C. Yoo, D. Macinnis /Journal of Business Research 58(2005)1397-1406 1401 were collected. Consumers subsequently indicated their judgment scale. Notably, these items are more"cognitive clings, evaluations of the ad, and beliefs about brand in nature, as they reflect judgment of the ad's credibility and attributes. Since some customers may have been familiar relevance(e.g, believable, realistic, valuable, and informa with the commercial even with these changes, we checked tive). As such, we regard them as indicators of our familiarity of the ad using five-point scale (not familiar at debility"construct. all-very familiar)at the end of the study. We retained one Belief strength(the b; factor) was also assessed Consum- hundred ninety subjects who indicated 1 or 2 on the ers' perceptions that the brand possessed five brand-relevant familiarity scale in subsequent ana attributes were measured on a 7-point scale ranging from “ extremely unlikely"to“ extremely likely”(bi).Thus, subjects rated the extent to which the advertised phone 4. Measures service(1)offered high quality service, (2)had 24-h operator service,(3)entailed considerable coverage across service Aad and Ab were each measured by four items (like areas, (4)offered a discount price, and (5)offered refunds for dislike, positive-negative, good-bad, and favorable wrong calls Evaluations of salient attributes of the advertised unfavorable)designed to assess consumers'attitudes toward brand(the ei factor)were also assessed via 7-point scales the brand and commercial, respectively. Each item was ranging from"extremely good"to"extremely bad"(Fish- scored on a seven-point semantic differential scale( Cron- bein and Ajzen, 1975). The belief strength indicators bachs alpha=96 and 93, respectively) represented a formative scale and the items were summed Feelings were assessed by a twenty-one item-scale. to form a composite index ubjects were told, "We are interested in your reactions to the ad, not how you describe it. Did this commercial make you feel.. ?"Responses to the twenty-one items were 5. Analysis and results scored on a seven-point scale(not at all-very much). Scale items were derived from cognitive response data collected in 5.1. Manipulation check a pilot study and feelings generated from Edell and Burke's (1987)feeling scales. Specifically, the items were designed As might be expected from the manipulation of ad to represent Edell and Burke's upbeat(e.g, delighted, execution format, an analysis of mean differences revealed elated, and stimulated), negative (e.g, sad, sorrowful, that consumers exposed to the emotional ad execution had distressed, irritated, angry, annoyed, offended, and significantly more positive feelings (r=4.82) than consum- depressed), and warm(e.g, sentimental, affectionate, ers exposed to the informational ad execution (X=3. 12: warmhearted, touched, and moved) feelings factors 1=6.32, p<.05). Moreover, consumers had significantly Subjects were also asked to indicate their evaluations of stronger beliefs when exposed to the informational ad the ad. Twenty items were designed to indicate ad (X=98.45)vS. the emotional ad execution (X=72. 84 evaluations. Subjects were told, "Now we are interested t=4.08, P<.05). These were the only variables for which in your evaluations or judgments about the ads, not just significant differences across the two experimental groups your feelings from the ad. Subjects indicated on seven were observed. Tests for homogeneity of variance showed point agreement scales the extent to which they agreed that that the variances of each variable across the two ad the adjective characterized the commercial. Items included conditions did not differ. Edell and Burke's (1987) evaluation(e.g, realistic, convincing, meaningful, valuable, and informative), activ- 5.2. Measure validation (e. g, exciting, energetic, amusing, playful, unique, and sepfoiatory factor analysis(principal component analy- and Sorbom, 1996). Because retaining each item as a imaginative), and gentleness(e. g, soothing, tender, lovely The model in Fig. I was analyzed by a maximum and gentle) factors. ikelihood estimation procedure using LISREL &(Joreskog sis) of feelings and evaluations revealed three factors with reflective indicator of its constructs would result in eigenvalues greater than one. The first factor was comprised identification problem, we used an adaptation of Anderson of positive emotions and emotion like evaluations that load and Gerbing's(1988) two-step approach to structural on Edell and Burke's activity and gentleness scales (e. g, equation modeling. soothing and amusing). The second represented negative First, we estimated a confirmatory measurement or factor emotions and negative emotion-like evaluations(e.g, sad, analytic model (Joreskog and Sorbom, 1996), including orrowful, distressed, irritated, angry, annoyed, offended, constructs of positive feelings, negative feelings, credibility, and depressed). The factors are subsequently labeled and beliefs. Summary statistics of confirmatory factor positive feelings and negative feelings. Notably, sad and analyses are summarized in Table 1 sorrowful feelings loaded on the positive feelings factor. We Second. we estimated the overall structural model return to this issue subsequently. The last factor is indicated Because retaining each item as a reflective indicator would by a set of items reflecting Edell and Burke's evaluation result in identification problems, we combined items into awere collected. Consumers subsequently indicated their feelings, evaluations of the ad, and beliefs about brand attributes. Since some customers may have been familiar with the commercial even with these changes, we checked familiarity of the ad using five-point scale (not familiar at all – very familiar) at the end of the study. We retained one hundred ninety subjects who indicated 1 or 2 on the familiarity scale in subsequent analyses. 4. Measures Aad and Ab were each measured by four items (like – dislike, positive – negative, good – bad, and favorable – unfavorable) designed to assess consumers’ attitudes toward the brand and commercial, respectively. Each item was scored on a seven-point semantic differential scale (Cron￾bach’s alpha =.96 and .93, respectively). Feelings were assessed by a twenty-one item-scale. Subjects were told, ‘‘We are interested in your reactions to the ad, not how you describe it. Did this commercial make you feel ... ?’’ Responses to the twenty-one items were scored on a seven-point scale (not at all – very much). Scale items were derived from cognitive response data collected in a pilot study and feelings generated from Edell and Burke’s (1987) feeling scales. Specifically, the items were designed to represent Edell and Burke’s upbeat (e.g., delighted, elated, and stimulated), negative (e.g., sad, sorrowful, distressed, irritated, angry, annoyed, offended, and depressed), and warm (e.g., sentimental, affectionate, warmhearted, touched, and moved) feelings factors. Subjects were also asked to indicate their evaluations of the ad. Twenty items were designed to indicate ad evaluations. Subjects were told, ‘‘Now we are interested in your evaluations or judgments about the ads, not just your feelings from the ad.’’ Subjects indicated on seven￾point agreement scales the extent to which they agreed that the adjective characterized the commercial. Items included Edell and Burke’s (1987) evaluation (e.g., realistic, convincing, meaningful, valuable, and informative), activ￾ity (e.g., exciting, energetic, amusing, playful, unique, and imaginative), and gentleness (e.g., soothing, tender, lovely, and gentle) factors. Exploratory factor analysis (principal component analy￾sis) of feelings and evaluations revealed three factors with eigenvalues greater than one. The first factor was comprised of positive emotions and emotion like evaluations that load on Edell and Burke’s activity and gentleness scales (e.g., soothing and amusing). The second represented negative emotions and negative emotion-like evaluations (e.g., sad, sorrowful, distressed, irritated, angry, annoyed, offended, and depressed). The factors are subsequently labeled positive feelings and negative feelings. Notably, sad and sorrowful feelings loaded on the positive feelings factor. We return to this issue subsequently. The last factor is indicated by a set of items reflecting Edell and Burke’s evaluation judgment scale. Notably, these items are more ‘‘cognitive’’ in nature, as they reflect judgment of the ad’s credibility and relevance (e.g., believable, realistic, valuable, and informa￾tive). As such, we regard them as indicators of our ‘‘credibility’’ construct. Belief strength (the bi factor) was also assessed. Consum￾ers’ perceptions that the brand possessed five brand-relevant attributes were measured on a 7-point scale ranging from ‘‘extremely unlikely’’ to ‘‘extremely likely’’ (bi). Thus, subjects rated the extent to which the advertised phone service (1) offered high quality service, (2) had 24-h operator service, (3) entailed considerable coverage across service areas, (4) offered a discount price, and (5) offered refunds for wrong calls. Evaluations of salient attributes of the advertised brand (the ei factor) were also assessed via 7-point scales ranging from ‘‘extremely good’’ to ‘‘extremely bad’’ (Fish￾bein and Ajzen, 1975). The belief strength indicators represented a formative scale and the items were summed to form a composite index. 5. Analysis and results 5.1. Manipulation check As might be expected from the manipulation of ad execution format, an analysis of mean differences revealed that consumers exposed to the emotional ad execution had significantly more positive feelings (X = 4.82) than consum￾ers exposed to the informational ad execution (X = 3.12; t = 6.32, p <.05). Moreover, consumers had significantly stronger beliefs when exposed to the informational ad (X = 98.45) vs. the emotional ad execution (X = 72.84, t = 4.08, p <.05). These were the only variables for which significant differences across the two experimental groups were observed. Tests for homogeneity of variance showed that the variances of each variable across the two ad conditions did not differ. 5.2. Measure validation The model in Fig. 1 was analyzed by a maximum likelihood estimation procedure using LISREL 8 (Joreskog and Sorbom, 1996). Because retaining each item as a reflective indicator of its constructs would result in identification problem, we used an adaptation of Anderson and Gerbing’s (1988) two-step approach to structural equation modeling. First, we estimated a confirmatory measurement or factor analytic model (Joreskog and Sorbom, 1996), including constructs of positive feelings, negative feelings, credibility, and beliefs. Summary statistics of confirmatory factor analyses are summarized in Table 1. Second, we estimated the overall structural model. Because retaining each item as a reflective indicator would result in identification problems, we combined items into a C. Yoo, D. MacInnis / Journal of Business Research 58 (2005) 1397 – 1406 1401