Effects of Affect on Judgment About Products

Guliz Ger, University of Illinois at Chicago
ABSTRACT - There are a number of marketing problems which have emotional aspects, such as product scares, ad affect, etc. In order to enhance the understanding of such problems, the present study investigates the role of subjective feeling states on a person's judgments within the theoretical framework of associative network model of memory. The influence of experimental manipulation of affect induced by a product scare story on evaluations of various products is examined. The results are consistent with a memory based model of affect.
[ to cite ]:
Guliz Ger (1986) ,"Effects of Affect on Judgment About Products", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 221-225.

Advances in Consumer Research Volume 13, 1986      Pages 221-225

EFFECTS OF AFFECT ON JUDGMENT ABOUT PRODUCTS

Guliz Ger, University of Illinois at Chicago

[I wish to thank Bobby Calder and Brian Sternthal for resources and thoughtful comments.]

ABSTRACT -

There are a number of marketing problems which have emotional aspects, such as product scares, ad affect, etc. In order to enhance the understanding of such problems, the present study investigates the role of subjective feeling states on a person's judgments within the theoretical framework of associative network model of memory. The influence of experimental manipulation of affect induced by a product scare story on evaluations of various products is examined. The results are consistent with a memory based model of affect.

EFFECTS OF AFFECT ON JUDGMENT ABOUT PRODUCTS

There are a number of pervasive marketing problems that are neither well studied nor understood. One such problem is product scares like McDonald's worms rumor (Tybout, Calder, and Sternthal, 1981) and the Tylenol scare. We to not know whether or when the negativity generated about product X will influence only that product or generalize to other products, such as Burger King or Anacin.

A similar problematic area arises with advertising issues such as at placement, and emotion in advertising. Sow and when the moot created by a prior program segment, a prior commercial, or affective context of a commercial will influence the response to the critical message are questions that are not well understood.

The problems mentioned above have something in common: an emotional aspect. However, until recently there has not been much research interest in emotions. Relying OD cognitive psychology, which has attended to its narrower task of explaining 'cool' thinking" (Simon 1982, p. 342), may fail to provide sufficient explanation for marketing issues with emotional aspects. Therefore, we need research on these marketing problems which are not well under stood. Furthermore, a theoretical framework is needed to enhance understanding of why and how affective states influence responses and hence to provide useful knowledge about managerial implications. The present study addresses this need for a basic understanding of how affect is processed and how it affects judgment.

THEORETICAL ISSUE

The purpose of the present study is to investigate the role of feeling states OD a person's judgment within a theoretical framework. The construct of interest is defined as "immediate affective state." Affect is used as a generic term here. A definition that is modified from Simon (1982) is: affect, emotion, and moods designate states and processes in the cognitive system, autonomic nervous system, and the endocrine system. Mood or feeling refer to affect that provides context for ongoing thought processes without noticeably interrupting them. This subtle affective state is the construct of interest here.

To investigate the role of affective states on judgment, a number of questions need to be addressed: 1) How are feeling states represented? and 2) What is the process by which affect influences judgment? Some recent theoretical views have emerged in the last few years concerning these questions.

Two approaches are discussed below. One is the "common processes" view (Clark and Isen 1982; Bower 1981; Bower and Cohen 1982). According to this view, affective and cognitive systems involve common processes, and are similarly represented in memory. Feeling states interact with and operate within the organization of thoughts. This approach argues that affective and cognitive systems are interdependent and provides a specification of how the two systems Interact. Common processes view 18 based on "accessibility hypothesis-: an affective state can function like an organizing unit as a cue to prime related cognitive material. mis hypothesis implies that affective tone may be an important dimension of cognitive organization. Different models of accessibility, derived from different models of memory (network models of Collins and Quillian 1969; Collins and Loftus 1975; and distributed memory motels of Eich, 1985; Murdock, 1983; Smith, et al. 1974; Tversky 1977 where items are represented as ordered sets of features) may provide explanations for how affect functions as a retrieval cue. Among these models of accessibility, a widely accepted one is "spreading activation in an associative network." Although there are many other models of retrieval, spreading activation is the example chosen here because of its prevalence in the literature.

According to the network model of memory, concepts are interconnected by associations. The associative network is a large connection of nodes, which are independent units. The spreading activation process, in the associative network, makes particular portions of the net selectively accessible for recall. When a concept is attended to, a node is stimulated. Activation spreads out along portions of the network associated with that node, in a decreasing gradient (Anderson and Bower 1973; Collins and Loftus 1975). In a recent version of this theory (Anderson 1983) spreading activation defines working memory. Spreading activation identifies and favors the processing of information most related to the immediate context or sources of activation.

Various nodes or elements-an element that encodes an environmental stimulus and a goal element serve as sources of activation. For example, the instructions of an experiment define a goal for the subject. Retrieval process involves refocusing which refers to selecting a subnode of a concept for focus of activation and focusing on a subset of facts about that concept. The notion of goal as a source of activation as well as the mechanism of refocusing on subnodes enable Anderson's ACT* model to activate almost any portion of the network selectively. Thus, modern spreading activation models are consistent with many patterns of activation, both specific and global.

Some nodes in the associative network represent feeling tones (Bower 1981). An aroused emotion spreads out activation, priming related concepts, words, and inference rules. An emotion becomes associated to coincident events and can later act like a retrieval cue. Research evidence that feeling states can cue retrieval of material in memory is provided by a number of studies (Fisher and Harrow 1934; Weingartner, et al. 1977; Isen, et al. 1978; Bower, et al. 1978; Teasdale and Fogarty 1979; Laird. et al. 1982).

Whereas Bower (1981) argues that affect is a node in the associative network, there is another way to view emotion that is prominent in the literature. It is termed "functional independence." According to this view affective and cognitive systems are represented differently, and affective states ag be independent of cognitions (Zajonc 1980; ZaJonc and Marcus 1982, 1984; Zajonc, Pietromonaco and Bargh 1982). The motor system can represent information independently of other forms of mental representations.

Functional independence approach represents the "somatic" view of emotions: discretely different patterns of neurophysiological activity, independent of cognitive appraisal, is capable of generating emotions (Izard 1982 and Leventhal 1979). This view is concerned ore with the somatic expression of emotion, and suggests that emotion 18 an experience which has immediate meaning for the person.

Wilson (1979) offers impressive evidence for a functional independence view. He reports that melodies presented five times were liked better than melodies never heard, even though subJects could not discriminate the format from the latter for familiarity. Although this effect appears congenial to a functional independence position it should be noted that a compelling explanation for why familiarity or exposure increases preferences has not been provided: "We have never been sure why exposure has positive effects" (ZaJonc and Marcus 1982, p. 125).

A recent study addressed these opposing views in aa attempt to test the spreading activation hypothesis (Johnson and Tversky 1983). They report that experimental manipulation of affect induced by a brief story of a tragic event produced a pervasive increase in the respondent's estimates of frequency of risks. This effect w independent of the similarity between the story and the estimated risk such that all estimates, not necessarily related to the mood inducing experience were mood congruent. For example, reading a story about a person tying from homicide not only influenced estimates of deaths from homicide but also estimates of deaths from floods. Johnson and Tversky (1983) suggest that a local (only the risk that matches the story topic being influenced) or a gradient defect (risks closely linked to the story being affected more than unrelated risks) would have indicated that the mood effect was dependent on memorial associations. They argue that the finding of global effects (all responses, unrelated and related to mood producing event are influenced by the mood), in the absence of specific effects (local or gradient mood effects) is supportive of the functional independence view. However, several alternative explanations for their findings can be provided.

There are a number of methodological and theoretical problems ia the Johnson and Tversky study. One problem pertains to the plausibility of a demand characteristics explanation. If some intervention influences non-treatment variables as well as treatment variables, then a suspicion for demand emerges. The subJects might Just be reacting to the demand of the situation and not to the intended intervention. Another methodological problem is the operationalization of the dependent measures: open ended frequency estimates may involve a very large within-group variance which may mask between-group differences.

A major theoretical problem is the limited variety o f the risk estimates. The span of risks does not provide enough range for a fair test of the spreading activation hypothesis. Another theoretical problem is with regards to data interpretation. Johnson and Tversky's results can be interpreted d thin the spreading activation framework: the story activates a general emotion node which is strongly linked to fatalities and different causes of death. When the feeling node is activated all the associations with that node are also activated. A homicide story activates a depressed" node which is connected also to the "flood" node. So, "flood" thoughts as well as other death thoughts are activated. In addition to the emotion node as a source of activation, goal element and other contextual cues may refocus activation on a number of elements such as superordinate nodes or specific subnodes. For example, giving a number of risk estimates before a particular one activates many nodes all of which are linked to "fatalities." The goal of estimating fatalities also activates the set of all events linked to the "fatalities" note. Estimating frequency of deaths due to a particular event, "flood", refocuses the goal element on both "fatality" node and "flood" subnode, and elements of these structures become sources of activation. Frequency estimates of "flood" would be biased by what is in the working memory: mood congruent flood episodes and other fatal events including homicide. Thus all estimates will be influenced by the homicide story, resulting in a global mood effect. Specific effects could have been obtained if homicide-flood link were very weak or not focused on or if the goal element were mainly focused on "flood."

Hence, both global and specific effects appear to be consistent with the spreading activation mechanism because of contest dependent sources of activation and the process of refocusing for further activation. The present view involves inclusion of emotion nodes in a network of associations where an Andersonian (1983) selective, context dependent activation mechanism operates. According to this interpretation, lack of a specific effect in Johnson and Tversky's study does not necessarily imply that the impact of mood was independent of the strength of association between story and risks and does not falsify the "common processes" view of affect. Especially in light of the difficulties involved in testing for global versus specific effects it seems premature to reject the spreading activation view. Thus, another study is needed to test the spreading activation notion. The current study investigates whether activation from a feeling node spreads along portions of the network of associations with that node, in a decreasing gradient. Specifically, the question of whether mood will influence all the judgments, even judgments unrelated to the cause of the mood (global effect) or only the judgments associated with the cause of the mood (specific effect) is examined. Co on process view would predict a specific effect (local or gradient) whereas the functional independence view would predict a global effect.

OVERVIEW OF THE EXPERIMENT

This study investigated the influence of affective states on judgments about products. The stimuli and measurement were designed to alleviate three problems identified with Johnson and Tversky's paradigm: 1) Numerous products were chosen to provide a wide range of associations for a fair test of the gradient hypothesis, 2) To reduce demand, cover stories attempted to conceal the relationship between mood induction and judgment task, and 3) To eliminate high within group variance that existed in risk perception and thus to improve statistical power. scaled measurements were used.

To induce affect, brief stories about lethal effects of toxic residue in a specific beverage were constructed. Three possible effects of exposure to a mood-evoking story on evaluations of various products (discussed below) were considered: compared to control group baseline, 1) a local decrease-for only the focal beverage in the story, 2) a generaLization gradient-decreased evaluations of beverages related to the focal one, and 3) a global decrease for all product (beverage and non-beverage) evaluations.

STIMULI

Stories

Two affective stories were written and pretested to be credible and worry/fear provoking. To maintain information content constant, the stories were exactly the same except for the product they referred to: bottled orange juice or wine. The neutral story was about test marketing for Citrus Sill orange juice; it had appeared in Advertising Age. All stories were about the same length, type-set, and presented as if they were clipped from a newspaper.

The Set of Products

Four groups of products were chosen to tap different associations with the target beverage. One set includes objects of different similarity with, and in the same category as the focal products-orange juice and wine, namely six beverages. Bottled orange Juice, grapefruit juice, instant coffee, bottled cola, beer and vine, in that order of proximity, were the beverages selected based on the results of cluster analysis and multidimensional scaling performed on beverage similarity data. However, category relationship is not the only type of association that may be retrieved. Thoughts evoked by the story may also be associated with nonbeverages. The second set of products included nine foods that either were ingredients of or complemented orange juice or vine.

The third set included four products that had no cognitive association to the target beverages but had an emotional relationship, i.e., they were products (such as household insecticide) that make people worried. The fourth set consisted of fillers, completely unrelated to the target beverages.

These products have different implications in terms of the hypotheses. If mood effects are local, only the evaluation of focal beverages should be influenced. Evaluation of beverages of differing degrees of proximity provides a test of the gradient notion within the object category. Second and third sets of products provide a test of the extended gradient notion; and the fourth set provides a test of global effects.

SAMPLE

Forty five male and female students at Northwestern's Graduate School of Management served as subjects. They were paid $5.00, each, for their participation. They were processed individually or in groups of two or three.

DESIGN AND PROCEDURE

When the subjects arrived at the laboratory they were told that they would participate in two surveys: one conducted for a newspaper to investigate readership and interest in different types of journalism, and the second to find out product opinions as a part of a market research. These instructions were intended to reduce demand characteristics.

Subjects were given a questionnaire entitled "Newspaper Reporting Study." Affect manipulation was introduced by varying the story subjects read in this questionnaire. Subjects were randomly assigned to one of the three story conditions: fear evoking story about orange juice or wine and neutral story. Filler questions about readership (presented before the story) and story ratings with respect to interest and quality of writing were also included. Subjects indicated how involved the story makes the reader (not at all: l/very: 7), and were asked to list five adjectives to describe how the story made the reader feel. The last two questions served as manipulation checks.

Following the completion of this questionnaire, subjects were presented with a second questionnaire entitled "Marketing Research Survey," containing the dependent measures. Participants were asked to evaluate each of the randomly ordered products on five seven-point semantic differential items: good/bad, reputable/ disreputable, valuable/worthless, pleasant/unpleasant, something I would be willing/unwilling to buy. Finally, subjects indicated their gender and frequency of target beverage usage (never: 0/everyday: 4). Thus, the experiment involved two independent variables: effect (between-subjects factor) and degree of proximity of products (within-subjects factor). Fifteen participants were assigned randomly to each of the three story conditions. Product evaluation served as the dependent variable.

MANIPULATION CHECK

The adjectives listed to describe feelings were classified by a scheme generated from Plutchik's (1970) categorization of emotions. Once they were coded, the number of adjectives indicative of fear, worry, and apprehension were recorded. X2 tests indicated chat experimental groups listed more fear/ worry adjectives than the control group. 26.7%/86.7% of the subjects who read the orange juice scare story and 40%/73.3% of those who read the wine scare story indicated that they felt afraid/concerned or apprehensive whereas nobody/6.7% of those in the control group felt so: X2(2)=7.2, p=0.03/X2(2)=22.3, p=0.0. Control subjects mostly responded with adjectives like informed, aware, indifferent, and bored.

T-tests indicated that subjects who read orange juice scare and wine scare stories felt more involved (X=5.07 and 5.00, respectively) than the control group (X=3.07): T(28)=4.19, p=0.0, and T(28)=3.85, p=0.001, respectively.

Thus, the number of fear/worry adjectives and the responses given to the involvement question revealed that the stories effectively induced the intended emotion.

RESULTS

Affective state did not have any effect on the evaluations of three sets of products: foods, emotionally related objects, or fillers. Hence, analyses of their ratings will not be discussed. The evaluative scales, composed of five items for each beverage, were reliable (Cronbach's a'8 were 0.79, 0.80, 0.80, 0.88, 0.89, 0.92). Mean of each subject's response on the five items was used as an evaluation score. Means and standard deviations of evaluative scores of each beverage, categorized by treatment, are shown in Table 1.

TABLE 1

MEANS AND STANDARD DEVIATIONS FOR BEVERAGE EVALUATIONS CATEGORIZED BY THE TYPE OF STORY

A profile analysis treating beverage evaluations as ordered within-subjects factor and type of story as between-subjects factor indicated a multivariate interaction between beverages and story: t 10,78)-1.84, p-0.068/ F(10,210)-2.37, p-0.01, with averaged test of significance using sequential sums of squares. This analysis suggested that affective state induced by a story did not influence ratings of all the objects, rather its effect depended on the type of beverage evaluated.

Separate analysis of variances of beverage evaluations revealed story effects only for grapefruit juice (F(2,42)-4.29, p-0.02) and coffee (F(2,42)-4.51, p-0.017). A priori contrasts indicated that subjects who had read either an orange juice scare or a wine scare story rated grapefruit juice more favorably than the control group: t 1,42)-8.02, p-0.007; and F(1,42)-4.26, respectively. Coffee evaluations were more positive in the wine scare story condition than in either orange juice scare story F(1,42)-4.21, p-0.046) or control (F(1,42)-8.56, p-0.006) groups. Thus, a respondent worried about wine evaluated grapefruit juice and coffee more favorably, and one worried about orange juice rated grapefruit juice more positively than a control subject.

Individual differences in consumption habits for the beverage in the story may mediate subjects' evaluative responses. Hence, frequency of drinking the target beverage was used as a covariate. Analysis of covariance did not significantly change the results for the beverage evaluations except for wine. Hence, analyses of the other dependent variables will not be discussed. When the effect of frequency of consumption of the focal beverage was partialled out, a story main effect was obtained on wine evaluations: t 2,41)-3.97, p-0.027. A priori contrasts indicated that subjects who read a wine scare story rated wine less favorable (adjusted X-3.17) than the control group (adjusted X-2.0): K 1,41)-7.62, p=0.009. Thus, when the effect of individual differences in consumption habits on vine ratings was removed statistically, a subject feeling apprehension towards wine (having read a wine scare story) evaluated wine less favorably than a respondent in a neutral state. This finding suggests a local wood effect.

DISCUSSION

This study failed to replicate Johnson and Tversky: a global effect was not obtained. Mood induction did not affect all the responses of a subject indiscriminately; evaluations of only three products that were in the same category as the target object were influenced. Category membership implies cognitive relationships. Thus, the results suggest that mood effects were somehow dependent on the association between the responses and the cause of the affective state. The existence of a story by beverage interaction supports this suggestion too.

The findings that the treatment influenced evaluations of only three products, and that of chose three two were in the opposite of the expected direction, rule out a demand explanation.

However, the overall pattern of the results were puzzling. First, a local effect was obtained only for wine but not for orange juice ratings. Orange juice may be a product towards which there are strong positive attitudes. In another study very favorable evaluations were detected; some subjects even described it as an "all American drink." A speculation may be that in the absence of any negative thoughts stored about orange juice, retrieval primed by a temporary negative affective state could not have made any mood congruent associations available.

Second, a gradient effect was not obtained. If structural proximity were the association which carried activation, orange juice scare/wine scare story should have made grapefruit juice/beer ratings less favorable than a neutral story but did not.

On the other hand, the story by beverage interaction, the local effect on wine evaluations, as well as the reverse effects on grapefruit juice and coffee suggest that mood effect is dependent on relationship between the target and the evaluated object, but that the nature of this relationship may be something other than categorical proximity. Characteristics of grapefruit juice and coffee suggest two types of association between the target and the evaluated object. One possible relationship may be that of substitutability. Even though coffee was not the closest beverage to wine in pre-experimental similarity analyses, both are mature, adult drinks that have an acquired taste (Levy 1984).

An alternative interpretation is also plausible. The data indicated that grapefruit juice and coffee were the beverages most negatively rated by the control group. A post hoc comparison, using Scheffe 0.95 confidence interval revealed that evaluations of these drinks were less favorable than the aggregated ratings of the other beverages. Thus, the relationship may be evaluative similarity to affect congruent target object thoughts. Then the subject might deliberately disregard, for example, bitter taste, and consider relative benefits of coffee over lethal vine. This may be done with an experimental goal of alleviating a negative state so that he could give objective product ratings.

Why were Johnson and Tversky's (1983) global effects not obtained here. It cannot easily be attributed to the differences in mood induction because any stable phenomenon should persist despite method variation. Besides, the induced moods were probably very similar: 1) both studies used stories, with exactly the sane cover, only the topics varied, and 2) both inductions seem to have lead to worry and fear. If at all, a slight difference may be in apprehension aspects in the present study versus depressive aspects in Johnson and Tversky's. To account for the discrepant results, one would have to argue that even slightly different moods are processed differently-a possible but nonparsimonius explanation.

Two other explanations seem more likely. One say be the greater likelihood of demand characteristics in Johnson and Tversky's study than in the present one. And the second is the existence of some unknown mediator, such as structure, levels, and kinds of associations in the memory.

Hence, it appears that global mood effects are not 80 reliable or stable. Mood can also have other effects, such as the local effect obtained here. Lack of a pervasive global effect leads to rejection of the functional independence view of affect. Although the results do not provide conclusive support for the common process view-memory-based model of affect, the findings of local mood effect on wine evaluation and story by beverage interaction are congenial with this view.

The results also suggests a more complicated process than was predicted. Further studies are needed to specify the underlying knowledge structure as a mediator, and provide stronger tests of the spreading activation mechanism. Goals and refocusing can be manipulated to create conditions under which specific versus global effects would be obtained. Furthermore, network models have mostly been tested in simple recognition or recall tasks. There may be a need to use different types of research paradigms to test these models for affective notes and in judgment tasks.

REFERENCES

Anderson, J. and Bower, G. (1973), "Human Associative Memory, Washington, D.C.: Winston.

Anderson, J. R. (1983), The Architecture of Cognition, Cambridge, Massachusetts: Harvard University Press.

Bower, G. U., Monterio, K. P., and Gilllgan, S. G. (1978), "Emotional Mood as a Context for Learning and Recall," Journal of Verbal Learning and Verbal Behavior, 17, 573-585.

Bower, G. E. (1981), "Mood and Memory," American Psychologist, 36, 129-148.

Bower, G. E. and Cohen, P. R. (1982), "Emotional Influences in Memory and Thinking: Data and Theory in M. S. Clark and S. T. Fiske (Eds.), Affect and Cognition: Seventeenth Annual Carnegie Symposium on Cognition, Hillsdale, NJ: Lawrence Erlbaum Associates.

Clark, M. S. and Isen, A. M. (1982), Toward Understanding the Relationship Between Feeling States and Social Behavior," in A. E. Hastorf and A. M. Isen (Eds.), Cognitive Social Psychology, New York. NY: Elsevier/North Holland, 73-108.

Collins, A. M. and Loftus, E. F. (1975), -A Spreading Activation Theory of Semantic Processing," Psychological Review, 82, 407-428.

Collins, A. M. and Quillian, M. R. (1969), Retrieval Time from Sematic Memory," Journal of Verbal Learning and Verbal Behavior, 8, 240-247.

Eich J. M. (1985), "Levels of Processing, Encoding Specificity, Elaboration, and Charm," Psychological Review, 92, 1-38.

Fisher, V. E. and Marrow, A. J. (1934), "Experimental Study of Moots," Character and Personality, 2, 201-208.

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

Izard, C. E. (1982), "Comments on Emotion and Cognition: Can There be a Working Relationship?," in M. S. Clark and s. T. Fiske (eds.), Affect and Cognition: Seventeenth Annual Carnegie Symposium on Cognition, Hillsdale, NJ: Lawrence Erlbaum Associates.

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

Laird J. D., Wagener, J. J., Halal, M., and Szegda, M. (1982), "Remembering What You Feel: Effects of Emotion on Memory," Journal of Personality and Social Psychology, 42, 646-657.

Leventhal, E. (1979), A Perceptual-Motor Processing Model of Emotion," in P. Pliner, R. R. Blankstein, and I. M. Spiegel (eds.), Advances in the study of Communication and Affect; Vol. 5: Perception of Emotion in Self and Others. New York, NY: Plenum Press.

Levy, S. J. (1984), "Synchrony and Diachrony in Product Perceptions, Working Paper, J. L. Kellogg Graduate School of Management, Northwestern University.

Murdock. B. B., Jr. (1983), A Distributed Memory model for Serial-Order Information," Psychological Review 90. 316-338.

Plutchik, R. (1970), Emotions, Evolution, and Adaptive Processes," in M. 3. Arnold (Ed.), Feelings and Emotions The Loyola Symposium, New York, NY: Academic Press.

Simon, a. A. (1982), "Affect and Cognition: Comments," in M. S. Clark and S. T. Fiske (Eds.), Affect and Cognition, Hillsdale, NJ: Lawrence Erlbaum Associates.

Smith. E. E., Shoben, E. J., and Rips, L. J. (1974), Structure and Process in Semantic Memory, Psychological Review, 81, 214-241.

Teasdale, J. D. and Fogarty, S. J. (1979), "Differential Effects of Induced Mood on Retrieval of Pleasant and Unpleasant Events from Episodic Memory," Journal of Abnormal Psychology, 88, 248-257.

Tversky, A. (1977), "Features of Similarity," Psychological Review, 84, 327-352.

Tybout, A. M., Calder, B. J. and Sternthal, B. (1981), "Using Information Processing Theory to Design Effective Marketing Strategies," Journal of Marketing Research, 18, 73-79.

Weingartner, R., Miller, a. , and Murphy, D. L. (1977), "Mood-State-Dependent Retrieval of Verbal Associations," Journal of Abnormal Psychology, 86, 276-284.

Wilson, W. R. (1979), "Feeling More Than We Can Know: Exposure Effects Without Learning," Journal of Personality and Social Psychology, 37, 811-821.

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

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

Zajonc, R. B. and Markus, H. (1984), "Affect and Cognition: The Hart Interface," in C. E. Izard, J. Kagan, and R. B. Zajonc (Eds.), Emotions, Cognitions, and Behavior Cambridge: Cambridge University Press.

Zajonc, R. B., Pietromonaco, P. and Bargh, J. (1982), "Independence and Interaction of Affect and Cognition," in N. S. Clark and S. T. Fiske (Eds.), Affect and Cognition: The Seventeenth Annual Carnegie Symposium on Cognition, Hillsdale, NJ: Lawrence Erlbaum Associates, 221-228.

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