Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A circumplex model of affect and its relation to personality : a five-language study Yik, Michelle Siu Mui 1999

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

Item Metadata

Download

Media
831-ubc_1999-390071.pdf [ 10.2MB ]
Metadata
JSON: 831-1.0089251.json
JSON-LD: 831-1.0089251-ld.json
RDF/XML (Pretty): 831-1.0089251-rdf.xml
RDF/JSON: 831-1.0089251-rdf.json
Turtle: 831-1.0089251-turtle.txt
N-Triples: 831-1.0089251-rdf-ntriples.txt
Original Record: 831-1.0089251-source.json
Full Text
831-1.0089251-fulltext.txt
Citation
831-1.0089251.ris

Full Text

A CIRCUMPLEX MODEL OF AFFECT AND ITS RELATION TO PERSONALITY: A FIVE-LANGUAGE STUDY by MICHELLE SIU MUI YIK B.S.Sc, The Chinese University of Hong Kong, 1991 M.Phil., The Chinese University of Hong Kong, 1993 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Department of Psychology We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA December 1998 ©Michelle Siu Mui Yik, 1998 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 f$yc^t l^/ (<p^^j The University of British Columbia Vancouver, Canada D a t e Dec S, W 8 DE-6 (2/88) II Abstract Are there aspects of affect that can be generalized across different languages? Are there consistent patterns of associations between self-reported affect and personality across groups speaking different languages? In the present dissertation, I explore these two questions in five different language samples. Studies of current self-reported affect in English suggest that Russell's (1980), Thayer's (1989), Larsen and Diener's (1992), and Watson and Tellegen's (1985) models of affect variables can be integrated and summarized by a two-dimensional space defined by Pleasant vs Unpleasant and Activated vs Deactivated axes. To assess the cross-language generalizability of this integrated structure, data on translations of the English affect scales (N for Spanish = 233, N for Chinese = 487, N for Japanese = 450, N for Korean = 365) were compared with the structure in English ON = 535). Systematic and random errors were controlled through multi-format measurements (Green, Goldman, & Salovey, 1993) and structural equation modeling. Individual measurement models as defined in English received support in all five languages, although revisions of these scales in non-English samples provided an even closer approximation to the two-dimensional structure in English. In all five languages, the two dimensions explained most, but not all, of the reliable variance in other affect variables (mean = 88%). The four structural models fit comfortably within the integrated two-dimensional space. In fact, the variables fell at different angles on the integrated space, suggesting a new circumplex structure. In prior studies conducted in English, the personality traits of Neuroticism and Extraversion were most predictive of affect and they aligned with the Pleasant Activated and Unpleasant Activated states. To clarify and extend the previous findings, participants in all five samples also completed NEO FFI (Costa & McCrae, 1992), a measure for the Five Factor Model of personality (FFM). Again, Neuroticism and Extraversion were most predictive of affect, Ill accounting for, on average, 10% of the variance. The remaining three factors of the FFM contributed, on average, 2%. In all five languages, the FFM dimensions did not align with the two predicted affective dimensions. Rather, they fell all around the upper half of the two-dimensional space. iv Table of Contents Abstract " List of Tables viii List of Figures xi Acknowledgments xii CHAPTER 1 INTRODUCTION 1 1.1 STRUCTURE OF AFFECT 2 1.1.1 DIMENSIONS OF AFFECT 2 1.1.1.1 Pleasantness 2 1.1.1.2 Activation 3 1.1.1.3 Pleasantness and Activation 3 1.1.2 FOUR STRUCTURAL MODELS OF AFFECT 4 1.1.2.1 Russell's Model 6 1.1.2.2 Watson and Tellegen's Model 6 1.1.2.3 Larsen and Diener's Model 7 1.1.2.4 Thayer's Model 7 1.1.2.5 Interim Summary 7 1.1.3 THE 45° ROTATION HYPOTHESIS 7 1.1.4 STUDIES ON INTEGRATING THE FOUR MODELS IN ENGLISH 11 1.1.5 PROBLEMS NOTED IN PAST RESEARCH 13 1.1.5.1 Measurement Errors 13 1.1.5.2 Time Frames 14 1.2 PREDICTING AFFECT FROM PERSONALITY 14 1.2.1 PAST RESEARCH FINDINGS 15 1.2.2 PROBLEMS NOTED IN PAST RESEARCH 16 1.2.2.1 Restricted Definition of Affect 16 1.2.2.2 Restricted Definition of Personality 17 1.2.2.3 Construct Overlap 17 1.2.2.4 Methodological Overlap 19 1.3 OVERVIEW OF THE PRESENT DISSERTATION 20 1.3.1 A CIRCUMPLEX MODEL OF AFFECT 20 1.3.2 RELATING AFFECT TO PERSONALITY 21 1.3.3 A SAMPLE OF LANGUAGES 21 CHAPTER 2 METHOD 22 2.1 AFFECT MEASURES 22 2.1.1 CMQ SCALES 22 2.1.2 THAYER'S SCALES 22 2.1.3 LARSEN AND DIENER'S SCALES 22 2.1.4 WATSON AND TELLEGEN'S SCALES 23 2.1.5 TRANSLATION 23 2.2 PERSONALITY MEASURES 24 2.3 PARTICIPANTS AND PROCEDURE ...24 2.3.1 THE ENGLISH SAMPLE 25 2.3.2 THE SPANISH SAMPLE 25 2.3.3 THE CHINESE SAMPLE 26 2.3.4 THE JAPANESE SAMPLE 26 2.3.5 THE KOREAN SAMPLE 26 2.4 DATA ANALYSES 26 CHAPTER 3 ENGLISH.. 28 3.1 STRUCTURE OF AFFECT 28 3.1.1 TEST FOR BIPOLARITY 28 3.1.2 MEASUREMENT MODELS 31 V 3.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes 32 3.1.2.2 Thayer's Constructs 36 3.1.2.3 Larsen and Diener's Constructs 37 3.1.2.4 Watson and Tellegen's Constructs 39 3.1.2.5 Interim Summary 41 3.1.3 RELATIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS 41 3.1.4 PREDICTING AFFECT CONSTRUCTS BY PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES 43 3.1.4.1 Thayer's Constructs •••44 3.1.4.2 Larsen and Diener's Constructs 46 3.1.4.3 Watson and Tellegen's Constructs 46 3.1.4.4 Interim Summary 48 3.1.5 THE FULL TWO-DIMENSIONAL SPACE 48 3.2 RELATING AFFECT TO PERSONALITY 50 3.2.1 DESCRIPTIVE STATISTICS FOR FFM 50 3.2.2 PREDICTING AFFECT FROM PERSONALITY 51 3.2.3 SINUSOID FUNCTIONS 54 CHAPTER 4 SPANISH 57 4.1 STRUCTURE OF AFFECT 57 4.1.1 A PRIORI SCORING 57 4.1.1.1 Test for Bipolarity .57 4.1.1.2 Measurement Models 59 4.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes 59 4.1.1.2.2Thayer's Constructs 60 4.1.1.2.3Larsen and Diener's Constructs 61 4.1.1.2.4Watson and Tellegen's Constructs 63 4.1.1.2.5Interim Summary 65 4.1.2 REVISED SCORING .65 4.1.2.1 Bootstrap Method 65 4.1.2.2 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes 67 4.1.2.3 Scales defining the other three structural models 69 4.1.3 RELATIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS 69 4.1.4 PREDICTING AFFECT CONSTRUCTS BY REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES 71 4.1.5 THE FULL TWO-DIMENSIONAL SPACE 73 4.2 RELATING AFFECT TO PERSONALITY 74 4.2.1 DESCRIPTIVE STATISTICS FOR THE FFM 74 4.2.2 PREDICTING AFFECT FROM PERSONALITY 75 4.2.3 SINUSOID FUNCTIONS 77 CHAPTER 5 CHINESE 79 5.1 STRUCTURE OF AFFECT 79 5.1.1 A PRIORI SCORING 79 5.1.1.1 Test for Bipolarity .79 5.1.1.2 Measurement Models 80 5.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes 81 5.1.1.2.2Thayer's Constructs 82 5.1.1.2.3Larsen and Diener's Constructs 83 5.1.1.2.4Watson and Tellegen's Constructs 85 5.1.1.2.5lnterim Summary 86 5.1.2 REVISED SCORING 87 5.1.2.1 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes. . . 87 5.1.2.2 Scales defining the other three structural models 89 5.1.3 RELATIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS 89 5.1.4 PREDICTING AFFECT CONSTRUCTS BY REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES 91 5.1.5 THE FULL TWO-DIMENSIONAL SPACE 93 5.2 RELATING AFFECT TO PERSONALITY 94 5.2.1 DESCRIPTIVE STATISTICS FOR THE FFM 94 5.2.2 PREDICTING AFFECT FROM PERSONALITY 95 5.2.3 SINUSOID FUNCTIONS 97 CHAPTER 6 JAPANESE 99 6.1 STRUCTURE OF AFFECT 99 6.1.1 A PRIORI SCORING 99 6.1.1.1 Test for Bipolarity 99 6.1.1.2 Measurement Models 100 6.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes 100 6.1.1.2.2Thayer's Constructs 102 6.1.1.2.3Larsen and Diener's Constructs 103 6.1.1.2.4Watson and Tellegen's Constructs 105 6.1.1.2.5Interim Summary 106 6.1.2 REVISED SCORING 106 6.1.2.1 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes 107 6.1.2.2 Scales defining the other three structural models 108 6.1.2.3 Relationships Among the Hypothesized Equivalents 109 6.1.3 PREDICTING AFFECT CONSTRUCTS BY THE REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES 111 6.1.4 THE FULL TWO-DIMENSIONAL SPACE 112 6.2 RELATING AFFECT TO PERSONALITY 114 6.2.1 DESCRIPTIVE STATISTICS FOR THE FFM 114 6.2.2 PREDICTING AFFECT FROM PERSONALITY 114 6.2.3 SINUSOID FUNCTIONS 117 CHAPTER 7 KOREAN 118 7.1 STRUCTURE OF AFFECT 118 7.1.1 A PRIORI SCORING 118 7.1.1.1 Test for Bipolarity 118 7.1.1.2 Measurement Models 119 7.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes 120 7.1.1.2.2Thayer's Constructs 121 7.1.1.2.3Larsen and Diener's Constructs 122 7.1.1.2.4Watson and Tellegen's Constructs 124 7.1.1.2.5Interim Summary 125 7.1.2 REVISED SCORING 125 7.1.2.1 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes 126 7.1.2.2 Scales defining the other three structural models 127 7.1.3 RELATIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS 128 7.1.4 PREDICTING AFFECT CONSTRUCTS BY REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES 130 7.1.5 THE FULL TWO-DIMENSIONAL SPACE 132 7.2 RELATING AFFECT TO PERSONALITY 133 7.2.1 DESCRIPTIVE STATISTICS FOR THE FFM 133 7.2.2 PREDICTING AFFECT FROM PERSONALITY 134 7.2.3 SINUSOID FUNCTIONS 137 CHAPTER 8 CROSS-LANGUAGE SIMILARITIES AND DIFFERENCES 138 8.1 THE TWO-DIMENSIONAL SPACE 139 8.1.1 REVISED MODELS 139 VII 8.1.2 A MULTI-SAMPLE ANALYSIS ON THE RELATIONS AMONG PLEASANT, UNPLEASANT, ACTIVATED, AND DEACTIVATED 140 8.2 INTEGRATION HYPOTHESIS 141 8.2.1 A CROSS-LANGUAGE COMPARISON OF VARIANCE EXPLAINED BY TWO BIPOLAR AXES 141 8.2.2 MULTI-SAMPLE ANALYSES ON THE VARIANCE EXPLAINED BY THE TWO BIPOLAR AXES 143 8.2.3 TESTING CIRCUMPLEXITY 146 8.3 THE 45° ROTATION HYPOTHESIS 147 8.4 RELATING AFFECT TO PERSONALITY 149 8.4.1 A CROSS-LANGUAGE EXAMINATION OF THE VARIANCE EXPLAINED BY FFM IN AFFECT 149 8.4.2 THE SINUSOID FUNCTIONS 152 CHAPTER 9 DISCUSSION AND CONCLUSIONS 154 9.1 IMPLICATIONS 155 9.1.1 CAN THE PROPER ROTATION BE DETERMINED BY EXTERNAL CORRELATES? 155 9.1.2 HOW SHALL AFFECT BE DEFINED? 159 9.1.3 WHAT ARE THE STATE-TRAIT RELATIONS? 161 9.2 LIMITATIONS OF THE PRESENT DISSERTATION 163 9.2.1 JUDGMENT SPACE FOR AFFECT 163 9.2.2 IMPOSED-ETIC APPROACH 163 9.2.3 A SELECTION OF LANGUAGES 164 9.2.4 FIT INDICES 165 9.3 CONCLUDING REMARKS 167 Bibliography 169 Appendix I The English version of all the affect items 180 Appendix II The English version of NEO FFI 186 Appendix III. English Questionnaire Packet 189 Appendix IV. Spanish Questionnaire Packet 200 Appendix V. Chinese Questionnaire Packet 216 Appendix VI. Japanese Questionnaire Packet 227 Appendix VII. Korean Questionnaire Packet 238 Appendix VIII. The Revised CMQ Scales 248 VIII List of Tables Table 1. Statistics for a Test of Bipolarity i n the English sample 30 Table 2. Fit Indices for the Measurement Models in the English sample 32 Table 3. CMQ in the English sample: Standardized Factor Loadings and Interfactor Correlations 33 Table 4. Thayer's Unipolar Constructs in the English sample: Standardized Factor Loadings and Interfactor Correlations 36 Table 5. Larsen and Diener's Unipolar Constructs in the English sample: Standardized Factor Loadings and Interfactor Correlations 38 Table 6. Watson and Tellegen's Unipolar Constructs in the English sample: Standardized Factor Loadings and Interfactor Correlations 39 Table 7. Estimated Correlations Among Thayer's, Larsen and Diener's, and Watson and Tellegen's Unipolar Constructs in the English sample 42 Table 8. Unipolar and Bipolar Affect Constructs explained by the Pleasant vs Unpleasant and Activated vs Deactivated Axes in the English sample 47 Table 9. Descriptive Statistics for the Five Factor Model in the English sample 51 Table 10. Fit Indices for the Structural Equation Models in the English sample 52 Table 11. Predicting Bipolar Affect Dimensions from the Five Factor Model in the English sample 54 Table 12. Statistics for a Test of Bipolarity in the Spanish sample 57 Table 13. Fit Indices for the Measurement Models i n the Spanish sample 59 Table 14. CMQ in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations 60 Table 15. Thayer's Unipolar Constructs in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations 61 Table 16. Larsen and Diener's Unipolar Constructs in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations 62 Table 17. Watson and Tellegen's Unipolar Constructs in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations 63 Table 18. Revised CMQ in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations 68 Table 19. Estimated Correlations Among Thayer's, Larsen and Diener's, and Watson and Tellegen's Unipolar Constructs in the Spanish sample 70 Table 20. Unipolar and Bipolar Affect Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Spanish sample 72 Table 21. Descriptive Statistics for the Five Factor Model in the Spanish sample 74 Table 22. Fit Indices for the Structural Equation Models in the Spanish sample 75 Table 23. Predicting Bipolar Affect Dimensions from the Five Factor Model in the Spanish sample 77 Table 24. Statistics for a Test of Bipolarity in the Chinese sample 79 Table 25. Fit Indices for the Measurement Models in the Chinese sample 81 Table 26. CMQ in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 82 Table 27. Thayer's Unipolar Constructs in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 83 Table 28. Larsen and Diener's Unipolar Constructs in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 83 Table 29. Watson and Tellegen's Unipolar Constructs in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 85 Table 30. Revised CMQ i n the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 88 ix Table 31. Estimated Correlations Among Thayer's, Larsen and Diener's, and Watson and Tellegen's Unipolar Constructs in the Chinese sample 90 Table 32. Unipolar and Bipolar Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Chinese sample 91 Table 33. Descriptive Statistics for the Five Factor Model in the Chinese sample 94 Table 34. Fit Indices for the Structural Equation Models in the Chinese sample 95 Table 35. Predicting Bipolar Affect Dimensions from the Five Factor Mode in the Chinese sample 96 Table 36. Statistics for a Test of Bipolarity in the Japanese sample ...99 Table 37. Fit Indices for the Measurement Models in the Japanese sample 101 Table 38. CMQ in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations 102 Table 39. Thayer's Unipolar Constructs in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations 103 Table 40. Larsen and Diener's Unipolar Constructs in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations 104 Table 41. Watson and Tellegen's Unipolar Constructs in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations 105 Table 42. CMQ Revised in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations 107 Table 43. Estimated Correlations Among Thayer's, Larsen and Diener's, and Watson and Tellegen's Unipolar Constructs in the Japanese sample 110 Table 44. Unipolar and Bipolar Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Japanese sample 111 Table 45. Descriptive Statistics for the Five Factor Model in the Japanese sample 114 Table 46. Fit Indices for the Structural Equation Models in the Japanese sample 115 Table 47. Predicting Bipolar Affect Dimensions from the Five Factor Model in the Japanese sample 116 Table 48. Statistics for a Test of Bipolarity in the Korean sample 118 Table 49. Fit Indices for the Measurement Models in the Korean sample 120 Table 50. CMQ in the Korean sample: Standardized Factor Loadings and Interfactor Correlations 121 Table 51. Thayer's Unipolar Constructs in the Korean sample: Standardized Factor Loadings and Interfactor Correlations 122 Table 52. Larsen and Diener's Unipolar Constructs in the Korean sample: Standardized Factor Loadings and Interfactor Correlations 123 Table 53. Watson and Tellegen's Unipolar Constructs in the Korean sample: Standardized Factor Loadings and Interfactor Correlations 124 Table 54. Revised CMQ in the Korean sample: Standardized Factor Loadings and Interfactor Correlations 126 Table 55. Estimated Correlation Among Thayer's, Larsen and Diener's, and Watson and Tellegen's Unipolar Constructs in the Korean sample 129 Table 56. Unipolar and Bipolar Affect Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Korean sample 130 Table 57. Descriptive Statistics for the Five Factor Model in the Korean sample 134 Table 58. Fit Indices for the Structural Equation Models in the Korean sample 135 Table 59. Predicting Bipolar Affect Dimensions from the Five Factor Model in the Korean sample 136 Table 60. RMSEA for the CMQ Constructs in five language samples 139 Table 61. Interfactor Correlations Among Pleasant, Unpleasant, Activated, and Deactivated in a Multi-sample Confirmatory Factor Analysis 141 Table 62. Variance Explained by the two bipolar axes: A Cross-Language Comparison 142 X Table 63. Bipolar Affect Dimensions explained by the Pleasant vs Unpleasant and Activated vs Deactivated Axes in Six Multi-sample Structural Equation Models 145 Table 64. Rank Order Correlations Among the Angular Positions of the 14 constructs in each language 146 Table 65. Eight Pairs of Hypothesized Equivalent Constructs: A Cross-Language Examination 148 Table 66. Predicting Affect from FFM: A Cross-Language Examination 149 Table 67. Percentage of variance accounted for by the Fitted Sinusoid Functions 153 xi List of Figures Figure 1. Four Structural Models of Affect. Structures shown are based on those presented by Russell (1980, p. 1164), Watson and Tellegen (1985, p. 221), Larsen and Diener (1992, p.39), and Thayer (1996, p. 150) 4 Figure 2. The 45° Rotation Hypothesis: 14 Unipolar Affect Constructs within a two-dimensional space 9 Figure 3. Path Diagram showing the structural model for the four CMQ unipolar constructs. For clarity of presentation, only those coefficients significant at .01 were shown. Dotted lines indicate the path there were estimated but the coefficient was not significant at .01 level. Error terms and correlation between error terms were also estimated but are not shown...35 Figure 4. Path Diagram showing the Structural Equation Model for Thayer's Energy Construct. Error terms and correlations between error terms were also estimated but are not shown.45 Figure 5. A Circumplex Representation of 14 Unipolar Affect Constructs in the English sample 49 Figure 6. Curves Relating Affect Constructs to Neuroticism and Extraversion in the English sample 56 Figure 7. A Circumplex Representation of 14 Unipolar Affect Constructs in the Spanish sample 73 Figure 8. Curves Relating Affect Constructs to Neuroticism and Agreeableness in the Spanish sample 78 Figure 9. A Circumplex Representation of 14 Unipolar Affect Constructs in the Chinese sample 93 Figure 10. Curves Relating Affect Constructs to Neuroticism and Openness in the Chinese sample 98 Figure 11. A Circumplex Representation of 14 Unipolar Affect Constructs in the Japanese sample 113 Figure 12. Curves Relating Affect Constructs to Neuroticism and Openness in the Japanese sample 117 Figure 13. A Circumplex Representation of 14 Unipolar Affect Constructs in the Korean sample 133 Figure 14. Curves Relating Affect Constructs to Agreeableness and Openness in the Korean sample 137 Figure 15. Projecting the FFM dimensions onto the Integrated Two-Dimensional Space in five language samples 156 XII Acknowledgments I thank Dr. Eric Eich, Dr. Liisa Galea, Dr. Ishu Ishiyama, Dr. Kerry Jang, Dr. Randy Larsen, Dr. Del Paulhus, and Dr. Jim Steiger, for their perceptive comments on my dissertation. I especially thank my research supervisor, Dr. Jim Russell, for his introduction of such a fascinating dissertation topic and for his unfailing support at all stages of the dissertation project. I thank Dr. Michael Bond, Dr. Jeff McCrae, and Dr. Shalom Schwartz for their advice on carrying out this five-language project. I also thank Dr. Michael Browne, Dr. Lisa Feldman Barrett, and Jeremy Miles for their statistical help. I thank Chang-kyu Ahn (Korea), Michael Bond (Hong Kong), Joey Cheung (Canada), Jose Miguel Fernandez-Dols (Spain), Irene Kim (Canada), Chung-Leung Luk (Hong Kong), Naoto Suzuki (Japan), Yuriko Takahasi (Canada), Catherine Tang (Hong Kong), and Dong-Sheng Zhou (Hong Kong), for translating the affect questionnaires and collecting the data. I thank Paula Castellon, Miguel Imperial, Yuka Mukai, and Melanie Parkin for assisting me at the final stage of my dissertation. I also thank my comrades in the Kenny Building, Anne Cheng, Kimmy Chiu, Tess Gerra, Cecilia Li, Dawn Macaulay, and Monica Mori for their cheers and tears. I thank Coffee Lo, Maggi Yik, John Lo, and Cindy Lo for their support of my study and for taking care of my mother during my stay in Canada. I thank Steven So for his understanding and patience during the ups and downs of my career; and my little princess, Stephanie So, who slept just enough to let me finish my dissertation. Finally, I dedicate this dissertation to my mother, Ms Wai-Wah Tong, who taught me to strive for excellence in my life. 1 CHAPTER 1 INTRODUCTION Happy and sad, excited and bored, and other feeling states (collectively called affect) are common in everyday language. The number of terms describing feeling states are estimated to be 500 in English (Averill, 1975), 700 in Chinese (Boucher, 1979; see also Russell & Yik, 1996), and 500 in Japanese (Hama, Matsuyama, & Lin, 1986). The lexicon in each language is certainly big and the content is important in its own right. Research evidence has shown that affect influences memory (Eich, 1995), cognition (Forgas, 1995), and behavior (Isen, 1993). In order to provide a common framework to measure and compare affect, it is desirable to specify the relations among all these terms. In English, studies have shown that the covariations of the affect terms can be summarized by a two-dimensional space (e.g., Russell, 1978, 1980; Watson & Tellegen, 1985). Further, the personality traits of Neuroticism and Extraversion were found to be most predictive of affect. Unfortunately, studies are mostly restricted to groups speaking English and expanding the research to other languages is much needed to advance our understanding of affect. Which aspects of self-reported affect can be generalized across different languages and which aspects are specific to groups speaking particular languages? The present dissertation focuses on examining the momentary affective ratings for a thin slice of time. Some theorists predict that there are more universal aspects than the culture-specific counterparts (e.g., Ekman, 1978); whereas others predict the reverse pattern (e.g., Averill, 1980; Harre, 1986; Levy, 1984; Potter, 1988; Wierzbicka, 1992). One approach to study this question is by looking at the dimensions underlying the structure of self-reported affect. In the present study, the two-dimensional space found in English is proposed to be generalizable to languages other than English and such a hypothesis is tested in four different languages. If such a space proves effective in organizing self-reported affect of different languages, it surely provides a useful framework for measuring and comparing human affect across languages. Are there any consistent patterns of associations between affect and personality across groups speaking different languages? Close links, in particular those between Neuroticism and 2 negative affective states and those between Extraversion and positive affective states, have been reported in English-speaking samples (Watson & Clark, 1992; Watson, Wiese, Vaidya, & Tellegen, 1998). Relatively few studies, however, have been conducted in non-English speaking samples (Allik & Realo, 1997; Church, Katigbak, Reyes, & Jensen, 1997, are exceptions). More studies are needed to sketch the precise relations between personality and affect in both English and especially, other languages. In this dissertation, I explore two such aspects in five different languages: The structure of affect (the relations among different affective dimensions) and the relations of affect to the enduring characteristics of personality. 1.1 STRUCTURE OF AFFECT 1.1.1 DIMENSIONS OF AFFECT Interrelationships among different types of self-reported affect have attracted a lot of attention in past research. Different models have proposed different dimensions to characterize the relations. What are the major dimensions underlying the structure of affect? Past research has emphasized one dimension over the other or both. 1.1.1.1 Pleasantness Pleasantness has been a prominent dimension in the discussion of affect in the past research. Evaluation (good-bad) was found to be a primary factor in Osgood, May, and Miron's (1975) study of affective meanings in 22 cultural groups. Pleasantness was found in the perception of facial expressions (Russell, Lewicka, & Niit, 1989; Schlosberg, 1954) and self-reported current affect (Russell, 1978, 1979). Weirzbicka's (1992) analyses of emotion concepts cross-culturally showed that good and bad are universal semantic primitives. Based on an analysis of 170 affect-related terms, McConville & Cooper (1992) found 5 first-order factors which were hierarchically related to a second-order hedonic factor. Similar finding was reported by Williams (1989, 1990) who concluded with a good / bad factor (see also Wessman & Ricks, 1966). 3 1.1.1.2 Activation Activation has also been referred to activity and arousal in the psychological literature. It is usually referred to the physiological activity (Schachter & Singer, 1962; Thayer, 1967). In the present study, the focus is on the subjective perception of activation. Thayer (1989) has heralded a program of work on the influence of activation on everyday life from diet to personality to neurochemistry. Beginning with the conceptualization of a single dimension of activation in affect, Thayer's analyses led not to a single activation-deactivation dimension, but to a two-dimensional space defined by Energy vs Tiredness and Tension vs Calmness (Thayer, 1996). 1.1.1.3 Pleasantness and Activation Research has also focused on considering both Pleasantness and Activation simultaneously. One reason is that pleasantness and activation may have interactive effects on cognition and behavior (Macaulay, 1997). Only by considering both dimensions could such effects be understood. Second, our language does not allow us to separate them easily. It is an empirical finding that both pleasant and unpleasant words vary in the degree of activation they denote (Averill, 1975; Bush, 1973; Neufeld, 1975, 1976; Russell, 1978; Whissell, 1981). Some pleasant words imply activation (elated, enthused), others imply deactivation (serene. contented). Some unpleasant words imply activation (upset, distressed), others imply deactivation (sad, depressed). In an exactly parallel fashion, words that denote activation and deactivation also vary in pleasantness. Some activation words are pleasant (thrilled, excited), some unpleasant (tense, jittery). Some deactivation words are pleasant (relaxed, calm), others unpleasant (exhausted, lethargic). As a result, any attempt to isolate one dimension runs the risk of confounding that dimension with the other. The apparent nature of pleasantness depends on what level of activation is assumed, and vice versa. Thus needless controversies can arise when one dimension is emphasized at the expense of the other. In the present study, I pursue this third approach - using Pleasantness and Activation simultaneously to describe the covariations among affect variables. However, I am not claiming 4 that these are the only two dimensions capturing the full complexities of human affect. Rather, I propose that the two-dimensional space defined by these two dimensions provides an efficient way to understand human affect. 1.1.2 FOUR STRUCTURAL MODELS OF AFFECT With a long period of divided opinions on which model represents the best description of affect, psychology is witnessing a near consensus on what the model is. What researchers do not agree on is the major dimensions underlying the self-reported affect. Various two-dimensional models have been proposed to describe the relations among different types of affect. Here I emphasize four and they are shown in Figure 1. Figure 1. Four Structural Models of Affect. Structures shown are based on those presented by Russell (1980, p. 1164), Watson and Tellegen (1985, p. 221), Larsen and Diener (1992, p.39), and Thayer (1996, p. 150). Russell's Model AROUSAL Distress Excitement MISERY Depression Relaxation PLEASURE SLEEP Watson & Tellegen's Model Engagement HIGH HIGH POSITIVE AFFECT Pleasantness Low Positive Affect Negative Affect Disengagement Larsen & Diener's Model Activated Unpleasant UNPLEASANT Unactivated Unpleasant HIGH ACTIVATION LOW ACTIVATION Activated Pleasant PLEASANT Unactivated Pleasant Thayer's Model 6 Tense-Energy Tense-Tiredness TIREDNESS TENSION CALMNESS ENERGY Calm-Energy Calm-Tiredness 1.1.2.1 Russell's Model Russell (1980) proposed a circumplex model of affect, characterized by two orthogonal axes, Pleasure and Arousal. In addition to self-reported affect, the circumplex model has been recovered from judgments of facial expressions, and similarity ratings of emotion terms (Russell, 1978; Russell & Bullock, 1985). Cross-cultural evidence on the model was found in the self-reported affect of Estonian, Greek, Polish, Chinese, Croatian, and Gujaranti (Russell, 1983; Russell, Lewicka, & Niit, 1989). The terminology of the descriptive model was updated in Feldman Barrett and Russell (1998) and scales were designed to measure the four major constructs of the model. 1.1.2.2 Watson and Tellegen's Model Watson & Tellegen (1985) performed an analysis of their own data on affect and a reanalysis of previous published data and concluded that there are two dimensions of affect: Positive Affect and Negative Affect. They accounted for 50% to 75% of common variance in affect-related terms. These two dimensions had been found in different time frames, different languages, and different set of affect-related terms (Almagor & Ben-Porath, 1989; Mayer & Gaschke, 1988; Meyer & Shack, 1989; Zevon & Tellegen, 1982), although evidence has shown 7 that these two dimensions were separated into four unipolar factors (Burke, Brief, George, Roberson, & Webster, 1989). Watson and Tellegen argued that Russell's dimensions are obtainable before rotation and theirs obtainable after rotation. These two pairs of dimensions lie 45° apart. Mayer and Gaschke (1988) proffered evidence on this claim. 1.1.2.3 Larsen and Diener's Model In line with Russell (1980) and Watson and Tellegen (1985), Larsen & Diener (1992) proposed a model defined by Pleasant vs Unpleasant and High Activation vs Low Activation dimensions in another attempt to incorporate findings in the affect literature. In addition to the four cornerstone constructs, Larsen and Diener proposed another four diagonal constructs which represent different combinations of the two dimensions. Research has been conducted on building up the external correlates of the eight constructs (Ketelaar, 1989; Larsen, 1989). 1.1.2.4 Thayer's Model Thayer's (1986) psychometric work on activation led not to a single activation vs deactivation dimension, but to a two-dimensional structure (as seen in Figure 1). Thayer's attempt to conceptualize this structure led to his labeling regions of the resulting structure as "Calm-Energy" and "Tense-Tiredness". A careful examination of the defining items suggests that they refer to pleasant and unpleasant states in general. 1.1.2.5 Interim Summary In earlier work, Yik, Russell, and Feldman Barrett (1998) re-oriented and rotated these four structures to emphasize their similarities over their differences. As seen in Figure 1, the four structural models do not look as different as their labels imply. Still, research is much needed to spell out the relations among these four structural models. In the next section, I turn to the topic on the hypothesized interrelationships among the four models: 1.1.3 THE 45°ROTATION HYPOTHESIS The four structures of Figure 1 have each been proposed as broad general descriptions of momentary affective states. Each has guided research, often in quite different domains. 8 Each has been the basis of specific measures of affect. Integration of these various structures is something to be desired, potentially uniting an impressive and diverse array of research. CD O CD Q_ CO "TO c g w c cu £ 5 co -*—» o i _ -*—' 10 c o O o i I— JO o Q. C Z> CO 'co cu o CL >» X c g "•4—» ro o DL o m cu CM cu D) i i . T3 CU -*—' CD .> O < CO "Q5 CO co Q. °8 Q LU I-O < *3 CU i i CO m c CD E CU U_ Q LU I-| H O < < CO < UJ _l QL z 3 O CD 0) CO •If .2 o)^ g CO tu ^ _ o < 5> t ? - w rr \ - _ oa b o ^ CO c JO tu CO > E> ra o O 00 c ro co ro a> o. c I- __ < co co <q z * -5 2 to ca i c ro E T3 O r-tN Q LU r -| O 55 c CO CO to _u D_ c T3 0 » CO > o a p ro I— co —) Z "•-< £ - 1 2 CO CO cu < -C c LU h- tu a- 2 Z ° « ID c cu co i _ ro T3 cu •4—» CO > o CO cu Q Q CO UJ == O 08 LU 13 Q E ro CQ • c CO E 32 cu Li-CU LL 10 Many writers in affect assume that integrating these four structures is extremely simple. Indeed some writers - including all the authors of these structures - have assumed that these structures are rotational variants of one another (Barlow, 1988; Hutchison et al., 1996; Lang et al. 1992; Mano, 1991). The assumption is captured in Figure 2 by placing all four structures within the same two-dimensional space with 45° between appropriate sets of dimensions. In place of Russell's (1980) labels are labels used by Feldman Barrett and Russell (1998), as proposed by Larsen and Diener (1992). Based in part on the conceptual analysis of Larsen and Diener, Feldman Barrett and Russell offered the specific structure of Figure 2 as a means to integrate elements from various previous accounts of mood, affect, and activation. Figure 2 is simply the Cartesian space formed from the Pleasantness and Activation dimensions. This simple schematic diagram captures the main features of the structures of Figure 1 as well as those offered by Feldman (1995), Green et al. (1993), Lang (1994), and Reisenzein (1994). Indeed, its roots go back to Schlosberg (1952, 1954) and Wundt (1912/1924). Further, the diagram of Figure 2 incorporates not only the dimensions of Pleasantness and Activation but also a circular ordering among the elements (Plutchik, 1980; Russell, 1980; Schlosberg, 1954). As shown in Figure 2, the 45° rotation hypothesis makes extremely strong predictions. Some dimensions are identical, despite quite different labels. For example, on this hypothesis, Thayer's (1989) Energy, Watson and Tellegen's (1985) Positive Affect, and Larsen and Diener's (1992) Activated Pleasant are all identical. Each unipolar construct has another construct that is its bipolar opposite. All individual dimensions are separated by 45° or multiples of 45°. What is more is that any two non-redundant dimensions account for all the substantive variance in all the other dimensions. The 45° rotation hypothesis is often assumed to be correct, based on the schematic similarity among the various structures in Figure 1. Each structure is labeled differently, to reflect what each author believes is fundamental about affective experience, but there is an underlying assumption that all structures are subsumed in the same descriptive space. There is 11 little empirical evidence, however, on how similar the four structures are, and what data exist have not been consistent (Burke, Brief, George, Robertson, & Webster, 1989; Feldman Barrett & Russell, 1998; Hutchison et al., 1996; Mayer & Gaschke, 1988; Nemanick & Munz, 1994; Russell, 1979; Russell & Mehrabian, 1977; Russell, Weiss, & Mendelsohn, 1989). For instance, Mayer and Gaschke (1988) found supportive evidence on the close relation between Russell's (1980) structure and Watson and Tellegen's (1985) structure. On the other hand, Hutchison et al. (1996) examined the equivalence of Watson and Tellegen's system with the pleasure-arousal system and obtained correlations between .09 and .45 in magnitude for variables assumed to be 45° apart (under ideal measurement conditions, two variables separated by 45° correlate .71). They therefore concluded that, "it may be misleading to assume that data obtained by measuring dimensions from one of the models can be used to make inferences about dimensions of the other model." (p. 785). Similarly, Burke et al. (1989) administered Watson and Tellegen's scales of Positive Affect and Negative Affect and what, on the structural hypothesis of Figure 2, would be their hypothesized bipolar opposites. Precisely contrary to that hypothesis, a model with four orthogonal unipolar factors fit the data better than did a model with two bipolar factors. 1.1.4 STUDIES ON INTEGRA TING THE FOUR MODELS IN ENGLISH To examine the pattern of relations among the four structures in Figure 1 and the viability of integrating the four structures within the same two-dimensional space as shown in Figure 2, Russell and his colleagues (Feldman Barrett & Russell, 1998; Yik, Russell, & Feldman Barrett, 1998) marshaled a series of studies testing the integration hypothesis. They demonstrated that the two dimensions - Pleasant vs Unpleasant and Activated vs Deactivated -represent an efficient organizing scheme accounting for the covariations of self-reported affect among the English-speaking people. Furthermore, they showed that all four structures in Figure 1 - Russell's (1980), Thayer's (1996), Watson and Tellegen's (1985), and Larsen and Diener's (1992) models - can be meaningfully integrated within the proposed two-dimensional space. 12 Finally, the constructs of these models were shown to form a circular ordering along the circumference of the two-dimensional space. The preceding findings are consistent with a circumplex model. A circumplex model suggests that the covariations of human feelings can be described by two dimensions and their permutations. Each feeling state is a hybrid of these two dimensions. Affective states fall at any angle on the circumference of the two-dimensional space. In factor analytic terms, the circumplex is not a simple structure and that feeling states are not required to cluster at certain angles. Rather, the feeling states can scatter anywhere on the circumference of the two-dimensional space. A circumplex model does offer strong predictions on the relations between an outside variable, such as a personality dimension of Extraversion, and the feelings states located on the circumference. Suppose Extraversion has a high correlation with happy. It must also be true that Extraversion will also be strongly associated with the affective states adjacent to happy. Further, it is expected that Extraversion will correlate negatively with affective states located opposite to happy, such as unhappy. Hence it follows that a sinusoid function can be fitted to the pattern of associations between any outside variable and the affective states. How well the sinusoid function fits the data speaks directly to the viability of the circumplex model. Evidence supportive of the two-dimensional space are, however, are limited to North American samples. This is not an unusual practice in studies on the structure of affect (Russell, Lewicka, & Niit, 1989; Sjoberg, Svensson, & Persson, 1979; Watson, Clark, & Tellegen, 1984, are exceptions). North American samples share the similar languages and socialization patterns. Thus the robust structure uncovered across different samples might be an artifact of similarity of language or culture. In order to demonstrate the generalizability of the two-dimensional space to integrate other structural models, the first necessary step is to examine the relationships among the four structural models in other language samples. In the present study, such an investigation will be carried out in Spanish, Chinese, Japanese, and Korean, in addition to English. If such integration proves successful in other languages, the two-13 dimensional space will provide a common level field to measure and compare affect across different languages. 1.1.5 PROBLEMS NOTED IN PAST RESEARCH 1.1.5.1 Measurement Errors Various factors have been reported to obscure the observation of bipolar factors in affective ratings. Green and Goldfried (1965) argued that Osgood's (1975) three bipolar factors (Evaluation, Potency, and Activity) of affective meanings in 22 different languages was artifactual because the adjective rating scales were bipolar which was an almost guarantee for the observation of bipolar factors in their results. In response to Green and Goldfried (1965), Bentler (1969) transformed all bipolar adjectives into single adjectives and partialled out the acquiescence bias. Bentler found bipolar factors. Russell (1979) conducted a comprehensive study on the biases in affect ratings, including acquiescence bias, affecting the bipolarity of factors. He observed a two-dimensional space characterized by two bipolar axes when these biases were taken into account. After controlling for the acquiescence bias in their self-rating data on 60 mood adjectives, Lorr, Shi, and Youniss (1989) found two axes which were identical to Russell's (1980). Green, Goldman, & Salovey (1993) proposed a multi-format procedure to control for errors introduced in the process of measurement. Each affect construct was measured with multiple formats and the resulting correlation matrix was submitted to structural equation modeling analysis. In the analysis, both random and systematic (such as errors associated with the same type of response format) errors can be controlled. Green et al. concluded that bipolarity surfaces after controlling for measurement errors: Pleasant affective states were found to be the bipolar opposite of Unpleasant affective states; Activated affective states opposite of Deactivated affective states. These two bipolar dimensions were found to be largely independent of each other. In view of the possible serious effects of measurement errors on obtaining bipolar dimensions in affective ratings, the present study adopts Green et al.'s (1993) multi-format 14 procedure and each affect construct is measured by three different response formats and structural equation modeling technique is used to analyze the data, which controls for systematic and random errors. 1.1.5.2 Time Frames Different time frames, ranging from "right now" to "in general", from state to trait measures, have been used in the past studies. As Russell (1979) noted, when participants are asked to report their affect over an extended period of time (for instance, a month), it is very likely for them to report a few affect-related episodes which are opposite in pleasantness or activation. Hence the results might obscure the appearance of bipolar affect dimensions. In view of the time frame problem which might obscure the observation of bipolar factors, Feldman Barrett and Russell (1998) and Yik et al. (1998) asked participants to provide ratings for the instant moment right before they began completing the questionnaires. Although this time frame assures that participants are rating an instant moment of time and thus gives a better chance to bipolarity, the fact that all participants were staying in the same location and doing similar things (i.e., waiting for the lab session) might reduce the variety of feeling states and thus promote the similarity of their feeling states. The repeatedly found two-dimensional space in the structure of affect might be an artifact of the experimental procedure. The present study is intended to examine the structure of momentary affective ratings sampled at different moments of the day. Participants were randomly assigned to one of the six instructions. In each instruction, they were asked to search their memory to pick a moment around a mealtime of the day before that they had the clearest memory of how they felt. The mealtimes used were "before breakfast," "after breakfast," "before lunch," "after lunch," "before dinner," or "after dinner." This sampling method is intended to control the nuisance variables that might have obscured the appearance of bipolar dimensions. 1.2 PREDICTING AFFECT FROM PERSONALITY Affect and personality have been heralded as amongst the most important variables in predicting human behaviors. Establishing linkages between personality and affect has attracted 15 considerable attention in the past. Linking affect and personality can shed light on both. The nature of a personality construct is determined by its external correlates and one hypothesis is that personality and affect are highly correlated (Izard, 1977; McCrae & Costa, 1991; Plutchik, 1980). Indeed, personality may influence behavior indirectly via its influence on affect. The study of affect too is enhanced through establishing its link to personality. Our understanding of its nature will depend on the degree to which it is situation- versus person- dependent. The basic dimensions of affect have been equated with those dimensions that most highly correlated with the basic dimensions of personality (Emmons & Diener, 1985; Meyer & Shack, 1989; Tellegen, 1985). However, the personality-affect links are susceptible to at least four problems. Affect was narrowly defined as positive and negative affective states; personality was largely restricted to Extraversion and Neuroticism. Worse still, items defining affect could also be found in those defining personality and both affect and personality instruments employed the same type of rating instructions. No prior research has addressed the contribution of personality to predict affect with the foregoing three problems taken into consideration. Further, studies in this area were restricted mostly to English-speaking samples. The present study is intended to examine the relations between personality and affect in English, and four other languages. 1.2.1 PAST RESEARCH FINDINGS The Eysenckian superfactors of Extraversion and Neuroticism have long been argued as temperamental traits which influence feelings and behaviors (see H. J. Eysenck, 1992; H. J. Eysenck & M. W. Eysenck, 1985). Similarly, Costa and McCrae (1980) proposed a model of happiness in which personality predisposes individuals to experience different types of feelings: People high in Extraversion are predisposed for positive affective states, and people high in Neuroticism are predisposed for negative affective states. Extraversion and Neuroticism thus became the focus of past research studies on personality and affect. Measures of positive affective states were found to correlate with Extraversion but not with Neuroticism, whereas measures of negative affective states were found to correlate with Neuroticism but not with Extraversion (Costa & McCrae, 1980, 1984; Diener & Emmons, 1984; 16 Izard, Libera, Putnam, & Haynes, 1993; Meyer & Shack, 1989; O'Malley & Gillett, 1984; McCrae & Costa, 1991; Thayer, Takahashi, & Pauli, 1988; Warr, Barter, & Brownbridge, 1983; Watson & Clark, 1992, 1997; Williams, 1981). In their 1984 study, Costa and McCrae demonstrated that the very same relations between personality and affect were held when spousal ratings rather than self-ratings of personality were used (see also McCrae & Costa, 1991). The robustness of the findings led Tellegen (1985; Watson & Clark, 1984) to argue that Extraversion and Neuroticism be renamed as "Positive Emotionality" and "Negative Emotionality" respectively. Studies have also been conducted to examine the predictive utilities of Agreeableness, Conscientiousness, and Openness to Experience on affect. Positive relations were reported between Openness to Experience and positive affective states (Costa & McCrae, 1984; McCrae & Costa, 1991; Watson & Clark, 1992). Both Agreeableness and Conscientiousness were found to correlate positively with positive affective states and negatively with negative affective states (McCrae & Costa, 1991; Watson & Clark, 1992). The present study is intended to consolidate the past findings and to explore new directions in the personality-affect research area in five different languages. 1.2.2 PROBLEMS NOTED IN PAST RESEARCH The convergence of previous findings suggests meaningful links between personality and affect. However, the findings are not without doubts. Here I emphasize four. 1.2.2.1 Restricted Definition of Affect Past research has narrowly defined affect as positive and negative affective states which sometimes refer to pleasant and unpleasant states (e.g., Bradburn's Affect Balance Scale [1969]) or sometimes refer to the idiosyncratic characterization by "Positive Affect" and "Negative Affect" states in Watson, Clark, and Tellegen's (1988) PANAS. Either focus represents a slim slice of the two-dimensional space shown in Figure 2. In particular, Bradburn's Scale captures the horizontal axis of Pleasant vs Unpleasant; Watson and Tellegen's (1985) Positive Affect and Negative Affect capture parts of the Pleasant Activated and Unpleasant Activated affective states. Apparently, they represent a very limited conception 17 of affect. Affect surely is more than these four constructs - in fact, Figure 2 suggests past research has used only the upper half of the space. 1.2.2.2 Restricted Definition of Personality With a few exceptions (e.g., McCrae & Costa, 1991; Watson & Clark, 1992), previous research defined personality as Extraversion and Neuroticism. These two factors are certainly important and continue to guide research. In addition to these two dimensions, personality has more to offer. In recent years, personality psychologists have witnessed the resurgence of the Five Factor Model as a consensual descriptive map for assessing personality (Costa & McCrae, 1992; Digman, 1990; Goldberg, 1993; McCrae & John, 1992; Wiggins & Trapnell, 1997). In addition to Extraversion (E) and Neuroticism (N), researchers agree that we need Agreeableness (A), Conscientiousness (C), and Openness to Experience (O) to provide a comprehensive description of personality. McCrae and Costa (1997) compared the American structure of personality with those of six other languages, including Chinese and Japanese, and concluded that the Five Factor Model replicated well. The authors argued the universality of the FFM in describing personality and proposed that the model represents an adequate taxonomy for the description of personality (see also McCrae, Costa, & Yik, 1996). As such, the Five Factor Model provides a useful organizing scheme to relate personality to other psychological phenomena, such as human affect. The availability of validated translations of the NEO PI R and NEO FFI (Costa & McCrae, 1992) greatly enhances the feasibility of studying relations between FFM and other psychological phenomena in different languages. 1.2.2.3 Construct Overlap The distinctions between personality and affect have never been made clear. As Pervin (1993, p. 301) noted,"... one could just as readily include affect as a part of personality as include personality as a part of affect." Indeed, empirical evidence has shown that personality and affect items jointly define common dimensions on a factor space (Meyer & Shack, 1989). 18 This conceptual unclarity was carried into the scale construction of each domain (see McCrae & Costa, 1991; Schmutte & Ryff, 1997; Watson & Clark, 1992). Here is an illustration. In a personality-affect study, suppose that a researcher decided to use NEO FFI (Costa & McCrae, 1992) and PANAS (Watson, Clark, & Tellegen, 1988) to operationalize personality and affect respectively. Suppose further that the researcher found significant correlations between Extraversion and Positive Affect, and between Neuroticism and Negative Affect. It would appear that personality and affect are linked. Nonetheless, when the researcher examined the scale content, she was surprised by the degree of overlap in the items defining each scale. On the PA scale are the terms active, enthusiastic, and inspired; on the Extraversion scale are the statements "I am a very active person," and "I am a cheerful, high-spirited person." On the NA scale are the terms jittery, afraid, scared, nervous, and ashamed; on the Neuroticism scale are the statements "I often feel tense and jittery," "I rarely feel fearful or anxious," "At times I have been so ashamed I just wanted to hide." Even when the terms are not exactly the same, similar ideas are used to construct the personality and affect scales. Thus the significant relations reported in the past may be largely exaggerated by the conceptual overlap between the two domains. Worse still, PANAS (Watson, Clark, & Tellegen, 1988) was frequently used in past research and the results may even be unique to the particular measure used. Given some personality and affect scales are theoretically similarly conceptualized, shared item content is unavoidable. Nonetheless, the narrower the definition is taken to define affect and personality, the higher the chance is to build in the content overlap between the two domains of interest, and the higher the chance to report robust relations between them. One way to tackle the construct overlap problem is to expand the scope of the personality and affect measures. Instead of restricting affect to positive and negative affective states, the present dissertation offers 14 affect constructs which fit nicely within a two-dimensional space. Instead of restricting personality to Neuroticism and Extraversion, the present dissertation offers the additional factors of Openness to Experience, Agreeableness, 19 and Conscientiousness. Broadening the scope of both domains gives the present study an opportunity to observe the variations of the personality-affect relations and thus a clearer picture of how the two domains are related. 1.2.2.4 Methodological Overlap In addition to the conceptual overlap, the significant personality-affect correlations are further enhanced by the methodological overlap - shared rating format. For instance, in the trait instruction of PANAS, respondents are instructed to "Indicate to what extent you generally feel this way, that is, how you feel on the average." The instruction is very much the same as the trait instruction used in most personality inventories. With the same trait rating format used in previous studies (e.g., Meyer & Shack, 1989; Watson & Clark, 1992), the personality-affect correlations are problematic in two ways. First is that the reported relations may be exaggerated, especially when the personality and affect measures highly overlap with each other in terms of item content. Second is that if both personality and affect ratings are based on the same time frame (i.e., in general, on average), affect may bias the personality ratings. Even though some studies claim to capture the transient affect in which respondents were asked to provide their affective ratings on the basis of "the past two weeks" (e.g., Costa & McCrae, 1980, 1984; McCrae & Costa, 1991; Warret al., 1983), the ratings provided seem to suggest a trait rather than state measure of affect. Studies using daily report or hourly alarm method were a step in the right direction but were not without problems (e.g., Diener & Emmons, 1985; Emmons & Diener, 1986; Thayer et al., 1988). It is true that affect ratings for each day were collected. However, the reported personality-affect correlations were frequently computed on the basis of "aggregated" affect ratings and it is hard to judge whether it is a trait or state measure of affect. I believe the methodological overlap is a real problem in past research. The daily report method is a good beginning in overcoming the methodological overlap but more vigorous efforts have to be directed to the problem. The present study measures personality in general (trait) and affective states for an instant moment of time. Indeed, participants are asked to report their affective feelings for a 20 particular moment of time of the day. The personality ratings provided are based on a life-time information whereas the affect ratings provided are based on an instant moment of time. Hence the two types of ratings do not share the same time frame. The resulting personality-affect relations will be less susceptible to the problem of methodological overlap. 1.3 OVERVIEW OF THE PRESENT DISSERTATION In the present study, people of five different language backgrounds provided ratings of personality and affect. Scales tapping the FFM and the full two-dimensional space of affect were used. The personality scales (NEO FFI; Costa & McCrae, 1992) are well-validated translations of the original English version; the affect scales originated from English and are translated into four different languages. I seek to answer two major questions. 1.3.1 A CIRCUMPLEX MODEL OF AFFECT The present project examines the hypothesis that the schematic structure presented in Figure 2 is capable of integrating four specific structures of self-reported current affect shown in Figure 1. The focal question is the relations of the Pleasant vs Unpleasant and Activated vs Deactivated axes to all dimensions 45° away from the axes in Figure 2 - in other words, the diagonal dimensions. (Integration Hypothesis) More specifically, I examine how well the diagonal dimensions and their respective unipolar constructs fit well into the proposed integrated structure: (a) that unipolar constructs shown in Figure 2 as 180° apart are approximately bipolar, (b) that nearly all of the reliable variance in these four constructs are accounted for by the Pleasant vs Unpleasant and Activated vs Deactivated axes, and (c) that all 14 unipolar constructs fall at different angles of the integrated space and conform to a circumplex model of affect. The second hypothesis that I am testing in this five-language study is the 45° Rotation Hypothesis. The hypothesis states that the diagonal affect constructs are interchangeable. For instance, Thayer's (1996) Energy is treated as equivalent to Watson and Tellegen's (1985) Positive Affect and Larsen and Diener's (1992) Activated Pleasant. In the present study, I examine the how well this hypothesis is supported in five different languages. 21 1.3.2 RELATING AFFECT TO PERSONALITY The present study examines the empirical linkages between personality and momentary affective ratings for a thin slice of time. The Five Factor Model was used as a common framework to define personality in five different languages. I examine how well momentary affect can be predicted from the enduring personality variables. Finally, to provide another angle on how well the 14 unipolar affect constructs conform to a circumplex model, I examine the curves relating each FFM personality dimension to all 14 affect constructs. It is predicted that each pattern of associations will fit a sinusoid function, which can be determined by the variance explained by the fitted function. 1.3.3 A SAMPLE OF LANGUAGES The present study involves five different languages belonging to four distinct language families. Language families are groups of languages which share a common historical origin, similar grammar and syntax. Both English and Spanish belong to the Proto-Indo-European language family, although the former is under the Germanic branch and the latter is under the Italic branch. Chinese belongs to the Sino-Tibetan family. Japanese is generally not classified under any language family, although many Westerners assume that Chinese and Japanese are similar. Keene (1955, p. 2, quoted in McCrae & Costa, 1997) observed,"... Chinese is a language of great compactness. Japanese, on the other hand, is polysyllabic, has no tones like the Chinese, and ... is a language of interminable sentences." Finally, Korean is usually classified under the Altaic family (such as Turkish and Mongolian). 22 CHAPTER 2 METHOD 2.1 AFFECT MEASURES All five samples received an affect questionnaire in three parts, corresponding to three different response formats: (a) an adjective list accompanied by a five-point Likert scale ranging from 1 "not at all" to 5 "extremely" (Adjective Format, abbreviated ADJ), (b) a list of statements with which participants were asked to indicate their degree of agreement, ranging from 1 "strongly disagree" to 5 "strongly agree" ("Agree-Disagree" format, abbreviated AGR), and (c) a list of statements with which participants were asked to indicate how well it described their feelings, ranging from 1 "not at all" to 4 "very well" ("Describes Me" format, abbreviated DES). 2.1.1 CMQ SCALES Participants completed measures of Pleasant, Unpleasant, Activated, and Deactivated constructs, developed by Feldman Barrett and Russell (1998), who found these scales to have good convergent and discriminant validities. These scales tap the cornerstones of Russell's (1980) model and the two-dimensional space proposed in the present study to integrate other constructs. Participants also completed the semantic differential scales of Pleasure and Arousal (Mehrabian & Russell, 1974). The English version of all items for these scales are given in Appendix I. 2.1.2 THA YER'S SCALES Participants also completed scales measuring Thayer's (1996) Energy, Tiredness, Tension, and Calmness variables. Adjectives were taken directly from the authors. "Agree-Disagree" and "Describes Me" scales were constructed with same or similar adjectives and were adopted from Yik et al. (1998). The English version of all items for these scales are given in Appendix I. 2.1.3 LARSEN AND DIENER'S SCALES Participants also completed scales measuring Larsen and Diener's (1992) Activated Unpleasant, Unactivated Unpleasant, Activated Pleasant, and Unactivated Pleasant Affect variables. Adjectives were taken directly from the authors. "Agree-Disagree" and "Describes 23 Me" scales were constructed with same or similar adjectives and were adopted from Yik et al. (1998). The English version of all items for these scales are given in Appendix I. 2.1.4 WATSON AND TELLEGEN'S SCALES Participants also completed scales measuring Watson and Tellegen's (1985) "Positive Affect" and "Negative Affect". The adjective items for "Positive Affect" and "Negative Affect" were taken from Watson et al.'s Positive Affect and Negative Affect Schedule (PANAS). "Agree-Disagree" and "Describes Me" scales were constructed with same or similar adjectives and were adopted from Feldman Barrett and Russell (1998). The English version of all items for these scales are given in Appendix I. 2.1.5 TRANSLATION All translations were made by native speakers of Spanish, Chinese, Japanese, and Korean, respectively, who have training in psychology. The backtranslation procedure was adopted in all languages. All affect measures were first translated into each language by a translator. After the translation was completed, a second translator (blind to the original English version) was employed to backtranslate the questionnaire into English. Discrepancies between the backtranslated version and the original English version were reviewed by Jim Russell and myself. Translations were revised on the basis of joint consultation of the translator, backtranslator, Jim Russell, and me. Translators were advised to maintain the meaning of the items even though that might imply minor changes in the literal content. One common theoretical backdrop of the Affect Circumplex is that all affect items are made up different degrees of the two axes - Pleasant vs Unpleasant and Activated vs Deactivated axes. This backdrop implies translation is difficult if not impossible. In order to help ensure sensible translations, I made up a handbook delineating first the theoretical backdrop of the circumplex model and second a synonyms worksheet in which I listed all the affect terms vector by vector in accordance with the circumplex. Under each vector, I gave a brief description of the nature of the vector and then listed the relevant affect terms. For instance, 24 under "Pleasant Activated vector," I listed excited and elated; under "Unpleasant Deactivated," I listed tired and bored. Beside each term, I provided a few synonyms. For instance, I put down ecstatic and joyful besides elated. Finally, translators were asked to provide one translation for one term and used that translation for that term only throughout the whole questionnaire. In other words, they had to provide one single translation to each synonym and used that translation throughout the questionnaire packet. 2.2 PERSONALITY MEASURES The NEO FFI is a 60-item questionnaire designed to measure the Five Factor Model of personality. Each of the five factors is represented by 12 items. Responses are made on a 5-point rating scale ranging from "Strongly Disagree" through "Neutral" to "Strongly Agree." Data on reliability and other psychometric properties are given in Costa and McCrae (1992). The English version of the items are given in Appendix II. The NEO FFI was used in all but Spanish samples to measure the FFM. In Spanish, participants completed NEO PI R (Costa & McCrae, 1992), a 240-item questionnaire. NEO FFI scores were then computed on the basis of the 60 (out of 240) relevant items. The original English and four translations were available for the present project. No translation was thus involved in the personality measures. All translations were made by native speakers with training in psychology. Backtranslation procedure was adopted. Questionnaire items were first translated by one native speaker of the language. The translation was then backtranslated by a second translator who was blind to the original English. Discrepancies between the backtranslated version and the original English version were reviewed by the test authors. The translation was then revised and subject to validation studies. 2.3 PARTICIPANTS AND PROCEDURE Participants in all five samples completed a battery of questionnaires in groups of 10 to 15 in a laboratory or in class. 25 On the front page of each battery were the same general instructions for Remembered Moments Questionnaire, including the request that "To obtain a good sample of accurate reports, we need to ask you to remember a particular moment. Please think back to yesterday. Specifically, recall the time "just before breakfast." (If you didn't have breakfast yesterday, simply recall that approximate time of day.) It is important that you remember a specific moment accurately. So, please search your memory and try to recall where you were, what you were doing at that time, who you were with, and what you were thinking. Now select a particular moment that is especially clear in your memory. (If you really have no recollection of the time "just before breakfast." please search your memory for the closest time that you do recall accurately.)" There were six versions of the instructions. The affect and personality scales of the six versions were identical, except that each had a different anchoring time. To facilitate memory search, one anchoring time was provided in each version of instruction. The six anchoring times were "before breakfast," "after breakfast," "before lunch," "after lunch," "before dinner," and "after dinner." Participants were randomly assigned to one of the six instructions. All subsequent affect questionnaires were to be answered with respect to that selected moment of the day before. On average, completion took approximately 20 minutes. Following the affect questionnaires was the 60-items NEO FFI. Participants were instructed to "Describe yourself as you are generally or typically." On average, the completion time was 10 minutes. 2.3.1 THE ENGLISH SAMPLE Participants were 535 undergraduates (241 men, 294 women) from University of British Columbia and their mean age was 19.61 (S.D. = 3.22). They were enrolled in various psychology courses and received credits in exchange for their participation. A sample battery of questionnaires is given in Appendix III. 2.3.2 THE SPANISH SAMPLE Participants were 233 undergraduates (96 men, 137 women) from Universidad Autonoma de Madrid and their mean age was 19.83 (S.D. = 4.29). Participation was voluntary 26 and test administration took place during class time. A sample battery of questionnaires is given in Appendix IV. 2.3.3 THE CHINESE SAMPLE Participants were 487 undergraduates (164 men, 323 women) from Chinese University of Hong Kong and City University of Hong Kong. Their mean age was 19.74 (S.D. = 2.04). They were enrolled in various psychology or marketing courses. Their participation was either voluntary or in exchange for credits. A sample battery of questionnaires is given in Appendix V. 2.3.4 THE JAPANESE SAMPLE Participants were 450 undergraduates (228 men, 222 women) from Doshisha University and their mean age was 19.69 (S.D. = 1.15). Participation was voluntary and test administration was taken place during class time. A sample battery of questionnaires is given in Appendix VI. 2.3.5 THE KOREAN SAMPLE Participants were 365 undergraduates (176 men, 189 women) from Pusan University and their mean age was 21.16 (S.D. = 4.28). Participation was voluntary and test administration was during class time. A sample battery of questionnaires is given in Appendix VII. 2.4 DATA ANALYSES Correlation matrices for manifest variables were submitted to confirmatory factor analyses and structural equation modeling using SEPATH in Statistica (Steiger, 1995). Completely standardized solutions were obtained. (Thus, both latent and manifest variables are scaled to a variance of 1.00.) To control systematic error variance, we estimated the correlations between the error terms for manifest scales with the same response format. For SEPATH, many different indices are available to assess the degree to which a structural equation model fits the observed data. Because most researchers agree that no single measure of fit should be relied on exclusively (Bollen & Long, 1993), we used four indices to assess model fit when data samples were analyzed separately. First, the chi-square statistic (%2) was used. This statistic tests the null hypothesis that the hypothesized model reproduces the correlation matrix for the manifest variables. The larger the chi-square, the more the 27 correlation matrix specified by the hypothesized model deviates from the correlation matrix for the manifest variables. The chi-square statistic is dependent on sample size, however, such that it can be significant even for models that fit the data relatively well (Bentler, 1990). Second, Steiger and Lind's (1980) RMSEA, which can roughly be regarded as a root mean square standardized residual, was used. RMSEA is adjusted for model complexity and is therefore useful in comparing two nested models. Greater values indicate poorer fit. Third, the Adjusted Goodness of Fit (AGFI), which provides a direct measure of goodness of fit, was used. The values fall between 0 (a complete lack of fit) and 1 (a complete fit). This index is similar to an adjusted R 2 statistic used in the general linear model (Tanaka, 1993). Fourth, the Comparative Fitness Index (CFI; see Bentler, 1990), which is a normed-fit index that evaluates the adequacy of the hypothesized model in relation to a baseline model, was used. CFI is computed on the basis of the most restricted baseline model (null model) in which all manifest variables are assumed to be uncorrelated (i.e., every variable is an indicator for its own latent construct). Possible values range from 0 to 1, with higher values indicating better fit. Occasionally we report multi-sample analyses. Both AGFI and CFI are single-sample based statistics designed for standard structural equation models and are not appropriate for a multi-sample analysis. For our multi-sample analyses and Browne's (1992) CIRCUM, we therefore relied on x2, and RMSEA. The present study relies primarily on RMSEA for the interpretation of model fit. RMSEA is less susceptible to the sample size bias and is adjusted for model complexity. The RMSEA is taken to indicate the degree of model fit, rather than to determine whether a model is accepted or rejected. Based on past findings in similar research studies (e.g., Yik et al., 1998), an RMSEA of less than .10 is taken to indicate the hypothesized model fits the data "well;" an RMSEA between .10 and .14 is taken to be "moderately well;" an RMSEA between .15 and .18 is taken to be "modestly;" and any RMSEA greater than .18 is taken to be "poorly." The above cut-off points are best regarded as tentative and simply provide a heuristic guideline to discuss each structural equation model in the findings. 28 CHAPTER 3 ENGLISH First, I examine how well the bipolarity of the dimensions of the four structures was supported in all four structures. Second, I examine how well the four structures as defined by the original authors are supported in the data. Third, I examine how well the two bipolar axes defining the proposed two-dimensional space explain the variance in other affect constructs. Fourth, I examine how well the affect constructs conform to a circumplex model. Finally, I examine the relations between the affect constructs and the FFM. 3.1 STRUCTURE OF AFFECT 3.1.1 TEST FOR BIPOLARITY One of the major hypotheses is that the constructs shown 180° apart in Russell's (1980), Thayer's (1989), and Larsen and Diener's (1992) structural models are bipolar opposites. I begin with an assessment of this hypothesis. Russell and Carroll (in press) noted a contradiction in previous analyses of bipolarity, which requires unipolar response formats and a correlation of - 1 . These two requirements cannot be met simultaneously. Even when random and non-random error has been completely eliminated, to achieve a correlation o f - 1 requires a strictly bipolar response format, and yet bipolar response formats are (rightly) deemed illegitimate and not used. Unipolar formats were used, but the more strictly unipolar the format, the farther from - 1 is the expected correlation. Further, ostensibly unipolar formats, such as those used in the present study, vary in just how unipolar or bipolar they are. According to Russell and Carroll, the types of format adopted in the present study is ambiguous, partly unipolar and partly bipolar. The implication of Russell and Carroll's (in press) analysis is that testing bipolarity is not as straightforward as once thought. One cannot simply calculate a correlation and see how close it is to - 1 . Testing bipolarity requires a number of additional assumptions, such as that the latent bipolar dimension is normally distributed. With these assumptions, and for this type of format, I suggest the following considerations for bipolarity: 29 1. The correlation fall within the range of - .47 to -1.00. The closer to - 1 , of course, the more confident one is in bipolarity. By correlation, I refer to the correlation estimated by a structural equation modeling procedure that takes into account both random and non-random error. 2. At least one of the variates, and possibly both, show a positive skew (when they are scored, as is done traditionally and as was done here, with the lowest score corresponding to neutral and the highest score to a high degree of the named variable, such as sadness). In general, the more positive skew seen in the two variables, the lower in magnitude is the correlation between them. 3. The bivariate frequency distribution tends to be triangular, with more scores falling in the triangle that includes 1,1 than in the triangle that includes 5, 5, when the rating scale is 1 to 5. I call this triangle the lower triangle. These three considerations are to some extent overlapping, and must be considered tentative. Still, for the kinds of response formats in common use, such rules of thumb may be the best I can do for now. Table 1 shows relevant statistics for a test of bipolarity. The six hypothesized bipolar opposites, two for Russell's (1980), Thayer's (1996), and Larsen and Diener's (1992) models, respectively, were examined. For example, consider Pleasant and its hypothesized bipolar opposite, Unpleasant. Pleasant showed negative skew in two of the three response formats, but Unpleasant showed positive skew in all three formats, much greater in magnitude than the negative skew of Pleasant. For all three formats, more cases fell in the lower than in the upper triangle of the bivariate frequency distribution. Finally, correlation between Pleasant and Unpleasant was estimated with a confirmatory factor analysis with correlated error terms. The result, - .92 , fell within the predicted range, and indeed was substantial in magnitude. Overall, the results fit our considerations for bipolarity for all pairs considered. The weakest estimated correlation (-.61) occurred with Larsen and Diener's Activated Pleasant versus Unactivated Unpleasant. Accounting for the attenuated correlation is possibly the 30 positive skew in six cells and the disproportionate number of cases in the lower triangle. These results do not allow me to declare all six pairs perfect bipolar opposites. Still, the results do fit the pattern expected of bipolar opposites. In the remaining analyses, I assume that it is justified in combining these pairs into single bipolar dimensions. Table 1. Statistics for a Test of Bipolarity in the English sample Proportion of Cases Format Variable Skew Hypothesized Skew Lower Upper Opposite Triangle Triangle Two Axes ADJ Pleasant .06 Unpleasant .95 93 3 AGR Pleasant - .44 Unpleasant .52 75 10 DES Pleasant - .28 Unpleasant .33 47 28 - .92 ADJ Activated .63 Deactivated .29 85 9 AGR Activated .18 Deactivated - .07 64 27 DES Activated .23 Deactivated .13 75 20 - . 7 3 Thaver ADJ Energy .70 Tiredness .07 82 10 AGR Energy .37 Tiredness - .02 62 18 DES Energy .59 Tiredness .10 72 17 - .77 ADJ Tension .65 Calmness .17 93 5 AGR Tension .34 Calmness .17 84 10 DES Tension .29 Calmness .03 70 24 - .76 Larsen & Diener ADJ Activated .70 Unactivated .67 97 3 Pleasant Unpleasant AGR Activated .27 Unactivated .23 85 11 Pleasant Unpleasant DES Activated .68 Unactivated .25 82 11 - .61 Pleasant Unpleasant ADJ Activated .71 Unactivated .03 94 6 Unpleasant Pleasant AGR Activated .14 Unactivated - .07 65 19 Unpleasant Pleasant DES Activated .43 Unactivated .08 79 16 - . 7 9 Unpleasant Pleasant 31 Watson & Tellegen ADJ High Positive .33 Low Positive .67 92 7 Affect Affect AGR High Positive .07 Low Positive .23 77 21 Affect Affect DES High Positive .35 Low Positive .25 76 19 Affect Affect ADJ High Negative 1.15 Low Negative .03 97 2 Affect Affect AGR High Negative .21 Low Negative - .07 79 19 Affect Affect DES High Negative .55 Low Negative .08 87 9 Affect Affect Note. N = 535. Watson & Tellegen's Low Positive Affect and Low Negative Affect are defined by Larsen & Diener's Unactivated Unpleasant and Unactivated Pleasant respectively. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Lower triangle indicates the percentage of participants who fall below the diagonal in a bivariate distribution of the hypothesized opposites; upper triangle indicates the percentage of participants who fall above the diagonal. <J> indicates the latent correlation between the hypothesized opposites in a confirmatory factor analysis. All <D coefficients are significant at .001 level. 3.1.2 MEASUREMENT MODELS In this section, I examine the performance of the various scales to assess the original structures as shown in Figure 1 from which the scales were derived. For all measurement models, the following parameters were estimated: (a) the factor loadings between the manifest variables and their respective latent construct, (b) the error variance associated with each manifest variable, (c) the correlation between the error variance using the same response format, and (d) the correlation between the latent constructs. Table 2. Fit Indices for the Measurement Models in the English sample Fit Indices Constructs Model 1* df RMSEA AGFI CFI CMQ Hypothesized Model 175.20 30 .09 .87 .98 Comparison 1263.66 36 .21 .54 .80 Model Thayer's Unipolar Constructs Hypothesized Model 117.19 30 .07 .91 .99 Comparison 903.76 36 .18 .62 .87 Model Larsen & Diener's Unipolar Constructs Hypothesized Model 168.90 30 .09 .87 .98 Comparison 909.21 36 .20 .56 .86 Model Watson & Tellegen's Unipolar Constructs Hypothesized Model 145.13 5 .23 .66 .96 Comparison 247.22 6 .25 .60 .93 Model Note. N = 535. 3.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes To assess how well the horizontal and vertical axes of Figure 2 were assessed by Feldman Barrett and Russell's (1998) Current Affect Questionnaire (CMQ), I specified a confirmatory factor analysis with four latent constructs (Pleasant, Unpleasant, Activated, and Deactivated), each indicated by its three response formats. As shown in Table 2, the hypothesized model fit the data well. Model estimates and descriptive statistics for the hypothesized model are given in Table 3. Cronbach alphas for all scales were high, ranging from .64 to .91. Factor loadings ranged from .71 to .96; all were statistically significant. Interfactor correlations indicated that the hypothesized bipolar opposites correlated - .92 (Pleasant and Unpleasant) and - . 7 3 (Activated and Deactivated), respectively. Table 3. CMQ in the English sample: Standardized Factor Loadings and Interfactor Correlations 33 Factor Loadings Construct Format P U A D M SD a Pleasant ADJ .89** 2.67 1.04 .91 Pleasant AGR .96** 3.12 1.05 .91 Pleasant DES .89** 2.68 .93 .86 Unpleasant ADJ .92** 2.05 1.04 .89 Unpleasant AGR .95** 2.30 1.16 .89 Unpleasant DES .92** 2.17 .98 .90 Activated ADJ .71** 2.28 .86 .73 Activated AGR .94** 2.75 .92 .70 Activated DES .86** 2.31 .79 .76 Deactivated ADJ .83** 2.57 .99 .64 Deactivated AGR .81** 2.82 .98 .64 Deactivated DES .82** 2.21 .68 .68 Interfactor Correlations Pleasant — Unpleasant -.92** Activated .01 .04 Deactivated - .07 .09 -.73** — Note. N = 535. P = Pleasant, U = Unpleasant, A = Activated, and D = Deactivated. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *E<.05 **p_<.01 Results for the CMQ constructs are also shown in Figure 3. As anticipated, Pleasant was related highly and negatively with Unpleasant; Activated was related highly and negatively with Deactivated. Next, I compared the hypothesized model to the one where the correlations among all four latent constructs were fixed to .00. This comparison model thus posits four orthogonal unipolar factors and is what Burke et al. (1989) had found superior to a two-dimensional bipolar model. As shown in Table 2, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than the comparison model: A%2 (6, N=535) = 1088.46, p < .001, and RMSEA changed from .09 to .21. 34 For use in subsequent structural equation models, I specified a confirmatory factor model with two latent constructs, corresponding to the bipolar axes of Pleasant vs Unpleasant and Activated vs Deactivated, respectively. Each latent construct was indicated by the bipolar versions of its three response format. (For instance, the bipolar version of the adjective format of Pleasant was computed as the difference between the mean Pleasant score and the mean Unpleasant score.) The semantic differential scale of Pleasure was specified to load on the Pleasant vs Unpleasant construct; the semantic differential scale of Arousal was to load on the Activated vs Deactivated construct. The latent correlation between the two axes was fixed to .00. The hypothesized model fit the data moderately well: x 2 C 6 , N=535) = 161.15, p < .001, RMSEA = .13, AGFI = .85, CFI = .97. Factor loadings were all statistically significant and model estimates were used in subsequent analyses. in q ro ^—' c ro CO o o «= sz c >> O) c CO o —^* o c c o 10 ro c CD c 00 cu CD o cLitE M — CD o O O ™ CD _c ro o • 3 L . o X2 Li. T3 c CD -•—. GO ro o uct E sh • CO es o c c o o I — CD L _ (0 ro -*—' o X) Q. CD led C x: •4—» led x: ro o -«—• ro E Q. o I — he es 3 O ate als rthe ndic; were o •— were CO rms del ine rms o o E "D 0) i _ «•—» o ro •4—* O v. i — Q CD O C D c CD str 0WI CD cu JO CD x: (0 X) O) CD L . ion c CD ion S ro o CD x: q L _ (0 L _ o E ro o ro T3 c c ro ro ro o b nifi ms X! D) ter ro co ter 0. (0 -«—• o c L _ 00 CD Ul O Figun coeffi level. CD > CD t> CD < < 36 3.1.2.2 Thayer's Constructs To assess the structure of Thayer's (1989) model of affect, I specified a confirmatory factor analysis with four latent constructs (Energy, Tiredness, Tension, and Calmness). Each latent construct was indicated by its three response formats. As shown in Table 2, the hypothesized model fit the data well. Model estimates and descriptive statistics are given in Table 4. Cronbach alphas were high, ranging from .72 to .91. Factor loadings ranged from .81 to .95; all were statistically significant. The pattern among the four constructs was consistent with Thayer's (1989) structure such that the hypothesized bipolar opposites correlated - .77 (Energy vs Tiredness) and - .75 (Tension vs Calmness), respectively. Table 4. Thayer's Unipolar Constructs in the English sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Energy Tiredness Tension Calmness M SD a Energy ADJ .88** 2.22 .91 .88 Energy AGR .93** 2.57 .95 .83 Energy DES .93** 2.02 .85 .90 Tiredness ADJ .94** 2.93 1.06 .89 Tiredness AGR .95** 2.90 1.14 .87 Tiredness DES .94** 2.36 .99 .91 Tension ADJ .89** 2.18 .90 .86 Tension AGR .94** 2.45 1.03 .84 Tension DES .94** 2.21 .92 .89 Calmness ADJ .81** 2.46 .88 .72 Calmness AGR .88** 2.58 .83 .76 Calmness DES .88** 2.31 .80 .77 Interfactor Correlations Energy — Tiredness -.77** — Tension .02 .05 Calmness -.14** .14** -.75** ~ Note. N = 535. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p_ < .01 37 As shown in Table 2, a four-factor comparison model (the correlations among latent constructs fixed to .00) fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did their comparison model: A%2 (6, N=535) = 786.57, p_ < .001, and RMSEA changed from .07 to .18. 3.1.2.3 Larsen and Diener's Constructs To assess the structure of Larsen and Diener's (1992) model of affect, I specified a confirmatory factor analysis with four latent constructs (Activated Pleasant, Unactivated Unpleasant, Activated Unpleasant, and Unactivated Pleasant). Each latent construct was indicated by its three response formats. As shown in Table 2, the hypothesized model fit the data well. Cronbach alphas were high, ranging from .67 to .89. Model estimates and descriptive statistics are given in Table 5. Factor loadings ranged from .84 to .93; all were statistically significant. The pattern among the four constructs was consistent with Larsen and Diener's (1992) structure such that the hypothesized bipolar opposites correlated - .61 (Activated Pleasant vs Unactivated Unpleasant) and - .78 (Activated Unpleasant vs Unactivated Pleasant), respectively. As shown in Table 2, a four-factor comparison model (the correlations among latent constructs fixed to .00) fit the data poorly. Indeed, the hypothesized model fit the data significantly better than their comparison model: A x 2 (6, N=535) = 740.31, p_ < .001, and RMSEA changed from .09 to .20. Table 5. Larsen and Diener's Unipolar Constructs in the English sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format AP UU AU UP M SD a Activated ADJ .84** 2.03 .83 .84 Pleasant Activated AGR .90** 2.42 .80 .79 Pleasant Activated DES .91** 1.92 .75 .79 Pleasant Unactivated ADJ .88** 2.17 .95 .89 Unpleasant Unactivated AGR .91** 2.56 1.01 .80 Unpleasant Unactivated DES .93** 2.15 .92 .74 Unpleasant Activated ADJ .86** 2.16 .92 .85 Unpleasant Activated AGR .90** 2.67 1.05 .67 Unpleasant Activated DES .93** 2.09 .88 .83 Unpleasant Unactivated ADJ .90** 2.65 .96 .89 Pleasant Unactivated AGR .92** 2.79 .96 .83 Pleasant Unactivated DES .92** 2.27 .86 .86 Pleasant Interfactor Correlations Activated Pleasant Unactivated -.61** Unpleasant Activated -.45** .29** Unpleasant Unactivated .36** - . 0 8 -.78** Pleasant Note- N = 535. AP = Activated Pleasant, UU = Unactivated Unpleasant, AU = Activated Unpleasant, and UP = Unactivated Pleasant. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p. < .05 ** fi < .01 39 3.1.2.4 Watson and Tellegen's Constructs To assess the structure of Watson and Tellegen's (1985) model of affect, I specified a confirmatory factor analysis with two latent constructs (Positive Affect and Negative Affect). Each latent construct was indicated by its three response formats. As shown in Table 2, the hypothesized model fit the data poorly. Model estimates and descriptive statistics are given in Table 6. Cronbach alphas were high, ranging from .87 to .90. Factor loadings ranged from .86 to .97; all were statistically significant. The latent correlation between Positive Affect and Negative Affect was estimated to be - .45 . Table 6. Watson and Tellegen's Unipolar Constructs in the English sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Positive Affect Negative Affect M SD a Positive Affect ADJ .86** 2.56 .76 .87 Positive Affect AGR .92** 2.80 .84 .89 Positive Affect DES .95** 2.21 .73 .90 Negative Affect ADJ .87** 1.84 .77 .88 Negative Affect AGR .94** 2.52 .99 .88 Negative Affect DES .97** 1.94 .77 .90 Interfactor Correlations Positive Affect — Negative Affect -.45** — — Note. N = 535. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p < .05 ** p_ < .01 As shown in Table 2, a two-factor comparison model (the correlations between latent constructs fixed to .00) fit the data poorly. The hypothesized model fit the data significantly better than their comparison model: A x * (1, N=535) = 102.09, p < .001, and RMSEA changed from .23 to .25. As noted in Table 2, the hypothesized Watson and Tellegen's (1985) model fit the data poorly. One possible explanation was that systematic errors were inherent in the model and needed to be accounted for (e.g., Green et al., 1993). I examined the residual matrix and found 40 that several error variances were quite big. To further control for the systematic errors, Larsen and Diener's (1992) Unactivated Unpleasant was used to constitute the bipolar opposite of Watson and Tellegen's Positive Affect; Larsen and Diener's Unactivated Pleasant the bipolar opposite of Watson and Tellegen's Negative Affect. Yik et al. (1998) showed that Larsen and Diener's variables represent reasonable bipolar opposites of Watson and Tellegen's variables. To re-assess the structure of Watson and Tellegen's (1985) model of affect, I specified a confirmatory factor analysis with four latent constructs (High and Low Positive Affect; High and Low Negative Affect). Each latent construct was indicated by its three response formats. The hypothesized model fit the data moderately well: x * (30, N=535) = 255.82, RMSEA = .12, AGFI = .81, CFI = .97. Factor loadings ranged from .86 to .96; all were statistically significant. The latent correlation between Positive Affect and its bipolar opposite was estimated to be - .72 ; and the correlation between Negative Affect and its bipolar opposite was estimated to be - .76. High Positive Affect and High Negative Affect correlated - .46 in the revised model, as compared with - .45 in the original model. Consistent with Green et al. (1993), controlling for systematic errors does not influence the parameter estimates_of the measurement models. A four-factor comparison model fit the data poorly: x 2(36, N=535) = 1084.76, p < .001, RMSEA = .22, AGFI = .56, CFI = .85. The hypothesized model fit the data significantly better than the comparison model:: A X 2 (6 , N=535) = 828.94, p_ < .001, and RMSEA changed from .12 to .22. The bipolarity statistics were examined in the revised Watson and Tellegen's model. As shown in Table 1, for High vs Low Positive Affect, all six response formats had positive skew. For High vs Low Negative Affect, five formats had positive skew. In all cases, more cases fell in the lower than in the upper triangle of the bivariate frequency distribution. Finally, correlation between the hypothesized bipolar opposites was estimated to be - .72 and - . 7 5 , respectively. Again, the figures fell within the predicted range of - .47 and -1 .00. Therefore, the results justified combining these three pairs of scores into bipolar variables. 41 3.1.2.5 Interim Summary The analyses so far indicated that the scales used in the present study assess the hypothesized constructs. Further, the constructs within each original structure are not independent of one another, but correlate approximately in the manner predicted by the original authors and summarized in Figure 2. In every case, the orthogonal factors fit the data significantly less well than a model with factors that were correlated. In short, each of the original structures has been examined, and the next question is the relations among those structures. 3.1.3 RELA TIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS One of the major hypotheses is that the various structures shown in Figure 1 are variants of one and the same structure. Because of the hypothesized equivalence of the corresponding dimensions in Thayer's (1989), Larsen and Diener's (1992), and Watson and Tellegen (1985) models, respectively (that is, no 45° rotation is involved), the most straight-forward comparison concerns these three structures. Thus I first examine the hypothesis that Thayer's Energy is equivalent to Larsen and Diener's Activated Pleasant and Watson and Tellegen's Positive Affect; Thayer's Tiredness equivalent to Larsen and Diener's Unactivated Unpleasant; Thayer's Tension equivalent to Larsen and Diener's Activated Unpleasant and Watson and Tellegen's Negative Affect; Thayer's Calmness equivalent to Larsen and Diener's Unactivated Pleasant. In this analysis, I specified a confirmatory factor analysis with 10 latent constructs (Thayer's four constructs; Larsen and Diener's four constructs; Watson and Tellegen's Positive Affect and Negative Affect). Each latent construct was indicated by its three response formats. This model fit the data well: x*(225, N=535) = 813.01, p < .001, RMSEA = .07, AGFI = .46, CFI = .98. Estimated correlations between hypothesized equivalents are shown in Table 7. The correlations testing our hypothesized equivalences (highlighted in bold) were all above .85, with a mean of .93. CM w c LU CL) O i ~ -*—' co c o O _ro o Q . 'c •D co "c CD cn _0 "CD h -T J c CO c o co +^  CO T J c CO CD c CD T J C CO c CD CO co *t_ CD >» CO x: r-D ) c o E < to c o o O T J CD -4—• CO £ H—» to 111 Is- a) CD Q_ 1 E ro co f- to to o 3 L_ -*—' CO c o O to ~1 CD c CD b c CD 10 l _ CO to o 3 w_ 00 c O o co "L-CD >. CO JZ T J <D ro ra .> to o ro CJ Q) c CL D T J £ <D ro m to ro co o < _ 0 Q. C Z) T J - ~ CD C ro g .> ro tj a) ro o_ c c =3 3 T J CD c _ ro ro co CO 0) < CO to CD c E ro O c o CO c CD co to 0 c T J 0 0 c LU T - CM CM T -10 O 3 L--•—» to c o O to 0 c 0 b o3 c 0 CO * * in CM * * m * * in oo * * CO « « in o> CO o * o CM * in CM * m * * o 00 in o co o T J 0 CO > is < T J 0 ro > ro c 3 c CO T J C0 t; a) ro 0 T J C 0 3 | CO > 0 c o 3 < C L CO C C D 3 c CO CO CO 0 0. co CM CO « * o * * * CJ) to O CO • r « * oo cn o * CN CM w "S 3 L-"to c o o to i 0 0 H * o • I O 0 £ Jfc ^< 0 •E ro (0 D ) O 0 0. 2 0 x: •*-« c to c 0 ro > ' 3 cr 0 L -C0 0 c 0 Xi T J 0 N CO 0 x: —^' o Q. > » x: to _ 0 CO o to c 0 0 0 x> CO T J o XI 0 to o to c g _ r o L -o o 0 CO 0 to c 0 0 0 X) £ CM ro Q) 3 co L L O s 1 II _ ro to 0 o Z :i £ •§ 3 to q VI Ql in q VI Ql * 43 To be sure that the correlations across the three different sets of constructs were significantly different from zero, I compared the hypothesized model with a comparison model in which the interfactor correlations and the correlations of error variance using the same response format among Thayer's (1989) constructs, those among Larsen and Diener's (1992) constructs, and those among Watson and Tellegen's (1985) constructs were estimated. However, the correlations between constructs of the three structures and the correlated error variance between structures were fixed to zero. This comparison model, which is a null hypothesis of no relation between the three structures, fit the data poorly: x2(353, N=535) = 8886.85, p < .001, RMSEA = .18, CFI = .65. The hypothesized model fit the data significantly better than did this comparison model: Ax 2 (128, N=535) = 8073.84, p_ < .001, and RMSEA changed from .07 to .18. 3.1.4 PREDICTING AFFECT CONSTRUCTS BY PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES In this section, I focus on using the Pleasant vs Unpleasant and the Activated vs Deactivated axes as exogenous variables. Each other construct, in turn, was treated as endogenous. I could thus test the hypothesis that the Pleasant vs Unpleasant and Activated vs Deactivated axes predict the majority of the variance in every construct of the other structures (once random and systematic errors are removed). In the following structural equation models, the factor loadings between exogenous constructs (Pleasant vs Unpleasant and Activated vs Deactivated) and their corresponding eight bipolar response formats (2 constructs x 3 formats plus semantic differential) were adopted from the confirmatory factor analysis examining the measurement model with two bipolar axes whose correlation was fixed to .00. In each analysis, I estimated (a) the loadings between the endogenous construct and its three response formats, (b) the error variance associated with each response format, (c) the regression weight of the endogenous construct on the exogenous constructs, and (d) the percentage of variance explained for each endogenous construct. 44 3.1.4.1 Thayer's Constructs A separate analysis was conducted for each of Thayer's four unipolar constructs (Energy, Tiredness, Tension, and Calmness). I specified a structural equation model with three latent constructs. Two ~ Pleasant vs Unpleasant and Activated vs Deactivated - were considered as exogenous. One of Thayer's constructs was considered endogenous. I also conducted parallel analyses with the bipolar dimensions (Energy vs Tiredness, Tension vs Calmness). Results for Thayer's Energy are shown in Figure 4. As anticipated, Energy was related positively to both Pleasant vs Unpleasant and Activated vs Deactivated dimensions. The variance of the disturbance term was .19. The latter results is equivalent to saying that 81% of the variance in Thayer's Energy was explained by the two dimensions; 19% remained unaccounted for. This same result is also equivalent to saying that the square of the multiple correlation for predicting Thayer's Energy was .81 or that R = .90. Results for the remaining constructs are summarized in Table 8. All unipolar constructs were substantially explained by the two axes. Variance explained ranged from 60% to 83%, with a mean of 73%. The pattern of relations between the exogenous variables and the endogenous variables was approximately as expected in Figure 2. Results for the two bipolar dimensions were even clearer; the mean variance explained was 81%. 46 3.1.4.2 Larsen and Diener's Constructs I repeated the same analysis sequence in the preceding section with Larsen and Diener's (1992) variables. Results are summarized in Table 8. All unipolar constructs were substantially explained by the two axes; percentages ranged from 69% to 78%, with a mean of 73%. The pattern of relations between the exogenous variables and the endogenous variables was approximately as expected in Figure 2. For the bipolar dimensions, the mean variance explained was 84%. 3.1.4.3 Watson and Tellegen's Constructs Finally, the same analysis sequence was repeated with Watson and Tellegen's (1985) variables. Results are summarized in Table 8. For unipolar constructs, the mean variance explained was 79%; for bipolar constructs, the mean variance explained was 85%. The pattern of relations between the exogenous variables and the endogenous variable was approximately as expected in Figure 2. 47 Table 8. Unipolar and Bipolar Affect Constructs explained by the Pleasant vs Unpleasant and Activated vs Deactivated Axes in the English sample Regression Weights Fit Indices Construct Pleasant-Unpleasant Activated-Deactivated VAF xa RMSEA AGFI CFI Thaver Energy .50 .75 81 285.44 .08 .91 •97 (1.9) Tiredness - .35 - .69 60 291.92 .09 .90 .97 (2.8) Tension - .64 .52 69 355.55 .09 .88 .96 (2.3) Calmness .57 - .71 83 281.33 .08 .90 .97 (2.0) Energy vs .44 .76 78 343.84 .09 .88 .96 Tiredness (1.9) Tension vs - .65 .64 84 419.29 .11 .86 .95 Calmness (1.5) Larsen & Diener Activated .70 .44 69 308.77 .09 .89 .96 Pleasant (2.5) Unactivated - .46 - .69 69 322.72 .09 .89 .96 Unpleasant (2.5) Activated - .82 .32 77 388.85 .10 .88 .95 Unpleasant (2.0) Unactivated .71 - .53 78 325.39 .09 .89 .96 Pleasant (1.9) Activated .64 .65 83 420.26 .11 .86 .95 Pleasant vs (1.6) Unactivated Unpleasant Activated - .80 .45 85 439.30 .11 .86 .95 Unpleasant vs (1.4) Unactivated Pleasant Watson & Telleqen Positive Affect .69 .56 79 450.24 .11 .85 .94 (1.9) Negative Affect - .84 .27 78 405.54 .10 .87 .95 (1.8) High vs Low .61 .69 84 445.07 .11 .86 .95 Positive Affect (1.6) High vs Low - .82 .43 86 455.27 .11 .85 .95 Negative Affect (1.3) 48 Note. N = 535. df for x2 = 58. VAF = Variance Accounted For. Figures in parentheses are the standard errors. All regression coefficients are significant at .001 level, which is equivalent to an overall alpha level of less than .05 (by Bonferroni procedure, 32 regression weights x .001 = .03). 3.1.4.4 Interim Summary Russell's (1980), Thayer's (1989), Larsen and Diener's (1992), and Watson and Tellegen's (1985) structures are alternative descriptions of the same two-dimensional space. Variables that, according to Figure 2, are hypothesized equivalents, are highly correlated. Much, although not all, of the variance in any one unipolar or bipolar construct could be accounted for by the vertical and horizontal axes of Figure 2. Thus, the various structures can be integrated into a two-dimensional space, and I now turn to specifying the exact nature of that integrated structure. 3.1.5 THE FULL TWO-DIMENSIONAL SPACE The analyses reported so far justify representing all 14 constructs (3 structures x 4 unipolar constructs; 1 structure x 2 unipolar constructs) simultaneously within the two-dimensional space as defined by Pleasant vs Unpleasant and Activated vs Deactivated axes. A circumplex is a structure in which all variables array in a circular fashion within a two-dimensional space, but requires no assumption that exactly 45° separates the variables. To examine how well our data conformed to a circumplex, I used a structural equation modeling program (CIRCUM) developed by Browne (1992; Fabrigar, Visser, & Browne, 1997). This program provides estimates of the location of each variable on the circumplex. First, a score was created for each of the 14 unipolar constructs by summing the z-scores of its three separate response formats. (The semantic differential scale was not used in this analysis.) A 14 x 14 correlation matrix was then computed from the sums and was submitted to the maximum likelihood estimation using CIRCUM. In the analysis, Pleasant was designated as the reference variable (its location was fixed at 0°). The locations of other unipolar variables were then estimated relative to the Pleasant variable. The communality 49 estimates of all variables were left free to vary. No constraints were put on the minimum common score correlation. Figure 5. A Circumplex Representation of 14 Unipolar Affect Constructs in the English sample Thayer's Tension 131° (127°/134°) Larsen & Diener's Activated Unpleasant 145° (142°/148°) Watson & Tellegen's Negative Affect 151° (148°/154°) Unpleasant " 179° (176°/182°) Larsen & Diener's Unactivated Unpleasant 240° (236°/244°) Activated 89°(86<793°) Thayer's Energy 59° (56°/63°) Thayer's Tiredness 242° (238°/246°) Watson & Tellegen's Positive Affect 47° (44°/50°) Larsen & Diener's Activated Pleasant 42° (39°/46°) Pleasant 0° Larsen & Diener's Unactivated Pleasant 311° (308°/314°) ^Thayer's Calmness Deactivated 300° (297°/304°) 269° (265°/273°) The analysis converged on a solution in 14 iterations. Four free parameters were specified in the correlation function equation; additional free parameters did not improve the model fit. The final model had a total of 45 free parameters and 60 degrees of freedom. The data fit the model moderately well: %2 (60, N=535) = 479.54, p_ < .001, RMSEA = .11. The results are shown in Figure 5. CIRCUM estimates the angle within the circle for each variable, as well as a 95% confidence interval for that angle. Pleasant was fixed at 0°. Scales that, according to the 45° hypothesis, would be identical differed by anywhere up to 20° from each other. No large differences occurred in Unpleasant Deactivated quadrant. But noticeable differences did occur within the other three quadrants. Thayer's Energy did not fall within the confidence intervals of Larsen and Diener's Activated Pleasant or Watson and Tellegen's High 50 Positive Affect; Thayer's Tension did not fall within the confidence intervals of Larsen and Diener's Activated Unpleasant or Watson and Tellegen's Negative Affect; and Thayer's Calmness did not fall within the confidence interval of Larsen and Diener's Unactivated Pleasant. CIRCUM thus showed that the hypothesized equivalent constructs are not interchangeable, even though they are highly correlated. Thus, variables fell at various angles within the space, not only at multiples of 45°. Nonetheless, the circumplex model fit the data only moderately well. The marginal levels of fit indices resemble to the results obtained in Yik et al. (1998). One possible explanation for the marginal levels of fit indices is that CIRCUM is not able to estimate systematic errors in the data. Past research has shown that controlling systematic and random errors is vital to uncover the underlying circumplex structure of affect (e.g., Green et al., 1993; Russell, 1979; Yik et al., 1998). Another possible explanation is that additional variance is explained by another substantive dimension, other than the two in the circumplex. 3.2 RELATING AFFECT TO PERSONALITY In this section, I attempt to relate the whole spectrum of affect constructs to the FFM dimensions. First, the psychometric properties of the FFM dimensions are examined. Second, the variance explained by the FFM dimensions in affect dimensions was examined to determine how well personality predicts the affect ratings for a thin slice of time. Finally, to test for the integrity and viability of the circumplex model of affect, the sinusoid function is fitted to the pattern of correlations between each FFM dimension and 14 affect constructs. 3.2.1 DESCRIPTIVE STATISTICS FOR FFM Before proceeding to building the empirical relations between affect and personality, the psychometric properties of the FFM were examined. As shown in Table 9, the Cronbach alphas for the FFM dimensions ranged from .74 to .87, indicating that the measures are internally consistent. Further, the FFM dimensions are not independent but correlate with each other to a 51 certain extent. These psychometric characteristics resembles to those found in a previous administration in similar samples (Costa & McCrae, 1992, 1995). Table 9. Descriptive Statistics for the Five Factor Model in the English sample Correlation Coefficients FFM Dimensions 1 2 3 4 5 Mean SD a 1. Neuroticism 1.00 2.90 .70 .87 2. Extraversion - .32* 1.00 3.48 .70 .81 3. Openness to Experience .02 .04 1.00 3.43 .54 .74 4. Agreeableness - .15* .24* .04 1.00 3.61 .49 .74 5. Conscientiousness - .26* .15* - .07 .12 1.00 3.55 .57 .84 Note. N = 535. Each mean score is an average of 12 items. Possible mean scores range from 1 to 5. *p<.05 3.2.2 PREDICTING AFFECT FROM PERSONALITY To explore the explanatory utility of FFM on affect, I specified a structural model with five manifest personality variables (Neuroticism, Extraversion, Openness to experience, Agreeableness, Conscientiousness) on the exogenous side and one latent construct on the endogenous side. The latent endogenous construct was one of the bipolar affect dimensions (Thayer's Energy vs Tiredness, Tension vs Calmness; Larsen and Diener's Activated Pleasant vs Unactivated Unpleasant, Activated Unpleasant vs Unactivated Pleasant; Watson and Tellegen's High vs Low Positive Affect, High vs Low Negative Affect), each indicated by its three bipolar response formats. There are two reasons for using the bipolar affect dimensions to define the endogenous constructs. First, all pairs of hypothesized bipolar opposites in the four structures passed the test of bipolarity, indicating that the resulting dimensions were approximately bipolar. Second, in the preceding analysis on predicting unipolar and bipolar affect construct from the Pleasant vs Unpleasant and Activated vs Deactivated axes, results for 52 unipolar and bipolar constructs were almost identical. In fact, results for bipolar constructs were even clearer. For each endogenous variable, I created two models. In the E and N model, regression weights for Openness, Agreeableness, and Conscientiousness were set to zero. In the Full model, regression weights for all FFM dimensions were estimated. In each structural equation model, the following parameters were estimated: (a) the loadings between the endogenous construct and its three response formats, (b) the error variance associated with each response format, (c) the regression weights of the endogenous construct on each manifest exogenous variable, and (d) the percentage of variance explained for the endogenous construct. Table 10. Fit Indices for the Structural Equation Models in the English sample Endogenous Fit Indices Constructs Model X a RMSEA AGFI CFI Change Two Axes Pleasant vs E & N Model 15.25 .02 .98 1.00 Unpleasant Full Model 9.54 .00 .98 1.00 5.71 Activated vs E & N Model 34.67 .06 .96 .98 Deactivated Full Model 31.67 .06 .95 .98 3.00 Thaver Energy vs E & N Model 14.17 .01 .98 1.00 Tiredness Full Model 10.41 .01 .98 1.00 3.76 Tension vs E & N Model 26.45 .04 .97 .99 Calmness Full Model 24.61 .05 .96 .99 1.81 Larsen & Diener Activated Pleasant E & N Model 19.10 .03 .98 1.00 vs Unactivated Unpleasant Full Model 11.06 .01 .98 1.00 8.04 53 Activated E & N Model 11.81 .00 .99 1.00 Unpleasant vs Unactivated Pleasant Full Model 9.42 .00 •99 1.00 2.39 Watson & Tellegen High vs Low E & N Model 27.82 .05 •97 .99 Positive Affect Full Model 12.47 .02 .98 1.00 15.35* High vs Low E & N Model 11.51 .00 .99 1.00 Negative Affect Full Model 8.53 .00 .99 1.00 2.98 Note. N = 535. df for E & N model = 13; df for Full model = 10; df for % 2 Change = 3. *p_<.01 Fit indices for the 16 structural equation model are summarized in Table 10. Both the E and N models and the Full models fit the data well. Since the E and N model and the Full model are nested, a difference between the two chi-square statistics can be computed and the significance of the x 2 change an be used to indicate whether the addition of Openness, Agreeableness, and Conscientiousness personality variables on the exogenous side improves the model fit significantly. Of eight tests on x 2 change, only one was statistically significant. Results of the eight Full models are summarized in Table 11. The second to last column presents the variance explained by Neuroticism and Extraversion and the last column presents the additional variance explained by Openness, Agreeableness, and Conscientiousness dimensions. The variance explained by Neuroticism and Extraversion ranged from 3 % to 14%, with a mean of 9%; the additional variance explained by Openness, Agreeableness, and Conscientiousness (which was computed by deducting the variance obtained in the E and N model from that in the Full model) yielded a mean of 1%. The FFM dimensions together explained more variance in the Pleasant and Unpleasant Activated affective states than others. The pattern of regression weights demonstrated that Neuroticism was the best predictor of the affect ratings for a thin slice of time, yielding significant relations with all eight affect dimensions. 54 Table 11. Predicting Bipolar Affect Dimensions from the Five Factor Model in the English sample Regression Weights VAF by VAF by Affect Construct N E 0 A C N, E O, A , C Two Axes Pleasant vs -.32* .04 .06 .06 .05 13 1 Unpleasant (2.8) Activated vs .08 .18* .05 - . 0 3 .06 3 1 Deactivated (1.5) Thayer Energy vs -.17* .04 .06 - .01 .07 4 1 Tiredness (1.7) Tension vs .34* .08 - .02 - .04 .04 10 0 Calmness (2.5) Larsen & Diener Activated Pleasant -.20* .07 .10 .01 .07 6 2 vs Unactivated (2.1) Unpleasant Activated .38* .07 - . 0 5 - .04 .00 13 1 Unpleasant vs (2.8) Unactivated Pleasant Watson & Telleaen High vs Low - .21* .06 .13 .01 .12 7 3 Positive Affect (2.2) High vs Low .38* .07 - .04 - .06 - .01 14 0 Negative Affect (2.8) Note. N = 535. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. VAF = Variance Accounted For. Figures in parentheses are the standard errors. No standard errors were available for the VAF by O.A.C because the VAF was computed as the difference between the VAF by N,E,0,A,C and the VAF by N,E. Significance level was set at .001 in order to achieve an overall alpha of less than .05 (by Bonferroni procedure, 40 regression weights x .001 = .04). *p_<.001 3.2.3 SINUSOID FUNCTIONS In accordance with the assumption of circumplex pattern of correlations, any outside variable should relate to the affect constructs falling on a circumplex in a predictable way. For instance, the correlations between Extraversion and the 14 affect constructs should fit closely to a sine function. In this section, a sine function is fitted to five series of correlations, each 55 between one FFM dimensions and all 14 unipolar affect constructs. The variance explained by the fitted function is used to evaluate the model fit. First, a score was created for each unipolar affect construct by summing across the z-scores of its three response formats. Correlations between each FFM personality dimension and the 14 sums was computed. Second, each of the five series of correlations was fitted by a sine function. Third, variance explained by the fitted function was computed for each series, which could be used to indicate the overall model fit. The general form of the sine function was specified as y = a + b*sin(x + d), where y was the dependent variable representing the correlations between the 14 affect constructs and one FFM dimension; x was the independent variable representing the angles in radian for 14 affect construct; both a and b were the location parameters which could be used to move the y data range back to - 1 and +1; d was the phase parameter for x values which did not start from 0. If a = 0, b = 1, and d = 0, the general form of the sine function would be reduced to the special function of y = sin(x), which was more commonly seen. Of the five fitted functions, variance explained was 99% for Neuroticism, 85% for Extraversion, 92% for Openness, 94% for Agreeableness, and 92% for Conscientiousness; the mean was 92%. Figure 6 displays the fitted sine functions for Neuroticism (the best) and Extraversion (the worst) dimensions. The 14 unipolar affect constructs are arrayed along the horizontal axis on the basis of their respective angular locations computed by CIRCUM; whereas the magnitude of correlations between each the personality dimension and the 14 affect constructs was plotted on the vertical axis. 56 Figure 6. Curves Relating Affect Constructs to Neuroticism and Extraversion in the English sample To sum up, the overall results lend support to the viability of the circumplex model of affect. Affect constructs falling at different angles on the circumplex model are associated with the personality dimensions (outside variables) in a meaningful way - the sine function. Although Neuroticism was found to be the best personality predictor of the affect for a thin slice of time in the preceding analysis, all five personality dimensions exhibit a predicted pattern of relations with the circumplex model of affect. The present findings speak to the integrity and viability of how well the circumplex model of affect was supported in the present data. 57 CHAPTER 4 SPANISH 4.1 STRUCTURE OF AFFECT First, the four structural models of affect are examined and all analyses are conducted on the scales defined on an a priori basis - the scoring used in the English sample is adopted unconditionally. Second, attempts are made to locate the problematic scales, on the basis of how well they are at tapping the intended latent constructs - the measurement models and reliability analyses based on the a priori scoring will be employed to detect anomalies. Third, scales are revised to improve their ability to define the intended structural models. The revised scales are then used in subsequent analyses. 4.1.1 A PRIORI SCORING In this section, I repeated the analysis sequence in the English sample to examine the performance of the scales assessing the four structural models. 4.1.1.1 Test for Bipolarity Six pairs of hypothesized bipolar opposites were subjected to a test of bipolarity. As did in the preceding chapter, three considerations (positive skew, lower triangle in a bivariate distribution, and the latent correlation) were used to assess the bipolarity. Table 12. Statistics for a Test of Bipolarity in the Spanish sample Proportion of Cases Format Variable Skew Hypothesized Opposite Skew Lower Triangle Upper Triangle Two Axes ADJ Pleasant - .37 Unpleasant 1.10 96 1 AGR Pleasant - .60 Unpleasant 1.10 87 3 DES Pleasant - . 3 5 Unpleasant .83 73 10 - .89 ADJ Activated .28 Deactivated .20 77 12 AGR Activated .14 Deactivated - .26 61 28 DES Activated .10 Deactivated .20 79 14 - .82 Thaver ADJ Energy - .10 Tiredness .28 79 15 AGR Energy - .14 Tiredness .38 80 12 DES Energy .23 Tiredness .60 88 6 - .71 58 ADJ Tension .84 Calmness - .42 91 6 AGR Tension .65 Calmness .00 82 11 DES Tension .54 Calmness - . 0 5 75 17 - .79 Larsen & Diener ADJ Activated .50 Unactivated .81 97 3 Pleasant Unpleasant AGR Activated .16 Unactivated .56 96 3 Pleasant Unpleasant DES Activated .21 Unactivated .72 90 5 - .66 Pleasant Unpleasant ADJ Activated .83 Unactivated - . 5 9 94 6 Unpleasant Pleasant AGR Activated .44 Unactivated - . 3 9 65 21 Unpleasant Pleasant DES Activated .90 Unactivated - .17 85 8 - .87 Unpleasant Pleasant Watson & Telleqen ADJ High Positive .15 Low Positive .81 97 3 Affect Affect AGR High Positive - .10 Low Positive .56 90 9 Affect Affect DES High Positive .24 Low Positive .72 91 4 - . 7 3 Affect Affect 1.29 ADJ High Negative Low Negative - . 5 9 97 3 Affect Affect AGR High Negative .56 Low Negative - . 3 9 79 18 Affect Affect DES High Negative .90 Low Negative - . 1 7 88 6 - .82 Affect Affect Note. N = 233. Watson & Tellegen's Low Positive Affect and Low Negative Affect are defined by Larsen & Diener's Unactivated Unpleasant and Unactivated Pleasant respectively. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Lower triangle indicates the percentage of participants who fall below the diagonal in a bivariate distribution of the hypothesized opposites; upper triangle indicates the percentage of participants who fall above the diagonal. <E> indicates the latent correlation between the hypothesized opposites in a confirmatory factor analysis. All <X> coefficients are significant at .001 level. Table 12 shows relevant statistics for six hypothesized bipolar opposites. In each pair, at least three formats yielded positive skew. In all pairs, more cases fell in the lower triangle than in the upper triangle of the bivariate distribution. Finally, the estimated latent correlations 59 ranged from - .66 to - .89, all fell within the predicted range (-.47 to -1.00). In subsequent analyses, it is justified to combine the bipolar opposites into bipolar dimensions. 4.1.1.2 Measurement Models The same analysis sequence used in English was repeated in the following. 4.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes I specified a confirmatory factor analysis with four latent constructs (Pleasant, Unpleasant, Activated, and Deactivated). As shown in Table 13, the hypothesized model fit the data well. Model estimates and descriptive statistics for the hypothesized model are given in Table 14. Cronbach alphas were quite high, ranging from .59 to .93. Factor loadings ranged from .73 to .97; all were statistically significant. Interfactor correlations indicated that the hypothesized bipolar opposites correlated highly and negatively as expected: the latent correlation between Pleasant and Unpleasant was estimated to be - . 8 9 and that between Activated and Deactivated was - . 8 1 . Table 13. Fit Indices for the Measurement Models in the Spanish sample Constructs Model Fit Indices df RMSEA AGFI CFI CMQ Hypothesized 85.97 30 .08 .86 .98 Model Comparison 615.85 36 .21 .51 .79 Model Thayer's Unipolar Hypothesized 100.33 30 .10 .83 .97 Constructs Model Comparison 438.46 36 .19 .57 .85 Model Larsen & Diener's Hypothesized 138.72 30 .12 .77 .96 Unipolar Constructs Model Comparison 549.48 36 .23 .44 .81 - - — "Model -Watson & Tellegen's Hypothesized 102.09 5 .29 .47 .93 Unipolar Constructs Model Comparison 131.28 6 .27 .53 .91 Model Note. N = 233. 60 Next, I compared the hypothesized model to the one where the correlations among all four latent constructs were fixed to .00. As shown in Table 13, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did the comparison model: A%2 (6, N=233) = 529.88, p_ < .001, and RMSEA changed from .08 to .21. Table 14. CMQ in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format P U A D M SD a Pleasant ADJ .93** 2.90 1.09 .93 Pleasant AGR .97** 3.24 1.07 .93 Pleasant DES .90** 2.61 .93 .89 Unpleasant ADJ .94** 1.82 .91 .85 Unpleasant AGR .91** 1.84 1.05 .86 Unpleasant DES .93** 1.89 .92 .83 Activated ADJ .73** 2.59 .93 .86 Activated AGR .89** 2.79 1.04 .84 Activated DES .88** 2.25 .74 .74 Deactivated ADJ .83** 2.57 .92 .63 Deactivated AGR .81** 2.87 .93 .59 Deactivated DES .87** 2.11 .70 .76 Interfactor Correlations Pleasant — Unpleasant -.89** Activated .01 .22** Deactivated -.18** -.25** -.81** Note. N = 233. P = Pleasant, U = Unpleasant, A = Activated, and D = Deactivated. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p < .01 4.1.1.2.2 Thayer's Constructs I specified a confirmatory factor analysis with Thayer's (1989) four variables. As shown in Table 13, the hypothesized model fit the data well. Model estimates and descriptive statistics are given in Table 15. Cronbach alphas were high, ranging from .76 to .89. Factor loadings ranged from .81 to .95; all were statistically significant. The pattern among the four constructs 61 was consistent with Thayer's (1989) structure such that the hypothesized bipolar opposites correlated - . 7 2 (Energy vs Tiredness) and - .80 (Tension vs Calmness), respectively. Table 15. Thayer's Unipolar Constructs in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Energy Tiredness Tension Calmness M SD a Energy ADJ .85** 2.59 .88 .87 Energy AGR .94** 2.71 .94 .83 Energy DES .90** 2.18 .78 .88 Tiredness ADJ .90** 2.64 .95 .85 Tiredness AGR .91** 2.46 .99 .76 Tiredness DES .94** 1.93 .83 .85 Tension ADJ .85** 2.06 .80 .82 Tension AGR .91** 2.35 .89 .77 Tension DES .95** 2.09 .85 .89 Calmness ADJ .81** 2.84 .91 .72 Calmness AGR .87** 2.78 .86 .76 Calmness DES .89** 2.38 ,77 .76 Interfactor Correlations Energy — Tiredness -.72** — Tension .20** - .11 Calmness - .11 .15* -.80** — Note. N = 233. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p < .01 As shown in Table 13, a four-factor comparison model (the correlations among latent constructs fixed to .00) fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did their comparison model: Ax2 (6, N=233) = 338.13, p_ < .001, and RMSEA changed from . 10 to . 19. 4.1.1.2.3 Larsen and Diener's Constructs I specified a confirmatory factor analysis with Larsen and Diener's (1992) four variables. As shown in Table 13, the hypothesized model fit the data well. Model estimates and 62 descriptive statistics are given in Table 16. Cronbach alphas were quite high, ranging from .67 to .83. Factor loadings ranged from .80 to .95; all were statistically significant. The pattern among the four constructs was consistent with Larsen and Diener's (1992) structure such that the hypothesized bipolar opposites correlated - .66 (Activated Pleasant vs Unactivated Unpleasant) and - .86 (Activated Unpleasant vs Unactivated Pleasant), respectively. As shown in Table 13, a four-factor comparison model (the correlations among latent constructs fixed to .00) fit the data poorly. Indeed, the hypothesized model fit the data significantly better than their comparison model: A X 2 (6, N=233) = 410.76, p_ < .001, and RMSEA changed from . 12 to .23. Table 16. Larsen and Diener's Unipolar Constructs in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format AP UU AU UP M SD a Activated ADJ .80** 2.35 .89 .83 Pleasant Activated AGR .95** 2.57 .88 .82 Pleasant Activated DES .84** 2.03 .69 .67 Pleasant Unactivated ADJ .86** 2.04 .80 .79 Unpleasant Unactivated AGR .91** 2.14 .92 .71 Unpleasant Unactivated DES .86** 1.81 .79 .71 Unpleasant Activated ADJ .90** 2.08 .88 .85 Unpleasant Activated AGR .91** 2.51 1.08 •75 Unpleasant Activated DES .93** 1.82 .81 .83 Unpleasant Unactivated ADJ .92** 2.99 .92 .85 Pleasant Unactivated AGR .90** 3.03 1.02 .88 Pleasant Unactivated DES .90** 2.48 .84 .82 Pleasant 63 Interfactor Correlations Activated — Pleasant Unactivated -.66** Unpleasant Activated -.30** .44** Unpleasant Note. N = 233 AP = Activated Pleasant, UU = Unactivated Unpleasant, AU = Activated Unpleasant, and UP = Unactivated Pleasant. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p_<.05 **p_<.01 4.1.1.2.4 Watson and Tellegen's Constructs I specified a confirmatory factor analysis with Watson and Tellegen's (1985) two variables. As shown in Table 13, the hypothesized model fit the data poorly. Model estimates and descriptive statistics are given in Table 17. Cronbach alphas were high, ranging from .81 to .90. Factor loadings ranged from .84 to .95; all were statistically significant. The latent correlation between Positive Affect and Negative Affect was estimated to be - .40 . I also examined a two-factor comparison model (the correlations between latent constructs fixed to .00) which was shown to fit the data poorly as well. Table 17. Watson and Tellegen's Unipolar Constructs in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Positive Affect Negative Affect M SD a Positive Affect ADJ .84** 2.53 .69 .81 Positive Affect AGR .90** 2.91 .83 .88 Positive Affect DES .91** 2.17 .63 .86 Negative Affect ADJ .89** 1.79 .74 .88 Negative Affect AGR .94** 2.27 .93 .85 Negative Affect DES .95** 1.79 .70 .90 Interfactor Correlations Positive Affect — Negative Affect -.40** — Note. N = 233. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05** p_ < .01 64 As in the English sample, Watson and Tellegen's (1985) model fit the data poorly in the Spanish sample. As was done previously, Larsen and Diener's Unactivated Unpleasant and Unactivated Pleasant were used to construct the bipolar opposites of Watson and Tellegen's Positive Affect and Negative Affect. To examine how well controlling for systematic errors improved the model fit, I examined a confirmatory factor analysis with four latent constructs (High and Low Positive Affect; High and Low Negative Affect). The hypothesized model fit the data moderately well: x 2(30, N=233) = 193.53, p < .001, RMSEA = .15, AGFI = .71, CFI = .94. Factor loadings ranged from .84 to .95; all were statistically significant. The latent correlation between Positive Affect and its bipolar opposite was estimated to be - .74 ; and the correlation between Negative Affect and its bipolar opposite was estimated to be - . 8 2 . The latent correlation between Positive Affect and Negative Affect in the revised model was estimated to be - .38, very close to what was estimated in the original model (-.40). Again, results showed that model fit did not influence the parameter estimates of the model. A four-factor comparison model fit the data poorly: x 2 (36, N=233) = 614.91, RMSEA = .26, AGFI = .34, CFI = .80. The hypothesized model fit the data significantly better than the comparison model:: Ax 2 (6, N=233) = 421.38, p < .001, and RMSEA changed from .15 to .26. Next, the hypothesized bipolar opposites in the revised model were examined for bipolarity. As shown in Table 12, for High vs Low Positive Affect, five out of six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution, the estimated latent correlation o f - .73 fell within the predicted range. For High vs Low Negative Affect, three out of six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated correlation between the hypothesized bipolar opposite was - .82 which fell within the predicted range. Therefore, it is justified to combine these variables into bipolar dimensions in subsequent analyses. 65 4.1.1.2.5 Interim Summary With the exception of Watson and Tellegen's model, the analyses reported so far indicated that the other three hypothesized models fit the data reasonably well and that the orthogonal factors fit the data significantly less well than a model with factors that were correlated. With the addition of the bipolar opposites to Watson and Tellegen's two variables, model fit improved but the model parameter estimates did not yield large changes, an observation consistent with Green et al. (1993). It seemed that systematic errors inherent in the models greatly influenced the model fit but did not change the model estimates. 4.1.2 REVISED SCORING Since the scales defining the four structures are borrowed from English, it is possible that some scales may not be adequate indicators of the intended underlying constructs, while others may not be able to demonstrate reasonable levels of internal consistency. These problems may (a) obscure the structural appearance of the models tested and (b) reduce the reliability of the latent construct the scale is intended to measure. In order to define the latent constructs and thus the structural models tested by culturally appropriate scales, efforts are therefore directed to revising the scales on the basis of measurement models and reliability analyses. 4.1.2.1 Bootstrap Method Given that the four affect constructs - Pleasant, Unpleasant, Activated, Deactivated -form the cornerstones of the two-dimensional space proposed in the present dissertation to integrate other structural models, it is of paramount importance to ensure that the pattern of correlations among the four constructs aligns with that expected of a two-dimensional model. That is, in order to test the integration hypothesis, the four constructs must form two bipolar axes that are independent of each other. Further, in order to ensure the proper placements of the four constructs within the two-dimensional space, attempts are made to ensure that the Pleasant and Unpleasant scales are neutral in terms of the Activation dimension (neither 66 activated nor deactivated) and that the Activated and Deactivated scales are neutral in terms of the Pleasantness dimensions (neither pleasant nor unpleasant). To cut to the bottom line, the revision procedure has two interrelated purposes: (a) the bipolar opposites (Pleasant vs Unpleasant, Activated vs Deactivated) should correlate as negatively as possible, and (b) other correlations, such as those between the Pleasant and Activated scales, should be as minimal as possible. To assess how well the items define the four core constructs within the two-dimensional space, a two-factor principal components analysis was performed on a 162 x 162 correlation matrix among all the items (formats ignored) defining the four structural models. The unrotated solution was examined and the two factors were interpreted as Pleasantness and Activation. Items related to the four core constructs were then correlated with these two factor scores and constructs of the other three structural models. Items with unexpected pattern of correlations were singled out for further investigation. For instance, a Pleasant item should correlate highly (negatively) with any Unpleasant scale but minimally with any Activated or Deactivated scale. In addition, the revision process was also based on the content of each response format. The content of each response format of the Pleasant and Unpleasant constructs must be neither Activated nor Deactivated; likewise, the content of each response format of the Activated and Deactivated constructs must be neither Pleasant nor Unpleasant. The way to ensure each format to conform to the right amount of Activation or Pleasantness was through the chemistry of its defining items. Here is an illustration. Suppose there are three items in the adjective format of Pleasant and they all yield positive correlations with the factor score of Pleasantness. Of the three, however, two have a positive correlation with, and one has a negative correlation with the factor score of Activation. In order to be sure that the Pleasant adjective format is neither Activated nor Deactivated, one (but not both) of the two high Activated items would have to be dropped. Thus the adjective format will maintain a more or less neutral status in the Activation dimension. 67 The revision is not a one-step procedure. Rather, it is operated on a bootstrap basis. First, the 12 scales (4 constructs x 3 response formats) were revised on the basis of the preceding criteria. Second, a confirmatory factor analysis with four latent constructs and correlated error variance was examined on the revised scales. The resulting latent correlation matrix among the four constructs (Pleasant, Unpleasant, Activated, and Deactivated) was examined. The analysis sequence was repeated until the negative correlations between the bipolar opposites were maximized and other off-diagonal correlations were minimized. The above revision procedure may be accused of capitalizing on chance. Nonetheless, steps are taken to minimize this possibility by (a) making sure that no items are switched from one scale to another and (b) revising scales by means of dropping items. In other words, a conservation approach is taken. Since the psychometric properties of these scales were never test before, the present revision procedure represents the first step in revising the scales in each language sample. 4.1.2.2 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes The 12 scales defining the four core constructs were revised in a series of bootstrap steps. The result was that the six scales defining the Pleasant and Unpleasant constructs remained unchanged; five of the six scales defining the Activated and Deactivated constructs were revised accordingly. (See Appendix VIII for the rescored scales.) The revised model fit the data well: x2(30, N=233) = 52.70, RMSEA = .05, AGFI = .91, CFI = .99. The comparison • model (with all latent correlations set to .00) fit the data poorly: x2 (36, N=233) = 521.61, RMSEA = .18, AGFI = .61, CFI = .80. Indeed, the revised model fit the data significantly better than did the comparison model:: A% 2 (6, N=233) = 468.91, p_ < .001, and RMSEA changed from .05 to .18. Model estimates for the revised model are given in Table 18. Factor loadings ranged from .66 to .97; all were statistically significant. The correlations of the bipolar opposites are huge in magnitude as compared with others which were all below .20. in magnitude. The revised CMQ scales were used in all subsequent analyses. 68 Next, the newly revised Activated and Deactivated scales were examined for their bipolarity. Of the six scales, all showed positive skew. More cases fell in the lower triangle than in the upper triangle in their bivariate distributions. Finally, the correlation between the hypothesized bipolar opposites was estimated to be - . 7 6 in a confirmatory factor analysis with correlated error variance. Again, it is justified to combine these two scales into bipolar dimensions in subsequent analyses. Table 18. Revised CMQ in the Spanish sample: Standardized Factor Loadings and Interfactor Correlations Construct Format Factor Loadings M SD Pleasant Unpleasant Activated Deactivated Pleasant ADJ .93** 2.90 1.09 Pleasant AGR .97** 3.24 1.07 Pleasant DES .90** 2.61 .93 Unpleasant ADJ .94** 1.82 .91 Unpleasant AGR .91** 1.84 1.05 Unpleasant DES .93** 1.89 .92 Activated ADJ .74** 2.59 .93 Activated AGR .88** 2.83 1.06 Activated DES .79** 2.27 .81 Deactivated ADJ .77** 2.37 1.05 Deactivated AGR .66** 2.52 1.13 Deactivated DES .86** 2.02 .71 Interfactor Correlations Pleasant ~ Unpleasant -.89** Activated .05 .17* Deactivated - . 0 8 .03 -.76** Note. N = 233. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p_<.05 **p<.01 For use in subsequent structural equation models, I specified a confirmatory factor model with two latent constructs, corresponding to the bipolar axes of Pleasant vs Unpleasant and Activated vs Deactivated, respectively. Each latent construct was indicated by the bipolar 69 versions of its three response formats. The semantic differential scale of Pleasure was specified to load on the Pleasant vs Unpleasant construct; the semantic differential scale of Arousal was to load on the Activated vs Deactivated construct. The latent correlation between the two axes was fixed to .00. The hypothesized model fit the data moderately well: %2 (16, N=535) = 53.37, RMSEA = .10, AGFI = .88, CFI = .98. Factor loadings were all statistically significant and model estimates were used in subsequent analyses. 4.1.2.3 Scales defining the other three structural models Reliability estimates for scales defining Thayer's (1989), Larsen and Diener's (1992), and Watson and Tellegen's (1985) structural models were examined. The purpose was to make sure that the scales were as internally consistent as possible and that they were reasonable indicator of the intended constructs. Efforts were made to locate any scales with a Cronbach alpha less than .60. All scales passed the .60 threshold. The same scales were also examined for their factor loadings in their measurement models. Efforts were made to locate any scale which yielded a loading of less than .70. Again, all scales passed the .70 threshold. To sum up, scales defining these three structural models are internally consistent and good indicants of the intended constructs. 4.1.3 RELATIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS As did in the preceding chapter, to examine the relations among the hypothesized equivalents as shown in Figure 2, I specified a confirmatory factor analysis with 10 latent constructs. This model fit the data well: (225, N=233) = 645.18, RMSEA = .08, AGFI = .79, CFI = .96. Estimated correlations between hypothesized equivalents are shown in Table 19. The correlations testing our hypothesized equivalences (highlighted in bold) were all above .85, with a mean of .92. Next, I compared the hypothesized model with a comparison model. This comparison model fit the data poorly: x 2(353, N=233) = 4023.51, RMSEA = .18, AGFI = .38, CFI = .64. The hypothesized model fit the data significantly better than did this comparison model: A.%2 (128, N=233) = 3378.33, p<001and RMSEA changed from .08 to .18. SZ CO 'c cc Q . CO CD J C c CO o 3 to c o O CO o Q . 'c CO "c cu OJ _CU "cu )— TJ C CC c o CO ro C CO to" cu c cu b TJ c cc c cu co co _ l to" cu >. CO J C h-c o E < CO c o » CO o o TJ aj CO E co UJ ai T— cu CD Q . X I E CO (0 H CO CO •c ' co c o O co CU c cu b c cu CO CO CO t5 3 ' (O c o o cu CO TJ CD •—• <5 CO .> co cu CD ^ Z> TJ £ cu <S m co co co > cu T J ~ CU C "m CO SJ co .> CD CO D_ C C Z> 3 T J ~ CD f= * ; CO CO to > CO < Q-co CO cu c E CO o c g co c cu co co cu c T J CU CU c UJ CO co o 3 l --*—' CO c o O cu c CD D o3 c cu co L— co O) o f in r * * CO CO o in CN « * 00 00 « * CD * 00 * CO cn CO oo f 00 o * in T J *-> T J CU C C CU T J m CD T J CD * i 03 •— CO * ; CO CO ~ CO Q . ~ Q . CO jo 00 °° . • I CN JU CO 0 0 CD CO » o 3 u_ ••-» CO c o o CO ~c cu ol a> "cu I 00 CD CN o { CN I' o> * * * m - r -I I * CJ o cu f 5' CD s> > * i CD CO OJ O CD D_ Z cu CO c JU CD > •3 CT cu co cu c CD T J cu N co cu J C o Q . >. CO _CD CO o CO c CD cu CD j Q CO T J o CD CO O CO c g ro cu i_ o o JD CO O CO C CD CD 5 » CD CN cu ,-: CD CD CO CD LL 00 CN 0) O z 3 Lu VI O QI * CD * T J O E " r o i_ in 3 0 0 VI 3 » QI CO * 71 4.1.4 PREDICTING AFFECT CONSTRUCTS BY REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES As did in the preceding chapter, I test the hypothesis that the Pleasant vs Unpleasant and Activated vs Deactivated axes predict all of the variance in every construct of the other three structures. Results are summarized in Table 20. Both unipolar and bipolar constructs were substantially explained by the two axes. For Thayer's unipolar constructs, the variance explained ranged from 63% to 82%, with a mean of 75%; for the two bipolar dimensions, the mean variance explained was 85%. For Larsen and Diener's unipolar constructs, the variance explained ranged from 73% to 83%, with a mean of 79%; for the two bipolar dimensions, the mean variance explained was 90%. For Watson and Tellegen's unipolar constructs, the mean variance explained was 82%; for bipolar constructs, the mean variance was 90%. The pattern of relations between the exogenous variables and the endogenous variable was approximately as expected in Figure 2. Consistent with results in the English sample, the four structures can be integrated into a two-dimensional space comfortably. Results for bipolar constructs parallel those for unipolar constructs, or are even clearer. 72 Table 20. Unipolar and Bipolar Affect Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Spanish sample Regression Weights VAF Fit Indices Construct Pleasant-Unpleasant Activated-Deactivated t RMSEA AGFI CFI Thaver Energy .50 .76 82 132.67 .08 .89 .97 (2.9) Tiredness - .27 - .74 63 109.57 .06 .92 .98 (4.4) Tension - .62 .64 79 157.84 .09 .88 .97 (2.9) Calmness .54 - .70 77 117.67 .06 .91 .98 (3.6) Energy vs .41 .81 83 143.40 .08 .89 .97 Tiredness (2.7) Tension vs - .61 .70 87 168.08 .09 .87 .96 Calmness (2.3) Larsen & Diener Activated .71 .48 73 153.24 .09 .88 .97 Pleasant (3.5) Unactivated - .68 - .58 80 108.15 .06 .91 .98 Unpleasant (3.1) Activated - .80 .45 83 143.79 .08 .89 .97 Unpleasant (2.5) Unactivated .73 - .50 78 151.26 .08 .89 .97 Pleasant (3.0) Activated .76 .60 93 131.97 .07 .90 .98 Pleasant vs (1.7) Unactivated Unpleasant Activated - .78 .50 86 153.93 .08 .88 .97 Unpleasant vs (2.0) Unactivated Pleasant Watson & Tellegen Positive Affect .73 .50 79 244.78 .11 .82 .94 (3.0) Negative Affect - . 8 5 .34 84 143.11 .08 .88 .97 (2.2) Positive Affect .75 .59 92 192.94 .09 .86 .96 (1.8) Negative Affect - .82 .45 88 141.85 .08 .89 .98 (1.8) 73 Note. N = 233. df for %2 = 58. VAF = Variance Accounted For. Figures in parentheses are the standard errors. All regression coefficients are significant at .001 level, which is equivalent to an overall alpha level of less than .05 (by Bonferroni procedure, 32 regression weights x .001 = .03). 4.1.5 THE FULL TWO-DIMENSIONAL SPACE In this section, I examine how well the 14 affect constructs in Spanish conformed to a circumplex model via CIRCUM. I followed the analysis sequence stated in the English chapter. Figure 7. A Circumplex Representation of 14 Unipolar Affect Constructs in the Spanish sample Activated 84° (78°/89°) Larsen & Diener's Activated Unpleasant 135° (130°/141°) Watson & Tellegen's Positive Affect 144° (139°/150°) Unpleasant 169° (164°/174°) Larsen & Diener's Unactivated Unpleasant 219° (2137224°) Thayer's Tiredness 235° (2297241°) Watson & Tellegen's Positive Affect 37° (32742°) Larsen & Diener's Activated Pleasant 36° (31741°) Pleasant 0° Larsen & Diener's Unactivated Pleasant 307° (3017312°) Deactivated 257° (2517263°) Thayer's Calmness 294° (2887300°) The analysis converged on a solution in 15 iterations. Three free parameters were specified in the correlation function equation; additional free parameters did not improve the model fit. The final model had a total of 44 free parameters and 61 degrees of freedom. The data fit the model moderately well: %2 (61, N=233) = 313.33, RMSEA = .13. The results are shown in Figure 7. Scales that, according to the 45° hypothesis, would be identical differed by anywhere up to 22° from each other. Noticeable differences occurred in all four quadrants. For 74 instance, Thayer's Energy did not fall within the confidence intervals of Larsen and Diener's Activated Pleasant or Watson and Tellegen's High Positive Affect; Thayer's Tension did not fall within the confidence intervals of Larsen and Diener's Activated Unpleasant or Watson and Tellegen's Negative Affect. Consistent with the results in the English sample, results indicated that the hypothesized equivalent constructs in accordance with the 45° Hypothesis are not interchangeable, even though they are highly correlated 4.2 RELATING AFFECT TO PERSONALITY In this section, I relate the whole spectrum of affect constructs to the FFM dimensions. I followed the analysis sequence stated in the English chapter. 4.2.1 DESCRIPTIVE STATISTICS FOR THE FFM First, the descriptive statistics of the FFM were examined. As shown in Table 21, the Cronbach alphas for the FFM dimensions ranged from .67 to .86, indicating that the measures are internally consistent. Further, the FFM dimensions are not independent but correlate with each other to a certain extent. These psychometric properties resembles to those found in another administration in a similar sample (see Benet-Martinez & John, 1998). Table 21. Descriptive Statistics for the Five Factor Model in the Spanish sample Correlation Coefficients FFM Dimensions 1 2 3 4 5 Mean SD a 1. Neuroticism 1.00 3.10 .72 .86 2. Extraversion - .41* 1.00 3.34 .54 .76 3. Openness to Experience - .05 .03 1.00 3.68 .45 .67 4. Agreeableness - .20* .35* .10 1.00 3.59 .47 .68 5. Conscientiousness - .13* .17* - . 0 5 .18* 1.00 3.32 .60 .84 75 Note. N = 233. Each mean score is an average of 12 items. Possible mean scores range from 1 to 5. * E < .05 4.2.2 PREDICTING AFFECT FROM PERSONALITY The purpose of this section was to explore the explanatory utility of FFM on affect. 16 structural equation models were examined. The analysis sequence in this portion was identical to that used in the English data. Fit indices for the 16 structural equation model are summarized in Table 22. Both the E and N models and the Full models fit the data well. Of eight %2 change tests, six were statistically significant, indicating that the Openness, Agreeableness, and Conscientiousness dimensions improved the overall model fit significantly. Table 22. Fit Indices for the Structural Equation Models in the Spanish sample Fit Indices Endogenous Constructs Model RMSEA AGFI CFI x < Change Two Axes Pleasant vs Unpleasant Activated vs Deactivated Thayer Energy vs Tiredness Tension vs Calmness E & N Model Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model 28.83 15.71 31.90 18.25 29.54 15.92 25.24 15.89 .07 .05 .08 .06 .07 .05 .06 .05 .92 .99 .94 1.00 13.12* .91 .96 .93 .98 13.65* .92 .98 .94 .99 13.62* .93 .98 .94 .99 9.35 76 Larsen & Diener Activated Pleasant vs E & N 17.18 .04 .95 .99 Unactivated Model Unpleasant Full 5.18 .00 .98 1.00 12.00* Model Activated Unpleasant E & N 19.48 .04 .94 .99 vs Unactivated Model Pleasant Full 10.55 .01 .96 1.00 8.93* Model Watson & Tellegen High vs Low Positive E & N 36.78 .09 .90 .97 Affect Model Full 21.55 .07 .92 .99 15.23* Model High vs Low Negative E & N 21.83 .05 .94 .99 Affect Model Full 13.50 .04 .95 1.00 8.33 Model Note. N = 233. df for E & N model = 13; df for Full model = 10; df for x 2 Change = 3. *p_<.01 Results of the eight Full models are summarized in Table 23. The variance explained by Neuroticism and Extraversion ranged from 1% to 14%, with a mean of 8%; the additional variance explained by Openness, Agreeableness, and Conscientiousness yielded a mean of 5%. Personality was more predictive of Pleasant and Unpleasant Activated feeling states than others. The pattern of regression weights demonstrated differential predictive utility in the eight affect dimensions. 77 Table 23. Predicting Bipolar Affect Dimensions from the Five Factor Model in the Spanish sample Regression Weights VAF by VAF by Affect Construct N E O A C N, E O, A, C Two Axes Pleasant vs - .23* .18 .02 - .04 .23* 14 5 Unpleasant (4.3) Activated vs .13 .10 .24* - . 0 9 .09 1 7 Deactivated (1.6) Thayer Energy vs Tiredness .01 .18 .16 - . 1 0 .19 3 6 (2.4) Tension vs Calmness .27* - . 0 5 .18 - .02 - .07 9 3 Larsen & Diener (3.6) Activated Pleasant vs - .07 .20 .13 - .07 .20 7 5 Unactivated (3.3) Unpleasant Activated Unpleasant .29* - . 0 9 .12 .01 - .14 12 4 vs Unactivated (4.1) Pleasant Watson & Tellegen High vs Low - .10 .18 .15 - .08 .22* 6 7 Positive Affect (3.2) High vs Low .29* - .10 .10 - . 0 0 - .14 12 4 Negative Affect (4.1) Note. N = 233. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. VAF = Variance Accounted For. Figures in parentheses are the standard errors. No standard errors were available for the VAF by 0,A,C because the VAF was computed as the difference between the VAF by N,E,0,A,C and the VAF by N,E. Significance level was set at .001 in order to achieve an overall alpha of less than .05 (by Bonferroni procedure, 40 regression weights x .001 = .04). *p_<.001 4.2.3 SINUSOID FUNCTIONS As did in the English chapter, a sine function was fitted to five series of correlations, each between one FFM dimension and 14 unipolar affect constructs. Of the five fitted functions, variance explained was 98% for Neuroticism, 98% for Extraversion, 96% for Openness, 91% for 78 Agreeableness, and 99% for Conscientiousness; the mean was 96%. Results for Neuroticism (the best) and Agreeableness (the worst) series are displayed in Figure 8. Again, the results lent support to the viability of the circumplex model of affect in Spanish. Affect constructs falling at different angles on the circumplex model are associated with the personality dimensions (outside variables) in a meaningful way - the sine function. Figure 8. Curves Relating Affect Constructs to Neuroticism and Agreeableness in the Spanish sample 0.50 -, CO c o — A Neuroticism (98%) CO CD o O 0.00 J o CD TJ 3 C OJ CO Agreeableness (91 %) 0.50 0 90 180 270 360 Angular Positions on the circumference of the Two-Dimensional Space (in degrees) I postpone the discussion of similarities and differences between the findings in the English and the Spanish samples until Chapter 8. 79 CHAPTER 5 CHINESE 5.1 STRUCTURE OF AFFECT In this section, I repeated the same analysis sequence stated in the Spanish sample (Chapter 4). 5.1.1 A PRIORI SCORING 5.1.1.1 Test for Bipolarity Table 24 shows relevant statistics for six hypothesized bipolar opposites. With the exception of Activated vs Deactivated, the remaining five pairs fared well in the test of bipolarity. Of the five pairs, at least three out of six variates yielded positive skew; more cases fell in the lower triangle than in the upper triangle of the bivariate distribution. Finally, the estimated latent correlations ranged from - .58 to - .88 , all fell within the predicted range (-.47 to -1.00). In subsequent analyses, it is justified to combine these five bipolar opposites into bipolar dimensions. Table 24. Statistics for a Test of Bipolarity in the Chinese sample Proportion of Cases Format Variable Skew Hypothesized Opposite Skew Lower Triangle Upper Triangle Two Axes ADJ Pleasant - .01 Unpleasant .82 91 4 AGR Pleasant - .39 Unpleasant .58 74 11 DES Pleasant - .31 Unpleasant .31 35 41 - . 8 8 ADJ Activated .29 Deactivated .18 72 19 AGR Activated .11 Deactivated - .01 68 24 DES Activated .25 Deactivated .15 64 30 - .27 Thaver ADJ Energy .39 Tiredness .29 84 16 AGR - Energy - .30 Tiredness .04 74 14 DES Energy .30 Tiredness - .02 62 26 - . 7 6 ADJ Tension .77 Calmness - .07 91 7 AGR Tension .19 Calmness .12 75 17 DES Tension .29 Calmness - .06 56 38 - . 5 8 80 Larsen & Diener ADJ Activated .13 Unactivated .42 92 6 Pleasant Unpleasant AGR Activated - . 0 5 Unactivated - .02 60 35 Pleasant Unpleasant 20 DES Activated .46 Unactivated .14 71 Pleasant Unpleasant ADJ Activated .65 Unactivated - .14 92 6 Unpleasant Pleasant 26 AGR Activated .00 Unactivated - .14 55 Unpleasant Pleasant 22 DES Activated .32 Unactivated - .12 72 Unpleasant Pleasant Watson & Telleqen ADJ High Positive .19 Low Positive .42 90 8 Affect Affect AGR High Positive .01 Low Positive - .02 68 28 Affect Affect DES High Positive .25 Low Positive .14 71 24 Affect Affect .95 ADJ High Negative Low Negative - .14 97 2 Affect Affect AGR High Negative .25 Low Negative - .14 86 12 Affect Affect DES High Negative .34 Low Negative - .12 77 19 Affect Affect - .76 Note. N = 487. Watson & Tellegen's Low Positive Affect and Low Negative Affect are defined by Larsen & Diener's Unactivated Unpleasant and Unactivated Pleasant respectively. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Lower triangle indicates the percentage of participants who fall below the diagonal in a bivariate distribution of the hypothesized opposites; upper triangle indicates the percentage of participants who fall above the diagonal. O indicates the latent correlation between the hypothesized opposites in a confirmatory factor analysis. All Q> coefficients are significant at .001 level. 5.1.1.2 Measurement Models To examine the how well each of the four structural models fit the Chinese data, I repeated the analysis sequence used in the preceding chapters. 81 5.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes I specified a confirmatory factor analysis with four latent constructs (Pleasant, Unpleasant, Activated, and Deactivated). As shown in Table 25, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did the comparison model: A%2 (6, N=487) = 972.52, p < .001, and RMSEA changed from . 13 to .26. Table 25. Fit Indices for the Measurement Models in the Chinese sample Fit Indices Constructs Model df RMSEA AGFI CFI CMQ Hypothesized Model 314.40 30 .13 .77 .94 Comparison 1286.92 36 .26 .37 .74 Model Thayer's Unipolar Constructs Hypothesized Model 197.81 30 .10 .84 .97 Comparison 931.04 36 .22 .49 .83 Model Larsen & Diener's Unipolar Constructs Hypothesized Model 53.89 30 .04 .95 1.00 Comparison 1097.61 36 .26 .37 .82 Model Watson & Tellegen's Unipolar Constructs Hypothesized Model 78.27 5 .17 .80 .98 Comparison 268.97 6 .27 .56 .92 Model Note. N = 487. Model estimates and descriptive statistics for the hypothesized model are given in Table 26. Cronbach alphas varied from scale to scale, ranging from .42 to .91. With the exception of the adjective format of Activated, factor loadings were generally substantial, ranging from .62 to .94; all were statistically significant. The interfactor correlation matrix indicated that although the hypothesized bipolar opposites correlated negatively as expected, the magnitude with Activated vs Deactivated was extremely low. The matrix also indicated that Activated was highly correlated with Pleasant and Unpleasant. Table 26. CMQ in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 82 Factor Loadings Construct Format P U A D M SD a Pleasant ADJ .88** 2.75 .96 .91 Pleasant AGR .92** 3.15 .89 .89 Pleasant DES .92** 2.80 .90 .89 Unpleasant ADJ .90** 2.03 1.03 .90 Unpleasant AGR .94** 2.24 1.07 .86 Unpleasant DES .90** 2.22 .95 .90 Activated ADJ -.18** 2.85 .73 .68 Activated AGR .73** 2.75 .67 .22 Activated DES .92** 2.17 .73 .72 Deactivated ADJ .62** 2.55 .75 .42 Deactivated AGR .76** 2.74 .80 .49 Deactivated DES .82** 2.38 .60 .65 Interfactor Correlations Pleasant — Unpleasant -.89** Activated -.64** .81** Deactivated .10 - .12* -.19** Note. N = 487. P = Pleasant, U = Unpleasant, A = Activated, and D = Deactivated. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p_<.05 **p_<.01 5.1.1.2.2 Thayer's Constructs I specified a confirmatory factor analysis with Thayer's (1989) four variables. As shown in Table 25, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did their comparison model: A/2 (6, N=487) = 733.23, p_ < .001, and RMSEA changed from .10 to .22. Model estimates and descriptive statistics for the hypothesized model are given in Table 27. Cronbach alphas ranged from .62 to .92. Factor loadings ranged from .75 to .95; all were statistically significant. The pattern among the four constructs was consistent with Thayer's 83 (1989) structure such that the hypothesized bipolar opposites correlated - .77 (Energy vs Tiredness) and - .57 (Tension vs Calmness), respectively. Table 27. Thayer's Unipolar Constructs in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Energy Tiredness Tension Calmness M SD a Energy ADJ .87** 2.37 .87 .88 Energy AGR .88** 2.62 .84 .79 Energy DES .92** 2.24 .79 .91 Tiredness ADJ .89** 2.78 .96 .86 Tiredness AGR .90** 2.75 .95 .73 Tiredness DES .91** 2.42 .86 .81 Tension ADJ .88** 1.98 .90 .88 Tension AGR .95** 2.49 1.05 .88 Tension DES .94** 2.19 .95 .92 Calmness ADJ .75** 2.82 .77 .62 Calmness AGR .84** 2.74 .71 .69 Calmness DES .76** 2.57 .68 .67 Interfactor Correlations Energy • — Tiredness -.77** — Tension -.56** .60** Calmness .15** - .06 -.57** — Note. N = 487. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p_ < .01 5.1.1.2.3 Larsen and Diener's Constructs I specified a confirmatory factor analysis with Larsen and Diener's (1992) four variables. As shown in Table 25, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than their comparison model: A%2 (6, N=487) = 1043.72, p < .001, and RMSEA changed from .04 to .26. Table 28. Larsen and Diener's Unipolar Constructs in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations 84 Factor Loadings Construct Format AP UU AU UP M SD a Activated ADJ .89** 2.52 .91 .90 Pleasant Activated AGR .90** 3.03 .80 .80 Pleasant Activated DES .89** 2.17 .77 .83 Pleasant Unactivated ADJ .87** 2.31 .93 .87 Unpleasant Unactivated AGR .93** 2.70 1.02 .82 Unpleasant Unactivated DES .92** 2.28 .92 .80 Unpleasant Activated ADJ .90** 2.02 .95 .91 Unpleasant Activated AGR .94** 2.71 1.15 .85 Unpleasant Activated DES .92** 2.05 .83 .84 Unpleasant Unactivated ADJ .83** 2.86 .77 .79 Pleasant Unactivated AGR .91** 2.92 .81 .77 Pleasant Unactivated DES .87** 2.52 .78 .80 Pleasant Interfactor Correlations Activated — Pleasant Unactivated -.76** Unpleasant Activated -.68** .70** Unpleasant Unactivated .56** -.50** -.79** Pleasant Note, N = 487. AP = Activated Pleasant, UU = Unactivated Unpleasant, AU = Activated Unpleasant, and UP = Unactivated Pleasant. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format * p < .05 ** p. < .01 Model estimates and descriptive statistics for the hypothesized model are given in Table 28. Cronbach alphas ranged from .80 to .91. Factor loadings ranged from .83 to .94; all were statistically significant. The pattern among the four constructs was consistent with Larsen and 85 Diener's (1992) structure such that the hypothesized bipolar opposites correlated - .76 (Activated Pleasant vs Unactivated Unpleasant) and - .79 (Activated Unpleasant vs Unactivated Pleasant), respectively. 5.1.1.2.4 Watson and Tellegen's Constructs I specified a confirmatory factor analysis with Watson and Tellegen's (1985) two variables. As shown in Table 25, both the hypothesized model and the comparison model fit the data poorly. Model estimates and descriptive statistics for the hypothesized model are given in Table 29. Cronbach alphas ranged from .83 to .91. Factor loadings ranged from .81 to .97; all were statistically significant. The latent correlation between Positive Affect and Negative Affect was estimated to be - .61 . Table 29. Watson and Tellegen's Unipolar Constructs in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Positive Affect Negative Affect M SD a Positive Affect ADJ .81** 2.45 .65 .83 Positive Affect AGR .93** 2.86 .72 .87 Positive Affect DES .91** 2.25 .65 .90 Negative Affect ADJ .88** 1.72 .75 .91 Negative Affect AGR .97** 2.30 .93 .88 Negative Affect DES .96** 1.97 .74 .91 Interfactor Correlations Positive Affect — Negative Affect -.61** — Note. N = 487. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p<.05 **p_<.01 As did in the preceding two language samples, I examined the possible role played by systematic errors in the model fit by using Larsen and Diener's (1992) Unactivated Unpleasant and Unactivated Pleasant variables to construct the bipolar opposite of Watson and Tellegen's 86 Positive and Negative Affect, respectively. The revised model fit the data moderately well: x2 (30, N=487) = 138.56, RMSEA = .08, AGFI = .89, CFI = .98. Factor loadings ranged from .82 to .97; all were statistically significant. The latent correlation between Positive Affect and its bipolar opposite was estimated to be - .74; and the correlation between Negative Affect and its bipolar opposite was estimated to be - .77. The correlation between Positive Affect and Negative Affect was estimated to be - .61 which was identical to what was obtained in the original model. A four-factor comparison model fit the data poorly: x 2 (36, N=487) = 1110.15, RMSEA = .26, AGFI = .35, CFI = .82. The hypothesized model fit the data significantly better than the comparison model:: Ax 2 (6, N=487) = 971.59, p < .001, and RMSEA changed from .08 to .26. The hypothesized bipolar opposites in the revised model were examined for bipolarity. As noted in Table 24, for High vs Low Positive Affect, five out of six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated latent correlation of - .75 fell within the predicted range. For High vs Low Negative Affect, three out of six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated correlation between the hypothesized bipolar opposite was - .77 , which fell within the predicted range. Therefore, it is justified to combine these variables into bipolar dimensions in subsequent analyses. 5.1.1.2.5 Interim Summary The analyses just reported revealed two problems with the a priori scoring. First, the Pleasant vs Unpleasant and Activated vs Deactivated axes were problematic suggesting that they might not be readily transportable from English to Chinese simply by translation. In the next section, revisions of these scales were discussed. Second, as in English and Spanish, Watson and Tellegen's model did not fit the data well. As was done in the preceding languages, the hypothesized bipolar opposites of their two variables were added and the model fit improved 87 to a large extent. Results seemed to suggest that controlling for systematic and random error was vital to give good model fit. 5.1.2 REVISED SCORING The same revision sequence adopted in the Spanish sample was repeated with the Chinese sample. 5.1.2.1 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes The result of a series of revisions was that the six scales defining the Pleasant and Unpleasant constructs remained unchanged; five of six scales defining the Activated and Deactivated constructs were revised accordingly. (See Appendix VIII for the rescored scales.) The revised model fit the data moderately well: x2 (30, N=487) = 217.25, RMSEA = .12, AGFI = .82, CFI = .96. The comparison model (with all latent correlations set to .00) fit the data poorly: X2 (36, N=487) = 877.39, RMSEA = .19, AGFI = .59, CFI = .80. Indeed, the revised model fit the data significantly better than did the comparison model:: Ax 2 (6, N=487) = 660.14, p_ < .001, and RMSEA changed from .12 to .19. Model estimates for the revised model are given in Table 30. Factor loadings ranged from .45 to .94; all were statistically significant. The correlations of the bipolar opposites are relatively substantial in magnitude as compared with others which were all below .20 in magnitude. The revised CMQ scales were used in all subsequent analyses. Next, the newly revised Activated and Deactivated scales were examined for their bipolarity. Of the six scales, four showed positive skew. In all three pairs, more cases fell in the lower triangle than in the upper triangle in their bivariate distributions. Finally, the correlation between the hypothesized bipolar opposites was estimated to be - .42 in a confirmatory factor analysis with correlated error variance. The correlation missed the lower boundary by .05. Accounting for the attenuated correlations are possibly the positive skew in four of six variates and the large proportion of cases falling in the lower triangle of the bivariate distribution. While not psychometrically justified, the combined Activated vs Deactivated bipolar dimension is still the best approximation of the bipolar vertical axis in the proposed integrated structure. 88 Table 30. Revised CMQ in the Chinese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Pleasant Unpleasant Activated Deactivated M SD Pleasant ADJ .88** 2.75 .96 Pleasant AGR .92** 3.15 .89 Pleasant DES .92** 2.80 .90 Unpleasant ADJ .91** 2.03 1.03 Unpleasant AGR .94** 2.24 1.07 Unpleasant DES .89** 2.22 .95 Activated ADJ .46** 3.08 .69 Activated AGR .78** 2.81 .69 Activated DES .45** 2.14 .72 Deactivated ADJ .79** 2.47 .81 Deactivated AGR .81** 2.74 .80 Deactivated DES .76** 2.26 .69 Interfactor Correlations Pleasant — Unpleasant -.88** Activated .16** .07 Deactivated .05 - .12* -.44** — Note. N = 487. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * E < .05 ** E < .01 For use in subsequent structural equation models, I specified a confirmatory factor model with two latent constructs, corresponding to the bipolar axes of Pleasant vs Unpleasant and Activated vs Deactivated, respectively. Each latent construct was indicated by the bipolar versions of its three response formats. The semantic differential scale of Pleasure was specified to load on the Pleasant vs Unpleasant construct; the semantic differential scale of Arousal was to load on the Activated vs Deactivated construct. The latent correlation between the two axes was fixed to .00. The hypothesized model fit the data moderately well: %2 (16, N=487) = 135.02, RMSEA = .13, AGFI = .85, CFI = .96. Factor loadings were all statistically significant and model estimates were used in subsequent analyses. 89 5.1.2.2 Scales defining the other three structural models Scales were examined for their internal consistency; all passed the .60 threshold. Scales were also examined for their factor loadings in their intended construct. Again, all scales passed the .70 threshold. To sum up, scales defining these three structural models are internally consistent and good indicants of the intended models. 5.1.3 RELA TIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS As did in the preceding languages, I specified a confirmatory factor analysis with 10 latent constructs (Thayer's Energy, Tiredness, Tension, and Calmness; Larsen and Diener's Activated Pleasant, Unactivated Unpleasant, Activated Unpleasant, and Unactivated Pleasant; Watson and Tellegen's Positive Affect and Negative Affect) to examine the relations among the hypothesized equivalents. This model fit the data well: x2 (225, N=487) = 733.65, RMSEA = .06, AGFI = - .34, CFI = .98. (Because of the huge number of paths for correlated error variance specified in the model, the AGFI was going to overly compensate for them. Thus the values were not useful to evaluate the model fit.) Estimated correlations between hypothesized equivalents are shown in Table 31. The correlations testing our hypothesized equivalences (highlighted in bold) were all above .90, with a mean of .96. Next, I compared the hypothesized model with a comparison model (a null hypothesis of no relation between the three structures), which fit the data poorly: x2 (353, N=487) = 8637.63, RMSEA = .18, AGFI = .41, CFI = .62. The hypothesized model fit the data significantly better than did this comparison model: Ax2 (128, N=487) = 7903.98, rj<.001, and RMSEA changed from .06 to .18. CD to CD CZ x : O CD x : -*—• c co o 3 i_ 00 c o O i_ ro o CL ' c CO " c CD CO _CD 0 1-T3 C ro c o w ro T J C ro co" CD c CD b T J C ro c CD co 1_ ro _ i to" 0 >% ro x : r -CO c o E < CO c o ' ro k_ o O T J 0 ro E -•—* co U J CO 0 CD C L X ) E ro ro 1- CO co o 3 1— oo c o O CO "l CD c 0 b o3 c 0 co i_ ro co o 3 L_ -4—" co c o O CO 0 >. ro x : T J (D *-« ro ro .> co m 0 ro ^ T J £ -»—' CO ro co > 0 o Q-T J t : 0 c ro ro co . > CO t j 0 CO CL C C z> = T J *i CD £ ti CO > ro o £ co co 0 c £ ro O c g to c 0 to to 0 c T J 0 >> CO l_ 0 c LU « * CO CM « * CO CO * * oo CO •8 3 t— 10 Ci o O CO ~t-0 c 0 b c 01 ro I s-r * * en oo CO * CO cn * * CD I s-I s-« * cn cn CO CO * « CO CO CN cn * * CO I s-r in T J 0 CZ | S „ , ~ ro co Q . ~ o. ro ro t i l c c i j c c l < Q . D D < D D Q . co o| 3 k--«—» 10 c o O to " c 0 D l 0 0 08 c o CO * * in • I * i CD * " I cn ^ to m cn * I s-CD r CO CM * I s-m £ o * * to 0 0 cn • I O CD ^ 0 $ > *J CO CO CO O 0 CL Z 0 x : < to c 0 CO > •3 cr 0 ro 0 c 0 X ) T J 0 N CO 0 SZ O C L > . x : co J D ro o to c 0 0 £ 0 X I to T J O X ) C 0 (O O to c o _ r o S> o o 0 ro o to • c 0 0 0 - Q CM 0 ^ ro CD to 5 2 CO = LI L L o 0 J^ r T J 00 o ^ E 11 _ Z | I 01 "S • t t 3 to 91 5.1.4 PREDICTING AFFECT CONSTRUCTS BY REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES In this section, I repeated the analysis sequence used in the preceding languages to test the hypothesis that the Pleasant vs Unpleasant and Activated vs Deactivated axes predict the majority of the variance in every construct of the other three structures (once random and systematic errors are removed). Results are summarized in Table 32. Table 32. Unipolar and Bipolar Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Chinese sample Regression Weights Fit Indices Construct Pleasant- Activated- VAF X2 RMSEA AGFI CFI Unpleasant Deactivated Energy .73 .48 77 236.65 .08 .91 .96 (2.5) Tiredness - .66 - . 4 5 64 217.51 .07 .92 .97 (3.1) Tension - . 8 8 .22 82 369.45 .11 .86 .94 (1.8) Calmness .50 - . 7 8 86 256.55 .08 .90 .95 (3.2) Energy vs .74 .49 79 248.56 .08 .91 .97 Tiredness (2.2) Tension vs - .82 .49 91 330.87 .10 .87 .95 Calmness (1.5) Larsen & Diener Activated .87 .33 86 255.97 .09 .90 .96 Pleasant (1.8) Unactivated - . 7 8 - . 3 3 72 252.20 .08 .90 .96 Unpleasant (2.5) Activated - .91 .18 86 304.07 .10 .88 .96 Unpleasant (1.5) Unactivated .79 - . 4 9 86 222.47 .08 .91 .97 Pleasant (2.1) 92 Activated .87 .35 89 316.29 .10 .88 .96 Pleasant vs (1.4) Unactivated Unpleasant Activated - .90 .33 93 341.06 .10 .86 .95 Unpleasant vs (1.1) Unactivated Pleasant Watson & Tellegen Positive Affect .81 .35 78 307.69 .09 .89 .95 (2.2) Negative Affect - .89 .18 83 344.00 .10 .86 .95 (1.6) High vs Low .84 .37 84 312.34 .09 .88 .96 Positive Affect (1.6) High vs Low - .90 .35 93 353.35 .11 .86 .95 Negative Affect (1.1) Note. N = 487. df for %2 = 58. VAF = Variance Accounted For. Figures in parentheses are the standard errors. All regression coefficients are significant at .001 level, which is equivalent to an overall alpha level of less than .05 (by Bonferroni procedure, 32 regression weights x .001 = .03). As in English and Spanish, all unipolar and bipolar constructs were substantially explained by the two axes. For Thayer's unipolar constructs, the variance explained ranged from 64% to 86%, with a mean of 77%; for the two bipolar dimensions, the mean variance explained was 85%. For Larsen and Diener's unipolar constructs, the variance explained ranged from 72% to 86%, with a mean of 83%; for the two bipolar dimensions, the mean variance explained was 91 %. For Watson and Tellegen's unipolar constructs, the mean variance explained was 81%; for the two bipolar dimensions, the mean variance explained was 89%. The pattern of relations between the exogenous variables and the endogenous variable was approximately as expected in Figure 2. Using the newly revised scales defining the Pleasant vs Unpleasant and Activated vs Deactivated axes, the proposed two-dimensional space was able to explain most if not all variance in constructs defining other three structural models. Thus, the four structures can be integrated nicely within the two-dimensional space. 93 5.1.5 THE FULL TWO-DIMENSIONAL SPACE In this section, I examine how well the Chinese data conformed to a circumplex. As did in the preceding two languages, a 14 x 14 correlation matrix was submitted to the maximum likelihood estimation using CIRCUM. Figure 9. A Circumplex Representation of 14 Unipolar Affect Constructs in the Chinese sample Activated 85° (80°/90°) Thayer's Tension 149° (145°/154°) Larsen & Diener's Activated Unpleasant 153° (149°/157°) Watson & Tellegen's Negative Affect 153° (149°/157°) Unpleasant 167° (163°/171°) Larsen & Diener's Unactivated Unpleasant 213° (2097217°) Thayer's Tiredness 225° (2207229°) Thayer's Energy 42° (38746°) Watson & Tellegen's Positive Affect 35° (31738°) Larsen & Diener's Activated Pleasant 25° (22728°) Pleasant 0° Larsen & Diener's Unactivated Pleasant 314° (3107318°) Thayer's Calmness 295° (2917300°) Deactivated 277° (2737282°) The analysis converged on a solution in 23 iterations. Three free parameters were specified in the correlation function equation; additional free parameters did not improve the model fit. The final model had a total of 44 free parameters and 61 degrees of freedom. The data fit the model moderately well: x2 (61, N=487) = 498.81, RMSEA = .12. The results are shown Figure 7. Hypothesized equivalent scales were apart up to 19° from each other. Noticeable differences occurred in all four quadrants. For instance, Thayer's Energy did not fall 94 within the confidence intervals of Larsen and Diener's Activated Pleasant or Watson and Tellegen's High Positive Affect. Consistent with the results in the preceding languages, results indicated that the hypothesized equivalent constructs in accordance with the 45° Hypothesis are not interchangeable, even though they are highly correlated. 5.2 RELATING AFFECT TO PERSONALITY In this section, I repeat the analysis sequence in the preceding languages in relating the affect constructs to personality. 5.2.1 DESCRIPTIVE STATISTICS FOR THE FFM First, the descriptive statistics of the FFM were examined. As shown in Table 33, the Cronbach alphas for the FFM dimensions ranged from .62 to .83, indicating that the measures are internally consistent. Further, the FFM dimensions are not independent but correlate with each other to a certain extent. These psychometric characteristics resemble to those reported in another administration in a similar sample (J. Zhang, personal communication, October 22, 1998). Table 33. Descriptive Statistics for the Five Factor Model in the Chinese sample Correlation Coefficients FFM Dimensions 1 2 3 4 5 Mean SD a 1. Neuroticism 1.00 3.18 .61 .83 2. Extraversion - .42* 1.00 3.16 .48 .72 3. Openness to Experience .04 .02 1.00 3.33 .45 .64 4. Agreeableness - .19* .19* - .07 1.00 3.26 .41 .62 5. Conscientiousness - .32* .25* .08 .11 1.00 3.35 .51 .80 Note. N = 487. Each mean score is an average of 12 items. Possible mean scores range from 1 to 5. *fi<.05 95 5.2.2 PREDICTING AFFECT FROM PERSONALITY To explore the explanatory utility of FFM on affect, I repeat the analysis sequence used in the preceding chapters. Table 34. Fit Indices for the Structural Equation Models in the Chinese sample Fit Indices Endogenous Constructs Model RMSEA AGFI CFI x z Change Two Axes Pleasant vs Unpleasant Activated vs Deactivated Thayer Energy vs Tiredness Tension vs Calmness Larsen & Diener Activated Pleasant vs Unactivated Unpleasant Activated Unpleasant vs Unactivated Pleasant E & N Model Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model 21.59 17.73 26.12 19.91 34.18 26.08 23.82 21.99 25.35 18.20 22.80 20.99 .06 .04 .04 .04 .06 .06 .04 .05 .04 .04 .04 .05 .97 1.00 .97 1.00 3.86 .96 .98 .96 .98 6.21 .95 .99 .95 .99 8.1 .97 .99 .96 .99 1.83 .97 .99 .97 1.00 7.15 .97 1.00 .96 .99 1.81 96 Watson & Telleaen High vs Low E & N 23.77 .05 .97 .99 Positive Affect Model Full 11.95 .02 .98 1.00 11.82* Model High vs Low E & N 17.28 .03 ,98 1.00 Negative Affect Model Full 14.55 .03 •97 1.00 2.73 Model Note. N = 487. df for E & N model = 13; df for Full model = 10; df for x 2 Change = 3. *p<.01 Fit indices for the 16 structural equation model are summarized in Table 34. Both the E and N models and the Full models fit the data well. Of eight x2 change tests, only one was statistically significant. Table 35. Predicting Bipolar Affect Dimensions from the Five Factor Mode in the Chinese sample Regression Weights VAF by VAF by Affect Construct N E O A C N, E 0 , A, C Two Axes Pleasant vs - .44* .03 .00 .08 .02 22 1 Unpleasant (3.4) Activated vs .20* .16 .03 - . 0 5 .12 3 2 Deactivated (1.8) Thaver Energy vs - .25* .11 .05 - .02 .11 12 2 Tiredness (2.9) Tension vs .47* .09 - . 0 3 - .04 .04 19 0 Calmness (3.3) Larsen & Diener Activated - .30* .12 .04 .03 .10 16 2 Pleasant vs (3.1) Unactivated Unpleasant Activated .46* •04 - . 0 2 - .06 .01 20 1 Unpleasant vs (3.4) Unactivated Pleasant 97 Watson & Tellegen High vs Low Positive Affect High vs Low Negative Affect - .31* .46* .09 .05 - .02 - .07 - .01 .06 .02 .14 16 (3.1) 21 (3.3) 0 3 Note. N = 487. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. VAF = Variance Accounted For. Figures in parentheses are the standard errors. No standard errors were available for the VAF by 0,A,C because the VAF was computed as the difference between the VAF by N.E.O.A.C and the VAF by N,E. Significance level was set at .001 in order to achieve an overall alpha of less than .05 (by Bonferroni procedure, 40 regression weights x .001 = .04). Results of the eight Full models are summarized in Table 35. The variance explained Neuroticism and Extraversion ranged from 3% to 22%, with a mean of 16%; the additional variance explained by Openness, Agreeableness, and Conscientiousness yielded a mean of 1%. The pattern of regression weights demonstrated Neuroticism was the best predictor of all eight affect dimensions. 5.2.3 SINUSOID FUNCTIONS As did in the preceding chapter, a sine function was fitted to each of the five series of correlations. Of the five fitted functions, variance explained was 100% for Neuroticism, 97% for Extraversion, 67% for Openness, 97% for Agreeableness, and 97% for Conscientiousness; the mean was 92%. Results for Neuroticism and Openness are displayed in Figure 10. As in English and Spanish, the results lent support to the viability of the circumplex model of affect in Chinese. Affect constructs falling at different angles on the circumplex model are associated with the personality dimensions (outside variables) in a meaningful way - the sine function. *p_<.001 Figure 10. Curves Relating Affect Constructs to Neuroticism and Openness in the Chinese sample Angular Positions on the circumference of the Two-Dimensional Space (in degrees) I postpone the discussion of similarities and differences between the findings in the Chinese and other languages until Chapter 8. 99 CHAPTER 6 JAPANESE 6.1 STRUCTURE OF AFFECT In this section, I repeated the same analysis sequence stated in the Spanish and Chinese samples (Chapters 4 and 5). 6.1.1 A PRIORI SCORING 6.1.1.1 Test for Bipolarity Table 36 shows relevant statistics for six hypothesized bipolar opposites. All six pairs fared well in the test of bipolarity. In each pair, at least three out of six variates yielded positive skew. In all pairs, more cases fell in the lower triangle than in the upper triangle of the bivariate distribution. Finally, the estimated latent correlations ranged from - .52 to - .77 , all fell within the predicted range (-.47 to -1.00). In subsequent analyses, it is justified to combine these six bipolar opposites into bipolar dimensions. Table 36. Statistics for a Test of Bipolarity in the Japanese sample Proportion of Cases Format Variable Skew Hypothesized Skew Lower Upper 0 Opposite Triangle Triangle Two Axes ADJ Pleasant .08 Unpleasant 1.34 91 4 AGR Pleasant - .50 Unpleasant 1.01 79 9 DES Pleasant - . 0 3 Unpleasant .88 76 13 - .77 ADJ Activated .76 Deactivated .30 84 8 AGR Activated .53 Deactivated - .02 80 16 DES Activated .79 Deactivated - . 0 5 88 10 - .52 Thaver ADJ Energy .68 Tiredness .02 78 16 AGR Energy .18 Tiredness - . 0 2 73 16 DES Energy .50 Tiredness - . 0 3 73 21 - .67 ADJ Tension 1.47 Calmness - .27 92 4 AGR Tension .90 Calmness - .47 85 5 DES Tension 1.50 Calmness - . 4 0 89 8 - . 7 5 100 Larsen & Diener ADJ Activated Pleasant AGR Activated Pleasant DES Activated Pleasant ADJ Activated Unpleasant AGR Activated Unpleasant DES Activated Unpleasant Watson & Telleqen ADJ High Positive Affect AGR High Positive Affect DES High Positive Affect ADJ High Negative Affect AGR High Negative Affect DES High Negative Affect .79 Unactivated .35 Unpleasant .21 Unactivated .17 Unpleasant .74 Unactivated .19 Unpleasant 1.56 Unactivated - . 3 0 Pleasant .91 Unactivated - .57 Pleasant 1.23 Unactivated - .26 Pleasant 1.02 Low Positive .35 Affect .19 Low Positive .17 Affect .67 Low Positive .19 Affect 1.88 Low Negative - . 3 0 Affect .85 Low Negative - .57 Affect 1.37 Low Negative - .26 Affect 94 5 83 13 84 11 - .64 94 4 74 11 83 10 - .77 95 4 86 12 84 11 - .59 95 3 76 17 86 9 - . 7 3 Note. N = 450. Watson & Tellegen's Low Positive Affect and Low Negative Affect are defined by Larsen & Diener's Unactivated Unpleasant and Unactivated Pleasant respectively. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Lower triangle indicates the percentage of participants who fall below the diagonal in a bivariate distribution of the hypothesized opposites; upper triangle indicates the percentage of participants who fall above the diagonal. O indicates the latent correlation between the hypothesized opposites in a confirmatory factor analysis. All O coefficients are significant at .001 level. 6.1.1.2 Measurement Models To examine the performance of the four structural models in Japanese, I repeated the analysis sequence used in the preceding chapters. 6.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes I specified a confirmatory factor analysis with four latent constructs (Pleasant, Unpleasant, Activated, and Deactivated). As shown in Table 37, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did the comparison model: A%2 (6. N=450) = 634.40, p < .001, and RMSEA changed from .11 to .20. Table 37. Fit Indices for the Measurement Models in the Japanese sample Constructs Model Fit Indices • I2 df RMSEA AGFI CFI CMQ Hypothesized 194.62 30 .11 .83 .96 Model Comparison 829.02 36 .20 .56 .80 Model Thayer's Unipolar Hypothesized 185.93 30 .11 .84 .96 Constructs Model Comparison 688.10 36 .19 .60 .85 Model Larsen & Diener's Hypothesized 170.45 30 .10 .85 .97 Unipolar Constructs Model Comparison 797.19 36 .21 .54 .85 Model Watson & Tellegen's Hypothesized 161.13 5 .25 .58 .93 Unipolar Constructs Model Comparison 161.38 6 .23 .65 .93 Model Note. N =450 . Model estimates and descriptive statistics for the hypothesized model are given in Table 38. Cronbach alphas varied from .35 to .93. Factor loadings ranged from .61 to .97; all were statistically significant. Interfactor correlations indicated that the hypothesized bipolar opposites correlated negatively as expected. However, Activated also correlated highly with Unpleasant. Table 38. CMQ in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations 102 Factor Loadings Construct Format P U A D M SD a Pleasant ADJ .90** 2.76 1.16 .93 Pleasant AGR .97** 3.21 1.12 .94 Pleasant DES .75** 2.41 .82 .62 Unpleasant ADJ .93** 1.78 .94 .86 Unpleasant AGR .90** 1.93 1.08 .85 Unpleasant DES .90** 1.85 .89 .80 Activated ADJ .78** 2.18 .97 .73 Activated AGR .86** 2.36 .89 .58 Activated DES .61** 1.78 .63 .44 Deactivated ADJ .71** 2.54 .91 .56 Deactivated AGR .62** 2.64 .87 .35 Deactivated DES .85** 2.31 .66 .64 Interfactor Correlations Pleasant Unpleasant -.78** Activated -.31** .63** Deactivated .00 - .10 -.49** Note. N = 450. P = Pleasant, U = Unpleasant, A = Activated, and D = Deactivated. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p_ < .01 6.1.1.2.2 Thayer's Constructs I specified a confirmatory factor analysis with Thayer's (1989) four variables. As shown in Table 37, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did their comparison model: A%2 (6, N=450) = 502.17, e < 001, and RMSEA changed from .11 to .19. Model estimates and descriptive statistics for the hypothesized model are given in Table 39. Cronbach alphas ranged from .51 to .88. Factor loadings ranged from .76 to .95; all were statistically significant. The pattern among the four constructs was consistent with Thayer's 103 (1989) structure such that the hypothesized bipolar opposites correlated - .66 (Energy vs Tiredness) and - .75 (Tension vs Calmness), respectively. Table 39. Thayer's Unipolar Constructs in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Energy Tiredness Tension Calmness M SD a Energy Energy Energy ADJ AGR DES .87** .80** .94** 2.18 2.42 2.00 .90 .80 .80 .88 .51 .90 Tiredness ADJ • .88** 3.14 .79 .62 Tiredness AGR .86** 2.80 1.00 .67 Tiredness DES .88** 2.43 .83 .66 Tension ADJ .88** 1.61 .76 .85 Tension AGR .84** 1.90 .90 .82 Tension DES .90** 1.55 .75 .88 Calmness ADJ .82** 3.05 1.04 .86 Calmness AGR .95** 3.12 1.01 .88 Calmness DES Interfactor Correlations .76** 2.43 .70 .69 Energy Tiredness -.66** Tension - .06 .05 Calmness .18** - .10* -.75** — Note. N = 450. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p<.05 **p<.01 6.1.1.2.3 Larsen and Diener's Constructs I specified a confirmatory factor analysis with Larsen and Diener's (1992) four variables. As shown in Table 37, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than their comparison model: Ax 2 (6, N=450) = 626.74, p < .001, and RMSEA changed from .10 to .21. 104 Table 40. Larsen and Diener's Unipolar Constructs in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format AP UU AU UP M SD a Activated ADJ .84** 2.16 .84 .79 Pleasant Activated AGR .90** 2.57 .88 .73 Pleasant Activated DES .80** 1.78 .68 .65 Pleasant Unactivated ADJ .86** 2.37 .90 .81 Unpleasant Unactivated AGR .90** 2.51 1.01 .74 Unpleasant Unactivated DES .91** 2.25 .92 .74 Unpleasant Activated ADJ .88** 1.66 .81 .86 Unpleasant Activated AGR .92** 1.98 1.02 .77 Unpleasant Activated DES .90** 1.62 .74 .81 Unpleasant Unactivated ADJ 89** 3.11 1.07 .91 Pleasant Unactivated AGR 92** 3.28 1.08 .90 Pleasant Unactivated DES 92** 2.59 .91 .83 Pleasant Interfactor Correlations Activated — Pleasant Unactivated -.63** Unpleasant Activated -.30** .40** Unpleasant Unactivated .39** -.27** -.78** Pleasant Note. N = 450. AP = Activated Pleasant, UU = Unactivated Unpleasant, AU = Activated Unpleasant, and UP = Unactivated Pleasant. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p<.05 **p_<.01 Model estimates and descriptive statistics for the hypothesized model are given in Table 40. Factor loadings ranged from .80 to .92; all were statistically significant. The pattern among the four constructs was consistent with Larsen and Diener's (1992) structure such that the 105 hypothesized bipolar opposites correlated - . 6 3 (Activated Pleasant vs Unactivated Unpleasant) and - .78 (Activated Unpleasant vs Unactivated Pleasant), respectively. 6.1.1.2.4 Watson and Tellegen's Constructs I specified a confirmatory factor analysis with Watson and Tellegen's two variables. As shown in Table 37, both the hypothesized model and the comparison model fit the data poorly. Model estimates and descriptive statistics for the hypothesized model are given in Table 41. Factor loadings ranged from .75 to .97; all were statistically significant. The latent correlation between Positive Affect and Negative Affect was estimated to be .03. Table 41. Watson and Tellegen's Unipolar Constructs in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Positive Affect Negative Affect M SD a Positive Affect ADJ .75** 1.87 .68 .84 Positive Affect AGR .80** 2.50 .81 .85 Positive Affect DES .93** 1.79 .58 .82 Negative Affect ADJ .85** 1.47 .66 .89 Negative Affect AGR .92** 2.05 1.01 .91 Negative Affect DES .97** 1.55 .67 .90 Interfactor Correlations Positive Affect — Negative Affect .03 — Note. N = 450. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p_<.05 **p_<.01 To examine the role played by systematic errors, as was done before, Larsen and Diener's (1992) Unactivated Unpleasant and Unactivated Pleasant variables were used to construct the bipolar opposite of Watson and Tellegen's Positive and Negative Affect, respectively. I examined the confirmatory factor analysis for the newly revised model. The hypothesized model fit the data moderately well: x2(30, N=450) = 318.55, RMSEA = .15, AGFI = .73, CFI = .94. Factor loadings ranged from .84 to .91; all were statistically significant. The latent correlation between Positive Affect and its bipolar opposite was estimated to be - .58; and 106 the correlation between Negative Affect and its bipolar opposite was estimated to be - .73 . The correlation between Positive Affect and Negative Affect was estimated to - .03 , identical to what was obtained in the original model. A four-factor comparison model fit the data poorly: x2 (36, N=450) = 833.44, RMSEA = .22, AGFI = .51, CFI = .84. The hypothesized model fit the data significantly better than the comparison model: Ax 2 (6, N=450) = 514.89, p. < .001, and RMSEA changed from .15 to .22. Next, the hypothesized bipolar opposites in the revised model were examined for bipolarity. As noted in Table 36, for High vs Low Positive Affect, all six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated latent correlation of - .59 fell within the predicted range. For High vs Low Negative Affect, three out of six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated correlation between the hypothesized bipolar opposite was - . 7 3 , which fell within the predicted range. Therefore, it is justified to combine these variables into bipolar dimensions in subsequent analyses. 6.1.1.2.5 Interim Summary The just reported analyses revealed problems in two structural models. First, as in Chinese, the scales defining the four cornerstone constructs of the proposed two-dimensional space were problematic, as indicated by low factor loadings for the intended constructs and low reliability estimates. Revisions of these scales were discussed in the next section. Second, Watson and Tellegen's model did not fit the data well. After the systematic errors were removed, the model fit improved to a large extent. 6.1.2 REVISED SCORING The same revision sequence adopted in the preceding languages was repeated with the Japanese sample. 107 6.1.2.1 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes The result of a series of revisions was that the 10 of the 12 scales defining the four constructs were revised accordingly. (See Appendix VIII for the rescored scales.) The revised model fit the data moderately well: x2 (30, N=450) = 173.40, RMSEA = .10, AGFI = .85, CFI = .95. The comparison model (with all latent correlations set to .00) fit the data poorly: x2(36, N=450) = 709.73, RMSEA =19, AGFI = .61, CFI = .79. Indeed, the revised model fit the data significantly better than did the comparison model:: Ax 2 (6, N=450) = 536.33, p_ < .001, and RMSEA changed from .10 to .19. Model estimates for the revised model are given in Table 42. Factor loadings ranged from .41 to .96; all were statistically significant. The correlations of the bipolar opposites are relatively substantial in magnitude as compared with others which were all below .20 in magnitude. The revised CMQ scales were used in all subsequent analyses. Table 42. CMQ Revised in the Japanese sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Pleasant Unpleasant Activated Deactivated M SD Pleasant Pleasant Pleasant ADJ AGR DES .87** .96** .82** 2.83 3.27 2.58 1.23 1.13 1.01 Unpleasant Unpleasant Unpleasant ADJ AGR DES .90** .89** .88** 1.85 1.93 1.85 1.05 1.08 .89 Activated Activated Activated ADJ AGR DES .58** .58** .41** 2.17 2.55 1.67 1.03 1.09 .87 Deactivated Deactivated Deactivated ADJ AGR DES .65** .56** .91** 2.74 2.64 2.23 1.01 .87 .77 Pleasant Unpleasant Activated Deactivated Interfactor Correlations -.81** .18** .08 - .11* .06 -.72** 108 Note. N = 450. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p<.05 **p<.01 Next, the newly revised constructs were examined for their bipolarity. Of the 12 scales, all showed positive skew. In both pairs of hypothesized opposites, more cases fell in the lower triangle than in the upper triangle in their bivariate distribution. Finally, the correlation between the hypothesized bipolar opposites was estimated to be - .81 for Pleasant vs Unpleasant and - .72 for Activated vs Deactivated, respectively, in a confirmatory factor analysis with correlated error variance. To conclude, it is justified to combine these two scales into bipolar dimensions. For use in subsequent structural equation models, I specified a confirmatory factor model with two latent constructs, corresponding to the bipolar axes of Pleasant vs Unpleasant and Activated vs Deactivated, respectively. Each latent construct was indicated by the bipolar versions of its three response formats. The semantic differential scale of Pleasure was specified to load on the Pleasant vs Unpleasant construct; the semantic differential scale of Arousal was to load on the Activated vs Deactivated construct. The latent correlation between the two axes was fixed to .00. The hypothesized model fit the data moderately well: x 2 C 6 -N=450) = 111.86, RMSEA = . 12, AGFI = .86, CFI = .97. Factor loadings were all statistically significant and model estimates were used in subsequent analyses. 6.1.2.2 Scales defining the other three structural models Scales were examined for their internal consistency; all but one of Thayer's scale passed the .60 threshold. Reliability analysis suggested that dropping one item would increase the alpha by .30. The scale was revised accordingly and the measurement model for Thayer's structural model was re-examined. Results indicated that the factor loading associating the revised scale to the intended construct increased by .09. And the resulting interfactor correlations among Thayer's four constructs were almost identical to the those obtained with the a priori scoring. 109 Scales were also examined for their factor loadings in their measurement models; all scales passed the .70 threshold. 6.1.2.3 Relationships Among the Hypothesized Equivalents As did in the preceding languages, I specified a confirmatory factor analysis with 10 latent constructs (Thayer's Energy, Tiredness, Tension, and Calmness; Larsen and Diener's Activated Pleasant, Unactivated Unpleasant, Activated Unpleasant, and Unactivated Pleasant; Watson and Tellegen's Positive Affect and Negative Affect) to examine the relations among the hypothesized equivalents. This model fit the data well: %2(225, N=450) = 902.85, RMSEA = .08, AGFI = .58, CFI = .96. Estimated correlations between hypothesized equivalents are shown in. The correlations testing our hypothesized equivalences (highlighted in bold) were all above .87, with a mean of .94. Next, I compared the hypothesized model with a comparison model (a null hypothesis of no relation between the three structures), which fit the data poorly: %2(353, N=450) = 8221.79, RMSEA = .19, AGFI = .38, CFI = .59. The hypothesized model fit the data significantly better than did this comparison model: Ax2(128, N=450) = 7318.94, p_<001, and RMSEA changed from .08 to .19. CD to CD c ro a. ro —> CD SZ c to o 3 L_ oo c o O i— CO o CL 'c Z3 to "c CD CO _CD "CD f-T J C CO c o to "co T J c CO to" "L_ CD C CD b T J C CO c CD to CO _ l _to" CD >» CO x: h-co c o E < to c g CO o O T J Q) "co tu; -*—» to LU CO •^ r CD CD CL X I E CO CO h- to to o 3 l_ -*—' CO c o o to CD c CD b 08 c CD co to o 3 1— to c o O CO "L_ CD > » CO T J (D £5 co . > CO tl ro " CD ro c 0_ 13 "S ro ^ co ro co > CD <=5 T J ~ 0) c •JS ro ro co > co tj i> CO C L C C 3 =5 T J - ~ (D £ •s ro co co .> co o £ < 0-co CO CD c E CO O c g to c CD CO to CD c T J CD E? CD c LU « CO CO CO * « CD * * CO to o 3 L-^—' to c o o to "L-CD c CD b 03 c CD CO cn CM « * l O cn « CO CD * « LO * * CM cn * m CO * * o * 0 0 cn « LO * CO CM * * LO CO -C c c t j c c a> <0_=D=J<r3=D0_ to o 1— •*-» to c o O to "c & "CD h-oa c o CO I 00 * £ cn ^ cn 5- * r * * T cn CO * CO I CO * * 0 0 |g 0 0 • I •« 0 U CD CD :-E co CO CO O CD CD co c J D CO > ' 3 cr CD L_ CO CD CZ CD X ) T J CD N "to CD x: o CL >» x: to o CO o to c CD CD 5 "55 X I 8> CO T J o X ) CD CO O CO c o L-o o 0) CO o CO c CD CD CD X I CO to CD CM 3 u_ o VI Ql • CD * O -o LO o * E 11 _: VI o to « 111 6.1.3 PREDICTING AFFECT CONSTRUCTS BY THE REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES In this section, as was done in preceding chapters, I test the hypothesis that the Pleasant vs Unpleasant and Activated vs Deactivated axes predict most if not all of the variance in every construct of the other three structures. Results are summarized in Table 44. Table 44. Unipolar and Bipolar Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Japanese sample Regression Weights Fit Indices Pleasant- Activated- VAF Construct Unpleasant Deactivated x 2 RMSEA AGFI CFI Thayer Energy ,64 .59 76 251.21 .09 .89 .96 (2.5) Tiredness - .53 - .68 74 232.52 .09 .90 .96 (2.8) Tension - .65 .55 72 213.95 .08 .91 .96 (2.8) Calmness .72 - .54 81 319.41 .11 .86 .94 (2.4) Energy vs .64 .69 89 299.84 .10 .87 .95 Tiredness (1.6) Tension vs - .74 .58 88 326.27 .11 .86 .95 Calmness (1.7) Larsen & Diener Activated .67 .55 75 227.28 .08 .90 .96 Pleasant (2.8) Unactivated - .62 - . 6 3 79 212.36 .08 .91 .97 Unpleasant (2.4) Activated - .82 .35 80 235.98 .08 .90 .96 Unpleasant (2.2) Unactivated .82 - .41 83 268.06 .09 .89 .96 Pleasant (2.0) Activated .72 .65 94 241.37 .09 .90 .96 Pleasant vs (1.3) Unactivated Unpleasant Activated - .87 .41 92 297.33 .10 .87 .96 Unpleasant vs Unactivated Pleasant 112 Watson & Tellegen Positive Affect .43 .70 68 (3.4) 369.15 .11 .84 .93 Negative Affect - .79 .35 74 (2.4) 255.55 .09 .89 .96 High vs Low .61 .74 93 296.65 .10 .87 .95 Positive Affect (1.5) High vs Low - .87 .40 92 283.44 .09 .88 .96 Negative Affect (12) Note. N = 450. df for %2 = 58. VAF = Variance Accounted For. Figures in parentheses are the standard errors. All regression coefficients are significant at .001 level, which is equivalent to an overall alpha level of less than .05 (by Bonferroni procedure, 32 regression weights x .001 = .03). All unipolar and bipolar constructs were substantially explained by the two axes. For Thayer's unipolar constructs, the variance explained ranged from 72% to 81%, with a mean of 76%; for the two bipolar dimensions, the mean variance explained was 89%. For Larsen and Diener's unipolar constructs, the variance explained ranged from 75% to 83%, with a mean of 79%; for the two bipolar dimensions, the mean variance explained was 93%. For Watson and Tellegen's unipolar constructs, the mean variance explained was 71%; for the two bipolar dimensions, the mean variance explained was 93%. The pattern of relations between the exogenous variables and the endogenous variable was approximately as expected in Figure 2. As in Chinese, the newly revised scales defining Pleasant vs Unpleasant and Activated vs Deactivated axes explained most of the variance in constructs of other three structural models. The proposed two-dimensional space was shown to integrate other structural models meaningfully. 6.1.4 THE FULL TWO-DIMENSIONAL SPACE In this section, I examine how well the Japanese data conformed to a circumplex. A 14 x 14 correlation matrix was submitted to the maximum likelihood estimation using CIRCUM. The analysis converged on a solution in 12 iterations. Three free parameters were specified in the correlation function equation; additional free parameters did not improve the model fit. The final model had a total of 44 free parameters and 61 degrees of freedom. The data fit the model 113 moderately well: x 2 (61, NM450) = 355.45, RMSEA = .10. The results are shown in Figure 11. Noticeable differences occurred in all four quadrants. For instance, Thayer's Energy did not fall within the confidence interval of Watson and Tellegen's High Positive Affect; Thayer's Tension did not fall within the conference interval of Larsen and Diener's Activated Unpleasant or Watson and Tellegen's Negative Affect. Figure 11. A Circumplex Representation of 14 Unipolar Affect Constructs in the Japanese sample Thayer's Tension 128° (1227133°) Watson & Tellegen's Negative Affect 143° (1387148°) Larsen & Diener's Activated Unpleasant 145° (1407150°) Unpleasant 167° (1627173°) Larsen & Diener's Unactivated Unpleasant 222° (2177227°) Activated 76° (70782°) Thayer's Tiredness 227° (2227233°) Watson & Tellegen's Positive Affect 58° (54763°) Larsen & Diener's Activated Pleasant 43° (39747°) Thayer's Energy 42° (37747°) Pleasant 0° Larsen & Diener's Unactivated Pleasant 320° (3167324°) Deactivated 262° (2577267°) Thayer's Calmness 308° (3037312°) Consistent with the results in the preceding three language samples, results indicated that the hypothesized equivalent constructs in accordance with the 45° Hypothesis are not interchangeable, even though they are highly correlated. 114 6.2 RELATING AFFECT TO PERSONALITY I repeated the analysis sequence used in the preceding chapters to predict affect from personality. 6.2.1 DESCRIPTIVE STATISTICS FOR THE FFM First, the descriptive statistics of the FFM were examined. As shown in Table 45, the Cronbach alphas for the FFM dimensions ranged from .63 to .85, indicating that the measures are internally consistent. Further, the FFM dimensions are not independent but correlate with each other to a certain extent. These psychometric characteristics resemble to those reported in a previous administration in a similar sample (Shimonaka, 1996). Table 45. Descriptive Statistics for the Five Factor Model in the Japanese sample Correlation Coefficients FFM Dimensions 1 2 3 4 5 Mean SD a 1. Neuroticism 1.00 3.49 .67 .85 2. Extraversion -.26* 1.00 3.20 .59 .81 3. Openness to .09 - .04 1.00 3.53 .46 .63 Experience 4. Agreeableness -.24* .41* .04 1.00 3.38 .48 .69 5. Conscientiousness -.24* .26* - .04 .21* 1.00 3.21 .54 .76 Note. N = 450. Each mean score is an average of 12 items. Possible mean scores range from 1 to 5. *p_<.05 6.2.2 PREDICTING AFFECT FROM PERSONALITY 16 structural equations were examined and their fit indices are summarized in Table 46. Both the E and N models and the Full models fit the data well. Of the eight x2 change tests, only one was statistically significant, indicating that the Openness, Agreeableness, and Conscientiousness dimensions improved the overall model fit significantly. Table 46. Fit Indices for the Structural Equation Models in the Japanese sample 115 Fit Indices Endogenous Constructs Model RMSEA AGFI CFI x ' Change Two Axes Pleasant vs Unpleasant Activated vs Deactivated Thayer Energy vs Tiredness Tension vs Calmness E & N Model Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model Larsen & Diener Activated Pleasant E & N Model vs Unactivated Unpleasant Activated Unpleasant vs Unactivated Pleasant Watson & Tellegen High vs Low Positive Affect High vs Low Negative Affect Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model 26.35 14.23 32.06 29.16 14.43 7.56 10.73 4.60 16.47 10.20 15.44 8.08 21.80 15.84 18.71 10.57 .05 .03 .06 .06 .01 .00 .00 .00 .02 .00 .02 .00 .04 .04 .03 .01 .96 .97 .95 .94 .98 .99 .98 .99 .98 .97 .97 .97 .98 .99 1.00 .96 .96 1.00 1.00 1.00 1.00 1.00 .99 1.00 1.00 1.00 12.12* 2.90 6.87 6.13 .98 1.00 .98 1.00 6.27 .98 1.00 7.36 5.96 8.14 Note. N = 450. df for E & N model = 13; df for Full model = 10; df for x2 Change = 3. *E<.01 116 Results of the eight Full models are summarized in Table 47. The variance explained by Neuroticism and Extraversion ranged from 3% to 7%, with a mean of 5%; the additional variance explained by Openness, Agreeableness, and Conscientiousness yielded a mean of 2%. The pattern of regression weights demonstrated complex relations between personality and affect. Table 47. Predicting Bipolar Affect Dimensions from the Five Factor Model in the Japanese sample Regression Weights VAF by VAF by Affect Construct N E O A C N, E O, A, C Two Axes Pleasant vs - . 1 5 .01 .02 .17* .04 5 2 Unpleasant (2.0) Activated vs .01 .20* .05 - . 0 3 .08 4 1 Deactivated (2.2) Thayer Energy vs - .12 .14 .05 .05 .11 6 2 Tiredness (2.3) Tension vs .15 .08 - . 0 3 - .12 - . 0 3 3 1 Calmness (1.6) Larsen & Diener Activated Pleasant - .13 .14 .07 .05 .08 7 1 vs Unactivated (2.3) Unpleasant Activated .19* .04 - .01 - .14 .01 4 2 Unpleasant vs (2.0) Unactivated Pleasant Watson & Tellegen High vs Low - . 1 3 .13 .08 .04 .08 6 1 Positive Affect (2.2) High vs Low .19* .04 - . 0 3 - .14 - .02 4 2 Negative Affect (2.0) Note. N = 450. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. VAF = Variance Accounted For. Figures in parentheses are the standard errors. No standard errors were available for the VAF by 0,A,C because the VAF was computed as the difference between the VAF by N,E,0,A,C and the VAF by N,E. Significance level was set at .001 in order to achieve an overall alpha of less than .05 (by Bonferroni procedure, 40 regression weights x .001 = .04). * p < .001 117 6.2.3 SINUSOID FUNCTIONS Of the five fitted functions, variance explained was 97% for Neuroticism, 94% for Extraversion, 71 % for Openness, 98% for Agreeableness, and 94% for Conscientiousness; the mean was 91%. Results for Neuroticism and Openness are displayed in Figure 12. Results lent support to the viability of the circumplex model of affect in Japanese. Affect constructs falling at different angles on the circumplex model are associated with the personality dimensions (outside variables) in a meaningful way - the sine function. Figure 12. Curves Relating Affect Constructs to Neuroticism and Openness in the Japanese sample. c o "••-» TO CD CD zz 'c TO 2 0.50 -, 2 o.oo o -0.50 Openness (71%) 90 180 270 360 Angular Positions on the circumference of the Two-Dimensional Space (in degrees) I postpone the discussion of similarities and differences between the findings in the Japanese and other languages until Chapter 8. 118 CHAPTER 7 KOREAN 7.1 STRUCTURE OF AFFECT In this section, I repeated the same analysis sequence used in Spanish, Chinese, and Japanese (Chapters 4 to 6). 7.1.1 A PRIORI SCORING 7.1.1.1 Test for Bipolarity Table 48 shows relevant statistics for six hypothesized bipolar opposites. All six pairs fared well in the test of bipolarity. Of six pairs, at least three out of six variates yielded positive skew. More cases fell in the lower triangle than in the upper triangle of the bivariate distribution in all pairs. Finally, the estimated latent correlations ranged from - . 5 5 to - .88 , all fell within the predicted range (-.47 to -1.00). In subsequent analyses, it is justified to combine these six bipolar opposites into bipolar dimensions. Table 48. Statistics for a Test of Bipolarity in the Korean sample Proportion of Cases Format Variable Skew Hypothesized Skew Lower Upper Opposite Triangle Triangle Two Axes ADJ Pleasant - .01 Unpleasant 1.46 95 2 AGR Pleasant - .51 Unpleasant .83 82 5 DES Pleasant - . 2 3 Unpleasant .83 79 8 - .88 ADJ Activated .48 Deactivated .52 90 4 AGR Activated .03 Deactivated - . 1 5 67 25 DES Activated .20 Deactivated .11 81 16 - . 5 5 Thaver ADJ Energy .18 Tiredness .21 80 13 AGR Energy .20 Tiredness .13 75 11 DES Energy .35 Tiredness .19 81 11 - .74 ADJ Tension 1.03 Calmness .07 95 2 AGR Tension .51 Calmness - . 0 9 85 9 DES Tension .78 Calmness - . 1 6 81 13 - .71 119 Larsen & Diener ADJ Activated .52 Unactivated .45 96 2 Pleasant Unpleasant 8 AGR Activated .17 Unactivated .35 88 Pleasant Unpleasant DES Activated .60 Unactivated .39 90 6 Pleasant Unpleasant ADJ Activated 1.11 Unactivated .04 96 3 Unpleasant Pleasant 16 AGR Activated .51 Unactivated - . 0 8 70 Unpleasant Pleasant 11 DES Activated .74 Unactivated - . 0 5 86 Unpleasant Pleasant Watson & Tellegen ADJ High Positive .44 Low Positive .45 97 3 Affect Affect AGR High Positive .10 Low Positive .35 90 8 Affect Affect DES High Positive .35 Low Positive .39 90 8 Affect Affect ADJ High Negative 1.39 Low Negative .04 97 2 Affect Affect AGR High Negative .45 Low Negative - . 0 8 81 17 Affect Affect DES High Negative .85 Low Negative - . 0 5 90 7 Affect Affect - .69 - .76 Note. N = 365. Watson & Tellegen's Low Positive Affect and Low Negative Affect are defined by Larsen & Diener's Unactivated Unpleasant and Unactivated Pleasant respectively. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Lower triangle indicates the percentage of participants who fall below the diagonal in a bivariate distribution of the hypothesized opposites; upper triangle indicates the percentage of participants who fall above the diagonal. indicates the latent correlation between the hypothesized opposites in a confirmatory factor analysis. All <t> coefficients are significant at .001 level. 7.1.1.2 Measurement Models To examine how well various scales assess the four structural models, the same analysis sequence used in the preceding chapters was repeated. Table 49. Fit Indices for the Measurement Models in the Korean sample 120 Fit Indices Constructs Model x 2 df RMSEA AGFI CFI CMQ Hypothesized Model 187.37 30 .11 .82 .95 Comparison 767.88 36 •21 .53 .78 Model Thayer's Unipolar Constructs Hypothesized Model 107.27 30 .08 .88 .98 Comparison 564.58 36 .19 .60 .85 Model Larsen & Diener's Unipolar Constructs Hypothesized Model 93.71 30 .07 .90 .98 Comparison 691.17 36 .23 .47 .82 Model Watson & Tellegen's Unipolar Constructs Hypothesized Model 9.23 5 .05 .97 1.00 Comparison 61.69 6 .15 .83 .97 Model Note. N = 365 . 7.1.1.2.1 Pleasant vs Unpleasant and Activated vs Deactivated Axes I specified a confirmatory factor analysis with four latent constructs (Pleasant, Unpleasant, Activated, and Deactivated). As shown in Table 49, the hypothesized model fit the data well. In contrast, the comparison model fi the data poorly. Indeed, the hypothesized model fit the data significantly better than did the comparison model: Ax2 (6, N=365) = 580.51, p_ < .001, and RMSEA changed from .11 to .21. Model estimates and descriptive statistics for the hypothesized model are given in Table 50 . Cronbach alphas ranged from .43 to .93. Factor loadings ranged from .43 to .97; all were statistically significant. Interfactor correlations indicated that the hypothesized bipolar opposites correlated negatively as expected. 121 Table 50. CMQ in the Korean sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format P U A D M SD a Pleasant ADJ .88** 2.73 1.06 .93 Pleasant AGR .97** 3.12 .99 .92 Pleasant DES .86** 2.62 .83 .83 Unpleasant Unpleasant Unpleasant ADJ AGR DES .88** .95** .91** 1.73 2.08 1.75 .92 1.09 .76 .86 .89 .78 Activated ADJ .43** 2.32 .76 .62 Activated AGR .81** 2.58 .72 .43 Activated DES .75** 2.03 .57 .52 Deactivated ADJ .49** 2.37 .83 .58 Deactivated AGR .84** 2.88 .86 .69 Deactivated DES Interfactor Correlations .87** 2.28 .63 .77 Pleasant — Unpleasant Activated -.87** - . 0 3 .26** Deactivated .30** -.31** -.54** — Note. N = 365. P = Pleasant, U = Unpleasant, A = Activated, and D = Deactivated. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * P < .05 ** p. < .01 7.1.1.2.2 Thayer's Constructs I specified a confirmatory factor analysis with Thayer's (1989) four variables. As shown in Table 49, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than did their comparison model: A%2 (6, N=365) = 457.31, p < .001, and RMSEA changed from .08 to .19. Model estimates and descriptive statistics the hypothesized model are given in Table 51. Cronbach alphas ranged from .51 to .91. Factor loadings ranged from .67 to .94; all were statistically significant. The pattern among the four constructs was consistent with Thayer's 122 (1989) structure such that the hypothesized bipolar opposites correlated - .74 (Energy vs Tiredness) and - .71 (Tension vs Calmness), respectively. Table 51. Thayer's Unipolar Constructs in the Korean sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Energy Tiredness Tension Calmness M SD a Energy ADJ .85** 2.46 .92 .91 Energy AGR .86** 2.58 .79 .64 Energy DES .92** 2.12 .79 .92 Tiredness ADJ .87** 2.82 .86 .82 Tiredness AGR .91** 2.69 .90 .73 Tiredness DES .90** 2.19 .77 .75 Tension ADJ .84** 1.82 .76 .81 Tension AGR .91** 2.35 .87 .82 Tension DES .91** 1.89 .72 .81 Calmness ADJ .67** 2.68 .75 .51 Calmness AGR .94** 2.81 .75 .78 Calmness DES .82** 2.43 .69 .79 Interfactor Correlations Energy — Tiredness -.74** — Tension -.22** .19** Calmness .31** 2*|** 71** Note. N = 365. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p < .01 7.1.1.2.3 Larsen and Diener's Constructs I specified a confirmatory factor analysis with Larsen and Diener's (1992) four variables. As shown in Table 49, the hypothesized model fit the data well. In contrast, the comparison model fit the data poorly. Indeed, the hypothesized model fit the data significantly better than their comparison model: Ay2 (6, N=365) = 597.46, p. < .001, and RMSEA changed from .07 to .23. 123 Table 52. Larsen and Diener's Unipolar Constructs in the Korean sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format AP UU AU UP M SD a Activated ADJ .85** 2.23 .94 .90 Pleasant Activated AGR .91** 2.55 .81 .81 Pleasant Activated DES .83** 1.87 .67 .72 Pleasant Unactivated ADJ .80** 2.22 .83 .82 Unpleasant Unactivated AGR .91** 2.40 1.00 .86 Unpleasant Unactivated DES .90** 2.06 .87 .81 Unpleasant Activated ADJ .88** 1.79 .85 .88 Unpleasant Activated AGR .89** 2.41 .99 .79 Unpleasant Activated DES .90** 1.81 .82 .88 Unpleasant Unactivated ADJ .84** 2.75 .86 .81 Pleasant Unactivated AGR .92** 2.91 .89 .83 Pleasant Unactivated DES .84** 2.35 .72 .66 Pleasant Interfactor Correlations Activated — Pleasant Unactivated — . ~ Unpleasant .69** Activated — .65** Unpleasant .48** Unactivated .38** -.42** -.76** Pleasant - N o t e - N = 365. AP = Activated Pleasant, UU = Unactivated Unpleasant, AU = Activated Unpleasant, and UP = Unactivated Pleasant. ADJ = Adjective format, AGR = "Agree-Disagree" format, and DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. * p_ < .05 ** p_ < .01 Model estimates and descriptive statistics for the hypothesized model are given in Table 52. Cronbach alphas ranged from .66 to .90. Factor loadings ranged from .80 to .92; all were 124 statistically significant. The pattern among the four constructs was consistent with Larsen and Diener's (1992) structure such that the hypothesized bipolar opposites correlated - .69 (Activated Pleasant vs Unactivated Unpleasant) and - .76 (Activated Unpleasant vs Unactivated Pleasant), respectively. 7.1.1.2.4 Watson and Tellegen's Constructs I specified a confirmatory factor analysis with Watson and Tellegen's (1985) two variables. As shown in Table 49, the hypothesized model fit well. In contrast, the comparison model fit he the data poorly. Indeed, the hypothesized model fit the data significantly better than did their comparison model: Ax* (1, N=365) = 52.67, p. < .001, and RMSEA changed from .05 to.15. Model estimates and descriptive statistics for the hypothesized model are given in Table 53. Factor loadings ranged from .80 to .96; all were statistically significant. The latent correlation between Positive Affect and Negative Affect was estimated to be - .40 . Table 53. Watson and Tellegen's Unipolar Constructs in the Korean sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Positive Affect Negative Affect M SD a Positive Affect ADJ .80** 2.17 .67 .83 Positive Affect AGR .92** 2.50 .72 .85 Positive Affect DES 91** 1.95 .58 .86 Negative Affect ADJ .86** 1.61 .71 .89 Negative Affect AGR .89** 2.41 .90 .88 Negative Affect DES .96** 1.77 .72 .92 Interfactor Correlations Positive Affect Negative Affect „ -.40** Note. N = 365. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p<.05 **p<.01 125 As was done in the preceding chapters, Larsen and Diener's (1992) Unactivated Unpleasant and Unactivated Pleasant variables were used to construct the bipolar opposite of Watson and Tellegen's Positive and Negative Affect, respectively. The newly revised model fit the data well: x* (30, N=365) = 75.77, RMSEA = .06, AGFI = .92, CFI = .99. Factor loadings ranged from .79 to .95; all were statistically significant. The latent correlation between Positive Affect and its bipolar opposite was estimated to be - .67 ; and the correlation between Negative Affect and its bipolar opposite was estimated to be - .75 . The correlation between Positive Affect and Negative Affect was estimated to be - .40 , identical to the original model. A four-factor comparison model fit the data poorly: %2 (36, N=365) = 628.98, RMSEA = .21, AGFI = .51, CFI = .84. The hypothesized model fit the data significantly better than the comparison model:: A x 2 (6, N=365) = 553.21, p_ < .001, and RMSEA changed from .06 to .21. The hypothesized bipolar opposites in the revised model were examined for bipolarity. As shown in Table 48, for High vs Low Positive Affect, all six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated latent correlation of - .67 fell within the predicted range. For High vs Low Negative Affect, four out of six formats showed positive skew. More cases fell in the lower triangle than in the upper triangle in the bivariate distribution. The estimated correlation between the hypothesized bipolar opposite was - .74, which fell within the predicted range. Therefore, it is justified to combine these variables into bipolar dimensions in subsequent analyses. 7.1.1.2.5 Interim Summary As in Chinese and Japanese, the scales defining the four cornerstones constructs of the proposed two-dimensional space were problematic and their revisions were to be discussed in the next section. The other three structural models, Thayer's, Larsen and Diener's, and Watson and Tellegen's structural models fare well in model fit. 7.1.2 REVISED SCORING The same revision sequence in the preceding chapters was repeated with the Korean sample. 126 7.1.2.1 Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes The result of a series of revisions was that nine of 12 scales were revised. (See Appendix VIII for the rescored scales.) The measurement model on the four constructs were re-examined. The revised model fit the data moderately well: %2{30, N=365) = 89.73, RMSEA = .07, AGFI = .90, CFI = .98. The comparison model (with all latent correlations set to .00) fit the data poorly: x 2 (36, N=365) = 651.56, RMSEA = .18, AGFI = .64, CFI = .78. Indeed, the revised model fit the data significantly better than did the comparison model:: A x 2 (6, N=365) = 561.83, p < .001, and RMSEA changed from .07 to .18. Table 54. Revised CMQ in the Korean sample: Standardized Factor Loadings and Interfactor Correlations Factor Loadings Construct Format Pleasant Unpleasant Activated Deactivated M SD Pleasant ADJ .88** 2.73 1.06 Pleasant AGR .97** 3.12 .99 Pleasant DES .86** 2.62 .83 Unpleasant ADJ .86** 1.75 .95 Unpleasant AGR .92** 2.13 1.15 Unpleasant DES .92** 1.83 .86 Activated ADJ .54** 2.15 .79 Activated AGR .61** 2.65 .72 Activated DES .67** 2.03 .69 Deactivated ADJ .67** 2.52 .90 Deactivated AGR .71** 3.01 .99 Deactivated DES .76** 2.20 .70 Interfactor Correlations Pleasant Unpleasant -.88** Activated .11 .10 -Deactivated .07 - . 1 0 - .67 Note. N = 365. ADJ = Adjective format; AGR = "Agree-Disagree" format; DES = "Describes Me" format. Possible mean scores range from 1 to 5 for Adjective and "Agree-Disagree" formats; 1 to 4 for "Describes Me" format. *p<.05 **p<.01 Model estimates for the revised model are given in Table 54. Factor loadings ranged from .54 to .97; all were statistically significant. The correlations of the bipolar opposites are 127 relatively substantial in magnitude as compared with others which were all below .20 in magnitude. The revised CMQ scales were used in all subsequent analyses. Next, the newly revised two pairs of hypothesized opposites were examined for their bipolarity. 10 out of 12 scales showed positive skew. In all 12 scales, more cases fell in the lower triangle than in the upper triangle in their bivariate distribution. Finally, the correlation between the hypothesized bipolar opposites was estimated to be - .89 for Pleasant vs Unpleasant and - .69 for Activated vs Deactivated, each was assessed in a confirmatory factor analysis with correlated error variance. Again, it is justified to combine these two scales into bipolar dimension in subsequent analyses. For use in subsequent structural equation models, I specified a confirmatory factor model with two latent constructs, corresponding to the bipolar axes of Pleasant vs Unpleasant and Activated vs Deactivated, respectively. Each latent construct was indicated by the bipolar versions of its three response formats. The semantic differential scale of Pleasure was specified to load on the Pleasant vs Unpleasant construct; the semantic differential scale of Arousal was to load on the Activated vs Deactivated construct. The latent correlation between the two axes was fixed to .00. The hypothesized model fit the data moderately well: %2 (16, N=365) = 93.70, RMSEA = .12, AGFI = .86, CFI = .97. Factor loadings were all statistically significant and model estimates were used in subsequent analyses. 7.1.2.2 Scales defining the other three structural models Scales were examined for their reliability estimates; all but the adjective scale defining Thayer's Calmness (.51), passed the .60 threshold. Scales were also examined for their factor loadings in their measurement models; all but the same scale defining Thayer's Calmness (.67) passed the .70 threshold. Item-total correlations suggested that dropping one item in the adjective scale would increase the alpha by .15. With the revised scale, the measurement model for Thayer's structural model was re-examined. Results indicated that fit indices were worsened and the 128 respective factor loading changed from .67 to .50. Therefore, it was decided to keep the a prior scoring for this scale. 7.1.3 RELA TIONSHIPS AMONG THE HYPOTHESIZED EQUIVALENTS As was done in the preceding languages, I specified a confirmatory factor analysis with 10 latent constructs (Thayer's Energy, Tiredness, Tension, and Calmness; Larsen and Diener's Activated Pleasant, Unactivated Unpleasant, Activated Unpleasant, and Unactivated Pleasant; Watson and Tellegen's Positive Affect and Negative Affect) to examine the relations among the hypothesized equivalents. This model fit the data well: %2 (225, N=365) = 594.49, RMSEA = .06, AGFI = .16, CFI = .98. Estimated correlations between hypothesized equivalents are shown in Table 55. The correlations testing our hypothesized equivalences (highlighted in bold) were all above .85, with a mean of .93. Next, I compared the hypothesized model with a comparison model (a null hypothesis of no relation between the three structures), which fit the data poorly: %2 (353, N=365) = 6290.75, RMSEA = .18, AGFI = .41, CFI = .60. The hypothesized model fit the data significantly better than did this comparison model: Ax2 (128, N=365) = 5696.26, p_ < .001, and RMSEA changed from .06 to .18. c CO a> o CD J C ••-» c co "5 3 i_ W c o O CO o Q . C =3 CO "c CD CJ) J D "55 1-TJ c CD c o to 00 TJ c CD co" CD C CD b TJ C CD C CD CO L_ CO _J co" CD >^  CO J C O ) c o E < c g CO c 1— o O TJ £ ro E •*-« CO UJ in" in CD CD Q. J Q E ro ro 1- CO co o 3 i _ -*—• CO c o O _co CD c CD Q 08 c CD CO L_ CD CO o 3 L— co c o O CO "k-CD >^ CO TJ CD % C JS CO > CO m CD CD C 0. D TJ £ tu ro -Si to co ro > J D o 9-TJ ~ . > CO ro o. c c =3 = TJ *s CD £ - J - ; (D ro to > ro < Q-co co CD c E ro O c g to c CD H co co CD c TJ CD E° CD C UJ CO CO c o o CO CD c CD b c CD CO I— ro co « * * * CD CD « co * * CO CO CM cn * * * CO CO « « CO * * « 00 m CO O ) CD CM * O TJ 0) CO > TJ CD • CO > *- TJ C CD CO T3 CO * i CO CD CO ™ c ~ .2 CO w p 4? 5 Jj> c3 jo C O o 3 L— •*-» CO c o O CO "c CDl Sf "CD °3 c : o CO O J ^ C C O C C J C D ro < Q - D 3 < D 3 0 _ l> * * 00 1 . • I * * CD *-CD CM '. CD * * CO CO 00 CM CO oo CD * * CM * 83 a, * * ? a> oo m r o O CD l i !"=! co CO D ) O CD CL Z CD C a) co > ' 3 CT CD 1— CO CD c CD X I TJ CD N 'co CD J C o Q. >< JC CO J D CO o co c CD CD CD X I ro TJ O X ) C CD CO O CO c o *"•—» J O 1— o o J D CO O CO c CD CD CD X I ro to 22 3 CD » O z CM T— CD 3 D) O LL VI M— O QI « CD *TJ O E ral in 3 0 •*-> on VI 1— QJ co * 7.1.4 PREDICTING AFFECT CONSTRUCTS BY REVISED PLEASANT VS UNPLEASANT AND ACTIVATED VS DEACTIVATED AXES To test the hypothesis that the Pleasant vs Unpleasant and Activated vs Deactivated axes predict most if not all the variance in every construct of the other three structures, I repeated the analysis sequence used in the preceding chapters. Table 56. Unipolar and Bipolar Affect Constructs explained by the Revised Pleasant vs Unpleasant and Activated vs Deactivated Axes in the Korean sample Construct Regression Weights VAF Fit Indices Pleasant-Unpleasant Activated-Deactivated x2 RMSEA AGFI CFI Thaver Energy .69 .59 83 235.30 .10 .87 .95 (2.6) Tiredness - .48 - .61 60 206.37 .08 .89 .96 (3.8) Tension - .74 .47 77 182.92 .08 .90 .97 (2.7) Calmness .70 - .50 75 188.46 .08 .89 .96 (3.1) Energy vs .62 .65 81 230.62 .09 .88 .96 Tiredness (2.5) Tension vs - .78 .53 89 253.95 .10 .86 .95 Calmness (1.8) Larsen & Diener Activated .67 .55 75 207.54 .09 .89 .96 Pleasant (3.1) Unactivated - .71 - .47 72 185.97 .08 .90 .97 Unpleasant (3.1) Activated - . 9 3 .16 88 200.56 .09 .89 .96 Unpleasant d-8) Unactivated .80 - .47 86 181.43 .08 .90 .97 Pleasant (2.2) 131 Activated .76 .56 88 253.28 .09 .88 .95 Pleasant vs (1.9) Unactivated Unpleasant Activated - .92 .33 95 262.69 .11 .86 .95 Unpleasant vs (1.1) Unactivated Pleasant Watson & Telleaen Positive Affect .63 .54 68 189.75 .08 .90 .96 (3.3) Negative Affect - .91 .20 86 243.15 .10 .87 .95 (1.8) High vs Low .74 .55 85 272.72 .10 .87 .95 Positive Affect (2.0) High vs Low - .91 .36 96 248.71 .10 .87 .96 Negative Affect (1.1) Note. N = 365. df for x* = 58. VAF = Variance Accounted For. Figures in parentheses are the standard errors. All regression coefficients are significant at .001 level, which is equivalent to an overall alpha level of less than .05 (by Bonferroni procedure, 32 regression weights x .001 = .03). Results of the 16 structural equations are summarized in Table 56. All unipolar constructs were substantially explained by the two axes. For Thayer's unipolar constructs, the variance explained ranged from 60% to 83%, with a mean of 74%; for the bipolar dimensions, the mean variance explained was 85%. For Larsen and Diener's unipolar constructs, the variance explained ranged from 72% to 88%, with a mean of 80%; for the bipolar dimensions, the mean variance explained was 92%. For Watson and Tellegen's unipolar constructs, the mean variance explained was 77%; for the bipolar dimensions, the mean variance explained was 91%. The pattern of relations between the exogenous variables and the endogenous variable was approximately as expected in Figure 2. To sum up, results showed that the newly revised Pleasant vs Unpleasant and Activated vs Deactivated axes were able to explain most of the variance in constructs of other structural models. As in Spanish, Japanese, and Korean, all four structures were able to fit comfortably within the proposed two-dimensional space. 132 7.1.5 THE FULL TWO-DIMENSIONAL SPACE In this section, I examine how well the Korean data conformed to a circumplex. A 14 x 14 correlation matrix was submitted to the maximum likelihood estimation using CIRCUM. The analysis converged on a solution in 22 iterations. Four free parameters were specified in the correlation function equation; additional free parameters did not improve the model fit. The final model had a total of 45 free parameters and 60 degrees of freedom. The data fit the model moderately well: %2 (60, N=365) = 442.22, RMSEA = .13. The results are shown in Figure 13. CIRCUM estimates the angle within the circle for each variable, as well as a 95% confidence interval for that angle. Pleasant was fixed at 0°. Scales that, according to the 45° hypothesis, would be identical differed by anywhere up to 28° from each other. Noticeable differences occurred in two quadrants. Thayer's Tension did not fall within the confidence interval of Watson and Tellegen's Negative Affect or Larsen and Diener's Activated Unpleasant; Thayer's Tiredness did not fall within the confidence interval of Larsen and Diener's Unactivated unpleasant. Consistent with the results in the preceding languages, results indicated that the hypothesized equivalent constructs in accordance with the 45° Hypothesis are not interchangeable, even though they are highly correlated. 133 Figure 13. A Circumplex Representation of 14 Unipolar Affect Constructs in the Korean sample Thayer's Tension 133° (1287138°) Activated 85° (80790°) Watson & Tellegen's Negative Affect 157° (1527162°) Larsen & Diener's Activated Unpleasant 161° (1567165°) Unpleasant 168° (1637173°) Larsen & Diener's Unactivated Unpleasant 218° (2137223°) Thayer's Tiredness 230° (2247236°) Watson & Tellegen's Positive Affect 46° (42750°) Thayer's Energy 43° (39747°) Larsen & Diener's Activated Pleasant 43° (39747°) Pleasant Larsen & Diener's Unactivated Pleasant 316° (3127320°) Thayer's Calmness 308° (3037313°) Deactivated 273° (2687278°) 7.2 RELATING AFFECT TO PERSONALITY In this section, I repeated the analysis sequence used in the preceding chapters to relate the whole spectrum of affect constructs to the FFM dimensions. 7.2.1 DESCRIPTIVE STATISTICS FOR THE FFM First, the descriptive statistics of the FFM were examined. As shown in Table 58, the Cronbach alphas for the FFM dimensions ranged from .67 to .85, indicating that the measures are internally consistent. Further, the FFM dimensions are not independent but correlate with each other to a certain extent. These psychometric characteristics resemble to those found in a previous administration in a similar sample (C. Ahn, personal communication, October 14, 1998). Table 57. Descriptive Statistics for the Five Factor Model in the Korean sample 134 Correlation Coefficients FFM Dimensions 1 2 3 4 5 Mean SD a 1. Neuroticism 1.00 3.28 .63 .84 2. Extraversion - .45* 1.00 3.40 .61 .85 3. Openness to - .07 .21* 1.00 3.30 .48 .67 Experience 4. Agreeableness - .22* .29* .19* 1.00 3.48 .48 .72 5. Conscientiousness - .32* .27* .08* .03 1.00 3.46 .58 .85 Note. N = 365. Each mean score is an average of 12 items. Possible mean scores range from 1 to 5. * fi < .05 7.2.2 PREDICTING AFFECT FROM PERSONALITY As did in the preceding chapters, I examined 16 structural models to explore the predictive utility of FFM on bipolar affect dimensions. Fit indices for the 16 structural equation model are summarized in Table 58. Both the E and N models and the Full models fit the data well. None of the eight %2 change tests were statistically significant. 135 Table 58. Fit Indices for the Structural Equation Models in the Korean sample Fit Indices Endogenous Constructs Model X' RMSEA AGFI CFI X Change Two Axes Pleasant vs Unpleasant Activated vs Deactivated Thayer Energy vs Tiredness Tension vs Calmness Larsen & Diener Activated Pleasant vs Unactivated Unpleasant Activated Unpleasant vs Unactivated Pleasant Watson & Tellegen High vs Low Positive Affect High vs Low Negative Affect E & N Model 20.48 Full Model 10.36 E & N Model 13.90 Full Model 13.15 E & N Model 12.76 Full Model 10.55 E & N Full Model Full Model E & N Model Full Model E & N Model Full Model E & N Model Full Model 24.09 15.38 E & N Model 26.86 23.28 18.09 9.00 22.28 17.80 17.04 8.75 .04 .01 .01 .03 .00 .01 .05 .04 .05 .06 .03 .00 .04 .05 .03 .00 .96 .98 .97 .97 .98 .97 .96 .96 .95 .95 .97 .98 .96 .96 .97 .98 1.00 1.00 1.00 .99 1.00 1.00 .99 1.00 .99 .99 1.00 1.00 .99 .99 1.00 1.00 10.12 .75 2.21 8.71 3.58 9.09 4.48 8.29 Note. N = 365. df for E & N model = 13; df for Full model = 10; df for x 2 Change = 3. *p_<.01 136 Results of the eight Full models are summarized in Table 59. The variance explained by Neuroticism and Extraversion ranged from 1% to 12%, with a mean of 10%; the additional variance explained by Openness, Agreeableness, and Conscientiousness yielded a mean of 2%. The pattern of regression weights indicated that all FFM dimensions were differentially related to the affect dimensions. Table 59. Predicting Bipolar Affect Dimensions from the Five Factor Model in the Korean sample Regression Weights VAF by VAF by Affect Construct N E O A C N, E O, A, C Two Axes Pleasant vs - .16 .16 - .06 .13 .12 12 2 Unpleasant (3.2) Activated vs - . 0 3 .09 .05 - .00 - .02 1 0 Deactivated (1.2) Thayer Energy vs - .16 .17 - .02 .07 .05 10 1 Tiredness (3.1) Tension vs .16 - .12 .10 - .12 - . 0 8 8 3 Calmness (2.8) Larsen & Diener Activated - .17 .18 .06 .07 .03 11 1 Pleasant vs (3.2) Unactivated Unpleasant Activated .17 - .16 .08 - .14 - .07 11 2 Unpleasant vs (3.1) Unactivated Pleasant Watson & Tellegen High vs Low - . 1 8 .16 .04 .08 .06 12 1 Positive Affect (3.3) High vs Low .17 - . 1 5 .08 - . 1 3 - . 0 8 11 2 Negative Affect (3.1) Note. N = 365. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. VAF = Variance Accounted For. Figures in parentheses are the standard errors. No standard errors were available for the VAF by 0,A,C because the VAF was computed as the difference between the VAF by N,E,0,A,C and the VAF by N,E. Significance level was set at .001 in order to achieve an overall alpha of less than .05 (by Bonferroni procedure, 40 regression weights x .001 = .04). *p<.001 137 7.2.3 SINUSOID FUNCTIONS Of the five fitted functions, variance explained was 98% for Neuroticism, 98% for Extraversion, 77% for Openness, 98% for Agreeableness, and 94% for Conscientiousness; the mean was 93%. Results for Agreeableness and Openness are displayed in Figure 14. Results lent support to the viability of the circumplex model of affect in Korean. Affect constructs falling at different angles on the circumplex model are associated with the personality dimensions (outside variables) in a meaningful way - the sine function. Figure 14. Curves Relating Affect Constructs to Agreeableness and Openness in the Korean sample <o c o m S> i _ o O o cu TJ -»-* C O) CO 0.50 ^ 0.00 4 -0.50 Openness (77%) Agreeableness (98%) 90 180 270 360 Angular Positions on the circumference of the Two-Dimensional Space (in degrees) Similarities and differences between Korean and other languages will be discussed in the following two chapters. 138 CHAPTER 8 CROSS-LANGUAGE SIMILARITIES AND DIFFERENCES What are the pan-cultural aspects of affect? Are there any consistent patterns of associations between these aspects and personality? The present dissertation sought to explore these two research questions in five different languages. Past research suggested that a two-dimensional space defined by Pleasant vs Unpleasant and Activated vs Deactivated axes is a viable structure to integrate constructs of other structural models of affect. These two axes explained most if not all the reliable variance in the affect constructs. Constructs which are 180° apart were also found to be bipolar opposites. Most important of all, the constructs fell at different angles on the two-dimensional space and the data conformed to a circumplex model of affect. Unfortunately, research findings are limited to primarily English-speaking populations. To what extent is this two-dimensional space restricted to English? How well is this space generalizable to other languages? In this dissertation, I assessed the generalizability of this two-dimensional space in Spanish, Chinese, Japanese, and Korean, in addition to English. Past research on personality and affect suggested that Neuroticism and Extraversion were the two most reliable predictors of affective states. Findings are not without problems. Both personality and affect were narrowly defined. There is substantial scale content overlap between the two domains. Trait rating instructions were employed in both domains. To overcome the foregoing problems, the present dissertation defined affect as the full spectrum of affect constructs which were shown to fit nicely within the proposed two-dimensional space and personality as the Five Factor Model. Personality was used to predict momentary affective ratings for a thin slice of time. More important, to explore if there exist consistent patterns of relations between affect and personality across languages, I examined how momentary affective ratings were predicted by enduring characteristics of personality in all five languages. 139 8.1 THE TWO-DIMENSIONAL SPACE 8.1.1 REVISED MODELS The proposed integrated structure was comprised of four cornerstone constructs -Pleasant, Unpleasant, Activated, and Deactivated. In each language, I specified a confirmatory factor analysis with four latent constructs to examine how well the scales assessed the four constructs and how well the proposed structure was supported. For interpretation purpose, I used RMSEA to assess how well the scales worked. Table 60. RMSEA for the CMQ Constructs in five language samples RMSEA Language A Priori Model Revised Model English .09 (.21) — Spanish .08 (.21) .05 (.18) Chinese .13 (.26) .12 (.19) Japanese .11 (.20) .10 (.19) Korean .11 (.21) .07 (.18) Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). Numbers in parentheses are the RMSEA for the comparison model where all latent correlations were fixed to .00. As shown in Table 60, when the a priori scoring (based on English) was used, all four CMQ structural models fit the data moderately well: All RMSEA were below .14. However, a careful examination of the interfactor correlations among the four CMQ constructs revealed that, except for Spanish, the pattern of relations among the latent constructs were not that expected of the proposed two-dimensional space. To illustrate, let's take the Korean data as an example. Although the hypothesized model fit the data moderately well and that the hypothesized bipolar opposites (Pleasant vs Unpleasant, Activated vs Deactivated) had substantial negative correlations, Deactivated correlated highly with both Pleasant and Unpleasant. In the proposed 140 two-dimensional space, only high negative correlations were expected of the hypothesized bipolar opposites. Other correlations should be as minimal as possible. In the four languages other than English, scales defining the CMQ constructs were revised accordingly and the corresponding indices of fit improved. The RMSEA for the newly revised models ranged from .05 to .12. In each of the four languages, the four cornerstone constructs were related in the way expected of the proposed structure. To ensure the best approximation of the proposed two-dimensional space in each language, the CMQ scales were revised with the consequence of improved model fit and expected relations among the constructs. All subsequent analyses were based on the revised CMQ models. 8.1.2 A MULTI-SAMPLE ANALYSIS ON THE RELATIONS AMONG PLEASANT, UNPLEASANT, ACTIVATED, AND DEACTIVATED As a consequence of the revisions made in each of the four languages, the resulting revised models looked very similar to the one in English. As shown in Chapters 3 to 7, the estimated interfactor correlations for the four defining constructs (Pleasant, Unpleasant Activated, and Deactivated) varied only slightly from one sample to the other. Because these constructs are the cornerstones of the two-dimensional space of our proposed integrated structural model, it would be desirable to spell out the relations among the four cornerstone constructs based on the five samples. Those estimates should give the best approximations to how well the proposed structure was supported across languages. To obtain the interfactor correlations based on all five languages, I examined a multi-sample confirmatory factor analysis using data from all five samples. The original CMQ in English, and the revised CMQ in other four languages were used in the present analysis. In this analysis, I specified a factor model with four latent constructs (Pleasant, Unpleasant, Activated, and Deactivated) for each language. The parameter estimates for factor loadings and error variance were estimated in each language; the interfactor correlations among the four constructs, however, were constrained to be equivalent across the five samples. 141 The hypothesized model fit the data well: %2 (174, Ns =535,233,487,450,365) = 805.12, and RMSEA = .09. Factor loadings between the response formats and their intended constructs (Pleasant, Unpleasant, Activated, and Deactivated) were all statistically significant in all five languages. The resulting interfactor correlation matrix is given in Table 61. High negative correlations were reported between the hypothesized bipolar opposites: Pleasant and Unpleasant correlated - .88 ; Activated and Deactivated correlated - .69 . All other correlations were below .10 in magnitude. To conclude, results from the multi-sample analysis indicated that in each language, the best approximations to the vertical and horizontal axes of the proposed two-dimensional space were available. Next, I turn to examining how well the proposed structure integrate other affect constructs. Table 61. Interfactor Correlations Among Pleasant, Unpleasant, Activated, and Deactivated in a Multi-sample Confirmatory Factor Analysis Interfactor Correlations Pleasant Unpleasant Activated Deactivated Pleasant Unpleasant - .88* Activated .09 .09 Deactivated - . 0 3 .00 - .69* Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). *p.<.001 8.2 INTEGRATION HYPOTHESIS 8.2.1 A CROSS-LANGUAGE COMPARISON OF VARIANCE EXPLAINED BY TWO BIPOLAR AXES One of the major hypotheses in the present dissertation was that the proposed two-dimensional space was able to integrate affect constructs of the other three structural models 142 shown in Figure 1. In all five languages, overall, the integration hypothesis received substantial support. To examine the variance explained by the two axes, in each language, I examined 16 structural equation models. Results are organized by affect constructs in Table 62. Table 62. Variance Explained by the two bipolar axes: A Cross-Language Comparison VAF by the two bipolar axes Constructs ENG SPA CHI JPN KOR Mean Thaver Energy 81 82 77 76 83 80 (1.9) (2.9) (2.5) (2.5) (2.6) Tiredness 60 63 64 74 60 64 (2.8) (4.4) (3.1) (2.8) (3.8) Tension 69 79 82 72 77 76 (2.3) (2.9) (1.8) (2.8) (2.7) Calmness 83 77 86 81 75 80 (2.0) (3.6) (3.2) (2.4) (3.1) Energy vs Tiredness 78 83 79 89 81 82 (1.9) (2.7) (2.2) (1.6) (2.5) Tension vs Calmness 84 87 91 88 89 88 (1.5) (2.3) (1.5) (1.7) (1.8) Larsen & Diener Activated Pleasant 69 73 86 75 75 76 (2.5) (3.5) (1.8) (2.8) (3.1) Unactivated Unpleasant 69 80 72 79 72 74 (2.5) (3.1) (2.5) (2.4) (3.1) Activated Unpleasant 77 83 86 80 88 83 (2.0) (2.5) (1.5) (2.2) (1.8) Unactivated Pleasant 78 78 86 83 86 82 (1.9) (3.0) (2.1) (2.0) (2.2) Activated Pleasant vs 83 93 89 94 88 89 Unactivated Unpleasant (1.6) (1.7) (1.4) (1.3) (1.9) Activated Unpleasant vs 85 86 93 92 95 90 Unactivated Pleasant (1.4) (2.0) (1.1) (1.2) (11) 143 Watson & Tellegen Positive Affect 79 (1.9) 78 (1.8) 84 (1.6) 86 (1.3) 79 (3.0) 84 (2.2) 92 (1.8) 88 (1.8) 78 (2.2) 83 (1.6) 84 (1.6) 93 (1.1) 68 (3.4) 74 (2.4) 93 68 (3.3) 86 (1.8) 85 (2.0) 96 (1.1) 74 Negative Affect 81 High vs Low Positive Affect 88 (1.5) High vs Low Negative Affect 92 (1.2) 91 Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). ENG = English, SPA = SPANISH, CHI = Chinese, JPN = Japanese, and KOR = Korean. VAF = Variance Accounted For. Figures in parentheses are the standard errors. As shown in Table 62, both unipolar and bipolar constructs were substantially explained by the two bipolar axes. For unipolar constructs, the mean variance explained ranged from 64% to 83%, with a mean of 77%. For bipolar dimensions, the mean variance explained ranged from 82% to 91%, with a mean of 88%. To conclude, the two bipolar axes in all five languages were able to explained most if not all the reliable variance in constructs of other structural models. The Integration Hypothesis received initial support in all five samples. 8.2.2 MULTI-SAMPLE ANALYSES ON THE VARIANCE EXPLAINED BY THE TWO BIPOLAR AXES Results in the preceding section showed that most of the reliable variance in other affect constructs were explained by the two bipolar axes. And results for bipolar affect dimensions are cleaner. In this section, I conducted six multi-sample structural equation models to obtain the best estimates of the variance explained by the two axes in six bipolar affect dimensions, across five different languages. First, to obtain the best estimates to define each bipolar axis based on the five languages, I specified a multi-sample confirmatory factor analysis with two latent constructs, each corresponding to one axes. Each latent construct was indicated by its three bipolar response formats. The semantic differential scale of Pleasure was specified to load on the 144 Pleasant vs Unpleasant axis; the semantic differential scale of Arousal was specified to load on the Activated vs Deactivated axis. The correlation between the two axes was fixed to .00. Estimates for factor loadings and error variance were constrained to be equivalent across all five samples. The hypothesized model fit the data moderately well: : x2 (127, Ns =535,233,487,450,365) = 902.58, and RMSEA =.12. The parameter estimates for factor loadings and error variance were used in the following structural equation models. A separate multi-sample structural equation model was analyzed for each bipolar affect dimension (Thayer's Energy vs Tiredness and Tension vs Calmness; Larsen & Diener's Activated Pleasant vs Unactivated Unpleasant and Activated Unpleasant vs Unactivated Pleasant; Watson and Tellegen's High vs Low Positive Affect and High vs Low Negative Affect). Two - Pleasant vs Unpleasant and Activated vs Deactivated - were considered as exogenous. Each was indicated by the three bipolar response formats and the semantic differential scale. Factor loadings and error variances on the exogenous side were fixed to the estimates in the preceding multi-sample confirmatory factor analysis. One of the bipolar dimensions was considered as endogenous. The factor loadings and error variance of the endogenous variable were estimated in each language in each analysis. The regression weights of the endogenous constructs on the exogenous constructs and the disturbance term were constrained to be equivalent across all samples. Results for the six structural equation models are summarized Table 63. As noted in individual samples before, all bipolar dimensions were substantially explained by the two axes. Variance explained ranged from 82% to 91%, with a mean of 88%. The pattern of relations between the exogenous variables and the endogenous variables was approximately as expected in Figure 2. 145 Table 63. Bipolar Affect Dimensions explained by the Pleasant vs Unpleasant and Activated vs Deactivated Axes in Six Multi-sample Structural Equation Models Regression Weights Fit Indices Construct Pleasant-Unpleasant Activated-Deactivated VAF x2 RMSEA Thaver 1657.94 .10 Energy vs Tiredness .57 .70 82 (1.0) Tension vs Calmness -•73 .59 88 1892.49 .11 (•8) Larsen & Diener Activated Pleasant .74 .59 89 1744.88 .11 vs Unactivated (.7) Unpleasant 1924.63 .12 Activated Unpleasant - .86 .40 90 vs Unactivated (•6) Pleasant Watson & Telleaen High vs Low Positive .70 .62 87 1937.12 .11 Affect (.8) High vs Low Negative - .87 .39 91 1880.30 .11 Affect (•6) Note: Ns =535,233,487,450,365. df for %2 = 298. VAF = Variance Accounted For. Figures in parentheses are the standard errors. All regression coefficients are significant at .001 level. A qualification has to be added to the preceding results. Although the variance explained was very substantial, the fit indices were not very good. There are two possible explanations for the high levels of fit indices. First, there may be other substantive factors (perhaps interaction of the two exogenous variables), in addition to Pleasant vs Unpleasant and Activated vs Deactivated constructs, which explain the remaining variance. Second, although the best available analysis technique was used to control systematic and random errors in the present data analyses, perhaps more systematic errors are still inherent in the models which accounted for the marginal levels of fit indices. To sum up, results just reported are consonant with the analyses conducted in individual samples that (a) the proposed integrated model was well defined by the four cornerstone constructs whose relations were as predicted and consistent with the results in 146 English; (b) the two axes predicted most of the reliable variance in each bipolar affect dimension. 8.2.3 TESTING CIRCUMPLEXITY In each language, I examined how well the 14 unipolar affect constructs conformed to a circumplex model and their corresponding locations on the circumference. In all five languages, results indicated that (a) the 14 constructs conform to a circumplex model (RMSEA ranged from .10 to .13) and (b) that the majority of their variance was explained by the two axes (communality indices ranged from .87 to 1.00). Constructs did not fall at multiples of 45°. Rather they fell at different angles on the circumference. To provide a cross-cultural comparison of the circumplexity results, I rank ordered the 14 affect constructs by their angular positions on the circumplex in each language. For instance, Pleasant was always designated as the reference variable and was given the rank of 1. Rank 2 was given to the construct which came closet to Pleasant in the anti-clockwise position, such as Larsen and Diener's Activated Pleasant states. The same assignment procedure went on until Rank 14 was given to the construct on the circumplex. One series of rank orders was computed in each language. Spearman's rank order correlations were computed between the five series. Results are given in Table 64. Table 64. Rank Order Correlations Among the Angular Positions of the 14 constructs in each language Rank Order Correlations English Spanish Chinese Japanese Korean English ~ Spanish 1.00 Chinese 1.00 1.00 Japanese .98 .98 .99 Korean .99 .99 .99 10C) -Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). As shown in Table 64, all rank order correlations were above .97, indicating that the 14 constructs were arrayed in more or less the same angular positions on the circumplex in five different languages. Even clearer were the cases with English, Spanish, and Chinese who had identical positions for all 14 constructs. However, a word of caution is in order. The rank orders can be the same even when the absolute locations of the affect constructs in each language samples are very different. Therefore, I computed correlations on the actual angles. Results showed that the correlations were even higher, ranging from .99 to 1.00, with a mean of 1.00. 8.3 THE 45° ROTATION HYPOTHESIS In this section, I examined how well the 45° Rotation Hypothesis was supported all five language samples. The relationships between eight pairs of hypothesized equivalents were examined in each language. Results by pairs of hypothesized equivalents were summarized in Table 65. Results showed that the mean correlations between hypothesized equivalents across five languages ranged from .90 to .98, with a mean of .93. Consistent with previous findings in English, although the hypothesized equivalents highly correlated, they were not interchangeable. These results received further support in four different languages, in addition to English. Similar results were noted in the angular positions on the circumplex between the hypothesized equivalents. In the Pleasant Activated quadrant of Figure 2, hypothesized equivalents differed up to 17° in English, up to 16° in Spanish, up to 17° in Chinese, up to 1 6 0 in Japanese, and up to 3° in Korean; the range was from 3° to 17°. In the Unpleasant Activated quadrant, the range was from 4° to 28°. In Unpleasant Deactivated quadrant, the range was from 2° to 16°. Finally, in the Pleasant Deactivated quadrant, the range was from 11° to 19°. 148 Table 65. Eight Pairs of Hypothesized Equivalent Constructs: A Cross-Language Examination Correlations 8 Pairs of Hypothesized Equivalents Variable Hypothesized Equivalent Range Mean 1. Thayer's Energy Larsen & Diener's Activated Pleasant . 8 5 - .94 .90 2. Thayer's Energy Watson & Tellegen's Positive Affect .88 - .95 .91 3. Thayer's Tiredness Larsen & Diener's Unactivated Unpleasant .88 - .95 .93 4. Thayer's Tension Larsen & Diener's Activated Unpleasant .87 - .95 .94 5. Thayer's Tension Watson & Tellegen's Negative Affect .91 - .98 .93 6. Thayer's Calmness Larsen & Diener's Unactivated Pleasant . 9 2 - .98 .95 7. Larsen & Diener's Activated Pleasant Watson & Tellegen's Positive Affect .90 - .96 .93 8. Larsen & Diener's Activated Unpleasant Watson & Tellegen's Negative Affect .98 - .99 .98 Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). Range of correlations indicate the possible correlations for each pair of hypothesized equivalents obtained in five different languages. To sum up, across five different languages, constructs, according to the 45° Rotation Hypothesis were equivalent, differed anywhere from 2° to 28°. Noticeable differences could be found in all four quadrants of the two-dimensional space. 149 8.4 RELATING AFFECT TO PERSONALITY 8.4.1 A CROSS-LANGUAGE EX AMI NA TION OF THE VARIANCE EXPLAINED BY FFM IN AFFECT In each language, I examined the contribution of the FFM dimensions in predicting each bipolar affect dimensions. In total, eight structural equation models were examined in each language. In Table 66,1 summarize the findings in five languages by affect dimensions. Table 66. Predicting Affect from FFM: A Cross-Language Examination Regression Weights Language N E O A C VAF by VAF by Total NLE O.A.C VAF Endogenous Variable: Pleasant vs Unpleasant ENG - .32* .04 .06 .06 .05 13 (2.8) 1 14 (2.8) SPA - .23* .18 .02 - .04 .23* 14 (4.3) 5 19 (4.7) CHI - .44* .03 .00 .08 .02 22 (3.4) 1 23 (3.4) JPN - . 1 5 .01 .02 .17* .04 5 (2.0) 2 7 (2.5) KOR - .16 .16 - .06 .13 .12 12(3.2) 2 14 (3.5) Mean -.26 .08 .01 .08 .09 13.2 2.2 15.4 Endogenous Variable: Activated vs Deactivated ENG .08 .18* .05 - . 0 3 .06 3(1.5) 1 4(1.6) SPA .13 .10 .24* - . 0 9 .09 1 (1.6) 7 8 (3.6) CHI .20* .16 .03 - . 0 5 .12 3(18) 2 5 (2.2) JPN .01 .20* .05 - . 0 3 .08 4 (2.2) 1 5 (2.3) KOR - . 0 3 .09 .05 - . 0 0 - .02 1 (1.2) 0 1 (1.3) Mean .08 .15 .08 -.04 .07 2.4 2.2 4.6 Endogenous Variable: Thayer's Energy vs Tiredness ENG -.17* .04 .06 - .01 .07 4(1.7) 1 5(1.8) SPA .01 .18 .16 - . 1 0 .19 3 (2.4) 6 9 (3.7) CHI - .25* .11 .05 - .02 .11 12 (2.9) 2 14 (3.0) JPN - .12 .14 .05 .05 .11 6 (2.3) 2 8 (2.5) KOR - . 1 6 - :17 - .02 .07 .05 10(3.1) 1 11 (3.1) Mean -.14 .13 .06 .00 .11 7.0 2.4 9.4 150 Endogenous Variable: Thayer's Tension vs Calmness ENG .34* .08 - .02 - .04 .04 10(2.5) 0 10(2.6) SPA .27* - . 0 5 .18 - .02 - .07 9 (3.6) 3 12 (4.2) CHI .47* .09 - . 0 3 - .04 .04 19(3.3) 0 19(3.3) JPN .15 .08 - . 0 3 - .12 - . 0 3 3(1.6) 1 4(1.9) KOR .16 - .12 .10 - .12 - . 0 8 8 (2.8) 3 11 (3.1) Mean .28 .02 .04 -.07 -.02 9.8 1.4 11.2 Endogenous Variable: Larsen & Diener's Activated Pleasant vs Unactivated Unpleasant ENG -.20* .07 .10 .01 .07 6(2.1) 2 8 (2.3) SPA - .07 .20 .13* - .07 .20 7 (3.3) 5 12 (4.1) CHI - .30* .12 .04 .03 .10 16 (3.1) 2 18(3.2) JPN - .13 .14 .07 .05 .08 7 (2.3) 1 8 (2.5) KOR - .17 .18 .06 .07 .03 11 (3.2) 1 12 (3.3) Mean -.17 .14 .08 .02 .10 9.4 2.2 11.6 Endogenous Variable: Larsen & Diener's Activated Unpleasant vs Unactivated Pleasant ENG .38* .07 - . 0 5 - .04 .00 13(2.8) 1 14 (2.8) SPA .29* - . 0 9 .12 .01 - .14 12 (4.1) 4 16(4.5) CHI .46* .04 - .02 - .06 .01 20 (3.4) 1 21 (3.4) JPN .19* .04 - .01 - .14 .01 4 (2.0) 2 6 (2.2) KOR .17 - .16 .08 - .14 - .07 11 (3.1) 2 13 (3.4) Mean .30 -.02 .02 -.07 -.04 12.0 2.0 14.0 Endogenous Variable: Watson & Tellegen's High vs Low Positive Affect ENG - .21* .06 .13 .01 .12 7 (2.2) 3 10 (2.5) SPA - .10 .18 .15 - . 0 8 .22* 6 (3.2) 7 13 (4.2) CHI - .31* .09 .06 .02 .14 16 (3.1) 3 19(3.3) JPN - .13 .13 .08 .04 .08 6 (2.2) 1 7 (2.4) KOR - .18 .16 .04 .08 .06 12 (3.3) 1 13(3.4) Mean -.19 .12 .09 .01 .12 9.4 3.0 12.4 Endogenous Variable: Watson & Tellegen's High vs Low Negative Affect ENG .38* .07 - .04 - . 0 6 - .01 14 (2.8) 0 14 (2.8) SPA .29* - . 1 0 .10 - . 0 0 - .14 12(4.1) 4 16(4.5) CHI .46* .05 - .02 - .07 - .01 21 (3.3) 0 21 (3.4) JPN .19* .04 - . 0 3 - .14 - .02 4 (2.0) 2 6 (2.3) KOR .17 - . 1 5 .08 - . 1 3 - . 0 8 11 (3.1) 2 13 (3.4) Mean .30 -.03 .02 -.08 -.05 12.4 1.6 14.0 Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). ENG = English, SPA = SPANISH, CHI = Chinese, JPN = Japanese, and KOR = Korean. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. VAF = Variance Accounted For. Figures in parentheses are the standard 151 errors. No standard errors were available for the VAF by O.A.C because the VAF was computed as the difference between the Total VAF and the VAF by N,E. *p_<.001 First, I examined the variance explained by Neuroticism and Extraversion, the two personality dimensions which were most frequently related to affective states in past research. Across eight affect dimensions, the mean variance explained by Neuroticism and Extraversion ranged from 2.4% to 13.2%, with a mean of 9.5%. These two personality dimensions aligned roughly with the two dimensions of the proposed two-dimensional space respectively. For the diagonal dimension of Pleasant Activated vs Unpleasant Deactivated dimensions (Thayer's Energy vs Tiredness, Larsen and Diener's Activated Pleasant vs Unactivated Unpleasant, Watson and Tellegen's High vs Low Positive Affect), both Neuroticism and Extraversion were found to be frequently related. For the diagonal dimension of Unpleasant Activated vs Pleasant Deactivated dimensions (Thayer's Tension vs Calmness, Larsen and Diener's Activated Pleasant vs Unactivated Pleasant, Watson and Tellegen's High vs Low Negative Affect), only Neuroticism was frequently related. Next, I examined the contributions of Openness, Agreeableness, and Conscientiousness in predicting affective states. The mean variance explained by ranged from 1.4% to 3.0%, with a mean of 2.1%. Compared with Neuroticism and Extraversion, these three personality dimensions were less predictive of affective states. Finally, the mean variance explained by all FFM dimensions ranged from 4.6% to 15.4%, with a mean of 11.6%. Of eight affect dimensions, Pleasant vs Unpleasant bipolar dimension was best explained by the FFM dimensions (mean = 15.4%), and Activated vs Deactivated dimension was least explained (mean = 4.6%). One of the frequent findings in the past research was that Neuroticism was related to negative affective states and Extraversion to positive affective states. Contrary to past findings, the present study provided support for the strong influence of Neuroticism on the momentary 152 affective ratings. Extraversion, on the other hand, was less predictive of affective ratings. More interestingly, similar findings were found in all five language samples. Compared with past findings, the reported personality-affect correlations in the present investigation were modest in size. I offer two explanations for the modest correlations. First, previous research relied heavily on zero-order of multiple correlations. Systematic or random errors may be inherent in the estimations of relations (see Jaccard & Wan, 1985). In the present study, structural equation modeling technique was used in an attempt to control for systematic and random errors in the relations reported, which thus gave more conservative estimates. Second, the affective ratings collected in the past research used either "generally feel" (e.g., Watson & Clark, 1992) or "your feelings in the last two weeks (e.g., McCrae & Costa, 1991). Like I discussed in Chapter 1, the personality-affect relations may be exaggerated by the shared trait rating instructions, as well as the scale content overlap between the two domains. In the present study, the momentary affective ratings for a thin slice of time were gathered and thus the shared rating instructions between personality and affect were minimized. Hence the reported relations between the two domains were found relatively modest, compared with those in the past research. 8.4.2 THE SINUSOID FUNCTIONS In each language, a sinusoid function was fitted to five series of correlations between a personality dimension and 14 affect constructs. Model fit was determined by the variance explained by the fitted function. Results across five languages were summarized in Table 67. As shown in Table 67, the variance explained by the fitted sinusoid functions were generally high. The mean variance explained across five languages for Neuroticism was 98%; the mean variance explained for Extraversion was 94%; the mean variance explained for Openness was 81%; the mean variance explained for Agreeableness was 96%; and the mean variance explained for Conscientiousness was 95%. With the exception of Openness, all mean variances were above 90%. A careful examination of Table 67 revealed that the low mean 153 variance explained for Openness was attributed to the relatively low percentages obtained in the three Asian samples - Chinese, Japanese, and Korean. Table 67. Percentage of variance accounted for by the Fitted Sinusoid Functions % Variance Explained by the Fitted Sine Function Language N E O A C English 99 85 92 94 92 Spanish 98 98 96 91 99 Chinese 100 97 67 97 97 Japanese 97 94 71 98 94 Korean 98 98 77 98 94 Mean 98 94 81 96 95 Note. Ns = 535 (English), 233 (Spanish), 487 (Chinese), 450 (Japanese), 365 (Korean). N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness. 154 CHAPTER 9 DISCUSSION AND CONCLUSIONS Regardless of the different labels assigned to each of the four structural models in Figure 1, there is an underlying assumption that they are all mappable onto one another. However, past research on testing this integration hypothesis was rather limited and the findings reported have not been consistent. For instance, Mayer and Gaschke (1988) found support for the close connections between Russell's (1980) model and Watson and Tellegen's (1985) model. On the other hand, Hutchison et al. (1996) offered evidence that the very same models are not interchangeable and concluded that, "it may be misleading to assume that data obtained by measuring dimensions from one of the models can be used to make inferences about dimensions of the other model." (p.785). To clarify past findings, measures of 14 variables drawn from four different structural conceptualizations were examined in English and four other languages. First, overall support was found for the structural validity of each model. Second and more importantly, the four models were found to be mappable onto one another and all 14 variables fit comfortably within the proposed two-dimensional space in all five languages. Most if not all variance in each variable was explained by the two bipolar dimensions - Pleasant vs Unpleasant and Activated vs Deactivated - defining the integrated space. In short, the proposed integrated space received encouraging support as a common ground to map out the relations among the four structural models in all five different languages. Previous research on personality and affect was hampered by restricted definitions of each domain , construct overlap, and methodological overlap. Further, they were restricted primarily to English-speaking samples. In view of the foregoing problems, the present study used comprehensive descriptive systems for both personality and affect and studied momentary affect (for one thin slice of time) as compared with long-term affect in past research (in the last month). The present findings showed that personality explained small but reliable portion of variance in momentary affective feelings (5% to 15%). The five personality dimensions played 155 distinctive roles in different languages in predicting affective states, although Neuroticism was found to be most frequently related to affective states in all but Korean. In Korean, none of the personality dimensions reliably predicted affective states. 9.1 IMPLICATIONS In the following, I discuss the implications of the present results: (a) Can the proper rotation within the two-dimensional space be determined by external correlates? (b) How shall affect be defined? and (c) What are the state-trait relations? 9.1.1 CAN THE PROPER ROTA TION BE DETERMINED BY EXTERNAL CORRELATES? One of the vexing controversies in the two-dimensional structural models of affect concerns the proper rotation of the axes. The two competing rotations are on the one hand, the horizontal and vertical axes as shown in Figure 2 and on the other, the diagonal axes, namely Pleasant Activated vs Unpleasant Deactivated and Unpleasant Activated vs Pleasant Deactivated. From the statistical point of view, within a two-dimensional space, any pair of non-redundant axes explained the same amount of variance as the other pairs. Therefore, comparing the variance explained by the two competing rotations does not help decide which rotation is more "proper" or "basic" than the other. Researchers have turned to external correlates for help. For instance, Meyer and Shack (1989, p. 692) suggested that".. . the salience, replicability, and experimental support for a particular set of dimensions can be used as the criteria for selecting an agreed-on basic structure" (emphasis in original). The authors provided evidence in support of Watson and Tellegen's (1985) rotation that in both state and trait ratings of affect, Neuroticism and negative affective states shared a common dimension in the factor space; Extraversion and positive affective states shared another dimension. In the hypothesized model in Figure 2, Extraversion and Neuroticism fell at 45° and 135°, respectively. In other words, the authors concluded that the diagonal dimensions at 45° and 135° constituted the "basic" rotation. In a more vigorous attempt, Larsen (1989) examined the personality correlates of the eight vectors of the Larsen and Diener's (1992) structural model. Larsen found that correlates 156 fell all around the circumplex. Consistent with previous findings, the Pleasant Activated and Unpleasant Activated vectors relate meaningfully to external correlates. Further, a number of meaningful correlates were found for the horizontal axis of Pleasant vs Unpleasant states and the vertical axis of Activated vs Deactivated states. To examine both Meyer and Shack's (1989) and Larsen's (1989) hypotheses, I projected the FFM dimensions onto the proposed two-dimensional space in all five languages. To this end, I specified a structural equation model in which the Pleasant vs Unpleasant and Activated vs Deactivated axes were treated as exogenous and each of the five personality dimensions, in turn, was treated as the endogenous variable. The regression weights of each personality dimension on the two bipolar axes were used as x- and y- coordinates to project the FFM onto the integrated two-dimensional space. Results are shown in Figure 15. Figure 15. Projecting the FFM dimensions onto the Integrated Two-Dimensional Space in five language samples. ENGLISH O (45°) N (175°) E (37°) A(0°) C (25°) SPANISH CHINESE JAPANESE 158 N (177°) O (63°) A (355°) KOREAN N (185°) O (68°) C (14°) 159 In all five languages, the FFM dimensions fell at different angles on the circumference of the two-dimensional space. Across five languages, Neuroticism fell between 166° and 195° on the circumference; Extraversion fell between 2° and 43°; Openness fell between 45° and 113 °; Agreeableness fell between 0° and 355°; and Conscientiousness fell between 14° and 35°. Each FFM dimension was associated with a relatively large range of angles, ranging from 21° for Conscientious and 68° from Openness. Contrary to what Meyer and Shack (1989) predicted, Openness in English was the only FFM dimension which fell on the 45° diagonal dimension. Consistent with Larsen's (1989) findings, personality dimensions lined up in a meaningful pattern around the upper half of the space leaving no hints on just which rotation is more "basic" than the other. Once again, personality correlates offer no clear guidance on just where the "basic" rotation lies in the two-dimensional space. Other researchers have started looking to biological mechanisms to solve the vexing problem of rotational indeterminancy (e.g., Heller, 1990). The logic behind is that any rotation which is associated with the biological mechanisms should be regarded as more "basic" than the other alternatives. Unfortunately, the present data, like other correlational data on which the debate is centered, offer no guidance for preferring one rotation over the other. Rather, the selection of the proper rotation has to be based on conceptual clarity. Larsen and Diener (1992) and Reisenzein (1994) have proposed some conceptual arguments in favor of the Pleasure and Arousal rotation. 9 . 1 2 HOW SHALL AFFECT BE DEFINED? Another vexing problem in the study of affect is how to define affect. In the present study, I offer the Cartesian space in Figure 2 as a possibility. Within this two-dimensional space, all affect terms are composed of differing dosages of Pleasant vs Unpleasant and Activated vs Deactivated dimensions. Past and present research showed that the two axes are bipolar in nature, that is, Pleasant is the bipolar opposite of Unpleasant, Activated the bipolar 160 opposite of Deactivated. Further, these two axes were found to be virtually independent of each other. In study of the structure of affect, three approaches were used in the past; one emphasized the horizontal axis of Pleasant vs Unpleasant (e.g., Bradburn, 1969), another emphasized the vertical axis of Activated vs Deactivated (e.g., Thayer, 1989), and another emphasized both (e.g., Larsen & Diener, 1992; Russell, 1980) Each approach has marshaled a research program and integrating all three of them is clearly desirable for the study of affect. As shown in the present study, the Cartesian space was able to integrate the aforementioned three approaches comfortably in all five samples. It represents a common space to measure and compare affect. In particular, the four structural models, originating from three different approaches, fit nicely within the proposed integrated space. Hence, the two-dimensional space provides a common ground to integrate affect constructs in different structural models and to map out the affective feelings in different languages. Nonetheless, the two bipolar axes of the proposed space explained most but not all 100% of variance in the affect constructs and yielded relatively high levels of RMSEA (around .11). I offer two explanations for these findings. First, although the best possible method available today to control systematic and random errors was used (i.e., structural equation modeling technique), the control of systematic errors was restricted the error variance using the same response formats. There may be other systematic errors common to different formats. Efforts are to be made to locate those systematic errors. Second, a third substantive dimension, as well as interaction effects between the two included dimensions, may account for the remaining variance (e.g. Russell, 1978; Smith & Ellsworth, 1985). In the Cartesian space, some regions are more "affective" than the others (just as some colors are more "colorful" than the others). However, to understand human affect, none of the regions should be omitted and a person's feelings can fall at any spot on the Cartesian space. Past research has focused on the affective feelings for salient life events, such as marriage, graduation, or lottery ticket winner. Those feelings are certainly important in its own right and 161 need to be studied. Given the high activation component in those life events, I suspect that the Activation dimension of the Cartesian space will be truncated because of the problem of restriction of range and therefore the space will be squished. In my opinion, all affective feelings are important for our understanding of the structure of affect, as all colors are important for our understanding of the vision. Therefore, the present study was intended to study all kinds of affective feelings which were hypothesized to tap the whole proposed Cartesian space. 9.1.3 WHA T ARE THE STATE-TRAIT RELATIONS? In the present study, attempts were made to study the relations between affective states and the enduring characteristics of personality on a comprehensive basis. Overall, the correlations were modest in magnitude. Different personality dimensions played distinctive roles in different languages.. Across the five languages, Neuroticism was most frequently related to affect. Two qualifications are to be noted to the findings. In the present study, I relied heavily on translations. For the affect measures, I adopted the sensitive backtranslation procedure to translate these scales into the four languages. For personality, one of the widely used personality inventory - NEO FFI (Costa & McCrae, 1992) was used. The four language versions available for the present study were validated and found to yield sound psychometric properties. However, psychometric properties do not guarantee complete accuracy of the translated measures. And accuracy of translation is the factor influencing my interpretation of the personality-affect connections reported in the present study. I was involved in the Chinese translation of the NEO PI R or NEO FFI. Our research team in Hong Kong encountered difficulties when we did the translations, in particular the Openness scale items. Of course, the items were translatable, but the meanings implied might not. However, what counts in defining Openness is the meanings implied. To illustrate, let me explain my thoughts with the item "I am sometimes completely absorbed in music I am listening to." I did not understand how I could be absorbed in music until after I stayed in North America for three years. Music certainly exists in all Chinese societies, but society exerts an influence on how music is perceived and used. As a Hong Kong Chinese, I use music as entertainment 162 as in the public transport or in parties. So, the item did not make much sense to me; I just did not know how to get absorbed in music. As a Canadian resident, I now use music to tune my mood and to set myself into deep thought. All in a sudden, the item makes sense to me. Only after living in Canada does the rating I provide for this item measures Openness. I am not saying that the translations for the personality scales are necessarily bad. What I did was to demonstrate that even if the translations are good, the meaning of the items may be altered by the society and thus the items may not tap the intended personality dimension in all participants properly. The present personality-affect connections are therefore best regarded as tentative. Perhaps, they are lower-bound estimates of the actual connections. More research efforts are needed to extend and replicate the present results with different and improved personality and affect measures. The present study defines personality as a set of enduring characteristics and assumes that traits possess explanatory power. This definition is one of the many approaches to define personality traits. The individual differences approach, the one taken in the present study, defines personality traits as psychological structures lying deep inside a person and as responsible for human behaviors, thoughts, and feelings (e.g., Cattell, 1965; Costa & McCrae, 1992). Traits are therefore regarded as explanatory constructs. However, not every trait psychologist agrees that traits have explanatory power. Buss and Craik (1983) proposed that traits are simply labels that sum up the frequency of behaviors. Saying someone is extraverted simply means that that person engages in social activities more often than do most people. The authors termed this approach as the act frequency approach. I believe that the traits of the FFM explain some of the variability in observed affect. Still, from Buss and Craik's (1983) point of view, the proper conclusion from the present (correlational) data is more cautious: Personality traits for the FFM are able to predict some of the variability in observed affect. Momentary affective ratings can be influenced by very many factors, such as diurnal variations in hormones, current situation, and weather, in addition to personality. Therefore, 163 being able to predict small but reliable variance in momentary affect represents an important contribution of personality. Again, the present study is the first of its kind to explore the relations between personality and momentary affective ratings in a cross-language context. Results need to be replicated and expanded in future studies, but they suggest that meaningful relations will be found. 9.2 LIMITATIONS OF THE PRESENT DISSERTATION 9.2.1 JUDGMENT SPACE FOR AFFECT The present dissertation relied completely on self-reported measures of affect. In psychology, there are at least three other measures, namely, peer reports, physiological measures, and behavioral observations. Using self-reported measures is certainly the first step in the scientific research of affect. However, replicating the present results with non-self-report measures is needed in future research. What is the judgment space for affect? The present dissertation looked at the judgments about one's own affect and results lent support to the proposed two-dimensional space. Other research has looked at the judgments about others' affect via facial expressions. Again, results supported the proposed two-dimensional space in English and Chinese (Russell, Weiss, & Mendelsohn, 1989; Schlosberg, 1952; Yik & Russell, 1998). Is it then justified to equate the judgement space to the scientific description of affect? Again, the present data do not offer guidance on how close the judgement space and the scientific descriptive map of affect are. 9.2.2 IMPOSED-ETIC APPROACH One of the major criticisms of the present dissertation is that scales originating in English were translated and administered to participants of different language backgrounds. Results were then compared across different languages. These constructs were "imposed upon" the participants in four different languages. This is called "imposed-etic" approach (Berry, 1969; Church & Katigbak, 1989). In the four languages, no prior study was conducted to find out which terms and dimensions were salient within that language. The possibility thus remains that 164 additional affect dimensions would emerge with additional items. This hypothesis needs to be tested in future studies. The problematic scales of Activated and Deactivated expose a problem of translation. Although I adopted a sensitive translation procedure (that is, scales were to be translated to yield equivalents which were neither pleasant nor unpleasant), results indicated that the Activated and Deactivated scales were either slightly pleasant or slightly unpleasant in the four languages. Perhaps, refining the content of these scales with indigenous inputs in each language may help define the proposed two-dimensional space more appropriately. 9.2.3 A SELECTION OF LANGUAGES The present study was phrased in terms of language groups. The question might be raised of the implication of this research for the relation of affect to culture. Many attempts have been made to define the concept of culture but no single definition can tell how much difference there must be to distinguish one culture from the other (Smith & Bond, 1993) and what the differences are based on. Cultures vary in terms of social institutions, norms, rules, marriages, and so on. Researchers started looking for other variables to distinguish one culture from the other. Some relied on the boundaries of nation-states (e.g., Hofstede, 1980; Smith & Bond, 1993) while others relied on values priorities people ascribed to (Schwartz, 1992; Chinese Culture Connection, 1987). Given that (a) the complicated issue of defining a culture, and (b) only one sample was recruited from each country involved in the present study, I decided to treat my five samples as representing five languages than five cultures. Language is at least an objective criterion descriptive of the samples and is the common denominator to all participants in each sample. The present study focussed on comparing research findings in English with those in four other languages - Spanish, Chinese, Japanese, and Korean. These languages are by no means exhaustive of the world languages. Expanding the current investigation to other languages is necessary to advance our understanding of cross-language similarities and differences. For instance, findings in the structure of affect in English and Spanish are 165 extremely similar, if not identical. Do the similarities reflect the generalizability of the integrated structure? Or is the integrated structure a product of Indo-European languages? By obtaining a more diverse sample of languages, we can get a bit closer to the answer. The present study provided evidence supportive of the integrated structure in three distinct languages, other than English and Spanish. Thus, this structure received initial support beyond Indo-European languages. However, the present dissertation is just a beginning. Replicating the present findings in other languages is much needed to advance our understanding of the structure of affect. 9.2.4 FIT INDICES In the present dissertation, in order to examine how well the hypothesized models fit the data, I relied on structural equation modeling techniques. Many different indices are available to assess the degree to which a structural equation model fits the observed data. Because different fit indices are suitable for evaluating different aspects of the model, such as parsimony, variance explained, and so on, no single measure of fit should be relied on exclusively (see Bollen & Long, 1993). Four indices were used to assess model fit in the present dissertation. In the following, I discuss the potential advantages and pitfalls in using each of the four indices. First, the chi-square goodness-of-fit statistic (x2) tests the null hypothesis that the hypothesized model reproduces the correlation matrix perfectly for the manifest variables. This chi-square statistic is computed as N - 1 times a "discrepancy function," which is a weighted measure of the difference between the observed and reproduced correlation matrices. In general, while keeping the sample size constant, the larger the chi-square, the more the correlation matrix specified by the hypothesized model deviates from the correlation matrix for the manifest variables. However, because of the "weighting" of the discrepancy function, this relation between the chi-square statistic and sample size may not be consistent for all circumstances. Under the condition of large sample sizes and the assumption of multivariate normal distribution, the maximum likelihood fit function will be distributed asymptotically as chi-square. However, with large sample sizes needed to justify the use of chi-square statistic, most 166 hypothesized models will be rejected, even when the residuals are small (Bentler, 1990; Cudeck & Browne, 1983). The present dissertation involved relatively large sample sizes, ranging from N for Spanish = 233 to N for English = 535. In almost all structural equation models, the chi-square statistics were so big that they suggested rejection of the hypothesized models, even when the residuals were small. Second, Steiger and Lind's (1980) RMSEA can roughly be regarded as a root mean square standardized residual. RMSEA is adjusted for model complexity and is therefore useful in comparing two nested models. Greater values indicate poorer fit. However, "fit" is measured by the discrepancy function, which is a "weighted" function of the difference between observed and reproduced correlation matrices. Because the weighting is itself a function of parameter values, the practice of using fixed cutoff points, say .05, for evaluating fit may be questionable (Steiger, 1998). When the variables are highly correlated, as in the present dissertation, the weighted function may tend to produce higher discrepancy function values than when the correlations are low. RMSEA is the quadratic function of the discrepancy function and may tend to produce values that are somewhat higher than those encountered under more typical circumstances. In the present study, the RMSEA for structural equation models predicting affect constructs by the two bipolar axes (Pleasant vs Unpleasant and Activated vs Deactivated states) ranged from .06 to .11; the RMSEA for structural equation models predicting affect by the FFM dimensions ranged from .00 to .09. Results indicate that models hypothesized with highly correlated variables tend to yield higher values of RMSEA whereas those hypothesized with less correlated variables tend to yield smaller values of RMSEA. Third, Adjusted Goodness of Fit (AGFI) provides a direct measure of goodness of fit by adjusting the Goodness of Fit downwards to compensate for model complexity. However, like the chi-square statistic, AGFI is also dependent on sample size (Bentler, 1990). Authors disagree on whether this index undercorrects (Mulaik et al., 1989) or overcorrects (Marsh, Balla, & McDonald, 1988) for model complexity. So, it is debatable whether a high AGFI indicates a good fit or not. Since the models tested in the present dissertation were generally complicated, 167 in particular those testing the relationships among the four structural models, it was as expected that their AGFI were relatively small. Fourth, Bentler's (1990) Comparative Fitness Index (CFI) is a fit index that evaluates the adequacy of the hypothesized model in relation to a baseline model and is normed to the values between 0 and 1. CFI is computed on the basis of the most restricted baseline model (null model) in which all manifest variables are assumed to be uncorrelated (i.e., every variable is an indicator for its own latent construct). The higher is the value, the better is the model fit. The problem with CFI is its comparison point. CFI defines a specific comparison point such as the mutually uncorrelated model and the logic implies that more sophisticated models cannot be hypothesized for the data if the baseline model fit the data well (Tanaka, 1993). So, even a CFI values of .99 may not indicate that the hypothesized model fit the data very well. CFI does not provide a conservative comparison base. In all, the four indices adopted to evaluate the goodness or badness of fit in the present study are each associated with underlying statistical assumptions. Each has associated strengths and weaknesses. The approach taken in the present dissertation was that hypothesized models should not be accepted or rejected on the basis of fixed cutoff points, such as an RMSEA of .05. Rather, all the hypothesized models should be viewed as candidates to understand what the underlying structures are and these alternative models are varying in terms of reasonableness, rather than strictly true or false. In evaluating the reasonableness of the models, I relied on all four indices simultaneously. These four indices are useful as overall guidelines and best used as comparative indices comparing nested models. Important is the degree of agreement among the indices on how well the model fit the data. 9.3 CONCLUDING REMARKS Progress has been made in the psychology of affect. Starting from divided opinions on how the covariations of affective feelings should be described, to the consensus on various two-dimensional models each focusing on their important dimensions, and to the present dissertation demonstrating how the proposed integrated space integrate various two-dimensional models in five different languages. The Cartesian space defined by two bipolar axes, Pleasant vs Unpleasant and Activated vs Deactivated axes, proves to be a sufficient space to describe self-reported momentary affect, and to integrate other structural models. Further, the circular ordering of the affect constructs on the circumference of the proposed space found in different languages is highly consistent with a circumplex model in which variables can fall at any angle on the circumference, rather than at the multiples of 45° Segmenting the two-dimensional space into 45° slices is a completely arbitrary practice and represents an approximation to the circumplex model. The empirical results in the present findings represents another approximation. Still, future research is much needed to progress beyond the present approximation. 169 Bibliography Allik, J., & Realo, A. (1997). Emotional experience and its relation to the Five-Factor Model in Estonian. Journal of Personality, 65, 625-647. Almagor, M., & BenPorath, Y. (1989). The two-factor model of self-reported mood: A cross-cultural replication. Journal of Personality Assessment. 53, 10-21 Averill, J . R. (1975). A semantic atlas of emotional concepts. JSAS: Catalog of selected documents in psychology, 5(4), 1103. Averill, J . R. (1980). A constructive view of emotions. In R. Plutchik & H. Kellerman (Eds.), Theories of emotion (Vol. 1, pp. 305-340). New York: Academic Press Barlow, D. H. (1988). Anxiety and its disorders: The nature and treatment of anxiety and panic. New York: Guildford. Benet-Martinez, V., & John, O. P. (1998). Los Cinco Grandes across cultures and ethnic groups: Multitrait-multimethod analyses of the Big Five in Spanish and English. Journal of Personality and Social Psychology, 75. 729-750. Bentler, P. M. (1969). Semantic space is (approximately) bipolar. Journal of Psychology, 71, 33-40. Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238-246. Berry, J. W. (1969). On cross-cultural comparability. International Journal of Psychology, 4, 119-128. Bollen, K. A., & Long, J. S. (Eds.). (1993). Testing structural eouation models. Newbury Park, CA: Sage. Boucher, J . D. (1979). Culture and emotion. In A. J . Marsella, R. G. Tharp, & T. V. Ciborowski (eds.), Perspectives on cross-cultural psychology (pp. 159-178). San Diego, CA: Academic Press. Bradburn, N. M. (1969). The structure of psychological well-being. Chicago: Aldine. Browne, M. W. (1992). Circumplex models for correlation matrices. Psychometrika, 57, 469-497. Burke, M. J . , Brief, A. P., George, J. M., Roberson, L, & Webster J. (1989). Measuring affect at work: Confirmatory analyses of competing mood structures with conceptual linkage to cortical regulatory systems. Journal of Personality and Social Psychology, 57, 1091-1102. Bush, L. E. II. (1973). Individual differences multidimensional scaling of adjectives denoting feelings. Journal of Personality and Social Psychology, 25, 50-57. Buss, D. M., & Craik, K. H. (1983). The act frequency approach to personality. Psychological Review, 90. 105-126. Cattell, R. B. (1965). The scientific analysis of personality. Baltimore: Penguin Books. Chinese Culture Connection (1987). Chinese values and the search for culture-free dimensions of culture. Journal of Cross-Cultural Psychology. 18. 143-164. Church, A. T., & Katigbak, M. S. (1989). Internal, external, and self-report structure of personality in a non-Western culture: An investigation of cross-language and cross-cultural generalizability. Journal of Personality and Social Psychology. 57. 857-872. Church, A. T., Katigbak, M. S., & Reyes, J! A. S., & Jensen, S. M. (1997). The structure and personality correlates of affect in a non-Western culture: Evidence for cross-cultural comparability. Manuscript submitted for publication. Costa, P. T., Jr., & McCrae, R. R. (1980). Influence of extraversion and neuroticism on subjective well-being: Happy and unhappy people. Journal of Personality and Social Psychology. 38. 668-678. Costa, P. T., Jr., & McCrae, R. R. (1984). Personality as a lifelong determinant of wellbeing. In C. Z. M. C. E. Izard (Ed.), Emotion in adult development (pp. 141-157). Beverly Hills: Sage. Costa, P. T., Jr., & McCrae, R. R. (1992). The Revised NEO Personality Inventory (NEO-PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources. Costa, P. T., Jr., & McCrae, R. R. (1995). Solid ground in the wetlands of personality. Psychological Bulletin. 117. 216-220. Costa, P. T., Jr., & McCrae, R. R. (1996). Mood and personality in adulthood. In C. M. S. H. McFadden (Ed.), Handbook of emotion, adult development, and aging (pp. 369-383). San Diego, California: Academic Press. Cudeck, R., & Browne, M. W. (1983). Cross-validation of covariance structures. Multivariate Behavioral Research. 18. 147-167. Diener, E., & Emmons, R. A. (1984). The independence of positive and negative affect. Journal of Personality and Social Psychology. 47. 1105-1117. Digman, J. (1990) Personality structure: Emergence of the five-factor model. Annual Review of Psychology. 41. 417-440. Eich, E. (1995). Searching for mood dependent memory. Psychological Science. 6. 67-75. Emmons, R. A., & Diener, E. (1985). Personality correlates of subjective well-being. Personality and Social Psychology Bulletin. 11. 89-97. Emmons, R. A., & Diener, E. (1986). An interactional approach to the study of personality and emotion. Journal of Personality. 54. 371-384. Eysenck, H. J. (1992). Four ways five factors are not basic. Personality and Individual Differences. 13. 667-673. Eysenck, H. J., & Eysenck, M. W. (1985). Personality and individual differences. A Natural Science Approach. New York: Plenum Press. Fabrigar, L. R., Visser, P. S., & Browne, M. W. (1997). Conceptual and methodological issues in testing the circumplex structure of data in personality and social psychology. Personality and Social Psychology Review. 1. 184-203. Feldman Barrett, L, & Russell, J. A. (1998). Independence and bipolarity in the structure of current affect. Journal of Personality and Social Psychology. 74. 967-984. Feldman, L. A. (1995). Variations in the circumplex structure of mood. Personality and Social Psychology Bulletin. 21. 806-817. Forgas, J. P. (1995). Mood and judgment: The affect infusion model (AIM). Psychological Review. 117. 39-66. Goldberg, L. R. (1993). The structure of phenotypic personality traits. American Psychologist. 48, 26-34. Green, D. P., Goldman, S. L, & Salovey, P. (1993). Measurement error masks bipolarity in affect ratings. Journal of Personality and Social Psychology. 64.1029-1041. Green, R. F., & Goldfried, M. R. (1965). On the bipolarity of semantic space. Psychological Monographs. 79 (6, Whole No. 599). Hama, H., Matsuyama, Y., & Lin, R. Y. (1986). Analysis of Chinese emotional words. Unpublished manuscript, Doshisha University. Heller, W. (1990) The neuropsychology of emotion: Developmental patterns and implications for psychopathology. In N. L. Stein, B. Leventhal, & T. Trabasso (Eds.). Psychological and biological approaches to emotion (pp. 167-214). Hillsdale, NJ: Lawrence Erlbaum. Hofstede, G. (1980). Culture's conseguences: International differences in work-related values. Beverly Hills, CA: Sage. Hutchison, K. E., Trombley, R. P., Collins, F. L, Jr., McNeil, D. W., Turk, C. L, Carter, L. E., Ries, B. J., & Leftwich, M. J. T. (1996). A comparison of two models of emotion: Can measurement of emotion based on one model be used to make inferences about the other? Personality and Individual Differences. 21. 785-789. Isen, A. M. (1993) Positive affect and decision making. In M. Lewis & J. M. Haviland (Eds.), Handbook of emotions (pp. 261-278). New York: Guilford. Izard, C. E. (1977). Human emotions. New York: Plenum Press. Izard, C. E., Libero, D. Z., Putname, P., & Haynes, O. M. (1993). Stability of emotion experiences and their relations to traits of personality. Journal of Personality and Social Psychology, 64. 847-860. Jaccard, J. , & Wan, C. K. (1995). Measurement error in the analysis of interaction effects between continuous predictors using multiple regression: Multiple indicator and structural equation approaches. Psychological Bulletin. 117, 348-357. Ketelaar, T. (1989). Examining the circumplex model of affect in the domain of mood-sensitive tasks. Unpublished master's thesis, Purdue University. Lang, P. J. (1994). The motivational organization of emotion: Affect-reflex connections. In S. H. M. Van Goozen, N. E. Van de Poll, & J. A. Sergent (Eds.), Emotions: Essays on emotion theory (pp. 61-93). Hillsdale, New Jersey: Lawrence Erlbaum. Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1992). A motivational analysis of emotion: Reflexcortex connections. Psychological Science. 3. 44-49. Larsen, R. J . (1989). Personality as an affect dispositional system. In L. A. Clark & D. Watson (Chair), Emotional Bases of Personality. Symposium conducted at the meeting of the American Psychological Association, New Orleans. Larsen, R. J. , & Diener, E. (1992). Promises and problems with the circumplex model of emotion. In M. S. Clark (Ed.), Review of Personality and Social Psychology: Emotion (Volume 13, pp. 25-59). Newbury Park, CA: Sage. Lorr, M., Shi, A. Q., & Youniss, R. P. (1989). A bipolar multifactor conception of mood states. Personality and Individual Differences. 10. 155-159. Mano, H. (1991). The structure and intensity of emotional experiences: Method and context convergence. Multivariate Behavioral Research. 26. 389-411. Mayer, J. D., & Gaschke, Y. N. (1988). The experience and meta-experience of mood. Journal of Personality and Social Psychology. 55. 102-111. Macaulay, D. L. (1997). Mood dependent memory: Extension and validation. Unpublished doctoral dissertation, University of British Columbia, Vancouver, British Columbia, Canada. Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodness-of-fit indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin. 103. 391-410. McConville, C , & Cooper, C. (1992). The structure of moods. Personality and Individual Differences. 13. 909-919. McCrae, R. R., & Costa, P. T., Jr. (1991). Adding Liebe und Arbeit: The full five-factor model and well-being. Personality and Social Psychology Bulletin. 17. 227-232. McCrae, R. R., & Costa, P. T., Jr. (1997). Personality trait structure as a human universal. American Psychologist. 52. 509-516. McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its applications. Journal of Personality. 60. 175-215. McCrae, R. R., Costa, P. T., Jr., & Yik, M. S. M. (1996). Universal aspects of Chinese personality structure. In M. H. Bond (Ed.), The handbook of Chinese psychology (pp. 189-207). Hong Kong: Oxford University Press. Mehrabian, A., & Russell, J. A. (1974). An approach to environmental psychology. Cambridge, MA: MIT Press. Meyer, G. J. , & Shack, J. R. (1989). Structural convergence of mood and personality: Evidence for old and new directions. Journal of Personality and Social Psychology. 57. 691-706. Mulaik, S. A., James, L. R., Van Alstine, J., Bennett, N., Lind, S., & Stilwell, C. D. (1989). Evaluation of goodness-of-fit indices for structural equation models. Psychological Bulletin. 105. 430-445. Nemanick, R. C. Jr., & Munz, D. C. (1994). Measuring the poles of negative and positive mood using the Positive Affect Negative Affect Schedule and Activation Deactivation Adjective Check List. Psychological Reports. 74. 195-199. Neufeld, R. W. J. (1975). A multidimensional scaling analysis of schizophrenics' and normals' perceptions of verbal similarity. Journal of Abnormal Psychology. 84. 498-507. Neufeld, R. W. J. (1976). Simultaneous processing of multiple stimulus dimensions among paranoid and non-paranoid schizophrenics. Multivariate Behavioral Research. 11. 425-441. O'Malley, M. N., & Gillette, C. S. (1984). Exploring the relations between traits and emotions. Journal of Personality, 52, 274-284. Osgood, C. E., May, W. H., & Miron, M. S. (1975). Cross-cultural universals of affective meaning. Urbana: University of Illinois Press. Pervin, L. A. (1993). Affect and personality. In M. L. J. M. Haviland (Ed.), Handbook of emotions (pp. 301-311). New York: The Guilford Press. Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New York: Harper & Row. Potter, S. H. (1988). The cultural construction of emotion in rural Chinese social life. Ethos, 16, 181-208. Reisenzein, R. (1994). Pleasure-activation theory and the intensity of emotions. Journal of Personality and Social Psychology, 67. 525-539. Russell, J . A. (1978). Evidence of convergent validity on the dimensions of affect. Journal of Personality and Social Psychology, 36, 1152-1168. 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. (1983). Pancultural aspects of the human conceptual organization of emotions. Journal of Personality and Social Psychology, 45, 1281-1288. Russell, J . A., & Bullock, M. (1985). Multidimensional scaling of emotional facial expressions: Similarity from preschoolers to adults. Journal of Personality and Social Psychology, 48, 1290-1298. Russell, J . A. & Carroll, J . M. (in press). On the bipolarity of positive and negative affect. Psychological Bulletin. Russell, J . A., Lewicka, M. & Niit, T. (1989). A cross-cultural study of a circumplex model of affect. Journal of Personality and Social Psychology, 57. 848-856. 176 Russell, J . A., & Mehrabian, A. (1977). Evidence for a three-factor theory of emotions. Journal of Research in Personality. 11. 273-294. Russell, J . A., Weiss, A., & Mendelsohn, G. A. (1989). Affect Grid: A single-item scale of pleasure and activation. Journal of Personality and Social Psychology. 57. 493-502. Russell, J . A., & Yik, M. S. M. (1996). Emotions among the Chinese. In M. H. Bond (Ed.), The handbook of Chinese psychology (pp. 166-188). Hong Kong: Oxford University Press. Schachter, S., & Singer, J. E. (1962). Cognitive, social, and physiological determinants of emotional state. Psychological Review. 69. 379-399. Schlosberg, H. (1952). The description of facial expressions in terms of two dimensions. Journal of Experimental Psychology. 44. 229-237. Schlosberg, H. (1954). Three dimensions of emotion. Psychological Review. 61. 81-88. Schmutte, P. S., & Ryff, C. D. (1997). Personality and well-being: Reexamining methods and meanings. Journal of Personality and Social Psychology. 73. 549-559. Schwartz, S. H. (1992). The universal content and structure of values: Theoretical advances and empirical tests in 20 countries. In M. Zanna (Ed.), Advances in Experimental Social Psychology (Vol. 25, pp. 1-65). New York: Academic Press. Shimonaka, Y. (1996). Development of the Japanese NEO-PI-R and NEO-FFI. Technical Report. Tokyo, Japan: Tokyo Metropolitan Institute of Gerontology. Sjoberg, L, Svensson, E..-& Persson, L. (1979). The measurement of mood. Scandinavian Journal of Psychology. 20. 1-18. Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal in emotion. Journal of Personality and Social Psychology. 48. 813-838. Smith, P. B., & Bond, M. H. (1993). Social psychology across cultures: Analysis and perspectives. New York: Harvester & Wheatsheaf. Steiger, J. H. (1995). Structural equation modeling in Statistica (Version 5). [Computer software]. Tulsa, OK: Statsoft, Inc. Steiger, J . H. (1998). Some cautions about the interpretation of the RMSEA. Unpublished manuscript, University of British Columbia, Canada. Steiger, J. H., & Lind, J. C. (1980, May). Statistically-based tests for the number of common factors. Paper presented at the annual Spring Meeting of the Psychometric Society in Iowa City, IA. Tanaka, J. S. (1993). Multifaceted conceptions of fit in structural equation models. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 10-39). Newbury Park, CA: Sage. Tellegen, A. (1985). Structures of mood and personality and their relevance to assessing anxiety, with an emphasis on self-report. In A. H. Tuma & J. D. Maser (Eds.), Anxiety and Anxiety Disorders (pp. 681-706). Hillsdale, New Jersey: Erlbaum. Thayer, R. E. (1967). Measurement of activation through self-report. Psychological Reports. 20, 663-678. Thayer, R. E. (1986). Activation-Deactivation Check List: Current overview and structural analysis. Psychological Reports. 58. 607-614. Thayer, R. E. (1989). The biopsychology of mood and arousal. New York: Oxford University Press. Thayer, R. E. (1996). The origin of everyday moods: Managing energy, tension, and stress. New York: Oxford University Press. Thayer, R. E., Takahashi, P. J., & Pauli, J. A. (1988). Multidimensional arousal states, diurnal rhythms, cognitive and social processes, and extraversion. Personality and Individual Differences. 9. 15-24. Warr, P., Barter, J., & Brownbridge, G. (1983). On the independence of positive and negative affect. Journal of Personality and Social Psychology. 44. 644-651. Watson, D., & Clark, L. A. (1984). Negative affectivity: The disposition to experience aversive emotional states. Psychological Bulletin. 96. 465-490. Watson, D., & Clark, L. A. (1992). On traits and temperament: General and specific factors of emotional experience and their relation to the Five-Factor Model. Journal of Personality. 6Q, 441-476. Watson, D., & Clark, L. A. (1997). Extraversion and its positive emotional core. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 767-793). San Diego: Academic Press. Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood. Psychological Bulletin. 98. 219-235. Watson, D., Clark, L. A., & Tellegen, A. (1984). Cross-cultural convergence in the structure of mood: A Japanese replication and a comparison with U.S. findings. Journal of Personality and Social Psychology. 47. 127-144. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology. 54. 1063-1070. Watson, D., Clark, L. A., & Tellegen, A. (1994). Basic principles of affective structure. In L. A. Feldman (Chair), Mood: Consensus and controversy. Symposium presented at the 102nd Annual Conference of the American Psychological Association, Los Angeles, CA. Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1998). The two general activation systems of affect: Structural findings, evolutionary considerations, and psychobiological evidence. Manuscript submitted for publication. Wessman, A. E., & Ricks, D. F. (1966). Mood and personality. New York: Holt, Rinehart & Winston. Whissell, C. M. (1981). Pleasure and activation revisited: dimensions underlying semantic responses to fifty randomly selected "emotional" words. Perceptual and Motor Skills. 53. 871-874. Wierzbicka, A. (1992). Defining emotion concepts. Cognitive Science. 16. 539-581. Wiggins, J. S. & Trapnell, P. D. (1997). Personality structure: The return of the big five. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 737-765). San Diego: Academic Press. Williams, D. G. (1981). Personality and mood: State-trait relationships. Personality and Individual Differences, 2, 303-309. Williams, D. G. (1989). Personality effects in current mood: Pervasive or reactive? Personality and Individual Differences. 10, 941-948. Williams, D. G. (1990). Effects of psychoticism, extraversion, and neuroticism in current mood: A statistical review of six studies. Personality and Individual Differences. 11. 615-630. Wundt, W. (1924). An introduction to psychology (R. Pintner, Trans.) London: Allen & Unwin. (Original work published 1912) Yik, M. S. M., & Russell, J. A. (1998). [The Chinese Affect Grid]. Unpublished raw data. Yik, M. S. M., Russell, J . A., & Feldman Barrett, L. (1998). Structure of current affect: Integration and beyond. Manuscript under revision. Zevon, M. A., & Tellegen, A. (1982). The structure of mood change: An idiographic/nomothetic analysis. Journal of Personality and Social Psychology, 43. 111 -122. Appendix I The English version of all the affect items Pleasure Unhappy - Happy Pleased - Annoyed Dissatisfied - Satisfied Contented - Melancholic Despairing - Hopeful Relaxed - Bored Semantic Differential Items Arousal Relaxed - Stimulated Excited - Calm Sluggish - Frenzied Jittery-Dull Sleepy - Wide awake Aroused - Unaroused 181 "Adjectives" Format with 5 response options (1= not at all, 2 = a little, 3 - moderately, 4 = quite a bit, 5 = extremely) CMQ Scales Pleasant happy pleased content satisfied Unpleasant miserable troubled unhappy dissatisfied Activated aroused hyperactivated activated Deactivated sleepy still quiet Thayer's Scales Energy active energetic vigorous lively full of pep Tiredness tired drowsy wide awake (-) wakeful (-) sluggish Tension jittery intense fearful clutched up tense anxious Calmness placid calm at rest Larsen & Diener's Scales Activated Pleasant enthusiastic elated excited euphoric peppy Unactivated Unpleasant dull tired drowsy sluggish bored droopy Activated Unpleasant distressed annoyed fearful nervous jittery anxious Unactivated Pleasant relaxed at rest calm serene at ease Watson & Tellegen's Scales Positive Affect Negative Affect interested excited strong enthusiastic proud inspired determined attentive active alert distressed upset guilty scared hostile irritable ashamed nervous jittery afraid "Agree-Disagree" Response Format with 5 response options (1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = strongly agree) CMQ Scales Pleasure I was happy. I was satisfied. I felt content. Activated Unpleasant I was unhappy. I was dissatisfied. I was miserable. Deactivated I was filled with energy and tension. My body felt activated. I was feeling stirred up. My body felt still. I was feeling quiet. Thayer's Scales Energy I was full of pep and energy. My mind was quick and alert. I felt energetic and vigorous. Tension For some reason, I was feeling stirred up and jittery. I felt tense. I felt "clutched up". I was feeling pretty fearful at that moment. Tiredness I was having trouble staying awake. I felt tired. My body was sluggish. Calmness I was feeling placid. All of me felt at rest. I was too calm to worry about anything. My pace was leisurely and quiescent. Larsen & Diener's Scales Activated Pleasant I was feeling elated. Life was exciting. I was enthused about what I was doing. I was in a state of euphoria. Activated Unpleasant I was feeling "jittery." I was annoyed by something. Right then, life felt like one big stress. Unactivated Unpleasant Everything seemed boring. I felt droopy and drowsy. Things seemed pretty dull right then. Unactivated Pleasant I was quite serene. I was relaxed. I was feeling calm and rested. Watson & Tellegen's Scales Positive Affect I felt interested in what I was doing at that moment. I felt pretty enthusiastic about my life right then. I was feeling inspired. Right then, life felt terrific. I was feeling energetic and positive. I felt determined. I felt very focused and on task. I was feeling lively and cheerful. Negative Affect For some reason, I had been feeling sort of nervous. I felt on edge. I felt worried. I was feeling pretty angry at that moment. I was bothered by something. Right then, life felt like one big struggle I was full of guilt and remorse. "Describes Me" Response Format with 4 response options (1 = not at all, 2 = not very well, 3 = somewhat, 4 = very well) CMQ Scales Pleasant Unpleasant My mood was positive. My mood was NOT good. Overall, I was satisfied. I was feeling troubled. I felt unhappy. Activated Deactivated I was full of energy and tension. I was keyed up. I was stirred up. I was feeling placid, low in energy. My internal engine was running slowly and smoothly. My body was in a quiet, still state. My mind and body were resting, near sleep. Thayer's Scales Energy Right then, I was brimming with vigor. I felt full of energy. I felt active and peppy. I felt very lively. Tension I felt jittery for some reason. I felt a lot of tension. My tension was quite intense. I was anxious. Tiredness I was so tired. I felt drowsy. I felt sluggish and slow. Calmness I felt placid and still. Right then, I was calm about things. My body was at rest. Larsen & Diener's Scales Activated Pleasant I felt enthusiastic. I felt ecstatic. I felt elated. Unactivated Unpleasant I felt droopy and drowsy. Things were dull and boring. Activated Unpleasant For some reason, I felt scared and afraid. I felt irritated at something. I felt distressed. At that moment, I felt nervous. Unactivated Pleasant Serene - that was what I felt. My mind was pleasantly at ease. I was relaxed. Positive Affect Negative Affect Right then, I was sharp and attentive. I felt proud of myself. I felt ecstatic. I felt very inspired. I felt enthusiastic. I felt set and determined about something right then. I felt powerful and strong. I felt alive and active. I felt guilty about something that I had said or done. I felt irritated at something. I felt distressed. I felt ashamed of myself. For some reason, I felt scared and afraid. I felt angry. I felt jittery [for some reason]. At that moment, I felt nervous. I felt disturbed and upset. 186 Appendix II The English version of NEO FFI This questionnaire consists of 60 statements. Use the following 5-point scale as a guide, write a number beside each statement to indicate how much you agree with it. Describe yourself as you are generally or typically. 1 2 3 4 5 Strongly Disagree Neutral Agree Strongly Disagree Agree 1. I am not a worrier. 2. I like to have a lot of people around me. 3. I don't like to waste my time daydreaming. 4. I try to be courteous to everyone I meet. 5. I keep my belongings clean and neat. 6. I often feel inferior to others. 7. I laugh easily. 8. Once I find the right way to do something, I stick to it. 9. I often get into arguments with my family and co-workers. 10. I'm pretty good about pacing myself so as to get things done on time. 11. When I'm under a great deal of stress, sometimes I feel like I'm going to pieces. 12.1 don't consider myself especially "light-hearted". 13.1 am intrigued by the patterns I find in art and nature. 14. Some people think I'm selfish and egotistical. 15.1 am not a very methodical person. 16.1 rarely feel lonely or blue. 17.1 really enjoy talking to people. 18.1 believe letting students hear controversial speakers can only confuse and mislead them. 19.1 would rather cooperate with others than compete with them. 20.1 try to perform all the tasks assigned to me conscientiously. 21.1 often feel tense and jittery. 22.1 like to be where the action is. 23. Poetry has little or no effect on me. 24.1 tend to be cynical and skeptical of others' intentions. 187 25.1 have a clear set of goals and work toward them in an orderly fashion. 26. Sometimes I feel completely worthless. 27.1 usually prefer to do things alone. 28.1 often try new and foreign foods. 29.1 believe that most people will take advantage of you if you let them. 30.1 waste a lot of time before settling down to work. 31.1 rarely feel fearful or anxious. 32.1 often feel as if I'm bursting with energy. 33.1 seldom notice the moods or feelings that different environments produce. 34. Most people I know like me. 35.1 work hard to accomplish my goals. 36.1 often get angry at the way people treat me. 37.1 am a cheerful, high-spirited person. 38.1 believe we should look to our religious authorities for decisions on moral issues. 39. Some people think of me as cold and calculating. 40. When I make a commitment, I can always be counted on to follow through. 41. Too often, when things go wrong, I get discouraged and feel like giving up. 42.1 am not a cheerful optimist. 43. Sometimes when I am reading poetry or looking at a work of art, I feel a chill or wave of excitement. 44. I'm hard-headed and tough-minded in my attitudes. 45. Sometimes I'm not as dependable or reliable as I should be. 46.1 am seldom sad or depressed. 47. My life is fast-paced. 48.1 have little interest in speculating on the nature of the universe or the human condition. 49.1 generally try to be thoughtful and considerate. 50.1 am a productive person who always gets the job done. 51.1 often feel helpless and want someone else to solve my problems. 52.1 am a very active person. 53.1 have a lot of intellectual curiosity. 54. If I don't like people, I let them know it. 55.1 never seem to be able to get organized. 56. At times I have been so ashamed I just wanted to hide. 57.1 would rather go my own way than be a leader of others. 58.1 often enjoy playing with theories or abstract ideas. 188 59. If necessary, I am willing to manipulate people to get what I want. 60.1 strive for excellence in everything I do. Appendix III. English Questionnaire Packet 190 Remembered Moments Questionnaire INSTRUCTIONS: PLEASE READ CAREFULLY This is a study of people's feelings and moods at various times. We are especially concerned with everyday, ordinary feelings. We are therefore hoping to study cases of actual feelings. It doesn't matter whether they are the dramatic emotions you read about in textbooks or not. To obtain a good sample of accurate reports, we need to ask you to remember a particular moment. Please think back to yesterday. Specifically, recall the time "just before breakfast." (If you didn't have breakfast yesterday, simply recall that approximate time of day.) It is important that you remember a specific moment accurately. So, please search your memory and try to recall where you were, what you were doing at that time, who you were with, and what you were thinking. Now select a particular moment that is especially clear in your memory. (If you really have no recollection of the time "just before breakfast," please search your memory for the closest time that you do recall accurately.) Have you selected the moment? We shall call that moment your REMEMBERED MOMENT. Please try to estimate the time of that moment. Write the time here: AM / PM Now, please try to remember how you were feeling at that moment. On the following pages are four questionnaires. They all concern how you felt at the REMEMBERD MOMENT. We realize that some of the items in the packet look similar to one another, but that is okay. Just answer all items spontaneously, as best you can. Respond to each item in order - don't go back and check your answers to previous questions. Your responses are completely confidential and are used for research purpose only. VERSION 1 October 5, 1997 191 RMQ VAN 2 First, please use the following scales to describe how you were feeling during your REMEMBERED MOMENT. Some of the word pairs may seem unusual, but you probably felt more one way than another. So, for each pair below, put a check-mark somewhere along the line. Example : : J_ : : : 1. Unhappv Happv 2. Relaxed Stimulated 3. Pleased Annoved 4. Excited Calm 5. Dissatisfied Satisfied 6. Sluqgish Frenzied 7. Contented Melancholic 8. Jitterv : : : : Dull 9. Despairinq Hopeful 10. Sleepy Wide awake 11. Relaxed Bored 12. Aroused Unaroused When you have finished, please be sure that there is one check on each line. 192 RMQ VAN 3 This scale consists of a number of words that describe feelings, mood, and emotions. Please indicate to what extent you felt each of these at the REMEMBERED MOMENT. Use the following scale to record your answers. 1 2 3 4 5 Not at all A little Moderately Quite a bit Extremely 1.. activated 26. energetic 51. pleased 2. active 27. enthusiastic 52. proud 3. afraid 28. euphoric 53. quiet 4. agitated 29. excited 54. relaxed 5. ; alert 30. _ _ fearful 55. relieved 6. annoyed 31. frenzied 56. sad 7. anxious 32. full of pep 57. satisfied 8. aroused 33. gloomy 58. scared 9. ashamed 34. guilty 59. secure 10. at ease 35. happy 60. serene 11. at rest 36. hostile 61. shaky 12. attentive 37. hyperactivated 62. . sleepy 13. blue 38. inspired 63. _ _ sluggish 14. bored 39. intense 64. soothed 15. calm 40. interested 65. still 16. clutched up 41. irritable 66. strong 17. content 42. jittery 67. tense 18. determined 43. lively 68. tired 19. dissatisfied 44. melancholy 69. tranquil 20. distressed 45. miserable 70. troubled 21. down 46. nervous 71. unhappy 22. droopy 47. panicky 72. upset 23. drowsy 48. peaceful 73. vigorous 24. dull 49. peppy 74. wakeful 25. elated 50. placid 75. wide awake Please double-check that you've put one number on each blank. 193 RMQ VAN 4 This questionnaire contains 61 statements about how you felt at the R E M E M B E R E D MOMENT. Please indicate how much you agree or disagree with each statement. Please use the following scale to record your answer. 1 2 3 4 5 Strongly Disagree Disagree Neutral Agree Strongly Agree 1. I was happy. 2. For some reason, I was feeling stirred up and jittery. 3. I was satisfied. 4. I felt interested in what I was doing at that moment. 5. I was feeling placid. 6. I was filled with energy and tension. 7. I was feeling "jittery". 8. For some reason, I had been feeling sort of nervous. 9. I was miserable. 10.1 was quite serene. 11.1 felt tired. 12.1 felt pretty enthusiastic about my life right then. 13.1 felt content. 14.1 was full of pep and energy. .15.1 was unhappy. 16.1 felt on edge. 17.1 was having trouble staying awake. 18.1 was feeling lively and cheerful. 19.1 was feeling elated. 20. Everything seemed boring. 21.1 was feeling inspired. 22.1 was feeling quiet. 23.1 was full of guilt and remorse. 24.1 was dissatisfied. 25.1 felt worried. 26.1 was feeling stirred up. 27.1 felt tense. 28. My body felt activated. 29. Right then, life felt terrific. 30.1 was too calm to worry about anything. 194 RMQ VAN 5 1 2 3 4 5 Strongly Disagree Disagree Neutral Agree Strongly Agree 31. Things seemed pretty dull right then. 32.1 felt very focused and on task. 33.1 was annoyed by something. 34.1 felt sad and blue. 35.1 was relaxed. 36.1 was feeling pretty angry at that moment. 37.1 was in a panic. 38. My mind was quick and alert. 39. My body was sluggish. 40. Life was exciting. 41.1 was feeling energetic and positive. 42.1 felt agitated. 43.1 was blissfully at ease. 44.1 felt droopy and drowsy. 45:1 felt energetic and vigorous. 46.1 felt "clutched up". 47.1 was bothered by something. 48.1 felt frenzied. 49. All of me felt at rest. 50. Right then, life felt like one big stress. ' 51.1 was feeling calm and rested. 52.1 felt determined. 53.1 was sadly slow. 54.1 was enthused about what I was doing. 55. My pace was leisurely and quiescent. 56. Right then, life felt like one big struggle. 57.1 was in a state of frenzied excitement. 58.1 was in a state of euphoria. 59.1 was feeling pretty fearful at that moment. 60. My body felt still. 61.1 was floating in a sea of tranquillity. Now please go back and make sure that you have put one number on each of the 61 blanks. 195 R M Q V A N 6 Please use the following response options to indicate how well each phrase describes your feeling at the REMEMBERED MOMENT. 1. 2 3 4 Not at all Not very well Somewhat Very well 1. My mood was positive. 2. I was full of energy and tension. 3. I felt jittery for some reason. 4. Right then, I was sharp and attentive. 5. For some reason, I felt scared and afraid. 6. Right then, I was brimming with vigor. 7. I felt guilty about something that I had said or done. 8. I felt enthusiastic. 9. My mood was NOT good. 10.1 felt proud of myself. 11.1 was feeling placid, low in energy. 12. Right then, I was calm about things. 13.1 felt irritated at something. 14. Serene - that was what I felt. 15.1 was so tired. 16.1 felt droopy and drowsy. 17. Overall, I was satisfied. 18.1 felt very inspired. 19.1 was keyed up. 20.1 felt a lot of tension. 21.1 felt distressed. 22.1 felt full of energy. 23.1 felt set and determined about something right then. 24.1 felt ecstatic. 25.1 was feeling troubled. 26. My mind was racing. 27.1 felt ashamed of myself. 28. My internal engine was running slowly and smoothly. 29.1 felt placid and still. 30.1 felt powerful and strong. 196 RMQ VAN 7 1 2 3 4 Not at all Not very well Somewhat Very well 31. My body was trembling with tension. 32. My mind was pleasantly at ease. 33.1 felt drowsy. 34. My mind was frantically agitated. 35. Things were dull and boring. 36.1 was stirred up. 37.1 felt alive and active. 38. My mind was soothed and unperturbed. 39. My tension was quite intense. 40.1 was surrounded with gloom and doom. 41.1 felt active and peppy. 42.1 felt angry. 43. My mood was melancholy and down. 44.1 felt elated. 45.1 felt unhappy. 46. My body was in a quiet, still state. 47. My body was at rest. 48.1 felt disturbed and upset. 49.1 was relaxed. 50.1 felt sluggish and slow. 51.1 was anxious. 52. My body was tranquil. 53. At that moment, I felt nervous. 54.1 felt very lively. 55. My mind and body were resting, near sleep. Now, please go back and make sure that you have put one number in each of the 55 blanks. Personal Information Today's Date: Sex: Age: Language you learned first in life: Language you are currently best with: What is your ethnic background (e.g., European, Chinese, etc) Country of Birth: Number of years living in Canada / USA: 198 N E O - FFI This questionnaire consists of 60 statements. Use the following 5-pdint scale as a guide, write a number beside each statement to indicate how much you agree with it. Describe yourself as vou are generally or typically. 1 2 3 4 5 Strongly Disagree Neutral Agree Strongly Disagree Agree 1. I am not a worrier. 2. I like to have a lot of people around me. 3. I don't like to waste my time daydreaming. 4. I try to be courteous to everyone I meet. 5. I keep my belongings clean and neat. 6. I often feel inferior to others. 7. I laugh easily. 8. Once I find the right way to do something, I stick to it. 9. I often get into arguments with my family and co-workers. 10. I'm pretty good about pacing myself so as to get things done on time. 11. When I'm under a great deal of stress, sometimes I feel like I'm going to pieces. 12. I don't consider myself especially "light-hearted". 13. I am intrigued by the patterns I find in art and nature. 14. Some people think I'm selfish and egotistical. 15. I am not a very methodical person. 16. I rarely feel lonely or blue. 17. I really enjoy talking to people. 18. I believe letting students hear controversial speakers can only confuse and mislead them. 19. I would rather cooperate with others than compete with them. 20. I try to perform all the tasks assigned to me conscientiously. 21. I often feel tense and jittery. 22. I like to be where the action is. 23. Poetry has little or no effect on me. 24. I tend to be cynical and skeptical of others' intentions. 25. I have a clear set of goals and work toward them in an orderly fashion. 26. Sometimes I feel completely worthless. 27. I usually prefer to do things alone. 28. I often try new and foreign foods. 29. I believe that most people will take advantage of you if you let them. 30. I waste a lot of time before settling down to work. 199 1 2 3 4 5 Strongly Disagree Neutral Agree Strongly Disagree Agree 31. I rarely feel fearful or anxious. 32. I often feel as if I'm bursting with energy. 33. I seldom notice the moods or feelings that different environments produce. 34. Most people I know like me. 35. I work hard to accomplish my goals. 36. I often get angry at the way people treat me. 37. I am a cheerful, high-spirited person. 38. I believe we should look to our religious authorities for decisions on moral issues. 39. Some people think of me as cold and calculating. 40. When I make a commitment, I can always be counted on to follow through. 41. Too often, when things go wrong, I get discouraged and feel like giving up. 42. I am not a cheerful optimist. 43. Sometimes when I am reading poetry or looking at a work of art, I feel a chill or wave of excitement. 44. I'm hard-headed and tough-minded in my attitudes. 45. Sometimes I'm not as dependable or reliable as I should be. 46. I am seldom sad or depressed. 47. My life is fast-paced. 48. I have little interest in speculating on the nature of the universe or the human condition. 49. I generally try to be thoughtful and considerate. 50. I am a productive person who always gets the job done. 51. I often feel helpless and want someone else to solve my problems. 52. I am a very active person. 53. I have a lot of intellectual curiosity. 54. If I don't like people, I let them know it. 55. I never seem to be able to get organized. 56. At times I have been so ashamed I just wanted to hide. 57. I would rather go my own way than be a leader of others. 58. I often enjoy playing with theories or abstract ideas. 59. If necessary, I am willing to manipulate people to get what I want. 60. I strive for excellence in everything I do. Appendix IV. Spanish Questionnaire Packet 2 0 1 CUESTIONARIO DE MOMENTOS RECORD ADOS INSTRUCCIONES: POR FAVOR, LEA ATENTAMENTE. El siguiente estudio trata sobre los sentimientos y estados de animo que tiene la gente en distintos momentos. Estamos especialmente interesados en sentimientos normales y cotidianos. Es decir, queremos estudiar sentimientos autenticos. No importa si coinciden o no con las dramaticas emociones que se pueden encontrar en los libros. Para obtener un informe lo mas ajustado posible, necesitamos pedirle que recuerde un momento en concreto. Por favor, piense sobre lo que hizo ayer. Concretamente, recuerde el momento "justo despues del desayuno" (Si ayer no tomo el desayuno. sencillamente recuerde aproximadamente dicho momento del dia.) Es importante que recuerde un momento en concreto con la mayor exactitud posible. Por favor, busque en su memoria e intente recordar donde estaba, que estaba haciendo en aquel momento, con quien estaba, y sobre que estaba pensando. Ahora seleccione un momento en particular que este especialmente claro en su cabeza. (Si no tiene un recuerdo claro del momento "justo despues del desayuno". por favor, busque en su memoria el momento mas cercano que usted pueda recordar con exactitud.) I Ha seleccionado el momento ? Denominaremos este momento su MOMENTO RECORDADO. Por favor, intente estimar la duracion de dicho momento. Escriba aqui la duracion: Ahora trate de recordar como se sentia en aquel momento. En las proximas paginas encontrara cuatro cuestionarios. Todos ellos tratan sobre como se sentia usted durante el MOMENTO RECORDADO. Somos conscientes de que algunas de las cuestiones presentes en el cuestionario le pareceran muy similares entre si, pero es necesario sea de este modo. Sencillamente, conteste a todas la cuestiones de forma espontanea tan bien como pueda. Por favor, responda a cada cuestion en orden -no vuelva atras para comprobar sus respuestas a cuestiones previas. Sus respuestas son completamente confidenciales y se utilizan solo con propositos de investigation academica. VERSION 2 202 En primer lugar, por favor utilice las siguientes escalas para describir como se sentia durante el MOMENTO RECORDADO. Alguno de los pares de palabras le puede parecer poco usual, pero probablemente durante aquel momenta usted sentiria mas de una forma que de otra. Por lo tanto, para cada par de palabras, coloque una marca en algun punto de la linea que los separa. Ejemplo : : V : : : 1. *Infeliz : : : : : Feliz 2. Relaiado : : : : : : : : Estimulado 3. *Complacido ' : : : : : : Disgustado* 4. Excitado : : : : : : : : Tranquilo* 5. Insatisfecho : : : : ' Satisfecho 6. *Remol6n : : : : : : : : Frenetico 7. Contento : : : : : : : Melancolico 8. •Inquieto : : : : Desanimado 9. Desesperado : : : : : Ilusionado* 10. Adormilado : : : : : : Vigilante* 11. Relaiado : : : : : : : : Aburrido 12. Activado : : : : : : : : Apagado* Una vez que haya terminado, por favor asegurese de que solo hay una marca en cada linea. 203 Esta escala consiste en un conjunto de palabras que describen sentimientos, estados de ammo, y emociones. Por favor, indique en que medida sintio durante el MOMENTO RECORDADO cada uno de los siguientes sentimientos o emociones. Por favor, utilice la siguiente escala para indicar sus respuestas 1 2 3 4 5 Nada Unpoco Moderadamente Bastante Extremadamente 1. activado 22. gozoso 43. complacido* 2. activo 23. energico 44. orgulloso 3. miedo 24. entusiasmado* 45. tranquilo 4. alerta 25. euforico 46. relajado 5. disgustado 26. excitado 47. satisfecho 6. ansioso 27. atemorizado 48. asustado 7. activado 28. impetuoso* 49. sereno 8. avergonzado 29. culpable 50. adormilado 9. a gusto 30. feliz 51. remolon* 10. descansado 31. hostil 52. quieto* 11. atento 32. hiperactivo 53. fuerte 12. aburrido 33. inspirado 54. tenso 13. tranquilo 34. intenso 55. cansado 14. atenazado 35. interesado 56. preocupado 15. contento 36. irritable 57. infeliz 16. resuelto* 37. inquieto* 58. fastidiado 17. insatisfecho 38. animado 59. vigoroso 18. estresado* 39. miserable 60. despierto* 19. alicaido 40. nervioso 61. vigilante* 20. somnoliento 41. dinamico* 21. desanimado 42. sosegado Por favor, compruebe de nuevo que ha puesto un numero en cada espacio 204 Este cuestionario contiene 53 frases sobre como se sentia en el M O M E N T O RECORD ADO. Por favor, indique su grado de acuerdo con cada una de las frases. Por favor, utilice la siguiente escala para indicar su respuesta. 1 2 3 4 5 Total Desacuerdo Neutral Acuerdo Total Desacuerdo Acuerdo 1. Estaba feliz. 2. Por alguna razon, me sentia agitado e inquieto*. 3. Estaba satisfecho. 4. Me sentia interesado en lo que estaba haciendo en ese momento. 5. Me sentia sosegado. 6. Estaba lleno de energia y tension. 7. Me sentia "inquieto"*. 8. Por alguna razon, habia sentido algiin tipo de nervios. 9. Me sentia miserable. 10. Estaba bastante sereno. 11. Me sentia cansado. 12. Me sentia muy entusiasmado sobre mi vida justo en ese momento. 13. Me sentia contento. 14. Me sentia lleno de impetu y energia. 15. Me sentia infeliz. 16. Me sentia al limite. 17. Estaba teniendo problemas para mantenerme despierto. 18. Me sentia animado y alegre. 19. Me sentia gozoso. 20. Todo me parecia aburrido. 21. Me sentia inspirado. 22. Me sentia tranquilo. 23. Estaba lleno de culpa y remordimiento. 24. Estaba insatisfecho. 25. Me sentia preocupado. 26. Me sentia agitado. 205 27. Me sentia tenso. 28. Mi cuerpo estaba activado. 29. Justo en ese momento, la vida me parecia fabulosa. 30. Estaba demasiado tranquilo para preocuparme por algo. 31. Las cosas me parecian bastante torpes justo en ese momento. 32. Me sentia bastante concentrado e inmerso en lo que estaba haciendo. 33. Estaba disgustado por algo. 34. Estaba relajado. 35. Me sentia bastante enfadado en aquel momento. 36. Mi mente estaba rapida y alerta. 37. Mi cuerpo estaba remolon*. 38. La vida era excitante. ^ 39. Me sentia energico y positive 40. Me sentia alicaido y somnoliento. 41. Me sentia energico y vigoroso. 42. Me sentia "atenazado" 43. Estaba molesto por algo. 44. Todo yo me sentia descansado. 45. Justo en ese momento, sentia la vida como un gran estres.* 46. Me sentia tranquilo y descansado. 47. Me sentia resuelto. 48. Estaba entusiasmado por lo que estaba haciendo. 49. Mi ritmo era pausado y reposado. 50. Justo en ese momento, sentia la vida como una gran lucha.* 51. Estaba en un estado de euforia. 52. Me sentia bastante atemorizado en aquel momento 53. Mi cuerpo se sentia quieto. Ahora, por favor, repase y asegurese de que ha puesto un numero en cada uno de los 53 espacios. 206 Por favor, utilice las siguientes opciones de respuesta para indicar en que medida cada frase describe adecuadamente su sentimiento durante el MOMENTO RECORDADO. 1 2 3 4 En absoluto No muy bien Bastante bien Muy bien 1. M i estado de animo era positive 2. Estaba lleno de energia y tension. 3. Me sentia inquieto por alguna razon. 4. Justo en ese momento, me sentia agudo y atento. 5. Por alguna razon, me sentia asustado y con miedo. 6. Justo en ese momento, estaba repleto de vigor. 7. Me sentia culpable sobre algo que habia dicho o hecho. 8. Me sentia entusiasmado. 9. M i estado de NO era bueno. 10. Me sentia orgulloso de mi mismo. 11. Me sentia sosegado*, bajo de energia. 12. Justo en ese momento, me sentia tranquilo. 13. Me senti irritado por algo. 14. Sereno-asi es como me sentia. 15. Me sentia tan cansado. 16. Me sentia alicaido y somnoliento. 17. En general, me sentia satisfecho. 18. Me sentia muy inspirado. 19. Me sentia emocionado. 20. Sentia mucha tension. 21. Me sentia estresado. 22. Me sentia lleno de energia. 23. Me sentia dispuesto* y resuelto* sobre algo justo en ese momento. 24. Me sentia extasiado. 25. Me sentia preocupado. 26. Me sentia avergonzado de mi mismo. 207 27. Mi maquinaria interna estaba funcionando lenta y suavemente. 28. Me sentia pausado y quieto. 29. Me sentia poderoso y fuerte. 30. Mi mente estaba muy a gusto. 31. Me sentia somnoliento. 32. Las cosas eran torpes y aburridas. 33. Estaba agitado. 34. Me sentia vivo y active 35. Mi tension era muy intenso. 36. Me sentia activo y dinamico. 37. Me sentia enfandado. 38. Me sentia gozoso. 39. Me sentia infeliz. 40. Mi cuerpo estaba tranquilo y en un estado de calma*. 41. M i cuerpo estaba descansado. 42. Me sentia alborotado y fastidiado. 43. Estaba relaj ado. 44. Me sentia remolon y lento. 45. Estaba ansioso. 46. En aquel momento, me sentia nervioso. 47. Me sentia muy animado. 48. Mi cuerpo y mente estaban descansando, a punto de dormirse. Ahora, por favor, repase y asegurese de que ha puesto un numero en cada uno de los 48 espacios. 208 Fecha actual: Sexo: Edad: Lengua nativa: Lengua que mejor maneja: Grupo etnico: Pais de nacimiento: Numero de afios viviendo en Espafla: 209 INVENTARIO DE PERSONALIDAD NEO-PI-R Forma S PT. Costa Jr. y R.R. Mc Crae, 1992 Adaptation de M.D. Avia, I994 Instrucciones Lea cuidadosamente, por favor, estas instrucciones antes de empezar. Marque sus respuestas en la hoja de respuestas y escriba solamente en el lugar que se le indica. NO escriba nada en este cuadernillo. En la hoja de respuestas adjunta marque, por favor, su nombre en el espacio indicado. Indique su sexo poniendo una marca en la casilla correspondiente bajo el tftulo "Sexo". Escriba la fecha y su numero de indentificacion, si es que le han dado uno. Marque "Yo mismo" en el espacio donde pone "Persona a quien se califica", puesto que Vd. se esta describiendo a si mismo. Escriba su edad y marque la casilla proxima a "S" en el espacio que se llama "Forma del NEO". Este cuestionario consta de 240 enunciados. Lealos con atencion y rodee con un cfrculo la respuesta que corresponda mejor a su acuerdo o desacuerdo con ella. Marque "TD" si la afirmacion es completamente falsa o Vd. esta en total desacuerdo con ella. TD D N A TA Marque "D" si la afirmacion es fundamentalmente falsa o Vd. esta en desacuerdo con ella. TD D N A TA Marque "N" si la afirmacion es aproximadamente lo mismo de cierta o de falsa, si no puede decidirse, o si se considera neutro en relation con esta afirmacion. TD D N A TA Marque "A" si la afirmacion es fundamentalmente cierta o si Vd. esta de acuerdo con ella. TD D N A TA Marque "TA" si la afirmacion es completamente cierta o Vd. estatotalmente de acuerdo con eHa TD D N A TA No hay respuestas correctas ni incorrectas, y no necesita ser un experto para rellenar este cuestionario. Describase a si mismo de forma sincera y exprese sus opiniones de la forma mas precisa posible. Responda a todos los items. Observe que las respuestas estan numeradas en la hoja de respuestas, y asegurese de que marca las respuestas en el espacio correspondiente al numero adecuado. Si se equivoca o cambia de opinion, jNO BORREI. Ponga una cruz en la respuesta incorrecta y rodee con un cfrculo la respuesta correcta. Cuando haya contestado los 240 items, responda a las tres preguntas A B y C de la hoja de respuestas. Pase la hoja y empiece con el item 1. 210 .. INU soy una persona que se preocupe mucho 2. La mayoria de la gente que conozco me cae muy simpatica 3. Tengo una imaginacibn muy activa 4. Tiendo a ser cinico y esceptico respecto a las intenciones de los demas 5. Se me conoce por mi prudencia y sentido comun 6. Con frecuencia me irrita la forma en que me trata la gente 7. Huyo de las multitudes 8. Los intereses esteticos y artisticos no son muy importantes para mi 9. No soy astuto ni taimado 10. Prefiero hacer planes abiertos mas que planificarme todo de antemano 11. Rara vez me siento solo o triste 12. Soy dominante, energico y asertivo 13. Sin emociones fuertes la vida careceria de interes para mi 14. Algunas personas creen que soy egofsta y egocentrico 15. Trato de realizar concienzudamente todas las cosas que se me encomiendan 16. Al tratar con los demas siempre temo hacer una patochada 17. Tanto en el trabajo como en Is diversion tengo un estilo pausado 18. Tengo unas costumbres y opiniones bastante arraigadas 19. Preferiria cooperar con los demas a competir con ellos 20. No me enfado por nada, soy un poco pasota 21. Rara vez me excedo en algo 22. A menudo anhelo tener experiencias emocionantes 23. Con frecuencia disfruto jugando con teorias o ideas abstractas 24. No me importa jactarme de mis talentos y logros 25. Soy bastante bueno en organizarme para terminar las cosas a tiempo 26. Con frecuencia me siento indefenso y quiero que otro resuelva mis problemas 27. Literalmente, nunca he saltado de alegria 28. Considero que dejar que los jovenes oigan a personas cuyas opiniones son polemicas solo puede confundirlos o equivocarlos 29. Los lideres politicos deberian ser mas conscientes del lado humano de sus programas 30. He hecho bastantes tonterias a lo largo de mi vida 31. Me asusto con facilidad 32. No me gusta mucho charlar con la gente 33. Intento que todos mis pensamientos sean realistas y no dejar que vuele la imaginacion 34. Creo que la mayoria de la gente tiene en general buena intention 35. No me tomo muy en serio mis deberes civicos, como ir a votar 36. Soy una persona apacible 37. Me gusta tener mucha gente alrededor 38. A veces me quedo totalmente absorto en la musica que escucho 39. Si es necesario, estoy dispuesto a manipular a la gente para conseguir lo que quiero 40. Tengo mis cosas bien cuidadas y limpias 41. A veces me parece que no valgo absolutamente nada 42. A veces no soy capaz de ser todo lo asertivo que debiera 43. Rara vez experimento emociones fuertes 44. Trato de ser cortes con todo el que conozco 45. A veces no soy tan formal ni fiable como debiera 211 46. Rara vez me siento cohibido cuando estoy con gente 47. Cuando hago cosas, las hago con energia 48. Creo que es interesante aprender y desarrollar nuevos hobbies 49. Puedo ser sarcastico y mordaz si lo necesito 50. Tengo unos objetivos claros y me esfuerzo por alcanzarlos de forma ordenada 51. Me cuesta resistir mis deseos 52. No me gustaria pasar las vacaciones en Las Vegas 53. Encuentro aburridas las discusiones filosoficas 54. Prefiero no hablar de mi o de mis exitos 55. Pierdo mucho tiempo hasta que me pongo a trabajar 56. Creo que soy capaz de enfrentarme a la mayor parte de mis problemas 57. A veces he experimentado intensa alegria o arrebato 58. Considero que las leyes y normas sociales deberian cambiar para reflejar las necesidades de un mundo cambiante 59. Soy poco sentimental y duro en mis actitudes 60. Pienso muy bien las cosas antes de llegar a una decision 61. Rara vez me siento con miedo o ansioso 62. Se me conoce como una persona calida y cordial 63. Tengo mucha fantasia 64. Creo que la mayoria de la gente se aprovechari'a de uno si se le dejara 65. Me mantengo informado y por lo general tomo decisiones inteligentes 66. Se me tiene por colerico y de genio vivo 67. En general prefiero hacer las cosas solo 68. Me aburre ver ballet o danza moderna 69. Aunque quisiera, no podrfa enganar a nadie 70. No soy una persona muy metodica 71. Rara vez estoy triste o deprimido 72. A menudo he sido un Kder en los grupos en que he estado 73. Para mi son importantes mis sentimientos sobre las cosas - 74. Algunas personas piensan de mi que soy frio y calculador 75. Pago mis deudas puntualmente y en su totalidad 76. En ocasiones he estado tan avergonzado que he querido esconderme 77. Probablemente mi trabajo sea lento pero constante 78. Cuando encuentro la manera de hacer algo, me aferro a ella 79. Me resulta dificil expresar rabia, aunque lleve razon 80. Cuando empiezo un programa para mejorar algo mfo, lo habitual es que lo abandone a los pocos dias 81. Me cuesta poco resistir la tentacion 82. A veces he hecho cosas por mera excitation, buscando emocionarme 83. Disfruto resolviendo problemas o puzzles 84. Soy mejor que la mayoria de la gente, y yo lo se 85. Soy una persona productiva, que siempre termina su trabajo 86. Cuando estoy bajo un fuerte estres, a veces siento que me voy a desmoronar 87. No soy un alegre optimista 88. Considero que deberiamos contar con las autoridades religiosas para tomar decisiones sobre cuestiones morales 89. Hagamos lo que hagamos por los pobres y los ancianos, nunca sera demasiado 90. En ocasiones primero actuo y luego pienso 212 91. A menudo me siento tenso e inquieto 92. Mucha gente cree que soy aigo frio y distante 93. No me gusta perder el tiempo sonando 94. Me parece que la mayoria de la gente con la que trato es honrada y fidedigna 95. Muchas veces no preparo de antemano lo que tengo que hacer 96. No se me considera una persona diffcil ni de mal genio 97. Si estoy solo mucho tiempo, siento mucha necesidad de la gente 98. Me despiertan curiosidad las formas que encuentro en el arte y en la naturaleza 99. Ser absolutamente honrado es un mala manera de hacer negocios 100. Me gusta tener cada cosa en su sitio, de forma que sepa exactamente donde esta 101. A veces he sentido una sensation profunda de culpa o pecado 102. En reuniones por lo general prefiero que hablen otros 103. Rara vez pongo mucha atencion en mis sentimientos del momento 104. Por lo general trato de pensar en los demas y ser considerado 105. A veces hago trampas cuando juego a hacer solitarios 106. No me averguenzo mucho si la gente se rie de mi y me toma el pelo 107. A menudo siento como si estuviera reventando de energia 108. Con frecuencia pruebo comidas nuevas y extranjeras 109. Si alguien no me cae simpatico, se lo digo 110. Trabajo mucho para conseguir mis metas 111. Cuando como las comidas que mas me gustan, tiendo a comer demasiado 112. Tiendo a evitar las peliculas demasiado violentas y terrorificas 113. A veces pierdo el interes cuando la gente habla de cuestiones muy abstractas y teoricas 114. Trato de ser humilde 115. Me cuesta forzarme a hacer lo que tengo que hacer 116. En situaciones de emergencia mantengo la cabeza fria 117. A veces reboso felicidad 118. En mi opinion, las distintas ideas sobre lo quo esta bien y lo que esta mal que tienen otras sociedades pueden ser validas para ellas 119. Los mendigos no me inspiran simpatia 120. Antes de emprender una action, siempre considero sus consecuencias 121. Rara vez me inquieta el futuro 122. Me divierte mucho hablar con la gente 123. Me gusta concentrarme en una fantasia o ensuefio y explorar todas sus posibilidades, dejandolas que crezcan y se desarrollen 124. Cuando alguien es agradable conmigo, me entran recelos 125. Estoy orgulloso de mi sensatez 126. Con frecuencia me producen aversion las personas con las que tengo que tratar 127. Prefiero los trabajos que me permiten trabajar solo, sin que me molesten los demas 128. La poesia tiene poco o ningun efecto sobre mi 129. Detestaria que alguien pensara de mi que soy un hipocrita 130. Parece que nunca soy capaz de organizarme 131. Cuando algo va mal, tiendo a culpabilizarme 132. Con frecuencia los demas cuentan conmigo para tomar decisiones 213 133. Experimento una gran variedad de emociones o sentimientos 134. No se me conoce por mi generosidad 135. Cuando me comprometo a algo, siempre se puede contar conmigo para llevarlo a termino 136. A menudo me siento inferior a los demas 137. No soy tan rapido ni tan animado como otras personas 138. Prefiero pasar el tiempo en ambientes conocidos 139. Cuando me han ofendido, solo trato de perdonar y olvidar 140. No me siento impulsado a conseguir el exito 141. Rara vez cedo a mis impulsos 142. Me gusta estar donde esta la action 143. Me gusta hacer puzzles de los que hacen estrujarte el cerebro 144. Tengo una opinion muy alta de mi mismo 145. Cuando empiezo un proyecto, casi siempre lo termino 146. Con frecuencia me resulta dificil decidirme 147. No me considero especialmente alegre 148. Considero que la fidelidad a los propios ideales y principios es mas importante que la mentalidad abierta 149. Las necesidades humanas deberian estar siempre por delante de consideraciones economicas 150. A menudo hago cosas de forma impulsiva 151. Me preocupo con frecuencia por cosas que podrfan salir mal 152. Me resulta facil sonreir y ser extrovertido con desconocidos 153. Si noto que mi cabeza empieza a divagar y a sonar, generalmente me ocupo en algo y empiezo a concentrarme en algun trabajo o actividad alternativo 154. Mi primera reaction es confiar en la gente 155. No parece que haya tenido exito completo en nada 156. Es dificil que yo pierda los estribos 157. Preferiria pasar las vacaciones en una playa muy frecuentada por gente que en una cabana aislada en el monte 158. Ciertos tipos de musica me producen una fascination sin limites 159. A veces consigo con artimahas que la gente haga lo que yo quiero 160. Tiendo a ser algo quisquilloso o exigente 161. Tengo una baja opinidn de m( mismo 162. Preferiria ir a mi aire a ser el lider de todos 163. Rara vez me doy cuenta del humor o los sentimientos que producen ambientes diferentes 164. A la mayoria de la gente que conozco le caigo simpatico 165. Me adhiero de forma estricta a mis principios eticos. 166. Me siento a gusto en presencia de mis jefes u otras figuras de autoridad 167. Habitualmente parece como si tuviera prisa 168. A veces hago cambios en la casa solo para probar algo diferente 169. Si alguien empieza a pelearse conmigo, yo estoy dispuesto a pelear tambien 170. Hago todo lo que puedo por conseguir lo maximo 171. A veces como hasta ponerme enfermo 172. Adoro la excitation de las montanas rusas 173. Tengo poco interes en especular sobre la naturaleza del universo o la condition humana 214 174. Siento que no soy mejor que los demas, independientemente de cual sea su condition 175. Cuando un proyecto se pone demasiado dificil, me siento inclinado a empezar uno nuevo 176. Puedo manejarme bastante bien en una crisis 177. Soy una persona alegre, animosa 178. Me considero de mentalidad abierta y tolerante con los estilos de vida de los demas 179. Creo que todos los seres humanos merecen respeto 180. Casi nunca tomo decisiones precipitadas 181. Tengo menos miedos que la mayoria de la gente 182. Tengo una fuerte relation emotional con mis amigos 183. De nino rara vez me divertia jugando a ser otra persona 184. Tiendo a pensar lo mejor de la gente 185. Soy una persona muy competente 186. A veces me he sentido amargado y resentido 187. Las reuniones sociales normalmente me resultan aburridas 188. A veces, cuando leo poesia o contemplo una obra de arte, siento un escalofrio o una ola de excitation 189. A veces intimido o adulo a la gente para que haga lo que yo quiero 190. No soy compulsivo con la limpieza 191. A veces las cosas me parecen demasiado sombrias y sin esperanza 192. En las conversaciones tiendo a ser el que mas habla 193. Me parece facil empatizar, sentir yo lo que los demas sienten 194. Pienso que soy una persona caritativa 195. Trato de hacer mi trabajo con cuidado, para que no haya que hacerlo otra vez 196. Si he dicho o hecho algo mal a una persona, me cuesta-mucho poder enfrentarme a ella de nuevo 197. Mi vida lleva un ritmo rapido 198. En vacaciones prefiero volver a un sitio conocido y fiable 199. Soy cabezota y testarudo 200. Me esfuerzo por llegar a la perfection en todo lo que hago 201. A veces hago las cosas impulsivamente y luego me arrepiento 202. Me atraen los colores llamativos y los estilos ostentosos 203. Tengo mucha curiosidad intelectual 204. Preferiria elogiar a otros que ser elogiado 205. Hay tantas pequenas cosas que hacer que a veces lo que hago es pasar de todas 206. Cuando parece que todo va mal, todavia puedo tomar buenas decisiones 207. Rara vez uso palabras como "jfantastico!" o "jsensational!" para describir mis experiencias 208. Creo que si la gente no tiene las cosas claras a los 25 anos, algo esta mal en ellos 209. Me inspiran simpatia los que son menos afortunados que yo. 210. Cuando voy de viaje, lo planifico cuidadosamente con antelacion 211. A veces me vienen a la cabeza pensamientos aterradores 212. Me tomo un interes personal por la gente con la que trabajo 213. Tendria dificultad para dejar que mi cabeza vagara sin control o direction 214. Tengo mucha fe en la naturaleza humana 215 215. Soy eficiente y eficaz en mi trabajo 216. Hasta las minimas molestias me pueden resultar frustrantes 217. Disfruto en las fiestas en las que hay mucha gente 218. Disfruto leyendo poesfas que se centran en sentimientos e imagenes, mas que en acontecimientos 219. Estoy orgulloso de mi astucia para tratar con la gente 220. Gasto un monton de tiempo buscando cosas que he perdido 221. Con demasiada frecuencia cuando las cosas van mal me siento desanimado y a punto de tirar la toalla 222. No me parece facil asumir el mando de una situation. 223. Cosas extrafias - como ciertos olores o los nombres de lugares lejanos-pueden evocarme fuertes estados de animo 224. Me aparto de mi camino por ayudar a los demas, si puedo 225. Tendria que estar muy enfermo para perder un dfa de trabajo 226. Cuando alguien que conozco hace tonterias, siento verguenza ajena 227. Soy una persona muy activa 228. Cuando voy a alguna parte sigo siempre el mismo camino .229. Con frecuencia me enzarzo en discusiones con mi familia y mis companeros 230. Soy un poco adicto al trabajo 231. Siempre soy capaz de mantener mis sentimientos bajo control 232. Me gusta ser parte del publico en los acontecimientos deportivos 233. Tengo una gran variedad de intereses intelectuales 234. Soy una persona superior 235. Tengo mucha auto-discipina 236. Soy bastante estable emocionalmente 237. Me rio con facilidad 238. Considero que la "nueva moralidad" de lo permisivo no es en absoluto moralidad 239. Antes preferiria ser conocido como " misericordioso" que como "justo" 240. Antes de contestar una pregunta, me lo pienso dos veces Appendix V. Chinese Questionnaire Packet 217 ^ l p i | 4 f f l M I . ( f o ^ ^ ^ ^ g , # B t t ; M M P B * « : W E J ^ M ^ M s t r i f e — - I S 4 t ^ « ; # * J a t — J t t * f l t M « j , m i t % ifc 3r E IB IB] ^ , ' * W ^ f ^ « r a « * P * I W ^ « ^ E ftttftillfWlftai 218 5fc## + —J*tt^ tf M T f f 2. «* ** 8 . MM — : — = — : — : io. _ = _ = _ = — = — = — : — : — : — : * * * * 2 1 9 1 2 3 4 5 1 . 1*1* 23 # 4 5 . 2 . l 24 46 . mm 3 . 25 47 . 4 . 26 H # 48 . 5 . 27 HI 4 9 . 6 . 28 & 5 0 . 7 . 29 5 1 . 8 . 30 52 . 9- ft* 1 2 - #£W 1 3 -is. ^mM.^. 20., 2 1 -2 2 . Hi^^JRl 3 1 . 32 . 33 . 34 . 35 36 . 37 . 38 . 39 . 40 . 4 1 . 42 . 43 . 44 . With mm. & * 53 . 54 . 5 5 . 56 . 57 . 58 , 59 60 6 1 mm 7C 220 • 1 2 3 4 5 # # E * 1. 3. & * # 5 * j f e . 5. 8 . * ^ « J S 0 , a « f j^-« f i ! .#« . i o . & * 4 M B H H # # . i 6 . a * T * * & j * * # . 1 8 . * N F P f f i l l . 2 0 . / f t : f r t t * # 4 » # f t $ f t H . 2 i . & 5 f c # 2 5 5 « }M . 22.&Jfc###. 2 3 . & # 5 K f t & 5 r * $ H * & m * . 1 2 3 4 5 T-mM + 31 mM ' # * 1*1 * 2 7 . f f c j f c * £ & f c . 2 9 . ^ ^ - ^ E ^ P — 3 i . m * " * L * H f ^ t t r * * P f f i * t W i ' * . 33. ffc.fc—*¥rffm'KS„ 34. ffc«l!HS*L. 35. ffc^ —*J5fc4**S*&. 3 7 . a ^ ^ l t i l # f e . 39. f f c * # & J ! r - * 3 ^ ' l f r ' & . 40. # f c * # & & & l £ * i S S J I * * i . . 4 2 . f J t ^ f i J # ^ ^ F ^ „ 43. 44. ffc$f!|ft'(BA48J^Ett&. 45. ^ ^ ^ p - ^ J ^ A ^ I ' J ^ — - J B ^ l j t f f i . 4 8 . a ^ # W f t a « - B % ^ E m ^ * . 49. ^ W « P # * « t # ^ ^ « # W . 51.0, $ B#^ E—-na WL m m a*&+. 53.ffcttJ!rtt:£#jhJ!*)t& + .-1 2 3 4 2. 0,%m^mnT^. 5. ftmm&mHg-'te* 6. ffe^IP—*l 355*7^. . 8. l o . f f c j S K l g fc. i4.'rt#--5t:*t*ilS;^f«l!lW. i 8 . & f t # M # j £ . 21 . ffeft#^0L 2 3 . a £ HP B% ff * * ft # ^ # 32- * * P S £ * # . 24. =ft*8f'J3£*o 25. a f t # « ' ^ . 1 2 3 4 28. f f c 5 M * * * M P ^ & « & . 29. a « f ! l ^ 7 * ^ 5 * J ? l S ! f t . 32. m^ * * P*W5 . « * ^ « t P 5 J . 33. ffeft#^J#^^ 0 34. a 5 t « - ^ 3 » ^ * L * ^ a i . 36. a * » « « ^ P # * f H » . 37. ffeft#<|£&„ 3 9 . a f t # ^ t t ^ o 4 2 . a * ^ J * ^ * J » ' ^ * * « L . 4 3 . & H $ I ! H S * . 4 4. a m m *r si & at J& « ' i t . 4 5 . a * # * A . -4 6 . & * B * J & l ! | * 3 r * . 4 7 . a * # ^ n r ^ » . 224 • W H H G J 4 M H 4 8 IB A iff *4 • 4^#J 0 W: • WLM : • 225 1 fe+S-ra* <NEOFFI> 1 2 3 4 5 2. f f c S t t f c ^ A ^ f f c r M . 3 . 0^ + . 4. &#w*«fc**i$4&-&auL«;A. 12. f 3 ^ : F * # f f c 6 B t t J f l t t f ^ fc**J*. 1 3 . A g ^ &$ #f «j « , flt» & ft a « i i j « & ft&. 16. 161'J 17. a ^ * ^ ^ ^ ' J A ^ ^ „ 19. a ^ ^ # A ^ # , # ^ A $ L # - „ 20. # J& ^ # — # # m w x # . 21. aa#i$i! i tf sMp^tt^jt. 22. a#itaJ!r^*'jafcw*friL+. 24.ffc-#-*JltS.I««.J8!lAW*l«. 2 7 . a - r ^ # m * ^ x # „ 29. a * - f t M A*'j m # , flt ^ A * P • 3 s . 226 2 1 2 3 4 5 (Hi* 3 1 . 33 . a * 4> a m m*km gi « • w ***t * ft. 3 6 . a « # # 0 # . A t H # a i f t # : & W ^ ? ' J ' l « . 37. a ^ — i B t t m , W * I « J A . 3 8 . a * f W * ^ a t « i r a « ' « t ^ i * , a ^ j f c a t ^ i M f rift^m. 3 9 . ^ - * A f t # a ^ « , # w r # . 4 0 . ^ a t i P T > c . ^ # i , j B i A ^ t e a ^ t e * * * * * * . 4 2 . a ^ * - m'& * & m m #. 4 4 . a t t f c & £ t t # W I » M P * « t f r t f * . 4 7 . a i & i k ^ # ' f t ^ ' & . 4 9 . a i t * * I t & JBK MI I ' J . 5 0 . a * — 1 B £ £ A J i f e # i x f H f c . * j - « j A . s i . a ^ ^ ^ ^ i ' j ^ ^ j , j t * 3 £ ^ A H f e » ^ f > t w r a « . 52 . a * — - f © + ^ ^ A „ 54. ^ ^ a ^ # « < : — A A , a « * « - t e 4 3 i a t . 55. a ^ m ^ M f c * ' * ^ # # # # ^ ^ „ 5 6. ^ B# a- t - * «, # m . 5 7 . a * ^ a t t a * ^ ^ ^ , ^ ^ j A ^ ^ # 0 5 9 . ^ » « 5 ," a - t - M f t » ^ J A I # I'J M M W . 6 o . a / L * ^ ^ i t * # M . Appendix VI. Japanese Questionnaire Packet 228 c © f f i f t l i f e * t e © " ? * . &7t-61*4$leftB©, « » © « 1 » l c l W * C ^ - 3 T ^ * - r . - t O f c t t f t f e - B - l i f t K O l B f l l ^ W ^ L / f c t ^ * # £ ? j m i W H T I 4 * y * 1 * A / . T I 4 f c y * t * A / . BtB<D;i<h£JguaiUT<fc£U. SSIC r s & f f & l © ftlW*JlnaiLT<*:3^ ( f c U l r > B g £ * t t - 3 T l * f c J » » o f c f c & , $.^7i:^^:(DI^P Bi^iE5SlcJSoai-rc«t^mgT-r. L,jfc#oT* SEISES < o T . -t©ft£c:ic:ujfcj&\ ffl*l/T^fc*, l t £ — H t t f o f c * * , f a r * * a . T i * fcfrB^WUT*T<£S^. T'I4, *«:fcOlB1lO+T»lcWfll*:. ^ ^ O K f e l 1*. (t>Lfcfcjfcj&**l«K: r g ^ n ^ j ©ft(DK1t;WRj'bfcl*Jl£' CI4, ftftfc^J£itlZl[t\^1t«aftIiaic-»ifil\l«IBT<0lE1t*Sl^tfiL'T< -tO»IWIifii^8i^*Tlcir2ALT</f*l\ T I « © » M t & < i : t ^ 0 ^ 5 l c J 8 i ; T ^ f c * » * l i ^ a i U T * T < ^ : S ^ . r®^ajLfc»IBj lc£©«fc3ic«i;Ti*fc#CHr*'t>0T*\ &jfc"5liKffltt o+©si a t Mcttiao t> © T » * c t i i ^ -3 T ^ * r. L * ^ u cni*w H * U * ^ . ± T © « B l c e « l l c S i L * J : 5 l c f t # * g < U T < / £ $ ^ . M S U#lC l i t t f f l l\fcU£1*A/ 0 version 4 229 RMQ 2 1. ^ f / j : 2. U^y ^ ^ U f c 3. 5 f t L U 4. I f L t 5. ^ £ 6. *;SSg£ 7. » £ J E U f c 8. tf<tf<UJt 9. H&MLfc 10. IS(**StJLfc 12. ffitittf 230 RMQ3 1 2 3 4 5 ±< 1. 22. 43 . 2. 23. 44. 3. 24. 45 . 4. 25. 46. 5. 26. 47. L-it 6. 27. 48 . 7. 28. S e t t l e tf-OO 49. 8. 29. 50. 9. 30. 51. 10. 31 . 52. 11. 53. 12. 33. I t fc54. SR38L.fc 13. 34. i k i / i x 55. 14. 35. 56. 15. 36. 57. 16. 37. 58. 17. 38. 59. 18. 39. 60. 19. 40. 61 . 20. 4 1 . 21 . 42. » t ) 'J * Lit 6» M(Df-x 7 £#Sil*TOftUJ&\ • t> 5 - f i i t S UT < /c$ t\ 231 RMQ4 1 2 3 4 5 i: t3 to i : 'fel^X.&l^ 1. ijMtTH^fc. 2. «»*asnTtf<tf<.Ljtfii;fc'ot. 3. HSUTUfco 4. * © » I B I C L T l ^ C t C | | l * * f B i ; T l ^ . 5. a ^ f c f t f t S r t f - s f c . 6. S * t S B » « C * - 6 T l ^ . 7. tf<tf<L,THfc. 8. <h*5H3fctt;&>, *M8*lcfcoTi*jfc. 9. frCtbtz-otto 10. chT-b^a^m -^to -^pfc. 11. ** l* : «GT^fc . 12. *0BS0£#©£SlCjfrfcUt&+UTltfc. 13. u s s y - e l ^ . 14. S f t t S & ^ l C ^ t l T l N f c . 15. ^mtd-ofco 16. Ob^oTtATt. 17. B*fc*LT^<fc5<!:SS&UTUfc. 18. ££«<fcUT»ft fcfc#£-3jfc . 19. 20. r^T*JiBBicffii;6nfc. 22. nwfcws/^jfc. 23. ^>*LS« ! : »1SC»^T^fc . 24. ^H^f-pfc. 25. &E£-pjfc. 232 RMQ 5 1 2 3 4 5 26. jfrSSLSnfcJRL^-aJfc. 27. fgSSUT^fc . 28. 1MB Gfc'^fc. 29. toim* A £ l * * « l & L , j 0 » o j t . 30. # M ? C £ 5 * i m ^ f a t > i f c E l ' T l * f c a > - 3 J f c . 31. * < D I $ t o W i # t t l c g 1 * # f c l \ £ : » - 3 f c . 32. m^LT, « B [ C t t y » A / W f c . 33. fBJj&MZ^lvcu&jE-aTlVfc. 34. y ^ y £ * L , T l ^ f c . 35. .-t<0«IHL $tmzW).&&bTVt£. 36. tt0*fcli8lttT, »»fJ&ifcJ&>-pfc. 37. ftwAttli^gfBfcoJfc. 38. A£li3nl3&tt£o 39. ;S*)lc&,SvrX> 40. « t t JH2 tUT^T I S *>-3 f c . 41. ? £ 2 i £ £ f t f c » - f e T U f c . 42. ^ f e f U T ^ t . 43. f B j ^ C - D ^ T f t « : ^ T t ^ . 44. & © * ^ T # 5 F » f c « G £ - 3 f c . 45. -ten*, A± l i -O0*S£BE0« f c5OB i ; e > n f c . 46. £ S * l v u v C £ & l , f c f t » 5 £ - 3 f c . 47. ^ t U T U t , 48. i t t O U T l ^ C ^ l c & ^ U W T t . 49. fAO^-7.l4i$D^< U <tUTL\T , a W f c f c f l r C & ^ i f c . 50. - to f t , A £ ( J - : 3 0 * # £ S H 0 « f c 3 l c « G & n j f c . 51. ffl«IC»-5fctt«£-3jfc. 52. 53. # # ^ » ± U T l ^ . « f c 5 l C « ' i ; f c . 233 RMQ 6 1 2 3 4 ± < W t t T f c l x 1. it©ia^iiaffiWT*r>fc. 2. J S ^ t S R a c i f f e X l x f c . 3 . fcifj&\ tf<tf< L T l x f c . 4 . -t<0H#, TZWtSiK. # t t l c a * * & - 3 T l x j f c . 5. i f 5 ix 5 f c t / x . T t x T 3 S 5 U * ^ f c . 6. ^ m i C & ^ t l T l X T t o 7. t ^ T L ^ p 7 t C < ! : , * ? T L , * t > f c C < f c l c o i \ T * * L < S i ; T l x f c . 8. & * L - T i x f c c 9 . -a fc. 10. e # £ B U I C J B - 3 - C l x f c . 1 1 . a ^ > * » T , T&JlWtefr-Dtt. 12. ^-(D<h^, * 0 ! ) E i y © « ! l ^ t t S - 6 a i X T l x y | § ^ l - X f c . 1 3 . {Ri^icDbsnyt^Dfc^fc. 14. ¥a — ;:ft#&©jsi;fct>©je-3fc. 15. ^ « l C « n T t x f c . 16. 5c*W«fc<, K f c f r o f c . 17. ± J K f l * J l C ^ £ U T t x f c o 1 8 . # S l C ^ > * t f U - v a > £ J § i ; T l x f c . 1 9 . r f p J T ' ^ ^ t x j <k lO<fc5fc* l#* : -3J fc . 2 0 . # f c y < D R a * s i ; T i x f c . 2 1 . fe'^SUA/Wfc. 2 2 . J S ^ l C ^ ^ ^ ^ n T t x f c . 2 3 . ^ ^ C i l C O t x T ^ + U T l x T , L T l x f c . 2 4 . vrntistzimtt-ott. 2 5 . ffl^TUfeo 234 R M Q 7 1 2 3 4 26. e ^ K ' - r ^ u i x t s D T t x f c . 27. ^ © r t ^ ^ X > v > l i ( 0 o < y ^ A - X l c S l l x T l X 7 t . 28. a^^T-H±UTlx^>^i;/£-p/i:. 29. MlMBGfc-afc. 30. &0*M£3!&L,< 0 5 W t . 31. B l f c ^ f c o 32. 4fc*(*?S*U&*fc<. SSE^-pfco 33. ^T l i i xe > n ^ : ^ o f c o 34. £ f t t t T 5 S 3 8 £ ? f c . 35. & © R » l i f r f c y » L , J & > o f c . 36. S5B"C7cft*<«k^-pfc. 37. B ^ T l x f c . 38. * # £ ^ ® £ : o J t . 39. ^ f c ' o t . 40. a<*j&*»jO>T, B iLUfcttB lCfc-pfc. 41. JMfcl i^l frfctt l i lC&^fc. 42. & S I U m^ l f igUTUJ t . 43. y ^ ^ ^ ^ L T t x f c o 44. ^mV®?<VtUfi:®Cfc?fZ. 45. SW*ftTI*&J&>oJfc. 46. t<Dffl£l. ttfcKCfcoWfc. 47. < h T f c £ £ £ £ < h L f c i £ i ; £ r > ; t . 48. & © i k < f c # t t l J t M / T l \ T , 8 K o W < 5 # $ i l c i f i ^ f c . RMQ 8 mxmm 14 m : £f HP • M M f§ : (m : B *A . +BBA«:<!f) : ££ftfcffl : (NEO-FFI) ^(7)SPp^lal i60Sero35:S*^/S^Tl^^•r. T K 0 £ » * B M / C . £ & * t l c £ t > J : < a T l i £ £ < h c 3 l c O E P Ic •* it E T-13. TJ * < T? fe 1. IMi/b-EttT-iifcU l 2 3 4 5 2. *&<DA£-mz[.\Z<Dt)WZfii l 2 3 4 5 3. S S l c ^ l t o T ^ I f f l ^ t t l c f « © l i » * T - « : 0 1 2 3 4 5 i 2 3 4 5 5. # 5 % * € r 5 A / « k U S t U M c U T l ^ * i 2 3 4 5 6. ^ S ^ C ! : * ( J : < ^ i 2 3 4 5 7. 1L\t?<lzmz> i 2 3 4 5 i 2 3 4 5 9. &m*>m»££<ntt&Tz> i 2 3 4 5 10. Rjra«!:*3y(C!Kl**^y|?^-5<ke>, a#<D^-;*(#i8)£^3©;!><<!:<Ofc' i 2 3 4 5 11. 7 h l / ^ 0 < t . r g > ^ * < * , * j lcfc-5«fce>lcJgC.5c<fc!t>*5 i 2 3 4 5 12. * 8 ( c f i l J 6 a > £ A I B J T I£fcU 1 2 3 4 5 13. s W f ^ a ^ e M < D * T ^ i t f c ^ / t * . i c » i * * r > ^ n * 1 2 3 4 5 14. UOCE£&ftm^T$ttOZ\E^frmtt^AMtt£mz>T^ZAtfi^Z> 1 2 3 4 5 15. A IRS (£*..»: 5«>A,) T-|ifcl^ 1 2 3 4 5 16. £Zf[s<Ur>1tV&.?-0\z&-oft:V-t2>Z£\*lbz>-/ZlZt£\,\ 1 2 3 4 5 17. A t g - r r o t f t r f c r n u ^ 1 2 3 4 5 18. ^ £ l c * f l , T r i . l^5l^57Ji:]RM.^X.73A<^)'5<!:^-^e»**»^•5©"C ,li^<. —3 1 2 3 4 5 19. A i 5 s y a b 5 « f c U i b t a ^ u * 5 * * < f f * ^ 1 2 3 4 5 1 2 3 4 5 21. «t<s5 igLfcy » ig ja i i i i c^:^fcy - r-5 1 2 3 4 5 22. iSSaO&atCSICOSCDtf f f**: 1 2 3 4 5 23. S#£§5^fc f i J* ) ) S C 7 a o 1 2 3 4 5 24. A © ^ * £ & f c ^ l £ < ^ < D B T - J l . A < * > £ 1 2 3 4 5 25. W $ 8 £ B S £ J # o T f c y . ^ t l l c i n j ^ T f i ^ t b ^ ^ y T f ' C l D i y i B ^ O ' S . 1 2 3 4 5 26. e » l i ^ < f f i { i l ^ » . ^ i e i ; - 5 C < ! : * < » ' < r » * 1 2 3 4 5 27. { e j A >f < & i S £ l i - A T ^-5t55*<»#£ 1 2 3 4 5 28. ^ lAvf t^ « ! ) £ i a U T*<5*#<fc<&-5 1 2 3 4 5 29. IS5?T;fcttt f , £ l \ T O ( T M I i & £ a i U f e C ? < ! : ^ 5 £ 3 e > 1 2 3 4 5 237 IC •* fc' 1t if T' t> fe X fe < X fe 30. J i £ A n T f ± ¥ £ & 4 D . 5 £ T i ; : , #rAififrfr% 1 2 3 4 5 31. &Zl,^tt)^&fiitfr&CZ>Ztitib->l£lzt3:\,\ 1 2 3 4 5 32. 7tSi*<«^nT» L^£LTl\e>;h.feO*;&*«fc<&.5 1 2 3 4 5 33. &mtfi&to%t.^*%ft*,W&toZ>*<DtLtfi. Vx\*%tttfi&t>?1tt\*mit3.\.\ 1 2 3 4 5 34. fMira<^£©Aj&>&ffa>*l-a^ 1 2 3 4 5 35. £#roB«££fi£-r-5«fc5lc#Ay|f<5 1 2 3 4 5 36. A © f ± n t l C « f c < K * f c T « 1 2 3 4 5 37. TcSfe. l i o e o t U f c A I H f c ' 1 2 3 4 5 38. i i ^ M & ^ J B r l i . #a>e>£3«*?lcScK'^*£ 1 2 3 4 5 39. &£>£fc<fr»ff iO<hJlTU<5A#0.5 1 2 3 4 5 40. ^-rS^*-C^UJi t t -5^.J iU*<fcoT^e . f±*€-? l*Slt « 1 2 3 4 5 41. *3*a*2>£<U;&>feU<>:. « b * > i J U , fc£Sttfc<fe-5Ci:#$S***-S 1 2 3 4 5 42. Sfc^S?T?l*fe^ 1 2 3 4 5 43. » * » ^ u a « f ^ f l * R T i ^ t , ^ < ^ < b ! t u « i i i r o s s y * ) e c ^ 1 2 3 4 5 44. &l;migttTt»T-|iSia>feU 1 2 3 4 5 45. AOWJftlC-tofcU. ftSJ^^fciJ bfeltntffe&fet^©!;: , t * * * 3 T l i & l i 1 2 3 4 5 46. e u < t t ? f c u . £ £ ^ f c * y r * C < b l i l f i : A ; i : f e O 1 2 3 4 5 47. iM2lV3<bf i?*UTl i « 1 2 3 4 5 48. ^ffi©*H^AfIA<fc*»nTO*ttffilcS^^to<ei-C<!:lcliia<t^<!:»i**<fel\ 1 2 3 4 5 49. &l*lV3fcf l fe©A*-J I l^ .5AIH"?&33 <!: 1 2 3 4 5 50. A g A U i t t * * ^ ! . *tl&*>VmfZ> 1 2 3 4 5 51. i ' 5 l / « t H 6 < T > ^©pDlS^iS*Mc»^UT !t)e.l\fc^<!:JSpC<t^«t<a&-5 1 2 3 4 5 52. WcSiSbftfeAIHrt: 1 2 3 4 5 53. ftftffSf/fctfSBO 1 2 3 4 5 54. JtOfeAICli , ^ a e t f T ^ s 1 2 3 4 5 55. A l i f c © * £ , t ^ t U t A l f f l t l J J I o T O f c O 1 2 3 4 5 56. fi#£^;t&Ay;tOi:.®e>(5i:. % r * » H \ C < t ^ t « l c « ) * 1 2 3 4 5 57. A©5feBSlcV2;-p«ty>b, tfb5lfe*<JB*?7<73^«fcO 1 2 3 4 5 58. tefcM£#;t^S8|£igUtJC 1 2 3 4 5 59. e^gfrfcro^icAtta/ttofee. A £ & -5c<!:!fe;t£>e>fc£o 1 2 3 4 5 60. ^ - S C i r l i - f ^ T I C f c O T . £ ( C C 5 ? U ) £K<#-3TA</ulf<3 1 2 3 4 5 Appendix VII. Korean Questionnaire Packet 239 (RMQ-5) ^s) 3711 § H Jt-SH-S-. °1 -3€-*lfe ^l?fls]fe HQs] ^ 4 7]^1 tfl^- oj^# 3 $ ^"a^i}. -fel H<& *1«2*1 7 W l 5tf<^i}. asflAi A J ^ .g*cfA) <y<Hi-Rr #3°ll ^ ^ u l ^ tftjl*. aj6j|Ai £*i ^ °lfe °VMfe*l t a ^ J I S H &<&m}. ^ -^«r7)l 2 ^ 7 1 -fgjfe *3-4H?fi ^ - 3 $ £ * b * 7l^«rS.^- JL *1. 0J>|S §Oph ^2|3H SA|ch 3*3. *IM &m *t zn cfl^ - 7l^oi giq.^ , tg-xjoi 3**MI 7l«i««ii ^ xlfe 4-g- 1^1 °1 1^ 47flsl ^^x|s-ci o j^qt^ . =L%^£ S.^ ZL ' 7 l ^ S ££*<>fl ±%t= #?|xl°l si^ q c f . - f a i t $ 7flfil <a-o]«-°l A|s a l*^4fe S t l^cKV t «• *flsH ^ c f i - f l ^ q t F . a^l cfls. aej_a sj-S-fr s.^ ^-g-ofl ^ Translated by Chang-kyu Ahn (Pusan National University) from The Remembered Monxtits Questionnaire of Michelle Yik in University of British Columbia , Department of Psychology, Vancouver, BC, Caniida V6T 1Z4 240 m — v-1. : : : : : : : : : : 2. : : : : : : : : : : *RSM5flfe 3. # 7 l £ : : : : : : : : - — : : ^7R! 4 X| 0] : : : : : : : : : : 4 ^ 5. **hii3* : : : : : : : : : -—: ^ i a ) ^ 6. : : : : : : : : : : 7. ••—-: : : : : : : : : - W l 8. « « Jpa*« r f e : : - — : — : : : : : — - : — : ^ 9. ^ a f t ! : : : : : : : : 3^3 <3 10. # Afe : : : : : : : : : <>Fr *WStfe 11. : : : : — - : : : : - — : : 12 zttJ^HSfe : : : : : — : : : : — : *WW 8fe - 2 -241 °1 ^ £fe 2=^, 7]&, :JW(#3H| ^ ^«Hr ^ l - s . ^s|cH fl^qr), '7]^^ £ ? V ^Al ^ A l o ] x^ g. ^ S-oJ^ V j - H f l f l J i A j A J i . ^ 3 7l^«l-7l ^ t^2) A]-§-«VAjA]iL. 1 2 3 4 5 3*1 <M=7> E M 1. 2 3. 4 a 6. 7. & 9. 10. 11. 12. i a 14 15. 16. 17. . 1& . 19. . 20. . 21. . _^7rAj _=ls^«rfe _*°1 _i^fe 22. 1-^  43 #7i-£-23. 44 7,}^e]^ 24 "S^^^l 45. 2-§-i& 25. ^ - a * ^ l f e 46. 26. Aiv^ 47, _ D j a A ^ 27. 48. 3i -Hr 28. °lfe 49. 29. 2=*lfc 50. # ife 30. 51 31. ^W>1 jfe 52. 3x13 32. ^^.^o] 53 #Q 33. <3#-fr 54. 34. #1$ 55. 35. 3*l$Kr 56. 36. *Hs^fe 57. 37. 91*1 ^ ^ « H r 58. W £ 38. 59. ^7] $fe 39. - f r ^ 60 40. 4M3<a 61. ?H<H 3l\ 41. A^71 Sife 42. Arff-fr 4 Sltt^ *7+# ^ftfe*] « g ^a^ A l i L w - 3 -242 1 2 3 4 5 . 1. «B*«tk 3. 4. Hfl7r a £#<>|) ^JI <yo|] ^.P|§ H^IJI siSicf. 5. * H M I * ^#4. 6. 7l£4 2 # A 3 . 7r=- ^ 4 . 7. 5J4. 8. St 7J-X) °l-fj-S, C f i A}^ol *b:£tf. 9. ^#4. io. °YT q&m. 11. 12. a. "3*1 14^ #cfl tj)sj| p]| o. ^ ^ ^ o ] ^ _ _ 13. W i S M ^ 4 . 14. 71*4 « e H ^t}. 15. !r*S$l4. 16. #4. 17. 34!*}-B]7l7r ^ Bjflr-I-18. *7) 9X3- #37fl ^£4. 19. ^ M S 4 . 20. Ji^cf. 21. °m <3#-g- i^°l$4. _ _ 22. S-g-SlTfl ^ ^ r . 23. 7]-s- 214. 24. *<ki3fltH 25. - 4 -243 26. *r-8-°l *SLS\JL sisq, 27. *P8-5S4. 28. 43 -g-°l *MW3:a 5tf-§-fr }=.m. 29. 43- a t «\$°] ^ ^1431 ^ 4 . 30. =L*% * H M-f- «*WH opf. ^  ^ ^ ^ i # J L ^ 4 . 31. S-Zrlio] "fl-f *|-¥-*Wm J i £ 4 32. "Vi- 3*1*1 Siui -iJ*rai ^8°1$4. 33. «J «J|-g-°l] ^ 4 . 34. i * * i^ol$4. 35. 4fe ZL nfl-f s\-y\ #4. 36. 43 *M-°1 4#skn. 7]*mzi\ #$4. 37. 43 %°] ^m) ^ £ 4 . 38. <y^ <H 4Jtf4. 39. 7l£*}ui ^ ^ 0]si4. 40. #°1 *fca #314. 41. $^°]5L *7] # 8 4 42. 4^ "}-§-dl 2°J^ §- ^ 5*4. 43. •H'S ^ ^7H7f l H ^ 4 44. i^ja. #84. 45. =m xfl ^ . g . 4^0) ^ -a-Mf^^ ^aa4. 46. ^ ShSrTfl *4)-g- ^ J O . »i4fe ^ o]oj4. 47. i&£& m& 4m ^°1$4 . 4S. 4)7} «roi £fe «H 40«fl-*i "S^aW}. 49. 43 -^fr 5lJL 3.SL$&. 50. ^ 1 ^43 ^ JMW 3MS4. 51. 31# ^Ne| xiSJ4. 52. a -f-A^ -g-i- ^#4. 53. juLm ±*\m. °W &°}7}*\ 537113 «F4£ ^ a ° l -§-3l8Sr*| s t t!SKW£. - 5 -244 1 2 3 4 3*1 4 i 1. 43 ^)*W4. 2. 7l^3f Tj^^g. 7ysi m 3. ^ 7W ol-fj-S «HJW $4. 4 ^lel^ji ^s}3°l 3 $ 4 5. S? o] o . ^ ^ o ] tfjj. ^ajflq.. 6. ^A] iffe. »7l7 r ^ £ 4 8. 4fe -g^^o^. i-^q, _ 9. 43 7)f-o] Sta-r).. 10. 4 7»HQofl tflsfl T ^ i e ^ c ] , 11. *r&*Ka 7]^o] 7]-e} -^^lg(rf. 12. ^ A] vfe <H«3 <a°U 0141 ^ $ 4 . 13. i-Hr ©IS «2«T*tS* $534. _ 14. *t-&#-n3W 4 s tfl7l- ^#"3 3H4. 15. 4^- y&m. 16. #31 #34. 18. «l-f <3#-§- i£#4. 19. 4fe a-¥-s)^ ojgjq 20. #e] 2#$14. 21. S | S « t | . . 22. 7 ] ^ - ^ 7Hf ^ 4 . 23. ^-Al o)v£ msj-o} ^ c^ojq 24. %&m. 25. 26. AJAjo] if-jo^m. 27. 43 ^91 <&&°] aej j i ^ ^ 7 ) ] #<47}-JL at&T}, _ 28. JLSL^V ^o]ojc}. - 6 -245 29. ^9X3. 7^ 7)1 i ^ 4 . 30. 4 3 «HM -fr^sKa. *&<&m] 31. 4fe ^m. 32. 5LZ ±v\m. 33. 4*°1 ^-8-3$!}. 34. ^71 $13. ^ S - ^ A S . H ^ ^ 4 . 35. ^ ^#4. 36. 7]-g-oi i ^ 4 . _ 37. 3r7r #4. 38. l-ETfl 3 £ 4 . 39. *-*J95l4. 40. 4 3 -§-°l a-S-sKa. JL8_$- #3184. 41. 4 3 -g-S-42. 4-§-°l ^£-3ai *3-%5a4. 43. 7j^ -o] #^^^4. _ 44 ^#*rJi ^e)7fl ± ^ £ 4 . 45. * ^ r « 4 46. o. "4l3*M0lsa4. 47. 4 + #7) SiTfl ^ 4 . _ 48. 4 3 713 #*Kr $3 S|-g-.°i2. ^°>7H 487113 «*M 4H£ '^f l<>l -§-#«fe*| *€44X|.SL. • -a^ w: • #• 4°V-• 7)ft *)•&• «H£ °H: _ ^ c i 4 , 3 ^ 4 7 p£ &°1 Aj-g-^ - tW:__j«foL <#> -^Ajoj oj^ aj B H ^ ^ : ^ <#> _tH?_ • ^ T ° W A> 4014 4m«l #4^-44. - 7 -246 N E O a S i f <^4 .m«f2 JZ^«r2r 051-510-2637) o] ^m-ir °]mfa. a #n ^ ¥<*\ &3.s= i - ^ - ^^-t °i«fl*fc*n 3 ^ 3 ) ^ - * # 3 3 . cj§ S 8 ) 2 , M¥ 2 0 | *fzm*l Q | A J H g& J=9|fe c«i£ s a s f l ^ A I S gAmsftMct. 4- vfl-g-t- m± if-, € - # 3 3-Aii). o]± ^ s . «sj^«H^^ll- °>2fl3 *°1 ^ l 0 ! 1 -< M 7| > 1. 444I- #°>W- < »Hf «Hf .*K14 op^ <y*i«* & ® ra ® ^ • * R ! 3 i i * i ^ n i ® ® m • @ • ^ ^ T l ^ W , «3*>7l <Ha1-S- •! ® ® • ® ® • * R 1 3 ^ l * } ^ & # ffl ® ® •^Hl^oH5- «J*1*M • ® m ® @ j & ® ffl ® @ 1. '-Rr S°l t^r ^ o]cf. & ® ffl © ® 2. «^e) AV -^i-i). 3 4 o^s} ^xfl^ ^ c] * 4 . & ® ffl ® ® 3. Al^M- ^3*Rr ^-1: #°>*M &^4-® ® ffl © © 4. xfl7> ^4-b JE.^  A)-^ 1-6(17)1 3**1 t^ji*} i i ^ ^ 4 . <g> ® m © ® 5. *ns*wi ^ H I « H § 4 . <s> ® m ® ® 6. 4€- -4##<>ii cfl*H i W i - ^ 4 . & ® ffl © ® 7. ^ -^fe-4. ® ® m ® ® 8. #*H3. 3*B*I 4 3 ^ a p M s^4. ® ® ffl ® ® 9. 7>^o]4 *Fr *4 . & ® ffl © ® lO. «)|oJ: ^ Ojo] a}. Al^ > iflofl ^ 4 . ® ® ffl © ® 11. i E ^ i f 2L^°V ^ > S . 7^)1 ^-7l^«H^)4. & ® ffl © ® 12. 4fe n-s ° W 4 . ® ® El © ® 13. ^ 3 2 /£°14 *|-#<H] **l7> «fl7> #4. & ® ffl © ® 14. o)^ 41- ^ >7l ^ ^ o } 4 ol7H?l A ^ o l ^ j i AJ2|^ 4. ® ® ffl © ® 15. <a-§- ^- «4| *t*l-3 !8-°W *I3*R=- ^c,14. ® ® IB © ® 16. 31-3-714 *W ^fc3-. ® ® m © ® 17. A V ^ # 3 olo>7lt|-b #^4. ® ® m © ® is. sa^  n ^ 4 r s^-oii *#TD ® ® m © ® 19. 4^- ^ ^ ^ o l 7 l a 4 ^ ^2:^°! ^^14. ® ® ffl © ® 2 0 . 4<MI 3)^ -^1- 7 H 3 i ^- i^^^-4. ® ® El © ® 21. #%o] oflnl^ l-jL, ^  2J#3$ ^ 02) ® B] © © 22. #71 XIJl 48*3-01 oJ- ^ o)| 7>7l§ #o>^4. & ® ffl © ® 23. A H - ^-tftj-cieji £ X ) 5 - ^ 4 . ® ® ffl © ® 24. 4-b 4€- 4^-1- ^ ^ ^ - 5 . iJL ^ ^ ^ . ^ *W' ^ ^ ^ l ^ " . -1 -247 ® ® m © © 2 5 . - g - ^ -=-£-1- . A H - j i tf^JlS. ^3811 #4. ® ® m ® © 2 6 . 7YB- lfl 7+A)ol 7>^7l- ^*)*]4. ® m ® 2 7 . i f l * 3 . «)-fe ? H #t}. ® El ® © 2 8 . -g-Ajoiq- <q -^ £-A|-g. ^  i L s j j i ^ 4 . ® ® m ® © 2 9 . A > ^ # ^ 7 R r $ 4<JM- ©l-g-^ fkfcn. A jz^ r } . ® m ® © 3o . •a-i: MA^W*] ^ T H - ^ « - ^ i * . ® ® m ® © 3 1 . -g^ig ^ ^ | * > 7 1 4 8fc4-® ® m ® © 3 2 . 4 f e # 7 l 7 V SU4. ® ® m ® © 3 3 . ^ 3*1 2 = ^ "fl7> &4. ® ® m ® © 34 . vfl7> ol-b t f l ^ - 3 # o ^ 4 -® ® m ® © 3 5 . ^S. ^-ir ^ <&<$*\ i i 3 ^ 4 . ® ® m ® © 3 6 . 4^r A>^#ol m - rfl«j-b BflS. «fl£-<*H ^7} ^ ^7} $ 4 . ® ® m ® © 3 7 . 4^-°] w #3J! 71M4-® ® m ® © 3 8 . J E ^ A]rfls}. Aj.3g.0Ji q . ^ ^ 4 3 7]&o] ^e l-a ^ 5!i4. ® ® m ® © 39 . <H£ A f ^ ^ 4 * A> -^J1 TflAj-^Cj A}^OS. A JZ J ;^ . ® ® m ® ® 4 0 . 4 ^ * ^ - ^ - l - € ^ « r ^ 3 1 *J-A> oj^ofl ^ 4 . ® ® m ® © 4 1 . ?Jol ^ 5 . ^ nf l^ -S-71S- ^Tfl S)51, i 7 l « « WlBlJl ^7}. & l } . ® ® m ® © 4 2 . 4fe 3|)*-«ru! ' 8 * * A V ^ - O I o}i)4. ® ® m ® © 4 3 . A i q a)]^ ^xj-^. „}| ,97]] ^ g . ^ . ^ ^ 4 . ® ® m ® © 4 4 . 4fe -&^^ °14 3«« 7^11 4^-°l ^ ^ l f e Al-^ol o} i !4 . ® ® m ® © 4 5 . n°\ ^ <£-$r ^-g-aj i^ja. -HI*} 5 .^ nJl7> §4 . ® ® ra ® © 4 6 . ^ 4 « r 7 l 4 3 7 1 ^ ^ 1 ^ 1 S f e 4 . ® ® m ® © 4 7 . 4c- ^ J f A l ^ ^ a l H>ffi7H Alvjtr. ® ® m ® © 4 8 . ©1^ > $78l4 -*Nf^ AjZl;«1-fe IgS. 3h*l7l- ^ 4 -® ® m ® © 4 9 . A}^l-ofl7ll A>3 © H A J siTfl tfl«l-^;a i c . ^ ^ 4 -® ® m ® © 5 0 . *»o> § P J ^ . ^nfloj, ^. ^E l^r fe ^olc]-. oft <£*\d> ® ® ® ® © 5 1 . 4 3 «-*n* ^ ^ 4-E- ^7> §4 . ® ® m ® © 5 2 . 4 b "11-f # ^ ^ ° J A>^o|t>. ® ® m ® © 5 3 . A13°1 J i 7 l A j c ] 7 ^ 4 . ® ® m ® © 5 4 . <H£ AV^=-o] Ajo.^^ ^r^-j! ® ® m ® © 5 5 . i ^ a l j ? _ 5 . ^ a l ^8^8- ^ ^ * ^ 4 . ® ® m ® © 5 6 . 7H^ 3 - 4*M wlBloi 5U4. ® ® m ® © 5 7 . A V ^ # 3 ^1£A1-7> S l 7 l J i 4 f e ifl VmS- * A V *>fe #4. ® ® m ® © 5 8 . <H«3 ^Aj-si ^ e i q oi^ -o)] ^ • o > i 7 l 4 ^«l-7l #o>^.c]-. ® ® m ® © 5 9 . -gSBr ^-t- ?i7l ^ * | - 4 ^ , A ^ # - i - 7 l^o ] ol-g-^-® ® m ® © 6 0 . p j A M -ff^*r31 ^ o ! 4 7 l ^afl ^ ^ 4 -- 2 -Appendix VIII. The Revised CMQ Scales The Spanish sample PLEASANT "Adjectives" Format content happy pleased satisfied "Agree-Disagree" Format I was happy. I was satisfied. I felt content. "Describes Me" Format My mood was positive. Overall, I was satisfied. ACTIVATED "Adjectives" Format activated aroused hyperactivated "Agree-Disagree" Format UNPLEASANT dissatisfied miserable troubled unhappy I was miserable. I was unhappy. I was dissatisfied. My mood was NOT good. I was feeling troubled. I felt unhappy. DEACTIVATED sleepy still I was filled with energy and My body felt still, tension. My body felt activated. "Describes Me" Format I was keyed up. I was stirred up. I was feeling placid, low in energy. My internal engine was running slowly and smoothly. My mind and body were resting, near sleep. 249 The Chinese sample PLEASANT "Adjectives" Format content happy pleased satisfied "Agree-Disagree" Format I was happy. I was satisfied. I felt content. "Describes Me" Format My mood was positive. Overall, I was satisfied. ACTIVATED "Adjectives" Format activated aroused UNPLEASANT dissatisfied miserable troubled unhappy I was miserable. I was unhappy. I was dissatisfied. My mood was NOT good. I was feeling troubled. I felt unhappy. DEACTIVATED quiet still "Agree-Disagree" Format I was filled with energy and tension. My body felt activated. I was feeling quiet. My body felt still. "Describes Me" Format I was full of energy and tension. I was feeling placid, low in energy. My body was in a quiet, still state. My mind and body were resting, near sleep. 250 The Japanese sample PLEASANT "Adjectives" Format content satisfied UNPLEASANT dissatisfied unhappy "Agree-Disagree" Format I was happy. I was satisfied. "Describes Me" Format Overall, I was satisfied. I was miserable. I was unhappy. I was dissatisfied. My mood was NOT good. I felt unhappy. ACTIVATED DEACTIVATED "Adjectives" Format aroused quiet hyperactivated sleepy "Agree-Disagree" Format I was filled with energy and I was feeling quiet, tension. My body felt still. "Describes Me" Format I was keyed up. I was feeling placid, low in energy. My body was in a quiet, still state. My mind and body were resting, near sleep. 251 The Korean sample PLEASANT "Adjectives" Format content happy pleased satisfied "Agree-Disagree" Format I was happy. I was satisfied. I felt content. "Describes Me" Format My mood was positive. Overall, I was satisfied. UNPLEASANT dissatisfied miserable unhappy I was miserable. I was dissatisfied. My mood was NOT good. I felt unhappy. ACTIVATED "Adjectives" Format DEACTIVATED aroused hyperactivated quiet sleepy "Agree-Disagree" Format I was feeling quiet. I was feeling stirred up. My body felt activated. "Describes Me" Format I was full of energy and tension. I was keyed up. I was feeling placid, low in energy. My mind and body were resting, near sleep. 

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.831.1-0089251/manifest

Comment

Related Items