Advertisement (IAN) The intensity of attitude toward the advertisement (IAw) was measured by the strength of the Aad regardless of the valence. The larger the absolute value of the A“, the more intense the Aad is. Attitude toward the Brand (Ab) The attitude toward the brand (Am) was operationalized on seven—point semantic differential scales 58 of good—bad, dislike—like, favorable—unfavorable, enjoyable— unenjoyable, useful-useless, low quality-high quality, beneficial—harmful. The directions of item adjectives were randomly altered and scores will be averaged. Brand Familiarity (BF) Brand name familiarity (BF) was measured by both recognition and recall of the brand name. In measuring the brand name recognition, a six-point confidence rating procedure was employed. Assuming familiarity lies along a continuum, the procedure divides the familiar—unfamiliar range into six intervals, with one extreme indicating "definitely familiar" and the other indicating "definitely unfamiliar." The confidence rating procedure obtains more information about the certainty of the subject's judgment than that can be obtained from a simple familiar-unfamiliar scale. Generally even number of scales are used, permitting all responses to be divided into one of the two familiar- unfamiliar categories if necessary (Snodgrass, Levy—Berger, and Haydon, 1985, pp 194-195). In measuring brand name recall, unaided recall was used. The question was ”You have seen a few times of an advertisement during the slide show (Getting Ready To Ride). Please recall the brand name that appeared in the commercial.” 59 Purchase Intention (PI) Purchase intention was measured by the likelihood of purchasing the product in the future. The likelihood was operationalized on seven-point semantic differentials of improbable-probable, likely—unlikely, and possible- impossible. Then, the three scores were averaged. CHAPTER FOUR RESULTS The results of the experiment are presented in this chapter. The chapter begins with reliability measurements followed by a manipulation check, and concludes with tests of hypotheses. Reliability Measurements In this research, multi—item scales were used to measure the attitude toward the advertisement (Am)! the attitude toward brand (Ab), and purchase intentions (PI). Cronbach's coefficient a was used to analyze reliabilities of the two measurements. Cronbach's coefficient a measures internal consistency of multi—item scale. Cronbach's coefficient a for the measurement of the attitude toward the brand (Ab) was .9072. and all of the seven semantic differentials were significant good measurements according to sensitivity analyses. Also, all seven items were significantly good in the attitude toward the advertisement (Am) measurement and the Cronbach's coefficient a for Aad was turned out to be as high as .9398. 60 61 The purchase intention (PI) scale also revealed a high coefficient a of .9363. Since the Cronbach's coefficient a's are high, all of A“, Ab, and PI measurements are internally consistent and, therefore, reliable for further analyses. Manipulation Check The brand name Chowa, a fictitious brand name for a chewing gum, initially should not have been familiar to the subjects because the brand name was newly created. Using an initially unknown brand name was important in this experiment because brand name familiarity was measured at each time of the measurement. All except one subject answered that they never heard the brand name Chowa before the experiment. Only one subject indicated that he or she heard it only once before. It is assumed that he or she had participated in the pretest and the subject was eliminated from the analysis. Therefore, the brand name Chowa was initially not familiar to any of the participated subjects whose answers were analyzed. The subjects were supposed to have no idea about the purpose of the experiment because no instruction about the purpose was given and, therefore, they had to pay attention to neither the program nor the advertisement alone. The subjects were asked on a seven-point scale in the 62 questionnaire if they had paid more attention to the program or more to the advertisement during the slide show. The mean of the answers was .127; the mode was 0; the median was 0. The percentage of each answer category was ideally distributed as shown in the Table 4. Table 4 Percentage of Each Attention Category Label category percent more to the program —3 13.3 —2 14.5 —1 9.7 0 17.6 1 15.2 2 12.7 more to the advertisement 3 17.0 Tests of Hypotheses This section explains the test results of the hypotheses that were derived in chapter three. New It was hypothesized that there is a positive relationship between the level of the advertisement (Lu) and the attitude toward the advertisement (Aw)- The 63 relationship was investigated with the Pearson correlation coefficient between the two constructs. The correlation for the all of the data was computed as .70 and was significantly greater than zero ( p < .01 ). The significantly positive correlation indicates that there is a positive relation between the level of advertisement (LM) and A“. A positive relationship between Lad and Aad means that as the level of the advertisement (LN) increases, the attitude toward the advertisement (Am) increases. The correlation between the level of the advertisement (Lu) and the attitude toward the advertisement (Aw) was also calculated for the three different times of measurement and the three correlations were compared. The computed correlations were summarized in the Table 5. Right after the advertisement exposure, the correlation between Lad and Aad was .75 (p < .01); four days later the correlation was .72 (p < .01); 10 days later the correlation was .59 (p < .01). All three correlations were significantly greater than zero but the correlations decreased with the passage of the time. As the Table 6 indicates, the correlation measured four days later (r = .75) decreased from that measured right after the advertisement exposure (I = .72) even if the decrease was not statistically significant. The correlation after ten days (r = .72) decreased significantly from that after four days (r = .59). The tests revealed that (a) there was a directional decrease from the 64 correlation right after ad exposure to the correlation after Table 5 Correlation between Lad and Aad time correlation p (one tail) overall .70 < .01 right after ad exposure .75 < .01 four days later .72 < .01 ten days later‘ .59 < .01 Table 6 Differences among Correlations between Lad and Aad Correlations *z-value p (two tail) Right after ad exposure and four days later .33* .94 Four days later and ten days later 8.11* < .01 1 + Vm 1 + Vf 1/2 In — 1/2 1n 1 - Vm 1 - Vf * z = 1 1 —_____ + ____.___ Nm — 3 Nf - 3 correlation coefficient sample size ,where Vm, Vf Nm, Nf 65 four days, and (b) there was a significant decrease between the correlations after four days later and ten days later (p < .01). Consequently, these figures indicate that the correlation between Lad and Aad weakens over time. In sum, the test results suggest a positive relationship between Lad and Aad and the relationship tends to decrease with the passage of time. Hypothesis 2a Hypothesis 2a holds that there is a positive relationship between the attitude toward the advertisement (AN) and the attitude toward the brand (Ab). A simple linear regression between Ab and Aad for overall data revealed the B coefficient of Aad (.596003) was statistically positively significant (p < .01). The significantly positive 8 coefficient indicates a positive relationship betweenAad and Ab. This relationship means the generalization between.Am and Ap exists for the given period of time (10 days) The relationship was further investigated at the time of each of the measurements. The regression models for each time period were tested and the results are shown in Table 7. All of the t-tests indicate that the Aad is a positively significant variable for predicting Ab. Consequently, the significant results indicate that attitude toward the 66 advertisement has a positive relationship with attitude toward the brand, and that this relationship was maintained for the given period of time (10 days). Table 7 Regression Models between Ab and Aad time regression * p R2 overall Y = .041165 + .596003 X1 < .01 .526 right after ad exposure Y = .134942 + .626987 X1 < .01 .541 four days later Y = —.111350 + .658253 X1 < .01 .645 ten days later Y = .052497 + .457587 X1 < .01 .361 where X1 attitude toward the advertisement (Am) Y = attitude toward the brand (Ab) * one tail test (H0: B1 <= 0) Hypothesis 2b Hypothesis 2b predicts that there is a positive relationship between brand familiarity (BF) and the attitude toward the brand (Ab). In order to test the hypothesis, a sets of regression models for each time period as well as overall data were tested. Furthermore, in order to see if there are any differences between brand name recognition 67 measurement and brand name recall measurement, two sets of separate regression models were built for the two different independent variables. The regression models were shown in Table 8 and Table 9. As the p—values indicate, none of the fin coefficients were significantly positive. Only the test for the regression model right after advertisement exposure marginally supported a positive relationship between Ab and brand name recall. Furthermore, none of the 3% coefficients was different from zero. Therefore, it is concluded that no evidence indicates either brand name recognition or unaided brand name recall has positive relationship with the attitude toward the brand. In other words, when brand name recognition or recall is the only independent variable to predict Ab, neither is a significant variable. 68 Table 8 Regressions Models between.A$ and Brand Name Recognition time regression p overall Y = -.140090 — .022747 X1 .60* .79** right after ad exposure Y = .254240 - .105352 X1 .71* .59** four days later Y = —.315573 — .016908 X1 .55* .90** ten days later Y = -.152916 — .021255 X1 .57* .87** where X5 = brand name recognition Y = attitude toward the brand (Ab) * one—tail test (H0:IL <= 0) ** two-tail test (H0:IL = 0) Table 9 Regression Models between A%.and Brand Name Recall time regression p overall Y = -.204633 + .011541 X1 .48* .96** right after ad exposure Y = -.326531 + .477791 X1 .08* .16** four days later Y = —.232804 — .241686 X1 .79* .43** ten days later Y = -.076923 - .248200 X1 .86* .28** where X1== brand name recall Y = attitude toward the brand (Ab) * one-tail test (Hozlh <= 0) ** two—tail test (H0:IL = 0) 69 Hypothesis 2c Hypothesis 2c explains that brand familiarity (BF) and the attitude toward the advertisement (Am) are jointly positively related to the attitude toward the brand (Ab). The hypothesis was tested with multiple linear regressions in which Aad and BF are independent variables and Ab is the dependent variable. Two sets of regression models for the brand name recognition and recall were also developed. The regression models were shown in Table 10 and Table 11. In all of the eight regression models in Table 10 and Table 11, the attitude toward the advertisement was a highly significant independent variable and neither brand name recognition nor recall was significant. This finding only partially supports hypothesis 2c that predicts that both of the independent variables are significant. As predicted, however, the test indicates that the attitude toward the advertisement has a strongly positive relationship with the attitude toward the brand at any time during the experiment. This is because the attitude toward the advertisement in each regression model at any given time of measurement was strongly significant. All eight tests gave p < .001. This result, therefore, strongly confirms the generalization by classical conditioning between attitude toward the advertisement and attitude toward the brand. Furthermore, this relationship was maintained over a limited time. may—7a.. , 70 Table 10 Regressions for Brand Name Recognition time regression overall Y = .037626 +.596033 x -+ .001388 x2 * (.044487) (.060041) 2 ** (<.001) (.491) R = .526 R2 = .520 Standard Error = .74161 n = 165 right after ad exposure Y = .493502 + .629100 X1-— .132148 X2 (.079275) (.129388) 2 (<.001) (.845) It = .550 R2 = .533 Standard Error = .83169 n = 55 ’ four days later Y = -.240909 + .662156 X1-+ .055004 X2 (.067705) (.083759) 2 (<.001) (.514) R = .648 R2 = .634 Standard Error = .67259 n = 55 ten days later Y = .066148 + .457465 X1-— .005314 X2 (.084372) (.104019) 2 (<.001) (.502) R == .361 R2 = .337 Standard Error = .68485 n = 55 where X5 = Attitude toward the advertisement (Amfi x: = brand name recognition 2 Y = Attitude toward the Brand (Ab) * (standard error of the coefficient) ** (p (Ho: 3 <= 0)) 71 Table 11 Regressions for Brand Name Recall time regression overall Y = .043901 + .596023 X1-+ -.004947 X2 * (.044470) (.116093) 2 ** (<.001) (.502) R = .526 R2 = .520 Standard Error = .74161 n = 165 right after ad exposure Y = .043313 +.617806 X1-+ .144289 x2 (.081160) (.236398) 2 (<.001) (.272) R == .545 R2 = .527 Standard Error = .83700 n = 55 four days later Y = —.138277 + .661691 x1-+ .055400 x2 (.068626) (.184596) 2 (<.001) (.765) R = .646 R2 = .632 Standard Error = .67478 n = 55 ten days later Y = .176295 + .456294 X1- .236183 X2 (.083014) (.182060) 2 (<.001) (.900) R. = .381 R2 = .358 Standard Error = .67404 n = 55 where X1 .Attitude toward the Advertisement (Amfi X2 brand name recall Y = Attitude toward the Brand (Ab) * (standard error of the coefficient) ** (p (no: 8 <= 0)) 72 Unexpectedly, however, none of eight tests for the brand familiarity indicate a positive relationship between the brand familiarity (BF) and the attitude toward the brand (Ab). The relationship between BF and A.b was further analyzed by an analysis of covariance (ANCOVA). In the ANCOVA, the dependent variable (Ab) was measured by the independent variable (the brand name recognition), with Aad as a covariate. The result of the ANCOVA indicated that the brand name recognition was not a significant variable. Table 12 shows the ANCOVA result. From the test results of regression models and ANCOVA, it was concluded that there is no evidence to indicate a positive relationship between BF and Ab. Table 12 ANCOVA: Brand Name Recognition Source SS DF MS F p within cells 88.92 160 .56 regression 98.01 1 98.01 176.35 .000 Recognition .17 3 .06 .10 .957 Hypothesis 3a It was hypothesized that there is a positive relationship between brand familiarity (BF) and the intensity of the advertisement (1w)' This hypothesis will be called as ”intensity hypothesis for BF" conveniently. In order to test the hypothesis, the correlation 73 between brand familiarity measured by brand name recognition and Iad measured by the absolute value of the level of the advertisement was computed (Table 13). For the overall data, the correlation coefficient between brand recognition and the Iad was calculated as .16; it was significantly positive (p = .019 at one tail test). At each period of time, the correlation between brand recognition and the Iad was measured for further analysis. The correlation coefficient right after the advertisement exposure was .20 and was marginally significant (p = .73); the correlation four days later was .31 and significant (p = .010); ten days later the correlation turned out to be -.10 and was not different from 0 (p = .468 at two-tail test). The differences between these correlations were statistically tested and are shown in Table 14. The difference between the correlation coefficient right after ad exposure and the correlation four days later was not statistically significant regardless of the increase from .20 to .31. The correlation, however, decreased from .31 to -.10 between the period of the fourth day and the tenth day after the advertisement exposure. 74 Table 13 Correlation Coefficients between Brand Familiarity and the Intensity of the Advertisement time overall right after four days ten days ad exposure later later Brand .16 .20 .31 -.10** Familiarity *(p=.019) (p=.073) (p=.010) (p=.734) * one—tail test (H0: r <= 0) ** p = .468 (when H0::r = 0) Table 14 Differences between Correlations Correlations Z—value p (two tail) Right after ad exposure and four days later -.60 .55 Four days later and ten days later 2.15 .03 A set of multiple comparisons for the means of each recognition measurement for each level of the advertisement were planned and performed in order to see if there are differences. condition were presented in Table 15. The means of brand name recognition under each From the "intensity hypothesis for BF," it is expected that the more intense the advertisement level, the higher the brand recognition. multiple mean comparisons for the expectation were performed by SPSS MANOVA contrast procedure and the results were shown in Table 16. Table 15 75 Brand Familiarity (Recognition) Level of right after four days ten days overall ad ad exposure later later strongly 3.000 2.455 2.636 2.697 positive (<.001) (1.036) ( .545) ( .684) weakly 2.636 2.545 2.727 2.636 positive ( .924) (1.036) ( .905) ( .929) neutral 3.000 1.545 1.909 2.152 (<.001) (1.508) (1.300) (1 278) weakly 2.182 2.545 3.000 2.576 negative (1.401) ( .934) (<.001) (1 001) strongly 2.727 2.818 2.636 2.727 negative ( .905) ( .405) ( .924) ( .761) average 2.709 2.382 2.582 The Multiple Mean Comparisons of Brand Name Recognition 76 Table 16 across the Level of Advertisement level of advertisement mean difference std. err. *p strongly negative vs. weakly negative .15152 .22661 .252 weakly negative vs. neutral .42424 .22661 .032 neutral vs. weakly positive —.48484 .22661 .017 weakly positive vs. strongly positive —.06061 .22661 .395 strongly negative vs. neutral .57576 .22661 .006 neutral vs. strongly positive —.54545 .22661 .009 negative vs. neutral .50000 .19625 .006 neutral vs. positive —.51515 .19625 .005 * one—tail test The multiple mean comparisons clearly indicated that the more intense the level of advertisement, the more familiar the brand name. intensity of the advertisement, the more the brand name will be familiar to consumers. In other words, This generalization is due to the following two sets of test results. the higher the 77 First, the contrasts between adjacent levels of advertisement indicated that when the level of the advertisement is more intense, the brand name is better recognized. The brand name under the condition of strongly negative advertisement is more highly recognized than that under the weakly negative advertisement condition (directionally supported); the brand name recognition under the condition of weakly negative advertisement is higher than that under the neutral advertisement condition (p = .032); the brand name recognition under the condition of weakly positive advertisement is higher than that under the neutral advertisement condition (p = .017); the brand name recognition under the condition of strongly positive advertisement is higher than that under the weakly positive advertisement condition (directionally supported); consequently, strong intensity of both positive and negative advertisement caused better brand name recognition than the neutral advertisement. Second, both negative and positive advertisement caused better brand name recognition than neutral advertisement. The test result indicate that brand name recognition under negative (composite of strong and weak intensity) advertisement condition is higher than that under neutral advertisement (p = .006); brand name recognition under positive (composite of strong and weak intensity) advertisement condition is higher than that under neutral 78 advertisement. To summarize, the "intensity hypothesis for BF" is supported by strong evidence. Hypothesis 3b The hypothesis stated that there is a positive relationship between brand familiarity (IAN) and the intensity of the advertisement (Im)° This hypothesis can be called as "intensity hypothesis for IAN." In order to test the hypothesis, the correlations between IAad and Iad at different time of measurement were compared, and the mean values of IAad in different levels of advertisement were contrasted. As Table 17 indicates, the correlations between IAad and Iad at any time periods were significantly positive. The correlation went up from .40 to .45 after four days; it decreased to .23 ten days later. Neither change, however, was statistically significant. Table 18 reveals the test results for the changes. No changes were significant at an a level of .05. That is, the positive relationship between IAad and Iad is relatively stable across the given time period (10 days). Hypothesis 3c It is expected that at a given level of intensity of an advertisement, brand familiarity (BF) is higher for 79 advertisement with a negative valence than for those with a positive valence. This hypothesis can be called as ”valence hypothesis for BF." To test this hypothesis, brand familiarities under the same intensity level of advertisement but with opposite valence needs to be compared. In the manipulation of the experiment, the intensity of strongly negative advertisement (mean = .85) and the intensity of weakly positive advertisement (mean = .64) were not statistically different. According to a pre—test for the difference between the two intensities (standard deviations are 1.16 and 1.22 respectably), no evidence indicated the two intensities are different when a is at a .05 level (p = .2758). Therefore, the mean values of brand familiarity under strongly negative advertisement and weakly positive advertisement conditions can be compared to find out if there is a difference between them. The mean value of brand name recognition under the strongly negative advertisement was 2.727 with a standard deviation of .761, and that under weakly positive advertisement was 2.636 with a standard deviation of .929. A contrast by SPSS MANOVA procedure indicates that the difference was directionally supported but was not statistically significant (p = .344). Moreover, the mean of brand familiarity (mean = 2.727 and std. dev. = .761) under strongly negative advertisement was higher than that (mean = 80 2.697 and std. dev. = .684) under strongly positive advertisement. Although the fact was only directionally supported (p = .448), it has a significant meaning. Since the intensity of the strongly positive advertisement (mean = 1.63 and std. dev. = .67) was significantly higher than that of the strongly negative advertisement (mean = .71 and std. dev. = 1.16) (p < .01), the strongly positive advertisement was expected to create significantly higher brand familiarity than strongly negative advertisement according to the "intensity hypothesis for BF." The strongly negative advertisement, oppositely, created higher brand name recognition than the strongly positive advertisement (directionally). Consequently, this fact supports the "valence hypothesis" indirectly. In sum, the "valence hypothesis for BF” was marginally supported, which predicts that brand familiarity is higher for advertisements with a negative valence than for those with a positive valence at a given level of intensity of an advertisement. The relationship between the "intensity hypothesis" and the "valence hypothesis" will be discussed in Chapter 5. 81 Table 17 Correlation Coefficients between the Intensity of Aad and the Intensity of the Advertisement time overall right after four days ten days ad exposure later later IAad .39 .40 .45 .23 *(p=.000) (p=.001) (P=.000) (p=.043) * one—tail test (H0: r <= 0) Table 18 Differences between Correlations Correlations Z—value p (two tail) Right after ad exposure and four days later -.32 .75 Four days later and ten days later 1.28 .20 Multiple mean comparisons were conducted to contrast mean differences for different intensity level of the advertisement. The test results are shown in Table 19. The comparisons of IAad in the adjacent categories indicated that the higher the I“, the higher the IAad is. The mean of Iad in the strongly negative category was higher than that 82 in the weakly negative category (p = .004); the mean in the weakly negative was higher than that in the neutral category (p = .006); the mean in the weakly positive category was slightly higher than that in the neutral category (directional support); the mean in the strongly positive category was slightly higher than that in the weakly positive category (directional support). The contrast between the negative category and neutral category indicated that the mean of IAad under negative advertisement condition was higher than that under neutral advertisement condition (p = .001). The contrast between neutral and positive category, however, received only directional support (p = .129). In sum, the "intensity hypothesis for IAM' was well supported by the correlation and multiple mean comparisons. 83 Table 19 Multiple Mean Comparisons of Intensity of Aad across the Level of Advertisement level of advertisement mean difference std. err. *p strongly negative vs. weakly negative .45454 '.17009 .004 weakly negative vs. neutral .43722 .17009 .006 neutral vs. weakly positive -.13400 .17009 .215 weakly positive vs. strongly positive -.06496 .17009 .351 strongly negative vs. neutral 1.61200 . .17009 < .001 neutral vs. strongly positive -.19913 .17009 .122 negative vs. neutral .66450 .14730 .001 neutral vs. positive —.16650 .14730 .129 Hypothesis 3d The hypothesis 3d states that at a given level of intensity of an advertisement (In)! the intensity of the attitude toward the advertisement (IAN) is higher for advertisements with a negative valence than for those with a positive valence. This hypothesis can be called as "valence hypothesis for Aw.' In order to test this hypothesis, (a) 84 intensities of the attitude toward the advertisement (IAw) under same intensity level of advertisement (In) but with opposite valence was contrasted, and (b) the effect of significantly more intense advertisement (positive valence) was contrasted with less intense advertisement (negative valence). As a t-test in the pre—test indicated, there was no evidence to suggest the intensity of strongly negative advertisement (.85) was different from that of weakly positive advertisement (.64) in this experiment. Therefore, the mean value of the intensity of the attitude toward the advertisement (IAw) under strongly negative advertisement condition was contrasted with that under weakly positive advertisement. The mean of IAad under a strongly negative advertisement condition was 1.697 and that a under weakly positive advertisement condition was .939 (Table 20). A contrast by MANOVA procedure suggested that the intensity difference (.75758) with standard error of .17009 was highly significant (p = .000). That is, IAad under a negative advertisement condition is higher than that under a positive advertisement condition, even if the intensities of the 4 advertisement (In) are same. In the other contrast, the IAad under a strongly negative advertisement condition (1.697) was significantly higher than that under a highly positive advertisement 85 condition (1.004). The difference between the two IAms (.693) with standard error of .17009 was statistically significant (p < .001). This was very strong evidence to support the "valence hypothesis for IAN" because IAad under strongly positive advertisement expected to create higher brand name recognition than that under a strongly negative advertisement condition. The expectation was based on the ”intensity hypothesis for IAw" because the intensity of the strongly positive advertisement was significantly higher than that of strongly negative advertisement (p < .01). In sum, the test results showed strong evidence to indicate the ”valence hypothesis for IAN” holds. The further discussion about the relation between the "valence hypothesis for IAw' and ”intensity hypothesis for 1AM” will be in Chapter 5. 86 Table 20 Intensity of the Attitude toward the Advertisement (IAN) Level of right after four days ten days overall ad ad exposure later later strongly 1.104 .857 1.052 1.004 positive ( .971) ( .636) ( .743) weakly .961 .935 .922 .939 positive ( .434) ( .760) ( .773) neutral .974 .701 .740 .805 ( .403) ( .773) ( .477) weakly .234 1.273 1.221 1.243 negative ( .551) (1.005) ( .517) strongly .987 1.779 1.325 1.697 negative ( .621) ( .735) ( .636) average 1.252 1.109 1 052 Hypothesis 4a This hypothesis indicates that the intensity of the attitude toward the advertisement (IAw) decreases over time. As Table 20 indicates the mean values at each time of the measurement decreased (1.252, 1.109, 1.052). An analysis of variance was conducted to see if the decreases are statistically significant (Table 21). The analysis showed that there was no evidence to indicate the mean values of the attitude toward the advertisement were different across time because the main effect of time was 87 not significant (p = .297) and neither was the interaction between Lad and time (p = .789). Consequently, the directional support and the statistical insignificance suggests that the intensity of the attitude toward the advertisement tends to decrease very slowly over the given time (10 days) in the current experimental situation. Table 21 ANOVA for the intensity of Aad Source SS DF MS F P level of advertisement 16.217 4 4.054 8.493 < .001 time 1.167 2 .584 1.223 .297 interaction 2.233 8 .279 .585 .789 residual 71.603 150 .477 total 91.220 164 .556 Hypothesis 4b Brand familiarity (BF) is hypothesized to decrease over time. An analysis variance, where brand name recognition was the dependent variable, was performed to test the hypothesis. The main effect (nondirectional hypothesis) revealed that there was no significant difference between 88 the means at the three different times of measurement. Brand recognition, however, decreased from 2.709 to 2.382 after four days (p = .03 at one-tail directional contrast, std. error = .17553). Unexpectedly, after ten days the mean of the brand name recognition went up to 2.582. The increase was not statistically significant (p = .18). In all levels of the advertisement, except the strongly negative category, the brand name recognition went up although it was not statistically significant. One possible reason for the directional increase could be that subjects who answered the questionnaire in earlier stages of the experiment had mentioned the brand name to other subjects who were going to answer the questions later. Post hoc analysis In this part, the effect of brand familiarity (BF) was reconsidered, since the hypothesis 2b and 2c were not supported unexpectedly. Why was the brand familiarity effect on Ab not significant? One possible reason can be that the familiarity effect could affect purchase intention directly without being mediated by Ab. .According to Batra and Ray (1985), familiarity is the major determinant of purchase intentions in low involvement situations. In this study, this possibility was tested by regression analyses in which brand familiarity (BF) was an independent variable and purchase intention (PI) was a dependent variable. 89 As Table 22 shows, brand recognition is marginally significant in two regression models. These marginal significance suggests further investigation for the relationship between PI and brand recognition with Ab. Table 23 reveals the regression models in which PI is the dependent variable and brand recognition and Abrare independent variables. The test results indicate that brand recognition is a significant variable at all times except four days after the advertisement exposure. The A%*was strongly significant at all times. Table 22 Regression models: PI and brand recognition time regression *p R overall Y —1.185347 + .182264 x1 .095 .011 right after ad exposure Y = -1.213574 + .208590 X1 .225 .011 four days later Y = -.868681 + .054281 X1 .400 .001 ten days later Y = —1.660240 + .344929 X1 .091 .033 *one-tail test 90 Table 23 Regressions for Brand Recognition Time Regression Overall Y = -1.049876 + .967025 x1-+ .204262 X2 * (.099212) (.110054) 2 ** (<.001) (.033) R = .37632 R2 = .36862 Standard Error = 1.35969 n = 165 Right after ad exposure Y = —1.475827 + 1.031517 X1-+ .317263 X2 (.139188) (.193597) 2 (<.001) (.054) R = .51892 R2 = .50042 Standard Error = 1.24126 n = 55 Four days later Y = -.597494 + .859350 X1-+ .068811 X2 (.177399) (.179879) 2 (<.001) (.352) R = .31178 R2 = .28531 Standard Error = 1.44983 n = 55 Ten days later Y = -1.492021 + 1.100077 X1-+ .368311 X2 (.227840) (.213791) 2 (<.001) (.056) R .33262 R2 .30695 Standard Error = 1.40777 n = 55 where X1 = Ab X? = brand recognition Y = purchase intention * (Standard error of the coefficient) ** (p (H0: 8 <= 0)) 91 The brand recall measurement was not significant independently but significant in the multiple regression model after ten days (Table 24 and Table 25). These results provide evidence to indicate there is a positive relationship between brand recognition and purchase intention when Ab is simultaneously considered. Therefore, it suggests that Am and brand name recognition together affect purchase intention. Table 24 Regression models: PI and brand recall time regression *p R overall Y = —.806306 + .157955 X1 .295 .002 right after ad exposure Y = —.952381 + .491597 X1 .159 .019 four days later Y = —.555556 — .361111 X1 .220 .011 ten days later Y = —.948718 + .339523 X1 .230 .010 *one—tail test 92 Table 25 Regressions for Brand Recall Time Regression Overall Y = -.609269 + .962881 X1-+ .146842 X2 * (.100098) (.214781) 2 ** (<.001) (.248) R = .36489 R?- = .35705 Standard Error = 1.37209 n = 165 Right after ad exposure Y = -.621385 + 1.013676 X11+ .007271 X2 (.145039) (.360020) 2 (<.001) (.492) R = .49408 R2 = .47462 Standard Error = 1.27291 n = 55 Four days later Y = -.357559 + .850483 X1-+-.155562 X2 (.178433) (.393387) 2 (<.001) (.347) R = .31191 R2 = .28544 Standard Error = 1.44969 n = 55 Ten days later Y = —.860504 + 1.146786 X1-+ .624155 X2 (.231078) (.385715) 2 (<.001) (.056) R = .32835 R2 = .30252 Standard Error = 1.41226 n = 55 where X1 = Ab X? = brand recall Y = purchase intention * (Standard error of the coefficient) ** (p (H0: B <= 0)) CHAPTER FIVE DISCUSSION AND CONCLUSIONS This chapter begins with a discussion of the findings. In the section that follows, the limitations of the research are examined. Then, contributions and managerial implications are discussed. Finally, a summary is given. Discussion of Findings The first hypothesis centered around the relationship between the level of advertisement and the attitude toward the advertisement. It was posited that there is a positive relationship between the level of advertisement (La) and the attitude toward the advertisement (Aw)- The Pearson correlation coefficients for the two constructs for the overall data indicated a relatively strong positive relation (r = .70). It means that if the level of the advertisement is higher, then a higher level of attitude toward the advertisement will be expected. The positive relationship was further investigated at each time of measurement. Since the attitude toward the advertisement was measured at three different times after 93 94 the advertisement exposure, three additional correlation coefficients were computed to investigate if the positive relationship exists at each time period and if any changes occurred in the relationship during the given period of time (10 days). The three correlation coefficients indicated that at each time period of measurement (right after the advertisement exposure, four days after the exposure, and ten days after the exposure), the positive relationship was maintained. The relationship appeared to decrease with the passage of time. Right after the advertisement exposure the relationship was the strongest (r = .75); four days later the relationship started decreasing (r = .72); ten days later the relationship has weakened significantly (r = .59). This finding suggests that the attitude toward the advertisement can be well manipulated by the level of the advertisement and the manipulated attitude toward the advertisement persists over time. The effect of the level of the advertisement on the attitude toward advertisement diminishes with a passage of time. The initial effect level can be maintained for a certain period of time (about four days in this research experiment), then the effect weakens more sharply after the point in time. Figure 11 represents the changes. A set of hypotheses within hypothesis 2 attempted to explain the effect of attitude toward the advertisement and 95 the brand familiarity on the formation of the attitude toward brand. Hypothesis 2a leads one to expect a positive relationship between the attitude toward the advertisement (Am) and the attitude toward the brand (Ab). Hypothesis 2a was formulated based on classical conditioning theory and expected to be significant because many studies have provided support. Particularly, since the brand name in this research was new and the product category was generally considered to be low involvement, a strong positive relation was expected. Without disappointment, all of the four simple regression models at different times of measurement revealed strong relationships. For a period of time (four days in this research experiment), the magnitude of the relation remained stable (R2 = .541 and .645), but after then the magnitude decreased rather sharply (R2 = .361). This finding confirmed the generalization between Aad and Ab and disconfirmed the balance theory in explaining the relationship. In a static time, it seems that both classical conditioning and balance theory were confirmed by the results because of the significant positive relationship between Aad and Ab. In a dynamic sense, however, the finding disconfirmed balance theory, though it still confirmed classical conditioning theory. According to classical conditioning, the generalization between Aad and Ab tends to weaken over time due to forgetting (and extinction). On the other hand, according to balance theory, the strength of the 96 association should remain the same unless the balance is broken by new associations of the same constructs. Since there was no chance to see the Chowa chewing gum or to watch commercials about the gum, balance theory predicts that the three correlation coefficients between Aad and Ab at different point in time should be same. This, however, was not the case. This finding further extended the dynamic understanding of the relationship between Aad and Ab. It is of interest to note that the decrease in the magnitude of the relationship does not seem to be linear. Figure 11 represents the change. It is also interesting to remember that a similar pattern was noticed in the relationship between the level of advertisement and the attitude toward advertisement in the analysis of the Hypothesis 1. Further research in the dynamic change of both relationships would open an avenue to longitudinal understanding of consumer information processing. 97 Figure 11 Dynamic Changes of the magnitude of the relationships correlation .74 .80 .75 .72 m b 59 Lad 8' Aad .31 - 10 BF & Iad time t1 t2 t3 Hypothesis 2b indicated a positive relationship between brand familiarity (BF) and the attitude toward the brand (Ab). It was posited based on the mere exposure hypothesis. The hypothesis expected that increased brand familiarity may increase liking for the brand. In order to test the hypothesis 2b, a set of regression models were also built and tested for the significance. The result, unexpectedly and rather surprisingly, none of the eight regression models (four brand recognition models and four brand name recall models) was significant. Hypothesis 2c considered simultaneously the attitude toward the advertisement (Am) and the brand familiarity (BF) to predict the attitude toward the brand (Ab). The joint effect on the attitude toward the brand (Ab) was 98 tested in multiple regression models in which Aad and BF were independent models. The regression models revealed that at any time of the measurement (overall, right after ad exposure, 4 days later, or 10 days later) the independent variable the brand familiarity was not statistically significant. The attitude toward the advertisement (Am)! on the other hand, was highly significant at any time of the measurement. In another attempt, the interaction term between Aad and A.b was included as well as the main effects. The regression analyses also suggested that Aad was the only significant independent variable and neither BF nor the interaction term was significant. This finding was unexpected because the literature suggested there would be a familiarity effect on the attitude toward brand formation. Zajonc (1968, 1980, 1986) proposed mere exposure effect and posited that the increased brand familiarity may increase liking for the brand. Obermiller's (1985) ”exposure without cognition model" suggested that familiarity may create positive feeling. However, the familiarity effect was not confirmed in the current research. The effect was not confirmed even directionally because the B coefficients in any of the eight regression models were hardly different from zero and the signs of the coefficient indicated no pattern. Moore and Hutchinson's (1983) familiarity—based-sleeper hypothesis also predicts that over time the direct influence 99 of ad affect on Ab should decay and the impact of brand familiarity should increase. In the current research, however, no increase or significant effect of familiarity effect on.A% was observed. Consequently, the familiarity— based—sleeper hypothesis was not confirmed. On the other hand, as a post hoc analysis in Chapter Four indicates, brand familiarity as well as attitude toward the brand seems to have a direct and positive effect on purchase intention over time. Brand familiarity effect does not seem to be mediated by attitude toward the brand at any times of measurements. The Hypothesis 3 set focused on the effect of the level of the advertisement (Lw) on both of the brand familiarity (BF) and the intensity of the attitude toward the advertisement (IAN). To give the conclusion first, the test results of the hypotheses suggest that there may be a "intensity—valence effect" over time. "Intensity-valence effect" refers to a phenomenon that as the intensity of advertisement (1“) increases, the brand familiarity (BF) and the intensity of the attitude toward the brand (1AM) increase at a diminishing rate with a condition that a negative advertisement tends to have a stronger effect than a positive advertisement. The "intensity-valence effect" is represented in Figure 12. 100 Figure 12 Intensity-Valence Effect BF, IAad , negative ///////,, positive Iad The magnitude of "intensity-valence effect" seemed to be maintain same for a period of time (four days in this study), but decreases quickly after the period. The rationale for the "intensity—valence effect” is discussed as following. "Intensity hypothesis for BF" (H3a) explains that there is a positive relationship between BF and I“. In other words, the higher the intensity of an advertisement, the higher the brand familiarity formed. This hypothesis was strongly supported by correlation coefficient analyses and multiple mean comparisons for different levels of the advertisement. Note that brand familiarity under strongly positive (negative) advertisement was only directionally higher than that under weakly positive (negative) advertisement. This fact may suggest that as the Iad increases, BF increases with a diminishing 101 rate. Weber's law can give rationale to support this phenomenon. Weber's law describes that the smallest increase in intensity of the object that is noticeable increases as the intensity of the initial object increases. ”Intensity hypothesis for IAN" (H30) was also supported and suggests that as the intensity of an advertisement (In) increases, the intensity of Aad (IAN) increase. Correlation coefficients and multiple mean comparisons support the hypothesis. Regarding the effect of the valence of an advertisement (Va) on the brand familiarity (BF) and the intensity of the attitude toward the advertisement (IAN), ”valence hypothesis for BF” and "valence hypothesis for IAw" were proposed. Multiple mean comparisons marginally supported "valence hypothesis for BF" and strongly supported the valence hypothesis for IAN. That is, the effect of the intensity of an advertisement (In) on brand familiarity (BF) and the intensity of Aad (IAN) is stronger when the valence is negative than when positive. Neither of the "intensity hypothesis" or "valence hypothesis" would be completed alone because an intensity of advertisement and a valence of advertisement together make the level of an advertisement. Otherwise, each hypothesis explains only one dimension of the effect and might be misleading in some cases. For example, when the levels of two advertisements are —1 and +1, if only the "intensity 102 hypothesis" is applied, then the same brand familiarity and intensity of the attitude toward the advertisement are expected, which may not be true according to the valence hypothesis. For another example, when the levels of two advertisements are —1 and +3, if only the ”valence hypothesis" is applied, then a higher BF and IAad is expected for the negative advertisement than for the positive advertisement, which also may not be true based the "intensity hypothesis." To be completed, therefore, on the intensity effect and the valence effect are integrated, and "intensity-valence effect” is suggested. The definition the effect was given earlier. Hypotheses 4a and 4b predicted a decrease of the intensity of attitude toward the advertisement (1AM) and brand familiarity over time. The test results only gave only directional support. Consequently, all of the above findings together suggest an affective information processing model. This of model explains that "intensity—valence effect" affects the positive relationships between Lad and Au! and between Lad and BF; there is a positive relation between Aad and Ab; Ab and BF jointly provide positive relationships with PI. Figure 13 represents the proposed affective information processing model. 103 Figure 13 Proposed Affective Information Processing Model Aad Add 11».ad . Ab Lad vad PI I ad BF Limitations of the Research As with most experimental research, this study has a limitation in external validity. Since the experiment was conducted in classrooms and the subjects had to come only for the experiment, the ad-watching environment could be different from the natural or real-world watching environment. In order to reduce this external validity problem, three attempts were made. First, the advertisements were produced by professionals. A professional artist designed the logo and package design of the product and combined background pictures with the 104 package of gum; a professional photographer transformed the advertisements into slides; an audio professional edited the background music in the slide show narration tape. Second, the advertisements were embedded in a slide show. This method is considered to be superior to another method in which subjects watch only commercials in a laboratory. Last, the subjects were not informed about the purpose of the study. They simply were told to watch a slide show. Another limitation is in the sampling process. Using student sampling offers convenience but also limits the generalization of the result to real world. The limitation is not critical in this research, though, because the main purpose of the research is not to generalize the result to real population, but to understand theoretically how consumer information processing occurs. Rather, this convenient sample was beneficial in controlling the experiment because the students were relatively homogeneous. Finally, since this experiment was conducted over ten days and some subjects finished the questionnaire earlier than others, if anyone discussed the research questionnaire with subsequent participants the answers could be biased. In particular, the reliability of the brand name recognition and recall measurement would suffer if the subjects knew the questions before hand. An attempt was made to reduce this possibility. The written questionnaire directions did request participants to not discuss the questionnaire with 105 anyone. Managerial Implications The findings in this research suggest several managerial implications. These implications can be applied mostly to a new or unfamiliar product under a low involvement condition. First, brand familiarity as well as attitude toward the advertisement are important factors. When a product is introduced, advertising managers should attempt to increase brand familiarity as well as the attitude toward the brand through the attitude toward the advertisement. Second, when the effect of an advertisement is measured, it is desirable to measure and analyze both of the attitude toward the brand and brand familiarity (brand recognition). Third, when brand familiarity is measured to predict purchase intentions, brand name recognition seems to be a better measurement than unaided brand name recall. It is because of the research result that brand name recognition was a significant variable in predicting purchase intention, whereas unaided brand name recall was not. Fourth, since the intensity and valence of advertisement are found to be important, the "intensity— valence effect" can be applied to create a desirable 106 advertisement level. When frequent advertisement repetition is not available, advertisement with strong intensity may be effective. Advertisement with negative valence can be effective in achieving brand familiarity. Fifth, a manager needs to understand the dynamic change of an advertisement effect. As a research finding suggests, if the magnitude of the advertisement effect persists for a period of time, without loosing too much effectiveness, less frequency could be cost effective. Future Research Future research is needed to confirm the "intensity~ valence effect" in different settings. The effect was fairly well supported in this study. However, the effect may be situation specific. More replication studies and retesting of the effect in different settings are desirable. Future research needs to be directed toward a better understanding of the effect of brand familiarity. Even if the current study suggests that brand familiarity affects purchase intentions directly, a clearer understanding is needed. The new research could attempt to answer the following questions. Why, in this research, did brand familiarity not have a relationship with the attitude toward the brand which was contrary to what literature suggested? Can brand familiarity directly affect the behavior? Does 107 brand familiarity have a more important role when a consumer has to evaluate some alternative products (relative comparison) than to evaluate a product likeness (absolute measurement)? A stream of research on the dynamic changes of advertising effect would be of interest and make important theoretical and managerial contributions. Finally, research can be done to confirm the whole model suggested by the current study in the context of causal structure. The relationship among the constructs (Lad, Aad, BF, Ab, and PI) can be retested by considering each relationship simultaneously. LISREL or PACKAGE could be used. Summary This study has contributed to a better understanding of Consumer information processing. In an attempt to explain the role of an advertisement in the formation of the attitude toward the advertisement and brand familiarity, the "intensity—valence effect" was suggested as the result of integrating supported hypotheses. The "Intensity-valence effect" explains that as the intensity of advertisement (1w) increases, the brand familiarity (BF) and the intensity of the attitude toward the brand (IAM) increase at a diminishing rate with the condition that a negative 108 advertisement tends to have a stronger effect than a positive advertisement. The magnitude of the "intensity— valence effect” seems to be maintain same for a period of time, then it seems to decrease quickly. It was confirmed that the attitude toward the advertisement (Am) has a positive relationship with the attitude toward the brand (Ab).- The strength of the relationship persisted for a period of time but after additional time it weakened rather sharply. No evidence found that brand familiarity (BF) has a positive relationship with the attitude toward the brand (Ab). Brand familiarity, rather, seems to have a significant effect on purchase intentions directly in conjunction with the attitude toward the brand. This study provided some understanding of how consumer information is processed. The new understanding, however, only raises new questions. APPENDICES APPENDIX A RESEARCH CONSENT FORM RESEARCH CONSENT FORM With regard to my participation in research: 1. I understand that participation is voluntary and not required as part of course requirements. I agree to sign up for a study ONLY WHEN I FULLY INTEND TO PARTICIPATE. 2. I understand the procedures by which my participation will count for some form of credit in the class listed below and an alternative way to earn class credit is available. 3. I understand that any credit I may earn via participation in research is not transferable to another class or another term. 4. I understand that, apart from my participation in a given study, my actual performance in that study will in no way affect my evaluation in a given course. 5. I understand that the data will be collected anonymously. 6. I understand that the data I provide a researcher as a result as a result of my participation in a given study may be used by other scientists for secondary analysis. Again data will be treated with the strict anonymity. 7. I understand that the participation involves attending two sessions and the total time required will be around 30 minutes. 8. I understand that should I have any questions, problems, complaints, or I desire further information, I have the right to contact the Research Coordinator. Given these understandings, I have freely consented to participate in scientific research being conducted during this term. Signed Date Name (Print) Student Number Class MTA317 Section Time Class Meets Teaching Assistant's Name 109 APPENDIX B QUESTIONNAIRE 110 Thank you very much for your participation. Your sincere answers will be very important for a research project at MSU. INSTRUCTIONS 1. Please read instructions for each question carefully. 2. Please answer every questions. 3. If you have any questions, please raise your hand. 4. Please do not discuss this questionnaire with anyone after you answer. 5. Do not write your name on this questionnaire. 111 1. Place a mark on the position between each adjective pair which best describes your opinion about the slide show (Getting Ready To Ride) as a whole. Example: bad dislike not irritating interesting favorable unenjoyable pleasant bad dislike not irritating interesting favorable unenjoyable pleasant irritating boring unfavorable enjoyable unpleasant good like irritating boring unfavorable enjoyable unpleasant 2. You have seen a few times of an advertisement during the slide show (Getting Ready To Ride). name that appeared in the advertisement. Please recall the brand (Please do not go to next page until you answer the question on this page. Once you go to the next page, you must not return to this page in order to change your answer.) 3. Regarding the advertisement in the slide show (Getting Ready To Ride) you have watched, how much the following 112 brand names are familiar to you? Chuncha: definitely familiar Chingwe: definitely familiar Chowa: definitely familiar Waddsha: definitely familiar (Please do not go to next page until you answer the questiOn on this page. Once you go to the next page, you must not definitely unfamiliar definitely unfamiliar definitely unfamiliar definitely unfamiliar return to this page in order to change your answer.) 113 4. Place a mark on the position between each adjective pair which best describes your opinion about the chewing gum "Chowa". bad : : : : : : good dislike : : : : : : like favorable : : : : : : unfavorable enjoyable : : : : : : unenjoyable useless : : : : : : useful low quality : : : : : : high quality beneficial : : : : : : harmful 114 5. Place a mark on the position between each adjective pair which best describes your overall reaction to the chewing gum ”Chowa" advertisement in the slide show (Getting Ready To Ride) that you have watched. bad : : : : : : good like : dislike irritating : not irritating interesting : : : : : : boring favorable : : : : : : unfavorable enjoyable : : : : : : unenjoyable pleasant : : : : : : unpleasant 6. Please circle around one of the words to complete the following sentence. The advertisement of the chewing gum "Chowa" in the slide show (Getting Ready To Ride) is overall a ( ) advertisement. I --------- | --------- l negative neutral positive 7. If you answered in question #7 as either negative or positive, then how strongly do you think so? (If your answer was neutral, then you don't have to answer this question #7.) 115 8. Place a mark on the position between each adjective pair which best describes the likelihood that you will purchase the chewing gum "Chowa” in the future. improbable : : : : : : probable likely : : : : : : unlikely possible : : : : : : impossible 9. Have you heard the brand name of chewing gum ”Chowa” before watching the slide show (Getting Ready To Ride)? never one or a few many two times times times 10. During the slide show (Getting Ready To ride), I paid attention More More To the : : : : : : To the Program Advertisement 116 11. Finally, please answer the question below. I am years old. I am a male, female. I am a freshman sophomore junior senior graduate other. Thank you very much for your cooperation in completing this questionnaire. BIBLIOGRAPHY BIBLIOGRAPHY Aaker, D.A. and D.E. Bruzzone (1985), "Causes of Irritation in Advertising," Journal of Marketin , 49 (Spring), 47—57. Allen, C.T. and T.J. Madden (1985), ”A Closer Look at Classical Conditioning,” Journal of Consumer Research, 12 (December), 301—315. Appel, V. (1971), ”On Advertising Wearout,” Journal of Advertising Research, 11 (February), 11—13. Axelrod, J.N. (1980), ”Advertising Wearout,” Journal of Advertising Research, 20 (October), 13—17. Batra, R. (1986),"Affective Advertising: Role, Processes, and Measurement," in The Role of Affect in Consumer Behavior, Peterson, R.A., W.D. Hoyer, W.R. Wilson (eds.) Lexington, MA: Lexington Books, 53—85. , and M.L. Ray (1983), "Operationalizing Involvement as Depth and Quality of Cognitive Response," Advances in Consumer Research, vol. 10, R.P. Bagozzi and A.M. Tybout, (eds.), Ann Arbor, MI: Association for Consumer Research, 309—313. nd (1985), "How Advertising Works at Contact, " in Psychological Processes and Advertising Effects: theory, research, and application, Linda F. Alwitt and Andrew A. Mitchell, (eds.), Hillsdale, NJ: Lawrence Erlbaum Associates, 13-43. and (1986), "Affective Responses Mediating Acceptance of Advertising," Journal of Consumer Research, 13 (September), 234—49. Belch, G.E. (1982), ”The Effects of Television Commercial Repetition on Cognitive Response and Message Acceptance,” Journal of Consumer Research, 9 (June), 56—65. 117 118 and M.A. Belch (1984), "An Investigation of the Effects of Repetition on Cognitive and Affective Reactions to Humorous and Serious Television Commercials,” Advances in Consumer Research, vol. 11, T.C. Kinnear, (ed.), Provo, UT: Association for Consumer Research, 4—10. Bierley, C., F.K. McSweeney, and R. Vannienwkerk (1985), "Classical Conditioning of Preferences for Stimuli,” Journal of Consumer Research, 12 (December), 316—23. Boderns, K.S. and B.B. Abbott (1988), Research Design and Methods: A Process Approach. Mountain View, CA: Mayfield Publishing Company Bower, G.H. and P.R. Cohen (1982), "Emotional Influences in Memory and Thinking," in Affect and Cognition, Clark, M.S. and S.T. Fiske (eds.), Hillsdale, NJ: Lawrence Erlbaul Associates. Broadbent, E. (1977), "The Hidden Preattentive Processes,” American Psychologist, 32, 108—118. Burke, M.C. and J. A. Edell (1986), ”Ad Reactions Over Time: capturing changes in the real world," Journal of Consumer Research, 13 (June), 56—65, 114—118. Calder, B.J. and B. Sternthal (1980), "Television Commercial Wearout: an information processing view,” Journal of Marketing Research, 17 (May), 173—186. Choi, Y. and E. Thorson (1983), ”Memory for Factual, Emotional, and Balanced Ads Under Two Instructional Sets,” Proceedings of the 1983 Conference of the American Academy of Advertising, 160—163. Clore, G.L. and J.A. Mccarthy (1985), "How the Head Hears the Heart: studies of judgment and emotion," Proceedings of the 1985 Conference of the American Academy of Advertising, Nancy Stephens, (ed.), 191—195. Cohen, J. (1977), Statistical Power Analysis For The Behavioral Sciences, Revised Edition, Orlando, FL: Academic Press. Cook, T.D. and B.R. Fray (1978), ”The Persistence of Experimentally Induced Attitude Change,” in L. Berkowitz (ed.), Advances in Experimental Social Psychology, vol. 11, New York: Academic Press, 1—57. 119 C. L. Gruder, K. M. Hennigan, and B.R. Fray (1979), “History of the Sleeper Effect: Some Logical Pitfalls in Accepting the Null Hypothesis," Psychological Bulletin, 86 (4), 662-79. Dabbs, J. and H. Leventhal (1966), "Effects of Varying the Recommendations in a Fear- -Arousing Communication, Journal of Personality and Social Psychology, 4 (November), 525-31 Domjan, D. and B. Burkhard (1986), The Principles of Learning and Behavior, Monterey, CA: Brooks/Cole Publishing Company. Dutta, S. and R.N. Kanungo (1975), Affect and Memory: a reformulation, Oxford: Pergamon Press. Edell, J.A. and M.C. Burke (1984), ”The Moderating Effect of Attitude Toward an Ad on Ad Effectiveness Under Different Processing Conditions," in Advances in Consumer Research, vol. XI, T.C. Kinnear, (ed.), Provo, UT: Association for Consumer Research, 644—49. and (1987), "the Power of Feelings in Understanding Advertising Effects,” Journal of Consumer Research, 14 (December), 421—33. Fishbein, M. (1963), ”An Investigation of the Relationship Between Beliefs About an Object and the Attitude Toward that Object," Human Relations, 16, 233—240. (1967). ”A Behavior Theory Approach to the Relations Between Beliefs About an Object and the Attitude toward that Object," in Readings in Attitude Theory and Measurement, Martin Fishbein, ed., New York: John Wiley & Sons, Inc. Gardner, M.P. (1985a), ”Does Attitude Toward the Ad Affect Brand Attitude Under a Brand Evaluation Set?,” Journal of Marketing Research, 12 (December), 281—300. (1985b), ”Mood States and Consumer Behavior: a critical review," Journal of Consumer Research, 12 (December), 281—300. and S. P. Raj (1983), "Responses to Commercials in Laboratory Versus Natural Settings: a conceptual framework," in R. P. Bagozzi & A.M. Tybout (eds.), Advances in Consumer Research, vol. X, Ann Arbor: Association for Consumer Research, 142—6. 120 Gelb, B.D. and C.M. Pickett (1983), "Attitude-Toward—the Ad: links to humor and to advertising effectiveness," Journal of Advertising, 12 (2), 34-42. Gillig, P.M. and A.G. Greenwald (1974), "Is it Time to Lay the Sleeper Effect to Rest?" Journal of Personality and Social Psychology, 299, 132-139. Golden, L.L. and K.A. Johnson (1983), ”The Impact of Sensory Preference and Thinking Versus Feeling Appeals on Advertising Effectiveness,” in Advances in Consumer Research, vol. 9, R.P. Bagozzi and A.M. Tybout, (eds.), Ann Arbor, MI: Association for Consumer Research, 203— 208. Gorn, G.J. (1982), "The Effects of Music in Advertising on Choice Behavior: A Classical Conditioning Approach," Journal of Marketing, 46, 94—101. Gresham, L.G. and T.A. Shimp (1985), "Attitude Toward the Advertisement and Brand Attitudes: a classical conditioning perspective,” Journal of Advertisin , 14 (1), 10—17, 49. Greyser, S.A. (1973), "Irritation in Advertising," Journal of Advertising Research, 13 (February), 3410. Grifitt, W. and P. Guay (1969), ”'Object' Evaluation and Conditionaing Affect,” Journal of Experimental Research in Personality, 4 (July), 1—8. Gruder, C. L., T.D. Cook, K.M. Hennigan, B.R. Flay, C. Alessis, and J. Halamaj (1978), ”Empirical Tests of the Absolute Sleeper Effect Predicted from the Discounting Cue Hypothesis. Journal of Personality and Social Psychology, 36, 1061—1074. Hannah, D. and B. Sternthal (1984), "Detecting and Explaining the Sleeper Effect," Journal of Consumer Research, 11 (September), 632—42. Harrell, G.D. (1986), Consumer Behavior, San Diego: Harcourt Brace Jovanovich Publishers. Harrison, A.A. (1977), "Mere Exposure," in L. Berkowitz (ed.), Advances in Experimental Social Psychology (vol. 10), New York: Academic Press. Heider, F. (1946), "Attitudes and Cognitive Organization," Journal of Psychology, 21, 107—12. 121 Holbrook, M.B. (1978), "Beyond Attitude Structure: toward the informational determinants of attitude," Journal of Marketing Research, 15 (November), 546—56. and E.C. Hirschman (1982), "The Experiential Aspects of Consumption: consumer fantasies, feelings, and fun," Journal of Consumer Research, 9 (September), 132—40. and R.A. Westwood (1985), ”The Role of Emotion in Advertising Revisited: Testing a Topology of Emotional Responses," paper presented at the Advertising and Consumer Psychology Conference, Needham Harper Worldwide, (July). Hovland, C.I., A.A. Lumsdaine, and F.D. Sheffield (1949), Experiments in Mass Communication, Princeton, NJ: Princeton University Press. Hovland, C.I., L.J. Irvine, and H.H. Kelley (1953), Communication and Persuasion, Hew Haven, CT: Yale University Press. Hutchinson, J.W. and D.L. Moore (in press), "Issues Surrounding the Examination of Delay Effects in Advertising," in T. Kinnear (ed.), Advances in Consumer Research, vol. II, Ann Arbor: Association for Consumer Research. Isen, A., T. Shalker, M. Clark, and L. Karp (1978), "Affect, Accessibility of Material in Memory, and Behavior: A Cognitive Loop?," Journal of Personality and Social Psychology, 36 (January), 1—12. Ittelson, W.H. (1973), "Environment Perception and Contemporary Perceptual Theory," in Environment and Cognition, Ittelson, W.H. (ed.), New York: Seminar Press. Johnson, H.H. and T.A. Watkins (1971),"The Effects of Message Repetition on Immediate and Delayed Attitude Change," Ps chonomic Science, 22 (2), 101—3. Kelman, H.C. and C.I. Hovland (1953), "Reinstatement of the Communicator in Delayed Measurement of Opinion Change,” Journal of Abnormal and Social Psychology, 48 (3), 327— 35. Keppel, G. (1982), Design and Analysis: A Researcher's Handbook, 2nd edition, Englewood Cliffs, NJ: Prentice— Hall. 122 Kroeber-Riel, W. (1979), "Activation Research: Psychobiological Approaches in Consumer Research," Journal of Consumer Research, 5, 240-50. (1983), "Emotional Product Differentiation by Classical Conditioning," in Kinnear, (ed.), "Advances in Consumer Research," Vol. 11, Ann Arbor, MI: Association for Consumer Research, 538-43. Krugman, H.E. (1972), "Why Three Exposures May Be Enough," Journal of Advertising Research, 12 (December), 11-14. Lee, H. (1987), "Moderating Roles of Involvement in Information Processing Routes and Message Acceptance for Differing Numbers of AD Repetitions, Ph.D. Dissertation, E. Lansing, Michigan State University. Lutz, R.J. (1975), "Changing Brand Attitudes Through Modification of Cognitive Structure," Journal of Consumer Research, 1 (March), 49-59. (1985), "Affective and Cognitive Antecedents of Attitude Toward the Ad: a conceptual framework," in Linda F. Alwitt and Andrew A. Mitchell, (eds.) Psychological Processesyand Advertising Effects: theoryr research and application. Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers, 45—65. , S.B. MacKenzie, and G.E. Belch (1983), "Attitude Toward the Ad as a Mediator or Advertising Effectiveness: Determinants and Consequences," in Advances in Consumer Research, vol. 10, R.P. Bagozzi and A.M. Tybout, eds., Ann Arbor, MI: Association for Consumer Research, 532—539. Machleit, K. A. and R.D. Wilson (1988), "Emotional Feelings and Attitude Toward the Advertisement: The Roles of Brand Familiarity and Repetition," Journal of Advertising, Vol. 17 (3), 27—35. MacKenzie, S.B. and R.J. Lutz (1983), "Testing Competing Models of Advertising Effectiveness: A Test of Competing Explanations,"Journal of Marketing Research, 23 (May), 130-143. , (1989), "An Empirical Examination of the Structural Antecedents of Attitude Toward the Ad in an Advertising Pretesting Context, " Journal of Marketing, Vol. 53 (2), (April), 48-65. 123 , and G. E. Belch (1986), "The Role of Attitude Toward the Ad as a Mediator of Advertising Effectiveness: a test of competing explanations," Journal of Marketin Research, 23 (May) 130—43. Maden, T.J., K. Debevec, and J.L. Twible (1985), "Assessing the Effects of attitude—Toward—the-Ad on Brand Attitudes: a multitrait—Multimethod Design," in Marketin Communications——Theor and Research, M.J. Houston and R.J. Lutz, (eds.), Chicago, IL: American Marketing Association, 109—113. Mazursky, D. and Y. Schul (1988), "The Effect of Advertisement Encoding on the Failure to Discount Information: Implications for the Sleeper Effect," Journal of Consumer Research, Vol 15, (June), 24—36. McSweeney, F.K. and C. Bierley (1984), "Recent Developments in Classical Conditioning,”Journal of Consumer Research, Vol. 11, (September), 619—31. Messmer, D.J. (1979), "Repetition and Attitudinal Discrepancy Effects on the Affective Response to Television Advertising," Journal of Business Research, 7 (1), 75—93. Mitchell, A.A. (1983), ”The Effects of Visual and Emotional Advertising: an information—processing approach," in Advertising and Consumer Psychology, Larry Percy and Arch G. Woodside, (eds.) Lexington, MA: D.C. Heath and Company, 197—217. and J.C. Olson (1981), ”Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitudes?” Journal of Marketin Research, 18 (August), 318—22. Mizerski, R.W. and J. D. White (1986), ”Understanding and Using Emotions in Advertising," Journal of Consumer Marketing, 3 (Fall), 57—69. Moore, D.L. and J.W. Hutchinson (1983), "The Effects of Ad Affect on Advertising Effectiveness," in R.P. Bagozzi and A.M. Tybout (eds.), Advances in Consumer Research, vol. X, Ann Arbor: Association for Consumer Research, 526—31. Mueller, D.J. (1986), Measuring Social Attitudes: A Handbook for Researchers and Practioners, New York, NY: Teachers College Press. 124 Mueller, C.W. and M.J. Watkins (1977), "Inhibition from Part—se Cuing: A cue Overload Interpretation," Journal of Verbal Learning and Verbal Behavior, 20, 251—66. Nord, W.R. and J.P. Peter (1980), "A behavior Modification Perspective on Marketing,"Journal of Marketin , 46 (Summer), 86—91. Obermiller, C. (1985), "Varieties of Mere Exposure: The Effects of Processing Style and Repetition on Affective Response," Journal of Consumer Research, 12, 17—30. Olson, D., M.J. Schlinger, and C. Young (1982), "How Consumers React to New-product Ads," Journal of Advertising Research, 22(3), 24-30. Pavlov, I.P. (1927), Conditioned Reflexes, Oxford, England: Oxford University Press. Park, C.W. and S.M. Young (1983), ”Types and Levels of Involvement and Brand Attitude Formation Process," in Advances in Consumer Research, Vol. 10, R.P. Bagozzi and A.M. Tybout, eds., San Francisco: Association for Consumer Research, 320—4. , and (1986), "Consumer Response to Television Commercials: the impact of involvement and background music on brand attitude formation," Journal of Marketing Research, 23 (February), 11—24. Petty, Richard E. and John T. Cacioppo (1981), Attitudes and Persuasion: Classic and.Contemporary Approaches, Dubuque, IA: W.C. Brown. , , and D. Schumann (1983), Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement,” Journal of Consumer Research, 10 (September), 135—46. Puto, C.P. and W.D. Wells (1984), "Informational and Transformational Advertising: the differential effects of time,” in Advances in Consumer Research, T. Kinnear (ed.), Vol. XI, Provo, UT: Association for Consumer Research, 638—43. Ray, M.L. (1973), "Psychological Theories and Interpretations of Learning," in Consumer Behavior: Theoretical Sources, Ward, S. and T.S. Robertson (eds.), Englewood Cliffs, NJ: Prentice—Hall. 125 (1977), "When Does Consumer Information Processing Research Actually Have Anything to Do With Consumer Information Processing?" in Advances in Consumer Research, Perrault, W.D.Jr. (ed.), Vol. 4, Atlanta: Association for Consumer Research, 372—5. (1982), Advertisin and Communication Mana ement, Englewood Cliffs, NJ: Prentice-Hall and R. Batra (1983), "Emotion and Persuasion in Advertising: What We Know and Don't Know About Affect," In Advances in Consumer Research, Bagozzi R.P. and A.M. Tybout (eds.) Vol.10, Ann Arbor, MI: Association for Consumer Research, 543—8. and W. L. Wilkie (1970), "Fear: the potential of an appeal neglected by marketing,” Journal of Marketing, 34 (January), 54—62. Ronis, D.L., M.H. Baumgardner, M.R. Leippe, J.T. Cacioppo, and A.G. Greenwald (1977), "In Search of Reliable Persuasion Effects: I. a computer—controlled procedure for studying persuasion," Journal of Personality and Social Psychology, 35, 548—569. Sawyer, A.G. (1981), "Repetition, Cognitive Responses, and Persuation.” In Co nitive Res onses to Persuasion, Petty R.E., T.M. Ostrom, and T.C. Brock (eds.), Hollsdale, NJ: Erlbaum, 237-62. Shimp, T.A. (1981), "Attitude toward the Ad as a Mediator of Consumer Brand Choice," Journal of Advertisin , 10, 9— 15. Silk, A.J. and T.J. Vavra, (1974), "The Influence of Advertising's Affective Qualities on Consumer Response, " in D. G. Hughes and M. L. Ray (eds.), BuyerZConsumer Information Processing, Chapel Hill, University of Carolina Press, 157— 186. Snodgrass, J.G., G. Levy—Berger, and M. Haydon (1985), Human Experimental Psychology, New York: Oxford University Press, Inc.. Srull, T.K. (1983), "Affect and Memory: the impact of affective reactions in advertising on the representation of product information and memory,” in R.P. Bagozzi and A.M. Tybout (eds.), Advances in Consumer Research, vol, X, Ann Arbor: Association for Consumer Research, 520—5. 126 Weiss, W. (1953), "A 'Sleeper' Effect in Opinion Change," Journal of Abnormal and Social Psychology, 48, 173—180. Watkins, M.J. (1979), ”Engrams as cuegrams and forgetting as cue overload: Acueing approach to the structure of memory," in Memor Or anization and Structure, Puff, C.R. (ed.), New York: Academic Press. and Watkins, O.C. (1975), ”Buildingup of Proactive Inhibition as a Cue—overload effect," Journal of Experimental Psychology: Human Learning and Memory, 1, 442—52 and (1979), "Cue—overload Theory and the Method of Interpolatated Attributes," Bulletin of the Psychonomic Society, 7, 289—91. Whittaker, J.O. and R.D. Meade (1968), ”Retention of Opinion Change as a Function of Differential Source Credibility: a cross—cultural study,” International Journal of Psychology, 3, 103—108. Zajonc, R.B. (1968), ”Attitudinal Effect of Mere Exposure," Journal of Personality and Social Psychology, 9, (Part 2), 1-28. (1980), "Feeling and Thinking: preferences need no inferences," American Ps cholo ist, 35, 151— 175. (1986), "Basic Mechanisms of Preference Formation," in The Role of Affect in Consumer Behavior, Peterson, Hoyer, and Wilson, (eds.), Lexington, MA: Lexington Books, 1—16. and H. Markus (1982), ”Affective and Cognitive Factors in Preferences," Journal of Consumer Research, 9, 123—31. Zielske, H.A. (1982), ”Does Day—After Recall Penalize 'Feeling' Ads?," Journal of Advertisin Research, 22 (February/March), 19-24. 127 Table x Attitude toward the advertisement Level of right after four days ten days overall ad ad exposure later later strongly 1.104 .805 —.117 .597 positive (.971) (.706) (1.315) (1.130) weakly .831 .156 -.636 .117 positive (.666) (1.230) (1.043) (1.150) neutral -.039 -.260 -.481 —.260 (1.097) (1.032) ( .759) ( .961) weakly —1.234 «.779 —.571 —.861 negative ( .551) (1.457) (1.244) (1.149) strongly -1.987 -1.779 —1.039 —1.602 negative ( .621) ( .735) (1.071) ( .905) average - 265 — 371 — 569 * The numbers represent mean values of A“. ** The numbers in the parentheses represents standard deviations. "rrtrrriirrrrrr