Creation and Validation of an Advertising Scale Based on the Individual Perception of the Emotional Or Informational Intent of the Advertisement

Philippe Jourdan, JPL Consulting/UPMF University
ABSTRACT - The purpose of this article is to propose a new classification scale for advertisements based on the individual perception of the informational or emotional intent of the Ad. After a review of the literature, we adopt the creation of a measurement scale based on induced affective and cognitive reactions. The method of creation and validation follows Churchill’s Paradigm to achieve a scale with 2 distinct constructs and 9 items. A classification of 10 different ads based on the scale grades illustrates by example the diagnostic value of the instrument. Caveats and managerial applications are also discussed.
[ to cite ]:
Philippe Jourdan (1999) ,"Creation and Validation of an Advertising Scale Based on the Individual Perception of the Emotional Or Informational Intent of the Advertisement", in NA - Advances in Consumer Research Volume 26, eds. Eric J. Arnould and Linda M. Scott, Provo, UT : Association for Consumer Research, Pages: 504-512.

Advances in Consumer Research Volume 26, 1999      Pages 504-512

CREATION AND VALIDATION OF AN ADVERTISING SCALE BASED ON THE INDIVIDUAL PERCEPTION OF THE EMOTIONAL OR INFORMATIONAL INTENT OF THE ADVERTISEMENT

Philippe Jourdan, JPL Consulting/UPMF University

[The author gratefully acknowledges helpful comments by Prof. Alain Jolibert, ACR editors, Eric J. Arnould and Linda Scott, as well as four anonymous reviewers. Special thanks to Valerie Jourdan for a critical examination of an earlier draft, to Nick Gilbert for a careful translation and to GfK Company for its contribution to the data collection.]

ABSTRACT -

The purpose of this article is to propose a new classification scale for advertisements based on the individual perception of the informational or emotional intent of the Ad. After a review of the literature, we adopt the creation of a measurement scale based on induced affective and cognitive reactions. The method of creation and validation follows Churchill’s Paradigm to achieve a scale with 2 distinct constructs and 9 items. A classification of 10 different ads based on the scale grades illustrates by example the diagnostic value of the instrument. Caveats and managerial applications are also discussed.

Work in advertising research is primarily concerned with measurement of persuasive capacity and explanation of its antecedents. In this respect, the numerous models which have been developed concentrate on the roles of sender and receiver. Numerous studies have indeed tried to characterize the objective informational or emotional content of an ad. Few however have turned on the way the receiver perceives subjectively the advertiser’s intent to persuade him through either informational or emotional cues. The objective of this article is therefore to construct a classificatin scale for ads according to their perceived informational or emotional characteristics. After a short review of the existing literature, we will adopt a definition for the constructs on which the generated items of the scale are based. We briefly present Churchill’s paradigm used as the validation method and afterwards summarize the results. Finally, a cluster analysis of 10 advertisements illustrates the diagnostic value of the scale use for advertising practitioners.

REVIEW OF METHODS MEASURING THE REACTION TO ADVERTISING STIMULI

Three types of classification methods have been put forward: an analytical classification by experts, an experimental investigation of emotions experienced and a categorisation of affective and cognitive responses.

Classification according to informational or emotional content

This method consists of classifying the advertising according to its dominant characteristic: informational or emotional. This is most often done by judges (Stewart and Furse, 1986; Pechmann and Stewart, 1989). Mixed results are frequently obtained: the dispersion of responses is in fact considerable when the same ad is evaluated. Two reasons explain this limited performance (and the fact that we reject it). The first is that most advertisements work on two levels: giving information by the central elements of the message they contain (copy) and arousing an emotion of varying nature and intensity through the elements of execution. Thus 64% of ads examined by 4 independent judges in research carried out by Stewart and Furse (1986) were placed in an intermediary category: "multiple stimuli" both rational and emotional. The second is linked to the individual differences in the way advertisements are processed. In this research, we distinguish between the types of advertising stimuli, which among these interest us more directly, and the individual or contextual variables that we attempt to control. In this way, the classification scale according to ad stimulus type should not take account of a dimension linked to the way the ad is processed: i.e. systematically versus heuristically (Pechmann and Stewart, 1989).

Classification by observation or by the emotional state report

We distinguish here between one approach centred on observation (by a judge) of the emotional reactions of the individual and a second that consists of questioning the nature and intensity of the emotions experienced. These two approaches have been adopted in numerous marketing studies (Batra and Ray, 1986; Aaker et al., 1986; Edell and Burke-Moore, 1987). The use of observation stems from the work of clinical or ethnological psychologists. Emotional reactions are most often recorded by measuring facial expressions, eye movement, gestures, intonation of the voice, verbal content, but also by rises in temperature, level of perspiration, ... etc. Its application in marketing has until now been very limited, as pointed out by Derbaix and Pham (1989) who, noting that emotion is a transitory state of varying intensity, underline that the coding of facial expressions is one of the most reliable measures. The exclusive use of the FAST technique method is however relatively complex when many different ads have to be evaluated (Ekman and Friesen, 1975). In the end, the external validity is limited by the existence of deep cultural and racial differences in the way the same emotions are expressed (Ekman, 1972). The second approach differs from the observation method in that the individual evaluates the emotions he has experienced. Verbal or non-verbal methods can be used. The most often used non-verbal method is that of the Warmth Monitor (Aaker et al., 1986) in which the individual traces the different psychological states experienced duing the showing of the ad with the help of a continuous line, the length of which is proportional to the stimulus intensity. This method, thus, does not allow for the qualification of the nature of the emotions (Abeele and MacLachlan, 1994), unlike the more explicit verbal methods, which, too, have undergone criticism. One concerns the difficulty of distinguishing mood, emotions and feelings within the psychological states recorded by the individual. The term "affect", sometimes unclearly employed in Anglo-Saxon literature, refers indistinctly to emotions, mood, feelings and motivations (Batra and Ray, 1986) and to attitudes, preferences and evaluations (Pieter and Van Raaij, 1988). A second criticism concerns cognitive constraints faced by the individual in the expression of his emotional state: the person being questioned might not actually know, be able to, or want to describe the often transitory emotions experienced during a commercial (Derbaix et Pham 1989). This last criticism is in fact linked to the diversity of emotions felt. Thus, even if the work of Rossiter, Percy and Donovan (1991) achieved a clear separation of the concepts of affective and emotional judgement of the ad, the authors used a measurement scale with no less than 58 types of emotions. The structuring of the primary emotions into eight classes (Plutchik, 1980), the combination of which explains the secondary emotional states, is an encouraging path of research, but their subjective, abstract and generic nature makes their use inadequate in a direct questioning method. In summary, methods of observation or of direct questioning of individuals concerning their emotions prove to be poorly adapted to our research. Moreover, they only measure one of the two dimensions that we are attempting to analyze: emotion.

Classification according to the induced cognitive or affective responses

Modelling the processes of advertising persuasion emphasizes the leading role of both cognitive and affective antecedents. Two large families of models can be distinguished. The first regroups the cognitive processes that are based on the beliefs of an individual concerning given information to explain his attitude towards the brand (Ab). The second takes into account the affective reactions upon exposure to the ad, in order to explain the effect of the advertising upon the attitude towards the brand. Whilst the research was initially based on the postulate of a distinct influence of cognitions and affect, the accepted reality today is more complex (MacInnis and Jaworski, 1990) as these two persuasion processes most often act in unison. The concept of attitude towards the advertising (Aad) has itself greatly evolved. The most recent works distinguish a cognitive dimension of Aad, sometimes described as "central" and an affective one or "peripheral" (Shimp, 1981; MacKenzie and Lutz, 1989), the activation of which depends on the involvement level of the subject. One of the merits of these models is to have placed the focus on the importance of the cognitive or affective reactions induced by advertising communication. The determination of the nature of the message is henceforth based essentially on their measure and their characterization with the help of appropriate scales. The real difficulty rests in the multiplication of the definitions given to the construct of information and emotion: the characterization of the advertising types opposes rational/emotional, thinking/feeling, factual/evaluative, informational/transformational advertisements (Berger, 1981; Hansen, 1981; Holbrook, 1978; Puto and Wells, 1984; Shimp, 1981; Vaughn, 1980) without an unambiguous definition specifying the outline of each term. We are opting nevertheless for a research method similar to that of these authors. Its implementation is preceded by the statement of a paradigm and a synthetic definition of the constructs.

DEFINITION OF THE CONSTRUCTS AND CHOICE OF THE STATEMENTS

The hypothesis that modification of the attitude towards a brand (Ab) proceeds simultaneously from cognitive and affective transfer of attitude towards the ad (Aad) is widely demonstrated (MacKenzie and Lutz, 1989). We assume therefore that informational or emotional characteristics are two distinct, but non-exclusive, advertising dimensions. The persuasion proceeds simultaneously, but to varying degrees, from one (or the other) of these characteristics (Holbrook and O’Shaughnessy, 1984). The prevailing content of an ad, either informational or emotional, therefore allows us to foresee which central or peripheral route to persuasion is likely to be activated. Our definition of the constructs of information and emotion emphasizes the perception by the audience of the underlying advertiser’s persuasive intent. Therefore, it goes beyond the description of the ad’s salient characteristics (Aaker and Norris, 1982): the objective content of the ad is of less interest than its perception by the exposed audience. Thus, we define:

I. Informational advertising which puts forward the characteristics or the benefits attached to the product or the brand. The message is centred on the important choice criteria for the consumer and the persuasion rests, among others, on the perceived value of the information, the source’s trustworthiness and the credibility of the claims.

II. Emotional advertising that modifies, in a temporary or durable way, the attitude towards the brand in associating it in the consumer’s mind to an intense psychic or psychological state experienced during the ad exposure. The persuasion rests on the characteristics of the emotion experienced (i.e. nature, intensity, duration, repetition, ...etc.) as well as on the transfer of the attitude concerning the ad (Aad) towards the attitude concerning the brand (Ab).

Note that the adopted definition of emotional advertising is independent from the nature (or the directionality) of the activated emotions. It is less restrictive than that of transformational advertising (Puto and Wells, 1984) which forms a sub-category of emotional ads. From the definition of the constructs, we define 25 statements in a first qualitative phase (see annex 1) of which certain are adapted from existing scales (Aaker and Norris, 1982; Puto and Wells, 1984). Supplementary statements are added to cover a wider definition of the constructs, in particular for emotional advertising. Finally, we retain 14 informational statements and 11 emotional ones.

METHODOLOGY FOR THE CREATION AND VALIDATION OF THE SCALE

In order to construct and validate an advertising classification scale according to perceived informational or emotional intent of the ad, we are following the Churchill’s paradigm (1979) with three restrictions. We do not proceed to a second data collection, which is only justified when the formulation of new statements (or the refinement of former ones) is necessary. The examination of the structure of the initial data only shows the need to eliminate the statements that are judged scarcely pertinent or inappropriate to the measurement of the constructs. We also do not use the Multitraits-Multimethods technique to measure trait validity as suggested by Campbell and Fiske (1959): this rather heavy and complex experimental procedure is not free from criticism and the results are not guaranteed (Peter, 1979). We therefore adopt a decision rule based on the formal indicators of convergent and discriminant validity of the models linking the measurement variables to the latent consructs (Fornell and Larcker, 1981). This approach, certainly imperfectBsince, unlike in the MTMM procedure, we do not measure the share of variance due to the methodBreveals itself to be easier to implement. Finally, we assess reliability before and after the factor structure has been determined and items are eliminated (Gerbing and Anderson, 1988).

The data collection is carried out during advertising pre-tests done by GfK, which guarantees a realistic context of simulated advertising exposure. Each ad is inserted twice into a TV program. At the end of the session, each person evaluates unassisted 2 ads on 25 statements with the help of a 5 point measurement scale (1: "does not apply at all" to 5: "completely applies"). Thanks to a filter question, people do not evaluate the ad if they do not remember having seen it. The questionnaire ends up with a direct measure on a 5 point scale of the underlying advertiser’s persuasive intention, either informational ("this ad brings me specific information ...") or emotional ("this ad elicits a particular emotion..."). In order to evaluate 10 different ads, 5 samples of approximately 120 people are necessary. The samples are not, in principle, matched (due to the pre-tests constraints) which might induce a bias. This point will be the object of particular attention in the analysis of the results. The chosen ads concern 7 different categories of durable and consumer goods. A pre-test conducted among marketing experts validates the varied informational or emotional appeal of the ads: 4 have a high content of information and 6 favor the use of emotion.

RESULTS

Refining the measurements

This preliminary stage fulfils two requirements: first and foremost, to estimate the stability across the unmatched sub-samples, then to refine the items. The end result is to confirm that, for a given commercial, the research instrument reproduces the same measurement whatever the sub-sample analyzed. A random half-split allows the construction of two sub-samples of 50% of the total sample (n=499 and n=500 respectively), all ten ads being equally represented. At this stage, we end up with a scale made up of two dimensions measured by 16 variables: the deletion of 9 items (see appendix 1) reinforces its stability across the sub-samples. A final factor analysis on the total sample allows the measure to be *refined+. The Cattell screetest again retains two factors. The explained variance increased to 58.2%, the variance of the third factor remaining marginal (5%). The communalities of the variables on the first two factors are satisfactory: all the values reach or pass the threshold of 50%. On each factor, the contribution is close to or greater than .50: we note that if a factor strongly explains a variable, the opposite factor explains it weakly. These results confirm the underlying structure of the data: the 16 measurement variables are highly correlated to two axes which we term perceived informational intent and perceived emotional intent.

Measuring the Reliability

We first calculate a formal internal consistency coefficient, Cronbach’s alpha (1951) before conducting three confirmatory factor analyses. The values that we obtain approach .90 on every factor (.90 and .88 respectively) confirming the reliability of the measurement instrument (Peter, 1979). Another way of measuring the internal consistency of the scale is to calculate the value of the multiple correlation coefficient (R2) for each statement. Two statements have a value inferior to .50: info5 (.35) and emo7 (.37). However, only the deletion of info5 does not have a significant effect on the value of the Cronbach’s alpha coefficient, thus recalculated: the interpretation of the confirmatory analyses will allow us o discard it definitively.

Three confirmatory factor analyses complete the assessment of the measurement instrument’s reliability. The method consists here of increasing the reliability of each construct separately before examining them conjointly.

Confirmatory analysis on the dimension of emotion

The construct of perceived emotional intent, considered as a latent variable, is first measured by 7 variables. The examination of the standardised residual values, as well as the modification index of the model suggested by Lisrel7 8 (J÷reskog and S÷rbom, 1989), lead us to put aside 2 measurement variables, emo2 and emo6 (see legend of appendix 1), thus significantly increasing the adjustment of the model to the data: DX2calc. =B261.71; p=.00. [In this research we are using the value of X2 to improve the fit of the model to the data following a procedure of sequential tests of nested models (Bentler and Bonnet, 1980). We are not directly interpreting the value of X2 of the final model because, as underlined by Fornell and Larcker (1981), this test is tainted with weaknesses. We are subsequently using three other indicators-GFI, SGFI and NFI - to evaluate the model's fit.] The two values GFI and AFI (1.00 and .99) are close to 1: the theoretical model takes well into account the major part of the variance of the data. The third, NFI or incremental index of Bentler and Bonnet (1980) is itself equal to 1.00: it measures the covariance share of the observed data reproduced by the model. These three indicators reveal that the model has a good adjustement quality to the data., which an examination of the average value of the residual terms backs up: the value of .25 is weak compared to the average values of variance and covariance of the observed variables (J÷reskog and S÷rbom, 1989).

Confirmatory analysis on the dimension of information

The construct of perceived informational intent, considered as a latent variable, is first measured by 9 variables. The improvement of the fit of the model, according to the previously described procedure, leads to the removal of three measurement variables: info4, info5 and info6 (see legend of appendix 1). The value of the c2 decreases heavily: Dc2calc. =B214.05; p=.00. The model thus made up of 6 measurement variables appears to adjust well to the data: the c2 is 8.63 (p=.47), the GFI, AGFI and NFI values are close to 1 (1.00; .99 and 1.00 respectively) and the average value of the residual terms is weak (.19).

Confirmatory analysis on the two dimensions of emotion and information

The last confirmatory analysis is based on the two dimensions of emotion and information taken together. In view of the previous stages, they are measured respectively by 5 and 6 measurement variables. Two measurement variables are removed after considering the modification indexes suggested by Lisrel7 8: emo5 and info3 (see legend of appendix 1). The final model, therefore containing 9 measurement variables, sensitively improves the quality of adjustment: Dc2 =B176.81; p=.00. The weak convergent validity of the two items emo5 and info3 with the latent variable to which they are theoretically linked is explained by the existence of a significant dependent relationship with the complementary latent variable: thus info3 is negatively correlated with the emotional dimension (l =B.88; p=.00) and emo5 negatively with the informational dimension (l=B.26; p=.00). The content analysis of the two items provides us with an explanation. Wording of emo5 is indeed ambiguous and ambivalent: it opposes emotion to information as the dominant advertising persuasion process in the ad ("...this ad appeals more to feeling than to reason"). The item info3 puts forward that, in the case of informational advertising, the product and its characteristics are "the main elements of the ad": analysis of the data reveals that the emotional dimension is also characterised by the absence of reference to the product or to its characteristics. We remove them in order to keep a structure with two totally independent dimensions. The model with 9 variables fits correctly to the data: GFI, AGFI, and NFI are in the region of 1 (1.00; .99 and 1.00 respectively) and the average value of the residual ters is low (.28). All of the dependent coefficients between the measurement variables and their constructs are positive and significantly different from 0 (p=.00). For the two constructs, the average variance achieved by the latent variables is superior to .50: the values are .54 and .55 respectively for x1 and x2, the weakest single values being .47 for the variables info1 and info2.

We conclude from the three successive confirmatory factor analyses that the model fits well to the data and presents a good convergent validity.

Measuring the Construct Validity

We examine in turn the two components of the construct validity, the trait validity and the criterion validity. For this, we establish a causal model for which the two dimensions of the scale form the exogenous variables (xi) whereas the two constructs of direct evaluation constitute the endogenous variables (hi) (see figure 1).

FIGURE 1

CAUSAL MODEL: RELATIONSHIPS BETWEEN AN INDIRECT EVALUATION USING THE ADVERTISING CLASSIFICATION SCALE AND THE EVALUATION BY DIRECT QUESTIONING

The fit index of the model to the data is good without having to remove or reformulate one or more statements. The value of c2 is weak (c2calc.=30.54; p= .17) and the values of the fit indexes are all close to 1 (GFI=.99; AGFI=.98 and NFI=.99). The standardised values g1,1 and g2,2 of the dependent coefficients between the exogenous and endogenous latent variables are high and different from 0 (.95 and 1.01 respectively): there exists a strong correlation between the indirect and direct evaluation of a same construct. On the other hand, the dependent coefficients between two latent variables that correspond to different constructs are nil: the two constructs of information and emotion are thus clearly distinct and measure different concepts.

The trait validity test is based on the calculation of the indicators of convergent and discriminant validity formulated by Fornell and Larcker (1981). The exogenous variables capture 52% (x1) and 55% (x2) respectively of the variance of their measurement variables. The endogenous variables respectively share 75% (h1) and 72% (h2) of their unique measurement variable’s variance. The convergent validity of each variable with its construct reaches or surpasses 50%. Only two of them have an inferior value to this threshold: emo7 (38%) and info2 (42%). However, as the quality of the fit of the theoretical model to the data deteriorates when they are eliminated, we decide to keep them for the next stage. The measurement instrument thus presents a satisfactory convergent validity. The discriminant validity of the causal model is confirmed provided that the variance shared between the constructs is inferior to that shared between the constructs and their measurement variables. The variance shared between the exogenous variables is: (f1,2)2=.15, a value well inferior to the average variance shared between the measurement variables and their latent variable: .535. The measurement tool thus presents a good discriminant validity. The scale presents a good trait validity since it fulfills the criteria of both convergent and discriminant validity: the scale, thus refined, accurately measures (convergent validity), with the use of the 9 retained items, two distinct constructs (discriminant validity), called perceived informational intent and perceived emotional intent of the ad.

The criterion validity (or nomological validity) consists of measuring the relationship between the studied construct and other constructs which are theoretically similar and thus logically correlated. Two approaches are implemented: the first, more formal, is based on the calculation of an indicator of nomological validity of the causal model. The second, more analytical, uses a linear regression between the direct measure (explained variable) and the indirect measure (explanatory variable) by means of the scale. Nomological validity of the causal model can be tested with the help of an extension of the redundancy coefficient of Stewart and Love (1968), tested by Miller’s F test (1975). This is analogous to the mean quare of the multiple correlation between each endogenous variable (hj) and all the explanatory exogenous variables (xi). In our case, this measure represents the percentage of information contained in the direct measurement variables (YL) explained by the constructs xi since each endogenous variable is linked to a single measurement variable. The significancy test of the coefficients R2(Y1/x1) (.76) and R2(Y2/x2) (.75) leads us to reject the null hypothesis for the two dimensions of information and emotion (p=.00). We conclude that the causal model, that confronts both the direct and indirect measures of a same construct, presents a good nomological validity for the two studied constructs. The results of the following two linear regressions reinforce this conclusion. The linear regression between the direct measurement of information and the factorial z-score of information is significant: Fcalc.=751.91, p=.00. The correlation coefficient (R2) is .51 and the regression coefficient is positive, clearly greater than 0: b=.71, p=.00. The mapping shows that only two ads slightly deviate from the regression line (see appendix 2). One of them, for example, "NescafT’s ProcTdT Plein Ar(me", has an evaluation score lower than the mean: the public judged that the ad did not convey specific information on the superiority of this brand of product. Its score on the scale is, however, higher than the average: the ad thus clearly possesses characteristics of informational advertising. One of the reasons for this apparent contradiction is surely linked to the "ProcTdT Plein Ar(me" seniority on the market (1976): this information is clearly present, but is scarcely innovative in relation to the product. The linear regression between the direct measurement of emotion and the factorial score of emotion is also significant: Fcalc.=697.64; p=.00. The correlation coefficient (R2) is .49 and the regression coefficient is positive, clearly greater than 0: b=.70, p=.00. The mapping shows an accurate adjustment of the ads to the line of regression (see appendix 2). Only two points deviate from it. One of them, "the Lavomatic" advertising for the Gore-Tex clothing, gains a higher emotional score when it is measured in a direct manner than when it is evaluated with the scale. The scenario here is very dynamic, with an alternation of strong and distorted images interspersed with written messages on the product benefits. The finalised scale is not very well adapted to the measurement of emotional reactions caused by a sonorous level, a musical rythm or a rapid series of visual or auditory stimuli: the galvanic measure of the reactions is certainly, in this precise case the most adequate solution! In brief, the two linear regressions confirm the nomological validity of the measurement scale. The way the ads fit the regression lines highlights the strengths (but also the weaknesses) of the measurement instrument: the complex nature of emotional activity, as well as the diversity of advertising creativity styles, constitutes an inherent limit to any measurement scale, inevitably "reducing" compared to the reality that it intends to measure.

The trait validity is henceforth associated with a good criterion validity. The instrument thus shows a good construct validity.

TABLE 1

MEAN FACTORIAL SCORES PER GROUP ON THE INFORMATIONAL AND EMOTIONAL AXIS

Developing the norms

Lastly, a principal components analysis allows the calculation of the factor score coefficients for each ad on each dimension of the scale. The first two factors henceforth account for more than 65% of the total variance. The internal consistency of the scale, assessed on the final factor structure is satisfactory: Cronbach’s Alpha reaches values of .86 and .83 for the informational and emotional factors respectively. In order to develop an initial series of norms, we form clusters of similar ads using an hierarchical algorithm under SPSS on the standardized factorial scores. The analysis of the agglomeration schedule (dendrogram) determines an optimal number of four groups of which mean cores on the two factorial axes are brought in table 1.

Group I: Information (+)BEmotion (=)

This group classes together the ads whose information score is higher than their emotion score, the latter however being at an average value. The analysis of the content of the ads reveals that the product is put to the fore, in particular emphasising the characteristics and the benefits that it offers. In this type of ad, the advertising persuasion prioritizes the information, the emotion being secondary to arouse and maintain attention as illustrated by the scene of a chick kept dry thanks to Gore-Tex.

Group II: Information (=)BEmotion (+)

The ads of group II have the opposite profile to that of the previous group: the emotional score is the highest of all the ads whereas the informational score is average. The ad for C(te d’Or chocolate shows a man who, thanks to tasting chocolate, escapes into an African setting that alternates between grandiose landscapes and exotic scenes. The ad for Maxwell coffee is presented as a slice of life in a New Orleans ambience: the sampling of coffee amongst people of different races. In this set of ads, emotion outweighs information, but the product remains, in spite of everything, present because it is the cue that triggers the aroused emotions.

Group III: Information (-)BEmotion (=)

This group is made up of four ads. These ads are characterized by an average score for emotion, but their distinctive feature is their very weak score for information. Examination of the scenarios shows that executional characteristics outweigh all other considerations: three are based on humourous but rather foolish scenes, the last one displays youth’s innocence in a Tibetan Valley. In these four ads, the product is not very present (sometimes hardly shown), and the advertising message is free from strong arguments (most notably, the absence of benefits from using the product) and emotion has either a low vividness, or is too weakly associated to the product or the brand.

Group IV: Information (=)BEmotion (-)

The two ads for Ariel and NescafT form group IV of which the profile is the opposite of group III. The measure on the scale reveals the absence of emotion, which the examination of the contents of the two ads confirms. No emotional context is added to scenes favouring a sober, rational speech with a scientific connotation. For both ads however, the information content, which is a priori high, is not found in the measure carried out. Is the information judged somewhat unbelievable in the case of Ariel washing powder? The viewer may penalize the artificial presence of C. Bravo who, while giving the impression of objectivity, pursues uniquely a commercial goal. This outline prompts the weak credibility amongst the public of advertising claims made by stars or personalities (Tripp et al., 1994). Concerning the NescafT ad, we have already stated that the informational content is judged out-of-date. Is this perhaps the reason why both advertisers, Ariel and NescafT cancelled very soon the broadcasting of those films?

The use of the measurement scale to classify ads according to their characteristics of information or emotion confirms the paradigm of our research: while the dimensions of information or emotion are mutually independent, they are not mutually exclusive. The four classes obtained however highlight the opposition of the two outlines: the first groups together the ads which present a dominant trait (of information or emotion), the other dimension being somewhat removed but never totally absent (groups I and II). Inversely, the second (groups III and IV) associates the ads whose common characteristic is to present a deficit of one or the other of the two dimensions, without the opposite dimension unbalancing this deficit: the absence of information is not compensated by a strong emotional content (and vice-versa). Analysis of the content of the films is coherent with the measure carried out: the validity of the measurement scale is thus reinforced.

CONCLUSIONS

At the onset of the research, we dispose of a reliable and valid measurement instrument for the classification of ads according to their characteristics of information or emotion. We voluntarily favored a measure of induced cognitive and affective reactions. We have, in effect, placed the emphasis on the perceived processing of the ad by a direct measurement among the audience. Of course, we might be criticized for too narrowly complying with the information processing theory: other theoretical frames have successfully been put forward to highlight psychological processes that we obviously do not take into account (Scott, 1994). The choice of an opinion and attitude scale, constitutes another limitation in the generalisation of this tool’s usage. The verbal method does not perfectly account for transitory emotions, or those, difficult to report, which the individual himself perhaps is not aware of. Nevertheless, the external validity of the scale could be prolonged by confronting it with other measurement techniques, such as observation or measurement of physiological or psychological reactions.

An initial use illustrates the diagnostic value of our scale for marketing practice. The stability and the validity of the obtained classification deserves to be tested on a greater number of ads. Last but not least, it is regrettable that no ad likely to trigger negative emotions (fear, disgust) was tested, as their persuasive impact lies on a rather distinctive process (Rossiter and al., 1991). A priori, the potential of persuasion of the first two groups is higher than that of the last two. A measure of advertising efficiency within the previous classification groups would allow a deepening of our diagnosis: does the deficit of one of the dimensions explain the weak performance of an ad? Does the interaction between the two constructs of perceived informational and perceived emotional intent reinforce the efficiency of the ad? Are our newly measured constructs predictive of the route to persuasion followed by the advertising message? These questions are of potential interest for both researchers and marketing practitioners.

APPENDIX 1

STATEMENTS OF THE CLASSIFICATION SCALE: INITIAL VERSION (25 ITEMS) AND FINAL VERSION (9 ITEMS)

APPENDIX 2

REFERENCES

Aaker, David A. and Donald Norris (1982), ''Characteristics of TV Commercials Perceived as Informative,’’ Journal of Advertising Research, 22 (2), 61-70.

Aaker, David A., Douglas M. Stayman., and Michael R. Hagerty (1986), ''Warmth in Advertising: Measurement, Impact and Sequence Effects,’’ Journal of Consumer Research, 12 (4), 365-381.

Abeele, Piet Vanden and Douglas L. MacLachlan (1994), ''Process Tracing of Emotional Responses to TV ads: Revisiting the Warmth Monitor,’’ Journal of Consumer Research, 4 (20), 586-600.

Batra, Rajeev and Michael L. Ray (1986), ''Affective Responses Mediating Acceptance of Advertising,’’ Journal of Consumer Research, 13 (2), 234-249.

Bentler, Peter M. and Douglas G. Bonnett (1980), ''Significance Test and Goodness-of-Fit in the Analysis of Covariance Structures,’’ Psychological Bulletin, 88 (3), 588-606.

Berger, D.A. (1981), ''Retrospective: FCB recall study,’’ Advertising Age, 26, 36-37.

Campbell, D.T. and D.W. Fiske (1959), ''Convergent and Discriminant Validation by the Multitrait-multimethod Matrix,’’ Psychological Bulletin, 56 (1), 81-105.

Cattell, B. and S. Vogelmann (1977), ''A Comprehensive Trial of the Scree nd KG criteria for Determining the Number of Factors,’’ Multivariate Behavioral Research, 12 (3), 289-325.

Churchill, Gilbert A. (1979), ''A Paradigm for Developing Better Measures of Marketing Constructs,’’ Journal of Marketing Research, 16 (1), 64-73.

Cronbach, L.J. (1951), ''Coefficient Alpha and the Internal Structure of Tests,’’ Psychometrika, 16 (3), 297-334.

Derbaix, Christian and M.T. Pham (1989), ''Pour un dTveloppement des mesures de l’affectif en marketing,’’ Recherche et Applications en Marketing, 4 (4), 71-87.

Edell, Julie A. and Marian Chapman Burke (1987), ''The Power of Feelings in Understanding Advertising Effects,’’ Journal of Consumer Research, 14 (3), 421-433.

Ekman, Paul (1972), ''Universal and cultural differences in facial expressions of emotions,’’ in Nebraska Symposium on Motivation, 19, Lincoln, University of Nebraska Press.

Ekman, Paul and Wallace V. Friesen (1975), Unmasking the face: a guide to recognizing emotions from facial clues, Englewood Cliffs, N.J.: Prentice-Hall.

Fornell, C. and D.F. Larcker (1981), ''Evaluating Structural Equations Models With Unobservables Variables and Measurement Errors,’’ Journal of Marketing Research, 18 (1), 39-50.

Hansen, F. (1981), ''Hemispheral lateralization: Implication for understanding consumer behavior,’’ Journal of Consumer Research, 1 (8), 23-36.

Holbrook, Morris B. (1978), ''Beyond Attitude Structure: Toward the informational determinants of attitude,’’ Journal of Marketing Research, 4 (15), 545-556.

Holbrook, Morris B. and John O’Shaughnessy (1984), ''The Role of Emotions in Advertising,’’ Psychology & Marketing, 1 (2), 45-64.

J÷reskog, Karl G. and Dag S÷rbom (1989), Lisrel 8: User’s Reference Guide, Chicago Il., SSI Scientific Software International.

MacInnis, Deborah J. and Bernard J. Jaworski (1990), ''Two Routes to Persuasion Model in Advertising: Review, Critique and Research Directions’’, in Review of Marketing, ed. V.A Zeithaml, Provo, Utah: American Marketing Association, 3-42.

MacKenzie, Scott B. and Richard J. Lutz (1989), ''An Empirical Examination of the Structural Antecedents of Attitude Towards the Ad in an Advertising Pretesting Context,’’ Journal of Marketing, 53 (2), 48-65.

Miller, John K. (1975), ''The Sampling Distribution and a Test for the Significance of the Bimultivariate Redundancy Statistic: a Monte Carlo Study,’’ Multivariate Behavioral Research, 10 (2), 233-244.

Nunnaly, Jum C. (1978), Psychometric Theory, 2nd edition, New York, NY: McGraw-Hill Book Company.

Pechmann, Cornelia and David W. Stewart (1989), ''The Multidimensionality of Persuasive Communications: Theoretical and Empirical Foundations,’’ in Cognitive and Affective Response to Advertising, ed. P. Cafferata and A. Tybout, Lexington, Mass.: Lexington Books, 31-65.

Peter, J.P. (1979), ''Reliability: a Review of Psychometric Basics and Recent Findings,’’ Journal of Marketing Research, 16 (1), 6-17.

Pieter, R.G.M. and W.M. Van Raaij (1988), ''Functions and Management of Affect: Applications to Economic Behavior,’’ Journal of Economic Psychology, 9 (3), 251-282.

Plutchik, Robert (1980), Emotions: a Psychoevolutionnary Synthesis, New York: Harper & Row.

Puto, Christopher P. and William D. Wells (1984), ''Informational and Transformational Advertising: the Differential Effects of Time,’’ in Advances in Consumer Research, Vol. 11, Ann Arbor, MI: Association for Consumer Research, 638-643.

Rossiter, John R, Larry Percy, and Robert J. Donovan. (1991), ''A Better Advertising PlanningGrid,’’ Journal of Advertising Research, 31 (5), 11-21.

Scott, Linda M. (1994), ''The Bridge from Text to Mind: Adapting Reader-Response Theory to Consumer Research,’’Journal of Consumer Research, 21 (3), 461-480.

Shimp, Terry A. (1981), ''Attitude Toward the Ad as a Mediator of Consumer Brand Choice,’’ Journal of Advertising, 10 (2), 9-15.

Stewart, David W. and W. Love (1968), ''A General Cannonical Correlation Index,’’ Psychological Bulletin, 70, 160-163.

Stewart, David W. and David H. Furse (1986), Effective Television Advertising: A Study of 1,000 Commercials, Lexington, Mass.: Lexington Books.

Tripp, Carolyn, Thomas D. Jensen, and Lee Carlson (1994), ''The Effects of Multiple Product Endorsements by Celebrities on Consumer’s attitudes and Intentions,’’ Journal of Consumer Research, 20 (4), 535-547.

Vaughn, Richard (1980), ''How Advertising Works: A Planning Model,’’ Journal of Advertising Research, 20 (5), 27-33.

----------------------------------------