Relationships Between Affect, Patronage Frequency and Amount of Money Spent With a Comment on Affect Scaling and Measurement

Linda L. Golden, University of Texas at Austin
Mary R. Zimmer, University of Texas at Austin
ABSTRACT - This paper investigates relationships between affect, patronage frequency and amount of money spent for three retail stores: Sears, K-Mart and Wards. Predictions for a statistically significant relationship between affect, patronage frequency and spending were supported. Examination of two different scaling techniques, modified semantic differential and graphic, yielded no statistically significant differences.
[ to cite ]:
Linda L. Golden and Mary R. Zimmer (1986) ,"Relationships Between Affect, Patronage Frequency and Amount of Money Spent With a Comment on Affect Scaling and Measurement", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 53-57.

Advances in Consumer Research Volume 13, 1986      Pages 53-57

RELATIONSHIPS BETWEEN AFFECT, PATRONAGE FREQUENCY AND AMOUNT OF MONEY SPENT WITH A COMMENT ON AFFECT SCALING AND MEASUREMENT

Linda L. Golden, University of Texas at Austin

Mary R. Zimmer, University of Texas at Austin

[This research was sponsored by a grant from the New York University Institute of Retail Management.]

ABSTRACT -

This paper investigates relationships between affect, patronage frequency and amount of money spent for three retail stores: Sears, K-Mart and Wards. Predictions for a statistically significant relationship between affect, patronage frequency and spending were supported. Examination of two different scaling techniques, modified semantic differential and graphic, yielded no statistically significant differences.

INTRODUCTION

The concept of affect has been widely researched and discussed in psychology, but, more recently, has been introduced into the consumer behavior literature. In the psychology literature, affect has often been conceptualized as a "positive mood state" and treated as an experimental (manipulated) variable. In this context, a "positive mood state" has been associated with a variety of different variables, including: an increased propensity to engage in helping social behaviors and being more generous (Berkowitz 1972, Isen 1970, Isen and Levin 1972); reducing risk taking behaviors (Isen and Patrick 1983); reducing hostility in negotiations (Carnevale and Isen 1983); more positive expectations, evaluations and judgments of stimuli (Isen and Shalker 1982, Isen, Shalker, Clark and Karp 1978, Schiffenbauer 1974); increased efficiency in decision-making (Isen and Means 1983); and, a tendency to discriminate less among categories (simplified decision-making), increased memory and more creative problem solving (Isen 1984).

In an atypical study that investigated "negative mood states," Johnson and Tversky (1983) found that a report of a tragic event produced a strong increase in estimates of the frequency of risks and other undesirable events. They concluded that people tend to make judgments compatible with current mood, even if the subject matter is unrelated to the cause of the mood.

What the psychological research suggests is that "affect matters." It matters for cognitive processes, decision-making processes and results, as well as behavior. Indeed, interest in the processes by which happiness leads to helping behaviors produced studies concerning the influence of affect on decision-making and other cognitive processes (Isen 1984).

The potential implications for consumer behavior are self-evident and, hence, have inspired the interest of consumer behavior researchers in new directions for the study of affect. Until recently, affect studies in consumer behavior have focused on affect as a component of attitude rather than as a global measure of "feeling."

AFFECT AND CONSUMER BEHAVIOR

A number of conceptual articles discussing affect have appeared in the consumer behavior literature. Cacioppo, Losch, Tassinary and Petty (1984) view affect as a ". . . system with motivational, perceptual, cognitive, physiological, motor expressive and subjective manifestations," and point out that by focusing solely on the "cognitive" consumer, researchers overlook potentially predominant, although crude, forces on consumer behavior. Zajonc and Markus (1982) also present the thesis that affective factors potentially play an important role in the development and maintenance of preferences relevant to consumer behavior. Others have discussed the role of affect in categorization (Cohen 1982), the difficulties of measuring the cognitive neuropsychological effects of affect (Ray and Batra 1983) and measurement aspects of manipulating and assessing buyers' moods (Gardner 1984).

Empirical investigations of affect (as an experimental variable) in the marketing and consumer behavior literature have been relatively few. (The discussions of the difficulties of measurement and manipulation are not unfounded.) Yet, the consumer research studies that have treated affect as an experimental variable have tended to support predictions parallel to the findings in psychology.

For example, Gorn (1982, experiment one) found that subjects were more likely to choose a pen that had been accompanied with "liked" music than "disliked" music. In a second experiment, Gorn's (1982) subjects selected a pen advertised with attribute information when they were in high consequence condition, but chose the pen advertised with music when they were in the low consequence condition. Thus, it was concluded that affect influenced choice more strongly than information when the situation was of low consequence.

These results are consistent with those that would be predicted with Petty and Cacioppo's Elaboration Likelihood Model of attitude change (Petty, Cacioppo and Schumann 1983). This model views simple affective cues as being more powerful determinants of attitudes when motivation and/or ability to process issue-relevant information is low (the peripheral route).

In a series of studies, Srull (1983) manipulated mood states and fount that positive mood states lead to more favorable product evaluations at the time of information acquisition. A contrast effect of judgments occurred when the moot state at the time of retrieval was inconsistent with the major evaluative implications of the stimulus information. This appears to be due to a tendency to recall more items inconsistent than consistent with mood at the time of retrieval.

In another paper, Srull (1984) reported a study that manipulated mood and distinguished between computational and retrieval information processing objectives. Affective (mood) states had a direct effect on product evaluations for persons in computational (figuring out their evaluations), but not retrieval (recalling evaluations) situations.

Moore and Hutchinson (1983) studied the relationships between affective reactions to advertising and advertising effects. Immediately after exposure to print advertisements subjects had greater change in brand consideration as affective reactions became more positive. This same effect occurred for measurements taken two days after exposure; however, after a seven day delay, products associated with advertisements eliciting positive or negative affect showed greater change in brand considerations than did neutral advertisements. Thus, at least for some situations, over time the "positive affect effect" may converge to have the same impact as "any affect effect" (excluding neutrality). (Final implication added by these authors, not Moore and Hutchinson.)

The efforts to model perceptual distortion on evaluative judgments (see Holbrook 1983) are reflective of the direction affect research has taken in consumer behavior previously. Affect (or attitude) has often been derived as the summation of evaluative Judgments across product attributes. Researchers are, thus, concerned with the halo effect of affect for product evaluation measurements. This "evaluative judgment orientation" is quite different than the current approach that is beginning to "halo" in consumer behavior from the psychology literature.

In spite of the "new affective directions" (and multiplicity of definitions and operationalizations of affect-from mood to emotions to physiological indicators) emerging in the consumer behavior literature, researchers continue to search for the meaning of affect and its relationship to attribute perceptions, attitudes and behavior. For example, Reibstein, Lovelock and Dobson (1980) focused on establishing cause and effect when they measured affect (via "overall satisfaction") toward buses and found that perceptions and behavior mutually influenced each other when affect mediated the relationship.

According to Ray and Batra (1983) consumer behavior seems to be returning to a conception of attitudes similar to that offered by Rosenberg and Hovland (1960): affective, cognitive and conative components. The idea that attitudes can only be changed by changing underlying beliefs (Fishbein and Ajzen 1975) is an "opposing" perspective and assumes that attitudes are unidimensional (affective), based on beliefs and lead to behavioral intentions. Research by Bagozzi and Burnkrant (1979), Bagozzi, Tybout, Craig and Sternthal (1979) and Bagozzi (1980, 1981) has suggested both cognitive and affective attitudinal components

Indeed, affect is a "must" component for any consumer behavior model, as evidenced by its inclusion in the models and even "partial models." Yet, what affect is, what it does to whom in what situations and why varies by the research domain and perspective. What we do know is that affect, in its variety of measurement and definitional forms, may, in certain situations, influence perceptions, memory, cognitive processing (no - it is not necessarily mutually exclusive with cognitions), attitudes (although it depends on how they are measured), intentions and behavior.

Research Purpose and Hypotheses

While numerous consumer behavior researchers have associated affect (via attitude models with only product attributes as measures of brand evaluation) with behavior, there has been a neglect of the simple concept of "like/ dislike" as a global construct (as opposed to attribute specific). Affect (in the attitudinal sense) is generally measured as the sum of its component evaluative parts. Brand image, be it a consumer package good or a retail outlet, is often "reduced" to the sum of its parts. Yet, the affect, or global feeling surrounding the whole, may be lost in the parts (or summation thereof).

The purpose of this study is very straightforward: to investigate the relationship between global affect (as measured by like/dislike) and two measures of behavior. Thus, the orientation of this paper is more reflective of the attitudinal approach to affect than the operationalization of affect as an induced mood state. The focus is three retail outlets: Sears, K-Mart and Wards. Behavior is measured in two ways: patronage frequency and dollar amount spent. Thus, there is a measure of "going" and a measure of "spending."

The concern in this study is not cause and effect but rather association. The "model" to be tested is that people do what they like and like what they do. And, we are focusing on the relationship between affect (operationalized in this study as "like/dislike") and behavior rather than inferring behavior from intentions or investigating the triadic relationship of attitudes, intentions and behavior.

The alternate hypotheses for all three stores are:

H1: There will be a statistically significant association between affect and patronage frequency.

H2: There will be a statistically significant association between affect and dollar amount spent, and

H3: There will be a statistically significant association between patronage frequency and dollar amount spent.

It is expected that more positive affect will be associated with greater patronage frequency and dollar amounts spent, and that more frequent patronage will be associated with larger dollar amounts spent at the stores. (Obviously the relationship between dollar amount spent and patronage frequency is likely to be causal--if you do not patronize the store, there would be no spending. Again, however, we are focusing on association rather than causation in our methodology.)

The justifications for these hypotheses are derived from the results of previous research linking affect and behavior (although this measure of affect is a direct, single, global measure) and the idea that shopping at a "well known" store (such as Sears, K-Mart and Wards) is likely to be an experiential decision in which affect and behavior directly influence each other in an intervening response system (Holbrook and Hirschman 1982). (That is, there may be a "circular" relationship between affect and behavior.) Shopping may be viewed as a "consumption activity" or process, in itself, involving "brand" (store) choice decisions and, hence, an experiential model is appropriate.

METHODOLOGY

There were three broad phases to the methodology: design and pre-test of the survey instrument, data collection and preliminary analysis. Each phase is discussed in this section.

Survey Design and Pre-Test

Questions were designed to measure affect toward shopping at Sears, R-Mart and Wards, patronage frequency and the amount of money spent at each of the three stores during the last year (1983). The affect question, "To what extent do you like or dislike shopping at [Sears/K-Mart/ Wards]?" was measured using two different scaling methods: semantic differential and graphic positioning. An example of each is shown below.

Semantic Differential:                           Sears          K-Mart         Wards

Dislike  1  2  3  4  5  6  7    Like                                                                     

Graphic Positioning:

FIGURE

Subjects receiving the semantic differential treatment were told to write the number from the scale that best represented their opinion in the blank provided for Sears, K-Mart and Wards. Those subjects receiving the graphic positioning scale were told to write the first letter of each store (S,K or W) above the point on the scale that best described their opinion. The questionnaire for each scale included a "response example."

The placement of the measure of affect varied by placing the affect question either before or after a set of open-end image questions (one for each store). Thus, the experiment included in the questionnaire was a 2 X 2 factorial design with two levels of scale type (semantic and graphic) and two levels of affect placement (before and after image).

The scale formats were pre-tested on 94 undergraduate marketing students at a large state university. The results indicated that the scales and instructions were understandable.

Patronage frequency was measured by a single question: "How frequently do you go to Sears, K-Mart and Wards?"

Response categories provided were: Never (1), Once a year or less often (2), Two to six times a year (3), Seven to twelve times a year (4), Two to three times a month (5), and Once a week or more often (6). Respondents wrote their response in the space provided beside the names of Sears, K-Mart and Wards.

The second behavioral measure investigated the amount of money spent at each of the three stores during 1983. Respondents were asked, "Approximately how much did you spend at Sears, K-Mart and Wards during 1983?" Again, responses were written in a space provided by the name of each store.

The questionnaire was pre-tested on a small convenience sample of individuals under actual field conditions. Minor wording changes were made for clarity.

Data Collection and Sample Description

A sample frame of 1600 adults from a nationwide consumer mail panel was selected to represent sex, region, population density and demographic criteria proportionate to the population of the United States.

The cover letter instructed the panel member to fill out the questionnaire him/herself in half the mailings with the other half of the cover letters instructing the panel member to ask his/her spouse to fill out the questionnaire. Thus, an attempt was mate both in the administration and sample selection to balance the sample by sex (due to the larger number of females likely to respond). Subjects were randomly assigned to one of the four treatment combinations.

Questionnaire returns were terminated six weeks after mail-out. The final sample consisted of 894 usable surveys resulting in a response rate of 56 percent. Approximately twenty surveys were not usable.

More women (58.8 percent) returned the survey than did men (41.2 percent). The annual household income distribution for the sample was: 29.3 percent earned less than $15,000 a year, 28.2 percent earned between $15,000 and $24,999 a year, 26.3 percent earned between $25,000 and $39,999 a year, and 16.2 percent earned $40,000 or more a year. Approximately ten percent (10.4) of the sample had less than a high school education, 39.4 percent had graduated from high school, 25.1 percent had some college, and 25.1 percent had four years of college or more. Respondents represented the following geographical regions: New England (6.0 percent), Middle Atlantic (16.1 percent), East North Central (18.5 percent), West North Central (8.7 percent), South Atlantic (16.9 percent), East South Central (6.4 percent), West South Central (9.7 percent), Mountain (5.8 percent), and Pacific (11.9 percent). Thus, the sample represented a cross-section of the United States' population.

Preliminary Data Analysis

The preliminary analysis involved a two-way analysis of variance to determine the effect of scale type (semantic differential or graphic) and affect placement (before or after image) on affect ratings for each store separately. There were no significant main or interaction effects for any of the three stores (within an alpha level of .12). Thus, the format and positioning of the affect scales did not influence responses at a statistically significant level and the four treatments were collapsed for the remaining analyses.

ANALYSES AND RESULTS

In order to determine the pairwise relationships between affect, patronage frequency, and dollar amount spent at each store, Pearson correlations were calculated as indices of linear association. All of the pairwise correlation coefficients were significant within an alpha level of .01 for all three stores, as presented in Table 1.

Since the numerical values assigned to question responses represent ranked categorical data, the treatment of these data as interval through Pearson correlations may tend to "artificially deflate r and systematically underestimate P" (Peterson 1982, p. 494). That is, the relationships may actually be "stronger" than these results represent.

The correlations between affect and patronage frequency and between affect and dollar amount spent include the total sample. However, because of the built in association between patronage frequency and dollar amount spent (no patronage, no spending), the "never shop" subjects were eliminated from the analyses.

TABLE 1

RESULTS OF PEARSON CORRELATIONS

As can be seen from inspection of Table 1, there is a considerable amount of consistency across stores within each of the three pairwise correlations. The strongest relationship emerges for affect and patronage, followed by patronage frequency and the amount of money spent. The association between affect and amount of money spent is not as strong as the other two relationships (lover explained variance).

In order to further investigate the relationship of affect and behavior, the data were submitted to partial correlation analyses with each of the behavioral measures treated as a partial. The subjects who never shopped at a store were eliminated from these analyses (for each store separately). The results of the partial correlation analyses are reported in Table 2.

TABLE 2

RESULTS OF PARTIAL CORRELATION ANALYSIS

The results presented in Table 2 further support the relatively strong relationship between affect and patronage frequency and the very weak relationship between affect and amount of money spent. Although the partial correlations for Sears and R-Mart are statistically significant, the amount of explained variation between affect and amount of money spent (controlling for patronage frequency) is virtually negligible. In contrast, the proportion of variability in affect associated with variations in patronage frequency (controlling for amount of money spent) is relatively substantial and consistently statistically significant for all three stores.

The data were submitted to a series of one-way analyses of variance to determine patterns of affect and expenditure means according to patronage groups. These analyses provide additional insight into the relationship between affect and patronage and patronage and dollar amount spent and treat the patronage frequencies as categorical data (as opposed to metric as in the correlational analyses).

The tabulation below presents the sample distribution of patronage frequency by store.

TABLE

Wards is characterized by relatively infrequent shopping patterns, as can be seen from the number of subjects who "never" shop there (approximately one-third). More subjects actually shopped at Sears with some degree of frequency than either at Wards or K-Mart; however, K-Mart evidenced the highest frequency of shopping patronage (approximately one-third of the sample shopped there two to three times a month or more often). Only eight subjects said they shopped at Wards once a week or more often). The results for this cell are problematic statistically; however, the cells were not collapsed in order to preserve the consistency of calibrations across stores. Paired comparisons for this cell will be presented but should be interpreted with extreme caution.

As is shown in Table 3, there is a statistically significant effect of patronage frequency on both affect and dollar expenditures for all three stores. The F-ratio for all one-way ANOVAS were highly significant (within an alpha level of .01). Again, the analyses for affect and patronage frequency included all the subjects and analyses for patronage frequency and amount of money spent included only subjects who shopped at the store (i.e., eliminating the "never shop" subjects for each store).

TABLE 3

RESULTS OF ONE-WAY ANALYSES OF VARIANCE FOR PATRONAGE FREQUENCY GROUPS

Table 4 displays the mean affect and dollars spent by patronage frequency category. The letters that "match" beside the mean ratings indicate means that are significantly different from each other at the .05 level of alpha. These are the results of a Duncan's Multiple Range Test for each paired comparison.

TABLE 4

MEAN RATINGS AND RANGE TEST FOR ANALYSES OF VARIANCE RESULTS

The results presented on Table 4 indicate that people do what they "like." Clearly, the relationship between affect and patronage frequency is a strong one, as evidenced by the amount of variation in affect ratings explained by shopping frequency (Table 3). Further, in terms of paired mean differences, forty-three out of forty-five mean affect comparisons were significantly different from each other.

The same strength and direction of relationship results from the association between dollar amount spent and patronage frequency. Twenty-five of thirty possible paired mean comparisons were significantly different from each other. The conventional wisdom that "the more people come into your store the more they spend" appears to hold.

CONCLUSIONS AND IMPLICATIONS

The results indicate that all three hypotheses were supported for each retail store. There were statistically significant relationships between affect and patronage (Hypothesis 1), affect and dollar amount spent (Hypothesis 2), and between patronage and dollar amount spent (Hypothesis 3). In addition, there were no statistically significant differences between the two scaling techniques for affect: semantic differential and graphic positioning.

Thus, a single item, global measure of affect was a fairly strong predictor of shopping frequency for a heterogeneous sample of consumers, validated across three retail stores, and supported by results with a relatively high amount of explained variance. However, with respect to the relationship between affect and dollar amount spent, although statistically significant, the explained variance is negligible. Patronage frequency appears to have a stronger relationship with dollar amount spent than does affect. People do, indeed, shop where they like and like where they shop.

This study measured affect on a continuum from dislike to like. Since affect may be conceived as a "system" that "lends its power to memory, to perception, to thought and to action no less than to drives," (Tomkins 1984, p. 164) there are many other "affective states" (beyond liking) appropriate for consumer behavior measurement (or manipulation). This research does suggest, however, that a single, easily applied measure of affect (operationalized as like/ dislike) can contribute to explaining a reasonably "large" amount of variation in behavior.

REFERENCES

Bagozzi, R. P. (1981), "An Examination of the Validity of Two Models of Attitude," Multivariate Behavioral Research, 16, 323-59.

Bagozzi, R. P. and R. E. Burnkrant (1979), "Attitude Organization and Attitude-Behavior Relationship," Journal of Personality and Social Psychology, 37, 913-929.

Bagozzi, R. P., A. M. Tybout, C. S. Craig, and B. Sternthal (1979), "The Construct Validity of the Tripartite Classification of Attitudes," Journal of Marketing Research, 16, 88-95.

Berkowitz, L. (1972), "Social Norms, Feelings, and Other Factors Affecting Helping and Altruism," in L. Berkowitz (ed.), Advances in Experimental Social Psychology, 6, New York: Academic Press.

Cacioppo, J. T., M. E. Losch, L. G. Tassinary, and R. E. Petty (1984), "Properties of Affect and Affect-Laden Information Processing," paper presented at the University of Texas at Austin Affect Symposium.

Carnevale, P. and A. M. Isen (1983), "Communication Processes in Bilateral Negotiation," Proceedings of the Academy of Management," Dallas, Texas.

Cohen, J. (1982), "The Role of Affect in Categorization: Toward A Reconsideration of the Concept of Attitude," in A. Mitchell (ed.), Advances in Consumer Research, IX, 94-100.

Fishbein, M. and I. Ajzen (1975), Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research, Reading, Mass.. Addison-Wesley.

Gardner, M. (1984), "The Consumer's Mood: An Important Situational Variable," in T. Kinnear (ed.), Advances in Consumer Research, XI, 525-529.

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

Holbrook, M. B. (1983), "Using a Structural Motel of Halo Effect to Assess Perceptual Distortion due to Affective Overtones," Journal of Consumer Research, 10, 247-252.

Holbrook, M. B. and E. C. Hirschman (1982), "The Experiential Aspects of Consumption: Consumer, Fantasies, Feelings and Fun, Journal of Consumer Research, 9, 132-140.

Isen, A. M. (1970), "Success, Failure, Attention and Reactions to Others: The Warm Glow of Success," Journal of Personality and Social Psychology, 15, 294-301.

Isen, A. M. (1984), "The Influence of Positive Affect on Decision Making and Cognitive Organization," in T. C. Kinnear (ed.), Advances in Consumer Research, XI, 534-536.

Isen, A. M. and P. F. Levin (1972), "The Effect of Feeling Good on Helping: Cookies and Kindness," Journal of Personality and Social Psychology, 21, 384-388.

Isen, A. M. and B. Means (1983), "The Influence of Positive Affect on Decision-Making Strategy," Social Cognition, 2, 15-31

Isen, A. M. and R. Patrick (1983), "The Effect of Positive Feelings on Risk Taking: When the Chips are Down," Organizational Behavior and Human Performance, 31, 194-202.

Isen, A. M. and T. E. Shalker (1982), "The Influence of Mood State on Evaluation of Positive, Neutral, and Negative Stimuli: When You "Accentuate the Positive," Do You "Eliminate the Negative"? Social Psychology Quarterly, 45, 58.

Isen, A. M., T. E. Shalker, M. Clark, and L. Karp (1978), "Affect, Accessibility of Material in Memory, and Behavior: A Cognitive Loop?" Journal of Personality and Social Psychology, 36, 1-12.

Johnson, E. J. and A. Tversky (1983), "Affect Generalization, and the Perception of Risk," Journal of Personality and Social Psychology, 45, 20-31.

Moore, D. L. and 3. U. Hutchinson (1983), "The Effects of Ad Affect on Advertising Effectiveness," in R. P. Bagozzi and A. M. Tybout (eds.), Advances in Consumer Research, X, 578-531

Peterson, R. A. (1982), Marketing Research, Plano, Texas: Business Publications, Inc.

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

Ray, M. L. and R. Batra (1983), "Emotion and Persuasion in Advertising: What We Do and Don't Know About Affect," in R. P. Bagozzi and A. M. Tybout (eds.), Advances in Consumer Research, X, 543-548.

Reibstein, D. J., C. H. Lovelock, and R. P. Dobson (1980), "The Direction of Causality between Perceptions, Affect and Behavior," Journal of Consumer Research, 6, 370-376.

Rosenberg, M. J. and C. I. Hovland (1960), "Cognitive, Affective, and Behavioral Components of Attitudes," in C. I. Hovland and M. J. Rosenberg (eds.). Attitude Organization and Change, New Haven: Yale University Press, 1-14.

Schiffenbauer, A. (1974), "Effect of Observer's Emotional State on Judgments of the Emotional States of Others," Journal of Personality and Social Psychology, 30, 31-35.

Srull, T. K. (1983), "Affect and Memory: The Impact of Affective Reactions in Advertising on the Representation of Product Information in Memory," in R. P. Bagozzi and A. M. Tybout (eds.), Advances in Consumer Research, X, 520-525.

Srull, T. R. (1984), "The Effects of Subjective Affective States on Memory and Judgment," in T. C. Kinnear (ed.), Advances in Consumer Research, XI, 530-533.

Tomkins, S. S. (1984), "Affect Theory," in K. R. Scherer and P. Ekman (eds.), Approaches to Emotion, Hillsdale, N.J.: Erlbaum, 163-196.

Zajonc, R. B. and H. Markus (1982), "Affective and Cognitive Factors in Preferences," Journal of Consumer Research, 9, 123-131.

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