Citation
Feeling like consuming

Material Information

Title:
Feeling like consuming
Creator:
Pham, Tuan, 1965-
Copyright Date:
1994
Language:
English

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Tuan Pham. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
34679167 ( OCLC )
AKS5516 ( LTUF )
0022866911 ( ALEPH )

Downloads

This item has the following downloads:


Full Text










FEELING LIKE CONSUMING:
AFFECT AS INFORMATION IN DECISIONS
ABOUT CONSUMPTION EPISODES
















By

TUAN PHAM


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1994
































To my parents, Pham Van Lai and Nguyen Thi Luu, for believing in me.














ACKNOWLEDGMENTS


I would not have been in the position of putting the final touch to this dissertation without the intellectual, material and personal support of several individuals and institutions. I am particularly thankful to Joel Cohen, my dissertation chair, for having been a great mentor. He guided me throughout this entire project, very supportive when I doubted, extremely patient when I erred. Working with him has clearly changed the way I look upon things. I know that one day I will find myself telling my own students: "Think about the big picture." For that, I am immensely grateful to Joel.

Richard Lutz and John Lynch both have been more than committee members to me. In the last four years, Rich's characteristic red ink comments on my work have always been insightful, motivating, and exceptionally prompt. John has been a constant help in clarifying my ideas and advising me on analysis issues. His rigor, both conceptual and methodological, and his dedication to his students will always be a source of inspiration. Dr. Margaret Bradley has been very generous and flexible with her time and comments. It has been a privilege to have such a remarkable group of scholars on my dissertation committee.

I cannot express how grateful I am to the Intercollegiate Center for Management Sciences (ICM, Brussels) for financially supporting my doctoral education. Without the ICM's generous three-year Doctoral Fellowship, this project would never have


iii








materialized. I am also indebted to the various professors who allowed me generous access to their students, especially Jack Faricy and John Hall.

I thank Christian Derbaix, Alain Bultez, and Louis Eeckhoudt, my former teachers, colleagues, and occasional tennis partners at the Catholic University of Mons for guiding my first steps into academia. Finally, I thank my classmates Corinne Faure, Susan Fournier, and especially Luk Warlop, for being for being great celebration companions.


iv















TABLE OF CONTENTS



ACKNOWLEDGMENTS................................ iii

ABSTRACT ............................................... viii

CHAPTERS

1. INTRODUCTION: REVISITING THE ROLE OF AFFECT IN
CONSUMER DECISION MAKING ........................... 1

Feeling Like Consuming. .................................... 1
An Overview of the Literature ............................... 3
Blueprint of a Conceptual Bridge ........................ 14

2. AFFECT AS INFORMATION IN DECISION MAKING: A PROCESSING
FRAM EW ORK .................................. 15

Conceptual Foundations. .................................... 15
The Affect Recruitment Heuristic ........................ 23
Appraising Consumption Episodes for Affect Management .......... 39

3. INTENDED CONTRIBUTION. .............................. 48

The Framework in "The Big Picture" ........................ 48
A Research Agenda on Affect Recruitment.................. 52

4. EXPERIMENTS 1 AND 2: THE DETERMINANTS OF AFFECT
RECRUITMENT. ........................................ 56

Purpose and Predictions. .................................... 56
Experiment 1: Method ................................... 60
Experiment 1: Results ........................... 65
Experiment 2: Method ................................... 74
Experiment 2: Results. ........................... ....... 76
Discussion of Experiments 1 and 2...................... 77


v









5. EXPERIMENT 3: THE INFORMATION VALUE OF ANTICIPATORY
AFFECTIVE RESPONSES ............................ 81

Purposes and Predictions .............................. 81
M ethod . . . . . . . . . . . . . . . . . . . . 84
R esults . . . . . . . . . . . . . . . . . . . . 88
D iscussion . . . . . . . . . . . . . . . . . . . 93

6. EXPERIMENT 4: INFERRING AFFECT RECRUITMENT THROUGH
PROCESS BASELINES.............................. 98

Purpose and Predictions: The Logic of Process Baselines .......... 98 M ethod. . . . . . . . . . . . . . . . . . . . .. 10 1
R esults. . . . . . . . . . . . . . . . . . . . 106
D iscussion. . . . . . . . . . . . . . . . . . . . 121

7. EXPERIMENT 5: THE MISATTRIBUTION OF ANTICIPATORY
AFFECTIVE RESPONSES............................. 128

Purpose and Predictions .................................... 128
Method. ............................................... 131
R esults. . . . . . . . . . . . . . . . . . .... 138
Discussion. ............................................. 146

8. CONCLUSIONS ................................... 149

On Bricks and Holes ...................................... 149
A New Agenda .................................... 156


APPENDICES

A. SHORTENED VERSION OF THE AFFECT INTENSITY SCALE ... 159

B. VISUAL IMAGERY SCALE (SHEEHAN 1967) .................. 161

C. EXPERIMENT 1: REPEATED MEASURES ANOVA OF BEHAVIORAL
INTENTIONS ................................. .........163

D. EXPERIMENT 1: EFFECTS OF AFFECT INTENSITY AND VISUAL
IMAGERY OF BEHAVIORAL INTENTIONS ................. 164

E. EXPERIMENT 2: REPEATED MEASURES ANOVA OF BEHAVIORAL
INTENTIO NS ..................................... 166


vi









F. EXPERIMENT 3: ANCOVA OF BEHAVIORAL INTENTIONS .... 167 G. PROTOCOL CONTENT: CODING SCHEME ................... 168

H. PROTOCOL CONTENT AS A FUNCTION OF MOOD AND
INSTRUCTIONS ................................... 172

I. EXPERIMENT 5: REPEATED ANCOVA OF BEHAVIORAL
INTENTIONS ..................................... 176

J. EXPERIMENT 5: UNIVARIATE ANCOVAS OF BEHAVIORAL
INTENTIONS ..................................... 177

REFERENCES .............................................. 181

BIOGRAPHICAL SKETCH ...................................... 195


vii








Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy FEELING LIKE CONSUMING:
AFFECT AS INFORMATION IN DECISIONS ABOUT CONSUMPTION EPISODES By

Tuan Pham

December 1994


Chairperson: Joel B. Cohen
Major Department: Marketing

Unlike previous research, which has either ignored the role of affect in consumer decision making or treated it as part of the decision context, this dissertation takes the premise that affect is a source of information. First, a brief review of previous research on affect and consumer behavior is presented. Then an integrative framework discusses the processes through which affect is used as an input in decisions about consumption episodes. The cornerstone of the framework is the affect recruitment heuristic, which suggests that to make decisions about consumption episodes, consumers may (1) access or construct concrete representations of the episodes, which (2) instantiate anticipatory affective responses, (3) that are used as inputs because of their unique information value. The framework specifies several determinants of affect recruitment, such as consumers' motives and their imagery abilities. The framework also suggests a process explanation of consumption decisions geared toward the management of affective states.


viii








Propositions of the framework about the affect recruitment heuristic were tested in five experiments. Experiments 1 and 2 focused on predictions of the framework about how different inputs influence the outcome of decisions about consumption episodes. To test the unique information value of anticipatory affective responses, Experiment 3 relied on contextual mood as a means of manipulating anticipatory affective responses to an episode without altering its evaluative content. The results suggest that feelings experienced during the decision process may indeed have unique information value. Experiment 4 tested the descriptive accuracy of the affect recruitment heuristic as a model of how consumers make decisions about consumption episodes. The heuristic appears to be a plausible paramorphic description of how consumers with consummatory motives rely on their feelings to make decisions about consumption episodes. An expectancy-value model did not describe this process equally well. Using a misattribution paradigm, Experiment 5 suggests that anticipatory affective responses are indeed instantiated in decisions about consumption episodes. The results of these five experiments are discussed and directions for future research are provided.


ix














CHAPTER 1
INTRODUCTION: REVISITING THE ROLE OF AFFECT IN CONSUMER DECISION MAKING.



Feeling Like Consuming


It is Sunday; the weather is beautiful. Marla is trying to decide how to spend the afternoon. She could play tennis with Mark; she has just bought the latest model of Nike shoes and is eager to try them. Or she could go clothes shopping in the mall with Diane; after shopping, they could have a large strawberry ice cream at Haagen-Dazs. Or maybe she should just go to the park, find a bench in the shade, and finish reading "The Pelican Brief." She "pictures" herself on the tennis court, panting and all sweaty. This thought makes her feel unpleasant. No, I don't feel like playing tennis, she concludes. But the thought of going to the mall excites her: she has a vivid image of herself buying a nice casual dress at the Gap and receiving Diane's compliments. Yes, I really feel like going to the mall, she says to herself. She picks up the phone and calls Diane...

Frank has a date tomorrow tonight. He has bought two tickets for the "Phantom of the Opera" on Broadway. But he hasn't decided yet on where to take her for dinner before the play. What would be a good place for the two of them to get to know each other? He remembers a restaurant he went to several months ago. He imagines what it would be like in the restaurant: he is with his date; they are talking to each other; the


I








2

atmosphere is just right; he can almost smell the food and hear the soft music in the background. This thought "feels good" to him. He concludes that it would be a great place to take his date. He opens the phone book and looks up the restaurant's number...

These two examples depict consumers making decisions about consumption episodes. Consumption episodes are events or activities in which the use of products or services is salient in the consumer's mind. Because alternative episodes (e.g., going to the mall vs. playing tennis) may involve the consumption of very different products or services, such decisions should be of substantial interest to marketing and consumer researchers. Yet, very few empirical studies have examined how these decisions are made. The focus of this dissertation is on decisions about consumption episodes.

This focus arose not only from the substantive relevance of these decisions, but also from the fact that their study may reveal important theoretical insights into the process through which consumers make a variety of decisions. As suggested in the opening examples, the central thesis of this dissertation is that the role of affect' in consumer decision making may be more fundamental than previously recognized. Affect has been largely ignored in the "traditional" consumer decision making literature (see for instance Bettman, Johnson, and Payne's 1991 review). And when affect was considered, it was mostly as a "background" source of influence, part of the decision context, as reflected by the growing literature on mood effects (e.g., Gardner 1985; Kahn and Isen 1993; Mano 1990; Miniard, Bhatla, and Sirdeshmukh 1992).



The term affect is used as a generic descriptor of valenced feeling states; emotions and moods being specific examples of affective states and responses (cf.
Cohen and Areni 1991).








3

Unlike previous research, this research program regards affect as a source of information that consumers use to make a variety of decisions. Once the informational status of affective responses is recognized, it becomes easier to incorporate these responses into information processing models of consumer decision making. An integrative processing framework that describes the informative role of affective responses in consumer decision making is proposed in Chapter 2. A central thesis of this framework is the pervasive use of an affect-recruitment heuristic in consumer decision making, particularly but not exclusively in decisions about consumption episodes. This heuristic involves three subprocesses: (1) consumers access concrete representations of the episodes; (2) they instantiate affective responses to these representations; and (3) they use the instantiated affective responses as inputs to the decision. To help position the framework into the literature, prior research on affect and consumer behavior is first briefly reviewed. The outline of the dissertation is presented next.



An Overview of the Literature

A comprehensive review of the literature on affect and consumer behavior would exceed the scope of this introductory chapter and duplicate the work of others (see Cohen and Areni 1991). The purpose of this section is to provide a brief overview of the existing literature, emphasizing its key insights as well as its shortcomings. Three somewhat independent streams research on affect can be identified in the consumer behavior literature. Their main characteristics (e.g., typical substantive domain, methodological paradigm, limitations) are summarized in Table 1.1.







TABLE 1.1
THREE STREAMS OF RESEARCH ON AFFECT AND CONSUMER BEHAVIOR

Focus Contextual Affective State Affective Response I Affective Mediation

Examples Gardner (1985) Mick and Demoss (1990) Batra and Ray (1986) Miniard et al. (1992) Havlena and Holbrook (1986) Holbrook and Batra (1987) Kellaris and Mantel (1994) Derbaix and Pham (1991) Edell and Burke (1986) Goldberg and Gorn (1987) Rook (1987) Allen et al. (1992)

Substantive Judgments Consumption behaviors (e.g., Advertising responses (e.g., domain purchase and product use) Aad, Ab)

Typical methods Experimental Open-ended questioning Correlational

Theoretical Social psychology Unclear Emotion literature underpinnings

Main axiological Uncovering biases Describing consumption Explaining incremental concern experiences variance

Affect as Dichotomous Highly differentiated response Reduced number of (positive/negative mood) dimensions (e.g., pleasure and arousal)

Main Contextual affective states Affective responses to Influence of affect is not only contributions have a variety of effects; consumption are diverse and contextual; Attempts to explain process rich Influence of affect may be unique

Limitations Fragmented Atheoretical Little insight into process Ignores informational and Ignores informational and Focuses on static settings motivational function of motivational function of affect affect








5

Affective States as Contextual Influences

An extensive amount of research has focused on the effects of contextual mood states on a variety of dependent variables, including: ad evaluation (Goldberg and Gorn 1987), product and brand evaluations (e.g., Axelrod 1963; Dommermuth and Millar 1967; Gardner and Wilhelm 1987; Gorn, Goldberg, and Basu 1993; Miniard et al. 1992), behavioral intentions (Alpert and Alpert 1990; Swinyard 1993), variety seeking (Kahn and Isen 1993), and time perception (Hornik 1993; Kellaris and Mantel 1994). This stream of research has its main theoretical underpinning in social psychology, where contextual mood has been found to have a variety of effects on social cognition and social behavior (see Isen 1984 and Gardner 1985 for reviews).

Most of these studies rely on an experimental paradigm where different mood states (positive, negative, or neutral) are induced among different groups of subjects2 that are then compared on the dependent variable(s) of interest. The main axiological concern of this stream of research can be described as "bias-driven." It is to show that, somehow, consumers in a positive or negative mood do not process information as they would if they were in a neutral state. Essentially, the contribution of this research amounts to a vast collection of empirical effects that show that contextual mood does matter. Another contribution of this research comes from its growing concern for explaining the processes that underlie the effects of affective states (e.g., Batra and Stayman 1990; Pham 1994; Srull 1984, 1987).



2 The particular mood induction procedure may vary: background music, television
programs, reading materials, small gifts, etc.








6

This stream of research has two main limitations. First, by focusing on various biasing effects of contextual mood, this research has yielded a highly fragmented knowledge base from which it is difficult to extract general principles of consumer behavior (beyond the fact that mood states do matter). One exception is the robust congruency bias that consumers' moods exert on their evaluations of marketing stimuli. In Dommermuth and Millar (1967), for instance, consumers exposed to a pleasant movie evaluated more favorably a product tasted while in a prior neutral mood than did consumers exposed to an unpleasant movie. Miniard et al. (1992) obtained similar results using background music as a mood induction procedure. Mood induced by the context of advertisements was also found to affect ad and brand evaluations in a congruent direction (e.g., Gardner and Wilhelm 1987; Goldberg and Gorn 1987).

The explanation generally offered for this pervasive effect is an accessibility explanation (Isen, Shalker, Clark, and Karp 1978). Mood states may increase the accessibility of mood-congruent material in memory (cf. Bower 1981). If moodcongruent cognitions about the stimulus object (e.g., a product or an ad) are more likely to be accessed, evaluations of this object may become biased toward the person's mood state. More recently, Schwarz (1990) and his colleagues (e.g., Schwarz and Clore 1983, 1988) have proposed a different explanation. They argue and provide compelling evidence that mood states bias evaluations of various stimulus objects because people use their mood as information about the objects. This mood-as-information explanation was recently tested and supported by Gorn et al. (1993) in a marketing context. As detailed


I








7

in Chapter 2, the theoretical framework of this dissertation expands on the idea that people's affective responses are used as information in a variety of decisions.

A second limitation of the affective context stream of research is that the induced mood states are typically independent of the judgment to be made or the task to be completed. In this research paradigm, the influence of affect --on consumer decision making, for instance--is regarded as merely contextual. This perspective ignores the possibility that affect may have a more central function in consumer decision making: being a source of information. A more minor criticism of this research is that it often reduces affect to a mere distinction between good mood and bad mood or between pleasant states and unpleasant states, whereas other research (see below) shows that affect may be considerably richer.



Affective Responses as Consumption Outcomes

A second stream of research focuses on affect as the outcome of consumption activities, that is, affect as a response to consumption. Consistent with Holbrook and his colleagues' plea to broaden the scope of consumer research (e.g., Holbrook 1986; Holbrook and Hirschman 1982; Hirschman and Holbrook 1982), a number of studies have examined affective responses to consumption as a dependent variable of intrinsic interest. The substantive domain of these studies is a variety of consumption phenomena. Some studies focus on one specific behavior or phenomenon: important purchases (Richins, McKeage, and Najjar 1992, study 2), impulse purchases (e.g., Gardner and Rook 1988; Rook 1987; Weinberg and Gottwald 1982), compulsive buying (e.g.,








8

O'Guinn and Faber 1989), self-gift (e.g., Mick and Demoss 1990), sport activities (e.g., Hill and Robinson 1991; Lehmann 1987), game playing (Holbrook, Chestnut, Oliva, and Greenleaf 1984), music listening (Baumgartner 1992- Bruner 1990), etc. Other studies have examined affective responses across consumption activities (e.g., Derbaix and Pham 1991; Havlena and Holbrook 1986; Havlena, Holbrook, and Lehmann 1989).

In essence, most of these studies are descriptive. For instance, in a typical study (Richins et al. 1992, study 1), respondents were asked to recollect a purchase that had made them happy, and describe the purchase circumstances and the feelings experienced while making the purchase and during the subsequent few weeks. The descriptions were content-analyzed into frequency counts of the type of purchase mentioned and the type of emotions elicited (e.g., happiness, excitement, anxiety). Derbaix and Pham (1991) used a similar methodology, combined with correspondence analysis, to uncover typical associations between a variety of consumption situations (e.g., "while shopping for groceries" and "after having been to the hairdresser's") and specific affective responses (e.g., "irritation" and "regret"). Because of their emphasis on description, these studies often rely on content analysis of responses to open-ended questions (e.g., Havlena and Holbrook 1986; Gardner and Rook 1988; Mick and Demoss 1990; Rook 1987). Some of these studies also use multivariate analyses (e.g., Batra and Holbrook 1990; Havlena and Holbrook 1986; Havlena et al. 1989).

Unlike the typical research on contextual mood effects, most studies on affective responses avoid reducing affect to being as either good or bad, pleasant or unpleasant. A contribution of this stream of research has been to show that affective responses to








9
consumption may be very rich (e.g., Holak and Havlena 1992) and diverse (e.g., Derbaix and Pham 1991; Richins et al. 1992).

A major limitation of this research is that it has been predominantly atheoretical (which is due in part to its exploratory character). Whereas research on the contextual effects of affective states is solidly grounded in social psychology, the theoretical underpinnings of studies on affective responses are unclear. Aside from informing us on the variety and richness of affective responses to consumption, these studies tell us little about (1) how these responses are elicited; (2) how diverse they really are and why;

(3) how invariant they are across consumption activities, etc. Unless these questions receive satisfactory answers, these studies' findings amount to little more than a collection of data points in a multivariate space of consumption phenomenon by type of affective responses. Recent theoretical models of affective responses (e.g., Cohen and Areni 1991; Frijda 1986; Lazarus 1991; Roseman 1979) may help better understand how-as opposed to what--affective responses are elicited by consumption phenomena. The processing framework detailed in Chapter 2 integrates several aspects of these models.

A second limitation of this stream of research is that by focusing on affective responses per se, it captures neither their information value nor their motivational significance. Operant learning principles would suggest that if affective responses are consistently experienced in relation to consumption behaviors, consumers would somehow incorporate these responses in their decision to consume. The conceptual framework detailed in Chapter 2 provides a process account of how affective responses may be used as information in consumers' decisions and how these responses may be








10

anticipated and incorporated in consumers' decisions. Also, as implied in Hirschman and Holbrook's (1982) concept of hedonic consumption, certain consumption behaviors may be enacted in order to experience desired affective responses (see also Gould 1991). The conceptual framework suggests several hypotheses about the process underlying an affect management motive.



Affect as a Mediating Factor

Whereas the affective response stream of research focused on affective responses per se, this stream of research focuses on how affective responses mediate (and possibly motivate) other variables of interest. Most of these studies have been in the advertising domain (e.g., Aaker, Stayman, and Hagerty 1986; Batra and Ray 1986; Burke and Edell 1989; Edell and Burke 1987; Holbrook and Batra 1987; Olney, Holbrook, and Batra 1991; Stayman and Aaker 1988), where researchers have examined how affective responses to advertisements mediate attitudes toward the ad (Aad) and attitudes toward the brand (Ab). More recently, similar studies have been conducted in other domains. For instance, Allen, Machleit, and Klein (1992) examined how affective responses to blood donation mediate future donation behavior; Holbrook and Gardner (1993) examined how affective responses to music mediate listening time; and Mick and Faure (1993) studied how affective responses to success and failure mediate the likelihood of self-gifts.

Most of these studies are correlational and based on a specific paradigm that involves (1) exposure to some stimuli (e.g., advertisements, musical pieces); (2) measurement of affective responses to these stimuli; (3) measurement of dependent








1

variables (e.g., Ab, listening time); and (4) some assessment of the stimulus' properties (e.g., ratings of the ad's quality by judges). Whereas the "contextual" stream of research emphasizes biases and the "response" stream of research emphasizes description, the "mediation" paradigm focuses on incremental variance explained. For example, Edell and Burke (1987, p. 428) summarized one experiment as follows: "We found that positive and negative feelings contributed to explaining variance in Aad and Ab even after accounting for the variance explained by judgments of the ad's characteristics." Similarly, Olney et al. (1991, pp. 440-441) introduced their study as focusing on "the variance in [advertising] viewing time explained by three attitudinal components . by two emotional dimensions (pleasure and arousal), and by the uniqueness of ad content." Because of this axiological focus (i.e., variance explained per degree of freedom), these studies have often examined affective responses along a reduced number of dimensions such as Mehrabian and Russell's (1974) pleasure and arousal factors.

The major contribution of the affective mediation stream of research has been to show that affective responses may have a more direct, hence noncontextual influence on various dependent variables of interest in marketing and consumer research. This stream of research also suggests that the influence of affective responses is not redundant with the influence of predictor variables examined in previous research. For example, it has been observed that feelings induced by an ad influence (1) brand attitudes independently of attitudes toward the ads (Burke and Edell 1989; Stayman and Aaker 1988); and (2) attitude toward the ad independently of the ad's content (Edell and Burke 1987; Holbrook and Batra 1987). Allen et al. (1992) also found that emotions experienced in response








12

to blood donation predicted future blood donation independently of peoples' attitudes toward this act. Therefore, one can infer that the influence of noncontextual affective responses on a variety of judgments and behaviors may be unique.

The main limitation of this research, however, is that it rarely seeks to explain the process through which affective responses exert their noncontextual influence. As Stayman and Aaker (1988, p. 371) acknowledge, the goal of this research is "to demonstrate that conditions exist under which feelings have direct effects . rather than to assess the process underlying the effect." For instance, it is not clear in the Allen et al. (1992) study exactly how the emotions experienced during one blood donation occasion came to influence people's subsequent donation behaviors. If these behaviors are preceded by some decision making process, the role of the initial emotion experience in this process needs to be elucidated. The proposed framework offers an attempt at specifying such processes.

The scope of this paradigm is also limited to situations in which stimulus exposure, criterion variables, and mediation relate to a single incident. For instance, in Holbrook and Gardner (1993), subjects were presented a series of musical pieces (stimulus exposure) which they were asked to rate on pleasure and arousal dimensions (mediation). The amount of time subjects actually listened to each piece (criterion variable) was recorded unobtrusively. In these designs, the mediational role of affective responses to the stimulus can easily be interpreted in terms of approach-avoidance principles. And when the criterion variable relate to an object that differs from the affect-eliciting stimulus--such as when attitude toward a product is expressed as a








13

function of affective responses to the embedding commercial--simple association principles may suffice. However, the mediating role of affective responses need not be restricted to single incident settings. As the Allen et al. (1992) study suggests, affective responses to one incident may influence behaviors relating to future incidents. How does affect influence decisions in more dynamic settings? The proposed framework examines setting in which consumers need to anticipate future outcomes.



Blueprint of a Conceptual Bridge

In summary, a growing body of research suggests that affect is important to the study of consumer behavior: (1) affective states have a variety of contextual effects on consumers' judgments; (2) affective responses to marketing stimuli are significant (and possibly unique) noncontextual predictors of criterion variables of interest in marketing and consumer research; and (3) affective responses are a rich component of many consumption experiences. Yet, the literature on consumer decision making has traditionally ignored the role of affective inputs (Bettman 1993), and the different streams of research on affect and consumer behavior have ignored the informational and motivational function of affect in decision making.

This dissertation intends to provide a conceptual bridge between the growing literature on affect and the better established literature on consumer decision making. The unifying theme of this program of research is that affect is not only a contextual source of bias or a response to consumption phenomena, it is also as source of information, a decision input. Chapter 2 articulates an integrative framework that








14

specifies the processes through which consumers use their affective responses in their decisions. Chapter 3 formalizes a research agenda for testing a central component of the framework: the affect recruitment heuristic. Chapters 4 through 7 report five experiments implementing this agenda. And Chapter 8 provides a general discussion of the findings, as well as future directions.














CHAPTER 2
AFFECT AS INFORMATION IN DECISION MAKING: A PROCESSING FRAMEWORK



This chapter introduces a processing framework that describes how consumers may rely on affective responses as a decision making input. The framework is presented in the context of decisions about consumption episodes. This chapter is divided into three sections. The first section of the chapter sets the conceptual foundations of the framework. The second section introduces the affect recruitment heuristic, the cornerstone of the framework. In the third section, the framework's logic is extended to provide a process explanation of consumption decisions geared toward the management of aversive affective states.

Conceptual Foundations


The foundations of this framework lie in recent cognitive theories of affect. It is postulated that experienced affective responses or feelings are information that consumers may rely on in decisions about consumption episodes. This information originales in the appraisal process underlying the elicitation of these affective responses. The nature of this information is described in terms of key components of the meaning structures typically underlying these affective responses.


15








16


Feelings as Information

Although previous research has stressed the biasing effects of affective states on judgments (see e.g, Gardner 1985 for a review), recent theories emphasize that affect also fulfills adaptive functions (e.g., Lazarus 1991; Plutchik 1980). One of these functions is to provide information about the state of the environment (e.g., Frijda 1986, 1988; Morris 1991). For instance, Schwarz and Clore (1988; Schwarz 1990) propose that to evaluate objects, people often ask themselves "How do I feel about it?" In Schwarz and Clore (1983), judgments of life satisfaction were collected among subjects either in positive or in negative mood. The authors expected that subjects' mood would bias their reported life satisfaction. Moreover, Schwarz and Clore reasoned that if people inspect their feelings because of their informative value, these feelings would not bias judgments when attributed to a cause other than the object of judgment. This prediction was supported. Subjects' mood states biased their judgments unless an external cause for these mood states (e.g., the weather) was made salient.

In Schwarz and Clore (1983), as in typical research on contextual affective states (see Chapter 1), the cause of subjects' feelings was independent of the object of judgment. However, the "How-do-I-feel-about-it?" heuristic applies also, and mostly, to situations where the source of people's feelings is actually the object to be judged. What this heuristic means is not so much that mood states are often misattributed and bias judgments, as that people often evaluate objects based on their feelings in response to these objects. For example, while at a cocktail party we might be cheerfully drinking our favorite single malt scotch and discussing movies seen recently. When asked








17

whether we like the party, we would inspect our feelings and infer from our cheerfulness that we probably do. Here, as probably in many cases, the feelings are correctly attributed. Therefore, a natural way in which consumers may evaluate consumption episodes is by inspecting their feelings in response to them. This postulate is the first building block of the present framework.


Postulate 1: Consumers evaluate consumption episodes, in part, by inspecting their
affective responses to these episodes.



The Origins of the Information Conveyed by Affective Responses

Postulate 1 implies that affective responses convey information to the consumer experiencing them. The origin of this information can be traced to the processes underlying the elicitation of these responses (Clore, Schwarz, and Conway 1994). Consistent with cognitive theories of affect (e.g., Arnold 1960; Frijda 1986; Lazarus 1991; Ortony, Clore and Collins 1988; Roseman 1984), it is postulated that affective responses to consumption episodes emanate from personally-relevant meaning structures generated by the appraisal of these episodes. Appraisal is the "evaluation of the significance of what is happening in the person-environment relationship for personal well-being" (Lazarus 1991, p.87). Appraisal theories emphasize that affective responses to external events are not reactions to the events per se but responses to interpretations of these events.








18

Postulate 2: Consumers' affective responses to consumption episodes are mediated by
their appraisals of these episodes, which generates personally-relevant
meaning structures .



Because there is a lawful relation between emotions and appraisal-generated meaning structures (Frijda 1988; Lazarus 1991), not only do different meaning structures result in different emotions, but the experience of an emotion informs us about the particular meaning of the eliciting event (Schwarz 1990). Keltner, Ellsworth, and Edwards's (1993) Experiment 4 provides dramatic support for this postulate. Anger or sadness was induced by having subjects model behavioral expressions typical of these emotions. Although they were not exposed to any stimulus information, subjects subsequently showed judgmental biases characteristic of the meaning structure underlying the emotion they were modeling. For instance, compared to subjects modeling anger, subjects modeling sadness were more likely to believe that future negative events in their lives would be caused by situational (an opposed to human) factors. As detailed below, situational agency is an appraisal characteristic of sadness emotions. Postulate 3: The information conveyed to the consumer by the experience of a
particular emotion (e.g., sadness or pride) corresponds to the appraisalgenerated meaning structure that typically underlies this emotion.






This postulate does not mean that affective responses are always based on controlled processes. Recent models of affective responses (Branscombe 1988; Cohen and Areni 1991; Hoffman 1986; Leventhal 1984) suggest that appraisal of consumption episodes may be performed in a controlled and elaborated mode, or
more automatically through the activation of affective schemata.








19


The Nature of the Information Conveyed by Affective Responses

Although it is beyond the scope of this chapter to specify comprehensively the meaning structure underlying each emotion, a good understanding of the information conveyed by affective responses can be reached by specifying several key components of these meaning structures (for more complete treatments, see Abelson 1983; Arnold 1960; Frijda 1986; Lazarus, 1991; Ortony et al. 1988; Roseman 1984; Scherer 1984; Smith and Ellsworth 1985). These core elements may be regarded both as major antecedents of emotional experiences (the traditional view) and as information implied by these emotional experiences (the feelings-as-information view). The mapping between these core element and the main types of affective responses is summarized in Tables 2. 1 and 2.2.

Desirability. That events such as consumption episodes are appraised as desirable or undesirable is posited by every appraisal theorist. Desirability results from the interaction between the presence/absence of a stimulus and its intrinsic rewarding/aversive property (Arnold 1960; Ortony et al. 1988; Roseman 1984). The presence of a rewarding stimulus in a consumption episode is conducive of joy (e.g., receiving a gift). The absence or loss of a rewarding stimulus is conducive of disappointment or sadness (e.g., not receiving an expected gift, Higgins 1987). The presence of an aversive stimulus may result in distress or disgust (e.g., smelling garbage while dining in a restaurant). The absence or removal of an aversive stimulus is associated with relief (e.g., arriving late at the airport and realizing that the plane has not left yet).








20


TABLE 2.1
TENTATIVE SPECIFICATION OF THE CORE MEANING STRUCTURES
OF THE MAIN TYPES OF PLEASANT AFFECTIVE RESPONSES
AFFECTIVE Joy Pride Hope Relief Attraction RESPONSE (Happiness (Liking, TYPE Elation) Love)

APPRAISAL
Desirability: Rewarding Rewarding Rewarding Aversive Rewarding Type & Presence & Present & Present & Present & Absent & Present of Stimulation or Aversive
& Absent

Agency Situation Self Situation Situation Other

Control High High Low Low High

Legitimacy Open High Open Open Open

Outcome Low Low High Low Open uncertainty

Note: Open means that the affective response is compatible with different appraisal along the dimension.


Agency. Different emotions will be experienced in a consumption episode depending on whether the focus of appraisal is the self (self-agency), a specific object or person (other-agency), or the consequences of the entire episode (situation-agency, e.g., De Rivera 1984; Frijda 1986; Roseman 1984; Scherer 1984; Ortony et al. 1988). Emotions such as anger or attraction are associated with the appraisal of specific objects (e.g., product) or people (e.g., a salesperson). Emotions such as pride (e.g., wearing expensive clothes), shame, and guilt (e.g., drinking too much) are based on appraisal of








21

the self. Emotions such as fear (e.g., being scared in a roller- coaster), joy (e.g., having obtained a good deal on a purchase), and sadness (e.g., being disappointed because one's favorite team has lost), are based on the appraisal of an entire event.


TABLE 2.2
TENTATIVE SPECIFICATION OF THE CORE MEANING
OF THE MAIN TYPES OF UNPLEASANT AFFECTIVE


STRUCTURES RESPONSES


AFFECTIVE Anger Fear Sadness Guilt Repulsion RESPONSE (Rage) (Anxiety) (Disappoint- (Shame, (Disgust,
TYPE ment, Embar- Hate) Despair) rassment)
APPRAISAL

Desirability: Aversive Rewarding Rewarding Rewarding Aversive Type & Presence & Present & Absent & Absent & Absent & Present of Stimulation or or Aversive Aversive
& Present & Present

Agency Other Situation Situation Self Other

Control High Low Low Low Low


Legitimacy High Low Open Low Open


Outcome Low High Low Low Low uncertainty


Note: Open means that the affective response is compatible with different appraisal along the dimension.


Control. Consumers' sense of control in a consumption episode results from an assessment of their available resources against the resource required for coping with the event (Lazarus 1991). A sense of high control in an episode is intrinsically rewarding.








22

It is because they perceived the consumption episode as "under control" that consumers voluntarily expose themselves to threatening or potentially harmful stimulation, such as riding roller-coasters or watching frightening movies (Mandler 1984a). A low sense of control is at the core of fear-like emotions (e.g., driving on a slippery highway), and sadness-like emotions (e.g., having one's house destroyed by a fire).

Legitimacy/deservingness. Appraisal of legitimacy or deservingness refers to whether one deserves or does not deserve a positive or a negative outcome in a consumption episode (Roseman et al. 1990). The experience of pride is based on the appraisal that one deserves a rewarding stimulation (high legitimacy) and receives it (e.g., an adolescent buying himself a nice stereo because he has graduated with honors from high school). When experiencing anger people have the sense of "being right" (high legitimacy) but do not receive the deserved reward (e.g., customers getting angry with a salesperson for not providing the expected level of service). In contrast, guilt and shame are experienced as result of low deservingness or high blameworthiness (e.g., drinking too much at a party). In other words, legitimacy/deservingness appraisal is an assessment the strength of one's position in a consumption episode with respect to moral or social standards.

Outcome uncertainty. Consumption episodes differ in the predictability of their outcome (e.g., Frijda 1986; Lazarus 1991; Roseman 1984; Smith and Ellsworth 1985). High outcome uncertainty is characteristic of emotions such as hope (e.g., awaiting the guests' reactions to a new food recipe), and fear-like emotions, such as anxiety (e.g., a wife dreading that her husband might not like her new dress). When the outcome of an








23

episode becomes certain, other emotions take place, such as embarrassment or disappointment (e.g., realizing that the guests don't like the new recipe) and joy or relief (e.g., the husband likes the new dress.)

In summary, the meaning structure of the main types of emotions (their informational content) appears to depend heavily on five kinds of appraisal: desirability, agency, control, legitimacy, and outcome uncertainty. The intensity of these emotions (e.g., apprehension versus fright, blue feelings versus despair) is a function of the significance of the consumption episode to the consumer's goal (Frijda et al. 1991). This significance depends in turn on (1) the personal relevance of (or involvement with) salient features of the episodes (e.g., a Chicago Bulls fan attending one of their games), and (2) the potential consequences of the episode with respect to these salient features (e.g., Game 7 of the playoff finals).



The Affect Recruitment Heuristic

Overview of the Process

Let us return to our first opening example: Marla is trying to decide how to spend her afternoon. She has several options: playing tennis with Mark, clothes shopping in the mall with Diane, or reading a novel in the park. In the typical consumer decision making paradigm, information about the alternatives is either present in the consumer's environment or has been memorized previously. In contrast, the information required for Marla's decision may not be available as such in her current environment or her memory. She does not "know" yet what will happen. For instance, how pleasant








24
will her tennis game actually be? Because of the uncertainty surrounding the actual content of each episode, she may have to anticipate how the alternative episodes will unfold.

The main theoretical account offered for such a process is Fishbein and Azjen's (1975) attitude formation model. Maria may access salient beliefs about each episode and integrate their evaluative implications into an overall evaluation of (or attitude toward) each episode. For example, she may access propositional beliefs such as "It will be hot on the tennis court," "I'll get a chance to wear my new Nike shoes" and "I haven't exercised much lately," and reach a moderately favorable evaluation of the episode. Alternatively, instead of computing a new evaluation, she may retrieve a previously formed evaluation (cf. Lingle and Ostrom 1979). For instance, when thinking about the tennis game episode, she may retrieve a summary evaluation, such as "It's nice to play tennis on Sundays," that she computed earlier.

There is an alternative way in which this consumer may evaluate each episode and make a decision. Instead of accessing propositional beliefs about the episode, she may access a concrete (analog-like) representation of the episode, possibly a representation that includes herself. That is, she may literally "picture" the episode to herself. While holding this "picture", she may experience anticipatory affective responses. Then she may use these anticipatory affective responses as inputs to her decision. This process is called the affect recruitment heuristic.

Previous research has suggested that people may use their affective response to current objects as a source of information about these objects (e.g., Keltner et al. 1993;








25

Gorn et al. 1993; Martin et al. 1992; Schwarz and Clore 1983). It is proposed that consumers may similarly use affective responses to hypothetical construals of consumption episodes (i.e., anticipatory affective responses) as a source of information about these episodes. Clore (1992) suggests a similar process in person impression formation. He proposes that to evaluate people about whom we have not previously formed an opinion, we may hold the concept of this person in consciousness, and assess if any feelings can be detected. The affect-recruitment heuristic, an extension of Postulate 1, may be formalized as follows:


Thesis: Consumers' decisions about future (or hypothetical) consumption episodes
are based in part on their anticipatory affective responses to construals of
these episodes.


Constructing Consumption Episodes

An important feature of decisions about consumption episodes is that the necessary information about the episodes may not be available in the consumers' present environment or in their memory. As a result, such decisions may necessitate some form of prediction about the episode's content. For instance, to decide whether to bring his date to a given restaurant, Frank may try to predict how the episode will unfold (e.g., whether the setting would allow them to carry an intimate conversation or whether she would find anything she likes on the menu). Although, previous research has assumed that the "expected" content of the episode would consist of a series of propositional beliefs, it is proposed that judgments and decisions about consumption episodes may also involve more concrete analog representations of the episode's content.








26

Various theoretical frameworks suggest that to predict how a consumption episode will unfold, a consumer may mentally simulate this episode in his or her mind. This idea can be traced back to Lewin's (1951) field theory, where behaviors are described as driven by forces in the psychological field operating at different levels of reality. Lewin suggests that in planning, the person operates at an intermediate level of reality-irreality. At this level, "locomotion," the modification of psychological facts (e.g., imagining the atmospheres of several restaurants), is easier than in actual reality (e.g., physically checking out the restaurants). More recently, Kahneman and Tversky (1982) advanced the notion of a "simulation heuristic." They suggest that predictive judgments about the likelihood of events are often made by mentally "running" different scenarios of these events and relying on the ease with which the different scenarios are simulated. Even more recently, Taylor and Schemer (1989) have suggested that the mental simulation of past, future, and hypothetical events serves both problem solving and emotional regulation functions. They advance a definition of mental simulation consistent with the mental construction of consumption episodes postulated by the present framework:

[Simulation means] the cognitive construction of hypothetical scenarios or the reconstruction of real scenarios. These may include (1) rehearsal of likely future events [...] (2) reconstruction of past events [...] (3) fantasies [...] and (4) mixtures of real and hypothetical events, such as the reconstruction of a past
event [...] with a new ending (Taylor and Schneider 1989, p. 175).


That is, consumers may construct consumption episodes by retrieving past episodes (e.g., remembering how exciting it was to attend a ball game) or imagining hypothetical scenarios (e.g., imagining how embarrassed they would be if their date hates the restaurant they have chosen). It is further proposed that the representational format








27

of these construals is often concrete and analogical rather than abstract and propositional (Epstein 1990, 1994; Taylor and Schneider 1989). Proposition 1: Construals of consumption episodes involved in the affect recruitment heuristic are often episodic and nonpropositional.


Although, the idea that consumers may actively construct concrete mental representations of consumption episodes to help their decisions is consistent with several theoretical frameworks, empirical evidence of a mental simulation process has been restricted to counterfactual thinking: the generation of alternatives to events that have already happened or are occurring (Gleicher et al. 1990; Rose 1994; Well and Gavanski 1989). Some empirical evidence that people processing information experientially may be more responsive to concrete than to abstract representations was recently reported by Kirkpatrick and Epstein (1992).



Anticipatory Affective Responses and the Appraisal of Constructed Episodes

Once a representation of a consumption episode has been accessed, this representation is appraised through the same processes as the ones involved in the appraisal of actual episodes (cf. Mandler 1984b). This appraisal process may instantiate an affective response. To stress the fact that the responses are actually instantiated or experienced at the time of the decision, they are called anticipatory affective response.








28

The term "anticipated" affective responses is reserved for more abstract predictions of future affective experiences (e.g., "It will be fun"; "I would feel guilty").2

Several lines of evidence suggest that mental construals of consumption episodes may elicit anticipatory affective responses. First, a common affect manipulation consists of having subjects bring into consciousness affectively charged episodes, either by recalling them or by imagining them. The effectiveness of this manipulation indicates that affective responses can be based on mental construals rather than actual events (Taylor and Schneider 1989). Second, Roseman (1991) showed that, by reading short stories whose contents were systematically manipulated along key appraisal dimensions, subjects were able to predict relatively accurately the affective experiences of the characters in the stories. This finding suggests that appraisal processes are relatively reliable and can be applied with flexibility to hypothetical consumption episodes. Finally, it has been shown that while reading fictitious stories, readers spontaneously represent the characters' emotional states (Gernbacher, Goldsmith, and Robertson 1992). This implies that not only consumers can predict how a hypothetical (constructed) consumption episode will make them feel, but they represent these anticipatory affective responses relatively spontaneously.

Proposition 2: Mental construals of consumption episodes may instantiate anticipatory affective responses. These anticipatory affective responses are generated by appraisal processes similar to those influencing affective responses to actual consumption episodes.


2 The distinction between "anticipatory" and "anticipated" affective responses
parallels distinctions between "experiential" and "noetic" representations (Stepper and Strack 1993) and between "knowledge by acquaintance" and "knowledge
about" (James 1890/1950).








29


Relying on Anticipatory Affective Responses

It is proposed that consumers attach unique information value to these anticipatory affective responses, and as a result, incorporate these responses into their decisions about consumption episodes. In other words, consumers do not only consider the evaluative content of their construals (e.g., imagining buying a nice outfit at the Gap store vs. imagining not being able to find anything), they also rely on how these construals make them feel (e.g., joyful vs. frustrated). This proposition is consistent with recent discussions in various literatures.

In the social cognition literature, Strack (1992) has recently suggested that social judgments can be reached either through an "informational" route or through an "experiential" route. In the latter route, judgments are based on intero- and proprioceptive cues, i.e. feelings. This route includes both emotional feelings (e.g., to make evaluative judgments) and nonemotional feelings (e.g., inferring that a book in uninteresting from one's fatigue while reading it). Clore (1992) similarly suggested that feelings, emotional (e.g., cheerfulness) and nonemotional (e.g., fatigue), guide a variety of judgments.

In the personality literature, Epstein (1990, 1994) postulates the existence of two processing systems, called "rational" and "experiential." Consistent with the present framework, he suggests that when a person is exposed to an emotionally significant event, his/her experiential system searches for similar events in memory and their associated affective traces. The recollected feelings are activated and guide subsequent processing and reactions. These feelings may be subtle, in which case they are referred








30

to as "vibes." Unfortunately, Epstein offers only remote empirical evidence of the alleged experiential processes (e.g., Epstein 1990, 1994; Epstein et al. 1992; Kirkpatrick and Epstein 1992).

Some evidence consistent with the proposition that anticipatory affective responses have unique information value can be found in the behavioral decision literature. Kahneman and Snell (1990) asked subjects to evaluate sequences of hedonic outcomes (e.g., receiving a series of increasingly painful injections). They observed that to make their evaluations, subjects did not seem to aggregate the separate hedonic values of these outcomes (e.g., adding the unpleasantness of each injection). Instead, they seemed to consult an instantaneous representation of the sequence, which Kahneman and Snell call "news," and rely on their emotional responses to the news.

It appears easier to imagine and evaluate one's emotional response to receiving news than to imagine responses to a prolonged sequence of events. The assessed response to news can therefore be used as a heuristic for the task of evaluating
complex outcomes (Kahneman and Snell 1990, p. 301).

According to Kahneman and Snell (1990) "news utility" serves as a proxy for "experience utility" (i.e., prediction of one's future hedonic experience). Stated differently, anticipatory affective responses are use to predict future affective experiences (i.e., anticipated affective response). Although evolving from a very different theoretical perspective, these authors' ideas are strikingly similar to those proposed in the present framework.

Proposition 3: Anticipatory affective responses may be directly incorporated into decisions about consumption episodes because of their unique information value.








31


The Determinants of Affect Recruitment

The first three propositions stress unique aspects of the affect recruitment heuristic. They do not specify the conditions under which this heuristic will be used, that is, the determinants of affect recruitment. To identify these determinants, the present framework relies on the following logic: If anticipatory affective responses are information, their use in decision and judgments should respect principles similar to those guiding the use of other informational inputs. These principles can be specified using Feldman and Lynch (1988) terminology. Their framework posits that the use of informational inputs in judgments is a positive function of their relative accessibility and diagnosticity, compared to other inputs (Feldman and Lynch 1988; Lynch, Marmorstein, and Weigold 1988). The determinants of the diagnosticity of anticipatory affective responses are examined below. The determinants of their accessibility are discussed next.

Active Goal and Diagnosticity. The perceived information value of an affective response (anticipatory or not) is defined as its diagnosticity (Feldman and Lynch 1988).' Substantial evidence shows that people are more likely to incorporate their affective states in their judgments when they perceive these states to be diagnostic for the judgments (see Schwarz and Clore 1983, Strack 1992, and especially Keltner, Locke, and Audrain




3 Strack (1992) articulates a similar concept, called "representativeness," and
distinguishes between content representativeness and process representativeness.
Content representativeness is the degree to which the information conveyed by the feeling experience is similar to the target of the judgment. Process representativeness is the degree to which the event that produced the feelings is
relevant for the judgment.








32

1993). It would be logical that, in judgments of hypothetical consumption episodes, the influence of anticipatory affective responses should also be moderated by their diagnosticity. Three factors may affect the diagnosticity of anticipatory affective responses.

First, consumer researchers have historically identified two classes of consumption motives. Alderson (1957) contrasted consumption behaviors undertaken for their own sake (e.g., reading a novel, watching a movie) with behaviors undertaken to achieve some other goal (e.g., buying groceries to prepare diner, ironing a shirt before going to a job interview). Whereas the former, called congenial behaviors, are intrinsically rewarding, the latter, called instrumental behaviors, are seldom rewarding in themselves (Alderson 1957). Holbrook and Hirschman (1982; Hirschman and Holbrook 1982) suggest that, when consumers' motives are congenial or experiential, emotional and aesthetic criteria are more important determinants of their behavior than when their motives are instrumental. Translated into the present framework, their proposition implies that anticipatory affective responses should be more diagnostic when the consumption episode is evaluated for congenial reasons than when it is evaluated for instrumental reasons. For example, Lehmann (1987) suggests that some consumers lift weights for instrumental reasons (e.g., weight reduction, rehabilitating an injury), while others pursue this activity for more congenial reasons (e.g., "it feels good"). Presumably, when deciding whether to lift weights on a particular occasion, anticipatory affective responses should be less diagnostic among the former group of consumers than among the latter.








33

Indirect evidence for this proposition is provided by Millar and Tesser (1986, 1992). Consistent with Alderson (1957), the authors draw a distinction between "instrumental" behaviors, which are "intended to accomplish a goal which is independent of the attitude object," and "consummatory" behaviors, which are "engaged for their own sake" (Millar and Tesser 1992, p. 282). They further propose that attitudes have a cognitive component (the attributes and beliefs about the attitude object) and an affective component (the emotions and feelings associated with attitude object). In Millar and Tesser (1986), subjects were assigned either an instrumental goal or a consummatory goal for playing with puzzles. After expressing either cognitively-based attitudes or affectively-based attitudes toward these puzzles, subjects were given the opportunity to play with them. Among subjects with a consummatory goal, the amount of time spent playing with the puzzles correlated more strongly with affectively-based attitudes than with cognitively-based attitudes. The pattern of correlations was reversed among subjects with an instrumental goal. In the present framework, these results suggest that affective responses to the puzzles were more predictive of playing behavior when it was driven by consummatory motives than when it was driven by instrumental motives. In contrast, the predictive ability of "reasons why" was stronger when subjects had instrumental motives than when they had consummatory motives. This is consistent with the proposition that anticipatory affective responses to consumption episodes should be more diagnostic when consumers' motives are congenial or experiential than when they are instrumental.








34

Second, certain events are associated with socially shared beliefs about how one should feel during these events. Hochschild (1979) refers to these beliefs as "feeling rules." For example, on a New Year's Eve consumers believe that they should feel happy and cheerful. On Thanksgiving Day, which is dominated by cultural themes of abundance and self-indulgence (Wallendorf and Arnould 1991), people should not feel guilty. Feelings rules may also prescribe negative affective states. For instance, the loss of a loved one is culturally associated with a period of mourning in which people are expected to feel sorrowful. Presumably, when consumption episodes are judged in the contexts of events strongly associated with feeling rules, anticipatory affective responses should be perceived to be particularly diagnostic. For example, how cheerful a consumer expects to be at a party should be more diagnostic for judging it if it celebrates New Year's Eve than if it is held for no special reason.

Third, consumers' concern about how consumption episodes will make them feel should be higher when their preexisting affective state is aversive than when it is not. Abundant research in the literatures on helping (e.g., Cialdini, Darby, and Vincent 1973; Manucia, Baumann, and Cialdini 1984), mood management (e.g., Zillmann 1988; Bryant and Zillnann 1984), and coping (e.g., Lazarus and Folkman 1984) suggests that individuals experiencing aversive affective states are particularly attentive to aspects of their environment that might help them alleviate those states. Therefore, anticipations about how consumption episodes would make them feel should be more diagnostic among consumers in aversive affective states than among consumers in neutral or pleasant affective states.








35

The results of a recent series of experiments by Forgas (1991) support this proposition. Subjects were first induced to be in a positive, neutral, or sad mood state. In a purportedly unrelated experiment, they were given the opportunity to select a partner to work as a team. The list of potential partners contained individuals described as socio-emotionally rewarding (e.g., friendly, likeable) but not competent (e.g., not intelligent, poor exam performance), and individuals described as competent (e.g., intelligent, good exam performance) but not socio-emotionally rewarding (e.g., unfriendly). Whereas happy and neutral (control) subjects selected their partners primarily on the basis of their competence, sad subjects selected their partners on the basis of the socio-emotional reward that their company may provide. Interestingly, the process by which subjects selected their partners was also different across mood conditions. Whereas happy subjects compared potential partners feature by feature (across individuals), sad subjects compared them individual by individual (across features). Therefore, sad subjects relied more than happy subjects on overall impressions of the described individuals to select their partners. These results are very consistent with the proposition that: (1) consumers experiencing aversive affective states should attribute more importance than consumers in neutral or positive states to the affective responses that consumption episodes might elicit, and (2) the process by which they might anticipate these affective responses is by constructing overall representations of each episode.








36

Proposition 4: The relative diagnosticity of anticipatory affective responses to consumption episodes (and their influence on judgments about these episodes) is higher under the following conditions:

(a) The underlying consumption motive is congenial;
(b) The social context is strongly associated with feeling rules;
(c) The preexisting affective state is aversive;



Accessibility. Some research suggests that the influence of mood states on judgments is moderated by their relative accessibility. For instance, Isen and Shalker (1982) found that subjects induced to be in a positive mood evaluated slides more positively than subjects in a control condition. However, this effect was somewhat stronger for ambiguous slides than for clearly pleasant or unpleasant slides (see also Miniard et al. 1992). It is likely that the higher the relative accessibility of anticipatory affective responses to consumption episodes, the stronger their influence on judgments about these episodes. Four determinants of the intensity of anticipatory affective responses can be identified.

First, strong anticipatory affective responses should be more accessible and exert stronger influence on consumers' judgments of consumption episodes than weak anticipatory affective responses. Suppose that consumers are trying to decide whether or not to attend a basketball game. It is probable that their anticipatory emotions will be stronger and more salient if it is "Game 7" of the playoffs' final than if it is a regular season game. Therefore, one would predict that anticipatory emotion should play a greater role in consumers' decision in the former than in the later case.








37

To the extent that vivid representations of consumption episodes should elicit stronger affective responses (e.g., Lang 1984), a second determinant of the accessibility of anticipatory affective responses might be the consumer's imagery abilities. Consumers who have strong imagery abilities (high "imagers") should have easier access to the affective responses that the consumption episode might elicit than consumers with weaker imagery abilities (low "imagers").

An important determinant of the salience of affective responses is the level of selffocused attention or self-awareness (Scheier and Carver 1977). People are more likely to regulate their behavior in accordance with their affective responses to the situational context when self-awareness is high than when it is low (Carver, Blaney, and Scheier 1979a, 1979b; Carver and Scheier 1990, Scheier and Carver 1977). Therefore, anticipatory affective responses to consumption episodes should be more accessible when self-awareness is high.

Finally, several factors may increase (decrease) the relative accessibility of anticipatory affective responses by decreasing (increasing) the accessibility of alternative inputs. In particular, certain episodes may be evaluatively complex, making it difficult to reach an overall attitudinal judgment (Allen et al. 1992). In contrast, affective responses that simply need to be "read out" may be readily accessible (Strack 1990). Therefore, the relative accessibility of anticipatory affective responses may be higher for evaluatively complex episodes than for simpler episodes (Schwarz 1990; Strack 1992). This prediction is consistent with Kahneman and Snell's (1991) observation that people seem to rely on anticipatory feelings to evaluate complex sequences of outcomes.








38

Proposition 5: The relative accessibility of anticipatory affective responses to consumption episodes and their influence on judgments is higher in the following conditions:

(a) The anticipatory affective response is intense;
(b) The consumer has high imagery abilities;
(c) The consumer is highly self-aware;
(d) The episode is evaluatively complex.



A Graphical Summary

A graphical summary of the framework in presented in Figure 1. Conceptually, affect-recruitment involves three sets of elements: subprocesses, contents, and determinants. Three types of subprocesses are hypothesized: (1) the construction of a mental representation of the episode to be judged; (2) an assessment of the significance of this representation for personal well being, that is, its appraisal (Lazarus 1991); and

(3) the judgment integration. Three types of content are fed into or generated by these subprocesses: (1) a mental representation (or construal) of the consumption episode, (2) an anticipatory affective response to this construal, and (3) alternative inputs for making the judgment, such as prior attitudes or salient beliefs about the episode. The determinants of the heuristic are specified in terms of accessibility and diagnosticity principles. The higher the relative accessibility of anticipatory affective responses (compared to other inputs), the more likely it is that consumers will rely on affect recruitment. Factors such as the consumer's imagery ability and the episode's evaluative complexity are hypothesized to influence affect recruitment by their effects on the accessibility of anticipatory affective responses. Similarly, the higher the relative diagnosticity of anticipatory affective responses, the more likely it is that consumers will








39

rely on affect recruitment. Initiation of the heuristic is hypothesized to be contingent upon the consumer's active goal. By altering this goal, hence the relative diagnosticity of anticipatory affective responses, factors such as the consumers' motives and the salience of feeling rules may increase or decrease the likelihood that affect recruitment will be used. As depicted in Figure 1, consumers are assumed to have substantial control over the heuristic. Figure 1 suggests that they actively try to recruit anticipatory affective responses when these have high information value. It is also possible that the process is more "passive" than suggested in Figure 1. Consumers may weight the relative diagnosticity of anticipatory affective responses only at the judgment integration stage. That is, consumers may assess the relevance of anticipatory affective responses only after these are accessed.



Appraising Consumption Episodes for Affect Management

Recently, researchers have become interested in how people manage their affective states, especially when these states were aversive (e.g., Morris and Reilly 1987; Zillmann 1988). It has been suggested that consumption may play an important role in affect management (Gardner and Scott 1990; Kacen 1994). However, the process by which consumers select among consumption alternatives to manage their affective states needs clarification. The proposed framework may also provide a process explanation of consumption decisions involved in affect management.









SUBPROCESSES


I


I


-r -----------------------


ibility


EC rIng p -r- O Azosmlev






tIpto


CONTENTS



1Determinants of Accessibility: Intensity; Self-Awareness; Imagery Ability; Evaluative Complexity FIGURE 1
THE AFFECT RECRUITMENT HEURISTIC


0P


I


IApprisa1 -Process Dimensions Schematic Desirability Controlled Agenc .


I


I1


Access


-Gon


ANTECEDANTS








41


The Motivational Significance of Affective States

The present framework posits that consumers' preexisting affective states may influence their processing goals when constructing and appraising consumption episodes. The processing goal underlying this process is the attainment of a desired affective state (Gould 1991), which for an aversive affective experience is the alleviation of this experience. Consumption episodes are constructed and appraised with respect to this goal. This process results in anticipatory affective responses that might be experienced as "this would make me feel better (or worse)." Presumably, the consumption episode will be evaluated favorably to the extent that the consumer anticipates feeling better in response to it.

Prior research on mood management has mainly examined the valence of preexisting affective states, assuming that negative states would increase the attractiveness of rewarding or positive stimulation. For instance, Gardner and Scott (1990) suggest that when negative affective states are experienced consumers are more likely to consume "feel-good" products (e.g., cookies). Similarly, Zillmann (1988, p. 1952) proposes that "persons in acutely aversive states select hedonically positive stimuli over hedonically negative stimuli." Therefore, Zillmann (1988) hypothesized that television viewers experiencing negative affect should select hedonically positive television programs, such as TV comedies. However, attempts to validate this intuitive proposition have produced inconsistent results (e.g., Helregel and Weaver 1989; Zillmann, Hezel, and Medoff 1980), possibly because these studies understate the motivational significance of affective states by considering only their valence.








42

As stated earlier in the paper, affective states may signal more than the pleasantness or unpleasantness of a situation. Equally-valenced affective states, such as anger and fear or pride and hope, provide different information about the state of the consumer's environment. To consumers under different aversive states (e.g., guilt vs. sadness), the phrase "feeling better" should have different meanings. As Table 2.2 indicates, each main negative emotion is underlain by a unique pattern of appraisal information. It is proposed that in the process of managing negative affective states, consumer focus on the aspects of consumption that may correct critical appraisal elements. This proposition is called the part regulation principle. Proposition 6: When experiencing an aversive affective state, consumers will appraise consumption episodes by focusing on the appraisal dimensions that best define their aversive experience. Implications of the Part-Regulation Principle

The part regulation principle makes more specific predictions than Zillmann's (1988) proposition that people in aversive states always seek hedonically rewarding stimulation. The predictions of the principle will be derived for four types of negative affective states.

Sadness. The distinctive feature of the meaning structure underlying sadness-like emotions is the loss or absence of a reward (Lazarus 1991; Ortony et al. 1988; Roseman 1984). Thus, the experience of sadness communicates the need for alternative rewarding stimulation (Schwarz 1990). When appraising consumption episodes, sad consumers should focus on the rewarding stimulation that they may provide. They should expect








43

to "feel better" in response to episodes that contain positive stimulation (e.g., self-gifts, funny movies) than in response to episodes that do not (e.g., shopping for groceries). Proposition 6a: Consumers appraising consumption episodes when experiencing sadness should focus on the rewarding stimulation that this episode may provide.

Some indirect evidence supports this proposition. First, Mick and DeMoss's data (1990, 1992) suggest that one of consumers' main motives for self-gifts is to cheer themselves up when feeling down. Also, in Forgas's (1991) series of experiments described earlier, it was observed that sad subjects were more likely than happy and neutral subjects to select their teammates because of their socio-emotionally rewarding qualities.

Anger. The experience of anger is characterized by the presence of aversive stimulation, of someone (or something) to blame (other agency), and high legitimacy (Lazarus 1991; Roseman 1984). For instance, a consumer would feel angry if an airline company's employee (other agency) is rude to her (aversive stimulation), especially because rudeness is unexpected from a customer-oriented service provider (high legitimacy). Because the absence of reward is not a core element in anger, rewarding stimulation should not be an effective means of alleviating this state. For instance, in Zillmann et al. (1981) subjects were made angry by an insulting confederate and were later given a chance to retaliate. It was hypothesized that angry subjects exposed to pleasant movie segments would retaliate less severely than angry subjects not exposed to this positive stimulation. Contrary to this prediction, showing pleasant movies did not significantly reduce the level of retaliation of angry subjects. Therefore, as suggested








44

by an analysis of the meaning structure underlying anger, rewarding stimulation per se does not seem to alleviate anger.

A more natural way of alleviating anger would be to get rid of the aversive stimulation. Consumption episodes might achieve this purpose to the extent that they occupy consumer's consciousness, that is, by helping mentally get rid of the stimulation. Therefore, when appraising consumption episodes, consumers in a state of anger may be more likely to focus on the distracting potential of the episodes. In Zillmann et al. (1980), subject who had been frustrated by a poor test performance and by an experimenter's disparaging remarks were invited to watch television. Three programs were available: a situation comedy, an action drama, and a game show. Although the researchers predicted that frustrated subjects would be more likely to select the situation comedy (because of its positive valence), frustrated subjects actually selected the game show. To the extent that game show may have been be more absorbing and distracting, Zillmann et al.'s unexpected finding would be consistent with the following proposition: Proposition 6b: Consumers appraising consumption episodes when experiencing anger-like emotions should focus on the distracting potential of these episodes.

In Marlatt, Kosturn, and Lang (1975), subjects were provoked by a confederate, then given or not given the opportunity to retaliate. The dependent variable was the amount of alcohol consumed in a seemingly unrelated wine test. It was found that subjects who were not given the chance to retaliate drank more wine than subjects who had the opportunity to retaliate. According to the present framework, subjects who were given the opportunity to retaliate "got rid" of the aversive stimulation by "getting even"








45

with the confederate. In contrast, subjects who were not allowed to retaliate had to get rid of the offense through other means. Because of its distracting (Hull et al. 1983) and tension-reducing effects (Marlatt et al. 1975), drinking alcohol might have been used to get rid of the aversive state induced by the confederate. Similarly, Kacen (1994) found that to alleviate bad moods (often anger-like feelings) consumers often rely on escape activities such as reading magazines or going to the movies.

Fear and Anxiety. The meaning structure underlying fear-like emotions such as anxiety is defined by high uncertainty over an outcome and low control over a situation (Izard 1977; Roseman 1984). Alleviating such states thus requires regaining control and a sense of mastery. Therefore, when appraising consumption episodes under conditions of fear or anxiety, consumers seeking to alleviate these states should focus on aspects of the episodes that would help them regain a sense of mastery. Proposition 6c: Consumers appraising consumption episodes when experiencing fear-like affective states should focus on the potential of these episodes to help them regain a sense of mastery.



In Friedland, Keinan, and Regev (1992), half the subjects were made anxious by the threat of painful electric shocks. Subjects were then given a choice between pairs of gambling options. The gambles were designed in such a way that one of the options would reinforce subjects' illusion of control over the outcome of the gamble. Subjects' locus of control was also measured. It was observed that compared to nonanxious subjects, anxious subjects with an external locus of control were more likely to prefer the option giving them illusory control over the outcome of the gamble.








46

Mandler's (1984a) proposition that the consumption of certain threatening stimulations, such as horror movies or riding roller-coasters, might actually provide the consumer with a sense of mastery yields an interesting prediction. Anxious consumers might actually be more likely than nonanxious to watch scary movies, but only to the extent that they believe that they can master this experience.

Guilt and Shame. The core meaning structure underlying guilt and shame is a sense of low self-worth (self-agency, low legitimacy/deservingness, Roseman 1984; Ortony et al. 1988). To alleviate such a state, one can either reaffirm one's self-worth or reduce one's self-consciousness. When appraising consumption episodes under guilt or shame, the consumer should focus on aspects of those episodes that may enhance selfworth, restore a sense of legitimacy, or reduce self-consciousness. Proposition 6d: Consumers appraising consumption episodes when experiencing guilt or shame should focus on the potential of these episodes to enhance self-worth, restore a sense of legitimacy, or reduce selfconsciousness.

Cunningham, Steinberg, and Grev (1980, experiment 2) provide evidence that restoring one's sense of legitimacy is of particular concern when experiencing guilt. To induce guilt feelings, subjects were led to believe that they had broken the experimenter's delicate camera. Subjects were later asked to contribute to a charity. In one condition, the rewarding aspect of contribution was emphasized ("to keep the children smiling"), whereas in the other its moral aspect was stressed ("You owe it to the children"). Compared to subjects not induced to feel guilty, guilty subjects were more likely to contribute, but only when the request stressed the moral (i.e. legitimacy) aspect of contributing. The fact that this effect vanished if subjects were in a good mood before








47

the camera incident suggests that the aversive aspect of the guilt feelings motivated their contributions.

In summary, the present framework takes as its central premise that affective responses have information value. Building on three postulates inspired by recent cognitive theories of emotions, the framework introduces the affect recruitment heuristic as a pervasive process through which consumers may reach decisions about consumption episodes. Three propositions define the key characteristics of this process. Other propositions identify potential determinants deduced from accessibility and diagnosticity principles. Finally, it is suggested that the framework may be extended to account for consumption decisions aimed at managing affective states.














CHAPTER 3
INTENDED CONTRIBUTION



The Framework in "The Big Picture"


As stated in Chapter 1, the present framework intends to provide a conceptual bridge between the consumer decision making literature and the otherwise disparate literature on affect and consumer behavior. The framework stresses the information value of affective responses and describes the processes through which they may used in decisions about consumption episodes. This chapter puts this framework in perspective and clarifies its links to existing frameworks.

The present framework owes much to the work of Schwarz and Clore (e.g., Schwarz and Clore 1983, 1988; Schwarz 1990; Clore 1992). The central idea that affect is information emanates from their impressive program of research. However, several differences between their work and the present framework need to be recognized. The first difference is a difference of focus. The "feeling-as-information" research program, which originally aimed at explaining the congruent influence of mood on judgments, has focused on the information value of affective states (e.g., Schwarz and Clore 1988). Although it is implicit in Schwarz and Clore's research that affective responses to the object of judgment also have information value, this issue deserves more explicit


48








49

examination. The information value of anticipatory affective responses to consumption episodes is the focus of the present framework.

The second major difference is that the present framework specifies the conditions under which anticipatory affective responses are likely to be used in judgments and decisions; Schwarz and Clore do not.' A third difference is that prior research has focused on the influence of affect on the evaluation of existing objects (e.g., Keltner, Locke, and Audrain 1993; Martin, Harlow, and Strack 1992; Schwarz and Clore 1983; see also Gorn et al. 1993), while the present framework extends the underlying logic to decisions about future consumption experiences. Indeed, the affect recruitment heuristic is here viewed as an integral part of consumer decision making processes.

Epstein (1990, 1994) has proposed a theory that distinguishes between two processing systems: the rational system and the experiential system. Many of the features of Epstein's experiential processing system (e.g., concrete representations, reliance on feelings) are consistent with the notion of affect recruitment. One difference between the two frameworks is that Epstein suggests that the experiential system operates mostly outside awareness. No such assumption is made in the present framework. Although affect recruitment may not be easily verbalized, it need not be unconscious.

The major difference between the two frameworks, however, is philosophical. By contrasting the "rational" and the "experiential" systems, Epstein essentially associates the latter with irrational processing. Most of the evidence he presents to support the




Except, of course, when they show that mood does not influence judgment when
attributed to a cause unrelated to the judgment object.








50

notion of experiential processing consists of observed departures from normative judgments (e.g., Denes-Raj and Epstein 1994; Epstein et al. 1992; Kirkpatrick and Epstein 1992). Epstein (1994) even cites superstitious thinking and irrational fear as evidence of a separate experiential system. In the present framework, affect recruitment is not considered "irrational." Although affect recruitment is presented as a "different route" to judgment and decision making (a position consistent with Strack 1992 and Clore 1992), no statement is made regarding the normative value of this route. In fact, it is argued that the use of anticipatory affective responses follows similar information processing principles as the use of inputs such as beliefs or prior attitudes.

Although the principles (e.g., accessibility and diagnosticity) governing the use of anticipatory affective responses may be similar to those governing the use of attitudinal inputs, it argued that the process of affect recruitment is distinct from the process described by typical expectancy-value models of attitude formation (e.g., Fishbein and Ajzen 1975). Affect recruitment is distinct in three respects. First, it is proposed (Proposition 1) that in affect recruitment, consumers' representations of consumption episodes are more concrete, analog-like (and less propositional) than described by expectancy-value models. Second, it is proposed (Proposition 2) that in affect recruitment, anticipatory affective responses are indeed instantiated (i.e., feelings are experienced); they are not just predictions (or beliefs) about future affective experiences. Third, it is proposed (Proposition 3) that consumers attach unique information value to these anticipatory responses when making decisions about consumption episodes. That








51

is, these responses influence consumers' decisions beyond and above the evaluative content of the consumption episode.

Interestingly, a model historically related to Fishbein's model--that of Triandis (1980)--is more congenial with the present framework. Like Fishbein, Triandis posits that behavior is a function of intention, and intention is a function of (a) social norms and

(b) an expectancy value assessment of the behavior's consequences. The major difference between Fishbein's and Triandis's models is that in the latter, intention is not the sole determinant of behavior. Triandis proposes that behavior is also determined by the person's habits and that this influence operates outside self-instruction (i.e., is not mediated by intention). A second difference between the two models, more relevant to our discussion, lies in Triandis's treatment of affect. Whereas in Fishbein's model, affect (treated as equivalent to attitude) is the product of an expectancy-valuation of the attitude object, in Triandis's model, affect and expectancy assessment are independent antecedents of intention (and behavior).2 Triandis defines affect toward the behavior as "the direct emotional response to the thought of the behavior" (Triandis 1980, p. 218) or as "the particular configuration of emotions [that] becomes activated at the thought of the behavior" (Triandis 1977, p. 16). This definition is very consistent with the concept of anticipatory affective responses. So is Triandis's suggestion that affect and expectancy-value assessment are independent determinants of behavioral intentions.






2 In other words, in Triandis's (1980) model, intention is a function of three terms:
social norm, affect, and the expected value of the behavior's consequences.








52

Finally, the framework differs from all of the above in that it builds on the notion of affect as information to suggest a process account of consumption decisions geared toward the regulation of affective states. Now that the framework has been positioned conceptually, it needs to be tested empirically.



A Research Agenda on Affect Recruitment



The purpose of this dissertation research is to test the cornerstone of the framework: the thesis of an affect recruitment heuristic. Empirical examination of decision processes underlying affect management consumption is left for future research. Again, three propositions define affect recruitment: (1) access to or construction of concrete representations of the episode; (2) instantiation of (anticipatory) affective responses; and (3) reliance on these responses, because of their unique information value. It is suggested that these three characteristics differentiate this process from the expectancy-value process typically assumed in consumer research.

In addition, the framework identifies a series of determinants of affect recruitment. It must be recognized that these determinants are more than limiting "boundary conditions." They help refine the predictions of the framework by placing more severe a priori constraints on the data. Many of these determinants were logically derived from the processing assumptions of the framework. For instance, the prediction that affect recruitment is more likely under consummatory motives than under instrumental motives was derived from the assumptions that (1) consumers attach








53

information value to anticipatory affective responses; and (2) this information value is contingent upon consumers' goals. Similarly, the prediction that imagery ability should increase the likelihood of affect recruitment was a logical implication of the proposition that affect recruitment may involve more concrete, analog representations of consumption episodes than expectancy-value processes. To the extent that finer predictions are supported by the data, more confidence can be placed in the processing assumptions of the framework, and in its theoretical status.

The substantive significance of the framework also needs to be examined. Substantively, the value of the framework rests heavily on its ability to predict consumer decision outcomes from decision inputs. This issue is examined in Experiments 1 and 2. These experiments focus on several theoretical determinants of affect recruitment. To the extent that these determinants moderate the influence of decision inputs on decision outcomes, it can be concluded that, from a substantive standpoint, the idea of affect recruitment is worth pursuing.

The next three experiments are aimed at testing the processing assumptions of the framework (Propositions 1-3). The proposition that affect recruitment may involve concrete representations is tested indirectly in Experiment 3. This test relies on the logic that if affect recruitment necessitates the construction of concrete representations of the episode, reliance on this heuristic should depend on the person's propensity to process information in an analog (e.g., visual) mode rather than in a more abstract propositional (e.g., verbal) mode. (A similar indirect test is conducted in Experiments I and 2).








54

Experiment 4 attempts to test this proposition in a more direct way through protocol analysis.

Proposition 2, that in affect recruitment affective responses are anticipatorily instantiated, is examined in Experiment 5. Because such responses may be too subtle to be detected through verbal measurement (cf. Epstein's notion of vibes), this experiment relies on a misattribution paradigm to infer the instantiation of anticipatory affective responses. If decisions based on affect recruitment involve the instantiation of affective responses, misattribution of these responses to an external source should result in the subtraction of part of these responses from the decision outcome. To the extent that one cannot subtract what one does not have, observation of a subtraction effect would suggest that affective responses were indeed instantiated.

The third proposition, which stresses the unique information value of anticipatory affective responses, is perhaps the most important one. If the feelings that consumers experience while holding a mental construal of the consumption episode in short-term memory have unique information value, they should influence the decision outcome even when the evaluative content of the episode is kept constant.3 If the information value of affect is not unique, and is redundant with an expectancy-evaluation of the episode (e.g., Fishbein and Ajzen 1975), only the evaluative content of the episode should matter. This critical test is conducted in Experiments 3 and 4.






3 Although, admittedly, an orthogonal manipulation of evaluative content and
anticipatory affective responses probably lacks ecological validity.








55

Finally, a processing framework should be assessed not only on the basis of its predictive accuracy, but also on the basis of its descriptive accuracy. Ideally, its descriptive accuracy would be compared to that of an established framework. Such a comparative assessment is conducted in Experiment 4. This experiment uses a logic of process baselines to assess affect recruitment and expectancy-value attitude formation as paramorphic representations of decisions about consumption episodes.














CHAPTER 4
EXPERIMENTS 1 AND 2:
THE DETERMINANTS OF AFFECT RECRUITMENT


Purpose and Predictions


An important first step in the validation of a processing framework is to examine the ability of this framework to predict how inputs to this process will causally affect its outputs. The more specific the predictions generated by a framework--should they be empirically supported--the more confidence one can have in the heuristic value of this framework. These first two experiments test predictions of the framework on how behavioral intentions toward a consumption episode (a decision outcome) are affected by two types of decision inputs: (1) the experiential content of the episode (pleasant or unpleasant) and (2) the cost associated with this episode.

It was postulated in Chapter 2 that to make a decision about consumption episodes, consumers may construct concrete representations of the episodes that may instantiate anticipatory affective responses. The pleasantness (or unpleasantness) of these responses (e.g., being relazed versus being frustrated) should logically depend on the affective content' of the constructed representation (e.g., imagining having a terrible dinner in a restaurant). This affective content should in turn depend on available and



As indexed, for example, by the appraisal dimensions discussed in Chapter 2
(e.g., desirability, agency, ...).


56








57

accessible information about the episode (e.g., a friend's account of a previous dinner in the same restaurant).

Two types of information were provided to subjects in these experiments: (1) the pleasantness (or unpleasantness) of a friend's previous experience with the episode (an input called experiential content); and (2) the friend's description of the cost associated with the episode. It was reasoned that the experiential content of the episode (conveyed by the friend) would be a strong determinant of the affective content of subjects' own construals of the episode (and presumably of their anticipatory affective responses). In other words, it was reasoned that the manipulated (i.e., observable) experiential content of the episode would be proximally related to subjects' unobserved anticipatory affective responses to the episode. In contrast, it was expected that the cost associated with the episode would be weakly related to anticipatory affective responses to the episode, although it could influence its overall evaluation.

Experiments 1 and 2 examined how four theoretical determinants of affect recruitment would moderate the influence of these two decision inputs (see Figure 1): (1) whether the motive underlying the decision was instrumental or consummatory; (2) whether the decision was made in a context associated with feeling rules (Hothschild 1979), that is, a context where there are cultural expectations about how one should feel (e.g., people are typically expected to feel happy and joyful on their birthday); (3) subjects' chronic affect intensities (Larsen 1984); and (4) subjects' visual imagery abilities. The first two determinants were conceptualized as impinging upon the information value (or diagnosticity) of anticipatory affective responses. Consistent with








58

Proposition 4a, it was predicted that anticipatory affective responses would be perceived as more relevant, as a decision input, in decisions guided by consummatory motives than in decisions guided by instrumental motives (see also Hirschman and Holbrook 1982 and Millar and Tesser 1989). Consequently, affect recruitment was expected to be more likely in the former case than in the latter. Therefore, it was hypothesized that the experiential content of an episode, because of its assumed proximal relation with anticipatory affective responses, would be more influential under consummatory motives than under instrumental motives. Conversely, the associated cost of the episode was expected to be less influential under consummatory motives than under instrumental motives. These two predictions were tested in Experiment 1. H la: The experiential content of a consumption episode will have a stronger influence
on consumers' decisions when consummatory motives are being pursued than
when instrumental motives are being pursued.


H lb: The cost associated with a consumption episode will have a smaller influence on
consumers' decisions when consummatory motives are being pursued than when
instrumental motives are being pursued.


Consistent with Proposition 4b it was predicted that when decisions are guided by consummatory motives, anticipatory affective responses would be perceived as even more relevant (or diagnostic) if the decision making context includes cultural expectations about how the decision maker should feel, i.e., feeling rules. If it is the decision maker's birthday, he/she should be more sensitive to how a given consumption episode will make him/her feel, because people are typically expected to feel happy and joyful on their birthdays. Consequently, feeling rules should further increase the influence of








59

the experiential content of the episode and decrease the influence of its associated cost-predictions tested in Experiment 2.

H2a: The experiential content of a consumption episode will have a stronger influence
on consumers' decisions when the decision context is associated with feeling
rules.


H2b: The cost associated with a consumption episode will have a smaller influence on
consumers' decisions when the decision context is associated with feeling rules.


Two other two determinants examined in Experiments 1 and 2, chronic affect intensity and visual imagery ability, were conceptualized as moderating the accessibility of anticipatory affective responses (see Propositions 5a and 5b). It was reasoned that both individual difference variables would contribute to make anticipatory affective responses to a given construal of the episode more accessible, thereby amplifying the effect of experiential content on behavioral intentions. Moreover, to the extent that the recruitment of anticipatory affective responses should be more likely when they are diagnostic (see Figure 1), it was predicted that the amplifying role of chronic affect intensity and visual imagery ability would manifest itself especially when consumers have consummatory motives or in the presence of feeling rules. A corollary prediction of the framework was that visual imagery ability and chronic affect intensity would not moderate the influence of associated cost, because this input was hypothesized to be processed relatively independently of affect recruitment. H3a: The influence of the experiential content of an episode on consumers' decisions
should be amplified by chronic affect intensity when anticipatory affective responses are diagnostic (consummatory motives or feeling rules) but not when
these responses are not diagnostic (instrumental motive).








60

H3b: The influence of the experiential content of an episode on consumers' decisions
should be amplified by visual imagery ability when anticipatory affective responses are diagnostic (consummatory motives or feeling rules) but not when
these responses are not diagnostic (instrumental motive).

H3c: The influence of the cost associated with a consumption episode on consumers'
decisions is not affected by chronic affect intensity and visual imagery ability.




Experiment 1: Method



SubJects and Overview of the Design

Subjects were 129 undergraduates at the University of Florida, enrolled in several business classes, who participated in return for course extra credit. Subjects were asked to report their intentions of engaging in two hypothetical consumption activities or episodes: test driving a sports car and having dinner in a French restaurant. Three factors were of primary interest in this experiment: (1) the experiential content of each episode (pleasant or unpleasant); (2) the cost associated with each episode (high or low); and (3) the motive for engaging in the episode (instrumental or consummatory). Experiential content and associated cost were manipulated between subjects, whereas motive was manipulated within subjects across the two episodes. The order of the two episodes and of the two motives (hence, their pairing) was completely randomized across subjects. In addition to these manipulated factors, two individual difference variables hypothesized to moderate affect recruitment were measured: subjects' chronic affect intensity and their visual imagery ability.








61


Stimulus Material

Each subject was presented with two scenarios pertaining to two consumption episodes. The first episode consisted of having dinner with a friend of the opposite sex in a new French restaurant in town. The second episode consisted of test-driving a Ferrari at a local dealership as part of a promotional offer for an automotive magazine. Four considerations dictated the selection of these two consumption episodes. First, each episode had to be both relevant for the population of subjects examined (college students) and realistically take place in the college town where the study was conducted. Second, although relevant for the population studied, the episodes had to be relatively novel (a new French restaurant; a Ferrari) to reduce the likelihood that subjects would primarily rely on prior attitudes to make their decisions. Third, each episode had to lend itself to independent (and ideally orthogonal) manipulations of anticipated affective responses and associated cost. Finally, each episode should realistically lend itself to enactment under consummatory or instrumental motives.

The information about each episode was conveyed via a written scenario in which a friend described his or her own experience. This word-of-mouth format was chosen because of its intrinsic ability to convey experiential information. It was reasoned that to make their decisions, subjects would rely on both idiosyncratic inputs (e.g., prior attitudes toward French restaurants) and on the information contained in the friend's account. In affect recruitment they would construct a concrete representation of the episode and use their anticipatory affective responses to this construal as an input to their decision. It was expected that the experiential content of the friend's account would be








62

a major determinant of subjects own anticipatory affective responses. Four versions of each episode were created (two levels of experiential content and two levels of associated cost).

Dinner Episode: Experiential Content. In the positive experience condition, the restaurant episode was described as follows:

I was there the other night. What a classy place! My dinner was wonderful.
You should see the dining room: white tablecloths, linen napkins, slightly dim light, relaxing music in the background . The food was incredible . .
Mmmmm!!! You should have seen how they presented my plate . wow!
With vegetables carved into flowers, you know what I mean! And the waiter was just as classy; he was so polite . he made me feel like a million bucks!
IManipulation of associated cost took place here] I was thinking about having
dinner there tonight. Would you go with me?


In the negative experience condition, the episode was described as:

I was there the other night. What a dump! My dinner was awful! You should see the dining room: cheap paper napkins, the light was too bright, and the music in the background was irritating. The food was terrible... Even the way my plate was presented... gross! And the waiter was so rude ... he made me feel second class. I had such an unpleasant experience. [Manipulation of associated cost took place here] I was thinking about having dinner there tonight. Would
you go with me?

Dinner Episode: Associated Cost. The associated cost was manipulated in the last statement of the friend's account. In the low cost condition, the restaurant was described as very inexpensive, whereas in the high cost condition the restaurant was described as very expensive.

Test-Drive Episode: Experiential Content. In the positive experience condition, the test-drive episode was described as follows:

I tried it yesterday. It was a red one, like in the movies. I was so excited! At idle speed, it sounded like a purring tiger. Then, I pushed on the accelerator ...








63

What an acceleration! I felt like I was in an airplane. I took it on 1-75. It was unbelievable. I was up to 100 mph before I even left the on-ramp! I must have gone for two miles at that speed. I was so thrilled... passing all these cars like they were driving at 40 mph. It felt so good. Everybody was staring at me.
When I returned it to the dealership, I even saw one of my neighbors. You should have seen the look on her face! [Manipulation of associated cost] The
promotion ends next week, why don't you go check it out?

In the negative experience condition the episode was described as:

I tried it yesterday. It was a red one, like in the movies... at idle speed, it sounded like a purring tiger. At first I was really excited, but you won't believe what happened. The dealer got into the car and told me: "You can only drive it around the parking lot." Can you believe that? That was ridiculous! I couldn't even press down the accelerator. I might as well have been in a Geo! The fastest I got was 20 mph. On top of that, everybody at the dealership was staring at me, I felt like an idiot! When I got out of the car, I even ran across one of my neighbors. She had been watching me the whole time! You should have seen the smirk on her face! I was so embarrassed. [Manipulation of associated cost The
promotion ends next week, why don't you go check it out?


Test-Drive Episode: Associated Cost. Again, the associated cost was manipulated

in the last statement of the friend's account. In the low cost condition, the friend

explains that in return for test-driving the car he or she only had to buy one issue of the

magazine sponsoring the promotion. In the high cost condition, the friend explains that

he or she had to buy a one-year subscription of the magazine.

Motives

Each episode description was accompanied by a situational context designed to

manipulate the motive under which subjects would evaluate the episode. The situational

context was presented in a separate frame above the description of the episode. Subjects

were instructed to read carefully this situational context prior to reading the description

of the episode. The situational context for the consummatory motive was identical for








64

the two episodes: "It is the weekend. You don't have to study or work. You just want to relax, enjoy, and have a good time."

The situational context for the instrumental motive was essentially the same for the two episodes. Subjects had to write a paper whose topic was related to the content of the episode. The instrumental situational context for the dinner episode was:

You have to turn in a paper in a few days for a class. The topic of the paper is "The management of foreign-cuisine restaurants." To collect "first hand"
material for the paper, you would like to observe how such a restaurant operates. The instrumental context for the test-drive episode was:

You have to turn in a paper in a few days for a class. The topic of the paper is "How should the interior of sports cars be designed (e.g., control panel, seats)."
To collect "first hand" material for the paper, you would like to check the interior
of a prestige sports car (e.g., Ferrari, Corvette).


Individual Differences Moderators

It has been suggested that factors that influence the accessibility of anticipatory affective responses will moderate the likelihood that consumers will use affect recruitment to make decisions about consumption episodes. Two individual-specific variables related to these factors were assessed at the end of the session. First, it has been shown that individuals differ in terms of how intensely they experience affective stimulation (e.g., Larsen, Diener, and Emmons 1986). This individual difference variable was measured through the Affect Intensity Measure (AIM) Questionnaire (Larsen 1984). Because of limited subject time, a shortened version of the AIM was used in which half (20) of the items in the original scale were randomly selected. The shortened AIM scale is reproduced in Appendix A. Second, visual imagery ability was assessed








65

through seven items taken from Sheehan's (1967) shortened version of the Betts Questionnaire (see Appendix B).



Dependent Measures

After reading the description of each consumption episode, subjects reported their likelihood of engaging in each consumption activity on a 9-point scale (1=1 would definitely not go; 9= I would definitely go). Next, they were instructed to write down everything that went through their minds as they were thinking about the activity, including thoughts, feelings, and memories. Finally, they completed various manipulation and demand checks.2



Experiment 1: Results



Manipulation Checks

Anticipated Affect. The anticipated affect associated with each episode was measured through a two-item semantic differential scale. Subjects were instructed to anticipate their feelings and emotions if they were to actually engage in each episode, then recorded their anticipated affective responses on two nine-point scales, anchored by



2 Deferring the manipulation checks until the end of the experimental session
probably limited their ability to reflect the true amplitude of the manipulations.
However, collecting these checks immediately after each decision would have sensitized subjects about the manipulations and possibly biased their responses to the second episode. A solution to this dilemma would have been to use hold out
samples.








66

"My feelings would be extremely pleasant/unpleasant" and "I would be in a very good/bad mood" (alpha dinner = .93; alpha test-drive = .89). Anticipated affective responses to each episode were submitted to a 3-way ANOVA (experience x cost x motive). As expected anticipated affective responses to the dinner episode were more pleasant in the positive experience condition (M = 7.64) than in the negative experience condition (M = 5.01, F (1, 121) = 55.62, p < .0001, W2 = .29). Similarly, anticipated affective responses to the test drive episode were more pleasant in the positive experience condition (M = 7.65) than in the negative experience condition (M = 6.86, F (1, 121) = 7.34, p < .01, U2 = .05). No other effect was significant. Although these effects may appear small, these checks may underestimate the actual strength of the manipulation since the measures were collected at the end of the session.

Associated Cost. Subjects were asked two questions in relation to the dinner episode. First, subjects were asked "How much do they charge for a dinner at the French restaurant?" (1 = very little; 9 = a whole lot). Second, subjects were asked "Considering your budget, how much would you have to give up to have dinner with your friend at the French restaurant?" (1= I wouldn't have to give up much; 9 = I would have to give up a great deal). Responses to these two questions were averaged to provide a measure of perceived cost associated with the dinner episode (alpha = .73). As expected, the perceived cost associated with the dinner episode was higher in the high cost condition (M = 6.70) than in the low cost condition (M = 3.29, F (1, 121) = 141.60, p < .0001, w2 = .50). In addition, an experience x cost interaction suggested that the effect of the cost manipulation was stronger in the negative experience condition








67

than in the positive experience condition (F (1, 121) = 19.34, p < .0001, w2 = .07). Similar questions were asked in relation to the test-drive episode. Subjects were asked "How much do they ask in return for test driving the Ferrari?" (1 = very little; 9 = a whole lot) and "How much would you have to give up to test drive the Ferrari under the conditions described by your friend?" (1= I wouldn't have to give up much; 9 = I would have to give up a great deal). Responses to these two questions were also averaged (alpha = .66) to provide a measure of perceived cost associated with the testdrive episode. Again, the perceived cost associated with the test-drive episode was higher in the high cost condition (M = 4.42) than in the low cost condition (M = 2.82, F(1, 121) = 27.85, p < .0001, W2 = .16). There was an unexpected main effect of experience. Subjects perceive the cost of this episode to be higher in the negative experience condition (M = 4.17) than in the positive experience condition (M = 3.04, F(1, 121) = 13.95, p < .001, W2 = .08). Perhaps the wording of this check, which did not restrict "cost" to its restrictive monetary meaning, prompted subjects to incorporate "affective costs" (e.g., feeling silly) in their answers. No other effect was significant.



Effects of Experiential Content and Associated Cost

Reported intentions of engaging in each episode were submitted to a 2 (motive) x 2 (experience) x 2 (cost) x 2 (order of motives) x 2 (order of episodes) split-plot ANOVA, with motive as a repeated factor and the other factors between subjects. The treatment means collapsed across episodes (and orders of motives and of episodes) are








68

reported in Table 4. 1. The overall repeated ANOVA results are reported in Appendix C.

Experiential Content. As predicted, there was an experience x motive interaction (F (1,113) = 5.41, p = .02). The effect of experiential content was stronger when subjects had a consummatory motive (enjoying a good time over the weekend; Mpos = 6.07; M.eg = 2.52) than when they had an instrumental motive (writing a paper, Mp0s =

7.01; M.e = 4.82). This result supports hypothesis la.

TABLE 4.1
BEHAVIORAL INTENTIONS (COLLAPSED ACROSS EPISODES)
AS A FUNCTION OF EXPERIENTIAL CONTENT, ASSOCIATED COST AND MOTIVE

Consummatory Instrumental motive motive Experiential content Low High Diff. Low High Diff.
cost cost low- cost cost lowhigh high

Negative 2.97 2.03 .94 5.84 3.69 2.15 Positive 6.55 5.43 1.12 7.86 6.06 1.80
Difference (pos- 3.58 3.40 2.02 2.37
neg)


Effect of Associated Cost. A significant cost x motive interaction was also uncovered (F (1,113) = 2.90, one-tailed p < .05). The effect of associated cost was stronger when subjects had an instrumental motive (Mhigh = 4.93; M1,, = 6.91) than when they had a consummatory motive (Mhigh = 3.87; Mlw = 4.87). This finding supports hypothesis Hlb.








69


Effects of Affect Intensity and Visual Imagery

It was proposed that high accessibility of anticipatory affective responses would increase the likelihood of affect recruitment. Factors that theoretically increase the accessibility of anticipatory affective responses should amplify the effect of experiential content on behavioral intentions. It was hypothesized that, when subjects had consummatory motives, visual imagery abilities and chronic affect intensity would interact with the experience content of the episodes to affect behavioral intentions. That is, two 3-way interactions were predicted: experience x visual imagery x motive and experience x affect intensity x motive.

Subjects' affect intensity and visual imagery ability scores were added (after centering) to the repeated measure ANOVA model described earlier. For parsimony, only the theoretically meaningful interactions between these two predictors, experiential content, cost, and motive were included in the analysis. The overall analysis is summarized in Appendix D.

As predicted by Hypothesis 3c, neither chronic affect intensity nor visual imagery ability interacted with the cost associated with the episode (both Fs < 1). While the experience x affect intensity x motive interaction was not significant (F(1,106) = 2.49, p = 0.12)--disconfirming H3a, the predicted experience x visual imagery x motive interaction was uncovered (F(1,106) = 5.81, p = 0.02). To investigate the nature of this triple interaction, a multiple regression model was estimated at each level of the motive (after dummy coding of the other effects). The results of these two regressions are reported in Tables 4.2 and 4.3.








70


TABLE 4.2
EFFECTS OF VISUAL IMAGERY AND EXPERIENTIAL CONTENT
ON INSTRUMENTAL BEHAVIORAL INTENTIONS


Dependent Variable: Intention Instrumental R2 = 0.378 Parameters Estimate T- P value

Intercept 5.62500 6.47 0.0001 Dumaff x visima -0.01074 -0.14 0.8885 Dumaff 2.88691 2.47 0.0150 Dumcos -3.33929 -2.63 0.0099 Dumaff x dumcos 1.08392 0.63 0.5312 Dummoti 0.62500 0.51 0.6120 Dumaff x dummoti -1.87097 -1.10 0.2717 Dumcos x dummoti 0.80357 0.45 0.6559 Dumaff x dumcos x dummoti -2.09652 -0.84 0.4015 Dumsena -0.84722 -0.71 0.4795 Dumaff x dumsena -0.90009 -0.54 0.5907 Dumcos x dumsena 3.43651 1.97 0.0513 Dumaff x dumcos x dumsena -1.44725 -0.59 0.5531 Dummoti x dumsena 1.59722 0.92 0.3618 Dumaff x dummoti x dumsena 1.46475 0.61 0.5453 Dumaff x dumcos x dummoti x dumsena 4.91883 1.40 0.1629

Dummaff = experience (0: negative; 1: positive); visima = (visual imagery); dumcos = cost (0: low; 1: high); dummoti = order of motives (0: instrumental first; 1: consummatory first); dumsena = order of episodes (0: test drive first;
1: dinner first).








71


TABLE 4.3
EFFECTS OF VISUAL IMAGERY AND EXPERIENTIAL CONTENT
ON CONSUMMATORY BEHAVIORAL INTENTIONS


Dependent Variable: Intention Consummatory R2 = 0.492 Parameter Estimate T-value p Intercept 3.37500 4.15 0.0001 Dumaff x visima 0.17199 2.40 0.0179 Dumaff 2.13427 1.95 0.0537 Dumcos -1.51786 -1.27 0.2051 Dumaff x dumcos -0.34615 -0.21 0.8308 Dummoti -0.25000 -0.22 0.8284 Dumaff x dummoti 0.98551 0.62 0.5356 Dumcos x dummoti 0.82143 0.49 0.6267 Dumaff x dumcos x dummoti 1.31652 0.56 0.5733 Dumsena 0.06944 0.06 0.9506 Dumaff x dumsena 2.43756 1.56 0.1216 Dumcos x dumsena 0.57341 0.35 0.7262 Dumaff x dumcos x dumsena 0.67357 0.30 0.7680 Dummoti x dumsena -1.48016 -0.91 0.3668 Dumaff x dummoti x dumsena -1.78367 -0.79 0.4318 Dumcos x dummoti x dumsena -0.30556 -0.13 0.8975 Dumaff x dumcos x dummoti x dumsena -1.76506 -0.54 0.5914

Dummaff = experience (0: negative; 1: positive); visima = (visual imagery); dumcos = cost (0: low; 1: high); dummoti = order of motives (0: instrumental first; 1: consummatory first); dumsena = order of episodes (0: test drive first;
1: dinner first).








72

When subjects had an instrumental motive, their visual imagery ability did not affect their behavioral intentions (t < 1). In contrast, when subjects had a consummatory motive, their visual imagery ability amplified the effect of experiential content of the episode on their behavioral intentions(t(1 ,111) = 2.40, one-tailed p < .01). This result supports Hypothesis H3b. It is consistent with the idea that when anticipatory affective responses are diagnostic (e.g., consummatory motives), their hedonic content is likely to influence consumers' decisions, especially when these anticipatory responses are highly accessible (e.g., high visual imagery ability).



Correlational Analysis

It was hypothesized that the higher weight assigned by consumers to the experiential content of an episode under consummatory motives was caused in part by stronger anticipations of affective responses to the episode.3 Specifically, it was predicted that anticipated affective responses to the episodes would be a stronger mediator of behavioral intentions when subjects had consummatory motives than when they had instrumental motives. Anticipated affective responses to each episode (test-drive and dinner) were entered as continuous predictor of behavioral intentions toward the corresponding episode. To examine the moderating role of motives, the motive under which the episode was evaluated was entered as an additional dummy coded predictor (0



3 As stated earlier, these anticipations were not assessed at the time of the decision.
Instead, anticipated affective responses were measured retrospectively at the end of the session, which weakens the interpretability of the observed responses as mediators. While these measures are our best estimates of subjects' anticipations
of affective responses, their limitation is fully acknowledged.








73

instrumental; I = consummatory) along with an interaction term between anticipated affect and motive. The results of these regression analyses are reported in Tables 4.4 and 4.5.

There was an expected interaction between anticipated affective responses to the test-drive episode and motive (t (1,125) = 1.78, one-tailed p < 0.05). Anticipated affective responses to the test-drive episode were a stronger predictor of behavioral intentions when subjects had a consummatory motive than when subjects had an instrumental motive. However, for the dinner episode the interaction was opposite to the prediction (t (1,125) = -2.53, p < 0.02). Subjects with an instrumental motive (writing a paper) weighted their anticipated affective response to the dinner episode even more heavily that those with a consummatory motive. The explanation of this effect is not clear. A discussion of this experiment's results follows the report of Experiment 2.



TABLE 4.4
BEHAVIORAL INTENTIONS TOWARD THE TEST DRIVE EPISODE
AS A FUNCTION OF ANTICIPATED AFFECT MOTIVE


Dependent Variable: Intention Test-Drive R2 = 0.241

Parameter Estimate T-value p

Intercept 2.83077 1.73 0.0853 Caraff 0.36684 1.70 0.0907 Dumcar -5.27253 -2.49 0.0141 Caraff x dumcar 0.50437 1.78 0.0770








74


TABLE 4.5
BEHAVIORAL INTENTIONS TOWARD THE DINNER EPISODE
AS A FUNCTION OF ANTICIPATED AFFECT MOTIVE


Experiment 2: Method

This experiment tests the proposition that when decisions are made for consummatory motives, the additional presence of feeling rules (i.e., cultural expectations about how one should feel) further increases the weight of anticipated affect in consumers' decisions about consumption episodes.



Subjects and Overview of the Design

Subjects were 85 undergraduates at the University of Florida, enrolled in several business classes, who participated in return for course extra credit. They were asked to report their intentions of engaging in the same two hypothetical consumption episodes as in Experiment 1 (test driving a sports car and having dinner in a French restaurant). The first two factors were the same as in Experiment 1: the experiential content of each episode and the cost associated with each episode. The third factor manipulated whether


Dependent Variable: Intention Dinner I R2 = 0.521 Parameter Estimate T-value p Intercept -0.66791 -0.81 0.4170 Resaff 1.06403 9.11 0.0001 Dumres 1.59983 1.48 0.1417 Resaff x dumres -0.39964 -2.53 0.0128








75

or not the decision was made in the context of feeling rules (a birthday). Experiential content and associated cost were again manipulated between subjects, whereas decision context was manipulated within subjects across the two episodes. The order of the two episodes was randomized across subjects. The decision context was presented in a fixed order: feeling rules were associated with the first episode but not with the second. A pretest had suggested that this manipulation would be stronger if the feeling rule context was presented first.



Experiential Content and Associated Cost

The manipulations of the experiential content of each episode and of its associated cost were identical to those used in Experiment 1.



Decision Context

Unlike in Experiment 1, all decisions were made for consummatory motives. One of the decisions was associated with additional feeling rules, whereas the other decision was not. For the "feeling rule" decision, subjects had to make their decisions in the context of their birthday. The description of the episode was preceded by a situational context reading as follows: "Today is special, it is your birthday! It is the weekend. You don't have to study or work. You just want to relax, enjoy, and have a good time." For the "no-feeling rules decision, the situation context was described as follows: "It is another weekend (not on your birthday). You don't have to study or work. You just want to relax, enjoy, and have a good time."








76


Experiment 2: Results



Behavioral Intentions

Reported intentions of engaging in each episode were submitted to a 2 (decision context) x 2 (affect) x 2 (cost) x 2 (order of episodes) split-plot ANOVA, with decision context as a repeated factor and the other factors between subjects. The mean behavioral intentions, collapsed across episodes (and order of episodes), are reported in Table 4.6. The overall repeated measures ANOVA results are reported in Appendix E.

The predicted experience x context (F (1,77) = 1.71, p = .19) and cost x context (F (1,77) = 1.17, p = .28) interactions both failed to reach significance. Only directional support for our predictions was obtained. The simple effect of experiential content was somewhat (but not significantly) stronger in the feeling rules context (MP0, = 5.70; Mg = 2.55, W2 = .27) than in the no-feeling rules context (M.S = 5.77; Mneg = 3.69, W2 = .12). Conversely, the simple effect of cost was somewhat (but not significantly) smaller in the feeling rules context (M0w = 4.26; Mhigh 4.02, W2 = .00) than in the no-feeling rules context (M,0 = 5.33; M,2 = 4.14, w2 = .04). Additional analyses that parallel those reported for Experiment I (e.g., effects of visual imagery, affect intensity) failed to yield any significant results and are not reported.








77


TABLE 4.6
BEHAVIORAL INTENTIONS (COLLAPSED ACROSS EPISODES)
AS A FUNCTION OF EXPERIENTIAL CONTENT, ASSOCIATED COST AND DECISION CONTEXT

No Feeling Rules Feeling Rules Experiential content Low High Diff. Low High Diff.
cost cost low- cost cost low-


high


high


Negative 4.18 3.15 1.03 2.73 2.35 0.38 Positive 6.52 5.05 1.47 5.86 5.55 0.31 Difference (pos- 2.34 1.90 3.13 3.20 neg)


Discussion of Experiments 1 and 2

Summary

The two experiments explored the effects of four theoretical antecedents of affect recruitment on decision outcomes indexed by behavioral intentions. Two of these antecedents were conceptualized as affecting the diagnosticity of anticipatory affective responses: (1) whether the motive underlying the decision was consummatory or instrumental and (2) whether the situational context was associated with feeling rules, such as in the case of a birthday. Two other antecedents were conceptualized as affecting the accessibility of anticipatory affective responses: (1) visual imagery ability and (2) chronic affect intensity. The role of these four theoretical antecedents was examined by studying how they moderated the influence of two types of decision inputs:

(1) the affective-experiential content of a consumption episode and (2) the cost associated with this episode.








78

It was observed that when subjects had a consummatory motive, their decisions were more influenced by the experiential content of the episode, and less influenced by the cost associated with it, than when subjects were pursuing an instrumental motive. If, as hypothesized, the experiential content (positive or negative) of the episode is proximally related to anticipatory affective responses (pleasant or unpleasant) to it, this finding would be consistent with the proposition that affect recruitment is more likely when anticipatory affective responses are regarded as a relevant decision input.

When subjects had a consummatory motive, the addition of feeling rules to the decision context slightly increased the influence of experiential content and decreased the influence of associated cost, but not significantly. This lack of significance may be more methodological than substantive. First, Experiment 1 suggests that the manipulation in the "no-feeling-rule" condition may have been strong enough to induce subjects to pay little consideration to the cost associated with the episode and focus instead on its experiential content (see Table 4. 1). Therefore, the nonsignificance of the feeling rules factor may be due to a ceiling effect. Second, it should be noted that both the "no feeling rules" and the "feeling rules" conditions mentioned that it was the weekend. It is conceivable that weekends in and of themselves are associated with feeling rules. That is, people--including college students-presumably have strong beliefs that weekends are for relaxation and enjoyment. In other words, the manipulation may have been weaker than anticipated.

Although chronic affect intensity did not have the expected effect, behavioral intentions were determined by the predicted complex interaction among visual imagery,








79

experiential content, and motive. When subjects had a consummatory motive, and affect recruitment was presumably diagnostic, visual imagery amplified the influence of affective/experiential information presented to them. Visual imagery did not have such an effect when subjects had an instrumental motive. This finding is consistent with the proposition that affect recruitment may involve the construction of a concrete, analoglike, representation of the episode. High visual imagery ability may increase the accessibility of anticipatory affective responses to this representation.



Limitations

Several limitations of these experiments deserve mention. Given the hypothetical nature of the decisions to be made, one may question whether the effects reported here would describe how consumers would make "real" decisions. To remedy this limitation, the experiments reported in the next two chapters will involve more substantive decisions.

Focusing on determinants (i.e., antecedents) of affect recruitment and decision outcomes introduced a more serious limitation of these two experiments. Although it was shown that these antecedents could have meaningful influences on the outcome of consumers' decisions, the underlying processes were not directly examined. For instance, it was suggested that the increased influence of experiential content under consummatory motives was mediated by a stronger weight attached to anticipated affective responses. Whereas a correlational analysis supported this explanation for the








80

test-drive episode, for reasons that are unclear, a similar analysis contradicted this explanation for the restaurant episode.

A related limitation is that varying the experiential content of an episode is only an indirect--albeit ecologically sensible--manipulation of anticipatory affective responses to the episode. It is therefore difficult to ascertain that anticipatory affect per se played a significant role in the observed findings. It is possible that the evaluative content of the episode was the unique determinant of consumers' decisions. To address these issues, the next two studies will attempt to manipulate the affective experience of consumers as they are making a decision, without altering the evaluative content of the episode to be decided upon.














CHAPTER 5
EXPERIMENT 3: THE INFORMATION VALUE OF ANTICIPATORY AFFECTIVE RESPONSES


Purpose and Predictions


The first two experiments examined the ability of the framework to predict how two decision inputs, experiential content and associated cost, influence behavioral intentions toward two consumption episodes: having dinner in a restaurant and test driving a car. Two theoretical determinants of affect recruitment, consumers' motives and visual imagery abilities, were found to moderate the influence of these inputs in a manner consistent with the concept of affect recruitment. It was observed, for instance, that imagery abilities amplified the effects of experiential content when subjects had consummatory motives but not when they had instrumental motives. The consistentcy between such findings and the predictions of the framework suggest that the idea of affect recruiment is at least worth pursuing. A more direct investigation of the hypothesized processes is in order.

A central proposition of this framework is that feelings that are (anticipatorily) experienced in response to a construal of the consumption episode (i.e., anticipatory affective responses) have unique information value. That is, the outcome of consumers' decisions about consumption episodes is not necessarily determined by the evaluative content of their mental representation of the episode; this outcome may also be affected


81


I








82

by consumers' momentary affective responses to this representation. To show that anticipatory affective responses, as decision inputs, are not simply redundant with the evaluative content of representations of the episode, it is necessary to manipulate the former independently of the latter (unlike in Experiments 1 and 2). In Experiment 3, subjects' mood was used to proximally manipulate their anticipatory affective responses to an episode without altering its evaluative content.

Previous research (e.g., Schwarz and Clore 1983) has shown that people often misattribute their current mood states to the object that they are judging. If to make decisions about consumption episodes, consumers "see how it feels" (i.e., rely on affect recruitment), their decisions should be biased by their preexisting mood state: the decision outcome should be more favorable if the preexisting mood is positive than if it is negative. Assuming that this manipulation does not alter the evaluative content of consumers' representation of the episode, this pattern of results would suggest that the influence of moods and emotions experienced during the decision process is at least partially independent of the influence exerted by the evaluative content of the episode. If this effect vanishes when anticipatory affective responses have low diagnosticity, one can additionally infer that it is because of their information value that proximally manipulated anticipatory affective responses influence decision outcomes (e.g., Schwarz and Clore 1983).

In this experiment, subjects who had been induced to be in a positive or a negative mood through a supposedly unrelated task were asked to assess the likelihood that they would be going to a movie over the following weekend. Consistent with








83

Proposition 4a, the diagnosticity of the anticipatory affective responses was manipulated by modifying subjects' motive for going to the movie. Half the subjects were assigned a consummatory motive (anticipatory feelings had high information value), whereas the other half were assigned an instrumental motive (anticipatory feelings had low information value). Because subjects to whom anticipatory affective responses are diagnostic are more likely to inspect how the consumption episode "feels," it was predicted that the mood manipulation would have a greater influence on the decisions of subjects with a consummatory motive than on decisions of subjects with an instrumental motive.

Hla When subjects have a consummatory motive, a positive preexisting mood will
result in more favorable intentions than a negative preexisting mood.


H lb When subjects have an instrumental motive, their behavioral intentions will be
less influenced by their preexisting mood than when they have a consummatory
motive.


In Experiment 1 visual imagery ability was found to be a significant moderator of behavioral intentions. Experiment 3 explored the role of consumers' processing style. It has been suggested that individuals vary in terms of their propensity to process information visually or verbally (Childers, Houston, and Heckler 1985; Richardson 1977). If, as hypothesized, affect recruitment involves concrete analog-like representations of consumption episodes, consumers' tendency to rely on affect recruitment may be correlated with their propensity to process information visually rather than verbally. That is, "visualizers" may be more likely to use a "see-how-it-feels" heuristic than "verbalizers." Therefore, among subjects with a consummatory motive,








84

visualizers may be more influenced by their preexisting mood states than verbalizers. Among subjects with an instrumental motive, processing style should have little effect because subjects should not be likely to inspect their feelings given their low diagnosticity.

H2a Among subjects with a consummatory motive, visualizers will be more influenced
by their preexisting mood than verbalizers.


H2b Among subjects with an instrumental motive, neither visualizers nor verbalizers
will be influenced by their preexisting mood.


Method

Subjects and Design

Eighty-eight undergraduate marketing students at the University of Florida participated in the study for extra course credit. They participated in sessions of 4-18 and were randomly assigned to one of four experimental conditions of a 2 x 2 design. All four conditions were run in each session. The first factor was subjects' preexisting mood (positive or negative) when making the decision. The second factor was their motive consummatoryy or instrumental) for going to the movie. In addition to these two manipulated factors, this experiment also assessed subjects' processing style along the visualizer-verbalizer dimension.

Procedure. Subjects were told that they would be participating in two separate studies. The "first" study was purportedly about scale development. Subjects first completed the Style of Processing Scale (Childers et al. 1985). Their mood was then manipulated by having them provide vivid descriptions of a recent event that had affected








85

them personally. They were told that these descriptions would be used to develop a scale to be used in future research. The "second" study was introduced next as one that investigates students' liking of various consumption activities. Subjects were asked to read carefully the description of a fictitious new movie and to report the likelihood that they would attend a preview for the movie over the weekend. All sessions were run on Tuesdays and Wednesdays. The delay between the study and the consumption episode was designed to ensure that subjects would not assess the episode as a mean to manage their experimentally induced mood state. Finally, other dependent measures, process measures, and demand checks were collected.

Mood. Subjects' mood was manipulated using a task similar to the one used by Schwarz and Clore (1983). Subjects were told that the experimenter was trying to develop a "Scale of People's Concerns." The scale was purportedly intended to assess issues that matter in people's lives so that they could be related to television offerings. To help identify items for the scale, subjects were asked to describe in writing a recent event that had affected them personally. In the negative mood condition they were instructed to describe an event that made them "feel really bad," whereas in the positive mood condition they were instructed to describe an event that made them "feel really good." Subjects were encouraged to be as vivid and detailed as possible. They were given 20 minutes to complete the task.

Stimuli and Motives. In the consummatory motive condition, the instructions emphasized the experience of the going to the movie as the main motive for assessing whether to go to the preview:








86


After a long week of school and hard work, every student deserves some good time on the weekend. So this weekend, you may want to enjoy life, at least for
a few hours.

This weekend there will be several sneak previews of the movie "Oklahoma Bookstore." It is a new comedy staring John Goodman (from ABC's "Roseanne") as the owner of a bookstore in a small town and Steve Martin as an eccentric customer. The New York Times' critics have described "Oklahoma Bookstore" as one of the most hilarious movies of the year. Siskel and Ebert gave it "Two Thumbs Up." The previews will be shown Friday to Sunday at
6:30, 8:30, and 10:15 at Litchfield theaters (in Butler Plaza).

If you want to have a good time (you deserve it!), how likely is it that you will
go to one of the previews of "Oklahoma Bookstore?"



In the instrumental condition, the instructions emphasized an extrinsic reason for

going on not to the movie preview:

Next week (on Tuesday and Wednesday), we will run a commercial study for a movie company. The study is not for credit, but you will receive 4 dollars if you
participate.

To participate, you need to go this weekend to one of the sneak previews of the movie "Oklahoma Bookstore." It is a new comedy staring John Goodman (from ABC's "Roseanne") as the owner of a bookstore in a small town and Steve Martin as an eccentric customer. The New York Times' critics have described "Oklahoma Bookstore" as one of the most hilarious movies of the year. Siskel and Ebert gave it "Two Thumbs Up." The previews will be shown Friday to
Sunday at 6:30, 8:30, and 10:15 at Litchfield theaters (in Butler Plaza).

Of course, if you attend one of these previews this weekend, you don't have to participate in next week's study. But you may. If you decide to participate, bring your ticket stub and we will reimburse it (in addition to the 4 dollars of
your participation).

Given that it would allow you to participate in next week's study (and earn 4 dollars), how likely is it that you will go to one of the previews of "Oklahoma
Bookstore?"








87

It was reasoned that subjects in this latter condition would be less likely to rely on their feelings to make their decisions because the movie experience itself should be less relevant. It should be noted, however, that this scenario does not truly substitute an instrumental motive for an experiential (or consummatory) one. Rather, it provides an instrumental motive in addition to an experiential motive.

Style of Processing. Subjects were administered Childers et al.'s (1985) Style of Processing Scale (SOP), which assesses people's "preference and propensity to engage in a verbal and/or visual modality of processing" (p. 130). The SOP scale, adapted from the Verbalizer-Visualizer Scale (Richardson 1977), contains 22 items: 11 items assessing propensity to process visually and 11 items assessing propensity to process verbally.

Measures. The main dependent variable was subjects' reported likelihood of going to one of the movie previews measured on a nine-point scale (1 = extremely unlikely; 9 = very likely). After completing this measure, subjects were asked to list all the thoughts and feelings that went through their minds as they were making the decision. They were then asked about the purpose of the study. Next they completed several other measures such as their attitudes toward going to the movies on the weekend and their attitudes toward comedies. Finally, as a manipulation check, subjects were asked to report what mood they were in after having described an event that affected them.








88


Results



Assessment of the Independent Variables

Mood Manipulation. At the end of the experimental session, subjects were asked to report how they were feeling after having described their personal experience and before making their decision. These feelings were assessed by five bipolar seven-point items, anchored by cheerful/depressed; sad/joyful; annoyed/pleased; happy/unhappy, in a good/bad mood. Principal component analysis uncovered a single valence factor accounting for 85 percent of the variance. As a result, the five items were summed into a single measure of mood pleasantness (theoretical range: 5-35, alpha = .96). These self reports were submitted to a 2 (mood) x 2 (motive) x 2 (style of processing; see below) ANOVA. As expected, subjects in the positive mood condition reported being in a more pleasant mood (M = 28.67) than subjects in the negative mood condition (M = 24.84, F (1,80) = 112.51, p < 0.0001, omega-squared = 0.55). No other effect was significant.

Motive. The effectiveness of the motive manipulation was assessed using an independent sample of 35 subjects. They were asked which of two perspectives would "most people have in mind when assessing their intention of seeing" the movie. Answers were collected on a seven-point scale anchored by "They will focus on how enjoyable or unpleasant this experience will be" and "They will focus on whether there is something to gain." The question was worded in the third person to reduce social








89

desirability pressures.1 These subjects were also asked to report, on a similar scale, what perspective they had in mind when making their decision. These two measures were highly correlated (r = .92), and were, therefore, averaged into a single index of subjects' motives. Low values of this index indicated a consummatory motive, whereas high values indicated an instrumental motive. The motive index was submitted to a oneway ANCOVA with motive instructions (instrumental or consummatory) as a between- subjects factor and social desirability as a covariate. The analysis revealed that subjects exposed to the instrumental instructions (M = 3.78) were more likely to base their decisions on whether there were "something to be gained" (and less likely to base their decisions on "how enjoyable or unpleasant this experience would be") than subjects exposed to the consummatory instructions (M = 2.42, F(1,32) = 4.02, one-tailed p < .03, omega-square = .08).

To assess potential confounds of this manipulation, subjects were also asked (1) How important it was for them to be accurate in their assessment and (2) How difficult the decision was. Neither of these measures was affected by the motive manipulation (both Fs < 1), suggesting that the manipulation did not induce different levels of involvement and task difficulty.






1 A pretest of the manipulation check had suggested that subjects in the
instrumental motive condition, which involved a monetary incentive, were reluctant to admit that their decision was not purely consummatory. Propensity to respond in a socially desirable way was, therefore, assessed in the main manipulation check study, using six items selected from the Crown and Marlow
(1960) scale (Ballard, Crino, and Rubenfeld 1984).








90

Processing Style. Although Childers et al. (1985) suggested that the visual and verbal components of the SOP scale each capture a single factor, with good internal consistency, the factorial structure suggested by these authors was not replicated in this study. The internal consistency of the eleven visual items of the SOP scale was only moderate (alpha = .69) and a principal component analysis suggested that four factors underlie these items. Although the internal consistency of the verbal items was higher (alpha = .75), a principal component analysis suggested that there were three underlying factors. Therefore, it appears that, in this study at least, the SOP scale had a higher dimensionality than suggested by Childers et al. (1985).2

Despite the scale's higher dimensionality, a visualizer-verbalizer score was computed for each subjects by summing responses to the visual items and subtracting responses to the verbal items, as recommended by Childers et al. (1985). Based on this score, a median split was performed to distinguish "visualizers" from "verbalizers" in















2 In retrospect, this apparent discrepancy may be explained by the fact that Childers
et al. (1985) did not use exploratory factor analysis but confirmatory factor analysis with LISREL. Based on the factor loadings, the authors concluded that a two-orthogonal-factor solution was satisfactory, even though the chi-square
statistic suggested a significant lack of fit.








91

the remaining analyses.3 The split was performed within each of the four experimental conditions (mood x motive) to avoid overly unequal cell sizes.



Effects on Behavioral Intentions

The likelihood of going one of to the movie previews was submitted to a 2 (mood) by 2 (motive) by 2 (processing style) ANCOVA, with the number of movies seen in the last year in the same theater as a covariate. The least-square means are reported in Table 5.1 and the overall ANCOVA results are reported in Appendix F. There was a main effect of motive (F (1,79) = 18.62, p < 0.0001). Subjects with an instrumental motive reported a higher likelihood of going (M = 7.25) than subjects with a consummatory motive (M = 4.93). This effect probably reflects the higher monetary incentive in the instrumental condition (four dollars and a free ticket).

Mood and Motive. The mood x motive interaction was significant (F(1,79)

2.71, one-tailed p = 0.05). A planned contrast reveals that in the consummatory condition subjects in a positive mood (M = 5.67) reported a higher likelihood of going




3 To analyze interactions between a continuous variable (e.g., processing style)
with categorical variables (e.g., motive and mood), Jaccard, Turrisi and Wan (1990) prefer the use of multiple regression, with dummy coding of the categorical variables (as in Experiment 1), to the dichotomization of the continuous variable. Nevertheless, the analyses reported here rely on dichotomization for two reasons. First, the dimensionality of the scale (two theoretical factors and seven empirical factors) calls into question treating the SOP index as a unidimensional continuous measure. Second, dichotomization better reflects the underlying logic of the scale, which is meant to categorize individuals (see Richardson 1977). For completeness, the approach recommended by Jaccard et al. (1990) was also used to analyze the data. The results were
mostly nonsignificant and difficult to interpret.




Full Text

PAGE 1

FEELING LIKE CONSUMING: AFFECT AS INFORMATION IN DECISIONS ABOUT CONSUMPTION EPISODES By TUAN PHAM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1994

PAGE 2

To my parents, Pham Van Lai and Nguyen Thi Luu, for believing in me.

PAGE 3

ACKNOWLEDGMENTS I would not have been in the position of putting the final touch to this dissertation without the intellectual, material and personal support of several individuals and institutions. I am particularly thankful to Joel Cohen, my dissertation chair, for having been a great mentor. He guided me throughout this entire project, very supportive when I doubted, extremely patient when I erred. Working with him has clearly changed the way I look upon things. I know that one day I will find myself telling my own students: "Think about the big picture." For that, I am immensely grateful to Joel. Richard Lutz and John Lynch both have been more than committee members to me. In the last four years, Rich's characteristic red ink comments on my work have always been insightful, motivating, and exceptionally prompt. John has been a constant help in clarifying my ideas and advising me on analysis issues. His rigor, both conceptual and methodological, and his dedication to his students will always be a source of inspiration. Dr. Margaret Bradley has been very generous and flexible with her time and comments. It has been a privilege to have such a remarkable group of scholars on my dissertation committee. I cannot express how grateful I am to the Intercollegiate Center for Management Sciences (ICM, Brussels) for financially supporting my doctoral education. Without the ICM's generous three-year Doctoral Fellowship, this project would never have in

PAGE 4

materialized. I am also indebted to the various professors who allowed me generous access to their students, especially Jack Faricy and John Hall. I thank Christian Derbaix, Alain Bultez, and Louis Eeckhoudt, my former teachers, colleagues, and occasional tennis partners at the Catholic University of Mons for guiding my first steps into academia. Finally, I thank my classmates Corinne Faure, Susan Fournier, and especially Luk Warlop, for being for being great celebration companions. IV

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS iii ABSTRACT CHAPTERS 1. INTRODUCTION: REVISITING THE ROLE OF AFFECT IN CONSUMER DECISION MAKING Feeling Like Consuming 1 An Overview of the Literature 3 Blueprint of a Conceptual Bridge 14 AFFECT AS INFORMATION IN DECISION MAKING: A PROCESSING FRAMEWORK 15 Conceptual Foundations 15 The Affect Recruitment Heuristic 23 Appraising Consumption Episodes for Affect Management 39 INTENDED CONTRIBUTION 48 The Framework in "The Big Picture" 48 A Research Agenda on Affect Recruitment 52 EXPERIMENTS 1 AND 2: THE DETERMINANTS OF AFFECT RECRUITMENT 56 Purpose and Predictions 56 Experiment 1: Method 60 Experiment 1: Results 65 Experiment 2: Method 74 Experiment 2: Results 76 Discussion of Experiments 1 and 2 77

PAGE 6

5. EXPERIMENT 3: THE INFORMATION VALUE OF ANTICIPATORY AFFECTIVE RESPONSES 81 Purposes and Predictions 81 Method 84 Results 88 Discussion 93 6. EXPERIMENT 4: INFERRING AFFECT RECRUITMENT THROUGH PROCESS BASELINES 98 Purpose and Predictions: The Logic of Process Baselines 98 Method 101 Results 106 Discussion 121 7. EXPERIMENT 5: THE MISATTRIBUTION OF ANTICIPATORY AFFECTIVE RESPONSES 128 Purpose and Predictions 128 Method 131 Results 138 Discussion 146 8. CONCLUSIONS 149 On Bricks and Holes 149 A New Agenda 156 APPENDICES A. SHORTENED VERSION OF THE AFFECT INTENSITY SCALE ... 159 B. VISUAL IMAGERY SCALE (SHEEHAN 1967) 161 C. EXPERIMENT 1: REPEATED MEASURES ANOVA OF BEHAVIORAL INTENTIONS 163 D. EXPERIMENT 1: EFFECTS OF AFFECT INTENSITY AND VISUAL IMAGERY OF BEHAVIORAL INTENTIONS 164 E. EXPERIMENT 2: REPEATED MEASURES ANOVA OF BEHAVIORAL INTENTIONS 166 VI

PAGE 7

F. EXPERIMENT 3: ANCOVA OF BEHAVIORAL INTENTIONS .... 167 G. PROTOCOL CONTENT: CODING SCHEME 168 H. PROTOCOL CONTENT AS A FUNCTION OF MOOD AND INSTRUCTIONS 172 I. EXPERIMENT 5: REPEATED ANCOVA OF BEHAVIORAL INTENTIONS 176 J. EXPERIMENT 5: UNIVARIATE ANCOVAS OF BEHAVIORAL INTENTIONS 177 REFERENCES 181 BIOGRAPHICAL SKETCH 195

PAGE 8

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FEELING LIKE CONSUMING: AFFECT AS INFORMATION IN DECISIONS ABOUT CONSUMPTION EPISODES By Tuan Pham December 1994 Chairperson: Joel B. Cohen Major Department: Marketing Unlike previous research, which has either ignored the role of affect in consumer decision making or treated it as part of the decision context, this dissertation takes the premise that affect is a source of information. First, a brief review of previous research on affect and consumer behavior is presented. Then an integrative framework discusses the processes through which affect is used as an input in decisions about consumption episodes. The cornerstone of the framework is the affect recruitment heuristic, which suggests that to make decisions about consumption episodes, consumers may (1) access or construct concrete representations of the episodes, which (2) instantiate anticipatory affective responses, (3) that are used as inputs because of their unique information value. The framework specifies several determinants of affect recruitment, such as consumers' motives and their imagery abilities. The framework also suggests a process explanation of consumption decisions geared toward the management of affective states.

PAGE 9

Propositions of the framework about the affect recruitment heuristic were tested in five experiments. Experiments 1 and 2 focused on predictions of the framework about how different inputs influence the outcome of decisions about consumption episodes. To test the unique information value of anticipatory affective responses, Experiment 3 relied on contextual mood as a means of manipulating anticipatory affective responses to an episode without altering its evaluative content. The results suggest that feelings experienced during the decision process may indeed have unique information value. Experiment 4 tested the descriptive accuracy of the affect recruitment heuristic as a model of how consumers make decisions about consumption episodes. The heuristic appears to be a plausible paramorphic description of how consumers with consummatory motives rely on their feelings to make decisions about consumption episodes. An expectancy-value model did not describe this process equally well. Using a misattribution paradigm, Experiment 5 suggests that anticipatory affective responses are indeed instantiated in decisions about consumption episodes. The results of these five experiments are discussed and directions for future research are provided.

PAGE 10

CHAPTER 1 INTRODUCTION: REVISITING THE ROLE OF AFFECT IN CONSUMER DECISION MAKING. Feeling Like Consuming It is Sunday; the weather is beautiful. Maria is trying to decide how to spend the afternoon. She could play tennis with Mark; she has just bought the latest model of Nike shoes and is eager to try them. Or she could go clothes shopping in the mall with Diane; after shopping, they could have a large strawberry ice cream at Haagen-Dazs. Or maybe she should just go to the park, find a bench in the shade, and finish reading "The Pelican Brief." She "pictures" herself on the tennis court, panting and all sweaty. This thought makes her feel unpleasant. No, I don't feel like playing tennis, she concludes. But the thought of going to the mall excites her: she has a vivid image of herself buying a nice casual dress at the Gap and receiving Diane's compliments. Yes, I really feel like going to the mall, she says to herself. She picks up the phone and calls Diane... Frank has a date tomorrow tonight. He has bought two tickets for the "Phantom of the Opera" on Broadway. But he hasn't decided yet on where to take her for dinner before the play. What would be a good place for the two of them to get to know each other? He remembers a restaurant he went to several months ago. He imagines what it would be like in the restaurant: he is with his date; they are talking to each other; the 1

PAGE 11

2 atmosphere is just right; he can almost smell the food and hear the soft music in the background. This thought "feels good" to him. He concludes that it would be a great place to take his date. He opens the phone book and looks up the restaurant's number. . These two examples depict consumers making decisions about consumption episodes. Consumption episodes are events or activities in which the use of products or services is salient in the consumer's mind. Because alternative episodes (e.g., going to the mall vs. playing tennis) may involve the consumption of very different products or services, such decisions should be of substantial interest to marketing and consumer researchers. Yet, very few empirical studies have examined how these decisions are made. The focus of this dissertation is on decisions about consumption episodes. This focus arose not only from the substantive relevance of these decisions, but also from the fact that their study may reveal important theoretical insights into the process through which consumers make a variety of decisions. As suggested in the opening examples, the central thesis of this dissertation is that the role of affect 1 in consumer decision making may be more fundamental than previously recognized. Affect has been largely ignored in the "traditional" consumer decision making literature (see for instance Bettman, Johnson, and Payne's 1991 review). And when affect was considered, it was mostly as a "background" source of influence, part of the decision context as reflected by the growing literature on mood effects (e.g., Gardner 1985; Kahn and Isen 1993; Mano 1990; Miniard, Bhatla, and Sirdeshmukh 1992). 1 The term affect is used as a generic descriptor of valenced feeling states; emotions and moods being specific examples of affective states and responses (cf. Cohen and Areni 1991).

PAGE 12

3 Unlike previous research, this research program regards affect as a source of information that consumers use to make a variety of decisions. Once the informational status of affective responses is recognized, it becomes easier to incorporate these responses into information processing models of consumer decision making. An integrative processing framework that describes the informative role of affective responses in consumer decision making is proposed in Chapter 2. A central thesis of this framework is the pervasive use of an affect-recruitment heuristic in consumer decision making, particularly but not exclusively in decisions about consumption episodes. This heuristic involves three subprocesses: (1) consumers access concrete representations of the episodes; (2) they instantiate affective responses to these representations; and (3) they use the instantiated affective responses as inputs to the decision. To help position the framework into the literature, prior research on affect and consumer behavior is first briefly reviewed. The outline of the dissertation is presented next. An Overview of the Literature A comprehensive review of the literature on affect and consumer behavior would exceed the scope of this introductory chapter and duplicate the work of others (see Cohen and Areni 1991). The purpose of this section is to provide a brief overview of the existing literature, emphasizing its key insights as well as its shortcomings. Three somewhat independent streams research on affect can be identified in the consumer behavior literature. Their main characteristics (e.g., typical substantive domain, methodological paradigm, limitations) are summarized in Table 1.1.

PAGE 13

o > < X UJ CO 05 UJ oo Z o u Q z < u W U. U< z o X u a: < UJ oo UJ C* UO C/3 < UJ oo LU UJ ac E3 5 _> Q <

PAGE 14

5 Affective States as Contextual Influences An extensive amount of research has focused on the effects of contextual mood states on a variety of dependent variables, including: ad evaluation (Goldberg and Gorn 1987), product and brand evaluations (e.g., Axelrod 1963; Dommermuth and Millar 1967; Gardner and Wilhelm 1987; Gorn, Goldberg, and Basu 1993; Miniard etal. 1992), behavioral intentions (Alpert and Alpert 1990; Swinyard 1993), variety seeking (Kahn and Isen 1993), and time perception (Hornik 1993; Kellaris and Mantel 1994). This stream of research has its main theoretical underpinning in social psychology, where contextual mood has been found to have a variety of effects on social cognition and social behavior (see Isen 1984 and Gardner 1985 for reviews). Most of these studies rely on an experimental paradigm where different mood states (positive, negative, or neutral) are induced among different groups of subjects 2 that are then compared on the dependent variable(s) of interest. The main axiological concern of this stream of research can be described as "bias-driven. It is to show that, somehow, consumers in a positive or negative mood do not process information as they would if they were in a neutral state. Essentially, the contribution of this research amounts to a vast collection of empirical effects that show that contextual mood does matter. Another contribution of this research comes from its growing concern for explaining the processes that underlie the effects of affective states (e.g., Batra and Stayman 1990; Pham 1994; Srull 1984, 1987). 2 The particular mood induction procedure may vary: background music, television programs, reading materials, small gifts, etc.

PAGE 15

6 This stream of research has two main limitations. First, by focusing on various biasing effects of contextual mood, this research has yielded a highly fragmented knowledge base from which it is difficult to extract general principles of consumer behavior (beyond the fact that mood states do matter). One exception is the robust congruency bias that consumers' moods exert on their evaluations of marketing stimuli. In Dommermufh and Millar (1967), for instance, consumers exposed to a pleasant movie evaluated more favorably a product tasted while in a prior neutral mood than did consumers exposed to an unpleasant movie. Miniard et al. (1992) obtained similar results using background music as a mood induction procedure. Mood induced by the context of advertisements was also found to affect ad and brand evaluations in a congruent direction (e.g., Gardner and Wilhelm 1987; Goldberg and Gorn 1987). The explanation generally offered for this pervasive effect is an accessibility explanation (Isen, Shalker, Clark, and Karp 1978). Mood states may increase the accessibility of mood-congruent material in memory (cf. Bower 1981). If moodcongruent cognitions about the stimulus object (e.g., a product or an ad) are more likely to be accessed, evaluations of this object may become biased toward the person's mood state. More recently, Schwarz (1990) and his colleagues (e.g., Schwarz and Clore 1983, 1988) have proposed a different explanation. They argue and provide compelling evidence that mood states bias evaluations of various stimulus objects because people use their mood as information about the objects. This mood-as-information explanation was recently tested and supported by Gorn et al. (1993) in a marketing context. As detailed

PAGE 16

7 in Chapter 2, the theoretical framework of this dissertation expands on the idea that people's affective responses are used as information in a variety of decisions. A second limitation of the affective context stream of research is that the induced mood states are typically independent of the judgment to be made or the task to be completed. In this research paradigm, the influence of affect -on consumer decision making, for instance-is regarded as merely contextual. This perspective ignores the possibility that affect may have a more central function in consumer decision making: being a source of information. A more minor criticism of this research is that it often reduces affect to a mere distinction between good mood and bad mood or between pleasant states and unpleasant states, whereas other research (see below) shows that affect may be considerably richer. Affective Responses as Consumption Outcomes A second stream of research focuses on affect as the outcome of consumption activities, that is, affect as a response to consumption. Consistent with Holbrook and his colleagues' plea to broaden the scope of consumer research (e.g., Holbrook 1986; Holbrook and Hirschman 1982; Hirschman and Holbrook 1982), a number of studies have examined affective responses to consumption as a dependent variable of intrinsic interest. The substantive domain of these studies is a variety of consumption phenomena. Some studies focus on one specific behavior or phenomenon: important purchases (Richins, McKeage, and Najjar 1992, study 2), impulse purchases (e.g., Gardner and Rook 1988; Rook 1987; Weinberg and Gottwald 1982), compulsive buying (e.g.,

PAGE 17

O'Guinn and Faber 1989), self-gift (e.g., Mick and Demoss 1990), sport activities (e.g., Hill and Robinson 1991; Lehmann 1987), game playing (Holbrook, Chestnut, Oliva, and Greenleaf 1984), music listening (Baumgartner 1992; Bruner 1990), etc. Other studies have examined affective responses across consumption activities (e.g., Derbaix and Pham 1991; Havlena and Holbrook 1986; Havlena, Holbrook, and Lehmann 1989). In essence, most of these studies are descriptive For instance, in a typical study (Richins et al. 1992, study 1), respondents were asked to recollect a purchase that had made them happy, and describe the purchase circumstances and the feelings experienced while making the purchase and during the subsequent few weeks. The descriptions were content-analyzed into frequency counts of the type of purchase mentioned and the type of emotions elicited (e.g., happiness, excitement, anxiety). Derbaix and Pham (1991) used a similar methodology, combined with correspondence analysis, to uncover typical associations between a variety of consumption situations (e.g., "while shopping for groceries" and "after having been to the hairdresser's") and specific affective responses (e.g., "irritation" and "regret"). Because of their emphasis on description, these studies often rely on content analysis of responses to open-ended questions (e.g., Havlena and Holbrook 1986; Gardner and Rook 1988; Mick and Demoss 1990; Rook 1987). Some of these studies also use multivariate analyses (e.g., Batra and Holbrook 1990; Havlena and Holbrook 1986; Havlena et al. 1989). Unlike the typical research on contextual mood effects, most studies on affective responses avoid reducing affect to being as either good or bad, pleasant or unpleasant. A contribution of this stream of research has been to show that affective responses to

PAGE 18

9 consumption may be very rich (e.g., Holak and Havlena 1992) and diverse (e.g., Derbaix and Pham 1991; Richins et al. 1992). A major limitation of this research is that it has been predominantly atheoretical (which is due in part to its exploratory character). Whereas research on the contextual effects of affective states is solidly grounded in social psychology, the theoretical underpinnings of studies on affective responses are unclear. Aside from informing us on the variety and richness of affective responses to consumption, these studies tell us little about (1) how these responses are elicited; (2) how diverse they really are and why; (3) how invariant they are across consumption activities, etc. Unless these questions receive satisfactory answers, these studies' findings amount to little more than a collection of data points in a multivariate space of consumption phenomenon by type of affective responses. Recent theoretical models of affective responses (e.g., Cohen and Areni 1991; Frijda 1986; Lazarus 1991; Roseman 1979) may help better understand how -as opposed to what — affective responses are elicited by consumption phenomena. The processing framework detailed in Chapter 2 integrates several aspects of these models. A second limitation of this stream of research is that by focusing on affective responses per se it captures neither their information value nor their motivational significance. Operant learning principles would suggest that if affective responses are consistently experienced in relation to consumption behaviors, consumers would somehow incorporate these responses in their decision to consume. The conceptual framework detailed in Chapter 2 provides a process account of how affective responses may be used as information in consumers' decisions and how these responses may be

PAGE 19

10 anticipated and incorporated in consumers' decisions. Also, as implied in Hirschman and Holbrook's (1982) concept of hedonic consumption, certain consumption behaviors may be enacted in order to experience desired affective responses (see also Gould 1991). The conceptual framework suggests several hypotheses about the process underlying an affect management motive. Affect as a Mediating Factor Whereas the affective response stream of research focused on affective responses per se this stream of research focuses on how affective responses mediate (and possibly motivate) other variables of interest. Most of these studies have been in the advertising domain (e.g., Aaker, Stayman, and Hagerty 1986; Batra and Ray 1986; Burke and Edell 1989; Edell and Burke 1987; Holbrook and Batra 1987; Olney, Holbrook, and Batra 1991; Stayman and Aaker 1988), where researchers have examined how affective responses to advertisements mediate attitudes toward the ad (Aad) and attitudes toward the brand (Ab). More recently, similar studies have been conducted in other domains. For instance, Allen, Machleit, and Klein (1992) examined how affective responses to blood donation mediate future donation behavior; Holbrook and Gardner (1993) examined how affective responses to music mediate listening time; and Mick and Faure (1993) studied how affective responses to success and failure mediate the likelihood of self-gifts. Most of these studies are correlational and based on a specific paradigm that involves (1) exposure to some stimuli (e.g., advertisements, musical pieces); (2) measurement of affective responses to these stimuli; (3) measurement of dependent

PAGE 20

11 variables (e.g., Ab, listening time); and (4) some assessment of the stimulus' properties (e.g., ratings of the ad's quality by judges). Whereas the "contextual" stream of research emphasizes biases and the "response" stream of research emphasizes description, the "mediation" paradigm focuses on incremental variance explained For example, Edell and Burke (1987, p. 428) summarized one experiment as follows: "We found that positive and negative feelings contributed to explaining variance in Aad and Ab even after accounting for the variance explained by judgments of the ad's characteristics." Similarly, Olney et al. (1991, pp. 440-441) introduced their study as focusing on "the variance in [advertising] viewing time explained by three attitudinal components . by two emotional dimensions (pleasure and arousal), and by the uniqueness of ad content." Because of this axiological focus (i.e., variance explained per degree of freedom), these studies have often examined affective responses along a reduced number of dimensions such as Mehrabian and Russell's (1974) pleasure and arousal factors. The major contribution of the affective mediation stream of research has been to show that affective responses may have a more direct, hence noncontextual influence on various dependent variables of interest in marketing and consumer research. This stream of research also suggests that the influence of affective responses is not redundant with the influence of predictor variables examined in previous research. For example, it has been observed that feelings induced by an ad influence (1) brand attitudes independently of attitudes toward the ads (Burke and Edell 1989; Stayman and Aaker 1988); and (2) attitude toward the ad independently of the ad's content (Edell and Burke 1987; Holbrook and Batra 1987). Allen et al. (1992) also found that emotions experienced in response

PAGE 21

12 to blood donation predicted future blood donation independently of peoples' attitudes toward this act. Therefore, one can infer that the influence of noncontextual affective responses on a variety of judgments and behaviors may be unique. The main limitation of this research, however, is that it rarely seeks to explain the process through which affective responses exert their noncontextual influence. As Stayman and Aaker (1988, p. 371) acknowledge, the goal of this research is "to demonstrate that conditions exist under which feelings have direct effects . rather than to assess the process underlying the effect. For instance, it is not clear in the Allen et al. (1992) study exactly how the emotions experienced during one blood donation occasion came to influence people's subsequent donation behaviors. If these behaviors are preceded by some decision making process, the role of the initial emotion experience in this process needs to be elucidated. The proposed framework offers an attempt at specifying such processes. The scope of this paradigm is also limited to situations in which stimulus exposure, criterion variables, and mediation relate to a single incident For instance, in Holbrook and Gardner (1993), subjects were presented a series of musical pieces (stimulus exposure) which they were asked to rate on pleasure and arousal dimensions (mediation). The amount of time subjects actually listened to each piece (criterion variable) was recorded unobtrusively. In these designs, the mediational role of affective responses to the stimulus can easily be interpreted in terms of approach-avoidance principles. And when the criterion variable relate to an object that differs from the affect-eliciting stimulus— such as when attitude toward a product is expressed as a

PAGE 22

13 function of affective responses to the embedding commercial-simple association principles may suffice. However, the mediating role of affective responses need not be restricted to single incident settings. As the Allen et al. (1992) study suggests, affective responses to one incident may influence behaviors relating to future incidents. How does affect influence decisions in more dynamic settings? The proposed framework examines setting in which consumers need to anticipate future outcomes. Blueprint of a Conceptual Bridge In summary, a growing body of research suggests that affect is important to the study of consumer behavior: (1) affective states have a variety of contextual effects on consumers' judgments; (2) affective responses to marketing stimuli are significant (and possibly unique) noncontextual predictors of criterion variables of interest in marketing and consumer research; and (3) affective responses are a rich component of many consumption experiences. Yet, the literature on consumer decision making has traditionally ignored the role of affective inputs (Bettman 1993), and the different streams of research on affect and consumer behavior have ignored the informational and motivational function of affect in decision making. This dissertation intends to provide a conceptual bridge between the growing literature on affect and the better established literature on consumer decision making. The unifying theme of this program of research is that affect is not only a contextual source of bias or a response to consumption phenomena, it is also as source of information, a decision input. Chapter 2 articulates an integrative framework that

PAGE 23

14 specifies the processes through which consumers use their affective responses in their decisions. Chapter 3 formalizes a research agenda for testing a central component of the framework: the affect recruitment heuristic. Chapters 4 through 7 report five experiments implementing this agenda. And Chapter 8 provides a general discussion of the findings, as well as future directions.

PAGE 24

CHAPTER 2 AFFECT AS INFORMATION IN DECISION MAKING: A PROCESSING FRAMEWORK This chapter introduces a processing framework that describes how consumers may rely on affective responses as a decision making input. The framework is presented in the context of decisions about consumption episodes. This chapter is divided into three sections. The first section of the chapter sets the conceptual foundations of the framework. The second section introduces the affect recruitment heuristic, the cornerstone of the framework. In the third section, the framework's logic is extended to provide a process explanation of consumption decisions geared toward the management of aversive affective states. Conceptual Foundations The foundations of this framework lie in recent cognitive theories of affect. It is postulated that experienced affective responses or feelings are information that consumers may rely on in decisions about consumption episodes. This information originales in the appraisal process underlying the elicitation of these affective responses. The nature of this information is described in terms of key components of the meaning structures typically underlying these affective responses. 15

PAGE 25

16 Feelings as Information Although previous research has stressed the biasing effects of affective states on judgments (see e.g, Gardner 1985 for a review), recent theories emphasize that affect also fulfills adaptive functions (e.g., Lazarus 1991; Plutchik 1980). One of these functions is to provide information about the state of the environment (e.g., Frijda 1986, 1988; Morris 1991). For instance, Schwarz and Clore (1988; Schwarz 1990) propose that to evaluate objects, people often ask themselves "How do I feel about it?" In Schwarz and Clore (1983), judgments of life satisfaction were collected among subjects either in positive or in negative mood. The authors expected that subjects' mood would bias their reported life satisfaction. Moreover, Schwarz and Clore reasoned that if people inspect their feelings because of their informative value, these feelings would not bias judgments when attributed to a cause other than the object of judgment. This prediction was supported. Subjects' mood states biased their judgments unless an external cause for these mood states (e.g., the weather) was made salient. In Schwarz and Clore (1983), as in typical research on contextual affective states (see Chapter 1), the cause of subjects' feelings was independent of the object of judgment. However, the "How-do-I-feel-about-it?" heuristic applies also, and mostly, to situations where the source of people's feelings is actually the object to be judged. What this heuristic means is not so much that mood states are often misattributed and bias judgments, as that people often evaluate objects based on their feelings in response to these objects. For example, while at a cocktail party we might be cheerfully drinking our favorite single malt scotch and discussing movies seen recently. When asked

PAGE 26

17 whether we like the party, we would inspect our feelings and infer from our cheerfulness that we probably do. Here, as probably in many cases, the feelings are correctly attributed. Therefore, a natural way in which consumers may evaluate consumption episodes is by inspecting their feelings in response to them. This postulate is the first building block of the present framework. Postulate 1: Consumers evaluate consumption episodes, in part, by inspecting their affective responses to these episodes. The Origins of the Information Conveyed by Affective Responses Postulate 1 implies that affective responses convey information to the consumer experiencing them. The origin of this information can be traced to the processes underlying the elicitation of these responses (Clore, Schwarz, and Conway 1994). Consistent with cognitive theories of affect (e.g., Arnold 1960; Frijda 1986; Lazarus 1991; Ortony, Clore and Collins 1988; Roseman 1984), it is postulated that affective responses to consumption episodes emanate from personally-relevant meaning structures generated by the appraisal of these episodes. Appraisal is the "evaluation of the significance of what is happening in the person-environment relationship for personal well-being" (Lazarus 1991, p. 87). Appraisal theories emphasize that affective responses to external events are not reactions to the events per se but responses to interpretations of these events.

PAGE 27

18 Postulate 2: Consumers' affective responses to consumption episodes are mediated by their appraisals of these episodes, which generates personally-relevant meaning structures. 1 Because there is a lawful relation between emotions and appraisal-generated meaning structures (Frijda 1988; Lazarus 1991), not only do different meaning structures result in different emotions, but the experience of an emotion informs us about the particular meaning of the eliciting event (Schwarz 1990). Keltner, Ellsworth, and Edwards's (1993) Experiment 4 provides dramatic support for this postulate. Anger or sadness was induced by having subjects model behavioral expressions typical of these emotions. Although they were not exposed to any stimulus information, subjects subsequently showed judgmental biases characteristic of the meaning structure underlying the emotion they were modeling. For instance, compared to subjects modeling anger, subjects modeling sadness were more likely to believe that future negative events in their lives would be caused by situational (an opposed to human) factors. As detailed below, situational agency is an appraisal characteristic of sadness emotions. Postulate 3: The information conveyed to the consumer by the experience of a particular emotion (e.g., sadness or pride) corresponds to the appraisalgenerated meaning structure that typically underlies this emotion. 1 This postulate does not mean that affective responses are always based on controlled processes. Recent models of affective responses (Branscombe 1988; Cohen and Areni 1991; Hoffman 1986; Leventhal 1984) suggest that appraisal of consumption episodes may be performed in a controlled and elaborated mode, or more automatically through the activation of affective schemata.

PAGE 28

19 The Nature of the Information Conveyed by Affective Responses Although it is beyond the scope of this chapter to specify comprehensively the meaning structure underlying each emotion, a good understanding of the information conveyed by affective responses can be reached by specifying several key components of these meaning structures (for more complete treatments, see Abelson 1983; Arnold 1960; Frijda 1986; Lazarus, 1991; Ortony et al. 1988; Roseman 1984; Scherer 1984; Smith and Ellsworth 1985). These core elements may be regarded both as major antecedents of emotional experiences (the traditional view) and as information implied by these emotional experiences (the feelings-as-information view). The mapping between these core element and the main types of affective responses is summarized in Tables 2. 1 and 2.2. Desirability That events such as consumption episodes are appraised as desirable or undesirable is posited by every appraisal theorist. Desirability results from the interaction between the presence/absence of a stimulus and its intrinsic rewarding/aversive property (Arnold 1960; Ortony et al. 1988; Roseman 1984). The presence of a rewarding stimulus in a consumption episode is conducive of joy (e.g., receiving a gift). The absence or loss of a rewarding stimulus is conducive of disappointment or sadness (e.g., not receiving an expected gift, Higgins 1987). The presence of an aversive stimulus may result in distress or disgust (e.g., smelling garbage while dining in a restaurant). The absence or removal of an aversive stimulus is associated with relief (e.g., arriving late at the airport and realizing that the plane has not left yet).

PAGE 29

20 TABLE 2.1 TENTATIVE SPECIFICATION OF THE CORE MEANING STRUCTURES OF THE MAIN TYPES OF PLEASANT AFFECTIVE RESPONSES AFFECTIVE RESPONSE TYPE APPRAISAL Joy (Happiness Elation) Pride Hope Relief Attraction (Liking, Love) Desirability: Rewarding Rewarding Rewarding Aversive Rewarding Type & Presence & Present & Present & Present & Absent & Present of Stimulation or Aversive & Absent Agency

PAGE 30

21 the self. Emotions such as fear (e.g., being scared in a rollercoaster), joy (e.g., having obtained a good deal on a purchase), and sadness (e.g., being disappointed because one's favorite team has lost), are based on the appraisal of an entire event. TABLE 2.2 TENTATIVE SPECIFICATION OF THE CORE MEANING STRUCTURES

PAGE 31

22 It is because they perceived the consumption episode as "under control" that consumers voluntarily expose themselves to threatening or potentially harmful stimulation, such as riding roller-coasters or watching frightening movies (Mandler 1984a). A low sense of control is at the core of fear-like emotions (e.g., driving on a slippery highway), and sadness-like emotions (e.g., having one's house destroyed by a fire). Legitimacy/deservingness Appraisal of legitimacy or deservingness refers to whether one deserves or does not deserve a positive or a negative outcome in a consumption episode (Roseman et al. 1990). The experience of pride is based on the appraisal that one deserves a rewarding stimulation (high legitimacy) and receives it (e.g., an adolescent buying himself a nice stereo because he has graduated with honors from high school). When experiencing anger people have the sense of "being right" (high legitimacy) but do not receive the deserved reward (e.g., customers getting angry with a salesperson for not providing the expected level of service). In contrast, guilt and shame are experienced as result of low deservingness or high blameworthiness (e.g., drinking too much at a party). In other words, legitimacy/deservingness appraisal is an assessment the strength of one's position in a consumption episode with respect to moral or social standards. Outcome uncertainty Consumption episodes differ in the predictability of their outcome (e.g., Frijda 1986; Lazarus 1991; Roseman 1984; Smith and Ellsworth 1985). High outcome uncertainty is characteristic of emotions such as hope (e.g., awaiting the guests' reactions to a new food recipe), and fear-like emotions, such as anxiety (e.g., a wife dreading that her husband might not like her new dress). When the outcome of an

PAGE 32

23 episode becomes certain, other emotions take place, such as embarrassment or disappointment (e.g., realizing that the guests don't like the new recipe) and joy or relief (e.g., the husband likes the new dress.) In summary, the meaning structure of the main types of emotions (their informational content) appears to depend heavily on five kinds of appraisal: desirability, agency, control, legitimacy, and outcome uncertainty. The intensity of these emotions (e.g., apprehension versus fright, blue feelings versus despair) is a function of the significance of the consumption episode to the consumer's goal (Frijda et al. 1991). This significance depends in turn on (1) the personal relevance of (or involvement with) salient features of the episodes (e.g., a Chicago Bulls fan attending one of their games), and (2) the potential consequences of the episode with respect to these salient features (e.g., Game 7 of the playoff finals). The Affect Recruitment Heuristic Overview of the Process Let us return to our first opening example: Maria is trying to decide how to spend her afternoon. She has several options: playing tennis with Mark, clothes shopping in the mall with Diane, or reading a novel in the park. In the typical consumer decision making paradigm, information about the alternatives is either present in the consumer's environment or has been memorized previously In contrast, the information required for Maria's decision may not be available as such in her current environment or her memory. She does not "know" yet what wiU happen. For instance, how pleasant

PAGE 33

24 will her tennis game actually be? Because of the uncertainty surrounding the actual content of each episode, she may have to anticipate how the alternative episodes will unfold. The main theoretical account offered for such a process is Fishbein and Azjen's (1975) attitude formation model. Maria may access salient beliefs about each episode and integrate their evaluative implications into an overall evaluation of (or attitude toward) each episode. For example, she may access propositional beliefs such as "It will be hot on the tennis court," "I'll get a chance to wear my new Nike shoes" and "I haven't exercised much lately," and reach a moderately favorable evaluation of the episode. Alternatively, instead of computing a new evaluation, she may retrieve a previously formed evaluation (cf. Lingle and Ostrom 1979). For instance, when thinking about the tennis game episode, she may retrieve a summary evaluation, such as "It's nice to play tennis on Sundays," that she computed earlier. There is an alternative way in which this consumer may evaluate each episode and make a decision. Instead of accessing propositional beliefs about the episode, she may access a concrete (analog-like) representation of the episode, possibly a representation that includes herself. That is, she may literally "picture" the episode to herself. While holding this "picture", she may experience anticipatory affective responses. Then she may use these anticipatory affective responses as inputs to her decision. This process is called the affect recruitment heuristic Previous research has suggested that people may use their affective response to current objects as a source of information about these objects (e.g., Keltner et al. 1993;

PAGE 34

25 Gorn et al. 1993; Martin et al. 1992; Schwarz and Clore 1983). It is proposed that consumers may similarly use affective responses to hypothetical construals of consumption episodes (i.e., anticipatory affective responses) as a source of information about these episodes. Clore (1992) suggests a similar process in person impression formation. He proposes that to evaluate people about whom we have not previously formed an opinion, we may hold the concept of this person in consciousness, and assess if any feelings can be detected. The affect-recruitment heuristic, an extension of Postulate 1, may be formalized as follows: Thesis : Consumers' decisions about future (or hypothetical) consumption episodes are based in part on their anticipatory affective responses to construals of these episodes. Constructing Consumption Episodes An important feature of decisions about consumption episodes is that the necessary information about the episodes may not be available in the consumers' present environment or in their memory. As a result, such decisions may necessitate some form of prediction about the episode's content. For instance, to decide whether to bring his date to a given restaurant, Frank may try to predict how the episode will unfold (e.g., whether the setting would allow them to carry an intimate conversation or whether she would find anything she likes on the menu). Although, previous research has assumed that the "expected" content of the episode would consist of a series of propositional beliefs, it is proposed that judgments and decisions about consumption episodes may also involve more concrete analog representations of the episode's content.

PAGE 35

26 Various theoretical frameworks suggest that to predict how a consumption episode will unfold, a consumer may mentally simulate this episode in his or her mind. This idea can be traced back to Lewin's (1951) field theory, where behaviors are described as driven by forces in the psychological field operating at different levels of reality. Lewin suggests that in planning, the person operates at an intermediate level of realityirreality. At this level, "locomotion," the modification of psychological facts (e.g., imagining the atmospheres of several restaurants), is easier than in actual reality (e.g., physically checking out the restaurants). More recently, Kahneman and Tversky (1982) advanced the notion of a "simulation heuristic. They suggest that predictive judgments about the likelihood of events are often made by mentally "running" different scenarios of these events and relying on the ease with which the different scenarios are simulated. Even more recently, Taylor and Scheiner (1989) have suggested that the mental simulation of past, future, and hypothetical events serves both problem solving and emotional regulation functions. They advance a definition of mental simulation consistent with the mental construction of consumption episodes postulated by the present framework: [Simulation means] the cognitive construction of hypothetical scenarios or the reconstruction of real scenarios. These may include (1) rehearsal of likely future events [...J (2) reconstruction of past events [...] (3) fantasies [...] and (4) mixtures of real and hypothetical events, such as the reconstruction of a past event [...] with a new ending (Taylor and Schneider 1989, p. 175). That is, consumers may construct consumption episodes by retrieving past episodes (e.g., remembering how exciting it was to attend a ball game) or imagining hypothetical scenarios (e.g., imagining how embarrassed they would be if their date hates the restaurant they have chosen). It is further proposed that the representational format

PAGE 36

27 of these construals is often concrete and analogical rather than abstract and propositional (Epstein 1990, 1994; Taylor and Schneider 1989). Proposition 1: Construals of consumption episodes involved in the affect recruitment heuristic are often episodic and nonpropositional. Although, the idea that consumers may actively construct concrete mental representations of consumption episodes to help their decisions is consistent with several theoretical frameworks, empirical evidence of a mental simulation process has been restricted to counter factual thinking: the generation of alternatives to events that have already happened or are occurring (Gleicher et al. 1990; Rose 1994; Well and Gavanski 1989). Some empirical evidence that people processing information experientially may be more responsive to concrete than to abstract representations was recently reported by Kirkpatrick and Epstein (1992). Anticipatory Affective Responses and the Appraisal of Constructed Episodes Once a representation of a consumption episode has been accessed, this representation is appraised through the same processes as the ones involved in the appraisal of actual episodes (cf. Mandler 1984b). This appraisal process may instantiate an affective response. To stress the fact that the responses are actually instantiated or experienced at the time of the decision, they are called anticipatory affective response.

PAGE 37

28 The term "anticipated" affective responses is reserved for more abstract predictions of future affective experiences (e.g., "It will be fun"; "I would feel guilty"). 2 Several lines of evidence suggest that mental construals of consumption episodes may elicit anticipatory affective responses. First, a common affect manipulation consists of having subjects bring into consciousness affectively charged episodes, either by recalling them or by imagining them. The effectiveness of this manipulation indicates that affective responses can be based on mental construals rather than actual events (Taylor and Schneider 1989). Second, Roseman (1991) showed that, by reading short stories whose contents were systematically manipulated along key appraisal dimensions, subjects were able to predict relatively accurately the affective experiences of the characters in the stories. This finding suggests that appraisal processes are relatively reliable and can be applied with flexibility to hypothetical consumption episodes. Finally, it has been shown that while reading fictitious stories, readers spontaneously represent the characters' emotional states (Gernbacher, Goldsmith, and Robertson 1992). This implies that not only consumers can predict how a hypothetical (constructed) consumption episode will make them feel, but they represent these anticipatory affective responses relatively spontaneously. Proposition 2: Mental construals of consumption episodes may instantiate anticipatory affective responses. These anticipatory affective responses are generated by appraisal processes similar to those influencing affective responses to actual consumption episodes. 2 The distinction between "anticipatory" and "anticipated" affective responses parallels distinctions between "experiential" and "noetic" representations (Stepper and Strack 1993) and between "knowledge by acquaintance" and "knowledge about" (James 1890/1950).

PAGE 38

29 Reiving on Anticipatory Affective Responses It is proposed that consumers attach unique information value to these anticipatory affective responses, and as a result, incorporate these responses into their decisions about consumption episodes. In other words, consumers do not only consider the evaluative content of their construals (e.g., imagining buying a nice outfit at the Gap store vs. imagining not being able to find anything), they also rely on how these construals make them feel (e.g., joyful vs. frustrated). This proposition is consistent with recent discussions in various literatures. In the social cognition literature, Strack (1992) has recently suggested that social judgments can be reached either through an "informational" route or through an "experiential" route. In the latter route, judgments are based on interoand proprioceptive cues, i.e. feelings. This route includes both emotional feelings (e.g., to make evaluative judgments) and nonemotional feelings (e.g., inferring that a book in uninteresting from one's fatigue while reading it). Clore (1992) similarly suggested that feelings, emotional (e.g., cheerfulness) and nonemotional (e.g., fatigue), guide a variety of judgments. In the personality literature, Epstein (1990, 1994) postulates the existence of two processing systems, called "rational" and "experiential." Consistent with the present framework, he suggests that when a person is exposed to an emotionally significant event, his/her experiential system searches for similar events in memory and their associated affective traces. The recollected feelings are activated and guide subsequent processing and reactions. These feelings may be subtle, in which case they are referred

PAGE 39

30 to as "vibes." Unfortunately, Epstein offers only remote empirical evidence of the alleged experiential processes (e.g., Epstein 1990, 1994; Epstein etal. 1992; Kirkpatrick and Epstein 1992). Some evidence consistent with the proposition that anticipatory affective responses have unique information value can be found in the behavioral decision literature. Kahneman and Snell (1990) asked subjects to evaluate sequences of hedonic outcomes (e.g., receiving a series of increasingly painful injections). They observed that to make their evaluations, subjects did not seem to aggregate the separate hedonic values of these outcomes (e.g., adding the unpleasantness of each injection). Instead, they seemed to consult an instantaneous representation of the sequence, which Kahneman and Snell call "news," and rely on their emotional responses to the news. It appears easier to imagine and evaluate one's emotional response to receiving news than to imagine responses to a prolonged sequence of events. The assessed response to news can therefore be used as a heuristic for the task of evaluating complex outcomes (Kahneman and Snell 1990, p. 301). According to Kahneman and Snell (1990) "news utility" serves as a proxy for "experience utility" (i.e., prediction of one's future hedonic experience). Stated differently, anticipatory affective responses are use to predict future affective experiences (i.e. anticipated affective response). Although evolving from a very different theoretical perspective, these authors' ideas are strikingly similar to those proposed in the present framework. Proposition 3: Anticipatory affective responses may be directly incorporated into decisions about consumption episodes because of their unique information value.

PAGE 40

31 The Determinants of Affect Recruitment The first three propositions stress unique aspects of the affect recruitment heuristic. They do not specify the conditions under which this heuristic will be used, that is, the determinants of affect recruitment. To identify these determinants, the present framework relies on the following logic: If anticipatory affective responses are information, their use in decision and judgments should respect principles similar to those guiding the use of other informational inputs. These principles can be specified using Feldman and Lynch (1988) terminology. Their framework posits that the use of informational inputs in judgments is a positive function of their relative accessibility and diagnosticity, compared to other inputs (Feldman and Lynch 1988; Lynch, Marmorstein, and Weigold 1988). The determinants of the diagnosticity of anticipatory affective responses are examined below. The determinants of their accessibility are discussed next. Active Goal and Diagnosticity The perceived information value of an affective response (anticipatory or not) is defined as its diagnosticity (Feldman and Lynch 1988). 3 Substantial evidence shows that people are more likely to incorporate their affective states in their judgments when they perceive these states to be diagnostic for the judgments (see Schwarz and Clore 1983, Strack 1992, and especially Keltner, Locke, and Audrain Strack (1992) articulates a similar concept, called "representativeness," and distinguishes between content representativeness and process representativeness. Content representativeness is the degree to which the information conveyed by the feeling experience is similar to the target of the judgment. Process representativeness is the degree to which the event that produced the feelings is relevant for the judgment.

PAGE 41

32 1993). It would be logical that, in judgments of hypothetical consumption episodes, the influence of anticipatory affective responses should also be moderated by their diagnosticity. Three factors may affect the diagnosticity of anticipatory affective responses. First, consumer researchers have historically identified two classes of consumption motives. Alderson (1957) contrasted consumption behaviors undertaken for their own sake (e.g., reading a novel, watching a movie) with behaviors undertaken to achieve some other goal (e.g., buying groceries to prepare diner, ironing a shirt before going to a job interview). Whereas the former, called congenial behaviors, are intrinsically rewarding, the latter, called instrumental behaviors, are seldom rewarding in themselves (Alderson 1957). Holbrook and Hirschman (1982; Hirschman and Holbrook 1982) suggest that, when consumers' motives are congenial or experiential, emotional and aesthetic criteria are more important determinants of their behavior than when their motives are instrumental. Translated into the present framework, their proposition implies that anticipatory affective responses should be more diagnostic when the consumption episode is evaluated for congenial reasons than when it is evaluated for instrumental reasons. For example, Lehmann (1987) suggests that some consumers lift weights for instrumental reasons (e.g., weight reduction, rehabilitating an injury), while others pursue this activity for more congenial reasons (e.g., "it feels good"). Presumably, when deciding whether to lift weights on a particular occasion, anticipatory affective responses should be less diagnostic among the former group of consumers than among the latter.

PAGE 42

33 Indirect evidence for this proposition is provided by Millar and Tesser (1986, 1992). Consistent with Alderson (1957), the authors draw a distinction between "instrumental" behaviors, which are "intended to accomplish a goal which is independent of the attitude object," and "consummatory" behaviors, which are "engaged for their own sake" (Millar and Tesser 1992, p. 282). They further propose that attitudes have a cognitive component (the attributes and beliefs about the attitude object) and an affective component (the emotions and feelings associated with attitude object). In Millar and Tesser (1986), subjects were assigned either an instrumental goal or a consummatory goal for playing with puzzles. After expressing either cognitively-based attitudes or affectively-based attitudes toward these puzzles, subjects were given the opportunity to play with them. Among subjects with a consummatory goal, the amount of time spent playing with the puzzles correlated more strongly with affectively-based attitudes than with cognitively-based attitudes. The pattern of correlations was reversed among subjects with an instrumental goal. In the present framework, these results suggest that affective responses to the puzzles were more predictive of playing behavior when it was driven by consummatory motives than when it was driven by instrumental motives. In contrast, the predictive ability of "reasons why" was stronger when subjects had instrumental motives than when they had consummatory motives. This is consistent with the proposition that anticipatory affective responses to consumption episodes should be more diagnostic when consumers' motives are congenial or experiential than when they are instrumental.

PAGE 43

34 Second, certain events are associated with socially shared beliefs about how one should feel during these events. Hochschild (1979) refers to these beliefs as "feeling rules." For example, on a New Year's Eve consumers believe that they should feel happy and cheerful. On Thanksgiving Day, which is dominated by cultural themes of abundance and self-indulgence (Wallendorf and Arnould 1991), people should not feel guilty. Feelings rules may also prescribe negative affective states. For instance, the loss of a loved one is culturally associated with a period of mourning in which people are expected to feel sorrowful. Presumably, when consumption episodes are judged in the contexts of events strongly associated with feeling rules, anticipatory affective responses should be perceived to be particularly diagnostic. For example, how cheerful a consumer expects to be at a party should be more diagnostic for judging it if it celebrates New Year's Eve than if it is held for no special reason. Third, consumers' concern about how consumption episodes will make them feel should be higher when their preexisting affective state is aversive than when it is not. Abundant research in the literatures on helping (e.g., Cialdini, Darby, and Vincent 1973; Manucia, Baumann, and Cialdini 1984), mood management (e.g., Zillmann 1988; Bryant and Zillmann 1984), and coping (e.g., Lazarus and Folkman 1984) suggests that individuals experiencing aversive affective states are particularly attentive to aspects of their environment that might help them alleviate those states. Therefore, anticipations about how consumption episodes would make them feel should be more diagnostic among consumers in aversive affective states than among consumers in neutral or pleasant affective states.

PAGE 44

35 The results of a recent series of experiments by Forgas (1991) support this proposition. Subjects were first induced to be in a positive, neutral, or sad mood state. In a purportedly unrelated experiment, they were given the opportunity to select a partner to work as a team. The list of potential partners contained individuals described as socio-emotionally rewarding (e.g., friendly, likeable) but not competent (e.g., not intelligent, poor exam performance), and individuals described as competent (e.g., intelligent, good exam performance) but not socio-emotionally rewarding (e.g., unfriendly). Whereas happy and neutral (control) subjects selected their partners primarily on the basis of their competence, sad subjects selected their partners on the basis of the socio-emotional reward that their company may provide. Interestingly, the process by which subjects selected their partners was also different across mood conditions. Whereas happy subjects compared potential partners feature by feature (across individuals), sad subjects compared them individual by individual (across features). Therefore, sad subjects relied more than happy subjects on overall impressions of the described individuals to select their partners. These results are very consistent with the proposition that: (1) consumers experiencing aversive affective states should attribute more importance than consumers in neutral or positive states to the affective responses that consumption episodes might elicit, and (2) the process by which they might anticipate these affective responses is by constructing overall representations of each episode.

PAGE 45

36 Proposition 4: The relative diagnosticity of anticipatory affective responses to consumption episodes (and their influence on judgments about these episodes) is higher under the following conditions: (a) The underlying consumption motive is congenial; (b) The social context is strongly associated with feeling rules; (c) The preexisting affective state is aversive; Accessibility Some research suggests that the influence of mood states on judgments is moderated by their relative accessibility. For instance, Isen and Shalker (1982) found that subjects induced to be in a positive mood evaluated slides more positively than subjects in a control condition. However, this effect was somewhat stronger for ambiguous slides than for clearly pleasant or unpleasant slides (see also Miniard et al. 1992). It is likely that the higher the relative accessibility of anticipatory affective responses to consumption episodes, the stronger their influence on judgments about these episodes. Four determinants of the intensity of anticipatory affective responses can be identified. First, strong anticipatory affective responses should be more accessible and exert stronger influence on consumers' judgments of consumption episodes than weak anticipatory affective responses. Suppose that consumers are trying to decide whether or not to attend a basketball game. It is probable that their anticipatory emotions will be stronger and more salient if it is "Game 7" of the playoffs' final than if it is a regular season game. Therefore, one would predict that anticipatory emotion should play a greater role in consumers' decision in the former than in the later case.

PAGE 46

37 To the extent that vivid representations of consumption episodes should elicit stronger affective responses (e.g., Lang 1984), a second determinant of the accessibility of anticipatory affective responses might be the consumer's imagery abilities. Consumers who have strong imagery abilities (high "imagers") should have easier access to the affective responses that the consumption episode might elicit than consumers with weaker imagery abilities (low "imagers"). An important determinant of the salience of affective responses is the level of selffocused attention or self-awareness (Scheier and Carver 1977). People are more likely to regulate their behavior in accordance with their affective responses to the situational context when self-awareness is high than when it is low (Carver, Blaney, and Scheier 1979a, 1979b; Carver and Scheier 1990; Scheier and Carver 1977). Therefore, anticipatory affective responses to consumption episodes should be more accessible when self-awareness is high. Finally, several factors may increase (decrease) the relative accessibility of anticipatory affective responses by decreasing (increasing) the accessibility of alternative inputs. In particular, certain episodes may be evaluatively complex, making it difficult to reach an overall attitudinal judgment (Allen et al. 1992). In contrast, affective responses that simply need to be "read out" may be readily accessible (Strack 1990). Therefore, the relative accessibility of anticipatory affective responses may be higher for evaluatively complex episodes than for simpler episodes (Schwarz 1990; Strack 1992). This prediction is consistent with Kahneman and Snell's (1991) observation that people seem to rely on anticipatory feelings to evaluate complex sequences of outcomes.

PAGE 47

38 Proposition 5: The relative accessibility of anticipatory affective responses to consumption episodes and their influence on judgments is higher in the following conditions: (a) The anticipatory affective response is intense; (b) The consumer has high imagery abilities; (c) The consumer is highly self-aware; (d) The episode is evaluatively complex. A Graphical Summary A graphical summary of the framework in presented in Figure 1. Conceptually, affect-recruitment involves three sets of elements: subprocesses, contents, and determinants. Three types of subprocesses are hypothesized: (1) the construction of a mental representation of the episode to be judged; (2) an assessment of the significance of this representation for personal well being, that is, its appraisal (Lazarus 1991); and (3) the judgment integration. Three types of content are fed into or generated by these subprocesses: (1) a mental representation (or construal) of the consumption episode, (2) an anticipatory affective response to this construal, and (3) alternative inputs for making the judgment, such as prior attitudes or salient beliefs about the episode. The determinants of the heuristic are specified in terms of accessibility and diagnosticity principles. The higher the relative accessibility of anticipatory affective responses (compared to other inputs), the more likely it is that consumers will rely on affect recruitment. Factors such as the consumer's imagery ability and the episode's evaluative complexity are hypothesized to influence affect recruitment by their effects on the accessibility of anticipatory affective responses. Similarly, the higher the relative diagnosticity of anticipatory affective responses, the more likely it is that consumers will

PAGE 48

39 rely on affect recruitment. Initiation of the heuristic is hypothesized to be contingent upon the consumer's active goal. By altering this goal, hence the relative diagnosticity of anticipatory affective responses, factors such as the consumers' motives and the salience of feeling rules may increase or decrease the likelihood that affect recruitment will be used. As depicted in Figure 1, consumers are assumed to have substantial control over the heuristic. Figure 1 suggests that they actively try to recruit anticipatory affective responses when these have high information value. It is also possible that the process is more "passive" than suggested in Figure 1. Consumers may weight the relative diagnosticity of anticipatory affective responses only at the judgment integration stage. That is, consumers may assess the relevance of anticipatory affective responses only after these are accessed. Appraising Consumption Episodes for Affect Management Recently, researchers have become interested in how people manage their affective states, especially when these states were aversive (e.g., Morris and Reilly 1987; Zillmann 1988). It has been suggested that consumption may play an important role in affect management (Gardner and Scott 1990; Kacen 1994). However, the process by which consumers select among consumption alternatives to manage their affective states needs clarification. The proposed framework may also provide a process explanation of consumption decisions involved in affect management.

PAGE 49

40 CO LU CO CO LU o o DC CL CO D CO CO < Q LU CJ LU CO Iz LU hz o o o

PAGE 50

41 The Motivational Significance of Affective States The present framework posits that consumers' preexisting affective states may influence their processing goals when constructing and appraising consumption episodes. The processing goal underlying this process is the attainment of a desired affective state (Gould 1991), which for an aversive affective experience is the alleviation of this experience. Consumption episodes are constructed and appraised with respect to this goal. This process results in anticipatory affective responses that might be experienced as "this would make me feel better (or worse )." Presumably, the consumption episode will be evaluated favorably to the extent that the consumer anticipates feeling better in response to it. Prior research on mood management has mainly examined the valence of preexisting affective states, assuming that negative states would increase the attractiveness of rewarding or positive stimulation. For instance, Gardner and Scott (1990) suggest that when negative affective states are experienced consumers are more likely to consume "feel-good" products (e.g., cookies). Similarly, Zillmann (1988, p. 1952) proposes that "persons in acutely aversive states select hedonically positive stimuli over hedonically negative stimuli." Therefore, Zillmann (1988) hypothesized that television viewers experiencing negative affect should select hedonically positive television programs, such as TV comedies. However, attempts to validate this intuitive proposition have produced inconsistent results (e.g., Helregel and Weaver 1989; Zillmann, Hezel, and Medoff 1980), possibly because these studies understate the motivational significance of affective states by considering only their valence.

PAGE 51

42 As stated earlier in the paper, affective states may signal more than the pleasantness or unpleasantness of a situation. Equally-valenced affective states, such as anger and fear or pride and hope, provide different information about the state of the consumer's environment. To consumers under different aversive states (e.g., guilt vs. sadness), the phrase "feeling better" should have different meanings. As Table 2.2 indicates, each main negative emotion is underlain by a unique pattern of appraisal information. It is proposed that in the process of managing negative affective states, consumer focus on the aspects of consumption that may correct critical appraisal elements. This proposition is called the part regulation principle Proposition 6: When experiencing an aversive affective state, consumers will appraise consumption episodes by focusing on the appraisal dimensions that best define their aversive experience. Implications of the Part-Regulation Principle The part regulation principle makes more specific predictions than Zillmann's (1988) proposition that people in aversive states always seek hedonically rewarding stimulation. The predictions of the principle will be derived for four types of negative affective states. Sadness The distinctive feature of the meaning structure underlying sadness-like emotions is the loss or absence of a reward (Lazarus 1991; Ortony et al. 1988; Roseman 1984). Thus, the experience of sadness communicates the need for alternative rewarding stimulation (Schwarz 1990). When appraising consumption episodes, sad consumers should focus on the rewarding stimulation that they may provide. They should expect

PAGE 52

43 to "feel better" in response to episodes that contain positive stimulation (e.g., self-gifts, funny movies) than in response to episodes that do not (e.g., shopping for groceries). Proposition 6a: Consumers appraising consumption episodes when experiencing sadness should focus on the rewarding stimulation that this episode may provide. Some indirect evidence supports this proposition. First, Mick and DeMoss's data (1990, 1992) suggest that one of consumers' main motives for self-gifts is to cheer themselves up when feeling down. Also, in Forgas's (1991) series of experiments described earlier, it was observed that sad subjects were more likely than happy and neutral subjects to select their teammates because of their socio-emotionally rewarding qualities. Anger The experience of anger is characterized by the presence of aversive stimulation, of someone (or something) to blame (other agency), and high legitimacy (Lazarus 1991; Roseman 1984). For instance, a consumer would feel angry if an airline company's employee (other agency) is rude to her (aversive stimulation), especially because rudeness is unexpected from a customer-oriented service provider (high legitimacy). Because the absence of reward is not a core element in anger, rewarding stimulation should not be an effective means of alleviating this state. For instance, in Zillmann et al. (1981) subjects were made angry by an insulting confederate and were later given a chance to retaliate. It was hypothesized that angry subjects exposed to pleasant movie segments would retaliate less severely than angry subjects not exposed to this positive stimulation. Contrary to this prediction, showing pleasant movies did not significantly reduce the level of retaliation of angry subjects. Therefore, as suggested

PAGE 53

44 by an analysis of the meaning structure underlying anger, rewarding stimulation per se does not seem to alleviate anger. A more natural way of alleviating anger would be to get rid of the aversive stimulation. Consumption episodes might achieve this purpose to the extent that they occupy consumer's consciousness, that is, by helping mentally get rid of the stimulation. Therefore, when appraising consumption episodes, consumers in a state of anger may be more likely to focus on the distracting potential of the episodes. In Zillmann et al. (1980), subject who had been frustrated by a poor test performance and by an experimenter's disparaging remarks were invited to watch television. Three programs were available: a situation comedy, an action drama, and a game show. Although the researchers predicted that frustrated subjects would be more likely to select the situation comedy (because of its positive valence), frustrated subjects actually selected the game show. To the extent that game show may have been be more absorbing and distracting, Zillmann et al.'s unexpected finding would be consistent with the following proposition: Proposition 6b: Consumers appraising consumption episodes when experiencing anger-like emotions should focus on the distracting potential of these episodes. In Marlatt, Kosturn, and Lang (1975), subjects were provoked by a confederate, then given or not given the opportunity to retaliate. The dependent variable was the amount of alcohol consumed in a seemingly unrelated wine test. It was found that subjects who were not given the chance to retaliate drank more wine than subjects who had the opportunity to retaliate. According to the present framework, subjects who were given the opportunity to retaliate "got rid" of the aversive stimulation by "getting even"

PAGE 54

45 with the confederate. In contrast, subjects who were not allowed to retaliate had to get rid of the offense through other means. Because of its distracting (Hull et al. 1983) and tension-reducing effects (Marlatt et al. 1975), drinking alcohol might have been used to get rid of the aversive state induced by the confederate. Similarly, Kacen (1994) found that to alleviate bad moods (often anger-like feelings) consumers often rely on escape activities such as reading magazines or going to the movies. Fear and Anxiety The meaning structure underlying fear-like emotions such as anxiety is defined by high uncertainty over an outcome and low control over a situation (Izard 1977; Roseman 1984). Alleviating such states thus requires regaining control and a sense of mastery. Therefore, when appraising consumption episodes under conditions of fear or anxiety, consumers seeking to alleviate these states should focus on aspects of the episodes that would help them regain a sense of mastery. Proposition 6c: Consumers appraising consumption episodes when experiencing fear-like affective states should focus on the potential of these episodes to help them regain a sense of mastery. In Friedland, Keinan, and Regev (1992), half the subjects were made anxious by the threat of painful electric shocks. Subjects were then given a choice between pairs of gambling options. The gambles were designed in such a way that one of the options would reinforce subjects' illusion of control over the outcome of the gamble. Subjects' locus of control was also measured. It was observed that compared to nonanxious subjects, anxious subjects with an external locus of control were more likely to prefer the option giving them illusory control over the outcome of the gamble.

PAGE 55

46 Mandler's (1984a) proposition that the consumption of certain threatening stimulations, such as horror movies or riding roller-coasters, might actually provide the consumer with a sense of mastery yields an interesting prediction. Anxious consumers might actually be more likely than nonanxious to watch scary movies, but only to the extent that they believe that they can master this experience. Guilt and Shame The core meaning structure underlying guilt and shame is a sense of low self-worth (self-agency, low legitimacy/deservingness, Roseman 1984; Ortony et al. 1988). To alleviate such a state, one can either reaffirm one's self-worth or reduce one's self-consciousness. When appraising consumption episodes under guilt or shame, the consumer should focus on aspects of those episodes that may enhance selfworth, restore a sense of legitimacy, or reduce self-consciousness. Proposition 6d: Consumers appraising consumption episodes when experiencing guilt or shame should focus on the potential of these episodes to enhance self-worth, restore a sense of legitimacy, or reduce selfconsciousness. Cunningham, Steinberg, and Grev (1980, experiment 2) provide evidence that restoring one's sense of legitimacy is of particular concern when experiencing guilt. To induce guilt feelings, subjects were led to believe that they had broken the experimenter's delicate camera. Subjects were later asked to contribute to a charity. In one condition, the rewarding aspect of contribution was emphasized ("to keep the children smiling"), whereas in the other its moral aspect was stressed ("You owe it to the children"). Compared to subjects not induced to feel guilty, guilty subjects were more likely to contribute, but only when the request stressed the moral (i.e. legitimacy) aspect of contributing. The fact that this effect vanished if subjects were in a good mood before

PAGE 56

47 the camera incident suggests that the aversive aspect of the guilt feelings motivated their contributions. In summary, the present framework takes as its central premise that affective responses have information value. Building on three postulates inspired by recent cognitive theories of emotions, the framework introduces the affect recruitment heuristic as a pervasive process through which consumers may reach decisions about consumption episodes. Three propositions define the key characteristics of this process. Other propositions identify potential determinants deduced from accessibility and diagnosticity principles. Finally, it is suggested that the framework may be extended to account for consumption decisions aimed at managing affective states.

PAGE 57

CHAPTER 3 INTENDED CONTRIBUTION The Framework in "The Big Picture" As stated in Chapter 1, the present framework intends to provide a conceptual bridge between the consumer decision making literature and the otherwise disparate literature on affect and consumer behavior. The framework stresses the information value of affective responses and describes the processes through which they may used in decisions about consumption episodes. This chapter puts this framework in perspective and clarifies its links to existing frameworks. The present framework owes much to the work of Schwarz and Clore (e.g., Schwarz and Clore 1983, 1988; Schwarz 1990; Clore 1992). The central idea that affect is information emanates from their impressive program of research. However, several differences between their work and the present framework need to be recognized. The first difference is a difference of focus. The "feeling-as-information" research program, which originally aimed at explaining the congruent influence of mood on judgments, has focused on the information value of affective states (e.g., Schwarz and Clore 1988). Although it is implicit in Schwarz and Clore's research that affective responses to the object of judgment also have information value, this issue deserves more explicit 48

PAGE 58

49 examination. The information value of anticipatory affective responses to consumption episodes is the focus of the present framework. The second major difference is that the present framework specifies the conditions under which anticipatory affective responses are likely to be used in judgments and decisions; Schwarz and Clore do not. 1 A third difference is that prior research has focused on the influence of affect on the evaluation of existing objects (e.g., Keltner, Locke, and Audrain 1993; Martin, Harlow, and Strack 1992; Schwarz and Clore 1983; see also Gorn et al. 1993), while the present framework extends the underlying logic to decisions about future consumption experiences. Indeed, the affect recruitment heuristic is here viewed as an integral part of consumer decision making processes. Epstein (1990, 1994) has proposed a theory that distinguishes between two processing systems: the rational system and the experiential system. Many of the features of Epstein's experiential processing system (e.g., concrete representations, reliance on feelings) are consistent with the notion of affect recruitment. One difference between the two frameworks is that Epstein suggests that the experiential system operates mostly outside awareness. No such assumption is made in the present framework. Although affect recruitment may not be easily verbalized, it need not be unconscious. The major difference between the two frameworks, however, is philosophical. By contrasting the "rational" and the "experiential" systems, Epstein essentially associates the latter with irrational processing. Most of the evidence he presents to support the Except, of course, when they show that mood does not influence judgment when attributed to a cause unrelated to the judgment object.

PAGE 59

50 notion of experiential processing consists of observed departures from normative judgments (e.g., Denes-Raj and Epstein 1994; Epstein et al. 1992; Kirkpatrick and Epstein 1992). Epstein (1994) even cites superstitious thinking and irrational fear as evidence of a separate experiential system. In the present framework, affect recruitment is not considered "irrational." Although affect recruitment is presented as a "different route" to judgment and decision making (a position consistent with Strack 1992 and Clore 1992), no statement is made regarding the normative value of this route. In fact, it is argued that the use of anticipatory affective responses follows similar information processing principles as the use of inputs such as beliefs or prior attitudes. Although the principles (e.g., accessibility and diagnosticity) governing the use of anticipatory affective responses may be similar to those governing the use of attitudinal inputs, it argued that the process of affect recruitment is distinct from the process described by typical expectancy-value models of attitude formation (e.g., Fishbein and Ajzen 1975). Affect recruitment is distinct in three respects. First, it is proposed (Proposition 1) that in affect recruitment, consumers' representations of consumption episodes are more concrete, analog-like (and less propositional) than described by expectancy-value models. Second, it is proposed (Proposition 2) that in affect recruitment, anticipatory affective responses are indeed instantiated (i.e., feelings are experienced); they are not just predictions (or beliefs) about future affective experiences. Third, it is proposed (Proposition 3) that consumers attach unique information value to these anticipatory responses when making decisions about consumption episodes. That

PAGE 60

51 is, these responses influence consumers' decisions beyond and above the evaluative content of the consumption episode. Interestingly, a model historically related to Fishbein's model— that of Triandis (1980)— is more congenial with the present framework. Like Fishbein, Triandis posits that behavior is a function of intention, and intention is a function of (a) social norms and (b) an expectancy value assessment of the behavior's consequences. The major difference between Fishbein's and Triandis's models is that in the latter, intention is not the sole determinant of behavior. Triandis proposes that behavior is also determined by the person's habits and that this influence operates outside selfinstruction (i.e., is not mediated by intention). A second difference between the two models, more relevant to our discussion, lies in Triandis's treatment of affect. Whereas in Fishbein's model, affect (treated as equivalent to attitude) is the product of an expectancy-valuation of the attitude object, in Triandis's model, affect and expectancy assessment are independent antecedents of intention (and behavior). 2 Triandis defines affect toward the behavior as "the direct emotional response to the thought of the behavior" (Triandis 1980, p. 218) or as "the particular configuration of emotions [that] becomes activated at the thought of the behavior" (Triandis 1977, p. 16). This definition is very consistent with the concept of anticipatory affective responses. So is Triandis's suggestion that affect and expectancyvalue assessment are independent determinants of behavioral intentions. 2 In other words, in Triandis's (1980) model, intention is a function of three terms: social norm, affect, and the expected value of the behavior's consequences.

PAGE 61

52 Finally, the framework differs from all of the above in that it builds on the notion of affect as information to suggest a process account of consumption decisions geared toward the regulation of affective states. Now that the framework has been positioned conceptually, it needs to be tested empirically. A Research Agenda on Affect Recruitment The purpose of this dissertation research is to test the cornerstone of the framework: the thesis of an affect recruitment heuristic. Empirical examination of decision processes underlying affect management consumption is left for future research. Again, three propositions define affect recruitment: (1) access to or construction of concrete representations of the episode; (2) instantiation of (anticipatory) affective responses; and (3) reliance on these responses, because of their unique information value. It is suggested that these three characteristics differentiate this process from the expectancyvalue process typically assumed in consumer research. In addition, the framework identifies a series of determinants of affect recruitment. It must be recognized that these determinants are more than limiting "boundary conditions." They help refine the predictions of the framework by placing more severe a priori constraints on the data. Many of these determinants were logically derived from the processing assumptions of the framework. For instance, the prediction that affect recruitment is more likely under consummatory motives than under instrumental motives was derived from the assumptions that (1) consumers attach

PAGE 62

53 information value to anticipatory affective responses; and (2) this information value is contingent upon consumers' goals. Similarly, the prediction that imagery ability should increase the likelihood of affect recruitment was a logical implication of the proposition that affect recruitment may involve more concrete, analog representations of consumption episodes than expectancy-value processes. To the extent that finer predictions are supported by the data, more confidence can be placed in the processing assumptions of the framework, and in its theoretical status. The substantive significance of the framework also needs to be examined. Substantively, the value of the framework rests heavily on its ability to predict consumer decision outcomes from decision inputs This issue is examined in Experiments 1 and 2. These experiments focus on several theoretical determinants of affect recruitment. To the extent that these determinants moderate the influence of decision inputs on decision outcomes, it can be concluded that, from a substantive standpoint, the idea of affect recruitment is worth pursuing. The next three experiments are aimed at testing the processing assumptions of the framework (Propositions 1-3). The proposition that affect recruitment may involve concrete representations is tested indirectly in Experiment 3. This test relies on the logic that if affect recruitment necessitates the construction of concrete representations of the episode, reliance on this heuristic should depend on the person's propensity to process information in an analog (e.g., visual) mode rather than in a more abstract propositional (e.g., verbal) mode. (A similar indirect test is conducted in Experiments 1 and 2).

PAGE 63

54 Experiment 4 attempts to test this proposition in a more direct way through protocol analysis. Proposition 2, that in affect recruitment affective responses are anticipatorily instantiated, is examined in Experiment 5. Because such responses may be too subtle to be detected through verbal measurement (cf. Epstein's notion of vibes), this experiment relies on a misattribution paradigm to infer the instantiation of anticipatory affective responses. If decisions based on affect recruitment involve the instantiation of affective responses, misattribution of these responses to an external source should result in the subtraction of part of these responses from the decision outcome. To the extent that one cannot subtract what one does not have, observation of a subtraction effect would suggest that affective responses were indeed instantiated. The third proposition, which stresses the unique information value of anticipatory affective responses, is perhaps the most important one. If the feelings that consumers experience while holding a mental construal of the consumption episode in short-term memory have unique information value, they should influence the decision outcome even when the evaluative content of the episode is kept constant. 3 If the information value of affect is not unique, and is redundant with an expectancy-evaluation of the episode (e.g., Fishbein and Ajzen 1975), only the evaluative content of the episode should matter. This critical test is conducted in Experiments 3 and 4. 3 Although, admittedly, an orthogonal manipulation of evaluative content and anticipatory affective responses probably lacks ecological validity.

PAGE 64

55 Finally, a processing framework should be assessed not only on the basis of its predictive accuracy, but also on the basis of its descriptive accuracy. Ideally, its descriptive accuracy would be compared to that of an established framework. Such a comparative assessment is conducted in Experiment 4. This experiment uses a logic of process baselines to assess affect recruitment and expectancy-value attitude formation as paramorphic representations of decisions about consumption episodes.

PAGE 65

CHAPTER 4 EXPERIMENTS 1 AND 2: THE DETERMINANTS OF AFFECT RECRUITMENT Purpose and Predictions An important first step in the validation of a processing framework is to examine the ability of this framework to predict how inputs to this process will causally affect its outputs. The more specific the predictions generated by a framework— should they be empirically supported— the more confidence one can have in the heuristic value of this framework. These first two experiments test predictions of the framework on how behavioral intentions toward a consumption episode (a decision outcome) are affected by two types of decision inputs: (1) the experiential content of the episode (pleasant or unpleasant) and (2) the cost associated with this episode. It was postulated in Chapter 2 that to make a decision about consumption episodes, consumers may construct concrete representations of the episodes that may instantiate anticipatory affective responses. The pleasantness (or unpleasantness) of these responses (e.g., being relazed versus being frustrated) should logically depend on the affective content of the constructed representation (e.g., imagining having a terrible dinner in a restaurant). This affective content should in turn depend on available and As indexed, for example, by the appraisal dimensions discussed in Chapter 2 (e.g., desirability, agency, ...). 56

PAGE 66

57 accessible information about the episode (e.g., a friend's account of a previous dinner in the same restaurant). Two types of information were provided to subjects in these experiments: (1) the pleasantness (or unpleasantness) of a friend's previous experience with the episode (an input called experiential content); and (2) the friend's description of the cost associated with the episode. It was reasoned that the experiential content of the episode (conveyed by the friend) would be a strong determinant of the affective content of subjects' own construals of the episode (and presumably of their anticipatory affective responses). In other words, it was reasoned that the manipulated (i.e., observable) experiential content of the episode would be proximally related to subjects' unobserved anticipatory affective responses to the episode. In contrast, it was expected that the cost associated with the episode would be weakly related to anticipatory affective responses to the episode, although it could influence its overall evaluation. Experiments 1 and 2 examined how four theoretical determinants of affect recruitment would moderate the influence of these two decision inputs (see Figure 1): (1) whether the motive underlying the decision was instrumental or consummatory; (2) whether the decision was made in a context associated with feeling rules (Hothschild 1979), that is, a context where there are cultural expectations about how one should feel (e.g., people are typically expected to feel happy and joyful on their birthday); (3) subjects' chronic affect intensities (Larsen 1984); and (4) subjects' visual imagery abilities. The first two determinants were conceptualized as impinging upon the information value (or diagnosticity) of anticipatory affective responses. Consistent with

PAGE 67

58 Proposition 4a, it was predicted that anticipatory affective responses would be perceived as more relevant, as a decision input, in decisions guided by consummatory motives than in decisions guided by instrumental motives (see also Hirschman and Holbrook 1982 and Millar and Tesser 1989). Consequently, affect recruitment was expected to be more likely in the former case than in the latter. Therefore, it was hypothesized that the experiential content of an episode, because of its assumed proximal relation with anticipatory affective responses, would be more influential under consummatory motives than under instrumental motives. Conversely, the associated cost of the episode was expected to be less influential under consummatory motives than under instrumental motives. These two predictions were tested in Experiment 1. Hla: The experiential content of a consumption episode will have a stronger influence on consumers' decisions when consummatory motives are being pursued than when instrumental motives are being pursued. Hlb: The cost associated with a consumption episode will have a smaller influence on consumers' decisions when consummatory motives are being pursued than when instrumental motives are being pursued. Consistent with Proposition 4b it was predicted that when decisions are guided by consummatory motives, anticipatory affective responses would be perceived as even more relevant (or diagnostic) if the decision making context includes cultural expectations about how the decision maker should feel, i.e., feeling rules. If it is the decision maker's birthday, he/she should be more sensitive to how a given consumption episode will make him/her feel, because people are typically expected to feel happy and joyful on their birthdays. Consequently, feeling rules should further increase the influence of

PAGE 68

59 the experiential content of the episode and decrease the influence of its associated costpredictions tested in Experiment 2. H2a: The experiential content of a consumption episode will have a stronger influence on consumers' decisions when the decision context is associated with feeling rules. H2b: The cost associated with a consumption episode will have a smaller influence on consumers' decisions when the decision context is associated with feeling rules. Two other two determinants examined in Experiments 1 and 2, chronic affect intensity and visual imagery ability, were conceptualized as moderating the accessibility of anticipatory affective responses (see Propositions 5a and 5b). It was reasoned that both individual difference variables would contribute to make anticipatory affective responses to a given construal of the episode more accessible, thereby amplifying the effect of experiential content on behavioral intentions. Moreover, to the extent that the recruitment of anticipatory affective responses should be more likely when they are diagnostic (see Figure 1), it was predicted that the amplifying role of chronic affect intensity and visual imagery ability would manifest itself especially when consumers have consummatory motives or in the presence of feeling rules. A corollary prediction of the framework was that visual imagery ability and chronic affect intensity would not moderate the influence of associated cost, because this input was hypothesized to be processed relatively independently of affect recruitment. H3a: The influence of the experiential content of an episode on consumers' decisions should be amplified by chronic affect intensity when anticipatory affective responses are diagnostic (consummatory motives or feeling rules) but not when these responses are not diagnostic (instrumental motive).

PAGE 69

60 H3b: The influence of the experiential content of an episode on consumers' decisions should be amplified by visual imagery ability when anticipatory affective responses are diagnostic (consummatory motives or feeling rules) but not when these responses are not diagnostic (instrumental motive). H3c: The influence of the cost associated with a consumption episode on consumers' decisions is not affected by chronic affect intensity and visual imagery ability. Experiment 1: Method Subjects and Overview of the Design Subjects were 129 undergraduates at the University of Florida, enrolled in several business classes, who participated in return for course extra credit. Subjects were asked to report their intentions of engaging in two hypothetical consumption activities or episodes: test driving a sports car and having dinner in a French restaurant. Three factors were of primary interest in this experiment: (1) the experiential content of each episode (pleasant or unpleasant); (2) the cost associated with each episode (high or low); and (3) the motive for engaging in the episode (instrumental or consummatory). Experiential content and associated cost were manipulated between subjects, whereas motive was manipulated within subjects across the two episodes. The order of the two episodes and of the two motives (hence, their pairing) was completely randomized across subjects. In addition to these manipulated factors, two individual difference variables hypothesized to moderate affect recruitment were measured: subjects' chronic affect intensity and their visual imagery ability.

PAGE 70

61 Stimulus Material Each subject was presented with two scenarios pertaining to two consumption episodes. The first episode consisted of having dinner with a friend of the opposite sex in a new French restaurant in town. The second episode consisted of test-driving a Ferrari at a local dealership as part of a promotional offer for an automotive magazine. Four considerations dictated the selection of these two consumption episodes. First, each episode had to be both relevant for the population of subjects examined (college students) and realistically take place in the college town where the study was conducted. Second, although relevant for the population studied, the episodes had to be relatively novel (a new French restaurant; a Ferrari) to reduce the likelihood that subjects would primarily rely on prior attitudes to make their decisions. Third, each episode had to lend itself to independent (and ideally orthogonal) manipulations of anticipated affective responses and associated cost. Finally, each episode should realistically lend itself to enactment under consummatory or instrumental motives. The information about each episode was conveyed via a written scenario in which a friend described his or her own experience. This word-of-mouth format was chosen because of its intrinsic ability to convey experiential information. It was reasoned that to make their decisions, subjects would rely on both idiosyncratic inputs (e.g., prior attitudes toward French restaurants) and on the information contained in the friend's account. In affect recruitment they would construct a concrete representation of the episode and use their anticipatory affective responses to this construal as an input to their decision. It was expected that the experiential content of the friend's account would be

PAGE 71

62 a major determinant of subjects own anticipatory affective responses. Four versions of each episode were created (two levels of experiential content and two levels of associated cost). Dinner Episode: Experiential Content In the positive experience condition, the restaurant episode was described as follows: I was there the other night. What a classy place! My dinner was wonderful. You should see the dining room: white tablecloths, linen napkins, slightly dim light, relaxing music in the background . The food was incredible . Mmmmm!!! You should have seen how they presented my plate . wow! With vegetables carved into flowers, you know what I mean! And the waiter was just as classy; he was so polite ... he made me feel like a million bucks! [Manipulation of associated cost took place here] I was thinking about having dinner there tonight. Would you go with me? In the negative experience condition, the episode was described as: I was there the other night. What a dump! My dinner was awful! You should see the dining room: cheap paper napkins, the light was too bright, and the music in the background was irritating. The food was terrible... Even the way my plate was presented... gross! And the waiter was so rude ... he made me feel second class. I had such an unpleasant experience. [Manipulation of associated cost took place here] I was thinking about having dinner there tonight. Would you go with me? Dinner Episode: Associated Cost The associated cost was manipulated in the last statement of the friend's account. In the low cost condition, the restaurant was described as very inexpensive, whereas in the high cost condition the restaurant was described as very expensive. Test-Drive Episode: Experiential Content In the positive experience condition, the test-drive episode was described as follows: I tried it yesterday. It was a red one, like in the movies. I was so excited! At idle speed, it sounded like a purring tiger. Then, I pushed on the accelerator ...

PAGE 72

63 What an acceleration! I felt like I was in an airplane. I took it on 1-75. It was unbelievable. I was up to 100 mph before I even left the on-ramp! I must have gone for two miles at that speed. I was so thrilled... passing all these cars like they were driving at 40 mph. It felt so good. Everybody was staring at me. When I returned it to the dealership, I even saw one of my neighbors. You should have seen the look on her face! [Manipulation of associated cost] The promotion ends next week, why don't you go check it out? In the negative experience condition the episode was described as: I tried it yesterday. It was a red one, like in the movies... at idle speed, it sounded like a purring tiger. At first I was really excited, but you won't believe what happened. The dealer got into the car and told me: "You can only drive it around the parking lot. Can you believe that? That was ridiculous! I couldn't even press down the accelerator. I might as well have been in a Geo! The fastest I got was 20 mph. On top of that, everybody at the dealership was staring at me, I felt like an idiot! When I got out of the car, I even ran across one of my neighbors. She had been watching me the whole time! You should have seen the smirk on her face! I was so embarrassed. [Manipulation of associated cost] The promotion ends next week, why don't you go check it out? Test-Drive Episode: Associated Cost Again, the associated cost was manipulated in the last statement of the friend's account. In the low cost condition, the friend explains that in return for test-driving the car he or she only had to buy one issue of the magazine sponsoring the promotion. In the high cost condition, the friend explains that he or she had to buy a one-year subscription of the magazine. Motives Each episode description was accompanied by a situational context designed to manipulate the motive under which subjects would evaluate the episode. The situational context was presented in a separate frame above the description of the episode. Subjects were instructed to read carefully this situational context prior to reading the description of the episode. The situational context for the consummatory motive was identical for

PAGE 73

64 the two episodes: "It is the weekend. You don't have to study or work. You just want to relax, enjoy, and have a good time." The situational context for the instrumental motive was essentially the same for the two episodes. Subjects had to write a paper whose topic was related to the content of the episode. The instrumental situational context for the dinner episode was: You have to turn in a paper in a few days for a class. The topic of the paper is "The management of foreign-cuisine restaurants." To collect "first hand" material for the paper, you would like to observe how such a restaurant operates. The instrumental context for the test-drive episode was: You have to turn in a paper in a few days for a class. The topic of the paper is "How should the interior of sports cars be designed (e.g., control panel, seats)." To collect "first hand" material for the paper, you would like to check the interior of a prestige sports car (e.g., Ferrari, Corvette). Individual Differences Moderators It has been suggested that factors that influence the accessibility of anticipatory affective responses will moderate the likelihood that consumers will use affect recruitment to make decisions about consumption episodes. Two individual-specific variables related to these factors were assessed at the end of the session. First, it has been shown that individuals differ in terms of how intensely they experience affective stimulation (e.g., Larsen, Diener, and Emmons 1986). This individual difference variable was measured through the Affect Intensity Measure (AIM) Questionnaire (Larsen 1984). Because of limited subject time, a shortened version of the AIM was used in which half (20) of the items in the original scale were randomly selected. The shortened AIM scale is reproduced in Appendix A. Second, visual imagery ability was assessed

PAGE 74

65 through seven items taken from Sheehan's (1967) shortened version of the Betts Questionnaire (see Appendix B). Dependent Measures After reading the description of each consumption episode, subjects reported their likelihood of engaging in each consumption activity on a 9-point scale (1=1 would definitely not go; 9— I would definitely go). Next, they were instructed to write down everything that went through their minds as they were thinking about the activity, including thoughts, feelings, and memories. Finally, they completed various manipulation and demand checks. 2 Experiment 1: Results Manipulation Checks Anticipated Affect The anticipated affect associated with each episode was measured through a two-item semantic differential scale. Subjects were instructed to anticipate their feelings and emotions if they were to actually engage in each episode, then recorded their anticipated affective responses on two nine-point scales, anchored by 2 Deferring the manipulation checks until the end of the experimental session probably limited their ability to reflect the true amplitude of the manipulations. However, collecting these checks immediately after each decision would have sensitized subjects about the manipulations and possibly biased their responses to the second episode. A solution to this dilemma would have been to use hold out samples.

PAGE 75

66 "My feelings would be extremely pleasant/unpleasant" and "I would be in a very good/bad mood" (alpha dinner = .93; alpha test-drive = .89). Anticipated affective responses to each episode were submitted to a 3-way ANOVA (experience x cost x motive). As expected anticipated affective responses to the dinner episode were more pleasant in the positive experience condition (M = 7.64) than in the negative experience condition (M = 5.01, F (1, 121) = 55.62, p < .0001, or = .29). Similarly, anticipated affective responses to the test drive episode were more pleasant in the positive experience condition (M = 7.65) than in the negative experience condition (M = 6.86, F (1, 121) = 7.34, p < .01, co 2 = .05). No other effect was significant. Although these effects may appear small, these checks may underestimate the actual strength of the manipulation since the measures were collected at the end of the session. Associated Cost Subjects were asked two questions in relation to the dinner episode. First, subjects were asked "How much do they charge for a dinner at the French restaurant?" (1 = very little; 9 = a whole lot). Second, subjects were asked "Considering your budget, how much would you have to give up to have dinner with your friend at the French restaurant?" (1= I wouldn't have to give up much; 9 = I would have to give up a great deal). Responses to these two questions were averaged to provide a measure of perceived cost associated with the dinner episode (alpha = .73). As expected, the perceived cost associated with the dinner episode was higher in the high cost condition (M = 6.70) than in the low cost condition (M = 3.29, F (1, 121) = 141.60, p < .0001, or = .50). In addition, an experience x cost interaction suggested that the effect of the cost manipulation was stronger in the negative experience condition

PAGE 76

67 than in the positive experience condition (F (1, 121) = 19.34, p < .0001, w 2 = .07). Similar questions were asked in relation to the test-drive episode. Subjects were asked "How much do they ask in return for test driving the Ferrari?" (1 = very little; 9 = a whole lot) and "How much would you have to give up to test drive the Ferrari under the conditions described by your friend?" (1= I wouldn't have to give up much; 9 = 1 would have to give up a great deal). Responses to these two questions were also averaged (alpha = .66) to provide a measure of perceived cost associated with the testdrive episode. Again, the perceived cost associated with the test-drive episode was higher in the high cost condition (M = 4.42) than in the low cost condition (M = 2.82, F(l, 121) = 27.85, p < .0001, co 2 = .16). There was an unexpected main effect of experience. Subjects perceive the cost of this episode to be higher in the negative experience condition (M = 4.17) than in the positive experience condition (M = 3.04, F(l, 121) = 13.95, p < .001, or = .08). Perhaps the wording of this check, which did not restrict "cost" to its restrictive monetary meaning, prompted subjects to incorporate "affective costs" (e.g., feeling silly) in their answers. No other effect was significant. Effects of Experiential Content and Associated Cost Reported intentions of engaging in each episode were submitted to a 2 (motive) x 2 (experience) x 2 (cost) x 2 (order of motives) x 2 (order of episodes) split-plot ANOVA, with motive as a repeated factor and the other factors between subjects. The treatment means collapsed across episodes (and orders of motives and of episodes) are

PAGE 77

68 reported in Table 4.1. The overall repeated ANOVA results are reported in Appendix C. Experiential Content As predicted, there was an experience x motive interaction (F (1,113) = 5.41, p = .02). The effect of experiential content was stronger when subjects had a consummatory motive (enjoying a good time over the weekend; M pos = 6.07; M neg = 2.52) than when they had an instrumental motive (writing a paper, M pos = 7.01; M neg = 4.82). This result supports hypothesis la. TABLE 4.1 BEHAVIORAL INTENTIONS (COLLAPSED ACROSS EPISODES) AS A FUNCTION OF EXPERIENTIAL CONTENT, ASSOCIATED COST AND MOTIVE

PAGE 78

69 Effects of Affect Intensity and Visual Imagery It was proposed that high accessibility of anticipatory affective responses would increase the likelihood of affect recruitment. Factors that theoretically increase the accessibility of anticipatory affective responses should amplify the effect of experiential content on behavioral intentions. It was hypothesized that, when subjects had consummatory motives, visual imagery abilities and chronic affect intensity would interact with the experience content of the episodes to affect behavioral intentions. That is, two 3-way interactions were predicted: experience x visual imagery x motive and experience x affect intensity x motive. Subjects' affect intensity and visual imagery ability scores were added (after centering) to the repeated measure ANOVA model described earlier. For parsimony, only the theoretically meaningful interactions between these two predictors, experiential content, cost, and motive were included in the analysis. The overall analysis is summarized in Appendix D. As predicted by Hypothesis 3c, neither chronic affect intensity nor visual imagery ability interacted with the cost associated with the episode (both Fs < 1). While the experience x affect intensity x motive interaction was not significant (F( 1,106) = 2.49, p = 0. 12)-disconfirming H3a, the predicted experience x visual imagery x motive interaction was uncovered (F(l,106) = 5.81, p = 0.02). To investigate the nature of this triple interaction, a multiple regression model was estimated at each level of the motive (after dummy coding of the other effects). The results of these two regressions are reported in Tables 4.2 and 4.3.

PAGE 79

TABLE 4.2 EFFECTS OF VISUAL IMAGERY AND EXPERIENTIAL CONTENT ON INSTRUMENTAL BEHAVIORAL INTENTIONS 70 Dependent Variable: Intention Instrumental

PAGE 80

71 TABLE 4.3 EFFECTS OF VISUAL IMAGERY AND EXPERIENTIAL CONTENT ON CONSUMMATORY BEHAVIORAL INTENTIONS Dependent Variable: Intention Consummatory

PAGE 81

72 When subjects had an instrumental motive, their visual imagery ability did not affect their behavioral intentions (t < 1). In contrast, when subjects had a consummatory motive, their visual imagery ability amplified the effect of experiential content of the episode on their behavioral intentions(t( 1,111) = 2.40, one-tailed p < .01). This result supports Hypothesis H3b. It is consistent with the idea that when anticipatory affective responses are diagnostic (e.g., consummatory motives), their hedonic content is likely to influence consumers' decisions, especially when these anticipatory responses are highly accessible (e.g., high visual imagery ability). Correlational Analysis It was hypothesized that the higher weight assigned by consumers to the experiential content of an episode under consummatory motives was caused in part by stronger anticipations of affective responses to the episode. 3 Specifically, it was predicted that anticipated affective responses to the episodes would be a stronger mediator of behavioral intentions when subjects had consummatory motives than when they had instrumental motives. Anticipated affective responses to each episode (test-drive and dinner) were entered as continuous predictor of behavioral intentions toward the corresponding episode. To examine the moderating role of motives, the motive under which the episode was evaluated was entered as an additional dummy coded predictor (0 As stated earlier, these anticipations were not assessed at the time of the decision. Instead, anticipated affective responses were measured retrospectively at the end of the session, which weakens the interpretability of the observed responses as mediators. While these measures are our best estimates of subjects' anticipations of affective responses, their limitation is fully acknowledged.

PAGE 82

73 = instrumental; 1 = consummatory) along with an interaction term between anticipated affect and motive. The results of these regression analyses are reported in Tables 4.4 and 4.5. There was an expected interaction between anticipated affective responses to the test-drive episode and motive (t (1,125) = 1.78, one-tailed p < 0.05). Anticipated affective responses to the test-drive episode were a stronger predictor of behavioral intentions when subjects had a consummatory motive than when subjects had an instrumental motive. However, for the dinner episode the interaction was opposite to the prediction (t (1,125) = -2.53, p < 0.02). Subjects with an instrumental motive (writing a paper) weighted their anticipated affective response to the dinner episode even more heavily that those with a consummatory motive. The explanation of this effect is not clear. A discussion of this experiment's results follows the report of Experiment 2. TABLE 4.4 BEHAVIORAL INTENTIONS TOWARD THE TEST DRIVE EPISODE AS A FUNCTION OF ANTICIPATED AFFECT MOTIVE Dependent Variable: Intention Test-Drive

PAGE 83

74 TABLE 4.5 BEHAVIORAL INTENTIONS TOWARD THE DINNER EPISODE AS A FUNCTION OF ANTICIPATED AFFECT MOTIVE Dependent Variable: Intention Dinner

PAGE 84

75 or not the decision was made in the context of feeling rules (a birthday). Experiential content and associated cost were again manipulated between subjects, whereas decision context was manipulated within subjects across the two episodes. The order of the two episodes was randomized across subjects. The decision context was presented in a fixed order: feeling rules were associated with the first episode but not with the second. A pretest had suggested that this manipulation would be stronger if the feeling rule context was presented first. Experiential Content and Associated Cost The manipulations of the experiential content of each episode and of its associated cost were identical to those used in Experiment 1 Decision Context Unlike in Experiment 1, all decisions were made for consummatory motives. One of the decisions was associated with additional feeling rules, whereas the other decision was not. For the "feeling rule" decision, subjects had to make their decisions in the context of their birthday. The description of the episode was preceded by a situational context reading as follows: "Today is special, it is your birthday! It is the weekend. You don't have to study or work. You just want to relax, enjoy, and have a good time. For the "no-feeling rules decision, the situation context was described as follows: "It is another weekend (not on your birthday). You don't have to study or work. You just want to relax, enjoy, and have a good time."

PAGE 85

76 Experiment 2: Results Behavioral Intentions Reported intentions of engaging in each episode were submitted to a 2 (decision context) x 2 (affect) x 2 (cost) x 2 (order of episodes) split-plot ANOVA, with decision context as a repeated factor and the other factors between subjects. The mean behavioral intentions, collapsed across episodes (and order of episodes), are reported in Table 4.6. The overall repeated measures ANOVA results are reported in Appendix E. The predicted experience x context (F (1,77) = 1.71, p = .19) and cost x context (F (1,77) = 1.17, p = .28) interactions both failed to reach significance. Only directional support for our predictions was obtained. The simple effect of experiential content was somewhat (but not significantly) stronger in the feeling rules context (M pos = 5.70; M neg = 2.55, or = .27) than in the no-feeling rules context (M pos = 5.77; M ni;g = 3.69, or = .12). Conversely, the simple effect of cost was somewhat (but not significantly) smaller in the feeling rules context (M )ow = 4.26; M high = 4.02, or = .00) than in the no-feeling rules context (M, ow = 5.33; M neg = 4.14, or = .04). Additional analyses that parallel those reported for Experiment 1 (e.g., effects of visual imagery, affect intensity) failed to yield any significant results and are not reported.

PAGE 86

77 TABLE 4.6 BEHAVIORAL INTENTIONS (COLLAPSED ACROSS EPISODES) AS A FUNCTION OF EXPERIENTIAL CONTENT, ASSOCIATED COST AND DECISION CONTEXT

PAGE 87

78 It was observed that when subjects had a consummatory motive, their decisions were more influenced by the experiential content of the episode, and less influenced by the cost associated with it, than when subjects were pursuing an instrumental motive. If, as hypothesized, the experiential content (positive or negative) of the episode is proximally related to anticipatory affective responses (pleasant or unpleasant) to it, this finding would be consistent with the proposition that affect recruitment is more likely when anticipatory affective responses are regarded as a relevant decision input. When subjects had a consummatory motive, the addition of feeling rules to the decision context slightly increased the influence of experiential content and decreased the influence of associated cost, but not significantly. This lack of significance may be more methodological than substantive. First, Experiment 1 suggests that the manipulation in the "no-feeling-rule" condition may have been strong enough to induce subjects to pay little consideration to the cost associated with the episode and focus instead on its experiential content (see Table 4. 1). Therefore, the nonsignificance of the feeling rules factor may be due to a ceiling effect. Second, it should be noted that both the "no feeling rules" and the "feeling rules" conditions mentioned that it was the weekend. It is conceivable that weekends in and of themselves are associated with feeling rules. That is, people— including college students— presumably have strong beliefs that weekends are for relaxation and enjoyment. In other words, the manipulation may have been weaker than anticipated. Although chronic affect intensity did not have the expected effect, behavioral intentions were determined by the predicted complex interaction among visual imagery,

PAGE 88

79 experiential content, and motive. When subjects had a consummatory motive, and affect recruitment was presumably diagnostic, visual imagery amplified the influence of affective/experiential information presented to them. Visual imagery did not have such an effect when subjects had an instrumental motive. This finding is consistent with the proposition that affect recruitment may involve the construction of a concrete, analoglike, representation of the episode. High visual imagery ability may increase the accessibility of anticipatory affective responses to this representation. Limitations Several limitations of these experiments deserve mention. Given the hypothetical nature of the decisions to be made, one may question whether the effects reported here would describe how consumers would make "real" decisions. To remedy this limitation, the experiments reported in the next two chapters will involve more substantive decisions. Focusing on determinants (i.e., antecedents) of affect recruitment and decision outcomes introduced a more serious limitation of these two experiments. Although it was shown that these antecedents could have meaningful influences on the outcome of consumers' decisions, the underlying processes were not directly examined. For instance, it was suggested that the increased influence of experiential content under consummatory motives was mediated by a stronger weight attached to anticipated affective responses. Whereas a correlational analysis supported this explanation for the

PAGE 89

80 test-drive episode, for reasons that are unclear, a similar analysis contradicted this explanation for the restaurant episode. A related limitation is that varying the experiential content of an episode is only an indirect— albeit ecologically sensible— manipulation of anticipatory affective responses to the episode. It is therefore difficult to ascertain that anticipatory affect per se played a significant role in the observed findings. It is possible that the evaluative content of the episode was the unique determinant of consumers' decisions. To address these issues, the next two studies will attempt to manipulate the affective experience of consumers as they are making a decision, without altering the evaluative content of the episode to be decided upon.

PAGE 90

CHAPTER 5 EXPERIMENT 3: THE INFORMATION VALUE OF ANTICIPATORY AFFECTIVE RESPONSES Purpose and Predictions The first two experiments examined the ability of the framework to predict how two decision inputs experiential content and associated cost, influence behavioral intentions toward two consumption episodes: having dinner in a restaurant and test driving a car. Two theoretical determinants of affect recruitment, consumers' motives and visual imagery abilities, were found to moderate the influence of these inputs in a manner consistent with the concept of affect recruitment. It was observed, for instance, that imagery abilities amplified the effects of experiential content when subjects had consummatory motives but not when they had instrumental motives. The consistentcy between such findings and the predictions of the framework suggest that the idea of affect recruiment is at least worth pursuing. A more direct investigation of the hypothesized processes is in order. A central proposition of this framework is that feelings that are (anticipatorily) experienced in response to a construal of the consumption episode (i.e., anticipatory affective responses) have unique information value That is, the outcome of consumers' decisions about consumption episodes is not necessarily determined by the evaluative content of their mental representation of the episode; this outcome may also be affected 81

PAGE 91

82 by consumers' momentary affective responses to this representation. To show that anticipatory affective responses, as decision inputs, are not simply redundant with the evaluative content of representations of the episode, it is necessary to manipulate the former independently of the latter (unlike in Experiments 1 and 2). In Experiment 3, subjects' mood was used to proximally manipulate their anticipatory affective responses to an episode without altering its evaluative content. Previous research (e.g., Schwarz and Clore 1983) has shown that people often misattribute their current mood states to the object that they are judging. If to make decisions about consumption episodes, consumers "see how it feels" (i.e., rely on affect recruitment), their decisions should be biased by their preexisting mood state: the decision outcome should be more favorable if the preexisting mood is positive than if it is negative. Assuming that this manipulation does not alter the evaluative content of consumers' representation of the episode, this pattern of results would suggest that the influence of moods and emotions experienced during the decision process is at least partially independent of the influence exerted by the evaluative content of the episode. If this effect vanishes when anticipatory affective responses have low diagnosticity, one can additionally infer that it is because of their information value that proximally manipulated anticipatory affective responses influence decision outcomes (e.g., Schwarz and Clore 1983). In this experiment, subjects who had been induced to be in a positive or a negative mood through a supposedly unrelated task were asked to assess the likelihood that they would be going to a movie over the following weekend. Consistent with

PAGE 92

83 Proposition 4a, the diagnosticity of the anticipatory affective responses was manipulated by modifying subjects' motive for going to the movie. Half the subjects were assigned a consummatory motive (anticipatory feelings had high information value), whereas the other half were assigned an instrumental motive (anticipatory feelings had low information value). Because subjects to whom anticipatory affective responses are diagnostic are more likely to inspect how the consumption episode "feels," it was predicted that the mood manipulation would have a greater influence on the decisions of subjects with a consummatory motive than on decisions of subjects with an instrumental motive. Hla When subjects have a consummatory motive, a positive preexisting mood will result in more favorable intentions than a negative preexisting mood. Hlb When subjects have an instrumental motive, their behavioral intentions will be less influenced by their preexisting mood than when they have a consummatory motive. In Experiment 1 visual imagery ability was found to be a significant moderator of behavioral intentions. Experiment 3 explored the role of consumers' processing style. It has been suggested that individuals vary in terms of their propensity to process information visually or verbally (Childers, Houston, and Heckler 1985; Richardson 1977). If, as hypothesized, affect recruitment involves concrete analog-like representations of consumption episodes, consumers' tendency to rely on affect recruitment may be correlated with their propensity to process information visually rather than verbally. That is, "visualizers" may be more likely to use a "see-how-it-feels" heuristic than verbal izers." Therefore, among subjects with a consummatory motive,

PAGE 93

84 visualizers may be more influenced by their preexisting mood states than verbalizers. Among subjects with an instrumental motive, processing style should have little effect because subjects should not be likely to inspect their feelings given their low diagnosticity. H2a Among subjects with a consummatory motive, visualizers will be more influenced by their preexisting mood than verbalizers. H2b Among subjects with an instrumental motive, neither visualizers nor verbalizers will be influenced by their preexisting mood. Method Subjects and Design Eighty-eight undergraduate marketing students at the University of Florida participated in the study for extra course credit. They participated in sessions of 4-18 and were randomly assigned to one of four experimental conditions of a 2 x 2 design. All four conditions were run in each session. The first factor was subjects' preexisting mood (positive or negative) when making the decision. The second factor was their motive (consummatory or instrumental) for going to the movie. In addition to these two manipulated factors, this experiment also assessed subjects' processing style along the visual izer-verbalizer dimension. Procedure Subjects were told that they would be participating in two separate studies. The "first" study was purportedly about scale development. Subjects first completed the Style of Processing Scale (Childers et al. 1985). Their mood was then manipulated by having them provide vivid descriptions of a recent event that had affected

PAGE 94

85 them personally. They were told that these descriptions would be used to develop a scale to be used in future research. The "second" study was introduced next as one that investigates students' liking of various consumption activities. Subjects were asked to read carefully the description of a fictitious new movie and to report the likelihood that they would attend a preview for the movie over the weekend. All sessions were run on Tuesdays and Wednesdays. The delay between the study and the consumption episode was designed to ensure that subjects would not assess the episode as a mean to manage their experimentally induced mood state. Finally, other dependent measures, process measures, and demand checks were collected. Mood Subjects' mood was manipulated using a task similar to the one used by Schwarz and Clore (1983). Subjects were told that the experimenter was trying to develop a "Scale of People's Concerns." The scale was purportedly intended to assess issues that matter in people's lives so that they could be related to television offerings. To help identify items for the scale, subjects were asked to describe in writing a recent event that had affected them personally. In the negative mood condition they were instructed to describe an event that made them "feel really bad," whereas in the positive mood condition they were instructed to describe an event that made them "feel really good." Subjects were encouraged to be as vivid and detailed as possible. They were given 20 minutes to complete the task. Stimuli and Motives In the consummatory motive condition, the instructions emphasized the experience of the going to the movie as the main motive for assessing whether to go to the preview:

PAGE 95

86 After a long week of school and hard work, every student deserves some good time on the weekend. So this weekend, you may want to enjoy life, at least for a few hours. This weekend there will be several sneak previews of the movie "Oklahoma Bookstore." It is a new comedy staring John Goodman (from ABC's "Roseanne") as the owner of a bookstore in a small town and Steve Martin as an eccentric customer. The New York Times' critics have described "Oklahoma Bookstore" as one of the most hilarious movies of the year. Siskel and Ebert gave it "Two Thumbs Up." The previews will be shown Friday to Sunday at 6:30, 8:30, and 10:15 at Litchfield theaters (in Butler Plaza). If you want to have a good time (you deserve it!), how likely is it that you will go to one of the previews of "Oklahoma Bookstore?" In the instrumental condition, the instructions emphasized an extrinsic reason for going on not to the movie preview: Next week (on Tuesday and Wednesday), we will run a commercial study for a movie company. The study is not for credit, but you will receive 4 dollars if you participate. To participate, you need to go this weekend to one of the sneak previews of the movie "Oklahoma Bookstore." It is a new comedy staring John Goodman (from ABC's "Roseanne") as the owner of a bookstore in a small town and Steve Martin as an eccentric customer. The New York Times' critics have described "Oklahoma Bookstore" as one of the most hilarious movies of the year. Siskel and Ebert gave it "Two Thumbs Up." The previews will be shown Friday to Sunday at 6:30, 8:30, and 10:15 at Litchfield theaters (in Butler Plaza). Of course, if you attend one of these previews this weekend, you don't have to participate in next week's study. But you may. If you decide to participate, bring your ticket stub and we will reimburse it (in addition to the 4 dollars of your participation). Given that it would allow you to participate in next week's study (and earn 4 dollars), how likely is it that you will go to one of the previews of "Oklahoma Bookstore?"

PAGE 96

87 It was reasoned that subjects in this latter condition would be less likely to rely on their feelings to make their decisions because the movie experience itself should be less relevant. It should be noted, however, that this scenario does not truly substitute an instrumental motive for an experiential (or consummatory) one. Rather, it provides an instrumental motive in addition to an experiential motive. Style of Processing Subjects were administered Childers et al.'s (1985) Style of Processing Scale (SOP), which assesses people's "preference and propensity to engage in a verbal and/or visual modality of processing" (p. 130). The SOP scale, adapted from the Verbal izer-Visualizer Scale (Richardson 1977), contains 22 items: 1 1 items assessing propensity to process visually and 1 1 items assessing propensity to process verbally. Measures The main dependent variable was subjects' reported likelihood of going to one of the movie previews measured on a nine-point scale (1 = extremely unlikely; 9 very likely). After completing this measure, subjects were asked to list all the thoughts and feelings that went through their minds as they were making the decision. They were then asked about the purpose of the study. Next they completed several other measures such as their attitudes toward going to the movies on the weekend and their attitudes toward comedies. Finally, as a manipulation check, subjects were asked to report what mood they were in after having described an event that affected them.

PAGE 97

Results Assessment of the Independent Variables Mood Manipulation At the end of the experimental session, subjects were asked to report how they were feeling after having described their personal experience and before making their decision. These feelings were assessed by five bipolar seven-point items, anchored by cheerful/depressed; sad/joyful; annoyed/pleased; happy/unhappy, in a good/bad mood. Principal component analysis uncovered a single valence factor accounting for 85 percent of the variance. As a result, the five items were summed into a single measure of mood pleasantness (theoretical range: 5-35, alpha = .96). These self reports were submitted to a 2 (mood) x 2 (motive) x 2 (style of processing; see below) ANOVA. As expected, subjects in the positive mood condition reported being in a more pleasant mood (M = 28.67) than subjects in the negative mood condition (M = 24.84, F (1,80) = 112.51, p < 0.0001, omega-squared = 0.55). No other effect was significant. Motive The effectiveness of the motive manipulation was assessed using an independent sample of 35 subjects. They were asked which of two perspectives would "most people have in mind when assessing their intention of seeing" the movie. Answers were collected on a seven-point scale anchored by "They will focus on how enjoyable or unpleasant this experience will be" and "They will focus on whether there is something to gain." The question was worded in the third person to reduce social

PAGE 98

89 desirability pressures. 1 These subjects were also asked to report, on a similar scale, what perspective they had in mind when making their decision. These two measures were highly correlated (r = .92), and were, therefore, averaged into a single index of subjects' motives. Low values of this index indicated a consummatory motive, whereas high values indicated an instrumental motive. The motive index was submitted to a oneway ANCOVA with motive instructions (instrumental or consummatory) as a betweensubjects factor and social desirability as a covariate. The analysis revealed that subjects exposed to the instrumental instructions (M = 3.78) were more likely to base their decisions on whether there were "something to be gained" (and less likely to base their decisions on "how enjoyable or unpleasant this experience would be") than subjects exposed to the consummatory instructions (M = 2.42, F(l,32) = 4.02, one-tailed p < .03, omega-square = .08). To assess potential confounds of this manipulation, subjects were also asked (1) How important it was for them to be accurate in their assessment and (2) How difficult the decision was. Neither of these measures was affected by the motive manipulation (both Fs < 1), suggesting that the manipulation did not induce different levels of involvement and task difficulty. A pretest of the manipulation check had suggested that subjects in the instrumental motive condition, which involved a monetary incentive, were reluctant to admit that their decision was not purely consummatory. Propensity to respond in a socially desirable way was, therefore, assessed in the main manipulation check study, using six items selected from the Crown and Marlow (1960) scale (Ballard, Crino, and Rubenfeld 1984).

PAGE 99

90 Processing Style Although Childers et al. (1985) suggested that the visual and verbal components of the SOP scale each capture a single factor, with good internal consistency, the factorial structure suggested by these authors was not replicated in this study. The internal consistency of the eleven visual items of the SOP scale was only moderate (alpha = .69) and a principal component analysis suggested that four factors underlie these items. Although the internal consistency of the verbal items was higher (alpha = .75), a principal component analysis suggested that there were three underlying factors. Therefore, it appears that, in this study at least, the SOP scale had a higher dimensionality than suggested by Childers et al. (1985). 2 Despite the scale's higher dimensionality, a visualizerverbal izer score was computed for each subjects by summing responses to the visual items and subtracting responses to the verbal items, as recommended by Childers et al. (1985). Based on this score, a median split was performed to distinguish "visualizers" from "verbalizers" in In retrospect, this apparent discrepancy may be explained by the fact that Childers et al. (1985) did not use exploratory factor analysis but confirmatory factor analysis with LISREL. Based on the factor loadings, the authors concluded that a two-orthogonal -factor solution was satisfactory, even though the chi-square statistic suggested a significant lack of fit.

PAGE 100

91 the remaining analyses. 3 The split was performed within each of the four experimental conditions (mood x motive) to avoid overly unequal cell sizes. Effects on Behavioral Intentions The likelihood of going one of to the movie previews was submitted to a 2 (mood) by 2 (motive) by 2 (processing style) ANCOVA, with the number of movies seen in the last year in the same theater as a covariate. The least-square means are reported in Table 5.1 and the overall ANCOVA results are reported in Appendix F. There was a main effect of motive (F (1,79) = 18.62, p < 0.0001). Subjects with an instrumental motive reported a higher likelihood of going (M = 7.25) than subjects with a consummatory motive (M = 4.93). This effect probably reflects the higher monetary incentive in the instrumental condition (four dollars and a free ticket). Mood and Motive. The mood x motive interaction was significant (F(l,79) = 2.71, one-tailed p = 0.05). A planned contrast reveals that in the consummatory condition subjects in a positive mood (M = 5.67) reported a higher likelihood of going To analyze interactions between a continuous variable (e.g., processing style) with categorical variables (e.g., motive and mood), Jaccard, Turrisi and Wan (1990) prefer the use of multiple regression, with dummy coding of the categorical variables (as in Experiment 1), to the dichotomization of the continuous variable. Nevertheless, the analyses reported here rely on dichotomization for two reasons. First, the dimensionality of the scale (two theoretical factors and seven empirical factors) calls into question treating the SOP index as a unidimensional continuous measure. Second, dichotomization better reflects the underlying logic of the scale, which is meant to categorize individuals (see Richardson 1977). For completeness, the approach recommended by Jaccard et al. (1990) was also used to analyze the data. The results were mostly nonsignificant and difficult to interpret.

PAGE 101

92 to a preview than subjects in the negative mood condition (M = 4.20, F(l,79) = 4.20, one-tailed p = 0.02). This finding is consistent with hypothesis la. In contrast, positive mood (M = 7.15) and negative mood subjects (M = 7.36) in the instrumental condition were equally likely to go to a preview (F < 1), a finding consistent with hypothesis lb. TABLE 5.1 MEAN SELF-REPORTED LIKELIHOOD OF GOING TO A PREVIEW Consummatory Instrumental motive motive Verbalizers Visualizers Verbalizers Visualizers Negative mood 4.53 3.87 7.07 7.66 Positive mood 4.73 6.61 6.80 7.49 Difference .20 2.74 -.27 -.17 (positive-negative) Processing Style Although the overall mood x motive x style of processing interaction was not significant (F(l,79) = 1.43, p = .23), it appears that the simple main effect of mood in the consummatory condition was essentially driven by visualizer subjects. A planned contrast shows that the simple effect of mood is significantly greater among visualizers with a consummatory motive (diff pos neg = 2.74) than in the other conditions pooled together (F(l,79) = 11.85, one-tailed p < .001). This finding supports Hypothesis 2a. Mood did not influence the likelihood of going to the movie preview of verbalizers and subjects with an instrumental motive, supporting H2b.

PAGE 102

93 Discussion The Information value of Anticipatory Affective Responses To show that anticipatory affective responses have unique information value, this experiment used a contextual mood induction to manipulate feelings experienced during the decision process. Conceptually, this manipulation allows us to alter how anticipatory affective responses to a consumption episode are experienced without altering the evaluative content of this episode. Consistent with this study's hypotheses, it was found that subjects' preexisting mood influenced their reported likelihood of going to the movie preview when their motive was consummatory but not when their motive was instrumental. This pattern of findings is consistent with the proposition that when anticipatory affective responses are diagnostic (e.g., because of a consummatory motive) consumers trying to reach a decision about a consumption episode are more likely to "see how it feels" and incorporate these feelings into their decision. This affect recruitment process is less likely to occur when anticipatory affective responses are perceived to be nondiagnostic (e.g., because of an instrumental motive). Beside the fact that the feelings experienced during the decision process may alter the outcome of this process, two inferences can be drawn from these results. First, to the extent that this effect was contingent upon subjects' motives (and presumably the diagnosticity of anticipatory affective responses), one can infer that these momentary feelings influence the decision because of their information value (e.g., Schwarz 1990; Schwarz and Clore 1983, 1988). Second, to the extent that this contextual manipulation

PAGE 103

94 of anticipatory affective responses did not affect the evaluative content of the episode, one can infer that the information value of these momentary feelings is unique. It was also found that, in the consummatory condition, the influence of preexisting mood was stronger among subjects with a propensity to process information visually than among subjects with a propensity to process information verbally. This finding is consistent with the thesis that affect recruitment should be more readily used by those consumers who can easily construct a vivid representation of the episode and instantiate anticipatory affective responses to this representation. Inclination to rely on affect recruitment as a decision heuristic, i.e. to "see how it feels," may be stronger among consumers with a propensity to process information visually rather than verbally. This finding also provides indirect support to the proposition that the type of representation involved in affect recruitment may be more concrete and imaginal than abstract and propositional. Confidence in these findings requires that three rival explanations be ruled out. Alternative Explanations Access to Different Evaluative Contents A critical premise of this experiment is that the mood manipulation did not alter the evaluative content of the episode. Although the external information provided was constant across conditions, the mood manipulation may have modified the content of subjects' internal representations of the episode. It has been proposed that mood influences a variety of judgments by increasing the accessibility of mood congruent material in memory (e.g., Isen 1984; Isen et al. 1978). While thinking about the movie preview, subjects in a positive mood may have

PAGE 104

95 accessed more positive thoughts and less negative thoughts than subjects in a negative mood. Therefore, the actual content of subjects' mental representations of the episode may have differed across mood conditions. However, this explanation does not explain the absence of a mood effect in the instrumental condition. If the mood effect observed in the consummatory condition was caused by an increased accessibility of mood congruent thoughts, a parallel effect should have occurred in the instrumental condition. It did not. The manipulation/confounding check study suggests that the absence of a mood effect in the instrumental condition was not due to a higher level of involvement or a higher degree of task difficulty. Ceiling Effect On the other hand, the absence of mood effect in the instrumental condition may be due to a ceiling effect. The extra monetary rewards of the instrumental condition (four dollars and a free ticket) may have raised the attractiveness of the consumption episode so much that no additional mood effect could be detected. To test for this explanation, the data were reanalyzed through the method of successive intervals, largely insensitive to ceiling and floor effects (Jones 1960). The original nine-point dependent measure was recoded as a three-category ordinal scale. For each mood x motive condition, an interval scale value was computed from the observed response frequencies across these three categories. The transformed data yielded a virtually identical pattern of results, suggesting that a ceiling effect is not a likely explanation for the absence of mood effect in the instrumental condition. Stronger Mood Manipulation among Visualizers The stronger mood effect among visualizers with a consummatory motive may indicate that the mood manipulation

PAGE 105

96 was stronger among visual izers than among verbal izers. When asked to describe a pleasant (unpleasant) personal experience, visualizers might have had more vivid recollections than verbalizers. As a result, visualizers might have been in a better (worse) mood than verbalizers before making the decision. This explanation, however, is not consistent with the absence of a mood x processing style interaction on the mood manipulation check (F < l). 4 Subjects' mood-induction protocols were also contentanalyzed. A judge, blind to the study's hypotheses and to the subject's style of processing and motive, coded the pleasantness (1 = Extremely sad/unpleasant; 9 = Extremely happy/pleasant) of each written description. As expected, these descriptions were more pleasant in the positive mood condition (M = 7. 12) than in the negative mood condition (M = 2.40, F(l,80) = 112.51, p < .0001). However, the mood x style of processing interaction was again insignificant (F < 1). Therefore, there was no evidence that subjects' style of processing might have intensified the mood manipulation. The Need for Additional Evidence Reanalysis of the data through successive-interval scaling suggests that the absence of mood effect in the instrumental condition was not caused by a ceiling effect. However, stronger evidence against a ceiling-effect explanation would require that the average level of the main dependent variable be lower in the instrumental condition that it was in the present experiment. 4 It is acknowledged, however, that the timing of this manipulation check (at the end of the experimental session) decreases its diagnostic value.

PAGE 106

97 This experiment allowed a closer examination of the processes underlying affect recruitment than the first two studies, which focused on the predictive ability of the framework. However, the process evidence collected in this experiment remains largely inferential. Further evidence needs to be collected to assess the accuracy of affect recruitment as a description of the actual processes underlying many decisions about consumption episodes. In particular, the descriptive accuracy of the proposed process framework needs to be compared to that of expectancyvalue models (e.g., Fishbein and and Ajzen 1975). This issue is addressed in Experiment 4.

PAGE 107

CHAPTER 6 EXPERIMENT 4: INFERRING AFFECT RECRUITMENT THROUGH PROCESS BASELINES Purpose and Predictions: The Logic of Process Baselines Experiment 3 suggests that consumers may assign unique information value to the feelings they experience during the decision process. Consumers appear to let the valence of these feelings affect their decisions, to the extent that these feelings are perceived to be relevant. Consumers' propensity to incorporate their current feelings into their decisions seems to depend on their processing style: this propensity may be stronger among consumers who tend to process information more visually than verbally. While these findings are consistent with the proposition that affect recruitment may adequately describe the processes underlying decisions guided by consummatory motives, more direct evidence of this descriptive accuracy is needed. The purpose of this experiment is twofold. First, this experiment examines the descriptive accuracy of affect recruitment as a model of how consumers with a consummatory motive make decisions about consumption episodes. Second, the experiment compares this descriptive accuracy with that of an expectancy-value attitude formation model (cf. Fishbein and Azjen 1975). To achieve this dual purpose, this experiment expands the design of Experiment 3 to include process baseline conditions. 98

PAGE 108

99 Subjects in Experiment 3 did not receive any direction about the process (or decision heuristic) they should follow to make their decisions. When no process constraint is imposed on them, consumers presumably self-select some (unknown) heuristic appropriate for the decision with which they are faced. One way of inferring the nature of this process is to have "baseline" subjects instructed to follow a specific (and known) decision process, and compare the outcomes of their decisions with those of subjects who did not receive such processing instructions. To the extent that both groups of subjects reach similar decisions, one can infer that the specified process imposed on the "baseline" subjects provides at least a paramorphic --and possibly isomorphic-representation of the unknown process followed by the uninstructed subjects. If, on the other hand, the two groups reach different decisions, it is unlikely that the unknown process followed by the uninstructed subjects resembles the one specified in the baseline condition. 1 In this experiment, subjects, who had been induced to be in a positive or a negative mood through a supposedly unrelated task, were asked to assess their intention of engaging in a consumption activity. As in Experiment 3, the contextual mood manipulation was expected to alter (through misattribution) anticipatory affective responses to the consumption episode without altering its evaluative content. In the two "open-process" conditions (conceptually replicating Experiment 3), subjects were "given" either a consummatory motive or an instrumental motive for Bassili (1989) and Fazio, Lenn and Effrein (1984) applied a similar logic to examine spontaneous inference making and spontaneous attitude formation.

PAGE 109

100 making their decisions. They did not receive any instruction about how to make their decision. Again, it was predicted that subjects with a consummatory motive would be influenced by their contextual mood as they sought to inspect how the episode "feels." In contrast, subjects with an instrumental motive would be little affected by their contextual mood. In the two baseline conditions, all subjects were provided with a consummatory motive. In addition, they were encouraged to reach their decisions by relying on one of two specific processes. In the baseline-attitude formation condition, subjects were encouraged to follow an expectancyvalue process: accessing salient beliefs and combining their evaluative implications into an overall evaluation of the episode. In the baseline-experiential condition, subjects were instructed to follow processes that mimic those hypothesized in affect recruitment. If, as hypothesized, consumers with consummatory motives rely spontaneously on an affect recruitment process, the decision outcomes of (uninstructed) subjects in the open-process-consummatory motive condition should resemble the decision outcome of baseline subjects explicitly instructed to mimic affect recruitment processes. By the same logic, if it is observed that the decision outcome in the open-process-consummatory condition differs from that in the baseline attitude formation condition, it can be deduced that an expectancy-value attitude formation process does not adequately represent the process underlying decisions of consumers with a consummatory motive. The hypotheses of this experiment were thus as follows:

PAGE 110

101 HI: Subjects' preexisting mood will have comparable effects on the decision outcome in the open-process-consummatory condition and in the baseline-experiential condition: in both conditions, a positive preexisting mood will result in more favorable behavioral intentions than a negative preexisting mood. H2: The mood effect observed in the in the open-process-consummatory and the baseline-experiential conditions will be significantly smaller in the open-processinstrumental and in the baseline-attitude formation conditions. TABLE 6.1

PAGE 111

102 manipulated the instructions provided to subjects. In the two open-process conditions, subjects were provided with a specific motive (consummatory or instrumental) for engaging in the consumption activity and received no direction about how to make their decisions. In the two "baseline" conditions, all subjects were provided a consummatory motive and were encouraged to make their decision by relying on a specific process. In the baseline -attitude formation condition, subjects were instructed to judge the episode through an expectancyvalue process, whereas in the baseline-experiential condition subjects were instructed to mimic affect recruitment processes. Procedure Subjects participated in groups of 1-9. They were told that they would be completing a series of unrelated tasks. First, subjects completed the same mood induction procedure as in Experiment 3. They were asked to write a detailed description of a pleasant or unpleasant personal experience. The purpose of the task was allegedly to identify items for a scale. Next, subjects received a second questionnaire in which they were asked to assess their intention of going to a movie within the next week. Unlike in Experiment 3, the movie was not specified. After reporting their behavioral intentions, subjects were asked to provide a detailed description of everything that went through their minds as they were thinking about whether to go see a movie. Next, as a manipulation check for the mood manipulation, subjects were asked to report their current feelings. The remaining part of the questionnaire consisted of other manipulation checks and demand characteristics checks. The experimental factors (mood and

PAGE 112

103 instructions) were manipulated within session, except for the baseline-attitude formation condition, which was run in separate sessions. 2 Contextual Mood The mood manipulation was similar to the one used in Experiment 3. Under the guise of collecting items for the development of a scale, subjects had to provide a detailed and vivid account of a pleasant or unpleasant experience. Whereas in Experiment 3 the affective content of the experience to be described was relatively generic (an event that made them feel "really bad" or "really good"), in this experiment the affective content of the experience to be described was more specific. In the negative mood condition, subjects had to describe an event that made them feel "really angry, irritated, or annoyed." In the positive mood condition, subjects had to describe an event that made them feel "really happy, joyful or cheerful." Pretest had suggested that anger-like and joy-like emotions would elicit a stronger mood effect than other negative (e.g., sadness) or positive (e.g., pride) feelings. Instructions Open-process-consummatory motive condition Subjects in this condition received the following instructions: 2 As explained later in the chapter, the baseline attitude formation condition involved a task slightly more complicated than the other conditions. There was a concern that providing additional instructions at the time of the task might disrupt the effect of the mood manipulation. To avoid this potential problem, subjects in this condition were given an example of how to complete the task at the beginning of the session, i.e. before the mood induction procedure. This different timing of the tasks required that this condition be run in separate sessions. (Mood was still manipulated within session.)

PAGE 113

104 This questionnaire is about how much college students enjoy various kinds of leisure activities. After a long week of school and hard work, students often want to just enjoy themselves and have a good time. After all, everyone needs some time off. One way of having a good time would be to go to the movies. How much do you intend to go see a movie within the next week? (Place a check on the scale.) Open-process-instrumental motive condition Subjects in this condition received the following instructions: This questionnaire is a pretest for a study that we will be running in two weeks from now (there will be several sessions on several days). The study (one hour) involves writing a short essay. Participants will receive 10 dollars. In addition, the 40 most interesting essays will be rewarded by an extra 20 dollars each. Since we expect 80 participants, this means that half of the participants (the ones who wrote interesting essays) will receive a total of 30 dollars ($10 + $20). Do you think you would be interested in participating in that study? Circle yes or no. YES NO The essay can be on either: (1) a movie seen within the next week; (2) a news article published within the next week; or (3) or a book read within the last two years. How much do you intend to go see a movie within the next week? (Place a check on the scale.) It was expected that most subjects would express interest in participating in the subsequent study and would assess their intention of going to a movie as a means of being able to write a successful essay. Because subjects could participate without seeing a movie (e.g., by having read a book in the last two years), it was expected that the monetary incentive would not be perceived as being directly associated with the movie episode, unlike in Experiment 3. The dissociation between the monetary incentive and

PAGE 114

105 the consumption episode itself reduced the chance of a ceiling effect problem since the actual cost of going to a movie was not lower in this condition than in the other conditions. 3 In addition, unlike in Experiment 3, going to the movie was not a requirement for participating in the subsequent study. Baseline-attitude formation condition Subjects in this condition were exposed to the same consummatory instructions as subjects in the open-process-consummatory motive condition ("This questionnaire is about how much college students enjoy various kinds of leisure activities..."). In addition to these consummatory motive instructions, subjects were instructed to assess their intention of going to a movie by using an expectancy-value attitude formation process: Think about why seeing a movie within the next week would be a good thing to do or why it might not be a good thing to do. List these pros and cons in the space below. Indicate the order in which each pro or con has occurred to you by writing a number (1,2,3..) next to it. Subject were provided two columns to list their positive and negative thoughts about the episode. They were then asked the same question as in the other conditions. Baseline-experiential condition Again subjects were exposed to the same consummatory motive instructions as subjects in the open-process-consummatory motive condition. In addition to these instructions, they were instructed to rely on processes that mimic affect recruitment to assess their behavioral intentions: Try to "picture" yourself at a movie. Try to imagine as vividly as you can what it would be like to be at this movie (the actors, the theater, the atmosphere, the 3 In Experiment 3, subjects in the instrumental conditions were told that the movie ticket would be reimbursed and that they would receive an additional four dollars.

PAGE 115

106 sound, etc.). Focus in particular on your feelings as you imagine yourself at this movie. Does it feel "good" or does it feel "bad?" Is it pleasant or unpleasant? Results Demand and Manipulation Checks Subjects were debriefed about the objectives of the study after the dependent measures were collected. Although no one could articulate the hypotheses being tested, two subjects expressed suspicion that the mood manipulation was related to the second questionnaire. These two subjects' data were discarded, leaving 136 observations. Consummatory vs. instrumental Motive The effectiveness of the motive manipulation was assessed with an independent sample of 25 subjects from the same student population. After making their decisions, these subjects were asked which of two statements better reflected the perspective from which they assessed their intention of going to a movie. They recorded their responses on a 7-point semantic differential scale anchored by "I focused on how enjoyable or unpleasant this experience would be" and "I focused on whether seeing it would help me attain a specific objective." Low values indicated a consummatory motive, whereas high values indicated an instrumental motive. As expected, subjects presented with the instrumental instructions (M = 3.53) were more likely to base their decisions on whether the episode would "help them attain a specific objective" (and less likely to base their decisions on "how enjoyable or unpleasant this experience would be") than subjects presented with the consummatory instructions (M = 2.08, F(l,23) = 6.31, one-tailed p < .01, omega-square = .17).

PAGE 116

107 To assess potential confounds of this manipulation, subjects were also asked (1) How important it was for them to be accurate in their assessment and (2) How difficult this decision was. Neither of these measures was affected by the motive manipulation (both Fs < 1), suggesting that the manipulation did not induce different levels of involvement and task difficulty. Attitude formation vs. experiential baselines To assess the effectiveness of the processing instruction manipulation in the baseline conditions, subjects were asked which of two statements better reflected how they assessed their intention of going to the movies. Again, they recorded their responses on a 7-point semantic differential scale anchored by "I relied on whether it felt good or bad picturing myself at the movie" and "I considered the positive and the negative aspects of going to see a movie." Subjects in the attitude formation conditions were slightly more likely to agree with the latter statement (M = 5.60) than subjects in the experiential conditions (M = 5.09). However, this effect was not significant (F < 1). As shall be seen, however, it appears that subjects in the baseline-experiential condition djd rely more on their current feelings than subjects in the attitude formation condition. It is possible that such processes are difficult to verbalize and that subjects failed to recognized that one of the endpoints of the scale was a rewording of the process they were encouraged to follow. Alternatively, because of social desirability, subjects in the baseline-experiential condition may have been reluctant to admit that they relied on their feelings to make their decisions. Mood It has been observed that measuring mood before the dependent variable of interest may weaken the effect of the mood manipulation by making salient to subjects

PAGE 117

108 that their mood had been manipulated. Therefore, the mood manipulation check was administered after subjects had reported their behavioral intentions and completed their protocols. Subjects were asked "How are you feeling right at this moment?" Their mood was assessed by four seven-point scales anchored by in a good/bad mood; joyful/sad; happy/unhappy; and angry /pleased (alpha = .90). That is, unlike in Experiment 3, where subjects were asked to reflect back on how they were feeling after having described their personal experiences, in this experiment subjects were asked how they were currently feeling after having completed the main dependent measures. A 2 (mood) x 4 (instructions) ANOVA of these mood reports did not reveal the expected main effect of the mood manipulation. Self-reported mood in the positive mood condition (M = 4.92) was not significantly more pleasant than in the negative mood condition (M = 4.68, F( 1,127) = 1.33, p = .25). It seems likely, however, that these reports, administered after the several dependent measures, underestimate the actual strength of the mood manipulation. As reported below, several mood effects were in fact uncovered. Behavioral Intentions Intentions of going to a movie within a week were analyzed through a series of planned interaction contrasts in a 2 (mood) x 4 (instruction) ANCOVA, summarized in Table 6.3. Because substantial subject variability was anticipated, three covariates were entered in the analysis: subject's self-reported mood coming into the study, the number of movies seen in the last six months, and subjects' self-assessed time and monetary

PAGE 118

109 constraints. 4 The raw means and the least-square means (adjusted for the covariates) are reported in Table 6.2. TABLE 6.2 INTENTIONS OF GOING TO SEE A MOVIE LEAST SQUARE MEANS (RAW MEANS) Conditions Open-processOpen-processConsummatory Instrumental BaselineBaselineAttitude Experiential Formation Positive mood Negative mood Difference (positivenegative) 5.16(5.44) 5.62(5.53) 4.14(3.94) 5.61(5.38) 1.02(1.50) 0.01(0.15) 4.89 (5.00) 7.34 (7.06) 5.42(4.53) 5.30(6.10) -0.53 (0.47) 2.04 (0.96) Again, if consumers with consummatory motives are likely to "spontaneously" inspect their feelings to make decisions about consumption episodes, the mood manipulation should exert a significant influence on behavioral intentions in the openprocess consummatory condition. That influence should resemble the one observed in the baseline experiential condition, where subjects were explicitly instructed to mimic affect recruitment processes. If consumers with instrumental motives are not as likely to "spontaneously" inspect how an episode makes them feel, the mood manipulation should not influence behavioral intentions in the open-process instrumental condition. The study being run during the two weeks preceding the final exams, much variability was anticipated with respect to the time constraint that subjects would experience. To assess the severity of this constraint, subjects were asked "How much would you have to give up (time, money, etc.) to see a movie over the next week?" They answered on a nine-point scale anchored by "I wouldn't have to give up much" and "I would have to give up a great deal."

PAGE 119

110 Futhermore, if the affect recruitment process is empirically distinct from an expectancyvalue process, the mood manipulation should not have a similar influence on behavioral intentions in the baseline attitude formation condition as in the open-process consummatory conditions. These predictions were reframed in terms of three critical contrasts. It was predicted that subjects in the open-process-consummatory condition would be influenced by their mood in the same way as subjects in the baseline-experiential condition. To test this prediction, a mood x instruction interaction contrast was calculated by assigning the following weights: +1, 0, 0, -1 to the open-processconsummatory, open-process-instrumental, baseline-attitude formation, and baselineexperiential conditions, respectively. As predicted, this contrast revealed a main effect of mood (F(l,125) = 7.24, p < 0.01) and no mood x instruction interaction (F < 1), suggesting that the mood effect was comparable in the open-process-consummatory and the baseline-experiential conditions. This finding is consistent with Hypothesis 1. Similarly, it was predicted that both subjects in the open-process-instrumental condition and subjects in the baseline-attitude formation condition would be little influenced by their mood. A second mood x instruction interaction contrast was calculated by assigning the following vector of weights: 0, 1, -1, to the open-processconsummatory, open-process-instrumental, baseline-attitude formation, and baselineexperiential condition, respectively. This contrast revealed neither a main effect of mood nor a mood x instruction interaction (both F's < 1), suggesting that the mood effect was comparably nonsignificant in these two conditions.

PAGE 120

Ill Table 6.3 BEHAVIORAL INTENTIONS: OVERALL ANCOVA AND PLANNED CONTRASTS Overall ANCOVA

PAGE 121

112 A third interaction contrast was constructed by pooling the open-processconsummatory condition with the baseline-experiential condition and pooling the openprocess-instrumental condition with the baseline-attitude formation condition (i.e., using the following vector of weights: 1,-1,-1, 1). As expected, a significant mood x instruction interaction was uncovered (F(l, 125) = 4.96, p < 0.03), reflecting the fact mentioned above that the mood manipulation had a significant effect in the open-processconsummatory and baseline-experiential conditions but not in the open-processinstrumental and baseline-attitude conditions. This finding is consistent with Hypothesis 2. Protocol Content Subjects' protocols were segmented into verbalizations ." Each verbalization consisted of a separate unit of thought or feelings (cf. Ericsson and Simon 1993). Verbalizations consisted mostly of complete sentences and grammatical clauses separated by transition words such as: and, then, but, if, because, etc. Subjects produced an average of 4.88 verbalizations. Two judges, blind to the experimental conditions, independently assigned each these verbalizations into a 33-category scheme. Inter-judge agreement was 73%. Disagreements were resolved through discussion. The low frequency categories then were merged for subsequent analysis, resulting in a 16-category scheme detailed in Appendix G. The first 10 categories consisted of five types of verbalizations with clear behavioral implications (positive or negative) for the consumption episode. The last six categories consisted of verbalizations with no clear evaluative implications.

PAGE 122

113 The number of verbalizations falling in each category was submitted to a 2 x 4 ANOVA. These analyses revealed little difference in terms of the thought and feeling content of the protocols. Appendix H lists the average number of verbalizations per content category and by experimental condition. Only the significant effects are reported here. Consequences The consequence category was the most substantial (3 1 % of the verbalizations). It included all verbalizations with clear evaluative implications for the episode and no intrinsic affective content (e.g., "I wouldn't be able to take my son"; "[The reason was that] I needed to take this girl out to try get over my girlfriend.)" The ANOVA revealed a main effect of mood (F(l 125) = 4.56, p < .04) and a main effect of instructions (F(3,125) = 2.86, p < .04). Subjects in the negative mood condition reported more consequence verbalizations (M = 1.761) than subjects in the positive mood condition (M = 1.318). It was expected that consequence verbalizations would be more frequent in the open-process-instrumental and baseline-attitude formation conditions than in the open-process-consummatory and baseline-experiential conditions. Contrary to this prediction, these verbalizations were most frequent in the open-processconsummatory condition (M = 2.029), followed by the baseline-attitude formation condition (M = 1.613). The subjects in the baseline-experiential condition, as expected, reported fewer of these verbalizations (M = 1.2). Unexpectedly, subjects in the instrumental condition also reported relatively few consequence verbalizations (M = 1.312). Although the overall mood x instruction interaction was not significant (F(3, 125) = 2.01, p = .12), it appears that relatively high frequency of consequence verbalizations

PAGE 123

114 in the open-process-consummatory and baseline-attitude formation conditions was driven by subjects in the negative mood condition (M = 2.765 and 3.214 for open-processconsummatory and baseline attitude formation, respectively). Analysis of the consequence verbalizations broken down according to their evaluative implication indicates that these effects are mainly attributable to the consequence verbalizations with negative implications. Subjects reported more than twice as many negative consequence verbalizations (M = 1.113) as positive consequence verbalizations (M = 0.429). Informal inspection of the negative consequence verbalizations suggests that they consisted mostly of subjects' preoccupations with their upcoming end of the class term (e.g., "I thought about the fact that finals are coming up;" "I have two papers that are due next week"). 5 An ANOVA of these verbalizations uncovered effects that parallel the ones obtained with the total number of consequence verbalizations (i.e. positive and negative). There was a main effect of instruction (F(3,125) = 5.02, p < .01). Subjects in the open-process-consummatory conditions reported the highest number of negative consequence verbalizations (M = 1.771). The next highest number of negative consequence thoughts was reported by subjects in the baseline-attitude formation condition (M = 1.032). Subjects in the open-processA fraction of negative consequence verbalizations consisted of negative statements about available movies (e.g., "There is not really a movie I want to see"; "Everything out is garbage"). The frequency of these statements differed across instruction conditions (F(l,126) = 3.21, p < .03). Subjects in the open-process consummatory condition were more likely to report such verbalizations (M = .286) than subjects in the open-process instrumental (M = .156), baseline experiential (M = 1 14), and baseline attitude formation (M = .031) conditions.

PAGE 124

115 instrumental (M = .844) and in the baseline-experiential conditions (M = .771) reported fewer or such thoughts. In summary, contrary to expectations, subjects in the open-process-consummatory and baseline-attitude formation conditions reported more nonaffective consequence verbalizations than subjects in the other two conditions. This effect was mainly the result of the number of negative consequence verbalizations generated by open-processconsummatory and baseline-attitude formation subjects in a negative mood, reflecting their preoccupations with their final examinations. Feelings This category includes all verbalizations with an affective/experiential content: (1) expressions of desire (e.g., "I want to see the movie 'Above the Rim'"; "I'm dying to watch a movie soon"), (2) sensory verbalizations (e.g., "I also thought about how cold a movie theater is"), (3) cravings (e.g., "I've been wanting to see it for a couple of weeks now"), (4) current feelings (e.g., "I got a little excited when one came to mind that I wanted to see"), (5) expected feelings (e.g., "How relaxing it would be to just kick back with some popcorn"), (6) reinstantiation of previous affective experience and affectively charged attitudes (e.g., "I was thinking about... the last time I went to see a movie and how much I liked it"). This category represents 17 percent of all reported verbalizations. An ANOVA of the feeling verbalizations revealed a main effect of instruction (F(3,125) = 5.56, p < .002): feeling verbalizations were reported more frequently in the two (processinstructed) baseline conditions (M attfonn = 1.314; M exper = 1.194) than in the two (process-uninstructed) open-process conditions. Subjects in the open-process-consummatory condition reported twice as many feeling verbalizations

PAGE 125

116 (M consum = .629) as subjects in the open-process-instrumental condition (M instru = .313); however, this effect was not significant (F(l, 125) = 1.25, p = .27). The analysis was pursued by breaking down these feeling verbalizations according to their evaluative implication. The number of feeling verbalizations with positive implication was slightly higher in the positive mood condition than in the negative mood condition (F( 1,125) = 3.39, p = .07). There was again a main effect of instruction (F(3,125) = 3.72, p < .02). Positive feelings were most frequently reported in the baseline-experiential condition (M = 1.114), followed by the baseline-attitude formation condition (M = 0.871). Again, there were more positive feelings in the open-process-consummatory condition (M = 0.600) than in the open-process-instrumental condition (0.281), but this effect was not significant (F( 1,125) = 1.47, p = .23). Analysis of the negative feelings revealed a main effect of instructions (F(3,125) = 3.51, p < .02). There were more negative feelings in the baseline-attitude formation (M = 0.323) and baseline-experiential (M = 0.200) conditions than in the open-process-consummatory (M = 0.029) and openprocess-instrumental (M = 0.031) conditions. Conclusions This category (8.5 % of the verbalizations) consists of expressions that reflect mostly the outcome of subjects' decision process (e.g., "So I figured I probably won't have the time to see a movie until after April 28, 1994"). An ANOVA showed a main effect of mood (F(l,125) = 5.33, p < .03): subjects in the negative mood condition verbalized more conclusions (M = .537) than subjects in the positive

PAGE 126

117 mood condition (M = .288). This effect was more pronounced for negative conclusions (F( 1,125) = 7.40, p < .01) than for positive conclusions (F < 1). In summary, the content of subjects' protocols did not show much variation across experimental conditions. It appears that the provision or nonprovision of processing instructions had a stronger influence on the content of the protocols than the motive or the direction of the processing instructions. Correlational Analysis Three main types of decision inputs were assessed in the study. Two of these three types of inputs were distinguished in the protocol analysis: (1) consequence verbalizations; and (2) feeling verbalizations. This distinction is consistent with that suggested by Triandis (1980, see the relevant discussion in Chapter 3). However, it is possible that feeling verbalizations (e.g., "I personally love going to movies with friends"; "I would enjoy going") do not adequately capture the feelings experienced by subjects as they were making their decisions. This feeling experience may be better reflected by subjects' self-reported feelings after they had reported their decisions and completed the protocols. This analysis examines how these three types of inputs jointly account for the behavioral intentions reported in the different experimental conditions. A "net consequence" index was computed by subtracting the number of negative consequence verbalizations from the number of positive consequence verbalizations reported by each subject. A similar "net feeling" index was computed by subtracting the number of negative feeling verbalizations from the number of positive feeling verbalizations. Both indices should be positively related to subjects' behavioral

PAGE 127

118 intentions. Self-reported post-decision feelings should also be positively related to behavioral intentions if subjects did rely on their feeling experience These three indices were entered as predictors of behavioral intentions in a series of multiple regressions. To test differences of regression slopes across conditions, dummy variables and their interactions with each predictor were included in each regression equation. The results of these regressions are reported in Table 6.3. In Equation 1, the pooled open-process consummatory and baseline experiential conditions (dummy instruction = 1) were compared to the pooled open-process instrumental and baseline attitude formation conditions (dummy instruction = 0). Consistent with the ANOVA results reported earlier, Equation 1 suggests that subjects in the open-process consummatory and baseline-experiential conditions placed greater weight on their current feelings than subjects in the open-process instrumental and baseline-attitude formation conditions (t = 2.76, p < .01). There was surprisingly no difference in the weight of feeling verbalizations (t < 1). The regression weight of the consequence content was slightly smaller in the open-process consummatory and baseline experiential conditions than in the open-process instrumental and baseline attitude formation conditions, but this effect was not significant (t = -1.21, p = .23). In Equation 2, the two open-process conditions were compared to each other (dummy =1 for consummatory motive = 1; dummy = for instrumental motive). The analysis suggests that the consequence content received a smaller weight among subjects with a consummatory motive than among subjects with an instrumental motive (t = 1.93, p = .06). In contrast, current mood received a stronger weight among subjects

PAGE 128

119 with a consummatory motive than among subjects with an instrumental motive (t = 1.96, p = .05). Again, there was no difference in the weight of feeling verbalizations (t < 1). Table 6.3 REGRESSION ANALYSIS OF BEHAVIORAL INTENTIONS AS A FUNCTION OF TYPES OF CONTENTS AND EXPERIMENTAL CONDITIONS Equation 1 (R 2 = 0.445): (Consummatory + Experiential) vs. (Instrumental + Attitude Formation)

PAGE 129

120 TABLE 6.3-CONTINUED REGRESSION ANALYSIS OF BEHAVIORAL INTENTIONS AS A FUNCTION OF TYPES OF CONTENTS AND EXPERIMENTAL CONDITIONS Equation 2 (R 2 = 0.449): Open-process consumm. vs. Open-process instrum.

PAGE 130

121 Discussion Inferences from the Behavioral Intention Data This experiment relied on a logic of process baselines to compare the hypothesized affect recruitment process with an expectancy-value process as models of how consumers make decisions when they have consummatory or instrumental motives. In the open-process consummatory and baseline experiential conditions, subjects who had been induced into a good mood expressed more favorable behavioral intentions than subjects who had been induced into a bad mood. This effect was not observed either in the open-process instrumental condition or in the baseline attitude formation condition. This pattern of results suggests two inferences. The parallel mood effects obtained in the open-process consummatory and baseline experiential conditions suggests that, with respect to the reliance on feelings experienced during the decision process, the process imposed on subjects in the latter baseline condition is a plausible paramorphic representation of the process consumers with consummatory motives may (spontaneously) follow when making decisions about consumption episodes. Subjects in the open-process consummatory condition were influenced by their mood as if they were judging the episode by "seeing how it feels". By the same logic, the absence of a parallel mood effect in the baseline attitude formation condition suggests that an expectancy-value model may not fully account for (i.e., describe) how consumers with a consummatory motive make decisions about consumption episodes. Finally, the absence of a parallel mood effect in the open-process instrumental condition, which replicates Experiment 3's main finding, suggests that there may be a

PAGE 131

122 genuine process difference in how consumers make decisions under instrumental versus consummatory motives. The weight attached to feelings experienced during the decision process may be stronger when a consummatory motive increases the information value of anticipatory affective responses. Alternative Explanations Confidence in these inferences requires that alternative explanations be ruled out. Such explanations have to jointly account for: (1) the presence of mood effects in the open-process consummatory and baseline experiential conditions and (2) the absence of such effects in the open-process instrumental and baseline attitude formation conditions. Ceiling Effects The absence of experimental effects is often attributed to range effects (i.e., floor or ceiling). As reported earlier, unlike in Experiment 3, the instrumental motive induction avoided making the cost of the episode systematically lower than in the other conditions. As a result, behavioral intentions (on a 1-9 scale) were not particularly high in the open-process instrumental (raw mean = 5.45; sd = 2.61) and in the baseline attitude formation (raw mean = 4.88; sd = 2.45) conditions. A ceiling effect was probably not the cause for the absence of effect in these conditions, especially since behavioral intentions were higher (raw mean = 6.41; sd = 2.23) in the baseline experiential conditions, where a mood effect was observed. Disruptive Task Instructions in the baseline attitude formation conditions intended to encourage an expectancy-value process The task itself (listing pros and cons), not the process, may have disrupted the mood effect. As reported earlier (Footnote 2), to reduce the disruptive effect of this task, subjects in this condition were

PAGE 132

123 provided additional instructions and an example, before the mood manipulation. All subjects indicated being comfortable with this task after these initial instructions. Overkill A related explanation is that the task may have induced subjects to rely on beliefs that they would not have accessed otherwise. These additional beliefs may have mitigated an effect that would have occurred had subjects relied on the most salient beliefs only. 6 To help rule out this explanation, subjects in the baseline attitude formation were asked to mention the order in which the pros and cons of the episode occurred to them. If the mood effect was limited to the first few beliefs accessed, there should be traces of mood congruency in the first few pros and cons listed. Subjects listed an average of 4.97 pros and cons. The valence of the first pro or con listed was examined across mood conditions. In the negative mood conditions, 10 subjects out of 14 (71.4%) first listed a pro. In the positive mood condition, the proportion was 12 subjects out of 18 (66.7%). There was no evidence the first thought accessed depended on contextual mood (x 2 (1 df) = 0.083, p = .773). Additional analyses examined the evaluative implication of the first few pros or cons listed. Three indices were computed by assigning +1 to pros and -1 to cons and summing over the first two, the first three, and the first four pros or cons listed. None of these indices was significantly affected by the mood factor (largest F(l,30) = 1.28. p = .27). In summary, it is unlikely that lack of mood effect in the baseline attitude 6. This explanation would be consistent with the proposition that mood effects on judgments are mediated by increased accessibility of mood congruent thoughts (Isen 1984; Isen et al. 1978).

PAGE 133

124 formation condition resulted from inducing subjects to access more beliefs than they would have otherwise. Inferences from the Process Measures Although feeling verbalizations were more frequent in the open-process consummatory condition than in the open-process instrumental condition, this effect was not significant. In fact, feeling verbalizations were most frequent in the two baseline conditions. Unexpectedly, subjects in the open-process consummatory condition reported more consequence verbalizations than subjects in the other conditions. In retrospect, this finding may not be that surprising. These consequence verbalizations mostly expressed subjects' concerns with the end of the class term. The consummatory motive manipulation may have inadvertently called subjects' attention to their upcoming examinations and assignments by suggesting that "After a long week of hard work, students often want to enjoy themselves and have a good time. After all everyone needs some time off." In contrast, the manipulations used in the baseline experiential and open-process instrumental conditions (where fewer consequence verbalizations were expressed) may have distracted subjects away from the upcoming end of the term. The content of subjects' protocols revealed few other experimental effects. When effects were uncovered, they reflected more the provision or nonprovision of processing instructions than which motive or which processing instruction was provided. The changes of representations that were hypothesized under different decision heuristics (e.g., reference to abstract rules versus exemplars; conclusions [possibly indicative of "reasoned action"] versus generalized feelings; consequence thoughts versus experiential

PAGE 134

125 thoughts) may be too subtle to be captured by the coding scheme used in this study, or perhaps more generally by the protocol method. Alternatively, it is plausible that different decision heuristics (e.g., affect recruitment versus expectancy-value) have a stronger influence on how decision inputs are used than on which type of input is likely to be accessed and how it is represented Two findings are consistent with a utilizationrather-than-representation interpretation. First, although subjects in the open-process consummatory condition may have thought more about the consequences (especially negative ones) of the episode, these subjects appear to have attributed a lesser weight to these consequences than subjects with an instrumental motive. Second, subjects in the open-process consummatory condition weighed their current feelings more heavily than subjects in the open-process instrumental condition. The weight of feeling verbalizations did not differ between the two motive conditions. The fact that motive moderated the weight of current feelings but not of feeling verbalizations may indicate that in affect recruitment it is the experience of affect that is recruited, not its more abstract (and verbalizable) representation. Limitations It should be stressed that these "paramorphic" inferences apply only to how subjects in this experiment were influenced by their contextual mood. Whereas it is appropriate to state that subjects in the open-process consummatory condition were influenced by their mood as if they were relying on how the episode "felt," it would be premature to conclude that the processes followed by the two groups of subjects were

PAGE 135

126 comparable in every respect. One could argue that although subjects in these two conditions were influenced by their mood in a similar way, they may still have followed different processes. The issue then becomes: What process did subjects in the openprocess consummatory condition actually follow, such that it would also explain the presence of a mood effect? Although the data discussed above make an expectancyvalue process rather implausible, they cannot positively establish that this process was indeed affect recruitment. Stronger evidence of affect recruitment would have been obtained if the protocol analysis had revealed strong parallelism between (1) the open-process consummatory condition and the baseline experiential condition, and between (2) the open-process instrumental condition and the baseline attitude-formation condition. The expected parallelism was not observed. 7 Both Experiments 3 and 4 were based on the rationale that a contextual mood manipulation would alter anticipatory affective responses to a consumption episode without changing the evaluative content of this episode. In other words, the manipulation deliberately confounded subjects' mood with their anticipatory responses to the episode. This confound was required to show the unique information value of feelings experienced during the decision process. The drawback of this manipulation is that it cannot Note, however, that studies relying on the logic of process baselines (e.g., Bassili and Smith 1986) usually focus on the parallelism (or lack of parallelism) of process outcomes -a strategy similar to that followed here with the behavioral intention data. They do not attempt to measure the inferred process. It is not clear what "process measures" actually tap when a specific process is imposed on the subjects.

PAGE 136

127 document the proposition that anticipatory affective responses are indeed experienced. This issue is addressed in Experiment 5.

PAGE 137

CHAPTER 7 EXPERIMENT 5: THE MISATTRIBUTION OF ANTICIPATORY AFFECTIVE RESPONSES Purpose and Predictions The framework suggests that to make decisions about consumption episodes, consumers may attempt to anticipate their future experience by (1) accessing or constructing a concrete representation of the episode, (2) instantiating anticipatory affective responses to this representation, and (3) using these instantiated feelings as a decision input. A central proposition (Proposition 2) is that in affect recruitment, anticipatory affective responses are indeed instantiated or experienced. This proposition could not be tested in Experiments 3 and 4 because the hypothesized anticipatory affective responses were deliberately confounded with subjects' mood. The purpose of this study is to test the instantiation of anticipatory affective responses in affect recruitment. The study also attempts to show that if consumers incorporate current feelings into their decisions (as they "mistakenly" did in Experiments 3 and 4), it is not because they attach information value to their mood per se ("How am I feeling now"), but because they attribute part of these feelings to the object of the decision ("How do I feel about it "). Stated differently, affect recruitment is the 128

PAGE 138

129 recruitment of (anticipatory) responses to the episode, not just the inspection of one's current mood. To provide evidence that (1) consumers do experience anticipatory affective responses to consumption episodes, and (2) it is these anticipatory feelings, not their current mood, that they are trying to incorporate into their decisions, this study relies on the following reasoning. First, if anticipatory affective responses are indeed experienced, they should be phenomenologically comparable to current valenced feelings (e.g., moods). Second, consumers should seek to subtract from their judgments any part of these feelings that is associated with a cause independent of the object of judgment (e.g., Martin, Seta, and Crelia 1990). Suppose that consumers are asked to make a decision about a pleasant consumption episode and that this decision is driven by consummatory motives (e.g., attending a high school reunion). It is hypothesized that consumers may construct concrete representations of the episode (e.g., imagining themselves chatting and laughing with old-time friends), which would instantiate pleasant anticipatory affective responses (e.g., cheerfulness). Suppose now that a plausible alternative source for these pleasant feelings is made salient (e.g., a popular musical tune). If anticipatory affective responses are indeed experienced as pleasant feelings, the consumers may (mis)attribute part of their anticipatory affective responses to this alternative source. According to the set/reset model of social judgment (Martin 1986; Martin et al. 1990), the consumers may subtract the misattributed feelings from their assessment of the consumption episode. As a result, the outcome of their decision may become less favorable than that of consumers who are

PAGE 139

130 not led to misattribute their anticipatory feelings. A mirror effect is predicted if consumers are asked to make a decision about an unpleasant episode and a plausible alternative cause for the unpleasant feeling experience is made salient. Subtraction of the misattributed portion of the unpleasant anticipatory affective responses would result in more favorable behavioral intentions. Two inferences would follow from such subtraction effects. First, one can infer that a particular decision input has indeed been accessed. There was "something" to be subtracted. Second, one can infer that this input is phenomenologically similar to the experience of valenced feelings. Otherwise, consumers would not misattribute part of this input to the alternative source of valenced feelings. HI Consumers making a decision about a pleasant (unpleasant) consumption episode for consummatory reasons will be less (more) favorable toward this episode, if a plausible alternative cause for their feelings is made salient, than if this alternative cause is not made salient. It has also been proposed that affect recruitment is less likely when decisions are guided by instrumental motives than when they are guided by consummatory ones. If there is less instantiation of anticipatory feelings in decisions guided by instrumental motives, the outcome of such decisions should be less sensitive to the type of attribution that consumers make about their feelings. H2 Decisions guided by instrumental motives will be less affected by whether or not a plausible alternative cause for their feelings is made salient.

PAGE 140

131 Method Overview of the Design Seventy-one undergraduates at the University of Florida participated in the study in exchange for extra course credits. They were asked to report their intention to engage in different consumption episodes described to them. In the two experimental conditions, a musical piece was played in the background while subjects were assessing their intentions. In a separately run control condition, subjects assessed their behavioral intentions with no musical background. 1 Three factors were manipulated. The first factor, manipulated between subjects, created plausible alternative attributions for feelings experienced when making the decisions. In the high salience condition, it was made salient that the background music could affect, positively or negatively, subjects' feelings. In the low salience condition, the music's potential influence on subjects' feelings was not made salient. Again, there was no music in the control condition. As detailed below, the other two factors were manipulated within subjects, across consumption episodes (see Table 7.1). It was expected that two consumption episodes would be judged primarily on the basis of consummatory motives, whereas the other two episodes would be judged primarily on the basis of instrumental motives. Two consumption episodes were expected to elicit low behavioral intentions, whereas the other The control condition was run after the two experimental conditions to aid in interpreting the results of the main study. Control subjects were selected from the same student population as were experimental subjects. Although experimental and control subjects did not differ in terms of age, classification, and other background factors, there were significantly more females in the control condition than in the experimental conditions. As a result, gender was entered as a covariate in the main analysis.

PAGE 141

132 two episodes were expected to elicit moderately high behavioral intentions. In the experimental conditions, the low behavioral intention episodes were presented against an unpleasant musical background, whereas the high behavioral intention episodes were presented against a pleasant musical background. The order of the musical backgrounds and the order of the motives were counterbalanced across subjects. TABLE 7.1 EXPERIMENT 5: SUMMARY OF THE WITHIN SUBJECT MANIPULATIONS Episode content

PAGE 142

133 of episodes, one was expected to elicit relatively low behavioral intentions (the folkloric dance show and the tax manual), whereas the other was expected to elicit moderately high behavioral intentions (the high school reunion and the computer course). The four episode descriptions are reported in Table 7.2. An additional purpose of the pretests was to assess the types of feelings, if any, that were associated with the various episodes. It was found that the folkloric dance show was mainly associated with negative anticipated feelings such as "feeling bored." In contrast, the high school reunion was associated with pleasant feelings such as happiness and excitement. It was predicted that the two decisions guided by consummatory motives (high school reunion and folkloric dance) would lead to the instantiation of anticipatory affective responses-pleasant for the reunion episode and unpleasant for the folkloric dance show episode. Subjects in the high salience condition were expected to (mis)attribute part of their feelings to the musical background and subtract the misattributed feelings from their decisions. As a result, it was predicted that subjects in the high salience condition would report less favorable behavioral intentions toward the high school reunion episode and more favorable behavioral intentions toward the folkloric dance episode than subjects in the low salience condition. It was further predicted that decisions about the two instrumental episodes (computer course and tax manual) would not be affected by the attribution manipulation, because they are less likely to involve the instantiation of anticipatory affective responses. An expectancy value process would be

PAGE 143

134 more likely, since instrumental motives imply a focus on extrinsic rewards (e.g., becoming more knowledgeable about computers). Procedure To ensure a minimum level of distraction, subjects were run in small sessions of 1-8 (average = 4) and were not allowed to speak to each other. They were assigned well-spaced seats in an area enclosed by large panels. The experimenter stayed out of sights throughout most of the session. The first part of the session in the experimental conditions was intended to manipulate subjects' attributions. Three musical segments were played (through a tape player) for 40 seconds each. Subjects were asked to rate each segment along several dimensions, which differed across attribution conditions. In the second phase of the experiment, it was explained that the experimenter was interested in "students' ability to make plans when there is background music, such as in department stores, restaurants, waiting rooms, etc." Subjects were told that, while working on their main task (assessing their plans about various activities), they would hear various pieces of music. They were instructed to "treat the music as just part of the background and keep on answering the questionnaire ." Following these instructions subjects were presented written descriptions of five consumption episodes, each on a different page. Subjects were to read each episode description and indicate their intention to engage in the consumption activity. While they were reading the description and making a decision, one of the three musical segments presented earlier was played again in the background (at a quieter level). To allow the experimenter sufficient time

PAGE 144

135 to switch to another musical segment when appropriate, subjects were told to wait for the experimenter's instructions before turning to the next page. The first episode decision, intended as a practice with the task, was fixed and was accompanied by a neutral musical segment. The four experimental episodes were presented in two consecutive pairs. One of the pairs was accompanied by a pleasant musical segment, whereas the other pair was accompanied by an unpleasant segment. The order of the two pairs of episodes (and the accompanying musical segments) was rotated between sessions. Within each pair of episodes, one episode was expected to be assessed primarily on the basis of consummatory motives, whereas the other was expected to be assessed primarily on the basis of instrumental motives. The order of the episodes within each pair was counterbalanced across subjects within each session. After reporting their five decisions, subjects reported their anticipated affective and evaluative responses to each episode. They then completed a series of manipulation and confounding checks. Finally, they completed a series of individual difference measures (e.g., age, self-reported knowledge of tax issues, gender, etc.) Subjects in the control condition did not complete the first task. They were only administered the second part of the experiment. Every other aspect of the study was identical to that in the experimental conditions, except that there was no musical background in the control condition.

PAGE 145

136 Table 7.2 DESCRIPTIONS OF THE EPISODES Computer Course Episode As you know, an increasing number of teachers (especially of senior level courses) encourage the use of spreadsheets and computer-generated graphs. Starting this Fall, the Circa lab will offer a free introductory course on using spreadsheets (Excel, Lotus) and generating graphics with Harvard Graphics and Power Point. The course consists of four one-hour weekly sessions. Please take the time to give us your most accurate assessment of your intention to take this course in the Fall. High School Reunion Your next high school reunion (most likely the fifth-year reunion) will probably take place in a few semesters from now. Many of your former classmates and friends will enjoy the unique experience of being together again: evoking old memories, dancing, having a drink, munching,... having fun. You might be interested in sharing this experience. Please take the time to give us your most accurate assessment of your intention to attend this reunion. Tax Manual Next Fall, the Fisher School of Accounting will publish, at a reasonable price ($ 22), a manual on "What everyone should know about taxes." The manual will cover topics such as how to itemize deductions, when to claim dependents, how to select the appropriate filing status (e.g., in case of dual incomes), etc. Please take the time to give your most accurate assessment of your intention to buying a copy of this manual. Folkloric Dance Show An enjoyable way of spending an evening is to attend live performances. In September, the International Student Association will sponsor a two-hour show of folkloric dances at the Center for Performing Arts. The show features a group of international UF students who will perform traditional dances from a variety of countries (e.g., India, China, Russia, Morocco). Admission is free. Please take the time to give your most accurate assessment of your intention to attend this show.

PAGE 146

137 Musical Segments and Feelings It was necessary to identify musical segments that would elicit affective responses similar to those associated with the two consummatory episodes. Pretests were conducted to assess affective responses to a variety of musical segments, ranging from instrumental Indian pieces to popular tunes. Three musical segments were selected based on these pretests. The first segment was a from a classical piece by Eric Satie, expected to elicit relatively neutral feelings (M = 4.39 on a 1-7 scale). The second segment selected was a piece of Aboriginal music, to which pretest subjects reacted negatively, expressing feelings such as being "bored" and "annoyed" (M = 2.56 on a 1-7 scale). The third segment was from the opening tune of the popular "Muppet Show." Pretest subjects had expressed largely positive affective responses to this tune, such as feeling "cheerful" and "joyful" (M = 5.89 on a 1-7 scale). In the main experiment, the unpleasant segment (Aboriginal music) was consistently associated with the low behavioral intention episodes (the folkloric dance show and the tax manual), whereas the pleasant segment (Muppet Show) was consistently associated with the moderately high behavioral intention episodes (the high school reunion and the computer course). Attribution Subjects' attributions were manipulated during the first phase of the study when subjects listened to the three musical segments for the first time. The segments were presented in a fixed order: the first segment was the neutral one; the second segment was the unpleasant one; and the third segment was the pleasant one. In the high salience condition, subjects' attention was directed to how each segment affected their feelings.

PAGE 147

138 They were to report their reactions to each segment on four semantic differential scales anchored by sad/joyful, bored /excited, depressed/cheerful, and annoyed/happy. To further increase the salience of the affective influence of the musical segments, these subjects were encouraged to compare how their feelings in response to each segment varied. In addition, subjects were told that in previous studies most subjects had reported negative feelings in response to the second segment and positive feelings in response to the third segment. In the low salience condition, subjects' attention was directed away from their feelings. These subjects were asked to rate each segment on four nonevaluative dimensions also presented in a semantic differential format: familiar/unfamiliar, slow/fast paced, simple/complex, modern/classical. To further distract their attention from their feelings, subjects in this condition were asked to assess, based on their ratings, which pair of segments was the most similar and which one was the most dissimilar. Results Motive Manipulation and Confounding Checks Subjects reported their motives for each episode using a seven-point scale anchored by "I focused on how enjoyable or unpleasant this experience would be" and "I focused on whether it would help me attain a specific objective." Low values indicated a consummatory motive whereas high values indicated an instrumental motive. Responses for the four episodes were submitted to a split-plot ANOVA, with episode as a repeated factor and attribution, order of the motives, and order of the musical pieces

PAGE 148

139 as between-subjects factors. The analysis revealed a significant main effect of episode (F(3,177) = 104.39, p < 0.0001, partial a> 2 = .52) and no significant interaction. A planned contrast confirmed that the high school reunion (M = 1.83) and the folkloric dance (M = 1.97) episodes were more perceived as more consummatory (and less instrumental) than the computer course (M = 5.75) and the tax manual (M = 5.18) episodes (F(l,59) = 217.17, p < 0.0001). To ensure that the motive manipulation was not confounded with an involvement manipulation, subjects rated how important or unimportant each activity was on a sevenpoint scale (1 = not important at all; 7 = very important). These ratings were analyzed in the same way as the motive ratings. The analysis uncovered a main effect of episode, suggesting that the episodes were not equally important (F(3,177) = 24.58). A planned contrast revealed that the high school reunion (M = 4.68) and the computer course (M = 4.79) were equally important episodes (F < 1). These two episodes thus differed in the type of underlying motive but not in importance. A second contrast revealed that the tax manual episode (M = 3.98) was perceived as more important than the folkloric dance episode (M = 2.46, F(l,59) = 21.98, p < .0001, or = .26), calling for caution when comparing these two episodes. Behavioral Intentions The mean behavioral intentions are reported in Table 7.3. They were analyzed in a split-plot ANCOVA with episode as a repeated factor, and attribution (high salience, low salience, control), order of motives, and order of musical pieces as between-subjects factors. Four individual difference variables were entered as covariates: gender, self

PAGE 149

140 reported knowledge about the material covered in the computer course, self reported knowledge about taxes, and the pleasantness of high school memories. The overall repeated measure ANCOVA results are summarized in Appendix I. The episode x attribution interaction was not significant (F(l, 165) = 1.22, p = .30). That is, unexpectedly, the attribution manipulation did not have signicantly different effects on behavioral intentions toward the different episodes. Univariate analyses were conducted for each episode, separately. They are reported in Appendix J. It was predicted that behavioral intentions toward a pleasant consummatory episode would be less favorable if it was made salient that the background music could be responsible for the pleasant feelings experienced during the decision process, than if this fact was not made salient. A planned comparison revealed, as expected, that behavioral intentions toward the high school episode were significantly less favorable in the high salience condition than in the low salience condition (F(l,55) = 4.65, one-tailed p < .02, partial or comparison = .08). The observed difference may be regarded as meaning (1) that behavioral intentions were lower in the high salience condition than in the low salience condition, or (2) that behavioral intentions were higher in the low salience condition than in the high salience condition. This truism reflects genuinely different interpretations of the findings. The interpretation discussed above is based on a subtraction effect (cf. Martin 1986; Martin et al. 1990): Subjects in the high salience condition substracted part of their anticipatory affective response to the episode and misattributed it to the musical background. Alternatively, one could advance an inclusion explanation (see Schwarz and

PAGE 150

141 Clore 1983 and Experiments 3 and 4 of this dissertation). The musical segment may in fact have elevated subjects' mood. Because the attribution manipulation did not make the actual source of their feelings (i.e., the music) salient, subjects in the low salience condition may have included their positive mood in their evaluation of the episode. Behavioral intentions in the control condition suggest that the effect of the attribution manipulation was to lower behavioral intentions in the high salience condition rather than to heighten behavioral intentions in the low salience condition. Behavioral intentions toward the high school episode were similar in the control condition (where there was no background music) and in the low salience condition. This suggests that the pleasant background music (played at a substantially quieter level during the decision making phase than during the initial attribution manipulation phase) probably did not elevate subjects' mood in the low salience condition. Therefore, the results are more consistent with a subtraction explanation than with an inclusion explanation. It was further predicted that the attribution manipulation would have little effect on decisions driven by instrumental motives. Behavioral intentions toward the computer course were slightly lower in the high salience condition than in the low salience condition. However, consistent with the predictions, this effect was not significant (F(l,55) = 1.37, p = .24, w 2 comparison < .01).

PAGE 151

142 TABLE 7.3 (LEAST SQUARE) MEAN BEHAVIORAL INTENTIONS Episode

PAGE 152

143 appraisal of the episode. A logical implication of this interpretation is that subjects' initial affective reactions to the musical segment (during the first phase) should be negatively correlated with their subsequent behavioral intentions toward the episode. Factor analysis of the four items measuring affective responses to the pleasant musical segment (the "Muppet Show" tune) uncovered a single valence factor, accounting for 82 percent of the variance. These four items were averaged to form an index of the pleasantness of the initial affective responses to the segment (alpha = .88). This index was then regressed as a predictor of behavioral intentions toward the high school reunion and the computer course episodes, along with two other predictors: the reported salience of the influence of the music on subjects' feelings and the interaction between these two predictors. The results of this analysis are summarized in Table 7.4. As expected, the more pleasant the initial affective responses to the Muppet Show segment, the lower the behavioral intentions toward the consummatory high school episode. By contrast, initial affective responses to this musical tune were not significantly related to behavioral intentions toward the instrumental computer course episode. This pattern of results is consistent with a subtraction explanation of the findings. Value or Affective Response? The attribution effect uncovered for the high school episode is interpreted as a subtraction of feelings experienced in response to the episode. These feelings are hypothesized to have a specific affective quality. To further document this specificity, additional measures were collected (with no musical background) after subjects had

PAGE 153

144 reported their behavioral intentions toward all four episodes. Subjects were asked to rate each episode on four seven-point items. Two items were intended to measure anticipated affective responses to the episode: "very boring/very fun, enjoyable" and "would put me in a good/bad mood." Two items were intended to measure evaluative responses "very useful/very useless" and "very interesting/uninteresting" (Pham, Hughes, and Cohen 1993). TABLE 7.4 REGRESSION ANALYSES OF AFFECTIVE RESPONSES TO MUSIC AS A PREDICTOR OF BEHAVIORAL INTENTIONS

PAGE 154

145 attribution and order of motive as nonrepeated factors, and pleasantness of high school memories as a covariate. A similar analysis was performed on affective and evaluative responses to the computer episode, using self-reported familiarity with spreadsheets and graphic softwares as a covariate. The mean responses are reported in Table 7.5. The ANCOVA results are reported in Appendix K. TABLE 7.5 (LEAST SQUARE) MEAN BEHAVIORAL INTENTIONS Episode

PAGE 155

146 had a significant influence on anticipated affective responses (F(2,64) = 3.48, p < .04) and a nonsignificant influence on evaluative responses (F(2, 64) = 1.55, p = .22). These results suggest that the misattributed responses may have a distinctive affective quality. Discussion Inferring Instantiation Through Subtraction To document the instantiation of anticipatory affective responses, this study reversed the logic of Experiments 3 and 4. In these two experiments, anticipatory affective responses to a consumption episode were subtly augmented by covertly altering subjects' moods. Subjects apparently misattributed part of their contextually manipulated mood to the consumption episode. In this study, part of the anticipatory affective responses to a consumption episode was apparently subtracted and attributed to an irrelevant but plausible source. Subjects who were informed that the background music could be responsible for the feelings experienced as they were deciding about a pleasant episode expressed lower behavioral intentions than subjects for whom the potential affective influence of the background music was not made salient. Subjects' attribution of their feelings did not affect their behavioral intentions when their decision was driven by instrumental motives. This pattern of results is consistent with the proposition that, in decisions driven by consummatory motives, consumers may anticipatorily experience affective responses

PAGE 156

147 to consumption episodes. The phenomenological experience of these anticipatory affective responses may be similar enough to that of valenced current feelings— such as the ones a popular musical tune would elicit— that these two responses may be confused in a decision making task. These inferences follow the logic that one cannot subtract what is not there. Unless some decision input was indeed accessed in the consummatory decision, the outcome of this decision should not have depended on subjects' attributions. Further, unless this input resembled the experience of valenced feelings, it is unlikely that subjects would have confused the two. Further evidence of the distinctive affective quality of this input is provided by the finding that the attribution manipulation influenced subjects' anticipated affective responses to the episode but not their evaluative responses. The weaker attribution effect for the computer course decision suggests that in instrumental decisions, affective responses may be less likely to be anticipatorily experienced. As a result, there may be little to be "subtracted" from the decision. The Folkloric Dance Show Behavioral intentions toward the folkloric dance show were not affected by subjects' attributions as expected. Two explanations may be offered. First, a pretest had suggested that anticipatory affective responses to this episode would be largely negative. In particular, subjects were expected to instantiate feelings of boredom when deciding whether or not to attend the show. The pretest had also suggested that the selected unpleasant musical segment would also be experienced as "boring." In fact, subjects in

PAGE 157

148 this experiment did not react as unfavorably to the segment as subjects in the pretest, especially along the bored/excited dimension (M = 3.7 on a 1-7 scale where 1 means "bored" and 7 means "excited"). The lack of effect with this musical segment might have been caused by a weaker than anticipated manipulation. If subjects did not find this segment to be "that boring," they were not likely to misattribute their negative anticipatory affective responses to the folkloric dance show to the musical background. Second, it has been suggested that negative affective responses are more differentiated than positive affective responses (e.g., Isen 1984). There are more "shades of pain" than "shades of joy" (Ellsworth and Smith 1988). Anticipatory affective responses to the folkloric dance episode may have been too differentiated to be confused with feelings induced by the music. Alternatively the feelings induced by the unpleasant musical segment may have been too specific. Either source of mismatch would weaken the effectiveness of the attribution manipulation.

PAGE 158

CHAPTER 8 CONCLUSIONS On Bricks and Holes This dissertation aimed at providing a conceptual bridge between the consumer judgment and decision making literature and the affect literature. It is now time to review the empirical "bricks" of this bridge and identify its conceptual "holes." Unlike previous research, which has either ignored the role of affect in consumer decision making or treated it as part of the decision context, this dissertation took the central premise that affect is a source of information. It was proposed that to make decisions about consumption episodes consumers often rely on an affect recruitment heuristic: they access or construct concrete representations of the episodes (Proposition 1), which instantiate anticipatory affective responses (Proposition 2) that are then used as a decision input because of their unique information value (Proposition 3). It was theorized that because of their status as information, the use of anticipatory affective responses in judgment and decision making would follow the same principles as the use of other informational inputs. The use of anticipatory affective responses should depend on their relative accessibility (compared to alternative inputs) and their relative diagnosticity or information value (cf. Feldman and Lynch 1988). This reasoning led to 149

PAGE 159

150 the identification of several theoretical determinants of affect recruitment, such as consumers motives and their imagery abilities and propensities (Propositions 4a to 5d). A robust finding of this dissertation was that consumers do not make decisions about consumption episodes in the same way when they have consummatory motives and when they have instrumental motives. Consistent with previous research (e.g., Hirschman and Holbrook 1982; Millar and Tesser 1986), it was found in Experiment 1 that subjects with a consummatory motive were more influenced by the experiential content of an episode and less influenced by the cost associated with the episode than subjects with an instrumental motive. Although this finding per se does not validate the process of affect recruitment, it is at least consistent with the proposition that the likelihood of affect recruitment may depend on consumers' motives (instrumental or consummatory; Proposition 5a). More direct evidence that different motives may induce genuine process changes was uncovered in Experiments 3 and 4. It was observed that the decisions of subjects with a consummatory motive were more likely to be influenced by their contextual mood than the decisions of subjects with an instrumental motive. In Experiment 5, decisions about going to a high school reunion-decisions presumably guided by consummatory motives-were somewhat less sensitive to how feelings experienced during the decision process were attributed than were decisions about taking a computer class-decisions presumably guided by instrumental motives. In summary, motive appears to be a major determinant of the process through which consumers make decisions about consumption

PAGE 160

151 episodes. Is the process underlying decisions guided by consummatory motives that of affect recruitment? There is evidence that consumers may attach unique information value to feelings experienced during the decision process—evidence consistent with Proposition 3. An important characteristic of Experiments 3 and 4 was that the contextual mood manipulation presumably did not alter the evaluative content of the episode. Yet, in both experiments, this manipulation influenced the behavioral intentions of subjects with consummatory motives but not of subjects with instrumental motives. To the extent that this effect depended on subjects' motives, one can infer that this effect occurred because subjects with consummatory motives attached information value to their feeling experience. Insofar as the evaluative content of the episode was presumably unchanged, one can infer that the information value of this feeling experience may be unique. Experiment 4 additionally showed that to make decisions, subjects with consummatory motives appear to weigh feelings experienced during the decision process as if they were "seeing" how the episode "felt." In other words, affect recruitment is a plausible paramorphic description of the actual process underlying decisions guided by consummatory motives-at least with respect to how consumers incorporate their feelings into their decisions. It was also found that an expectancyvalue process did not model this effect equally well. This suggests that decisions about consumption episodes may be based on processes distinct from the attitude formation process typically assumed in consumer research.

PAGE 161

152 There was also some evidence in support of Proposition 2. It was found in Experiment 5 that behavioral intentions toward a pleasant episode were lower when subjects were led to believe that feelings experienced during the decision process could have been induced by a musical background. This finding was interpreted as indicating that subjects to whom the music's potential influence on feelings was made salient subtracted part of their anticipatory affective responses to the episode, misattributing it to the musical background. To the extent that one cannot subtract what one does not have, this subtraction (e.g., Martin et al. 1990) effect suggests that anticipatory affective responses can indeed be instantiated through decisions about consumption episodes. Additional correlational analyses further suggested that the phenomenal experience of this input was distinctively affective and not merely evaluative. The evidence for Proposition 1 that affect recruitment involves the construction or retrieval of concrete representations of the episode, is more tenuous. It was found in Experiment 1 that experiential content of the episode was most influential when subjects had high visual imagery abilities and a consummatory motive. Similarly it was observed in Experiment 3 that only subjects with a propensity to process information visually rather than verbally, and a consummatory motive, were influenced by their contextual mood when making a decision about going to a movie. Both findings suggest that affect recruitment may be facilitated (or amplified) when analog processing is more likely. Neither of these results, however, indicates that affect recruitment necessitate a concrete representation of the episode. The protocol analysis conducted in Experiment 4 failed

PAGE 162

153 to reveal any significant change in how the episode was represented when subjects had consummatory versus instrumental motives. The weakness of the empirical support for Proposition 1 may be due to the lowinvolvement character of the various decisions examined in this research (e.g., a fictitious dinner at a French restaurant or going to a movie). Conceivably, more involving decisions, such as planning a vacation trip or, as in one of the opening example, planning a romantic date, may provide stronger evidence that affect recruitment often involves concrete construals of consumption episodes. Alternatively, the proposition that consumers often construct concrete analog-like representations of the episode may need to be reconsidered. One could argue that, to make decisions, consumers may use anticipatory affective responses to abstract representations as well. Returning to our opening example, Maria may experience an unpleasant anticipatory affective response to the propositional belief "It will be too hot on the tennis court" and use this input to reject the tennis episode. The position that affect recruitment does not exclusively involve concrete representations is softer than that advocated by Epstein (1990, 1994), who postulates two independent processing systems. This softened position is more congenial with Triandis's (1980) model, in which no assumption is made about the representational format of the "thoughts" that activate anticipatory affective responses. A possible revised position could be as follows. Access to or construction of a concrete representation of the episode is not a necessary condition of affect recruitment. However, concrete construals may increase the likelihood that anticipatory affective

PAGE 163

154 responses are used by making such responses more accessible, as suggested by the moderating role of imagery-related variables observed in Experiments 1 and 3. A related issue concerns the amount of control that the consumer has on the way information is processed in decisions about consumption episodes. The affect recruitment heuristic implies that when anticipatory affective responses have information value, consumers actively construct representations of the episode to "recruit" anticipatory affective responses to these representations. As illustrated in Figure 1 (Chapter 2), the diagnosticity determinants of affect recruitment were assumed to operate early in the process-a position consistent with Strack's (1992) distinction between "experiential route" and "informational route" to judgment. It is also possible that consumers rely on the heuristic in a more passive way. The selective use of anticipatory affective responses may also be determined later in the process, perhaps at the judgment integration stage. Rather than (actively) retrieving, constructing or instantiating different types of inputs depending on their information value, consumers may (more passively) selectively weight the various inputs that are accessible to them. Instantiated (accessed) anticipatory affective responses would receive a greater weight in judgments and decisions when such responses are perceived to have information value. This selective weighing interpretation of affect recruitment would be consistent with the protocol results of Experiment 4. Although subjects' verbalizations did not reveal substantial content differences (e.g., consequence thoughts versus experiential thoughts; conclusions versus generalized feelings), there were significant differences in how different types of content (consequence thoughts and current feelings)

PAGE 164

155 predicted behavioral intentions. Subjects with a consummatory motive seemed to have weighted their current feelings more heavily than subjects with an instrumental motive. Conversely, subjects with an instrumental motive appear to have weighted consequence thoughts more heavily than subjects with a consummatory motive. According to this amended view of the process, affect recruitment may be more passive-reactive than initially postulated. It is possible, however, that the affect recruitment process appeared reactive because this research has focused on a specific type of decision outcome. All five experiments examined behavioral intentions toward single alternative episodes. 1 It is conceivable that choices among alternative consumption episodes would reveal a more constructive (active) view of affect recruitment. Undoubtedly, this issue is worthy of further investigation. It was suggested that the study of more involving decisions may provide a more powerful test of Proposition 1. It may help provide stronger support for other propositions as well. A noticeable feature of this program of research is that the observed effects were often weak. Several factors may be responsible for the weakness of these effects. First, the level of control necessary to rule out attitude formation processes-for instance, keeping the evaluative content of the episode constant in Experiments 3 and 4, and using mood as a proximal manipulation of anticipatory affective responses-placed severe constraints on the effect-sizes that could be empirically Similarly, the typical research on feelings as information examines judgments of single objects, such as life satisfaction (e.g., Schwarz and Clore 1983; Keltner, Locke and Audrain 1993).

PAGE 165

156 uncovered. It is likely that manipulations of anticipatory affective responses that would allow the evaluative content of the episode to vary would have resulted in stronger (although maybe less interpretable) effects. Second, this research focused on relatively inconsequential decisions. For instance, it is likely that the instantiation of anticipatory affective responses (Proposition 2) would be easier to observe with more involving decisions. Finally, the weakness of the effects may reflect the lack of sensitivity of available instruments to capture the hypothesized processes using a measurement approach (as opposed to a manipulation approach). For example, by requiring the conversion of memory contents into a verbal response format, protocol analysis may be inherently unable assess whether consumers accessed or constructed concrete analog-like representations of consumption episodes. 2 A New Agenda Apart from the issues raised above, which stress the need to "cement" holes left by the dissertation, this program of research needs to be extended in several directions. First, this research has focused primarily on the information conveyed by affective responses of different valence. However, as stressed in Chapter 2, the informational One could argue that the effects shown in this research are weak because the domain of processes that cannot be fully described by expectancy value models may itself be limited-which would account for the popularity of these models. It would inappropriate, however, to compare the affect recruitment heuristic with an expectancyvalue formulation uniquely on the basis of variance explained, for the theoretical frameworks involve different levels of explanation of consumer decision making. Should models be assessed solely on sheer parsimony and variance explained, consumer decision making would probably still be examined within a primitive cost-benefit framework.

PAGE 166

157 content of affective responses is not limited to their desirability (pleasant versus unpleasant). Affective responses may also convey information such as the degree of outcome uncertainty (e.g., fear versus sadness) and the person's sense of control in an episode (e.g., pride versus hope). For example, Gallagher and Clore (1985) observed that subjects induced (through hypnosis) into a state of anger were more likely to blame another person in an unrelated incident than subjects induced into a state of fear. As reported in Table 2.2 (Chapter 2), anger conveys a higher sense of legitimacy than fear, presumably increasing people's tendency to blame other people. Future research should therefore examine how anticipatory affective responses of similar valence but different content (e.g., pride versus joy) intervene in consumer decision making. Second, this research has studied only a subset of the potential determinants of affect recruitment. Whereas evidence was found for the moderating roles of consummatory versus instrumental motives and imagery-related factors, additional evidence needs to be gathered about the roles of feeling rules, self-awareness, aversive states, affect intensity, and evaluative complexity. Future research should also seek to identify additional determinants of the use of affective responses in consumer decision making. Third, this dissertation focused on decisions about consumption episodes. These decisions were examined not only because of their substantive relevance to marketing and consumer researchers, but primarily because of their potential to reveal theoretical insights into the role of affect in decision making. Future research should examine the role of affective responses in other decision domains (e.g., alternative occupations).

PAGE 167

158 Before any empirical effort is undertaken, the framework needs to be elaborated further in order to identify task factors (e.g., Payne 1982) that would explain when decisions are likely to involve affect recruitment. For instance, it was suggested (Proposition 5d) that a potential determinant of affect recruitment is the evaluative complexity of the episode. More generally, affect recruitment may be more likely in any decision that involves a high degree of evaluative complexity (e.g., Schwarz 1990; Schwarz and Clore 1988). For instance, when faced with a difficult career choice (going into industry versus staying in academia) a faculty member may picture himor herself performing each of these job roles to see how it feels. A systematic attempt to identify other task factors determining affect recruitment would be a useful extension of the framework; it would enhance the framework's applicability across decision domains. Finally, it was proposed that considering the information value of affective states would help examine consumption decision processes aimed at managing those affective states-an issue of increased interest among consumer researchers (e.g., Gardner and Scott 1990; Gould 1991; Kacen 1994). The framework suggested that such decisions may be guided by a part-regulation principle. This principle may engender a whole program of research. Let's see how it feels...

PAGE 168

APPENDIX A SHORTENED VERSION OF THE AFFECT INTENSITY SCALE The following questions refer to the emotional reactions to typical life-events. Please indicate how YOU react to these events by placing a number from the following scale in the blank space preceding each item. Please base your answers on how YOU react, not on how you think that others react or how you think a person should react. ALMOST ALMOST NEVER NEVER OCCASIONALLY USUALLY ALWAYS ALWAYS 12 3 4 5 6 1. __ When I accomplish something difficult I feel delighted or elated. 2. When I feel happy it is a strong type of exuberance. 3. __ __ I enjoy being with other people very much. 4. __ My emotions tend to be more intense than those of most people. 5. I get overly enthusiastic. 6. My heart races at the anticipation of some exciting event. 7. Sad movies deeply touch me. 8. My friends say I'm emotional. 9. The sight of someone who is hurt badly affects me strongly. 10. _ Seeing a picture of some violent car accident in a newspaper makes me feel sick to my stomach. 11. _ When I'm happy I feel very energetic. 12. _ When I receive an award I become overjoyed. 13. When I do something wrong I have strong feelings of shame and guilt. 159

PAGE 169

160 14. When things are going good I feel "on top of the world." 15. When I get angry it's easy for me to still be rational and not overreact. 16. _ My negative moods are mild in intensity. 17. When I feel happiness, it is a quiet type of contentment. 18. My friends would probably say I'm tense or "high-strung" person. 19. _ When someone compliments me, I get so happy I could "burst." 20. _ When I am nervous I get shaky all over.

PAGE 170

APPENDIX B VISUAL IMAGERY SCALE (SHEEHAN 1987) Rating instructions With respect to the mental picture suggested in each of the questions of the inventory, is the image which comes before your mind: [lj Perfectly clear and as vivid as the actual experience [2] Very clear and comparable in vividness to the actual experience [3] Moderately clear [4] Not clear or vivid but recognizable [5 1 Vague and dim [6| So vague and dim as to be hardly discernible [7| No image present at all, you only know that you are thinking of the object Items Think of some relative or friend whom you frequently see, considering carefully the picture that rises before your mind's eye. Rate the images suggested by each of the following questions in terms of clearness and vividness as specified on the card. 1. The exact contour of face, head, shoulders, and body ( ) 2. Characteristic poses of head, attitudes or body, etc. ( ) 3. The precise carriage, length of step, etc. in walking ( ) 4. The different colors worn in some familiar costume ( ) 161

PAGE 171

162 Think of seeing each of the following, considering carefully the picture which comes to your mind's eye; and rate each image in terms of clearness and vividness as specified on the card. 5. The sun as it is sinking below the horizon ( ) 6. A fish moving suddenly and fast through clear water ( ) 7. The stars on a clear winter night ( )

PAGE 172

APPENDIX C EXPERIMENT 1: REPEATED MEASURES ANOVA OF BEHAVIORAL INTENTIONS Source"

PAGE 173

APPENDIX D EXPERIMENT 1: EFFECTS OF AFFECT INTENSITY AND VISUAL IMAGERY ON BEHAVIORAL INTENTIONS Source"

PAGE 174

165 APPENDIX D-CONTINUED EXPERIMENT 1: EFFECTS OF AFFECT INTENSITY AND VISUAL IMAGERY ON BEHAVIORAL INTENTIONS AS A FUNCTION OF MOTIVE AND EXPERIENTIAL CONTENT Source 3

PAGE 175

APPENDIX E EXPERIMENT 2: REPEATED MEASURES ANOVA OF BEHAVIORAL INTENTIONS Source"

PAGE 176

APPENDIX F EXPERIMENT 3: ANCOVA OF BEHAVIORAL INTENTIONS Source 2

PAGE 177

APPENDIX G PROTOCOL CONTENT: CODING SCHEME Positive Consequences Verbalizations with clear positive implications for the episode and no intrinsic affective content (e.g., "[The reason was that] I needed to take this girl out to try get over my girlfriend"; "I probably will have some time between exams"). Positive Consequences Verbalizations with clear negative implications for the episode and no intrinsic affective content (e.g., "I wouldn't be able to take my son"; "Movies are too expensive"). Positive Feelings Verbalizations with an affective/experiential content and positive implications for the episode. The category includes: (1) expressions of desire (e.g., "I want to see the movie 'Above the Rim'"; "I'm dying to watch a movie soon"), (2) sensory verbalizations (e.g., "I also thought about how cold a movie theater is"), (3) cravings (e.g., "I've been wanting to see it for a couple of weeks now"), (4) current feelings (e.g., "I got a little excited when one came to mind that I wanted to see"), (5) expected feelings (e.g., "How relaxing it would be to just kick back with some popcorn"), (6) ^instantiation of previous affective experience and affectively charged attitudes (e.g., "I was thinking about... the last time I went to see a movie and how much I liked it"). 168

PAGE 178

169 Negative Feelings Same as above but with negative implication for the episode (e.g., "I would feel guilty if I did [go to the movies]"; "The image of sitting in uncomfortable seats. ..makes me cringe") Positive attitudes Summary evaluation (expressed in a nonemotional way) with positive implication for the episode (e.g., "My girlfriend and I like going to the movies and plan to go often"; "That doesn't mean I don't like to go to the movies because I do"). Negative attitudes Summary evaluation (expressed in a nonemotional way) with negative implication for the episode (e.g., "Going to the movies isn't really one of my favorite things to do anyway"; "I'd rather go to a bar than a movie"). Positive Nondecisions Statements referring to previous plans (e.g., I thought about maybe going on a date Friday [so we would probably see a movie]) or habits/base-rate (e.g., "I probably average one movie a week") with positive implication for the episode. Negative Nondecisions Statements referring to previous plans (e.g., "I will be going out of town for 3 days so will I bother) or habits/base-rate (e.g., "I usually go out to bars on the weekends. I have only been to the movies here at college about three times") with negative implication for the episode.

PAGE 179

170 Positive Conclusions Expressions that reflect mostly the (positive) outcome of the decision process (e.g., "So, this is why I may be going to the movies after all"; "So one night this weekend [we] will do the club thing and another [we] will see a movie"). Negative Conclusions Expressions that reflect mostly the (negative) outcome of the decision process (e.g., "So, I figured 1 probably won't have time to see a movie until after April 28, 1994"; "I know that I'll be unable to [watch a movie soon]) Abstract Attributes/Rules Listing of abstract attributes, reasoning rules/statements that do not specify their evaluative implications for the episode (e.g., "I [love] movies that I can relate to"; "Do I have free time or should I study"). Episodic/Exemplar Verbalizations References to concrete exemplars (specific persons, contexts, times, objects) without specific evaluative implication for the episode (e.g., "I thought about my girlfriend [because 1 usually take her with me]"; "I thought about how I almost went to see [a movie] last weekend, but never got around"). Process Verbalizations Descriptions of the process followed to reach a decision, with no evaluative implication (e.g., "When I was making my decision I based it on the fact that I never go out to see a movie").

PAGE 180

171 Task Verbalizations Statements or reflections about the decision making task, other than descriptions of the process (e.g., "I thought how funny it was that you would ask [my intention of going to a movie]"; "A first question: Is this week really a good example of an average week"). Movie Search Statements indicating a memory search for available movies (e.g., "I thoughts of the [movies] I had to chose from"; "What is playing at the movies"). Background Verbalizations Statements that explain other verbalizations but that do not have any evaluative implication per se (e.g., "I always see movies with my girlfriend"; "We have the same interest in movies"; "The movie has been out for a while").

PAGE 181

APPENDIX-H PROTOCOL CONTENT AS A FUNCTION OF MOOD AND INSTRUCTIONS Experimental conditions

PAGE 182

173 APPENDIX H-CONTINUED PROTOCOL CONTENT AS A FUNCTION OF MOOD AND INSTRUCTIONS Experimental conditions

PAGE 183

174 APPENDIX H--CONTINUED PROTOCOL CONTENT AS A FUNCTION OF MOOD AND INSTRUCTIONS Experimental conditions

PAGE 184

175 APPENDIX H-CONTINUED PROTOCOL CONTENT AS A FUNCTION OF MOOD AND INSTRUCTIONS Experimental conditions

PAGE 185

APPENDIX I EXPERIMENT 5: REPEATED ANCOVA OF BEHAVIORAL INTENTIONS Source"

PAGE 186

APPENDIX J EXPERIMENT 5: UNIVARIATE ANCOVAS OF BEHAVIORAL INTENTIONS HIGH SCHOOL EPISODE Source 11

PAGE 187

178 COMPUTER COURSE EPISODE Source"

PAGE 188

179 FOLKLORIC DANCE SHOW EPISODE Source"

PAGE 189

180

PAGE 190

REFERENCES Aaker, David A., Douglas M. Stayman, and Michael R. Hagerty (1986), "Warmth in Advertising: Measurement, Impact, and Sequence Effects," Journal of Consumer Research 12 (March), 365-381. Abelson, Robert P. (1983), "Whatever Became of Consistency Theory?," Personality and Social Psychology Bulletin 9 (1), 37-54. Alderson, Wroe (1957), Marketing Behavior and Executive Action Homewood, IL: Richard D. Irwin. Allen, Chris T., Karen A. Machleit, and Susan Schulzt Kleine (1992), "A Comparison of Attitudes and Emotions as Predictors of Behavior at Diverse Levels of Behavioral Experience," Journal of Consumer Research 18 (March), 493-504. Alpert, Judy I. and Mark I. Alpert (1990), "Music Influences on Mood and Purchase Intentions," Psychology and Marketing 7 (2), 109-133. Arnold, Magda B. (1960), Emotion and Personality Vol. 1. New York: Columbia University Press. Axelrod, Joel N. (1963), "Induced Moods and Attitudes toward Products," Journal of Advertising Research 3 (June), 19-24. Bassili, John N. and Marilyn C. Smith (1986), "On the Spontaneity of Trait Attribution: Converging Evidence for the Role of Cognitive Strategy," Journal of Personality and Social Psychology 50 (2). 239-245. Batra, Rajeev and Michael L. Ray (1986), "Affective Responses Mediating Acceptance of Advertising," Journal of Consumer Research 13 (September), 234-249. Batra, Rajeev and Douglas M. Stayman (1990), "The Role of Mood in Advertising Effectiveness," Journal of Consumer Research 17 (Sep), 203-214. 181

PAGE 191

182 Baumgartner, Hans (1992), "Remembrance of Things Past: Music, Autobiographical Memory, Emotion," in Advances in Consumer Research Vol. 19, eds. John F. Sherry and Brian Sternthal, Provo, UT: Association for Consumer Research, 613620. Bern, Daryl J. (1972), "Self-Perception Theory," in Advances in Experimental Social Psychology Vol. 6, ed. Leonard Berkowitz, New York: Academic Press, 1-62. Bettman, James R. (1993), "The Decision Maker Who Came In from the Cold," in Advances in Consumer Research Vol. 20, eds. Leigh McAlister and Michael L. Rotschild, Provo, UT: Association for Consumer Research, 7-11. Bettman, James R., Eric J. Johnson, and John W. Payne (1991), "Consumer Decision Making," in Handbook of Consumer Behavior eds. Thomas S. Robertson and Harold H. Kassarjian, Englewood Cliffs, NJ: PrenticeHall, 50-84. Bower, Gordon H. (1981), "Mood and Memory," American Psychologist 36 (2), 129148. Branscombe, Nyla R. (1988), "Conscious and Unconscious Processing of Affective and Cognitive Information," in Affect. Cognition, and Social Behavior eds. Klaus Fiedler and Joseph Forgas, Toronto: CJ Hogrefe, 3-25. Bruner, Gordon C, Jr. (1990), "Music, Mood, and Marketing," Journal of Marketing 54 (4), 94-104. Bryant, Jenning and Dolf Zillmann (1984), "Using Television to Alleviate Boredom and Stress: Selective Exposure as a Function of Induced Excitational States," Journal of Broadcasting 28 (1), 1-20. Buck, Ross (1985), "Prime Theory: An Integrated View of Motivation and Emotions," Psychological Review 92 (3), 389-413. Burke, Marian Chapman and Julie A. Edell (1989), "The Impact of Feelings on AdBased Affect and Cognition," Journal of Marketing Research 26 (February), 6983. Carver, Charles, Paul H. Blaney, and Michael F. Scheier (1979a), "Focus of Attention, Chronic Expectancy, and Responses to Feared Stimulus," Journal of Personality and Social Psychology 37 (7), 1186-1195. Carver, Charles S., Paul H. Blaney, and Michael F. Scheier (1979b), "Reassertion and Giving Up: The Interactive Role of Self-Directed Attention and Outcome Expectancy," Journal of Personality and Social Psychology 37 (10), 1859-1870.

PAGE 192

183 Carver, Charles S. and Michael F. Scheier (1990), "Principles of Self-Regulation: Action and Emotion," in Handbook of Motivation and Cognition Vol. 2, eds. E. Tory Higgins and Richard M. Sorrentino, New York: Guildford Press, 3-52. Childers, Terry L., Michael J. Houston, and Susan E. Heckler (1985), "Measurement of Individual Differences in Visual Versus Verbal Information Processing," Journal of Consumer Research 12 (September), 125-134. Clore, Gerald L. (1992), "Cognitive Phenomenology: Feelings and the Construction of Judgments," in The Construction of Social Judgments eds. Leonard L. Martin and Abraham Tesser, Hillsdale, NJ: Lawrence Erlbaum Associates Inc, 133-163. — Norbert Schwarz, and Michael Conway (1994), "Affective Causes amd Consequences of Social Information Processing," in Hanbook of Social Cognition 2nd edition, eds. Robert Wyer and Thomas Srull, Hillsdale, NJ: Lawrence Erlbaum Associates Inc, 323-417. Cohen, Joel B. and Charles S. Areni (1991), "Affect and Consumer Behavior," in Hanbook of Consumer Behavior eds. Thomas S. Robertson and Harold H. Kassarjian, Englewood Cliffs, NJ: Prentice-Hall, 188-240. Cunningham, Michael R., Jeff Steinberg, and Rita Grev (1980), "Wanting and Having to Help: Separate Motivations for Positive Mood and Guilt-Induced Helping," Journal of Personality and Social Psychology 38 (2), 181-182. Denes-Raj, Veronika and Seymour Epstein (1992), "Conflict Between Intuitive and Rational Processing: When People Behave Against Their Better Judgment," Journal of Personality and Social Psychology 66 (5), 819-829. Derbaix, Christian and Michel Tuan Pham (1991), "Affective Reactions to Consumption Situations: A Pilot Investigation." Journal of Economic Psychology 12, 325-355. De Rivera, Joseph (1984), "The Structure of Emotional Relationships," in Review of Personality and Social Psychology ed. Philip Shaver, Beverly Hills, CA: Sage, 116-145. Dommermuth, William P. and William J. Millar Jr. (1967), "Consumption Coincidence in Product Evaluation," Journal of Marketing Research 4 (November), 388-390. Edell, Julie A. and Marian Chapman Burke (1987), "The Power of Feelings in Understanding Advertising Effects," Journal of Consumer Research 14 (December), 421-433.

PAGE 193

184 Ellsworth, Phoebe and Craig A. Smith (1988), "Shades of Joy: Patterns of Appraisal Differentiating Pleasant Emotions," Cognition and Emotion 2 (4), 301-331. Epstein, Seymour (1990), "Cognitive-Experiential Self-Theory," in Handbook of Personality: Theory and Research ed. Lawrence A. Pervin, New York: Guildford Press, 165-192. — (1994), "Integration of the Cognitive and the Psychodynamic Unconscious," American Psychologist 49 (8), 709-724. Ericsson, K. Anders and Herbert A. Simon (1993), Protocol Analysis: Verbal Report as Data Revised Edition, Cambridge, MA: MIT Press. Faure, Corinne and David Glen Mick (1993), "Self-Gifts Through the Lens of Attibution Theory," in Advances in Consumer Research Vol. 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT: Association for Consumer Research. Fazio, Russel H., T. M. Lenn, and E. A. Effrein (1984), "Spontaneous Attitude Formation," Social Cognition 2, 217-234. Feldman, Jack M. and John G. Lynch Jr. (1988), "Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention, and Behavior," Journal of Applied Psychology 73 (3), 421-435. Fishbein, Martin and leek Ajzen (1975), Belief. Attitude. Intention and Behavior: An Introduction to Theory and Research Reading, MA: Addison-Wesley. Fiske, Susan T. and Mark A. Pavelchak (1986), "Category-Based versus PiecemealBased Affective Responses: Developments in Schema-Triggered Affect," in Handbook of Motivation and Cognition Vol. 1, eds. Richard M. Sorrentino and E. Tory Higgins, New York: Guildford Press, 167-203. Forgas, Joseph P. (1991), "Affective Influences on Partner Choice: Role of Mood in Social Decisions." Journal of Personality and Social Psychology 61 (5), 708-720. Friedland, Nehemia, Giora Keinan, and Yechiela Regev (1992), "Contolling the Uncontrolable: Effects of Stress on Illusory Perceptions of Controllabililty," Journal of Personality and Social Psychology 63 (6), 923-931. Frijda, Nico H. (1986), The Emotions Cambridge: Cambridge University Press. — (1988), "The Laws of Emotion," American Psychologist 43 (May), 349-358.

PAGE 194

185 Frijda, Nico H., Andrew Ortony, Joep Sonnemans, and Gerald L. Clore (1992), "The Complexity of Intensity: Issues Concerning the Structure of Emotion Intensity," in Review of Personality and Social Psychology Vol. 13, ed. Margaret S. Clark, Newbury Park, CA: Sage Publications, 60-89. Gallagher, Dennis and Gerald Clore (1985), "Effects of Fear and Anger on Judgments of Risk and Evaluations of Blame," Paper presented at the Proceedings of the Midwestern Psychological Association, May 4. Gardner, Meryl Paula (1985), "Mood States and Consumer Behavior: A Critical Review," Journal of Consumer Research 12 (December), 281-300. Gardner, Meryl P. and Ronald Paul Hill (1988), "Consumers' Mood States: Antecedents and Consequences of Experiential versus Informational Strategies of Brand Choice," Psychology and Marketing 5 (2), 169-182. Gardner, Meryl Paula and Dennis W. Rook (1988), "Effects of Impulse Purchases on Consumers' Affective States," in Advances in Consumer Research Vol. 15, ed. Michael J. Houston, Provo, UT: Association for Consumer Research, 127-130. Gardner, Meryl P. and John Scott (1990), "Product Type: A Neglected Moderator of the Effect of Mood," in Advances in Consumer Research Vol. 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard Pollay, Provo, UT: Association for Consumer Research, 585-589. Gardner, Meryl Paula and Frederick O. Wilhelm (1987), "Consumer Responses to Ads with Positive vs. Negative Appeals:Some Mediating Effects of Context-Induced Mood and Congruency Between Context and Ad, in Current Issues and Research in Advertising Vol. 10, eds. James H. Leigh and Claude R. Martin, :, 81-98. Gleicher, Faith, Kathryn A. Kost, Sara M. Baker, Alan J. Strathman, Steven A. Richman, and Steven J. Sherman (1990), The Role of Counterfactual Thinking in Judgments of Affect Personality and Social Psychology Bulletin, 16 (2), 284295. Goldberg, Marvin E. and Gerald J. Gorn (1987), "Happy and Sad TV Programs: How They Affect Reactions to Commercials," Journal of Consumer Research 14 (December), 387-403. Gorn, Gerald J., Marvin E. Goldberg, and Kunal Basu (1993), "Mood, Awareness, and Product Evaluation," Journal of Consumer Psychology 2 ((3)), 237-256.

PAGE 195

186 Gould, Stephen J. (1991), "The Self-Manipulation of My Pervasive, Perceived Vital Energy through Product Use: An Introspective-Praxis Perspective," Journal of Consumer Research 18 (September), 194-207. Havlena, William J. and Morris B. Holbrook (1986), "The Varieties of Consumption Experience: Comparing Two Typologies of Emotion in Consumer Behavior," Journal of Consumer Research 13, 394-404. Havlena, William J., Morris B. Holbrook, and Donald R. Lehmann (1989), "Assessing the Validity of Emotional Typologies," Psychology and Marketing 6 (2), 97-112. Helregel, Brenda K. and James B. Weaver (1989), "Mood-Management During Pregnancy Through Selective Exposure to Television," Journal of Broadcasting and Electronic Media 33, 15-33. Higgins, E. Tory (1987), "Self-Discrepancy: A Theory Relating Self and Affect," Psychological Review 94 (3), 319-340. Hill, Ronald Paul and Harold Robinson (1991), "Fanatic Consumer Behavior: Athletics as a Consumption Experience," Psychology and Marketing 8 (2), 79-99. Hirschman, Elisabeth C. and Morris B. Holbrook (1982), "Hedonic Consumption: Emerging Concepts, Methods and Propositions," Journal of Marketing 46 (Summer), 92-101. Hochschild, Arlie Russell (1979), "Emotion Work, Feeling Rules, and Social Structure," American Journal of Sociology 85 (3), 551-575. Hoffman, Martin L. (1986), "Affect, Cognition, and Motivation," in Handbook of Motivation and Cognition eds. Richard M. Sorrentino and E. Torry Higgins, New York: Guilford Press, 244-279. Holak, Susan L. and William J. Havlena (1992), "Nostalgia: An Exploratory Study of Themes and Emotions in the Nostalic Experience," in Advances in Consumer Research Vol. 19, eds. John F. Sherry and Brian Sternthal, Provo, UT: Association for Consumer Research, 380-386. Holbrook, Morris B. (1986), "Emotion in the Consumption Experience: Toward a New Model of the Human Consumer," in The Role of Affect in Consumer Behavior: Emerging Theories and Applications eds. Robert A. Peterson, Wayne D. Hoyer, and William R. Wilson, Lexington, MA: D.C. Heath and Company, 17-52.

PAGE 196

187 Holbrook, Morris B. and Rajeev Batra (1987), "Assessing the Role of Emotions as Mediators of Consumer Responses to Advertising," Jo urnal of Consumer Research 14 (December), 404-420. Holbrook, Morris B. and Meryl P. Gardner (1993), "An Approach to Investigating the Emotional Determinants of Consumption Durations: Why Do People Consume What They Consume for How Long as They Consume It?," Journal of Consumer Psychology 2 (2), Forthcoming. Holbrook, Morris B. and Elizabeth C. Hirschman (1982), "The Experiential Aspects of Consumption: Consumer Fantasies, Feelings, and Fun," Journal of Consumer Research 9 (September), 132-140. Holbrook, Morris B., Robert W. Chestnut, Terence A. Oliva, and Eric A. Greenleaf (1984), "Plays as a Consumption Experience: The Roles of Emotions, Performance, and Personality in the Enjoyeant of Games," Journal of Consumer Research 11, 728-739. Hornik, Jacob (1993), "The Role of Affect in Consumers' Temporal Judgments," Psychology and Marketing 10 (3), 239-255. Hull, J. G., R. W. Levenson, R. D. Young, and K. J. Sher (1983), "The SelfAwareness Reducing Effects of Alcohol Consumption," J ournal of Personality and Social Psychology 44, 461-473. Isen, Alice M. (1984), "Toward Understanding the Role of Affect in Cognition," in Handbook of Social Cognition Vol. 3, eds. Robert S. Wyer and Thomas K. Srull, Hillsdale, NJ: Erlbaum, 179-236. Isen, Alice M. and Thomas E. Shalker (1982), "The Effects of Feeling State on Evaluation of Positive, Neutral, and Negative Stimuli: When You 'Accentuate the Positive,' Dou You 'Eliminate the Negative'?," Social Psychology Quarterly 45 (1), 53-63. Isen, Alice M., Thomas E. Shalker, Margaret Clark, and Lynn Karp (1978), "Affect, Accessibility of Material in Memory, and Behavior: A Cognitive Loop?," Journal of Personality and Social Psychology 36 (1), 1-12. Izard, Caroll E. (1977), Human Emotions New York: Plenum Press. James, William (1890/1950a), The Principles of Psychology Vol. 1. New York: Dover. — (189071950b), The Principles of Psychology Vol. 2. New York: Dover.

PAGE 197

188 Jones, Lyle V. (1960), "Some Invariant Findings Under the Method of Successive Intervals," in Psychological Scaling: Theory and Applications eds. Harold Gulliksen and Samuel Messick, New York: John Wiley and Sons,Inc, 7-20. Kacen, Jacqueline J. (1994), "Phenomenological Insights in Mood and Mood-Related Consumer Behaviors," in Advances in Consumer Research Vol. 21, eds. Chris T. Allen and Deborah Roedder John, Provo, UT: Association for Consumer Research, 519-525. Kahn, Barbara E. and Alice M. Isen (1993), "The Influence of Positive Affect on Variety Seeking among Safe, Enjoyable Products," Journal of Consumer Research 20 (September), 257-270. Kahneman, Daniel and Jackie Snell (1990), "Predicting Utility," in Insights in Decision Making ed. Robin M. Hogarth, Chicago: The University of Chicago Press, 295310. Kahneman, Daniel and Amos Tversky (1982), "The Simulation Heuristic," in Judgment Under Uncertainty: Heuristics and Biases eds. Daniel Kahneman, Paul Slovic, and Amos Tversky, Cambridge: Cambridge University Press, 201-208. Kellaris, James J. and Susan Powell Mantel (1994), "The Influence of Mood and Gender on Consumers' Time Perceptions," in Advances in Consumer Research Vol. 21, eds. Chris T. Allen and Deborah Roedder John, Provot, UT: Association for Consumer Research, 514-518. Keltner, Dacher, Phoebe C. Ellsworth, and Kari Edwards (1993), "Beyond Simple Pessimism: Effect of Sadness and Anger on Social Perception," Journal of Personality and Social Psychology 64 (5), 740-752. Keltner, Dacher, Kenneth D. Locke, and Paul C. Audrain (1993), "The Influence of Attributions on the Relevance of Negative Feelings to Personal Satisfaction," Personality and Social Psychology Bulletin 19 (1), 21-29. Kirkpatrick, Lee A. and Seymour Epstein (1992), "Cognitive-Experiential Self-Theory amd Subjective Probability: Further Evidence for Two Conceptual Systems," Journal of Personality and Social Psychology 63 (4), 534-544. Lang, Peter J. (1984), "Cognition in Emotion: Concept and Action," in Emotions, Cognition, and Behavior, eds. Carroll E. Izard, Jerome Kagan, and Robert B. Zajonc, New York: Cambridge University Press, 192-226.

PAGE 198

189 Larsen, Randy J. (1984), "Theory and Measurement of Affect Intensity as an Individual Difference Characteristic," Dissertation Abstracts International 5, 2297B (University Microfilms No. 84-22112). Larsen, Randy J., Ed Diener, and Robert A. Emmons (1986), "Affect Intensity and Reactions to Daily Events," Journal of Personality and Social Psychology 51 (4), 803-814. Lazarus, Richard S. (1991). Emotion and Adaptation New York: Oxford University Press. Lazarus, Richard S. and Susan Folkman (1984), Stress. Appraisal, and Coping New York: Springer Publishing Company. Lehmann, Donald R. (1987), "Pumping Iron III: An examination of Compulsive Lifting," in Advances in Consumer Research eds. Melanie Wallendorf and Paul Anderson, Provo, UT: Association for Consumer Research, 129-131. Leventhal, Howard (1984), "A Perceptual -Motor Theory of Emotion," in Advances in Experimental Social Psychology Vol. 17, ed. Leonard Berkowitz, New York: Academic Press, 117-182. Lewin, Kurt (1951). Field Theory in Social Science: Selected Theoretical Papers New York: Harper. Lingle, John H. and Thomas M. Ostrom (1979), "Retrieval Selectivity in Memory-Based Impression Judgments." Journal of Personality and Social Psychology 37 (2), 180-194. Lynch, John G., Howard Marmorstein, and Michael F. Weigold (1988), "Choices from Sets Including Remembered Brands: Use of Recalled Attributes and Prior Overall Evaluations," Journal of Consumer Research 15 (September), 169-183. Mandler, George (1984a), "Consciousness, Imagery, and Emotion with Special Reference to Autonomic Imagery," Journal of Mental Imagery 8 (4), 87-94. — (1984b), Mind and Body New York: W.W. Norton & Company. Mano, Haim (1990), "Emotional States and Decision Making," in Advances in Consumer Research eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT: Association for Consumer Research, 577-584.

PAGE 199

190 Manucia, Gloria K., Donald J. Baumann, and Robert B. Cialdini (1984), "Mood Influence on Helping: Direct Effects or Side Effects?." Journal of Personality and Social Psychology 46 (2), 357-364. Marlatt, G. Alan, Carole F. Kosturn, and Alan R. Lang (1975), "Provocation to Anger and Opportunity for Retaliation as Determinants of Alcohol Consumption in Social Drinkers," Journal of Abnormal Psychology 84, 6. Martin, Leonard L. (1986), "Set/Reset: Use and Disuse of Concepts in Impression Formation." Journal of Personality and Social Psycholog y, 51 (3), 493-504. Martin, Leonard L., Thomas F. Harlow, and Fritz Strack (1992), "The Role of Bodily Sensations in the Evaluation of Social Events," Personality and Social Psychology Bulletin 18 (4), 412-419. Martin, Leonard L., John J. Seta, and Rick A. Crelia (1990), "Assimilation and Contrast as a Function of People's Willingness and Ability to Expend Effort in Forming an Impression," Journal of Personality and Social Psychology 59 (1), 27-37. Mehrabian, Albert and James Russell (1974), An Approach to Environmental Psychology : Cambridge, MA: MIT Press. Mick, David Glen and Michelle DeMoss (1990), "Self-Gifts: Phenonomelogical Insights from Four Contexts," Journal of Consumer Research 17 (December), 322-332. — (1992), "Further Findings on Self-Gifts: Product, Qualities, and Socioeconomic Correlates," in Advances in Consumer Research eds. John F. Sherry and Brian Sternthal, Provo, UT: Association for Consumer Research, 140-146. Mick, David Glen and Corinne Faure (1993), "An Attributional Model of Self-Gift Purchase Likelihood in Achievement Contexts: Theoretical and Managerial Insights," Working Paper, University of Florida. Millar, Murray G. and Abraham Tesser (1986), "Effects of Affective and Cognitive Focus on the Attitude-Behavior Relation." Journal of Personality and Social Psychology 51 (2), 270-276. — (1992), "The Role of Beliefs and Feelings in Guiding Behavior: The Mismatch Model," in The Construction of Social Judgments eds. Leonard L. Martin and Abraham Tesser, Hillsdale, NJ: Lawrence Erlbaum Associates Inc, 277-300.

PAGE 200

191 Miniard, Paul W., Sunil Bhatla, and Deepak Sirdeshmukh (1992), "Mood as a Determinant of Postconsumption Product Evaluations: Mood Effects and Their Dependency on the Affective Intensity of the Consumption Experience," Journal of Consumer Psychology 1 (2), 173-195. Morris, William N. (1992), "A Functional Analysis of the Role of Mood in Affective Systems," in Review of Personality and Social Psychology: Emotion Vol. 13, ed. Margaret S. Clark, Newbury Park: CA: Sage, 256-293. Morris, William N. and Nora P. Reilly (1987), "Toward a Self-Regulation of Mood: Theory and Research," Motivation and Emotion 11 (3), 215-249. O'Guinn, Thomas C. and Ronald J. Faber (1989), "Compulsive Buying: A Phenomenological Exploration," Journal of Consumer Research 16 (September), 147-157. Olney, Thomas J., Morris B. Holbrook, and Rajeev Batra (1991), "Consumer Responses to Advertising: The Effects of Ad Content, Emotions, and Attitude toward the Ad on Viewing Time." Journal of Consumer Research 17 (March), 440-453. Ortony, Andrew, Gerald L. Clore, and Allan Collins (1988), The Cognitive Structure of Emotions Cambridge: Cambridge University Press. Payne, John W. (1982), "Contingent Decision Behavior," Psychological Bulletin 92 (2), 382-402. Pham, Michel Tuan (1994), "Cue Selection and Representation Effects of Arousal on Persuasion," Working Paper, University of Florida, May. Pham, Michel Tuan, G. David Hugues, and Joel B. Cohen (1993), "Validating a DialTurning Instrument for Real-Time Measurement of Affective and Evaluative Responses to Advertising," Marketing Science Institute Report No 93-116, September. Plutchik, Robert (1980), Emotion. A Psvchoevolutionarv Synthesis New York: Harper and Row. Richardson, Alan (1977), "Verbalizer-Visualizer," Journal of Mental Imagery 1 (1), 109-126. Richins, Marsha L., Kim Kr McKeage, and Debbie Najjar (1992), "An Exploration of Materialism and Consumption-Related Affect," in Advances in Consumer

PAGE 201

192 Research Vol. 19, eds. John F. Sherry and Brian Sternthal, Provo,UT: Association for Consumer Research, 229-236. Roese, Neale J. (1994), "The Functional Basis of Counterfactual Thinking," Journal of Personality and Social Psychology 66 (5), 805-818. Rook, Dennis W. (1987), "The Buying Impulse," Journal of Consumer Research 14 (September), 189-199. Roseman, Ira J. (1984), "Cognitive Determinants of Emotion," in Review of Personality and Social Psychology: Emotions. Relationships, and Health Vol. 5, ed. Phillip Shaver, Beverly Hills, CA: Sage, 1 1-34. — (1991), "Appraisal Determinants of Descrete Emotions," Cognition and Emotion 5 (3), 161-200. Roseman, Ira J., Martin S. Spindel, and Paul S. Jose (1990), "Appraisals of EmotionEliciting Events: Testing a Theory of Discrete Emotions, Journal of Personality and Social Psychology 59 (5), 899-915. Schachter, Stanley and Jerome E. Singer (1962), "Cognitive, Social, and Physiological Determinants of Emotional State." Psychological Review 69, 5(September), 379399. Schaller, Mark and Robert B. Cialdini (1990), "Happiness, Sadness, and Helping," in Hanbook of Motivation and Cognition Vol. 2, eds. E. Tory Higgins and Richard M. Sorrentino, New York: The Guilford Press, 265-296. N. Scheier, Michael F. and Charles S. Carver (1977), "Self-Focused Attention and the Experience of Emotion: Attraction, Repulsion, Elation, and Depression," Journal of Personality and Social Psychology 35 (9), 625-636. Scherer, Klaus R. (1984), "Emotion as a Multicomponent Process, A Model and Some Cross Cultural Data, in Review of Personality and Social Psychology: Emotions. Relationships, and Health Vol. 5, ed. Phillip Shaver, Beverly Hills, CA: Sage, 37-63. Schwarz, Norbert (1990), "Feelings as Information: Informational and Motivational Functions of Affective States," in Handbook of Motivation and Cognition Vol. 2, eds. E. Torry Higgins and Richard M. Sorrentino, New York: Guilford Press, 527-561.

PAGE 202

193 Schwarz, Norbert and Gerald Gore L (1983), "Mood, Misattribution, and Judgments of Well-Being: Informative and Directive Functions of Affective States," Journal of Personality and Social Psychology 45, 513-523. Schwarz, Norbert and Gerald L. Clore (1988), "How Do I Feel About It? The Informative Function of Affective States," in Affect. Cognition, and Social Behavior eds. Klaus Fiedler and Joseph Forgas, Toronto: C.J. Hogrefe, 44-62. Sheehan, Peter Wilson (1967), "A Shortened Form of Betts' Questionnaire Upon Mental Imagery," Journal of Clinical Psychology 23, 386-389. Smith, Craig A. and Phoebe Ellsworth (1985), "Pattern of Cognitive Appraisal in Emotion," Journal of Personality and Social Psychology 48 (4), 813-838. Srull, Thomas K. (1984), "Methodological Techniques for the Study of Person Memory and Social Cognition," in Hanbook of Social Cognition Vol. 2, eds. Robert S. Wyer and Thomas K. Srull, Hillsdale, NJ: Lawrence Erlbaum Associates Inc, 172. Srull, Thomas K. (1987), "Memory, Mood, and Consumer Judgment," in Advances in Consumer Research eds. Melanie Wallendorf and Paul F. Anderson, Provo, UT: Association for Consumer Research, 404-407. Stayman, Douglas M. and David A. Aaker (1988), "Are All the Effects of Ad-Induced Feelings Mediated by Aad?," Journal of Consumer Research 15 (December), 368-373. Stepper, Sabine and Fritz Strack (1993), "Proprioceptive Determinants of Emotional and Nonemotional Feelings," Journal of Personality and Social Psychology 64 (2), 211-220. Strack, Fritz (1992), "The Different Routes to Social Judgments: Experiential versus Informational Strategies," in The Construction of Social Judgments eds. Leonard L. Martin and Abraham Tesser, Hillsdale, NJ: Lawrence Erlbaum Associates Inc, 249-275. Swinyard, William R. (1993), "The Effects of Mood, Involvement, and Quality of Store Experience on Shopping Intentions," Journal of Consumer Research 20 (September), 271-280. Taylor, Shelley E. and Sherry K. Schneider (1989), "Coping and the Simulation of Events," Social Cognition 7 (2), 174-194. Triandis, Harry C. (1977), Interpersonal Behavior Monterey, CA: Brooks/Cole.

PAGE 203

194 — (1980), "Values, Attitudes, and Interpersonal Behavior," in Nebraska Symposium on Motivation 1979 Vol. 27, eds. Herbert E. Howe and Monte M. Page, Lincoln, NB: University of Nebraska Press, 195-259. Wallendorf, Melanie and Eric J. Arnould (1991), '"We Gather Together': Consumption Rituals of Thanksgiving Day," Journal of Consumer Research 18 (June), 13-31. Weinberg, Peter and Wolfgang Gottwald (1982), "Impulsive Consumer Buying as a Result of Emotions," Journal of Business Research 10, 43-57. Weiner, Bernard (1985), "An Attributional Theory of Achievement Motivation and Emotion," Psychological Review 92 (4), 548-573. Wells, Gary L. and Igor Gavanski (1989), "Mental Simulation of Causality," Journal of Personality and Social Psychology 56 (2), 161-169. Westbrook, Robert A. (1987), "Product/Consumption-Based Affective Responses and Postpurchase Processes," Journal of Marketing Research 24 (August), 258-270. Westbrook, Robert A. and Richard L. Oliver (1991), "The Dimensionality of Consumption Emotion Patterns and Consumer Satisfaction," Journal of Consumer Research 18 (June), 84-91. Zajonc, Robert B. (1980), "Feeling and Thinking, Preferences Need No Inferences," American Psychologist 35 (2), 151-175. — (1983), "Discussion of Abelson's Talk on Cartwright's Founders' Day," Personality and Social Psychology Bulletin 9 (1), 55-59. Zajonc, Robert B. and Hazel Markus (1982), "Affective and Cognitive Factors in Preferences," Journal of Consumer Research 9 (September), 123-131. Zanna, Mark P. and John K. Rempel (1988), "Attitudes: A New Look at an Old Concept," in The Social Psychology of Knowledge eds. Daniel Bar-Tal and Arie W. Kruglanski, Cambridge: Cambridge University Press, 315-334. Zillmann, Dolf (1988), "Mood Management: Using Entertainment to Full Advantage," in Communication. Social Cognition, and Affect eds. Lewis Donohew and Howard E. Sypher, Hillsdale, New Jersey: Lawrence Erlbaum, 147-171. Zillmann, Dolf, R. T. Hezel, and N. J. Medoff (1980), "The Effects of Affective States on Selective Exposure to Televised Fare," Journal of Applied Social Psychology 10, 323-339.

PAGE 204

BIOGRAPHICAL SKETCH (Michel) Tuan Pham was born September 29, 1965 in Leopoldville, Zaire. He grew up in Mons, Belgium, devoting most of his free time to sports, especially judo (competing for the CSPJ-Mons team) and tennis. He attended the Catholic University of Mons (FUCaM), where in 1987 he received a Licence in Applied Economics, with High Distinction. For the next three years, he worked as an instructor at his alma mater, developing his taste for research and teaching. In June 1989, he was awarded a threeyear doctoral fellowship by the Intercollegiate Center in Management Science, Brussels. He entered the Ph.D. program in marketing at the University of Florida in Fall 1990, and obtained a Master of Arts in August 1992. He was the 1993 American Marketing Association Doctoral Consortium Fellow from the University of Florida. He has accepted a position as Assistant Professor of Business at Columbia University. 195

PAGE 205

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. ;1 B. Cohen, Chair Distinguished Service Professor of Marketing I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. tit*aJtt&. Margaret M. Bradley / Associate Research Scientist of Psychology I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Richard J. Lutz Professor of Marketing I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. John G. Lynch, Jr. • 'r Graduate Research Professor of Marketing

PAGE 206

This dissertation was submitted to the Graduate Faculty ot the Department of Marketing in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December 1994 Dean, Graduate School

PAGE 207

LD 1780 1991 UNIVERSITY OF FUORIDA 3 1262 08285 330 9