FGDFGDFG

An Analysis of the Relationships Between Various Models and Measures of Emotional Intelligence.
Benjamin Robert Palmer
Centre for Neuropsychology School of Biophysical Sciences and Electrical Engineering Swinburne University of Technology December 2003
Table of Contents
Table of Contents
Declaration Acknowledgements Publications arising from the thesis List of Abbreviations Abstract List of Figures List of Tables vii viii ix x xi xv xvi
CHAPTER 1: Prologue 1.1 1.2 1.3 Introduction Aims Arrangement of cha pters
1 1 3 5
CHAPTER 2: Emotional intelligence: Conceptions, measures and research 2.1 2.2 2.3 2.4 2.5 2.6 Overview Introduction Conceptions of EI Measures of EI Research on EI Directions for future research
6 6 6 8 12 14 17
CHAPTER 3: The Mayer-Salovey-Caruso Emotional Intelligence Test 3.1 Overview 3.1.2 Introduction 3.1.3 Conceptual, developmental and correlational criteria for ability measures of EI 3.1.4 Scoring methods 3.1.5 Reliability 3.1.6 Factor structure 3.1.7 The Mayer-Salovey-Caruso Emotional Intelligence Test 3.1.8 Reliability and factor structure of the MSCEIT 3.1.9 Summary 3.1.10 Objectives of the present study
20 20 20 21 23 25 26 27 29 31 31 ii
Table of Contents
3.2
Method 3.2.1 Participants 3.2.2 Materials
33 33 33 33 34 36 37 37 38 40 41 42 43 45 48 53
3.2.2.1 The Mayer-Salovey-Caruso Emotional Intelligence Test 3.2.2.2 MSCEIT V2 3.2.2.3 MSCEIT scoring 3.2.3 Procedure 3.3 Results 3.3.1 Descriptive statistics 3.3.2 Reliability 3.3.3 EI and age 3.3.4 EI and gender 3.3.5 MSCEIT inter-correlations 3.3.6 Confirmatory factor analysis 3.4 Discussion 3.4.1 Conclusion
CHAPTER 4: The Bar-On Emotional Quotient Inventory (EQ-i). 4.1 Overview 4.1.2 Bar-Ons (1997a) model of emotional intelligence 4.1.3 The Bar-On Emotional Quotient Inventory 4.1.4 Factor structure of the Bar-On EQ-i 4.1.5 Overview of the current study 4.2 Method 4.2.2 Participants 4.2.3 Materials 4.2.4 Procedure 4.3 Results 4.3.2 Exploratory factor analysis 4.3.3 Confirmatory factor analysis 4.3.4 Demographic differences 4.5 Discussion
55 55 55 58 61 65 66 66 67 68 68 69 73 76 78
iii
Table of Contents
CHAPTER 5: The Trait Meta Mood Scale (TMMS) 5.1 Overview 5.1.2 The Trait Meta-Mood Scale 5.1.3 Overview of the current study 5.2 Method 5.2.2 Participants 5.2.3 Materials 5.2.4 Procedure 5.3 Results 5.3.2 Exploratory factor analysis 5.3.3 Confirmatory factor analysis 5.4 Discussion
83 83 83 85 86 86 87 87 87 88 91 93
CHAPTER 6: The emotional intelligence measure by Schutte et al. (SEI; 1998) 96 6.1 Overview 6.1.2 Introduction 6.1.3 Overview of the present study 6.2 Method 6.2.2 Participants 6.2.3 Materials 6.2.4 Procedure 6.3 Results 6.3.2 Confirmatory factor analysis 6.3.3 Descriptive statistics 6.3.4 Gender differences 6.3.5 EI and demographic variables 6.4 Discussion 96 96 100 102 102 103 103 104 104 107 108 109 110
CHAPTER 7: Assessing components of emotional intelligence with the TAS -20 116 7.1 Overview 7.1.2 Introduction 7.1.3 Overview of the current study 7.2 Method 116 116 119 120
iv
Table of Contents
7.2.2 Participants 7.2.3 Materials 7.2.4 Procedure 7.3 Results 7.3.2 Descriptive statistics 7.3.3 Confirmatory factor analysis 7.3.4 Relationships with demographic variables 7.4 Discussion
120 121 122 122 122 124 127 130
CHAPTER 8: A taxonomic model of emotional intelligence 8.1 Overview 8.1.2 Introduction 8.1.3 A taxonomy for EI 8.1.4 Summary 8.1.5 Overview of the present study 8.2 Method 8.2.2 Participants 8.2.3 Materials 8.2.4 Procedure 8.3 Results 8.3.2 Scoring 8.3.3 Descriptive statistics 8.3.4 Inter-correlations 8.3.5 Confirmatory factor analysis 8.3.6 Exploratory factor analysis 8.3.7 Results summary 8.4 Discussion 8.4.2 Conclusion
134 134 135 136 148 150 152 152 153 153 154 154 154 156 159 163 169 170 175
CHAPTER 9: Conclusions, Limitations, and Directions for Future Research 179 9.1 9.2 Overview Conclusions 9.2.2 Reliability 179 179 179
Table of Contents
9.2.3 Factor structure 9.2.4 Dimensional communality 9.3 9.4 Limitations Directions for future research
180 183 185 189
References
193
vi
Declaration
Declaration
I declare that this thesis contains no material which has been accepted for the award of any other degree or diploma and to the best of my knowledge contains no material previously published or written by another person except where due reference is made in the text of the thesis.
________________________________
Benjamin Robert Palmer. December 1st 2003.
vii
Acknowledgements
Acknowledgements First and foremost I would like to acknowledge the supervision of Professor Con Stough. His advice and assistance in the conceptualisation; data collection and interpretation; and write- up of the thesis has been invaluable and contributed greatly to the completion of the thesis. I would like to thank Professor Stough for his encouragement; for being a constant source of motivation and inspiration throughout; and for providing a comfortable and relaxed working environment. Professor Stough has also been a trusting and sincere friend. I extend to him my sincere thanks and gratitude for all that he has done. Secondly, I would like to acknowledge the contribution made by all those who participated in the project. The battery of emotional intelligence measures typically took more than five hours of your valuable time and without this contribution the results and findings of the thesis would not have been possible. I would also like to acknowledge the contribution of my fellow students and the staff at the Centre for Neuropsychology. You have been a constant source of inspiration and encouragement, and provided a fun, stimulating and productive work environment. I would also like to acknowledge the inspiration I have gained from Professor Peter Salovey and John Mayer, their seminal article on the construct of emotional intelligence was the impetus for my interest and devotion to the area. Finally, I would like to thank my loving parents and friends for your continued support, encouragement and faith throughout the completion of the thesis, without which it would not have been possible.
The author would like to acknowledge the following individuals for their contribution to the project:
Dr. Ramesh Manocha for his assistance in collecting data for the project helping to make the sample of participants as representative as possible.
Mr. Gilles Gignac and Dr Timothy Bates for their valuable advice and assistance concerning the structural equation models reported in the thesis.
Professor Peter Salo vey and Professor John Mayer for their advice and assistance concerning the results pertaining the MSCEIT in Chapter 3. viii
Publications
Publications arising from the thesis

Palmer, B.R., Manocha, R., Gignac, G., & Stough, C. (2003). Examining the factor structure of the Bar-On Emotional Quotient Inventory with an Australian general population sample. Personality and Individual Differences, 35, 11911210.
Palmer, B.R., Gignac, G., Bates, T. & Stough, C. (In Press). Examining the structure of the Trait Meta Mood Scale. Australian Journal of Psychology.
ix
List of Abbreviations
List of Abbreviations
MSCEIT Bar-On EQ-i TMMS SEI TAS-20 SEM CFA EFA EI
Mayer-Salovey-Caruso Emotional Intelligence Test Bar-On Emotional Quotient Inventory Trait Meta Mood Scale Self- Report Emotional Intelligence (Schutte et al., 1998) Twenty-Item Toronto Alexithymia Scale Structural Equation Modelling Confirmatory Factor Analysis Exploratory Factor Analysis Emotional Intelligence
Abstract
Abstract
Since Salovey and Mayers (1990) seminal article on emotional intelligence (EI), the construct has received wide spread interest and attention. A number of different models and measures of EI have been developed and a growing body of research in the area is emerging. However, relatively few independent studies have examined the psychometric properties of existing measures and many of the claims made by test authors concerning test reliability, factor structure and validity are yet to be substantiated. Furthermore, little is known about the relationships between existing measures and questions concerning how best to conceptualise and measure EI are at present unanswered. In the current study a population sample (n = 330) completed a battery of EI measures that were representative of the different approaches to the conceptualisation and measurement of EI including the Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT; Mayer, Salovey, Caruso & Sitarenios, 2003), the Bar-On Emotional Quotient Inventory (Bar-On EQ-i Bar-On, 1997), the Trait Meta Mood Scale (TMMS Salovey et al., 1995), the EI scale developed by Schutte et al., (1998) and the Twenty-Item Toronto Alexithymia Scale (TAS-20; Bagby, Parker & Taylor, 1994). The reliability and factor structure of each of these measures were examined in turn. The results show that the internal reliability of these measures of EI have improved markedly over earlier tests, specifically, those examined by Davies, Stankov and Roberts (1998). All the measures examined in the current series of studies yielded internal reliability coefficients above a = .80 at the full-scale level, and between a = .80 and a = .90 at the subscale level. Both exploratory and confirmatory factor analytic methodologies were employed in the series of studies reported in the thesis in order to
xi
Abstract
assess the factor structure of the measures examined. With the exception of the Bar-On EQ-i, (Bar-On, 1997) the findings of these factor analyses were consistent with the underlying theories of the various measures examined and their respective previous research findings. The factor analytic results showed evidence for the four abilities purported to be measured by the MSCEIT (Mayer et al., 2003), the three components of the TMMS (Salovey et al., 1995) and the TAS-20 (Bagby et al., 1994), and the four factors of the scale by Schutte et al., (1998) as previously found by Petrides and Furnham (2000). The dimensional structure of the EQ- i was not found to be consistent with the underlying theory it has been purported to measure or the factor analytic results of Bar-On (1997a). It was concluded that further factor analytic research findings are needed in order to clarify the dimensional structure of the EQ- i. In general, the factor structure of the different EI measures examined were found to be robust and consistent with their respective underlying theories. Following examination of the internal reliability and factor structure of each of the five tests in turn, a final factor analytic study was conducted in order to examine the relationships and dimensional communality amongst them. From a systematic review of the variables (by definition) assessed by these different measures a five-factor taxonomic model for EI was hypothesised involving; (1) the capacity to perceive and express ones own emotions; (2) the capacity to perceive and understand the emotions of others; (3) the capacity to utilise or reason with emotions in thought; (4) the capacity to effectively manage ones own emotions; and (5) the capacity to effectively manage the emotions of others. Confirmatory factor analysis via structural equation modelling was then used to assess whether this model (and a range of alternatives) provided a statistically significant fit with the data including; (a) a three-factor model representing the original model of EI conceptualised by Salovey and Mayer (1990); (b) a five-factor xii
Abstract
method variance test model where each factor represented one of the tests in the battery; (c) a two factor model representing self-report and ability measured EI. None of the models assessed were found to provide an acceptable fit with the data according to standard model fit statistics, however, the hypothesised five-factor taxonomic model identified in the review was found to be the best fitting model in comparison to the three-factor model and both the five-factor and two- factor method variance models. An exploratory factor analysis was conducted and interpreted as providing further partial support to the hypothesised model. This finding suggests that there is some common variance shared between the various models and measures of EI, and that the hypothesised five- factor model may better represent the different approaches to EI (as a definition of the construct), than the theoretical distinctions that have been made between them (e.g., trait and ability EI, Petrides & Furnham, 2001). However, the correlation between self-report and ability measured EI (r = .39 as found with the two- factor method variance model), suggests that these two approaches to the measurement of EI share only 15% common variance. It is argued that the hypothesised model of EI may not have provided a statistically significant fit with the present data as variables assessing the ability to express emotions were not adequately represented (i.e., too few items assess this common facet of the construct); and variables assessing the ability to utilise or reason with emotions in thought are not adequately reliable. Further limitations of the current study and recommendatio ns for future research are discussed. Given that the five dimensions of the hypothesised model can be systematically identified from a cross section of EI tests, and that partial support for this model was found by the current study, it was concluded that the hypothesised model is representative of the dimensional communality amongst models and measures of EI and therefore provides a taxonomic
xiii
Abstract
model of EI. It is concluded that this model of EI should be assessed by future research along with other theoretic ally justified taxonomies for EI, and the relatively goodness of fit should similarly be examined in order to substantiate whether the taxonomic model identified by the current study best describes the communality amongst different models and measures of the construct. While not distracting from more specific models and measures of EI, it is argued that the taxonomic model of EI established by the current study is useful in that it provides a common definition and understanding about the nature of the cons truct as well as a model upon which to base comprehensive measures of the construct. Based on the findings of the current study it is argued that EI can be commonly defined as a conceptually related set of abilities to do with ones own and others emotions, specifically; the ability to perceive and express ones own emotions; the ability to perceive and understand the emotions of others; the ability to allow emotions to direct ones reasoning; the ability to manage ones own emotions; and the ability to manage the emotions of others.
xiv
List of Figures
List of Figures
Figure 4.1 Figure 4.2
Bar-Ons (1997a) second order five- factor model of EI Exploratory six- factor model of the Bar-On EQ-I determined by the current study Scree plot of eigen values derived from the parallel and principal component analysis Structural equation model depicting the mediating effect of Clarity on the relationship between Attention and Repair Oblique Four-Factor Model of the SEI (Schutte et al., 1998) Oblique Three-Factor Model of the TAS-20 (Bagby et al., 1994b) Scree plot of eigenvalues derived from the parallel and principal component analysis
74 75
Figure 5.1
88
Figure 5.2
92
Figure 6.1 Figure 7.1 Figure 8.1
106 126 164
xv
List of Tables
List of Tables
Table 3.1
Means, Standard Deviations and Reliabilities for the MSCEIT V2 according to American Consensus Scores, Expert Scores and Australian Consensus Scores Descriptive Statistics for MSCEIT Branch and Overall EI Scores by Gender Intercorrelations Among Australian Consensus (Below the Diagonal) and Expert-Based (Above the Diagonal) MSCEIT Subscale Scores MSCEIT Parameter Estimates and Goodness-of-fit Statistics for the One-Two-, and Four-Factor Models The 15 Components of the Bar-On EQ-I Means and Standard Deviations for the Bar-On EQ-i Scales Factor loadings pertaining to the Bar-On EQ-i (direct-oblimin) Component Loadings Derived From a PCA of the Five Higher Order Facets Proposed by Bar-On (1997a) Differences in emotional intelligence as a function of Age, Sex, Mental Illness Factor (pattern matrix) loadings of principal components analysis with oblimin rotation for the TMMS Goodness of fit indices for the five different models of the TMMS assessed Model fit statistics for the SEI (Schutte et al., 1999) Means, Standard Deviations and Reliablities of the SEI (Schutte et al., 1998) Means and Standard Deviations on the SEI by Gender Descriptive Statistics for the TAS-20 Model fit statistics for the TAS-20 (Bagby et al., 1994) Means and Standard Deviations on the TAS-20 According to Gender, Years of Education and History of Mental Illness Measures of Emotional Intelligence The hypothesised five- factor general taxonomy for EI
39
Table 3.2
42
Table 3.3
43
Table 3.4
46
Table 4.1 Table 4.2 Table 4.3 Table 4.4
56 68 70 76
Table 4.5
77
Table 5.1
90
Table 5.2
91
Table 6.1 Table 6.2
105 107
Table 6.3 Table 7.1 Table 7.2 Table 7.3 Table 8.1 Table 8.2
108 123 125 128 138 141 xvi
List of Tables
Table 8.3 Table 8.4 Table 8.5 Table 8.6 Table 8.7 Table 8.8
Alternative Factor-Models Assessed via SEM Descriptive Statistics for the various measures of EI Intercorrelations amongst the total scores of the various measures Intercorrelations amongst the subscales of the various measures Fit statistics for the hypothesised models assessed Hypothesised Five-Factor Higher-Order Model Parameter Estimates (in parentheses) of the Observed Tasks on the Latent Variables, and Estimated Intercorrelations Pattern of Factor Loadings for the Five-Factor Oblique Rotated Exploratory Solution and Factor Intercorrelations
151 155 156 157 160 162
Table 8.9
166
xvii
Chapter 1; Prologue
CHAPTER 1 Prologue
There can be no knowledge without emotion. We may be aware of a truth, yet until we have felt its force, it is not ours. To the cognition of the brain must be added the experience of the soul. Arnold Bennett (1932) 1.1 Introduction
Very rarely do psychological constructs receive as widespread attention as the recently conceptualised construct of emotional intelligence (EI). EI has appeared on the cover of Time magazine (Gibbs, 1995), is the topic of the most widely read social science book in the world (Goleman, 1995) and many other popular books, magazine and newspaper articles (Mayer, Salovey & Caruso, 2000a). Much of this interest stems from the face validity of the construct. When EI was systematically conceptualised by Salovey and Mayer (1990), it was proposed as a construct that may not only broaden what has traditionally been considered intelligent, but as something that might be learned or taught, and as something that might account for individual differences in many important human values such as psychological well-being, life satisfaction, the quality of interpersonal relationships and occupational success. Indeed perhaps in acknowledgement of such, the American Dialect Society chose EI as the most useful new words or phrases of 1995 (Mayer, Salovey & Caruso, 2000b). From a construct perspective, there may be more to the attraction to EI than its mere face value (Mayer, et al., 2000b). Although the term emotional intelligence may be viewed somewhat as an oxymoron to historical thinkers who have considered emotions as disruptive to mental activity (Salovey & Mayer, 1990), EI has also been described as old-age wisdom neatly and precisely defined (Goleman, 1995). Seemingly,
Chapter 1; Prologue
some of the components of the EI construct are intuitive. If one asks the layperson whether they know people who are more and less adept at expressing how they feel, understanding the emotions of others, or managing themselves and their peers in times of distress, more often than not the answer is unambiguously yes. Indeed, the notion that our emotions contribute adaptively to our reasoning is reflected not only in the literature of emotion researchers (e.g., LeDoux, 1998), but also in common proverbs such as think before you speak, which refers to reasoning with emotions, or weighingup the emotional consequences of what one is about to say before saying it in order to arrive at a more congenial or purposeful outcome. Despite these intuitive elements, EI is as seemingly complex as its constituent terms. Indeed no two concepts in psychology have received as much attention and yet managed to resist consensual clarification than both emotion (Ekman, & Davidson, 1994) and intelligence (Sternberg & Kaufman, 1998). Over the last decade of research on EI this complexity has become most evident in attempts to measure the construct. For example, it has been argued that there are no right and wrong ways to feel in anyone given situation (Mayer & Salovey, 1997), therefore determining truly right and wrong answers for tests of EI, (one of the current standards for tests of intelligence), is proving difficult (Zeidner, Mathews & Roberts, 2001). Nonetheless, although EI is a construct that is still in its infancy, significant advances have been made toward the establishment of reliable and valid measures of EI in both objective and self- report formats (Mayer, Caruso & Salovey, 2000c). Establishing reliable and valid measures is vital to both extracting the potential value of EI and the sustainability of the construct itself (Salovey, Bedell, Detweiler & Mayer, 2000). Reliable and valid measures of EI will not only aid us in understanding the nature and cause of individual differences in EI, but will form the basis of targeted
Chapter 1; Prologue
training and remedial programs aimed at improving related areas such as life satisfaction, the quality of interpersonal relationships and occupational success. The history of psychology provides many examples of promising constructs that have failed to contribute to the discipline due to a lack of devisable measures. Indeed, it has been proposed that EI may represent a narrower attempt to assess aspects of social intelligence, which in the past has proven difficult to distinctly measure from intelligence (IQ; Mayer & Salovey, 1993).
1.2
Aims:
This dis sertation focuses on the measurement of EI. Since its formal beginnings in the early 1990s (Salovey & Mayer, 1990) several different models and measures of EI have been developed providing alternative theoretical frameworks for
conceptualising and measuring the construct (Mayer et al., 2000a). This work has resulted in models that have conceptualised EI either as a set of mental abilities and corresponding measures that attempt to assess it as such (e.g., Mayer, Caurso & Salovey, 1999); or models that comprise emotion related abilities together with personality traits and dispositions, and corresponding self-report measures purported to assess emotionally competent behaviours (e.g., Bar-On, 1997a). Preliminary, albeit tangible evidence for the reliability and validity of these approaches is mounting, however, further psychometric analyses are needed (Salovey, et al., 2000). Specifically, aspects of reliability and validity need to be independently replicated and further assessed across disparate population samples. As such, one aim of the dissertation was to examine the psychometric properties of five of the different measures of EI currently available.
Chapter 1; Prologue
A second aim of the dissertation was to examine the communality amongst various models and measures of EI and to establish whether a general taxonomy for the construct could be determined. While the development of various models and measures of EI has provided alternative approaches to the conceptualisation and measurement of the construct, it has also caused some confusion about the nature and boundaries of EI. Variables ranging from emotion related mental abilities to emotional competencies and personalty traits have been placed under the banner of EI, and it is not surprising that recent reviews of the area have branded the construct as popular but elusive with fuzzy boundaries (i.e., Pfeiffer, 2001). Such issues are not uncommon, yet typically problematic in psychology often making it difficult to properly ascertain essential elements of construct validity (Roodenburg, 2003). While leading authors in the area have sought to clarify this ambiguity by theoretically contrasting various approaches to the construct and providing definitive categories on that basis (Mayer et al., 2000a; Petrides & Furnham, 2000), little attention has been paid to the potential communality amongst them. Despite this however, some authors have noted that the various models and measures of EI tend to be complimentary rather than contradictory (Ciarrochi, Chan & Caputi, 2000), whilst others have proposed that there may be a general taxonomy for the construct (Goleman, 2001a). A taxonomy for EI would be useful in that it would provide a common definition for the construct and help define its more definitive dimensions much like the comprehensive taxonomy of personality traits, the widely known Five-Factor Model (FFM; Digman, 1990; Costa & McCrae, 1992a). Moreover, the identification of a comprehensive taxonomy for EI would provide a model upon which researchers could build comprehensive measures of the construct. Despite the development of numerous models and measures of EI it has been argued that comprehensive measures of EI that
Chapter 1; Prologue
cover the various operationalisations of the construct do not currently exist (Petrides & Furnham (2001). To ward this goal, the present dissertation examines the communality amongst five measures of EI including the Mayer-Salovey-Carsuo Emotional Intelligence Test (MSCEIT, Mayer, Salovey & Caruso, 2000d); the Bar-On Emotional Quotient Inventory (EQ-i; Bar-On, 1997a); and other related measures including: the Trait-Meta Mood Scale (TMMS; Salovey, Mayer, Goldman, Turvey & Palfai, 1995); the Twenty-Item Toronto Alexithymia Scale (TAS-20; Bagby, Parker & Taylor, 1994a,b); and the EI scale by Schutte et al., (SEI; 1998). The goal of this research was to determine a taxonomy for EI that would provide a consensual definition and basis for comprehensive measures of the construct.
1.3
Arrangement of chapters
Chapter 2 presents a brief and general review of the area, outlining the original conception of EI, the development of alternative models and measures, and corresponding research findings to date concluding with the potential value of establishing a general taxonomy for EI. This review is brief and general in content as each of the five different measures latter assessed are reviewed in more detail respectively. Chapters 3, 4, 5, 6, and 7 review each of the five models and measures of EI previously mentioned, and examine the reliability, factor structure and aspects of validity respectively. Chapter 8 examines the relationships between these various models and measures, and examines the communality amongst them via factor analyses. A taxonomic model for EI is identified and discussed. Chapter 9 presents a summary of the current research, discussing its limitations and directions for further study.
Chapter 2; Emotional Intelligence
CHAPTER 2 Emotional intelligence: Conceptions, measures and research
2.1
Overview
Chapter two presents a brief review of the EI literature, discussing the original conceptualisation of EI by Salovey and Mayer (1990), and alternative models and measures of EI that have followed. Research findings that have been found with the various approaches to the measurement of EI are also presented and the potential value of establishing a taxonomic model for EI is also discussed. As previously mentioned, this revie w is relatively brief and general in nature as each of the five different measures administered and analysed in the dissertation are reviewed in more detail in the subsequent chapters.
2.2
Introduction
Emotions are not simply something that we feel; they are a source of information. For example, feeling tired conveys information about our energy levels or fatigue; feeling cold provides us with information about the temperature of our environment; mutual feelings of warmth and trust convey informatio n about our level of friendship or affinity with another (Frijda, 1988). As such, we tend to reason with emotions and the information they provide (Mayer, Caruso & Salovey, 1999). For example, we reason about the emotional tones in voices and music, the emotional expression in people's faces and body language, and in literature and artwork. We often reason about the emotional consequences of our actions (like whether or not to say something that might hurt someone) and about our own emotional states (such as when
we attempt to keep frustration at bay when persisting with a challenging task). We reason with our emotions in these ways because using emotional information helps us to understand our position and relationships with the world, and to respond adaptively (Mayer, et al., 1999). This notion underlies a traditional view that considers emotions as motivating forces that adaptively focus cognitive activities and subsequent action (e.g., Leeper, 1948), a view that was the impetus for the development of the construct of EI (Salovey & Mayer, 1990). In their seminal article, Salovey and Mayer (1990), proposed that emotions and emotional reasoning may be a part of, and may contribute to, logical thought and intelligence in general. Furthermore, Salovey and Mayer (1990) asserted that the capacity to reason with emotions and emotional information may be considered a mental ability in which individuals may differ much like other mental abilities. In a framework that brought together a diverse literature on the appraisal and communication of emotions, Salovey and Mayer (1990) conceptualised EI as a set of mental abilities to do with emotions and the processing of emotional information. These included the capacity to identify and express emotions; the capacity to effectively regulate and manage emotions; and the capacity to utilise or reason with emotions in thought (Salovey & Mayer, 1900). Within this framework Salovey and Mayer (1990) proposed that individuals differed in these abilities, and that these differences were potentially important because; (a) emotional abilities might account for variance in important life criteria such as psychological well-being, life satisfaction and the quality of interpersonal relationships; and (b) because such differences underpinned skills that could possibly be learned or taught. It was these latter notions that caught the attention of Daniel Goleman (1995) who wrote a popular book Emotional Intelligence: why it can matter more than IQ.
Golemans (1995) book elaborated Salovey and Mayers (1990) conceptualisation of EI in layperson terms placing particular emphasis on the links between EI and important life criteria. This book quickly became a best seller and developed a considerable amount of interest in EI from the general pub lic. It is now the most widely read social science book in the world (Gardner, 1999). With this influence, the popularity of EI has spread leading to a plethora of other books and magazine and newspaper articles relating EI to success in various facets of life. In particular, the role of EI in successful leadership and work place relations (Cooper & Sawaf, 1997; Weisinger, 1997), parenting (Gottman, 1997; Shaprio, 1997), and self- improvement (Segal, 1997; Simmons & Simmons, 1997; Steiner & Perry, 1997). Popular literature on EI has contributed to the area in highlighting the potential contribution of the construct. However, it has also confounded the psychometric status of EI by discussing its face validity as fact rather than supposition (Mayer, et al., 2000a). As a result, many
practitioners and laypersons have been potentially mislead to believe that a great deal has been established and is known about the construct of EI, like for example, that EI is a strong predictor of success in life (Goleman, 1998) when in fact, much remains unknown about EI and much of the face value of the construct remains to be substantiated by research.
2.3
Conceptions of EI
Since Salovey and Mayers (1990) conceptualisation of EI, the construct has also received considerable attention in the scientific literature. A number of alternative models have been developed providing several theoretical frameworks for conceptualising and measuring the construct (e.g., Bar-On, 1997a; Cooper & Sawaf,
1997; Goleman, 1995; 1998; 2001b; Mayer & Salovey, 1997). Mayer and Salovey (1993; 1997) have evolved their original model of EI as a theory of intelligence, that is, as a conceptually related set of mental abilities to do with emotions and the processing of emotional information. Their latest ability model comprises a four-by-four matrix of emotional abilities placed within; a hierarchical structure; a developmental perspective; and the overall psychometric structure of intelligence g (Mayer & Salovey, 1997). In contrast, latter theorists (e.g., Bar-On, 1997a; Cooper & Sawaf, 1997; Goleman, 1995; 1998; 2001b) have conceptualised much broader models of EI that comprise emotional abilities mixed together with a variety of personality traits and dispositions such as impulsiveness, trustworthiness and optimism (Mayer, et al., 2000a). The development of alternative models of EI has resulted in some debate as to what should and should not be part of the construct (Mayer et al., 2000b). Mayer et al., (2000b) have noted that broad models of EI tend to mix together abilities with related (and sometimes conflicting) personality traits from the three major sub-systems of personality including motivation-type traits, emotion-related traits, and those associated with somewhat ill defined areas of behaviour (Mayer, et al.2000b). For example, Golemans (1998) model of EI includes traits such as Achievement Drive, Empathy, and Service Orientation. As such Mayer et al. (2000b), have argued that such conceptions do not refer explicitly to the terms 'emotion' or 'intelligence', and are ambiguous conceptions of the construct. Indeed latter models of EI (e.g., Bar-On, 1997a; Cooper & Sawaf, 1997; Goleman, 1995; 1998) are not as theoretically well clarified as Mayer and Saloveys (1997) ability model of EI, often describing well established personality traits as abilities (e.g., Optimism is the ability to look at the brighter side of life, Bar-On, 1997a, p.27) and failing to take into account the
distinction between different, sometimes conflicting parts of personality (e.g., achievement drive and conscientiousness, Goleman, 1998; as noted by Mayer et al., 2000b). The development of alternative models has also caused some confusion concerning the nature of EI. Broader models of EI (e.g., Bar-On, 1997a; Cooper & Sawaf, 1997; Goleman, 1995; 1998; 2001b) encompass many personality variables as a part of the construct rather than potential correlates as originally proposed by Salovey and Mayer (1990). As a result some reviewers, as previously mentioned, have concluded that EI appears to be a popular but elusive construct with fuzzy boundaries (Pfeiffer, 2001). Towards clarifying this ambiguity a number of leading authors in the area have theoretically contrasted the different models of EI and placed them into coherent categories. For example, Mayer, et al. (2000a), have recently separated the different models of EI into two general categories, namely, ability and mixed models. Mayer et al. (2000a) have classified ability models as those that described EI as an intelligence in the traditional sense (i.e., Mayer & Salovey 1997). In contrast, Mayer et al. (2000a) have classified mixed models as those that described EI as a compound conception of emotion-related abilities drawn from personality traits and dispositions (e.g., Bar-On, 1997a; Goleman, 1995; 1998; 2001b). Others (i.e., Petrides & Furnham, 2000), have provided similar definitive categories that incorporate the different conceptions and measurement approaches to EI currently developed. Petrides and Fur nham (2000) for example, differentiate between; trait EI, identified as models that draw heavily on personality variables (e.g., Bar-On, 1997a; Goleman, 2000b); and information-processing EI, identified as models that have attempted to incorporate EI into the overall psychometric structure of intelligence (i.e., Mayer, et al., 1999).
10
This work has been useful in organising the growing body of literature on EI. However, the definitive categories identified are based on theoretical distinctions and whether different models and measures of EI represent two distinct approaches (or constructs) as suggested by Petrides and Furnham (2001), has not yet been empirically examined. Furthermore, it could be argued that a mixture of the terms used to denote the categories identified by Mayer et al. (2000a) and by Petrides and Furnham (2000) may better represent the alternative categories. The term mixed models is somewhat ambiguous and it could be argued this category is better defined as trait EI, given the consensual distinctions in psychology between abilities and traits. Similarly, it could be argued that the term information-processing EI is also somewhat ambiguous and that the term ability EI (Mayer et al (2000a) is more congruent with the existing theory that defines this domain (i.e., Mayer & Salovey, 1993; 1997; Salovey & Mayer, 1990) . Despite this theoretical clarification the term EI is still used to denote various operationalisations of the construct. Moreover, at present there is no consensua l definition of EI, and the nature and boundaries of the construct remain unclear. While leading authors in the area have sought to contrast and categorise the various models of EI no one has yet tried to systematically identify the relationships and communality that may exist amongst them. Goleman (2001b) has recently speculated that the predominant models and measures of EI share some common elements, specifically, abilities or competencies concerned with the capacity to recognise and regulate emotions in oneself and others. Furthermore, Goleman (2001b) has suggested that a parsimonious definition of EI would involve four higher order factors including: (1) the capacity to recognise emotions in the self (Self- Awareness); (2) the capacity to regulate emotio ns in the self (Self-Management); (3) the capacity to recognise emotions in others (Social Awareness); and (4) the capacity to regulate emotions in others (Relationship
11
Management). There is much value in identifying common or core elements of EI. The identification of such would provide a common definition and a taxonomic model for EI much like the comprehensive taxonomy of personality traits, the five-factor model (FFM; Digman, 1990; McCrae & John, 1992). Moreover, a taxonomy for EI would provide a model upon which researchers could build measures that assess the definitive dimensions of the construct much like the NEO PI-R (Costa & McCrae, 1992).
2.4
Measures of EI
The development of different models of EI has been followed by the development of several different measures designed to assess the construct (Bar-On, 1997a; Boyatzis, Goleman & Rhee, 2000; Mayer et al., 1999; Salovey, Mayer, Goldman, Turvey, & Palfai, 1995; Schutte, Malouff, Hall, Haggerty, Cooper, Golden & Dornheim, 1998). To-date self-report and associated informant-rated measures (e.g., Bar-On, 1997a; Boyatzis et al., 2000) and objective or performance-based measures of EI for which there are more and less correct answers to items based on expert opinion and consensual responses (e.g., Mayer, et al., 1999) have been developed. Self- report measures are purported to index cross situation consistencies in behaviour (Bar-On, 2000; Boyatzis et al., 2000). These authors argue that the degree to which individuals display emotionally competent beha viours is a manifestation of their actual EI. In contrast, performance-based measures like other measures of mental ability are purported to index individuals actual emotional abilities. How best to assess EI has not yet been empirically determined, however, it has been argued that the most direct assessment is expected to be gained from performance-based measures (Mayer, et al., 2000c).
12
Mayer et al. (2000c) have argued that EI assessed via self-report measures is filtered through the self-concept and as such provides an indication of ones beliefs about EI, or perceived EI, rather than their actual capacity. As such, Mayer et al. (2000c), have argued that performance measures will offer the most direct assessment of EI and are likely to exhibit better predictive validity than self- report measures as has been shown in the area of ability assessment (Paulhus, Lysy & Yik, 1998). However, others have noted that for performance measures of EI to offer an index of actual ability they must involve veridical scoring criteria, which will prove difficult to devise in this area (Petrides & Furnham, 2001; Zeidner, et al., 2001). As noted by Mayer and Salovey (1997) there is no black and white or right and wrong ways to feel in any one give situation, rather there may only be more and less correct ways of feeling or emotionally responding. Although the most recent performance-based measure of EI, the MSCEIT (Mayer et al., 2000d), involves both consensual and expert-based scoring criteria that exhibit similar psychometric properties, Zeidner et al. (2001) have argued that this approach does not meet the veridical scoring criteria that is standard for cognitive ability tests (e.g., items of graduated difficulty that only 10% of the population, the highly emotionally intelligent, can answer correctly). As such, extant performance measures of EI may not exhibit the predictive qualities of standard ability tests and there may be an additive or alternative utility to self- report measures of EI. Mayer et al. (2000c) have noted that self-report measures may assess more internal experiences directly related to emotional thinking, which may be difficult to obtain with performance measures. Furthermore, self-report scales in comparison to objective measures tend to be relatively brief and easy to administer. Therefore, valid self- report measures based on well-conceptualised models of EI may be of added value. In this
13
sense, these two approaches to the measurement of EI may be complementary rather than contradictory (Ciarrochi et al. 2000).
2.5
Research on EI
A growing body of research is providing preliminary, albeit tangible evidence for the validity of EI measures. Over a series of studies, Mayer and colleagues (Mayer, DiPaolo & Salovey, 1990; Mayer & Geher, 1996; Mayer et al., 1999), have developed a performance-based approach to the assessment of EI culminating in the Mayer, Salovey, Caruso Emotional Intelligence Test (MSCEIT; Mayer et al., 2000d). There are few independent psychometric analyses of the MSCEIT as it is the latest version of their performance-based measures. However, there is some information on the reliability and validity of the MSCEIT published in the technical manual (Mayer et al., 2000d). In addition, given the overlap in design with the previous measure (i.e., the Multifactor Emotional Intelligence Scale, MEIS; Mayer, et al., 1999), the MSCEIT it is likely to exhibit similar if not better psychometric properties. Pervious research with the MEIS has found that it is generally reliable and that it correlates with several theoretically related criteria (Ciarrochi et al., 2000; Mayer et al., 1999). As evidence that the MEIS is measuring cognitive ability (EI), Mayer et al. (1999) have shown that the abilities measured form a positively interrelated set, correlate with other intelligence constructs and show age related differences with adults scoring significantly higher than adolescents. Mayer et al. (1999) have also found that women tend to score higher than men and that EI is related to measures of empathy, life satisfaction and parental warmth. Similar results have been reported in independent analyses. Ciarrochi et al. (2000) reported similar gender difference and found that scores
14
on the MEIS were similarly related with empathy and life satisfaction. Furthermore, Ciarrochi et al. (2000), found that EI (as measured by the MEIS) was related yet distinct from normal personality. Recent research findings reported by Salovey, Mayer, Caruso and Lopes (In Press) suggest that scores on the MEIS are not associated with social desirability or mood and that scores predict end-of- year grades amongst colleague students over and above measures of cognitive ability and personality. Salovey et al., (in press) also report that higher EI may be associated with more pro-social behaviours amongst students, and related to workplace success within customer service environments. Finally, independent criterion group studies have shown that higher EI is associated with less alcohol and tobacco use amongst adolescents (Trinidad & Johnson, 2002). These preliminary results provide promising evidence for the utility of performance-based measures of EI, however, more independent psychometric evaluations of this approach are needed, and much of that established needs to be replicated with the MSCEIT (Salovey et al., in press) given that these research findings have been established with the MEIS. Research examining self- report measures of EI has shown that they are similarly reliable and related to criteria such as psychological well-being, empathy, life satisfaction, alexithymia, leadership success and adaptive interpersonal functioning (e.g., Bar-On, 1997; Palmer, Walls, Burgess & Stough, 2001; Parker, Taylor & Bagby, 2001; Salovey, et al., 1995; Salovey, Stroud, Woolery & Epel, 2002; Schutte et al., 1998; Schutte, et al., 2001). Moreover research studies have found that self-reported EI is related to psychophysiological measures of adaptive coping (Salovey et al., 2002), and can predict goal orientation (Martinez-Pons, 1997), mood recovery (Salove y et al., 1995), life satisfaction (Palmer, Donaldson, & Stough, 2002), affect intensity and depression (Dawda & Hart, 2000).
15
Despite these findings, at present there is some doubt surrounding the validity of existing self-report measures of EI (Ciarrochi, et al., 2000; Davies, Stankov & Roberts, 1998; Petrides & Furnham, 2000). While some measures have been shown to correlate with theoretically related variables (i.e., Bar-On, 1997; Schutte et al., 1998), it has been suggested that these relationships may exist because they are essentially tapping personality traits known to predict these variables (Newsome, Day & Catano, 2000; Petrides & Furnham 2000; Mayer, et al., 2000c). Indeed many self-report measures of EI include variables that can be mapped onto the big five dimensions of personality as recently illustrated by McCrae (2000), and research has found considerable overlap between self-reported EI and normal personality (e.g., the Bar-On EQ-i; Dawada & Hart, 2000; Newsome et al., 2000). Although the content of some self-report measures of EI overlaps considerably with personality, at face value the content of others such as the Trait Meta Mood Scale (TMMS; Salovey et al., 1995), appear to show less content overlap with personality, and research has fo und that scores on such measures can predict theoretically related criteria (e.g., life satisfaction), over and above personality traits (Palmer et al. 2002). Furthermore, although there are high correlations between some self-report measures of EI and normal personality, research has also suggested that these measures may still be measuring distinct constructs. For example, when the Bar-On EQ-i was examined concurrently with the Eysenck Persoality Profiler (Eysenck, Barrett, Wilson & Jackson, 1992) and the NEO PI-R (Costa & McCrae, 1992b) in factor analyses by Petrides and Furnham (2001), the subscales of the EQ- i were found to load on a separate trait EI factor leading the authors to concluded that self-reported EI may be a distinct composite trait construct located at the primary level of hierarchical trait structures. However, this notion has not yet been substantiated by research. Furthermore, there is a general lack of
16
research examining whether self-report measures of EI can predict theoretically related life criteria over and above normal personality.
2.6
Directions for future research
In comparison to well- established measures of personality and intelligence (IQ) relatively little is known about the measurement of EI (Salovey et al., 2000). Test authors have examined some of the more rudimentary psychometric properties of their tests and results have provided preliminary evidence of reliability, factor structure and aspects of concurrent and discriminant validity. Although independent studies are emerging (e.g., Ciarrochi et al., 200; Davies et al., 1998; Palmer, Manocha, Gignac & Stough, 2003; Petrides & Furnham, 2000; 2001), there is often a general lack of independent replications (Bar-On, 2000). In addition, very few studies have examined the predictive validity of EI measures over and above existing psychological constructs such as personality and IQ. As noted recently by Salovey et al (in press), much more research of this type is also needed. In their recent review of the area, Mayer et al (2000a) asserted that any prediction over existing constructs in the 1%-5% range would be worthwhile. Research examining the relationships between various models and measures of EI would also be of value to the area. As previously discussed, models of EI have been placed into two coherent categories ability models that pertain to individuals emotional intellect and corresponding measures that index actual emotional abilities (ie., Mayer, Caruso & Salovey, 1999); and trait models of EI, that pertain to emotionally competent behaviour (e.g., empathy), and corresponding measures that are purported to index the degree to which such behaviours are displayed (e.g., Never, Seldom, Sometimes,
17
Usually, Always), (e.g., Bar-On, 1997; Boyatzis et al., 2000). Petrides and Furnham (2001) have recently noted that these two aspects of EI may not be mutually exclusive and may therefore co-exist. Others have noted that these two approaches to the conceptualisation and measurement of EI appear to be complementary rather than contradictory (Ciarrochi et al., 2000), yet very little data exists on the relationships between them. Research examining the relationships between various models and measures of EI will provide insight into whether self- reported EI provides an estimate of individuals actual abilities and to what extent. Furthermore, research examining a battery of measures that cover the various operationalizations of the construct may help contribute towards a consensual definition and a comprehensive taxonomy for EI. The establishment of a comprehensive taxonomy may further provide the basis for comprehensive measures of EI, which some have argued not currently exist (Petrides & Furnham, 2001). The present dissertation aims to contribute to the area by providing independent psychometric analyses of five of the more utilised (by research) measures of EI as outlined in chapter 1. Specifically, the dissertation examines the reliability and factor structure of the MSCEIT, EQ-i, TMMS, TAS-20 and SEI with an Australian general population sample. Also examined are the relationships between scores on these tests and demographic variables such as age, gender, relationship status, level of education, income, and history of mental illness. The present dissertation also aims to contribute to the area by examining the relationships between these various models and measures of EI and establishing whether a taxonomy for the construct can be identified via factor analysis. In the next chapter Mayer and Saloveys work in the area is discussed in more detail and the psychometric properties of the MSCEIT (Mayer et al., 2000d) is
18
examined. The subsequent chapters similarly review and examine the psychometric properties of the EQ-i, the TMMS, the SEI, and the TAS-20 respectively.
19
Chapter 3; The MSCEIT
CHAPTER 3
The Mayer-Salovey-Caruso Emotional Intelligence Test
3.1
Overview
This chapter presents an analysis of the scoring methods, reliability and factor structure of the most recent ability test of EI, the Mayer, Salovey, Caruso Emotional Intelligence Test (MSCEIT; Mayer et al., 2000d), with a sample (n = 431) drawn from the general population. Consistent with previous findings, there was a high level of convergence between the alternative scoring methods (i.e., expert and consensus determined scores correlated r = .97). The full scale reliability of the MSCEIT was high (total score split- half r = .91), with split- half reliability coefficients ranging from r = .78 to r = .91 at the area level (Experiencing and Strategic EI respectively), and r = .73 to r = .90 at the branch level. However, reliability at the subscale level was found to be moderate to poor with six of the eight subscales producing coefficient alphas below a = .70. A confirmatory factor analysis indicated that the dimensional structure of the MSCEIT is consistent with the underlying theoretical model of EI it has been designed to measure. The findings of the current study inform the debate about whether correct answers exist for ability measures of EI. In addition, the findings suggest that the MSCEIT has improved upon the earlier performance measure of EI (the MEIS), in terms of the validity of scoring methods, reliability and factor structure.
3.1.2
Introduction
As discussed in Chapter 2, Mayer and Salovey (1997) have conceptualised emotional intelligence (EI) as a set of mental abilities concerned with emotions and the
20
processing of emotional information. As such it has been argued that the most valid assessment of EI will be gained from ability-based scales that involve (like other tests of cognitive ability), items for which there are more and less correct answers, that assess the capacity to reason with and about emotions (Mayer, et al., 2000c). Over a series of studies Mayer and colleagues have designed and examined the reliability and validity of a number of ability-based measures of EI (Mayer, et al., 1999; Mayer, et al., 1990; Mayer & Geher, 1996). This work has culminated in their most recent ability-based test of EI, the Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT, Mayer, et al., 2000d). As previously mentioned, independent psychometric evaluations of the MSCEIT are few in number as it has only been available for a short period of time. However, there are conceptual, developmental and correlational criteria inherent within the theoretical framework of the ability model from which it can be evaluated. Furthermore, there are research findings recently reported by the authors (Mayer, Salovey, Caruso & Sitarenios, 2003), and research findings with previous measures against which it can be compared (Ciarrochi, et al., 2000; Mayer, et al., 1999; Roberts, Zeidner & Matthews, 2001).
3.1.3
Conceptual, developmental and correlational criteria for ability measures of EI
Mayer and Saloveys (1997) ability model of EI comprises four conceptually related abilities arranged hierarchically from the most basic to the more psychologically complex, including; (1) the ability to perceive emotions; (2) the ability to utilise emotion to facilitate reasoning; (3) the capacity to understand the meaning of emotions and the information they convey; and (4) the ability to effectively regulate and manage
21
emotion. Within this hierarchical organisation, the abilities are proposed to develop sequentially implying that they are a function of age and cognitive maturation (Mayer & Salovey, 1997). Consistent with this theoretical framework, measures of EI such as the MSCEIT are expected to; (a) show a positive manifold of correlatio ns amongst the subscales designed to assess these four major areas; (b) show a consistent factor structure that comprises a general factor of EI and four correlated primary factors; and (c) show age related differences that reflect the developmental perspective of the model (Mayer, et al., 1999). As evidence that EI is an intelligence, in addition to the above, ability measures are expected to positively correlate with established measures of mental ability (as mental abilities typically do), such as those that index verbal intelligence (Mayer et al., 1999). Finally, EI has been theoretically related to several important life criteria that ability measures have been hypothesised to predict (Mayer et al., 2000a; Salovey & Mayer, 1990). These include variables such as psychological well-being, life satisfaction, empathy, the quality of interpersonal relationships, success in occupations that involve considerable reasoning with emotions (e.g., leadership, sales and psychotherapy), and scholastic and academic success. Mayer and colleagues (Mayer et al., 1990; Mayer & Geher, 1996; Mayer et al., 1999; 2003) and others (e.g., Ciarrochi et al. 2000; Roberts et al. 2001), have assessed the validity of ability measures of EI according to these conceptual, developmental and correlational criteria. Research with the predecessor measure to the MSCEIT, the Multi-Factor Emotional Intelligence Scale (MEIS; Mayer et al., 1999), has provided preliminary evidence that EI meets some of the underlying conceptual, developmental and correlational criteria of Mayer and Saloveys (1997) ability model (Ciarrochi, et al., 2000; Mayer et al., 1999; Roberts, et al., 2001). This research has shown for example, that the four abilities measured form a positively interrelated set, and tha t they correlate
22
with other measures of established ability (verbal IQ, Mayer et al., 1999; and general intelligence, (Roberts et al., 2001). The research by Mayer et al., (1999) also demonstrated age related differences with an adult criterion group scoring significantly higher on the MEIS than did an adolescent criterion group. Research with the MEIS has also demonstrated that scores on the test are meaningfully correlated with theoretically related criteria such as life satisfaction, empathy, and parental warmth (Ciarrochi et al., 2000; Mayer et al., 1999). Importantly, the study by Ciarrochi et al. (2000), demonstrated that scores on the MEIS were related to criterion measures (e.g., life satisfaction) even after controlling for IQ and personality. Altho ugh these studies (Ciarrochi et al., 2000; Mayer et al., 1999; Roberts et al., 2001) provided promising evidence for the validity of the MEIS, the findings also revealed some psychometric problems with the test. These problems concerned the scoring methods, reliability and factor structure of the MEIS (Roberts et al., 2001).
3.1.4
Scoring methods
While there may be no right and wrong ways to feel in any one given situation, Mayer, et al. (2000c), have argued that more and less correct answers can be determined for EI test items. This notion is based on the assumption that there are certain universals with emotions and emotionally signalled information (e.g., universal facial expressions of emotions as evidenced by the work of Ekman, 1973; 1999). Moreover, it is based on previous research findings that have shown that the pooled responses of large normative samples can accurately determine more and less correct answers to test items (Legree, 1995; Mayer & Geher, 1996). Mayer et al. (1999) have compared the extent to which three scoring methods for the MEIS converge, arguing that convergence would provide some evidence that the answers to test items can accurately be judged to be correct and 23
incorrect. The three scoring methods for the MEIS included consensual, expert and target scoring methods, although the target scoring method has been subsequently dropped from further analyses on both empirical and theoretical grounds (see Mayer, Caruso & Salovey, 2000b). In the consensual scoring method more and less correct answers to items are assigned according to normative averages. For example, if there are five possible choices to an item on the MEIS and fifty percent of the normative sample endorse a particular response to that item (e.g., response A), then individual participants who choose that particular response receive a score of .50 for that item and so on. In this scoring method high scores on the MEIS are achieved by consistently choosing responses to items that have been endorsed by the majority of the normative sample. In the expert scoring method correct answers to items of the test are set by experts in emotion (in the case of the MEIS the experts were Mayer and Caruso) and participants receive scores for each item of the test according to the extent to which their choices correspond to those of the expert group. Finally, in the target scoring method, correct answers to test items are set by the person (or target) who wrote the item. In the case of the MEIS some items in the test were established by asking people to describe certain situations and their corresponding feelings about the situation. In this scoring method participants receive scores for correctly identifying the feelings of the target in the described situation. These three scoring methods (consensus, expert and target) have been shown to converge with correlations ranging from r = -0.16 to r = 0.95, with over half of the correlations between the subscales of the MEIS exceeding a correlation of r = .52 (Mayer et al., 1999). This level of convergence has lead Mayer et al. (1999), to conclude that there appears to be more and less correct answers to the items of the MEIS
24
according to all of the scoring methods. However, others have questioned whether the magnitude of these correlations suggests that there are reliable common elements to the different forms of scoring (Roberts et al. 2001). In their own analyses of the MEIS Roberts et al. (2001) examined the correspondence of factor scores derived from expert and consensus based scoring methods and found that they correlated at r = 0.26. Roberts et al. (2001) also found that the consensus and expert scoring methods differed in their relationships with sex and ethnic groups leading these authors to conclude that the two scoring methods (consensus and expert) did not appear to provide alternative indices of the same underlying construct. The finding however, of a relatively low level of convergence between expert and consensus scoring methods has lead Mayer and colleagues to the conclusion that there does appear to be more and less correct answers to EI test items (Mayer, Salovey, Caruso & Sitarenios, 2001). They have also concluded that the level of convergence (or lack thereof according to Roberts et al., 2001), was partly due to a poorly defined expert criterion and that there was merit in developing a more well defined expert group (e.g., one that was more representative of experts in the field of emotions, Mayer et al., 2001).
3.1.5
Reliability
While research with the MEIS has demonstrated that it is internally reliable at the full-scale level (e.g., Mayer et al., 1999 report split- half reliability of r = .96), the reliability of some of the subscales (particularly those measuring the higher order abilities of Understanding and Managing Emotions) have been found to be low across a number of studies. For example, Ciarrochi et al. (2000) reported coefficient alphas ranging from a = .35 for the Blends subscale, to a = .66 for the Relativity subscale that measure the capacity to understand emotions. Similarly, Roberts et al. (2001), report 25
coefficient alphas ranging from a = .26 for the Blends subscale to a = .69 for the Relativity subscale. While Mayer et al. (2001) have argued that the full-scale and branch level reliabilities of the MEIS render the test acceptable for use by others in research and applied contexts (Mayer et al., 2001), they have also noted that the scores that individuals obtain at the subscale level need to be accurate reflections of their ability. As such, Mayer et al. (2001) have recognized the importance of improving the reliability of ability-based tests of EI at the subscale level for use in more applied contexts.
3.1.6
Factor structure
Although separate factor analyses of the MEIS (i.e., Ciarrochi et al., 2000; Mayer et al. 1999; Roberts et al., 2001) have produced broad similarities in results, they have also presented some inconsistencies between the factor structure of the MEIS and the underlying four-branch model of EI it has been designed to assess. Exploratory factor analysis of the MEIS by Mayer et al. (1999), found evidence for a general factor of EI and three correlated primary factors. The main discrepancy between the factor analytic results and the underlying four-branch theory of EI was that the subscales assessing the ability to assimilate emotions in thought and the ability to understand emotions all loaded on a single factor Emotional Understanding. Exploratory factor analysis of the MEIS by Ciarrochi et al. (2000) similarly found evidence for a general factor of EI, however, this general factor was best represented by two primary factors. In this analysis the subscales assessing the ability to perceive emotions loaded on the first primary factor and subscales assessing the other three branches of the model (Assimilation, Understanding and Management) loaded on the second factor. As such, the results of this study failed to support Mayer et al.s (1999) findings that the MEIS 26
subscales measure three related dimensions of EI. It should be noted however, that the discrepancy in these factor analytic results could be attributed that the relatively lower internal reliabilities of the MEIS subscales that w found in this study, and the fact ere that Ciarrochi et al. (2000) only used 10 of the 12 MEIS subscales. Finally, exploratory factor analysis of the MEIS by Roberts et al. (2001) revealed a similar pattern of results to those of Mayer et al (1999) provid ing evidence for a general factor of EI representing three primary factors. Unlike the factor analytic results of Mayer et al.s (1999) study the subscales assessing the ability to assimilate emotions in thought did not load significantly on any of the factors, and the factors identified were Emotional Perception, Understanding and Management. These findings therefore failed support the second factor identified in the study by Mayer et al. (1999) combining the subscales assessing the ability to assimilate emotions and the ability to understand emotions. Roberts et al. (2001) concluded that their findings suggest that the subscales assessing the ability to assimilate emotions in thought (branch two of the EI model), were both factorially complex and under-represented by the MEIS. Subsequent confirmatory analyses via structural equation modelling by Roberts et al. (2001), revealed that a four-factor model was found to provide a more reasonable fit with the data than the three-factor model put fourth by Mayer et al. (1999). A first order fourfactor model (not including a general factor of EI) was found to provide the most reasonable fit, a result not entirely consistent with the underlying theory or previous factor analytic results (Roberts et al., 2001).
3.1.7
The Mayer-Salovey-Caruso-Emotional Intelligence Test
The MSCEIT (Mayer et al., 2000d) has been designed to improve upon the MEIS in the areas of scoring, reliability and factor structure. Mayer, et al., (2000d) have 27
developed a more comprehensive expert scoring criterion for the MSCEIT; have attempted to build upon the reliability of the test at the subscale level through item selection methods; and have developed a test with a facture structure more consistent with their underlying theory of EI using a collection of new subscales. While the expert-scoring criterion for the MEIS was based on responses to the test set by Mayer and Caruso, the expert-scoring criterion for the MSCEIT is based on responses to the test items from 21 members of the International Society of Research in Emotion (ISRE). Recent analyses by Mayer et al. (2003) of the MSCEIT
standardization data (n = 2112) demonstrated a higher level of convergence between expert and consensus scoring methods (r = .908) than that found with the MEIS. Mayer et al. (2003) also reported that there was higher inter-rater reliability in identifying correct alternatives to the test items amongst the expert group than a matched sample from the standardisation group. Additionally, the standardisation group as a whole obtained significantly higher scores on the Emotional Perception and Emotional Understanding subscales when scored with the expert scoring criterion than when scored with the consensus scoring criterion. With such, Mayer et al. (2003) concluded that the expert scoring method may provide a more accurate criterion for identifying correct answers to the test items particularly in the areas where the scientific study of emotion may have provided the expert group greater institutionalised knowledge concerning emotions. It was argued that the experts may have better emotional
perception and emotional understanding, given that a great deal of emotion research has focused on coding emotional expressions (e.g., Ekman & Friesen, 1975; Schere, Banse, & Wallbott, 2001), and delineating emotional understanding (e.g. Ortony, Clore & Collins, 1988; cf: Mayer et al., 2003). However, these findings are yet to be replicated in independent analyses. Moreover, while Roberts and colleagues have acknowledged
28
that the correla tion between expert and consensus scoring methods presents a significant advance in establishing reliability, they also assert that this latest result still leaves a number of questions unresolved (Zeidner, et al., 2001). Zeidner et al. (2001) question whether a group of emotion experts necessarily have high EI and thus provide truly veridical or more correct answers to items assessing emotional abilities. Zeidner et al. (2001) also note that the MSCEIT does not comprise items of graduated difficulty across the full range of abilities assessed (unlike traditional ability measures), and as such, question whether the MSCEIT may be more effective in screening for emotional stupidity rather than in discriminating levels of EI at the upper end of the range (Zeidner et al., 2001, p.268). Indeed the consensual scoring method precludes the existence of items on the test for which only 10% or 20% of the population can answer correctly (e.g., those with very high EI) and the high correlation between the consensus and expert scoring methods suggests that the latter method does not provide items that are only correctly identified by a small percentage of the population. In addition, the high level of convergence between the expert and consensus scoring methods found with the MSCEIT should be replicated. If the magnitude of correlation between the two methods is replicated, such findings may refocus, as argued by Mayer et al. (2001), questions concerning the validity of the scoring protocols to What does consensus mean?, and Is this form of determining a correct answer much different than that used in cognitive intelligence tests? (p.236).
3.1.8
Reliability and factor structure of the MSCEIT
Like the MEIS, the MSCEIT has been designed to assess the four conceptually related abilities of Mayer and Saloveys (1997) ability model of EI. Scores on the
29
MSCEIT represent three categories; (1) an Overall EI score reflecting a general level of EI; (2) two area scores, Experiencing EI reflecting the ability to identify emotions and to assimilate emotions in thought, and Strategic EI reflecting the ability to understand and manage emotions; and (3) four branch scores (each measured by two sub-tests) that assess the four primary abilities of Mayer and Saloveys (1997) model. Reliability analyses of the MSCEIT with the standardisation sample suggest that it has good internal consistency at the full-scale, area and branch level. Mayer et al. (2003) report split-half reliabilities ranging from r = .93 to r = .91 at the full-scale level, split-half reliabilities ranging from r = .90 to r = .86 at the area level, and split- half reliabilities ranging from r = .91 to r = .76 at the branch level (according to the consensus and expert scoring criteria). The reliability of the eight individual subscales were higher
than those of the MEIS (ranging from a = .64 to a = .88), however, approximately half of the subscales have coefficient alphas below the a = .7 criterion (Mayer et al., 2003). Factor analyses of the MSCEIT suggest that its factor structure better represents the underlying theory of Mayer and Saloveys (1997) ability model. Mayer et al. (2003) assessed whether 1, 2 and 4 oblique factor models of the MSCEIT provided a statistically significant fit with the standardisation data via structural equation modelling. The general factor model, two factor Experiential and Strategic model, and four primary factor model, were all found to exhibit reasonably good model- fit statistics suggesting that each model provides viable representations of the tests underlying factor structure. However, the four- factor model (that was most consistent with the underlying theory) was found to provide the best representation of the data in comparison to the two and one factor models respectively. These factor analytic results suggest that the dimensional structure of the MSCEIT may be more consistent with the underlying theory of EI than its predecessor the MEIS.
30
3.1.9
Summary
Research findings with the MSCEIT suggest that its psychometric properties are considerably better than those of its predecessor the MEIS, particularly with respect to the scoring, reliability and the factor structure of the test. The level of convergence between the consensus and expert scoring methods further demonstrates that more and less correct answers to the test items may exist, and findings with the expert criterion suggest that more and less correct answers to the test may exist with respect to a more objective criterion (particularly for Emotional Perception and Understanding). Moreover, the overall psychometric quality of the MSCEIT appears better than the MEIS with relatively higher reliability coefficients, and a factor structure more consistent with the underlying theory. Nonetheless, these findings need to be replicated, particularly given that the MSCEIT represents an entirely new collection of tasks and items (Mayer et al., 2003).
3.1.10 Objectives of the present study
In their initial research study with the MSCEIT, Mayer et al. (2003) examined three of its fundamental psychometric properties; (1) the level of convergence between the consensus and expert scoring methods; (2) the reliability of the MSCEIT; and (3) its factor structure. The current study similarly examines the relationship between the consensus and expert scoring methods, and the reliability and factor structure of the MSCEIT with an Australian general population sample. The current study also expands upon Mayer et als. (2003) original work by examining the relationship between
31
consensus scores determined with Mayer et als. standardisation data and consensus scores determined with the present sample, and their respective relationships with expert-based scores. In addition, the current study examines; (a) differences in MSCEIT scores according to gender; and (b) the relationship between scores on the MSCEIT and age. The objective of these analyses was to repliate Mayer et als. (2003) findings, and to examine the extent to which the consensual norms determined with Mayer et als. standardisation data provide a rele vant scoring criterion for another Western society, specifically the Australian population. On the basis of Mayer et als. (2003) research findings it was hypothesised that; (1) there would be a strong relationship between consensus and expert-based scores on the MSCEIT; (2) that the sample would obtain somewhat higher test scores when scored with the expert scoring method on the Perceiving and Understanding branch scores reflecting the higher inter-rater reliability and superiority of the expert scoring criterion found by Mayer et al. (2003); (3) that the MSCEIT would exhibit high internal consistency reliability at the full-scale and branch level; (4) that the factor analytic results obtained by Mayer et al (2003) would be replicated; (5) that females would obtain significantly higher test scores than males as has been found with the MEIS (Ciarrochi et al., 2000; Mayer et al., 1999); and (6) that there would be positive relationships between scores on the MSCEIT and age, supporting previous research with the MEIS demonstrating age related differences (Mayer et al., 1999) and those reported in the MSCEIT technical manual (Mayer et al., 2000a).
32
3.2
Method
3.2.1
Participants
The sample comprised 451 participants (297 females, 150 males, 4 unreported) ranging in age from 18 to 79 years with a mean age 37.39 years (SD = 14.13). The participants were drawn from the general population across the two most populated Australian states, Victoria and New South Wales via advertisements. The ethnic composition of the sample was diverse comprising; 62%(279) White Caucasian Australians; 17%(71) White Caucasian Emigrants; 8%(38) Asian/Pacific Islanders; and 13%(61) other/not reported. The levels of education amongst the sample was also diverse; 2.2% (10) reported to have comple ted primary school education only; 22%(99) reported to have completed secondary school education only; 17%(76) reported to have completed a tertiary certificate/diploma; 24%(106) reported to have completed an undergraduate degree; and 19%(86) reported to have completed a postgraduate degree; (16%(73) not reported). In summary although there was a gender imbalance in the sample it was relatively diverse with respect to age, ethnicity, and levels of education.
3.2.2
Materials
3.2.2.1 The Mayer-Salovey-Caruso Emotional Intelligence Test
Participants completed the MSCEIT Research Version 1.1 (RV1.1), a 292 item test comprising 12 subscales designed to measure the four major abilities of Mayer and Saloveys (1997) ability model of EI. For the purposes of replication however, the test
33
publisher Multi-Health Systems (MHS), scored the present data according to the shorter MSCEIT Version 2 (V2) scoring algorithms. The MSCEIT V2 was comprised from the MSCEIT RV 1.1 by reducing the number of test items and subscales. More specific details concerning this reduction procedure can be found in the MSCEITs technical manual (Mayer et al., 2000d). Data that represents the MSCEIT V2 can be obtained from the MSCEIT RV 1.1 (as done by the current study) because no new items were created for MSCEIT V2, thus MSCEIT V2 is essentially inherent within MSCEIT RV 1.1. Moreover, although the MSCEIT V2 was designed to make the test-taking experience smoother for the testtaker (Mayer et al. 2000d, p.64), and to increase the research and practical utility of the test by making it shorter, Mayer et al. report a high degree of correspondence between the two forms. For example, correlations between factor-based scores of the MSCEIT RV 1.1 and V2 are reported to range from r = .96 for Understanding Emotions to r = .80 for Emotional Perception (Mayer et al., 2000d). Thus although the test used in the present study may have been somewhat more laborious for the participants to complete, data collected with the MSCEIT RV 1.1 can be scored using the MSCEIT V2 scoring algorithms to produce representative data. MHS returned the database with raw responses that represented MSCEIT V2 data (141 items), consensus scored item and subscale scores, and expert scored item and subscale scores.
3.2.2.2 MSCEIT V2
The 141- item MSCEIT V2 comprises 8 subscales, 2 pertaining to each of the four branches of the ability model (Mayer & Salovey, 1997); (1) Perceiving Emotions (Perceiving), (2) Using Emotions to Facilitate Thought (Using), (3) Understanding
34
Emotions (Understanding), and (4) Managing Emotions (Management). Each subscale comprises a number of so-called item parcels that contain a number of individual items. For example, in the Faces subscale, participants viewed photographed facial expressions (each photograph representing an item parcel), and were asked to indicate the extent to which different emotions (which form the individual items for a parcel) were inherent in the facial expressions presented. In the Faces test five individual items combine to form an item parcel as they each ask about a different emotion related to the same face. Some subscales contain free-standing items in that they comprise only one response per stimulus. To reduce correlated measurement error and to ensure that results generalise across response methods, response formats are varied across the subscales (Mayer et al., 2003). Perceiving Emotions is measured by the Faces and Pictures subscales. In the Faces test participants viewed four photographed faces and were asked to indicate the degree to which five specific emotions were inherent in the stimulus on a five point rating scale. The Pictures test is similar except that different landscapes and abstract designs are presented as the stimuli and the response scale consists of cartoon faces depicting varying degrees of the specific emotions. Facilitation was measured by the Sensations and Facilitations subscales. In the Sensations test participants were asked to imagine specific emotions and to indicate the extent to which they matched different sensations (e.g., Imagine feeling frustrated, how much is the feeling of frustration like the following sensations; hot, slow, green etc). In the Facilitation test participants were asked to indicate the extent to which specific emotions assisted cognitive tasks or behaviours (e.g., the extent to which contentment, fear and happiness might be helpful to feel when negotiating with a salesperson to reduce the price on a product). Understanding Emotions was measured by the Blends and Changes subscales. The
35
Blends test required participants to identify emotions that combine to form more complex feelings (e.g., that sadness, guilt and regret combine to form (a) grief, (b) annoyance, (c) depression etc). The Changes test required participants to identify emotions that result from the intensification of certain feelings (e.g., a person felt more and more ashamed and began to feel worthless, then the person felt (a) overwhelmed, (b) depressed, (c) ashamed, etc). Finally, Managing Emotions was measured by the Emotional Management and Emotional Relationships subscales. In the Emotional Management test participants were asked to indicate how effective specific actions were in regulating certain moods and emotions (e.g., reducing anger, prolonging joy, keeping frustration at bay). Similarly, in the Emotional Relationships test participants were asked to indicate how effective the actions of a person were in regulating or managing the emotions of another person. More detailed information concerning the subscales of the MSCEIT are contained in the technical manual (Mayer et al., 2000d).
3.2.2.3 MSCEIT scoring
The MSCEIT yields eight subscale scores determined by summing the weights for each item (as determined by either the consensus or expert scoring method discussed in the introduction); four branch scores determined by summing the two corresponding subscale scores that measure each branch; and two area scores (1) Experiential EI, that represents individual differences in the lower order abilities, perceiving and using emotions in thought; and (2) Strategic EI, that represents individual differences in the higher order abilities, understanding and managing emotions. In addition, an Overall score can also be calculated, representing a general emotional intelligence score analogous to IQ. MHS scored the present data using the consensus weights determined
36
from the MSCEIT standardization sample, which is referred to as the American Consensus Scores (n = 2112, Mayer et al., 2003), and the expert weights determined from the sample of experts drawn from the ISRE which is referred to as the Expert Scores. For the purpose of comparison consensus weights using the present sample via the procedure outlined by Mayer et al. (2000d) were also generated, which is referred to as the Australian Consensus Scores.
3.2.3
Procedure
Participants responding to advertisements about the study collected pencil and paper MSCEIT item booklets and scannable answer sheets and were briefed about the purpose of the study. The briefing sessions were conducted according to the instructions for remote administrations of the MSCEIT outlined in the technical manual (Mayer et al., 2000d) with the administrator emphasising that the MSCEIT was to be completed independently without input from others and in its entirety. Upon completion participants returned the item booklets and answer sheets, and received a small stipend for participating.
3.3
Results
Prior to conducting the analyses a missing value analysis was performed to evaluate the validity of the participants responses. Mayer et al. (2000d) consider participants responses to be invalid if 10% or more of a given subscales items are missing. The missing value analyses found 20 of the participants responses to be invalid as per this criterion and were omitted from the data. As such, there results
37
reported hereafter are based on the sample for which all data at the subscale level was complete.
3.3.1
Descriptive statistics
The reliabilities, means and standard deviations of the present sample according to the American and Australian consensus criteria, and expert criterion are presented in Table 3.1.
38
Table 3.1 Means, Standard Deviations and Reliabilities for the MSCEIT V2 according to American Consensus Scores, Expert Scores and Australian Consensus Scores ACSa MSCEIT Subscale Perceiving M .47 SD .08 .09 .10 .05 .08 .06 .07 .08 .08 .06 .05 .09 .06 .06 .05 r* .90 .80 .86 .73 .63 .49 .71 .63 .50 .76 .59 .55 .91 .78 .91 M .47 .50 .44 .43 .46 .41 .53 .53 .52 .41 .41 .42 .45 .47 .46 AUSCb SD .08 .10 .09 .06 .08 .07 .07 .08 .10 .06 .06 .10 .05 .06 .05 r* .90 .81 .86 .80 .64 .54 .73 .60 .56 .74 .61 .53 .91 .79 .91 M .49 .53 .46 .40 .42 .39 .62 .63 .60 .41 .39 .43 .45 .51 .48 ESc SD .11 .15 .11 .05 .06 .06 .11 .12 .14 .07 .06 .11 .06 .07 .06 r* .89 .84 .85 .67 .48 .48 .69 .60 .54 .66 .48 .51 .90 .76 .89
Faces .48 Pictures .45 Using .42 Facilitation .45 Sensations .39 Understanding .52
Changes .54 Blends .49 Managing .40
Management .39 Relationships .41 Experiencing EI Strategic EI Overall EI .44 .46 .45
Note: The means and standard deviations reported are unscaled as per Mayer et al. (2003). a. American Consensus Scores b. Australian Consensus Scores c. Expert Scores * Coefficient alpha reliabilities are reported at the subscale level due to item homogeneity and split- half reliabilities (with Spearman Brown correction) are reported at the Branch, Area and Overall test levels due to item heterogeneity as per Mayer et al. (2003). The means and standard deviations of the present sample were highly comparable across the three scoring methods and very similar to those reported by Mayer et al.
39
(2003). Moreover the present sample were found to obtain somewhat higher test scores according to the expert scoring method in the areas where Mayer et al. (2003) reported a higher degree of convergence amongst the expert group (i.e., Perceiving and Understanding Emotions. For example, the mean Understanding branch score of the present sample was significantly higher according to expert scores than the American consensus scores t(430) = 41.53, p<.001, Cohens d = .91. This finding provides further evidence that the expert scoring criteria may prove superior to the general consensus in these areas. The correlation between the mean scores of the present sample and those reported by Mayer et al. (2003), was r = .74, and r = .89 for Australian consensus and expert-based scores respectively, confirming that the profile of mean scores for both scoring criteria were highly comparable with the previously reported results, a finding consistent with other studies in the area (Roberts et al., 2001).
3.3.2
Reliability
Table 3.1 also reports the internal reliability of the MSCEIT at the full scale (Overall EI), area, branch and subscale level. Split-half reliability coefficients (with Spearman Brown correction) have been determined for the full scale, area and branch scores as the items that combine to form these scores are heterogeneous, while coefficient alphas have been determined for the 8 subscales as the items at the subscale level all share the same response format (as per Mayer et al., 2003). As shown in Table 3.1, the MSCEIT full-scale split- half reliabilities were high according to the American and Australian consensus scores, and expert scores respectively. The split-half reliabilities across the area and branch scores were all above the criterion of r = .70 (Nunnally, 1967) for consensus determined scores, however, the reliability coefficients
40
of the expert based scores were somewhat lower, particularly for the Using and Managing branch scores that were below the criteria of .70. The reliabilities of the MSCEIT subscales were varied ranging from a high of a = .86 for the Pictures subscale to a low of a = .48 for the Facilitation, Sensations and Management subscales. In general, the reliabilities of the MSCEIT subscales were somewhat lower than those reported by Mayer et al. (2003). Collectively the findings of the current study suggest that while the reliabilities of some of the MSCEIT subscales are low, the MSCEIT is reliable at the full-scale, area and branch level. A
3.3.3
EI and age
Although there were no significant relationships between Overall MSCEIT scores and age, there were a number of small relationships found between the MSCEIT subscales and age in the present sample. Across all three scoring criteria there was a small negative correlation between age and scores on the Faces subscale (ranging from r = -.12, p<.05 to r = -.101, p<.05; according to American Consensus and Expert scores respectively). In contrast, there were small positive relationships between age and scores on the Facilitation subscale (ranging from r = .19, p<.001 to r = .16, p<.01 according to the Australian Consensus and Expert scores), and the Management subscale (ranging from r = .13 p<.01 to r = .12, p<.05 according to the American Consensus and Expert scores). These findings are consistent with those reported in the MSCEIT technical manual (Mayer et al. 2000d). While these correlations were all significant, the magnitude of the correlations explain little more than 1% of the variance suggesting that there was no real or meaningful increase in EI associated with age in the present adult population sample.
41
3.3.4
EI and gender Table 3.2 presents descriptive statistics for the MSCEIT branch and Overall EI
scores for female and male participants. Table 3.2 Descriptive Statistics for MSCEIT Branch and Overall EI Scores by Gender* MSCEIT Expert Scores M Perceiving EI Females Males Using EI Females Males Understanding EI Females Males Managing EI Females Males Overall EI Females Males .51 .47 .41 .38 .63 .58 .42 .38 .49 .45 SD .10 .10 .04 .06 .10 .13 .07 .07 .05 .07 American Consensus Scores M SD .48 .45 .43 .39 .54 .50 .41 .37 .46 .43 .07 .09 .05 .06 .06 .09 .06 .07 .04 .06 Australian Consensus Scores M SD .48 .45 .45 .41 .54 .50 .43 .39 .47 .44 .07 .09 .05 .07 .07 .09 .06 .07 .04 .06
* Females n = 286; Males n = 143
Consistent with previous findings, females scored significantly higher than the males on the MSCEIT using all three scoring criteria. According to Expert-based scores, female participants scored significantly higher than the male participants by approximately half a standard deviation (e.g., for Overall EI, F(1, 427) = 44.80, p<.001, d = .65; 15 by .65 = 9.75 EIQ points). Similarly, in terms of the Australian and American consensus weights, female participants scored approximately two thirds of a standard deviation above the male participants (F(1, 427) = 54.97, p<.001, d = .73; F(1, 427) = 51.23, p<.001, d =.69) respectively. These findings are consistent with those previously found
42
with the MEIS (Ciarrochi et al., 2000, Maye r et al., 1999), and those reported in the MSCEIT technical manual (Mayer et al., 2000d).
3.3.5
MSCEIT inter-correlations
Table 3.3 presents the intercorrelations amongst the Australian consensus and expert-based scores on the 8 subscales and Overall MSCEIT score.
Table 3.3 Intercorrelations Among Australian Consensus (Below the Diagonal) and Expert-Based (Above the Diagonal) MSCEIT Subscale Scores* MSCEIT Test 1. Faces 2. Pictures 3. Facilitation 4. Sensations 5. Changes 6. Blends 7. Management 8. Relationships 9. Overall EI F 0.94 0.32 0.13 0.31 0.23 0.21 0.20 0.30 0.60 P 0.33 0.99 0.29 0.29 0.17 0.16 0.19 0.17 0.56 FA 0.14 0.28 0.92 0.30 0.25 0.16 0.36 0.21 0.54 S 0.30 0.23 0.29 0.96 0.34 0.19 0.29 0.29 0.61 C 0.18 0.14 0.25 0.34 0.95 0.51 0.26 0.33 0.65 B 0.17 0.15 0.18 0.21 0.51 0.95 0.21 0.30 0.60 M 0.16 0.18 0.26 0.28 0.30 0.27 0.86 0.35 0.55 R 0.25 0.18 0.21 0.30 0.32 0.28 0.33 0.90 0.64 O 0.61 0.54 0.47 0.56 0.67 0.65 0.50 0.62 0.97
* correlations below r = .15 are significant at the p<.01; correlations above r = .15 are significant at the p<.001 level. The correlation between consensus and expert-based scores for each subscale is presented in boldface down the main diagonal of the table respectively.
Consistent with previous findings (Mayer et al., 2003) and the theory of EI that the MSCEIT has been designed to assess (Mayer & Salovey, 1997), there was a positive manifold of correlations amongst the subscales according to the different scoring methods (Australian consensus and expert-based scores are presented in Table 3.3). In addition, each subscale correlated mostly highly with the subscale with which it
43
combines with to produce a branch score (e.g., the Faces and Pictures subscales which measure Perceiving emotions), with the exception of the subscales measuring Using Emotions (which also exhibited low reliability). Very strong relationships were also found between scores determined by the different scoring methods. American and Australian consensus determined Overall EI scores correlated r = .99 with correlations among the subscales ranging from r = .96 for the Sensations subscale to r = .99 for the Faces subscale. This level of convergence (i.e., 98% of the variance) suggests that the American consensus weights determined from the MSCEIT standardization sample (Mayer et al., 2003) may be cross culturally applicable at least within other Western societies. As shown in Table 3.3, similar results were found between Australian consensus and expert-based scores. Finally, American consensus and expert-based Overall EI scores (not shown in Table 3.3) correlated r = .97, with correlations between subscale scores ranging from r = 1.0 for the Sensations subscale to r = .88 for the Management subscale. The correlation between the two correlation matrices (that is, the correlation matrix below the diagonal Australian consensus, and the correlation matrix above the diagonal expert as shown in Table 3.3) was also very high (r = .93), suggesting a high degree of correspondence (86% of the variance), between the pattern of intercorrelations based on these two scoring criteria. Collectively these findings suggest that the two scoring criteria for the MSCEIT are highly related, a finding consistent with those reported previously (Mayer et al., 2003). Ind eed, the correlation between the American consensus scores correlation matrix of the present study, and that reported by Mayer et al., (2003) was r = .60, and the correlation between the expert-based scores correlation matrix of the present study and that reported by Mayer et al., (2003) was r = .74. The replicability of the intercorrelations among the MSCEIT subscales across two quite different data sets
44
(particularly those generated with the expert criterion), suggests that they are robust. This property was also demonstrated with the MEIS (Roberts et al., 2001).
3.3.6
Confirmatory factor analysis
Mayer et al., (2000; 2003) used confirmatory factor analyses to test a one factor model that loaded all eight MSCEIT subscales; a two- factor model that represented the area scores on the MSCEIT where the Perceiving and Using emotions subscales loaded on one factor and the Understanding and Managing subscales loaded on the other; and a four-factor model representing the four branch scores of the MSCEIT were each factor loaded the two designated subscales that comprise the branch scores. These models were tested with the present sample using the same program utilised by Mayer et al. (2003), that is, AMOS (Arbuckle, 1999), and with the same model parameters, that is, for each respective model tested; error variances were uncorrelated; latent variables were correlated (i.e., oblique factors); and all other paths were set to zero. In addition, consistent with Mayer et als. (2003) approach, with the four- factor model the two within-area latent variable co-variances (i.e., between Perceiving and Using and between Understanding and Managing) were additionally constrained to be equal [to each other] so as to reduce a high covariance between the Perceiving and Facilitating branch scores. (p. 103). For the purpose of replication, these models were assessed using the American consensus and expert-based scores. Consistent with the methodology employed by Mayer et al. (2003), three model fit indices were chosen to assess the degree to which the models represented the dimension structure of the MSCEIT, the normed fit index (NFI; Bentler & Bonett, 1980), the Tucker Lewis index (TLI; Tuker & Lewis, 1973), and the root- mean-square
45
error of approximation (RMSEA; Steiger, 1990). Table 3.4 presents the parameter estimates of the observed tasks on the latent variables (based on the American consensus determined scores), and the goodness-of-fit statistics for the models tested for both American and Expert-based scores.
Table 3.4 MSCEIT Parameter Estimates and Goodness-of-fit Statistics for the One-Two-, and Four-Factor Models* Model tested Two- factor I .51 .46
Subscale Branch 1: Perceiving Faces Pictures Branch 2: Using Facilitation Sensations Branch 3: Understanding Changes Blends Branch 4: Managing Management Relationships Fit Indices X2 df NFI TLI RMSEA Fit Indices X2 df NFI TLI RMSEA
One-factor I .47 .40
Four- factor I .59 .54 II .45 .47 .48 .56 .61 .60 II III .62 .68 .83 .52 .57 .60 IV .51 .48 .58 .54 .54 .58 Goodness-of-fit Statistics (Consensus Scores) 97.78 83.05 37.59 20 19 15 .83 .86 .93 .80 .83 .92 .09 .09 .06 Goodness-of-fit Statistics (Expert Scores) 93.03 64.78 31.91 20 19 15 .83 .89 .94 .80 .91 .95 .09 .07 .05
* Consistent with Mayer et al. (2003), the boldfaced italicised roman numerals indicate the factors specified in each model with the estimated factor loadings (based on American consensus scores) for each of the tasks associated with that specific factor
46
presented beneath; all other factor loadings were fixed at zero. NFI = normed fit index; TLI = Tucker Lewis index; RMSEA = root- mean-square error of approximation. The subscales are ordered according to the four-branch model of EI the MSCEIT has been designed to measure (Mayer & Salovey, 1997). Consistent with the previous confirmatory analyses (Mayer et al., 2003), progressively better model- fit statistics were found from the one- to the four-factor models (i.e., based on consensus scores four vs. two factors, X2 (4) = 83.05 37.59 = 45.46, p < .01; two vs. one factors, X2 (1)= 97.78 83.05 = 14.73, p < .01). Also consistent with previous analyses the model fit statistics for consensus an expert-based scores were highly similar although the two and four- factor modes produced slightly better fit statistics according to expert-based scores. Unlike the previous findings however, only the four- factor provided a good fit with the present data according to all three model fit statistics, some of the model- fit statistics for the two- and one-factor models were below the acceptable ranges as shown in Table 3.4. As a final analysis of the extent to which the four- factor model shown in Table 3.4 provided the best fit to the present data, a non-sense four-factor model was also tested such that each branch score was measured by theoretically dissimilar subscales. Specifically, a model in which the first factor was measured by the Management and Pictures subscale; the second factor was measured by the Changes and Sensations subscales; the third factor was measured by the Facilitation and Blends subscales; and finally, the fourth- factor was measured by the Faces and Relationships subscales. This model was not found to provide an acceptable fit with the data according to all of the fit indices (X2 = 96.29, df = 15, NFI = .82, TLI = .71 and RMSEA = .11). This finding was consistent with the similar nonsense model tested by Mayer et al. (2003), and further demonstrated that the
47
four-factor model consistent with the underlying theory of EI the MSCEIT has been designed to assess provided the best fit with the present data.
3.4 Discussion
On the whole, the results of the current study replicate those reported by Mayer et al. (2003). Perhaps one of the most important of which, are the replicable findings concerning the alternative scoring criteria. The issue of convergence between the consensus and expert scoring criteria was identified as one of the most central psychometric issues facing the predecessor to the MSCEIT, the MEIS (Roberts et al., 2001). In recognition of such, Mayer and colleagues sought to improve the validity of the expert-scoring criterion for the MSCEIT by employing a more optimal group of judges to serve as experts (Mayer et al., 2000a). In the present study consensus and expert determined scores were highly correlated (r = .97) replicating the relationship between the two scoring methods previously reported by Mayer et al. (2003; i.e., r = .98). Moreover, there was a strong relationship between the pattern of intercorrelations based on the two scoring criteria (r = .93), further illustrating the relatively high degree of correspondence between them. To put this level of convergence into perspective (as did Roberts et al., 2001), the 95% confidence interval for a true score would be 1.96 multiplied by the standard error of measurement SEM, (which based on the present studies results would be at worst .73 (r = .862 for Management) and at best .24 (r = .985 for Pictures). Converted to represent standard IQ type scores with a mean of 100 and a standard deviation of 15, the SEM suggests that a persons score would be likely to vary between 10.1 IQ points (.73x15) at worst (i.e., on the Management subscale), and 3.6 IQ points (.24x15) at best (i.e., on the
48
Pictures subscale) depending on which scoring criteria was used. Roberts et al., (2001) noted that high SEM values adversely impact the comparative utility of an instrument and its predictive validity severely constraining its utility. The findings of the present study suggest that a marked improvement in scoring convergence has occurred with the scoring criteria of the MSCEIT. Roberts et al. (2001) demonstrated that a test score would be likely to vary between 79 and 121 IQ points on the MEIS according to the different scoring criteria. The level of convergence between the scoring criteria for the MSCEIT suggest that consensus and expert scoring assess decidedly similar constructs and that the utility of the test may have improved as a result. Not surprisingly, but of equal importance, the two scoring criteria showed a very similar pattern of factor analytic results, relationships with age and similar differences between females and males, further demonstrating the level of convergence between them. The present sample also achieved higher expert-based test scores in comparison to consensus-based test scores in areas where the expert group have been previously found to demonstrate higher inter-rater reliability in identifying correct answers (Mayer et al., 2003; i.e., Perceiving and Understanding emotions). This result further substantiates the findings of Mayer et al. (2003), that the expert criterion is superior to the consensus criterion in terms of determining more and less correct test answers (at least in the areas where research has possibly established clear criteria for answers) and Mayer et als. conclusion that the expert criterion may be the criterion of choice for ability tests if such findings are further replicated. Finally, there was a strong relationship (r = .99) between scores determined using American consensus weights and consensus scores determined with the present sample (Australian consensus weights), suggesting that American consensus weights may be applicable as a scoring criterion for use in research and applied contexts within othe r Western societies. However,
49
further research of this type is needed in order to substantiate this notion. Collectively, the findings of the present study concerning the MSCEIT scoring criteria replicate the higher level of convergence between the two alternate methods (in comparison to the MEIS), representing a significant improvement in establishing reliability (Roberts et al., 2001). The theory underlying the MSCEIT purports that for the construct of EI to be considered an intelligence, measures should meet three standard intelligence criteria; (1) the variables measured need to be operationalized as abilities; (2) the variables should show a positive manifold of correlations; and (3) there should be age related differences. The findings of the current study support the first and second of these criteria. The variables measured have been operationalized as abilities (Roberts et al., 2001) and a positive manifold of correlations among the subscales was found according to all three scoring methods consis tent with previous findings (Mayer et al., 2003). The age related criteria was not supported, while there were small significant relationships between scores on the MSCEIT and age, the magnitude of these relationships suggest that there were very little differences in EI associated with age. However, it should be noted that the relationship between age and EI (or lack thereof in the present study) does not nullify a developmental trend, rather it suggests that there is perhaps less variance in EI associated with age within adult populations. Indeed Mayer et al. (1999) have found that adult criterion groups tend to exhibit higher EI than adolescent criterion groups. Further research examining age related differences in MSCEIT scores is needed in order to determine whether EI develops with age and the potential underlying developmental trend. Although Mayer et al. (1999) have stated three standard criteria that measures of EI should meet in order for the construct to be considered an intelligence, this should
50
not be a definitive list and other criteria should also be debated. Roberts et al. (2001) have recently asserted that measures of EI should comprise one of the main criterion for intelligence tasks, that is, veridical scoring criteria in which answers to test items are judge either correct or incorrect via an unequivocal formal rule-bound system. Other criteria for EI measures might stem from areas in which intelligence (in comparison to EI) has been more widely studied such as biological differences (Wickett, Vernon & Lee, 2000) and individual differences in information processing speed (Deary & Stough, 1996). It should also be recognised that the nature and definition of intelligence (IQ) is still a matter widely debated (Sternberg, 2000), thus criteria for EI that are not standard for intelligence should also be explored and debated. Consistent with previous research, in the current study females were found to score higher than males on the MSCEIT by approximately half a standard deviation (Ciarrochi et al., 2000; Mayer et al., 1999). Importantly, this finding was consistent across the two scoring methods. In their evaluation of the MEIS, Roberts et al. (2001) found that gender differences in EI varied as a function of the scoring criteria used. The consistency of gender differences across the scoring methods found in the present study further illustrates the level of convergence between the two scoring methods for the MSCEIT. The consistency of gender differences across disparate population samples requires further research. Mayer et al. (1999) have speculated that women may be socialized to read emotions more carefully or to be more in-tune with them, or indeed that they may be better biologically equipped to process emotional information. Research on EI in general has yet to address gender differences in more detail despite the potential incremental understanding about the nature of the construct that might result. A better understanding about the nature and causes of gender differences in EI
51
might, for example, help delineate the underlying biology, heritability and nature of cultural and environmental influences on EI. In general, the reliability of the MSCEIT was found to be consistent with that reported previously (Mayer et al., 2003) according to all scoring methods. Split-half reliabilities for the Overall scale, area and branch scores were high for consensusdetermined scores and relatively good for expert-based scores. However, somewhat lower reliability coefficients (ranging from a high of a = .86 to a low of a =. 48) were found across the subscale scores in the present study. It should be noted that in the present study, MSCEIT V 2.0 data was carved out of participants responses to MSCEIT V1.1 (by MHS) that is more than twice the size of V2.0 (i.e., 292 items verses 141). Although the items are identical, participants completed a longer version of the test thus more independent reliability analyses using the shorter MSCEIT V2.0 are needed in order to more fully substantiate the tests reliability. Nonetheless, the findings of the current study suggest that some further revising of the MSCEIT, aimed at establishing acceptable internal consistency reliability across all the subscales, may be required. The confirmatory factor analyses of the MSCEIT were consistent with those reported by Mayer et al. (2003) and the underlying theory of EI the test has been designed to measure (Mayer & Salovey, 1997). The replication of factor analytic findings with relatively disparate population samples suggest that the dimensional structure of the MSCEIT may be more robust than its predecessor (the MEIS), where three independent analyses provided somewhat different results, although more independent analyses are required to substantiate this notion. The factor analytic results of the present study did not support the one- factor or gei model using conventional tests of fit, yet such results have also been found in intelligence research. General factor
52
models in intelligence have also failed to provide good model fit with various data sets. For example, using the Wechsler Adult Intelligence Scale (WAIS-III) standardization data, Ward, Ryan, and Axelrod (2000) obtained fit estimates of AGFI = .77, RMSR = .55, CFI = .90, and PGFI = .58. Saklofske, Hildbrand, and Gorsuch (2000) found very similar results with the WAIS-III using a Canadian sample: RMSEA = .11. Taken as a whole, these estimates suggest that the g factor model for intelligence does not always provide a good fit with various data sets. Although the g factor was not found to be a good fit of the data in these studies, intelligence researchers note that it would be erroneous to claim that the g factor does not exist (Jensen, 1998). Analogous with such, although the one-factor EI model did not provide a good fit with the present data, the pattern of intercorrelations and parameter estimates provide evidence for the existence of a general EI factor or gei.
3.4.1
Conclusion
In summary, the results of the current study suggest that the MSCEIT has improved over its predecessor measure the MEIS with respect to the validity of its scoring protocols, reliability and factor structure. One of the most contentious issues with the MEIS concerned the lack of convergence between the consensus and expert scoring protocols. While Roberts et al. (2001, p.224) concluded that it was difficult to see how the degree of correspondence could be improved to an acceptable level, the present findings indicate that there is significant improvement. While the present findings need further replication, they further inform the debate concerning the scoring of EI ability measures. As mentioned in the introduction, Mayer et al. (2001) noted that if the magnitude of correlation between the two methods was replicated, such findings may refocus questions concerning the validity of scoring protocols to What does 53
consensus mean?, and Is this form of determining a correct answer much different than that used in cognitive intelligence tests? (p.236). Determining correct answers to items of ability tests using a consensus scoring method is different than that traditionally used in cognitive intelligence tests (albeit not all cognitive ability subscales as noted by Mayer et al., 2001), in that consensus determined answers preclude items of graduated difficultly (i.e., items where only a small percent of the population, the highly emotionally intelligent, are capable of identifying the correct answer). However, Mayer, et al. (2000b) have argued that the emotionally intelligent should be able to identify the more correct answers to test items (according to the consensus weights), more consistently than the emotionally mediocre and that high and low scores of such do index individual differences in EI. As discussed in the introduction to this chapter, this approach to determining individual differences in EI is proving to be useful in terms of predicting real life criteria. Research to date with performance based measures of EI have shown that they may predict important human attributes such as empathy and life satisfaction (Ciarrochi et al., 2000; Mayer et al., 1999), end-of-year college grades amongst college students (Salovey, et al., in press), and problem behaviours in early adolescents (Trinidad & Johnstone, 2002 ). Much more research of this nature is needed however, in order to further establish the utility of the EI construct. Of particular value would be research examining whether or not (and to what extent), measures of EI (both self-report and performance-based) can predict incremental variance in theoretically related life criteria (e.g., psychological well-being and various facets of success in life) over and above established tests of intelligence and personality. Such research will help inform the wider debate amongst the EI literature concerning the efficacy of self-reported verses performance-based measures of EI (Mayer et al., 2000c).
54
Chapter 4; Bar-On EQ-i
CHAPTER 4 The Bar-On Emotional Quotient Inventory (EQ-i)
4.1
Overview
In this chapter the factorial validity of the Bar-On Emotional Quotient Inventory (EQ-i; Bar-On, 1997a) is examined. Bar-On (1997a) has claimed that the dimensional structure of the EQ-i comprises a general factor of EI, five-second order factors and 15 primary factors that closely represent his hierarchical model of emotional and social intelligence. It is argued that the there are several anomalies in the factor analytic methodology employed by Bar-On (1997a). In addition there are several difficulties in his interpretation of his factorial analyses that precent a clear understanding of the dimensional structure of the EQ-i. In contrast to the results provided by Bar-On (1997a), a series of exploratory and confirmatory factor analyses are reported that provide evidence for a general factor of EI and six primary factors. Differences between the results reported by Bar-On (1997a) and those of the current study are attributed largely to the more appropriate factor analytic methodology employed. Implications and directions for future research are discussed.
4.1.2
Bar-Ons (1997) model of emotional intelligence
Bar-Ons model of EI (1997a) comprises an array of emotional and social abilities, skills and personality traits, thus constituting a so-called mixed model of EI (Mayer et al., 2000a). While Bar-On (1997a) places this model under the banner of EI, it could be regarded as a somewhat broader construct that he more generically refers to as emotional and social intelligence (Bar-On 2000, p.363). Bar-On (1997a) 55
purports to have identified 15 determinants of successful emotional functioning from a review of the mental health literature, which have been operationally defined and conceptualised into the 15 components of his model. These components and their definitions (taken directly from the technical manual of the EQ-i, Bar-On, 1997a, pp. 17-21) are presented in Table 4.1.
Table 4.1 The 15 Components of the Bar-On EQ-i Component Emotional Self-Awareness (ES) Assertiveness (AS) Definition The ability to recognise ones feelings. The ability to express feelings, beliefs and thoughts, and to defend ones rights in a non-destructive manner. The ability to respect and accept oneself. The ability to realise ones potential capacities. The ability to be self-directed and self-controlled in ones thinking and actions and to be free of emotional dependency. The ability to be aware of, to understand, and to appreciate the feelings of others. The ability to establish and maintain mutually satisfying relationships. The ability to demonstrate oneself as a cooperative, contributing, and constructive member of ones social group. The ability to identify and define problems as well as to generate and implement potentially effective solutions. The ability to assess the correspondence between what is experienced and what objectively exists. The ability to adjust ones emotions, thoughts, and behaviour to changing situations and conditions. The ability to withstand adverse events and stressful situations. The ability to resist or delay an impulse, drive, or temptation to act. The ability to feel satisfied with ones life, to enjoy oneself and others, and to have fun. The ability to look at the brighter side of life and to maintain a positive attitude, even in the face of adversity.
Self- Regard (SR) Self- Actualisation (SA) Independence (IN)
Empathy (EM) Interpersonal Relationship (IR) Social Responsibility (RE)
Problem Solving (PS)
Reality Testing (RT) Flexibility (FL) Stress Tolerance (ST) Impulse Control (IC) Happiness (HA) Optimism (OP)
56
These 15 components of Bar-Ons model are described in greater detail in the EQ-i Technical Manual (Bar-On, 1997a). Within Bar-Ons (1997a) model, the 15 components are theoretically arranged into five broader or major conceptual components. These include; Intrapersonal Emotional intelligence (RAeq), representing abilities, capabilities, competencies and skills pertaining to the inner self, i.e., the ES, AS, SR, SA and IN components; Interpersonal Emotional intelligence (EReq), representing interpersonal skills and functioning i.e., EM, IR, RE; Adaptability Emotional Intelligence (ADeq), representing how successfully one is able to cope with environmental demands by effectively sizingup and dealing with problematic situations, comprising PS, RT and FL; Stress Management Emotional Intelligence (SMeq), representing the ability to manage and cope effectively with stress comprising the ST and IC components; and General Mood Emotional Intelligence (GMeq), representing the ability to enjoy life and maintain a positive disposition which comprises the HA and OP components (Bar-On, 1997a). These five major components of EI are theoretically related to a general factor of EI thus constituting a hierarchical model comprising overall EI, five-composite components, and 15 specific components at the bottom of the hierarchy (Bar-On, 1997a). Within this model, EI is defined as an array of non-cognitive capabilities, competencies, and skills that influence ones ability to succeed in coping with environmental demands and pressures (Bar-On, 1997a, p. 16). The 15 components of the model are described as non-cognitive variables that resemble personality factors (Bar-On, 1997b, p.6). It is also proposed that the components of the model develop over time (with age), change throughout life, and can be improved through training and remedial programs.
57
4.1.3
The Bar-On Emotional Quotient Inventory
The EQ-i has been designed to assess Bar-Ons (1997a) model of EI. Consistent with Bar-Ons (1997a) proposed theoretical structure of EI, the EQ- i comprises 15 subscales pertaining to the 15 components of the model, which render 15 sub-scale scores, 5 EQ composite scale scores and an overall or total EQ score. As with other self-report measures of EI, the EQ-i is described to provide an index of cross-situational consistencies in (emotionally and socially) competent behaviour and as such, provides an estimate of an individuals EI (Bar-On 2000). The EQ-i was one of the first measures of EI to be developed and is one of the more popular self-report measures of EI currently available. The EQ-i has been translated into twenty-two languages and normative data has been established in more than fifteen countries. In addition, the EQ-i comprises indices of social response bias, and response validity indicators (i.e., omission rate and an inconsistency index), that are purported to increase the accuracy of test scores. Finally, the EQ- i is a relatively brief and easy scale to administer and interpret. Results are computer generated, interpreted by the test publisher Multi- Heath Systems (MHS), and are represented in layperson terms numerically, graphically and textually in feedback reports. Nonetheless, as with all measures of EI further validation of the scale is needed (Bar-On, 2000). In particular further research into the predictive validity and dimension structure of the EQ- i is needed as discussed below. Psychometric analyses of the EQ-i reported in the technical manual (Bar-On, 1997a), indicate that it has good internal reliability and test-retest reliability. Across seven population samples, the 15 sub-scales are reported to have average-to-high internal consistency coefficients with Cronbach alphas ranging from a = .69 for RE to a = .86 for SR. Similar stability coefficients have also been reported. With a South African sample (n = 44) the average stability coefficient of the 15 sub-scales after a one58
month period was found to be r = .85, and with a smaller sub-set of this sample (n = 27) was found to be r = .75 after a four- month period. A wide number of correlation studies are also reported in the technical manual (Bar-On, 1997a), in support of the validity of the EQ- i as a measure of the ability to succeed in coping with environmental demands and pressures, and psychological well-being. For example; total EQ-i scores have been shown to correlate positively with measures of emotional stability (e.g., r = . 72 with the Emotional Stability factor of the Sixteen Personality Factor Questionnaire, 16PF; Cattell, Cattell & Cattell, 1993); overall life satisfaction (e.g., r = .41 with the Kirkcaldy Quality of Life Questionnaire, Kirkcaldy 1995); acculturation ( e.g., r = .34 with the Short Acculturation Scale, SAS; Marin et al., 1987); attribution style (e.g., r = .37, with the Attributional Style Questionnaire, ASQ; Peterson et al., 1982); and proxy measures of job performance and satisfaction (e.g., r = .51 with the Sense of Competence Questionnaire, SCQ; Wagner & Morese, 1975). Furthermore, total EQ-i scores correlate negatively with measures of poor emotional health (e.g., r = -.85 with the Ninety Symptom Check List, SCL-90; Derogatis, 1973); and depression (e.g., r = -.56 with the Beck Depression Inventory, BDI; Beck & Steer, 1987). Overall these results provide preliminary evidence for the construct validity of the EQ-i. One question currently concerning the EQ-i, however, is its discriminant validity from personality traits such as neuroticism and general affect (Mayer, et al. 2000a; Newsome, et al., 2000). A recent study by Newsome et al. (2000) reports correlations between the personality facets of the 16PF (Cattell et al., 1993) and the five EQ-i composite scale scores ranging from r = .1 to r = .77. The highest correlation in this study was between the total EQ scale score of the EQ- i and the Anxiety facet of the 16PF (r = -.77) leading these researchers to conclude that the EQ-i is largely a measure of neuroticism (Newsome et al. 2000; p. 1014). Dawda and Hart (2000)
59
have also demonstrated considerable overlap between the EQ-i and personality. These authors found the total EQ scale score of the EQ-i correlated with the NEO FFI (Costa & McCrae, 1992b); for males Neuroticism (N) (r = -.62), Extraversion (E) (r = .52), Openness (O) (r = -.12), Agreeableness (A) (r = .43), and Conscientiousness (C) (r = .51); and for females N (r = -.72), E (r = .56), O (r = -.17), A (r = .43), and C (r = .33). Given this overlap with personality, it is possible that the EQ- i may be predicting theoretically related life criteria in preliminary validity studies (e.g., life satisfaction, workplace performance, psychological well-being) because the EQ-i is measuring personality traits and dispositions known to account for these important human values. As discussed in Chapter 2, exploratory factor analyses that have examined the EQ-i concurrently with the Eysenck Personality Profiler (EPP; Eysenck, Barrett, Wilson & Jackson, 1992) and the NEO PI-R (Costa & McCrae, 1992b), have fo und the subscales of the EQ- i define a distinct trait EI factor amongst the major dimensions of personality (Petrides & Furnham, 2001). Petrides and Furnham (2001) have argued that these results provide evidence for the discriminant validity of the EQ- i from normal personality. However, further discriminant findings are needed to substantiate this claim. The study by Petrides and Furnham (2001) did not examine whether the distinct trait EI factor comprising the sub-scales of the EQ-i accounted for additional variance in theoretically related criteria such as life satisfaction, job performance and psychological well-being over and above normal personality. As such, it could be argued that whether the EQ-i accounts for additional variance in theoretically related criteria over and above well-established personality traits and dispositions has not yet been empirically substantiated and needs to be addressed by future research.
60
4.1.4
Factor structure of the Bar-On EQ-i
While further validation of the discriminant and predictive validity of the EQ-i is needed, it could be argued that further validation of its factor structure is also required. To-date, Bar-On, (1997a) and Petrides and Furnham (2001) are the only researchers to have examined the factor structure of the EQ-i. Empirical support is claimed by Bar-On (1997a), with a large representative population sample (n = 3, 831) through both exploratory and confirmatory factor analyses, that the hierarchical factor structure of Bar-Ons model of EI is measured by the 15 component sub-scales of the EQ-i. However, there are some anomalies in the interpretation of the results and factor solutions chosen to represent the data that need to be addressed. Bar-On (1997a), performed a principal components factor analysis of the EQ-i and reported using the eigen value-greater-than-one, together with the scree criteria to determine the correct number of factors in the data set. However, neither the number of factors with eigen values greater than one, nor the number of factors suggested by the scree plot were reported. Rather, Bar-On (1997a) reported that there were 13 factors in the data set and that the variance revealed by each of the factors was, in order, 23.1%, 4.8%, 3.6%, 2.8%, 2.6%, 2.0%, 1.7%, 1.5%, 1.3%, 1.3%, 1.2%, 1.1% and 1%. With large sample sizes of 300 or more, factor loadings as small as .15 (indicating that 2.25% of the variance is accounted for by the factor) are typically significant at the p < .01 level (Kline, 1994). Factors that explain less than 2.25% of the variance in a data set are typically considered to be insignificant and not to be accounting for anything meaningful in the data set (Cattell, 1978). As such, although Bar-On (1997a)
interpreted 13 factors in the data set, according to Cattells (1978) criterion, it could be argued that of these only 5 accounted for meaningful variance in the data.
61
Bar-On (1997a) reported that a number of factor solutions were examined for interpretability. Specifically, 12, 13, and 14 factor solutions (although, in a more recent book chapter (Bar-On 2000), this number increases to include a 15 factor solution), and it was reported that a 13- factor varimax rotated solution afforded the most meaningful interpretation theoretically (Bar-On 2000, p. 369). That an orthogonal varimax rotation provided the most meaningful interpretation is not consistent with BarOns (1997a) theoretical structure of EI. As previously mentioned, within Bar-Ons (1997a) model of EI similar components are logically clustered together into five major composite components of the model, representing a homogeneous construct of EI (ie., a general factor of EI). Orthogonal- rotated factor solutions preclude the emergence of a general factor (Kline, 2000). Moreover, many of the key components of the Bar-Ons (1997a) model are conceptually very similar (e.g., Happiness and Optimism, Empathy, Social Responsibility and Interpersonal Relationship) and indeed highly correlated (e.g., Bar-On, Brown, Kirkcaldy & Thome 2000; Dawda & Hart 2000; Newsome et al., 2000). In an orthogonal rotation, factor axes are kept at 90 degrees to each other and are thus uncorrelated. Oblique-rotated factor solutions of psychological variables typically provide a better fit to psychological theory than orthogonal- rotated solutions because psychological variables are typically (although not always) correlated (e.g., personality factors; Cattell & Kline, 1977). Bar-On (1997a) does not report examining oblique-rotated factor solutions. Whether the orthogonal varimax-rotated solution
provided the best fit with the theoretical structure of Bar-Ons (1997a) model is therefore unknown and should be evaluated by future research. Bar-On (1997a) labelled the 13- factors that emerged from the data set by examining the highest loading items on each factor (factor loadings of .4 and higher across and within factors). There were some differences between Bar-Ons (1997a)
62
theoretical structure of EI, and that, which emerged from the exploratory analysis. Firstly, two factors, rather than one single factor, emerged from the items pertaining to the Impulse Control sub-scale of the EQ-i. Secondly, the items pertaining to the SelfRegard, Self- Actualisation, Optimism and Happiness sub-scales loaded on two factors (factor 1 & 10), rather than four single factors. Similarly, the items from the Assertiveness and Independence sub-scales loaded on one factor (factor 6) rather than two single factors. Finally, Bar-On (1997a) reported that the Empathy and Self- Regard sub-scales emerged as separable factors in the data set, however, that they were highly correlated r = .80, a surprising result given that the solution was rotated orthogonally). In summary, Bar-On concluded that the major discrepancy between the theoretical 15factor structure of the EQ- i and the 13-factor structure that emerged from factor analysis was due to the fact that five factors emerged from seven original sub-scales. A series of confirmatory analyses by way of structural equation modelling were performed by Bar-On (1997a), to determine whether sub-scales that loaded on single factors could be treated as separable factors. It was reported that the results of these analyses confirmed that Self- Regard, Self- Actualisation, Optimism and Happiness could be treated as separable factors as could the Assertiveness and Independence subscales. Confirmatory analyses were also applied to assess whether a one-factor, second order model involving a general factor of emotional intelligence and the hypothesised five composite factors fitted the data. Multiple fit statistics supporting this model were reported (i.e., GFI = .971, Adjusted GFI = .892, NFI = .956, and CFI = .982). On this basis, Bar-On (1997a) claimed that there was empirical support that the dimensional structure of the EQ- i was consistent with the hierarchical structure of his model. However, other possible structural equation models of the data were not reported.
63
Given the interrelatedness of the 15-subscales and the dominance of the general factor of EI inherent within the scale, it is likely that a large number of models comprising a higher-order general factor may have fit Bar-Ons (1997a) data. MacCallum, Wegner, Uchino and Fabrigar (1993) have demonstrated that a large number of varying second-order models can be shown to fit correlation matrixes that comprise a general factor that accounts for a large proportion of the variance in a data set. Indeed Petrides and Furnham (2001) have recently found, a number of models to fit the EQ-i using SEM, including a model that comprised a general factor, the five composite factors and the fifteen key components; as well as a model that comprised a general factor and the fifteen key composites of Bar-Ons (1997a) model. Collectively these confirmatory model fit statistics do not necessarily provide substantiative evidence in support of the dimensional structure of the EQ-i. In order to substantiate the dimensional structure of the EQ-i using SEM the differential validity of various comparative models must be tested against external criterion variables (e.g., theoretically related criteria such as psychological well-being, the quality of interpersonal relationships, workplace success etc). Where several models are found to provide acceptable model fit statistics, the model that is more predictive of external criterion variables is typically considered superior and to better represent dimension structure (Byrne, 2001). In summary it could be argued that the dimensional structure of the EQ- i is unclear. While Bar-On (1997a) claims to have found empirical support through both exploratory and confirmatory analyses that the dimensional structure of the EQ-i represents his hierarchical model of emotional and social intelligence, the results of these analyses are not conclusive and have not been comparatively assessed. Furthermore, there are a number of issues concerning the interpretation of the
64
exploratory results that need to be addressed such as the factor solution that best represents the dimensional structure of the EQ- i.
4.1.5
Overview of the current study
Given that the factor structure of the EQ-i is unclear, exploratory factor analyses were performed at the outset in the current study. Principal axis factoring was performed and the joint Parallel Analysis (Horn, 1965)/ Scree Test (Cattell, 1966) criteria were used to determine the factor solution that best represented the data. Confirmatory factor analyses using SEM were then performed to compare the goodness of fit of the exploratory factor solution identified in the current study and the second order dimensional structure of the EQ-i proposed by Bar-On (1997a) comprising Overall EI (or a general factor) and the five composite components (Intrapersonal EI; Interpersonal EI; Adaptability EI; Stress Management EI; General Mood EI). Although the effects are small in magnitude, Bar-On (1997a), reports significant differences in EI according to age and gender. Bar-On (1997a) has shown that individuals aged 40-49 have significantly higher overall EI than individuals aged 20-29 (M = 102.7 and M = 96.8 respectively F = .46. 3, p <.001 ), supporting the claim that EI increases with age. Similarly, although Bar-On reports no gender differences in overall EI, consistently (i.e., over a number of population samples reported in the Technical EQ-i Technical Manual), females tend to have significantly higher Interpersonal EQ than males, although, males tend to have significantly higher Intrapersonal EQ, Adaptability EQ, and Stress Management EQ than females. Finally, the EQ-i reports several studies comparing the EI of clinical groups and matched normative groups showing that those who suffer from mental illness tend to exhibit lower EI than normal populations. 65
Given these findings, whether the factor solution identified in the current study and the second order dimensional structure of the EQ-i proposed by Bar-On differed in terms of their relationships with age, gender and two criterion groups; (1) individuals reporting no history of mental illness; and (2) individuals reporting a history of, and/or current mental illness was examined. It was hypothesised that there would be a positive correlation between Age and EI; that females would score higher on EI than males; and that individuals reporting no history of mental illness would have a significantly higher EI than those reporting a history of, and/or current mental illness.
4.2
Method
4.2.2
Participants
The sample comprised 377 participants (270 females, 103 males, 4 unreported) ranging in age from 15 to 79 with a mean age 39.44 years (SD = 13.83) drawn from the general population via advertisements detailing the research. Participants received a small monetary reward for completing the EQ-i. The ethnic composition of the sample comprised; 71%(260) White Caucasian Australians; 19%(70) White Caucasian Emigrants; 9%(34) Asian/Pacific Islanders; Other/Not Reported 1%(9). In comparison to the Australian population, the sample was slightly above average in education; 2.7%(10) had completed primary school education only; 22%(82) had completed secondary school education only; 20%(73) had completed a tertiary certificate/diploma; 28%(105) had completed an undergraduate degree; 22%(81) had completed a postgraduate degree; (6%(22) not reported). Of the 377 participants 58 reported having a history of mental illness, 10 reported having a current mental illness whilst the majority (288) reported having no known history of mental illness. While there was a
66
gender imbalance in the sample, the age, ethnic, and educational composition was relatively diverse.
4.2.3
Materials
The Bar-On Emotional Quotient Inventory
The EQ- i is a self- report inventory comprising 133 (items) declarative statements phrased in the first-person singular. Participants are required to indicate the degree to which each statement is true of the way they typically think, feel or act on a 5point scale (1 = Very Seldom or Not true of me, 5 = Very often true of me or True of me). The items of the EQ-i are summed to yield scores on 15 lower-order sub-scales, 5 higher-order composite scores, and an overall emotional intelligence score. Of the 133 items that comprise the EQ-i, 8 items comprise a Positive Impression Scale, and 7 items comprise a Negative Impression Scale. These two scales are designed to determine whether a participant is responding in an overly positive or overly negative fashion. In addition, there is an Inconsistency Index, calculated by summing the differences in scores between responses on ten pairs of similar items designed to assess random responding. Finally, item 133 I responded openly and honestly to the above sentences, also assesses random responding. If a participants response to this item is either 2 or 1 (Seldom true of me or Very seldom or Not true of me), their results are considered invalid. Participants responses are also consid ered invalid if a certain percentage of items are not answered (Omission Rate). If the Omission Rate is higher than 6% for a given participant, their EQ-i results are considered invalid (Bar-On, 1997a).
67
4.2.4
Procedure
Participants were administered the EQ- i according to the instructions outlined in the EQ- i Technical Manual (Bar-On, 1997a).
4.3
Results
Prior to conducting the factor analyses, the validity of participants responses were examined. Four of the participants responses were found to be invalid as per the Omission Rate criterion, and 15 participants responses were found to be invalid as per the inconsistency index criterion. These participants responses were removed from subsequent analyses. In order to facilitate comparisons among population samples the test publisher (MHS) converts raw scores to standard scores such that each scale score has a mean of 100 and a standard deviation of 15. Accordingly standardised means and standard deviations of the present sample are provided in Table 4.2.
Table 4.2 Means and Standard Deviations for the Bar-On EQ-i Scales
Intrapersonal Interpersonal Adaptability Mean SD 94.80 16.17 96.48 15.54 94.46 15.33
Stress Management 94.56 15.58
General Mood 93.92 17.04
Total EQ 93.70 15.93
As shown in Table 4.2, the present sample scored slightly lower than the North American normative sample (M = 100; SD = 15) for the EQ Total score as well as the EQ composite scales however, the difference in average performance for all composite scales fall within one standard deviation of the mean and are therefore considered to be within the normal range according to the EQ- i Technical Manual (Bar-On, 1997a). 68
4.3.2
Exploratory factor analysis Principal axis factoring was applied to the items pertaining to the 15 sub-scales
of the EQ-i (validity items omitted). The scree test suggested that six factors should be extracted from the data set, a result confirmed by a parallel analysis using the procedure provided by OConnor (2000). In total, the six factors accounted for 40.3% of the variance (22.8%, 5.0%, 4.3%, 3.1%, 2.8%, and 2.3%, respectively) in the data set. In the un-rotated solution most items loaded on a single factor providing evidence for a general factor of EI, a finding consistent with the interrelatedness of the EQ-i sub-scales reported by Bar-On (1997a). According to the scree test and parallel analysis, six factor (orthogonal- and oblique-rotated) solutions were examined in order to further interpret the factor structure of the EQ-i. Examination of the factor solutions revealed that the six-factor oblique (Direct Oblimin with Kaiser Normalisation) rotated solution best represented the present data. Both the orthogonal and oblique factor solutions produced highly similar results, however, the oblique-rotated solution was clearer, involving a more even spread of items loading on each factor. The resulting factor loadings are presented in Table 4.3.
69
Table 4.3 Factor loadings pertaining to the Bar-On EQ-i (direct-oblimin)* Item #
Q56 Q91 Q2 Q70 Q47 Q85 Q77 Q129 Q40 Q51 Q31 Q64 Q122 Q114 Q54 Q26 Q24 Q100 Q110 Q106 Q127 Q11 Q53 Q95 Q132 Q38 Q17 Q125 Q36 Q99 Q98 Q90 Q124 Q84 Q44 Q72 Q16 Q105 Q113 Q39 Q55 Q120 Q62 Q69 Q61 Q82 Q119 Q117 Q73 Q42 Q102 Q130 Q86 Q13 Q97 Q66 Q83 Q27 Q58 Q107 Q60 Q80 Q45 Q15 Q89 Q29 Q81 Q1
Factor 1
.725 .712 .678 .676 .622 .599 .592 .563 .561 .542 .516 .515 .509 .491 .478 .476 .465 .460 .455 .448 .393 .390 .366 .356 .352 .345 .312 .303 .301 .195 -.149 .151 -.200 .347 .246 .311 .352 .323 .180 -.123 .167 .138
Factor 2
.167 -.153 .169 -.147 -.162
Factor 3
.175
Factor 4
-.125
Factor 5
.157 .131 .119
Factor 6
-.206
.105 .273 .159 .191 -.104 .112 .190 .101 .115 .259 .227 .164 -.153 .319 .133 -.164 .210 -.241 .258 .190 .182 .226 .160 .348 -.236 .277 .101 .121 .226 .173 .203 .211 .115 .229 .215 .125 .210 .276 .173 .148 .181 .118 -.172 .165 .265 .152 .251 .320 .217 .223 .175 .214 .230 -.124 .217 .196 .118 -.148 .283 .180 .141 .238 .623 .618 .617 .609 .581 .561 .477 .465 .115 -.125 .123 .167 .136 -.118 .122 .147 .149 .351 .270 -.112 -.101 .268 -.318 .174 .114 .244
-.393 .119 .183 -.301 -.196 .187 -.138 -.388 .183 -.300 .106 -.294 .112 -.153 -.650 -.531 -.501 -.483 -.478 -.449 -.440 -.421 -.417 -.407 -.390 -.389 -.379 -.357 -.336 -.333 .317 -.309 -.193 .116 -.282 -.313 .182 -.102 .250 .121 .182 -.145 -.110 -.128 -.154
.252 .133 .109
.122 .232 .239
.163 -.124 .162 .121 -.178 -.155 -.134
.200 -.102 .228 .141 .143 .229 .235 .144 .175 -.187 .158
.178 -.312 -.134 .119 .647 .638 .606 .605 .586 .565 .548 .413 -.363 .350 .335 .314 .281 .111 .160 .132 .162 -.129
.163 .122 .221 .315 .303 .129 .145
.194 -.112 .281
-.126
.322 .104
-.242 .108
70
Q20 Q6 Q108 Q78 Q48 Q67 Q8 Q4 Q32 Q88 Q75 Q37 Q112 Q7 Q23 Q52 Q116 Q9 Q10 Q63 Q22 Q35 Q68 Q21 Q18 Q96 Q111 Q126 Q30 Q46 Q128 Q131 Q92 Q104 Q93 Q103 Q19 Q87 Q59 Q76 Q28 Q33 Q49 Q14 Q3 Q43 Q74 Q118 Q121
.153 .201 .132 .254 -.167 -.101 .238 .312 .135 -.130 .244 .266 -.181 -.195 .196 .133 .296 .138 -.225 .190 -.128
.199 .151
.419 .411 .396 .394 .384 .379 .375 .375 .362 .335 .313 .309 .212 -.107 -.131
.262 -.174 .189 .204 .201 .263 .151 .171 .287 .202 .772 .757 .732 .609 .569 .498 .488 .487 .439 .406 .384 .383 .372 .370 .368 .320 .301 .295 -.105 .224 .125 .266 .137 .149 .249 .349 .273 .250 -.117 .210 .324 .190 .114
-.105 .129 .201 .201 .287 .242 -.139 .137 .218 .288 -.276 .228 .102 .151 .194 -.106 .191 .245 .134 .182 .280 .253 .194 -.140 .177 .169
-.224 -.194 -.183 .111 -.179 .261 -.215 -.108 -.155 .194 -.136 .135 .208 .255 -.189 .267 .157 .194 .110 -.218 -.134 .150 -.127 .242 .140 .188 .282 .116 .318 -.259 .296 .198
.104 -.212 .238 .108 .106 .146 .215 .111 .219 .106 .167 .382 .172 .125 .214 .108 .249 .245 .195 .249 .200 .186 .203
.160 -.112 .162 .166
.102 .214 .233 .230 -.123 -.132 .571 .508 -.487 .437 .426 .409 .407 .403 -.379 .369 .357 .339 .336 .287 .286 .285 .266 .243
.135 .148
.115 .181 .145
*Note. Factor loadings have been sorted ascending. Items loading >|.4| are in bold face, and those items loading <|.1| have been omitted. The process employed by Bar-On (1997a) was used to identify and label the factors shown in Table 4.3, that is, items loading .4 or higher. Consistent with Bar-Ons (1997a) exploratory factor analyses of the EQ- i, the first factor that emerged in the analysis accounted for more than half of the variance (22.8%) in the data set, and had items loading .4 and higher primarily from the Self-Regard and Happiness sub-scales with some items from the Optimism, Self- Actualisation and Stress Tolerance subscales. Bar-On (1997a p.50,51) describes people who score high on the Self- Regard 71
sub-scale as those who have a good sense of self-esteem, feel positive about themselves, and know who they are. Similarly, people who score high on the Happiness sub-scale of the EQ- i are described by Bar-On as those who have a happy disposition and are pleasant to be with. Although Bar-On labelled this factor SelfContentment because it appears to measure emotionality in general, this factor was labelled Emotional Disposition. Items loading .4 or higher on the second factor that emerged in the present analysis were from the Interpersonal Relationship, Social Responsibility and Empathy sub-scales of the EQ- i. According to Bar-On (1997a) these three sub-scales all measure aspects of interpersonal skills and functioning thus this factor was labelled Interpersonal EQ. The third factor that emerged was almost an exact replication of the Impulse Control sub-scale of the EQ-i with one item loading .4 and higher from the Reality Testing sub-scale, as such this factor was labelled Impulse Control. Items loading .4 and higher on the fourth factor were predominantly from the Problem Solving sub-scale of the EQ-i, with one item from each of the Self Actualisation, Stress Tolerance and Optimism sub-scales, thus this factor was labelled Problem Solving. The fifth factor that emerged in the present analyses had items loading .4 and higher primarily from the Emotional Self- Awareness sub-scale of the EQ-i with one item from each of the Interpersonal Relationship, Assertiveness and Reality Testing sub-scales of the EQ-i, thus this factor was labelled Emotional Self- Awareness. Items loading on the sixth factor were primarily from the Flexibility and Independence sub-scales of the EQ-i with one item each from the Stress Tolerance and Self- Regard sub-scales. Flexibility is purported to assess how flexible individuals are in their thoughts and actions, while independence is purported to assess how self-reliant individuals are, and how independent they are in their thinking and actions. As such, this factor was labelled this
72
Character (i.e., flexible/inflexible, independent/dependent). Intercorrelations between these six factors ranged from .01 to .37. In summary, the results of the exploratory factor analysis failed to support the 13-factor structure of the EQ-i previously reported by Bar-On (1997a), and his theoretical 15- factor model of emotional and social intelligence. Rather, the results of the current study suggest that EQ-i comprises a general factor of EI and six primary factors.
4.3.3
A series of confirmatory factor analyses were conducted to examine whether the six-factor model determined by the current study better represented the dimensional structure of the EQ- i in comparison to the second-order five- factor model proposed by Bar-On (1997a) using structural equation modelling (AMOS 4.01; Arbuckle, 1999). Bar-Ons five-factor model, as illustrated in Figure 4.1, was found to provide a reasonably good fit with the present data set (CFI = .97, RFI = .95, RMSEA = .15). Similarly, the six- factor model derived from the exploratory factor analysis, as illustrated in Figure 4.2, also had a good fit (CFI = .98, RFI = .96, RMSEA = .12).
73
emotional self-awareness
.67
assertiveness
.77 .82 .81 .65

intrapersonal
self-regard
selfactualization
independence
empathy
.87 .80 .74

interpersonal
.95
social responsibility interpersonal relationship
.63
reality testing
.75 .66 .70

adaptability
flexibility
.98
general EI
problem solving
.92
stress tolerance
.87 .42
stress management
.99
impulse control
optimism
.88 .76
general mood
happiness
Figure 4.1 Bar-Ons (1997a) second order five- factor model of EI
74
selfregard
0.82
happiness
0.77 0.87
emotional disposition
optimism
0.79
selfactualization stress tolerance Interpersonal relationship social responsibility
0.75
0.74 0.79 0.87

interpersonal
0.97
empathy
0.63 0.7 0.59

impulse control
impulse
0.53 0.95
q97
problem solving
0.71 0.82
problem solving
0.74
q81, q20, q108
0.84
emotional self-awarness
0.86 0.91
emotional self-awareness
q23, q22, q68
flexibility
0.73 0.74 0.57

character
independence
q93
Figure 4.2 Exploratory six- factor model of the Bar-On EQ-I determined by the current study.
75
To statistically test which model provided a better overall fit of the data the difference in the chi-square values were determined (Breckler, 1990). The five-factor model yielded a chi-square value of 742.92 with 85 degrees of freedom. In contrast, the six-factor model from the exploratory factor analysis yielded a chi-square value of 664.76 with 113 degrees of freedom. As such, the six- factor model determined by the current study provided a significantly better fit than Bar-Ons (1997a) original fivefactor model (chi-square difference was 78.16, df 28, p< .05).
4.3.4
Demographic differences
Finally whether the six- factor model determined by the current study differed from the second-order five-factor model determined by Bar-On (1997a) in terms of the relationships with Age, Gender, and History of Mental Illness was examined. To determine if EI correlated with Age, Pearson correlations were performed on the five factors proposed by Bar-On (1997a) and the six factors found by the current study. Correlations were also computed with general EI (gei), which are component scores derived from the first un-rotated component of the PCA analysis performed on the five factors of EI proposed by Bar-On. Table 4.4 presents the component loadings on this general factor.
Table 4.4 Component Loadings Derived From a PCA of the Five Higher Order Facets Proposed by Bar-On (1997a). intrapersonal interpersonal Adaptability stress management general mood Loading .90 .76 .89 .79 .89
76
As shown in Table 4.4, all of the component loadings are positive. The first component accounted for 72.0% of the total variance. 1 Table 4.5 presents differences in EI as a function of Age, Gender, and History of Mental Illness.
Table 4.5 Differences in emotional intelligence as a function of Age, Sex, Mental Illness. Five Factor Model Six Factor Model Intra Inter Adapt Stress Mood Emo Inter Imp Prob Emo Disp Self .09 .11 .16 .20 .10 .13 .12 .22 .15 .02 -.22 -.86 -.13 .04 -.16 -.08 -.87 -.04 -.09 -.50
Char g .04 .04 .13
Age (r) Sex (d) MI (d)
.33 -.42 -.35 -.44 -.47 -.63 -.68 -.34 -.04 -.46 -.28 -.22 .50 Significant correlations (r) are in bold as are the Effect sizes (d), p < .05; Intra = Intrapersonal, Inter = Interpersonal; Adapt = Adaptability; Stress = Stress Management; Mood = General Mood; Emo Disp = Emotional Disposition; Imp = Impulse Control; Prob = Problem Solving; Emo self = Emotional SelfAwareness; Char = Character; g = General Emotional Intelligence;
A positive correlation between Age and EI was expected. As shown in Table 4.5 the correlations were positive, although, small in magnitude. To test the hypothesis that females would have a significantly higher in EI than males a series of independent ttests were performed. As can be seen in Table 4.5, the effect sizes (d) were in the hypothesized direction. In contrast to gender differences reported by Bar-On (1997a), females were found to score on overall or general EI than males. Consistent with gender differences reported by Bar-On (1997a) however, females were found to score higher on Interpersonal EQ and Emotional Self- Awareness. Fina lly, differences in EI as a
function of Mental Illness were pervasive across all factors in both models with the exception of Impulse Control as shown in Table 4.5.
The correlation between g component scores derived from a PCA on the five facets and a PCA on all of the items was .99, p < .001.
77
To determine if the six factor model determined by the current study could classify Gender (male =1, female = 2) and History of Mental Illness (no = 1, yes = 2) better than the five factor model determined by Bar-On (1997a), a series of discriminant function analyses were performed. The first analysis included Interpersonal EQ from Bar-Ons (1997a) five- factor model, which was able to classify 68.6% of the participants sex accurately (Wilks Lambda = .87, p < .001, Canonical r = .36). Similarly, the second analysis included Interpersonal and Emotional Self- Awareness from the six- factor model, which was able to classify 67.6% of the participants sex accurately (Wilks Lambda = .87, p < .001, Canonical r = .37). Regarding Mental Illness (MI), the first analysis included all five of the components of Bar-Ons (1997a) five factor model, which was able to classify 67.7% of the participants mental history/status accurately (Wilks Lambda = .94, p < .001, Canonical r = .25). Similarly, the second analysis included five of the six factors in the six- factor model (Impulse Control was excluded), which was able to classify 68.0% of the participants mental history/status accurately (Wilks Lambda = .92, p < .001, Canonical r = .29). In conclusion, these results suggest that neither of the two models were superior to the other with respect to predicting ge nder and history of mental illness.
4.4
Discussion
The results of the current study failed to support previous claims by Bar-On (1997a) that the dimensional structure of the EQ-i comprises 13 factors that closely match the theoretical decomposition of his model of emotional and social intelligence. In contrast, the current study found that the dimensional structure of the EQ-i comprised a general factor of EI and six primary factors.
78
It could be argued that the findings of the current study offer a more conc lusive interpretation of the EQ-is dimensional structure. The current study determined the number of factors inherent in the EQ-i using joint scree test/ parallel analysis criteria which has been shown to be the most accurate method for determining the correct number of factors in a correlation matrix (Fava, Velicer, & Eaton, 2000; Zwick & Velicer, 1986). In addition, Bar-On (1997a) used an orthogonal (varimax) rotation procedure to interpret the 13-factor solution found in his analyses of the EQ-i. When correlations of r = .3 or higher exist between variables it has been argued that obliquerotated solutions offer a clearer interpretation than orthogonal rotated solutions from an empirical standpoint, providing superior simple structure, superior factor replicability, and correlations between the factors which provide insightful information about the construct being analysed (Reise, Waller & Comrey, 2000). In the current study an oblique rotation procedure was used to interpret the dimensional structure of the EQ-i. Indeed three of the correlations between the six factors identified in the current study were greater than r = .3 suggesting that the oblique-rotated solution offered the most meaningful interpretation of the current data. The first factor that emerged in the exploratory analyses of the current study, Emotional Disposition, had items loading primarily from the Self- Regard and Happiness sub-scales of the EQ- i with some items from the Optimism, SelfActualisation and Stress Tolerance sub-scales. The second factor that emerged, Interpersonal EQ, had items loading from the Interpersonal Relationship, Social Responsibility and Empathy sub-scales of the EQ-i. The Impulse Control, Problem Solving and Emotion Self-Awareness sub-scales of the EQ-i emerged relatively clearly as separable dimensions (factors 3, 4 and 5), whilst the sixth factor that emerged in the current study loaded items from the Flexibility and Independence sub-scales of the EQ-
79
i. This six- factor model is considerably different from Bar-Ons (1997a) second-order five- factor model of EI that describes how the 15 sub-scales of the EQ-i cluster to provide broader definitive dimensions (Intrapersonal EQ; Interpersonal EQ;
Adaptability EQ; Stress Management EQ; and General Mood EQ). The only subscales of the EQ- i that were found in the current study to cluster together according to BarOns (1997a) five- factor model were Interpersonal Relationship, Social Responsibility and Empathy that form the broader Interpersonal EQ dimension. None of the other subscales of the EQ-i were found to cluster according to Bar-Ons second order five-factor model. Confirmatory factor analyses using structural equation modelling were conducted to compare the goodness of fit of the six- factor model determined in the exploratory analyses of the current study, and Bar-Ons (1997a) second order five-factor model of EI. Both models were found to have adequate fit, although a statistical comparison of the two models demonstrated the six- factor model determined by the current study was superior. It should be emphasized, however, that a large number of models that would have fit the correlation matrix, as has been demonstrated to be a common characteristic of SEM (MacCallum, Wegener, Uchino, & Fabrigar, 1993). In this case, the dominance of the general factor would likely have supported a large number of models with a higher-order general factor. The strength of the general factor can be appreciated by the PCA that found all five facets to load positively on one factor that accounted for 72% of the variance. The problem of equivalent models (MacCallum et al., 1993) is likely not solvable using comparative fit model fit statistics. Rather, the differential validity of any model must be tested using external criterion variables. A series of discriminant function analyses did not find either Bar-Ons (1997a) second order five- factor model or the six- factor model determined by the current study
80
to be more predictive of gender or mental illness. However, there were differences in the pattern of effects. For example, in the current study Bar-Ons (1997a) second order five- factor model only had the Interpersonal EQ dimension shown a significant difference between males and females. In contrast, the six- factor model identified by the current study had both the Interpersonal EQ and Emotional Self- Awareness dimensions showing significant differences between males and females. The items comprising Emotional Self- Awareness were in the Intrapersonal facet of the five-factor model; however, the effect was reduced to non-significance, because of the other items grouped into Intrapersonal (e.g., Assertiveness, and Self- Actualisation). Thus, although neither model offered more predictive validity, it could be argued that the six- factor model offered a more insightful interpretation when analysing gender differences. It should be noted that the six- factor solution of the EQ- i identified by the current study needs to be replicated in order to substantiate whether this solution offers a more accurate description of the dimensional structure of the EQ- i. Indeed, there is a considerable portion of error variance with the six- factor solution identified by the current study. However, the findings of the current study do suggest that researchers should factor analyse the scale and use total scores and factor scores given that the items of the EQ-i may not comprise either the 15 or 5 composite components of Bar-Ons (1997a) model of EI. Further factorial analyses of the EQ-i are needed in order to confirm its factor structure. Given the meaningful pattern of convergent validity with the EQ-i and measures of psychological well-being however, (as outlined in the introduction of this chapter), the EQ- i does appear to provide a general index of individual differences in psychological health, and perhaps therefore, individuals capacity to succeed in coping with environmental demands and pressures.
81
At a broader level the findings of the current study highlight the need for the psychometric properties of existing measures of EI to be substantiated by independent research. The dimensional structure of tests is one of the most rudimentary of these psychometric elements (Kline, 2000), and there is little research to-date that has examined the dimensional structure of existing measures of EI.
82
Chapter 5; The TMMS
CHAPTER 5
The factorial and construct validity the Trait Meta-Mood Scale
5.1
Overview
Chapter 5 examines the factorial and construct validity of the Trait Meta Mood Scale (TMMS; Salovey et al., 1995) using exploratory factor analysis and Structural Equation Modelling. In a general population sample (n =310), the three- factors (Attention, Clarity and Repair) identified by Salovey et al. (1995) were replicated in exploratory analyses, although support for a fourth factor was also found. Support was also found (applying SEM) for the construct validity of the emotional management competency as a three-phase process assessed by the TMMS, as originally identified by Martinez-Pons (1997). Overall, the findings support the Attention, Clarity and Repair decomposition of the TMMS. Implications and directions for further validation of the TMMS are discussed.
5.1.2
The Trait Meta-Mood Scale
The TMMS has been developed from research examining the reflective processes that accompany mood states (e.g., Mayer & Gaschke, 1988; Mayer, Salovey, Gomberg-Kaufman & Blainey, 1991; Mayer & Stevens, 1994). This research has identified an ongoing process associated with moods whereby individuals continually reflect upon their feelings, monitoring, evaluating, and regulating them termed the meta- mood experience (Salovey et al, 1995 p.127). Drawing from this work, the TMMS is comprised of 30 items that define three sub-scales based on factor analysis by Salovey et al. (1995). These are labelled: Attention to Feelings (e.g., I pay 83
Chapter 5; The TMMS
a lot of attention to my feelings); Clarity of Feelings (e.g., I am usually very clear about my feelings); and Mood Repair (e.g., Although I am sometimes sad, I have a mostly optimistic outlook). As a measure of EI, the TMMS assesses one of the higher order, more complex abilities of EI proposed by Mayer and Salovey (1997), concerned with the regulation and management of emotions. However, the TMMS measures the perceived ability to effectively regulate and manage emotions as opposed to actual capacity or mental ability (Salovey et al., 2000). Preliminary psychometric analysis of the TMMS by Salovey et al (1995) suggests that the TMMS may provide a reliable and valid self-report index of the ability to manage emotions. Salovey et al. (1995), report that each of the three sub-scales of the TMMS measure coherent and internally consistent variables (coefficient a = .86, .88, .82 for Attention to Feelings, Clarity of Feelings and Mood Repair respectively), and report evidence of both convergent and discriminant validity. Attention to Feelings has been shown to correlate positively with the tendency to attend to aspects of ongoing consciousness as measured by the Private Self-Consciousness and Public SelfConsciousness sub-scales of the Self-Consciousness Scale (SCS; Scheier & Carver, 1985), r = .42, r = .36 p<.01 respectively). Clarity of Feelings has been shown to correlate negatively with ambivalence over emotional expression and with normal depression, (r = .-25, r = .-.27, p<.05 respectively). Finally, Mood Repair has been shown to negatively correlate with normal depression, and to positively correlate with optimism and beliefs about the changeability of negative moods as measured by the CES-D r = -.37, p<.01 (Radloff, 1997), the Life Orientation Test r = .57, p<.01 (LOT; Scheier & Carver, 1985), and the Expectancies for Negative Mood Regulation r = .53, p<.01 (NMR; Catanzaro & Mearns, 1990). Collectively these psychometric properties provide preliminary evidence for the reliability and validity of the TMMS as a measure
84
Chapter 5; The TMMS
of the capacity to monitor and manage emotions, however, further independent psychometric analyses of the scale are needed. To-date, there have been several independent psychometric evaluations of the TMMS confirming the relationships between scores on the scale and ambivalence over emotional expression and normal depression (Lee, & Lee, 1998; Martinez-Pons, 1997). Independent studies have also found scores on the test to be positively related to measures of life satisfaction (Martinez-Pons, 1997; Palmer, Donaldson & Stough, 2002). However, Salovey et al. (1995) are the only researchers to have examined the factor structure of the TMMS. Martinez-Pons (1997) proposed a functional sequence of emotional management. Specifically, that clarity of feelings (Clarity) would not be possible without attention to feelings (Attention); and that the capacity to repair negative moods and emotions (Repair) would not be possible unless emotions were experienced clearly (Clarity). Using path analytic methodology, Martinez-Pons (1997) found evidence for this proposed functional sequence of the construct assessed by the TMMS. However, the three factors Attention, Clarity and Repair and the functional sequence proposed by Martinez-Pons (1997) have not been further examined. Further replication is necessary in order to establish the factorial validity of the TMMS.
5.1.3
To assess the factorial validity of the TMMS an exploratory principal components factor analysis was performed with a sample taken from the general population. In addition, structural equation modelling was used to examine the
functional sequence of the emotional management construct assessed by the TMMS as proposed by Martinez-Pons (1997). It was expected that the three- factor model and 85
Chapter 5; The TMMS
functional sequence would be confirmed by these analyses. It was argued that the resolution of the three- factor structure and functional sequence would provide additional support for the factorial validity of the TMMS and the construct validity of the emotional management ability of EI it has been purported to assess.
5.2 Method
5.2.2
Participants The sample comprised 310 participants, (221 females, 86 males, 3 unreported)
ranging in age from 15 to 79 years with a mean age of 39.4 years (SD = 13.8). The ethnic composition of the sample was relatively representative of the Australian population with the majority being White/Caucasian Australians (72.5%),
White/Caucasian immigrants (17%), and Asian or of Pacific Islander decent (9.5%). In comparison to the Australian population, the sample was slightly above average in education; 2.1% had completed primary school education only, 22.9% had completed secondary school education only, 20.5% had completed a tertiary certificate/diploma, 30.8% had completed an undergraduate degree, and 21.2% had completed a postgraduate degree. Participants were recruited f r the study via advertisements in o local newspapers and received a small stipend for participating.
86
Chapter 5; The TMMS
5.2.3
Materials The current study used the 30-item version of the TMMS as recommended by
Salovey et al. (1995). Participants responded on a five-point scale where 1 = strongly disagree and 5 = strongly agree. Total and sub-scale scores were derived by summing the respective item responses.
5.2.4
Procedure
Participants self-administered the questionnaire in their own time and were debriefed afterwards where possible.
5.3 Results
The means and standard deviations (in brackets) for the Attention, Clarity and Repairs sub scales were 51 (7.9), 42.5 (7.9), 23.2 (4.3) respectively. The means and standard deviations of the present sample were very similar to a sample of U.S. Air Force recruits, but significantly higher than a sample of 1st year psychology students from Australia on the Clarity, M = 35.21 SD(8.65) t(1,408) = 7.87, p<.05, d = .90, and Repair M = 20.53 SD(5.0) t(1,408) = 5.18, p<.05 d = .60 sub scales (Davies, Stankov & Roberts 1998). The means and standard deviations of the present sample provide some of the first Australian general population norms for the TMMS and indicate that there may be differences between different samples.
87
Chapter 5; The TMMS
5.3.2
Exploratory factor analysis
To determine the correct number of factors to extract from the data set, a parallel analysis (Principal Components Analysis) was performed using the procedure provided by OConnor (2000). The parallel analysis suggested that four factors should be extracted as illustrated in Figure 5.1.
8 7 6
Eigenvalue estimate
5 4
3 2
1 0 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00
PCA Parallel analysis
Eingenvalue number
Figure 5.1 Scree plot of eigen values derived from the parallel and principal component analysis.
88
Chapter 5; The TMMS
These four factors accounted for 47.87% of the variance in the data set with factors 1 through 4 accounting for 23.54%, 11.71%, 6.64%, and 5.98% of the variance respectively. To further assess the dimensional structure of the TMMS these four factors were rotated according to an oblique rotation (Direct Oblimin with Kaiser Normalisation). The resulting factor (pattern matrix) loadings are shown in Table 5.1.
89
Chapter 5; The TMMS
Table 5.1 Factor (pattern matrix) loadings of principal components analysis with oblimin rotation for the TMMS Item 30 25 16 5 28 6 15 22 20 11 29 Factor 1 .848 .820 .752 .736 .699 .682 .576 .574 .434 .338 Factor 2 Factor 3 Factor 4 -.111 .197 .119 -.122 .172 .335 .246 .720
-.117 .219 .227 .284 .136
17 .705 23 .184 .655 -.123 -.106 24 .645 -.425 21 .174 .629 -.327 18 .581 .116 2 .560 -.183 .206 7 .556 .254 -.250 27 .553 .273 4 .551 10 .544 .209 -.150 3 .457 .242 12 .296 .183 -.229 26 -.135 .798 1 .741 13 -.125 .707 8 .174 .534 .369 14 .212 .666 9 .170 .188 .580 19 .197 .378 .516 Note. Factor loadings have been sorted ascending and those >|.3| are in bold text.
The first three factors closely replicated the Clarity, Attention and Repair factors of the TMMS as reported by Salovey et al. (1995). The first factor corresponded to the Clarity of Feelings sub-scale, the highest loading item being I almost always know exactly how I am feeling. The second factor replicated the Attention to Feelings subscale the highest loading item being It is usually a waste of time to think about your 90
Chapter 5; The TMMS
emotions. The third factor corresponded to the Mood Repair sub-scale with the highest loading item being No matter how badly I feel, I try to think about pleasant things. The results of the present study also suggested the existence of a possible fourth factor within the TMMS. Six items loaded significantly on this fourth factor, however, only two of these items (14, and 9) loaded significantly on the fourth factor alone. The internal consistency estimates (Cronbachs alpha) f r these four factors were largely o respectable: Clarity (a = .87), Attention (a = .84) and Mood Repair (a = .71), with the exception of the fourth factor (a = .62) not achieving the criterion of = .70 (Nunnally, 1967), mostly due to the small number of items (two) defining this factor.
5.3.3
Given the discrepancy between the present findings and previous dimensional analyses of the TMMS (i.e., four- verses-three primary factors), the goodness of fit of a range of structural models for the data was assessed using confirmatory factor analysis (AMOS; Arbuckle, 1999). Five models were tested: (1) a General Factor model; (2) an orthogonal three-factor model; (3) an oblique three- factor model; (4) an orthogonal four-factor model; and (5) an oblique four- factor model. The various model fit indices for these five different models are presented in Table 5.2.
Table 5.2 Goodness of fit indices for the five different models of the TMMS assessed General Factor X2 Kline adjust CFI RMSEA 1727.37 4.27 .951 .103 Three Orthogonal Factors 1037.64 2.56 .977 .070 Three Oblique Factors 952.75 2.37 .980 .067 Four Orthogonal Factors 897.01 2.25 .982 .064 Four Oblique Factors 781.92 1.99 .986 .057 91
Chapter 5; The TMMS
As can be seen in Table 5.2, the General Factor model did not satisfy the Kline adjustment critical value of less than 3 (Kline, 1998). Also, the RMSEA value of .103 is larger than the allowable value of .05 (Byrne, 2001). In contrast, all four latter models assessed demonstrated good fit based on the Kline adjustment values. Further, the CFI values for these four models were all comfortably above the recognized critical value of .95 (Byrne, 2001). Finally, the RMSEA values for all four models approach the critical value of .05 (Byrne, 2001). Thus, taken as a whole, the general factor model can be rejected outright, in favour of three and/or four factor models. Further, the models based on oblique (correlated) factors represented the data more accurately (as suggested by their respective fit indices as shown in Table 5.2) than orthogonal (uncorrelated) factors. All of the correlations between the factors were positive and statistically significant as shown in Figure 5.2.
Clarity r = .39** Attention (r = .12*), r = -.05ns * = p <.05; ** = p <.001 r = .50** Repair
Figure 5.2 Structural equation model depicting the mediating effect of Clarity on the relationship between Attention and Repair
92
Chapter 5; The TMMS
The correlation between Attention and Clarity (r = .39) and Clarity and Repair (r = .50) are both larger than the correlation between Attention and Repair (r = .12), suggesting that this smaller relationship may be mediated by the Clarity of feelings subscales as hypothesised by Martinez-Pons (1997). To further test this hypothesis, a structural equation model was tested to determine if Clarity mediated the relationship between Attention and Repair. As shown in Figure 5.2, the hypothesis of mediation was supported, with the correlation of r =.12, p < .05 being reduced to a non-significant value of r = -.05ns, replicating the results of Martinez-Pons (1997). With the path from Attention to Repair omitted, the fit indices differed from the three oblique factors model minimally by only decimal points (?2 = 953.18; Kline Adjustment 2.37; CFI = .98; RMSEA = .067).
5.4 Discussion
This study is the first to examine the factorial validity of the TMMS with a large general population sample, and provides some of the first Australian norms on the test. The findings of the this study support the Attention, Clarity and Repair sub-scale structure of the TMMS as reported by Salovey et al. (1995). Evidence for a fourth factor was found, however, only two items loaded independently on this fourth factor. Confirmatory analyses suggested that both three and four-factor models provided a good fit with the present data. The discrepancy between the three- factor dimensional structure of the TMMS reported by Salovey et al (1995) and the exploratory four-factor model of the present study may be attributable to differences in the samples assessed. Subject sample size, variable sample size, and variations in the gender, social classes, levels of education and other such sampling parameters can all affect factor pattern reproduction (Velicer & Fava, 1998). Indeed, the dimensional analysis of the TMMS 93
Chapter 5; The TMMS
reported by Salovey et al. (1995) was based on a sample size of 200 Americans as apposed to the 310 Australian participants in the present study. Furthermore, the means of the present sample were found to differ significantly from a sample of 1st Psychology students on the Clarity and Repair sub-scales. Previous research with university student samples has shown cross-cultural differences on the TMMS (Ghorbani, Bing, Watson, Davison & Mack, 2002). The findings of the current study and those reported by Ghorbani et al. (2002), highlight the need to establish sub-population and cross cultural norms for the scale. The findings of the current study also provided support for the three-phase functional sequence of the emotional management competency of EI assessed by the TMMS (Martinez-Pons, 1997). In the confirmatory analyses, oblique (correlated) three and four- factor models provided better fit statistics than orthogonal (uncorrelated) factor models. Indeed, the correlations between the Attention, Clarity and Repair factors assessed in the oblique three-factor model were all positive and significant. Furthermore, the relationship between the Attention and Repair factors was found to be mediated by Clarity. This finding provides support for the notion proposed by MartinezPons (1997) that clarity of feelings is not possible without attention to feelings, and that emotional repair is not possible unless one is clear about the moods and emotions they are experiencing. The dimensions assessed by the TMMS, have a strong theoretical basis from research on the reflective processes that accompany mood states, and empirical evidence for Attention, Clarity and Repair as components of emotional monitoring and management is mounting. However, future research needs to replicate the extent to which the TMMS is related to other measures of mood and emotional management, and to determine the extent to which it correlates with measures of personality such as the
94
Chapter 5; The TMMS
NEO PI-R (Costa & McCrae, 1992b) and other emotional dispositions (ie. general affectivity and optimism). Comparative research examining the value of the TMMS and other measures of emotional monitoring and management over and above normal personality and other constructs (stress tolerance, adaptability, charisma) in predicting theoretically related life criteria (e.g., life satisfaction, interpersonal success) would also be of much value. A comparison study examining the differential predictive validity of the TMMS and the performance-based, Emotional Management sub scales of the MSCEIT (Mayer, et al., 2000d), may help delineate whether this aspect of EI is best assessed with self-report or performance-based measures.
95
Chapter 6; The SEI
CHAPTER 6 The emotional intelligence measure by Schutte et al. (SEI; 1998)
6.1
Overview
Research with the self-report measure of emotional intelligence (SEI) by Schutte et al. (1998) is proliferating despite few independent analyses of its fundamental psycho metric properties and a general lack of normative data. The current study provides normative data for the SEI and examines its reliability and factor structure in an Australian general population sample. The relationship between scores on the SEI and age, and differences in scores on the SEI among several criterion groups defined by gender, history of mental illness, parental upbringing and relationship status were also examined. The findings indicate that the SEI is internally reliable and confirmatory factor analyses support a four- factor solution previously identified with student samples. Significant gender differences in scores on the SEI were found, however, no significant differences were found amongst the other criterion groups. The potential utility of generating subscale scores for the SEI, and implications for further research with the scale are discussed.
6.1.2
Introduction
Schutte et al. (1998), have developed a self-report measure of EI (SEI), designed to assess Salovey and Mayers (1990) original conceptualisation of the construct. Consistent with such, it is purported to assess the capacity to perceive and express ones own emotions, the capacity to perceive the emotions of others, the capacity to effectively regulate and manage emotions (in the self and others), and the 96
Chapter 6; The SEI
capacity to utilise or reason with emotions in thought (Schutte et al., 1998). In their initial validation of the SEI, Schutte et al (1998) found that scores on the scale were meaningfully correlated with a number of related constructs including Alexithymia, the Attention, Clarity and Repair subscales of the Trait Meta Mood Scale (Salovey, Mayer, Goldman, Turvey & Palfai, 1995), optimism-pessimism, depression, and impulsivity. Furthermore, scores on the SEI appeared distinct from measures of normal personality and cognitive ability, predicted first-year colleague grades, and were found to be significantly higher amongst a group of psychotherapists than a group of individuals in a substance abuse treatment program and female prisoners. Finally, females were found to score significantly higher than males on the SEI, a finding consistent with other emerging EI scales at the time (e.g., Bar-On, 1997; Mayer & Geher, 1996; Mayer et al., 1999). Since their initial evaluation of the SEI ( chutte et al., 1998), Schutte and S colleagues have presented further evidence in support of the validity of the SEI. Scores on the SEI have been shown to be meaningfully correlated with aspects of adaptive interpersonal functioning (Schutte, et al., 2001), performance on difficult and frustrating cognitive tasks (Schutte, Schuettpelz & Malouff, 2000), positive mood states, and self-esteem (Schutte, Malouff, Simunek, McKenley & Hollander, 2002). Furthermore, independent analyses have found that; (a) scores on the SEI moderated the relationship between stress and mental health variables (Ciarrochi, Dean & Anderson, 2002); and (b) that EI can be reliably measured using the SEI in adolescent samples, and that adolescents scores on the test are meaningfully related with skill at identifying emotional expressions, amount of social support, extent of satisfaction with social support and mood management behaviour (Ciarrochi, Chan & Bajgar, 2001).
97
Chapter 6; The SEI
Collectively this research has provided preliminary evidence for the validity of the SEI, however, further psychometric analyses are needed. Much of this previous research has involved university student samples (e.g., Ciarrochi et al., 2002, Schutte et al., 1998), and population samples that have been considerably small in size (e.g., in the studies by Schutte et al., 2000; 2001; 2002, the sample sizes ranged from a high of n = 77 to a low of n = 24). Furthermore, research examining the relationship between scores on the SEI and important real life criteria (e.g., cognitive performance, interpersonal functioning, and emotional well-being; Schutte et al., 2000, 2001, 2002), has somewhat preceded the establishment of adequate normative data and further analyses of the fundamental psychometric properties of the SEI (e.g., reliability, factorstructure and convergent and divergent validity). Petrides and Furnham (2000), have recently highlighted several weaknesses in the initial factor analytic analyses of the SEI (Schutte et al., 1998), and the content of the scale appears to substantially overlap with measures of optimism and positive affect (Ciarrochi et al., 2002). Although the scale has shown divergent validity from the five major dimensions of personality in preliminary analyses with a relatively small sample (Schutte et al., 1998), others have argued that scores on the test may correlate with theoretically related life criteria because it is essentially measuring established traits and dispositions. For example, optimistic individuals with high general levels of positive affectivity also tend to exhibit more adaptive interpersonal qualities (Berry & Willingham, 1997; Sprecher, 1999; Vittengl & Holt, 2000). Importantly, one study has demonstrated that scores on the SEI moderated the relationship between stress and mental healt h even after controlling for related constructs (Ciarrochi et al., 2002). In summary, the SEI appears to be a promising self- report measure of EI that warrants further research. In particular,
98
Chapter 6; The SEI
further psychometric scrutiny of its reliability, factor structure, and convergent and divergent validity with general population samples is required. As previously mentioned, the factor structure of the SEI has been recently examined by Petrides and Furnham (2000) with a sample of university students (n = 260). In exploratory analyses of the 33- item SEI, a four- factor, orthogonally rotated (varimax) solution was found to provide the clearest interpretation accounting for 40.4% of the variance in the data set. At face value, the items loading on the fourfactors were closely related to the 3 conceptual components of Salovey and Mayers (1990) model of EI. The first factor, labelled optimism/mood regulation, comprised items such as I expect that I will do well on most things I try and I motivate myself by imagining a good outcome to tasks I take on, which may assess self-perceived ability to regulate ones own emotions. The second factor labelled appraisal of emotions comprised items such as By looking at their facial expressions, I recognise the emotions people are experiencing and I easily recognise my emotions as I experience them, which may assess the ability to perceive ones own emotions and the emotions of others. The third factor labelled social skills comprised items such as Other people find it easy to confide in me and I help other people feel better when they are down, which may assess the ability to regulate the emotions of others. Finally, the fourth factor labelled Utilization of emotions comprised items such as When I am in a positive mood, I am able to come up with new ideas and When I feel a change in emotions, I tend to come up with new ideas, which may assess the capacity to utilise or reason with emotions in thought. These factor analytic results have been replicated in another sample of university students (n = 302) by Ciarrochi et al. (2002). Ciarrochi et al (2002) reported that the SEI comprised the same four factors found by Petrides and Furnham (2000),
99
Chapter 6; The SEI
with minor differences in the loading of two items, however, Ciarrochi et al., labelled the factors in a manner more consistent with the conceptual components of Salovey and Mayers (1990) model of EI (i.e., Emotion Perception, Managing Self Emotions, Managing Others Emotions and Utilising Emotions). In both the analyses by Petrides and Furnham (2000), and Ciarrochi et al (2002), only four items loaded clearly on the fourth factor Utilising Emotions. Indeed while the first three factors exhibit acceptable internal consistency reliabilities (ranging from a low of a = .66 for Managing Others Emotions, to a high of a = .80 for Emotion Perception) the fourth factor exhibits low internal reliability (a = .58; Ciarrochi et al) suggesting that more items may need to be conceptualised to assess this component of the scale reliably. Both these factor analyses of the SEI have involved university student samples and need to be replicated with larger and more diverse population samples.
6.1.3
Overview of the present study
The present study examined the reliability and factor structure of the SEI with a sample representative of the Australian population, and examined the relationship between scores on the scale and age, sex, history of mental illness, parental upbringing and relationship status. Confirmatory factor analyses via structural equation modelling was used to determine the extent to which previous factor solutions of the SEI (i.e., the one- factor solution provided by Schutte et al., 1998; and the four- factor solution provided by Petrides and Furnham, 2000), provided a statistically significant fit with the present data. It was expected that a four- factor model comprising Optimism, Appraisal, Social Skills and Utilization of Emotions would provide the best fit with the present data, consistent with previous research examining the factor structure of this
100
Chapter 6; The SEI
test with university student samples (Ciarrochi et al., 2001; 2002). Also consistent with previous research, females were expected to score significantly higher than males on the SEI. In their original conceptualisation of EI, Salovey and Mayer (1990) purported that EI may be based on underlying skills that can be learned and that EI may therefore develop with age and experience. As evidence for this hypothesis scores on the SEI were expected to correlate with age. Salovey and Mayer (1990) also purported that EI might contribute to mental health. As mentioned in the introduction, Schutte et al. (1998) have found that scores on the SEI are meaningfully correlated with depression, whilst others (i.e., Ciarrochi et al., 2002) have demonstrated that scores on the SEI moderate the relationship between stress and mental health. As such, people reporting no known history of mental illness were expected to score significantly higher than those reporting a history of mental illness in the present sample. Previous research with the SEI has also demonstrated that scores on the scale are meaningfully correlated with aspects of adaptive interpersonal functioning. Specifically, those who report greater EI display more cooperative responses toward partners and also report greater marital satisfaction (Schutte et al., 2001). As such the relationship between relationship status an EI was explored. Individuals either married or in a relationship were expected to have significantly higher scores on the SEI than individuals who were not in a relationship. Finally, whether individuals who reported a dual parental upbringing (between the age of 0-18) reported greater EI than individuals who reported a single parent upbringing or who reported that their parents divorced during that time was also examined. Research has demonstrated that children from divorced relationships tend to experience more psychological, social and academic difficulties, and that broken families are characterised by less intimate parent-child relationships (Amato, 1993;
101
Chapter 6; The SEI
Demo & Acock, 1988; Sun, 2001). As such, individuals reporting a dual parental upbringing were expected to score significantly higher on the SEI than those reporting a single parental upbringing or whose parents had divorced during that time.
6.2
Method
6.2.2
Participants
The sample comprised 367 participants (107 males, 257 females, 3 unreported), ranging in age from 15 to 78 with a mean age of 38.3 years ( D = 13.7). The S participants were drawn from the general population across the two most populated Australian states, Victoria and New South Wales via advertisements. The majority of the sample were white Caucasian Australians (69.5%), the remainder were (18.3% White Caucasian Emigrants; 9.3% Asian Emigrants; 1 % other; 2.2% unreported). Of the 367 participants, 66 (18%) reported having a history of mental illness, which was representative of the prevalence rates of mental illness amongst the Australian population (which is 18% according to the Australian Bureau of Statistics, National Survey of Mental Health and Wellbeing of Adults, 1997). Finally, 175 participants (47.7%) reported being single at the time of testing, while 186 (50.7%) reported being either married or in a de- facto relationship (1.6% unreported). Similarly, 18.2% (67) reported that they were either raised by a single parent or that their parents divorced during their upbringing (which was classified as 0-18 years old), while the remainder 77.9% (286), reported that their parents were married during their upbringing (3.8% other/unreported).
102
Chapter 6; The SEI
6.2.3
Materials
Self- Report Emotional Intelligence scale (SEI; Schutte et al., 1998). The Self- Report Emotional Intelligence scale (SEI) by Schutte et al. (1998) comprises 33 items, 3 negatively keyed, (that are self-referencing statements), and asks respondents to indicate on a 5-point scale the extent to which they agree/disagree with each, where 1 = strongly disagree and 5 = strongly agree. Based on initial factor analyses, Schutte et al. (1998) originally intended for the scale to simply provide an overall score obtained by summing the 33 items, the internal reliability of which was reported as a = .90. However, more recent factor analyses have suggested that four primary factors better represent the dimensional structure of the SEI (Petrides & Furnham, 2000; Ciarrochi et al., 2002), with reliability coefficients ranging from a low of a = .58 to a high of a = .80. Research findings with the SEI to-date have provided preliminary evidence of construct, predictive and discriminant validity as outlined in the introduction of this chapter.
6.2.4
Procedure
An SEI test booklet and a scannable answer sheet was constructed from the instructions and items provided by Schutte et al. (1998). Participants who responded to advertisements about the study collected the SEI pencil and paper booklets and scannable answer sheets and were briefed about the purpose of the study. While the instructions in the SEI item booklets were self explanatory, the purpose of the briefing sessions were to emphasise that the SEI was to be completed independently without input from others and in its entirety. Upon completing the questionnaire participants
103
Chapter 6; The SEI
returned the item booklets and answer sheets, and received a small monetary reward for participating.
6.3
Results
6.3.2
Confirmatory factor analyses
Confirmatory factor analyses were performed to statistically compare the extent to which different exploratory factor solutions of the SEI provided a fit with the present data. The one- factor model interpreted by Schutte et al. (1998) was compared against the four- factor model interpreted by Petrides and Furnham (2000). Specifically, four different models were tested via structural equation modelling using AMOS (Arbuckle, 1999). The four models were: a general factor, four oblique factors, four orthogonal factors and a higher order hierarchical model. The hierarchical model consisted of a general factor and the four lower-order factors. Three fit indices were chosen to compare the degree of fit for each model. Specifically; (1) the Kline adjustment (?2 /df) which has a criterion of less than or equal to three; (2), the Comparative Fit Index (CFI; Bentler, 1990) for which values between .90 to .95 are deemed to be a satisfactory fit of the data, whereas values more than .95 are deemed to be a good fit of the data; and (3) the Root Mean Square Error of Approximation (RMSEA) for which values between .08 and .05 are deemed to be a satisfactory fit, whereas values less than .05 are deemed to be a good fit of the data (McDonald & Ringo Ho, 2002). Table 6.1 lists the fit indices obtained for each model.
104
Chapter 6; The SEI
Table 6.1 Model fit statistics for the SEI (Schutte et al., 1999). Fit Statistic General Four Orthogonal Four Oblique Hierachical Factor Factors Factors Model CMIN a 1723.32 1888.17 1312.69 1723.32 1 Kline adj. 3.48 (495) 3.81 (495) 2.68 (489) 3.48 (495) CFIb .97 .97 .98 .97 c RMSEA .082 .088 .068 .082 1: Degrees of Freedom in parentheses. a = Chi-square statistic; b = Comparative fit index; c = Root Mean Square Error of Approximation.
As can be seen in Table 6.1, the best fitting model was the oblique four- factor model, a model most consistent with the underlying theory of the SEI (i.e., Salovey & Mayer, 1990). It should be noted, however, that all four models fit the data in a satisfactory way, particularly according to the CFI and RMSEA statistics. However, the four orthogonal factor model was a worse fitting model than either the general or hierarchical models (which were the next best fitting models) ?2 = 1888.17 1723.32 = 164.85, with 6 degrees (495-489) of freedom, thus, p < .05. The oblique four- factor model, and the intercorrelations amongst the factors are illustrated in Figure 6.1.
105
Chapter 6; The SEI
10 3 23 14 optimism 21 12 28 2 31
0.63
9 15 22 32
0.83
appraisal
5 19 29 25 18
0.68
0.77
11 4 13 30
0.58
social skill
26 6 24 16 1 8
0.73
33
17 utilization 27 7 20
Figure 6.1 Oblique Four-Factor Model of the SEI (Schutte et al., 1998).
106
Chapter 6; The SEI
As shown in Figure 6.1, all of the correlations between the factors are substantial, ranging in size from r = .58 to r = .83. Although not displayed, all of the item loadings were positive, ranging in size from .45 to .73 with a mean of .57 and a standard deviation of .09. In summary the results of confirmatory factor analyses support the four- factor solution of the SEI previously reported by Petrides and Furnham (2000).
6.3.3
Given the need for further population data on the SEI, the means, standard deviations, and coefficient alpha reliabilities of the four factors of are presented in Table 6.2.
Table 6.2 Means, Standard Deviations and Reliablities of the SEI (Schutte et al., 1998). Subscale Total Score Optimism Appraisal Social Skills Utilization M 129.16 35.58 34.38 43.60 15.60 SD 15.82 5.43 5.60 5.77 2.39 a .92 .84 .84 .78 .67
The factor names presented in Table 6.2 are consistent with those chosen by Petrides and Furnham (2000). Previous research on the SEI have not reported descriptive statistics for the scale (e.g., Ciarrochi et al., 2002; Petrides & Furnham, 2000), thus it is difficult to meaningfully compare the means and standard deviations of the present sample to previous studies. However, Schutte et al. (1998; 2001; 2002) 107
Chapter 6; The SEI
have reported the means and standard deviations found in their work (at the full-scale level only), and they are similar to those found in the present study (at least for the normal population data reported in this work). As shown in Table 6.2, the reliability of the SEI was found to be high at the full-scale level and relatively good for the Optimism, Appraisal and Social Skills factors. The reliability to the Utilization factor was found to be low, a finding cons istent with previous research (Ciarrochi et al. 2000; Petrides & Furnham, 2000).
6.3.4
Gender differences
Table 6.3 presents the means and standard deviations for males and females in the present sample on the SEI Table 6.3 Means and Standard Deviations on the SEI by Gender Subscale Total Score Optimism Appraisal Social Skills Utilisation Malesa Females b T-Test M SD M SD 123.64 15.23 131.74 15.21 t(362) = 4.63, p<.001 34.87 32.51 40.99 15.26 5.18 4.93 6.12 2.27 35.95 35.24 44.78 15.78 5.47 5.58 5.19 2.36 Not significant t(362) = 4.38, p<.001 t(362) = 6.01, p<.001 Not significant Effect Size Cohens d -.53 -.20 -.52 -.67 -.22
a, n = 107; b, n = 257.
As shown in Table 6.3 females were found to have significantly higher selfreported EI than males by the order of about half a standard deviation. More specifically, females were found to score significantly higher than males on the Appraisal and Social Skills factors. Interestingly, males and females scores did not significantly differ on Optimism (or managing ones own emotions), or Utilising 108
Chapter 6; The SEI
emotions. In summary the gender differences in EI of the present sample were consistent with previous findings (Schutte et al., 1998; Ciarrochi et al., 2002).
6.3.5
EI and demographic variables
The relationships between scores on the SEI and age, history of mental illness, relationship status and parental upbringing were assessed. There was no significant relationship between scores on the SEI and age. There was a subset of the present sample that reported having a history of mental illness (n = 66; 18%). To assess whether any differences in EI existed between this group and those reporting no known history of mental illness, a similar (in size, age and gender) subset of the sample was selected for comparison. While there was no significant difference in the overall (or total) EI score of the two groups, individuals reporting a history of mental illness (M = 33.77, SD = 6.06) scored significantly lower on the Optimism factor of the SEI than individuals reporting no known history of mental illness (M = 35.76, SD = 5.37), t(153) = 2.159, p<.05, d = .35, although the effect size (Cohens d) of the this difference was not large (Cohen, 1998). When the EI scores of the entire sample of individuals reporting no known history of mental illness (n = 299) was compared to the subset reporting a history of mental illness a very similar difference in mean scores was found although (as could be expected), the effect size was slightly larger ( = .39). No d significant differences were found in the mean EI scores of those reporting to be married or in a relationship to those reporting to be single; and those reporting a dual parental upbringing to those reporting a single parent upbringing or an upbringing in which their parents were divorced.
109
Chapter 6; The SEI
6.4
Discussion
The factor analytic results of the present study confirm those of previous factor analyses of the SEI with a general population sample (Ciarrochi et al., 2002; Petrides & Furnham, 2000). Petrides and Furnham (2000) concluded that researchers should factor analyse the SEI before using it, as the four- factor solution in their sample appeared unstable and had not been replicated. That it has now been replicated, and found to best represent the dimensional structure of the SEI across a number of disparate samples suggests that researchers could generate subscale scores for the SEI according to the items comprising each of the four factors as presented in Figure 6.1 (although further factor analyses of the scale would more fully substantiate this no tion). This is no small point as part of the appeal of more comprehensive tests of EI (e.g., the Bar-On, EQ- i, 1997; MEIS, Mayer et al., 1999), is that they measure multiple dimensions of EI. More importantly, the subscale results on multi- faceted tests often offer more insightful information than overall scores (such as the case for intelligence tests). Indeed in the present study overall scores on the SEI did not differ by gender or history of mental illness, however, there were significant differences between these criterion groups according to the factor scores illustrating the incremental insight gained from comparing subscale results. Different authors have provided somewhat different names for the factors identified in the SEI. The factor names presented in the current study are consistent with those chosen by Petrides and Furnham (2000). Ciarrochi et al (2002) labelled these factors more consistently with the underlying theory of EI (i.e., Optimism was labelled Managing Self Emotions; Appraisal was labelled Perception of emotions; Social Skills was labelled Managing Others Emotions; both authors labelled the fourth-
110
Chapter 6; The SEI
factor Utilization). What the factors should be labelled may be debated until further empirical evidence on the SEI is available. For example, the items assessing Optimism appear to have some overlap in content with optimism scales as mentioned in the introduction (e.g., I expect good things to happen), however, some also appear to be tapping emotional regulation and management strategies more explicitly (e.g., I motivate myself by imaging a good outcome to tasks I take on). Further research examining the relationship between the SEI and optimism and other personality traits and dispositions is needed in order to label the factors more definitively. The findings of the present study pertaining to the reliability of the SEI were also consistent with those reported previously (Ciarrochi et al. 2001; 2002; Schutte et al., 1998). The SEI appears reliable at the full-scale level and the reliabilities of the Optimism, Appraisal and Social Skills factors are relatively good. As with previous studies, the reliability of the Utilization factor was found to be low. Consequently researchers examining the relationship between scores on the scale and criterion variables may choose to exclude this factor from analyses (as did Ciarrochi et al., 2001), or alternatively, attempt to conceptualise better items to assess individuals selfperceived capacity to utilise emotional information in thought. Overall the SEI appears to be a reliable short-self-report measure of EI with a consistent underlying factor structure. While Petrides and Furnham (2000) rightly cautioned further research with it at the time of their analyses, the findings of the current study together with those reported previously suggest that further research with the SEI is warranted. The current study failed to find any significant relationships between scores on the SEI and age in an adult population sample. However, in order to substantiate whether self- reported capacity to perceive, manage and utilise emotions in thought increases with age (or not), larger adult population samples are needed in which
111
Chapter 6; The SEI
substantial population subgroups broken down by age can be comprised for more appropriate comparisons. Further to this, adolescent samples need to be compared with adult samples to help ascertain the nature of the developmental trend (or lack thereof) underlying self-reported EI. It may be that self-reported EI does not increase with age within adult populations, however, that adults do score significantly higher than adolescents, suggesting that development in self-reported EI increases in the adolescent years and then remains relatively sable over time much like other established personality traits (Costa & McCrae, 1992b). As previously discussed, scores on the SEI were found to differ according to gender and history/no history of mental illness. Consistent with pervious research on the SEI (e.g., Ciarrochi et al. 2002; Schutte et al., 1998), and other tests of EI (i.e., the Bar-On EQ-I, Bar-On, 1997; MEIS, Mayer et al., 1999) females were found to score significantly higher than males (although only on some of the SEI factors identified). Furthermore, individuals reporting no known history of mental illness were found to score significantly higher on the Optimism factor than individuals reporting a history of mental illness. Future research, however, needs to establish whether the SEI can predict mental illness over and above personality traits and dispositions (such as optimism and positive affect). Nonetheless these findings highlight the potential utility and incremental insight that may be gained from generating subscale scores on the SEI according to the four-factors identified by the cur rent study and in previous research (Ciarrochi et al., 2002; Petrides & Furnham, 2000) over just using the total score on the scale. While scores on the SEI were found to differ by gender and mental illness, no significant differences in scores on the SEI were found according to parental upbringing or relationship status. Though mostly theoretical, there is a wealth of
112
Chapter 6; The SEI
literature linking adolescent differences in emotional expressiveness, emotional regulation strategies and emotional understanding to the level of parental warmth and the supportive role of the caregiving environment during upbringing (Saarni, 1990; 1999; Volling, McElwain, Notaro, & Herrera, 2002). There is also a wealth of literature showing that children of divorce tend to experience more psychological, social and academic difficulties and that broken families are characterised by less intimate parentchild relationships (Amato, 1993; Demo & Acock, 1988; Sun, 2001). While the current study failed to find any significant differences in the E of adults that reported to be I raised by married parents and adults who reported to be either raised by a single parent or whose parents divorced during their upbringing, this finding (or lack thereof), may be attributable to the rather proxy measures used to assess the level of parental warmth during upbringing. A better understanding of whether dual-parent upbringings, and/or the level of parental warmth during upbringing fosters EI (and to what extent) will be gained from a more in-depth assessment of these variables and their relationships with EI. Another possibility underlying the current findings (or lack thereof in this context), may be due to the fact that the SEI is a self- report measure of EI and therefore measures perceived EI as apposed to actual abilities. Leading authors in the area of EI have argued that performance-based tests of EI will be more predictive of real life criteria than self- report measures of EI because they assess actual abilities to do with emotions rather than perceived emo tional abilities (Mayer, et al., 2000c). Research examining differences in EI among criterion groups (such as those that have had intimate and less intimate parent-child relationships during upbringing), accounted for by different types of EI measures, will help to substantiate such arguments and (in the
113
Chapter 6; The SEI
context of upbringing), provide a better understanding about factors that influence the development of EI. The current study also failed to find any significant differences in the EI of adults that reported being single or in a relationship. This finding (or lack thereof) could also be attributed to the problems inherent in the current methodology (i.e., the use of proxy measures of relationship status/quality), or more simply, there may indeed be no differences in self-reported EI amongst people who are single and in relationships. Nonetheless, further research examining differences in EI amongst such criterion groups would help delineate what EI contributes to in terms of real life criteria. Indeed, using more in depth measures of relationship status, Schutte et al. (2001) has shown that participants who report greater EI display more cooperative responses toward partners, have higher scores for close and affectionate relationships, and report greater marital satisfaction than those who report low EI. This study, however, failed to control for variables that the SEI bears considerable content overlap with such as optimism and positive affect (Ciarrochi et al. 2002). As mentioned in the introduction of this chapter, optimistic individuals with high general levels of positive affectivity also tend to exhibit more adaptive interpersonal qualities (Berry & Willingham, 1997; Sprecher, 1999; Vittengl & Holt, 2000). As noted by Petrides & Furnham (2000), the predictive validity of self-report (or trait) measures of EI needs to be demonstrated against established measures of personality and related dispositions such as optimism and affectivity. A greater understanding of the nature, origin and contribution of EI to theoretically related variables will only be gained from research that includes a battery of EI measures and a battery of measures for which EI must exhibit incremental predictive validity against (e.g., established personality factors with self- report
114
Chapter 6; The SEI
measures of EI and intelligence factors for performance-based measures of EI; Petrides & Furnham, 2000). Research to-date on the SEI suggests that it is a reliable self- report measure of EI with a consistent factor structure that correlates meaningfully with theoretically related variables (Ciarrochi et al., 2001; 2002; Schutte et al., 1998; 2001; 2002). As such, it appears to be a useful measure to include in amongst a battery of others (e.g., broader self- report measures of EI such as the Bar-On EQ-i, Bar-On, 1997; and performance-based tests such as the MEIS, Mayer et al., 1999), in such multimethod research.
115
Chapter 7; The TAS-20
CHAPTER 7 Assessing components of the emotional intelligence construct with the TAS-20
7.1
Overview
Given the conceptual overlap between the latent variables assessed by the Toronto Alexithymia Scale (TAS-20) and the more recently introduced construct of emotional intelligence (EI), it has been argued that the TAS-20 provides a brief measure of EI or a criterion measure of convergent validity. This study examined the utility of the TAS-20 in this context with an Australian general population sample providing some of the first Australian general population norms for the scale. Scores on the TAS20 were not found to be normally distributed, however, further examination of the data revealed that the degree of skew is not likely to affect the results of multivariate research with the scale. The TAS-20 was found to have good internal reliability (full scale a = .86) and the results of confirmatory factor analyses supported the three-factor derived scales (Difficulty Identifying Feelings, Difficulty Describing Feelings, and Externally Oriented Thinking). Scores on the TAS-20 were also found to be significantly higher in males than females, and were associated with fewer years of education. The findings are discussed in terms of the utility of the TAS-20 as a criterion measure of convergent validity in EI validity studies or as an actual measure of components of EI.
7.1.2
Introduction
The term alexithymia, meaning no words for feelings, was coined by Sifenos (1973), to describe a series of affect related deficits that were originally observed in, and thought to be typical of, patients with classical psychosomatic disorders. Sifenos (1973), 116
reported that a common characteristic of many psychosomatic patients was that they appeared to have difficulty in identifying and describing their feelings and a cognitive style that appeared to be unimaginative, utilitarian and externally oriented. A general consensual that now exists amongst researchers and clinicians is that alexithymia represents a cognitive deficit in the processing of affect, the salient features of which include difficulty identifying subjective feelings, difficulty describing subjective feelings and an externally oriented cognitive style, that is, a cognitive style that is characterised by a preoccupation with the minute details of external events, rather than by feelings, fantasies, and other aspects of inner experience (Taylor & Bagby, 2000, p.43). These facets of alexithymia considerably overlap with some of the facets of EI (Parker, Taylor & Bagby, 2001; Taylor & Bagby, 2000; Saklofske, Austin, & Minski, 2003). As outlined in Chapter Two and Three, EI was originally operationalised by Salovey and Mayer (1990), as a set of abilities to do with emotions and the processing of emotional information, including; (1) the ability to identify and express emotions; (2) the ability to effectively regulate and manage emotions; and (3) the ability to utilise or reason with emotions in thought. The first of these EI abilities conceptually overlaps with the difficulty identifying and describing subjective feelings facets of alexithymia, while the third EI ability (to utilise or reason with emotions in thought), conceptually overlaps with the externally oriented thinking style facet of alexithymia. Indeed, socalled alexithymics are said to have difficultly linking their feelings and emotions with memories, fantasies, or specific situations (Taylor, Bagby & Parker, 1991). The twenty- item Toronto Alexithymia Scale (TAS-20; Bagby, et al., 1994) is one of the most widely used measures of alexithymia. It comprises three subscales that assess; (1) Difficulty Identifying Feelings; (2) Difficulty Describing Feelings; and (3) Externally Oriented Thinking. The TAS-20 has been found to have relatively good
117
psychometric properties (for a review see Taylor, Bagby & Luminet, 2000). Research studies have shown that the TAS-20 has satis factory internal consistency and test-retest reliability, and good factorial, convergent, and discriminant validity (Bagby, Parker & Taylor, 1994b; Bagby, et al., 1994a). Moreover, scores on the TAS-20 have been found to show high agreement with observer ratings of the facets of alexithymia (Bagby, et al., 1994a). Given the conceptual overlap between alexithymia and EI, it has been argued that the TAS-20 may provide a criterion measure of convergent validity for newly developed measures of EI (Mayer, et al., 1990; Parker, et al., 2001; Schutte et al. 1998), or indeed an actual measure of some components of the construct of EI (Davies, et al., 1998; Palmer, et al., 2002). Importantly, it may provide a measure of the ability to express emotions, which existing measures of EI do not adequately assess (Petrides & Furnham, 2001). In order for the TAS-20 to be used as a criterion measure in EI validity studies, or as an actual measure of components of the construct of EI, comparative general population data on the test is need. The TAS-20 has been predominantly used to classify the clinical features of a variety of psychiatric and psychosomatic disorders (Bagby et al., 1994a; Taylor & Bagby, 2000). As such, research with the TAS-20 has predominantly examined the prevalence and distribution of scores on the test in clinical population samples. However, there are an emerging number of studies that have examined the TAS-20 with population samples (Kooiman, Spinhoven, & Trijsburg, 2002) providing general population data on the test. Research on the TAS-20 with a general population sample has shown that scores on the test are normally distributed (Salminen, Saarijaervi, Aeaerelae, Toikka, & Kauhanen, 1999), however, this finding has not yet been substantiated and needs to be further examined by research.
118
Other research studies on the TAS-20 with general population samples have shown that it is internally reliable (Kooiman et al., 2002; Parker, et al., 2001), and the three- factor structure of the TAS-20 has been replicated in a number of studies (Loas et al., 2001; Joukamaa et al., 2001; Parker et al., 2001). While such psychometric properties have been established in a number of countries such as Canada (Parker et al., 2001), Sweden (Simonsson-Sarnecki et al., 2000), Engla nd and France (Zech, Luminet, Rime, & Wagner, 1999), Finland (Salminen et al., 1999), and the United States (Lane, Sechrest, & Riedel, 1998), at present, very little Australian population data on the TAS20 has been collected. The aim of the current study was to contribute to this emerging body of research by examining the distribution of scores, the reliability and factor structure of the TAS-20 with an Australian population sample. Research studies that have examined the TAS-20 in normal population samples have also shown that scores on the test are associated with several demographic variables such as sex, levels of education and income. For example, Salminen et al. (1999) found that scores on the TAS-20 were associated with advanced age, low levels of education and low socio-economic status. Similarly, Lane et al. (1998), also found that scores on the TAS-20 were associated with older age, male sex, lower socioeconomic status and fewer years of education.
7.1.3
The current study examined the distribution of scores on the TAS-20, its reliability and factor structure, and the relationship between scores on the TAS-20 and several demographic variables including, sex, education, income and history of mental illness. Consistent with previous research, scores on the TAS-20 were expected to be normally distributed (Salminen et al., 1999). The TAS-20 was also expected to show 119
good internal reliability and support for the three-factor solution comprising Difficulty Identifying Feelings, Difficulty Describing Feelings and Externally Oriented Thinking was expected to be found in confirmatory factor analyses. Moreover, males were expected to score significantly higher than females (Salminen et al., 1999). Finally, scores on the TAS-20 were also expected to be negatively associated with education and income (Salminen et al., 1999), and those participants reporting a history of mental illness were expected to score significantly higher on the TAS-20 than those reporting no known history of mental illness, consistent with previous research that has found the features of alexithymia to be prevalent in a variety of psychiatric and psychosomatic disorders (Bagby et al., 1994a,b; Taylor & Bagby, 2000).
7.2
Method
7.2.2
Participants
The sample comprised 367 participants (107 males, 257 females, 3 unreported), ranging in age from 15 to 78 years with a mean age of 38.30 years (SD = 13.70). The participants were drawn from the general population across the two most populated Australian states (Victoria and New South Wales), through advertisements to participate. The majority of the sample were white Caucasian Australians (69.5%), the remainder were; White Caucasian Emigrants (18.3%); Asian Emigrants (9.3%); other (1 %); unreported (2.2%). Of the 367 participants, 66 (18%) reported having a history of mental illness, which was representative of the prevalence rates of mental illness amongst the Australian population (which is 18% according to the Australian Bureau of Statistics, National Survey of Mental Health and Wellbeing of Adults, 1997). The
120
sample was diverse in terms of years of education; 96(28%) reported completing a secondary education only; 69(20%) reported completing a tertiary certificate or diploma; 105 (30%) reported completing an undergraduate degree; and the remainder 77(22%) reported to have completed a postgraduate degree; (20 unreported). The annual incomes of the sample were also diverse; 72(20%) reported an annual income between $0-10 thousand dollars; 56(16%) between $10-20 thousand dollars; 56(16%) between $20-30 thousand dollars; 53 (15%) between $30-40 thousand dollars, 45 (13%) between $40-50 thousand dollars; and 70 (20%) reported an annual income greater than $50 thousand Australian dollars. In summary, the present sample was diverse allowing for the comparison of scores on the test in terms of sex, years of education, annual income and history of mental illness.
7.2.3
Materials
The Twenty-Item Toronto Alexithymia Scale (TAS-20; Bagby et al., 1994a,b) is a self-report measure comprising a five-point Likert-type response format where 1 = strongly disagree and 5 = strongly agree. The TAS-20 comprises three subscales; 7 items assess Difficulty Identifying Feelings; 5 items assess Difficulty Describing Feelings; and 8 items assess Externally Oriented Thinking; five items are negatively keyed. The TAS-20 has been found to have satisfactory internal reliability, good convergent and discriminant validity, and factor analyses have revealed a three-factor structure consistent with the subscales of the scale (Bagby et al., 1994a,b; Luminet, Bagby, Wagner, Taylor & Parker, 1999).
121
7.2.4
Procedure
A TAS-20 test booklet was constructed from the instructions and items provided by Bagby et al. (1994b), and a scannable answer sheet was comprised. Participants who responded to advertisements about the study collected the TAS-20 and answer sheets and were briefed about the purpose of the study. While the instructions in the TAS-20 booklets were self explanatory, the purpose of the briefing sessions were to emphasise that the TAS-20 was to be completed independently without input from others and in its entirety. Upon completion participants returned the item booklets and answer sheets, and received a small monetary reward for participating.
7.3
Results
7.3.2
The distribution of scores on the TAS-20 were examined at the outset. Total TAS-20 and subscale scores were generated by summing their respective items as outlined by Bagby et al. (1994b) and are presented in Table 7.1.
122
Table 7.1. Descriptive Statistics for the TAS-20 Statistic Mean Median Standard Deviation Skewness Standard Error of Skewness Kurtosis Standard Error of Kurtosis Coefficient a Pearson Correlations DDF2 EOT3 Nonparametric Correlations (Spearmans Rho) DDF2 EOT3 DIF1 15.95 14.00 6.99 1.04 .128 .707 .255 .82 .598 .383 .608 .292 DDF2 11.29 11.00 4.63 .460 .128 -.599 .255 .80 EOT3 13.36 13.00 3.90 .705 .128 .638 .255 .65 TAS4 40.60 40.00 12.60 .781 .128 .619 .255 .86
.383
.380
N = 367; 1 = Difficulty Identifying Feelings; 2 = Difficulty Describing Feelings; 3 = Externally Oriented Thinking; 4 = TAS-20 Total Score.
As shown in Table 7.1, scores on the TAS-20 were not normally distributed. The Total TAS-20 scores and scores across the three subscales were all positively skewed and produced kurtosis values above and below z ero. Skewness and Kurtosis values greater than two times the standard errors of skewness and kurtosis are considered to represent data that is skewed to a significant degree (Tabacknick & Fidell, 1996) and this was the case for the Total TAS-20 and subscale scores in the present study. However, Tabacknick and Fidell (1996) caution the interpretation of formal inference tests of normality with large samples (e.g., greater than 200), as the standard errors for both skewness and kurtosis decrease according to the sample size, increasing the likelihood of rejecting the null hypothesis when there may be only minor deviations from normality. Indeed as shown in Table 7.1, the mean and median scores on the TAS123
20 differed at worst by only a quarter of a standard deviation (i.e., for the Difficulty Identifying Feelings subscale that was the most skewed in the data set). Furthermore the Pearson and Spearman intercorrelations between the subscales (where Pearson correlations are sensitive to violations of normality whilst Spearman correlations are less sensitive), produced highly similar results as also shown in Table 7.1. This finding suggests that despite the statistically significant skewness of the data, it was not likely to make a substantive difference in the other analyses. For example, Waternaux (1976) has highlighted that underestimates of variance associated with kurtosis values that are significantly above and below zero do not occur with samples of 200 or more. The means and standard deviations of the present sample (as shown in Table 7.1), were similar to a sample of Swedish undergraduate psychology students (Simonsson-Sarnecki et al., 2000), but significantly lower than a Canadian general population sample (Parker et al., 2001) on the Externally Oriented Thinking subscale M = 19.39, SD = 4.99, t(1095) = 4.67, p<.001. The internal reliability of the total TAS-20 scale and the Difficulty Identifying and Describing Feelings subscales was found to be high, however, the internal reliability of the Externally Oriented Thinking subscale was relatively lower in comparison, consistent with previous research (Kooiman, et al., 2002). Finally, the intercorrelations between the subscales are consistent with previous research and the underlying theory of alexithymia that describes the facets of the construct as conceptually distinct but related dimensions (Taylor & Bagby, 2000).
7.3.3
Confirmatory factor analysis via structural equation modelling using AMOS (Arbuckle, 1999), was performed to test the fit of the three-factor model of the TAS-20
124
found in previous analyses (Bagby et al., 1994b). Specifically, three different models were tested; a general factor model, a three-oblique factor model comprising Difficulty Identifying Feelings, Difficulty Describing Feelings, and Externally Oriented Thinking, and a three-orthogonal factor model comprising the same three factors. The threeoblique factor model was expected to provide the best fit with the present data. Three fit indices were chosen to compare the degree of fit for each model. Specifically; (1) the Kline adjustment (?2 /df); (2), the Comparative Fit Index (CFI; Bentler, 1990); and (3) the Root Mean Square Error of Approximation (RMSEA). Table 7.2 lists the fit indices obtained for each model.
Table 7.2 Model fit statistics for the TAS-20 (Bagby et al., 1994). Fit Statistic General Three Orthogonal Three Oblique Factor Model Factor Model Factor Model a CMIN 683.18 665.71 405.90 Kline adj. b 4.02 3.92 2.43 c CFI .96 .96 .98 RMSEAd .09 .09 .06 a = Chi-square statistic; b = Kline Adjustment; c = Comparative fit index; d = Root Mean Square Error of Approximation.
As shown in Table 7.2, the best fitting model was the three oblique factor model, a finding consistent with previous factor analyses and the underlying theoretical model of alexithymia (Bagby et al., 1994b). In contrast, none of the other tested models fit the data well according to the Kline adjustment. The three oblique factor model, and the intercorrelations amongst the factors are presented in Figure 7.1. All of the correlations are substantial, ranging in size from .54 to .73 Although not displayed, all of the item loadings were positive, ranging in size from .13 to .77 with a mean of .56 and a standard deviation of .18. 125
1 3 6 Identifying 7 9 13 14
.73
2 4 Describing 11 12 17
.54
.58
5 8 10 15 External 16 18 19 20
Figure 7.1
Oblique Three-Factor Model of the TAS-20 (Bagby et al., 1994b)
126
7.3.4
Relations hips with demographic variables
Table 7.3 presents the means and standard deviations of scores on the TAS-20 according to gender (males and females), years of education, annual income and history of mental illness.
127
Table 7.3 Means and Standard Deviations on the TAS-20 according to Gender, Years of Education and History of Mental Illness Criterion Group Gender Females Males Years of Education Secondary Education Tertiary Certificate/Diploma Undergraduate Degree Postgraduate Degree Annual Income 1 0-10K 10-20K 20-30K 39-40K 40-50K 50K+ History of Mental Illness No Known History 72 56 56 53 45 70 41.46 (12.99) 38.50 (10.04) 38.49 (12.13) 40.27 (13.06) 46.20 (14.68) 39.81 (11.80) 17.08 (7.00) 14.57 (5.22) 14.80 (6.56) 15.96 (7.28) 18.42 (8.17) 15.24 (7.18) 11.31 (5.04) 11.14 (4.61) 10.84 (4.85) 11.27 (4.59) 12.83 (4.52) 10.59 (3.72) 13.07 (3.85) 12.79 (3.07) 12.85 (3.76) 13.04 (4.32) 14.96 (4.71) 13.97 (3.56) 255 105 39.27 (12.56) 44.01 (12.10) 45.26 (12.38) 39.00 (12.52) 39.12 (11.80) 37.73 (12.96) 15.65 (6.94) 16.72 (7.07) 17.88 (7.00) 15.39 (7.19) 15.08 (6.33) 14.97 (7.16) 10.65 (4.47) 12.90 (4.62) 13.12 (4.92) 10.87 (4.31) 10.72 (4.16) 10.05 (4.47) 12.97 (3.94) 14.39 (3.61) 14.25 (3.74) 12.74 (3.83) 13.32 (3.73) 12.71 (4.03 n M (SD) TAS2 M(SD) DIF3 M(SD) DDF4 M(SD) EOT5
96 69 105 77
40.47 15.71 11.27 13.49 (12.92) (7.07) (4.69) (4.02) History Reported 65 41.22 17.17 11.36 12.67 (11.29) (6.58) (4.45) (3.23) 1: K = Thousand Dollars; 2 = TAS-20 Total Score; 3 = Difficulty Identifying Feelings; 4 = Difficulty Describing Feelings; 5 = Externally Oriented Thinking.
296
As shown in Table 7.3, males scored significantly higher than females overall (i.e., on the total TAS-20 score; t(358) = 3.28, p<.01, d = .38) by approximately one-
128
third of a standard deviation although the effect size (Cohens d) of the this difference was not large (Cohen, 1998). More specifically, males were found to score significantly higher than females on the Difficulty Describing Feelings (t(358 = 4.28, p<.001, d = .50), and the Externally Oriented Thinking subscales (t(358) = 3.18, p<.01, d = .42), but not on the Difficulty Identifying Feelings subscale. Scores on the TAS-20 were also found to be associated with years of education. More specifically, those who reported to have completed a secondary school education only scored significantly higher on the total score than those who reported to have completed a tertiary certificate or diploma (t(339) = 3.21, p<.01, d = .50), undergraduate degree (t(339) = 3.50, p<.01, d = .51), and post graduate degree (t(339) = 3.95, p<.001, d = .60). Interestingly this finding only existed between those who had only completed secondary school education and the other groups. For example, there was no significant difference in scores on the TAS-20 between those who reported to have completed a tertiary certificate or diploma and those who reported to have completed a postgraduate degree (t(339) = .614, p ns.). As shown in Table 7.3, scores on the TAS-20 did not appear to be meaningfully associated with annual income. For example, there were no significant differences in scores on the TAS-20 between those who reported an annual income between 10 and 20 thousand dollars and those who reported an annual income over 50 thousand dollars (t(342) = .586, p ns.). Surprisingly, no significant differences in TAS-20 scores were found between those who reported to have no known history of mental illness and those who reported having a history of mental illness.
129
7.4
Discussion
The findings of the current study failed to confirm earlier findings that scores on the TAS-20 are normally distributed in the general population (Salminen et al., 1999). In contrast, scores on the TAS-20 and its subscales were all positively skewed. Taylor and Bagby (2000) have claimed that alexithymia (as measured by the TAS-20), is a personality construct, as opposed to a categorical variable, that is normally distributed in the general population. The findings of the current study highlight that further research examining the distribution of scores on the TAS-20 is needed in order to substantiate this claim. However, the findings of the current study suggest that TAS-20 data that is not normally distributed is not likely to influence the patterns of correlations in multivariate research such as in EI validity studies that have used the TAS-20 either as a criterion measure of convergent validity (e.g., Schutte et al., 1998) or as an actual measure of components of the cons truct of (e.g., Davies et al., 1998). Scores on the TAS-20 in the current study were all skewed in the same direction (i.e., positively skewed) and to about the same degree, and there was little difference in the Pearson and Spearman determined inter-correlations amongst the subscales. Moreover, despite the skewness of scores on the TAS-20 there was very little difference between the mean and median scores on the scale. This finding suggests that mean scores on the TAS-20 are a good indicator of the central tendency of scores, thus interpretation of mean differences (e.g., males vs. females) are valid. The means and standard deviations of the present sample provide some of the first Australian general population norms for the TAS-20. These were found to be similar to a sample of Swedish undergraduate psychology students, but significantly lower than a Canadian general population sample (Parker et al., 2001), on the Externally
130
Oriented Thinking subscale. This finding highlights the need to establish popula tion norms on the test as they appear to differ (at least on some subscales), across different cultures as has been previously found (Le, Berenbaum, & Raghavan, 2002). Consistent with previous research, the TAS-20 was found to have acceptable internal reliability although the reliability of the Externally Oriented Thinking subscale is low in comparison to the other subscales (Parker et al., 2001). The three- factor structure of the TAS-20 was replicated by the current study confirming the stability and usefulness of its three- factor derived subscales (Bagby et al., 1994a,b). A three-oblique factor model consistent with previous analyses and the underlying theoretical model of alexithymia was found to provide the best fit to the present data than both a one- factor and a threeorthogonal factor model. The validity of the three- factor structure of the TAS-20 has been demonstrated in both clinical and non-clinical population samples and in a number of language-translated versions highlighting the utility of its subscale scores (Taylor, et al., 2000). Consistent with previous research scores on the TAS-20 were found to be associated with male gender and fewer years of education (Lane et al., 1998; Salminen et al., 1999). While males and females reported similar ability in identifying their feelings, males reported more difficulty in describing their feelings and less emotional reasoning. This finding is consistent with a wealth of literature that describes how males are socialised to be less emotionally expressive than females and to be less concerned with feelings and other aspects of inner experience than females (Brody, 1999). However, this finding could also be used to question the validity of the TAS-20. Do scores on the test provide a true index of individual differences in the processing of emotions, or are they influenced by and therefore reflect to some extent societal norms, and related gender roles and stereotypes? Given that the TAS-20 is a self-report
131
measure, individuals responses to the test are filtered through the self-concept and may therefore index beliefs about ones ability to process emotions as opposed to actual abilities (Mayer, et al., 2000c). However, there is strong support for the convergent validity of the TAS-20 and each of its three factor derived subscales (Taylor & Bagby, 2000). Moreover, scores on the TAS-20 have been shown to be in high agreement with observer ratings of alexithymia suggesting that it does index (to some extent) individual differences in emotional processing. The current study also examined differences in TAS-20 scores between two criterion groups, those reporting a history of mental illness and those reporting no known history of mental illness. Surprisingly, no significant differences between these two groups were found. This finding is in stark contrast to previous research that has shown that scores on the TAS-20 are related to variables such as maladaptive coping styles (Parker, Taylor & Bagby, 1998), vulnerability to stress (Bagby et al., 1994a) and psychiatric illness (Taylor, Bagby & Parker, 1997). The present findings (or lack thereof) are most probably attributable to the rather proxy measure of mental illness employed by the current study, that is, asking participants to report whether or not they have had a history of mental illness. Such a proxy measure potentially mixes together those who have had, and who are currently experiencing mental illness, with those who may be medicated and those who are not. Moreover it does not take into account those who may have had a mental illness or history thereof, who may have decided not to report such, and those who may have a mental illness they were possibly unaware of. A more stringent test for establishing such criterion groups would be to use a measure such as the Composite International Diagnostic Interview (CIDI; Robins et al., 1988), a comprehensive interview for adults which can be used to assess current and lifetime prevalence of mental disorders through the measurement of symptoms and their impact
132
on day-to-day activities. Nonetheless, previous research has shown that the scores on broader measures of EI do differ between those who report to have suffered from a mental illness in comparison to those who report no known history of mental illness (Palmer, Manocha, Gignac & Stough 2003). As a measure of EI, the TAS-20 may indexes individual differences in the perceived ability to identify and express inner feelings, and the capacity to utilise or reason with emotions in thought. Importantly, it provides a measure of perceived ability (or lack thereof) to express emotions, which extant EI measures do not adequately assess. For example, neither the MSCEIT (Mayer, et al., 2003), the Bar-On EQ-I (BarOn, 1997a), the TMMS (Salovey, et al., 1995), or the EI scale by Schutte et al. (1998) have sub-scales solely concerned with the expression of emotions. Of the 133 items within the Bar-On EQ-i only four are concerned with the expression of emotions (e.g., Its fairly easy for me to express feelings; Bar-On, 1997a, p.181) tha t are subsumed by the Emotional Self- Awareness sub-scale. Similarly the EI scale by Schutte et al. (1998) has only two items that assess the capacity to express emotions that are subsumed by the Emotion Perception factor identified within the scale (Ciarrochi, et al., 2001; Petrides & Furnham 2000). The Difficulty Describing Feelings subscale of the TAS-20 is internally consistent and has been shown to meaningfully correlate with criterion measures of convergent and divergent validity (Taylor et al., 2000). Given that no objective measures of the ability to express emotions currently exist (Petrides & Furnham, 2001), the TAS-20 may provide an interim measure of this important component of the EI construct. The results of the current study (and much previous research), suggest that the TAS-20 has relatively good psychometric properties and therefore presents at least a brief self-report measure of EI or indeed a good criterion measure for use in the validation of newly developed measures of EI.
133
Chapter 8; Relationships between measures of EI
CHAPTER 8 A taxonomic model for emotional intelligence
8.1
Overview
In this study a population sample completed a battery of EI measures that were representative of the tests currently available to measure EI. From a systematic review of variables measured by these tests, five common dimensions of EI were identified and a comprehensive model of EI was hypothesised including (1) Emotional Recognition and Expression; (2) Understanding Emotions External; (3) Emotions Direct Cognition; (4) Intrapersonal Manage ment; and (5) Interpersonal Management. Confirmatory factor analyses via structural equation modelling were then conducted to assess whether this model (and a range of alternatives) provided a statistically significant fit with the data. None of the models assessed were found to provide an acceptable fit with the data according to standard model fit statistics, however, the hypothesised model identified in the review was found to be the best fitting model in comparison to the other models assessed. An exploratory factor analysis was conducted and interpreted as providing further partial support to the hypothesised model. It is argued that the hypothesised model of EI did not provide a statistically significant fit with the present data as variables assessing the ability to express emotions were not adequately represented (i.e., too few items assess this common facet of the construct); and variables assessing the ability to utilise or reason with emotions in thought (Emotions Direct Cognition) are not adequately reliable. Given that the five dimensions of the hypothesised model can be systematically identified from a cross section of EI tests, and that partial support for this model found by the current study, it was concluded that the hypothesised model is representative of the communality amongst models and measures of EI. Directions for future research of this nature are discussed. 134
8.1.2
Introduction
While a number of recent papers have theoretically placed alternative models and measure of EI into more definitive categories (Mayer et al., 2000a; Petrides & Furnham, 2000), little attempt has been made to systematically identify the communality that may exist amongst them. It was concluded in Chapter Two that it would be useful to examine the relationships between different models and measures of EI, and to establish whether common dimensions of the construct could be identified. It was argued that the identification of common dimensions of EI would provide a useful definition for EI and a model upon which comprehensive measures of the construct could be based. Goleman (2001b) has recently speculated that the predominant models and measures of EI share some common elements, specifically, abilities or competencies concerned with the capacity to recognise and regulate emotions in oneself and others. Goleman (2001b) has further suggested that a parsimonious definition of EI would involve four higher order factors of EI: (1) the capacity to recognise emotions in the self (Self- Awareness); (2) the capacity to regulate emotions in the self (SelfManagement); (3) the capacity to recognise emotions in others (Social Awareness); and (4) the capacity to regulate emotions in others (Relationship Management). Although Goleman (2001b) has claimed that these four dimensions of EI can be identified as providing a common taxonomy for EI, no systematic review of the EI literature was presented in support of this hypothesis. Furthermore, no definition of what might constitute a common dimension of the EI construct was provided.
135
8.1.3
A taxonomy for EI
Toward establishing common dimensions of EI, the current study presents factor analyses involving five measures of EI representing a good cross section of alternative approaches to the conceptualisation and measurement of the construct. These included the MSCEIT (Mayer, et al., 2000d); the Bar-On EQ-i (Bar-On, 1997a); and other related measures including: the TMMS (Salovey, et al., 1995); the TAS-20 (Bagby, et al., 1994a,b); and the EI scale by Schutte et al., (SEI; 1998). Collectively these measures of EI represent (although not absolutely) the breadth of variables currently being placed under the banner of EI. A more comprehensive battery of EI models and measures may have included the Emotional Competency Inventory (Boyatzis, et al., 2000) and the EQMAP being offered QMetrics (http://www.qmetricseq.com), based on the model of EI put forth by Cooper and Sawaf (1997), however, these measures were not made available for the current research purpose by the companies that distribute them. Nonetheless, these so-called mixed models bear considerable overlap with other mixed models and measures of EI (Mayer et al., 2000a), notably the Bar-On EQ-i (BarOn, 1997a), which was included in the current battery. The measures included in the current battery are some of the most widely research and used in the area. While validity data for these measures in comparison to well established measures of personality and intelligence (IQ) is relatively sparse (Salovey, et al., 2000), each of these measures of EI exhibit preliminary evidence of reliability and validity and as such provide a platform from which to determine the dimensional communality amongst the different approaches to EI. One way to hypothesize common dimensions of the construct of EI is to; (a) define a criterion for what constitutes a common dimension of the construct; (b) systematically compare the components of various models and measures that cover the 136
breadth of variables being placed under the banner of the construct; and (c) from the comparison identify the common variables amongst them according to the predefined dimensional criterion. Table 8.1 lists the components of EI assessed by the MSCEIT (Mayer et al., 2000d), the Bar-On EQ-i (Bar-On, 1997a), the TMMS (Salovey et al., 1995); the TAS-20 (Bagby et al., 1994a,b); and the sub-scale factors of the EI scale by Schutte et al. (1998) as determined by both Petrides and Furnham (2000) and Ciarrochi, et al., (2001).
137
Table 8.1 Measures of Emotional Intelligence Measure MSCEIT Emotional Perception (Faces & Pictures tests) Emotional Facilitation (Facilitation & Synesthesia tests) Understanding Emotion (Changes & Blends tests) The ability to perceive emotions in oneself and others as well as in objects, art, stories and the like The ability to generate, use, and feel emotion as necessary to communicate feelings, or employ them in other mental processes. The ability to understand emotional information, how emotions combine and progress through relationship transitions, and to reason about such meanings The ability to be open to feelings, to modulate them in oneself and others so as to promote personal understanding and growth The ability to recognise ones feelings. The ability to express feelings, beliefs, and thoughts and defined ones rights in a nondestructive manner. The ability to respect and accept oneself and basically good. The ability to realise ones potential capacities The ability to be self-directed and selfcontrolled in ones thinking and actions and to be free of emotional dependency. The ability to be aware of, to understand, and to appreciate the feelings of others The ability to establish and maintain mutually satisfying relationships that are characterised by intimacy and by giving and receiving of affection The ability to demonstrate oneself as a cooperative contributing and constructive member of ones social group The ability to identify and define problems as well as to generate and implement potentially effective solutions The ability to assess the correspondence between what is experienced and what objectively exists The ability to adjust ones emotions, thoughts, and behaviour to changing situations and conditions. 138 Subscales Definitions
Managing Emotion Management & Relationships tests). EQ-I Emotional Self-Awareness Assertiveness
Self- Regard Self- Actualisation Independence
Empathy Interpersonal Relationship
Social Responsibility
Problem Solving
Reality Testing
Flexibility
Table 8.1 Continued Stress Tolerance
Impulse Control Happiness Optimism
The ability to withstand adverse events and stressful situations without falling apart by actively and positively coping with stress. The ability to resist or delay an impulse, drive or temptation to act The ability to feel satisfied with ones life, to enjoy oneself and others, and to have fun The ability to look at the brighter side of life and to maintain a positive attitude, even in the face of adversity. Ability to attend to moods and emotions Ability to discriminate clearly among subjective feelings Ability to regulate moods and emotions Difficulty identifying subjective feelings Difficulty describing feelings Cognitive style characterised by a preoccupation with the minute details of external events, rather than by feelings, fantasies, and other aspects of inner experience (Taylor & Bagby, 2000). Ability to identify emotions within the self and others Ability to effectively regulate and manage ones own emotions Ability to effectively regulate and manage the emotions of others The ability to utilise or reason with emotions in thought
TMMS Attention Clarity Repair TAS-20 Difficulty Identifying Feelings Difficulty Describing Feelings Externally Oriented Thinking
SEI Emotional Perception Emotional Management Self Emotional Management Others Utilisation
Comprehensive taxonomies in psychology have been identified in the past via a lexical type approach, where higher-order factors are identified from the major themes that recur in domain adjectives (e.g., descriptors of personality traits used to identify the comprehensive Five-Factor Model of personality; Costa & McCrae, 1992). In accordance with the principles of this methodology, it was proposed by the current study that a common dimension of EI could be defined as one that comprised variables that were shared by at least two of the main theoretical models and measures of the 139
construct. According to this criterion Golemans (2001b) four- factor model of EI does not take into account other variables that are common to models and measures of EI. Most notably, Golemans (2001b) model does not comprise a factor concerned with the use of emotions in thought, an element inherent in a number of EI models as shown in Table 8.1. For example, the Utilisation factor of the SEI (Schutte et al., 1998), the Facilitation subscale from the MSCEIT (Mayer et al., 2003), and the Externally Oriented Thinking subscale form the TAS-20 (Bagby et al., 1994). Based on the criterion for a common dimension of EI established by the current study, a systematic comparison of the variables (and their meaning) presented in Table 8.1, lead to the hypothesis of a five-factor model representing the communality amongst the alternative measures of EI assessed, as presented in Table 8.2 and discussed thereafter.
140
Table 8.2 The hypothesised five- factor general taxonomy for EI General emotional intelligence (1) Emotional Recognition and Expression (2) Perception and Understanding of Emotions External Empathy (EQ-i) Faces (MSCEIT) Pictures (MSCEIT) Blends (MSCEIT) Changes (MSCEIT) (3) Emotions Direct Cognition (4) Intrapersonal Management (5) Interpersonal Management
Emotional SelfAwareness (EQ-i) Attention (TMMS) Clarity (TMMS) Emotional Perception (SEI) Difficulty Identifying Feelings (TAS-20) Difficulty Describing Feelings (TAS-20)
Sensations (MSCEIT) Facilitation (MSCEIT) Utilisation (SEI) Externally Oriented Thinking (TAS-20) Problem Solving (EQi) Reality Testing (EQi) Flexibility (EQ-i)
Repair (TMMS) Managing Self Emotions (SEI) Stress Tolerance (EQi) Impulse Control (EQ-i) Happiness (EQi) Optimism (EQi) Self- Regard (EQ-i) Assertiveness (EQ-i) Independence (EQ-i)
Managing Others Emotions (SEI) Interpersonal Relationship (EQ-i) Social Responsibility (EQ-i) Management (MSCEIT) Relationships (MSCEIT)
The first hypothesised common dimension of EI presented in Table 8.2 Emotional Perception and Expression, concerns the ability to perceive and express ones own emotions. Salovey and Mayer (1990) and all subsequent theorists (Bar-On, 1997a; Cooper & Sawaf, 1997; Goleman 1995; Schutte et al., 1998) have conceptualised the capacity to perceive inner subjective emotions as a component of EI. Indeed all the measures presented in Table 8.1 comprise sub-scales purported to assess the capacity to perceive emotions. These include: the Emotional Self- Awareness subscale of the Bar-On EQ- i (Bar-On, 1997a); the Attention and Clarity sub-scales of the TMMS (Salovey et al., 1995); the Emotional Perception factor inherent within the scale 141
by Schutte et al., (1998); and finally the Difficulty Identifying Feelings sub-scale of the TAS-20 (Bagby et al., 1994). Other sub-scales that may load on the first hypothesised factor are those concerned with the capacity to express emotions. In their original conceptualisation of EI and in their later revision (Mayer & Salovey, 1997; Salovey & Mayer, 1990), Mayer and Salovey have argued that individuals differ in their capacity to express how they feel to others. Mayer and Salovey (1997) have also argued that the ability to express inner feelings and emotions is highly related if not dependant on the capacity to perceive emotions and have operationalised these two variables of EI together (e.g., the first branch of Mayer and Saloveys 1997 model of EI concerns the the ability to perceive accurately, appraise and express emotion (p.10). As such, the Difficulty Describing Feelings sub-scale of the TAS-20 which comprises items such as It is difficult for me to find the right words for my feelings and I am able to describe me feelings easily, was also expected to load on this first factor of EI, Emotional Perception and Expression. The second hypothesised dimension of EI presented in Table 8.2 Understanding of Emotions External concerns the capacity to perceive and understand the emotions of others. Most models and measures of EI comprise variables concerned with the capacity to perceive and understand the emotions of others (e.g., Bar-On, 1997a; Cooper & Sawaf, 1997; Goleman, 1995;1998; 2001b; Mayer & Salovey, 1997; Schutte et al., 1998), although the terms used to denote this area of EI somewhat differ. For example, Goleman (2000b) terms this area of EI, Social Awareness, that is, the capacity to recognise emotions in others, while Bar-On (1997a) describes Empathy as the ability to be aware of, to understand, and to appreciate the feelings of others (p.18). Subscales from the alternative models and measures presented in the current study under this hypothesised dimension included: the Empathy subscale from the EQ-i; and 142
the Emotional Perception (Faces and Pictures) and Understanding (Blends and Changes) subscales from the MSCEIT. As discussed in Chapter Three, the Faces and Pictures tasks from the MSCEIT require participants to view a series of faces (Faces), landscapes and abstract designs (Pictures) and to rate the degree to which specific emotions are present in each. Mayer et al. (2000d) purport that these subscales assess the capacity to perceive both the emotions of others and inner subjective feelings. Citing previous research that has shown that the ability to accurately perceive the emotions of others is related to the ability to accurately perceive inner emotions (e.g., Zuckerman, Lipets, Koivumaki & Rosenthal, 1975; Zuckerman, Hall, DeFrank & Rosenthal, 1977), Mayer et al. (2000d) have argued that the Emotion Perception sub-scales of the MSCEIT also serve as a proxy measure of the ability to perceive inner emotions and feelings states. However, this argument has not yet been empirically substantiated by research. Furthermore, there is some conflicting empirical evidence suggesting that the capacity to perceive inner feelings may be related to, yet sufficiently distinct from the capacity to perceive and understand the emotions of others as measured by tasks such as the Faces and Pictures subscales of the MSCEIT. In exploratory factor analyses of earlier EI tests (and other personality and intelligence variables), Davies, et al. (1998), found that objective measures of emotional perception that involved identifying emotions in faces, colours, music and sounds, all loaded together to define an Emotion Perception factor, however, other measures of emotional perception including the Attention and Clarity sub-scales of the TMMS and the Difficulty Identifying Feelings sub-scale of the TAS-20 loaded on separate dimensions. This finding may have resulted from the different measurement approaches (self-report and performance-based assessment), that is, because the former is assessing actual abilities and the latter is assessing self- reported or perceived abilities (Petrides & 143
Furnham, 2001). However, it may also have resulted because the performance-based tests predominantly assessed the perception of others emotions, or emotions external, that is, in faces, colours, music and sounds, while the self-report measures were predominantly assessing the perception of inner emotions. The battery of EI measures in the Davies et al. (1998) study did not include any self-report measures of the capacity to perceive the emotions of others. Whether individuals capacity to perceive the emotions inherent in faces and pictures provides a proxy index of their capacity to perceive their own inner feelings and emotions is questionable. Indeed Mayer et al. (2000a) have noted that it may be difficult to tap inner emotional experiences with performance-based measures of EI. As such, it is argued that these subscales would define a factor of EI concerned with the perception and understand ing of emotions in others. Other sub-scales that may load on this second hypothesised dimension of EI include the Blends and Changes subscales from the MSCEIT. The Blends task from the MSCEIT requires respondents to identify which emotions combine to form other emotions (e.g., that malice is a combination of envy and aggression). Similarly, the Changes task requires respondents to identify what emotion results from the intensification of another (e.g., that depression often results from the intensification of sadness and fatigue). These subscales are more explicitly concerned with emotional knowledge, that is, an understanding of emotions and the information they convey which has been conceptualised by Mayer and Salovey (1997) as a distinct component of EI. However, in comparison with other models of EI these subscales conceptually correspond with subscales such as Empathy from the EQ-i that, is purported to assess the ability to be aware of and understand the feelings of others. As such, in the current study the Blends and Changes tasks from the MSCEIT have been hypothesised to load on this second common dimension of EI, Perception and Understanding Emotions External. 144
The third hypothesised dimension of EI presented in Table 8.2, and labelled as Emotions Direct Cognition, concerns the capacity to utilise (or reason with) emotions in thought. Salovey and Mayer (1990) originally proposed that emotions can direct reasoning adaptively (e.g., Emotions prioritise thinking by directing attention to important information and can be generated as aids to judgement and memory concerning feelings, p. 11), leading to flexible planning, motivation and creative thinking (Salovey & Mayer, 1990). Moreover, Mayer and Salovey (1997) have proposed that individuals differ in their capacity to incorporate emotional information in thought and to use it to facilitate thinking. Most of the leading measures of EI have subscales concerning either the use of emotions in thought, or the adaptive outcomes of such (e.g., effective problem-solving, flexible decision- making etc), which may load together to form a common dimension of EI. Components of the MSCEIT that assess the use of emotions in thought include the Sensations and Facilitation subscales. Other subscales that assess the use of emotions in thought include the Utilization of Emotions factor of the EI scale by Schutte et al., (1998), and the Externally Oriented Thinking subscale from the TAS-20 (Bagby et al., 1994a,b). The Externally Oriented Thinking subscale from the TAS-20 involves items such as I find the examination of my feelings useful in solving problems (Bagby et al., 1994, p.27). Given the conceptual correspondence between these subscales, they may load with the Sensations and Facilitation subscales from the MSCEIT t form the hypothesized common factor o Emotions Direct Cognition. Other sub-scales that may load on the hypothesised Emotions Direct Cognition dimension of EI include the Problem Solving, Reality Testing and Flexibility sub-scales from the Bar-On EQ- i (Bar-On, 1997a). As
previously discussed, Salovey and Mayer (1990; Mayer & Salovey, 1993; 1997) proposed that emotions can direct cognition adaptively and that the use of emotions in thought may lead to effective problem solving, flexible planning and creativity. 145
Although the items that comprise the Problem Solving, Reality Testing and Flexibility subscales from the Bar-On EQ-i have little to do with the use of emotions in thought, they are purported to assess these potential correlates of such. The next hypothesised common dimension of EI presented in Table 8.2, Intrapersonal Management concerns the capacity to effectively regulate and manage ones own emotions. Most models and subsequent measures of EI have conceptualised the capacity to effectively regulate and manage ones own emotions as a component of the EI construct, although various authors have chosen somewhat different terms for this aspect of EI (Bar-On, 1997a; Cooper & Sawaf, 1997; Goleman, 1995; 1998; 2000b; Mayer & Salovey, 1993; 1997; Salovey & Mayer, 1990; Schutte et al., 1998). Some components of the various measures presented in Table 8.1 explicitly concern the capacity to effectively regulate and manage ones own emotions. For example, the Repair sub-scale from the TMMS (Salovey et al., 1995), has been designed to assess self-perceived beliefs about the ability to repair negative moods and emotions and maintain positive ones. Similarly the Managing Emotions (Self) factor identified within the EI scale by Schutte et al., (1998), comprises items such as I have control over my emotions and When I experience a positive emotion, I know how to make it last (p. 172). Sub-scales from the Bar-On EQ-i that conceptually correspond with the capacity to effectively regulate and manage ones own emo tions include the Stress Tolerance, Impulse Control, Happiness, Optimism, and Self- Regard subscales. Although the Assertiveness and Independence subscales from the EQ- i are not explicitly concerned with the intrapersonal management of emotions, it could be argued that Assertiveness and Independence (the ability to be self-directed and self controlled in ones thinking and actions and to be free of emotional dependency, Bar-On, 1997a, p.18) may be
146
potential correlates of intrapersonal management and thus oad on this hypothesised l dimension of EI. The final hypothesised dimension of EI presented in Table 8.2, Interpersonal Management, concerns the capacity to effectively regulate and manage the emotions of others and in interpersonal relationships. Many of the measures that have been designed to assess EI comprise components concerned with the capacity to regulate and manage the emotions of others and interpersonal relationships. As shown in Table 8.2, these include the Managing Emotions (Others) factor from the Schutte et al., (1998) scale, the Interpersonal Relationship and Social Responsibility subscales from the Bar-On EQ-i (Bar-On, 1997a), and the emotional Management and Relationships subscales from the MSCEIT (Mayer et al., 2000d). The emotional Manage ment subscale from the MSCEIT involves vignettes describing situations where people had to regulate their own emotions and asks respondents to rate the effectiveness of alternate regulation behaviours; Similarly, the Relationships subscale requires respondents to indicate how effective different thoughts and behaviours would be in achieving an interpersonal outcome. As with the Emotional Perception subscales from the MSCEIT, Mayer et al. (2000d) have argued that these subscales measure ones capacity to effectively regulate the emotions of themselves and those of others. That is, that individuals who use more effective regulatory thoughts and behaviours over less effective ones (according to the items of their test), have more insight into effective emotional regulation skills and therefore better regulate and manage their own and others emotions. However, whether the emotional Management and Relationships subscales from the MSCEIT measure an individuals capacity to effectively regulate inner subjective emotional experiences is similarly questionable. Some individuals may recognise more and less effective emotional regulation strategies and thus provide the more correct answers to the items 147
of these MSCEIT subscales, but this may not necessarily provide an index of how effectively they regulate and manage their inner emotional states. There is some preliminary evidence to suggest that an individuals capacity to effectively regulate their own emotions may be related to, yet distinct from, their capacity to effectively regulate and manage the emotions of others. Both the Bar-On EQ-i and the EI scale by Schutte et al. (1998) involve subscales concerned with the capacity to regulate and manage emotions within the self and within others. Factor analyses of these scales have shown that the items that assess these two aspects of emotional regulation and management do define separate factors (Bar-On, 1997a; see Petrides and Furnham, 2000 for factor analyses of the scale by Schutte et al., 1998). Given this evidence, the fact that the Emotional Management and Relationships subscales from the MSCEIT involve assessing the emotional regulation strategies of others, and that it may be difficult to assess inner processes related to emotions with performance-based measures, it is argued that these two subscales would load with other measures to define a common factor (or dimension) concerned with the regulation and management the emotions of others (Interpersonal Management).
8.1.4
Summary
In summary, from a systematic review of the EI literature and a comparison of the variables currently being placed under the banner of the construct, five common dimensions of EI have been theoretically identified as; (1) Emotional Perception and Expression; (2) Understanding Emotions External; (3) Emotions Direct Cognition; (4) Intrapersonal Emotional Management; and (5) Interpersonal Emotional Management, as shown in Table 8.2. As shown in Table 8.2, a higher order general factor of EI was also hypothesised, congruent with other leading theories of EI (e.g., Bar-On, 1997a; Mayer 148
& Salovey, 1993; 1997; Salovey & Mayer, 1990), and the existing research that has found evidence of general factors inherent with the measures assessed (e.g., Bar-On; Mayer et al., 2003). Petrides and Furnham (2001) have recently noted that a complete consensus with regards to what should and should not be a part of the EI construct is unlikely stating that such would be like asking what sports should be in the Olympics; neither question can be answered objectively (p. 428). Indeed the purpose of the present study is not to determine what should and should not be included under the banner of EI, rather, the purpose of the current study is to examine the dimensional communality amongst measures of EI, and on that basis to attempt to provide a comprehensive model of EI that represents the communality amongst the variables currently put forward by alternative models of the construct. Petrides and Furnham (2001) have also noted that the different measurement approaches to EI (self-report and performance-based measures) are likely to produce different results even if the same variables are being assessed as self- report measures assess behaviour tendencies and self-perceived abilities, while performance-based measures assess actual ability. Furthermore, they have argued that the variables assessed by self- report measures highly correlate and may therefore not define distinct factors. However, previous research has shown that measures of EI that comprise a number of quite highly correlated variables (e.g., the average inter-scale correlation of the 15 EQ-i subscales has been shown to be r = .50, Bar-On, 1997a), can define separable factors in a statistical sense (e.g., Bar-On, 1997a; Palmer et al., 2003; Salovey et al., 1995). If Petrides and Furnhams (2001) hypothesis was to be supported, the different measures analysed in the current study may best define two factors that represent the different measurement approaches to EI. That is, all the performance-based sub-scales may define one factor (information processing EI) and the self-report sub-scales may define 149
the other (trait EI), as per their terminology (Petrides & Furnham, 2001). It may also be the case that despite the conceptual communality shown in Table 8.2, that there may not be empirical relationships between the various measures used in the current study, that is, that each test may comprise more specific and error variance with itself than common variance with the test battery. If this were the case a five-factor method variance model may best represent the data where each factor represents each of the tests used in the current study.
8.1.5
Overview of the present study
In order to assess the extent to which the hypothesised five-factor model of EI presented in Table 8.2 represented the communality identified amongst the different models and measures, confirmatory factor analyses were performed. A number of alternative models were also assessed via confirmatory factor analyses to investigate whether the five-factor model best represented the communality amongst the various models and measures. Specifically, the hypothesised five-factor model along with three alternative models were assessed; (1) a two- factor method variance model representing self-reported EI and performance-based EI; (2) a five- factor method variance model were the subscales of each measure load together to form test factors; and (3) a model based on the original conceptualisation of EI by Salovey and Mayer (1990). On the basis of the alternative conceptualisations of EI addressed in the introduction, and the preliminary research findings presented, separate dimensions measuring the perception and management of emotions have been hypothesised (factors 1 and 2, and 4 and 5 respectively as shown in Table 8.2). That is, that the capacities to perceive ones own and others emotions are conceptually related yet distinct dimensions of the construct as are those concerned with the management of emotions. In their seminal article on the 150
construct of EI (and in the later revisions, i.e., Mayer & Salovey, 1997), Salovey and Mayer (1990) argued the case for one factor concerned with the perception of emotions (within the self and others) and for one factor concerned with the regulation and management of emotions (with the self and others). On this basis a three- factor model was also assessed in which factors 1 and 2 in Table 8.2 were collapsed to form an Emotional Perception factor, and factors 4 and 5 were collapsed on each other to form an Emotional Management factor. Three model fit indices were chosen to assess the degree to which these models represented the dimensional communality amongst the measures assessed; the Normed fit index (NFI; Bentler & Bonett, 1980); the Comparitive fit index (CFI; Bentler, 1990), and the Root-Mean-Square Error of Approximation (RMSEA; Steiger, 1990). It was hypothesised that the five- factor model identified by the current study would provide the best degree of fit with the current data thus best representing the dimensional communality amongst the models and measures of EI assessed. The alternative models assessed are summarised and presented in Table 8.3
Table 8.3 Alternative Factor-Models Assessed via SEM Model Three-Factor Model 2-Factor Method (Salovey & Mayer, Variance Model 1990). (1) Emotional (1) Self-Report Recognition and Measured EI Expression (2) Utilisation of (2) Ability Emotions Measured EI (3) Emotional Management
Five-Factor Model
5-Factor Method variance Model (1) MSCEIT
(1) Emotional Recognition and Expression (2) Understanding Emotions External (3) Emotions Direct Cognition (4) Intrapersonal Management (5) Interpersonal Management
(2) Bar-On EQ- i (3) TMMS (4) SEI (5) TAS-20 151
8.2
Method
8.2.2
Participants
The sample originally comprised 399 participants drawn from the general population. However, some participants did not complete the entire test battery and some participants responses were found to be invalid (e.g., random responding) as per the response validity criteria outlined in the technical manual of the Bar-On EQ- i (BarOn, 1997a), and MSCEIT (Mayer et al., 2000d), and in Chapters 3 and 4 respectively. The remainder of the sample, for which all data at the subscale level was complete, comprised 330 participants, 90 males and 238 females (2 unreported), ranging in age from 18 to 78 years old (M = 38.56; SD = 13.72). The sample was drawn via advertisements across the two most populated Australian states (Victoria and New South Wales) and was relatively diverse according to ethnicity and education. The ethnic composition of the sample comprised 229 (71%) White Caucasian Australians, 62 (19%) White Caucasian Emigrants, and 29 (9%) Asian/Pacific Islanders, (3, 1% Other/unreported). Of the 330 participants, 7(2%) reported to have completed primary school education only; 78 (24%), reported to have completed a secondary school education only, 63 (19%), reported to have completed a tertiary certificate/diploma; 92 (28%) reported to have completed an undergraduate degree; and 71 (21%) reported to have completed a postgraduate degree; 19 (6%) unreported. Although there was a gender imbalance in the sample, the demographics of the sample were relatively diverse with respect to age, ethnicity and levels of education.
152
8.2.3
Materials
Emotional Intelligence Research Booklet A test battery research booklet was developed for the purpose of the current study. The booklet comprised (in order), the MSCEIT (Research version 1.1; Mayer et al., 2000d), the Bar-On EQ-i (Bar-On, 1997a), the TAS-20 (Bagby et al., 1994a,b), the TMMS (Salovey et al., 1995) and the SEI by Schutte et al., (1998). More detailed information concerning each test can be found in the materials section of chapters 3 through 7 respectively. Participants responded to the items in the booklet on a scannable answer sheet that was also developed for the current study. As outlined in chapters 3 through 7, each test has been found (with subsets of the present sample) to exhibit relatively good internal consistency reliability (with the exception of some MSCEIT subscales), and factorial validity (with the exception of the Bar-On EQ-i). In addition, similar means, standard deviations, and relationships with age and gender to those reported by the respective test authors and others have typically been found.
8.2.4
Procedure
Participants responding to advertisements about the study collected the test booklets and scannable answer sheets and were briefed about how to complete the various tests. In the briefing sessions, participants were instructed to complete one test in its entirety at a time, in a room and at a time that was free from distraction. Where possible participants were also asked to complete each subsequent test in the same room and at the same time of day. During the briefing sessions the administrator emphasised that the test booklet was to be completed independently without input from others, and participants were given time to examine the instructions to make sure that they were clear about how to complete each test. Upon completion participants returned the test 153
booklets and answer sheets and received a small monetary reward for participating. The study was approved by the Swinburne University Human Research Ethics Committee and informed consent was provided by all participants.
8.3
Results
8.3.2
Scoring
Test pub lisher Multi-Health Systems (MHS) scored the raw data form the BarOn EQ-i and the MSCEIT as outlined in Chapters 3 and 4 respectively. The MSCEIT data was scored according to the MSCEIT Version 2 scoring algorithms as outlined in Chapter 3 according to the American consensus and expert weights. Given the potential advantage of the MSCEIT expert scoring protocol (as discussed in Chapter 3), expert scores for the MSCEIT were used in the factor analyses presented hereafter. The TAS20, TMMS and the SEI were all scored according to their respective scoring procedures, all of which involve the reverse scoring of negatively keyed items and the summing of respective subscale items to produce subscale scores.
8.3.3
Descriptive statistics The means, and standard deviations of the present sample for each test is
presented in Table 8.4.
154
Table 8.4. Descriptive Statistics for the various measures of EI Measure and Subscale Bar-On EQ-i Emotional Self-Awareness Assertiveness Self- Regard Self- Actualisation Independence Empathy Interpersonal Relationship Social Responsibility Problem Solving Reality Testing Flexibility Stress Tolerance Impulse Control Happiness Optimism MSCEIT Faces Pictures Facilitation Sensations Changes Blends Management Relationships TAS-20 Difficulty Identifying Feelings Difficulty Describing Feelings Externally Oriented Thinking TMMS Attention Clarity Repair Self-Report EI Scale (Schutte) Managing Self Emotions Emotional Perception Managing Others Emotions Utilising Emotions 31.00 27.00 12.00 65.00 49.00 30.00 52.31 39.61 23.86 7.41 4.71 3.91 52.88 69.34 59.30 58.83 66.17 63.88 72.73 56.39 128.25 131.38 140.90 140.15 140.06 137.06 142.76 136.37 104.22 96.51 104.36 100.45 101.55 101.09 99.82 97.67 14.35 12.38 14.42 14.31 12.06 13.52 13.07 14.12 Minimum Maximum M SD
60.82 55.52 36.24 50.53 35.15 58.79 42.12 59.62 48.51 52.88 50.80 37.07 50.47 43.04 42.89
131.85 126.33 121.95 121.54 125.55 123.23 126.84 120.48 126.77 128.15 134.14 130.16 127.74 121.99 123.61
101.36 96.47 94.27 98.40 96.62 99.86 98.42 97.93 97.10 94.90 98.61 94.27 98.27 96.81 96.08
16.20 15.59 16.34 15.54 16.26 13.91 16.20 14.04 14.67 15.70 15.23 16.21 14.71 15.61 15.39
7.00 4.00 8.00
35.00 23.00 35.00
13.84 11.10 17.94
6.07 4.57 4.97
19.00 20.00 28.00 9.00
45.00 45.00 54.00 20.00
36.27 34.85 44.28 15.85
4.62 5.28 4.97 2.15
155
As shown in Table 8.4, the means and standard deviations of the present sample were very similar to those that have been reported with the larger samples that have been collected and reported for each test in Chapters 3 through to 7 respectively.
8.3.4
Inter-correlations
Table 8.5 presents the intercorrelations of total scores amongst the various measures, and Table 8.6 presents the intercorrelations amongst the subscales of the different measures.
Table 8.5 Intercorrelations amongst the total scores of the various measures Measure MSCEIT Bar-On TAS-20 TMMS MSCEIT 1 Bar-On .28 1 TAS-20 -.30 -.58 1 TMMS .27 .48 -.36 1 SEI .14 .53 -.33 .66 Note: All the correlations shown are significant at p<.01 SEI
156
Table 8.6 Intercorrelations amongst the subscales of the various measures F F 1 P .27 FA .07 SE .19 CH .11 BL .10 MA .14 ER .18 ES .04 AS -.02 SR .08 SA .06 IN -.05 EM .10 IR .10 RE .11 PS .09 RT .21 FL .13 ST .10 IC .10 HA .19 OP .15 A .07 C .12 R .09 DIF -.16 DDF -.15 EOT -.06 EP .06 MSE .10 MOE .06 UT -.10 P 1 .18 .17 .05 .11 .12 .10 .15 .13 .19 .16 .15 .04 .13 .10 .05 .09 .00 -.07 .05 -.04 .02 .13 .14 -.01 -.14 -.09 -.15 .06 .03 .09 .04 FA SE CH BL MA ER ES AS SR SA IN EM IR
1 .15 1 .10 .15 .14 .07 .06 .12 .08 .18 .12 .12 .02 .06 .06 .03 .13 .09 .01 .07 .05 .15 .04 .14 .10 .11 .05 .06 .09 .02 .00 .04 -.07 .03 .05 .08 -.04 .04 .02 .04 .13 .14 .14 .10 -.01 -.02 -.14 -.08 -.09 -.12 -.15 -.12 .06 .13 .03 -.06 .09 .02 .04 -.02
1 .39 .06 .17 .02 -.01 .00 .00 .00 .04 -.02 -.05 .00 -.01 -.01 .02 .01 -.02 .06 .12 .10 -.03 -.11 -.07 -.14 .08 -.04 .04 .08
1 .07 .15 .11 .09 .02 .14 .08 .21 .08 .08 .04 .01 .07 -.03 .08 .06 .11 .21 .07 .01 -.12 -.16 -.25 .09 -.01 .09 .05
1 .17 .26 .14 .15 .19 .17 .27 .23 .27 .16 .23 .14 .15 .14 .27 .21 .13 .25 .22 -.10 -.17 -.12 .21 .20 .21 .04
1 .27 .12 .14 .24 .11 .26 .21 .27 .19 .10 .14 .10 .11 .16 .17 .17 .11 .12 -.18 -.21 -.19 .20 .10 .14 .08
1 .62 .48 .56 .38 .43 .68 .43 .38 .55 .41 .43 .15 .49 .46 .31 .48 .30 -.46 -.71 -.33 .43 .34 .44 .07
1 .60 .59 .61 .23 .53 .21 .46 .52 .42 .58 .09 .53 .60 .20 .43 .33 -.37 -.56 -.26 .32 .40 .32 .08
1 .66 .50 .23 .54 .26 .41 .58 .46 .60 .27 .67 .68 .09 .38 .46 -.39 -.40 -.18 .27 .47 .28 .04
1 .51 .35 .57 .34 .47 .49 .45 .49 .15 .61 .68 .23 .39 .43 -.39 -.46 -.30 .34 .51 .39 .17
1 .15 .33 .16 .48 .45 .49 .56 .20 .38 .57 .11 .25 .24 -.27 -.39 -.35 .24 .31 .17 .11
1 .59 .75 .35 .26 .30 .23 .24 .37 .37 .27 .32 .27 -.22 -.33 -.22 .41 .26 .44 .16
1 .49 .37 .42 .44 .45 .12 .69 .52 .24 .36 .36 -.35 -.54 -.23 .38 .39 .48 .10
157
Table 8.6 Continued RE RE 1 PS .34 RT .35 FL .24 ST .28 IC .29 HA .37 OP .41 A .10 C .33 R .32 DIF -.25 DDF -.32 EOT -.12 EP .31 MSE .25 MOE .36 UT .03 PS 1 .50 .37 .54 .30 .35 .56 -.01 .29 .32 -.26 -.31 -.24 .26 .40 .24 .13 RT FL ST IC HA OP A C R
DIF DDF EOT
EP
1 .41 .56 .48 .47 .55 .01 .41 .35 -.48 -.43 -.16 .30 .39 .24 -.03
1 .51 .27 .43 .51 .15 .26 .25 -.25 -.36 -.30 .29 .31 .24 .12
1 .32 .53 .75 .01 .36 .43 -.37 -.35 -.14 .37 .50 .23 .11
1 .20 .24 .01 .16 .13 -.29 -.16 -.10 .14 .08 .06 -.07
1 .61 .13 .35 .49 -.39 -.42 -.12 .29 .48 .36 .05
1 .06 .39 .53 -.39 -.38 -.26 .35 .56 .34 .17
1 .34 .12 -.10 -.24 -.31 .40 .20 .45 .26
1 .44 -.34 -.35 -.15 .43 .53 .48 .16
1 -.19 -.20 -.03 .30 .65 .43 .21
1 .58 .39 -.23 -.27 -.19 .02
1 .46 -.34 -.28 -.37 .00
1 -.21 -.12 -.21 -.02
1 .49 .60 .33
Table 8.6 Continued. MSE MOE UT MSE 1 MOE .59 1 UT .37 .35 1 Note: MSCEIT: F = Faces; P = Pictures; FA = Facilitation; SE = Sensations; CH = Changes; BL = Blends; MA = Emotional Management; ER = Emotional Relationships. Bar-On EQ-i: ES = Emotional Self- Awareness; AS = Assertiveness; SR = SelfRegard; SA = Self- Actualisation; IN = Independence; EM = Empathy; IR = Interpersonal Relationship; SR = Social Responsibility; PS = Problem Solving; RT = Reality Testing; FL = Flexibility; ST = Stress Tolerance; IC = Impulse Control; HA = Happiness; OP = Optimism. TMMS: A = Attention; C = Clarity; R = Repair TAS-20: DIF = Difficulty Identifying Feelings; DDF = Difficulty Describing Feelings; EOT = Externally Oriented Thinking. SEI: EP = Emotional Perception; MSE =
158
Managing Self Emotions; MOE = Managing Others Emotions; UT = Utilising Emotions in Thought.
As shown in Table 8.5 there was a positive manifold of significant correlations amongst the various measures assessed (higher scores on the TAS-20 represent lower capacity, thus the notion of a positive manifold). As could be expected the intercorrelations were generally stronger in magnitude between the self-report measures in comparison to those with the performance-based MSCEIT as shown in Table 8.6. Interestingly the MSCEIT was found to correlate most highly with the TAS-20 (r = .30). The relatively stronger correlations between the self- report measures probably reflects not only content overlap, but also some overlap attributable to the common selfreport assessment method of these tests. In general the correlations between the different tests were somewha t lower in magnitude than those reported previously. For example, the correlation between the total score of the Bar-On EQ- i and the TAS-20 was r = -.58, whereas Parker et al., (2001) report a correlation between these two tests of r = -.72. Similarly, in the current study the correlation between the total scores on the Bar-On EQ-i and the TMMS was r = .48 and the correlation between the Bar-On EQ-i and the MSCEIT was r =.24 as shown in Table 8.4, whereas Bar-On (2000) reports correlations between these measures of r = .58 and r = .46 respectively (Mayer et al., 2000d report a total EQ-i MSCEIT correlation of r = .36). The strongest correlation between the various measures was between the SEI and the TMMS (r = .66).
8.3.5
Confirma tory factor analyses were performed to statistically compare the extent to which the different hypothesised factor solutions provided a fit with the present data. 159
Specifically, the five- factor model (both oblique and orthogonal variants), hypothesised by the current study was compared against a general factor model; Salovey and Mayers (1990) original three- factor model (both oblique and orthogonal variants); and the twoand five-factor method variance models as previously discussed. These models were tested via structural equation modelling using AMOS (Arbuckle, 1999). As mentioned in the introduction, three fit indices were chosen to compare the degree of fit for each model; the Normed fit index (NFI; Bentler & Bonett, 1980); the Comparative Fit Index (CFI; Bentler, 1990), and the root- mean-square error of approximation (RMSEA; Steiger, 1990). Table 8.7 lists the fit indices obtained for each model.
Table 8.7 Fit statistics for the hypothesised models assessed Fit 5OB e 5OR f 3OB g 3OR h 2OB i 2OR j 5TOBk Statistic CMIN a 1838.56 2847.73 2206.87 2736.44 2345.03 2369.82 2026.25 Df 453 463 454 464 494 495 486 NFIb .649 .456 .579 .477 .580 .375 .637 c CFI .706 .494 .629 .517 .631 .627 .693 RMSEAd .096 .125 .107 .122 .107 .107 .098 Note: df = Degrees of Freedom. a = Chi-square statistic; b = Normed fit index; c = 5TOR l 2494.17 496 .553 .602 .111
Comparative fit index; d = Root Mean Square Error of Approximation; e = Five- factor higher order model (oblique); f = Five-Factor higher-order model (orthogonal); g = Three=factor higher-order model (oblique); h = Three- factor higher-order model (orthogonal); i = Two- factor method variance model (oblique); j = Two- factor method variance model (orthogonal); k = Five- factor method variance model (oblique); l = Five- factor method (orthogonal).
As can be seen in Table 8.7, none of the hypothesised models fit the present data very well according to all three model fit statistics. None of the CFI or NFI values were 160
above 0.90 and all of the RMSEA values exceeded 0.08 (McDonald & Ringo Ho, 2002). Nonetheless, the fit indices for the hypothesised oblique five- factor higher order model were closest to these recommended values according to all three model fit statistics. To test whether the hypothesised oblique five- factor higher-order model presented in Table 8.8 (with the estimated intercorrelations), provided a statistically better fit with the present data than the next best fitting model (which was the fivefactor method variance test model) the difference in chi square values of these models was calculated as per the procedure outlined by Byrne (2001). According to the difference in chi square values of these models, the hypothesised oblique five-factor model was found to provide a statistically better fit with the present data than the fivefactor method variance test model (i.e., X2 (33) = 2026.25 - 1838.56 = 187.69, p<.01). As such, it could be argued that the hypothesised five- factor model proposed by the current study provided the best fit with the present data. Interesting, the correlation between the oblique two- factor test model representing self-reported and ability EI was r = .39. This finding was consistent with previous research that has examined the relationship between single self-report and performance-based EI m easures (i.e., the Bar-On EQ-i and MSCEIT by Mayer et al., 2000d, as aforementioned, was r = .36).
161
Table 8.8 Hypothesised Five-Factor Higher-Order Model Parameter Estimates (in parentheses) of the Observed Tasks on the Latent Variables, and Estimated Intercorrelations. General emotional intelligence Emotional Recognition and Expression (ERE) ES (.88) Attention (.34) Clarity (.57) EPS (.41) DIF (-.57) DDF (-.78) Understanding Emotions External (UE) EM (.98) Faces (.11) Pictures (.05) Blends (.21) Changes (.04) Emotions Direct Cognition (EDC) Sensations (.08) Facilitation (.06) Utilisation (.16) EOT (-.36) PS (.63) RT (.71) FL (.61) Intrapersonal Management (IM) Repair (.58) MSE (.63) ST (.80) IC (.32) HA (.73) OP (.87) SR (.79) AS (.73) IN (.65) Interpersonal Management (TM) MOE (.56) IR (.76) RE (.70) Management (.32) Relationships (.30)
ERE 1.00 .496 .821 .695 .862
UE 1.00 .486 .376 .892
Estimated Intercorrelations EDC IM
TM
1.00 .958 .764
1.00 .738
1.00
Note: ES = Emotional Self- Awareness, EM = Empathy, PS = Problem Solving, RT = Reality Testing, FL = Flexibility, ST = Stress Tolerance, IC = Impulse Control, HA = Happiness, OP = Optimism, SR = Self- Regard AS = Assertiveness, IN = Independence, IR = Interpersonal Relationship, RE = Social Responsibility, (Bar-On EQ-i). EPS = Emotional Perception, Utilisation, MSE = Managing Self-Emotions, MOE = Managing Others Emotions (SEI). DIF = Difficulty Identifying Feelings and DDF = Difficulty Describing Feelings, EOT = Externally Oriented Thinking (TAS-20). Attention, Clarity, Repair (TMMS). Faces, Pictures, Sensations, Blends, Changes, Facilitation, Management, Relationships (MSCEIT).
162
8.3.6
Exploratory factor analysis
Given that none of the hypothesised models fit the present data very well according to the standard model fit indices (i.e., none of the CFI or NFI values were above 0.90 and all of the RMSEA values exceeded 0.08 as recommended by McDonald & Ringo Ho, 2002), an exploratory factor analysis was conducted in order to seek a more appropriate solution. A Principal Components exploratory factor analysis was conducted where scores on the sub scales of the various measures presented in Table 8.1 were entered as items. A parallel analysis using the procedure provided by OConnor (2000) was conducted to determine the correct number of factors to extract from the data. In the un-rotated solution most of the items loaded on a single factor providing evidence for a general factor of EI consistent with the hypothesised models. The parallel analysis suggested that four factors should be extracted from the data set as shown in Figure 8.1. As such, a number of rotated factor solutions around that which the parallel analysis suggested were examined in order to find a solution that best represented the present data, specifically, three, four, and five oblique and orthogonal factor solutions.
163
10
Eigenvalue
2 PCA 0 1 2 3 4 5 6 7 8 9 Parallel PCA
Eigenvalue number
Eigenvalues: PCA: 10.02, 2.43, 2.10, 1.63, 1.40, 1.28, 1.18, 1.06, .939 Parallel: 1.63, 1.58, 1.49, 1.45, 1.42, 1.36, 1.31, 1.25, 1.20, 1.17
Figure 8.1 Scree plot of eigenvalues derived from the parallel and principal component analysis.
The five-factor oblique (Direct Oblimin) rotated solution was found to best represent the present data. Both the oblique (Direct Oblimin) and orthogonally (Varimax) rotated factor solutions produced highly similar results. However, the oblique rotated factor solution involved a relatively good spread of item loadings across the factors and was the most meaningful to interpret theoretically (e.g., according the different theories of EI as per Bar-On, 1997a, and Mayer & Salovey, 1997), and the 164
confirmatory results where the oblique factor models typically provided better model fit statistics (i.e., closer to the recommended model fit statistic values) than the orthogonal models assessed. The five- factor solution was chosen over three and four factor solutions as a significant amount of variance was explained by the 5th factor (i.e., 4.24%) and the five-factor solution produced a clearer and more interpretable pattern of factor loadings. The pattern of factor loadings for the five- factor oblique rotated solution is presented in Table 8.9.
165
Table 8.9 Pattern of Factor Loadings for the Five-Factor Oblique Rotated Exploratory Solution and Factor Intercorrelations Factors 3 (ER) .138 -.258 -.103 -.158 .168 .171 -.140 .271 .428 .151 -.285 .172 -.103 .130 -.140 .164 -.103 .262 -.277 .125 -.111 .163 .271 -.119 .368 .278 .121 .224 .848 .808 .522 .390 -.217 -.144 .342 .365 .144 .166 -.204 .155 -.188 -.317
Item (subscale) ST OP SR IN AS RT SA PS HA FL R DIF IC CH F BL P SE FA ER UT MOE MSE EP A C RE EM IR MA DDF EOT ES
1 (IM) .861 .827 .824 .775 .734 .716 .659 .625 .620 .586 .471 -.435 .332
2 (UE)
4 (TM)
5 (EE) .110
.235
-.151 .148 .626 .548 .544 .489 .462 .412 .369
.518 .130 -.176 .306
.174
.723 .656 .599 .546 .526 .395 .165
.124 -.142 .219 -.129 -.404
.355 -.367 -.180 .381 IM 1 .11 .18** .39** -.154**
-.110 .208 -.214
.383 Intercorrelations UE ER TM 1 .04 .19** -.14**
-.120 -.266 .131 .633 .632 -.484 EE
IM UE ER TM EE
1 .13* -.14*
1 -.07
166
Note: Factor loadings have been sorted ascending. The highest loading items on each factor are presented in bold face and item loadings <.1 have been omitted. IM = Interpersonal Management, UE = Understanding Emotions External, ER = Emotional Recognition, TM = Interpersonal Management, EE = Emotional Expression; ES = Emotional Self-Awareness, EM = Empathy, PS = Problem Solving, RT = Reality Testing, FL = Flexibility, ST = Stress Tolerance, IC = Impulse Control, HA = Happiness, OP = Optimism, SR = Self- Regard AS = Assertiveness, IN = Independence, IR = Interpersonal Relationship, RE = Social Responsibility, (Bar-On EQ-i); EP = Emotional Perception, UT = Utilisation of Emotions, MSE = Managing Self Emotions, MOE = Managing Others Emotions (SEI). DIF = Difficulty Identifying Feelings and DDF = Difficulty Describing Feelings, EOT = Externally Oriented Thinking (TAS-20). A = Attention, C = Clarity, R = Repair (TMMS). F = Faces, P = Pictures, SE = Sensations, BL = Blends, CH = Changes, FA = Facilitation, MA = Management, ER = Relationships (MSCEIT).
These five factors accounted for 53.3% of the variance (30.4%, 7.4%, 6.3%, 4.9% and 4.2% respectively), in the data set. The first factor that emerged in the exploratory analysis comprised 11 of the 15 EQ-i subscales together with the Repair subscale from the TMMS and the Difficulty Identifying Feelings subscale from the TAS-20. This factor could be interpreted as representing the Bar-On EQ- i given the preponderance of EQ-i subscales loading on this factor. However, it could also be interpreted as the Intrapersonal Management factor of the hypothesised five-factor model proposed by the current study. That is, the highest loading items on this factor were those concerned with the management of ones own emotions (i.e., Stress Tolerance, Optimism, Self-Regard from the Bar-On EQ- i). Moreover, the other
167
subscales that loaded on this factor were also concerned with the management of ones own emotions (i.e., the Repair subscale from the TMMS). The second factor that emerged in the analysis comprised the Changes, Faces, Blends, Pictures, Sensations, Facilitation and Relationships subscales from the MSCEIT, and as such could be interpreted as representing this scale. Another interpretation however, could be that this factor represents that perception and understanding of emotions external (the second factor of the hypothesised five-factor model). It was argued in the introduction of this chapter that these MSCEIT subscales may assess the ability to perceive and understand emotions external as apposed to the ability to perceive and understand inner subjective feelings. More research evidence is needed in order to substantiate this notion, however the current findings are consistent with this assertion. That is, the two highest loading items on this factor were the Changes subscale (that is purported to index individuals understanding of emotions), and the Faces subscale that assess the ability to perceive emotions inherent in pictures of facial expressions. The third factor that emerged in the exploratory factor analysis comprised all of the subscales from the SEI (Schutte et al., 1998) together with the Attention and Clarity subscales from the TMMS. As such this factor could be interpreted as representing the SEI. Alternatively, it could also be interpreted as representing Emotional Recognition, that is, the ability to perceive ones own emotions. Almost one third of the 33 SEI items (9) are concerned with the ability to recognise ones own emotions. Moreover, the other subscales loading on this factor were the Attention and Clarity subscales from the TMMS, which are also concerned with the capacity to recognise ones own emotions, specifically, how much attention individuals pay to their own emotions and how clearly they tend to experience them.
168
The fourth factor that emerged in the exploratory factor analysis comprised the Social Responsibility, Empathy and Interpersonal Relationship subscales from the BarOn EQ-i together with the Management subscale from the MSCEIT. As argued in the introduction of this chapter, all these subscales are concerned with the capacity to regulate or manage the emotions of others. As such, this factor could be interpreted as the Interpersonal Management factor of the hypothesised five- factor model described in the current study. The highest loading items on the fifth factor that emerged from the analysis were the Difficulty Describing Feelings and Externally Oriented Thinking subscales from the TAS-20 together with the Emotional Self-Awareness subscale from the Bar-on EQ- i. This factor could be interpreted as representing the TAS-20, given that it is predominantly defined by the TAS-20 subscales. However, another interpretation may be that this factor could represent that ability to express emotions. The highest loading item on this factor was the Difficulty Describing Feelings subscale from the TAS-20. Moreover the Emotional Self- Awareness subscale from the Bar-On EQ-i, which also loaded on this factor, comprises four items concerned with the expression of emotions such as Its fairly easy for me to express feelings (Bar-On, 1997, p.181).
8.3.7
Results summary
In summary, neither the hypothesised five- factor model nor the five-factor method variance model emerged clearly in the exploratory analyses. While some factors were predominantly defined by one test in the battery (e.g., the first-factor comprised 11 of the 15 EQ-I subscales), in general, the subscales of the various tests did spread over the factors that emerged in the analyses (with the exception of the SEI). On that basis, a second interpretation of the exploratory results is that the dimensional communality amongst the different models and measures of EI may best be described by a general 169
factor and five second order factors that represent (1) Intrapersonal Management; (2) the Perception and Understanding of Emotions External; (3) Emotio nal Perception (within the self); (4) Interpersonal Management; and (5) Emotional Expression.
8.4
Discussion
None of the hypothesised models put forth in the current study were found to provide a statistically acceptable fit with the present data according to the standard model fit statistics. However, the results of the confirmatory factor analyses did present some interesting findings that are worthy of discussion. Firstly, the model fit statistics for the oblique factor models were generally better than those for the orthogonal factor models that were tested. This finding suggests that EI may best be conceptualised as a set of related yet distinct variables (be they abilities, competencies, emotion-related personality traits or otherwise), a finding consistent with Salovey and Mayers (1990) original conception of the construct, and later theories (e.g., Bar-On, 1997a). Indeed there was a positive manifold of correlations (with the exception of the TAS-20 tasks) amongst the various measures assessed, and mo st of the items loaded on a single factor in the un-rotated exploratory factor analysis providing evidence for a general EI factor. As such, it could be concluded that EI may best be conceptualised as a unifactorial construct. Secondly, the hypothesised five-factor model proposed in this Chapter was found to provide a statistically better fit with the present data than the five-factor and two- factor method variance models assessed. This finding suggests that there is some common variance shared between the various models and measures of EI, and that the hypothesised five- factor model may better represent the different approaches to EI (as a definition of the construct), than the theoretical distinctions that have been made 170
between them (e.g., trait and ability EI, Petrides & Furnham, 2001). The correlation between self-reported and ability measured EI (r = .39 as found with the two-factor method variance model), suggests that these approaches share approximately 15% common variance. This result may better represent the communality between self-report and ability measured EI than that which can be made from the correlating one selfreport measure with one ability EI measure (e.g., Bar-On EQ-i and MSCEIT). There is both specific and error variance associated with each test, spuriously lowering the correlation between such tests. A more stringent test of the relationship between these two measurement approaches is to correlate factors that represent self-report and ability measured EI via Structural Equation Modelling as done by the current study. This method controls for the specific and error variance associated with each of the tests used, thus producing a more accurate correlation coefficient (Byrne, 2001), that is, one that reflects the relationship between constructs as apposed to two different tests of a construct. While the distinction between trait (self- report) and ability models and measures of EI may be conductive to conceptually understanding the large literature on EI (Petrides & Furnham, 2001), the findings of the current study suggest that it may be premature to describe these as two fundamentally distinct constructs. Indeed Ciarrochi et als. (2000) conclusion that the different approaches to the conceptualisation and measurement of EI tend to be complementary rather than contradictory (p.540), better reflects the confirmatory findings of the current study pertaining to this issue. The results of the exploratory factor analysis provided further insight into the confirmatory findings. Neither the hypothesised five- factor model nor the five-factor method variance model was clearly supported in the exploratory analyses, although partial support for each was evident. Five- factors were found to best represent the present data in the exploratory analyses that were defined to some extent by each of the various tests. For example, the first factor comprised 11 of the 15 EQ-i subscales, the 171
second factor comprised seven of the eight MSCEIT subscales, the third factor comprised all of the SEI subscales, and the fifth factor comprised 2 of the TAS-20 subscales. This finding suggests that there is a substantial amount of specific and error variance associated with each of the tests involved in the current study. It also provides partial support for the five- factor method variance test model, which was one of the better fitting models assessed in the confirmatory analyses. However, none of the tests clearly defined the five- factors interpreted in the exploratory analyses with the exception of the second factor that loaded all of the SEI subscales (although the Attention and Clarity subscales from the TMMS also loaded on this factor). In contrast, the subscales from each test did spread to a certain extent over the five- factors. A second interpretation of the exploratory results based on the spread of the subscales over the five- factors was that the first factor represented Intrapersonal Management, that is, the ability to effectively regulate and manage ones own emotions; the second factor represented Emotional Perception, that is, the ability to perceive ones own emotions; the third factor represented Interpersonal Management, that is, the ability to effectively regulate and manage the emotions of others; the fourth factor represented Emotional Expression, that is, the ability to express emotions; and finally, that the fifth factor represented Understanding Emotions, that is, the ability to perceive and understand the emotions of others. This interpretation is somewhat consistent with the hypothesised five- factor model proposed in the current study. However, in the hypothesised model, variables to do with the perception and expression of ones own emotions were expected to form a single Emotional Perception and Expression factor; and a factor to do with the use of emotions in thought, (Emotions Direct Cognition, that did not emerge in the exploratory results) was expected. This second interpretation of the exploratory solution provides partial support for the hypothesised five- factor model, which, as aforementioned, was found to be the best fitting model in the confirmatory 172
analyses. However, the exploratory results also highlight some of the potential reasons as to why this model may not have provided a statistically acceptable fit with the present data. A systematic comparison of the variables (by definition) measured by the different tests of EI assessed by the current study lead to a theoretically derived fivefactor general taxonomy for EI. Partial support for this hypothesised model was found by the current study and the findings of both the confirmatory and exploratory analyses can inform future research in the area. One of the hypothesised factors that did not emerge in the exploratory factor analysis was the Emotions Direct Cognition factor, which was expected to load subscales from the various measures concerned with the capacity to utilise or reason with emotions in thought. A possible reason as to why this factor did not emerge may concern the reliability of the various subscales that measure this aspect of EI. Most of the subscales hypothesised to define this common factor have been found to exhibit relatively low internal consistency coefficients in comparison to the other variables assessed. The subscales from the MSCEIT, the SEI and the TAS-20 that were hypothesised to comprise this factor were all found to exhibit reliability coefficients below the criterion of a = .70 (Tibachnick & Fidell, 1996), as shown in Chapters 3, 6 and 7 respectively. As such, it could be argued that this area of EI is not reliably assessed by existing measures of EI. As a result, while Emotions Direct Cognition (or the use of emotions in thought) can be theoretically identified as a common facet of EI amongst various models and measures of the construct, it may not yet manifest empirically in factor analyses. Future research (following advances in the measurement of this area of EI) is needed in order to substantiate whether it empirically represents a common dimension of the EI construct. Another factor that was interpreted from the exploratory results of the current study concerned the ability to express emotions, the highest loading subscale being the 173
Difficulty Describing Feelings subscale from the TAS-20. As mentioned in the introduction of this chapter, in their original conception of EI Salovey and Mayer (1990) proposed that individuals might differ in the capacity to express feelings. They also proposed however, that the ability to express inner feelings and emotions would be highly related if not dependent on the capacity to perceive feelings conceptualising the ability to appraise and express emotions as a single component of the construct. Indeed one cannot imagine being able to express feelings accurately without them being first perceived. The findings of the current study however, suggest that these two elements may be conceptually related yet sufficiently distinct components of the construct, a finding consistent with the three- factor model of alexithymia that involves separate factors concerned with difficulty in identifying and describing feelings (Bagby et al., 1994a). A hierarchical model involving the ability to perceive emotions followed by a separate factor concerned with the ability to express emotions would allow for the theorising of Salovey and Mayer (1990), however, such a model would also take into account the findings of the current study (i.e., a distinct expression factor). As with the potential Emotions Direct Cognition factor, further advances in the measurement of the ability to express emotio ns is also needed. While most models of EI comprise the ability to express emotions as a part of EI, this area has been somewhat neglected by those who have designed measures of the construct (Petrides & Furnham, 2001). Neither the MSCEIT, the Bar-On EQ-i, or the EI scale by Schutte et al. (1998) have sub-scales solely concerned with the expression of emotions. For example, the EI scale by Schutte et al. (1998) has only two items that assess the capacity to express emotions that are subsumed by the Emotion Perception subscale. Furthermore, the EQ-i comprises only four items concerned with the expression of emotions (e.g., Its fairly easy for me to express feelings; Bar-On, 1997, p.181), which are subsumed by the Emotional Self- Awareness subscale. While a distinct Emotional Expression factor was 174
interpreted from the exploratory results of the current study, this factor may not emerge in other analyses because it is not adequately represented by existing measures of EI. Further replicative research (possibly following the development of better measures of emotional expression) is needed in order to substantiate whether Emotional Expression represents a separate definitive dimension in a common taxonomy of EI.
8.4.2
Conclusion
A five- factor taxonomic model for EI was theoretically derived from a systematic comparison of the variables assessed by the different models and measures utilised in the current study. While this model was not found to provide a statistically acceptable fit with the present data according to standard model fit criteria, it was found to be the best fitting model in comparison to a number of others assessed including; a five- factor method variance (or test) model representing the different measures assessed; a three factor model representing Salovey and Mayers (1990) original conceptualisation of EI; and a two- factor method variance model representing ability and trait (or performance-based and self- report measured) EI. As discussed by McDonald and Ringo Ho (2002), conclusions drawn on the basis of SEM results should not be purely data-driven. While model misfit can be taken to imply that the hypothesised model is not supported by the data, there are a number of unresolved problems (as outlined by McDonald & Ringo Ho, 2002) with criterion indices of model fit. McDonald and Ringo Ho (2002) recommend that competing models should be specified a priori (as done by the current study), and the relative goodness of fit reported. Accordingly, it was stated in the introduction of this chapter that the model that provided the best degree of fit with the present data would be taken to best represent the dimensional communality amongst the models and measures of EI 175
assessed. As such, given the theoretical justification for the hypothesised five-factor model, and the fact that it was the best fitting model in comparison to the others assessed, it is concluded that the five- factor model best represents the communality amongst the various measures of EI assessed and therefore provides a taxonomy for EI. The hypothesised five- factor model for EI may not have provided a statistically significant fit with the present data for a number of reasons including; (1) the reliability (or lack thereof) of subscales measuring the use of emotions in thought (Emotions Direct Cognition, which did not emerge in the exploratory factor analytic results); (2) the considerable amount of specific and error variance associated with each measure as evidenced by the exploratory findings were each measure roughly defined distinct test factors; and (3) the possibility that the first factor of the hypothesised model Emotional Perception and Expression, should be split into two factors representing the ability to perceive emotions, and the ability to express emotions as interpreted from the exploratory findings. Further support for the five- factor taxonomic model of EI
hypothesised by the current study may only be found following further advancements in the measurement of particular facets of the construct (such as the use of emotio ns in thought) and the establishment of more comprehensive measures of others (such as the expression of emotions). As with any set of multivariate data there will almost always be more than one plausible structural model (McDonald & Ringo Ho, 2002). The results of the exploratory factor analysis (and others particularly those pertaining to the TAS-20 Bagby et al., 1994 as previously discussed), suggest that another plausible taxonomic model for EI might similarly involve five-factors, however, that these five factors might include by definition; Emotional Perception (the ability to perceive emotions within the self); Emotional Expression, (the ability to accurately express ones own emotions); Understanding Emotions External (the ability to perceive and understand the emotions 176
of others and those inherent in external environments); Intrapersonal Management (the ability to effectively regulate and manage inner subjective feelings); and Interpersonal Management (the ability to effectively regulate and manage the emotions of others). This model of EI should be assessed by future research along with other theoretically justified taxonomies for EI, and the relatively goodness of fit should similarly be examined in order to substantiate whether the taxonomic model identified by the current study best describes the communality amongst different models and measures of the construct. A more comprehensive battery of measures may also be needed to better identify a general taxonomy for EI, that is, one that is more representative of the various models and measures currently being placed under the banner of the construct. Such a battery might include, for example, the Emotional Competency Inventory (Boyatzis et al., 2000) and the newly developed Trait Emotional Intelligence Questionnaire (TEIQue; Petrides & Furnham, 2003) that had not been released at the time the current study was conducted. The findings of the current study suggest that there are both conceptual and empirical relationships between different models and m easures of EI. The correlation between self-report and ability measured EI suggests that individuals with greater EI abilities as measured by the MSCEIT (Mayer et al., 2003), also tend to report greater EI and visa versa. Furthermore, the current findings suggest that while categorical distinctions between various models such as mixed/ trait and ability / information processing EI are useful in organising the various approaches to the construct, that these two categories should not be considered fundamentally distinct constructs. In contrast, it could be concluded that existing models of EI tend to compliment, rather than contradict one another (Ciarrochi et al., 2000), and that further research toward establishing a taxonomy for EI is thus warranted.
177
The findings of the current study suggest that a taxonomy for EI will most likely comprise a general factor represented by a number of related facets. While not distracting from more specific models and measures of EI, such a taxonomy would be useful in that it would provide a common definition and understanding about the nature of the construct. Based on the findings of the current study EI can be commonly defined as a conceptually related set of abilities to do with ones own and others emotions, specifically; the ability to perceive and express ones own emotions; the ability to perceive and understand the emotions of others; the ability to allow emotions to direct ones reasoning; the ability to manage ones own emotions; and the ability to manage the emotions of others.
178
Chapter 9; Conclusions
CHAPTER 9
Conclusions, Limitations, and Directions for Future Research
9.1
Overview
This dissertation has provided Australian normative data for the Bar-On EQ-i (Bar-On, 1997a), the MSCEIT (Mayer et al., 2003), the TMMS (Salovey et al., 1995), the TAS-20 (Bagby et al., 1994a,b), and the SEI (Schutte et al., 1998). In addition it has examined the reliability and factor structure of these measures, and their respective relationships with a number of demographic variables such as age, gender, relationship status, and history of mental illness. Finally, the dissertation has examined the dimensional communality amongst these five measures and has proposed a common definition and taxonomic model for EI. In this Chapter, some conclusions about the measurement of EI in these main areas are made, limitations of the research are discussed, and recommendations for future research are made.
9.2
Conclusions
9.2.2
Reliability
Internal reliability is an important psychometric test characteristic that evaluates the extent to which items in a test are coherently measuring the variable(s) assessed. Davies et al. (1998) found that many of the early measures of EI (both ability and self-report) lacked good internal reliability and concluded that this psychometric
179
property of EI measures needed improvement. The results of this dissertation suggest that the internal reliability of the more recently developed measures of EI have improved markedly over earlier tests (e.g., the Bar-On EQ-i and the MSCEIT). All the measures examined by the current series of studies yielded internal reliability coefficients above a = .80 at the full-scale level, and between a = .80 and a = .90 at the subscale level. However, the internal reliability of the subscales measuring with the ability to utilise or reason with emotions in thought were low (i.e., the Utilisation factor of the SEI, the Externally Oriented Thinking subscale of the TAS-20, and the Sensations and Facilitations subscales from the MSCEIT). Subscales measuring this facet of EI yielded internal reliability coefficients below the a = .70 criterion (Tabachnick, & Fidell, 1996). As such, it could be argued that this facet of EI (which was identified in Chapter 8 as a common dimension of EI) is not reliably assessed by existing measures and that more reliable measures of the capacity to utilise or reason with emotions in thought need to be developed in future research. In summary, it is concluded that the internal reliability of EI measures has improved over earlier measures.
9.2.3
Factor Structure
Factor analysis provides a means by which to assess the extent to which the structure of a scale is consistent with the underlying theory and number of facets that it measures. Both exploratory and confirmatory factor analytic methodologies were employed in the series of studies reported in the thesis in order to assess the factor structure of the measures examined. Where previous independent research had shown that the factor structure of the test was consistent with the underlying theory and factor
180
analyses conducted by the test authors (or others), confirmatory factor analyses were computed to further substantiate or replicate these findings (e.g., TAS-20, the SEI and the MSCEIT). Where no previous independent analyses had been conducted (e.g., the TMMS), or where the previous methodologies or results were unclear (as with the EQi), exploratory factor analyses were conducted in order to examine the factor structure of the test. With the exception of the EQ-i, the findings of these factor analyses were consistent with the underlying theories of the various measures examined and their respective previous research findings. The factor analytic results showed evidence for the four abilities purported to be measured by the MSCEIT, the three components of the TMMS and the TAS-20, and the four factors of the SEI as previously found by Petrides and Furnham (2000), and Ciarrochi et al. (2001; 2002). The dimensional structure of the EQ-i was not found to be consistent with the underlying theory it has been purported to measure or the factor analytic results of Bar-On (1997a). However, as mentioned in Chapter 4, differences in factor analytic findings can result from many different parameters (such as sample size and type), and further factor analytic research findings are need in order to clarify the dimensional structure of the EQ- i. In general, the factor structure of the different EI measures examined in this dissertation were found to be robust and consistent with their respective underlying theories. An important next step for research on EI is to examine the extent to which the factor structure of EI measures can be located in different cultures. So far research has shown that the factor structure of some measures of EI (e.g., the MSCEIT, TMMS, SEI,) are consistent across similar Western cultures (e.g., the United States, the United Kingdom and Australia). For example, the four factors of the MSCEIT have been found with North American and Australian population samples (Mayer et al., 2003; and
181
the findings presented in Chapter 3). Similarly the three factors of the TMMS and the four-factors of the SEI have also been found with North American, British and Australian samples (Ciarrochi et al., 2001; 2002; Palmer et al., In Press; Petrides & Furnham 2000; Salovey et al., 1995). Arguably, this research suggests that the facets of EI are universal variables within Western cultures. Clearly there are many important questions for cross cultural research to answer. For example, are Americans more emotionally intelligent than Australians? Are there universal age and gender differences in EI etc? However, research has yet to examine whether the facets of EI are universal variables across different cultures. While it has been established that basic emotions are expressed and recognised in similar ways across cultures, there is an emerging body of research also showing cultural differences in emotional processes (Matsumoto, 1996). Cultural differences in emotional processes are being identified in three domains. First, cultural differences in emotion display rules (ho w emotions are expressed) have been identified. For example, Matsumoto et al. (1998) have shown that Japanese express anger to a greater extent with people of a lower status than Americans, whereas Americans express disgust and sadness more freely to intimate friends and family than Japanese. Second, cultural differences in the intensity of emotional expressions have also been shown. For example, research has found cultural differences in the verbalisation of emotions, and non-verbal expression of emotions (in particular the expression of anger by men) across countries such as Greece and the United Kingdom, where the verbal and nonverbal expression of emotions is higher in former (Edelmann et al., 1989; Scherer & Wallbott, 1994). Finally, and perhaps most relevant for the construct of EI, research has found cultural differences in the ability to accurately identify emotions. For example, Matsumoto (1991) has found that the recognition accuracies for negative emotions such
182
as fear, anger and disgust were higher among Americans than Japanese, whereas the accuracy with which Americans and Japanese identified happiness and surprise were similar. Collectively this research suggests that there may be cultural variability in some of the emotional processes underlying certain facets of EI and that there may need to be some adjustment in the conceptualisation and operationalization of EI for certain cultures. At present, cross-cultural research studies on EI are rare and whether facets of EI are universal variables across cultures is unknown. As measures of EI are translated into different languages it will be important to examine the stability of the factor structure of EI measures across cultures, and to examine whether the same variables can be located in different cultures and whether the associations among facets of EI are similarly structured.
9.2.4
Dimensional Communality
From the outset of the thesis (see Chapter 2), it was noted that while a number of leading authors in the area had compared and contrasted various approaches to the conceptualisation and measurement of EI (e.g., Petrides & Furnham, 2000; 2001; Mayer et al., 2000a,b,c), very little research had examined the relationships and potential communality amongst them. It was argued that a taxonomic model for EI would be useful in that it would provide a common definition for EI and identify the more definitive dimensions of the construct much like the comprehensive taxonomy of personality traits, the widely known Five-Factor Model (FFM; Digman, 1990; Costa & McCrae, 1992a). Toward this goal a criterion for what might constitute a common facet of EI was established and a five- factor taxonomic model for EI was hypothesised from
183
a systematic review of the variables (by definition) assessed by the five different measures of EI involved in the study. Confirmatory and exploratory factor analyses were conducted to test this hypothesis against a number of other plausible models and to assess the communality amongst the MSCEIT, EQ-i, TMMS, SEI and TAS-20. While none of the models assessed were found to fit the data according to standard model fit statistics, the hypothesised five- factor taxonomic model was found to be a better fitting model than a five-factor method variance model and a two-factor method variance model representing self- report and ability measured EI (as two separate constructs). It was argued that this finding demonstrated that there was some communality amongst various approaches to the conceptualisation and measurement of EI and that although the categorical distinctions between various approaches were useful in organising the literature in the area, that these did not appear as two distinct constructs (e.g., trait and informational processing). On the basis of these findings it was concluded that the hypothesised taxonomic model of EI best represented the dimensional communality amongst the various models and measures of EI assessed, and that EI could be therefore commonly defined as a conceptually related set of abilities to do with ones own and others emotions. Specifically, (1) the ability to perceive and express ones own emotions; (2) the ability to perceive and understand the emotions of others; (3) the ability to allow emotions to direct ones reasoning; (4) the ability to manage ones own emotions; and (5) the ability to manage the emotions of others. This taxonomic model of EI is similar to that proposed by Goleman (2001b), who hypothesised that EI could be commonly defined as the ability to recognise and regulate emotions in oneself and others. However, the model proposed by Goleman (2001b) does not included the capacity to utilise or reason with emotions in thought, which was theoretically identified in Chapter 8 as a common facet of EI models and
184
measures. The taxonomic model for EI identified in Chapter 8 is also similar to Mayer and Saloveys (1997) four- factor ability model of EI. However, it differs in that it involves distinct facets to do with the perception of ones own and others emotions rather than a single Emotional Perception factor, and distinct facets to do with the management of ones own and others emotions rather than a single Emotional Management facet. It was argued in Chapter 8 that the ability to perceive ones own and others emotions, and the ability to manage ones own and others emotions might be conceptually related yet empirically distinct dimensions of the EI construct respectively. The findings of the current study support this later assertion, (i.e., the five- factor taxonomic model was found to be a better fitting model than a three factor model involving single Emotional Perception and Emotional Management factors as per Mayer & Salovey, 1997) however, this finding needs to be further examined in future research. Indeed, there are a number of methodological issues and limitations with the large factor analytic study presented in Chapter 8 that need to be addressed by future research, and the five-factor taxonomic model of EI proposed in the thesis needs to be replicated before it can be concluded as best represent ing the dimensional communality amongst various models and measures of EI.
9.3 Limitations
There are a number of methodological issues with the large factor analytic study presented in Chapter 8 that need to be taken into consideration when interpreting the results and addressed by future replicative research. The first limitation concerns the battery of measures used to determine the taxonomic model of EI. The current study utilised a battery of measures that covered both ability and mixed models of EI
185
(Mayer et al., 2000a), however, as previously discussed, it did not entirely cover the breadth of variables currently being placed under the banner of EI. In particular, the current battery could have been better represented by some of the broader models and measures of EI currently developed such as the Emotional Competency Inventory (Boyatzis et al., 2000). Future research should utilise a battery of EI measures that is as representative as possible of the various approaches to EI. At present only one performance-based measure of EI, the MSCEIT (Mayer et al., 2000d) has been developed and the current research findings are also limited in this context. More conclusive research findings will follow the development and use of further performance-based measures of EI. A second limitation of the current study concerns the method of administration of the test battery. In the current study participants were handed test booklets and response sheets and asked to complete the tests at home. As such, the conditions in which the test battery was completed were not well controlled. While briefing sessions were conducted to emphasis optimal conditions for completing the test battery (e.g., in a quiet setting that is free from distractions and sources of bias) and the testing procedure (e.g., to complete each test in its entirety and independently), the extent to which the participants adhered to such instructions was not supervised. As such, it is possible that some participants may not have completed the test battery in appropriate test conditions and external factors may have impeded on responses (e.g., time of day, whether tests were started and finished on separate days at different times etc). A more optimal approach would have been to conduct group-testing sessions where the test conditions could be better controlled and participants could have been supervised during the completion of the tests. Where possible, future research of this nature should
186
administer an EI test battery in groups and control the testing cond itions for all participants. The current study was also limited by the psychometric properties of some of the tests used and the nature and scope of existing measurement tools. For example, subscales to do with the use of emotions in thought were found to have relatively low reliability in comparison to others. Thus, while a common factor to do with the use of emotions in thought could be theoretically identified according to the stipulated criterion, it did not emerge in the exploratory analyses. If more reliable measures of this commonly proposed facet of the construct are developed, a factor to do with the use of emotions in thought may emerge in future analyses examining the communality amongst various measures of EI. Similarly, while the capacity to express emotions is commonly proposed as a facet of EI (e.g., Bar-On, 1997a; Mayer & Salovey, 1997), this aspect of EI is not adequately assessed by existing measures (Petrides & Furnham, 2001). As such emotional expression may not emerge as a common facet of EI in future research examining the communality amongst various measures simply because it is not adequately represented. Finally, as previously mentioned, the MSCEIT is the only performance-based measure of EI currently available. As such the battery of measures used by the current study was heavily biased with self-report measures. While the correlation between self- report and ability measured EI (as found with the two- factor method variance model, i.e., r = .39) was discussed as best representing the relationship between these measurement approaches, more conclusive findings will only be gained from examining the relationship between factors that are equally represented by these two measurement approaches. A final limitation worth discussion concerns the gender imbalance in the sample that was used to establish the taxonomic model of EI presented in Chapter 8. To-date
187
research on EI in general, appears to have assumed that measures of EI have measurement invariance across gender. That is, research has been reporting gender differences in the mean scores on EI measures (e.g., Bar-On, 1997a; Ciarrochi et al., 2000; Palmer et al., 2002; Mayer et al., 1999) without first examining whether the numerical scores on the measure of EI are comparable for males and females. As such, these findings are potentially misleading unless the test (used to determine gender differences) has measurement invariance. Indeed while gender differences were found with EQ- i and the MSCEIT in Chapter 4 and Chapter 3 respectively, research has yet to establish whether the empirical relationships between the items on these tests and the dimensions of EI respectively assessed are invariant for males and females. As such the findings concerning gender differences in the current series of studies (and others, e.g., Ciarrochi et al., 2000; Mayer et al., 1999) should be interpreted with caution. The primary approach to addressing measurement invariance involves the study of similarities and differences in the covariance patterns of item- factor relations. (Windle, Iwawaki & Lerner, 1988, p.551). Confirmatory factor analysis and Item Response Theory (IRT) are the most contemporary and substantive methods currently available to address measurement invariance and are widely used in large-scale achievement testing programs (Reise, Widaman & Pugh, 1993). However, all these methods require a sample size of five times the number of variables (or items) being assessed (Byrne, 2001). As such, in the current series of studies there were not enough male participants in the sample (i.e., there was not five times the number of males per variable measured) to conduct measurement invariance analyses (e.g., separate Exploratory Factor Analyses for males and females, CFA or IRT) thus measurement invariance was not addressed.
188
Establishing measurement invariance of a test across distinct groups is critical in determining the universality of the variables being measured, the validity of group differences, and the pattern of correlations of the test with external variables (Reise et al., 1993). It could be argued that the findings pertaining to the taxonomic model of EI hypothesised and assessed in Chapter 8 may be biased by the gender imbalance in the sample and may therefore be more representative for females than males. Future research needs to not only to replicate the findings of the current study but also to establish whether the taxonomic model of EI is invariant for males and females. Furthermore, the universality of the variables comprised by the taxono mic model of EI also need to be assessed across cultures and other distinct criterion groups.
9.4 Directions for future research
This thesis employed a factor analytic approach to establish the dimensional commonality amongst various models and measures of EI and has proposed a taxonomic model for EI on that basis. This model was established on, and reflects, a criterion for what might constitute a common dimension of EI (that is, a common dimension of EI is one that comprises variables that are shared by at least two of the main theoretical models and measures of the construct). Future research could either utilise this criterion in hypothesising alternative taxonomic models of EI or could substantiate other possible criteria from which to theoretically derive taxonomic models of the construct. Indeed future research should assess the hypothesised fivefactor taxonomic model of EI against other alternative taxonomic models. The construct of EI is still very much in its infancy in comparison to constructs such as personality and intelligence (IQ), and much more research is needed in order to
189
substantiate the nature and boundaries of the construct (Roberts et al., 2001). This thesis has focused on the measurement of EI and the findings of the thesis highlight a number of areas where better measures of the construct are needed. As previously mentioned the ability to express emotions can be systematically identified as a common facet amongst various models of the construct yet this area of EI is under represented and not well assessed by existing measures. While there are a number of items in selfreport measures of EI (e.g., the EQ- i) pertaining to the capacity to express emotions only the TAS-20 has a subscale solely concerned with the capacity (or lack thereof) to express emotions. Furthermore, although Mayer and Salovey (1997; Mayer et al., 1999) have argued that the ability to perceive emotions (in faces, landscapes and designs) serves as a proxy measure of the capacity to express emotions, this notion has not yet been substantiated by research. At present there are no objective or
performance-based measures of the capacity to express emotions and it is difficult to hypothesize what such a measure might entail, however, an objective measure of this capacity may greatly contribute to the area. The capacity to utilise or reason with emotions in thought is another area of EI where better measures are needed. As previously discussed, existing measures of this facet of EI are not internal reliable. As originally hypothesised by Salovey and Mayer (1990) the capacity to utilise or reason with emotions in thought may be a determinant of individual differences in a number of important variables such as flexible planning and creative thinking. As such, this is a potentially valuable facet of EI and more reliable measures of the capacity to utilise emotions in thought may also greatly contribute to the area. Future research should also attempt to devise alternative performance-based measures of EI. The MSCEIT and previous performance-based measures of EI developed by Mayer and colleagues (Mayer et al., 1990; 1999; Mayer & Geher, 1996)
190
are all static in nature (i.e., participants view static pictures, read vignettes, etc). The advent of computer based testing mediums provides the opportunity to devise more dynamic measures of the ability to process emotions and emotional information. More dynamic measures of the ability to perceive others emotions may for example, involve participants listening to sound files of different tones of voice or watching short movie clips of people interacting where the task may be to determine different body language and facial expressions. In this context measures of the ability to manage others emotions may similarly involve determining more and less effective emotional management strategies after watching several alternatives play out in short movie clips. Computer based testing mediums (e.g., Web or CDROM) may provide an alternative medium through which to assess emotional capacities and in particular may provide the opportunity to create more dynamic measures of the construct. Finally, as previously mentioned, this thesis employed a factor analytic approach in establishing the dimensional commonality amongst various models and measures of EI. Future research should explore alternative means of establishing the nature and boundaries of EI. An understanding of the biological basis of EI and differences therein (e.g., brain structures and genes) that manifest individual differences in EI may be particularly valuable in terms of defining the construct. Whilst the research agenda examining the brain mechanisms and basis of emotions is well developed (LeDoux, 1998), research has not yet focused explicitly on the potential biological mechanisms and basis of EI (Mayer & Salovey, 1993). The factor analytic results of the thesis suggest that EI is commonly defined as involving the capacity to perceive and express ones own emotions; the capacity to perceive and understand the emotions of others; the capacity to utilise or reason with emotions in thought; the capacity to effectively manage ones own emotions; and the
191
capacity to effectively manage the emotions of others. If this taxonomic model for EI is further established by future research an important next step would be to devise valid and reliable measures that assess these common dimensions of EI. Here it could be argued that it may be useful to devise both self- report and performance-based measures of the model. It has been argued that it may be difficult to tap inner process to do with emotions with performance-based measures (Mayer et al., 2000c). Indeed, it is questionable whether insight is gained to an individuals ability to manage inner subjective moods and emotions from their scores on scales that assess their ability to decipher more and less correct emotional management strategies in vignettes. As such, it could be argued that these different approaches to the measurement of EI (self- report and performance-based) may complement one another and both these mediums of measurement should be further researched and developed.
192
References
References: The National Survey of Mental Health and Wellbeing of Adults (SMHWB). (1997). Australian Bureau of Statistics, www.abs.gov.au Amato, P.R. (1993). Childrens adjustment to divorce: Theories, hypotheses and empirical support. Journal of Marriage and the Family, 55, 23-38. Arbuckle, J. L. (1999). Amos 4.01. Chicago: SmallWaters Corp. Bagby, R.M., Parker, J.D. & Taylor, G.J. (1994b). The Twenty-Item Toronto Alexithymia Scale TAS-20: Item selection and cross-validation of the factor structure. Journal of Psychosomatic Research, 38, 23-32. Bagby, R.M., Taylor, G.J., & Parker, J.D.A., (1994a). The Twenty-Item Toronto Alexithymia Scale: Part II, Convergent, discriminant, and concurrent validity. Journal of Psychosomatic Research, 38, 33-40. Bar-On, R. (1997a). Emotional Quotient Inventory: Technical manual. Toronto: MultiHealth Systems. Bar-On, R. (1997b). Development of the Bar-On EQ- i: A measure of emotional and social intelligence. Paper presented at the 105th annual convention of the American Psychological Association, Chicago, U.S.A. Bar-On, R. (2000). Emotional and social intelligence: Insights from the Emotional Quotient Inventory. In R. Bar-On & J.D.A., Parker, (Eds). The handbook of emotional intelligence. (pp.363-388) San Francisco: Jossey-Bass Inc. Bar-On, R. Brown, J.M., Kirkcaldy, B.D., & Thome, E.P. (2000). Emotional expression and implications for occupational stress; an application of the Emotional Quotient Inventory 0). Barrett, P., & Kline, P. (1982). Factor extraction: an examination of three methods. Journal of Personality and Group Behaviour, 2, 94-98.
193
References
Beck, A.T., & Steer, R.A. (1987). Beck Depression Inventory Manual. San Antonio, TX: The Psychological Corporation. Bentler, P.M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238-246. Berry, D.S., & Willingham, J.K. (1997). Affective traits, responses to conflict and satisfaction in romantic relationships. Journal of Research in Personality, 31, 564-576. Boyatzis, R.E., Goleman, D., & Rhee, K.S. (2000). Clustering competence in emotional intelligence: Insights from the Emotional Competency Inventory. In R. Bar-On & J.D.A., Parker, (Eds). The handbook of emotional intelligence. (pp.343-362) San Francisco: Jossey- Bass Inc. Brody, L.R. (1999). Gender, emotion and the family. Boston: Harvard University Press Byrne, B. M. (2001). Structural Equation Modelling with AMOS: basic concepts, applications, and programming. Manwah, New Jersey: Lawrence Erlbaum Associates. Carroll, J.B. (1993). Human cognitive abilities: A survey of factor-analytic studies. New York: Cambridge University Press. Catanzaro, S.J., & Mearns, J. (1990). Measuring generalized expectancies for negative mood regulation: Initial scale development and implications. Journal of Personality Assessment, 54, 546-563. Cattell, R.B. (1978). The Scientific Use of Factor Analysis. New York: Plenum. Cattell, R.B., Cattell, A.K., & Cattell, H.E. (1993). Sixteen personality factor questionnaire (5th ed.). Champaign, IL: Institute for Personality and Ability Testing.
194
References
Cattell, R.B. and Kline, P. (1977). The Scientific Analysis of Personality and Motivation. London: Academic Press. Ciarrochi, J., Chan, A.Y.C., & Bajgar, J. (2001). Measuring emotional intelligence in adolescents. Personality and Individual Differences, 31, 1105-1119. Ciarrochi, J.V., Chan, A.Y.C., & Caputi, P. (2000). A critical evaluation of the emotional intelligence construct. Personality and Individual Differences, 28, 539-561. Ciarrochi, J., Deane, F.P., & Anderson, S. (2002). Emotional intelligence moderates the relationship between stress and mental health. Personality and Individual Differences, 32, 197-209. Cohen, J. (1988). Statistical Power Analysis for the Behavioural Sciences. Hillsdale, NJ: Erlbaum. 2nd ed. Cooper, R.K. & Sawaf, A. (1997). Executive EQ: Emotional intelligence in leadership and organizations. New York: Grosset/Putnum. Costa, P.T., Jr., & McCrae, R.R. (1992a). Normal personality assessment in clinical practice: The NEO Personality Inventory. Psychological Assessment, 4, 5-13. Costa, P.T., Jr., & McCrae, R.R. (1992b). The Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI) Professional Manual. Odessa, FL: Psychological Assessment Resources. Dawda, D., & Hart, S.D. (2000). Assessing emotional intelligence: reliability and validity of the Bar-On Emotional Quotient Inventory ((1997A; 2000)) in university students. Personality and Individual Differences, 28, 797-812. Dayan, Y. (2000). Stability reliability of the EQ- i. Unpublished manuscript Davies, M., Stankov, L. & Roberts, R. (1998). Emotional intelligence: In search of an elusive construct. Journal of Personality and Social Psychology, 75, 989-1015.
195
References
Demo, D.H., & Acock, A.C. (1988). The impact of divorce on children. Journal of Family Issues, 15, 619-648. Derogatis, L.R. (1973). SCL-90: Administration, scoring and procedures. Baltimore, MD: John Hopkins University School of Medicine. Digman, J.M. (1990). Personality structure: Emergence of the five- factor model. Annual Review of Psychology, 41, 417-440. Edelmann, R. J., Asendorpf, J., Contarello, A., Zammuner, V., Georgas, J., & Villanueva, C. (1989). Self-reported expression of embarrassment in five European cultures. Journal of Cross-Cultural Psychology, 20, 357-371. Ekman, P. (1973). Dawin and facial expressions: A century of research in review. New York: Academic Press. Ekman, P. (1999). Facial expressions. In T. Dalgleish, & M.J. Power (Eds). Handbook of Cognition. 301-320. Ekman, P., & Davidson, R.J. (1994). The nature of emotion: Fundamental questions. New York: Oxford University Press. Ekman, P., & Friesen, W.V. (1975). Unmasking the face: A guide to recognizing the emotions from facial cues. Englewood Cliffs; NJ: Prentice Hall. Eysenck, H.J., Barrett, P.T., Wilson, G.D., & Jackson, C.J. (1992). Primary trait measurement of the 21 components of the P-E-N system. European Journal of Psychological Assessment, 8, 109-117. Fava, J. L., Velicer, W. F., & Eaton, C. A. (2000). Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components. In R. D. Goffin, & E. Helmes, (Eds). Problems and Solutions in Human Assessment: Honoring Douglas N. Jackson at Seventy. (pp. 41-71) New York, NY, US: Kluwer Academic/Plenum.
196
References
Frijda, N. H. (1988). The laws of emotion. American Psychologist, 43, 349-358. Gardner, H. (1999). Intelligence Reframed. New York: Basic Books. Gibbs, N. (1995). The EQ factor; New brain research suggests that emotions, not IQ, may be the true measure of human intelligence. Time, 60-68. Ghorbani, N., Bing, M.N., Watson, P.J., Davison, H.K., & Mack, D.A. (2002). Selfreport emotional intelligence: Construct similarity and functional dissimilarity of higher-order processing in Iran and the United States. International Journal of Psychology, 37, 297-308. Goleman, D. (1995). Emotional intelligence: why it can matter more than IQ. New York; Bantam Books. Goleman, D. (1998). Working with emotional intelligence. London: Bloomsbury Publishing. Goleman, D. (2001a). Emotional intelligence: Issues in paradigm building. In C. Cherniss & D. Goleman (Eds). The emotionally intelligent workplace: How to select for, measure, and improve emotional intelligence in individuals, groups and organisations. (pp. 13-26). New York: Jossey Bass. Goleman, D. (2001b). An EI-based theory of performance. In C. Cherniss & D. Goleman (Eds). The emotionally intelligent workplace: How to select for, measure, and improve emotional intelligence in individuals, groups and organisations. (pp. 27-44). New York: Jossey Bass. Gottman, J. (1997). The heart of parenting: How to raise an emotionally intelligent child. New York: Simon & Schuster. Henner, T. (1998). Comparing EQ- i and TMMS scale scores. Unpublished manuscript.
197
References
Horn, J.L. (1988). Thinking about human cognitive abilities. In J.R. Nesselroade & R.B. Cattell (Eds). Handbook of multivariate experimental psychology (2nd ed., 645685). New York: Plenum. Jensen, A. R. (1998). The g factor: The science of mental ability. Westport CT: Praeger Joukamaa, M., Miettunen, J., Kokkonen, P., Koskinen, M., Julkunen, J., Kauhanen, J., Jokelainen, J., Veijola, J., Lksy, K., Jrvelin, M.R. (2001). Psychometric properties of the Finnish 20- item Toronto Alexithymia Scale. Nordic Journal of Psyhiatry, 55, 123-128. Kirkcaldy, B. (1995). The Kirkcaldy Quality of Life Questionnaire. Unpublished test. Kline, P. (1994). An Easy Guide to Factor Analysis. London: Routledge. Kline, R.B. (1998). Principles and practice of structural equations modelling. New York: The Guildford Press. Kline, P. (2000). Handbook of Psychological Testing (2nd ed.). London: Routledge. Kooiman, C.G., Spinhoven, P., & Trijsburg, R.W. (2002). The assessment of alexithymia: A critical review of the literature and a psychometric study of the Toronto Alexithymia Scale-20. Journal of Psychosomatic Research, 53, 10831090. Lane, R.D., Sechrest, L., & Riedel, R. (1998). Sociodemographic correlates of alexithymia. Comprehensive Psychiatry, 39, 377-385. Le, H.N., Berendaum, H., & Raghavan, C. (2002). Culture and alexithymia: Mean levels, correlates, and the role of parental socialization of emotions. Emotion, 2, 341-360. LeDoux, J. (1998). The Emotional Brain. London: Weidenfeld & Nicolson.
198
References
Lee, S.J., & Lee, H.K. (1998). The research on the validation of the Trait Meta-Mood Scale: The domain exploration of emotional intelligence. Korean Journal of Social & Personality Psychology, 11, 95-116. Leeper, R.W. (1948). A motivational theory of emotions to replace emotions as disorganised response. Psychological Review, 55, 5-21. Legree, P.I. (1995). Evidence for an oblique social intelligence factor established with a Likert-based testing procedure. Intelligence, 21, 247-266. Loas, G., Corcos, M., Stephan, P, Pellet, J., Bizouard, P., Venisse, J.L., Perez-Diaz, F., Guelfi, J.D., Flament, M., & Jeammet, P. (2001). Factorial structure of the 20item Toronto Alexithymia Scale: Confirmatory factorial analyses in nonclinical and clinical samples. Journal of Psychosomatic Research, 50, 255-261. Luminet, O., Bagby, R.M., Wagner, H., Taylor, G.J., & Parker, J.D.A. (1999). Relation between alexithymia and the five- factor model of personality: A facet-level analysis. Journal of Personality Assessment, 73, 345-358. MacCallum, R. C., Wegener, D. T., Uchino, B. N., & Fabrigar, L. R. (1993). The problem of equivalent models in applications of covariance structure analysis. Psychological Bulletin, 114, 185-199. Marin, G., Sabogal, F., Marin, B., Otero-Sabogal, R., & Perex-Stable, E. (1987). The development of a Short Acculturation Scale for Hispanics. Hispanic Journal of Behavioural Sciences, 9, 183-205. Martinez-Pons, M. (1997). The relation of emotional intelligence with selected areas of personal functioning. Imagination, Cognition and Personality, 17, 3-14. Matsumoto, D. (1991). Cultural influences on facial expressions of emotion. Southern Communication Journal, 56, 128-137.
199
References
Matsumoto, D. (1996). Unmasking Japan: Myths and realities about the emotions of the Japanese. Stanford, CA: Stanford University Press. Matsumoto, D., Takeuchi, S., Andayani, S., Kouznetsova, N., & Krupp, D. (1998). The contribution of individualism-collectivism to cross-national differences in display rules. Asian Journal of Social Psychology, 1, 147-165. Mayer, J.D., Caruso, D. & Salovey, P. (1999). Emotional intelligence meets traditional standards for an intelligence. Intelligence, 27, 267-298. Mayer, J., Caruso, D.R., & Salovey, P. (2000c). Selecting a measure of emotional intelligence: The case for ability scales. In R. Bar-On & J.D.A., Parker, (Eds). The handbook of emotional intelligence. (pp.320-342) San Francisco: JosseyBass Inc. Mayer, J.D., DiPaolo, M., Salovey, P. (1990). Perceiving affective content in ambiguous visual stimuli: A component of emotional intelligence. Journal of Personality Assessment, 54, 772-781. Mayer, J.D., & Gaschke, Y.N. (1988). The experience and meta-experience of mood. Journal of personality and social psychology, 55, 102-111. Mayer, J.D., & Geher, G. (1996). Emotional intelligence and the identification of emotion. Intelligence, 22, 89-113. Mayer, J.D., & Salovey, P. (1993). The intelligence of emotional intelligence. Intelligence, 17, 433-442. Mayer, J. & Salovey, P. (1997). What is emotional intelligence? In P. Salovey & D. Sluyter (Eds.). Emotional development and emotional intelligence: Implications for educators. (pp3-31). New York: Basicbooks, Inc.
200
References
Mayer, J.D., Salovey, P., & Caruso D.R. (2000a). Models of emotional intelligence. In R.J. Sternberg (Ed.). Handbook of Intelligence. (pp396-420) New York: Cambridge. Mayer, J.D., Salovey, P., & Caruso, D.R. (2000b). Emotional intelligence as zeitgeist, as personality, and as a mental ability. In R. Bar-On & J.D.A., Parker, (Eds). The handbook of emotional intelligence. (pp.92-117) San Francisco: JosseyBass Inc. Mayer, J.D., Salovey, P., & Caruso, D.R. (2000d). The Mayer, Salovey, & Caruso Emotional Intelligence Test: Technical manual. Toronto: Multi-Health Systems. Mayer, J.D., Salovey, P., Caruso, D.R., & Sitarenios, G. (2001). Emotional intelligence as a standard intelligence. Emotion, 1, 232-242. Mayer, J.D., Salovey, P., Caruso, D.R., & Sitarenios, G. (2003). Measuring emotional intelligence with the MSCEIT V2. 0. Emotion, 3, 97-105. Mayer, J.D., Salovey, P., Gomberg-Kaufman, S., & Blainey, K. (1991). A broader conception of mood experience. Journal of Personality and Social Psychology, 60, 100-111. Mayer, J.D., & Stevens, A.A. (1994). An emerging understanding of the reflective (meta-) experience of mood. Journal of Research in Personality, 28, 351-373. McCrae, R.R. (2000). Emotional intelligence from the perspective of the five- factor model of personality. In R. Bar-On & J.D.A., Parker (Eds). The handbook of emotional intelligence. (pp.263-276) San Francisco: Jossey-Bass Inc. McDonald, R.P., & Ringo Ho, M-H. (2002). Principles and practice in reporting structural equation analyses. Psychological Methods, 7, 64-82. Nunnally, J.C. (1967). Psychometric theory. London; McGraw-Hill.
201
References
Newsome, S., Day, A.L., & Catano, V.M. (2000). Assessing the predictive validity of emotional intelligence. Personality and Individual Differences, 29, 1005-1016. O'Connor, P. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicier's MAP test. Behavior Research Methods, Instruments & Computers, 32(3), 396-402. Ortony, A., Clore, G.L., & Collins, A.M. (1988). The cognitive structure of emotions. Cambridge. England: Cambridge University Press. Palmer, B.R., Donaldson, K. & Stough, C. (2002). Emotional intelligence and life satisfaction. Personality and Individual Differences, 33, 1091-1100. Palmer, B.R., Gignac, G., Bates, T. & Stough, C. (In Press). Examining the structure of the Trait Meta Mood Scale. Australian Journal of Psychology Palmer, B.R., Manocha, R., Gignac, G., & Stough, C. (2003). Examining the factor structure of the Bar-On Emotional Quotient Inventory with an Australian general population sample. Personality and Individual Differences, 35, 11911210. Palmer, B., Walls, M., Burgess, Z., & Stough, C. (2001). Emotional intelligence and effective leadership. Leadership and Organization Development, 22(1), 1-7. Parker, J.D.A., Taylor, G.J., & Bagby, R.M. (1998). Alexithymia: Relationship with ego defense and coping styles. Comprehensive Psychiatry, 39, 91-98. Parker, J.D.A., Taylor, G.J., & Bagby, M. (2001). The relationship between emotional intelligence and alexithymia. Personality and Individual Differences, 30, 107115. Paulhus, D.L., Lysy, D.C., & Yik, M.S.M. (1998). Self-report measures of intelligence: Are they useful as proxy IQ tests? Journal of Personality, 66, 525-555.
202
References
Peterson, C., Semmel, A., von Baeyer, C., Abramson, L.Y., Metalsky, G.I., & Seligman, M.E.P. (1982). The Attribution Style Questionnaire. Cognitive Therapy Research, 16, 287-300. Petrides, K., & Furnham, A. (2000). On the dimensional structure of emotional intelligence. Personality and Individual Differences, 29, 313-320. Petrides, K. V., & Furnham, A. (2001). Trait emotional intelligence: Psychometric investigation with reference to established trait taxonomies. European Journal of Personality, 15, 425-448. Petrides, K.V., & Furnham, A. (2003). Trait emotional intelligence: behavioural validation in two studies of emotion recognition and reactivity to mood induction. European Journal of Personality, 17, 39-57. Pfeiffer, S.I. (2001). Emotional intelligence: Popular but elusive construct. Roeper Review, 23, 138-143. Radloff, L.S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401. Reise, S. P., Waller, N. G., & Comrey, A. L. (2000). Factor analysis and scale revision. Psychological Assessment, 12(3), 287-297. Reise, S.P., Widsman, K.F., & Pugh, R.H. (1993). Confirmatory factor analysis and item response theory: Two approaches for exploring measurement invariance. Psychological Bulletin, 114, 552-566. Roberts, R.D., Zeidner, M., & Matthews, G. (2001). Does emotional intelligence meet traditional standards for an intelligence? Some new data and conclusions. Emotion, 1, 196-231.
203
References
Robins, L.N., Wing, J.K., Wittchen, H.U., Helzer, J.E., Babor, T.F., Burke, J., Famer, A., Jablenski, A., Pickens, R., Regier, D.A., Sartorius, N., & Towle, L.H. (1988). The Composite International Diagnostic Interview. An epidemiological instrument suitable for use in conjunction with different diagnostic systems and in different cultures. Archives of General Psychiatry, 45, 1069-1077. Roodenburg, J. (2003). Cognitive style: A psycholexically-derived personality-centred model. European Journal of Personality, 17, 119-141. Saarni, C. (1990). Emotional competence: How emotions and relationships become integrated. In R.A. Thompson (Ed.). Newbraska Symposium on Motivation: Vol 36. Socioemotional development (pp.115-182). Lincoln, NE: University of Nebraska Press. Saarni, C. (1999). The developme nt of emotional competence. New York: The Guilford Press. Saklofske, D.H., Austin, E.J., & Minski, P.S. (2003). Factor structure and validity of a trait emotional intelligence measure. Personality and Individual Differences, 34, 707-721. Saklofske, D. H., Hildebrand, D. K., & Gorsuch, R. L. (2000). Replication of the Factor Structure of the Wechsler Adult Intelligence Scale Third Edition with a Canadian sample. Psychological Assessment, 12, 436-439. Salminen, J.K., Saarijaervi, S., Aeaerelae, E., To ikka, T., & Kauhanen, J. (1999). Prevalence of alexithymia and its association with sociodemographic variables in the general population of Finland. Journal of Psychosomatic Research, 46, 7582.
204
References
Salovey, P., Bedell, B.T., Detweiler, J.B., & Mayer, J. (2000). Current directions in emotional intelligence research. In M. Lewis & J.M. Haviland-Jones (Eds.), Handbook of Emotions (2nd Edition). (504-520) New York: Guilford Press. Salovey, P. & Mayer, J.D. (1990). Emotional intelligence. Imagination, Cognition and Personality, 9, 185-211. Salovey, P., Mayer, J.D., Caruso, D., & Lopez, P.N. (In Press). Measuring emotional intelligence as a set of abilities with the MSCEIT. In S.J. Lopez & C.R. Snyder (Eds.). Handbook of positive psychology assessment. Washington, DC: American Psychological Association. Salovey, P., Mayer, J., Goldman, S., Turvey, C. & Palfai, T. (1995). Emotional attention, clarity and repair: Exploring emotional intelligence using the Trait Meta-Mood Scale. In J.W. Pennebaker, Emotion, disclosure, and health (pp. 125-154) Washington, DC: American Psychological Association. Salovey, P., Stroud, L.R., Woolery, A., & Epel, E.S. (2002). Perceived emotional intelligence, stress reactivity, and symptom reports: Further explorations using the Trait Meta-Mood Scale. Psychology & Health, 17, 611-628. Scheier, M.F., & Carver, C.S. (1985). Optimism, coping, and health: Assessment and implications of generalized outcome expectancies. Health Psychology, 4, 219247. Scherer, K.R., Banse, R., & Wallbott, H.G. (2001). Emotion inferences from vocal expression correlate across languages and cultures. Journal of Cross-Cultural Psychology, 32, 76-92. Scherer, K. R., & Wallbott, H. G. (1994). Evidence for universality and cultural variation of differential emotion response patterning. Journal of Personality and Social Psychology, 66, 310-328.
205
References
Schutte, N., Malouff, J.M., Bobik, C., Costn, T.D., Greeson, C., Jedlicka, C., Rhodes, E., & Wendorf, G. (2001). Emotional intelligence and interpersonal relatio ns. The Journal of Social Psychology, 141, 523-536. Schutte, N., Malouff, J., Hall, L., Haggerty, D., Cooper, J., Golden, C. & Dornheim, L. (1998). Development and validation of a measure of emotional intelligence. Personality and Individual Differences, 25, 167-177. Schutte, N., Malouff, J.M., Simunek, M. McKenley, J. & Hollander, S. (2002). Characteristic emotional intelligence and emotional well-being. Cognition and Emotion, 16, 769-785. Schutte, N., Schuettpelz, E., & Malouff, J.M. (2000). Emotiona l intelligence and task performance. Imagination, Cognition and Personality, 20, 347-354. Segal, J. (1997). Raising your emotional intelligence. New York: Holt. Shapiro, L.E. (1997). How to raise a child with a high E.Q.: A parents guide to emotional intelligence. New York: HarperCollins Publishers. Sifenos, P.E. (1973). The prevalence of alexithymic characteristics in psychosomatic patients. Psychotherapy and Psychosomatics, 22, 255-262. Simmons, S., & Simmons, J.C. (1997). Measuring emotional intelligence with techniques for self- improvement. Arlington, TX: Summit Publishing Group. Simonsson-Sarnecki, M., Lundh, L.G., Trestad, B., Bagby, R. M., Taylor, G.J., & Parker, J.D.A. (2001). A Swedish translation of the 20- item Toronto Alexithymia Scale: Cross-validation of the factor structure. Scandinavian Journal of Psychology, 41, 25-30. Sprecher, S. (1999). I love you more today than yesterday: Romantic partners perceptions of changes in love and related affect over time. Journal of Personality and Social Psychology, 76, 46-53.
206
References
Sternberg, R. J., & Kaufman, J.C. (1998). Human abilities. Annual Review of Psychology, 49, 479-503. Stiner, C., & Perry, P. (1997). Achieving emotional literacy: A program to increase your emotional intelligence. New York: Avon. Sun, Y. (2001). Family environment and adolescents well-being before and after parents marital disruption: A longitudinal analysis. Journal of Marriage and Family, 63, 697-713. Tabachnick, B.G., & Fidell, L.S. (1996). Using multivariate statistics. : New York: HarperCollins Colleague Publishers (3rd ed.). Taylor, G.J., & Bagby, M. (2000). An overview of the alexithymia construct. In R. BarOn & J.D.A., Parker, (Eds). The handbook of emotional intelligence. (pp.40-67) San Francisco: Jossey- Bass Inc. Taylor, G.J., Bagby, M., & Luminet, O. (2000). Assessment of alexithymia: Self-report and observer-rated measures. In R. Bar-On & J.D.A., Parker, (Eds). The handbook of emotional intelligence. (pp.301-319). San Francisco: Jossey-Bass Inc. Taylor, G.J., Bagby, M., & Parker, J.D.A. (1991). The alexithymia construct: A potential paradigm for psychosomatic medicine. Psychosomatics, 32, 153-164. Taylor, G.J., Bagby, M., & Parker, J.D.A. (1997). Disorders of affect regulation: Alexithymia in medical and psychiatric illness. Cambridge: Cambridge University Press. Trinidad, D.R., & Johnson, C. (2002). The association between emotional intelligence and early adolescent tobacco and alcohol use. Personality & Individual Differences, 32, 95-105.
207
References
Velicer, W.F. (1976). Determining the number of components from the matrix of partial correlations. Multivariate Behavioural Research, 12, 3-22. Velicer, W.F., Eaton, C.A., & Fava, J.L. (2000). Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components. In R.D. Goffen & Helmes, E. (Eds). Problems and solutions in human assessment: Honoring Douglas N. Jackson at seventy. (pp. 41-71). Norwell, MA, US: Kluwer Academic Publishers. Velicer, W.F. & Fava, J.L. (1998). Affects of variable and subject sampling on factor pattern recovery. Psychological Methods, 3, 231-251. Vittengl, J.R., Holt, C.S. (2000). Getting acquainted: The relationship of self-disclosure and social attraction to positive affect. Journal of Social and Personal Relationships, 17, 53-66. Volling, B.L., McElwain, N.L., Notaro, P.C., & Herrera, C. (2002). Parents emotional availability and infant emotional competence: Predictors of parent-infant attachment and emerging self- regulation. Journal of Family Psychology, 16, 447-465. Wagner, F.R., & Morese, J.J. (1975). A measure of individual sense of competence. Psychological Reports, 36, 451-459. Ward, L. C., Ryan, J. J., & Axelrod, B. N. (2000). Confirmatory factor analyses of the WAIS-III standardization data. Psychological Assessment, 12, 341-345. Waternaux, C.M. (1976). Asymptotic distribution of the sample roots for a nonnormal population. Biometrika, 63, 639-645. Weisinger, H. (1997). Emotional intelligence at work. New York: Jossey-Bass.
208
References
Windle, M., Iwawaki, S., & Lerner, R.M. (1988). Cross-cultural comparability of temperament among Japanese and American preschool children. International Journal of Psychology, 23, 547-567. Zech, E., Luminet, O., Rime, B., Wagner, H. (1999). Alexithymia and its measurement: Confirmatory factor analyses of the 20-item Toronto Alexithymia Scale and the Bermond-Vorst Alexithymia Questionnaire. European Journal of Personality, 13, 511-532. Zeidner, M., Matthe ws, G., & Roberts, R.D. (2001). Slow down, you move too fast: Emotional intelligence remains an elusive intelligence. Emotion, 1, 265-275. Zwick, W. R., & Velicier, W. F. (1986). Comparison of five rules for determining the number of components to retain. Psychological Bulletin, 17, 253-269.
209

FGDFGDFG

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

FGDFGDFG

Enviado por

Direitos autorais:

Formatos disponíveis

An Analysis of the Relationships Between Various Models and Measures of Emotional Intelligence.

Benjamin Robert Palmer

Method 3.2.1 Participants 3.2.2 Materials

120 121 122 122 122 124 127 130

180 183 185 189

Benjamin Robert Palmer. December 1st 2003.

Publications arising from the thesis

MSCEIT Bar-On EQ-i TMMS SEI TAS-20 SEM CFA EFA EI

Figure 4.1 Figure 4.2

Figure 6.1 Figure 7.1 Figure 8.1

106 126 164

Table 4.1 Table 4.2 Table 4.3 Table 4.4

Table 6.1 Table 6.2

108 123 125 128 138 141 xvi

151 155 156 157 160 162

Chapter 2; Emotional Intelligence

CHAPTER 2 Emotional intelligence: Conceptions, measures and research

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Directions for future research

Chapter 2; Emotional Intelligence

Chapter 2; Emotional Intelligence

Chapter 3; The MSCEIT

The Mayer-Salovey-Caruso Emotional Intelligence Test

Chapter 3; The MSCEIT

Conceptual, developmental and correlational criteria for ability measures of EI

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

The Mayer-Salovey-Caruso-Emotional Intelligence Test

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Reliability and factor structure of the MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

3.1.10 Objectives of the present study

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

3.2.2.1 The Mayer-Salovey-Caruso Emotional Intelligence Test

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

3.2.2.3 MSCEIT scoring

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Changes .54 Blends .49 Managing .40

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

Chapter 3; The MSCEIT

* Females n = 286; Males n = 143