UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Mood, motivation, and task me Zerbe, Wilfred Joachim 1987

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1988_A1 Z47.pdf [ 13.18MB ]
Metadata
JSON: 831-1.0098296.json
JSON-LD: 831-1.0098296-ld.json
RDF/XML (Pretty): 831-1.0098296-rdf.xml
RDF/JSON: 831-1.0098296-rdf.json
Turtle: 831-1.0098296-turtle.txt
N-Triples: 831-1.0098296-rdf-ntriples.txt
Original Record: 831-1.0098296-source.json
Full Text
831-1.0098296-fulltext.txt
Citation
831-1.0098296.ris

Full Text

MOOD, MOTIVATION, AND TASK MEMORY by WILFRED J . ZERBE B.A. 1978, University Of B r i t i s h Columbia M.A. 1982, University Of B r i t i s h Columbia A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY i n THE FACULTY OF GRADUATE STUDIES Department Of Commerce And Business Administration We accept t h i s thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA November 1987 © Wil f r e d J . Zerbe, 1987 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Commerce and Business Administration The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 D a t e March 4, 1988 DE-6(3/81) i i Abstract Theorists in organizational behavior have generally ignored emotional determinants of behavior. A task of this dissertation was to extend the use of emotions for understanding organizational behavior in general and work motivation in particular. Two theories, expectancy theory and network theory, are used to make predictions about the relationship between mood and perceptions of the relationship between effort and performance. According to expectancy theory, the effort that people choose to expend at tasks i s a function of their belief about the degree to which effort and performance covary. Network theory predicts that memories are connected by a network of associations. The accessibility for recall of a memory i s a function of the activation of these associations. In this way positive events are more accessible for recall when individuals are in a positive mood state because of associations based on the affective valence of memories. Such accessibility of events for recall has been shown to be a determinant of probability judgements. On this basis i t was predicted that mood would bias individuals' judgements of the probability that specific levels of effort lead to specific levels of performance. In other words, that mood affects expectancy. Specifically, i t was predicted that individuals in an elated mood would report higher expectancy than individuals in a depressed mood. Mood was defined as a self-evaluative feeling state. Two other hypotheses were formed: that mood would influence how cause for behavior i s attributed, and that individual differences in self esteem would moderate the relationship between mood and expectancy. Three studies were performed to provide a foundation for the testing of these hypotheses. In a fourth study they were tested. Study One assessed the psychometric properties of measures of mood states, individual differences, i i i and task perceptions. Study Two concerned the experimental induction of mood. Mood manipulations used in the experimental literature were reviewed and one, a musical procedure, was chosen. The validity of this manipulation was then tested by having participants listen to the music of an elated, neutral, or depressed mood induction procedure. The results of Study Two provided strong evidence for the valid i t y of the manipulation. Both self-report measures of mood and an unobtrusive behavioral measure were significantly affected. The results of Study Two also showed the u t i l i t y of a conceptualization of mood as comprising two components: arousal and pleasure. It was shown that depression is characterized by low arousal and displeasure, and elation by high arousal and pleasure. Study Three reviewed the conceptualization and measurement of expectancy. It was argued that expectancy i s properly conceptualized as the perceived covariation between effort and performance. This requires measurement of the relationship between multiple levels of effort and multiple levels of performance and calculation from these measures of an index of perceived effort—performance covariation. Most prior measurement has only considered the relationship between high effort and high performance. Further, i t was argued that such appropriate measurement allows predictions to be made about expectancy across individuals, in contrast to the argument that expectancy theory i s a within-subjects theory. Previous authors have used such an approach to measure expectancy but have not demonstrated i t s validity. Study Three undertook such validation. Participants completed one of two experimental tasks: one with high objective expectancy, the other with low objective expectancy. As predicted, scores on the perceived covariation measure of expectancy were significantly higher in i v the high objective expectancy task. Measures of related constructs were influenced in a manner consistent with this finding. It was concluded that strong support for the expectancy measure existed. On the foundation of Studies One, Two, and Three, Study Four undertook to test the formal hypotheses of the dissertation. In each of three experimental sessions, participants completed a business decision-making task, underwent either an elation, neutral mood, or depression induction procedure, and then completed measures of their mood state, expectancy, and other task perceptions. The results of Study Four indicated that significant differences in mood resulted from the manipulation. However, none of the experimental hypotheses were supported. Mood did not influence expectancy or task attributions. A number of alternate explanations for this finding were considered, including failure of the mood manipulation, measurement error, and lack of s t a t i s t i c a l power. Of these, i t was concluded that while Study Four lacked power to detect a large effect, this did not f u l l y explain the failure to support the experimental hypotheses. Also compelling was the argument that the mood manipulation was not sufficiently powerful. V Table of Contents Abstract i i List Of Tables .v List Of Figures x i i Acknowledgement x i i i I. INTRODUCTION 1 II. WHAT IS EMOTION? 5 Causes Of Emotion 6 Fundamental Emotions And The Evolutional Perspective 7 William James And The Primacy Of The Periphery 7 Centralist Theories 10 Conflict Theories 11 Structural Theories 12 Reconciling Theories Of Emotion 17 What Is Mood? 18 The Effects Of Mood On Thought And Behavior 21 Affect And Behavior 22 Mood And Thought 24 Network Theory 28 Mood And Arousal 30 Arousal And Memory 32 III. EMOTION AND ORGANIZATIONS 34 Emotion As Dependent Variable 35 Job Satisfaction As Emotion 35 Emotional Expression As Part Of The Work Role 37 Emotion As An Independent Variable 41 v i The Expression Of Affect In Decision-making 42 Job Satisfaction As Mood 43 Mood And Performance Appraisal 45 IV. AFFECT AND MOTIVATION 47 Expectancy Theory 47 Measuring Expectancy 49 Mood And Expectancy 54 Research Hypotheses 59 Hypothesis One: Mood And Expectancy 59 Hypothesis Two: Mood And Causal Attributions 63 Hypothesis Three: Self-esteem And Mood 64 Proposed Research 71 V. STUDY ONE: PSYCHOMETRIC PROPERTIES OF MEASURES 73 Overview 73 Subjects 73 Procedure 74 Measures And Results 75 Individual Differences: Self-esteem And Impression Management 75 Task Perceptions 76 Causal Attributions 82 Multiple Affect Adjective Checklist 85 Pleasure And Arousal 89 Dimensionality Of Mood Measures 89 Parallel Scale Construction 94 Correlations Between Measures 95 VI. STUDY TWO: WHAT IS MOOD? 103 Manipulation Of Mood 103 v i i Method 110 Procedure .110 Measures I l l Adjective Checklist I l l Semantic Differential I l l Response Latency I l l Results 113 Evaluation Of Assumptions 113 Multivariate Analyses Of Variance 116 Summary 122 Discussion: What Is Mood? 124 VII. STUDY THREE: WHAT IS EXPECTANCY? 127 Conceptualizing And Measuring Expectancy 127 Validating Expectancy Measurement 132 Method 134 Subjects And Design 134 Manipulation Of Expectancy 135 Perceptual-motor Task 135 Cognitive Reasoning Task 135 Measures 137 Effort-performance Covariation 137 Control Over Performance 139 Perceived Correlation 140 Expected Performance 140 Task Perceptions And Causal Attributions ...140 Results 140 Evaluation Of Assumptions 141 v i i i Multivariate Analysis Of Variance 143 Discussion 149 VIII. STUDY FOUR: MOOD AND EXPECTANCY 151 Overview 151 Experimental Design 151 Method 154 Subjects 154 Procedure 154 Decision-making Task 157 Mood Manipulation 162 Dependent Measures 163 Sessions Two And Three 168 Results 169 Evaluation Of Assumptions 169 Analysis Of Manipulation Checks 172 Results Of Manipulation Checks 174 Measures Of Expectancy And Task Perceptions 184 Individual Differences ..190 Discussion: Alternate Explanations 192 Manipulation Failure 192 Measurement Error 197 Lack Of Experimental Control 198 Lack Of Stastical Power 199 The Null Hypothesis Is True 204 Conclusion 206 APPENDIX A: Experimental Materials, Study One 209 APPENDIX B: Musical Selections 224 ix APPENDIX C: Experimental Materials For Study Three 226 APPENDIX D: Consent Form And Questionnaire, Study Four 230 APPENDIX E: Verbal Expectancy Items, Study Four 237 APPENDIX F: Cell Means, A l l Dependent Variables, Study Four 238 References 249 X List of Tables Table 1. Fundamental Or Primary Emotions Listed By Three Leading Theorists (from Mandler, 1984, P. 36) 8 Table 2. Emotional Reactions Resulting From Structural Evaluations, According To Roseman (1984) 14 Table 3. Items Assessing Task Perceptions, Grouped By Constuct 78 Table 4. Factor Loadings, Communalities (h 2), Percent Of Variance For Factor Analysis On Task Perception Items ....81 Table 5. Descriptive Statistics And R e l i a b i l i t y Estimates For Measures Of Individual Differences, Task Perceptions, And Causal Attributions. 83 Table 6. Items Used From MAACL Short Form, Study One 87 Table 7. Means, Standard Deviations And Reli a b i l i t y Estimates For Measures Of Mood State, Study One 88 Table 8. Factor Loadings, Communalities For Maximum Likelihood Factor Extraction On MAACL Items 91 Table 9. Factor Loadings, Communalities For Maximum Likelihood Factor Extraction On Semantic Differential Items 93 Table 10. Means, Standard Deviations And Internal Consistencies Of The Split-half MAACL Scales, Study One 96 Table 11. Correlations Between Dependent Variables, Study One 99 Table 12. Correlations For Dependent Variables, Study Two 118 Table 13. Univariate And Stepdown F-tests, Study Two 119 Table 14. Means And Standard Deviations, Treatment And Gender Groups, Study Two 120 Table 15. Univariate And Stepdown F-tests, Study Three. 145 xi Table 16. Means And Standard Deviations For Dependent Variables, Study Two 146 Table 17. Correlations Between Dependent Variables, Study Three 148 Table 18. Univariate And Stepdown F-tests, Order By Session (Mood) Effect On Manipulation Checks, Study Four 175 Table 19. Pooled Within-cell Correlations For Manipulation Checks, Study Four 176 Table 20. Means And Standard Deviations For Manipulation Checks, By Mood Effect, Study Four ....178 Table 21. Univariate And Stepdown F-tests, Session Effect On Manipulation Checks, Study Four 181 Table 22. Means And Standard Deviations For Manipulation Checks, By Session Effect, Study Four ". 182 Table 23. Means And Standard Deviations For Dependent Variables By Mood Effect, Study Four 186 Table 24. Univariate And Stepdown F-tests On Dependent Variables, Session Effect, Study Four 187 Table 25. Means And Standard Deviations For Dependent Variables By Sessions, Study Four > 188 Table 26. Within-cells Correlation Matrix, Study Four 189 x i i List of Figures Figure 1. Structural Model Of Emotions, From Kemper 16 Figure 2. Circumplex Model Of Emotions 31 Figure 3. Contingency Table Model Of The Relationship Between Effort And Performance 61 Figure 4. Causal Attribution Measures, Study One 84 Figure 5. Instructions For The MAACL 112 Figure 6. Terminal Display For The Item "Active" 112 Figure 7. Assessment Of Relationship Between Effort And Performance, Study Two 138 Figure 8. Split-Plot Design, Mood Treatment By Session And Order Conditions 155 x i i i Acknowledgement I owe my appreciation to the many people who guided this manuscript to completion. Most central i s my supervisor, Dr. Ralph Stablein, whose intellect and personal support I called on often. For shepherding me through the research that is reported here, as well as through my years in the doctoral program at the University of British Columbia, I am most grateful. I would also like to thank the members of my supervisory committee, Dr. Eric Eich, Dr. Peter Frost, and Dr. Craig Pinder, for their honest criticism and encouragement. A dissertation i s but one of the steps in most doctoral programs. Appreciation and thanks are therefore due not only to those who helped bring this manuscript to fruition, but also to those who supported and encouraged the work that came before i t . I would lik e , therefore, to thank my fellow students, and the faculty in the Industrial Relations and Management Area at U. B. C. Finally, I would like to thank my parents, Edmund and Alma Zerbe, and most importantly, my wife, Rose-Marie Jaeger, for helping me, motivating me, and supporting me in my education. 1 I. INTRODUCTION "Thinking, feeling, and acting." For 150 years the idea that human experience could be understood to include these three realms has dominated psychological thinking (Hilgard, 1980). Some proponents of this t r i p a r t i t e conception intended that every event be represented in a l l three spheres (Hilgard, 1980; Isen, 1984). Thus, an event in the thinking domain would influence thinking, feeling and acting; an emotion or feeling would influence thinking, feeling and acting; an action would influence thinking, feeling and acting: in this conception each is inexorably linked to the others. Others, the Faculty Psychologists of the 18th century, divided psychological experience into separate realms of cognition, emotion, and motivation. Modern theorists have rejected Faculty Psychology, but have nonetheless tended to segregate the influences of the three domains (Hilgard, 1980). For example, we now commonly talk about thinking without also talking about feeling. Thoughts and feelings are seen as quite separate. This separation, and an emphasis on the study of cognition, has paved the way for the modern neglect of emotion. We regularly examine cognitive explanations of behavior without considering the role of emotion. We have, in general, ignored emotional determinants of behavior and the relationship between feeling and thinking. Of late, though, we have been called to attend to the role of emotion. For example: What is the role of emotion in cognition? We leave i t to the poet, the playwright, the novelist. As people we delight in art and music. We fight, get angered, have joy, grief, happiness. But as students of mental events we are ignorant of why, how? (Norman, 1980) Emotions have received renewed attention in the study of organizations 2 and work settings. Researchers and theorists have examined, among other-issues, how work settings establish rules for emotional feeling and expression (Hochschild, 1983; Rafaeli St Sutton, 1987; Sutton & Rafaeli, 1986); the effects of emotional expression on the performance of decision-making groups (Guzzo & Waters, 1982); the role emotions play in forming organizational culture and f a c i l i t a t i n g organizational development (Kahn, 1986; Van Maanen & Kunda, 1985; Weick, 1985); the effect of mood on performance appraisal (Park, Sims & Motowidlo, 1986; Sinclair, in press), and have attempted to use the construct of mood to shed light on the relationship between job satisfaction and performance, particularly of prosocial behaviors (Bateman & Organ, 1983; Brief & Motowidlo, 1986; Motowidlo, 1984). A task of the proposed dissertation i s to extend investigation of the implications of emotion for understanding organizational behavior. Specifically, the question at hand concerns how mood influences the processing of information about jobs, such as individuals' expectations about the link between effort and performance, perceptions about the likelihood that tasks w i l l be successfully accomplished, the attribution of responsibility for job performance, task satisfaction, and so on. The dissertation w i l l attempt to show how feelings can influence such job-related cognitions. It w i l l also discuss, though not examine empirically, where and when emotions arise in organizational settings, and how emotions can and should be managed. Extensions of the work proposed w i l l also be discussed, including the influence of emotion on variables relevant to behavior in organizations other than those at hand, and how a program of research building on the foundation of the dissertation might proceed. There is growing acknowledgement of the affective quality of workplaces. In a survey of emotional experiences in everyday l i f e , Scherer and Tannenbaum 3 (1986) found that work was second only to family as a source of emotions. Work was the largest source of positive experiences and the second largest source of negative experiences. Such emotional experiences are li k e l y to have significant implications for individuals in organizations and for organizations themselves. Yet we know l i t t l e about such impact. What, for example, i s the practical impact of emotion on, say, performance? Although emotion is often conceptualized as motivating, emotion i s absent in most theories of work motivation. However, i f we acknowledge the inherent connections between thinking, feeling, and acting, i t i s evident that by studying emotion, we can add to what we know about motivation, to our a b i l i t y to predict effort, attention, and persistence in job settings. Such study of emotions has theoretical as well as practical importance for organizations. Increasing our understanding of the determinants of organizational behavior i s in t r i n s i c a l l y of value. The goal of this dissertation i s to increase our understanding of the relationship between emotion and motivation. The following chapters contain the foundation and justification for the development of specific research hypotheses, and present a test of those hypotheses. Chapter Two reviews theories of emotion and considers conceptual issues. Approaches which have examined the consequences of emotion are distinguished from those seeking to establish i t s causes. The principal theories of'the causes of emotion are reviewed and a definition of mood i s presented. Theory and empirical findings about the effects of mood on thought and behavior are reviewed. In Chapter Three the study of emotions in organizations i s examined. Chapter Four defends the choice of motivation as an area in which further investigation of the role of mood should be productive and poses specific hypotheses relating mood to individuals' perceptions of the relationship 4 between effort and performance. Chapters Five through Seven describe a series of studies which provide the foundation for the testing of these hypotheses. In Chapter Eight they are tested directly. In each of these chapters the implications of findings are discussed. The dissertation that follows addresses the following questions: First, what i s mood? To understand this, i t i s helpful to examine approaches to the study of emotion. Having examined how emotions have been studied in organizational behavior, mood and motivation are chosen as the domain of interest, and the relationship between mood and expectancy as the specific research question to be addressed. The second question addressed i s therefore, What is expectancy? How is i t conceptualized and how is i t appropriately measured? With the answers to these two questions, the ultimate question i s addressed: Is there a relationship between mood and expectancy? 5 II. WHAT IS EMOTION? What i s emotion? James (1884) Our understanding of mood i s closely related to what we know about emotion. To provide a background for a discussion of mood and i t s consequences in later sections, in this section theories of emotion are reviewed. On can, in fact, evade explicit definition of emotion by allowing people to supply their own definition. Most people have such a definition, they can t e l l you what emotion " i s " . Yet when you ask academic students of emotion "what i s emotion?", you discover disagreement and lack of consensus. Plutchik (1980) l i s t s 28 different definitions of emotion in the psychological literature. Kleinginna and Kleinginna (1981) l i s t 92 definitions. They also l i s t 9 skeptical statements (e.g., "There does not seem to be a satisfactory way to define emotion, aside from i t s manifestations in act or verbal statements of feeling", Cofer, 1972). Arnold (I960) summarized inquiry into emotion: " i t i s one of the most d i f f i c u l t and confused fields in the whole of psychology" (pp. 10-11). There i s "a curious and overwhelming confusion" (Hillman, 1961, p. 5) in the theory of emotion. This confusion i s li k e l y because different theoretical perspectives have fostered different definitions and theoretical emphases. Some writers emphasize expressive reactions, others instrumental behavior, some identify emotions primarily as biological processes, others as social processes. Some emphasize the role of the central nervous system, some the autonomic nervous system, some emphasize cognitive processes and others feeling states. How then should we proceed? How can we wade into the sea of theories of 6 emotion without drowning or muddying the waters further? One way is by recognizing that theoretical disagreement i s in part a result of different approaches and traditions, and i s not inherent to the topic. Thus a f i r s t step toward reconciling theoretical differences and choosing an appropriate course i s to review these approaches and traditions. The study of emotion can, for example, be simply divided into two approaches: emotions studied as dependent variables and as independent variables. That i s , there are theories of emotion that try to specify the determinants of emotion, what i t s causes or antecedents are, and there are theories of the consequences of emotion, what i t s effects are, and how i t influences thought and behavior. In this chapter, these two approaches are reviewed. Presented f i r s t are the many and varied theories of the causes of emotion. Also presented i s an attempt at reconciling differences between these theories. Finally, mood i s defined and the effects of mood on thought and behavior are reviewed. Causes of Emotion Many writers have distinguished between traditions in the study of emotion. Fraisse (1968) described two "faces" of emotion. The f i r s t , "mental" face sees physiological events as the consequences of psychic events. In contrast, the "organic" face views physiological events, rather than thoughts, as the precursors of emotion. Hochschild (1983) makes a similar distinction between "interactionist" and "organismic" approaches. Others (Mandler, 1984; Schachter & Singer, 1962) have distinguished between "central" theories concerned with central nervous system mechanisms, and "peripheral" theories concerned with peripheral reactions and autonomic nervous system responses. Candland's (1977) system of description has three parts: the 7 biological, evolutional, and cognitive traditions. Fundamental Emotions and the Evolutional Perspective Berscheid (1983) points out that emotion was one of the f i r s t psychological phenomena "to be removed from the arena of philosophical speculation" (p. 119). This introduction of emotion to scientific inquiry was accomplished by Darwin (1872), who explored the adaptive significance of emotional expression. A number of theorists have continued this work (e.g., Izard, 1971, 1977; Plutchik, 1962; Tomkins, 1962, 1963, 1980). According to Berscheid (1983), Darwin has l e f t two legacies: The f i r s t was an insistence that no theory of emotion would be li k e l y to be valid i f i t did not illuminate what purpose emotion serves for the survival of the species. Today, v i r t u a l l y a l l theorists of emotion agree that the experience and expression of emotion has served, and probably continues to serve, an important function in the survival of humans. From the evolutionary perspective, then, emotion is not an "irrational" and frivolous component of human behavior. To the contrary, i t s rationality and dignity derive directly from i t s service in the l i f e and death struggle of each individual to survive and to survive as comfortably and as happily as possible (Berscheid, 1983, p. 120). The second legacy was Darwin's recognition that the expression of emotion was associated with strong "nervous excitation." This was taken up, as we shall see, by the advocates of the biological tradition. Contemporary proponents of the evolutional perspective postulate that emotional expression and experience results from innate neural processes that produce discrete, fundamental emotions. Although there i s much agreement on the processes leading to fundamental emotions, there is less agreement over their number and kind. Table 1 illustrates this. William James and the Primacy of the Periphery Before James, theories of emotion were largely centralistic; that i s , TOMKINS IZARD PLUTCHIK Fear Fear Fear Anger Anger Anger Enjoyment Joy Joy Disgust Disgust Disgust Interest Interest Anticipation Surprise Surprise Surprise Contempt Contempt Shame Shame Sadness Sadness Distress Guilt Acceptance Table 1. Fundamental or Primary Emotions Listed by Three Leading Theorists (from Mandler, 1984, p. 9 they posited a mind that interpreted events, the interpretation providing the emotional feeling (Candland, 1977). James took up investigation of the association between excitation and emotion that Darwin suggested and turned theories of emotion toward physiological explanations. As Candland states, though, this movement toward physiology was as l i k e l y a result of the increasing use of postulative-deductive scientific procedures in studying behavior. The novelty of James1 theory i s that i t shifted attention from the central mind to the peripheral structures of the body. According to James (1884), our natural way of thinking about...emotions i s that the mental perception of some fact excites the mental affection called the emotion, and that this latter state of mind gives rise to the bodily expression. My thesis on the contrary is that the bodily changes  follow directly the PERCEPTION of the exciting fact,and that our  feeling of the same changes as they occur IS the emotion (p. 189). The sequence James proposed was perception, then arousal or emotion, then cognition. This has generally come to be called the James-Lange theory, recognizing the similar contribution of Lange (1885). Lange's position was, in fact, more extreme. Lange insisted that emotions were the consequences of certain "vaso-motor effects." Mandler (1984) points out that James has frequently been interpreted as having said that emotions followed visceral changes and nothing more. In fact, James did not confine himself to visceral antecedents and argued that Lange had placed too much emphasis on the vaso-motor factor. Mandler recognizes that interpretation of the James-Lange theory is problematic with respect to how the perceptions of external events produce the bodily effects. He cites James' insistence that 10 environmental events can give rise to bodily, visceral changes without any awareness of the meaning of these environmental events and without any interpretation of them. For the modern reader, who sees nothing strange or surprising in perceptual, cognitive processes occurring in response to environmental events without conscious accompaniment, the central argument about the visceral-emotional sequence that James and Lange asserted seems to f a l l apart at that point (Mandler, 1984, p.22). The allowance that complex cognitive processes can intervene between the environment and bodily reaction, and hence that f e l t emotion results in part from some interpretive event that generates the bodily response, undoes the James-Lange position that bodily responses have primacy. Centralist Theories Much research has been devoted to testing the sequence proposed by James. Prominent i s the work of Cannon (1927). His empirical findings did not support the equality of bodily change and visceral feeling as presumed by the biological tradition. Cannon showed that different emotions are accompanied by similar, not different bodily states; that separation of the viscera and central nervous system does not preclude emotional behavior; that the autonomic nervous system responds slowly and diffusely in contrast to more rapid emotional reactions; and that manipulation of bodily states through a r t i f i c i a l means (e.g., drugs) does not necessarily affect emotional- states. More recently, Schachter and his colleagues (Schachter, 1971; Schachter & Singer, 1962) proposed a cognition—arousal or labelling explanation of emotion, in the tradition of Cannon. Schachter and his colleagues argue that emotional state results from the interaction of physiological arousal and a cognition as to the cause of the arousal. Meyer (1956), who anticipated the work of Schachter, concluded that physiological reactions are undifferentiated, becoming differentiated into a specific emotional experience only as a result of cognition of the stimulus situation. Arousal does not 11 evoke any specific emotion, i t serves only to intensify emotional states while cognitions determine their quality. Both physiological arousal and cognitive evaluation are necessary for the production of emotional states but neither i s sufficient. Unexplained arousal also stimulates a need to evaluate i t s causes. If this search identifies an emotional stimulus as the li k e l y cause, then emotional experience w i l l follow. Most of the empirical tests of Schachter's theory have examined the consequences of this unexplained arousal. Centralist theories, such as those of Cannon and Schachter, emphasize the importance of structures in the central nervous system, i.e. the mind, as determining the quality of emotional experience. Physiological reactions are granted to be necessary but not sufficient for emotion to follow. Conflict Theories Part of the mental or centralist position has adopted a particular explanation for the source of autonomic nervous system arousal necessary for emotion, namely conflict or interruption. Conflict theories "concern themselves with specific mechanisms whereby current behavior i s interrupted and 'emotional' responses are substituted....when an important activity of the organism i s blocked, emotion follows" (Hunt, 1941). Paulhan's (1930) main thesis was that we observe the same f a c t — t h e arrest of tendency—whenever any affective phenomena take place. Even positive emotions are the result of disconfirmed expectations or arrested actions. The particular emotion experienced depends on the tendency that i s arrested and the conditions under which i t occurs. Meyer (1956), for example, stated that emotion is "aroused when a tendency to respond i s arrested or inhibited." Meyer gave credit to Dewey (1894, 1895) for having fathered the conflict theory of emotion, though he 12 says that Paulhan's work predates Dewey's. Dewey said that there i s no emotion when actions run their course. Emotions arise only when other irrelevant reactions resist integration with those on their way to completion. "Such resistance means actual tensions, checking, interference, inhibition, or conflict....(Such) conflict constitutes the emotion..without such conflict there is no emotion, with i t there i s " (Angier, 1927). In contrast to Dewey and Paulhan, Mandler (1984) does not assume that conflict has specific emotional consequences, but rather that i t has undifferentiated visceral consequences. The emotional content—the quality of the emotion that follows— is set by specific cognitive circumstances of the interruption and possibly i t s consequences. Thus, both positive and negative emotions are seen as following interruption, and the same interruptive event may indeed produce different emotional states or consequences, depending on the surrounding situational and intrapsychic cognitive context (p. 46). Mandler's theory, as most recently expressed (Mandler, 1984), i s the most comprehensive conflict theory of emotion. Mandler incorporates modern theories of attentional and cognitive structures, such as schemas, into his analysis of the process by which individuals appraise stimuli and assign meaning to their feeling states. He also posits interruption in the realm of thinking and perception (in addition to behavior) as sources of arousal. So, for example, discrepant perception is that which does not f i t our perceptual expectations and schema and so interrupts the dominant expectation, the currently active schema. Similarly, discrepant thoughts such as prior engagements just remembered can interrupt the ongoing stream of thought and mental a c t i v i t i e s . Structural Theories Mandler proposes very few sources of arousal, interruption being the most 13 important for emotion. Cognitive evaluations, however, can have many sources, of which he posits three classes: innate preferences, personal and cultural predication, and structural value. These are the sources of value, the other half of the cognition-plus-arousal-equals-emotion formula. Of most interest here are structural theories of affect. As Mandler describes, the cognitive structuralists start with a small set of cognitive variables, usually including goals, and then construct possible outcomes for various combinations of these cognitive states. The outcomes of these cognitive constellations are then mapped, often with some hesitation, onto the putative emotional labels of the common language. These constructions eventuate in a relatively large number of emotional or affective outcomes, ranging from a dozen or so to sixteen or to as yet indeterminate larger numbers (p. 211). Roseman (1984), for example, presents five dimensions which are used to evaluate a real or imagined stimulus. The result of this evaluation determines the quality of the emotional reaction. Roseman's dimensions are (1) whether the event i s consistent with the motives of the individual, that i s , whether i t i s positive or negative; (2) whether the stimulus i s appetitive (oriented toward giving or withholding rewards) or aversive (oriented toward giving or withholding punishments); (3) whether the event or stimulus i s perceived to be caused by the self, by an other, or by circumstances; (4) whether the occurrence of the event i s certain or uncertain; and (5) whether the individual i s weak or strong relative to the stimulus or cause. Liking, for example, i s the result of a stimulus that i s positive, that i s consistent with the individual's motives, and is other-caused, and i s either certain or uncertain, and in relation to which the individual i s either strong or weak. Table 2 shows how combinations of levels of these dimensions are mapped onto natural language labels for emotional states. Kemper (1978, 1984) has put forward a social interactional theory of 14 Circumstance-Caused Unknown Uncertain Certain Uncertain Certain Positive Negative Mot ive-Con s i s t ent Appetitive | Aversive Motive-Inconsistent Appetitive | Aversive Surf irise He )pe iar Joy Relief Sorrow Discomfort Disgust He >pe Frustration Strong Joy Relief Other-Caused Uncertain Certain Uncertain Certain Liking Disliking Weak Anger Strong Self-Caused Uncertain Certain Uncertain Certain Pride Shame, Guilt Weak Regret Strong Table 2. Emotional reactions resulting from structural evaluations, according to Roseman (1984). 15 emotion based on the structural qualities of the social environment. Kemper's is a sociological theory, i t examines "relational or organizational concepts" of "properties of relationships or groups" as determinants of emotion. The starting point for this theory i s the postulate that human relationships generally occur within a context of interdependence and division of labor between actors. Further, these social relationships have two fundamental dimensions of interaction content: power and status. Granted this view of relationship, the following proposition i s possible: A very large class of human emotions results from real,  anticipated, recollected, or imagined outcomes of power and status  relations. This means that i f we wish to predict or understand the occurrence of many human emotions we must look at the structure and process of power and status relations between actors (Kemper, 1978, p. 371, emphasis in original). Kemper's theory proposes three classes of emotions: structural emotions, anticipatory emotions, and consequent emotions. Structural emotions result from the relatively stable power and status relations between an actor and an other. Power and status can be excess, adequate, and insufficient. Figure 1 depicts structural emotions. Guilt, for example, i s the emotion that results from a structural position of excess power over another in an interaction. Anticipatory emotions are responses to how the actor views the future state of the relationship: whether the future is viewed with optimism or pessimism and with confidence or lack of confidence. Finally, consequent emotions are the result of interaction episodes. They conclude the chain that links structural and anticipatory emotions and interaction outcomes. Power and status can increase, decrease, or stay the same as a result of interaction. The combinations of levels of structural, anticipatory, and consequent factors in Kemper's scheme produce 1701 outcome c e l l s . While this suggests that a social interactional approach has the requisite variety for analysis of emotion, i t also presents a daunting task of mapping interaction outcomes onto 16 Own Power Status fi* Adequate 'idem Guilt, anxiety Security Fear—anxiety -Shame Happy •Depression Fear—anxiety Other's Power Status Anger, contempt, shame Happy Guilt—shame, anxiety Figure 1. Structural model of emotions, from Kemper (1978). 17 emotion labels. By means of simplifying assumptions, though, Kemper is able to reduce the number of possible outcomes to less than 170. Reconciling Theories of Emotion The theories reviewed in this chapter may have spanned the domain of theories but have by no means exhausted i t . It i s f a i r to say that there are as many theories of emotion as there are theorists. The interested reader is referred to Strongman's (1977) detailed review of the positions of particular emotional theorists. Given these multiple conceptions of emotion, how should we proceed, how can they be reconciled? It i s tempting to choose one that i s "best", that maximizes some empirical criterion, say, like predicting more of the variance in some measure of emotionality. Fell (1977) observes, however, that each theory omits some crucial aspect of emotion, yet strives to be the foundation from which others are derived and from which the others must situate their data. For each theory the adoption of a particular criterion and definition of emotion determines that i t w i l l best explain a particular set of the aspects of emotion. Thus every theory i s the best explanation of i t s own criterion and no theory i s the best explanation of a l l c r i t e r i a . The multiple conceptions of emotion can be reconciled by recognizing that at present a universal theory of emotion may not be realizable. A view held by a number of theorists (e.g., A v e r i l l , 1980; Izard, 1977; Leventhal, 1979; Plutchik, 1980), i s that emotion i s a syndrome— a pattern of co-occurring responses. Cognitive processes, physiological changes, emotional displays, feeling states, are parts of a phenomenon that has many components and many expressions. Each i s appropriate, but each i s only one indicator of emotion. Therefore, rather than seeking a universal theory of a phenomenon with 18 multiple, disparate conceptualizations that do not justify such generality, we should accept existing theories of emotion as middle-range theories. Middle-range theories attempt to explain only segments of the universe (Pinder & Moore, 1979, 1980). The theory that i s best, therefore, depends on the question at hand. It is the one that best explains the segment of the universe of facets of emotion that one wishes to explain. If, for example, we are interested in how power and status affects emotion in organizations we might consult Kemper's theory. If, on the other hand, we are interested in the stressful effects of emotional conflict on worker health, we would be wise to use a theoretical approach that includes physiological aspects of emotion. Finally, i f we are interested in the consequences of mood states on the thought and behavior of individuals in organizations, we should consult theories of mood. What is Mood? Apart from or as part of the concept of emotion what do we mean by "mood"? I propose f i r s t of a l l , the use of the term "affect" to denote the general concept of emotion. Although some authors use the term affect to mean less intense or less specific states (e.g., Mandler, 1984), I concur with Berscheid (1983) who says that affect i s "an antiseptic term, but one that encompasses without prejudice the entire range of quality and intensity of human emotion and feeling, from mild i r r i t a t i o n to raging hatred to blinding joy to placid contentment" (p. 110). "Affect" and "emotion" both represent the general concept, although "emotion" i s a more evocative term, suggesting a more richly colored and varied construct. And what i s this general concept? Notwithstanding the recognition that 19 defining emotion is fraught with hazards, and that no one definition i s likely to be sufficient for a l l purposes, the following position i s adopted, largely from Mandler (1984): Affect is the evaluative aspect of human information processing, as opposed to i t s ideational or descriptive aspects or what we usually c a l l cognition. Affect i s the evaluation of information as good or bad, positive or negative, pleasant or unpleasant, arousing or nonarousing, for example. Cognition, on the other hand, is the descriptive component of information, such as i t s color, size, shape, temperature, etc. We commonly differentiate beliefs, which are ideational concepts, from attitudes, which are affective concepts, although they are not pure emotion. Attitudes are an evaluation of the goodness of a state of affairs, as well as a description of i t . At the outset, we commented on the division of psychological experience into thinking, feeling, and acting, and implied that such separation is inappropriate. Indeed, we have presented and proposed for adoption a conception of emotion that emphasizes cognitive processes as sources of affective quality. And yet there is vocal debate over the primacy of affect over cognition, which is very much a rewording of the debate about the sequence of peripheral and central structures as determinants of emotion, and a demonstration of the separation of thinking and feeling. Zajonc (1980), for example, argues for the primacy of affect (or "preferences need no inferences") against Lazarus (1982) who defends the role of cognition. Mandler (1984) says that the heart of this debate i s how preferences (affect) and inference (cognition) attain consciousness. He suggests that Zajonc's taunt be rephrased as "conscious preferences need no conscious inferences." Candland (1977), on the other hand, questions the very usefulness of arguments about whether emotion or cognition "comes f i r s t " : 20 there i s no special advantage in assuming that there i s a temporal sequence among the three processes, and much waste has come from believing that the appropriate way to untangle our confusions about emotion i s to sort out the temporal sequence. We have created a monstrous problem by assuming that explanation requires uncovering a temporal sequence....(p. 66) Candland supports a conception in which there i s a feedback loop among the elements of emotional experience. Through cognitive and physiological components emotions can feedback and modify perceptions and thus emotional reaction; emotional experience can modify cognitive appraisal and thus modify continuing and future emotional experience, and similarly emotional experience can modify physiological reactions and thus modify continuing experience. It i s appropriate to emphasize that, like in the intended conceptualization of the tr i p a r t i t e mind, emotion and cognition are intertwined. Cognition in humans i s the processing of information in the mind. Unless we believe that emotions are determined outside of the mind or do not involve the processing of information then we must agree that emotion has cognitive-aspects. Again, the crucial distinction i s between the ideational or descriptive aspects of human information processing and the affective or evaluative or emotional aspects (Mandler, 1984). Most often when we contrast cognition and emotion we intend to contrast ideation and evaluation. Both evaluation and ideation occur in the mind and are, therefore, cognitive processes. Emotion and evaluation do, however, have discernible effects on thought and behavior. Moods are self-evaluative emotional states. On the spectrum of emotional experience, moods are less intense values. Berscheid describes moods as " l i t t l e emotions", like emotions in tone but with less intense or no physiological arousal. We can differentiate moods from strong emotion by defining moods as non-21 interrupting states. The effects of moods have been argued to result from automatic processes. More intense emotion demands greater attention, i s associated with more specific response patterns and can interrupt ongoing behavior (Simon, 1967; Mandler, 1984). Moods do not have specific targets, they have pervasive effects, influencing how evaluations are formed and are reflected in the tone of other evaluations, coloring judgements in a congruent way. Moods are temporally limited, they shift over time. They are manifested in subjective experience and in self-reports. In sum, moods are feeling states that are reflected i n subjective experience and self-reports, and that broadly and pervasively affect the character of ongoing behavior and evaluations without interrupting them. The Effects of Mood on Thought and Behavior The consequences of positive and negative mood for thought and behavior are widespread and significant, as we shall see. The effects of mood diffe r from those of strong emotion. Strong positive or negative emotion can alter which behavior or thoughts are continued: i t can interfere with or disrupt ongoing thought and behavior. Indeed emotion has often been maligned on this count (Young, 1961; c.f. Arnold, 1970; Easterbrook, 1959; Leeper, 1970). Mood, on the other hand, does not change which behaviors or thoughts are continued but may affect their tone. So, for example, being in an anxious mood may not disrupt one's choice to engage in a task but may change one's focus of attention. Isen (1984) argues that such effects of mood are more pervasive and of more everyday consequence than the effects of intense emotion. The pervasiveness of the effects of mood i s of obvious consequence to organizational settings. While the interrupting effects of emotion are worthy of study in work settings, interruption i s more obvious. Moods, on the 22 other hand, may change the frequency, duration, or intensity of work behavior without changing i t s occurrence. Intense emotional experience may cause individuals to withdraw from work, such withdrawal i s important. Also important i s the effect of influences on the character and tone of ongoing work. In the following sections, we w i l l review in brief the empirical evidence surrounding the effects of mood on thought and behavior. Much of this material i s drawn from Blaney (1986) and Isen (1984), to which the interested reader i s also referred. Affect and Behavior Positive affect can have broad effects on behavior. Feeling good tends to make one more lik e l y to help others. Finding a coin in a public telephone or succeeding on a test or thinking about positive events i s associated with a tendency to help others, with an important qualification: recent studies have noted that mood protection can occur. Someone feeling good may be less l i k e l y to help someone else i f helping w i l l damage their positive feeling state (Isen & Simmonds, 1978). Similarly, people who feel good may be more likely to behave as they please, helping more when they want to but less when there i s a reason to avoid i t (Forest, Clark, Mills & Isen, 1979). Positive affect i s also associated with a tendency to reward oneself (Mischel, Coates & Raskoff, 1968) and to display increased preference for positive than negative information about the self (Mischel, Ebbesen & Zeiss, 1973, 1976). People who feel good are more willing to i n i t i a t e conversations with others (Batson, Coke, Chard, Smith & Taliaferro, 1979; Isen, 1970) express greater liking and hold more positive conceptions of others and be more receptive to persuasive communications (Veitch & G r i f f i t t , 1976). Positive affect i s associated with greater tendency to take slight risks, but not large ones (Isen, Means, 23 Patrick, & Nowicki, 1982). When risk i s substantial subjects who are feeling good tend to be more conservative (Isen & Patrick, 1983). Negative states have less consistent influences on behavior. Sometimes a symmetry with positive affect i s found such that negative affect has opposite effects, sometimes no effect i s found, sometimes an effect similar to that for positive affect i s found. Isen (1984) suggests that the assumed symmetry between positive and negative affect i s more a convention of language than a reflection of their function. Negative states seem to evoke conflicting tendencies, either to engage in thought and behavior that is compatible with negative mood, or tendencies to change or eliminate the .unpleasantness, including by engaging in affect-incompatible (positive) behavior. For example, negative affect sometimes reduces helping behavior (e.g. Cialdini & Kenrick, 1976; Weyant, 1978), and sometimes increases i t (e.g. Cialdini, Darby & Vincent, 1973; Weyant, 1978). Sometimes negative states seem to have no effect at a l l , as i f the competing tendencies had neutralized each other. Researchers who have found that negative feelings increase positive behavior have proposed that this i s the result of a tendency for negative affect to cause attempts at mood improvement or repair, parallel to a tendency toward mood protection under positive affect (Cialdini et a l . , 1973; Isen, Horn & Rosenhan, 1973; Weyant, 1978). Mood improvement may have more complex effects because tendencies toward positive behavior oppose negative affect, while mood protection effects are generally compatible with positive affect. Isen (1984) points out that both mood repair and protection are compatible with self-regulation theory (Bandura, 1977), which proposes that "experiencing positive affect gives rise to strategies designed to maintain that desirable state, and that negative affect results in strategies aimed at changing the undesirable state" (Isen, 1984, p. 198). 24 Another explanation for "mood repair" effects i s that they represent controlled rather than automatic processes. By the definition proposed here, these effects are not mood effects at a l l . According to this explanation, the negative affect which results in increased helping rather than decreased helping has been of sufficient intensity to shift processing from automatic to controlled. Rather than influence the character of helping behavior by reducing i t , thought may be interrupted such that controlled behavior occurs aimed at improving one's emotional state. Mood and Thought Mood has been argued to influence not only behavior but also the way information i s processed. Creativity, for example, has been shown to be fac i l i t a t e d by positive mood (Isen, Daubman & Gorgoglione, in press; Mednick, 1962; Mednick, Mednick, & Mednick, 1964). Positive mood has been shown to increase the use of an intuitive strategy or heuristic (Isen, Means, Patrick, & Nowicki, 1982), and the use of an efficient simplifying decision strategy (Isen & Means, 1983). Whether such strategies are effective depends on the nature of the decision making task. Isen and colleagues (1982) found that performance was impaired on the tasks they used. Using a more complex decision-making task, Isen and Means (1983) found that decisions made under positive affect were made more quickly and efficiently, although the f i n a l decisions of persons made to feel good did not differ from those of control subjects. Three mental processes describe the effects of mood on thought. These are (1) mood congruence at retrieval, (2) mood congruence at encoding, and (3) affect-state-dependent learning. The f i r s t process, mood congruence at retrieval, refers to improved 25 recall for material which has an affective tone congruent with mood during re c a l l . Underlying this process i s the argument that mood serves as a retrieval cue for memory, like category names or other organizing units. A large number of studies using varied mood inductions have shown that positive words w i l l be recalled more easily during positive moods and negative words wi l l be recalled more easily during negative moods (Blaney, 1986; Isen, 1984). Judgement, evaluation, expectations, decision-making, and behavior follow from these processes. For example, positive feeling associated with receiving an unexpected g i f t results in subsequent raised reports of the performance and service records of consumer goods (Isen, Shalker, Clark & Karp, 1978). Being in a positive mood state causes people to express higher expectations for future success (Feather, 1966). Sometimes, though, while positive affect at retrieval improves recall of positive material, negative affect at retrieval does not improve recall of negative material (e.g. Isen et a l . , 1978; Teasdale & Fogarty, 1979; Teasdale & Taylor, 1981; Nasby & Yando, 1982; Natale & Hantas, 1982). The second process, mood congruence at encoding, refers to the formation of stronger associations in memory when the affective tone of material to be remembered matches the mood state at encoding. So, for example, positive words w i l l be recalled better when learned while in a positive mood and negative words w i l l be recalled better when learned in a negative mood (e.g., Teasdale & Russell, 1983). Bower, Gilligan, and Montiero (1981) found that recall for facts compatible with a positive state was superior when individuals were happy while learning the material, and likewise individuals who learned material while sad showed superior recall of information compatible with a negative affective state. The effects of negative affect on cognitive processes are, like those on 26 behavior, not always entirely clear. Bower (1981; Bower et a l . , 1981) reported that sadness at the time of encoding fa c i l i t a t e d the recall of sad material; Nasby and Yando (1982) did not find such an effect. The third process i s affect-state-dependent-learning, which refers to improved recall for affectively neutral material when mood at recall matches mood during encoding. So, neutral words learned while someone is in a positive mood are more lik e l y to be remembered when that person i s again in a positive mood than when he or she is in a different mood (e.g., Eich & Metcalfe, in press). Affect-state-dependent learning i s the tendency for material learned when an individual i s in a specific mood state to be recalled better when the subject i s again in that state. For example, Bower, Montiero and Gilligan (1978) found recall of a l i s t of words to be better when mood during learning was the same as mood during r e c a l l . Other authors have, however, failed to replicate the effect (Isen, Shalker, Clark & Karp, 1978; Nasby & Yando, 1982). Isen and colleagues have argued that affect-state-dependent learning effects are attributable to the effect of affective state on the retrieval of mood congruent material. That i s , a retrieval rather than learning effect i s responsible. Bower and colleagues argue the reverse, that retrieval effects are the result of state-dependent-learning. Teasdale and Russell (1983) have, however, shown that retrieval effects can occur that are not attributable to affect during learning. They attempt to resolve the discrepancy in the literature by noting differences between studies in the tone of target words, suggesting that the retrieval effect might be state-dependent after a l l . Nevertheless, Isen (1984) concludes that state-dependent-learning as an effect of induced mood should be considered to be a specialized phenomenon, distinct from retrieval effects. In a review of the literature on affect and memory, Blaney (1986) 27 concludes that evidence for affect- state-dependency i s mixed. A number of researchers have failed to replicate Bower and colleagues' (1978) results. That i s , they have not been able to show experimentally that recall i s improved when mood at retrieval matches mood at encoding or learning. In contrast, autobiographical studies do tend to demonstrate such an effect. Individuals have better recall for positive l i f e events when in a positive mood and similarly have better recall for negative l i f e events when in a negative mood (e.g. Bower, 1981). Such autobiographical studies support the affect-state-dependent learning hypothesis while experimental studies appear not to. Bower has since questioned the valid i t y of his original (Bower et a l . , 1978) finding, suggesting that i t was a " s t a t i s t i c a l accident" (Bower, 1985). He has, however, proposed the hypothesis that affect-state-dependent learning does occur when there i s "causal belonging" between the emotion and the event. That i s , "in order to establish effective associations between the emotion and the event to be remembered, the subject would have to causally attribute his emotional reaction to that event" (Bower, 1985, pp. 16-17). According to this hypothesis, the recall of affectively neutral material would not be affected when mood occurs only as a background during learning. Eich and Metcalfe (in press) have proposed a more general hypothesis of affect-state-dependent learning. They argue that state-dependent re c a l l occurs only for items that individuals generate themselves, as opposed to items that they read. Eich and Metcalfe demonstrated that a mismatch between mood state at the time of encoding and retrieval hampered the recall of self-generated items but not read items: 28 events that are generated through internal processes such as reasoning, imagination, and thought are more closely connected to or colored by one's current mood than are those that derive from external sources, and as a consequence, internal events are more apt to be rendered inaccessible for retrieval in the transition from one mood state to another than are external events (Eich and Metcalfe, in press, p. 24). Network Theory The dominant theoretical perspective relating affect and memory and explaining mood congruence and state dependence is network theory (Blaney, 1986). Bower's statement of i t i s as follows: The semantic-network approach supposes that each distinct emotion...has a specific node or unit in memory that collects together many other aspects of emotion that are connected to i t by associative pointers....Each emotion unit i s also linked with propositions describing events from one's l i f e during which that emotion was aroused....These emotion nodes can be activated by many stimuli—by physiological or symbolic verbal means. When activated above a threshold, the emotion unit transmits excitation to those nodes that produce the pattern of autonomic arousal and expressive behavior commonly assigned to that emotion....Activation of an emotion node also spreads activation throughout the memory structures to which i t i s connected, creating subthreshold excitation at those event nodes....Thus...excitation (of) the sadness node...will maintain activation of that emotion and thus influence later memories retrieved (Bower, 1981, p. 135). Blaney (1986) points out that while Bower's statement favors state-dependent effects and i s focused on mood at input, i t can, with modification, describe the effects of mood congruence. A relevant concern i s whether the three seemingly separate processes of mood-state dependence, mood congruence at encoding, and mood congruence at retrieval can be related within a single framework. The revision of the conditions under which state dependence occurs, as suggested by Bower (1985) and Eich and Metcalfe (in press), help us do this. Mood congruence effects at retrieval can be modelled by a semantic-network theory as follows: affectively toned (non-neutral) mental contents are 29 connected to congruent emotion nodes. So, for example, the word "good" i s associated with the emotion for happy and the word "bad" i s associated with the node f or sad. Mood state, as a r e s u l t of natural or experimental induction, a c t i v a t e s the emotion node, spreading a c t i v a t i o n to connected memory u n i t s , creating subthreshold e x c i t a t i o n , and making those memory units more a c c e s s i b l e . Thus, I am more l i k e l y to remember the word "good" when i n a p o s i t i v e mood. At encoding, a c t i v a t i o n of emotion nodes and consequent spreading a c t i v a t i o n strengthens the development of ass o c i a t i o n s . I am more l i k e l y to learn the word "good" when i n a p o s i t i v e mood because that memory unit i s ac t i v a t e d . State-dependent learning can be understood, e s p e c i a l l y i n l i g h t of Bower's "causal belonging" hypothesis, and Eich and Metcalfe's of i n t e r n a l generation, as the process by which previously neutral material acquires some a f f e c t i v e tone or valence. When mental events are generated i n t e r n a l l y a x connection between the a c t i v e emotion node and the event i s established. So, i f I experience a p a r t i c u l a r event as pleasant, the elements of that event become associated with p o s i t i v e emotion. If I generate a word as s o c i a t i o n while i n a good mood that memory unit becomes linked to the congruent emotion node and takes on a p o s i t i v e a f f e c t i v e tone. Or, i f I a t t r i b u t e my p o s i t i v e mood to my performance on a task then that performance becomes connected to emotion nodes congruent with p o s i t i v e mood. R e c a l l for the previously neutral a s s o c i a t i o n or event i s subsequently f a c i l i t a t e d by the process of mood congruent r e t r i e v a l . The fundamental d i f f e r e n c e between mood congruence and mood-state-dependent learning i s the extant valence of the material to be encoded or r e t r i e v e d . In state dependence processes i t i s neutral, i n mood congruence i t i s not. 30 Mood and Arousal Network theory proposes that mental contexts are connected on the basis of their affective tone or valence, basically whether they are positive or negative. Thus positive moods prime positive material and negative moods prime negative material. Implicit in such a discussion i s the idea that moods vary on a continuum from positive to negative. It i s also assumed that this hedonic valence i s the only continuum on which moods l i e and the only basis by which priming in memory occurs. In fact, there i s theoretical and empirical support for the argument that moods also vary along a continuum from arousal to nonarousal, or of arousal—sleepiness. Most theories of emotion e x p l i c i t l y include, or at the least acknowledge, a role for physiological arousal. Arousal is often seen as a necessary part of emotional experience. Further, Russell (1979, 1980) has argued that the expression of emotion is best characterized by two bipolar dimensions. The f i r s t dimension corresponds to what we have called valence, Russell terms i t pleasure— displeasure. The second dimension is arousal—sleepiness. Russell (1980) showed that the relationships between terms describing the complete range of emotion f a l l meaningfully around the perimeter of the space defined by these two dimensions. That i s , they form a circumplex, as shown in Figure 2. In other words, the complete range of human emotional expression can be described in terms of these two dimensions, arousal—sleepiness and pleasure— displeasure. A corollary of this assertion i s that both dimensions are necessary to describe mood states. Depression and anxiety, for example, are similar in valence: they are both unpleasant. Yet they are evidently distinct mood states, a difference which i s captured by recognizing that anxiety i s characterized by more arousal and that depression i s characterized by sleepiness. A R O U S A L DISTRESS MISERY DEPRESSION EXCITEMENT •* PLEASURE C O N T E N T M E N T SLEEPINESS Figure 2. Circumplex model of emotions, from Russell (1980). 32 Arousal and Memory Network theory proposes that mental contents are mutually associated to the degree that they are similar in valence or pleasure—displeasure. It also posits that mental contents are primed by mood states with which they have congruent valence. If, however, moods also vary on an arousal dimension, mental contents might also be primed by mood states with which they are congruent in terms of arousal. That i s , i f a particular unit of memory i s associated with a mild amount of arousal, the presence of a mood state characterized by mild arousal may prime that unit, making i t more accessible to r e c a l l . Just such a mechanism has been proposed by Clark (1982). Clark does not take a position on whether arousal i s necessary to, separate from, or an integral part of what emotion i s . Rather, she assumes that "when a person feels good or bad as a result of some event or thought, that person often experiences increased autonomic arousal" (p.263). This' implies that the valence and arousal dimensions of mood coincide, both good and bad mood are accompanied by the presence of arousal. Russell argues that arousal and valence are orthogonal, although he makes no claims about the physiological concomitants of his arousal dimension. Both good and bad mood can be accompanied by the absence as well as the presence of arousal. According to Russell, the arousal dimension is necessary to distinguish between moods, between a state of pleasant calm and elation, for example, or between anxiety and depression. Thus, whereas Russell argues for separation of the two dimensions, Clark argues for their combination. Clark's conclusion may be a result of a bias in the literature she reviews toward mood inductions that tend to produce excitement rather than calm and anxiety rather than depression. In any case, i t i s evident that mood involves both arousal and valence. 33 What i s of interest here i s the role of arousal in priming recall of mental contents. As Clark points out, the concept of arousal has been l e f t out of most discussions of the influence of mood. She contends that arousal intensifies the priming of memory by positive or negative mood that i s similarly toned. In a study of the effects of arousal or the favorability of judgements, Clark (1981) found that arousal caused by exercise increased the effect of a positive mood induction on evaluations. Arousal did not, however, affect ratings in the absence of pleasant mood. Thus, i t seems that physiological arousal did not have an independent effect on r e c a l l . That i s , increased arousal did not increase ratings independently of mood valence. Clark points out, though, that these results are "consistent with the idea that experiencing arousal in the present w i l l help to prime material previously stored in memory linked with arousal" (p. 280). That i s , although i t was not demonstrated, arousal may, by i t s e l f , cue congruent material in memory, or cue material that i s associated with arousal. In summary, then, i t i s possible that both components of mood, arousal— sleepiness and pleasure—displeasure, may serve to prime associated material in memory. Evidence for an effect of valence i s strong. Certainly mental contents can be more easily placed along the latter dimension than the former. Memories can be pleasant, unpleasant, or neutral. More d i f f i c u l t and more idiosyncratic, i s the classification of memories as associated with arousal or sleepiness. 34 III. EMOTION AND ORGANIZATIONS Emotions themselves are not foreign to organizations or to organizational behavior. As Park, Sims and Motowidlo (1986) point out, affective variables such as satisfaction (Locke, 1976), valence (Vroom, 1964; Mitchell, 1974) and preference (Cyert and March, 1963) have historically been central to our understanding of organization. The early history of the Human Relations movement included discussion of emotion but, following trends in psychology and sociology, this was replaced by an emphasis on the cognitive determinants of behavior. Now, again following i t s intellectual parents, organizational behavior is beginning to refocus attention on emotion. The study of emotion in organizational behavior can, like the study of emotion in general, be simplistically categorized as representing two approaches. Fir s t i s the treatment of emotion as a dependent variable caused by, or a function of, the work setting. One such conception of emotions i s as intrapsychic outcomes that indicate employee well-being and happiness (Rafaeli & Sutton, 1985). Stress research and, more obviously, the job satisfaction literature are examples. Another conception of emotion as a dependent variable focuses on displayed rather than f e l t emotions, particularly as fulfillments of work role expectations. Exemplified by the work of Hochschild (1983) and Rafaeli and Sutton (1987), this approach examines how emotional display comes to form part of the work role and what the consequences of such managed emotion are. This approach adopts a sociological approach to the generation of the rules that govern emotional expression. It does grant, though, that displayed emotion can influence psychological emotional experience. The second approach treats emotion as an independent variable, one that 35 influences processes or variables of organizational consequence. Rather than treat emotion as an outcome of work or as a required role display, emotion is an independent variable that influences thought and behavior. Conceptual frameworks drawn from cognitive psychology and u t i l i z i n g affect have been applied to performance appraisal (Park, Sims, & Motowidlo, 1986; Sinclair, in press), task perceptions and performance (Motowidlo, Packard & Manning, 1985; Park, Sims & Motowidlo, 1984), organizational citizenship (Bateman & Organ, 1983), and turnover (Motowidlo & Lawton, 1984). Emotion as dependent variable Rafaeli and Sutton (1985) contend that emotion has most often been examined as a dependent variable in organizational behavior, predominantly in the vast literature on job satisfaction. Locke (1976), for example, defines job satisfaction as "a pleasurable or positive emotional state resulting from the appraisal of one's job or job experiences" (p. 1300). A large literature has addressed the causes of job satisfaction, such as work rewards, supervision and coworkers, working conditions, task characteristics, individual needs and values, and others. Job satisfaction as emotion. It i s misleading, however, to equate job satisfaction with emotion. Job satisfaction i s also, perhaps most often, defined as a "person's attitude toward work" (e.g. Tosi, Rizzo & Carroll, 1986). Attitudes are affective constructs, that i s , they have an evaluative, "feeling" component. But they are also ideational constructs, representing beliefs about jobs. Job satisfaction i s at least partly such an ideational construct. For example, job satisfaction can include comparison of aspects of one's job with some 36 standard. Such a comparison can be quite distinct from how one feels at work. When asked i f I am satisfied with my salary I may t e l l you that I am not, because I believe that similar jobs are better paid elsewhere. Yet this belief, this dissatisfaction may be quite independent of my emotional state: I may be happy at work. On the other hand, I may be unhappy because I am satisfied with my work. Job satisfaction can be an inappropriate indicator of emotion because of this confounding of affect and ideation. Suppose that we find that some variable influences job satisfaction. Does this mean that emotions have been affected, beliefs have been affected, or both? In the absence of unconfounded measurement we cannot t e l l . In summary, i t is li k e l y that job satisfaction as an organizational outcome taps some of the emotional response to work. But because job satisfaction i s an impure measure of emotional response i t i s unclear exactly what job satisfaction t e l l s us about emotions on the job. For example, what are the sources of emotions on the job? Do they match the dimensions of job satisfaction? That i s , are pay, promotions, supervision, coworkers, or the work i t s e l f sources of emotions at work? Are some job dimensions more li k e l y to arouse an emotional response than others? When we examine the dimensions of job satisfaction i t seems evident that we should exclude non-work causes. The same may not be true for job emotions. It makes sense that an individual's feeling state off the job is l i k e l y to affect his or her feeling state on the job. The literature on job stress recognizes this: non-work-related stressful incidents can have effects on the job. It i s clear that while affective variables such as job satisfaction have received much study, they cannot t e l l as much as we might like about emotions on the job. Future organizational research on emotion per se would help answer these and other 37 questions. Emotional expression as part of the work role The influence of role expectations on the expression of emotion i s a second example of the treatment of emotion as a dependent variable in organizations. Although the expression of feelings i s affected by rules in a l l contexts, this approach has particular relevance to organizational behavior. The workplace contains unique and powerful interpersonal, organizational, and economic forces that author such "feeling rules". Hochschild (1983) calls the management of emotion as a condition of. employment "the commercialization of human feeling". She examined the suppression and distortion of f e l t emotions among flig h t attendants. Most recently, Rafaeli and Sutton (1987) have proposed a conceptual framework for the expression of emotion as part of the work role. This framework i s limited to the display or expression of emotion in contrast to the subjective experience of individual feelings. The emotional displays that Rafaeli and Sutton analyze are "control moves" (Goffman, 1969), or "the intentional effort of an informant to produce expressions that he thinks w i l l improve his situation i f they are gleaned by the observer" (Goffman, 1969, p. 12). Rafaeli and Sutton point out that employees express emotions to promote the interests of others as well, such as clients, superiors, coworkers and subordinates. Rafaeli and Sutton's framework has three parts: the sources of role expectations, the characteristics of displayed emotions, and the outcomes of displayed emotions for the organization and the individual. They propose two sources for the role expectations that create, influence and maintain the expression of emotion. These they c a l l organizational context and emotional 38 transactions. The organizational context comprises the organizational practices that influence the display of emotion: recruitment and selection, socialization, and rewards and punishments. The feedback and cues that employees receive from the "targets" of emotional expression, or the emotional transactions themselves, are another source of role expectations. Customers, clients, superiors, coworkers, and subordinates can provide verbal and nonverbal cues that influence expression. The expression of emotion i t s e l f , as a consequence of role expectations, is the second component of Rafaeli and Sutton's framework. They offer two dimensions of emotions conveyed on the job. First, they posit that expressed emotions can be placed on a continuum from positive through neutral to negative. Smiling i s how positive affect i s expressed, negative affect includes frowning. Second, emotions are said to vary in the extent to which "they enhance the self-esteem of the target. That i s , jobs may require the expression of emotion to support others (social workers, for example), other jobs require neutrality (judges and referees), and some ask employees to degrade the self-esteem of others (e.g., d r i l l sargeants, immigration o f f i c i a l s ) . The f i n a l component of the framework concerns the outcomes of expressed emotion. Both positive and negative outcomes are possible for the organization and for the individual. Organizations benefit, for example, when the positive emotional expression results in increased sales. Hochschild (1983), for example, emphasizes the negative impact of emotion work on psychological and physical well-being. She argues that displayed emotions are harmful when they are inconsistent with true feelings. The result can be a loss of emotional control, physical illness, substance abuse, and absenteeism. The literature on burn out (Maslach, 1978) also supports the costs of 39 emotional labor. Rafaeli and Sutton point out, though, that expressing role appropriate emotion can have positive effects. Learning how to express appropriate emotion may protect physicians, for example, from the negative effects of f e l t emotion. Workers who learn positive emotion management on the job may benefit off the job. Finally, Rafaeli and Sutton provide a provocative discussion of the relationship between expressed emotion and experienced emotion, assessing the consequences of the mismatch between expression, experienced or "true feelings", and feeling rules. Rafaeli and Sutton's contribution to the examination of emotion in organizations i s significant. They have delineated how emotional display can arise from organizations and why: emotional display has outcomes that are salient to the organization and to individuals in organizations. Future research can build on the foundation they have provided, as has already begun (e.g., Sutton & Rafaeli, 1986; Denison & Sutton, in press), and should lead to refinement of their framework. At present, though, and in anticipation of more empirical development, theoretical refinements to the conceptual framework proposed by Rafaeli and Sutton are possible. Of greatest immediate importance i s conceptual c l a r i f i c a t i o n . Role required emotional display i s best understood to be those actions that are under a person's volitional control, the emotions that one chooses to display. This i s certainly part of the conception of expressed emotion as control moves or "intentional effort", as offered by Rafaeli and Sutton. Such a conception implies that emotional display i s functionally independent of emotional experience and that one can choose which emotion to express separate from one's feelings. At times in their discussion, however, Rafaeli and Sutton do not maintain 40 the distinction between displayed and f e l t emotion. For example, they say that expressed emotions can be characterized as positive or negative depending on how they appear. They also say, though, that emotions with the same appearance can be either positive or negative depending on how they are experienced by the role actor. They cite this ambiguity as a limitation of the positive—negative continuum. By adhering to a conceptualization that separates emotional expression as i t appears, from emotional experience as i t feels, this ambiguity can be avoided. Rather than confusing the conceptualization of the construct of expressed emotion by allowing for i t to have different meaning depending on the experience of the encoder, c l a r i t y can be gained by constraining the construct to mean emotions as expressed to and ascribed meaning by a decoder. The point of view of the object of expression rather than the subject should be taken. The second dimension of expressed emotion proposed by Rafaeli and Sutton, whether the display enhances or degrades the esteem of the object of the emotion, takes such a point of view. Rafaeli and Sutton say that a limitation of this dimension i s that individual differences moderate the effect on a person's self-esteem. A more serious criticism of this second dimension i s to ask "why this distinction?" What is special about the effects of displayed emotion on an observer's self-esteem? Other factors could be named, such as those proposed by Kemper (1978, 1984) as the essential dimensions of an interpersonal theory of emotion: power and status. Displayed emotion i s l i k e l y to have many effects on the recipient. In the absence of an exhaustive l i s t of such effects or a rationale for choosing one, I suggest that no single effect of displayed emotion be put forward as especially characteristic. How, then, should the intentional display of role-required emotions be 41 characterized? As Rafaeli and Sutton note, faci a l expressions can be labelled with great r e l i a b i l i t y and accuracy (Ekman, 1984; Leathers and Emigh, 1980). A l i s t of discrete emotions associated with such expression might, therefore, serve as characteristic. A number of researchers have proposed between six and twelve monopolar factors of affect, such as sadness, anger, anxiety, elation, and tension (e.g., Borgatta, 1961; Clyde, 1963; McNair & Lorr, 1964; Nowlis, 1965). More recently, however, methodological refinements have supported the conclusion that affect, including f a c i a l expression, can be represented by the dimensions of pleasure/displeasure and arousal/sleepiness. These could serve as characteristic dimensions of expressed emotion. That i s , the range of required emotional display could be expressed as combinations of these two dimensions. Emotion as an independent variable The effect of emotion on variables in organizations has been studied in only a limited way. As with the study of emotion as a dependent variable, organizational scholars may be unsure what emotion i s . In the following studies, which represent some of the studies of the effects of emotion in organizations, we shall see emotion conceptualized loosely as expression or display, similar to that discussed in the previous section, and with l i t t l e consideration for the process by which such expression has an effect. We shall also review studies that have attempted to use the construct of mood to shed light on the consequences of job satisfaction for performance, and which use job satisfaction as a proxy for mood. As suggested above, such use i s inappropriate because job satisfaction confounds affect with ideation. Finally, we w i l l review the few studies that have used the construct of mood directly. These have examined i t s implications for performance appraisal in 42 organizations. The expression of affect in decision-making Guzzo and Waters (1982) examined the expression of affect in decision-making groups. The considered what happens when groups are called on to make decisions about problems that are l i k e l y to e l i c i t emotional responses. They cite past propositions that emotional expression i s detrimental to effective decision-making: that high levels of emotion hinder clear thinking and communication (Maier, 1963), that group members may lack the s k i l l s necessary for dealing effectively with affect in groups (Argyris, 1966), and that norms against expressing emotion may suppress the sharing of information important for decision-making (Argyris, 1966; Maier, 1963). Guzzo and Waters examined how the timing of the expression of affect influenced decision-making performance. Maier (1963) prescribed the expression of affect early i n the decision-making process. Guzzo and Waters found, however, that groups that followed instructions to delay expression made better decisions than those that expressed i t early. The unregulated expression of affect among control subjects led to decisions of intermediate quality. Guzzo and Waters speculated that early expression of affect may drain the group of productive energy that might otherwise be spent generating alternatives, or that the group's focus of attention may be restricted to a narrow range of alternatives. Of additional interest i s Guzzo and Waters' finding that group participants instructed not to express affect apparently ignored these instructions. Guzzo and Waters concluded that expression of affect i s unavoidable when an emotionally arousing problem exists, that some expression of feeling i s essential to decision-making, and that norms against such 43 expression cannot be completely enforced in some settings. It should be noted that Guzzo and Waters used a decision problem likely to create negative affect, and about which individuals could be expected to feel anxiety and displeasure. Negative mood has been related to narrowed focus of attention (Easterbrook, 1959), thus delay of i t s expression may have been beneficial to the generation of alternative problem solutions. Positive mood, in contrast, can f a c i l i t a t e decision-making (e.g., Isen, Daubman, & Gorgoglione, in press; Mednick, 1962; Mednick, Mednick, & Mednick, 1964). Had the decision problem been one involving the expression of positive emotion early rather than late expression may have led to better decisions. Isen and colleagues point out though, that while mood influences the strategies used to process information, whether such strategies are effective depends on the nature of the task (Isen, Means, Patrick, & Nowicki, 1982; Isen & Means, 1983). Job satisfaction as mood Some researchers have attempted to show evidence of a job s a t i s f a c t i o n — performance linkage by equating job satisfaction with mood. According to Motowidlo: Although mood i s conceptually and operationally different from job satisfaction the two constructs are s t i l l intimately related. People who find their work situations satisfying should generally be in more positive moods than people who do not. Accordingly, the affective response implicit in job satisfaction would have causal effects on behaviors at work (p. 911). Motowidlo (1984) found that job satisfaction was associated with the performance of considerate job behaviors. In a similar vein, Bateman and Organ (1983) argue that prosocial behaviors are most l i k e l y to occur when a person experiences a generalized mood state characterized by positive affect. To the extent that job satisfaction, as conventionally 44 measured, reflects this positive affective state, i t i s lik e l y that more satisfied persons display more of the prosocial, citizenship behaviors (p. 588). Bateman and Organ found significant relationships between satisfaction and prosocial organizational behavior, although the results of a cross-lagged panel regression analysis did not support the predicted direction of causality. Bateman and Organ speculate as to why a causal connection between "mood" and citizenship behavior was not found, including the possibility of methodological weakness or a common antecedent variable. It should be noted that cross-lagged analyses have been c r i t i c i z e d as not validly demonstrating causal effects (Rogosa, 1980). Two additional explanations are possible. First i s that job satisfaction may not reflect mood. As we said, job satisfaction confounds emotional and ideational reactions to work. Further, mood, as defined earlier and as generally conceptualized in studies of i t s effect on behavior, is*understood to be relatively variable over time. Moods shift and change. Mood measurement, by adjective checklists or rating scales, usually specify current mood or, for example, "your feelings right now". Job satisfaction, on the other hand, i s generally understood to be f a i r l y stable. One would not expect job satisfaction to vary from one part of the day to another. A second explanation for Bateman and Organ's result i s suggested by the finding that positive mood does not always lead to prosocial behavior. Recall that while individuals in positive moods tend to perform positive behaviors they may avoid behaviors that threaten to disrupt their good mood (Forest, Clark, Mills, & Isen, 1979; Isen & Simmonds, 1978). Thus the relationship between mood and organizational citizenship behavior may be moderated by the perceived consequences of those behaviors for the mood of the individual. The obvious implication for organizations i s that i t is insufficient to promote 45 organizational citizenship by improving the affective state of workers when the consequences of organizational citizenship are aversive. Mood and performance appraisal Perhaps because of the links between the performance appraisal literature and theories of social cognition, examination of the effects of mood on performance appraisal have been more appropriately conceptualized. Park, Sims, and Motowidlo (1986) reviewed models of the relationship between affect and cognition and considered implications for organizational behavior. In particular, they focused on the role of affect in performance appraisal processes. Because performance appraisal involves a complex of memory-based cognitive tasks, affect is l i k e l y to be an important factor. They draw on models such as Bower's (1981), .in which mood biases the encoding and recall of information. For example, individuals in positive moods are said to be more li k e l y to attend to and remember positive information. This implies that the mood of a rater w i l l influence how performance is rated. This was, in fact, what Sinclair (in press) demonstrated. Sinclair provided raters with information about the performance of a target individual, experimentally manipulated the mood of the raters, and then asked them to evaluate the target individual. Raters in a positive mood provided more favorable evaluations and recalled more positive information than did raters in a negative or neutral mood. However, consistent with the argument that elated individuals make broader categorizations (Easterbrook, 1959; Isen & Daubman, 1984), raters in a depressed mood made more accurate evaluations and displayed less halo error. Many of our theories of individual level organizational behavior are 46 based on individuals as active, although perhaps bounded, information gatherers and processors. The effects of mood on performance appraisal are suggestive of a variety of significant implications for social judgements in organizations. Similarly, differences i n the reca l l and use of information as a result of mood has obvious implications for such topics as organizational decision making, the use of personal interviews in personnel selection, goal setting, and cognitive theories of motivation. Clearly, the effect of emotion i n general and mood in particular in these areas i s worthy of investigation. 47 IV. AFFECT AND MOTIVATION Individual level job performance continues to be the variable of primary interest to the micro side of organizational behavior. As Staw (1984) states, theories of work motivation can be considered most directly relevant to performance, which they are devoted primarily to predicting. Following Campbell and Pritchard (1976) and Mitchell (1982), motivation is defined as the psychological processes that explain the direction, amplitude, and persistence of purposeful voluntary behavior. Expectancy Theory Expectancy or valence—instrumentality—expectancy (VIE) formulations have dominated theories of work motivation. Expectancy formulations agree that motivational force is the product of (a) the expectancy that effort w i l l lead to task accomplishment, (b) the'instrumentality between task accomplishment and i t s outcomes, and (c) the valence of the outcomes. Although c r i t i c i z e d on conceptual and empirical grounds, expectancy theory continues to have the support of motivation theorists. This support may be because i t has strong intuitive appeal, or because the formulation of VIE theory makes i t resistant to empirical rebuttal. For example, although Campbell and Pritchard (1976) report a maximum correlation of about .30 between motivation and independent ratings of effort in correlational f i e l d and experimental studies of the f u l l valence—instrumentality— expectancy models, because VIE theory does not specify the content of outcomes to be considered such weak results can be defended on the grounds that researchers have not included in the study a l l the outcomes relevant to the subjects (Mitchell, 1974). On the other hand, the use of other experimental designs or 48 dependent measures can yield highly supportive results. Wanous, Keon & Latack (1983) revealed an average within-subjects correlation between valence times instrumentality and organizational attractiveness of .72 across 16 within-subjects expectancy studies. Campbell and Pritchard (1976) conclude that while the expectancy framework w i l l continue to have strong heuristic value, i t s practical value would be enhanced by greater attention to i t s theoretical and empirical d i f f i c u l t i e s . In particular they argue for in depth investigation of the individual components of VIE theory rather than tests of a " f u l l " model with "superficial measures of poorly understood variables" (p. 95). Further, they urge more research into the process by which individuals choose to expend a certain level of effort. What, they ask, i s expectancy and how does i t relate to other variables? The goal of this dissertation i s to begin to answer this question with emphasis on the emotional state of individuals. That i s , how are mood and expectancy related? There i s certainly theoretical speculation supporting the proposition that mood influences expectancy. Staw (1984), for example, points out that as a model of individual decision making expectancy theory i s subject to well known limitations on human information processing. The careful screening of alternatives and assessment of rewards that expectancy theory assumes may occur in special circumstances only. Staw advocates increased emphasis on examination of the heuristics and biases that influence decisions in routine situations. One of these i s the availability heuristic (Tversky and Kahneman, 1973), which proposes that, under conditions of uncertainty and limited information processing capacity, individuals estimate probabilities of events occurring based on how easily the event comes to mind. In other words, subjective judgements of the probability of an event depend on how available instances of 49 the event are in memory. To the extent that mood state cues or primes material in memory, making i t more accessible, mood should influence estimates of probability. Included among these is the perceived probability that effort leads to task outcomes, that i s : expectancy. Measuring Expectancy The construct expectancy i s , in Vroom's (1964) original formulation, a subjective probability. Specifically, i t i s the perceived probability associated with the strength of the relationship between individual effort and job outcomes, such as task performance. According to Vroom, i t is the perceived covariation between effort and performance. Effort i s , of course, only one possible manifestation of motivation. Vroom's formulation has also been applied to job choice and job persistence. The discussion here is constrained to the case of effort, although similar lines of argument can be developed for choice and persistence. Vroom described expectancy as the degree to which a person believed that high effort would lead to high performance. This description may explain the high degree of consensus in how expectancy has been measured (Mitchell, 1974). For example, Garland (1984) measured "expectancy" by asking individuals to estimate the likelihood that they could beat a specified performance standard. Similar measurements of "expectancy" have asked subjects to indicate the probability that they could perform an intended behavior (Snyder, Howard, & Hammer, 1978). Others have asked for ratings of the relationship between working hard and performing well (Muchinsky, 1977). In spite of apparent consensus, this way of measuring expectancy is not true to the original concept. Expectancy i s , and should therefore be measured as, the covariation between effort and performance. That i s , expectancy 50 should comprise the perceived probability that high levels of effort lead to high levels of performance, as well as that low levels of effort lead to low levels of performance, and that low levels of effort do not lead to high levels of performance or vice versa. In other words, expectancy measurement should span the rows and columns of the perceived effort-performance covariance matrix. This is not a new admonition. Hollenbeck (1984) proposes a "matrix method" for expectancy research that includes measurement of expectancy at several levels of effort and performance. Hollenbeck's method also proposes, moreover, that instrumentality be measured as the relationship between levels of performance and levels of relevant outcomes. Instrumentality should comprise the link between each of high, medium, and low performance and high, medium, and low levels of outcomes, such as wages and feelings of accomplishment. Further, valence should be measured for-each level of relevant outcomes. Matrix multiplication can then be applied to calculate the motivational force associated with each level of the relevant outcomes. Hollenbeck's method makes at least two significant contributions. First is the recognition that expectancy should be measured at multiple levels. Second i s the demonstration that motivational force exists for levels of relevant outcomes rather than some overall outcome state. In other words, the resultant of expectancies between levels of effort and performance, instrumentalities between levels of performance and outcomes, and valences of levels of outcomes, is a motivational force score, or propensity to direct effort, toward each of a number of performance levels. Hollenbeck points out that force scores are nothing more than the valences of particular effort levels. Unfortunately, no theoretical rationale exists for relating effort to these multiple force scores. Say, for example, that a person has force scores 51 of 10 and 9 for high effort and low effort. The traditional maximization rule predicts that people "will exhibit the single level of effort that corresponds to the single strongest force score" (Hollenbeck, 1984, p. 586). Yet we would be unlikely to predict the same behavior for a person with respective scores of 10 and 1. Hollenbeck points out that "just as the valence of high performance has no particular meaning until i t i s compared to the valence of low performance, the force of high effort has no particular meaning until i t i s compared to the force of low effort" (p. 586). Hollenbeck suggests a probabilistic interpretation of the link between force and effort, in which individuals exert effort in proportion to the relative attractiveness of different levels of effort. Multiple force scores thus predict the distribution of effort over time. Kennedy, Fossum, and White (1983) describe the procedure of obtaining motivational force scores for several effort levels as the "choice model": i t predicts that individuals choose the effort level with the highest product of expectancy, instrumentality, and valence. They cite a number of reasons supporting the choice model as the model of choice. Primarily, the choice model is the one Vroom intended. Kennedy and colleagues cite Vroom's example of three individuals: the f i r s t has high expected payoffs for both high and low effort, the second has high expected payoffs for high effort but not for low effort, and the third has high expected payoffs for neither high nor low effort. "The second person i s the only one predicted to choose high effort, because only he benefits from putting forth high effort. Choice and single alternative (high effort) predictions would be contradictory, since the single-alternative model would have had the f i r s t person also choosing high effort" (Kennedy et a l . , 1983, p. 125). This issue is at the heart of the argument that expectancy is a 52 "within-subjects theory" (Mitchell, 1974). Expectancy theory predicts how individuals w i l l choose from among a set of effort levels at which level of effort he or she desires to work. Kennedy and colleagues add that the strength of a person's motivation at one level of effort i s only meaningful when compared to that person's motivation for other effort levels. This i s the same as saying that, a l l other things being equal, one person's expectancy for one level of effort must be measured alongside his or her expectancies for other effort levels. Expectancy theory thus requires assessment within individuals of the degree to which each of a set of effort levels i s perceived to lead to each level of set of outcomes (Mitchell, 1974; Muchinsky, 1977). There has been much debate about this "within/between issue". Some researchers have argued that a l l between-subjects tests of expectancy theory are invalid. This position rests on two objections: that between-subjects studies do not take into account individual differences, and that between-subjects studies do not assess the choices that individuals make between effort levels, tasks, or outcomes. The f i r s t objection i s , I propose, not founded on a requirement of expectancy theory but is instead one of relative methodological advantage. It i s an issue about the fact that factors which influence motivation may not be included in the expectancy theory model and so, unless stated, measured, and accounted for are included as error variance. Within-subjects analyses use each subject as his or her own control, thus controlling for such unmeasured variation. However, a between-subjects study which completely measured such individual difference variables and controlled for them would, in principle, yield the same results as a comparable within-subjects analysis. Note that the fundamental issue here i s one of measurement. A within-subjects study which implicitly matches subjects across levels of the variable of interest w i l l always outperform a between-subjects 53 study in which a l l the factors relevant to the variable of interest are not measured. Empirical comparisons of within- versus across-subjects designs have generally revealed greater effects for the former (e.g. Kennedy, Fossum & White, 1983; Muchinsky, 1977; Wanous, Keon & Latack, 1983). But this result w i l l hold for most theories, and as such does not invalidate between-subjects tests of expectancy theory. It merely demonstrates that when variation attributable to factors extraneous to the theory i s reduced, the theory predicts more of the measured variation. We should also note that the finding that within-subjects tests of expectancy theory outpredict between-subjects tests implies that the model i s underspecified. That i s , i t omits explicit statement of components, probably related to individual differences, that have predictive va l i d i t y . (Few theories are completely specified, though, and most would be too unwieldy i f we attempted i t . ) However, some within—subjects tests of expectancy have used performance rather than motivation as the criterion. Expectancy theory proports to predict motivation, not performance. It i s generally held that performance i s the resultant of motivation and other factors, including a b i l i t y . So when performance i s used as the predictive criterion in expectancy theory tests, within-subjects studies are bound to outperform between-subjects studies. Individual differences in a b i l i t y , among other factors not included in expectancy theory, are being accounted for. In using a performance criterion as the proxy for motivation within-subjects studies are improperly ascribed some theoretical ascendance. The second objection to between-subjects studies, that they violate assumptions about the nature of expectancy theory i s , I propose, founded on improper conceptualization and operationalization of the expectancy construct. As we have discussed, expectancy should be conceptualized as effort-54 performance covariation, and should be measured as the perceived probability, within individuals, that effort leads to performance across multiple levels of effort and performance. This captures the essence of expectancy theory that people choose high effort as opposed to low effort when the former leads to performance. By understanding and measuring expectancy as within-subjects effort-performance covariation the within-subjects aspect of expectancy theory, that individuals compare choices about actions and their consequences, is maintained. It is then possible to make between subjects predictions, such as that individuals in work roles where objective effort-performance covariation i s higher w i l l perceive greater expectancy (and hence be more motivated) than w i l l other individuals in work roles with lower objective effort-performance covariation. Without the capability of making predictions across people, expectancy theory seems to have few implications for job design. Mood and Expectancy Empirical results suggest a relationship between mood and expectancy. The f i r s t such suggestion comes from the literature on depression rather than mood. Alloy and Abramson (1982) compared the judged contingency between responses and outcomes of self-selected depressed and non-depressed individuals. Students were asked to judge the degree of control their responses exerted over outcomes. Alloy and Abramson found that subjective representations of contingency generally mirror objective contingencies. However, when responses and outcomes are noncontingently related, non-depressed individuals falsely infer control from reinforcement. In other words, non-depressed individuals overestimate the contingency between their actions and outcomes. Depressed individuals on the other hand, are r e a l i s t i c 55 in their judgements whether or not reinforcement i s noncontingently high or low. Apparently non-depressed individuals falsely take credit for their "success". While dispositionally depressed individuals are l i k e l y to experience depressed mood, i t i s also possible that they possess dispositional differences in cognitive style. These, rather than differences in transient mood states, may be responsible for differences in judgements of response-outcome relationships. To assess this, Alloy, Abramson and Viscusi (1981) tested the impact of induced mood on contingency judgements. They found that induction of depressive affect in naturally non-depressed individuals resulted in reduced judgements of control over non-contingent outcomes. Induction of elation among naturally depressed individuals resulted in increased judgements of control. Thus transient mood states do influence the relationship between objective and perceived response-outcome contingency when no contingency exists. Alloy, Abramson, and Viscusi (1981) do not specify the mechanism by which mood state influences judgements of the contingency between responses and outcomes. Differences in the accessibility of events for recall provides one such mechanism (Jennings, Amabile, & Ross, 1982). As we have said already, Kahneman and Tversky propose that people judge the probability of events based on how easily exemplars of the event come to mind, that i s , how available they are in memory. Research on how people construct judgements of the contingency between variables suggests that, when the event i s desirable (e.g., successful task performance), and when co-occurrences of the variables are presented sequentially (i.e., rather than in tabular form), individuals display a bias in favor of the overall probability of the outcome (Allan & Jenkins, 1980; Alloy & Abramson, 1979; Alloy & Tabachnik, 1984; Jenkins & Ward, 1965; 56 Wasserman, Chatlash & Neunaber, 1983). That i s , estimates of the covariation between a response and an outcome are based on the perceived likelihood of the outcome occurring. Thus, to the extent that mood state primes the retrieval from memory of desirable task outcomes, individuals' judgements of the contingency between responses (such as effort) and task performance w i l l be affected. Studies of self-esteem also suggest an effect of mood on expectancy. Cl i n i c a l theories of depression (Beck, 1967) include low self-esteem as a central component of depressive symptoms. At the same time, people with low self-esteem have been shown to have lower expectancies (Brockner, 1979; Campbell & Fairey, 1985; Coopersmith, 1967), but i t is unclear whether or in what combination this effect i s due to the ideational as opposed to emotional components of self-esteem. That i s , individuals with high dispositional self-esteem may believe themselves to be more competent, more able to accomplish goals, or they may be predisposed to more favorable mood states, which may bias estimates of their competence. Conversely, people with low self-esteem may be more susceptible to negative mood and subsequently have lowered expectations for success-. The Beck depression inventory (Beck, 1967), which measures self-esteem (Brockner & Guare, 1983) taps mood state as well as perceived favorability of future events. In sum, the influence of self-esteem on expectancies i s consistent with a mood effect but may be the result of a non-affective cognitive mechanism. More direct evidence comes from studies that have induced mood. Brown (1984) found that subjects in a positive mood state were more confident of future task success. After task performance that was successful or nonsuccessful this expectation remained significantly higher in positive affect subjects and was higher for subjects experiencing success. 57 Furthermore, the interaction of mood and task outcome was significant. After success individuals in the elated condition had more positive anticipation for future success than individuals who failed, while individuals in a negative mood did not show this difference. Teasdale and Spencer (1982, 1984) investigated the effects of mood on estimates of past success and probability of future success. Subjects completed 72 t r i a l s of an "unconscious-decision-making task", in which they had to choose one of a pair of words and received 50% random feedback as to whether they had chosen "correctly" or "incorrectly". After the task depressed or elated mood was induced. Unhappy subjects gave lower estimates of the number of successful t r i a l s than did elated subjects. Similarly, estimates of the probability of future success were lower for subjects in whom depressed mood had been induced. Wright and Mischel (1982) also found that positive mood increased expectations of success on tasks as well as satisfaction with previous performance and self-rated a b i l i t y . Individuals in a positive mood also described themselves in more positive terms. A successful task outcome also improved expectations for the future and satisfaction with performance. Although the interaction effect of mood and outcome on satisfaction f e l l just short of significance, i t was apparent that negative mood state had l i t t l e effect on satisfaction after success, but exacerbated dissatisfaction after failure. It is likely that successful performance helped counteract the effect of negative mood on satisfaction. Wright and Mischel point out though, that feedback about success did not overcome the effect of mood on performance expectations. Despite repeated feedback indicating good performance, negative affect subjects repeatedly underestimated expected performance.. Similarly, positive-affect subjects did not adjust their expectations to match the 58 negative feedback they received. In contrast to Alloy and Abramson (1982; Alloy, Abramson & Viscusi, 1981) negative affect subjects were not more accurate than non-depressed subjects. This incongruity may, though, reflect differences between the measures of expectations used. Wright and Mischel's findings are consistent with the operation of an accessibility mechanism as underlying mood effects, perhaps especially under conditions of uncertainty. They found that the estimates of the frequency of positive task outcomes of positive-affect subjects were significantly higher than those of negative-affect subjects. Positive-affect fa c i l i t a t e d the rec a l l of successful outcomes. Subjects in the successful outcome condition also recalled more successful outcomes. In particular, though, after experiencing positive outcomes, negative-affect subjects underrecalled their positive outcomes. Similarly, after failure positive-affect subjects overrecalled their positive outcomes. In summary, mood appears to be an important influence on individual judgements under uncertainty. While i t thus warrants examination for i t s effect on expectancies i t has not received such attention as yet. Feather (1984a), in summarizing the present status and future directions of expectancy theory, says that while affect has not been neglected i t certainly needs increased consideration. "When affect does appear in expectancy-value models i t often seems to appear almost as an appendage or epiphenomenon, rather than as an integral part of the theory" (1984a, p. 406). Most often affect i s considered as a coincidental outcome of achievement, such as in Weiner and colleagues' analysis of the affective consequences of success and failure (Weiner, Russell & Lerman, 1978, 1979; c.f., Feather, 1984b). 59 Research Hypotheses The preceding review of theory and research, supports investigation of the effect of mood on memory and motivation. In the following sections, specific hypotheses about the relationship between mood states, task perceptions, and individual differences are proposed. Principal among these i s the effect of mood on subjective effort-performance covariation, i.e., expectancy. Hypothesis One: Mood and Expectancy Individuals in a pleasant mood w i l l report higher subjective effort-performance covariation than w i l l individuals in a negative mood. The' studies reviewed above (e.g., Alloy & Abramson, 1979; Alloy, Abramson & Viscusi, 1981; Brown, 1984; Wright & Mischel, 1982) are consistent with and highly suggestive of an effect of mood on expectancy. Although they have even used the term "expectancy", they have not examined expectancy as conceptualized in theories of work motivation, that i s , as the subjective belief that performance and effort covary. Alloy and Abramson (1979) and Alloy, Abramson and Viscusi (1981), for example, examined the effect of mood on the perceived contingency between pressing or not pressing a button and the onset of a light. Brown (1984) and Wright and Mischel (1982) assessed mood effects on expectations that task performance would be successful, as opposed to on expectations that task performance was a function of task effort. The mechanism proposed to underlie this mood effect i s enhanced accessibility in memory: the mood state of an individual cues for recall material in memory that i s affectively congruent with that mood state. When 60 in a positive mood, individuals are more li k e l y to remember positive task outcomes, such as successful task performance. To the extent that individuals perceive themselves as working hard, this increased accessibility of recollections of high performance w i l l increase individuals' perception of the connection between working hard and performing well. To state i t another way, consider a model of individuals' representations of the relationship between effort and performance as a contingency matrix, like that in Figure 3. Each c e l l contains the number of "co-occurrences" of the marginals, such as the number of times that high performance and high effort co-occur, or each c e l l can be taken to contain the probability that the marginals w i l l coincide. Thus, c e l l a, for example, i s the probability that high effort and high performance coincide. Covariation between effort and performance increases as the size of ce l l s a and d increase relative to cells b and c. Thus perfect covariation is when low effort only and always leads to low performance and high effort only and always leads to high performance. According to the mechanism we have described, in which mood influences the recall of congruent memories, when individuals are in a positive mood, they are more l i k e l y to remember the occurrences in cells a and b, or to raise their estimates of the probability that the marginals of those cells coincide. Conversely, when in a negative mood, individuals w i l l raise their estimates of the magnitude of cells c and d. When individuals see themselves as working hard, then, positive mood inflates c e l l a. Negative mood, by the same reasoning, inflates c e l l c. Objectively, we know that covariation depends on how much bigger c e l l a i s than c e l l c, although also on how much bigger c e l l d is than c e l l b. Research evidence suggests that individuals combine contingencies to obtain subjective estimates of covariation in two ways. One way is by comparing the incidence of cells a and d, the "confirming cases", to 61 Effort HIGH LOW HIGH LOW Figure 3. Contingency table model of the relationship between effort and performance. 62 that of cells b and c, the "disconfirming cases" (Shaklee and Tucker, 1980). In this way, a l l four cells are used as in the calculation of real or objective covariation. Alternatively, individuals focus only on cells a and d, the confirming cases (Ward and Jenkins, 1965), and perhaps even only on c e l l a (Smedslund, 1963). Which of these ways best describes how subjective perceptions of covariation are formed has yet to be resolved. But in each of them covariation depends on the relative magnitude of c e l l a. To the extent then that positive mood inflates c e l l a by influencing the accessibility of memories of high performance, perceived effort performance covariation w i l l increase. The implications of mood for job performance are based on the presumption that expectancy i s prospective. That i s , i t i s forward looking. Individuals are hypothesized to make choices about future levels of effort based on their evaluations of expectancy, instrumentality, and valence. When in a good mood, individuals are more likely to remember past successes and increase their estimates of the covariation between effort and performance. Given that valued rewards and performance also covary (instrumentality and valence are high), higher expectancy w i l l result in individuals choosing a higher level of effort. The influence of mood may extend to other task cognitions, such as task interest, d i f f i c u l t y , satisfaction, and perceived task effort. These influences w i l l also be tested by including measures of these cognitions as dependent variables in the studies undertaken below. Other measures, used in the past to measure "expectancy" but which do not measure effort—performance covariation, may also be influenced by mood. Therefore, measures such as the subjective probability of success and the perceived correlation between effort 63 and performance w i l l also be used as dependent variables. Hypothesis Two: Mood and Casual Attributions Individuals in a positive mood state w i l l attribute successful and unsuccessful performance more to internal causes than w i l l individuals in a negative mood state. Weiner, Russell and Lerman (1978) have proposed that affective reactions to success or failure depend on how cause i s attributed. Some emotional reactions to success and failure are posited to be attribution-dependent, i.e., they depend on how cause i s attributed, and other reactions are outcome-dependent but attribution independent. For example, Weiner, Russell, and Lerman (1979) report that happiness i s outcome dependent but attribution independent: happiness i s a reliable result of success no matter to what success i s attributed. Similarly, sadness i s a reliable result of failure. In contrast, pride requires that success be attributed to internal causes. Gratitude results when success i s attributed to external causes. Porac, Nottenburg, and Eggert (1981) have extended this approach to an organizational context. In this study, we are not interested in effects of attributions on mood but rather in the effects of mood on other variables. The present research provides an opportunity to examine how mood influences causal ascriptions. Specifically, i t i s hypothesized that positive mood increases internal attributions for both success and failure relative to negative mood. The rationale for this hypothesis i s related to the explanation for self-serving biases in attribution theory. This body of research has shown that individuals tend to make internal, self-enhancing attributions for positive outcomes and external self-protective attributions for negative outcomes, (see 64 Bradley, 1978, for a review). That i s , they take credit for success and avoid blame for failure. Positive mood may have two influences on this process: i t may increase the perception of success and thus increase the tendency to take credit for i t . Positive mood might also increase internality of attributions by biasing expectancy or effort-performance covariation. If individuals perceive a stronger relationship between effort and performance when in a good mood then they are more li k e l y to attribute success or failure to effort an internal cause. Negative mood, on the other hand, may increase the perceived magnitude of failure and thus the avoidance of blame. Negative mood may also bias effort-performance expectancy downward so that success or failure i s attributed to factors other than effort. In fact, biased expectancies may represent a component of the mechanism by which cause i s attributed. So, i t is hypothesized that positive mood w i l l be associated with internal attributions for performance. Hypothesis Three: Self-esteem and Mood Dispositional self-esteem w i l l moderate the effect of mood on expectancies. Specifically, the effect of mood on expectancies w i l l be stronger for individuals lower in self-esteem. There i s reason to believe that the effect of mood on expectancy may depend in part on an individuals' self-esteem. The f i r s t suggestions for such an effect stems from the finding that global self-esteem i s positively related to prior expectations of success (Campbell & Fairey, 1985; Coopersmith, 1967; Shrauger, 1972). That i s , people who are more self-accepting believe that they w i l l perform better. This, in turn, i s lik e l y to result in higher expectancy. People with higher self-esteem may report higher effort-performance covariation. This may be the result of some ideational process, in which high self-esteem i s associated with more positive self-belief, or of 65 a predisposition among individuals with high self-esteem toward more favorable temporary mood states. Either way, self-esteem i s positively associated with absolute levels of expectancy. This may have the result of imposing a ceiling on the relationship between mood and expectancy among high self-esteem individuals, reducing the strength of the relationship. The second effect i s also possible. It i s based on the finding that individuals with high self-esteem may be resistant to mood manipulations. High self-esteem individuals have been shown to be less sensitive to a negative mood manipulation (Brockner, Hjelle & Plant, 1985). Whether low self-esteem individuals would be less sensitive or more sensitive to positive mood inductions i s not evident. That i s , they may be more or less easily put into a good mood. Nevertheless, high self-esteem individuals may be chronically in a better mood, and may also be resistant to temporary mood inductions and their effects. Thus the effect of mood on expectancies i s li k e l y to be weaker for individuals with high self-esteem. Self-esteem i s of particular relevance to the study of mood and expectancy because i t has been demonstrated to be related to both mood and expectancy. Other variables also of interest are those related to expectancy theory. These include locus of control, or a generalized belief about the source of rewards or reinforcement, and the concept of self-efficacy. In a recent examination of conceptual and methodological issues surrounding self-esteem. Demo (1985) identified two separate dimensions of self-esteem: experienced self-regard measured by self-report and presented self-regard measured by specific others. The latter, presented self, involves behaviors consistent with role requirements and situational demands but not necessarily consistent with the actual or the desired self. It focuses on the 66 level of self-regard communicated to others. Experienced self-esteem, in contrast, i s one's attitude toward oneself, degree of self-respect and perception of self as a person of worth. Experienced self-esteem is most often viewed as a global positive or negative self-assessment. In this view, self-esteem i s a personality t r a i t that i s relatively stable over time and situations. Although there are apparently multiple measures of self-esteem and approaches to i t s measurement, Demo (1985) showed that these cluster around the concepts of experienced as opposed to presented self-esteem. The methodological question of how best to measure self-esteem is thus primarily dependent on which conceptual dimension i s of interest. Demo concluded that the Rosenberg Self-Esteem Scale (Rosenberg, 1979) and the Coopersmith Self-Esteem Inventory (Coopersmith, 1967) were shown to be valid measures of experienced self-esteem. The former appeared, of the two, to have the better r e l i a b i l i t y . It i s experienced self-esteem that i s of interest in the investigation proposed here. Our principal variables are perceptions held by individuals, such as effort-performance covariation or expectancy, and experienced mood state. We are interested in the influence of individual differences in global experienced self-esteem on these perceptions and the relations between them. Experienced self-esteem has also been linked theoretically and empirically to expectancy percepts. In his expectancy theory model of motivation, Lawler (1971) says that individual beliefs about the probability that effort w i l l lead to performance are influenced by "the subject's self esteem . . ., that i s , his general (SR) beliefs about his a b i l i t y to cope with and control his environment (p. 107)." Studies by Brockner (1979), Campbell and Fairey (1985), Coopersmith (1967), Lied and Pritchard (1976) and Schrauger (1972) 67 have found that high self-esteem individuals generally have higher performance expectancies than do low self-esteem individuals. Individuals with more favorable self-regard are more li k e l y to expect that their effort w i l l be successful. Self-efficacy i s a construct that i s related to the components of expectancy theory, although i t i s not, s t r i c t l y speaking, a stable individual difference li k e self-esteem. According to self-efficacy theory, the behavior of individuals, including effort expended on and persistance at tasks, i s determined by their sense of personal mastery or efficacy (Bandura, 1977; 1982; 1986). "An efficacy expectation is the conviction that one can successfully execute the behavior required to produce (certain) outcomes" (Bandura, 1977, p.193). Efficacy expectancy i s contrasted with outcome expectancy, which i s "defined as a person's estimate that a given behavior w i l l lead to (the) outcomes" (Bandura, 1977, p.193). The correspondence between these concepts and the components of expectancy theory is striking. (Especially so in light of seeming claims to innovation on the part of efficacy theorists (e.g., Goldfried & Robins, 1982).) Outcome expectancies are obviously very similar to instrumentality beliefs. Efficacy expectancies are similarly very close to effort-performance expectancy as conceptualized by Vroom (1964). The difference, perhaps, li e s in Vroom's emphasis on the determinants of effort, whereas Bandura allows that successful performance i s a function of a b i l i t y ( s k i l l ) as well as effort. Theories of work motivation assume that performance of the behavior is possible and are concerned with motivating effort. Self-efficacy theory grew out of therapeutic efforts to reduce individual fears about an i n a b i l i t y to perform certain behaviors. Nonetheless, both approaches come together in positing that behavior is a function of belief that the behavior is possible. 68 Bandura differentiates between self-efficacy and self-esteem in at least two ways. Fi r s t , he argues for the independence of self-worth from a sense of mastery. Second, he says that "judgements of personal efficacy do not operate as dispositional determinants independently of conceptual factors" (1978, p. 411). Let us consider each of these arguments in turn. Bandura says that self-esteem and self-efficacy represent two different aspects of self-referent thought. "Self-esteem pertains to the evaluation of self-worth, which depends on how the culture values the attributes one possesses and how well one's behavior matches personal standards of worthiness. Perceived self-efficacy is concerned with the judgement of personal capabilities. Judgments of self-worth and self-capability have no uniform relation. Individuals may regard themselves as highly efficacious in an activity from which they derive no pride (skilled combat soldier) or judge themselves inefficacious at an activity without suffering a loss of self-worth (e.g., inept skater). However, in many of the activities people pursue, they cultivate self-efficacies in what gives them a sense of self-worth. Thus, both self-esteem and self-efficacy contribute in their own way to the quality of human l i f e (Bandura, 1986, p.410). Dispositional self-esteem i s not necessarily associated with judgements of self-efficacy, and self-efficacy does not necessarily influence self-esteem. However, a sense of mastery and competence in a valued activity does contribute to self-regard and individuals are generally able to self-select a c t i v i t i e s to pursue. Unlike global self-esteem, perceived self-efficacy i s not a disposition to view the self in a particular way. For Bandura, efficacy must be assessed at a microanalytic level, with a degree of specificity in the assessment of self-beliefs that matches the specificity of the behavior. If we wish to understand tendencies to perform specific levels of a particular behavior we 69 must measure beliefs about the conviction that those specific levels of that particular behavior can be successfully executed. Although self-efficacy beliefs can generalize in a limited way, Bandura decries the notion of generalized self-efficacy. "It i s no more informative to speak of self-efficacy in global terms than to speak of non-specific social behavior" (1986, p.411). Nevertheless, the idea that individuals possess stable differences i n their degree of self-efficacy across a variety of situations i s appealing. Individuals who have a history of numerous and varied mastery experiences are lik e l y to have positive self-efficacy expectations in general. Scherer and colleagues (Scherer, Maddux, Mercandante, Prentice-Dunn, Jacobs & Rogers, 1982) predicted "that individual differences in general self-efficacy exist and that these differences have behavioral correlates. An individual's past experiences with success and failure in a variety of situations should result in a general set of expectations that the individual carries into new situations. These generalized expectancies should influence the individual's expectations of mastery in the new situations" (1982, p.664). To test their prediction, Scherer and colleagues constructed a self-report measure of general self-efficacy and correlated scores on this instrument with measures of other personality characteristics. Among other results, they found that general self-efficacy was strongly associated with self-esteem (r = .51). While they took this as evidence of convergent validity, the magnitude of this correlation raises concerns about divergent validity, especially in view of Bandura's distinction between self-efficacy and self-esteem. Scherer and colleagues also found a significant association between socially desirable responding (r = .43) and general self-efficacy, raising the spectre of response bias, although a belief in one's efficacy is l i k e l y to be seen as a 70 socially desirable characteristic. These findings, and the novelty of this measure, caution against i t s unbridaled use. As yet, evidence is sparse on what i t does or does not measure. The concept of generalized self-efficacy i s not part of Bandura's theory. Were Bandura to posit a stable individual difference related to self-efficacy expectations, i t i s more li k e l y that he would name i t self-esteem. A construct which was conceptualized as a stable individual difference reflecting a generalized expectation i s locus of control. Rotter (1966) conceived of locus of control as a person's generalized expectancy to perceive reinforcement along a continuum from internal to external. Internal locus of control i s the perception that reinforcement i s contingent on one's own behaviors as opposed to a perception of an external locus of control, that reinforcement is the result of forces beyond one's control and due to chance, fate, or powerful others. The conception of perceived control was part of Rotter's social learning theory. According to this theory, a person's actions are predicted by expectations and values. Specifically, the potential of a person to perform a certain behavior in a situation i s a function of the expectancy that the behavior w i l l be reinforced in that situation and the value of reinforcement in that situation. Locus of control as a dispositional factor or stable individual difference i s the belief that reinforcement i s as a result of "behavior, s k i l l s , or internal dispositions" (Rotter, 1966, p.4). Rotter's theory is an expectancy theory. In contrast to Vroom, however, Rotter did not distinguish between performance expectancy and outcome expectancy. Rotter's focus was clearly on the concept of reinforcement. Yet to the extent that locus of control represents dispositional determinants of perceived expectancy, expectancy for behavior as well as reinforcement, i t is also involved with generalized expectations for the performance of behavior. 71 This i s evident from the content of some of the items of Rotter's measure, which refer to effort, hard work, and the a b i l i t y to do things. As well, elaboration of the components of locus of control has lead to development of subscales that make explicit effort-performance expectancy. Lefcourt's (1981) Multidimensional- Multiattributional Causality Scale includes assessment of beliefs regarding specific attributions of causality to a b i l i t y and effort. Paulhus and Christies' (1981) Spheres of Control measure separates the three behavioral domains of personal efficacy, interpersonal control, and sociopolitical control. Personal efficacy, or control in the sphere of personal achievement i s related to beliefs about a b i l i t y and effort. The locus of control construct thus includes both expectancy for control over reinforcement and expectancy for control over behavior. In summary, while the concept of self-efficacy i s related to effort-performance expectancy, i t i s neither conceptualized as a stable individual difference, nor has substantial empirical evidence been developed to suggest that i t i s a stable individual difference. In contrast, self-esteem i s related to both mood and expectancy and might be expected to influence the relationship between them. Similarly, locus of control i s related to the expectancy and instrumentality components of expectancy theory. Therefore, i t is proposed that locus of control also be measured and i t s possible moderating effects investigated, in addition to testing the hypothesized effects of self-esteem. Proposed Research The previous sections explicate three novel hypotheses relating mood and expectancy. To test these and related hypotheses a laboratory experiment was undertaken. This is described below as Study Four. To provide a foundation 72 for Study Four, three other studies were completed. In Study One, measures of constructs identified in the previous chapter were chosen and administered. This allowed evaluation of their psychometric properties and examination of their inter-relationship. In Study Two mood induction procedures that have been used in experimental settings are reviewed and their s u i t a b i l i t y for testing the hypotheses posed here i s evaluated. A specific induction procedure was thereby chosen and i t s vali d i t y was then tested. In Study Three, the issue of valid measurement of expectancy or the covariation between effort and performance i s discussed. To effect such validation, perceptions of perceived covariation were collected from participants in two tasks which differed in objective expectancy. Finally, Study Four uti l i z e d the measures developed and validated in Studies One and Three, and the manipulation chosen in Study Two, to test the principal hypotheses of this dissertation. These four studies can be summarized as collectively addressing the following questions: What is mood?, What i s expectancy?, and How are they related? 73 V. STUDY ONE: PSYCHOMETRIC PROPERTIES OF MEASURES Overview Study One was undertaken to assess the interrelationships between and dimensionality and r e l i a b l i t y of measures of task perceptions, mood states, and individual differences. Participants completed measures of individual differences, undertook a target task, and completed measures of perceptions of that task and of their mood state. In the following sections, the nature of the participants in Study One and i t s procedure are described more f u l l y . Each measure or set of measures is then described separately with the results of the study for that measure. Subjects Participants were second year business students, recruited to participate in a study of goal-setting and productivity. Participants met in group settings where they were asked f i r s t to complete a "questionnaire about themselves" which contained the individual difference measures, then "to participate in a simulated quality control task", and f i n a l l y to complete a questionnaire to assess their "reactions to and perceptions of the task". This f i n a l questionnaire contained the remaining measures. Participants were told that they were part of a baseline, no goal-setting treatment. Participants were 99 males and 89 females, ranging in age from 19 to 35 (median age = 20). Each received course credit for participating. Confidentiality of responses was promised and demonstrated by having participants place their completed questionnaires in unmarked envelopes. Informed consent forms were completed and collected separately. This procedure was made clear at the outset of the study. The informed consent 74 form i s contained in Appendix A. Following the session, participants were debriefed and dehoaxed. That i s , the purpose of the study and i t s procedures was elaborated and the absence of any deception was emphasized. The procedures used in this study were reviewed and approved by an appropriate institutional ethical review committee. Procedure Participants f i r s t completed the questionnaire containing the measures of self-esteem and impression management. When a l l questionnaires had been completed, they were placed by the respondent in an unmarked envelope. This questionnaire i s shown in Appendix A. Participants were then asked to undertake a task described as a simulated quality control task. They were told that in quality control tasks "a product must be matched against a standard". Participants were provided with sheets of numbers, asked to check the number at the l e f t of each row on each sheet, and then to c i r c l e each number in the row that matched the number at the l e f t . Participants worked at the task for a series of 10 one-minute periods. The purpose of this task was to provide a basis for the measures of task perceptions that followed. The instructions for the task are shown in Appendix A. Following the task, participants completed a questionnaire containing measures of task perceptions, namely satisfaction, d i f f i c u l t y , interest, challenge, performance, effort, and internal work motivation. Also contained in this questionnaire were two measures of mood states, an adjective checklist and a semantic differential measure. The complete questionnaire is contained in Appendix A. The entire procedure, from introduction to debriefing, took approximately 75 90 minutes. Measures and Results In the following sections, each measure i s described and i t s psychometric properties are assessed. Two properties are of interest: dimensionality and r e l i a b i l i t y . Because the measures used were adopted in whole or in part from existing scales, the concern with respect to dimensionality i s one of confirmation, as opposed to exploration. For example, rather than determining how many factors appear to underlie the 48 items in the Multiple Affect Adjective Checklist, we are interested in confirming that the three factors hypothesized by Zuckerman and Lubin (1965) are present. The r e l i a b i l i t y of a measure is i t s freedom from measurement error, specifically from variation over t r i a l s (Guttman, 1945). Measurement error attenuates correlations between variables; unreliable measures degrade an analysis while reliable ones enhance i t . In this study, the criterion for acceptable r e l i a b i l i t y i s 0.70, following Nunnally's (1978) rule of thumb for scales used in research. Two estimates of r e l i a b i l i t y are reported: Internal consistency or coefficient alpha ( a ) i s reported because i t i s most commonly used. Also reported is the highest of Guttman's (1945) six estimates of r e l i a b i l i t y . Guttman showed that each of these estimates is a lower bound on r e l i a b i l i t y , therefore true test-retest r e l i a b i l i t y i s at least equal to the highest estimate. This highest estimate w i l l be referred to as the "Guttman lower bound" or "GLB". Individual differences: Self-esteem and Impression Management. The measure of self-esteem chosen for use was the Rosenberg Self-Esteem Scale (RSE) (Rosenberg, 1965). The measure of impression management chosen was the Impression Management scale of the Balanced Inventory of Desirable Responding 76 (BIDR-IM) developed by Paulhus (1984). A five-point response format was used for the BIDR-IM and the RSE. Individuals were asked to indicate for each statement whether their usual attitude, feeling or behavior was best reflected by the word rarely, occasionally, sometimes, frequently, or usually. As in, for example, the item "I feel that I have a number of good qualities." The RSE consists of ten items which measure self-acceptance, or liking and approving of self. Robinson and Shaver (1973) cite high r e l i a b i l i t y and validity for the RSE, including a test-retest correlation over two-weeks of 0.85 (Silber & Tippet, 1965). Most recently, Demo (1985) substantiated the r e l i a b i l i t y and validity of the RSE in a multimethod, multitrait study. In this study, the internal consistency of the RSE was 0.90. The Guttman lower bound on true r e l i a b i l i t y was 0.91. The BIDR-IM was developed to measure impression management. Impression management or "other deception" is self-presentation that the presenter knows to be false (Paulhus, 1984; Zerbe & Paulhus, 1987). The BIDR-IM contains 20 items about socially desirable but s t a t i s t i c a l l y infrequent behaviors. A high score indicates a tendency to act in a socially acceptable manner, to present a favorable, though false, impression. In this study, one item was deleted from the BIDR-IM because i t created undesirable reactions in a pi l o t sample of students (the item was "I sometimes pick my nose"). In this study the internal consistency of the remaining 19 items was 0.68. The Guttman lower bound was 0.75. Paulhus (1987) reports that measures of impression management are unlikely to have very high internal consistency when respondents are unsure as to exactly what impression they should present. Both the RSE and BIDR-IM exceeded Nunnally's criterion for r e l i a b i l i t y , supporting their use in the research that follows. 77 Task perceptions. Sixteen items were chosen to measure task satisfaction, task performance, task d i f f i c u l t y , task challenge, task interest, task effort, and internal work motivation. Some of these items were adapted from the Job Diagnostic Survey (Hackman and Oldham, 1980), others were written for this study. Participants were asked to indicate their agreement with each item using a five-point scale from "Strongly Agree" to "Strongly Disagree". The sixteen items used are shown in Table 3. To confirm that the seven concepts l i s t e d above were contained in the sixteen items, their underlying factor structure was investigated via factor analytic procedures using the P4M program in the BMD s t a t i s t i c a l software series (Dixon, 1981). Before reporting the results of the factor analyses, however, the practical issues underlying the procedure are considered. Specifically, factor analysis i s sensitive to problems created by outlying cases and variables with skewed distributions. (Tabachnik & F i d e l l , 1983). Univariate outliers were defined as cases with standard scores i n excess of 2.58. Of the 16 items across 188 participants, 24 scores exceeded this criterion. This represents 0.8 %, under the 1 % of scores we would expect due to chance. Therefore, no cases were excluded as univariate outliers. Multivariate outliers were identified by examining the Mahlanobis distance (D2) of each case to the centroid of the sample, computed by BMDPAM. The Mahlanobis distance is distributed as a chi-square variable, and so a c r i t i c a l value for extreme cases can be computed. Eight cases exceeded the c r i t i c a l value for D2 at p 5 .01 and were excluded from subsequent analyses. Thus the f i n a l sample comprised 180 cases. Inspection of the correlation matrix revealed several sizable correlations, making factor analysis appropriate. Examination of the squared multiple correlations (SMC) of each variable with a l l other variables revealed Satisfaction with Task Performance: 1. Generally speaking, I am unsatisfied with my performance on the proofreading task. 2. A l l in a l l , I am very satisfied with my performance on the proofreading task. 3. Compared to other people, I don't think I did very well on the proofreading task. 4. My performance on the proofreading task was high. Task Dif f i c u l t y and Challenge: 5. I found the proofreading task d i f f i c u l t . 6. I found the proofreading task to be easy. 7. The proofreading task was challenging. 8. I didn't find the proofreading task very challenging. Task Interest: 9. The proofreading task was not very interesting. 10. I enjoyed working on the proofreading task. Task Effort: 11. I expended a high level of effort on the proofreading task. 12. Overall, I didn't try very hard on the proofreading task. Internal Work Motivation: 13. I feel a great sense of personal satisfaction when I do well. 14. My opinion of myself goes up when I do well. 15. My own feelings generally are not affected much one way or another by how well I did on the proofreading task. 16. I feel bad and unhappy when I've done poorly. Table 3. Items assessing task perceptions, grouped by construct. 79 that 12 of the 16 items shared variance in excess of 40% with the other variables. The remaining four items, a l l assessing internal work motivation, had SMC's of less than .25 indicating that they contributed less to the analysis. In i t s e l f this does not warrant elimination of these variables. It does, however, suggest that care be taken in subsequent analysis. Principal components extraction was used to estimate the number of factors. Five factors had eigenvalues greater than 1. The scree test (Cattell, 1966) suggested four or five factors. Inspection of the varimax rotated factor pattern for the five factor principal components solution revealed simple structure with items assessing a similar construct loading together for a l l constructs except internal work motivation. Principal factors extraction of five factors revealed that the f i f t h factor was not well defined nor internally consistent. The squared multiple correlation of the factors predicted from scores on the observed variables did not reach .70, Tabachnik and Fidell's (1983) suggested criterion for internal consistency. The i n s t a b i l i t y of this factor was demonstrated further by the results of maximum likelihood factor extraction, in which the f i f t h factor was very poorly structured. On the basis of these results, the four items assessing internal work motivation were excluded from subsequent analyses. Principal components extraction of the remaining 12 items showed four factors with eigenvalues greater than one. The scree test supported four factors as well. Inspection of the varimax rotated factor structure from principal components analysis revealed simple structure for a l l factors. Comparison of the results of principal components, principal factors, and maximum likelihood factor analysis revealed l i t t l e variation in rotated structure, strong evidence for the s t a b i l i t y of the solution. As indicated by SMC's, a l l factors were 80 internally consistent and well-defined by the variables; the lowest SMC for factors from variables was .785. Further, variables were well defined by the factor solution, communality values, shown in Table 4, tended to be high, and a l l variables participated in the solution. Inspection of the matrix of residual correlations revealed a l l to be near-zero. Maximum likelihood extraction was chosen for the fi n a l solution because of i t s appropriateness to cases where the common factor model i s held to be true, as i s the case here where the objective i s the discovery of few common factors. Also, maximum likelihood factor analysis i s generally considered to be superior to principal factors extraction. Maximum likelihood factor analysis assumes multivariate normality. Multivariate normality i s d i f f i c u l t to test (Tabachnik & F i d e l l , 1983). In the present circumstance the likelihood that the data are multivariate normal i s increased by the finding that the data are more or less univariate normal. Inspection of Table 4 reveals that the two items assessing satisfaction and the two items assessing performance define the f i r s t factor, suggesting that these constitute a single construct. Similarly, the items assessing task d i f f i c u l t y and those assessing challenge load together on the second factor. The third factor is defined by two items assessing task interest and the f i n a l factor by two items assessing task effort. No items are complex, that i s , none show a factor loading of more than .45 on more than one factor. This represents a less than 20% variance overlap, the cutoff for inclusion of an item in the definition of a factor proposed by Comrey (1973). The conclusion drawn from this analysis i s that the set of items assessing task perceptions taps four dimensions, corresponding to (1) satisfaction with task performance, (2) task d i f f i c u l t y and challenge, (3) task interest, and (4) task effort. Four scales were constructed accordingly. FACTOR I II III IV h 2 Item 1. Satl .64 -.09 -.02 .02 .42 2. Sat2 .85 -.03 .07 .08 .74 3. Perfl .61 -.07 -.02 .23 .42 4. Perf2 .71 -.11 -.07 .32 .62 5. Dif'fl -.18 .58 -.09 .01 .37 6. Diff2 -.09 .64 .09 .02 .43 7. Chall .06 .75 .38 .17 .73 8. Chal2 -.02 .77 .24 .17 .67 9. Intrl -.09 .30 .68 .07 .57 10. Intr2 .06 .05 .99 .07 .99 11. E f f l .20 .13 .10 .79 .69 12. E££2 .21 .12 .06 .78 .67 Percent of 18.0 17.0 14.6 12.0 Variance Table 4. Factor loadings, communalities (h 2), percent of variance for four-factor maximum likelihood factor extraction on task perception items. Factor loadings subject to Varimax rotation. 82 A f i f t h scale, comprised of the internal work motivation items, was also constructed, although these items do not appear to form a coherent dimension as indicated by the factor analyses above and the r e l i a b i l i t y estimation below. Table 5 shows the means, standard deviations, internal consistency coefficients, and Guttman lower bounds for these scales. With the exception of the internal work motivation scale, these measures have good r e l i a b i l i t y . The lack of coherency in the internal work motivation scale i s again shown by i t s low r e l i a b i l i t y . On the basis of these results, i t was decided that the four reliable scales would be retained for subsequent use and the f i f t h scale discarded. Causal attributions. The Causal Dimension Scale (CDS) (Russell, 1982) was used to assess how individuals perceived the cause of their performance on the proofreading task. The CDS contains 9 items that are presented in a semantic differential format as shown in Figure 4. Three items measure each of internality, stability, and controllability. These three subscales correspond to the causal dimensions described by Weiner (1979). Internal locus of causality refers to whether the cause i s attributed internally, to something about the attributor or to some external cause, outside the attributor. Stability i s defined as whether the cause is constant over time or variable over time. Controllability refers to whether the cause could be changed by the attributor or by someone else. Russell (1982) recommended that individuals be provided with a reason or cause for an outcome and then be asked to use the CDS to rate the cause. He presented evidence for the valid i t y of the CDS: causes that contained objective combinations of internality, s t a b i l i t y and controllability were accurately reflected in ratings on the CDS. Further, the CDS measure differentiated between these three dimensions of causal attributions. Russell (1982) reported r e l i a b i l i t y Items Mean Stand. a GLB Scale Dev. Self-esteem 10 43.40 5.95 .904 .913 Impression-Management 19 73.60 7.44 .681 .745 Task perceptions: Satisfaction with 4 14.18 2.60 .788 .789 performance Di f f i c u l t y and 4 9.49 3.29 .781 .800 challenge Task Interest 2 4.86 2.01 .820 .820 Task Effort 2 8.23 1.32 .831 .831 Internal Work 4 15.48 1.86 .240 .256 motivation Causal attributions: Internality 3 19.72 3.92 .686 .707 Stability 3 14.03 5.86 .806 .807 Controllability 3 19.59 4.00 .474 .484 T Table 5. Descriptive statistics and r e l i a b i l i t y estimates for measures of individual differences, task perceptions, and causal attributions. 84 The next set of questions ask you to consider the reasons behind your actual performance on the proofreading task. For each scale, c i r c l e the number that best describes your impression or opinion of the cause of your performance. Is the cause of your performance something that : Reflects an aspect of yourself 9 8 7 6 5 4 3 2 1 Reflects an aspect of the situation Is the cause of your performance: Controllable by you or other people 9 8 7 6 5 4 3 2 1 Uncontrollable by you or other people Is the cause of your performance something that i s : Permanent 9 8 7 6 5 4 3 2 1 Temporary Is the cause of your performance something: Intended by you or other people 9 8 7 6 5 4 3 2 1 Unintended by you or other people Is the cause of your performance something that i s : Outside of you 9 8 7 6 5 4 3 2 1 Inside of you Is the cause of your performance something that i s : Variable over time 9 8 7 6 5 4 3 2 1 Stable over time Is the cause'of your performance: Something about you 9 8 7 6 5 4 3 2 1 Something about others Is the cause of your performance something that i s : Changeable 9 8 7 6 5 4 3 2 1 Unchanging Is the cause of your performance something for which: No one is responsible 9 8 7 6 5 4 3 2 1 Someone i s responsible Figure 4. Causal Attribution Measures, Study One 85 coefficients of 0.87, 0.84, and 0.73 for the locus of internality, stability and controllability scales, respectively. In this study respondents were not asked to name a cause for their performance but rather to "consider the reasons behind [their] actual performance on the proofreading task." They then used the CDS to describe their impression of the cause of their performance. The internal consistency coefficients and Guttman bounds for the internality, s t a b i l i t y , and controllability scales are shown in Table 5. The s t a b i l i t y scale has acceptable r e l i a b i l i t y , while that of the internality scale i s marginal and that of the controllability scale is unacceptable. It is possible that Russell (1982) found higher r e l i a b i l i t y because he provided explicit causes which respondents were asked to rate. Because i t was his aim to show convergent and divergent validity, i s i s l i k e l y that the causes he provided were more easily reliably classified than the causes for performance which subjects attributed in this study, which were not e x p l i c i t l y provided nor identified. On the basis of Study One, i t was decided to retain the internality and stability scales and discard the controllability scale in further Studies. Multiple Affect Adjective Checklist. Two measures of mood state were administered. The f i r s t was the brief version of the Multiple Affect Adjective Checklist (MAACL) (Zuckerman & Lubih, 1965). The MAACL i s , in i t s f u l l version, an 89 item measure comprising three scales : anxiety, depression, and h o s t i l i t y . The MAACL measures emotional state by means of verbal reports, that i s , by asking respondents to report whether adjectives .describe their feelings "now." Zuckerman and Lubin (1965) report numerous demonstrations of the validity of the component scales, such as relationships with other measures, differences between normal and c l i n i c a l populations, and 86 differences resulting from situational inducements. Zuckerman and Lubin have also constructed a brief version of the MAACL with anxiety, depression and h o s t i l i t y scales of 10, 24 and 14 items, respectively. The items which best discriminated between the f u l l scales were chosen for the brief version. Zuckerman and Lubin report v a l i d i t y evidence for the brief scales close to that of the f u l l scales. Correlations between the brief and f u l l scales are reported by Zuckerman and Lubin as 0.82, 0.93 and 0.92 for the anxiety, depression and h o s t i l i t y scales respectively. The items comprising the brief scales are shown in Table 6. Zuckerman and Lubin report split-half r e l i a b i l i t i e s of 0.79, 0.92, and 0.90 for f u l l scale measures of state anxiety, depression and h o s t i l i t y in a sample of college students. Retest r e l i a b i l i t i e s over seven days were low (0.15 to 0.21) reflecting the fluctuating nature of mood in members of the normal population. While Zuckerman and Lubin (1965) used a yes/no, "checklist" response format, in this study a four-point response scale from "definitely do not feel" to "definitely do feel" was used, as suggested by Russell (1979). This four-point response format for self-descriptive adjectives has been used by Meddis (1972), and shown by Russell (1979) to result in a better response distribution than alternative formats, and to avoid biases due to tendencies" to check most or few adjectives. In this study, participants were asked to indicate, using the four-point scale, how well each of the 48 adjectives described their "feelings right now." (One item, "gay", was changed to "happy", in keeping with more contemporary usage.) The means, standard deviations and internal consistencies of the brief scales of anxiety, depression, and h o s t i l i t y in this study and their Guttman lower bounds are shown in Table 7. As shown, the r e l i a b i l i t i e s for the Anx e t y Depres 5 i on H o s t i i ty P l u s Mi nus Pl u s Mi nus P l u s Mi nus a f r a i d ( 1 ) f e a r f u l ( 1 ) f r i g h t e n e d ( 2 ) nervous(1) p a n i c k y ( 2 ) shaky(1) tense(2) upset( 1) worryi ng(2) calm(2) a l o n e ( 1 ) awful(2) b l u e ( 1 ) d i s c o u r a g e d ( 2 ) f o r l o r n ( 2 ) gloomy( 1) ho p e l e s s ( 2 ) 1onely(1) l o s t ( 2 ) low(1) mi s e r a b l e ( 1 ) r e j e c t e d ( 2 ) s u f f e r i n g ( 1 ) sunk(2) t e r r i b l e ( 1) tormented(2) unhappy(1) wi 1ted(2) ac t i ve(1) a l i v e ( 2 ) f i n e d ) happy(2) heal thy(1) merry(2) a n g r y ( 2 ) c r u e l ( 1 ) d i s a g r e e a b l e ( 1 ) mad(2) a g r e e a b l e ( 1 ) am i a b l e ( 2 ) c o o perat i ve( 1) k i n d l y ( 2 ) p o l i te(1 ) sympathet i c ( 1 ) t e n d e r ( 2 ) u n d e r s t a n d i n g ( 1 ) devoted(2) warm(2) T a b l e 6. Items used from MAACL s h o r t form. Numbers i n parentheses r e f e r t o items used i n p a r a l l e l V e r s i o n 1 or p a r a l l e l V e r s i o n 2. Items Mean St. Dev. a GLB MAACL Anxiety 10 16.48 4.72 .827 .850 Depression 24 40.63 10.97 .931 .951 Hostility 14 27.79 4.99 .765 .801 Semantic Differential Pleasure 6 35.54 7.40 .826 .828 Arousal 6 27.63 8.33 .840 .845 Table 7. Means, Standard Deviations and Rel i a b i l i t y Estimates for measures mood state, Study One. 89 ho s t i l i t y and anxiety scales exceed 0.80, with the r e l i a b i l i t y for the depression scale exceeding 0.90. Pleasure and Arousal. The second measure of mood state administered was Russell and Mehrabian's (1977) measures of arousal-sleepiness and pleasure-displeasure. Each of these i s measured by 6 nine-point semantic differential scales on which respondents are asked to indicate the point that best describes their "feelings right now." Table 7 shows the means, standard deviations, and internal consistencies of the pleasure and arousal scales. The r e l i a b i l i t y estimates of both scales exceed .80. Dimensionality of Mood Measures. Both sets of mood measures used in this study are extant measures with a history of use in vali d i t y and other studies. Zuckerman and Lubin have, for example, demonstrated the divergent and convergent v a l i d i t y of the MAACL scales. It would not be appropriate, therefore, to rearrange the items comprising the MAACL scales, or those of the Mehrebian and Russell semantic differential measure, on the basis of this study. However, the responses of participants do permit us to confirm the dimensionality of the MAACL and semantic differential measures. Factor analytic techniques were used to do this. Again, before factor analysis proceeded the suitability of the data set was evaluated. Inspection of the distribution of scores on the 48 items comprising the MAACL revealed 89 scores across 188 cases for which the standard score was in excess of 2.58 standard units above or below the mean. This represents 0.98%, or about the 1% of scores, as would be expected due to chance. Therefore, no cases were excluded as univariate outliers. Multivariate outliers, those cases with significant (p < .01) Mahlanobis distances, were identified using BMDPAM. Seventeen cases were so identified 90 and excluded from the analyses. Thus the f i n a l sample comprised 171 cases. Inspection of the correlation matrix revealed many sizeable correlations. Each variable shared at least 40% variance with a l l other variables, as indicated by SMC's. Therefore, the factor analysis proceeded. Principal components extraction revealed 10 factors with eigenvalues greater than one. The scree test indicated three or four factors. Because of the presence of three a p r i o r i scales, a three factor solution was selected. Maximum likelihood factor extraction revealed that a l l factors were distinguishable and well defined by the items, the lowest of the SMC's for factors from items was .83. The results of principal components, principal factors, and maximum likelihood factor analysis were very similar, an indicator of the stability of the solution. Maximum likelihood extraction was chosen as most appropriate. The f i t of the solution to the a p r i o r i three factor structure can be assessed by examining Table 8. The items have been ordered according to the scale that Zuckerman and Lubin intended they be part of. Overall, the f i t i s only f a i r . As can be seen, while most of the depression items load on the f i r s t factor, most of the anxiety items load on the second, and most of the hos t i l i t y items load on the third factor, some of the loadings are relatively small and a number of items do not load on any factor. Using a factor loading criterion of .45 for inclusion of an item in a factor, of the 24 items intended to form the depression scale, 19 were included in the Depression factor ("hits"), 1 item was included in the third, Hostility factor ( a "mis-hit") while the remaining 4 items were "misses" rather than "mis-hits": they were not included in any factor. Similarly 7 of 10 anxiety items were "hits, 2 were "mis-hits", loading on the Depression factor, and 1 was a miss. Eight of the 14 h o s t i l i t y items loaded together on the third factor, 2 were mis-91 I II III h 2 Depressior l Anxiety Hostility 1 active 0.38 -0.29 0.27 0.32 4 alive 0.41 -0.31 0.35 0.39 5 alone 0.52 0.22 0.01 0.32 8 awful 0.74 0.06 0.07 0.56 9 blue 0.79 0.05 -0.04 0.63 15 discouraged 0.64 0.31 0.10 0.52 17 fine 0.40 0.14 0.32 0.28 18 forlorn 0.53 0.00 0.01 0.28 20 gloomy 0.75 0.09 0.10 0.58 21 happy 0.44 -0.08 0.44 0.40 22 healthy 0.42 -0.10 0.36 0.32 23 hopeless 0.64 0.25 0.18 0.50 25 lonely 0.63 0.31 -0.03 0.50 26 lost 0.65 0.32 0.00 0.54 27 low 0.79 0.11 0.05 0.64 29 merry 0.37 0.00 0.49 0.37 30 miserable 0.79 0.21 0.14 0.69 34 rejected 0.69 0.21 -0.08 0.52 36 suffering 0.64 0.31 0.02 0.51 37 sunk 0.70 0.21 0.11 0.54 41 terrible 0.76 0.20 0.08 0.62 42 tormented 0.60 0.36 0.09 0.49 44 unhappy 0.67 0.23 0.13 0.53 47 wilted 0.60 0.18 -0.02 0.40 2 afraid 0.28 0.60 -0.15 0.46 10 calm 0.07 0.29 0.04 0.09 16 fearful 0.33 0.57 -0.12 0.45 19 frightened 0.31 0.53 -0.05 0.38 31 nervous 0.12 0.69 -0.04 0.50 32 panicky 0.14 0.61 0.03 0.40 35 shaky 0.22 0.52 0.04 0.32 40 tense 0.22 0.53 0.14 0.34 45 upset 0.69 0.34 0.05 0.60 48 worrying 0.46 0.45 -0.06 0.42 3 agreeable 0.13 -0.10 0.53 0.31 6 amiable 0.17 -0.05 0.36 0.16 7 angry 0.47 0.26 0.06 0.29 11 cooperative 0.09 0.05 0.50 0.26 12 cruel 0.41 0.14 0.13 0.20 13 devoted -0.08 -0.15 0.42 0.21 14 disagreeable 0.44 0.07 0.19 0.23 24 kindly 0.15 -0.04 0.61 0.40 28 mad 0.57 0.27 0.17 0.42 33 polite 0.03 0.04 0.48 0.24 38 sympathetic -0.18 0.08 0.52 0.31 39 tender -0.12 0.09 0.63 0.42 43 understanding -0.04 0.07 0.63 0.40 46 warm 0.12 0.06 0.68 0.47 T Table 8. Factor loadings, communalities (h 2) for maximum likelihood factor extraction on MAACL items. Factor loadings are subject to Varimax rotation. 92 hits, loading on the Depression factor, and the remaining four were misses. Thus, overall, 34 out of 48 items were included in the proper factor, 5 were included in some other factor, and 9 items were not included in the solutions. As Table 8 shows, communalities for these 9 misses were low, as one would expect. A similar set of analyses was performed for the 12 items that make up the semantic differential measures of pleasure and arousal. According to Russell and Mehrebian (1977) these are orthogonal components of mood state. A two factor solution i s thus anticipated. Of the 12 scores for each of 188 cases, 7 scores were identified as extreme, well below 1% of cases. Twelve multivariate outlying cases were identified using BMDPAM and excluded from further analyses. Examination of the correlation matrix and SMC's among items supported factor analysis, the smallest SMC was .36. Having met the practical limitations of the technique, factor analysis proceeded. Principal components extraction revealed two factors with eigenvalues greater than one. A scree test also supported at least two factors. Again, the results of principal components, principal factors and maximum likelihood extractions were very similar. Maximum likelihood extraction was chosen. Inspection of the SMC's for factors from items revealed that each factor was well defined, the lowest SMC was .87. Factor loadings are shown in Table 9. Using a loading of .45 as a criterion for inclusion of an item in a factor, i t is evident that each item loads on only one factor and loads with other items in the same scale. That i s , the pleasure items load together, defining the f i r s t factor, and the arousal items load together, defining the second factor. Russell and Mehrebian's proposed structure i s thus very well supported. 93 I II h 2 Pleasure Arousal Unhappy—Happy 0.71 0.14 0.52 P1ea sed—Annoyed 0.83 0.01 0.69 Unsatisfied—Satisfied 0.75 0.07 0.57 Contented—Melancholic 0.72 0.12 0.53 Despairing—Hopeful 0.60 0.06 0.36 Relaxed—Bored 0.56 0.10 0.33 Relaxed—Stimulated -0.16 0.73 0.56 Excited—Calm -0.06 0.70 0.49 Sluggish—Frenzied 0.26 0.71 0.57 J i t t e r y — D u l l 0.09 0.66 0.44 Sleepy—Wide Awake 0.42 0.67 0.62 Aroused—Unaroused 0.32 0.75 0.67 Table 9. Factor loadings, communalities (h 2) for maximum likelihood factor extraction on Semantic Differential items. Factor loadings are subject to Varimax rotation. 94 Parallel scale construction The measures used in this Study were administered so that their psychometric properties could be evaluated prior to their use in later studies. Of most concern i s the use of self-report measures of mood state as a check on the effectiveness of the experimental manipulation of mood. The high r e l i a b i l i t i e s of the adjective checklist and semantic differential scales reported above certainly support their use. In Study Four below, however, i t w i l l be desirable to measure mood twice during one experimental session: immediately following manipulation of mood and later following the measurement of the principal dependent variables. Measurement at the latter instant as well as at the former checks both the fact and the duration of the manipulation. However, repeating the 60 mood items would constitute a very lengthy manipulation check indeed. Therefore, construction of shorter, parallel measures of each mood state was undertaken. The c r i t e r i a for acceptability of these parallel scales i s that they correlate highly with each other, and that the scales of most interest meet Nunnally's (1978) r e l i a b i l i t y criterion. In Study Four, the mood induction to be used w i l l manipulate mood along a depression-elation dimension. Therefore, the scales of most interest for checking this manipulation are the MAACL depression scale, and both the arousal and pleasure scales of the semantic differential measure. Parallel anxiety, depression, and h o s t i l i t y scales were constructed for the MAACL by including alternate items, subject to the constraint that the versions contain equal proportions of positively and negatively scored items. Thus the parallel forms contained 5, 12, and 7 items respectively. Table 6 identifies which MAACL items were included in each version. Parallel pleasure and arousal scales were similarly constructed for the semantic differential 95 scales. Each scale thus contained 3 items. The responses of participants in Study One were used to evaluate these scales. Table 10 shows the means, standard deviations, and r e l i a b i l i t i e s for each version, the Guttman split-half coefficients, and the correlations between forms. Because of the decrease in items per scale the r e l i a b i l i t i e s can be expected to drop. Fortunately, the r e l i a b i l i t y coefficients of the depression scales remains quite high, exceeding .85. The correlation between forms i s also high. While the r e l i a b i l i t i e s of the anxiety and h o s t i l i t y scales f a i l to reach .70 in three of four instances, they are not of central concern. Therefore, the parallel forms of the MAACL scales were adopted. Because the pleasure and arousal scales contain only 6 items each, construction of shorter, parallel scales was not completed. The reduction in items per scale to 3 is l i k e l y to impact significantly on r e l i a b i l i t y and, in any event, use of the f u l l scales w i l l not increase the questionnaire length substantially. Correlations between measures While not the purpose of Study One, i t is of interest to examine the correlations between measures. For example, what are the correlations between measures of mood? How do they compare with the intercorrelations between MAACL scales reported by Zuckerman and Lubin? What are the relationships between the MAACL and semantic-differential measures of mood? Hypothesis Two states that individuals in a positive mood state w i l l attribute performance to internal rather than external causes. We can examine the correlation between mood and attributions to shed some light on this issue. Hypothesis Three is based partly on the notion that individuals with high self-esteem may be more V e r s i o n One V e r s i o n Two Guttman ( i Sorrel at i on Items Mean St Dev Alpha Guttman Mean St.Dev Alpha Guttman Spl i t Between Hal f Forms MAACL Anx i e t y 5 7.72 2 56 . 729 .734 8.77 2.50 .661 .680 . 848 . 736 Depress i on 12 20.48 6 09 .883 .902 20. 15 5 . 19 .852 .869 .934 .888 Host i 1 i ty 7 13.52 2 74 .651 .678 14.27 2.74 .559 .606 . 797 .662 S e m a n t i c - D i f f e r e n t is il P1easure 3 18 .06 3 77 .738 .739 17.48 4 . 36 .725 .731 . 784 .653 Ar o u s a l 3 13.43 4 75 .674 .678 14.21 4 .09 .749 .750 . .867 .774 T T a b l e 10. Means, s t a n d a r d d e v i a t i o n s and i n t e r n a l c o n s i s t e n c i e s of the S p l i t - H a l f MAACL s c a l e s . Study One. 97 resistant to mood induction. The correlation between self-esteem and static mood states i s therefore of interest. Also of interest are the relations between task perceptions, and between task perceptions and mood states. And fi n a l l y , the correlation between impression management and other measures i s an indication of their susceptibility to socially desirable responding. In the following sections the probabilities associated with individual correlation coefficients are not reported. This i s because the probability associated with each coefficient taken singly i s an inappropriate reflection of the probability associated with the coefficient given that a large number of tests have been performed. That i s , individual probabilities do not control experimentwise error rate. In such a post hoc analysis of very many coefficients, controlling for experimentwise error would diminish stastical power to such an extent that no coefficient would be declared significant. The relationships discussed below must be taken as highly speculative and subject to the cautions that accompany such a "fishing expedition". Zuckerman and Lubin (1965) found very large correlations between their f u l l scales of anxiety, depression and hos t i l i t y , and smaller but significant ones between their brief scales. Specifically, in a sample of 40 c l i n i c a l and non-clinical respondents, they found correlations between the brief scales of anxiety and depression of .82, between anxiety and h o s t i l i t y of .31 and between depression and h o s t i l i t y of .47. Recall that Zuckerman and Lubin u t i l i z e d a checklist rather than the rating scale response format used here. Checklist response formats may be more susceptible to method variance artifacts than rating scale responses and thus inflate correlations. In this sample of 188 students, the correlations between rating scales were more moderate, .54 between anxiety and depression, .15 between anxiety and h o s t i l i t y and .41 between depression and h o s t i l i t y . Note that the relative 98 magnitude of these correlations matches those reported by Zuckerman and Lubin. Examination of the correlations between the semantic-differential and the adjective checklist scales reveals that anxiety i s negatively correlated with pleasure (r = -.34) and positively with arousal (r = .28) as one would expect. Depression i s highly negatively correlated with arousal (r = -.20). This agrees with the location of depression in the Russell (1980) circumplex model as low in both pleasure and arousal. The h o s t i l i t y scale, which i s less well defined by Zuckerman and Lubin in terms of i t s conceptual relationship to anxiety or depression, i s negatively correlated with both pleasure (r = -.49) and with arousal (r = -.21). Support for Hypothesis Two, that positive mood i s associated with attribution to internal as opposed to external causes, can be sought using the pleasure and depression scales as indicators of positive mood and the internality scale of Russell's (1982) CDS. As Table 11 shows, depression i s negatively associated with internality (r = -.12), and pleasure i s positively associated (r = .18). A similar pattern i s present for s t a b i l i t y of attributions (r = -.12; r = .21, respectively). Controllability i s significantly related to pleasure (r = .22) although not to depression ( r = -.08). These results are consistent with Hypothesis Two. Individuals in a positive mood make more internal attributions. Stated in another way, individuals who attribute their performance to internal causes tend to be in a good mood. The design of Study One precludes conclusions as to causality, so we cannot determine whether attributions influence moods, as Weiner, Russell, and Lerman (1979) have shown, or whether moods influence attributions, as posed in Hypothesis Two. It should also be noted that internality, stability, and controllability of attributions are positively related to satisfaction with performance (r = (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (1) S e l f - e s t e e m (2) Impression Management . 239 (3) S a t i s f a c t i o n with task performance . 172 - .078 (4) Task d i f f i c u l t y and cha11enge - .061 .013 -.118 (5) Task i n t e r e s t .056 . 146 - .038 .416 (6) Task e f f o r t .051 - .015 . 364 .229 . 184 (7) I n t e r n a l work m o t i v a t i o n -.151 -.092 - .014 - .015 .046 .220 (8) P e r c e i v e d i n t e r n a 1 i ty .060 - .009 .268 - .032 - .002 . 179 .318 (9) s t a b i 1 i ty . 178 . 160 .240 -. 167 - .045 . 198 .060 .409 ( 10) c o n t r o l l a b i 1 i t y .040 - .006 .207 -.038 . 150 . 171 .068 .208 .082 (11) A n x i e t y -. 247 - .001 - . 216 .113 .021 .009 . 198 - .079 - .094 - . 136 (12) D e p r e s s i o n -.411 -. 137 -.265 .007 - . 156 -.231 .097 - . 122 - . 122 - .085 .538 (13) H o s t i l i t y - . 146 - .229 .050 - .017 -.311 -.115 - .057 - .028 - .067 - .075 . 145 - .409 (14) P l e a s u r e .334 . 194 .209 .093 .340 .274 - .027 . 184 . 206 .219 -.336 --.677 -- .491 (15) Aro u s a l - .078 .091 .035 .261 .369 ' .298 .181 . 100 .038 .019 . 284 • - .200 -- .209 .213 T a b l e 11. C o r r e l a t i o n s between Dependent V a r i a b l e s , Study One. 100 .27, .24, .21, respectively). Taking credit for performance by making an internal attribution i s related to the whether that performance is successful or not, as well as to mood. In Study One, the mean satisfaction with performance was 14.2, rated on a scale from 4 to 20. Overall, then, successful performance was related to internality of attributions and to positive mood, in keeping with Weiner et al.'s model. To investigate whether the relationship between attributions and mood differs as a function of satisfaction with performance, a subgroup analysis was performed. The sample was divided between those scoring above and below the midpoint of the satisfaction with performance measure. The correlations between attributions and mood among those individuals l i k e l y to have experienced their performance as successful (N = 146) were then compared to those who had li k e l y experienced i t as less than successful (N = 42). In the former group, internality, stability and controllability of attributions were positively related to pleasure (r = .16, .16, .19, respectively) but were unrelated to depression (r = -.02, -.09, -.06, respectively). Thus, among people who are successful, internality i s associated with positive mood, at least as measured by the pleasure scale. For people scoring below the midpoint on satisfaction, the only significant association was between pleasure and controllability (r = .26). In this sample, then, attributions for failure were unrelated to mood, individuals who were in a good mood were no more or less l i k e l y to report that they were responsible for their low performance. Overall, then, support is evident for Hypothesis Two when performance is successful. When satisfaction with performance is low, Hypothesis Two i s not supported. Also not supported is Weiner, Russell and Lerman's contention that internal attributions for failure are associated with negative mood. Thus, 101 further investigation i s appropriate, to test the direction of causality inherent in Hypothesis Two and to evaluate the effect of mood on attributions without reference to success or failure. Although static correlations between measures of mood and self-esteem cannot answer the question of whether self-esteem affects resistance to mood change, the correlations in Table 11 do show that mood and self-esteem are related. As we would expect, high self-esteem is positively associated with positive mood, negatively associated with negative mood ( anxiety, r = -.25; depression, r = -.41, p < .001; ho s t i l i t y , r = -.15) and unrelated to arousal (r = -.08). With respect to task perceptions, significant correlations are present between self-reported effort and satisfaction with performance (r = .36), task d i f f i c u l t y and challenge (r = .23) and task interest (r = .18). Task interest was significantly related to d i f f i c u l t y and challenge (r = .42) but unrelated to satisfaction. Satisfaction with task performance was marginally negatively associated with task d i f f i c u l t y and challenge (r = -.12). The pattern of relationships between mood measures and task perceptions is interesting. Satisfaction with performance was significantly negatively associated with anxiety and depression, positively associated with pleasure, and unassociated with h o s t i l i t y or arousal. Task d i f f i c u l t y and challenge was related only to arousal, and the association was positive. Task interest was negatively related to depression and hos t i l i t y , and positively related to pleasure and arousal. Task effort was negatively associated with depression and positively related to pleasure and arousal. Drawing broadly on the labels attached to the mood measures to describe the task perceptions, the pattern of correlations suggests the following conclusions: satisfaction with performance i s moderately related to the 102 pleasantness of mood, whereas task d i f f i c u l t y and challenge i s related to i t s arousal component. Task interest and effort are associated with both positive mood and arousal. Finally, a precautionary note. In this study, sizeable correlations were present between impression management and measures of task interest (r = .15), stab i l i t y of attributions (r = .16), h o s t i l i t y (r = .23) and pleasure (r = .19, p < .01). Impression management, which is the conscious presentation of a socially desirable front, was measured. Relationships between scales which correlate highly with i t are subject to special scrutiny as subject to measurement artifacts (Zerbe & Paulhus, 1987). Problems associated with response bias are not, however, a necessary result of such zero order correlations (Ganster, Hennessey, & Luthans, 1983). From Study One we can conclude that most of the measures used have good psychometric properties. A few have unacceptable internal consistency and have been discarded; they w i l l not be used further. The remainder have acceptable r e l i a b i l i t y . Further, shorter versions of scales which were constructed for the adjective checklist mood measure were shown to have good r e l i a b i l i t y . Finally, investigation of the associations between measures revealed that most were related in the way we would expect and that while some support was evident for hypotheses, further research was clearly necessary. 103 VI. STUDY TWO: WHAT IS MOOD? Study Two was undertaken to demonstrate the experimental manipulation of mood. A valid and reliable mood induction i s required to test the hypothesized effects of mood on expectancy beliefs and task perceptions in an experimental context. This chapter reviews the use of mood manipulations in the literature and evaluates their suitability for the present research. A manipulation i s chosen for use, namely a musical mood manipulation, and an experiment undertaken to show that the manipulation has appropriate effects on various measures of mood. Manipulation of Mood The increasing interest in the relationship between emotion and thought or behavior has lead to the development of a number of mood induction procedures. These have, for the most part, been developed and validated within the context of a particular investigation. Typically a manipulation i s chosen which has face v a l i d i t y (e.g., a "funny" film, failure on a task, "sad" music) or for which a theoretical rationale for an effect on mood exists (e.g., reading self-evaluative statements, remembering or imagining l i f e events). The induction i s supported by the administration of affect-sensitive measures, such as self-reports (e.g., adjective checklists, bipolar adjective scales), or psychomotor tasks (e.g., writing and counting speed). These manipulation checks are themselves constructed on the basis of theoretical or empirical differences between mood states. For example, individuals with depressed mood reliably exhibit slower psychological and motor responding, and c l i n i c a l l y depressed individuals, individuals in a neutral mood state, and elated individuals differ in the extent to which they endorse affective 104 adjectives as self-descriptive. Across differing studies, mood manipulations show convergent val i d i t y : the effects of different inductions generally agree, and the effects of inductions on affect-sensitive tasks match differences between c l i n i c a l and normal populations. The procedures employed in the experimental literature in order to induce mood have included providing feedback as to success and failure (e.g. Feather, 1966; Isen, 1970), hypnosis (e.g. Bower, 1981; Natale & Hantas, 1982), providing an unexpected g i f t (e.g. Isen, Shalker, Clark & Karp, 1978), music (Clark, 1983; Sutherland, 1982), asking individuals to imagine or remember emotion producing events (Nasby & Yando, 1982), watching funny or sad films (Isen & Gorgoglione, 1983), and the Velten technique, which involves reading statements designed to induce mood (Velten, 1968) (for a review of mood induction procedures see Goodwin & Williams, 1982). Some of these procedures are more suitable than others to the investigation of expectancy. Similarly, some manipulations are more suitable to particular research designs. Within-subjects designs, in which each subject receives a l l experimental treatments, are vulnerable to hypothesis guessing and demand characteristics. Because multiple treatments are experienced by participants, they can more easily guess what a study is about and perhaps behave to confirm the hypothesis they have formed. A mood induction procedure employing different films could be susceptible to such artifacts because of the films' obvious content. Similarly, providing an unexpected g i f t might be a reliable way to produce positive mood. Such a procedure has an organizational analogue in the provision of bonuses (c.f., Boggiano & Hertel, 1983). However, a comparable negative induction is not obvious. Taking away a g i f t might have unpredictable effects on mood, as li k e l y producing anger as sadness. 105 In general the c r i t e r i a for a desirable mood induction i s that i t produces mood, that i s has good construct validity, and that i t does not foster threats to internal va l i d i t y . Inducing mood through hypnosis, for example, i s unreliable: some individuals are resistant to hypnotic suggestions, leading to a special variety of self-selection. Hypnosis has also been c r i t i c i z e d as artifactually producing mood-maintenance effects (Isen, 1984). Success or failure on a task has high face val i d i t y as a source of mood effects in organizational settings. In an experimental exploration of the effects of mood on task expectancies such a procedure is undesirable, however, because i t confounds induced mood with information about the probability of task accomplishment. As Kavanagh and Bower (1985) put i t , "feedback about one's success or failure at a task can modify one's emotional state as well as provide 'cognitive information' about one's capability at that specific task" (p.510). In the Velten technique, used most often to experimentally induce mood, subjects read a set of self-referent statements. Those in the elation induction condition, for example, progress from neutral to elated. For instance, from "Today i s neither better nor worse than any other day", to "Things w i l l be better and better today", and "God, I feel great". In the depression induction condition statements progress from neutral to depressive. Although the Velten technique produces a close analogue to naturally occurring mood (Clark, 1983), i t has been c r i t i c i z e d as containing strong demands to act elated or depressed (Buchwald, Strack, & Coyne, 1981; Polivy & Doyle, 1980), and as f a i l i n g to induce mood in some people (Clark, 1983). A more damaging criticism of the Velten procedure i s that i t has a large cognitive aspect (Isen, 1982). The original statements include many phrases that concern beliefs about success (e.g., "I know I've got what i t takes to 106 succeed", and "I am discouraged and unhappy about myself"). Clearly the potential for ideational as well as affective changes is present in such statements. In defence of this critique, Isen points out that effects on thought and behavior of the Velten, hypnosis, success-failure, and gif t procedures generally agree. However, when the dependent variable of interest i s related to achievement or task performance, as in the proposed study, the cognitive statements in the Velten technique represent an unacceptable confound. A mood induction procedure that i s gaining prominence as an alternative to the Velten technique is a musical induction procedure. This involves the use of suggestive music alone (Pignatiello, Camp, & Rasar, 1986) or as a background to subjects' own efforts to develop a particular mood, as is often done (e.g., Clark & Teasdale, 1982; Eich & Metcalfe, in press; Richards, 1981; Sutherland, Newman & Rachman, 1982; Sutton & Teasdale, 1982; Teasdale & Spencer, 1982, 1984). The musical induction procedure has been shown to have good construct validity, produce stronger shifts in mood than the Velten procedure, and to reliably produce mood in most people. Clark (1983) evaluated the Velten and musical mood induction procedures by comparing their effects to those of naturally occurring depressed mood. For example, musical induction of sadness results in higher levels of self-reported sadness or despondency relative to the elation induction (Clark & Teasdale, 1982; Sutherland, Newman & Rachman, 1982 Sutton & Teasdale, 1982; Teasdale & Spencer, 1982, 1984), as does naturally occurring depression relative to elation. The musical induction procedure has also been shown to influence measures of psychomotor retardation, as occurs in naturally occurring depression, such as counting out loud (Clark & Teasdale, 1982; Richards, 1981; Sutton & Teasdale, 1982; Teasdale & Spencer, 1982) and writing 107 speed (Richards, 1981). Subjects in the sadness condition of the musical induction procedure recall more negative words than positive words, while subjects in the elation condition recall more positive words (Clark & Teasdale, 1982). Teasdale and Spencer (1982, 1984) used the musical induction procedure to show that depressed subjects gave lower estimates of the probability of future success and lower re c a l l for past successes than elated subjects. Eich and Metcalfe (in press) have recently used the musical induction procedure to show mood dependent memory effects. Subjects who generated word associations in a particular musically induced mood condition were more l i k e l y to recall them when in the same mood than when their mood was altered. The Velten and musical procedures have been directly compared. Clark (1983) combined the data from Clark and Teasdale (1982), who used the musical induction procedure, and Teasdale and Russell (1982), who used the Velten procedure. These two studies came from the same laboratory, used the same self-report measures and drew subjects from the same student population. Clark reported significantly greater effects of the musical procedure on ratings of despondency and happiness. Eich and Metcalfe (in press) report that in a pil o t study, the Velten technique produced significant but only short lived shifts in self-ratings. When using the Velten technique they found no evidence of mood effects on memory. In contrast, when using the musical technique they were able to show mood effects. Finally, the music induction procedure appears to produce mood more reliably than the Velten procedure. Sutherland et a l . (1982) found that 100% of subjects who underwent a musical mood induction met a predetermined mood change criterion compared to only 68% in a study employing the Velten procedure. Clark (1983) cites Clark and Teasdale's (1982) finding that 87% of 108 subjects report, in a post-experimental questionnaire, that they experienced a genuine change of mood as a result of the musical procedure, as compared to Polivy and Doyle's (1980) finding that only 50% of subjects reported genuine mood change as a result of the Velten technique. In the studies reviewed above, the musical mood induction procedures employed used music in addition to verbal instructions to "to try to develop" the specified mood, usually by "thinking about past events". Clearly specifying a desired mood state raises the possibility that demand characteristics are produced, as Clark (1983) warns. He also says, though, that the music technique might not be effective without such additional effort. Pignatiello, Camp, and Rasar (1986) have, however, developed a musical mood induction that does not include such instructions. They used non-lyrical selections from classical, popular and musical score soundtracks. These were rated by judges, including music therapists, according to how depressing/elating they were. Three inductions were created using selections high in interrater r e l i a b i l i t y . The selections chosen were ordered such that each induction condition started with the same neutral selection and then became either successively more elating, more depressing, or remained neutral. This ordering was based on two principles used in therapeutic settings to alter the mood of individuals: the "Iso" principle and "vectoring" (Shatin, 1970). The "Iso" principle refers to the process of matching the mood of the music to the mood of the subject in order to alter mood with gradual changes in the music. These gradual changes are termed "vectoring", and are the directed movement of music towards the desired goal, such as from sadness to cheerfulness, or as in the case of Pignatiello and colleagues' induction, from a neutral to a happy mood. 109 Pignatiello and colleagues report two assessments of the valid i t y of their induction. In the f i r s t experiment a significant effect of mood on self-reported depression was found. No influence was found on a measure of writing speed or time estimation. A second experiment, however, included a writing speed pretest. Again, significant differences were found for self-reported depression. When pretest writing speed was used as a covariate a significant effect of mood on this psychomotor task was revealed. No effects on time estimation were found. Finally, when subjects who had indicated that they had guessed that the experiment dealt with mood were excluded from the analysis, the same results were evident. In summary, the music induction i s the preferred method for experimentally inducing mood (Clark, 1983; Eich & Metcalfe, in press). It has been shown to produce a strong analog to naturally occurring mood, has produced effects on affect-sensitive tasks, and when used in the absence of verbal instructions to achieve a particular mood state, i s less open to the criticism that demand characteristics are present. It i s therefore proposed for use i n the present study. Specifically, the induction developed by Pignatiello and colleagues w i l l be used. To further demonstrate the validity of this manipulation in the subject population of interest, Study Three was undertaken. Specifically, i t i s predicted that individuals in the Depression condition of the musical mood induction w i l l self-report greater depression than w i l l individuals in the Elation condition. The scores of individuals in the Neutral condition are expected to be intermediate. In keeping with the circumplex model of emotion (Russell, 198x), i t is predicted that the Elation induction w i l l result in high scores on pleasure and arousal and that the Depression induction w i l l result in low scores on self-reported pleasure and 110 arousal. Method Procedure Twenty-five second year business students participated in the study. Each was recruited as part of a study of "Music in the Workplace", and received course credit for his or her participation. Of the participants, 13 (52%) were men, 12 (48%) were women, and their median age was 20 years. Each participant was randomly assigned to one of three mood induction conditions. On arrival to the experimental sessions, each participant was seated in a comfortable chair facing two loudspeakers and provided with the following instructions: Music in the Workplace Informed Consent Form The study in which you are being asked to participate i s part of an investigation of the use of music in the workplace. Many organizations play music as a background to work. The study you are in i s examining the use of music during work and people's perceptions of the use of music. Later this term, we w i l l be asking participants to help with two parts of this study. In the f i r s t part they w i l l be asked to complete a simulated business decision-making task on a computer terminal while music i s played in the background. In the second part, we would like participants to listen to some music and then we w i l l ask them some questions about the music. In today's session, we would like your help with this second part only. So, f i r s t we would like you to si t back and listen to some music that might be used at work, and then we w i l l ask you some questions about your reactions to the music. The music you w i l l listen to wi l l probably evoke different reactions in different people. These honest reactions are what we are interested i n . Your participation in this study i s voluntary, you are free to discontinue participation at any time without penalty. Your responses w i l l be used only for the purposes of this study and w i l l be kept confidential. If you wish to participate,, please sign below, indicating that you have, read this form and give your informed consent to participate in the study. I l l Each p a r t i c i p a n t then l i s t e n e d to one of the three mood induction tapes developed by P i g n a t i e l l o et a l (1986). Each recording l a s t e d 20 minutes. The musical selections included i n each induction are l i s t e d i n Appendix B. Measures Following the mood induction, s e l f - r e p o r t measures of mood state were administered on a computer terminal. A l l p a r t i c i p a n t s were, as business students, f a m i l i a r with the use of the computer keyboard. The dependent measures were presented item by item on the computer video monitor. For each set of items corresponding to one of the measures described below, p a r t i c i p a n t s were f i r s t shown i n s t r u c t i o n s for the measure and provided with an example. Pa r t i c i p a n t s were then asked to cle a r the d i s p l a y and respond to the f i r s t item. Following each response, the screen would be cleared and the next item presented. The program administering the items captured the response to each item as i t was entered. Figures 5 and 6 show the i n s t r u c t i o n s to the f i r s t set of items and the d i s p l a y of the f i r s t item, "active", r e s p e c t i v e l y . Adjective C h e c k l i s t . The f i r s t measure of self-reported mood was the b r i e f version of the M u l t i p l e A f f e c t Adjective C h e c k l i s t (MAACL) (Zuckerman & Lubin, 1968), as described i n Study One. Semantic D i f f e r e n t i a l . The second measure of self-reported mood was Mehrebian and Russell's (1970) 12 item semantic d i f f e r e n t i a l measure of pleasure-displeasure and arousal-sleepiness, also described i n Study One. Response latency. In addition to capturing the numerical response to each s e l f - r e p o r t item, the computer program timed the latency of response. As each item was presented, the computer recorded the number of milliseconds from 112 Music Reaction Questionnaire The following set of questions are designed to measure your reactions to the music you just heard. We would like you to use the words and scales you w i l l see to describe your feelings right now. You w i l l be presented with words that describe feelings. Please use the following scale to indicate how well these words describe your feelings. 1 2 3 4 definitely do not slightly definitely do not feel feel feel feel If the word definitely describes how you feel at the moment you read i t , respond with a 4. . If the word only slightly describes how you feel at the moment, respond with a 3, and so on. Work quickly, do not spend a long time on one word. Ready to go? Figure 5. Instructions for the MAACL active 1 2 3 4 definitely do not slightly definitely do not feel feel feel feel How well does active describe the way you feel? Figure 6 . Terminal display for the item "Active". 113 completion of the item display to completion of the subject's response. This constituted an unobtrusive measure of depressed mood. Placed between the two self-report mood measures were 5 bogus items intended to reinforce the cover story. These asked about the suitability of the music to various kinds of work settings. Results A multivariate analysis of variance (MANOVA) was performed on the data using the SPSSX s t a t i s t i c a l software (SPSS Inc., 1983). Before presenting the results of the MANOVA procedure, the va l i d i t y of assumptions underlying the technique w i l l be evaluated. Evaluation of Assumptions. The assumptions of MANOVA are (1) multivariate normality, (2) homogeneity of variance-covariance matrices, (3) linearity of relationships among dependent variables, and (4) freedom from multicollinearity and singularity. The mathematical model underlying MANOVA is based on the multivariate normal distribution. It requires that the sampling distribution of the means of the various dependent variables in each c e l l are normally distributed. In univariate analysis, the sampling distribution of means can be expected to approach normality for large samples. MANOVA has also been shown to be robust to modest violation of normality i f the violation i s created by skewness rather than by outliers (Mardia, 1971). Tabachnik and Fi d e l l (1983) say that robustness i s ensured with a sample size that produces 20 degrees of freedom for error in the univariate case so long as sample sizes are equal and two tailed tests are used. In this study robustness was assured by equating the sample size in each 114 c e l l of the design. One male participant was discarded randomly from the Depression treatment group. This resulted in a total sample size of 24, 4 participants in each combination of three mood treatments and 2 gender conditions. Nevertheless, normality was assessed by comparing the skew of each distribution of scores to the standard error of skewness. None of the six dependent variables was significantly skewed. Concomitantly, examination of extreme scores revealed no values beyond 3 standard deviations from the mean of each group (to mitigate the effects of low power, the search for outliers and skew was also conducted in the three groups formed by combining gender conditions.) The presence of multivariate outliers was evaluated by examining the Mahlanobis distance (Dz) from each case to the centroid of i t s group. The Mahlanobis distance i s distributed as a chi-square variable, therefore a c r i t i c a l value for extreme cases can be computed. No significant multivariate outliers were found. Significance tests in MANOVA are robust to both heterogeneity of variance and non-normality i f sample sizes are equal and exceed 20 degrees of freedom for error in the univariate case (Hakstian, Roed and Lind, 1979; Tabachnik and F i d e l l , 1983). In this study robustness to violation of the assumption of homogeneity of variance-covariance matrices was guaranteed by equal sample size. Examination of the ratio of largest to smallest variances for each dependent variable across groups revealed some ratios in excess of Tabachnik and Fidell's (1983) suggested criterion of 20:1. None of the univariate homogeneity of variance tests across the six groups was significant, although these were based on small samples. Homogeneity of variance tests across the three treatment groups only, a more powerful test, also showed no significant result, although that for h o s t i l i t y approached significance (p= .062). The multivariate test of homogeneity of dispersion across the three mood treatment 115 groups was not significant (Box's M = 72.43, F(42,1397) = 1.00). The MANOVA model assumes that the interrelationships among a l l DV's are linear within each c e l l . Deviation from linearity w i l l reduce the power of sta t i s t i c a l tests in that linear combinations of dependent variables w i l l not show maximum relationship with the independent variables. The significance test produced by MANOVA are tests of linear relationships. They do not capture nonlinear relationships but at the same time provide unbiased tests of linear relationships. Therefore, non-linearity w i l l result in weaker significance tests, i t biases tests in the direction of greater conservativeness. Linearity was assessed by examining the plots of observed versus predicted residuals. No evidence of gross curvilinearity was found. Multicollinearity occurs when two variables are perfectly or nearly perfectly correlated and show a similar pattern of correlations with other dependent variables. Singularity occurs when one score i s a linear or nearly linear combination of others. Multicollinearity and singularity pose similar problems for multivariate analyses. Specifically, they prohibit or render unstable matrix inversion. Portions of the multivariate solution that follow multiplication by an unstable inverted matrix are also unstable. One method of dealing with multicollinearity or singularity i s to delete the offending variable or variables. Because one variable i s a linear combination of others, this does not result in any loss of information. Multicollinearity and singularity were assessed by examining the squared multiple correlation (SMC) of each variable predicted by a l l others. No SMC's approached 1.0. Similarly, the determinant of the within-cells correlation matrix was non-zero. Homogeneity of regression, or the assumption that the slope of the 116 regression of dependent variables on covariables is equal across cel l s , i s a requirement of analysis of covariance and stepdown analysis in MANOVA. In this study, test of homogeneity of regression were run for a l l variables, since each serves as covariates for a l l others in stepdown F-tests. Homogeneity of regression was achieved for a l l components. Given succesful evaluation of the assumptions underlying the procedure, MANOVA proceeded. Multivariate analyses of variance. A 2x3 multivariate analysis of variance was performed on the six dependent variables: anxiety , depression, hostility, pleasure, arousal and latency. Independent variables were mood treatment (elation, neutral, and depression) and gender (male and female). Total N was 24 with an equal number of subjects i n each c e l l . Wilk's criterion indicated a significant effect of the mood treatment on the combined dependent variables, F(12,26)= 5.96, p < .001. The main effect of gender was not significant, F(6,13)= 1.11, p > .05. The gender by group interaction was also not significant, F(12,26)= 2.02, p > .05. To simplify the discussion that follows, and to provide increased degrees of freedom for subsequent analyses because the gender effect was non-significant, the data for males and females were combined. The multivariate test of significance for the one-way effect of treatment group produced an overall F value of 6.02 (d.f= 12,32; p < .001). This reflected a strong association between the effect of the mood induction treatment and the combined dependent variables (Wilk's A= 0.094; 7?2= .906). Because the omnibus MANOVA shows significant multivariate effects, the univariate effects can be examined. Significant univariate effects of the 117 mood induction treatment were found for the measures of latency, arousal, pleasure, depression, and h o s t i l i t y . The univariate test of significance for the six dependent variables are shown in Table 12. These results indicate that the mood manipulation had effects on the variables of interest, namely response latency, self-reported depression, pleasure and arousal. The pattern of means, shown in Table 13, is as hypothesized. Univariate procedures do not, however, take into account correlations between the dependent variables. Examination of the pooled within-cell correlation matrix, shown in Table 14 reveals, as expected, that the various measures of mood are largely intercorrelated. Bartlett's test of sphericity was significant (x2(15)= 43.46, p<.001) supporting rejection of the hypothesis that the correlation matrix i s an identity matrix or that the variables are independent. Therefore, stepdown analysis i s appropriate. Stepdown analysis determines the contribution of each dependent variable while controlling for i t s relationship to other variables. Stepdown analysis also controls for Type I error rate. In this study, the dependent variables were entered into the analysis in the following order: latency, arousal, pleasure, depression, anxiety and h o s t i l i t y . In the stepdown analysis procedure, the f i r s t variable is tested in an univariate ANOVA. The second variable i s then tested in an analysis of covariance with the f i r s t variable as the covariate. The third variable is tested with the f i r s t two as covariates, and so on. Specification of the order of entry or variables into stepdown analysis i s on the basis of the importance of the dependent variables to the hypotheses of interest, as determined by the researcher (Tabachnik & F i d e l l , 1983). Clearly, some orders w i l l result in a larger number of significant dependent variables than others. Because stepdown analysis accounts for correlations 118 Latency Arousal Pleasure Depression Anxiety Latency Arousal .028 Pleasure -.201 .337 Depression .280 -.128 -.544 Anxiety .311 .451 -.301 .503 Hostility .037 .203 -.594 .383 .577 Table 12. Correlations for dependent variables, Study Two. 119 Univarie ite Stepc [own F d.f. P F d.f. P Dependent Variable Latency- 4.37 2/21 .026 4. 37 2/21 .026 .29 Arousal 9.95 2/21 .001 7. 43 2/20 .004 .43 Pleasure 22.12 2/21 .000 8. 86 2/19 .002 .50 Depression 45.09 2/21 .000 6. 10 2/18 .009 .40 Anxiety 0.69 2/21 .513 1. 95 2/17 .173 .19 Hostility 11.20 2/21 .000 0. 62 2/16 .552 .07 T Table 13. Univariate and Stepdown F-tests, Study Two. 120 Elal :ion Neuti -al Depres ssion Mean St. Dev. Mean St. Dev. Mean St. Dev. Group Measure Combined (n=8) Anxiety 16 .38 4 .44 14 .50 4 .47 17 .37 5 .88 Depression 34 .63 5 .40 39 .25 5 .29 66 .25 9 .91 Hostility 25 .00 5 .81 23 .37 2 .20 32 .62 3 .70 Pleasure 41 .88 4 .32 40 .38 6 .30 25 .62 5 .39 Arousal 29 .87 8 .29 19 .75 5 .87 15 .37 5 .50 Latency 234 .47 38 .67 258 .40 46 .13 312 .04 71 .05 Males (n=4) Anxiety 15 .75 5 .68 13 .50 4 .12 17 .25 2 .06 Depression 38 .25 4 .78 40 .25 6 .39 66 .75 5 .97 Hostility 22 .75 4 .03 24 .25 2 .87 32 .00 3 .16 Pleasure 42 .75 2 .75 40 .25 6 .13 23 .75 6 .50 Arousal 25 .50 5 .26 20 .25 8 .22 14 .75 6 .18 Latency 212 .74 32 .49 231 .64 35 .91 341 .48 79 .54 Females (n=4) Anxiety 17 .00 3 .56 15 .50 5 .19 17 .50 8 .73 Depression 31 .00 3 .16 38 .25 4 .65 65 .76 13 .89 Hostility 27 .25 6 .99 22 .50 1 .00 33 .25 4 .57 Pleasure 41 .00 5 .83 40 .50 7 .42 27 .50 4 .04 Arousal 34 .25 9 .03 19 .25 3 .50 16 .00 5 .59 Latency 256 .21 34 .25 285 .15 42 .03 282 .59 56 .04 T Table 14. Means and Standard Deviations, Treatment and Gender Groups, Study Two. 121 between variables and sets experimentwise Type I error rate, the procedure permits such specification. The results of the stepdown analysis are shown in Table 12. Response latency was strongly associated with mood induction treatment, stepdown F(2,21) = 4.37, p < .05. Strength of association (rj 2) between the independent variable and response latency, as indicated by the ratio of hypothesis to total sum of squares, was .29. As Table 13 shows, mean latency of response was greatest in the depression induction group (312.0) and least in the elation induction group (234.5). After the pattern of differences measured by latency was entered, a difference was also found for self-reported arousal, stepdown F(2,20) = 7.43, p < .01, rj 2 =.43. Self-reported arousal was greatest in the elation group (29.9) and least in the depression group (15.4). The third variable entered, self-reported pleasure, was significantly related to the mood treatment, stepdown F (2,19) = 8.86, p < .01, T}2 = .50. Self-reported pleasure was highest i n the elation induction condition (41.88), although not very much more so than in the neutral condition (40.38) while the depression group was much lower (25.62). Self-reported depression, the fourth variable entered, also made a unique contribution to the overall difference between mood treatment groups, stepdown F (2,18)= 6.10, p<.01, T J 2 = .40. As in the case of self-reported pleasure, the mean for the elation condition was at one extreme (34.63) but did not differ greatly from the neutral condition (39.25) while the depression induction group reported the greatest depressed mood (66.25). Anxiety, which would not have contributed to the difference between groups i n a univariate context, was not significant in the stepdown analyses. Finally, the measure of ho s t i l i t y , which showed a significant univariate F, was not significant when variance i t shared with variables entered into the stepdown analysis before i t was accounted for. 122 Summary It i s evident from the results of the multivariate analysis of variance that significance differences in self-reported mood state resulted from the musical mood induction procedures. No differences in responses were found between males and females. It was also evident from the results that the association between the mood treatment and scores on the combined mood measures was very strong. The valid i t y of the manipulation i s strongly supported. Examination of group means and of the results of stepdown analyses revealed that the mood induction treatment had the expected effect on the measures. The pattern of responses reflected greatest negative affect in the depression induction as indicated by self-reported depression, pleasure and by response latency, a behavioral measure of psycho-motor depression. Least negative affect was indicated in the elation induction condition. The neutral mood induction produced intermediate scores on these measures, although they were closer to those in the elation condition. This suggests that while there i s strong evidence that the elation, neutral, and depression induction conditions produce an ordering of mood from unhappy to happy, the mood induced by the neutral treatment was not unlike that of the elation condition. Conversely, i t could be argued that i t was more d i f f i c u l t to produce elation relative to neutral mood, than i t was to produce depressed mood. The results for the three measures not yet discussed, anxiety, hostility, and arousal, both confirm the valid i t y of the musical mood manipulation and suggest an elaboration. First, the induction had no effect on the self-report measure of anxiety and, after accounting for the contribution of other variables, had no effect on the measure of ho s t i l i t y . This i s evidence of discriminant validity. In combination with evidence that the mood 123 manipulation does influence convergent measures, this provides strong support for the construct validity of the mood manipulation. That i s , the induction does produce relative elation and depression. In our sample, however, i t appears that the mood produced by the neutral induction condition does not diffe r very much in terms of self-reported depression from that in the elation condition. The results for the measure of self-reported arousal suggest an elaboration, however. Recall that the semantic differential measures of pleasure-displeasure and arousal-sleepiness come out of a theoretical framework that proposes that these dimensions are orthogonal and that descriptors of moods form a circumplex marked by these two dimensions. Elation, in this framework, i s characterized by high pleasure and high arousal. Depression i s characterized by low pleasure and low arousal. The results for the elation and depression induction condition f i t this model. Viewed in this way, the neutral induction condition in this study i s shown to have been neutral in terms of arousal-sleepiness, but somewhat pleasant. That i s , examining the mean scores on the arousal and pleasure measures we find that the mean score for self-reported arousal for the neutral induction i s intermediate between that of the elation and depression conditions, being slightly closer to depression than elation. On the pleasure dimension, however, the mean pleasure score of the neutral group i s close to that of the elation group. In other words, rather than producing a mood state that i s neutral in a l l respects, the neutral induction in this study produced a state that Russell would c a l l "calm"; unaroused but pleasant. This finding affirms Russell's contention that in order to describe mood states, two dimensions are necessary; one is not sufficient. Most discussions and most examinations of the validity and impact of mood inductions consider only differences on a dimension of pleasantness. 124 In summary, i t i s evident that the musical mood induction procedure developed by Pignatiello and colleagues produces mood states of elation and depression, as indicated by self-report and behavioral measures. Alternative explanations, such as demand characteristics or hypothesis guessing are unlikely explanations of the effect on the unobtrusive, behavioral latency measure, or of the intermediate impact of the neutral induction. Further use of this manipulation i s therefore appropriate. Discussion: What is Mood? In Chapter Two, which reviewed theories of and approaches to emotions, the position of the fundamental emotion theorists was outlined. This position holds that there are a discrete number of qualitatively different emotions. Joy, for example, i s a different emotion than sadness. In contrast i s the circumplex model of emotions. By holding that the relationships between emotions can be mapped on a two-dimensional space defined by pleasure and arousal, Russell (1980) places joy and sadness on one continuum. Distress and contentment are not qualitatively different, but are different ends of one dimension, representing different combinations of pleasure and arousal. Study Two showed that the musical mood manipulation did produce the hypothesized states of elation and depression. It also affirmed the circumplex model. Both pleasure and arousal measures were affected by the inductions. In other words, both dimensions were needed to characterize the mood induced in Study Two; both pleasure and arousal are necessary components of mood. Mood researchers sometimes adopt a unitary approach to mood, overlooking the arousal component of mood, although i t can illuminate otherwise cloudy research results. Consider, for example, two studies of mood and performance 125 appraisal. In the f i r s t (Sinclair, in press), raters were provided with information about a target individual. The mood of raters was then manipulated using the Velten procedure, which produces happy and sad mood. Raters were then asked to evaluate the target individual. Sinclair found that raters in a positive mood were more positive in their evaluations and recalled more positive information about the target than did raters in a negative mood. He also found that raters in a depressed mood made more accurate evaluations and displayed greater dispersion of ratings across dimensions and hence less halo error, consistent with the argument that elated individuals make broader categorizations (Easterbrook, 1959; Isen & Daubman, 1984). The second study, by Srinivas and Motowidlo (1987), investigated the effect of stress on performance ratings. The authors predicted and found that a stressful event produced negative mood. Although they did not find an effect of mood of the favorability of ratings, Srinivas and Motowidlo did find that ratings made by people who underwent a stressful experience showed less dispersion across performance dimensions. In other words, individuals in a more negative mood displayed more halo, they made narrower categorizations. On the surface, then, the findings of Sinclair oppose those of Srinivas and Motowidlo. In the former study negative mood was associated with greater dispersion of ratings, in the latter negative mood was associated with less dispersion. This seeming contradiction can be resolved i f we consider the relative level of arousal in the two studies. In the f i r s t study the Velten procedure produced elation and depression, depression being characterized in the circumplex model as low pleasure and low arousal. Indeed the Velten depression induction contains statements like "I'm so tired." In contrast, the negative condition in the Srinivas and Motowidlo study 126 i s characterized by high arousal. On the face of i t stressful experiences are more arousing than non-stressful ones. Further, Srinivas and Motowidlo report stronger associations between their treatment and ho s t i l i t y or anxiety than depression. It i s possible, then, that the differences in dispersion in the two studies were a result of the arousal component of mood. The greater dispersion in Sinclair's negative mood group may have been a function of the lower arousal relative to the positive mood group. In the same way the greater dispersion in Srinivas and Motowidlo's positive mood, low stress group may have been a result of lower arousal. In sum, then, arousal is a possible explanation for what otherwise are contradictory results. So, mood i s not a unitary construct. By assuming that i t varies only along one dimension from positive to negative we overlook much of the domain of mood i t s e l f , as well as much of the domain.to which i t might be relevant. Future research in organizational behavior would be wise to acknowledge a more complete conception of mood, recognizing that the manipulation of mood produces many effects and that measures which capture these effects should be employed. 127 VII. STUDY THREE: WHAT IS EXPECTANCY? Study Three was undertaken to develop and validate a measure of perceived effort-performance covariation or expectancy. In the following sections, the theoretical and empirical background to expectancy measurement i s reviewed. Conceptualizing and measuring expectancy Expectancy, as conceptualized by Vroom (1964), i s the perceived covariation between effort and performance. As we discussed earlier, to measure this construct i t is necessary to assess not only a person's belief that high effort w i l l lead to high performance, but also his or her belief about the extent to which low effort leads to high performance, high effort leads to low performance, and so on. Expectancy can be measured in keeping with Vroom's conceptualization by assessing individual's perceptions of the probability that multiple levels of effort are associated with multiple levels of performance, and then combining those multiple perceptions in a way that reflects covariation. This approach to measurement of expectancy was pioneered by Ilgen, Nebecker & Pritchard (1981). However, i t has not been validated, as we shall see. Ilgen and colleagues point out that while Vroom's conception of expectancy was one of effort-performance covariation, many studies have overlooked this in favor of single probability estimates of the link between high effort and high performance. Ilgen, Nebecker and Pritchard's approach was to measure the degree of association between a l l levels of effort and performance. They asked respondents to indicate the probability that exerting a given level of effort would result in a given level of performance. Ilgen, Nebecker and Pritchard also measured the association between effort and performance by asking for estimates of the frequency of effort—performance 128 combinations and by using verbal indicators of probability. For each of these three ways of measuring the link between levels of effort and performance two composite scores were constructed. The f i r s t was an "expected value index", reflecting the number of work units an individual expected to complete. Given a set of n x m probabilities p(i,j) relating each of m levels of effort C(j) to each of n levels of performance R(i) where C(j) and R(i) are the values or weights assigned to each level of effort or performance, the expected value index (EVI) i s calculated as follows: EVI = Z R(i) Z p(i,j) C(j) (Z Z p(i,j) C(j)) The second composite score was a coefficient reflecting "the degree of linear covariation" between effort and performance. This uses the same matrix of responses, treating the values as frequencies in a bivariate distribution. Based on these frequencies, covariation between the two variables was calculated in the same way that a correlation coefficient i s calculated. Using the notation from above, and where E(R) and E(C) are the weighted mean levels of perceived performance R and effort C, respectively, the perceived covariation measure (COV) i s calculated as follows: COV = Z Z (R(i) - E(R)) (C(j) - E(C)) p(i,j) • Z (R(i) - E(R)) 2 p(i) • Z (C(j) - E(C)) 2 p(j) Consider, for example, the assessment of the perceived covariation between three levels of effort and four levels of performance. Respondents are asked twelve questions about the probability that a specified level of effort w i l l lead to a specified level of performance, for each combination of effort and performance levels. These are the p ( i , j ) . Unit weights are 129 assigned to effort and performance levels. That i s , the lowest effort and lowest performance levels are assigned a weight of 1, the next highest levels a weight of 2, and so on. Alternatively, real values can be used as weights, as in the study by Ilgen, Nebecker, and Pritchard. These are the C(j) and R(i), respectively. It i s important to note that calculation of the product moment correlation between two variables from a matrix of probabilities requires that the c e l l entries are joint probabilities. That i s , probabilities of the joint occurrence of the specified levels of the variables, such as for example, that high effort and high performance co-occur. The probabilities p(i,j) above, however, are conditional probabilities. They represent the probability of a specified level of performance occurring given a specified level of effort. The perceived covariation coefficient i s not, therefore, s t r i c t l y equivalent to a correlation coefficient. The two coefficients are equal when the conditional probabilities and the joint probabilities are equal, which occurs when the marginals are equal, that, i s , when each level of effort i s perceived to be equally l i k e l y . By asking individuals to report conditional probabilities we are asking for the joint probabilities that would hold i f the marginal probabilities were equal, i f high, medium, and low effort were equally probable. Variation in the marginal probabilities results in restriction of the range of the correlation coefficient. For example, i f the perceived probability of low effort i s zero then the range of possible correlation between effort and performance i s much less than i f the perceived probability of effort i s much more than zero. Different individuals may hold different perceptions of the probability of expending low effort or high effort. Yet these same individuals may have similar perceptions of the degree to which low or high 130 effort lead to high performance, or in general of the relationship between effort and performance. By constraining individuals to report conditional probabilities this problem i s avoided. This approach ensures that the perceived covariation coefficient has the same possible range for each respondent. Effort Level 1 2 3 1 40 10 0 Performance 2 25 20 10 Level 3 15 25 20 4 5 10 35 For the matrix shown above, treating the c e l l values as probabilities in a bivariate distribution, f i r s t dividing each value by the sum of a l l values so that the total probability i s unity, the strength of relationship between effort and performance i s .61. This score reflects the degree of linear covariation between effort and performance. That i s , i t reflects expectancy. Inspection of the matrix reveals that low effort i s most l i k e l y to lead to low performance and not l i k e l y to lead to high performance. High effort effort i s most l i k e l y to lead to high performance and is not perceived to be at a l l l i k e l y to lead to low performance. Effort Level 1 2 3 1 25 10 10 Performance 2 25 20 15 Level 3 25 25 20 4 10 10 20 For the second matrix shown above, in which i t i s evident that a low level of effort i s equally l i k e l y to result in the three lowest levels of performance and high effort i s perceived to be equally l i k e l y to lead to the highest and next to highest level of performance and may even lead to low performance, the strength of relationship between effort and performance is 131 .22. In contrast to the covariation index, the expected value index is not a measure of covariation. It does not reflect perception of the extent to which effort and performance covary, or the degree to which high effort leads to high performance and not low performance, or the degree to which low effort leads to low performance and not high performance. Consider the case where an individual believes with complete confidence (100% probability) that he or she wi l l achieve a high level of performance i f he or she exerts a high or medium or low level of effort, as follows: 0 0 0 0 0 0 100 100 100 In this case effort and performance do not covary. High performance is associated with a l l effort levels, and expectancy should have a value" of zero. In this case the expected value index is numerically equal to the value assigned to the high level of performance, ( i f unit weights are used, 3). Consider a second case where a person believes with complete confidence that no matter what level of effect i s expended that he or she w i l l achieve a medium level of performance, as follows: 0 0 0 100 100 100 0 0 0 Again, subjective effort-performance covariation i s absent, yet in this case the expected value index is different from that in the previous case; i t is equal to the value assigned to the medium level of performance, ( i f unit weights are used, 2). Consider yet a third case where subjective-effort performance covariation i s perfect: an individual believes that the probability i s 100% that i f he or she expends a low level of effort, a low 132 level (and only a low level) of performance w i l l result and that i f he or she expends a medium level of effort that a medium level of performance w i l l result, and that high effort w i l l result in high performance, as follows: 100 0 0 0 100 0 0 0 100 In this third case the expected value index i s equal to that of the second case, yet they are obviously quite different. The expected value index constructed by Ilgen, Nebecker and Pritchard i s just that, a measure of expected performance. It i s not a measure of effort- performance covariation or expectancy. The correlational measure, in contrast does capture the covariation between effort and performance. It is this measure that we should seek to validate. Validating expectancy measurement In general, the construct validity of a measure i s supported by demonstrations that the measure f i t s the nomological net surrounding i t . That i s , that i t converges with other measures of the same construct, diverges from measures of other constructs, i s responsive to changes in the presence or absence of the construct and due to variations in persons or settings, and is insensitive to changes in the presence of unrelated constructs. Ilgen, Nebeker and Pritchard (1981) sought to demonstrate the validity of a measure of expectancy by comparing scores in two settings: one with low d i f f i c u l t y and another with high d i f f i c u l t y . They manipulated the amount of work required to meet a standard. In the f i r s t condition workers were required to complete five items in a unit of work as compared to seven items per unit in the second condition. Ilgen, Nebecker and Pritchard argued that the low d i f f i c u l t y condition had objectively high expectancy and that the high d i f f i c u l t y 133 condition had objectively low expectancy. This operationalization of objective expectancy is inappropriate. If expectancy is effort-performance covariation then objective expectancy should be higher in jobs in which the connection between effort and performance is stronger.Tasks i n which performance i s primarily a function of effort should have high expectancy. Jobs with low expectancy are jobs where performance i s a matter of a b i l i t y , or luck, or how well the person on the assembly line ahead of you does his or her job. Low expectancy is where performance is high or low independent of effort. Task d i f f i c u l t y can sometimes increase d i f f i c u l t y and sometimes decrease i t . A task may be so d i f f i c u l t that performance i s uniformly low. Or a task may be so easy that performance is uniformly high. In both these instances, expectancy is constrained. Similarly, two tasks may diff e r in d i f f i c u l t y yet have the same effort-performance relationship. Consider task A, in which 10 units of effort produce 10 units of performance. Increasing effort to 15 units increases performance to 15 units. In task B, an easier task, 10 units of effort produce 20 units of performance and increasing effort to 25 units increases performance to 25 units. In each task a 5 unit increase in effort i s associated with a 5 unit increase in performance. The e f f o r t — performance relationship i s identical in the two tasks and yet task A i s more d i f f i c u l t than task B. The expected performance given 10 or 15 units of effort i s lower in task A. Ilgen, Nebecker and Pritchard's manipulation is like our hypothetical Tasks A and B. In their high task d i f f i c u l t y , "low expectancy" condition, each unit of work required the completion of seven items. In the low d i f f i c u l t y , "high expectancy" condition, only five items were required per unit. The underlying relationship between effort and performance, however, was essentially unchanged. 134 Ilgen and colleagues found that the d i f f i c u l t y manipulation significantly affected scores on the expected value index. Expected performance was significantly higher in the low d i f f i c u l t y , "high expectancy" condition. In contrast, their correlational index "showed l i t t l e responsiveness to the expectancy manipulation" (p.215). They interpreted these results as evidence for the vali d i t y of the expected value index as a measure of expectancy, over and above the correlational index. As we have discussed here, however, Ilgen, Nebecker and Pritchard's operationalization did not create differences in objective effort—performance covariation. They did manipulate d i f f i c u l t y , which was reflected in the perceptions of participants that they could expect to complete more units of work when fewer items were required per unit. The lack of responsiveness of the correlational index i s an indication that they did not manipulate expectancy, rather than a sign that the measure lacks validity. In fact, this lack of responsiveness is an indication of divergent vali d i t y . In sum, then, Ilgen, Nebecker and Pritchard have developed a measure of expectancy that i s true to Vroom's conceptualization. At present, however, evidence as to the validity of such a multiple-level correlational index of effort—performance covariation i s lacking. Study Two was undertaken to address this. Method Subjects and Design Participants in the study were 221 second year business students. Each received course credit for participating. Participants completed one of two tasks, a cognitive reasoning and decision making task or a perceptual motor s k i l l s task. The perceptual motor task was one in which a strong objective 135 effort-performance connection existed whereas the cognitive reasoning task had a weak objective effort- performance connection. Following completion of the task measures of the perceived link between effort and performance were taken. Manipulation of Expectancy Perceptual-motor task. The high objective expectancy condition was created by using a task relying on the perceptual and motor s k i l l s of participants. Specifically,- the task chosen was a "proofreading or quality control" task, briefly described above in Study One. Participants were told that "in quality control a product must be matched against a standard." Participants were provided with a booklet, each page of which contained 25 rows of digits between 0 to 9, each row having 20 d i g i t s . These numbers were chosen from a random number table. In the l e f t margin of each page was a column of digits corresponding to the rows. The task assigned to the participants was to check the number at the l e f t margin of each row, then c i r c l e each number in the row that matched i t . Following a practice session, participants worked at the task for a series of 10 work periods which averaged one minute in length. They were asked to do their best but to work carefully as "only correctly completed rows count." At the end of each period participants were asked to draw a line under the last row completed, count the number of rows completed and write the number of rows completed in the right hand margin. They were then instructed to begin again at the top of the next page. The complete instructions and a sample of the task provided are shown in Appendix C. Cognitive reasoning task. The low objective expectancy condition was created by using a task relying on reasoning a b i l i t y and conceptual understanding. The task was the "Brand Managers' Allocation Problem", or 136 "Marketing Game" (Mclntyre 1979, 1982). This task involved the allocation of a fixed promotional budget across three markets, with the objective of maximizing the total profit earned. Each market was represented by an independent response function that related promotional expenditures to profit. Participants were told that they were the newly hired brand manager and had to decide how to divide their promotional budget among three markets for each of 10 periods. Before beginning, they were shown the market allocations and market profit results of the previous manager for five previous periods, although the previous budget was not fixed and promotion was allocated evenly across markets. So, for example, a participant may have been assigned a promotional budget of $54,000 for each of 10 periods. Before beginning, they were shown the profit that resulted in each market from equal division of different promotional budgets over each of five previous periods. They were then asked to divide the $54,000 among the three markets, were shown the profit that resulted in each market, and the procedure was repeated. The instructions provided to participants are shown in Appendix C. Performance on the Proof-reading task was primarily a function of two aspects of effort: attention to the number provided as standard and criterion, and energy directed at completing as many rows as possible. Performance was not a function of a b i l i t y or understanding of how to achieve high performance but rather of effort expended. In contrast, performance on the Marketing Game depended on participants' marginal analysis s k i l l s and a b i l i t i e s ; on examining past allocation decisions and profit results, inferring the relationship between promotion and profit, and applying that understanding to the next decision. Unlike the Proof-reading task performance on the Marketing Game depended on individual differences in s k i l l s and a b i l i t i e s . Performance was 137 thus constrained and dependent on factors over and above effort. Effort at the Marketing Game was present primarily in the form of concentration and time spent on a decision. Thus the degree to which performance was a result of effort was lower in the Marketing Game than in the Proof-reading task. Measures Three measures of the relationship between effort and performance were taken: a multiple effort and performance level measure of covariation, a single item measure of control over performance, and a single item measure of the perceived relationship between working hard and performing well. Participants' expected performance was also measured. Finally, measures of task perceptions and causal attributions were taken. Perceived covariation. Participants were asked, for each combination of five levels of performance and'three levels of effort, to estimate the probability that they would achieve the specified level of performance i f they were to expend the specified level of effort. For both the proofreading task and the Marketing Game the level of performance chosen divided the range of possible performance into five equal intervals. For example the performance intervals for the proofreading task were 1 to 10 rows, 11 to 20 rows, 21 to 30 rows, 31 to 40 rows, and more than 40 rows completed. For the Marketing Game, the intervals were $1400 or less profit, $1400 to $1600 profit, $1601 to $1800 profit, $1801 to $2000, and more than $2000 pro f i t . For both tasks the effort levels specified were "high", "medium", and "low". Figure 7 shows the assessment of the perceived probability that high effort would lead to $1801 to $2000 profit in the Marketing Game. For each participant, then, 15 perceived probabilities were recorded. These were combined into a composite score reflecting the covariation between 138 The next set of questions are about your participation in the Marketing Game. We'd like you to think about what would happen i f you were to do the Game again, under similar circumstances. We would like you to estimate the probability of achieving different amounts of total profit. Remember that you earned somewhere between $1000 and about $2000 profit each period. The following questions w i l l ask you to estimate the probability that you could earn on average per period : $ 1401 or less between $ 1401 and $ 1600 between $ 1601 and $ 1800 between $ 1801 and $ 2000 more than $ 2000 The following questions also ask you to think about what would happen i f you were to expend a HIGH, MEDIUM, or LOW degree of effort. We are not interested in the actual amount of effort you expend, but in what would happen i f you were to do the game again under similar circumstances. If you were to expend a HIGH level of effort, what i s the probability that you would earn between $ 1801 and $ 2000 on average per period? (Answer in percentage terms) Figure 7. Assessment of relationship between effort and performance, Study Two. 139 effort and performance by treating the values in the matrix as frequencies in a bivariate distribution. Unit weights were used for the three level of effort (e.g., 1, 2, 3) and for the five performance levels (e.g., 1, 2, 3, 4, 5). The composite score was then transformed using the Fisher r to z transformation to normalize the distribution of coefficients. Prior to transformation the distribution was significantly skewed, skew = -.684, p < .01, following transformation skew was .092, p > .05. Control over performance. A second measure was constructed using the concept of "control", as described by Alloy and Abramson (1979). Alloy and Abramson argue that the relation between a response and an outcome, or the dependence of an outcome on a response, is best construed as one of controllability. Most importantly, control conveys the technical meaning of covariation in natural, everyday language (Jenkins and Ward, 1965). Participants were provided with the following instructions: On the following scale, indicate your judgement of the amount of control you had over your performance on the [Proofreading task/ Marketing Game], at 100 i f you had complete control and at 0 i f you had no control. Complete control means that the [ number of rows you complete/ profit you achieve] i s determined by how hard you try. No control means that how hard you try or don't try has nothing to do with your performance. Another way to look at having no control is that the [number of rows completed/profit achieved ] in any period i s totally determined by factors such as chance or luck, rather than the effort you expended. Intermediate control means that your effort has some influence but does not completely determine the profit you achieve. Underneath these instructions was displayed a scale with endpoints of 0 and 100 with increments of 10 also indicated. 140 Perceived correlation. A third measure mirrored those used most often in studies of expectancy theory. It asked participants to indicate, on a 10 point-scale with end-points of 0 and 9 labelled "no relationship" and "strong relationship", the number that best represented "the relationship between working hard on the task and performing well". Expected performance. The fifteen probability estimates collected to measure perceived covariation, above, were used to calculate the expected value index developed by Ilgen, Nebecker and Pritchard, as described above. Task Perceptions and Causal Attributions. Measures were also completed of four task perceptions and two causal attributions. The task perceptions measured were (1) satisfaction with performance, (2) task d i f f i c u l t y and challenge, (3) task interest, and (4) task effort. The Causal Dimension Scale (Russell, 1982) was used to measure two dimensions of causal attributions: internality and stability, he measures of task perceptions and causal attributions were those described in Study One. As shown in Study One, these measures have good r e l i a b i l i t y . Results In this Study, we w i l l consider the following pattern of results to be in support of the validity of the perceived covariation measure of expectancy: It i s hypothesized that scores on the measures of perceived covariation, control over performance and perceived correlation w i l l be significantly higher in the high expectancy condition. The scores on the remaining measures do not bear as directly on the validity of the measure, although we would predict that cause would be attributed more to effort in the high expectancy condition. That i s , attributions of greater internality and less s t a b i l i t y should be 141 made. Before proceeding with the multivariate analyses, the variables used wi l l be evaluated with respect to practical limitations of the technique. Evaluation of Assumptions The analysis procedure appropriate to the case of one independent variable with two levels and multiple dependent variables i s Hotelling's T 2. Hotelling's T 2 is a special case of multivariate analysis of variance (MANOVA), which i s designed to test hypotheses about group differences on a set of measures. The following discussion w i l l be described in terms of MANOVA. In this study, the data set comprised 188 subjects in the high-expectancy task condition and 33 subjects in the low-expectancy task condition . The differential ease of administering the paper-and-pencil motor s k i l l s task in group settings, as opposed to the individually, computer-administered cognitive-reasoning task, accounted for this large difference in group size. Twelve of the subjects in the high expectancy group did not provide probability estimates for a l l 15 effort—performance level combinations. The perceived covariation and expected value measures'could therefore not be computed, so data for these participants were discarded. Thus scores for 176 subjects were available for analysis in the high expectancy group. Inspection of the distributions of scores on the ten dependent variables revealed significant skew on only one measure: that of perceived control. Before taking action to resolve this, the presence of outliers was investigated. Univariate outliers were identified by examining extreme z-scores. Of the 2090 observations (10 dependent variables time's 209 cases), 8 or 0.4% had z-scores in excess of 3.09. This was slightly more than chance 142 expectation. The most extreme score did not, however, exceed 3.75 standard deviations from i t s mean. Nine cases were identified as multivariate outliers, five in the high-expectancy group and four in the low-expectancy group. Following deletion of these cases, no variable was significantly skewed. The assumption of multivariate normality was thus met. As a preliminary check for homogeneity of variance-covariance matrices, sample variances for each of the 10 dependent variables were examined. ' For no DV did the ratio of largest to smallest variance across the two groups approach 20:1, the criterion suggested by Tabachnik and F i d e l l . Univariate homogeneity of variance tests, however, revealed three significant results at a= .01 (Bartlett-Box F), for the measures of perceived covariation, control, and correlation. A multivariate test of homogeneity of dispersion was, not surprisingly, also significant (Box's M= 115.4, p < .01). For the perceived covariation measure the larger variance belonged to the larger group, thus the significance test would have been more conservative. For the other two variables, however, the larger variance belonged to the smaller group. The direction of bias produced in the test of multivariate significance by this heterogeneity i s therefore unknown. To resolve this uncertainty, equality of sample sizes was achieved by randomly selecting 29 cases for analysis from the high-expectancy group, to equal the number of cases in the low expectancy group. Significance tests in MANOVA are robust to both heterogeneity of variance and non-normality i f sample sizes are equal and exceed 20 degrees of freedom for error in the univariate case (Hakstian, Roed and Lind, 1979; Tabachnik and F i d e l l , 1983) Evidence of non-linearity of relationships among dependent variables was sought by examining the plots of observed values versus standardized residuals for each dependent variable. No evidence of gross curvilinearity was found, 143 therefore no transformation of variables was undertaken. In this study, multicollinearity was assessed by examining the squared multiple correlation (SMC) between each variable and a l l other dependent variables. No variables showed a SMC approaching 1.0. Similarly, the determinant of the within-cell correlation matrix was non-zero. Finally, an assumption of homogeneity of regression, or that the slope of the regression of dependent variables on covariables is equal across c e l l s , i s made in stepdown analysis in MANOVA. In this study, tests of homogeneity of regression were run for a l l variables, since each serves as a covariate for a l l others in the Stepdown F-tests that follow. Homogeneity of regression was achieved for a l l components. Multivariate analysis of variance. A multivariate analysis of variance was performed on the ten dependent variables: perceived covariation, perceived control, perceived correlation, expected performance, satisfaction with task performance, task d i f f i c u l t y and challenge, task effort, task interest, and perceived internality and sta b i l i t y of the cause of performance. The independent variable was objective task expectancy (high and low). The MANOVA procedure in SPSSX was used for the analyses. Because samples sizes were equal (n=29), no adjustment for nonorthogonality was necessary. Total N was 58 following the deletion of outliers and the sampling procedure described above. Hotelling's T 2 criterion showed a highly significant difference between the two groups, F(10, 47) = 14.21, p < .001. These results reflected a strong association between objective task expectancy and the combined dependent variables (A = .25; r)2 = 0.75). 144 The influence of the difference in objective effort-performance relationship between the two groups on each dependent variable was examined next. The univariate test of significance for the eleven dependent variables are shown in Table 15. Significant univariate effects are present for perceived covariation, perceived control, perceived correlation, expected performance, task effort, task interest, and internality of attributions. As Table 16 shows, the average perceived covariation was .41 in the proofreading, high expectancy group, compared to .28 in the marketing game, low expectancy group. By i t s e l f , this result confirms the validity of the composite measure of expectancy. In keeping with this result, mean perceived control was also significantly higher in the high expectancy group than in the low expectancy group. Mean expected task performance was, however, significantly lower in the high expectancy group. This means that the hypothesis formulated by Ilgen, Nebecker and Pritchard, that the expected value index measures expectancy, i s disconfirmed. Instead of showing that scores on the expected value index were higher in the high expectancy group, which would be evidence of validity, or that scores were unaffected by objective expectancy, which is the null hypothesis, Study Three supports rejection of the null hypothesis in the opposite direction to that supposed. The univariate results also showed that participants rated the cause of their performance as significantly more internal in the high expectancy condition than in the low expectancy condition. Overall, these results provide strong evidence for the vali d i t y of the expectancy manipulation and the perceived covariation measure. Finally, task effort was higher in the high expectancy, Proof-reading task, but participants found the Marketing Game more interesting. Although the univariate statistics are of interest, they overlook 145 Ur livariat :e SU jpdown F d.f. P F d.f. P Measure Perceived covariation 7 .15 1/56 .016 7 .15 1/55 .127 .071 control 8 .19 1/56 .006 7 .23 1/55 .131 .071 correlation 5 .56 1/56 .022 .32 1/54 .571 .006 Expected performance 70 47 1/56 .000 54 .32 1/53 .000 .506 Perceived internality 4 .83 1/56 .032 3 .93 1/52 .053 .070 sta b i l i t y .34 1/56 .558 2 .93 1/51 .092 .054 Satisfaction with task performance 0 59 1/56 .444 .18 1/50 .674 .003 Task d i f f i c u l t y and challenge 3 .61 1/56 .062 2 .54 1/49 .117 .049 Task effort 21 10 1/56 .000 8 .33 1/48 .006 .148 Task interest 36 19 1/56 .000 4 .60 1/47 .037 .089 T Table 15. Univariate and Stepdown F-tests, Study Three. 146 Measure Perceived covariation Perceived control Perceived correlation Expected performance Perceived internality s t a b i l i t y Satisfaction with task performance Task d i f f i c u l t y and challenge Task effort Task interest Proofreading Task (High Expectancy) Mean 0.41 81.72 7.17 2.12 20.89 14.72 14.10 10.00 8.48 5.13 St. Dev. 0.23 16.49 1.60 0.38 3.07 5.76 2.67 2.55 1.18 1.72 Marketing Game (Low Expectancy) Mean 0.28 70.93 6.10 2.98 18.83 13.86 13.58 11.20 6.82 7.62 St. Dev. 0.16 11.85 1.84 0.39 4.02 5.36 2.43 2.27 1.54 1.40 Table 16. Means and standard deviations for dependent variables, Study Three. 147 relationships between the dependent variables. The matrix of correlations between dependent variables, shown in Table 17, reveals intercorrelations among many of the measures. Bartlett's test of sphericity was significant (x 2(45) = 140.4, p < .001) supporting rejection of the hypothesis that the correlation matrix is an identity matrix and hence that the variables are independent. To evaluate the contribution of each dependent variable with the effects of i t s correlation with other variables controlled, a stepdown analysis was performed. In this study, the measures relating effort and performance were of most interest and were entered into the analysis f i r s t , followed by the measures of expected performance, causal attributions, and task perceptions, as shown in Table 15. The f i r s t variable entered was the composite measure of perceived effort-performance covariation. Scores on this measure were significantly associated with objective task expectancy, stepdown F ( l , 56)= 7.15, p < .02. As indicated by the ratio of hypothesis sum of squares to error plus hypothesis sum of squares, the strength of association was .13. After the pattern of differences in perceived covariation was entered, a significant difference remains for perceived control, stepdown F ( l , 55) = 7.23, p < .01, n2 = .13. Given i t s high correlation with perceived control, i t i s not surprising that the unique contribution of the perceived correlation measure is non-significant, stepdown F ( l , 54) = .325, p > .05. The fourth variable entered, expected performance, made a strong unique contribution to the multivariate difference between expectancy conditions, stepdown F (1,53) = 54.33, p < .001, rj 2 = .51. Stepdown analysis revealed that internality of attributions made a marginally significant unique contribution, stepdown F = 3.93, p < .053, 772 = .07. As stated above, the mean internality score was higher in the high expectancy condition. Of the remaining measures, only task effort and task interest made significant unique (1) (2) (3) (4) (5) (6) (7) (8) (9) ( 10) (1) P e r c e i v e d c o v a r i a t i o n (2) P e r c e i v e d c o n t r o l . 1 14 (3) P e r c e i v e d c o r r e l a t i o n .249 .523 (4) Expected performance -.243 -.252 - . 170 (5) S a t i s f a c t i o n w i t h task performance - .042 . 332 . 126 - .056 (6) Task d i f f i c u l t y and chal1enge - . 106 - .088 - . 104 .097 - .253 (7) Task e f f o r t .340 .207 .231 - .412 .383 . 297 (8) Task i n t e r e s t - . 133 - . 145 .009 .525 - .023 .461 - .277 (9) P e r c e i v e d i n t e r n a l i t y .096 .640 .291 - .055 . 399 -. 194 .215 -.071 ( 10) s t a b i 1 i ty .027 .457 .373 - .044 .533 - . 189 .200 .093 . 584 T a b l e 17. C o r r e l a t i o n s between dependent v a r i a b l e s . Study Three. 149 contributions to the overall effect. Task effort was significantly higher in the high expectancy condition, stepdown F = 8.34, p < .01, 772 = .15. Task interest was significantly higher in the low expectancy condition, stepdown F = 4.60, p < .05, 7?2 = .09. Discussion Study Three provided strong evidence for the vali d i t y of the correlational index of perceived covariation as a measure of expectancy. The immediate implication of this i s that this measure is not only appropriate to further research but that i t should be used in place of other measures. The perceived covariation measure i s closest to the conceptualization of expectancy as intended by Vroom and is responsive to a manipulation that matches that conceptualization of expectancy. . The f i r s t task of future research should be to replicate Study Three using similar tasks and measures and, more importantly, using other tasks. A key demonstration would be to show that the perceived covariation measure i s responsive to differences in objective expectancy within a single task. Such a difference might be created by manipulating task interdependency or the role of chance.. It is not clear, however, to what extent expectancy can be manipulated without changing the essential nature of tasks. That i s , to what extent is the objective expectancy of tasks malleable? It may be that gross differences in expectancy only exist across tasks. This question is key to the design of jobs. Perceived effort-performance covariation i s an important factor in determining the effort that individuals choose to expend at work, and objective expectancy is probably the principal determinant of expectancy percepts. It follows then, that to design jobs to motivate people we should know more about manipulating objective expectancy. Better understanding of 150 the concept and how to measure i t is only one step in that direction. Study Three showed that self-reported task effort was significantly higher in the high expectancy task. This cannot be taken as evidence of the va l i d i t y of the expectancy manipulation, the expectancy measure, or expectancy theory because instrumentality and valence are unknown. The obvious question i s whether effort and expectancy are s t i l l positively associated when the other components of expectancy theory are held constant. That i s , does the perceived covariation measure reliably predict effort? An affirmative answer would lend support to both the measure and the theory. A negative answer would cast doubt on both the theory and the measure. One could argue that expectancy theory as conceptualized by Vroom has as yet not been tested because appropriate measures have not been used. Clearly, the f i n a l word on expectancy theory is not i n . 151 VIII. STUDY FOUR: MOOD AND EXPECTANCY Overview Study Four tested the principal hypotheses about the influence of mood on expectancy. A laboratory experiment was conducted with students as participants. In each experimental session students participated in a simulated business task, underwent a mood induction procedure, and then completed measures of mood state, expectancy, and task cognitions. The experimental design chosen was a within-subjects design in which each participant received a l l levels of the independent variable. Further, the order of presentation of the three levels of the mood manipulation was counterbalanced by creating six between-subjects order conditions. Thus, the complete design was a mixed between-within, or split-piot design (Kirk, 1968). The design has particular advantages and disadvantages and poses some unique analysis issues. Experimental Design Analysis of variance (ANOVA) procedures employed to test differences among mean scores can conveniently be summarized in terms of how the sums of squared differences between the scores of individuals and their group means" are partitioned. For example, in an experimental design where each subject i s assigned to one of a number of treatment groups (a one-way between-subjects design), the total sum-of-squares can be partitioned into those associated with differences between groups and those associated with differences within groups, e.g. the sum-of-squares total equals sum-of-squares groups plus sum-of -squares subjects-within groups. The F test of the significance of the 152 difference between groups is a function of the ratio of the between-group to subjects-within-group sum-of-squares. The within-group sum-of-squares is used to estimate the population (error) variance. In within-subjects ANOVA, each subject receives more than one treatment. This i s also known as a repeated measures design. The source of differences among means for levels of the independent variable is the same as the between-subjects design, above. The error term can, however, be partitioned further, e.g. sum-of-squares total equals sum-of-squares groups plus sum-of-squares subjects plus sum-of-squares subjects-by-group. Differences between individuals( sum-of-squares subjects) can be viewed as a systematic source of variance in scores. If individuals are measured repeatedly, these individual differences can be examined. Since this variation i s systematic, i t can be subtracted from the error term. Or, stated differently, the best estimate of population (error) variance i s the group by subject interaction. This error term is smaller than that in a between-subjects design to the degree that the variation due to subjects i s different from zero. That i s , i f the sum-of-squares subjects i s non-zero then the error term in a within-subjects analysis is smaller than in a between-subjects analysis. A smaller between-group effect i s therefore needed to achieve a significant F-ratio, provided, of course, that the loss of degrees of freedom in the within-subjects case does not offset any gain. Within-subjects designs have disadvantages that may mitigate any gains in s t a t i s t i c a l power. First, there is the potential for effects due to the fact of repeated measurement, such as learning effects or effects of hypothesis guessing. Second, the degrees of freedom for error are reduced from those in a between-subjects design. This loss in power from diminished degrees of freedom may overcome gains in power from reduced error variance. The relative 153 efficiency of the within-subjects design depends on the degree to which variation between subjects within groups has been reduced relative to the variation between groups. We can conceptualize each subject in a within-subjects design as serving as his or her own control. The relative power of a within-subjects design i s a function of the degree to which such control for individual differences i s relevant. In this study both the dependent and independent variables are perceptions held by individuals: of qualities of tasks and of their own feeling state. These are li k e l y to be highly individual. That i s , people are lik e l y to vary greatly in the perceptions they have of tasks and of their own mood. Across individuals this variation may be much larger than differences over occasions within individuals. Controlling for these differences across levels of the independent variable i s therefore highly advantageous. In a within subjects design these between-subjects effects are estimated and removed from the error term, increasing the power to detect a between-treatment group difference. A further reason for using a within-subjects design i s the possibility of additional error variance due to differential success or failure experiences between subjects. Task success and failure are reliable manipulations of mood. Some participants w i l l experience the experimental task as more successful than w i l l others. Such an inadvertant mood manipulation would add to error variance in a between-subjects design. In a within-subjects design, however, this effect i s held constant. The different tasks used on repeated sessions are of equal d i f f i c u l t y . Thus each participant i s likely to have the same success or failure experience in each of the three mood treatment sessions. In a between-subjects design the experience of different 154 individuals i s lik e l y to be different even i f task d i f f i c u l t y i s held constant. Because each participant in a within-subjects design receives each level of the independent variable, the possi b i l i t y of session effects i s created. That i s , scores may vary systematically from the f i r s t to subsequent experimental sessions. Further, the order of presentation of levels of the independent variable may influence scores over sessions. To control for these concerns in this study, each participant was randomly assigned to one of the six possible orders of presentation of the three mood treatments. The complete design is thus a mixed-between-within, or split-plot design. This i s diagrammed in Figure 8. Method Subjects. Participants in the study were 94 UBC second-year business students. Each received course credit in return for his or her participation. Sixteen participants were excluded from the analysis for a number of reasons. Data from six participants was not complete due to their failure to attend sessions or because of computer data-capturing malfunctions. The data for six additional participants were discarded because of procedural errors in which they received the same mood treatment during two sessions. Complete data was thus collected for 82 participants over the three sessions. Four of these cases were discarded such that the f i n a l design contained 13 cases in each of the 6 order conditions. That i s , the f i n a l sample comprised 78 participants. Procedure. Invitations to participate in the study were made to students during regularly scheduled course time. This invitation described the study as an investigation of music in the workplace and informed participants as to the benefits of participation, the commitment required, and what procedures 155 Session Subjects One Two Three Order 1 nOl 1 nl3 Elation Depression Neutral Order 2 nl4 1 n26 Elation Neutral Depression Order 3 n27 1 n39 Depression Elation Neutral Order 4 n40 1 n52 Depression Neutral Elation Order 5 n53 ' 1 n65 Neutral Elation Depression Order 6 n66 1 n78 Neutral Depression Elation Figure 8. Split-Plot Design, Mood Treatment by Session and Order Conditions. 156 would be employed. It also ensured that c o n f i d e n t i a l i t y of responses would be maintained and that p a r t i c i p a t i o n was voluntary. Participants were asked to sign a form i n d i c a t i n g that they had gave t h e i r informed consent to p a r t i c i p a t e . The form used i s shown i n Appendix D. At the time the i n v i t a t i o n to p a r t i c i p a t e was issued, A l l procedures employed i n Study Four were judged e t h i c a l by i n s t i t u t i o n a l review, the questionnaire containing measures of i n d i v i d u a l d i f f e r e n c e v a r i a b l e s was d i s t r i b u t e d f o r c o l l e c t i o n the next c l a s s . Arrangements were made to contact each person at a l a t e r time to schedule h i s or her p a r t i c i p a t i o n i n the experimental sessions. P a r t i c i p a n t s were tested i n d i v i d u a l l y i n each of the three experimental sessions. These sessions were scheduled on d i f f e r e n t days so that mood induced i n one session would not carry over to the next. Other scheduling constraints imposed were that a l l three sessions occur within seven days, to reduce a t t r i t i o n , " a n d that a l l three sessions for each p a r t i c i p a n t take place before noon or a f t e r noon. This l a t t e r constraint was imposed to reduce v a r i a b i l i t y due to time-of-day e f f e c t s . Although these l a t t e r two constraints were adhered to as much as possible, they were relaxed where necessary to ensure a subject's p a r t i c i p a t i o n . The procedure f o r each session was l a r g e l y i d e n t i c a l . Each of the pa r t i c i p a n t s completed a business decision-making task, then underwent a mood induction, and f i n a l l y completed measures of the manipulation checks and dependent v a r i a b l e s . The major exception to t h i s procedure was that the f i r s t session included a p r a c t i c e period i n the decision-making task. Session One. On a r r i v a l at the f i r s t experimental session, p a r t i c i p a n t s were greeted by one of three experimenters. The three experimenters were male, between 20 and 30 years of age and were rehearsed to ensure that the i n t e r a c t i o n of each subject with each experimenter was comparable. 157 Participants were then seated in front of a computer terminal and presented with a description of the procedure for Session One, as follows: Music in the Workplace SESSION ONE The study in which you are being asked to participate i s part of an investigation of the use of music in the workplace. Many organizations play music as a background to work. The study you are in is examining the use of music during work and people's perceptions of the use of music. We would like your help with two parts of this study. In the f i r s t part you w i l l be asked to complete a simulated business decision-making task on a computer terminal while music is played in the background. In the second part of today's session we would like you to listen to some music and then we w i l l ask you some questions about the music. Both parts of the session w i l l provide us with baseline information about the decision-making task and about people's reactions to the music. Because we are interested in the use of different types of music we would li k e you to come back twice more to listen to other selections. At that time we would again lik e to collect baseline information on another version of the decision-making task. Today's session w i l l include a practise period and so should take about one hour, the following two sessions should take less than an hour each. The decision-making task is based on a r e a l i s t i c business problem. The music you w i l l listen to w i l l probably evoke different reactions in different people. These honest reactions are what we are interested i n . Your participation in this study i s voluntary, you are free to discontinue participation at any time without penalty. Your responses w i l l be used only for the purposes of this study and w i l l be kept confidential. Decision-making task. To provide a basis for measurement of participant's expectancies they were asked to complete an experimental task. Expectancy research has used experimental tasks such as processing catalogue orders (e.g., Ilgen, Nebecker & Pritchard, 1981), and has asked individuals to rate the expectancy associated with educational or occupational choices (e.g., Snyder, Howard & Hammer, 1978; Muchinsky, 1977). Studies of the influence of mood on memory have usually used much more a r t i f i c i a l tasks, such as mental rotation, concept formation, or contingency judgement tasks (Alloy, Abramson & 158 Viscusi, 1981; Brown, 1984; Wright & Mischel, 1982). As a study cf organizational motivation, i t was desirable to use a more r e a l i s t i c task with generalizability to organizational settings in this study. Further, based on findings that mood effects were most l i k e l y when the objective contingency between behavior and outcomes was low (Alloy & Abramson, 1979; Alloy, Abramson & Viscusi, 1981), a task with low expectancy was preferable. A task which met these requirements was the "Brand Manager's Allocation Problem" (Mclntyre, 1979, 1982), the same task used in the low expectancy condition of Study Two. This task involves the allocation of a fixed promotional budget across three territories with the objective of maximizing the resultant total profit. Each of the three territories i s represented by an independent response function that relates promotional expenditures to profit, of the following form: Profit = t min + ts * (aX ** b / ( l + aX ** b)) - e - X where, X = promotional dollars allocated to the territory, t min = sales in the territory when X = 0, ts = saturation sales minus t min in the territory, a, b = parameters, e = random error term (uniformly distributed between -20 and 20%). The decision-making task was to allocate the fixed promotional budget across the three territories to maximize total profit. Participants were told that they would be evaluated based on the total profit earned over .the ten decision periods. Each period was independent, that i s , there was no promotional carryover from one period to another, of which participants were 159 informed. After each allocation decision the subjects would automatically receive an updated "History Display". This report showed the promotional allocations made to each territory, profit by territory, and total profit for each decision. Before making the f i r s t decision, subjects received a fiv e -period history report. These five decisions were presumably made by the previous manager whom the subject was replacing. In the i n i t i a l five-period history report the budget varied by plus or minus 10% from period to period but was always allocated equally between the territ o r i e s . Following the practise session, any further questions were answered. Participants were then told that they would now begin the actual decision-making simulation. It was emphasized that the task was identical but that the particular problem was a different one: participants were told that they had been given three new markets and a different promotional budget. The actual simulation was 10 periods long. Participants were again provided with allocations and profit results for five "previous" periods made by the "previous manager". To enhance participants' beliefs that the study was about "music in the workplace" so that the subsequent mood induction would be plausible but disguised, music was also played in the background during the task. Participants were told they were to "concentrate on the task as i f i t were part of their normal job". A selection of non-lyrical, neutral music (Ackerman, 1981; "Passage: Pieces for guitar") was played at low volume. The results of a pretest indicated that this music was perceived as neutral and had no significant effect on the mood of participants. The task was presented interactively on a computer terminal video monitor. Participants were told that the computer would record their profit earned for each allocation made. The profit results shown after each 160 allocation decision were the actual profit earned in those markets, given their response functions and the degree of randomness ("external market factors") present. The instructions provided to participants were as follows: THE MARKETING GAME: Session One Practice You w i l l play the role of a newly hired brand manager for National Foods. Your job is to decide how to spend money on promoting your company's product in your region. You w i l l be given a promotional budget and have to decide how best to divide that budget between the three markets in your region. The more of the promotional budget you spend in a market the more profit you w i l l earn in that market. But in some markets the same amount of promotion earns more profit. Your goal i s to earn as much profit in your region as you can over the next ten periods. So you must decide for each of these periods, how to allocate or divide your budget between the three markets to earn as much profit as possible. The way you wi l l do this i s by making your allocation for one period, then you can look at the results of that allocation before you go on to make the allocation for the next period. To help you get started in your new job, you w i l l be shown the "history report" of the previous manager. This report w i l l look lik e this: H I S T O R Y D I S P L A Y A L L F I G U R E S I N $ 000'S PERIOD: 1 2 3 4 5 PROMOTION MARKET 1: 7 10 8 11 9 MARKET 2: 7 10 8 11 9 MARKET 3: 7 10 8 11 9 PROFIT MARKET 1: 221 315 244 332 279 MARKET 2: 97 103 99 102 105 MARKET 3: 443 634 494 686 563 TOTAL PROFIT 761 1052 837 1120 947 HIT RETURN TO CONTINUE This shows that in period One the previous manager divided a budget of $21 ( a l l figures are in thousands) evenly between Market 1, Market 2, and Market 3. This resulted in a profit of $221 in Market 1, $97 in Market 2, and $443 in Market 3 for a total of $761. In period 2 the previous manager divided a budget of $30 evenly among the three markets. 161 The previous manager had a different budget in each period. YOUR budget wi l l be EQUAL for each period. During this practice session you w i l l have $54 to allocate each period to the three markets in your region. You must spend a l l of this budget each period. The practice session w i l l be THREE periods long. Your performance w i l l be evaluated on the basis of the total profit you earn over the next three periods. During this practice session and the real session later, the computer w i l l prompt you each period. It w i l l show you the history report, including the decisions and results of the previous manager, then i t w i l l ask you to make a decision. The prompt w i l l look like this: YOUR BUDGET IS: $ 54 WHAT IS YOUR ALLOCATION TO MARKET 1, MARKET 2, MARKET 3 FOR PERIOD: 6 Enter your decision by entering three numbers on the same line, separated by spaces. For example: ?20 14 20 The computer w i l l then show you the updated history report with the profit results of your decision and go on to the next period. A recent market survey, commissioned by National Foods, has determined that although other market factors also influence profit, the single most important decision that you can make is how to divide your promotional budget. This survey has also determined that although other factors may influence profit within a period, the periods are independent. There is no carryover of promotion from one period to the next. Any questions? Following an opportunity to have any questions about the procedure or the Marketing Game answered, each participant was asked a series of multiple choice questions to ensure understanding of the decision-making task. Care was taken that these questions were posed for classification rather than evaluation. Any misunderstandings were corrected by the experimenter. When the participant was prepared to proceed, the experimenter initiated the practice version of the decision-making task, ensured that the program was 162 running properly and l e f t the room to allow the participant to complete the practise. The practice session included three decision periods and took participants from five to ten minutes. Four versions of the decision-making task were used. Each version had a different promotional budget (e.g., 54, 56, 62 or 75 thousands of dollars), and a different response function relating promotion to profit. One of these cases was used for a l l participants during the practice session. The remaining three cases were assigned randomly to participants such that in each session, a given participant received a new case and such that a l l possible orderings of these three cases occurred an equal number of times. The three cases used in the experimental sessions were of equal objective d i f f i c u l t y , as defined by Mclntyre (1979) as the degree to which the profit resulting from the optimal budget allocation differed from that resulting from an equal allocation, as a proportion of the latter. That i s , Di f f i c u l t y = (Profit optimal - Profit equal)/ Profit equal Mood manipulation. Following the decision-making task, the subjects were told that the second part of the session would begin. The following instructions were read to each participant: We are interested in the use of different types of music in the workplace and in how people react to music. What we would like you to do next i s s i t back and listen to some music that might be used at work. Listen carefully and think about the images that the music brings to mind. Afterward we w i l l ask you what you think of the music, and what images the music brings to mind. Upon receiving these instructions, the subjects were seated in a comfortable lounge chair, which was placed in front of two stereo loudspeakers. Through these speakers was played, at a comfortable listening volume, one of the elated, neutral, or depressed mood induction audio tapes 163 developed by Pignatiello et a l . (1986). Each induction took twenty minutes. The particular mood induction used in a given session was predetermined randomly so that equal numbers of subjects were assigned to each order condition and so that subjects received each The experimenter was kept blind to the treatment condition The treatment condition to be used in each session was not known by the experimenter. This was achieved through the use of multiple audiotapes which were given a unique code number and assigned to participants by a third party. Dependent Measures. Following the musical induction, participants completed manipulation checks, measures of the principal dependent variables, and bogus measures designed to support the "music in the workplace" rationale for the study. The dependent measures were presented item by item on the computer terminal. Typically, the participants would be shown instructions for each measure and provided with an example i f necessary. They could then clear the display screen and respond to the f i r s t item. The screen would then be cleared and the following item presented. Details of these measures follow in the order they were administered. 1. First manipulation check: Checklist. Subjects were presented with instructions for the MAACL, as shown in Figure 5. The computer then presented, one item at a time, the 24 items comprising the f i r s t split-half version of the MAACL short form, measures of anxiety, depression and h o s t i l i t y as described in Study One. Participants were asked to indicate on a four point scale how well the words described their current feelings. Items were presented in alphabetical order. An example of the display for the item "active" i s shown in Figure 6. 2. First manipulation check: Semantic d i f f e r e n t i a l . Participants were 164 asked to complete the 12-item semantic differential measure of pleasure— displeasure and arousal—sleepiness (Russell & Mehrabian, 1977) to describe their current feelings. 3. Latency. In addition to capturing the numerical response to each mood measure item, the latency of response was also captured, as described in Study Two. 4. Bogus measures. Participants were then asked a number of questions to enhance the face validity of the mood manipulation procedure and the cover story. Subjects were asked to rate, on a seven point scale from "Very Unsuitable" to "Very Suitable", the suitability of the music to (1) work in general, (2) repetitive work, (3) interesting work, (4) work requiring concentration, and (5) work done in groups. They were also asked to rate their liking of the music played during the decision-making task, and how much they f e l t that the music enhanced or detracted from their performance on the task. 5. Perceived correlation. The measure of perceived correlation described in Study Three was used. It asked participants to indicate on a 10-point scale the relationship between working hard and performing- well. 6. Expectancy. Participants were then presented with 15 items (each combination of 3 levels of effort and 5 levels of performance) to assess their expectancy, as described in Study Three. These 15 scores were combined into a composite measure of expectancy by treating them as frequencies i n a bivariate distribution and calculating a correlation coefficient. The combined score was subjected to a Fisher r to z transformation to normalize the distribution of scores. 165 7. Recall. Participants' recall of outcomes in the task was assessed by asking them to indicate how many periods out of the 10 in which they participated that their profit was within each of five ranges of performance. The levels of performance corresponded to those for the expectancy measure. These five scores were combined into a composite measure using the same weights as for the expectancy measure. 8. Expectations for future performance. Participants were asked to indicate what they expected their performance would be i f they were to make decisions for 10 more periods and were to expend the same amount of effort. They were asked how many times their profit would be in each of the five ranges of performance. These five scores were combined into one composite score. 9. Perceived control over performance. Participants were then asked some questions "about their participation in the Marketing Game." They were f i r s t asked to indicate their judgement of the amount of control they had over their performance, as described in Study Two. 10. Causal attributions. The Causal Dimension Scale (Russell, 1982) described in Study Three was used to assess perceived internality, s t a b i l i t y , and controllability of attributions. 11. Task perceptions. Participants were presented with 24 items assessing task effort, interest, d i f f i c u l t y , satisfaction, performance, challenge and motivation as described in Study One. Included among these items also were 8 items which were verbal (as opposed to numerical) statements of the concept of effort-performance covariation. These eight new items are shown in Appendix E. Following the results of the factor analysis performed 166 in Study One, the four items measuring internal work motivation were discarded. The twelve items exclusive of the verbal expectancy items were combined into the following scales: (1) satisfaction with task performance, (2) task d i f f i c u l t y and challenge, (3) task effort, and (4) task interest. Thus, with the verbal expectancy measure, five task perception scales were administered. 12. Second manipulation check: Checklist. To ensure that the mood manipulation had endured the measurement of the dependent variables a second manipulation check was administered. This consisted of the second parallel form of the MAACL brief version, as described in Study One. 13. Second manipulation check: Semantic d i f f e r e n t i a l . Participants were again presented with the 12-item measure of pleasure—displeasure and arousal—sleepiness. 14. Response latency. Again, the latency of response to the measures constituting the second manipulation check was captured. 15. Imagery. At the end of the second and third experimental sessions participants were asked to verbally report the images the musical mood manipulation brought to mind. These were recorded by the experimenter, in part to continue the cover story for the manipulation, and in part to assess what, in fact, participants thought of during the manipulation. Imagery concerned with achievement settings, for example, could induce a change in ideational set that might represent a possible confound. 16. Recall Accuracy. During the experimental task the computer recorded the actual profit performance for each t r i a l . The number of times profit was in each of five ranges of performance, corresponding to the expectancy and 167 recall measures, was counted. These five scores were combined into one composite measure of performance which was subtracted from the recall measure to obtain a measure of recall accuracy. An accuracy score of zero represents perfect r e c a l l , positive scores represent optimistic r e c a l l and negative scores represent pessimistic r e c a l l . The questionnaire administered prior to the experimental sessions contained the following measures. These were used as covariates i n the analyses that follow. 1. Self-esteem. Global self-esteem was measured using Rosenberg's (1965) 10-item Self-esteem Scale, as described in Study One. 2. Impression Management. Impression management was measured by Paulhus' (1984) Balanced Inventory of Desirable Responding-Impression Management Scale (BIDR-IM) as described in Study One. In this study the internal consistency of the BIDR was .733, i t s Guttman lower bound was .811. 3. Locus of control. Locus of control was measured using the Spheres of Control Battery (Paulhus & Christie, 1981). The conceptual model underlying this battery holds that perceived control spans three domains of interaction: personal efficacy, interpersonal control, and sociopolitical control. Accordingly, the Spheres of Control Battery has three subscales, each has 10 items. Higher scores on these scales indicates greater perceived internality of control. Paulhus and Christie report alpha r e l i a b i l i t i e s of .75, .77, and .81 for the three scales, respectively, and substantial evidence as to the vali d i t y and u t i l i t y of the subscales. In this study internal consistencies and Guttman lower bounds of .57 and .61 were found for the personal efficacy scale, .81 and .82 for the interpersonal control scale, and .76 and .79 for 168 the sociopolitical control scale. Inspection of the item intercorrelation matrix revealed no obvious reason for the low r e l i a b i l i t y of the personal efficacy scale in this sample. Because i t failed to meet the criterion for r e l i a b i l i t y of .70 adopted here, i t was not used further. At the end of each session each participant was asked to wait for a few minutes while the experimenter ostensibly checked to make sure that the computer program captured the participant's responses. During this time the music from the "elated" mood induction procedure used by Eich and Metcalfe (in press) to manipulate mood was played. Participants were also provided with a selection of magazines to read. This procedure was employed to moderate the possible depressed post-experimental state of participants. Before being allowed to leave, each participant was thanked and reminded of his or her next appointment. Because of the length of session one the imagery questionnaire was not administered. Sessions Two and Three The second and third sessions were identical to Session One, with the following exceptions. First, the practice session on the decision-making task was omitted in sessions Two and Three. Participants were told that they would begin the actual decision-making problem without a practice session but that, in order to ensure that they remembered the essential aspects of the task, they would be asked a number of questions about the task. The same questions were used as in Session One. This allowed some time to pass between the time the participant entered the experimental session and when the task began, and ensured that this time was spent in a similar way by a l l participants. This increased the likelihood that participants began the task in Sessions Two and 169 Three in a similar neutral mood state. Second, i t was emphasized that although the decision making task in a l l sessions was identical, the decision problem was a different one. A different response function was used, of equal d i f f i c u l t y but with a different budget. Third, the mood induction condition employed was, of course, different in each session. Fourth, the shorter length of Sessions Two and Three allowed the open-ended measure of mood imagery to be administered. Finally, at the end of each participant's third session, the experimenter asked a set of questions to assess participant's suspicions about the study, whether they had guessed the true hypothesis or some other hypothesis, and whether any inter-subject contamination had occurred. Each participant was thanked and f u l l y debriefed via a written summary once a l l subjects had participated. Results Before proceeding with the multivariate analyses, the variables used wi l l be evaluated with respect to practical limitations of the technique. Evaluation of Assumptions Studies in which measures of few dependent variables are obtained for each of several levels of the independent variable may be analyzed by multivariate or univariate procedures. That i s , scores can be conceptualized as the same measure on multiple occasions and analyzed by repeated-measures Analysis of variance (ANOVA) procedures. Alternatively, scores can be conceptualized as separate tests of the same subject where each dependent variable i s measured on the same scale. Multivariate ANOVA (MANOVA) can then 170 be performed on the between-subjects independent variables for the set of within-subjects dependent measures. While the univariate procedures are less complex, they also require a highly restrictive set of assumptions concerning population treatment variances and covariances, (e.g. the assumption of homogeneity of covariance, or that the correlations among levels of the within-subjects variable are constant over a l l combinations of levels). MANOVA procedures are preferable where such assumptions are unwarranted. More importantly, MANOVA procedures protect against Type I error in testing multiple dependent measures, as i s the case here. As described earlier, MANOVA i s robust to modest violations of the assumption of multivariate normality insofar as the violation i s created by skewness rather than by outliers (Mardia, 1971). In this study, the data set comprised 13 subjects in each of 6 between-subjects order c e l l s . For one subject, one observation was missing. A multiple regression procedure (BMDPAM) was used to predict the missing value from complete cases. The missing observation was replaced with this estimated value. Thus sample sizes are equal. With 13 cases in each of 6 between-subjects cell s , the degrees of freedom for error in the univariate case is 14, below Tabachnik and Fidell's suggested criterion for robustness. Therefore, the extent of skewness and the presence of outliers was tested for. The skewness of the distribution each variable in standard units was computed. Of the 25 dependent variables in each of eighteen order by session cel l s , in 11 instances (or 2.4% of the 468 instances examined) skewness was in excess of ± 2.58 standard units. In no case was the same variable skewed in different order c e l l s . Thus skewness was not concentrated on any one variable across order c e l l s . Univariate outliers were identified by examining extreme z-scores for variables on a casewise basis. 171 Of the 5850 observations examined (25 variables on 3 occasions for 78 cases), 56 or about 1% had z-scores in excess of ± 2.58. Univariate outliers, therefore, are not numerous. The presence of multivariate outliers was evaluated by examining the Mahlanobis distance of each case to the centroid of i t s group. No cases exceeded the c r i t i c a l value for Mahlanobis distance at p < .01. Overall, the distribution of scores for the dependent variables can be summarized as moderately skewed for a small proportion of cases but containing no outliers. Transformations might be appropriate for those variables with skewed distributions. This would, however, make interpretation more complex and would be required for not just skewed variables, but their repeated-measure counterparts. In light of the absence of outliers and the demonstration by Mardia (1971) that MANOVA i s robust to violation of normality when caused by skewness rather than outliers, then, no variables were transformed. Examination of the results of MANOVA following square-root transformation of skewed variables resulted, in fact, in results identical to those from the analysis of untransformed variables. Analysis of the distributions of variables following square-root transformation showed that this eliminated significant skewness. As a preliminary check of homogeneity of variance-covariance matrices, sample variances for each of the 78 dependent variables were examined across the 6 order groups. For no dependent variable did the ratio of the largest to smallest variance approach 20:1, the criterion suggested by Tabachnik and Fid e l l (1983). Univariate homogeneity of variance tests on the 78 variables revealed five significant results at p = .05, including one significant result, at p = .01 (Bartlett-Box F). Such a result i s to be expected given the number of tests performed. In any event, the robustness of significance tests in 172 MANOVA is guaranteed for equal sample sizes (Hakstian, Roed & Lind, 1979). Evidence of non-linearity of relationships among dependent variables was sought by examining the plots of observed values versus standardized residuals for each dependent variable. No evidence of gross curvilinearity was found, therefore variables were retained in original form. In the analyses that follow, MANOVA i s performed f i r s t on each set of manipulation check measures, and separately on the remaining measures. The presence of multicollinearity and singularity was investigated by examining the SMC of each variable with a l l other variables within each of these analyses. In no instance was a SMC close enough to 1.0 to warrant concern. Inspection of the eigenvalues and determinants of the within-cells correlation matrix revealed that they were sufficiently different from zero such that neither multicollinearity not singularity was judged to be a problem. Analysis of Manipulation Checks Recall that the manipulation checks consisted of six measures: self-reported anxiety, depression, h o s t i l i t y , pleasure, arousal, and a behavioral measure of response latency. Recall also that parallel versions of these measures were administered before and after the other dependent measures. Thus the manipulation checks consist of two sets of 6 variables. These w i l l be referred to as the " f i r s t " and "second" manipulation checks. Overall, multivariate analysis of variance would be appropriate to a test of differences between a l l 12 measures. However, the interpretation of subsequent analyses, particularly stepdown F-tests, would be problematic. Because stepdown analyses partial out the effects of other variables, and because we know that the parallel versions of each mood manipulation check 173 measure are very highly correlated, i f we included both sets of manipulation checks in a single stepdown analysis, we would not expect to find any effects for the second set of variables. Therefore the f i r s t and second manipulation checks are considered separately. However, to guard against increased experimentwise error inherent in such multiple testing, multivariate tests using a l l twelve measures simultaneously are also reported. The experimental design, as discussed above, included the within-subjects session variable, which was whether the measure was administered during the f i r s t , second, or third experimental session, and the between-subjects order variable, denoting which of the six possible orderings of the mood manipulation that the participants experienced over the three sessions. Each MANOVA was thus a 6 x 3 mixed, between-within analysis of 6 dependent variables. In both SPSSX and BMDP this type of design i s treated as a repeated-measures analysis of 18 dependent variables, with the dependent variables grouped to capture the session variable. It i s important to note that the influence of the mood treatment on the dependent measures i s captured in the order by session interaction. Because the order effect represents i n which session the participant receives each level of the mood induction, the effect of the mood induction on dependent measures i s that of the order by session interaction. One consequence of this i s to complicate inspection of group means. There are 78 order by session cells for each dependent variable. Each of these c e l l means for each dependent variable i s presented in Appendix F. To fa c i l i t a t e inspection of main and interaction effects, the tables in the text that follows have been collapsed according to the treatment effect of interest and show the mean score for the 78 participants for each variable in each treatment. 174 Results of Manipulation Checks. Wilk's criterion indicated a non-significant effect of order on the f i r s t and second manipulation checks taken simultaneously, F(60,289.42) = 0.79, p > .05. Order of mood treatment presentation over the three experimental sessions did not significantly affect the response of participants to the mood manipulations, as indicated by the 12 manipulation checks. The effect of interest, the order by session interaction, was significant for the f i r s t and second manipulation checks taken simultaneously, as indicated by Wilk's criterion, F (120, 1046.95) = 2.16, p< .001, rj2= .82. When the f i r s t manipulation checks are considered alone, the effect of order i s non-significant, F(30, 270)= 0.69, p> .05. The order by session interaction was significant, F(60, 733.32)= 3.11, p< .001, r?2= .70. Similarly, the second manipulation checks, considered separately, reveal a non-significant effect of order, F(30,270)= .69, p> .05, and a significant order by session interaction F(60, 733.32)= 2.22, p< .001, 7?2= .59. Thus, the mood manipulation had a significant effect on the manipulation checks administered both before and after the other dependent measures. The results also indicate a strong association between the mood treatment and the mood measures. To evaluate the contribution of each mood measure to the overall session by order interaction, stepdown F-tests were performed. The results of univariate analyses are shown in Table 18. Among the f i r s t manipulation checks, significant univariate order by session interaction effects were found for self-reported anxiety, depression, h o s t i l i t y , pleasure, and arousal. The latency measure was not significant. Such univariate analyses overlook correlations between the variables. Examination of the pooled, averaged within cells correlations, shown in Table 19, reveals that the measures of 175 Dependent Variable Univariate d.f. First Manipulation Check: Latency Arousal Pleasure Depression Anxiety Hostility .53 11.66 4.41 11.40 4.62 2.17 Second Manipulation Check: Latency Arousal Pleasure Depression Anxiety Hostility 1.39 2.29 2.59 7.37 1.76 1.73 10/144 10/144 10/144 10/144 10/144 10/144 10/144 10/144 10/144 10/144 10/144 10/144 .863 .000 .000 .000 .000 .023 .187 .016 .006 .000 .074 .080 .53 11.58 1.99 3.29 1.25 2.00 1.39 2.27 2.29 5.02 1.06 1.59 Stepdown d.f. 10/144 10/143 10/142 10/141 10/140 10/139 10/144 10/143 10/142 10/141 10/140 10/139 .863 .000 .039 .001 .262 .037 .187 .017 .016 .000 .397 .116 Table 18. Univariate and Stepdown F-tests, Order by Session (Mood) Effect on manipulation checks, Study Four. Latency Arousal Pleasure Depression Anxiety First Manipulation Checks: Arousal -.130 Pleasure -.129 .308 Depression .103 -.432 -.702 Anxiety .028 . .014 -.369 .430 Hostility .075 .178 .214 -.177 .192 Second Manipulation Checks : Arousal -.168 Pleasure -.043 .022 Depression .250 -.295 -.472 Anxiety .112 .072 -.358 .460 Hostility .110 -.068 -.390 .515 ,.297 Table 19. Pooled within-cell correlations for manipulation checks, Study Four. 177 mood are intercorrelated. Bartlett's test of sphericity was significant for the correlation matrices of the f i r s t and second manipulation checks ( x 2 (15)= 192.9, p< .001; x 2 (15)= 158.4, p< .001, respectively), supporting rejection of the hypothesis that the matrices are identity matrices and hence that the variables are independent. The results of the stepdown analyses are shown in Table 18. The order of entry into analysis was as follows: latency, arousal, pleasure, depression, anxiety, and ho s t i l i t y . Entered f i r s t , the stepdown test for response latency is equivalent to the univariate test, and so showed no effect. The second variable entered, self-reported arousal, showed a significant effect, stepdown F(10, 143)= 11.58, p< .001. The strength of association between the order by session effect and arousal, as indicated by the ratio of hypothesis sum of squares to hypothesis plus error sum of squares, was 0.45. Table 20 shows a steady trend from highest arousal in the elation treatment (31.0) to least arousal in the depression treatment (19.5), with the score for the neutral treatment (23.6) f a l l i n g closer to the depression treatment. After the pattern of differences measured by latency and arousal were entered, a difference was found for self-reported pleasure, stepdown F(10, 142)= 1.99, p< .05, r>2 = .11. Pleasure was greatest in the elation condition (37.1), slightly less in the neutral condition (36.0), and least in the depression condition (32.2) The fourth variable entered, depression showed a significant effect, stepdown F(10, 141)= 3.29, p< .001, rj2= .19. Self-reported depression was greatest in the depression treatment group (26.4) and least in the elation group (19.1). The mean score for the neutral group (21.9) was, as in the case for pleasure, closer to that of the elation condition. The f i f t h variable entered, anxiety, was not significantly affected, stepdown F ( l , 140)= 1.25, p> .05. Finally, h o s t i l i t y was significantly affected, stepdown F ( l , 139)= 2.00, 178 Elation Mean St. Dev. First Manipulation Check: Anxiety Depression Hostility Pleasure Arousal Latency 7.03 19.12 17.00 37.14 31.02 144.18 1.95 4.58 1.40 6.48 10.34 51.22 Second Manipulation Check: Anxiety Depression Hostility Pleasure Arousal Latency 8.19 19.91 15.92 31.76 29.46 107.85 2.13 4.39 71 73 35 27.73 Neutral Mean 7.19 21.98 16.70 36.06 23.65 145.63 8.02 21.65 15.73 31.56 28.15 110.21 St. Dev. 2.19 6.58 1.75 7.20 8.16 57.17 2.23 5.86 1.58 3.27 3.33 27.94 Depression Mean 8.11 26.41 16.53 32.21 19.51 146.03 8.53 23.92 16.10 31.00 27.88 107.69 St. Dev. 2.31 7.31 1.56 8.66 6.18 52.56 2.075 6.365 1.884 3.797 3.154 24.751 Table 20. Means and Standard Deviations for Manipulation Checks, by Mood Effect, Study Four. 179 p< .05. Scores on the ho s t i l i t y scale were highest in the elation condition (17.0), lower in the neutral condition (16.7) and least in the depression condition (16.5). Although s t a t i s t i c a l l y significant, these effects are small. The mood treatment had a much smaller effect on the second set of manipulation checks. Significant univariate effects were evident for arousal, pleasure and depression. In stepdown analyses, these three measures were shown to contribute uniquely to the overall order by session interaction. Latency, as before, did not significantly differentiate between treatment groups. Self-reported arousal was uniquely and significantly related to treatment, stepdown F(10, 143)= 2.27, p< .05, r)2= .14, as was pleasure, stepdown F(10, 142)= 2.29, p< .05, r)2= .14, and depression, stepdown F(10, 141)= 5.02, p< .001, 7)z= .26. Table 20 shows a congruent pattern of scores for the three variables, pleasure and arousal were highest in the elation group (31.8 and 29.5, respectively), intermediate i n the neutral group (31.6 and 28.1) and lowest in the depression group (31.0 and 27.9). The scores for depression mirrored these, being highest in the depression treatment (23.9), intermediate in the neutral treatment (21.6) and lowest in the elation treatment (19.9). These mean differences, although s t a t i s t i c a l l y significant, were notably smaller than those among the f i r s t set of manipulation checks. Analysis of the session variable revealed a significant multivariate effect on the combined dependent variables for both the f i r s t and second set of manipulation checks, F(12, 278) = 31.76, P < .001, and F(12, 278) = 9.07, p < .001, respectively. Wilk's criterion indicated that a strong association existed between the session variable and the combined dependent variables for the manipulation checks administered before the other measures (T?2 = .82), while a more moderate association existed for the second set of manipulation 180 checks (r)2 = .48). Examination of the univariate tests reveals significant session effects among the f i r s t manipulation checks for latency, arousal, depression and h o s t i l i t y . These are shown in Table 21. Cell means, shown in Table 22., show that the latency of participants' responses dropped markedly over the three sessions, from 203.5 to 125.1 to 107.3 seconds. Self-reported arousal increased somewhat (cell means of 22.3, 25.4, and 26.5 for sessions 1, 2, and 3) while depression and h o s t i l i t y decreased (23.9, 21.9, and 21.7 for depression; 17.0, 16.7, and 16.5 for h o s t i l i t y ) . However, when stepdown analyses were performed on these variables, the results showed that once the pattern of differences measured by the latency variable had been entered, the remaining variables did not contribute to the session effect. The stepdown results are also shown in Table 21. A similar pattern was evident for the second manipulation checks, but while the univariate test on self-reported arousal almost exceeds a significance criterion of .05, the only significant result was for response latency. Again, though, stepdown analysis revealed that once latency was entered into the analysis, the remaining variables did not make a unique contribution. Examination of c e l l means showed that, as for the f i r s t latency measure, response latency decreased markedly over the three sessions (means = 123.6, 107.2, and 95.0). Overall, the results for the manipulation checks indicate that the mood induction procedure produced significant shifts in the mood states of individuals as predicted. As was found in Study Two, the elation and depression inductions were characterized by self-reports of high and low pleasure, respectively, high and low arousal, respectively, and low and high depression, respectively. The neutral induction results were most like those 181 Dependent Variable Univariate First Manipulation Check: Latency Arousal Pleasure Depression Anxiety Hostility 314.15 6.74 .89 4.33 1.18 3.40 d.f. Second Manipulation Check: Latency Arousal Pleasure Depression Anxiety Hostility 63.08 2.95 0.03 1.26 0.97 2.54 2/144 2/144 2/144 2/144 2/144 2/144 2/144 2/144 2/144 2/144 2/144 2/144 .000 .002 .412 .015 .310 .036 .000 .055 .962 .286 .378 .082 Stepdown 314.15 0.05 0.53 0.72 1.03 0.97 63.08 0.11 0.13 0.49 0.28 1.17 d.f. 2/144 2/143 2/142 2/141 2/140 2/139 2/144 2/143 2/142 2/141 2/140 2/139 .000 .951 .587 .487 .356 .380 .000 .894 .875 .609 .755 .311 Table 21. Univariate and Stepdown F-tests, Session Effect on manipulation checks, Study Four. 182 Session One Mean Std. Dev. First Manipulation Check: Anxiety-Depression Hostility Pleasure Arousal Latency 7.58 23.84 17.02 34.50 22.26 203.48 2.16 7.06 1.55 8.09 8.92 46.09 Second Manipulation Check: Anxiety Depression Hostility Pleasure Arousal Latency 8.43 22.33 16.20 31.50 27.97 123.55 2.31 6.08 2.05 3.27 3.33 30.06 Session Two Mean 7.23 21.98 16.73 35.12 25.42 125.10 8.21 21.75 15.82 31.38 28.48 107.18 Std. Dev. 2.16 7.00 1.66 8.12 10.35 26.22 2.05 5.80 1.50 3.74 3.52 21.29 Session Three Mean 7.52 21.69 16.48 35.79 26.50 107.26 8.10 21.39 15.73 31.44 29.03 95.02 Std. Dev. 2.29 6.59 1.51 7.08 9.16 23.18 2.09 5.58 1.56 2.85 3.11 19.86 Table 22. Means and Standard deviations for manipulation checks, by Session Effect, Study Four. 183 for the elation induction for measures of pleasure and depression, but more like those for the depression induction on the arousal measure. The size of these effects was smaller than in Study Two, however, and no effect on the behavioral response latency measure was found. Significant effects were evident primarily for the manipulation checks administered f i r s t , directly after the musical induction. Those administered second, after the other measures had been collected, were smaller. The strength of association between the treatment and the combined dependent variables was lower for the second set of measures (TJ 2 = .59 vs. n2 = .82), and mean differences were very small. Informal analysis of the content of the imagery questionnaires completed revealed that in no instances did any participant report images concerned with task evaluation of performance or that might in any way represent a confound to the study of expectancy. Most often images concerned the content of the music, such as in reports of images of "marching bands", "busy c i t i e s " , "string quartets", "pastoral scenes", "rainy days", and "funerals". Participants often used words descriptive of moods and feelings. There was a high degree of agreement between the affective tone of the descriptions and the mood treatment which the participant had just completed. This further supports the effectiveness of the mood manipulation procedure. Worthy of comment is the finding that the behavioral response latency measure was not significant, unlike the finding of Study Two. That i s , no significant difference between mood treatment groups was evident on the psycho motor measure. One possible reason for this may be the increased familiarity and experience with the computer keyboard over the three experimental sessions for participants in Study Four.- Participants in Study Four used the keyboard for the experimental task as well as the dependent measures. The sensitivity 184 of the latency measure to the effects of depression may have been overshadowed by the effects of practice. Evidence for this i s found in the multivariate effect of the session variable on the manipulation checks which was revealed by stepdown analyses to be due to the latency measure. Over the three experimental sessions, the response latency of participants decreased by almost 50%. Measures of Expectancy and Task Perceptions A multivariate, repeated-measures analysis of variance was performed on the 13 remaining dependent measures. These consisted of the principal effort-performance covariation measure, measures of perceived control and correlation, measures of rec a l l , accuracy, and expected performance, of causal attributions and task perceptions. The multivariate test of significance of the order effect was not significant, F (65,287.49) = 1.12, p > .05. The order of presentation of the mood effect had l i t t l e apparent effect on the combined dependent measures. The means and standard deviations of each measure for a l l order cells are shown in Appendix F. As before, the mood treatment effect i s tested by the session by order interaction effect. For the combined dependent measures, the overall multivariate test was not significant, F (130, 1072) = .838, p >.05. Thus the results do not support Hypothesis One or Two. That i s , there was no effect of mood treatment on expectancy (Hypothesis One), nor on internality of attributions (Hypothesis Two). It is possible that adjustment for individual differences might reveal some effect, therefore Hypothesis Three will be tested below. Although the absence of a significant omnibus MANOVA precludes most further analysis, the reader might be interested to know that for not one 185 of the thirteen dependent variables was the univariate order by session test significant. The effect of mood was, apparently, not present or not detected. The means and standard deviations of the dependent variables, aggregated by mood treatment, are shown in Table 23. The averaged multivariate effect of the session variable on the combined dependent variables was significant, F(26, 264) = 2.35, p < .001, r?2 = .34. Examination of the univariate tests revealed significant effects for perceived control, verbal expectancy, performance r e c a l l , recall accuracy, task effort, task interest, and task d i f f i c u l t y and challenge. The univariate results are shown in Table 24. Inspection of the means in Table 25 shows that mean perceived control over performance declines over sessions as does the verbal measure of expectancy. Individuals recall higher levels of performance in later sessions and, in fact, this recall i s more accurate in subsequent sessions. Self-reported task effort declined over the three sessions and individuals reported the task to be less d i f f i c u l t and challenging, and less interesting i n later sessions. To investigate the unique effects of the session main effect on the individual dependent variables, a stepdown analysis was performed. Examination of the averaged within-cells correlation matrix, shown in Table 26, reveals that the dependent measures are intercorrelated. Bartlett's test of sphericity was significant, F (13,144) = 10692.9, p < .001, supporting rejection of the hypothesis that the matrix i s an identity matrix and that the variables are independent. The a p r i o r i ordering of the importance of the dependent variables was as follows: (1) perceived covariation, (2) control, and (3) correlation, (4) verbal expectancy, (5) performance re c a l l , (6) recall accuracy, (7) performance expectation, (8) internality, (9) stability, of causality, (10) task effort, (11) task interest, (12) satisfaction with task 186 Measure Perceived Covariation Control Correlation Satisfaction with performance Effort Task d i f f i c u l t y and challenge Task Interest Expectancy-verbal Attributed Internality Stability Controllability Performance recall expectation Recall accuracy Elated Mean .22 52.24 4.79 13.29 6.84 11.94 6.93 28.10 15.87 12.23 17.92 3.45 3.56 .22 Std.Dev. .18 19.41 2.15 2.03 1.33 2.74 1.52 5.13 4.47 4.30 3.98 .51 .66 .39 Neutral Mean .24 53.59 4.92 13.12 6.65 11.62 6.64 27.16 16.05 12.15 17.64 3.46 3.59 .22 Std.Dev. .16 20.06 1.95 2.21 1.60 2.69 1.53 5.43 4.60 4.20 3.69 .60 .47 .57 Depressed Mean .24 53.98 4.83 13.48 6.73 11.67 6.66 27.89 15.43 12.30 17.28 3.42 3.60 .21 Std.Dev. .18 21.42 2.23 2.39 1.53 2.58 1.63 4.58 4.79 3.98 3.64 .61 .51 .52 Table 23. Means and standard deviations for dependent variables by mood effect, Study Four. 187 Univaria ite Stepdc >wn Measure F d.f. P F d.f. P Perceived covariation control correlation 1.56 8.37 .77 2/144 2/144 2/144 .215 .000 .463 1.99 10.44 1.44 2/144 2/143 2/142 .140 .000 .241 .027 .127 .018 Verbal expectancy 4.56 2/144 .012 1.21 2/141 .301 .017 Performance reca l l expectation 3.11 .51 2/144 2/144 .048 .603 1.59 1.26 2/140 2/139 .206 .288 .022 .018 Recall accuracy 6.54 2/144 .002 .05 2/138 .950 .000 Attributions of internality s t a b i l i t y 2.17 .33 2/144 2/144 .117 .720 2.31 2.61 2/137 2/136 .735 .077 .043 .037 Task effort 11.74 2/144 .000 3.32 2/135 .039 .047 Task interest 6.86 2/144 .001 4.75 2/134 .010 .066 Satisfaction with task performance 1.99 2/144 .140 .02 2/133 .980 .000 Task d i f f i c u l t y and challenge 3.05 2/144 .050 .99 2/132 .372 .815 T Table 24. Univariate and Stepdown F-tests on dependent variables, Session effect, Study Four. 188 Perceived covariation Perceived control Perceived correlation Verbal Expectancy Performance recall Performance expectation Recall accuracy Internality Stability Task Effort Task Interest Satisfaction with task performance Task Di f f i c u l t y and challenge Session One Mean .25 58.60 5.00 28.67 3.33 3.54 .35 16.01 12.10 7.19 7.05 13.57 12.15 Std.Dev. .14 18.47 2.16 4.42 .72 .64 .64 4.34 4.03 1.25 1.55 2.02 2.99 Session Two Mean .22 51.85 4.67 27.58 3.46 3.61 .21 16.12 12.44 6.74 6.64 13.34 11.53 Std.Dev. .18 20.55 2.17 5.03 .56 .56 .50 4.83 4.48 1.48 1.57 2.23 2.45 Session Three Mean .22 49.36 4.87 26.89 3.54 3.60 .09 15.21 12.14 6.29 6.55 12.98 11.56 Std.Dev. .20 20.73 1.99 5.55 .37 .44 .23 4.65 3.95 1.57 1.54 2.36 2.50 Table 25. Means and Standard Deviations for dependent variables by Sessions, Study Four. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (1) S a t i s f a c t i o n w i t h task performance (2) Task e f f o r t . 1 16 (3) Task d i f f i c u l t y and chal1enge - . 177 .274 (4) Task i n t e r e s t .111 . 4 18 . 336 (5) Ver b a l expectancy -.052 . 191 . 153 .209 (6) P e r c e i v e d i n t e r n a l i t y .060 .200 .077 . 133 .485 (7) s t a b i 1 i ty - .003 . 108 - .077 - . 194 .051 . 137 (8) P e r c e i v e d c o v a r i a t i o n - . 152 . 103 . 128 . 136 .395 .207 .010 (9) c o n t r o l . 321 .238 .077 .218 .513 . 522 .081 . 193 ( 10) c o r r e l a t i o n . 206 .229 . 108 . 192 .447 . 358 - .009 . 229 .514 (11) Performance r e c a l l . 189 .098 - .057 .026 .095 . 154 - .027 - .009 . 125 .026 ( 12) Performance e x p e c t a t i o n .311 - .020 - . 129 - .030 .225 .209 - . 104 .052 .203 .080 . 6 16 (13) Recal1 a c c u r a c y . 104 - .072 .017 .016 - .049 - . 107 - .004 .039 .008 .060 -.833 - . 360 T a b l e 26. W i t h i n - c e l l s c o r r e l a t i o n m a t r i x , Study Four. 190 performance, and (13) task d i f f i c u l t y and challenge. The results of the stepdown analysis are shown in Table 24. Unadjusted session group means are shown in Table 25. Table 24 shows that, when the pattern of differences measured by variables entered earlier into the stepdown analyses are accounted for, significant unique contributions to the multivariate test are made by perceived control, performance re c a l l , and task effort. These session effects are l i k e l y the result of increased experience with the experimental task and reduced novelty of the experimental procedure. Individual Differences To test whether individual difference variables moderated the effect of mood on self-esteem a sub-group analysis was performed. That i s , participants were divided into those with high scores and those with low scores, and this factor was added as an independent variable to the multivariate analysis of variance. The use of a within-subjects design precluded the treatment of the individual difference measures as continuous covariates. Because the individual difference measure is assumed to be stable, i t i s measured only once and is therefore constant across levels of the within- subjects factor. So no covariate effect on the mood factor i s possible. Within each order condition group the six participants highest on an individual difference variable were placed i n one subgroup and the six lowest scorers were placed in the other. Data for the remaining participant in each order group was discarded. Ties were broken randomly. This procedure maintained equality of c e l l size, thus avoiding problems caused by non-orthogonality and vulnerability to violation of assumptions. The resulting tests are very conservative because of the power loss inherent in adding a 191 factor, and because some members of a "high" subgroup in one order condition may actually have lower individual difference scores than members of a "low" subgroup i n another order condition, and vice versa. The self-esteem subgroup analysis revealed no significant effects. The multivariate main effect of self-esteem was not significant, neither were the interaction effects with order, session or order by session, the latter representing the mood effect. Thus individual differences in self-esteem did not have any influence on the mood — expectancy relationship. Hypothesis Three received no support. A similar analysis for the interpersonal control scale of the Spheres of Control measure revealed no main effects or simple interactions, however the interpersonal control by order by session interaction was marginally significant, F(130,1170) = 1.23, p = .051 by P i l l a i s ' criterion, F(130,880) = 1.26, p = .035 by Wilks' criterion. The presence'of a multivariate effect suggests at least one significant univariate effect: exactly one was found. The univariate three way interaction was significant for the perceived correlation measure. Inspection of the c e l l means indicated that the form of the order by session interaction (the mood effect) was different for each control subgroup in the following way: For individuals high in interpersonal control, scores on the perceived correlation measure were high in the elation and depression conditions and lower in the neutral condition. For individuals low in interpersonal control, scores were low in the elation and depression conditions and higher in the neutral condition. The lack of any other effects and the marginal nature of the multivariate significance test suggest that l i t t l e meaning should be attached to this result. It i s , quite possibly, due to chance alone. Finally, no effects were found in the analysis of subgroups formed on the 192 sociopolitical control measure. Discussion: Alternate Explanations A number of alternative explanations exist for the failure of Study Four to demonstrate a relationship between mood and expectancy. Among these are failure of the musical induction procedure to successfully manipulate mood, insensitivity of the expectancy measure, lack of experimental control, insufficient s t a t i s t i c a l power to detect a relationship or the absence of any true relationship. Support for each of these po s s i b i l i t i e s w i l l be considered in turn and ways of surmounting them w i l l be explored. Manipulation Failure It i s possible that a relationship exists between mood and expectancy, as hypothesized, but that i t was not revealed because mood was not successfully manipulated. In other word, there was no mood effect because there was no mood change. The a p r i o r i support for this alternate explanation i s mixed. Recall that Study Two showed that the manipulation was successful in influencing both self-reported mood and an unobtrusive, behavioral response latency indicator of depression. The self-report measures have a long history of use and substantial support for their v a l i d i t y . Taken alone, Study Two suggests that we can have confidence in the success of the manipulation. However, Study Four contained a task that was not present in Study Two. The self-report manipulation checks used there were significantly influenced by the manipulation, but the latency measure was not. Further, the size of the manipulation effect was much smaller on the measures taken after the principal dependent measures. From these results we can conclude f i r s t that a 193 significant difference in self-reported mood did exist following the mood induction. Second, while the difference in self-reported mood at the time of the second manipulation check was also s t a t i s t i c a l l y significant, the mean differences were small. Overall, the mood induction procedure appeared to change mood as hypothesized but this manipulation was greatly diminished over the measurement period. These mood shifts, although s t a t i s t i c a l l y significant, may not have been of sufficient magnitude to cause the effects hypothesized. Eich and Metcalfe (in press) found that the Velten induction did not produce mood state dependent learning, although i t evidently influenced manipulation checks. The magnitude of the mood differences induced in Study Four, although significantly different across conditions, may not have been sufficient to drive the memory effects. The complexity of the experimental procedure may have contributed to this. Consider, for instance, differences between Study Two and Study Four. Study Two employed only the musical induction and subsequent manipulation checks. Study Four also employed a r e a l i s t i c business decision making task, which may have served to distract participants and moderate their attention to the induction. Comparison of the mean mood scores of Study Four (Table 20) with those of Study Two (Table 14) (adjusting for the fact that the MAACL scales in Study Four had half as many items) shows that the former were smaller than the latter. For example, the mean scores for pleasure in the Elation and Depression conditions of Study Two were 41.8 and 25.6, respectively, compared to 37.1 and 32.2 in Study Four. Comparable depression scores in the Elation and Depression conditions were 17.3 and 33.1 in Study Two, versus 19.1 and 26.4 in Study Two. It should be noted that there were other differences between Studies Two and Four, such as the between— versus within—subjects design. Nevertheless, i t is possible that 194 the additional elements of Study Four moderated the mood manipulation. The additional cognitive requirements of the decision- making task may have served as a distraction. Alternatively, the repeated exposure to the mood inductions procedure, though not to a particular treatment condition may have reduced the novelty and hence interest and attention to the musical selections. To test the possibility that carry over effects from session to session interfered with the mood manipulation or i t s effects, a between-subjects analysis was performed on the two sets of manipulation checks and on the set of dependent measures. The results indicated a significant overall effect of the mood manipulation on the f i r s t set of manipulation checks, F(12,140) = 2.87, p < .001, n2 = '36. Examination of the univariate tests showed significant effects for scores on depression, F(l,75) = 10.27, p < .001, and on arousal, F(l,75) = 6.83, p < .002. The means in the Elation, Neutral, and Depression conditions were 20.54, 22.73, and 28.27, respectively, for depression, and 26.31, 22.69, and 17.81, respectively, for arousal. The overall effect of the induction of the second manipulation checks was not significant, F(12,140) = 1.64, p < .088. For the combined dependent variables the multivariate effect was not significant, F(26,126) = .836, p = .694. None of the univariate tests on the dependent variables were significant. These results indicate that the presence of carry-over effects i s not a lik e l y explanation for failure to find an effect of mood in Study Four. When presented with only one induction for a f i r s t time, individuals reported changes in their mood as measured before the dependent measures, but did not evidence any effects on the dependent measures nor on the second set of manipulation checks. Thus some additional evidence as to the weakness of the mood manipulation i s also shown. It should be noted that the nominal relative power of the between-subjects analysis, above, and the mixed-between-within 195 analysis in Study Four is v i r t u a l l y identical. In the within-subjects case, however, the effect size might be expected, a p r i o r i , to be larger. It i s , nonetheless, rather low. The preceding sections considered why the complete experimental procedure employed in Study Four failed to demonstrate an effect. Further insight into the impact of the mood induction might be gained from internal analyses. Analyses, that i s , performed only on those participants who were, according to their mood self-reports, influenced by the manipulation in a predicted pattern. Such analyses have the potential to reveal effects otherwise diluted by the failure of some participants to respond to the manipulation. However, such selection limits our a b i l i t y to make causal statements about the effect of the indpendent variable. The responses of the participants in Study Four to the mood induction were examined on the basis of their scores on the f i r s t depression, arousal, and pleasure manipulation checks. These variables are the best indicators of the success of the mood induction. Each participant's pattern of scores over the three sessions was examined. Participants whose scores f i t the pattern predicted by their order treatment were noted. For example, a participant whose scores on depression were highest in the session in which they recieved the depression induction, lowest after the elation induction, and intermediate following the neutral induction were classified as f i t t i n g the predicted pattern for that measure. Each pattern was examined for each of the three manipulation checks. Of the 78 participants, 44 evidenced f i t for the depression measure, 43 for the arousal measure, and 30 for the pleasure measure. Two sets of analyses were performed: One set on the 44 participants who reported the predicted pattern of scores on the depression measure, and one set on the 18 participants who f i t the predicted pattern on a l l three manipulation checks. These represent the least and most stringently 196 selected individuals. The results of both sets of analyses were similar and support the results of earlier analyses. In each internal analysis the multivariate test on the f i r s t set of manipulation checks was significant, obviously in part due to the selection procedure, F(60,345.61) = 3.58, p = .000 for N = 41, and F(60,94.12) = 3.08, p = .000 for N = 18. The mean scores for depression in the Elation, Neutral, and Depression conditions were, respectively, 17.95, 22.19, and 29.68 for N = 41, and 18.00, 22.06, and 32.41 for N = 18. The mean scores for arousal in the Elation, Neutral, and Depression conditions were, respectively, 33.90, 22.17, and 17.41 for N = 41, and 37.23, 23.47, and 16.35 for N = 18. The mean scores for pleasure in the Elation, Neutral, and Depression conditions were, respectively, 38.54, 36.54, and 28.78 for N = 41, and 39.88, 34.58, and 22.71 for N = 18. Despite the selection, however, for both analyses no multivariate effects of mood on the dependent measures was found, F(130,480) = .959, p = .606 for N = 41, and F(130,96) = .984, p = .538 for N = 18. For none of the dependent measures was their a significant univariate effect. Similarly, their were no multivariate or univariate effects of mood on the 15 probabilities that comprise the composite covariation measure. It is clear, then, that the mood manipulation as evident in the self-reported responses to the manipulation checks, did not have an effect on the dependent measures. In light of a l l of the above, what can be reasonably concluded about the manipulation of mood in Study Four? It i s evident that differences in self-reported mood greater than chance were found. That i s , s t a t i s t i c a l l y significant effects of mood on the manipulation checks were present. However, these differences may not be practically significant or meaningful. In support of this argument is the finding of small differences among the mean mood scores. Countering this argument i s the fact that the significant mood 197 effects were not, as we shall discuss more completely in a later section, the result of excess power. Further, a review of the size of mood effects found in the experimental literature reveals studies whose effect sizes as indicated on mood scales of the sort used here were smaller than found here yet which were sufficient to produce hypothesized effects (e.g., Sutherland, Newman, & Rachman, 1982; Teasdale & Russell, 1983). In sum their i s evidence for both the position that the mood induction in Study Four was sufficent to produce the hypothesized effects as well as for the position that the effects produced were of l i t t l e meaningful significance. Future research might seek to strengthen the mood manipulation, perhaps by adding measures designed to focus individuals' attention. Study Four included a step in this direction by asking participants to think of the images that the music brought to mind and subsequently asking about those images. Strengthening the manipulation in this and other ways might also increase the spectre of hypothesis guessing by participants. Alternatively a procedure like that used by Eich and Metcalfe (1987) could be employed. They required that subjects meet a mood change criterion before proceeding with tests of mood effects. The effectiveness of the manipulation could then be verified by using unobtrusive measures taken before and after the dependent measures. Measurement Error A relationship between mood and expectancy may not have been found because the expectancy measure was not sensitive to real changes in perceived expectancy. That i s , there may have been expectancy change which was not measured. L i t t l e support exists for this explanation. Study Three showed that 198 scores on the expectancy measure matched objective differences in effort-performance covariation between tasks. The expectancy measure was shown to be valid. Further, Study Four included a verbal, rating scale measure of participants' agreement with statements indicative of effort-performance covariation. This measure was also not influenced by mood. In order to determine whether mood might have had an effect on the individual components of the matrices used to compute the covariation measure, analysis of variance was performed on the scores for the 15 c e l l s . The analysis was performed on the non-normalized scores (i.e., prior to dividing them by their total so that they sum to unity). The overall multivariate effect was not significant, F(150,1108.22) = .886, p = .825. Inspection of the 15 univariate tests revealed no significant effects. The same result was evident from analyses of the normalized scores. Mood did not, apparently, influence judgements of the probability associated with any c e l l of the matrix, just as i t had no influence on the composite measure derived from those probabilities. In sum, then, there is good evidence from Study Three for the valid i t y of the expectancy dependent variable. Recall, though, that this conclusion was qualified by other differences between the conditions in Study Three, and by the need for replication of Study Three. To the extent that the perceived covariation measure of expectancy is valid, the effect of the mood induction on perceived effort-performance covariation would have been reflected in the expectancy measure. Lack of Experimental Control When an experiment succeeds from the researcher's point of view, or hypothesized differences are found to reliably exist, then i t i s incumbent on 199 the researcher to consider alternative explanations for the effect or effects. Of particular concern are confounded alternative causes, or experimental procedures that covary with the intended treatment, such that effects could be attributable to either or both. Experimental procedures that are essentially random, which do not vary systematically with the treatments, are ignored because they serve to inflate error variance. They reduce the study's a b i l i t y to find an effect. When an experiment succeeds i t has apparently done so despite such error, and the error i s of no interest. When an experiment f a i l s , the sources of random error are of interest. Unfortunately, they are by nature random and so are d i f f i c u l t to identify. In this study, although three experimenters were used, the gender of the experimenter was controlled. Care was taken to ensure that experimenter-participant interaction was minimized and standardized. It i s unlikely that this contributed substantially to error variance. In the within-subjects, repeated-measures design used, participants completed the decision making task three times. A different decision problem was used each time but the decision problems were of essentially equal d i f f i c u l t y . The particular decision problem used in each session was determined randomly to prevent them from being confounded with the mood manipulation or session. Therefore, differences in the perceptions of the task due to the decision problem would have been included as error variance. However, i t is unlikely that the particular decision problem had an effect on expectancy. In sum, few obvious, uncontrolled sources of random error were present in the study. Future research would find l i t t l e upon which to improve. Lack of Statistical Power 200 It i s possible that a relationship exists between mood and expectancy which was not detected because the experimental design lacked s t a t i s t i c a l power. St a t i s t i c a l power i s defined as the probability of rejecting a false null sypotheses. Stat i s t i c a l power has three codeterminants: (1) the significance criterion, or alpha level chosen by the researcher as the probability of rejecting a true null hypothesis, (2) the r e l i a b i l i t y of sample results, which is dependent on the size of a sample, and (3) the effect size or magnitude of the phenomenon in a population (Cohen, 1977; Mazen, Graf, Kellogg & Hemmasi, 1987). In other words, the larger alpha i s (e.g., .10 vs. .05), the more like l y a false null hypothesis w i l l be rejected and the higher the power. The larger the sample size i s , the greater the r e l i a b i l i t y , and the higher the probability of rejecting a false null hypothesis. The larger the effect size, the more li k e l y the effect w i l l manifest i t s e l f and be detected. If these three parameters are known, power can be precisely determined. Conversely, the sample size, effect size, and alpha necessary to set power at a certain level can be determined. As Mazen et a l . say, the ideal procedure would be to a p r i o r i set alpha and the level of power, estimate effect size, and then solve for the necessary sample size. Power can, of course, be calculated after the fact of an experiment. Most d i f f i c u l t to estimate among the three parameters is effect size. Often past, well-conceived research from which the proportion of variance explained can be gleaned does not exist. Cohen has for this reason, provided conventions for effect size, corresponding to small, medium, and large effects. For example, for the standardized difference between two means, the three values are .20, .50, and .80. The u t i l i t y of power analysis i s that i t reveals the likelihood of 201 finding an effect in a particular research contest. Cohen (1977) recommends that the probability of f a i l i n g to reject a false null hypothesis (the likelihood of Type II error) be set at .20, and hence that power be set at .80. Investigations should, according to Cohen, have at least an 80% chance of rejecting a false n u l l . Power analysis i s not the norm in present day research, however (Brewer, 1972; Chase & Chase, 1976; Mazen et a l . , 1987). Power analysis can also be used to sustain the null hypothesis. If sufficient power exists (say, .90) to find an effect so small that the researcher i s -willing to declare i t of no meaning, and this effect is not found, then i t i s possible to conclude that the likelihood of not finding an effect where one exists i s significantly small and thus the hypothesis that an effect exists can be rejected. Such a conclusion is limited to the particular study, however. Possibly for this reason, and because accepting the null runs counter to the logic of f a l s i f i c a t i o n as usually employed, power analysis is seldom used in this way, when i t i s performed at a l l . Cohen (1977) has published tables which show power for most sample sizes, alpha values, and effect sizes. He provides indices of effect size for most types of significance tests and formulae which adjust the sample size parameter so that correct estimates of power are made in the case of complex research designs. In a mixed between-within design, such as Study Four, the main and interaction effects have differing degrees of freedom and hence differing power. The index of effect size for significance tests in univariate analysis of variance (he does not consider multivariate ANOVA) is T? 2, the proportion of sample variance explained by an effect. The levels of effect size corresponding to small, medium and large for r\2 are .01, .06, and .14. In Study Four, s t a t i s t i c a l power was .94 to detect a large mood (session 202 by order interaction) effect, .50 to detect a medium mood effect, and .10 to detect a small mood effect. These values were determined for a 6 by 3, between by within design, where alpha = .05 and c e l l size i s 13. Power to detect a small, medium, and large order effect was .08, .33, and .75, respectively. For the session effect corresponding power values were .17, .78, and .99. Mazen and colleagues surveyed the power values in a sample of the 1984 management literature, which we can use to provide a context for the power of Study Four. They found that the average power to detect small effects was .31, although the median power was .25 and the mode .14. The corresponding values for medium effects were .77, .89 and .99, and for large effects .91, .99 and .99. Compared to the management literature, then, Study Four had lower than typical power. It should be noted, though, that only about 10% of the significance tests included in Mazen et al.'s survey were F-tests. Most of the remainder were tests of correlation or regression coefficients. It i s lik e l y that sample sizes in the experimental designs surveyed by Mazen et a l . were smaller than those in other designs, and hence the average power of experimental designs may be lower than the average they reported. The st a t i s t i c a l power of Study Four i s thus l i k e l y closer to the average in the experimental management research literature. Further, surveys of s t a t i s t i c a l power in other disciplines report average power lower than that in the management literature. Cohen (1962) found average power to detect small, medium and large effects of .18, .48, and .83 among a sample of studies in the 1960 Journal of Abnormal and Social Psychology. This may be because other disciplines, such as psychology, employ experimental designs proportionately more often. Or i t may be that more recent studies have higher power — the survey reported in the psychological 203 literature i s much older. Nevertheless, the s t a t i s t i c a l power of Study Four was below that typical in management research and psychology, even though i t was close to that in the psychological literature. It was not sufficient to permit us to declare that we had high power to detect a small but meaningful effect and hence that the null hypothesis i s sustained. How could the power of Study Four be improved? First, sample size could be increased. In order to achieve .80 power to detect a small, medium and large effect of mood, given the same alpha of .05, 136, 23, and 9 subjects would be respectively required. Second, steps could be taken to increase effect size. Effect size i s a function of both mean differences and error variance. As discussed above, steps were taken to minimize random error. Differences between the means of treatment groups depend on the treatment effect and the real relationship between the treatment effect and the real relationship between the treatment and dependent measures. The mood induction used in Study Four was shown to be valid by Study Two. A within-subjects design was chosen to control for factors that might moderate the real relationship between mood and expectancy. It seems, then, that increasing effect size much over Study Four might not be possible. To achieve power of .80, with sample size equal to 13 and alpha =.05 an effect size equivalent to ?72= .101 would be necessary. This corresponds to Cohen's convention for a large effect size. Finally, power might be increased by relaxing alpha. The probability of rejecting a true null hypothesis is conventionally limited to 5%. However, i t could be argued that this i s too stringent. Especially in the case of research on new topics, using new measures and variables, a less stringent criterion might be appropriate to declare that a real, non-chance effect 204 exists at alpha = .10. Unfortunately, the findings of Study Four would not meet this criterion. The power of Study Four to detect a small, medium and large effect where alpha = .10, and sample size of 13 would be .18, .63, and .97 respectively. Clearly, relaxing alpha improves power, especially for small and medium effect sizes, but does not increase i t .80, the level sought. In summary, lack of s t a t i s t i c a l power i s a plausible explanation for the failure of Study Four to find an effect. With the same design, power could be increased in the ways described above. Alternatively, future research could employ a different design. A between subjects design with the same c e l l sample size of 13 and alpha =.05, would have power of .08, .25, and .57 to detect a small, medium and large difference. A between subjects design with the same number of participants as Study Four, namely 78, with 26 in each treatment would have corresponding power of .11, .48, and .89. Finally, a between subjects design with the same number of observations," namely 234, with 78 subjects i n each treatment, would have respective power of .26, .94, and more than .995. The Null Hypothesis is True It i s possible that mood does affect expectancy judgements but that the conditions of Study Four did not f a c i l i t a t e demonstrating such a relationship. For example i t might be that expectancy was measured too soon after task performance. If we understand mood to influence subjective perceptions then i t follows that i t w i l l emerge when objective influences are weaker. When recall i s for a task farther in the past, mood may have an influence on memories of performance and hence expectancy judgements. It might also be the case that the level of objective expectancy was too high in the Marketing Game. That i s , studies of the influence of mood on contingency judgements 205 reveal that mood matters when objective contingency i s low. It may be that the objective effort performance relationship in Study Four was not low enough. When no relationship exists then individuals in a positive mood may perceive one. Explanations like the preceding, which speculate as to the conditions under which an effect might be found, must be considered in the context of a study of organizational motivation. That i s , although we might be able to find an effect, in doing so we are moving farther away from the conditions found in organizations. We can remove expectancy measurement temporally from performance of an experimental task. In organizations, however, expectancy perceptions arise in the midst of ongoing work. We can produce an experimental task that has no expectancy. However, work tasks in organizations do have expectancy. The f i n a l explanation for the failure of Study Four to show a relationship between mood and expectancy i s that the null hypothesis i s true, that no relationship exists. The cognitive processes involved in the formation of expectancy percepts may be resistant to mood effects. It may be that estimates of the relationship between effort and performance are robust to changes in mood state. People experience success and failure in tasks, they do not abandon tasks because failures cause them to believe that no success i s possible. In this way i t i s possible that expectancy is robust to changes in performance. If so, i t i s unlikely that transient mood states can influence expectancy judgements. Related to this i s the possibility that the process of forming expectancy judgements requires so much attention and cognitive effort that mood i s destroyed or severely diminished when the questions that comprise the covariation variable are presented. Studies which have used t r i v i a l 206 experimental tasks may have been able to show an effect because of that t r i v i a l i t y . When a person i s bored at a task their mood might be easily manipulated and might influence their task perceptions. But when involved in a more interesting experimental task their mood and perceptions might be immune to manipulation. It may be an irony that more interesting, emotionally involving tasks are thus less influenced by previously induced mood. The theoretical chain which we described in Chapter Four is long. It may be too long. Recall that we postulated that i f mood influences recall of positive and negative events, and recall influences probability judgements, and probability judgements combine to form expectancy, then mood should influence expectancy. In a chain like this each link i s required for the connection between mood and expectancy to be made. Any weakness in relationship between parts of the chain, as a result of truly weak or limited relationships, or of experimental imprecision, threatens the valid i t y of the model and our a b i l i t y to support i t experimentally. Conclusion If Study Four had shown a significant relationship between mood and expectancy, then we would be discussing the implications of such a relationship for organizations. We would also be discussing threats to the internal v a l i d i t y of Study Four and implications for future research. Instead, we are le f t with uncertainty as to whether an effect might be detected in another study, whether mood influences expectancy under only very specialized conditions which might not easily generalize to organizational settings, or whether mood does not affect expectancy at a l l . So what does this dissertation mean for the study of emotions in organizations and for the study of expectancy? First, i t has taken a 207 theoretical framework from psychology and applied i t to organizational behavior. That i s , the network model of how mood influences memory and judgement has been used to make predictions about effects on variables of interest to organizational researchers. The lack of success in the domain chosen does not negate the potential applicability in other domains. Future research could apply similar theoretical frameworks to other variables, such as performance appraisal, employee selection, risk-taking, organizational decision-making, or prosocial organizational behavior. Second, i t has shown that mood can be successfully manipulated in an experimental context applicable to studying organizations and that mood can be validly measured in such contexts. Similar experimental procedures, or procedures using different mood manipulations can be applied to future research questions. The measures of mood used can be applied in experimental settings like those used here, or to non-experimental research designs including questionnaire research. Third, this dissertation has provided a rationale for the measurement of expectancy in a manner consistent with Vroom's conceptualization, and has validated such measurement. In doing so i t has examined the relationship between objective task expectancy and task d i f f i c u l t y and shown them to be unrelated contrary to much of the literature. It has also argued that multiple effort and performance level measurement of expectancy allows comparison of expectancies to be made across individuals. Expectancy theory is shown to not be limited only to "within-subjects" predictions. The question of how judgements of the contingency between effort and performance are formed are raised also. That i s , what i s i t about tasks that causes individuals to perceive high or low expectancy? While much research i s proceeding on this topic in psychology, l i t t l e attention i s being directed to 208 i t in organizational behavior, although i t has direct relevance to job design. Conceptualizing expectancy as covariation and measuring i t accordingly opens a number of directions for expectancy research. Finally, although i t showed only that mood could be manipulated Study Four provided a basis for future investigations of mood and expectancy. A number of possible reasons for the failure to find an effect were explored and alternative approaches considered. On this foundation i t may yet be demonstrated that mood does unequivocally affect expectancy or, unequivocally, that i t does not. 210 PART ONE Remember that the information collected here w i l l be confidential and used only for the purposes of the study. Age: Sex: Please read each of the following statements. Where there i s a blank , decide what your normal or usual attitude, feeling or behavior would be: 1 2 3 4 5 RARELY OCCASIONALLY SOMETIMES FREQUENTLY USUALLY Write the number that describes your usual attitude or behavior in the blank. For example: A. I t e l l others that Vancouver i s a fine city. If you RARELY t e l l others this, write a 1 in the blank. If you OCCASIONALLY t e l l others this, write a 2 in the blank, and so on. Answer a l l the items. If you have d i f f i c u l t y with one do not leave i t blank, answer as best you can. 1. When faced with a problem I try to forget i t . 2. I throw my l i t t e r into waste paper baskets on the street. 3. I need frequent encouragement from others for me to keep working at a d i f f i c u l t task. 4. I feel that I'm a person of worth, at least on an equal basis with others. 5. I like jobs where I can make decisions and be responsible for my own work. 6. I have received too much change from a cashier and not said anything. 7. I change my opinion when someone I admire disagrees with me. 8. When I hear people talking privately I avoid listening. 9. If I want something I work hard to get i t . 10. I feel that I have a number of good qualities. 211 1 2 3 4 5 RARELY OCCASIONALLY SOMETIMES FREQUENTLY USUALLY 11. I prefer to learn the facts about something from someone else rather than have to dig them out for myself. 12. I have taken things that didn't belong to me. 13. I w i l l accept jobs that require me to supervise others. 14. I t e l l l i e s i f I have to. 15. I have a hard time saying "no" when somone tries to s e l l me something I don't want. 16. A l l in a l l , I am inclined to think that I am a failure. 17. I like to have a say in any decisions made by any group I'm i n . 18. I keep my promises, no matter how inconvenient i t might be to do so. 19. I consider the different sides of an issue before making any decisions. 20. I take a sick-leave from work or school even though I wasn't really sick. 21. What other people think has a great influence on my behavior. 22. I am able to do things as well as most other people. 23. Whenever something good happens to me I feel i t i s because I've earned i t . 24. I like to gossip about other people's business. 25. I enjoy being in a position of leadership. 212 RARELY OCCASIONALLY SOMETIMES FREQUENTLY USUALLY 26. I have done things that I don't t e l l other people about. 2 7 • 1 n e e d someone else to praise my work before I am satisfied with what I've done. 2 8 • 1 feel I do not have much to be proud of. 2 9 • 1 am sure enough of my opinions to try and influence others. 3 0 • 1 s a y only good things about my friends behind their backs. 31. When something i s going to affect me I learn as much about i t as I can. 3 2 • 1 put off un t i l tomorrow what I should do today. 3 3 • 1 decide to do things on the spur of the moment. 3 4 • 1 take a positive attitude toward myself. 35. For me, knowing I've done something well is i s more important than being praised by someone else. 3 6 • 1 declare everything at customs. 3 7 • 1 l e t other peoples' demands keep me from doing things I want to do. 3 8 • 1 think I have some pretty awful habits. 3 9 • 1 stick to my opinions when someone disagrees with me. 40. On the whole, I am satisfied with myself. 213 1 2 3 4 5 RARELY OCCASIONALLY SOMETIMES FREQUENTLY USUALLY 41. I • do what I feel like doing not what other people think I ought to do. 42. I t e l l the truth. 43. I get discouraged when doing something that takes a long time to achieve results. 44. I am _ _ _ _ _ late for appointments. 45. When part of a group I prefer to let other people make a l l the decisions. 46. I wish I could have more respect for myself. 47. When I have a problem I follow the advice of friends or relatives. 48. I obey t r a f f i c laws even i f I'm unlikely to get caught. 49. I enjoy trying to do d i f f i c u l t tasks more than I enjoy trying to do easy tasks. 50. When I was a child I obeyed my parents. 51. I prefer situations where I can depend on someone else's a b i l i t y rather than just my own. 52. I feel useless at times. 53. Having someone important t e l l me I did a good job i s more important than feeling I've done a good job. 54. When I'm involved in something I try to find out a l l I can about what i s going on even when someone else i s in charge. 55. I am polite to others including my friends and family. 56. I think I am no good at a l l . 57. I have cheated on a test or assignment in any way. 214 PART TWO The task we would like you to complete contains aspects of a proofreading or quality control task. In quality control a product must be matched against a standard. In the task you w i l l complete you w i l l be shown rows of numbers. Your task i s to check the number at the lef t of each row, and then c i r c l e each number in the row that matches the number at the l e f t of the row. For example: 3 4 6 9 3 9 0 4 4 9 9 I 9 2 I 2 6 4 I 8 6 4 I I 7 9 4 3 0 5 2 6 7 6 6 2 5 9 A 0 Please work carefully, only correctly completed rows count. Try the following rows yourself: 4 3 9 9 7 2 2 2 2 0 9 7 1 5 0 0 6 4 5 6 8 7 9 1 0 2 A 2 1 6 0 7 8 k k 6 9 6 1 8 9 0 0 5 8 2 0 k 7 1 1 8 7 9 1 7 7 7 3 1 1 k 2 2 0 6 3 5 1 2 6 7 0 8 7 9 9 5 7 6 8 1 8 1 7 k 2 6 0 7 3 8 0 8 7 6 6 5 5 2 6 2 0 2 8 7 6 6 3 0 0 0 7 2 5 6 9 8 8 k 7 6 2 7 9 7 5 6 1 7 0 8 6 3 2 k 8 8 0 7 2 3 7 6 2 2 2 3 6 0 8 6 8 k 6 3 7 9 3 1 6 7 8 7 6 0 3 8 6 5 8 5 5 7 1 9 1 9 8 8 0 0 6 6 6 9 0 1 1 6 5 7 9 5 9 5 8 7 6 5 5 2 9 3 Next you w i l l be asked to work at this task for a series of 10 work periods. The periods w i l l be a minute long on average, although some may be longer than others. Do your best during each of these periods but remember to work carefully. Only correctly completed roes count. For each period, you w i l l be told when to begin and when to stop working. At the end of each period, draw a line under the last line you completed and count the number of rows you completed during that period. Write the number of rows you completed in the right margin and wait until you are instructed to begin again. When you are told to begin working again turn to the next page and start immediately at the top. Any questions? 215 PART THREE For the next few questions we would like you to estimate the probability of certain outcomes. We would like you to answer in percentages and make sure that your answers add up to 100%. For example, i f asked to estimate the chances of i t raining in Vancouver in the next month on less than 7 days?: between 7 and 13 days?: between 14 and 20 days?: on more than 21 days?: You might answer in the following way: on less than 7 days?: between 7 and 13 days?: between 14 and 20 days?: on more than 21 days?: Note that the answers add up to 100%. Any questions? We would like you to think about the proofreading task. We are not interested in your actual level of effort or your actual performance. Instead we would like you to think about what would happen i f you were to repeat the simulation under similar circumstances. If you were to complete 10 more periods, and you were to expend a HIGH degree of effort, what are the chances that you would complete on average each of the following numbers of rows! between 1 and 10 rows?: between 11 and 20 rows?: between 21 and 30 rows?: between 31 and 40 rows?: more than 40 rows?: 100% 216 If you were to complete 10 more periods, and you were to expend a MEDIUM degree of effort, what are the chances that you would complete on average each of the following numbers of rows: between 1 and 10 rows?: between 11 and 20 rows?: between 21 and 30 rows?: between 31 and 40 rows?: more than 40 rows?: 100% If you were to complete 10 more periods, and you were to expend a LOW degree of effort, what are the chances that you would complete on average each of the following numbers of rows: between 1 and 10 rows?: between 11 and 20 rows?: between 21 and 30 rows?: between 31 and 40 rows?: more than 40 rows?: 100% 217 On the following scale, indicate your judgement of the amount of control you had over your performance, at 100 i f you had complete control and at 0 i f you had no control. Complete control means that the number of rows you complete i s determined by how hard you try. No control means that how hard you try or don't try has nothing to do with your performance. Another way to look at having no control i s that the number of rows completed in any period i s totally determined by factors such as chance or luck, rather than by the effort you expended. Intermediate control means that your effort has some influence but does not completely determine the number of rows you complete. Circle the number that best represents your judgement: 0 : i0 : 20 : 30 : 40 : 50 : 60 : 70 : 80 : 90 : 100 If you were to complete 10 more periods, for how many of those periods would you complete on average each of the following numbers of rows: Note that your answer should be in numbers of periods and should add up to 10. between 1 and 10 rows?: between 11 and 20 rows?: between 21 and 30 rows?: between 31 and 40 rows?: more than 40 rows?: 10 218 For the following descriptions please use the folowing scale to indicate how attractive each is to you. Circle the number that best describes your feeling. For example: A l l things considered, how attractive would i t be to you ... ...to take part in outdoor a c t i v i t i e s . 1 2 3 4 5 6 7 Very Neutral Very Unattractive Attractive If this i s Very Unattractive to you, c i r c l e the 1. If this i s Somewhat Unattractive to you, c i r c l e the 2, and so on. Any questions? Please begin. A l l things considered, how attractive would i t be to you ... ...to expend a HIGH level of effort on the proofreading task? 1 2 3 4 5 6 7 Very Neutral Very Unattractive Attractive ...to complete between 1 and 10 rows on the proofreading task? 1 2 3 4 5 6 7 Very Neutral Very Unattractive Attractive ...to complete between 31 and 40 rows on the proofreading task? 1 2 3 4 5 6 7 Very Neutral Very Unattractive Attractive 219 Using the following scale, cir c l e the number that best represents the relationship between working hard on the proofreading task and performing well. For example, i f you think there is no relationship between working hard on the proofreading task and performing well, c i r c l e the 0. If you think there i s a very strong relationship, c i r c l e the 9. If you think the relationship i s somewhere in between, c i r c l e the appropriate number. 0 1 2 3 4 5 6 7 8 9 No Strong Relationship Relationship As best you can rec a l l , how many rows did you complete in each period of the proofreading task? How many rows did you complete in the FIRST period?: How many rows did you complete in the SECOND period?: How many rows did you complete in the THIRD period?: How many rows did you complete in the FOURTH period?: How many rows did you complete in the FIFTH period?: How many rows did you complete in the SIXTH period?: How many rows did you complete in the SEVENTH period?: How many rows did you complete in the EIGHTH period?: How many rows did you complete in the NINTH period?: How many rows did you complete in the TENTH period?: 220 For the next set of questions, we are interested in how much you agree with the statements that follow. Using the following scale as a guide, write a number beside each statement to indicate how much you agree with i t . 1 2 3 4 5 Strongly Disagree Neutral Agree Strongly Disagree Agree 1. I feel a great sense of personal satisfaction when I do well. . 2. I expended a high level of effort on the proofreading task. 3. The proofreading task was not very interesting. 4. I found the proofreading task d i f f i c u l t . 5. Generally speaking, I am unsatisfied with my performance on the proofreading task. 6. The proofreading task was challenging. 7. Compared to other people, I don't think I did very well on the proofreading task. 8. My opinion of myself goes up when I do well. 9. Overall, I didn't try very hard on the proofreading task. 10. I enjoyed working on the proofreading task. 11. My own feelings generally are not affected much one way or another by how well I did on the proofreading task. 12. A l l in a l l , I am very satisfied with my performance on the proofreading task. 13. I didn't find the proofreading task very challenging. 14. My performance on the proofreading task was high. 15. I feel bad and unhappy when I've done poorly. 16. I found the proofreading task to be easy. 221 The next set of questions ask you to consider the reasons behind your actual performance on the proofreading task. For each scale, c i r c l e the number that best describes your impression or opinion of the cause of your performance. Is the cause of your performance something that: Reflects an aspect 9 8 7 6 5 4 3 2 1 Reflects an aspect of of yourself the situation Is the cause of your performance: Controllable by 9 8 7 6 5 4 3 2 1 Uncontrollable by you you or other people or other people Is the cause of your performance something that i s : Permanent 9 8 7 6 5 4 3 2 1 Temporary Is the cause of your performance something: Intended by you 9 8 7 6 5 4 3 2 1 Unintended by you or other people or other people Is the cause of your performance something that i s : Outside of you 9 8 7 6 5 4 3 2 1 Inside of you Is the cause of your performance something that i s : Variable over time 9 8 7 6 5 4 3 2 1 Stable over time Is the cause of your performance: Something about you 9 8 7 6 5 4 3 2 1 Something about others Is the cause of your performance something that i s : Changeable 9 8 7 6 5 4 3 2 1 Unchanging Is the cause of your performance something for which: No one is 9 8 7 6 5 4 3 2 1 Someone is responsible responsible 222 indicate no» well J o £ ^ese f -ordf^f f 6 9 3 0 1 1 o £ t h e " o r d s b e l ° » to quickly, d o not spend alot o f t S e on S e ^ o r l 7 0 " ^ m ' W ° r * definitely do not feel 2 do not feel slightly feel definitely feel 1. active _ 2. afraid _ 3. agreeable _ 4. alive _ 5. alone 6. amiable 7. angry 8. awful 9. blue 10. calm 11. cooperative 12. cruel 13. devoted 14. disagreeable 15. discouraged 16. fearful _ 17. fine _ 18. forlorn _ 19. frightened _ 20. gloomy _ 21. happy _ 22. healthy 23. hopeless 24. kindly 25. lonely 26. lost 27. low 28. mad 29. merry 30. miserable 31. nervous 32. panicky _ 33. _ 34. _ 35. _ 36. _ 37. _ 38. _ 39. . 40. . 41. . 42. . 43. 44. 45. 46. 47. 48. polite rejected shaky suffering sunk sympathetic tender tense terrible tormented understanding unhappy upset warm wilted worrying 223 For the following pairs of words, please indicate the point on the scale that describes your current feelings. For each pair circle the number on the scale that best describes your feelings right now. 1 Unhappy 2 3 4 5 6 7 8 9 Happy 1 Relaxed 2 3 4 5 6 7 8 9 Stimulated 1 Pleased 2 3 4 5 6 7 8 9 Annoyed 1 Excited 2 3 4 5 6 7 8 9 Cain 1 Unsatisfied 2 3 4 5 6 7 8 9 Satisfied 1 Sluggish 2 3 4 5 6 7 8 9 Frenzied 1 Contented 2 3 4 5 6 7 8 9 Melancholic 1 Jittery 2 3 4 5 6 7 8 9 Dull 1 Despairing 2 3 4 5 6 7 8 9 Hopeful 1 Sleepy 2 3 4 5 6 7 8 9 Wide Awake 1 Relaxed 2 3 4 5 6 7 8 9 Bored 1 2 3 4 5 6 7 8 9 Aroused Unaroused That's i t . Thanks a lot. 224 APPENDIX B: Musical Selections Elated Tape: (20 minutes and 11 seconds) (1) "Intermezzo", Leopold Stokowski conducts the National Philharmonic Orchestra, Great performances Carmen and L'Arlesienne Suites, CBS, MY 37260. (time: 2:58) (2) "An American in Paris" (Gershwin), Leonard Bernstein conducts the New York Philharmonic Symphony, Columbia Records, M 31804. (first 3:14) (3) "Ode to Joy" (Schiller), Karl Bohm conducts the Weiner Philharmoniker, Beethoven's Symphonie No. _9, Duetsche Grammophon, 2707073, started recording 3:37 into piece, recorded for 2:23. (4) "Guadalcanal March" (Rodgers), Robert Bennett conducts, Victory at sea, RCA, VCS-7064. (time: 2:58) (5) "Le Basque", (Galway), Annie's song and other Galway favorites, RCA, ARL1-3061. (time: 1:50) (6) "Les Torreadors", (Bizet), Carmen Suite, Mercury, MG 50374. (time: 2:14) (7) "Overture", (Conti), Rocky II, United Artists, LA 972-1. (omit f i r s t 0:18, record 1:36, then omit until break at 4:41, resume recording for 1:42). Depressed Tape: (19 minutes and 39 seconds) (1) "Intermezzo" (same as for elated tape) (2) "Egmont Overture", (Beethoven), Josef Krips conducts the London Symphony Orchestra, Everest, 3119. (first 1:16) (3) "A Song to the Evening Star", (Wagner), Young Listener's Library, (L i l l i a n Baldwin, ed.), Sound Book Press Society, Inc., MSB 33103B. (time: 3:15). (4) "Overture-Fantasy" from Romeo and Juliet, (Tchaikovsky), Scheherzade  rhapsodic mood music, Charles Gerhardt conducts, RCA. (fi r s t 2:26) (5) "Introduction" from Scottish Fantasy, (Bruch) Op. 46, Sir Malcolm Sargent conducts the New Symphony Orchestra of London featuring Heiffetz as vi o l i n i s t , RCA, LSC-2603. (f i r s t 2:28) (6) "Sonata No. 7 in D major", Op. 10, No. No.3, second movement, Beethoven's "Piano Sonatas" (Vol. 3), Orpheus, B 118. ( f i r s t 2:37) (7> "Marche Funebre", Sonata No.2 in Bb minor. Op. 35, (Chopin), (50th anniversary complete ed.), Westminister, XWN 18882. ( f i r s t 2:15) 225 (8) "Symphony No.6 in Bm", (Pathetique), Op. 74, 4th movement, Otto Klemperer conducts the Philharmonica Orchestra, Angel, 35787. (last 2:09) Neutral Tape: (19 minutes and 36 seconds) (1) "Intermezzo" same as for the elated and depressed tapes (2) "Canon in D major", (Pachelbel), Jean-Francois Paillard conducts the Jean-Francois Chamber Orchestra, Musical Heritage Society, Inc., 1060. (first 4:08) (3) "Symphonic Variations for Piano and Orchestra", (Franck), Massimo Freccia conducts, RCA. ( f i r s t 3:10) (4) "Othello Overture", (Dworak), Op. 93, Istvan Kertesz conducts the London Symphony Orchestra, London, CS 6527. (omit f i r s t 3:20, record 2:53) (5) "Les Parfums de la Nuit", (DeBussy), Iberia, Lorin Maazel conducts the Cleveland Orchestra, London, CS 7128. (time: 3:48) (6) "The Homecoming", (Hardy), courtesy of WAJY-FM. (time: 2:28) 226 PART TWO The task we would l ike you to complete contains aspects of a proofreading or quality control task. In quality control a product must be matched against a standard. In the task you wi l l complete you wi l l be shown rows of numbers. Your task is to check the number at the left of each row, and then circ le each number in the row that matches the number at the left of the row. For example: 5 3 4 6 9 3 9 0 4 4 9 9 1 9 2 1 2 6 4 1 8 2 4 1 1 7 9 4 3 0 5 2 6 7 6 6 2 5 9 4 0 Please work carefully, only correctly completed rows count, following rows yourself: Try the 4 3 9 9 7 2 2 2 2 0 9 7 1 5 0 0 6 4 5 6 8 7 9 1 4 0 2 4 2 4 1 6 0 7 8 4 4 6 9 6 1 8 9 0 0 5 8 2 0 4 7 1 1 8 7 9 1 7 7 7 3 4 1 1 4 2 2 0 6 3 5 1 2 6 7 4 0 8 7 9 9 5 4 7 6 8 1 8 1 7 4 2 6 0 7 4 3 8 0 8 7 6 6 5 5 2 6 2 0 2 8 7 6 6 3 0 0 0 7 2 5 6 9 8 8 4 7 6 2 7 9 7 5 6 1 7 0 8 6 3 2 4 8 8 0 7 2 3 7 6 2 2 2 3 6 0 8 6 8 4 6 3 7 9 3 1 6 7 8 7 6 0 3 8 6 5 8 5 5 7 7 9 1 9 8 8 0 0 6 6 6 9 0 1 1 6 5 7 9 5 9 5 8 7 6 5 5 2 9 3 Next you w i l l be asked to work at this task for a series of 10 work periods. The periods wi l l be a minute long on average, although some may be longer than others. Do your best during each of these periods but remember to work carefully. Only correctly completed roes count. For each period, you w i l l be told when to begin and when to stop working. At the end of each period, draw a line under the last line you completed and count the number of rows you completed during that period. Write the number of rows you completed in the right margin and wait unti l you are instructed to begin again. When you are told to begin working again turn to the next page and start immediately at the top. Any questions? 227 THE MARKETING GAME: Session One Practice You w i l l play the role of a newly hired brand manager for National Foods. Your job i s to decide how to spend money on promoting your company's product i n your region. You w i l l be given a promotional budget and have to decide how best to divide that budget between the three markets i n your region. The more of the promotional budget you spend i n a market the more p r o f i t you wil earn i n that market. But i n some markets the same amount of promotion earns more p r o f i t . Your goal i s to earn as much p r o f i t i n your region as you can over the next ten periods. So you must decide for each of these periods, how to allocate or divide your budget between the three markets to earn as much p r o f i t as possible. The way you w i l l do this i s by making your a l l o c a t i o n for one period, then you can look at the results of that a l l o c a t i o n before you go on to make the a l l o c a t i o n for the next period. To help you get started i n your new job, you w i l l be shown the "history report" of the previous manager. This report w i l l look l i k e t h i s : H 1 S T 0 R Y D I S P L Jl A L L F I G U R E S PERIOD: 1 2 3 4 5 PROMOTION MARKET 1: 7 10 8 11 9 MARKET 2: 7 10 8 11 9 MARKET 3: 7 10 8 11 9 PROFIT MARKET 1: 221 315 244 332 279 MARKET 2: 97 103 99 102 105 MARKET 3: 443 634 494 686 563 TOTAL PROFIT 761 1052 837 1120 947 Y I N $ 000'S HIT RETURN TO CONTINUE This shows that i n period One the previous manager divided a budget of $21 ( a l l figures are i n thousands) evenly between Market 1, Market 2, and Market 3. This resulted i n a p r o f i t of $221 i n Market 1, $97 i n Market 2, and $443 i n Market 3 for a t o t a l of $761. In period 2 the previous manager divided a budget of $30 evenly among the three markets. 228 The previous manager had a d i f f e r e n t budget i n each pe r i o d . YOUR budget w i l l be EQUAL fo r each period. During t h i s p r a c t i c e session you w i l l have $54 to a l l o c a t e each period to the three markets i n your regio n . You must spend a l l of t h i s budget each period. The p r a c t i c e session w i l l be THREE periods long. Your performance w i l l be evaluated on the basis of the t o t a l p r o f i t you earn over the next three periods. During t h i s p r a c t i c e session and the r e a l session l a t e r , the computer w i l l prompt you each period. I t w i l l show you the h i s t o r y r e p o r t , i n c l u d i n g the de c i s i o n s and r e s u l t s of the previous manager, then i t w i l l ask you to make a d e c i s i o n . The prompt w i l l look l i k e t h i s : YOUR BUDGET IS: $ 54 WHAT IS YOUR ALLOCATION TO MARKET 1, MARKET 2, MARKET 3 FOR PERIOD: 6 ? Enter your decision by entering three numbers on the same l i n e , separated by spaces. For example: ?20 14 20 The computer w i l l then show you the updated h i s t o r y report with the p r o f i t r e s u l t s of your d e c i s i o n and go on to the next p e r i o d . A recent market survey, commissioned by National Foods, has determined that although other market fac t o r s also i n f l u e n c e p r o f i t , the s i n g l e most important d e c i s i o n that you can make i s how to d i v i d e your promotional budget. This survey has a l s o determined that although other f a c t o r s may i n f l u e n c e p r o f i t w i t h i n a per i o d , the periods are independent. There i s no carryover of promotion from one peri o d to the next. Any questions? 230 Remember that the information c o l l e c t e d here w i l l be c o n f i d e n t i a l and used only for the purposes of the study. Age: Sex: Student Number: (for i d e n t i f i c a t i o n purposes only) Please read each of the f o l l o w i n g statements. For each statement i n d i c a t e the degree to which you agree with the statement, using the f o l l o w i n g s c a l e : 1 2 3 4 5 6 7 STRONGLY MODERATELY SLIGHTLY NEUTRAL SLIGHTLY MODERATELY STRONGLY DISAGREE DISAGREE DISAGREE AGREE AGREE AGREE Write the number that describes how much you agree with the statement i n the blank. For example: A. Vancouver i s the f i n e s t c i t y i n North America. If you STRONGLY AGREE with t h i s , w r i t e a 1 i n the blank. If you MODERATELY AGREE with t h i s , w r i t e a 2 i n the blank, and so on. Answer a l l the items. If you have d i f f i c u l t y w i t h one do not leave i t blank, answer as best you can. 1. When I get what I want i t ' s u s u a l l y because I worked hard f o r i t . 2. I always throw my l i t t e r i n t o waste paper baskets on the s t r e e t . 3. Even when I'm f e e l i n g s e l f confident about most t h i n g s , I s t i l l seem to lack the a b i l i t y to c o n t r o l i n t e r p e r s o n a l s i t u a t i o n s . 4. I have received too much change from a cashier and not sa i d anything. 5. By taking an a c t i v e part i n p o l i t i c a l and s o c i a l a f f a i r s we, the people, can c o n t r o l world events. 6. I f e e l that I'm a person of worth, at l e a s t on an equal basis with others. 7. When I make plans I am almost c e r t a i n to make them work. 8. When I hear people t a l k i n g p r i v a t e l y I avoid l i s t e n i n g . 9. I have no tro u b l e making and keeping f r i e n d s . 10. I have taken things that didn't belong to me. 231 1 2 3 4 5 6 7 STRONGLY MODERATELY SLIGHTLY NEUTRAL SLIGHTLY MODERATELY STRONGLY DISAGREE DISAGREE DISAGREE AGREE AGREE AGREE 11. The average citizen can have an influence on government decisions. 12. I feel that I have a number of good qualities. 13. I prefer games involving some luck over games requiring pure s k i l l . 14. I sometimes t e l l lies i f I have to. 15. I'm not good at guiding the course of a conversation with several others. 16. I always keep my promises, no matter how inconvenient i t might be to do so. 17. It is d i f f i c u l t for people to have much control over the things politicians do in office. 18. A l l in a l l , I am inclined to feel that I am a failure. 19. I can learn almost anything i f I set my mind to i t . 20. I have taken a sick-leave from work or school even though I wasn't really sick. 21. I can usually establish a close personal relationship with someone I find sexually attractive. 22. I like to gossip about other people's business. 23. This world is run by the few people in power and there is not much the l i t t l e guy can do about i t . 24. I am able to do things as well as most people. 25. My major accomplishments are entirely due to hard work and intelligence. 26. I have done things that I don't t e l l other people about. 27. When being interviewed I can usually steer the interviewer toward the topics I want to talk about and away from those I wish to avoid. 28. I say only good things about my friends behind their backs. 29. With enough effort we can wipe out poltical corruption. 232 1 2 3 4 5 6 7 STRONGLY MODERATELY SLIGHTLY NEUTRAL SLIGHTLY MODERATELY STRONGLY DISAGREE DISAGREE DISAGREE AGREE AGREE AGREE 30. I feel I do not have much to be proud of . 31. I usually don't make plans because I have a hard time following through on them. 32. I sometimes put off unti l tommorrow what I should do today. 33. If I need help in carrying out a plan of mine, i t ' s usually d i f f i c u l t to get others to help. 34. I always declare everything at customs. 35. One of the major reasons we have wars is because people don't take enough interest in p o l i t i c s . 36. I take a positive attitude toward myself. 37. Competition encourages excellence. 38. I think I have some pretty awful habits. 39. If there's someone I want to meet I can usually arrange i t . 40. I always t e l l the truth. 41. There is very l i t t l e we, as consumers, can do to keep the cost of l iving from going higher. 42. On the whole, I am satisfied with myself. 43. The extent of personal achievement is often determined by chance. 44. I am sometimes late for appointments. 45. I often find i t hard to get my point of view across to others. 46. I always obey t raff ic laws even i f I'm unlikely to get caught. 47. When I look at i t carefully I realize that i t i s impossible to have any really important influence over what polit icians do. 48. I wish I could have more respect for myself. 49. On any sort of exam or competition I l i k e to know how well I do relative to everyone else. 233 1 2 3 4 5 6 7 STRONGLY MODERATELY SLIGHTLY NEUTRAL SLIGHTLY MODERATELY STRONGLY DISAGREE DISAGREE DISAGREE AGREE AGREE AGREE 50. I have never cheated on a test or assignment in any way. 51. In attempting to smooth over a disagreement I usually make i t worse. 52. When I was a child I obeyed my parents. 53. I prefer to concentrate my energy on other things rather than in solving the world's problems. 54. I certainly feel useless at times. 55. Despite my best efforts I have few worthwhile accomplishments. 56. I am always polite to others including my friends and family. 57. I find i t easy to play an important part in most group situations. 58. In the long run we, the voters, are responsible for bad government on a national as well as a local level. 59. At times I think I am no good at a l l . 234 APPENDIX E: Verbal expectancy items, Study Four. 1. If I tried harder on the Marketing Game my performance would improve 2. Whether I do better or worse does not depend on trying harder. 3. When doing the Marketing Game, i f I increase my effort, my performance is likely to go up. 4. There's a strong connection between my level of effort and my level of performance. 5. I'm likely to do as well or better on the Marketing Game even i f I were to reduce my effort. 6. When i t comes to my performance level, i t really doesn't matter much whether I work hard or not. 7. My performance on the Marketing Game would go down i f I were to decrease my effort. 8. My performance on the Marketing Game wouldn't be affected much i f I tried harder. APPENDIX F: Cell means, a l l dependent variables, Study Four. First Manipulation Checks Variable; Anxiety Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev, EDN 6.92 1.25 8.30 2. 56 7 .53 2.18 END 8.07 2.49 8.00 2. 08 7 .23 1.87 DEN 8.53 2.29 6.84 1. 57 7 .84 3.18 DNE 8.30 2.52 6.07 1. 32 6 .30 1.43 NED 7.00 2.04 6.38 1. 93 8 .53 2.14 NDE 6.69 1.70 7.76 2. 58 7 .69 2.39 Entire sample 7.58 2.16 7.23 2. 16 7 .52 2.29 Variable: Depression Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev, EDN 19.84 4.66 27.38 6.14 22.07 7.49 END 21.23 2.31 22.61 6.04 20.69 4.46 DEN 27.15 7.25 17.53 5.09 23.76 6.78 DNE 29.38 9.00 18.00 5.53 16.30 4.40 NED 22.46 5.28 20.30 4.93 27.76 5.94 NDE 23.00 7.64 26.07 8.21 19.53 4.42 Entire sample 23.84 7.06 21.98 7.00 21.69 6.59 Variable: Hostility Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev, EDN 17.15 1.40 15.92 1.44 15.76 1.23 END 17.15 1.34 15.92 2.01 16.53 1.94 DEN 17.00 1.29 16.92 1.11 17.00 1.35 DNE 16.92 1.65 17.61 1.55 16.46 1.45 NED 17.30 1.54 17.23 1.92 16.07 1.65 NDE 16.61 2.14 16.76 1.30 17.07 1.18 Entire sample 17.02 1.55 16.73 1.66 16.48 1.51 Variable: Pleasure Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 36.84 5. 14 28.07 8 .29 36.07 5 .89 END 34.69 6. 93 35.07 7 .17 35.53 6 .60 DEN 32.84 8. 16 37.76 6 .84 34.23 7 .41 DNE 31.00 11. 43 39.38 7 .28 39.61 6 .35 NED 35.76 5. 80 35.61 5 .72 31.00 6 .35 NDE 35.84 9. 41 34.84 9 .47 38.30 7 .59 it i r e sample 34.50 8. 09 35.12 8 .12 35.79 7 .08 Variable: Arousal Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 25.53 11 .20 17.61 6 .25 25.23 7 .57 END 27.07 10 .08 23.00 6 .70 23.84 7 .32 DEN 18.69 6 .70 37.69 6 .57 21.30 7 .30 DNE 16.92 5 .21 27.00 10 .87 33.84 8 .44 NED 23.46 7 .06 27.53 11 .24 20.30 3 .25 NDE 21.92 8 .89 19.69 6 .07 34.46 9 .24 it i r e sample 22.26 8 .92 25.42 10 .35 26.50 9 .16 Variable: Response Latency Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 194.13 45.81 128.61 30.15 109.40 20.26 END 206.38 32.59 125.03 26.83 104.78 22.97 DEN 209.36 46.95 121.81 27.19 106.31 21.30 DNE 191.69 48.18 113.71 16.74 98.99 24.82 NED 204.52 33.47 134.64 18.77 114.92 19.86 NDE 214.81 66.26 126.80 33.81 109.15 29.87 Entire sample 203.48 46.09 125.10 26.22 107.26 23.18 Second Manipulation Checks Variable: Anxiety Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev EDN 8.38 2.10 8.92 1.97 8.23 1.73 END 9.23 2.38 8.69 1.97 8.00 1.41 DEN 8.23 2.04 8.38 2.56 8.07 2.75 DNE 8.30 2.32 6.69 1.18 7.07 1.49 NED 8.38 3.15 8.23 1.96 9.38 2.56 NDE 8.07 1.93 8.38 2.02 7.84 1.95 Entire sample 8.43 2.31 8.21 2.05 8.10 2.09 Variable: Depression Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 21.46 3.77 25.46 4.70 23.00 5.95 END 21.23 4.90 22.69 6.34 20.38 3.50 DEN 23.84 6.86 19.23 4.78 22.84 6.10 DNE 24.07 8.32 18.00 3.76 16.84 3.10 NED 20.53 5.41 21.30 4.23 25.92 5.70 NDE 22.84 6.49 23.84 7.49 19.38 4.17 Entire sample 22.33 6.08 21.75 5.80 21.39 5.58 Variable: Hostility Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev EDN 17.38 2.43 16.76 1.53 16.00 1.47 END 16.30 1.65 15.69 1.60 15.15 1.57 DEN 15.84 1.95 15.46 1.56 16.07 2.06 DNE 16.61 2.69 15.53 1.50 15.23 1.16 NED 15.76 1.36 15.69 .85 16.46 1.39 NDE 15.30 1.60 15.76 1.69 15.46 1.45 Entire sample 16.20 2.05 15.82 1.50 15.73 1.56 Variable: Pleasure Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 30.92 2.28 29.53 4.48 31.15 2.60 END 31.84 3.15 31.46 3.52 32.69 3.25 DEN 31.07 3.68 32.15 3.07 29.23 2.48 DNE 30.07 4.64 32.46 4.57 33.00 2.64 NED 32.61 1.75 31.15 2.99 31.07 2.72 NDE 32.46 3.20 31.53 3.55 31.53 2.10 Entire sample 31.50 3.27 31.38 3.74 31.44 2.85 Variable: Arousal Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 28.23 2.71 27.23 3.26 28.69 3.42 END 29.00 4.67 28.38 4.03 29.38 4.07 DEN 28.69 3.70 30.15 3.15 27.84 3.50 DNE 26.15 2.37 28.23 3.65 30.07 2.36 NED 27.84 3.18 29.38 3.92 28.30 1.60 NDE 27.92 2.66 27.53 2.72 29.92 2.98 Entire sample 27.97 3.33 28.48 3.52 29.03 3.11 Variable: Response Latency Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 131.96 35.59 114.45 23.24 100.62 21.58 END 117.18 27.90 108.59 22.06 90.74 20.94 DEN 118.25 26.03 98.76 22.41 95.08 21.70 DNE 120.48 26.99 103.52 20.87 92.65 17.59 NED 119.36 23.08 114.42 16.00 98.90 17.81 NDE 134.10 38.88 103.34 21.55 92.15 21.08 i t i r e sample 123.55 30.06 107.18 21.29 95.02 19.86 Dependent Variables Variable: Satisfaction with Task Performance Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev EDN 12.84 1.67 13.15 1.72 12 .15 1.67 END 13.61 2.02 12.38 2.66 12 .84 2.40 DEN 13.30 2.46 13.38 2.32 13 .61 2.32 DNE 15.30 1.18 14.23 2.27 13 .38 2.32 NED 13.76 1.58 13.38 1.60 12 .76 3.00 NDE 12.61 2.10 13.53 2.60 13 .15 2.44 Entire sample 13.57 2.02 13.34 2.23 12 .98 2.36 Variable: Internal Work Motivation Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std . Dev EDN 16.92 1.89 16.61 1.75 16.69 1 .70 END 16.23 2.04 16.84 1.86 17.38 2 .43 DEN 16.92 1.11 16.92 1.25 17.07 1 .38 DNE 16.84 2.26 17.00 1.95 16.84 2 .76 NED 16.53 1.19 17.07 1.80 17.07 1 .80 NDE 17.69 1.97 17.92 1.70 17.23 2 .77 Entire sample 16.85 1.79 17.06 1.73 17.05 2 .14 Variable: Task Effort Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. De EDN 7.00 1.15 6.15 1.21 5.76 1.48 END 6.84 1.28 6.38 1.66 5.92 1.49 DEN 7.23 1.09 6.84 1.46 6.30 1.93 DNE 7.46 1.71 6.84 1.72 6.38 1.66 NED 7.23 1.09 7.07 1.11 6.46 1.50 NDE 7.38 1.26 7.15 1.67 6.92 1.38 Entire sample 7.19 1.25 6.74 1.48 6.29 1.57 Variable: Task D i f f i c u l t y and Challenge Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 11.84 3.36 11.07 2.66 11.00 2.67 END 11.76 3.65 10.53 1.85 10.30 1.75 DEN 12.15 2.67 11.23 2.74 11.00 3.18 DNE 11.92 2.98 12.00 2.27 12.15 2.03 NED 12.15 2.57 11.69 2.62 11.92 2.72 NDE 13.07 3.01 12.69 2.32 13.00 1.77 Entire sample 12.15 2.99 11.53 2.45 11.56 2.50 Variable: Task Interest Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 6.76 1.09 5.84 .98 5.92 1.03 END 6.69 1.43 6.38 1.60 6.00 1.52 DEN 6.69 2.01 6.38 1.98 6.00 1.73 DNE 7.38 1.89 6.76 1.53 7.15 1.67 NED 6.92 1.49 7.00 1.77 6.61 1.60 NDE 7.84 1.06 7.46 1.05 7.61 .86 Entire sample 7.05 1.55 6.64 1.57 6.55 1.54 Variable: Verbal Expectancy Session One Order Mean Std. Dev. EDN 28.23 5.05 END 27.61 4.29 DEN 29.53 3.61 DNE 28.92 4.53 NED 27.53 4.35 NDE 30.23 4.76 Entire sample 28.67 4.42 Session Two Session Three Mean Std. Dev. Mean Std. Dev. 26.46 5.02 25.38 5.83 26.84 5.38 24.92 4.49 27.38 4.55 27.38 6.31 25.61 5.33 25.69 7.01 29.30 5.40 27.61 4.53 29.92 3.83 30.38 3.45 27.58 5.03 26.89 5.55 Variable: Internality Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 13 .84 5 .32 14. 61 6.44 14.00 5.61 END 15 .92 3 .75 15. 15 4.91 14.00 3.91 DEN 15 .69 3 .66 16. 61 4.01 15.84 4.98 DNE 16 .61 4 .87 17. 30 3.54 15.23 5.35 NED 16 .84 3 .91 16. 92 4.44 15.53 4.33 NDE 17 .15 4 .29 16. 15 5.44 16.69 3.72 it i r e sample 16 .01 4 .34 16. 12 4.83 15.21 4.65 Variable: Stability Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 11 .69 5 .51 12.53 4.78 10.69 4.30 END 11 .84 3 .82 11.69 3.56 12.23 3.29 DEN 10 .84 2 .91 11.84 4.39 11.76 5.15 DNE 12 .07 3 .42 12.61 3.96 12.38 3.33 NED 12 .30 4 .81 13.30 4.49 13.46 2.75 NDE 13 .84 3 .26 12.69 6.01 12.30 4.64 it i r e sample 12 .10 4 .03 12.44 4.48 12.14 3.95 Variable: Controllability Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. De EDN 17.00 6.37 17.92 4.19 17.53 5.34 END 18.23 3.67 17.15 3.28 16.15 3.10 DEN 17.38 3.59 19.15 2.07 18.30 3.19 DNE 18.23 3.26 17.92 3.49 17.23 3.56 NED 15.92 2.53 16.76 3.56 15.76 3.29 NDE 19.00 3.62 18.23 4.16 19.15 3.46 Entire sample 17.62 4.02 17.85 3.50 17.35 3.80 Variable: Perceived Covariation Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. D€ EDN .17 .16 .21 .23 .22 .22 END .28 .12 .27 .10 .21 .20 DEN .30 .16 .25 .18 .25 .19 DNE .26 .14 .19 .22 .16 .26 NED .23 .15 .18 .17 .19 .16 NDE .26 .08 .20 .18 .28 .13 Entire sample .25 .14 .22 .18 .22 .20 Variable: Perceived Control Order Session One .Session Two Session Three Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 54 .07 17 .91 53.23 24 .67 46 .87 21 .53 END 50 .38 22 .02 43.46 19 .72 43 .84 19 .80 DEN 59 .23 16 .93 54.23 17 .54 55 .84 23 .59 DNE 70 .84 10 .30 58.30 14 .10 51 .53 19 .51 NED 53 .84 20 .22 49.23 22 .53 44 .07 20 .76 NDE 63 .23 16 .60 52.69 23 .68 54 .00 19 .78 it i r e sample 58 .60 18 .47 51.85 20 .55 49 .36 20 .73 Variable: Perceived Correlation Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev EDN 4.84 2. 70 4.92 2.43 4.76 2.31 END 4.92 1. 84 4.38 2.06 4.61 1.85 DEN 5.07 2. 46 4.30 2.46 5.38 2.10 DNE 4.92 2. 39 4.76 1.58 4.46 2.06 NED 5.46 1. 89 4.46 2.10 4.23 1.87 NDE 4.76 1. 87 5.23 2.55 5.76 1.64 Entire sample 5.00 2. 16 4.67 2.17 4.87 1.99 Variable: Performance Recall Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev, EDN 3.34 .53 3.18 1.08 3.37 .46 END 3.10 .85 3.45 .47 3.54 .27 DEN 3.21 .70 3.46 .45 3.71 .30 DNE 3.73 .29 3.63 .31 3.62 .30 NED 3.03 1.12 3.53 .31 3.36 .46 NDE 3.58 .37 3.52 .28 3.66 .27 Entire sample 3.33 .72 3.46 .56 3.54 .37 Variable: Performance Expectation Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. EDN 3.51 .50 3.65 .35 3.43 .41 END 3.43 .95 3.64 .46 3.57 .51 DEN 3.47 .71 3.56 .52 3.76 .31 DNE 3.84 .26 3.72 .29 3.77 .28 NED 3.40 .58 3.35 1.04 3.36 .65 NDE 3.60 .66 3.72 .39 3.73 .24 Entire sample 3.54 .64 3.61 ,56 3.60 .44 Variable: Recall Accuracy Session One Session Two Session Three Order Mean Std. Dev. Mean Std. Dev. Mean Std. Dev, EDN .35 .28 .40 1.06 .17 .32 END .50 .71 .26 .44 .08 .19 DEN .39 .50 .25 .30 .03 .06 DNE .06 .19 .05 .21 .04 .16 NED .72 1.15 .13 .23 .18 .32 NDE .07 .21 .15 .16 .00 .11 Entire sample .35 .63 .21 .50 .08 .22 244 References Allan, L.G., & Jenkins, H.M. (1980) The judgement of contingency and the nature of the response alternative. Canadian Journal of Psychology, 34, 1-11. Alloy, L.B. & Abramson, L.Y. (1979) Judgement of contingency in depressed and nondepressed students: Sadder but wiser? Journal of Experimental Psychology: General, 108, 441-485. Alloy, L.B. Sc Abramson, L.Y. (1982) Learned helplessness, depression, and the ill u s i o n of control. Journal of Personality and Social Psychology, 42, 1114-1126. Alloy, L.B., Abramson, L.Y. St Viscusi, D. (1981) induced mood and the i l l u s i o n of control. Journal of Personality and Social Psychology, 41, 1129-1140. Alloy, L.B., St Tabachnik, N. (1984) Assessment of covariation by humans and animals: The joint influence of prior expectations and current situational information. Psychological Review, 91, 112-149. Angier, R.P. (1927) The conflict theory of emotion. American Journal of Psychology, 39, 390-401. Argyris, C. (1966) Interpersonal barriers to decision making. Harvard Business Review, 44, 84-97. Arnold, M.B. (1960) Emotion and personality (Volumes 1 and 2). New York: Columbia University Press. Arnold, M.B. (1970) Perennial problems in the f i e l d of emotion. In M.B. Arnold (Ed.), Feelings and Emotion (pp.169-186). New York: Academic Press. 245 Averi l l , J.R. (1980) A constructionist view of emotion. In R. Plutchik & H. Kellerman (Eds.), Emotion: Theory, research and experience (Vol. 1). New York: Academic Press. Bandura, A. (1977) Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191-215. Bateman, T.S. & Organ, D.W. (1983) Job satisfaction and the good soldier: The relationship between affect and employee "citizenship". Academy of Management Journal, 26, 587-595. Batson, CD., Coke, J.S., Chard, F., Smith D. & Taliaferro A. (1979) Generality of the "Glow of Goodwill": Effects of mood on helping and information acquisition. Social Psychology Quarterly, 42, 176-179. Beck, A.T. (1967) Depression: C l i n i c a l , experimental and theoretical aspects. New York: Harper & Row. Berscheid, E. (1983) Emotion. In H.H Kelley, E. Berscheid, A. Christensen, J. Harvey, T. Huston, G. Levinger, E. McClintock, A. Peplau, & D.R. Peterson (Eds.), Close Relationships (pp. 110-168). San Francisco, Ca l i f . : Freeman. Blaney, P. H. (1986) Affect and memory: A Review. Psychological Bulletin,99, 229-246. Boggiano, A.K. & Hertel, P.T. (1983) Bonuses and bribes: Mood effects in memory. Social Cognition, 2, 49-61. Borgatta, E. I. (1961) Mood, personality and interaction. Journal of General Psychology, 64, 105-137. 246 Bower, G.H. (1981) Mood and memory. American Psychologist, 36, 129-148. Bower, G.H., Gilligan, S.G. & Montiero, K.P. (1981) Selectivity of learning caused by affective states. Journal of Experimental Psychology: General, 110, 451-473. Bower, G.H., Montiero, K.P. & Gilligan, S.G. (1978) Emotional mood as a context for learning and re c a l l . Journal of Verbal Learning and Verbal Behaviour, 17, 573-585. Bower, G.H. (1985) Review of research on mood and memory. Paper presented at the symposium on Affect and Cognition at the meeting of the Cognitive Psychology section of the British Psychology Association, Oxford. Bradley, G.W. (1978) Self-serving biases in the attribution process: A reexamination of the fact or fiction question. Journal of Personality and Social Psychology, 36, 56-71. Brief, A. P. & Motowidlo, S. J. (1986) Prosocial organizational behaviors. Academy of Management Review, 11, 710-725. Brockner, J. (1979) The effects of self-esteem, success—failure, and self-consciousness on task performance. Journal of Personality and Social Psychology, 37, 1732-1741. Brockner, J. (1983) Low self-esteem and behavioral p l a s t i c i t y : Some implications. In L. Wheeler & P.R. Shaver (Eds.) Review of Personality and Social Psychology (Vol. 4.). Beverly H i l l s : Sage. Brockner, J., & Guare, J. (1983) Improving the performance of low-self-esteem individuals: An attributional approach. Academy of Management Journal, 26, 642-656. 247 Brockner, J., Hjelle, L., & Plant, R.W. (1985) Self-focused attention, self-esteem, and the experience of state depression. Journal of Personality, 53, 425-434. Brown, J. (1984) Effects of induced mood on causal attributions for success and failure. Motivation and Emotion, 8, 343-353. Buchward, A.M., Strack, S. & Coyne, J.C. (1981) Demand characteristics and the Velten mood induction procedure. Journal of Consulting and Clin i c a l Psychology, 49, 478-479. Campbell, J.P. & Fairey, P.J. (1985) Effects of self-esteem, hypothetical explanations, and verbalization of expectancies on future performance. Journal of Personality and Social Psychology, 48, 1097-1011. Campbell, J.P. & Pritchard, R.D. (1976) Motivation theory in industrial and organizational psychology. In M.D. Dunnette (Ed.), Handbook of Industrial and Organizational Psychology (pp. 62-130). Chicago: Rand McNally. Candland, D.K. (1977) The persistent problems of emotion. In D.K. Candland, J.P. F e l l , E. Keen, A.I. Leshner, R. Plutchik & R.M. Tarpy (Eds.), Emotion (pp.1-84). Monterey, C a l i f . : Brooks/Cole. Cannon, W.B. (1927) The James-Lange theory of emotions: a c r i t i c a l examination and an alternative theory. American Journal of Psychology, 39, 106-124. Cialdini, R.B., Darby, B., & Vincent, J. (1973) Transgression and altruism: A case for hedonism. Journal of Experimental Social Psychology, 9, 502-516. 248 Cialdini, R.B. & Kenrick, D.T. (1976) Altruism as hedonism: A social development perspective on the relationship of negative mood state and helping. Journal of Personality and Social Psychology, 34, 907-914. Clark, M.S. (1981) Enhancing the link between feeling states and  judgements through arousal. Unpublished manuscript, Carnegie-Mellon University. Clark, M.S. (1982) A role for arousal in the link between feeling states, judgments, and behavior. In M.S. Clark & S.T. Fiske (Eds.) Affect and Cognition, London: Erlbaum. Clark, D.M. (1983) On the induction of depressed mood in the laboratory: Evaluation and comparison of the Velten and Musical procedures. Advances in Behaviour Research Therapy, 5, 27-49. Clyde, D.J. (1963) Manual for the Clyde Mood Scale. Coral Gables, Fla.: Biometric Laboratory, University of Florida. Cofer, C.N. (1972) Motivation and emotion. Glenview, I l l i n o i s : Scott, Foresman. Coopersmith, S. (1967) The Antecedents of Self-Esteem. San Francisco: Freeman. Cyert, R.M., & March, J.G. (1963) A behavioral theory of the firm. Englewood C l i f f s , N.J.: Prentice-Hall. Darwin, C. (1872) Expression of emotion in man and animals. New York: Appleton. Demo, D.H. (1985) The measurement of self-esteem: Refining our methods. Journal of Personality and Social Psychology. 48, 1490-1502. 249 Denison, D.R. & Sutton, R.I. (in press) Using emotion and meaning in organizational diagnosis: Lessons form a team of operating room nurses. In J.R. Hackman (Ed.), Groups that Work. San Francisco: Jossey-Bass. Dewey, J. (1894) The theory of emotions: I Emotional attitudes. Psychological Review, 1, 553-569. Dewey, J. (1895) The theory of emotions: II The significance of emotions. Psychological Review, 2, 13-32. Dixon, W.J. (Ed.) (1981) BMDP Sta t i s t i c a l Software. Berkeley: University of California Press. Easterbrook, J.A. (1959) The effect of emotion on cue utilization and the organization of behaviour. Psychological Review, 66, 183-201. Eich, E. & Metcalfe, J. (in press) Mood dependent memory for internal vs. external events. Journal of Experimental Psychology: Learning, Memory & Cognition. Ekman, P. (1984) Expression and the nature of emotion. In K.R. Scherer & P. Ekman (Eds.) Approaches to emotion. Hillsdale, N.J.: Erlbaum. Feather, N.T. (1966) Effects of prior success and failure on expectations of success and subsequence performance. Journal of  Personality and Social Psychology, 3, 287-298. Feather, N.T. (1984a) Expectancy-value approaches: Present status and future directions. In N.T. Feather (Ed.), Expectations and  Actions: Expectancy-Value Models in Psychology (pp. 395-420). Hillsdale, N.J: Erlbaum. 250 Feather, N.T. (1984b) Actions in relation to expected consequences: An overview of a research program. In N.T. Feather (Ed.), Expectations and Actions: Expectancy-Value Models in Psychology (pp. 53-96). Hillsdale, N.J: Erlbaum. F e l l , J.P. (1977) The phenomenological approach to emotion. In D.K. Candland, J.P. F e l l , E. Keen, A.I. Leshner, R. Plutchik & R.M. Tarpy (Eds.), Emotion (pp.253-285). Monterey, Calif.: Brooks/Cole. Forest, D., Clark, M., Mills, J. & Isen, A.M. (1979) Helping as a function of feeling state and nature of the helping behaviour. Motivation and Emotion, 3, 161-169. Fraisse, P. (1968) Les Emotions. Traite de Psychologie Experimentale, 5, Paris: Presses Universitaires. Ganster, D.C., Hennessey, H.W., Luthans, F. (1983) Social Desirability Response effects: Three Alternative Models. Academy of Management Journal, 26, 321-331. Garland, H. (1984) Relation of effort-performance expectancy to performance in goal-setting experiments. Journal of Applied Psychology, 69, 79-84. Goffman, E. (1969) Strategic Interaction. Philadelphia: University of Pennsylvania Press. Goodwin, A.W. & Williams, J.M.G. (1982) Mood induction research: Its implications for c l i n i c a l depression. Behaviour Research and Therapy, 20, 373-382. Guttman, L. (1945) A basis for analyzing test-retest r e l i a b i l i t y . Psychometrika, 10, 255-282. 251 Guzzo, R.A. & Waters, J.A. (1982) The expression of affect and the performance of decision-making groups. Journal of Applied  Psychology, 6 7 , 6 7 - 7 4 . Hackman, J.R., & Oldham, G.R. (1980) Work Design. Reading, Mass.: Addison-Wesley. Hakstian, A.R., Roed, J.C., Si Lind, J.C. (1979) Two-sample T 2 procedure and the assumption of homogeneous covariance matrices. Psychological Bulletin, 8 6 , 1 2 5 5 - 1 2 6 3 . Hilgard, E.R. (1980) The trilogy of mind: Cognition, affection and conation. Journal of the History of the Behavioural Sciences, 16, 1 0 7 - 1 1 7 . Hillman, J. ( 1961) Emotion: A comprehensive phenomenology of theories  and their meanings for therap. Evanston 1 1 1 . : Northwestern University Press. Hochschild, A. (1979) Emotion work, feeling rules and social structure. performance. American Journal of Sociology, 8 5 , 5 5 1 - 5 7 5 . Hochschild, A. R. ( 1983) The Managed Heart: Commercialization of Human Feeling. University of California Press: Berkeley. Hollenbeck, J. (1984) A matrix method for expectancy research. Academy of Management Review, 4 , 5 7 9 - 5 8 7 . Hunt, W.A. (1941) Recent developments in the f i e l d of emotion. Psychological Bulletin, 3 8 , 2 4 9 - 2 7 6 . Ilgen, D.R., Nebecker, D.B., Si Pritchard, R.D. (1981) Expectancy theory measures: An empirical comparison in an experimental simulation. Organizational Behavior and Human Performance, 2 8 , 1 8 9 - 2 2 3 . 252 Isen, A.M. (1970) Success, failure, attention and reactions to others: The warm glow of success. Journal of Personality and Social  Psychology, 15, 297-301. Isen, A.M. (1984) Toward understanding the role of affect in cognition. In R.S. Wyer Jr. and T.K. Skrull (Eds.), Handbook of Social  Cognition (Vol. 3) (pp. 179-236). Hillside, N.J.: Erlbaum. Isen, A.M., & Daubman, K.A. (1984) The influence of affect on categorization. Journal of Personality and Social Psychology, 47, 1206-1217. Isen, A.M., Daubman, K.A. & Gorgoglione, J.M. (In press) The influence of positive affect on cognitive organization. In R. Snow & M. Farr (Eds.), Aptitude, Learning and Instruction: Affective and Conative Processes. Hillsdale, N.J.: Erlbaum. Isen, A.M., & Gorgoglione, J.M. (1983) Some specific effects of four affect induction procedures. Personality and Social Psychology . Bulletin, 9, 136-143. Isen, A.M., Horn, N., & Rosenhan, D.L. (1973) Effects of success and failure on children's generosity. Journal of Personality and Social Psychology, 27, 239-247. Isen, A.M. & Means, B. (1983) The influence of positive affect on decision making strategy. Social Cognition, 2, 18-31. Isen, A.M., Means, B., Patrick, R. & Nowicki, G. (1982) Positive affect and decision making. In M.S. Clark & S. Fiske (Eds.), Affect and Cognition. Hillsdale, N.J.: Erlbaum. 253 Isen, A.M. & Patrick, R. (1983) The effects of positive feelings on risk-taking: When the chips are down. Organizational Behavior and  Human Performance, 31, 194-202. Isen, A.M., Shalker, T., Clark, M. & Karp, L. (1978) Affect, accessibility of material in memory and behaviour: A cognitive loop? Journal of Personality and Social Psychology, 36, 1-12. Isen, A.M. & Simmonds, S.F. (1978) The effect of feeling food on a helping that i s incompatible with good mood. Social Psychology, 41, 345-349. Izard, C E . (1971) The face of emotion. New York: Plenum. Izard, CE. (1977) Human Emotions. New York: Plenum. James, W. (1884) What is emotion? Mind, 9, 188-205. Jenkins, H.M., & Ward, W.C (1965) Judgement of contingency between responses and outcomes. Psychological Monographs, 79, (1, Whole No. 594). Jennings, D.L., Amabile, T.M., & Ross, L. (1982) Informal covariation assessment: Data-based versus theory-based judgements. In D. Kahneman, P. Slovic, & A. Tversky (Eds.), Judgement under uncertainty: Heuristics and biases (pp. 211-230). Cambridge: Cambridge University Press. Kahn, W.A. (1986) Towards a sense of organizational humor: Implications for organizational diagnosis. Paper presented at the annual meeting of the Academy of Management, Chicago. Kavanagh, D.J., & Bower, G.H. (1985) Mood and self-efficacy : Impact of joy and sadness on perceived capabilities. Cognitive Therapy and Research, 9, 507-525. 254 Kemper, T.D. (1978) A Social Interactional Theory of Emotions. Wiley: New York. Kemper, T.D. (1984) Power, status, and emotions: A Sociological contribution to psychophysiological domain. In K.R. Scherer & P. Ekman (Eds.), Approaches to emotion (pp. 369-384), Hillsdale, N.J.: Erlbaum. Kennedy, C.W., Possum, J.A. & White, B.J. (1983) An empirical comparison of within-subjects and between-subjects expectancy theory models. Organizational Behaviour and Human Performance, 32, 124-143. Kirk, R.E. (1968) Experimental Design: Procedures for the Behaviour Sciences. Belmont, California: Brooks Cole. Kleinginna, P.R., & Kleinginna, A.M. (1981) A categorized l i s t of emotion definitions, with suggestions for a consensual definition. Motivation and Emotion, 5, 345-379. Lange, C. (1885) Om Sindsbevaegelser. Copenhagan. Lazarus, R.S. (1982) Thoughts on the relations between emotion and cognition. American Psychologist, 37, 1019-1024. Leathers, D.G., & Emigh, T.H. (1980) Decodong facia l expressions: A new test with decoding norms. Quarterly Journal of Speech, 66, 418-436. Leeper, M.R. (1970) The motivational and perceptual properties of emotions indicating their fundamental character and role. In M.B. Arnold (Ed.), Feelings and Emotions (pp.151-168). New York: Academic Press. 255 Leventhal, H. (1979) A perceptual-motor processing model of emotion. In P.Pliner, K. Blankstein, & I.M. Spigel (Eds.), Advances in the Study of Communication and Affect (Vol. 5). New York: Plenum. Locke, E.A. (1976) The nature and causes of job satisfaction. In M. Dunnette (Ed.) Handbook of Industrial and Organizational Psychology. Skokie, 111.: Rand McNally. Maier, N.R.F. (1963) Problem solving discussions and conference: Leadership methods and s k i l l s . New York: McGraw-Hill. Maier, N.R.F. (1970) Problem solving and creativity: In individuals and groups. Monterey, C a l i f . : Brooks/Cole. Mandler, G. (1984) Mind and Body. New York: Norton. Mardia, K.V. (1971) The effect of nonnormality on some multivariate tests and robustness to nonnormality in the linear model. Biometrika, 58, 105-121. Maslach, C. (1978) How people cope. Public Welfare, Spring, 56-58. Mclntyre, S.H. (1979) The leverage impact of judgement based marketing models. Unpublished Doctoral Dissertation, Stanford University Graduate School of Business, Stanford, California. Mclntyre, S.H. (1982) Experimental study of the impact of judgement based marketing models. Management Science, 28, 17-33. McNair, D.N. & Lorr, M. (1964) An analysis of mood in neurotics. Journal of Abnormal and Social Psychology, 69, 620-627. Meddis, R. (1972) Bipolar factors in mood adjective checklists. British Journal of Social and Cli n i c a l Psychology, 11, 178-184. 256 Mednick, S.A. (1962) The associative basis of creative process. Psychological Review, 69, 220-232. Mednick, M.T. Mednick, S.A. S. Mednick, E.V. (1964) Incubation of creative performance and specific associative priming. Journal of Abnormal and Social Psychology, 69, 84-88. Meyer, L.B. (1956) Emotion and Meaning in Music. Chicago: University of Chicago Press. Mischel, W., Coates, B. & Raskoff, A. (1968) Effects of success and failure on self-gratification. Journal of Personality and Social Psychology, 10, 381-390. Mischel, W., Ebbesen, E. & Zeiss, A. (1973) Selective attention to the self: Situational and dispositional determinants. Journal of Personality and Social Psychology, 27, 129-142. Mischel, W., Ebbesen, E. & Zeiss, A. (1976) Determinants of selective memory about the self. Journal of Consulting and C l i n i c a l Psychology, 44, 92-103. Mitchell, T.R. (1974) Expectancy models of job satisfaction, occupational preference and effort: A theoretical, methodological, and empirical appraisal. Psychological Bulletin, 81, 1053-1077. Mitchell, T.R. (1982) Motivation: New directions for theory, research, and practice. Academy of Management Review, 7, 80-88. Motowidlo, S.J. (1984) Does job satisfaction lead to consideration and personal sensitivity? Academy of Management Journal, 27, 910-915. Motowidlo, S.J. & Lawton, G.W. (1984) Affective and cognitive factors in soldier's reenlistment decisions. Journal of Applied Psychology , 69, 157-166. 257 Motowidlo, S.J., Packard, J.S. & Manning, M.R. (1985) Occupational stress: Its causes and consequences for job performance. Unpublished manuscript, Department of Organizational Behavior, Pennsylvania State University. Muchinsky, P.M. (1977) A comparison of within- and across-subjects analyses of the expectancy-valence model for predicting effort. Academy of Management Journal, 20, 154-158. Nasby, W. & Yando, R. (1982) Selective encoding and retrieval of affectively valent information: Two cognitive consequences of children's mood states. Journal of Personality and Social Psychology, 43, 1244-1253. Natale, M. & Hantas, M. (1982) Effect of temporary mood states on selective memory about the self. Journal of Personality and Social Psychology, 42, 927-934. Nelson, R.E. & Craighead, W.E. (1977) Selective recall of positive and negative feedback, self-control behaviours, and depression. Journal of Abnormal Psychology, 86, 379-388. Norman, D.A. (1980) Twelve issues for Cognitive Science. Cognitive Science, 4, 1-32. Nowlis, V. (1965) Research with the mood adjective checklist. In S.S. Tomkins & C E . Izard (Eds.), Affect, cognition, and personality. New York: Springer. Nunnally, J.C (1978) Psychometric Theory. New York: McGraw-Hill. Park, O.S., Sims, H.P., Jr., & Motowidlo, S.J. (1984) Mood and cognition: Variations and causal relationships in the achievement context'. Unpublished manuscript, Department of Organizational Behavior, Pennsylvania State University. 258 Park, O.S., Sims, H.P., & Motowildo, S.L. (1986) Affect in organizations: How feelings and emotions influence managerial judgement. In H.P. Sims & D.A. Gioia (Eds.), The Thinking  Organization (pp. 215-237). San Fransisco: Jossey-Bass. Paulhan, F. (1930) The Laws of Feeling. (Translated by C.K. Ogden.) London: Kegan Paul, Trench, Trubner. Paulhus, D.L. (1984) Two-component models of socially desirable responding. Journal of Personality and Social Psychology, 46, 598-609. Paulhus, D.L. (1987) Personal communication. Pignatiello, M.F. Camp, C.J., & Rasar, L.A. (1986) Musical mood induction: An alternative to the Velten technique. Journal of  Abnormal Psychology, 95, 295-297. Pinder, C.C., & Moore, L.F. (1979) The resurrection of taxonomy to aid the development of middle range theories of organizational behavior. Administrative Science Quarterly, 24, 99-118. Pinder, CC. & Moore, L.F. (Eds.) (1980) Middle Range Theory and The  Study of Organizations. Boston: Martinus Nijhoff. Plutchik, R. (1962) The Emotions: Facts, Theories and a New Model. New York: Random House. Plutchik, R. (1977) Cognitions in the service of emotions: An evolutionary perspective. In D.K. Candland, J.P. F e l l , E. Keen, A.I. Leshner, R. Plutchik & R.M. Tarpy (Eds.), Emotion (pp. 189-210). Monterey CA: Brooks Cole. 259 Plutchik, R. (1980) Emotion: A Psychoevolutionary Synthesis. New York: Harper & Row. Polivy, J. & Doyle, C. (1980) Laboratory induction of mood states through the reading of self-referent mood statements: Affective changes or demand characteristics. Journal of Abnormal Psychology, 89, 286-290. Porac, J.F., Nottenburg, G. St Eggert, J. (1981) On extending Weiner 1s attributional model to organizational contexts. Journal of Applied Psychology, 66, 124-126. Rafaeli, A. & Sutton, R.J. (1987) The expression of emotion as part of the work role. Academy of Management Review, 12, 23-37. Richards, C. (1981) The use of music to induce mood. Unpublished 3rd year undergraduate project. Dept. of Experimental Psychology, University of Oxford. Robinson, J.P. St Shaver, P.R. (1973) Measures of social psychological attitudes. Survey Research Center Institute for Social Research. Roseman, I.J. (1984) Cognitive determinants of emotion: A structural theory. In P. Shaver (Ed.) Review of personality and social psychology (Vol. 5). (pp. 11-36). Beverly H i l l s : Sage. Rosenberg, M. (1965) Society and the adolescent self-image. Princeton, N.J.: Princeton University Press. Russell, D. (1982) The causal dimension scale: A measure of how individuals perceive causes. Journal of Personality and Social Psychology, 42, 1137-1145. 260 Russell, J.A. (1979) Affective space is bipolar. Journal of Personality and Social Psychology, 37, 345-356. Russell, J.A. (1980) A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161-1178. Russell, J.A., & Mehrebian, A. (1977) Evidence for a three factor theory of emotions. Journal of Research in Personality, 11, 273-294. Schachter, S. (1971) Emotion, Obesity and Crime. New York: Academic Press. Schachter, S., Singer, J.E. (1962) Cognitive, social and physiological determinants of emotional state. Psychological Review, 69, 379-399. Scherer, K. R. & Tannenbaum, P. H. (1980) Emotional experiences in everyday l i f e : A survey approach. Motivation and Emotion, 10, 295-314. Shaklee, H., & Tucker, D. (1980) A role analysis of judgements of covariation between events. Memory & Cognition, 8, 459-467. Shatin, L. (1970) Alteration of mood via music: A study of the vectoring effect. Journal Psychology, 75, 81-86. Shrauger, J.S. (1972) Self-esteem and reactions to being observed by others. Journal of Personality and Social Psychology, 23, 192-200. Silber, E. & Tippett, Jean. (1965) Self-esteem: C l i n i c a l assessment and measurement validation. Psychological Reports, 16, 1017-1071. Simon, H. A. (1967) Motivational and emotional controls of cognition. Psychological Review, 74, 29-39. 261 Sinclair, R.C (in press) Mood, categorization breadth, and performance appraisal: effects of information acquisition and affective state on halo, accuracy, information retrieval and evaluations. Personality and Social Psychology Bulletin. Snyder, R.A., Howard, A., St Hammer, T.H. (1978) The predictive power of within- versus across-subjects scores in expectancy research. Journal of Psychology, 100, 285-292. Staw, B.M. (1984) Organizational behaviour: A review and reformulation of the field's outcome variables. Annual Review of Psychology, 35, 627-666. Strongman, K.T. (1978) The psychology of emotion (2nd Ed.). Chichester: Wiley. Sutherland, G., Newman, B., & Rachman, S. (1982) Experimental investigations of the relations between mood and intrusive, unwanted cognitions. British Journal of Medical Psychology, 55, 127-138. Sutton, R.I. & Rafaeli, A. (1986) Emotion work and financial performance. Unpublished manuscript. Stanford University. Sutton, L. & Teasdale, J.D. (1982) Unpublished manuscript. Department of Psychiatry, University of Oxford. Tabachnik, B.G., & Fi d e l l , L.S. (1983) Using multivariate s t a t i s t i c s . New York: Harper & Row. Teasdale, J.D. & Fogarty, S.J. (1979) Differential effects of induced mood on retrieval of pleasant and unpleasant events from episodic memory. Journal of Applied Psychology, 88, 248-257. 262 Teasdale, J.D. & Russell, (1983) Differential aspects of induced mood on the recall of positive, negative, and neutral words. British  Journal of Cli n i c a l Psychology, 22, 163-171. Teasdale, J.D. & Spencer, P. (1982) Effects of induced elation-depression on estimates of past successes and subjective probability of future success. Unpublished manuscript, Department of Psychiatry, University of Oxford. Teasdale, J.D. & Spencer, P. (1984) Induced mood and estimates of past success. British Journal of C l i n i c a l Psychology, 23, 149-150. Teasdale, J.D. & Taylor, R. (1981) Induced mood and accessibility of memories: An effect of mood state or of induction procedure? British Journal of Cl i n i c a l Psychology, 20, 39-48. Tomkins, S. (1962) Affect, Imagery, and Consciousness. Vol. I: The  Positive Affects. New York: Springer. Tomkins, S. (1963) Affect, Imagery, and Consciousness. Vol.11: The  Negative Affects. New York: Springer. Tomkins, S. (1980) Affect as amplification: Some modifications in theory. In R. Pultchik and H. Kellerman (Eds.), Theories of  Emotion (Vol.1) (pp. 141-163). New York: Academic Press. Tosi, H.L., Rizzo, J.R., & Carroll, S.J. (1986) Managing Organizational  Behavior. Marshfield, MA.: Pitman. Tversky, A. & Kahneman, D. (1973) Availab i l i t y : A heurisitic for judging frequency and probability. Cognitive Psychology, 5, 207-232. 263 Van Maanen, J. & Kunda, G. (1985) "Real feelings": Organizational time outs and the expressions of emotions. Paper presented at the Annual Meeting of the Academy of Management, San Diego. Veitch, R., & G r i f f i t t , W. (1976) Good news—bad news: Affective and interpersonal effects. Journal of Applied Social Psychology, 6, 69-75. Velten Jr., E. (1968) A laboratory task for induction of mood states. Behavior Research and Therapy, 6, 473-482. Vroom, V.H. (1964) Work and Motivation. New York: Wiley. Wanous, J.P., Keon, T.L. s> Latack, J.C. (1983) Expectancy theory and occupational/organizational choices: A review and test. Organizational Behavior and Human Performance, 32, 66-86. Ward, W.C., & Jenkins, H.M. (1965) The display of information and the judgement of contingency. Canadian Journal of Psychology, 19, 231-241. Wasserman, E.A., Chatlosh, D.L., & Neunaber, D.J. (1983) Perception of causal relations in human factors affecting judgements of response-outcome contingencies under free-operant procedures. Learning and  Motivation, 14, 406-432. Weick, K.E. (1985) The Emotions of Organizing. Paper presented at the Annual Meeting of the Academy of Management, San Diego. Weiner, B. (1979) A theory of motivation for some classroom experiences. Journal of Educational Psychology, 71, 3-25. Weiner, B., Russell, D. & Lerman, D. (1978). Affective consequences of causal ascriptions. In J.H. Harvey, W.J. Ickes & R.F. Kidd (Eds.), New Directions in Attribution Research (Vol. 2). Hillsdale, N.J.: Erlbaum. 264 Weiner, B., Russell, D. & Lerman, D. (1979) The cognition-emotion process in achievement-related contexts. Journal of Personality and Social Psychology, 37, 1211-1220. Weyant, J.M. (1978) Effects of mood states, costs, and benefits of helping. Journal of Personality and Social Psychology, 36, 1169-1176. Wright, J. & Mischel, W. (1982) Influence of affect on cognitive social learning person variables. Journal of Personality and Social Psychology, 43, 901-914. Young, P.T. (1961) Motivation and Emotion. New York: Wiley. Zajonc, R.B. (1980) Feeling and thinking: Preferences need no inferences. American Psychologist, 35, 151-175. Zerbe, W.J., & Paulhus, D.L. (1987) Socially desirable responding in organizational behavior: A reconception. Academy of Management Review, 12, 250-264. Zuckerman, M., & Lubin, B. (1965) Manual for the Multiple Affect Adjective Checklist. San Diego, CA.: Educational Testing Service. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0098296/manifest

Comment

Related Items