warmth in advertising: measurement, impact, and sequence effects

18
Journal of Consumer Research, Inc. Warmth in Advertising: Measurement, Impact, and Sequence Effects Author(s): David A. Aaker, Douglas M. Stayman, Michael R. Hagerty Reviewed work(s): Source: Journal of Consumer Research, Vol. 12, No. 4 (Mar., 1986), pp. 365-381 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/254299 . Accessed: 22/02/2012 23:26 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR to digitize, preserve and extend access to Journal of Consumer Research. http://www.jstor.org

Upload: independent

Post on 11-Dec-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Journal of Consumer Research, Inc.

Warmth in Advertising: Measurement, Impact, and Sequence EffectsAuthor(s): David A. Aaker, Douglas M. Stayman, Michael R. HagertyReviewed work(s):Source: Journal of Consumer Research, Vol. 12, No. 4 (Mar., 1986), pp. 365-381Published by: The University of Chicago PressStable URL: http://www.jstor.org/stable/254299 .Accessed: 22/02/2012 23:26

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR todigitize, preserve and extend access to Journal of Consumer Research.

http://www.jstor.org

Warmth in Advertising:

Measurement, Impact, and Sequence Effects

DAVID A. AAKER DOUGLAS M. STAYMAN MICHAEL R. HAGERTY*

The impact of the feeling of warmth created by a commercial, as measured by the ' warmth monitor," is explored in a series of three studies. The first study examines the relationship between warmth and arousal as measured by skin response. The other studies explore the relationship between warmth and advertising responses such as liking of the ad and purchase likelihood through testing ads with warm and other execution strategies. They also test the effects of sequences of commercials on warmth responses and on the impact of the ad.

T he last two decades have seen an emphasis upon information processing models of advertising.

However, it is increasingly being realized that such models are incomplete or inadequate (Bagozzi 1982; Holbrook and Hirschman 1982) and that it is important to model the execution-focused effects of advertising (Petty, Cacioppo, and Schumann 1983). One such effect is the affective/experiential response of audience mem- bers (Batra and Ray 1985).

Interest in affect has been spawned by recent work on emotions and on the relationship between affect and cognition. Zajonc (1980; Zajonc and Markus 1982) has proposed separate affective and cognitive processes. Isen (Isen et al. 1982), Johnson and Tversky (1983), and others have recently demonstrated that even relatively mild affective states can substantially influence cogni- tive processing and social behavior. Finally, the under- standing of emotions has recently resurged with interest centered on facial movement (Ekman, Friesen, and Ellsworth 1972; Izard 1977), cognitive theories (Laza- rus, Kanner, and Folkman 1980; Mandler 1975), and neurochemistry (Brown and Wallace 1980).

Several approaches to studying affect in persuasion have been proposed. For example, Kroeber-Riel (1984)

demonstrated conditioning using emotion-laden stim- uli. However, the predominant approach in the con- sumer behavior literature has focused upon the attitude toward the advertisement (Aad) as a mediating variable (Lutz 1985; Silk and Vavra 1974). One hypothesis is that improving Aad will improve attitude toward the brand via some conditioning process (Mitchell and Ol- son 1981) or will affect attention, perception (Bower 1981), or the amount of information processed (Ray and Batra 1983). The competing "J-shaped" model postulates that disliked commercials will be more ef- fective than neutral ones by stimulating attention and processing, creating brand familiarity, or distracting the audience from counter-arguing (Moore and Hutchinson 1985).

One problem with the Aad-focused research is that little effort has been expended in making distinctions between different execution strategies. For example, three equally liked commercials-one using slapstick humor, another employing a serious informative copy, and a third with warm, sentimental copy-may be ef- fective in completely different ways, as may two equally disliked commercials-one which is considered boring and the other irritating.

The intent of this study is to focus upon one of the several dimensions that underlie the Aad variable: the feeling of warmth engendered by some commercials. Warmth has been used as an emotion in a number of contexts, appearing in the literature as a part of lists of emotions (Bush 1972), as an experience associated with emotional terms (Davitz 1969), and as a part of em- pathetic emotional response (Coke, Bateson, and McDavis 1978). Evidence for the potential relevance of warmth in advertising comes from three audience per- ception studies of television advertising. In each study

*David A. Aaker is the J. Gary Shansby Professor of Marketing Strategy at the School of Business Administration, University of Cal- ifornia, Berkeley, CA 94720. Douglas M. Stayman is Assistant Pro- fessor in the Department of Marketing Administration, University of Texas, Austin, TX 78712. Michael R. Hagerty is Assistant Professor of Marketing, School of Business Administration, University of Cal- ifornia, Berkeley, CA 94720. Special thanks go to four anonymous JCR reviewers, whose comments and suggestions improved the final product enormously. The authors also gratefully acknowledge the helpful comments of Donald Bruzzone, Trudy Kehret-Ward, and William Wells.

(? JOURNAL OF CONSUMER RESEARCH * Vol. 12 * March 1986

365

hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight

366 THE JOURNAL OF CONSUMER RESEARCH

one of the four major perceptual dimensions can be interpreted as involving perceived warmth.

Aaker and Bruzzone (1981) found that their warmth factor was associated with commercials that utilized sentimental / family - kids / friends - feelings / feel - good - about-yourself creative approaches. Wells, Leavitt, and McConville (1971) include adjectives such as gentle (also in Aaker and Bruzzone's warmth factor), tender, soothing, serene, and lovely in the factor they term "sensuousness." Schlinger (1979) found an "empathy" factor associated with commercials involving affection- ate couples, warm relationships, mother-child inter- actions, attractive products, vacation settings, or ap- pealing characters such as Pillsbury's "soft and cuddly" doughboy. The construct emerging from these three dissimilar studies, although complex, has some consis- tent characteristics and associations.

In the following sections, a definition of warmth is proposed, elaborated, and discussed in terms of other emotion research. The objectives to be addressed and the hypotheses to be tested are then presented, followed by a description of the "warmth monitor," an approach to providing a continuous measure of felt warmth. Three sets of experiments that test the hypotheses are then discussed.

THE "WARMTH" CONSTRUCT

Warmth is here defined to be a positive, mild, volatile emotion involving physiological arousal and precipi- tated by experiencing directly or vicariously a love, family, or friendship relationship. The proposed defi- nition positions warmth on the two dimensions most often used in describing emotions, valence and arousal (Daly, Lancee, and Polivy 1983). Warmth is concep- tualized to be positive and to involve arousal. Thus, a detached expression of love or friendship without con- current involvement and arousal would not generate warmth. However, the arousal level associated with warmth is moderate. Therefore, an experience in which the involvement, depth of feeling, and arousal was ex- tremely high would be too intense to be warmth. As a mild, positive emotion, it differs qualitatively from the strong, negative emotions such as fear, anger, and dis- gust that have been the objects of the bulk of empirical work in emotion research (Lazarus et al. 1980, pp. 202-212).

A notable aspect of the definition is the suggestion that the direct or vicarious experience of a love, family, or friendship relationship is involved. Thus, it follows in the tradition of Darwin (1872) and many modern psychologists who view emotions in a social context. For example, Averill (1980, p. 312) defines emotions as transitory social roles. Thus there will usually be a social object like a person or persons, animal, organi- zation (e.g., fraternity, team, or club), or institution (e.g., country). Further, this social object will usually

be linked to another social object in a relationship that involves emotions such as love, pride, joy, sentimen- tality, tenderness, or happiness as well as warmth.

In advertising, warmth can involve a relationship be- tween two characters in a commercial such as a scene between a father and son in a Lowenbrau commercial. In that case, the experience is vicarious and, thus, prob- ably less intense than portrayed by the actors. The in- tensity depends in part upon how easily the viewer can empathize with the scene. Or it could recall a similar personal experience which is then "relived" by the viewer (Clynes 1980). A commercial could also involve a relationship between the audience member and a character in the commercial. The commercial character might be the object of pride, love, sympathy, or physical attraction.

One way in which emotions have been depicted is through analysis of associated or similar emotions (e.g., Plutchik 1980). A study by Smith and Ellsworth (1985) suggests that warmth is associated with or similar to happiness and pride. Smith and Ellsworth asked 16 stu- dents to recall in detail past experiences associated with each of 15 different emotions. Most of the subjects for the "happy" experience remembered scenes similar to those found in the warm ads in the perception studies reviewed above. These scenes included times spent with other people, being reunited with close friends or rel- atives, being at a party with friends, or going out on a pleasant date. When positioning the 15 studied emo- tions in a six-dimensional space, Smith and Ellsworth found that pride was very similar to happiness. Inter- estingly, commercials with a pride/"feel-good-about- yourself" theme loaded high on the warmth dimension in the Aaker and Bruzzone (1981) perception study (of the top 27 warm commercials, six had a pride theme).

Smith and Ellsworth's emotional space provides ad- ditional insights into warmth. Pride and happiness were characterized as extremely pleasant and as involving very little effort, a high level of certainty about the sit- uation, and a strong desire to pay attention. Pride was distinguished primarily by a closer association with a personal sense of responsibility and control. The pleas- antness and low effort dimensions conform to the con- ception of warmth above as having positive valence and moderate arousal.

Following Holbrook and O'Shaughnessy's (1984) ty- pology, warmth is therefore viewed as an emotion which is acute, specific, and reactive. Warmth is acute in that it is short-lived, capable of being created or changed quickly. Specificity means that warmth is created by a specific target in the environment (as opposed to moods, which are general predispositions to respond in specific ways). Specificity also implies that warmth changes quickly (as the perceived environment changes), fol- lowing the suggestion by many researchers (cf., Wess- mann and Ricks 1966, Chap. 1) that emotions change rapidly, in seconds or minutes, as opposed to more long lasting moods. Reactivity further identifies warmth as

hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight

WARMTH IN ADVERTISING 367

being created by a reaction to the environment rather than by a goal-directed behavior or response.

STUDY OBJECTIVES AND HYPOTHESES

This article introduces the "warmth monitor," a method that provides a continuous measure of felt warmth, and has three interrelated objectives. First, it tests two hypotheses that are suggested by the definition of warmth just proposed. Second, it assesses how feel- ings created by prior commercials will affect the warmth engendered by a commercial and the other dependent variables. Third, it studies the relationship between warmth and other relevant dependent variables such as Aad, brand image, memory of copy points, and purchase likelihood.

Hypotheses Related to the Proposed Warmth Definition

Warmth was conceptualized in the proposed defini- tion as being a stimulus-specific, volatile emotion in- volving physiological arousal. Two testable hypotheses relating to this definition are:

HI: Warmth is accompanied by physiological arousal.

H2: Warmth can be created or changed during a single television commercial.

The first hypothesis tests the existence of accompa- nying physiological arousal. If this is not supported, there is doubt that warmth could be considered an emotion as defined by most theorists (Lazarus et al. 1980; Stout and Leckenby 1984). The second hypothesis tests the stimulus specificity of warmth as well as its volatility.

Warmth and Sequence Effects

The second objective of the study is to explore se- quence effects-how feelings created by prior commer- cials impact upon subsequent ads. Specifically, will the impact of a warm commercial be enhanced if it is pre- ceded by a humorous rather than a warm commercial? What about being preceded by an irritating commercial?

Study of sequence effects has two purposes. First, the effects of prior ads can provide experimentally induced variance in response to subsequent ads that allows for more sensitive analysis of effects for all of the hy- potheses. Second, sequence effects have practical as well as theoretical interest (especially with firms now buying time for sets of commercials and thus controlling se- quence), and have been understudied, especially with respect to "feeling" commercials where cognitive mod- els are least appropriate (i.e., Brooker 1981; Webb 1979;

Yuspeh 1977). Brooker (1981) did study sequence ef- fects using humor and fear prior commercials but used factual test ads. He found that attitude and liking for the factual ad increased when preceded by a fear ad.

Adaptation theory (Helson 1964) suggests that the feeling induced by a prior commercial creates a refer- ence point. A prior commercial that induces a lower specific feeling than a subsequent commercial should create a lower reference point and therefore enhance the feeling induced by the subsequent commercial. Thus, a warm commercial would appear warmer when it is preceded by a commercial that creates a feeling other than warmth (i.e., a non-warm informative, hu- morous, or irritating ad) because the reference point to which it is being compared is lower. Warmth will there- fore have an inverse, monotonic relationship with the warmth level of the prior commercial. Potential support for this effect comes from assimilation-contrast theory (Sherif and Hovland 1961), which postulates a contrast effect when a warm test commercial is seen in contrast to a prior commercial that is perceived to be less warm.

Alternatively, it might be expected that a commercial would be more effective if the audience had a high "going-in" feeling level because the average feeling would be higher. A conditioning effect should then be stronger. However, this effect is more likely only if feel- ing levels tend to be relatively stable across stimuli, since then the going-in level would have the greatest lingering effect through the commercial. Since warmth is ex- pected to be stimulus-specific and volatile, it is proposed that:

H3: If a warm commercial is preceded by another, similar, warm commercial instead of a com- mercial that generates a low feeling of warmth, the following impact measures of the ad will be lower:

1: Trend in warmth

2: Aad

3: Purchase likelihood

4: Recall

Warmth and Post-Exposure Measures

Based upon the conditioning phenomenon suggested by Mitchell and Olson (198 1), Gorn (1982), and others, it can be hypothesized that:

H4: A more positive experience (a higher warmth level regardless of sequence) will become as- sociated with the commercial and the brand and thus impact positively upon:

1: Aad

2: Purchase likelihood

hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight
hebdaboy
Highlight

368 THE JOURNAL OF CONSUMER RESEARCH

Findings by Kroeber-Riel (1979) and others suggest that arousal is accompanied by increased levels of cog- nitive activity leading to increased recall. In addition, Choi and Thorson (1983) found emotionality of ads related to recall. Therefore, following Hypothesis 1 (that warmth is accompanied by arousal):

H5: Warmth generated by a commercial will be positively associated with recall.

THE WARMTH MONITOR

Several physiological and self-report measures have been employed to measure affective responses. Physi- ological responses such as skin response, pupil dilation, and heart rate, while used with some success in the psy- chological literature (e.g., Lacey 1967; Mewborn and Rogers 1979), are difficult to administer and cannot pinpoint specific emotions (Cacioppo and Petty 1981; Stewart and Furse 1982). Self-report measures such as adjective check lists (Cattell 1960; Nowlis 1965) and response protocols (Batra and Ray 1985) are retrospec- tive and intrusive. Further, the respondent may not know whether to report modal, average, or end feelings.

Several continuous and concurrent self-report devices have been employed to measure agreement or liking during exposure. For example, subjects have been asked to press buttons indicating agreement or disagreement with a message (Hovland, Lumsdaine, and Scheffield 1949) or like or dislike of a radio program (Peterman 1940). In advertising, ASI research uses a dial in theater tests to indicate like-dislike and Market Facts' new TRACE technique involves taking respondents back through a commercial asking them to explain why par- ticular responses were given. Academics have also re- cently used dial turning devices to assess degree of af- fective response (Friestad and Thorson 1984). The warmth monitor is in this measurement tradition but, of course, focuses upon measuring a feeling response- warmth-rather than liking.

The warmth monitor involves a respondent moving a pencil down paper while viewing a commercial, mov- ing it to the left and right to reflect how warm his or her feelings are at any given time. As one commercial ends and another begins the respondent draws a hori- zontal line. The directions and a sample are shown in Figure A. Note that the respondent is instructed to keep a constant rate of speed down the page and not look down while watching the commercials. Experience with using the warmth monitor on several hundred subjects suggests that it is easy to understand and use. In fact, with practice most respondents do not need to look down at all and others do so only rarely. Subjects are typically asked to practice the warmth monitor on three sets of two commercials each. The practice serves to make respondents comfortable with the warmth mon- itor, familiar with the scale anchors (to reduce cognitive

FIGURE A

WARMTH MONITOR DIRECTIONS AND SAMPLE

We want to monitor your feelings of warmth as you watch some sequences of two commercials each. Please chart how warm you feel: how tender, warmhearted, and sentimental. Note the labels on the warmth monitor which go from "absence of warmth" on the left to "emotional" on the right.

Please move your pencil down the paper as you watch the com- mercials, moving it to the left and right to reflect the warmth you feel at any given moment. Draw a line across your chart at the breaks between commercials as the sample illustrates.* REMEMBER: When filling out the warmth monitor, we just want to know how warm you feel, not how warm or good you think the com- mercial is. There is no right or wrong answer. Just chart how you feel as best you can.

SAMPLE WARMTH MONITOR

Absence of Warmhearted Emotional Warmth Neutral Tender (moist eyes)

PARKAY

MCDONALD'S

Subjects are also shown a sample warmth monitor which is described, and told to try to keep a constant rate of speed down the page and not look down at the paper while watching the commercials.

processing), and capable of moving down the page at a steady, comfortable rate.

The warmth monitor has four scale anchors. The ex- treme right is "emotional (moist eyes)," which repre- sents an extreme but not unreachable advertising-in- duced feeling. The next anchor, "warmhearted tender," is less intense and involves words related to the warmth construct-warmhearted and tender-that are also in the instructions. It was difficult to introduce a left an- chor that would not introduce another dimension to the scale. For example, prior audience perception stud- ies suggested that one possible antonym-irritating- in fact tapped another dimension (Aaker and Bruzzone 1981). Holbrook and Westwood (1983) also found dif- ficulty in replicating the emotional opposites suggested by Plutchik (1980). This difficulty follows Cattell's (1960) argument that bipolarity of emotions is difficult to establish, leading Cattell to conclude that emotions are best viewed as unipolar. Therefore, "absence of

WARMTH IN ADVERTISING 369

warmth," was used to provide a "zero"' point of refer- ence. The neutral anchor introduces a base-line level that is not associated with any particular absence or stimulation of warmth.

The physical height of the raw tracing is measured at five equally spaced points (except where noted) during the commercial. Thus, the rate at which respondents move down the page is standardized. A zero height is assigned to "absence of warmth" and an 80 to the max- imum "moist eyes."

Reliability, Sensitivity, and Validity of the Warmth Monitor

Several analyses and tests to examine the reliability, sensitivity, and validity of the warmth monitor are re- ported in more detail in the Appendix. Test-retest cor- relations of 0.81 were found between first and second exposures of warm commercials when they were pre- ceded by a similar commercial (i.e., both were preceded by a humorous commercial). A degree of convergent and discriminant validity was shown, because warmth change had significantly higher correlations (averaging 0.45 over 10 commercials) with a post-exposure warmth scale than with post-exposure humorous, informative, or irritating scales (for which the correlations averaged 0.22, 0.16 and -0.36, respectively).' In one measure of sensitivity, the standard deviation of the warmth mon- itor for a set of over 1,200 observations involving 10 commercials averaged 8.6 on the 80-point scale.

Face validity for the warmth monitor was demon- strated in a study in which subjects indicated that the warmth monitor reflected their mood better than two alternative measurement approaches. The warmth monitor itself was shown to have little reactive impact upon arousal during exposure to a commercial or upon post-exposure measures of commercial impact.

STUDY 1: WARMTH AND AROUSAL

The first hypothesis, motivated by the proposed def- inition of warmth, was that warmth is accompanied by physiological arousal. To test Hypothesis 1, warmth monitor output was correlated with a companion mea- sure of skin response. The warmth monitor is used as the self-report measure of warmth because it is contin- uous and thus can provide within-subjects correlations of warmth over a single stimulus. Skin response is used as the physiological measure for a number of reasons. First, skin response has a clearer interpretation than other measures in that it is unidirectional-rising for any type of arousal, while other measures rise for some

emotions and fall for others (see Kroeber-Riel 1979, pp. 242-243). Therefore, with all other effects constant, we can expect positive correlations between skin re- sponse and the warmth monitor for warm stimuli (both will rise) and zero or negative correlations for non-warm stimuli (only skin response rises). Second, Rogers (e.g., Mewborn and Rogers 1979) has found positive corre- lations between skin response and a self-report measure of fear in a persuasive communication. Third, Kroeber- Riel (1 979) found skin response sensitive to advertising stimuli.

Design and Procedure The warmth monitor and skin response were mea-

sured for 30 subjects in two phases who viewed com- mercials drawn from a pool of 16 ads.2 In the first phase 24 subjects viewed four commercials each, and in the second phase six subjects viewed eight commercials each. Skin response was measured using the Narco Biosystems Physiograph recording of second-by-second averages of ohms of skin resistance with electrodes at- tached to the first and third fingers of the nondominant hand (the dominant hand was used for warmth monitor responses). It was not necessary to adjust resistance amplitudes for individual differences in initial resistance levels (c.f., Kroeber-Riel 1979, pp. 243-244) because the data were only used for within-subjects analyses; thus analyses were independent of baseline effects.

Fourteen test commercials were selected from a larger set of 40 commercials which, by design, included rep- resentatives of four commercial types: warm, humor- ous, informative, and irritating. The 40 commercials (selected on the basis of Bruzzone Research Corporation ratings, BRC; see Aaker and Bruzzone 1981) were shown to 92 student respondents who did not partici- pate in the subsequent study. Students rated each com- mercial on the four dimensions representing each ad type as well as liking of the ad using 9-point scales (from "not at all" to "very" on the appropriate adjectives). The 14 selected test commercials, four of each type (ex- cept only two informative), were all perceived as high on one dimension and low on the others. The ratings are given in Table 1. Scheffe's test was used to test whether the rating of the adjective of interest (e.g., warm

'Thanks to the large number of observations, the power of this test was quite high. Cohen (1969, p. 122) computes that the probability of finding a significant difference when a "small" true difference really exists is above 90 percent.

2The use of real commercials in these studies was desirable, if not necessary, for external validity reasons. To be meaningfully tested, several of the hypotheses required stimuli that approximate real commercials. In addition, it is unlikely that experimenter-produced commercials could generate the level of feeling available from real ads. There are, of course, inevitable confounding elements in the use of real ads caused by product class, length, casting, script, and count- less other subtle commercial characteristics. The use of replicated commercials of each type in Study 1, replicated warm test commercials in Studies 2 and 3, and replicated manipulation commercials in Study 3 were introduced to minimize confounding. However, it should be recognized that the trade-off between external and internal validity remains.

370 THE JOURNAL OF CONSUMER RESEARCH

TABLE 1

MEAN RATINGS FOR 14 TEST COMMERCIALSa

Perceptions that the commerfcias are

Commercial type Warm Humorous Irritating Informative

Warm 6.17 3.75 3.75 3.34 Humorous 4.78 6.83 3.23 4.65 Irritating 2.12 2.67 6.43 4.52 Informativeb 2.37 2.25 3.20 7.21

a Perceptions were obtained from 92 respondents who scaled each commercial on a nine. point Ukert scale.

bShell and Amercan Express were not hIcluded in the 40 test comrrwcials, but were ranked in the top 1% of a sample of prime time commercials using BRC measures.

for McDonald's) was higher than the other ratings for that ad in Table 1. Forty of the 42 possible comparisons (three each for the 14 ads tested) were significant using a experimentwise error rate of 0.05. Two additional in- formative commercials were chosen solely on the BRC data due to the inability to find appropriate ads from the original set of 40. A total of nine subjects, six from phase one and three from phase two, saw each of the 16 commercials.

Results

The average within-subjects correlations across the nine subjects for each commercial between the warmth monitor and skin response are given in Table 2. Each commercial was divided into 10 equally spaced inter- vals, with the warmth monitor and skin response mea- sures taken at the beginning of each interval and at the end of the commercial. Thus the correlation between the warmth monitor and skin response for a commercial was over 11 data points distributed evenly through the commercial.

It was expected that for the four warm ads there should be a substantial positive correlation between the warmth monitor and skin response. In fact, the average correlations-0.73, 0.65, 0.76, and 0.54-are all sub- stantial and significant. The average over the four com- mercials is 0.67 (which should be compared to a test- retest reliability of 0.81 reported in the Appendix).

It was also expected that arousal for humorous and informative commercials, if it occurs at all, will be as- sociated with constructs other than warmth. Thus, low correlations were expected between warmth and skin response for the humorous and informative commer- cials. The Table 2 correlations for informative com- mercials are very close to zero, in part because there was little variation in skin response. However, the av- erage correlations for the four humorous commercials were all significantly positive and averaged 0.31, sug- gesting that the humor in these ads may generate some warmth even though the commercials were not origi-

TABLE 2

CORRELATIONS BETWEEN SKIN RESPONSE AND THE WARMTH MONITOR

Average Commercials within-subjects correlation

Warm

Lowenbrau8 .73 Gallo 1 .65 McDonald's .76 Gallo 2 .54

Average across warm adsb,c .67

Humorous

Parkay 1 .42 Busch .31 Parkay 2 .36 Peugeot .16

Average across humorous ads .31

Irritating

Clorox .01 Stayfree -.10 Calvin's -.12 Jack-in-the-Box .06

Average across irritating ads -.04

Informative

Shell .19 American Express -.16 Bell Telephone .26 PG&E -.21

Average across informative ads .02

n = 9 in each advertisement. bn = 36 for the set of commercials. ' For all differences between the averages of the commercial sets (.67, .31, -.04, and .02),

except those between irritating and informative, p < 0.05.

nally rated as being especially warm (4.78 on the 9- point scale).

There was an expectation that irritation might have some tendency to be negatively correlated with warmth. The highest irritation levels, which would be expected to generate the highest arousal levels, should precipitate a movement to the far left (less warmth) on the warmth monitor. Further, any rightward (more warmth) move- ment on the warmth monitor to the neutral position or beyond was expected to accompany reduced irritation and thus less arousal. However, the Table 2 correlations for the irritating commercials, while variable, are es- sentially zero, again in part because there was little variation in skin response for this set of commercials.

In summary, Hypothesis 1 was largely confirmed, and it appears justified to view warmth as accompanied by physiological response. The low correlations for irri- tating and informative commercials suggest that some

WARMTH IN ADVERTISING 371

spurious factor such as physical activity is not creating the high correlations. We now turn to a discussion of Studies 2 and 3, which address the remaining objectives and hypotheses.

STUDY 2

Design The purpose of Study 2 was to test Hypothesis 2 (the

specificity and volatility of warmth), Hypothesis 3 (se- quence effects), Hypothesis 4 (warmth and ad/brand attitudes), and Hypothesis 5 (warmth and recall). Sub- jects were shown two sets of three commercials. One set contained two warm 30-second manipulation ads followed by a warm test ad, while the other set had two humorous 30-second manipulation ads followed by a different warm test ad. Commercials were chosen to be either warm or humorous based upon perceptions of a national consumer survey conducted by Bruzzone Re- search Corporation (BRC; Aaker and Bruzzone 1981). All commercials chosen were in the top 1 percent of ads on the appropriate scale. The two test commer- cials-a Kodak 30-second commercial featuring Mi- chael Landon and his son and a McDonald's 60-second commercial portraying a family reunion-were chosen both for their high scores on the warmth factor and for the relevance of products to subjects. The four manip- ulation commercials were selected because of their high scores on the warmth factor (Lowenbrau and Bell Tele- phone) or on the entertaining/humorous factor (Parkay and Chef's Blend).

Sixty-seven undergraduate marketing students rep- resenting 70 percent of a subject pool whose partici- pation was required as part of an introductory market- ing class were randomly assigned to one of two sequence conditions. One condition had the two warm manip- ulation commercials before the McDonald's test com- mercial and the two humorous manipulation commer- cials before Kodak, while the second condition had the warm ads before Kodak and the humorous ads before McDonald's (n = 32 and 35, respectively). Order was counterbalanced by half of the subjects in each condi- tion viewing the McDonald's commercial sequence first and the other half viewing the Kodak sequence first. There was an interlude of approximately 30 seconds between sequences while the experimenter switched tapes and subjects turned to a clean warmth monitor page.

In addition to applying the warmth monitor, several post-exposure measures were obtained to test Hy- potheses 3, 4, and 5. Aad was first measured for all six commercials. Attribute evaluations, purchase inten- tions, and recall of copy points were then (in that order) obtained for the two test brands. Commercial liking (Aad) was measured using a 6-point scale ("did not like the ad at all" to "liked the ad very much"). Attribute evaluations were measured using 6-point scales ("very

descriptive" to "not at all descriptive"). For both test products, four attributes were selected to be most rel- evant to the test commercials. The selection was based on experimenter judgment and five focus groups (with different students not exposed to the test commercials) in which attributes of the relevant product class affecting choice were elicited and discussed. Purchase intentions were measured by asking subjects to identify which of four brands in a product category (one being the test brand) he or she would be most likely to purchase (e.g., McDonald's, Burger King, Wendy's or Jack-in-the- Box). The brands were selected based on a pre-test that showed that the test brand (McDonald's for fast food and Kodak for cameras) would have approximately a 0.5 probability of being selected. Copy point recall was tested using a 1 0-item test in which subjects were asked to check recalled copy points (five were correct and five not correct). Recall was coded for analysis as items checked (or not checked) correctly minus items checked (or not checked) incorrectly.

Procedure Subjects were run in groups limited to five at a time

(the minimum was three) on weekday evenings over a two-week period in a college lounge room, with couches and carpeting resembling a living room situation. A comfortable setting and small group were used to sim- ulate an in-home experience. Subjects were told that the purpose was to get their honest reaction to the ads. They then familiarized themselves with the warmth monitor through reading the instructions, asking ques- tions, and using it with two sequences of three com- mercials each; these were selected to be neither very warm nor humorous (based on the BRC data and ex- perimenter judgment). Subjects then viewed the two sequences of three commercials separated by approxi- mately a 30-second break and completed the post-ex- posure questionnaire.

Results Results for Study 2 were tested using ANOVA for

sequence and order effects. Only sequence effects are discussed because no significant order or interaction effects were found. Figure B gives mean warmth mon- itor results for the two test commercials after either the warm or humorous manipulation ads. The two lines for each commercial represent the mean warmth level during the ad for the two different types of manipulation commercials. For example, the lower line for the Kodak test ad represents the mean warmth monitor levels at the end of each 7.5 second interval during the ad for the 32 subjects who saw Kodak after the humorous commercials.

In a manipulation check, initial warmth levels for the test ads with prior warm manipulation ads were found to be significantly higher (p < 0.001) than when

372 THE JOURNAL OF CONSUMER RESEARCH

FIGURE B

EFFECTS OF HUMOROUS AND WARM PRIOR COMMERCIALS ON WARMTH FOR STUDY 2

Warmth Level Warmth Level

60 KODAK 60 MCDONALD'S

55 - 55 - X

50- 50K 45 -

. 45 -

40 - 40 -

35 _0> 35 -

30< 30 4

O I I I l l o 0~~~~~~~~~~ 0 7.5 15 22.5 30 0 15 30 45 60

Seconds into Commercial Seconds into Commercial

-a-- Humorous

- -&-- -Warm

NOTE: For differences at the start, p < 0.001.

the prior manipulation ads were humorous. To examine sequence effects (Hypothesis 3) the 5 warmth monitor measurement points (3 quartiles and two end points) were treated as a repeated measure in the analysis. The interaction between sequence and time provides a test for Hypothesis 3.1 -that there will be differences in the trend in warmth responses for the test ads after the warm versus the humorous manipulation ads. For both test commercials the interaction was significant at the 0.001 level. As is apparent from Figure B, this interaction is the result of a more positive trend in warmth that oc- curred when a test commercial followed the humorous manipulation ads than when it followed the warm ma- nipulation ads. Interestingly, the end levels are not dif- ferent across sequence for either test ad (p > 0.10).3

Two additional observations can be made from Fig- ure B. First, a rapid and substantial change in warmth occurred during a matter of seconds, supporting Hy- pothesis 2-that warmth is volatile and stimulus-re- sponsive. Warmth in the McDonald's commercial that followed the humorous ads changed significantly (p < 0.001; t-test) during the first 15 seconds and again during the second 15 seconds (p < 0.001; t-test). After 30 seconds most of the change in warmth had occurred. In the Kodak commercial that followed the warm ads the response was even faster, a similar change (p < 0.001) occurring in a 7.5-second interval. These re- sults suggest that warmth is volatile-capable of chang- ing within seconds.

TABLE 3

EFFECT OF SEQUENCE ON CHANGE OF WARMTH, LIKING OF COMMERCIAL, INTENTIONS, AND RECALL (STUDY 2)

McDonald's after Kodak after Impact of

advertisement Humorous Warm Humorous Warm

Change of warmth +14.9 -5.9d +1.8 -14.0d

Liking (1 -dislike) (6-like) 4.7 3.8c 3.6 3.0b

Recall (0-10) 5.9 4.9a 6.6 5.0c

Product choice (binary) 40% 30% 44% 34%

N 35 32 32 35

*p < 0.10. bp < 0.05.

Cp < 0.01. dp < 0.001.

Second, note the drop in warmth during the last quarter of the McDonald's ad. The McDonald's ad was very warm and emotional, containing tears, laughter, touching, happiness, and sadness. The identification of the sponsor, which occurred for the first time during the last few seconds, seemingly caused the warmth drop. This example illustrates the potential diagnostic insight offered by the warmth monitor.

Analysis of the post-exposure measures provides tests of Hypothesis 3.2-sequence effects on Aad, Hypothesis 3.3-purchase likelihood, and Hypothesis 3.4-recall. There was no overall impact of sequence upon the at- tribute ratings. Only one of the eight t-tests (four attri- butes for each test brand) was significant at the 0.10 level. However, as shown in Table 3, a very clear se- quence effect supporting the adaptation or contrast ef- fect emerged for the other three dependent variables, supporting the hypotheses. The test commercials were significantly better liked, received higher recall scores, and (although not significant) the test brands were cho- sen more often when the test commercial was preceded by the humorous than by the warm ads. The effect of sequence without corresponding changes in attribute ratings is consistent with Mitchell and Olson (1981), who found an ad execution effect independent of at- tribute belief change.

Table 4 presents the correlations of three warmth monitor measures (the end level, average level, and change defined as end less start) with the post-exposure measures, testing Hypotheses 4 and 5. The between- subjects correlations represent the mean of the corre- lations for each treatment group averaged across the ads. The within-subjects correlations are only mean- ingful for liking because data were obtained for only the two test ads on the other measures. The correlations

3The anchors were selected to be extreme enough to avoid possible floor or ceiling effects. In fact, the average value never got above 60 or below 20 on the 80-point scale in the three studies.

WARMTH IN ADVERTISING 373

TABLE 4

CORRELATIONS BETWEEN WARMTH MONITOR MEASURES AND EFFECTS (STUDY 2)

Liking of Likelihood Warmth measure commerciala RecalIa of purchasea

Between subjectsb End state .30c .06 .11 Change .25c .08 .07 Average .25c .04 .13

Within subjectsb End state .476d e 6e

Change .54d _e _e

Average .28c 6e _e

For all 6 commercials for liking and two warm test commercials for recall and likelihood.

Between-subjects correlations are the mean across commercials. Within-subjects cor-

relations are the mean across subjects. c p < 0.01.

dp < 0.001.

* Not meaningful due to measuring the variable for only two commercials per subject.

between warmth and liking for the six commercials in the study were substantial and significant, indicating a strong relationship between warmth and liking of ads, supporting Hypothesis 4.1. The low correlations for purchase likelihood, which fail to support Hypothesis 4.2, may be due to the relative insensitivity of the binary purchase likelihood measure. The low correlations for recall, which fail to support Hypothesis 5, suggest that warmth (and perhaps its correlate, arousal) may not be significantly related to recall, at least for warm ads. The correlation patterns indicate that all three measures of warmth (end, change, and average) are similar in terms of explaining variance in liking, although average warmth is significantly lower than the other two for the within-subjects correlations.

Although these results provide some insight into the usefulness of the warmth monitor, the volatility of warmth levels, and the presence of sequence effects in advertising, there are several apparent problems. First, the manipulation commercials were not matched on variables other than the warmth-humor dimensions. This allows for possible confounding effects such as product category/ad execution confounds. In particular, there are possible confounding effects with liking. For example, the second humorous manipulation com- mercial was less well liked than the comparably placed warm ad-4.04 versus 3.06 on a six-point like/dislike scale. Therefore, the sequence effects found in the study may have been due to differences in liking of the ma- nipulation ads rather than differences in warm versus humorous execution. Second, the manipulation and test commercials were selected based primarily upon the BRC data, which were based upon a different popula- tion. Third, the use of only warm and humorous ma- nipulation ads and only warm test ads restricts the gen- eralizability of the results to other execution strategies, especially since both are positive feelings.

STUDY 3 The third study, designed to replicate and extend the

second, involved several important differences. First, rather than relying on the BRC data, commercials were selected based upon perceptions of subjects comparable to those in the study. The set of 40 candidate commer- cials mentioned in the description of Study 1 were rated on warmth, humor, irritation, informativeness, and liking by 92 members of the subject pool in a prior term. Second, we felt it was important to control the manipulation (prior) commercials for liking by match- ing the warm and humorous manipulation commercials on liking based upon the prestudy ratings. As discussed previously, in the second study the manipulation com- mercials differed on liking as well as on execution, which provided a potential confounding effect.

Third, irritating manipulation commercials were in- cluded to allow the introduction of negative feelings. There were two purposes for this: (1) to provide a better test of Hypotheses 2, 4, and 5 by generating more vari- ation in warmth, and (2) to provide a liking manipu- lation, since the humor and warmth manipulation ads were now controlled for liking. Fourth, a single prior (manipulation) commercial was used instead of two ads. The first study demonstrated that emotion levels were sufficiently volatile that a single 30-second commercial was adequate to establish prior feeling levels. However, to reduce the possibility that effects were due to extra- neous characteristics of the manipulation commercials such as product category, casting, or setting, two ma- nipulation commercials of each type (warm, humorous, and irritating) were used. Fifth, the experiment was again replicated using two warm test commercials; however, in addition, the study was also extended through using a humorous and an irritating test com- mercial. The goal was to assess whether the findings for warm ads held over other types of commercials.

Design

Six manipulation commercials (two humorous, two warm, and two irritating) and four test commercials (two warm, one humorous, and one irritating) were used. Two warm test commercials were used because of the focus upon the warmth construct and to better replicate Study 2. The 24 possible manipulation-test commercial sequences (6 manipulation ads X 4 test ads) were divided into six groups, each having four sequences (a sequence being a manipulation ad followed by a test ad) involving the four test ads (no two sequences in one group had the same manipulation commercial). Thus, each test ad was seen after each manipulation ad by one-sixth of the subjects.

Subjects were 101 members of the same subject pool as Study 2, but in a subsequent school term (over 80% of pool members participated). Subjects were randomly assigned to one of the six commercial groups. Therefore,

374 THE JOURNAL OF CONSUMER RESEARCH

each subject saw eight commercials (four sequences) including all four test ads. The four test commercials were the McDonald's commercial used in Study 1 (warm), a Gallo wine commercial depicting a reunion of a couple (warm), a Peugeot commercial using tennis player Vitas Gerulitas and his father (humorous), and a Calvin Klein commercial with Brooke Shields (irri- tating). The two warm manipulation commercials were the Lowenbrau and Bell Telephone ads from Study 1. The two humorous commercials were the Parkay mar- garine ad from Study 2 and a Busch beer commercial. The two irritating manipulation ads were for Clorox Bleach and Stayfree, a feminine hygiene ad that was rated equally irritating by both sexes in the prestudy. The ratings of the commercials are given in Table 5.4

All 10 commercials were 30 seconds long except for the McDonald's and Gallo test commercials and the Busch beer manipulation ad (which were 60 seconds long). The decision to include ads of varying lengths was made to allow for better selection on the primary criteria. Because the primary analyses were done sep- arately for each test commercial, different lengths of test ads had no effect. Note that the two warm test com- mercials, which replicate Study 2, were of equal length.

Measures

Several measures were added in Study 3. First, a pur- chase likelihood question using a 6-point bipolar scale ("very likely" to "not at all likely" to purchase the brand if a purchase were made in the product class) was added. It was felt that a 6-point scale, through allowing greater variation in response, would be more sensitive than the binary forced choice measure used in Study 2. Second, subjects rated the test commercials on the dimensions of warmth, humor, irritation, and informativeness to provide manipulation checks and as additional mea- sures of feeling to test for convergent and discriminant validity of the warmth monitor (see the discussion in the Appendix). These ratings were obtained after other post-exposure measures to avoid any biasing effect on the prior measures. Third, as a check for effects of prior exposure a question was added at the end of the study to determine whether subjects recalled having seen the commercials before (all ads except Busch beer had aired nationally, but none had been on the air during the year preceding the study).

Procedure

The procedure was similar to that in Study 2. Groups of from two to five subjects were run on weekday eve-

TABLE 5

RATINGS FOR STUDY 3 COMMERCIALS ON ADJECTIVE SCALES

Irrita- Inform- Commercial type Warmth Humor tion ative Liking

Test Warm 6.90 3.14 4.13 2.56 4.85 Humorous 5.32 6.83 2.27 5.90 6.86 Irritating 2.27 2.79 6.39 2.59 3.51

Manipulation Warm 6.48 2.54 3.62 3.45 5.34 Humorous 4.57 6.74 3.89 3.98 5.61 Irritating 1.86 2.25 6.87 4.78 2.48

Mean for 40 commercials tested 4.04 3.77 4.08 4.89 4.87

NOTE: N = 92.

nings over a three-week period in the same lounge room used in Study 2. Subjects were randomly assigned to one of the six treatment groups, and the order of se- quences within each group of commercials was ran- domly changed to assure that each sequence was viewed in each possible position. Subjects were again told that they would be shown a series of commercials and that the intent was only to get their honest reactions. They then familiarized themselves with the warmth monitor through practicing with three sequences of two com- mercials each. These practice commercials were selected to be not high on warmth, humor, or irritation. As shown in the chart below, a viewing of two sequences of commercials was immediately followed by a ques- tionnaire. The third and fourth sequences were then shown after a short one- to two-minute break, followed by the administration of a second questionnaire:

P Practice warmth monitors

XI First sequence (a manipulation and test ad)

X2 Second sequence

01 Questionnaire on XI and X2 commercials

X3 Third sequence

X4 Fourth sequence

02 Questionnaire on X3 and X4 commercials

Results

Before examining the results for the study an initial test was made for possible demand effects created by the intermediate rating procedure. Since only three of 36 measures differed in first (Qi) and second (02) mea- surement positions (p < 0.05, t-test), it was decided

4Scheffe's test was used to determine whether the rating of the ad- jective of interest (e.g., warm for McDonald's) was higher than the other adjective ratings for that ad in Table 5. All 30 of the possible comparisons (three each for the 10 ads used) were significant using an experimentwise error rate of 0.05.

WARMTH IN ADVERTISING 375

that all results could be combined for analysis. Figure C gives mean warmth monitor results for the four test commercials (Table 6 gives associated n's). As in Figure B, the lines represent the mean warmth level during the test ad for each of the different types of manipulation ads. For example, the lowest line for the McDonald's ad (the upper left graph) represents the mean warmth monitor level during the McDonald's commercial for subjects who saw it after one of the two irritating ads.

A series of ANOVA tests were conducted on the Fig- ure C results. As in Study 2, the five warmth monitor measurement points were used in the analysis as a re- peated measure dependent variable. In the first test, both the main effect of manipulation ads (MAD) and the interaction of MAD with time (T; the repeated measure of warmth) were significant (p < 0.001) for all four test commercials. The main effect shows that the manipu- lation ads were successful in creating different warmth levels. The interaction indicates support for Hypothesis 3.1 -that there was a different trend in warmth after the different manipulation ads.

In the second test, there was no significant difference in the effect of the different manipulation ads of each type as assessed by the interaction of the nested manip- ulation ad (MAD; e.g., Parkay and Busch Beer) within manipulation type (MTYP; e.g., humorous) with time [MAD(MTYP) X T]. For all four test ads, the interac- tion was nonsignificant (p > 0.10), suggesting that the two manipulation ads of each type had similar effects and that extraneous characteristics of the manipulation ads were not responsible for the effects.5

A number of other observations can be made from Figure C. First, the rapid change in warmth was even more pronounced than in Study 2 because the prior warmth level generated by the irritating commercial was lower. In the two warm test commercials, the movement exceeded 15 units on the 80-point scale during the first 15-second period of the commercial, and in the hu- morous commercial the movement exceeded 10 units following the irritating commercials (versus seven and eight for McDonald's and Kodak in Study 2 following the humorous and warm ads, respectively). These movements were all significant at the 0.001 level (t- test). In the humorous and irritating commercials a sig- nificant (p < 0.001 and p < 0.05 for humorous and irritating) movement was observed after 7.5 seconds following the irritating and warm commercials, respec- tively. Thus, the support found in Study 2 for Hypoth- esis 2-that warmth is specific and volatile-was rep- licated. Second, the end warmth level was not signifi- cantly affected by the prior commercial for any of the test ads. Thus, the prior commercials' effect on warmth dissipates after about eight to 30 seconds. Third, the drop in warmth during the last quarter of the Mc-

FIGURE C

EFFECTS OF HUMOROUS, WARM, AND IRRITATING PRIOR COMMERCIALS ON WARMTH FOR STUDY 3

Warmth Level Warmth Level 60 - 60

MCDONALD'S GALLO 55 - (Warm) 55 _ (Warm)

50 50

45 - 45 -

40 - 40 -

35 - 35

I1 / 30 - 30 _

25 _25 d

o 1 I I I I _ _ _ _ _ _ _ _ _ _ _ 0 15 30 45 60 0 15 30 45 60

Seconds into Commercial Seconds into Commercial

Warmth Level Warmth Level 60 60

PEUGEOT CALVIN KLEIN 55 _ (Humorous) 55 _ (Irritating)

50 _ 50 _

45 - 45 -

40 - 40 -

35 _ 35

30 - ' 30 -

_0- 25 - 25 _ -

0 ~~~~~~~~~0 0 7.5 15 22.5 30 0 7.5 15 22.5 30

Seconds into Commercial Seconds into Commercial

Humorous - - Warm - --Q-- Irritating

NOTE: Irritating different from warm and humorous at p < 0.0001.

Donald's ad, presumably caused by sponsor identifi- cation, was replicated.

Fourth, the warm and humorous manipulation com- mercials generated approximately the same warmth monitor levels even though the humorous ads received lower warmth ratings as reported in the Table 5 pretest (4.57 versus 6.48; p < 0.001, t-test). There are4a number of possible explanations. First, the warmth monitor, by focusing subjects' attention on one construct- warmth, may artificially increase warmth responses to the humorous ads. In contrast, the pretest provided the respondent with five items (warmth, humor, irritation, informativeness, and liking), and warmth was not sim- ilarly highlighted. Second, the large group and austere classroom conditions of the pretest (recall that condi-

5Further, the mean warmth monitor responses were about the same for each of the two manipulation ads of each type. For example, the average McDonald's test ad values were 51.8 and 53.1 for the two humorous manipulation ads.

376 THE JOURNAL OF CONSUMER RESEARCH

TABLE 6

EFFECT OF SEQUENCE ON POST EXPOSURE MEASURES (STUDY 3)a

Likelihood Recall of Liking of of copy

Sequence N ad purchase points

McDonald's (warm) after

Warm 32 3.88d 3.09c 5.08b

Humorous 34 4.65 3.59 5.94 Irritating 35 4.80 3.89 5.72

Gallo (warm) after Warm 34 3.65d 2.74c 4.76c Humorous 32 4.25 3.28 5.88 Irritating 35 4.43 3.34 5.78

Peugeot (humorous) after

Warm 35 4.71 2.54 7.54 Humorous 31 4.26d 2.26 6.58c Irritating 35 4.80 2.63 7.48

Calvin Klein (irritating) after

Warm 34 2.56 3.50 4.88 Humorous 35 3.00 3.17 5.14 Irritating 32 3.28b 3.91 b 5.62

a Each sequence represents the mean of the prior commercials of each type (warm, etc.) used before each of the four test commercials (e.g., McDonald's after warm represents the 16 subjects who viewed McDonald's after Lowenbrau and the 16 subjects who viewed McDonald's after Bell Telephone).

bDifferent from two other means at p < 0. 10. Different from two other means at p < 0.05. Different from two other means at p < 0.02.

tions of the pretest were quite different from that of the study) may have suppressed warm responses to the hu- morous ads. However, there were sequence effects be- tween warm and humorous manipulation commercials, as will be discussed in the following section. Third, the Study 2 sequence effects on the warmth monitor levels may have been in part the result of differences in liking of the manipulation commercials (which was controlled in Study 3).

Sequence Effects on Post-Exposure Measures

Table 6 presents the impact of sequence on the post- exposure measures. The data represent the mean scores for each measure across the subjects who viewed the test commercial after one of the two manipulation commercials of each type. For example, the first row represents the mean levels for liking, likelihood, and recall for the McDonald's test commercial for subjects who saw it after either the Lowenbrau or Bell Telephone commercial (the two warm manipulation ads).

ANOVA was again used to test Hypotheses 3.2, 3.3, and 3.4, which predicted an effect of sequence on post- exposure measures consistent with an adaptation or

contrast explanation. The results for each test ad fol- lowing manipulation ads of the same type (e.g., a warm test following warm manipulation ads) were contrasted with those for each test ad following ads of a different type (e.g., a warm test ad following humorous or irri- tating ads). The effect of individual manipulation ads (MAD) within type of manipulation ads (MTYP) was again nonsignificant for all tested measures, indicating that both manipulation ads of each type had equivalent effects on the post-exposure measures.

A very clear sequence effect supporting Hypotheses 3.2, 3.3, and 3.4 was found. Both warm test commercials had significantly lower scores on all three post-exposure measures when they were preceded by another warm commercial than by either a humorous or an irritating commercial (p < 0.05, except recall for McDonald's, p < 0. 10). Interestingly, it made little difference whether the prior ad was humorous or irritating. Since the hu- morous ad is matched on liking with the warm manip- ulation ads, the result is not explained by a difference in liking. Consistent with Study 2, there were few dif- ferences found in the attribute ratings. Of the 16 attri- butes tested (four for each of the four test ads) only two differed at p < 0.10-about the chance level.

Further support for Hypotheses 3.2, 3.3, and 3.4 was obtained from the humorous (Peugeot) and irritating (Calvin Klein) test commercials. For the humorous commercial, the liking, likelihood, and recall measures were all higher when the prior commercial was not hu- morous (although the difference is not statistically sig- nificant for likelihood). For the irritating commercial the liking, likelihood, and recall were all higher when the prior commercial was irritating (although the dif- ference is not statistically significant for recall), as pre- dicted by Hypothesis 3 (when the fact that there is a polarity reversal with irritation is considered).

Providing further evidence of the sequence effect, Table 7 gives the mean post-experiment ratings from the 6-point scales for each of the different test ad types (warmth for McDonald's, humor for Peugeot, etc.). The humorous test ad (Peugeot) was rated less humorous when it followed another humorous commercial rather than a warm or irritating one (p < 0.01). A similar con- trast effect was found for the irritating (Calvin Klein) ad. As would be expected from the results in Figure C where warmth increased for the Calvin Klein ad only when it followed another irritating commercial, the ir- ritation ratings for Calvin Klein were lowest when it followed the irritating manipulation commercials (p < 0.10). The ratings for the warm commercials are in a consistent direction but are not statistically significant.

The observed sequence effects from Study 3 are con- sistent with an adaptation or contrast effect explanation. For the positive feeling ads (warm and humorous) a similar prior manipulation commercial created lower scores on the dependent measures than did different prior ads. This indicates that the test ad was more ef- fective in creating the desired effect when the audience

WARMTH IN ADVERTISING 377

TABLE 7

POSTEXPOSURE FEELING MEASURES BY COMMERCIAL TYPE (STUDY 3)a

Warmth Humor Irritation Prior

commercial Gallo McDonald's Peugeot Calvin Klein

Warm 4.38 4.89 4.83 3.91 Humorous 4.59 5.15 4.10c 3.94 Irritating 5.00 5.29 4.94 3.34b

* Measures from 6-point postexposure scales. bDifferent from two other means at p < 0.10, t-test. c Different from two other means at p < 0.01; t-test.

had "adapted" to a different prior ad approach. Further, the irritating (negative feeling) appeal created less irri- tation if the audience had "adapted" to a prior irrita- ting ad.

Relating Warmth to Commercial Impact As in Study 2, correlations between the warmth

monitor measures and the post-exposure variables pro- vide tests of Hypotheses 4 and 5-the relationship be- tween warmth and commercial impact. The results, given in Table 8, represent the mean between-subjects correlations for each experimental group-e.g., Low- enbrau (manipulation)/McDonald's (test)-and the within-subjects correlations across all test commercials. Ad liking and likelihood of buying are both significantly correlated with warmth, supporting Hypotheses 4.1 and 4.2. As in Study 2, the correlation between warmth and recall is essentially zero, failing to support Hypothe- sis 5.

Note that the correlations are substantially larger than those in Study 1. The higher within-subjects correlations for liking are probably caused by the inclusion of the irritating ads, which were almost always lower in both warmth monitor rating and liking than the other com- mercials. If the irritating commercials are deleted, the within-subjects correlations with liking fall to about the same levels as those found in Study 1 (about 0.50).

Warmth Monitor's Explanatory Value To further test the usefulness of the warmth monitor,

the warmth monitor rating was added to the post-ex- posure adjective check list rating of warmth in a step- wise manner to examine its incremental explanatory power. With liking as the dependent variable, the check list measure accounts for 45 percent of the variance, rising to 51 percent when the warmth monitor change (end less start) measure is added (significant at p < 0.01). With purchase likelihood as the dependent variable, the post-exposure measure accounts for 2.8 percent of the variance, rising to 6.8 percent when change in warmth is included (p < 0.05). These increases indicate that the

TABLE 8

CORRELATIONS BETWEEN WARMTH MONITOR MEASURES AND EFFECTS (STUDY 3)

Likelihood Liking of of

Warmth measure commercial0 Recalib purchaseb

Between subjectsc End state .55h .01 .21 Change .47h .00 .17e Average 44h 09 .16e

Within subjectsd End state .749 .06 .10e Change .679 .03 .16' Average .459 .06 .07

a For all 10 commercials.

bFor 4 test commercials. c Mean across commercials. dMean across subjects. *p < 0.10.

fp < 0.01.

0p < 0.001.

hFor each commercial, p < 0.001, except Change-Peugeot (0.003), Change-McDonald's (0.03), and Average-Peugeot (0.13).

warmth monitor can improve upon the sole predictive power of a post-exposure adjective rating.

The three warmth measures have similar predictive power, although average warmth had a within-subjects correlation with liking that was significantly lower than that found between ad liking and end or change in both studies (p < 0.01, t-test using Fischer R to Z transfor- mation). Further, the mean correlation between average warmth and the post-exposure adjective warmth mea- sure in Study 3 was 0.32 for the four test ads, versus 0.45 and 0.42 for end and change, respectively.

There was some evidence that all three measures have some unique predictive power. Using liking as the de- pendent variable, the mean percent variance explained over the four test ads rises from 15 percent with only average as an independent variable and 27 percent with only change to 34 percent when both are included (p < 0.01 for three of the four test ads). Variance explained also rises from 30 percent with end as the independent variable to 34 percent with end and change (p < 0.05 for three of the four test ads).

DISCUSSION

The Warmth Construct Warmth is here defined as a positive, mild, volatile

emotion involving physiological arousal and precipi- tated by experiencing directly or vicariously a love, family, or friendship relationship. The studies reported in the paper provide three specific results that support this conceptualization. First, as Hypothesis 1 predicted, warmth was shown to have a physiological component as demonstrated by the average correlation of 0.67 be-

378 THE JOURNAL OF CONSUMER RESEARCH

tween skin response and the warmth monitor for warm ads. Second, as Hypothesis 2 predicted, the construct as measured by the warmth monitor was found to be quite volatile. In Studies 2 and 3, felt warmth levels changed substantially within even the first 15 seconds of a commercial. Finally, the positive correlations with Aad of 0.5 to 0.7 indicate that warmth is also positive.

Sequence Effects of a Prior Commercial

Studies 2 and 3 demonstrate strong support for Hy- potheses 3.1, 3.2, 3.3, and 3.4-that a warm ad preceded by another warm ad would be less effective than a warm ad preceded by a non-warm ad. The adaptation/contrast explanation is thus supported. The effects for warm ads were replicated in both studies by the use of two warm test ads. In addition, the Study 3 results for the hu- morous and irritating test ads provide additional sup- port for Hypotheses 3.1, 3.2, 3.3, and 3.4. The humor- ous ad generated higher liking, recall, and purchase likelihood when not preceded by a humorous ad-as did the irritating ad when preceded by another irritating ad (presumably because it was then less irritating).

Relating Warmth to Commercial Impact

Support for Hypothesis 4-that warmth is related to A,d and purchase likelihood-was also found. In both Studies 2 and 3, significant and substantial correlations were found both within and between subjects for warmth and liking of the ad. Study 3, with a more sen- sitive purchase likelihood measure, contained signifi- cant correlations between warmth and purchase like- lihood. There was little evidence that any one of the three warmth measures-end state, change in warmth, and average warmth-were substantially superior with respect to these correlations. Each did provide at least some independent information.

The lack of support for Hypothesis 5-a relationship between warmth and recall-suggests that at least some form of arousal is not related to recall. However, there are two possible reasons why other researchers (e.g., Choi and Thorson 1983) have found effects of feelings on recall. First, by comparing emotional to nonemo- tional stimuli they probably created a greater range of feelings than occurred in this research, which merely compares level of warmth with recall. Second, Thorson found that ads with dramatic changes in feeling ("poi- gnant" ads) are most closely related with recall (Friestad and Thorson 1984). In contrast, the ads used in this research were chosen to be consistently of one feeling type and may have lacked sufficient within-ad feeling changes.

The Warmth Monitor The warmth monitor, developed to study the warmth

construct, has several obvious limitations. It can only

measure one construct at a time (although that construct need not necessarily be warmth), the respondent needs to learn and use the measure, and as with any scale, there are interpretation difficulties associated in part with individual differences in interpreting anchors and ability to accurately record felt warmth. However, the results in these studies have shown that continuous measurement of audience feelings can help to under- stand both the nature and effect of specific feelings; thus it is a useful pursuit. Further, such measures can be relatively inexpensive, easy to use, and portable (espe- cially in comparison to some alternative methods, such as those used by pretesting agencies).

Future Research

Clearly there is much to learn about warmth and other affective effects of advertising. What is the inter- action of warmth and repetition? Most advertising with a substantial feeling component involves heavy repe- tition. What is the relationship between warmth and other feelings/emotions such as humor or excitement? What is the role of warmth in other media?

The impact of warmth on the effectiveness of com- mercials found in this research raises a number of issues. First, are the results for the humorous and irritating ads in Study 3 replicable, and will other feeling ads such as exciting or fear ads produce similar results that support the adaptation or contrast hypothesis? Second, what is the best measure of warmth? This research provides evidence that average, end, and change levels each yield some independent information. Third, what is the pro- cess through which warmth and other responses affect commercial impact? Is warmth's impact through Aad as an intervening variable or does it operate by some other, perhaps more direct, mechanism? Finally, what is the role of such factors as past experience, empathy, and personality in creating warmth? Answers to such ques- tions should contribute to our understanding of the role of emotional response in persuasive communication.

APPENDIX

Reliability, Sensitivity, and Validity of the Warmth Monitor

Test-Retest Reliability

Test-retest reliability was measured during the course of the third experiment when 101 subjects were exposed to the two warm test commercials (McDonald's and Gallo) a second time. About 30 minutes intervened be- tween exposures, during which time post-exposure questionnaires were completed and a break was taken. The resulting average correlation between test and retest exposures when both were preceded by a similar com- mercial (i.e., both were preceded by a humorous com-

WARMTH IN ADVERTISING 379

mercial) was 0.81. The 0.81 correlation was reasonably high, especially since the retest involves a second ex- posure which can be a different experience from a first exposure.

Convergent and Discriminant Validity The measurement of warmth by means of post-ex-

posure scales as well as by the warmth monitor in Study 3 provided a convergent and discriminant validity test. The correlation across 1O 1 subjects between the change in warmth during the commercial as measured by the warmth monitor and ratings of warmth in the post- experiment adjective scale task averaged over the 10 commercials was 0.45. By contrast, the average corre- lation between the warmth monitor and ratings of hu- morous, informative, and irritating are 0.22, 0.16, and -0.36. The R = 0.67 association found in Study 1 be- tween warmth as measured by skin response also pro- vides an indication of convergent validity.

Sensitivity The warmth monitor exhibited a reasonable level of

sensitivity. The standard deviation of the warmth mon- itor during the 10 commercials used in the third study involving 101 subjects was 8.6 on the 80-point scale, a level of variation about the same as that exhibited by the McDonald's commercial in Figure A. Further, a deviation from linearity was computed by fitting a straight line between the warmth monitor beginning and ending positions and observing the sum of the larg- est positive and negative deviation from this line. In the same data base, the average value of this deviation was 8.9.

Face Validity Face validity for the warmth monitor was demon-

strated in a study wherein 28 subjects were exposed to three sets of three commercials. The warmth monitor was used to measure warmth for one commercial set. In another set, an adjective check list (ACL: tender, sentimental, affectionate, warmhearted, and kindly) was administered after each commercial. In the third set, the ACL was administered only after all three com- mercials were viewed so that subjects applied the ACL retrospectively to the first two commercials in the set. The order was systematically varied.

The warmth monitor was clearly considered the best measure by the subjects who were asked to allocate 1 00 points among the three approaches reflecting how well each "actually measured your mood."6 A total of 75

percent preferred the warmth monitor (significantly more than the equal preference hypothesis of 33 per- cent, p < 0.001), whereas only 21 percent preferred the use of the ACL after each commercial, and only 4 per- cent preferred the ACL at the end of the three com- mercials. Subjects commented upon the value of a con- tinuous measure (rather than a summary measure) and upon the intrusiveness of translating feelings into words.

Reactivity Tests

A final issue with any concurrent measure is the ex- tent to which the act of measurement influences the construct being measured. Two experiments were con- ducted to assess the reactivity of the warmth monitor. One assessed the effect of the warmth monitor on post- exposure measures while the other assessed the impact of the warmth monitor on arousal during exposure.

In the first experiment, two sets of subjects were shown commercials using the procedure for Study 3 (n = 13 and n = 18 for the experimental and control groups, respectively).7 The experimental group used the warmth monitor while the control group did not. Of the 124 measures in the post-exposure questionnaires, only 23 differed at the 0.10 level (t-test), indicating that the warmth monitor did have some effect on the post- exposure measures (12-13 would be expected to differ by chance), although it is clear that few variables were affected. In addition, there was no apparent pattern to the variables that were affected.

The second experiment was part of the second phase of Study 1 described earlier in which six subjects viewed eight commercials each (three each viewing half of the 16 test commercials) with both skin response and the warmth monitor used to measure response. Six addi- tional subjects were used as a control group. They viewed the commercials (each viewing the same half of the 16 test ads) with only skin response used to measure response. If the warmth monitor had affected feeling responses, then it was expected that, on average, the correlations between skin response for subjects treated alike (i.e., skin response only or skin response and the warmth monitor) would be higher than correlations be- tween subjects treated differently. In fact, it was lower. The mean correlation within groups was 0.205 (repre- senting the mean across the 16 commercials for three subjects in each condition with each other, or 16 X 3 X 2 = 96 observations) and between groups was 0.216

6At the time of this study the warmth monitor was termed the mood monitor. However, mood was used consistently in the face validity study. It seems unlikely that the word "mood" would have so great an effect that a different measure would have been preferred had either "warmth" or "emotion" been used instead.

7Even though the sample size was small for each of the 124 t-tests, the power of the overall test was quite high. From Cohen (1969, p. 30) we computed that the probability of finding a significant result on any one test, assuming a "medium" true effect size, was over 50 percent. Hence, if a medium-sized reactivity effect had truly existed, we would expect over 50 percent of the t-tests to be significant. Since far fewer were actually significant we conclude that the reactivity effect was nonexistent or very small.

380 THE JOURNAL OF CONSUMER RESEARCH

(representing the mean across the 16 commercials for the 3 X 3 = 9 possible between-groups comparisons, or 144 observations).

[Received March 1984. Revised September 1985.]

REFERENCES Aaker, David A. and Donald E. Bruzzone (1981), "Viewer

Perceptions of Prime-Time Television Advertising," Journal ofAdvertising Research, 21 (October), 15-23.

Averill, James R. (1980), "A Constructionist View of Emo- tion," in Emotion: Theory, Research and Experience, Vol. 1, eds. Robert Plutchik and Henry Kellerman, New York: Academic Press, 305-339.

Bagozzi, Richard P. (1982), "A Field Investigation of Causal Relations Among Cognitions, Affect, Intentions, and Behavior," Journal of Marketing Research, 19 (Novem- ber), 562-584.

Batra, Rajeev and Michael L. Ray (1985), "How Advertising Works at Contact," in Psychological Processes and Ad- vertising Effects, eds. Linda F. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: Lawrence Erlbaum, 13-44.

Bower, Gordon H. (1981), "Mood and Memory," American Psychologist, 36 (February), 129-148.

Brooker, George (1981), "A Comparison of the Persuasive Effects of Mild Humor and Mild Fear Appeals," Journal ofAdvertising Research, 10 (4), 9-40.

Brown, T.S. and P.M. Wallace (1980), Physiological Psy- chology, New York: Academic Press.

Bush, Lynn E. (1972), "Empirical Selection of Adjectives De- noting Feelings," Journal of Supplement Abstract Service, 2, 67.

Cacioppo, John T. and Richard E. Petty (1981), "Social Psy- chological Procedures for Cognitive Response Assess- ment: The Thought-Listing Technique," in Cognitive Assessment, eds. T.U. Merluzzi, C.R. Glass, and M. Ge- nest, New York: Guilford, 309-342.

Cattell, Raymond B. (1960), "The Dimensional (Unitary- Component) Measurement of Anxiety, Excitement, Stress and Other Mood Reaction Patterns," in Drugs and Behavior, eds. L. Uhr and J.G. Miller, New York: Wiley, 438-461.

Choi, Y. and Esther Thorson (1983), "Memory for Factual, Emotional, and Balanced Ads Under Two Instructional Sets," in Proceedings of the 1983 ConJference of the American Academy of Advertising, ed. A.D. Fletcher, Knoxville, TN: University of Tennessee.

Clynes, Manfred (1980), "The Communication of Emotion: Theory of Sentics," in Emotion: Theory, Research, and Experience, eds. Robert Plutchik and Henry Kellerman, New York: Academic Press.

Cohen, Jacob (1969), Statistical Power Analysis for the Be- havioral Sciences, New York: Academic Press.

Coke, Jay S., D. Daniel Bateson, and Kathrine McDavis (1978), "Empathetic Mediation of Helping: A Two-Stage Model," Journal of Personality and Social Psychology, 36, 752-766.

Daly, Eleanor M., William J. Lancee, and Janet Polivy (1983), "A Conical Model for the Taxonomy of Emotional Ex- perience," Journal of Personality and Social Psychology, 45, 443-457.

Darwin, Charles (1872), The Expression of the Emotions in Man and Animals, London.

Davitz, Joel R. (1969), The Language of Emotion, New York: Academic Press.

Ekman, Paul, Wallace V. Friesen, and Phoebe Ellsworth (1972), Emotion In The Human Face, Elmsford, New York: Pergamon.

Friestad, Marian and Esther Thorson (1984), "The Effects of Emotion on Recall and Evaluation of Televised Pro- motional Messages," unpublished manuscript, School of Journalism and Mass Communication, University of Wisconsin, Madison, WI 53706.

Gorn, Gerald J. (1982), "The Effects of Music in Advertising on Choice Behavior: A Classical Conditioning Ap- proach," Journal of Marketing, 46 (Winter), 94- 101.

Helson, Harry (1964), Adaptation Level Theory, New York: Harper & Row.

Holbrook, Morris B. and Elizabeth C. Hirschman (1982), "The Experiential Aspects of Consumption: Consumer Fantasies, Feelings, and Fun," Journal of Consumer Re- search, 9 (September), 132-140.

and Robert A. Westwood (1983), "The Structure of Emotional Responses to Advertising," working paper, Graduate School of Business, Columbia University, New York, NY.

and John O'Shaughnessy (1984), "The Role of Emo- tion in Advertising," Psychology and Marketing, 1 (2), 45-64.

Hovland, Carl I., Arthur A. Lumsdaine, and Fred D. Sheffield (1949), Experiments on Mass Communication, Prince- ton: Princeton University Press.

Isen, Alice M., Thomas E. Shalker, Margaret Clark, and Lynn Karp (1978), "Affect, Accessibility of Material in Mem- ory and Behavior: A Cognitive Loop?" Journal of Per- sonality and Social Psychology, 36 (1), 1-12.

Izard, Carroll E. (1977), Human Emotions, New York: Plenum.

Johnson, Eric and Amos Tversky (1983), "Affect, General- ization, and the Perception of Risk," Journal of Person- ality and Social Psychology, 45, 20-31.

Kroeber-Riel, Werner (1979), "Activation Research: Psycho- biological Approaches in Consumer Research," Journal of Consumer Research, 5 (March), 240-250.

(1984), "Emotional Product Differentiation by Clas- sical Conditioning," in Advances in Consuimer Research, Vol. 11, ed. Thomas C. Kinnear, Ann Arbor, MI: As- sociation for Consumer Research, 538-543.

Lacey, J.I. (1967), "Somatic Response Patterning and Stress: Some Revisions of Activation Theory," in Psychological Stress. Issues in Research, eds. M.H. Appley and R. Trunbull, New York: Appleton.

Lazarus, Richard S., Allen D. Kanner, and Susan Folkman (1980), "Emotions: A Cognitive-Phenomenological Analysis," in Emotion: Theory, Research, and Experi- ence, eds. Robert Plutchik and Henry Kellerman, New York: Academic Press, 189-217.

Lutz, Richard J. (1985), "Affective and Cognitive Antecedents of Attitude Toward the Ad: A Conceptual Framework," in Psychological Processes and Advertising Effects, eds. Linda F. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: Lawrence Erlbaum, 45-64.

Mandler, George (1975), Mind and Emotions, New York: John Wiley.

WARMTH IN ADVERTISING 381

Mewborn, R.C. and R.W. Rogers (1979), "Effects of Threat- ening and Reassuring Components of Fear Appeals on Physiological and Verbal Measures of Emotion and At- titudes," Journal of Experimental Social Psychology, 15, 242-253.

Mitchell, Andrew A. and Jerry C. Olson (1981), "Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitude?" Journal of Marketing Research, 18 (August), 318-332.

Moore, Danny L. and J. Wesley Hutchinson (1985), "The Influence of Affective Reactions to Advertising," in Psy- chological Processes and Advertising Effects, eds. Linda F. Alwitt and Andrew A. Mitchell, Hillsdale, NJ: Law- rence Erlbaum, 65-87.

Nowlis, Vincent (1965), "Research with the Mood Adjective Check List," in Affect, Cognition, and Personality, eds. Sylvia E. Tomkins and Carroll E. Izard, New York: Springer, 352-389.

Peterman, Jack N. (1940), "The Program Analyzer-A New Technique in Studying Liked and Disliked Items in Ra- dio Programs," Journal ofApplied Psychology, 23, 725- 741.

Petty, Richard E., John T. Cacioppo, and David Schumann (1983), "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement," Journal of Consumer Research, 10 (September), 135- 146.

Plutchik, Robert (1980), Emotion: A Psychoevolutionary Synthesis, New York: Harper & Row.

Ray, Michael L. and Rajeev Batra (1983), "Emotion and Per- suasion in Advertising: What We Do and Don't Know About Affect," in Advances in Consumer Research, Vol. 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Arbor, MI: Association for Consumer Research, 532- 539.

Schlinger, Mary Jane (1979), "A Profile of Responses to Commercials," Journal of Advertising Research, 19 (2), 37-46.

Sherif, M. and Carl I. Hovland (1961), Social Judgment: As- similation and Contrast Effects in Communication and Attitude Change, New Haven, CT: Yale University Press.

Silk, Alvin J. and Terry G. Vavra (1974), "The Influence of Advertising's Affective Qualities on Consumer Re- sponse," in Buyer/Consumer Information Processing, eds. David Hughes and Michael L. Ray, Chapel Hill, NC: University of North Carolina Press, 157-186.

Smith, Craig A. and Phoebe C. Ellsworth (1985), "Patterns of Cognitive Appraisal in Emotion," Journal of Person- ality and Social Psychology, 48 (4), 813-838.

Stewart, David W. and David H. Furse (1982), "Applying Psychophysiological Measures to Marketing and Adver- tising Research Problems," Current Issues and Research in Advertising, 1-38.

Stout, Patricia and John D. Leckenby (1984), "The Redis- covery of Emotional Response in Copy Research," paper presented at the Annual Conference of the American Academy of Advertising, Denver, CO.

Webb, Peter H. (1979), "Consumer Initial Processing in a Difficult Media Environment," Journal of Consumer Research, 6 (December), 225-236.

Wells, William D., Clark Leavitt, and Maureen McConville (1971), "A Reaction Profile for TV Commercials," Jour- nal of Advertising Research, 11 (December), 11 - 15.

Wessman, Alden E. and David G. Ricks (1966), Mood and Personality, New York: Holt, Reinhart & Winston.

Yuspeh, Sonia (1977), "Program Environment Copy Test," J. Walter Thompson Co. research report, New York.

Zajonc, Robert B. (1980), "Feeling and Thinking: Preferences Need No Inferences," American Psychologist, 35 (Feb- ruary), 151-175.

and Hazel Markus (1982), "Affective and Cognitive Factors in Preferences," Journal of Consumer Research, 9 (September), 123-131.