The Power of : Predicting Intention

JON D. MORRIS This robust structural modeling study, with over 23,000 responses to 240 advertising The University of Florida messages, found that affect when measured by a visuai measure of emotional [email protected] response dominates over cognition for predicting conative attitude and action. CHONGMOO WOO The University of Florida chongmoo@uf I .ed u FOR DECADES marketing and advertising profes- require cognitive processing. Thus, one clear sionals and researchers have been struggling with solution to these issues would be the development JAMES A. GEASON two important questions: What is more predictive and use of a nonverbal measure of affect. A non- The University of Florida of consumer intent—thoughts or ? And, in verbal measure would offer the potential for rep- [email protected] the tripartite of cognitive, affective, conative atti- resenting attitudinal properties without cognitive tudes does cognitive attitude dominate and does processing. JOOYOUNG KIM it mediate the relationship between affect and The University of Florida intent? AFFECTIVE AND COGNITIVE-BASED ATTITUDE [email protected] Affect is clearly one component of attitude and a For many years, there was a tendency of focusing force in persuasion. Petty and Cacioppo (1981) on cognitive-based attitude, suggesting that, with have defined attitude as, "a general and enduring advertising involvement, cognition predominates positive or negative about some person, over affective processing and that affective reac- object or issue." Although this definition obvi- tions are always mediated by cognition (Green- ously assigns an affective component to attitude, wald and Leavitt, 1984; Tsal, 1985). In fact, the persuasion research has been dominated by the derivation and strength of the attitude toward the message learning approach, assigning the affective ad (Aad) process is based on the relationship be- processes a relatively minor role. This is surprising tween attitude toward the ad and attitude toward given the extensive use of emotional appeals in the brand (Abd), and the determination that Abd advertising. Thus the questions remain: Would predicts purchase intention (Mitchell and Olson, better prediction of behavior be achieved if more 1981; Lutz, MacKenzie, and Belch, 1983; MacKen- emphasis were placed on the nonrational and zie and Lutz, 1986). Fishbein (Fishbein and emotional determinants of behavior? Would atti- Middlestadt, 1995) also heralded the notion of cog- tude research improve if affect were shown to be a niti\'e-based attitude by suggesting that a consum- major component of conative attitude? er's attitude is a function of (cognitive) beliefs and those beliefs predict intentions of behavior. One possible reason for this enigma might be the measurement tools. Attitude measures rely al- Studies examining the role and relationship of most entirely on cognitive scales, requiring ad- as the mediator of responses to advertis- vanced verbal skills and a cerebral analysis by re- ing (Edell and Burke, 1987; Holbrook and Batra, spondents of surveys. These methods rely on the 1987), however, have found that cognition can assumption that respondents are capable of ac- drive affect. In fact, some researchers (Brown and cessing the individual components of attitudes, Stayman, 1992; Cohen and Areni, 1991; Petty et al., judge their feelings, and translate them into re- 1991) have argued that affect can directly influence sponses on typical Likert scales. Although verbal attitude and that cognitive-based models fail to measures can represent many distinct aspects of properly measure feelings associated with the emotion, they do not produce a true dichotomy sources of information (Edell and Burke, 1987; between affect and cognition, because they too Schwarz, 1997). Failing to understand the role of

May. June 2002 JDUnilflL OFflDUERTISlOG fiESEIIIICH 7 PREDICTING INTENTION

by focusing on cognitive process tion, and their antecedents: cognitive, af- pendent, bipolar dimensions (, only impedes the ability for understand- fective, conative measures (Hilgard, 1980; , or Dominance) first proposed ing various consumer behaviors (Allen, McGuire, 1989), in which previous studies by Osgood, Suci, and Tannenbaum Machleit, and Kleine, 1992). have produced conflicting results and (1957) (evaluation, activity, and potency) The introduction of emotional response conclusions about their relationships. and later refined by Mehrabian and Rus- adds a more robust paradigm for analyz- So we set out to determine which of the sell (1974) (pleasure, arousal, and domi- ing advertisements (Batra and Ray, 1986). previously reviewed variables hold the nance). In this process, all emotional re- The Advertising Research Founda- answer to intention and which are diag- sponses are combinations, in varying de- tion copy-testing project (Brown and Stay- nostic as well as predictive. We had a rela- grees, of these three basic emotions man, 1992; Haley and Baldinger, 1991) tively natural setting at our disposal: a se- (Russell and Mehrabian, 1977). Evidence found that liking of an advertisement is a ries of monatic copytests conducted shows that three independent, bipolar di- good predictor of effectiveness. The di- within a pool of balanced clutter televi- mensions reliably and sufficiently define rectness of the liking questions is clear, sion commercials. Most of the tests, de- all emotional states (Mehrabian and Rus- sell, 1974). but more insightful attitudinal informa- scribed in detail in the next section, in- tion toward the advertisement can be cluded samples from between 230 to 280 Pleasure/displeasure ranges from ex- leamed by expanding the measurement respondents each, from various demo- treme to extreme unhappiness. beyond the simple valance score (Allen, graphic backgrounds. AdSAM®, a non- Arousal/nonarousal constitutes a physi- Machleit, and Kleine, 1992; Holbrook and verbal emotional response modeling atti- ological continuum connoting a level of Hirschman, 1982). In fact, the ARF project tude toward the ad (like)—A,,^^, cognitive, physical activity, mental alertness, or fren- found fhat "emotions can have a direct and conative measures were used in the zied excitement at one extreme, with inac- influence on behavior tbat is not captured analysis. tivity, mental unalertness, or sleep at the or summed up by attitude judgments" other end. Dominance/submissiveness re- (Allen, Machleit, and Kleine, 1992). In Benefits of using a nonverbal fers to a feeling of power, control, or in- addition, reviews of the role of affect in measure: AdSAM- fluence versus the inability to influence a marketing suggest that affect is not depen- AdSAM-''' is based on the Self-Assessment situation or a feeling of lack of control. dent on cognitive variables (Machleit and Manikin [SAM] (Lang, 1980) and was de- Subjects use the PAD scales to report how Wilson, 1983). veloped to measure emotional response to they feel (Mehrabian and de Wetter, 1987). Further support for the influence of af- marketing communications stimuli. Ad- Because the three dimensional PAD ap- fect has been found in studies of , SAM^i' is a research tool that employs a proach is capable of characterizing diverse (Petty, Schumann, Richman, and Strath- database of 232 emotional adjectives, emotional responses in consumption situ- man, 1993), judgment (Pham, Cohen, scored with SAM, to gain insight and di- ations (Holbrook and Batra, 1988; Mehra- Prancejus, and Hughes, 2001), susceptibil- agnose the relationships among attitude, bian and Russell, 1974), it was used in the ity (Fabrigar and Petty, 1999), and studies cognition, brand , and purchase in- present study. Verbal emotional response linking affect and behavioral prediction tention. In this study, AdSAM^, or non- measures, however, are difficult to em- (Smith, Haugtvedt, and Petty, 1994). verbal affective scores from advertising ploy in advertising research. When adjec- These call for additional research to deter- copytests were compared to the cognitive tive checklists or semantic differential mine the role of affect and to find methods scores (Morris, 1995). Purchase intention scales are used to assess emotional re- for eliminating the measurement bias as- and brand interest comprised the conative sponse, the precise meaning of the emo- sociated with affect measures that rely on measures and served as the dependent tional words may vary from person to cognitive techniques to assess emotions variables. A structural equation model person. Por example, or may (Erevelles, 1998). was used to examine the relationships be- mean one emotion to one person but tween cognitive and affective attitude and something slightly different to someone conative attitude. else. This may vary the outcome of the THE PRESENT STUDY subject's real emotional response. Also With this in mind, we set out to examine SAM is a graphic character that follows pnjblematic are the use of open-ended the relationships among the key variables the PAD theory of affective response. This questions that request respondents de- that surround communications and con- theory adequately describes the full spec- scribe their emotional responses bo the ad- sumer activity, namely: attitude, inten- trum of human emotions in three inde-

8 JDURnflL OF flOUERTlSlfie HESEflRCH May . June 2002 PREDICTING INTEMTION

vertisements {Stout and Rust, 1986; Stout been criticized for attempting to general- categories, advertising media, and num- and Leckenby, 1986). Both approaches re- ize student samples to general popula- ber of advertisements tested are listed in quire a significant amount of cognitive tions (Brown and Stayman, 1992), and in- Table 1). The majority of these surveys processing. In contrast, the nonverbal deed we found, in earlier studies, that stu- were mail intercept studies, and target measure, SAM, eliminates the cognitive dent and nonstudent samples produce qualified respondents were randomly as- processing associated with verbal mea- different results in cross-cultural analyses signed to treatment cells. Although, the sures (Edell and Burke, 1987) and is quick of emotional response (Morris, Bradley, samples might be deemed less represen- and simple to use (Morris and Waine, Sutherland, and Wei, 1993). In addition, tative, since they were chosen using mall 1994; Lang, 1980). Correlations of .937 for the size of the sample for most attitude/ intercepts, a form of non-probability sam- pleasure, .938 for arousal, and .660 for behavior studies has also been criticized pling, this method of gathering data has dominance were found between ratings (Brown and Stayman, 1992). In fact, nearly been shown to be efficient (Bush and Hair, generated by SAM and by the semantic every study that we reviewed for this 1985) and representative when compared differential scales used by Mehrabian analysis has had samples of less than 120. (}r Goodness-of-fit tests) to randomly se- and Russell (Morris, Bradley, Sutherland, lected sample data (Vincent, Thompson, and Wei, 1993; Morris, Bradley, Lang, SAMPLE AND PROCEDURE and Pagan, 2001). In this study, the and Waine, 1992; Morris and Waine, The purpose of this study is to report on sample sizes were of such magnitude and 1994). SAM uses a nine-point scale for the analysis of the relationship among so geographically varied that the chance each of the dimensions. On each of the measures of cognitive, affective, and of sampling error has been greatly three scales, respondents were required to conative attitude in response to various reduced. mark the dot below the manikin or be- television, radio, and print advertise- tween the manikins that best represented ments. The sample was comprised of DEPENDENT VARIABLE AND MEASURES their feelings after seeing the advertise- 23,168 respondent evaluations of 240 ad- Following the exposure to the advertise- ment. (See Figure 1.) vertisements in 13 product categories. Re- ments, subjects responded to multiple- spondents in each test scored questions of item scales assessing cognitive, affective, Benefits of nonstudent sampling purchase intent or intent to visit the and conative attitude and to demographic Though there have been several contro- dealer, change in brand interest, as well as questions (see Table 2). Of interest to this versial issues regarding methodological cognitive and affective attitude. Affective analysis were studies in which, following problems in attitude research (e.g., Fish- attitude was measured by AdSAM'*^. exposure to the advertisements, respon- bein and Middlestadt, 1995; Schwarz, During the course of a multiyear con- dents were either asked about (heir likeli- 1997), two overriding factors that may af- tract, a major tj. S. copy-testing firm col- hood of buying, or, depending on the fect the outcome of most attitude/ lected AdSAM® emotional response data, product, visiting the stores. In many cases, intention studies are the quality and quan- cognitive, and conative data across a num- a question about the change in brand in- tity of the sample. Many studies have ber of product categories (the product terest also was asked. The "intent" and the brand interest questions were measured on five-point ordinal scales. For the cognitive attitude measures, Pleasure items developed by the copy-testing firm were employed to gauge belicvability and knowledge. AdSAM"'* was utilized as an Arousal affective attitude measure. In addition, a n\-e-point, Aad measure was employed to measure liking of the advertisement. Ac- Dominance cordingly, the raw scores of the cognitive, affective, and conative attitude scale items were used as indicators of those con- structs in the analysis described tn the fol- Figure 1 SAM (Self-Assessment Manikin) lowing sections.

May . June 2002 JOllRflHL DF RDUEfiTISIfie RESKH PREDICTING INTENTION

item had low communality with omni- Independent Variable and Number of Responses P''''''' '"^^^"S- since dominance often proves helpful as a diagnostic too!, and Independent experience has shown its value is related Variables Product Category Ad Media Copy-Testing Format to the vicariousness of the experience. Alcoholic Beverage TV Finished Film dominance was included in the analysis. "' A scree plot inspection and a forced three- Apparel Radio Rough Finished Film factor extraction with varimax rotation were performed based on the trichotomy ^^J}}^_ .1^....!!^^^ theory of attitude structure. Computers 1.5 Page In summary, the seven items were Fast Food Doublefage Spread" strongly loaded on the "correct" factors (two cognition items loaded strongly on Food ^ , , ., cognitive attitude, the three affect items ^i^^T^R^^'^^ loaded on affective attitude, and two Other Financial Institutions conative items loaded on conative atti- Pharmaceutical tude). Thus, there is tentative evidence for the convergent and discriminate validity Restaurant " ^ of the seven items used in the study. Retail Stores ^. ^ ^ u • u, u uu ,. Given that there might be a concern that I^.^.P...^!^^..?!^!!'.^.^.'?.'.^^ using identical items to measure specific Total 13 Product Categories 3 Media 6 Formats attitude might introduce unwanted mea- 83 Brands 240 Advertisements surement error (Heath and Gaeth, 1994), the demonstration of convergent and dis- 28.720 Responses 34,602 32,911 ^ criminate validity is important. The final three factors solution account for 64.20 VALIDITY CHECK followed by varimax rotation revealed percent of the original variance. The responses to the eight items in the only two inter-correlated factors with ei- cognitive, affective, and conative scales eenvalues greater than 1.0. The Aad mea- ^ 6 & RESULTS were first factor analyzed. Due to the tri- sure loaded equally on the affect and the partite character of attitude {Hilgard, conative measure, intent. The variable was Descriptive statistics and 1980), it is important that evidence for the considered confounded and eliminated. assumption check three individual dimensions be found. Ini- The three AdSAM* scales loaded to- „ , ,^ B. , The descriptive results provide a sum- tially, a principal component extraction gether, albeit the AdSAM® dominance mary of variables that are important in J ADI F 2 subsequent analyses. The primary inde- „ . . — . . ^ , , , pendent variable in this study is affective Descnptive Statistics of Measurement Item • ^ . u u • attitude. Across the three measurement Measures Item Mean S.D. N 'tems, mean score for affective measures „" '" ' . , ., , , ,^. > r^,- -»^ ^-^ -lo/^ varied from a low of 4.95 to a high of 6.76 Cognitive Attitude Knowledge (Ql) .85 .36 23,160 ° ona nine-unit bipolar scale ranging from 1 Belief (Q2) .86 .35 23,160 t, o 9n. Tabl-r u.e 2-. and-. TablT ui e 3^ displa.i- , y Adescrip- Affective Attitude .!^.'.®.^.^.'^.'^® .'.9."^^ ?:7,? h^^. ^.?:?:.^.9. tive statistics and the correlation matrix Arousal (Q4) 4.95 2.29 23,160 used as input data for LISREL. As shown Dominance (Q5) 6.06 2.23 23,160 '" """able 3, the total seven scale items were inter-correlated. These results showed Conative Attitude Brand Interest Change (Q6) 3.73 1.09 23,160 that items measuring the same construct, PM^C':^3Se..l':'te"t'0".!QZ.) 3.57 1.11 23,1^0 e.g., conative attitude, were more highly

10 JOUeORL OF flflOEmilG RESEfleCH May . June 2002 PREDICTING INTENTION

TABLE 3 tions among the seven measurement Correlation Matrix of Measurement Items items. Nonredundant residuals with abso- lute values over .05 were 47 percent. The Ql Q2 Q3 Q4 Q5 Q6 Q7 model demonstrates a good model fit be- Ql 1.00 tween observed correlation and assumed Q2 .23* 1.00 correlation since the nonredundant re- siduals with absolute values over .05 is Q3 .15* .10* 1.00 below 50 percent. Q4 .11* .08* .38* 1.00 Q5 .02* .00 .15* .03* 1.00 Structural equation modeling Q6 .21* .16* .41* .30* .08* 1.00 Before a comparison of the coefficients of the cognitive-affective-conative (tri- Q7 .18* .14* .37* .27* .07* .70* 1.00 chotomy) attitude path, the psychometric "p < .05 properties (e.g., dimensionality and reli- ability) of the measures were examined correlated with each other than they were sample ratio was 1 to 5,849 and satisfied again. The three factors derived via the with any of the other items. the criteria suggested by Nunnally (1978). principal components analysis dominated Kaiser-Meyer-Olkin's measure of sam- the solution and reflected the structure of Assumption check pling adequacy was .69, and Bartlett's test the responses. An exploratory procedure, Prior to the main analysis, several under- of sphericity index also showed signifi- LISREL 8.30 (Joreskog and Sorbom, 1993) lying assumptions for structural equation cant ;j-value at the .05 significance level. was used as a confirmatory factor test of modeling were checked. The underlying Thus, there was substantial evidence for the trichotomy solution of the seven atti- assumptions for the SEM analysis were the planned factoring of the seven items tude measurement items. The dimension- similar to the factor analysis: an adequate used in the study (Kaiser, 1974). ality of the trichotomy model was as- variable-to-sample ratio, normality, lin- Extracted communalities were .41 to .97 sessed through an examination of the as- earity, no extreme multicollinearity, and across all measurement items, demon- sociated fit indices. The LISREL indices sampling adequacy (Hair, Anderson, Tan- strating that there were no extreme mul- [i.e., RMSEA (Root Mean Square Error of tham, and Black, 1998). The variable-to- ticollinearity or strong linear combina- Approximation), NFl (Normed Fit Index), NNFI (Non-Normed Eit Index), GEI (Goodness of Fit Index), and AGFI (Ad- AdSAM Pleasure <- -42 justed Goodness of Eit Index)] all pro- vided evidence of acceptable levels of fit AdSAM Arousal .72 for the cognitive-affective-conative atti- tude model. Overall goodness-of-fit indi- AdSAM Dominance ces (RMSEA = .03, NFI = .99, NNEi = .98, GFi = 1.00, AGFI = .99) were satisfactory (Bagozzi and Yi, 1988), demonstrating that the model is statistically plausible and can reasonably reproduce the correlation ma- Brand [merest .34 trix. In this stable environment, three di- rect and one indirect path coefficients Intent lo Buy/Visit ^ .29 were created. (See Figure 2.) LISREL 8.30 was also used For a simul- Belief ,84 taneous estimation of the measurement and structural model. All indicators Figure 2 SEM Path Diagram of Cognitive-Affective-Conative loaded (exclusively) on the appropriate la- Attitude tent constructs, and all /-values associated

May . June 2002 JDUHnflL OF HOUERTISIHG flESEIIfiCH li PREDICTING INTENTION

with those loadings were statistically sig- the direct affective-conative attitude path tude, several regression analyses were nificant (p <.O5). The LISREL results in coefficient was the highest and exceeded conducted, by product category and me- Figure 2 show that, as hypothesized, gen- the total (direct + indirect) path coeffi- dia, to determine sources of these differ- eral evaluation of conative attitude is posi- cient of cognitive-conative attitude spec- ences. The cognition scales were used as tively predicted by cognitive and affective trum. These causal sequences of atti- the independent variables for one model attitude. The direct path coefficients from tudes leading to purchase intention and and the affective scales for the other cognitive attitude to affective attitude brand interest are importajit measures of model. The estimates were made for each (.25), cognitive to conative attitude (.28), advertising effectiveness (Deogun and Be- conative attitude: purchase intent and and affective to conative attitude (.49) atty, 1998). brand interest. This allowed us to com- were significant. In addition, the indirect pare the R^s of the two factors across path coefficient from cognitive attitude to Additional regression analyses product category/advertising medium/ conative attitude via mediation of affec- Since the SEM demonstrates a stronger advertisement copy-testing format condi- tive attitude (.25 x .49 ^ .12) was signifi- link between affect and conative attitude tions. The results of stepwise regressions cant. Among the three path coefficients, than between cognition and conative atti- used to estimate the total variances asso- ciated with the variable groups are re- TABLE 4 ported in Tables 4 and 5. Variations of Cognitive-Affective to Conative Attitude by Table 4 shows the effect of cognitive Product CategojT ^ ^"^ "^^"'^''^" '"^*^^^"''" '^"''"'^' '"''"'^" for each product category. Column 3 of Affective Attitude Cognitive Table 4 reports the R^s of affective attitude -Pictoriai AdSAM Attitude-Verbai regressed on conative attitude by product Product Category Conative Attitude Measures' R^* Measures' ff* category. Column 4 reports R^s for cogni- Aicohol Beverage Brand Interest 30.10** 6.60 tive attitude regressed on conative atti- ^ •' '"'"'"' !,o"cn T on tude by product category. All regressions Purchase Intent 12.60 7.20 •' ^ ^ ' and betas are significant at ^ < .05. Affec- Apparel Purchase Intent 14.40 4.70 , i • - i f in t f Autos Brand interest 20.30 6.80 ^i^g variance in conative attitude to adver- Purchase intent 16.40 5.20 tisements in the various product catego- Banks Purchase Intent 19.70 3.30 '^''^^- Cognitive attitude explains 2 to 13 percent of conative attitude for the same Computers Brand Interest 12.60 5.30 ^ . ^ „ product categories. Overall, cognitive and ^^l".^.*^.^.?.^..!!??.^.?! ..'.„.. .: affective attitude is positively related to Fast Food Purchase Intent 10.90 2.80 conative attitude and the two dimensions Food Purchase intent 17.30 6.50 explain different amounts of variance to- '".'" '• n u , * . o nr. o on ward the dependent variable. The picto- 011 Companies Purchase Intent 8.60 2.30 ^ ^ rial, affective measure AdSAM® had more Financial Institutions Brand Interest 17.00 4.90 .u ^u u i explanatory power than the verbai cogni- ^l^!'.'?.'^.^.?.^..!!^!".^.'?.! }:^:?P.. .'3\:. tive measures across 12 of the 13 product Pharmaceutical Brand Interest 18.90 5.80 categories. Purchase Intent 15.40 7.90 The relationship between the advertise '"''''" -, ^^ ing medium and attitude variables was Restaurant Purchase intent 19.30 7.20 ^ also assessed (see Table 5). All R s of !^ai!.Stor^s Br^ndjnterest 20.40 13.00 ^^^ regression equation in Columns 3 and Purchase Intent 18.40 8.60 4 ^^.[-o significant (/' < .05). For two of Telephone Companies Purchase Intent 16.20 2.60 the three media, affective attitude was more predictive of conative attitude. Af- *Adjnslcd R- ofMidliptt' Rc-irei^f-ion n-Hli diiuiiiiy viiruMe ^*Ail R's wercfrou, the sigmfica.,1 F let. ' fective attitude explains 15 to 22 percent 12 JOURURL OF RDUERTISIOG Kmm May . June 2002 PREDiCTING INTENTION

TABLE 5 Table 6 shows the effects of cognitive Variations of Cognitive-Affective to Conative Attitude in and affective attitude on conative attitude in each of the six advertisement copy- Ad Medium testing formats. It summarizes 24 multiple Affective Attitude Cognitive regressions, 2 for each conative attitude -Pictorial AdSAM Attitude-Verbal domain. All 24 regressions are significant Ad Medium Conative Attitude Measures' ff^ Measures' ft'* at p <.O5. Affective attitudes explain 5 to 37 percent of the variance in conative at- TV Brand Interest 19.70** 6.60 titude hy advertisement copy-testing for- Purchase Intent 15.10 4.80 mat. Cognitive attitudes explain 0 to 13 Radio Purchase Intent 22.20 25.50 percent of the conative attitude. Print Brand Interest 20.90 0.30 As conceptualized by the SEM analysis, Purchase Intent 18.50 10.80 the additional regression analyses indicate that cognitive and affective attitude are as- 'Aiijustcil R"^ of MuUipk Regression with dummy variabk *M// R-s wen-from the ^igiiificmil F h'sl. sociated with conative attitude, but that affective attitude has more explanatory power toward conative attitude in all but of the variance in conative attitude to ad- egory, which contained less than 1 percent one product category, in all advertising vertisements across media. Cognitive atti- of the responses. In addition, cognitive at- media except radio and in all copy-testing tude explains 0 to 26 percent of conative titude explained only 0.3 percent of the formats. attitude for the three media categories. variance toward conative attitude in the This is mostly driven by the radio cat- print media. DISCUSSiON TABLE 6 The tripartite of human experience of cog- nitive, affective, conativo attitude or Variations of Cognitive-Affective to Conative Attitude in Ad thought, feeling, and planned action, al- Copy Testing Format though not logically compelling, is preva- lent in Indo-European thought (being Affective Attitude Cognitive found in Hellenic, Zoroastrian, and Hindu Ad Copy Testing -Pictorial ADSAM® Attitude-Verbal philosophy) to suggest that it corresponds Format Conative Attitude Measures' ff^ Measures' f^* to something basic in our way of concep- Finished Film Brand Interest 17.40** 6.10 tualization {McGuire, 1989; pp. 40-41). Purchase Intent 13.60 4.70 Perhaps the greatest enigma is the rela- Rough Finished Film Brand Interest 22.20 8.60 tionship of these attitudes. Heretofore, Purchase Intent 17.20 6.00 many researchers, using the variable "lik- ing of an advertisement" (Aad) to mea- Animatic Brand Interest 30.90 7.30 sure affect, have insisted that a direct link Purchase Intent 27.20 6.20 exists between affect and cognition, that Full-page Brand Interest 36.90 0.30 cognition predominates over affective Purchase Intent 30.20 13.30 processing, and that affective reactions are always mediated by cognition (Green- 1.5-Page Brand Interest 7.60 3.90 wald and Leavitt, 1984; Tsal, 1985). Even Purchase Intent 4.80 4.40 more curious is the insistence that cogni- Double-Page Spread Brand Interest 9.20 6.00 tion and affect are separate and distinct Purchase Intent 7.60 4.70 elements to persuasion (Petty and Ca-

'Aiiju^lcd R-" of Multiple Regre^^ioii wilh diiiinin/ viiriiil'li' cioppo, 1981; Mitchell, 1986; Petty, Schu- "All R^s were from the si};iiifica)it F test. mann, Richman,and Strathman, 1993). We

May . June 2002 JOUROflL OF flDUEIlTISIflG RESEflflCH 13 PREDICTING INTENTION

Emotional response is a powerful predictor of intention mediate consumer attitude or the ability to pfocess advertising messages, but and brand attitude, and ... is a valuable tool for strate- clearly this Qiodel has helped to show the importance of affective attitude. (2& gic planning, message testing, and brand tracking.

JON D. IVIonfii5 is a professor of advertising in the view these arguments as extreme and our the Self-Assessment Manikin, are direct College of Journalism and Communications at the research clearly found that cognition and emotional reactions since the measure is University of Florida. Previously, he wori^ed for several affect are interdependent and that an nonverbal. Emotions (i.e., thankful, victo- advertising agencies, including Nicholson-Morris in emotional response measure allows for af- rious, unexcited, or embarrassed) as deter- Louisvijie, KY. and Doyie Dane Bernbach and Dancer fective elaboration. mined with AdSAM''^, are both descrip- Fitzgerald Sample in New Yorii City. He earned his In this robust study of over 23,000 re- tive and directive. This information is Ph.D. from the University of Florida. His research has sponses to 240 advertising messages, we missing with any Likert scale of "like the appeared in the Journal of Advertising Research, found that affect dominates over cogni- ad." Moreover, this study has shown that Educational Technology, the internationai Journal of tion for predicting conative attitude and these emotional reactions are strongly instructional Media, and in the Proceedings of tiie action. Moreover, we learned that liking predictive of behavioral intention. American Academy of Advertising and The Association (Aad) may be a confounded variable, that The tripartite model, cognitive, affec- for Consumer Research, among others. He deveioped affect is not mediated by cognition, and tive, and conative attitude, has been used a model, called AdSAM, for analyzing emotional that brand attitude (interest) is not neces- pervasively in psychology (Hilgard, 1980), response to marketing communications. sarily a precursor to intention (intent to nowhere more than in attitude research buy). Affect as measured by emotional re- (McGuire, 1989), but, in this study, the re- CHONGMOO WOO is an advertising Ph.D. student sponse was sbown to be the dominant di- lationship of these dimensions to one an- co-majoring in statistics at the University of Florida. mension, accounting for more of the vari- other has been shown to be different than iHe received his B.A. in advertising and PR, M.A. in ance toward conative attitude than cogni- previously thought. Affective attitude as adverttsing and PR at the Chung-Ang University, Seoul, tion. Emotional response, as measured by measured by emotional response offers an Korea, and M.A. in advertising at the University of SAM pleasure, arousal, and dominance, alternate view of this paradigm. Emo- Florida. Previousiy he was an advertising lecturer at had a stronger relationship to affective at- tional response offers a composite mea- the Kangnam University and Sookmyung Women's titude than the information-seeking vari- sure for predicting conative attitude University in Korea. His research interests are in ables knowledge and belief toward (brand interest and purchase) as well as advertising copy testing, probability distribution for conative attitude. Contrary to some previ- separate indices of affect for diagnostic media selection, measures of advertising affect, and ous assertions that cognition is the domi- purposes. logit modeling. nant variable for predicting intention, Researchers should become more confi- when compared to affect, our results show dent that measuring emotions would JAMES A. GEASON is a doctoral candidate majoring in that affect accounts for almost twice the help to determine consumer intentions. advertising in the College of Journalism and variance toward conative attitude. Emo- Marketers, who are skeptical about the Communications at the University ot Flonda. He tional response is a powerful predictor of importance of affect in the marketing received his B.A. from Washington and Lee University intention and brand attitude, and given communications mix, should have those in Lexington. Virginia, and an M.A. in Mass the diagnostic capabilities that are missing feelings allayed. Communication from the University of Florida. He in other measures of affect (Aad), it is a Beyond cognitive-affective-conative at- teaches marketing and advertising at Santa Fe valuable tool for strategic planning, mes- titude research, there is a need to explore Community College in Gainesville, Florida. His sage testing, and brand tracking. the broader information-processing impli- research interest is in the area of emotional response Unlike attitude toward the ad (Aad), cations of this study including the contex- measurement. emotional response offers a direct method tual effects (Norris and Colman, 1992; of analyzing the complex feelings that Page, Thorson, and Heido, 1990) and con- KIM is an advertising Ph.D. student at the comprise human reactions to advertising. sumer involvement. The tripartite model University of Florida. He received his M.A. in integrated In fact, the responses gathered with SAM, is but one of multiple constructs that may Marketing Communications from the University of

14 JDUBOflL OFflDUERTISinG HESEIIfiCfl May . June 2002 PREDICTING INTENTICN

Colorado-Boulder, and B.A. In economics from Hong-Ik C08B-WALGREEN C. J., C. A. RUBLE, and N. Analysis, 5th ed. Upper Saddle River, NJ: University in Seoul Korea. DONTHU. "Brand Equity, Brand Preference, ad Prentice Hall, 1998. Purchase Intent." journal of Advertising 24, 3 (1995): 25-40. HALEY, RUSSELL I., and ALLEN. L. BALDINGER. REFERENCES "The ARF Copy Research Validity Project." KN, J. B., and C. S. ARENI. "Affect and lournal of Advertising Research 31, 2 (1991): 11- Consumer Behavior." In Handbook of Consumer Ai.t.EN, C. T., K. A. MACHLEIT, and S. KLIIINE. 32. "A Comparison of Attitudes and Emotions as Behavior, A. Robertson and H. Kassarjian eds. Predictors of Behavior at Diverse Leveis of Englewood Cliffs, NJ: Prentice-Hall, 1991. Behavioral Experience." jaiirua! of Consiuncr HEATH, TIMOTHY B., and G. J. GAETH. "Theory Resenrch 18, 4 (1992): 493-504. DEOGUN, NIKHIL, and S. G. BHATTY. "Coke's and Method in the Study of Ad and Brand Marketing Chief, Sergio Zyman, to Quit." Attitudes: Toward a Systemic Model." In At- Wall Street ]ournal. March 19, 1998. tention, Attitude, and Affect in Response to Ad- BAGOZ/I, RICHARD D., and Y. Yi. "On the vertising, E. M. Clark, T. C. Brock, and D. W. Evaluation of Structural Equation Models." DROGE, C. "Shaping the Route to Attitude Stewart, eds. Hillsdale, NJ: Lawrence Erlbaum joumal of the Academy of Marketing Science 16, Change: Centra! versus Peripheral Processing Associates, 1994. 1 (1988): 74-94. through Comparative versus Noncomparative Advertising." joitrnal of Marketing Research 26, BAti?A, R., and M. L. RAV. "Affective Re- HILGARD, E. R. "The Trilogy of Mind: Cogni- 2 (1989): 193-204. sponses Mediating of Advertis- tion, , and Gonatlon," journal of the ing." journal af Consumer Research 13, 2 (1986): History of the Behavioral Sciences 16, 2 (1980): EDELL, JULIE A., and M. C. BURKK. "The Power 234-50. 107-17. of Feelings in Understanding Advertising Ef- fects." lournal of Consumer Research 14, 3 BoDUK H. O., D. BRIM3ERC, and E. COUPEY. (1987): 421-33. HOLBROOK, M. B., and E. C. HiRSCiiMAN. "The "Belief, Affect, and Attitude: Alternative Mod- Experiential Aspects of Consumption: Con- els of the Determinants of Attitude." journal of EREVELLES, S. "The Role of Affect in Market- sumer Fantasies, Feelings, and Fun." journal of Coimmier Psychalo^n 9, 1 (2000): 17-28. ing." Joiirnal of Business Research 42, 2 (1998): Consumer Research 9, 2 (1982): 132-140. 199-215. BROWN, STEViiN P., and D. M. STAYMAN"Ante- cedents and Consequences of Attitude toward , and R. BAIRA, "Assessing the Role of GAR, L. R. and R. E. PKTTY "The Role of the Ad: A Meta-analysis." journnl nf Consumer Enuitions As Mediators of Consumer Re- the Affective and Cognitive Bases of Attitudes Research 19 (1992): 34-51. sponse to Advertising." journal of Consuitier in Susceptibility to Affective and Cognitively Research 14, 3 (1987): 404-20. Based Persuasion." Persoiialiti/ and Social Psi/- M. C, and J. A. EDIELL. "The Impact of chology Bulletin 25, 3 (1999): 363-81.

Feelings on Ad-Based Affect and Cognition." JORKSKOG, KARL G., and D. SORBUM. LISREL 8: journa] of Marketing Research 26, 1 (1989): 69- N, M., and S. E. MIUDLESTADT. "Noncog- Structural Equation Modeling wilh the SIMPLIS 83. nitive Effects on Attitude Formation and Command Latigtiage. Ghicago, IL: Scientific Change: Fact or Artifact?" journal of Consumer Software International, Inc., 1993. BURTON, Scor, and D. R. LICHTENSTEINI. "The Psychology 4, 21 (1995): 181-202. Effect of Ad Claims and Ad Content on Atti- KAISER, H. F. "An Index of Factorial Simplic- tude toward the Advertisement." jounial of GREENWALD, A. G., and C. LEAVITT. "Audience ity." Psychametrika 39, 1 (1974): 31-36. Advertising 17, 1 (1988): 3-11. Involvement in Advertising: Four Levels." journal of Consumer Research 11, 1 (1984): 581-

BUSH, A. j., and J. F. HAIR. "An Assessment of 92. KRISNAN, H. S., and S. SHAPIRO. "Prospective the Mall-intercept as a Data Collection and Retrospective Memory for Intentions: A Method." journal of Marketing Research 22, 2 HAIR, JR., JOSEPH F., R. E. ANDERSON, R. L. Two-Component Approach." journal of Con- (1985): 158-67. TATHAM, and W. G. BLACK. Multivariate Data sumer Psychology 8, 2 (1999): 141-66.

May . June 2002 JDURflOL OF fIDUERTISIIlG RESERRCH 15 PREDICTING INTENTION

LANG, P. ]. "Behavioral Treatment and Bio- MEHRABIAN, A., and R. DE WETTER "Experi- Magazine Advertisements." journal of Adver- Behtwioml Assessment: Computer Applica- mental Test of an Emotion-based Approach to tising 21, 3 (1992): 37-46. tions in Technology." In Mental Health Care Fitting Brand Names to Products." journal of Delievciy Systems, J. B. Sidowski, J. H. John- Applied Psychology 11, 1 (1987): 125-30. NUNNALLY, j. C. Psychoinetric Theory, 2nd ed. son, and T. A. Williams, eds. Norwood, NJ: New York, NY: McGraw-Hill, 1978. Ablex, 1980. MiTCiiKM, A. A. "Tlie Effect of Verbal and Visual Components of Advertisenients and Osc,t«D, C. E., G. J. SUCI, and P. H. TANNEN- Brand Attitudes and Attitude toward the Ad- LORD K. R., M. U±, and P. L. SAUER. "The BAUM. Thf Measureinent of Meaning. Chicago, vertisement," journal of Consmner Research 13, Combined Influence Hypothesis: Central and IL: University of Illinois Press, 1957. 1 (1986): 12-24. Peripheral Antecedents of Attitude toward the Ad." joumal of Advertisijig 24, 1 (1995): PAGE, THOMAS ]., E. THORSON, and M. P. 73-84. , and J. C. OLSOK. "Are Product Attrib- HEIDO. "The Memory Impact of Commercials ute Beliefs the Only Mediator of Advertising Varying in Emotional Appeal and Product Effects on Brand Attitudes?" journal of Market- LUTZ, R. J., S. B. MACKHNZIL, and G. E. BELCH. Involvement." In Emotion in Advertising, Stuart ing Research 16, 1 (1981): 318-22. "Attitude toward the Ad as a Mediator of J. Agres, Julie A. Edell, and Tony Dubitsky, Advertising Effectiveness: Determinants and eds. New York, NY: Quorum Books, 1990. Consequences." In Adzmnces in Consumer Re- MORRIS, J. D. "Observations: SAM: The Self- search, Richard P. Bagozzi and Alice M. Ty- Assessment Manikin; An Efficient Cross-Cul- PETTY, R. E., and J. T. CACIOPPO. "Attitudes and bout, eds. Ann Arbor, Ml: Association for tural Measurement of Emotional Response." Persuasion: Classic and Contemporary Ap- Consumer Research, 1983. journal of Advertising Research 35, 6 (1995): 63- 68. proaches." Dubuque, IA: William C. Brown, 1981.

MACHLIHT K., and R. D. WII.'JO.M. "Emotional , and C. WAINE. "Managing the Cre- Feelings and Attitude Toward the Advertise- ative Effort: Pre-production and Post-produc- PFHY. R. E., D. W. SCHUMANN, S. A. RICHMAN, ment: The Roles of Brand Familiarity and tion Measures of Emotional Response." Work- and A. J. STRATHMAN. "Positive Mood and Repetition." joimtai of Advertising 17, 3 (1983): ing Paper, University of Florida, Gainesville, Persuasion: Different Roles for Affect Under 27-35. 1994. High- and Low-Elaboration Conditions." jour- nal of Personaliti/ and Social Psychology 64, 1 (1993): 5-20. MACKFNZIK, S. B., and R. J. LUTZ. "The Role of , M. BRADLHV, C.A. WAINE, and J. B. Attitude toward the Ad as a Mediatt)r of Ad- LANC. "Assessing Affective Reactions to Ad- vertising Effectiveness: A Test of Competing; vertisements with the Self-Assessment Mani- PETTY, R. E., E. GLEICHER, and S. M. BAKER. Explanations." journal of Mai'keting Research kin (SAM)." Southern Marketing Association "Multiple Roles for Affect in Persuasion." In 23, 2 (1986): 130-43. Conference, 1992. Emotion and Social judgments, J. P. Forgas, ed. Oxford, England: Pergamon, 1991.

F, W. J. "The Structure of Individual . , , }. SUTHERIAMD, and L. WEL Attitudes and Attitude Systems." In Attitude "Assessing Cross Cultura! Transferability of i'HAM, M. T., J. B. CtiHEN, J. W. PKACEIUS, and Structure and Function, Pratkanis, A. R., S. J. Standardized Global Advertising: An Emo- G. D. HUGHES. "Affect Monitoring and the Breckler, and A. G. GKKI:NWALU, eds.. Hills- tional Response Approach." Presented at the Primacy of Feelings in Judgment." journa! of dale, NJ: Lawrence Eribaum Associates, Pub- National Conference of the Association of Consumer Research 28, 2 (2001): 167-88. Ushers, 1989. Education in journalism and Mass Communi- cations, Kansas City, 1993. RUSSELL, J.A., and A. MFMRABIAN. "Evidence MrHRARlAN, A., and J.A. RussFi^i.. An Approach for a Three-factor Theory of Emotions." jour- to Environmental Psi/chology. Cambridge, MA: s, CLAIRE E., and A. M. COLMAN, "Con- nal of Research in Personality U, 3 (1977): 273- M.I.T. Press, 1974. text Effects on Recall and Recognition of

16 OF flDU[[lTISinG May . June 2002 PREDICTING INTENTION

SCHWARZ, N. "Moods and Attitude Judg- Measurement of Affect Enhance Behavioral Zajonc and Markus." journal of Consumer Re- ments: A Comment on Fishbein and Middle- Prediction?" Psychology and Marketing 11, 4 search 12, 3 (1985): 358-62. stadt." journal of Consumer Psychology 6, I (1994): 359-74. (1997): 93-9S.

STOUT, P. A., and J. D. LHCKI'NRY. "Measuring T, V. C, W. THOMi-yoN, and J. PAGAN. P, T. A. "Attitude toward the Ad as a Emotional Response to Advertising." journal- "Winter Migration of Retirees to Texas." In Mediator of Consiuner Brand Choice." journal of Advertising 15, 4 (1986): 35-42. Conference Paper Series of 2000-01 International of Advertising 10, 2 (1981): 9-]5. Conference on Applied Dcmograpln/. Center for , and R. T. Rtsr. "The Effect of Music Family and Demographic, College of Business SMITH, R. E., and W. R. SWINYARD. "Attitude- on Emotional Response to Advertising." In Administration, University of Texas-l*an Behavior Consistency; The Impact of Product Proceedings of the T986 Confere)ice of the Ameri- American, Edinberg, TX, 2000-01. Trial vs. Advertising." journa! of Marketing can Academy of Advertising, Ernest F. Larkin Research 2(1, 3 (1483): 257-67. ed., 1986.

ZEIUIN, D. M., and R. A. WK^II.V(.X)U. "Mea- SMITH, S. M., C: P. HAUGTVEDT, and R. E. TsA\,, Y. "On the Relationship between Cogni- suring Emotional Response." journal of Adver- PETTY. "Attitudes and Recycling: Does the tive and Affective Process: A Critique of tising Research 26, 5 (1986): 34-44.

Strategic Insig

Dick Westwood and the Strategy & Tactics team of professionals provide the insights your marketing needs:

- How your market actually works - What your brand's equities really ore - How to build a path to more growth

Whether it's a segmentation study, a brand image study, a brand extendibility study— or any other area of strat research— we provide the extra level of marketing insight that makes the work pay off for you.

Call us for an initial consultation. Bring your issues and objectives. We'll design a research program to bring your marketing to the next Level. Strategy & Tactics, Ud. 109 Forest Street • Stamford, CT 06901 Tel 203-977-8148 • Fax 203-977-8575 E-mail • [email protected]

May. June 2002 JOUROflL DF RDUERTISinG dESERRCH 17