|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
THE VALIDITY OF EMOTIONAL INTELLIGENCE AND ITS ABILITY TO PREDICT
ERIC ALLAN ROSSEN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2007 Eric Allan Rossen
I would like to acknowledge many people for helping me during my doctoral work. I
would especially like to thank my advisor and mentor, Dr. John Kranzler, for his generous
commitment to my professional development. Throughout my doctoral work, he encouraged me
to develop independent thinking and research skills. He greatly assisted me with scientific
writing and contributed to my success (and sanity) as a graduate student.
I am also very grateful for having an exceptional doctoral committee and wish to thank Dr.
Nancy Waldron, Dr. James Algina, and Dr. Manfred Diehl for their continual support and input.
I want to give special mention to Dr. Thomas Oakland and Dr. Diana Joyce for their willingness
to closely supervise, support, and guide me through several academic endeavors throughout my
I extend many thanks to my colleagues and friends for managing to keep me grounded and
motivated. I would like to thank Lauren Gibbons for enduring the task of reviewing several
revisions of this project. I would also like to thank the good people at iTunes for providing me
with entertainment during countless hours in front of the computer.
Finally, I would like to thank my family. My mother and father were a constant source of
support, both emotionally and financially. I am grateful to my brother for his support. I am also
grateful to my grandparents for their encouragement throughout my academic career.
TABLE OF CONTENTS
A CK N O W LED G M EN TS ................................................................. ........... ............. 3
LIST O F TA BLE S ............................................................................................. .............
L IST O F FIG U RE S ............................................................................... 9
ABSTRAC T ............................... ..................... 10
1 REV IEW OF TH E LITER A TU RE ............................................................... .......... .. .. 12
Precursors of El in the Intelligence Literature .... ............................. .............13
P sy ch om etric g ................................................................13
Criticism s of Psychom etric g................................................. .............................. 14
C attell-H orn G f-G c T theory ..................................... ............. .............. .......................15
Three-Stratum Theory ...................................... ......... .............. 15
S o cial In telling en c e ................................. ..........................................................16
Multiple Intelligences ....... .... .. .... .............................17
Triarchic Theory of Intelligence........................ ...................... ........... ............... 18
Emergence of Contemporary Research on El ......................................................... 19
E l T heory and R research .......... ............................................................ ...... ...... ...... .... 2 1
M ixed M odels and Their M easurem ent ........................................ ....... ............... 21
Criticism of M ixed M odel Theories of EI................................................. ............... 25
A b ility M models of E .......... .. ...... .................. .... ...... .......................... ..... ..... ...... 2 7
M easurem ent of A ability M models of E .............................. .................. .....................29
Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT) .......................................30
F acto r S tru ctu re ........................... ... ............................................. 3 0
Gender, Racial/Ethnic, and Cross-Cultural Differences......................................31
Relationship with Self-Report El Measures ............ .............................................31
Relationship w ith Age and Experience ........................................ ....................... 32
V validity ................................................................... ............. 33
P redicting Im portant O utcom es............................................................ .....................34
C criticism of A ability M odels.................................................. ............................... 38
S u m m a ry ............... .......... .. ................................................................................4 2
P purpose of the C current Study ......................................................................... ...................42
2 M E T H O D S .........................................................................44
P articip an ts .........................................................................4 4
Instrum ents ..................... .. ............ .......................................... ........ 44
Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT) ..............................44
International Personality Item Pool (IPIP) ..........................................46
Wonderlic Personnel Test (WPT) ..................................................47
C riterion M measures ........................ .. ........................ .. .... ........ ........ 48
P ro c e d u re ................... ................... ...................1..........
D ata A analysis ................................................... 51
3 R E SU L T S .............. ... ................................................................53
D descriptive Statistics ................................................................... 53
O utcom e V ariables ................................................................53
Predictor V ariables ........................................54
Mean Score Differences by Race/Ethnicity, Gender, and Year of Study ....................55
Correlational Analyses.......................................... 55
Factorial Analysis of Variance (AN OVA) ..........................................................................57
Confirm atory Factor A analysis ...................................................................58
M multiple Regression ............................................. 60
A cadem ic Success ....................... .......................... .. .................. 60
Psychological W ell-Being .............................................. ................. .. ............ 61
Positive R relations w ith O others ............................................................. 61
Peer A ttachm ent .............................. ............................. .. 62
Cigarette U se ......................... .................. ..... .. .. ............. ..... 62
A alcohol U se .......................................................... ........................ 62
4 D ISCU SSIO N .............................................................. .................... 83
Research Question 1: Evidence of Validity ................................. .................... 83
Convergent Evidence of Validity ................ .. .............................. ......... ....................83
D ivergent Evidence of V alidity ............................................................... .. ...... 84
Research Question 2: EI, Age, and Experience ........................ .................................. 85
Research Question 3: Group D differences ....................................................... 86
Research Question 4: Predicting Real-Life Outcomes ........................................ ....86
A ca d em ic S u cce ss ..................................................................................................... 8 6
P psychological W ell-B being ............................................................... .......... ................87
Peer A ttachm ent ......................... ......... ...............................................88
C ig arette U se .............................................................................8 9
A alcohol U se .................................................................. ........................ 89
Research Question 5: Latent Structure of the MSCEIT .............. ..................................89
Summary and Implications for Future Research ............................................... ......91
Limitations .................. ................. ........ ... 92
A DEM OGRAPHIC QUESTIONNAIRE .................................................... .....................94
B INTERNATIONAL PERSONALITY ITEM POOL (IPIP) PERSONALITY SCALE.........95
C INVENTORY OF PARENT AND PEER ATTACHMENT (IPPA) SECTION III..........100
D ALCOHOL USE DISORDERS IDENTIFICATION TEST (AUDIT) ...............................103
E SU RV EY O F CIG A RE TTE U SE ............................................................. .....................105
L IST O F R E FE R E N C E S ......... .. ............. ..........................................................................106
B IO G R A PH IC A L SK E T C H ............ ........................................................... .......................... 124
LIST OF TABLES
3-1 Descriptive statistics for outcome variables. ........................................ ............... 63
3-2 Descriptive statistics for predictor variables................................................................... 64
3-3 Mean scores on the MSCEIT by race/ethnicity ...................................... ............... 65
3-4 M ean Scores on the M SCEIT by Gender ............................................... ............... 65
3-5 M ean scores on the M SCEIT by year of study ........................... .....................66
3-6 Pearson product-moment correlations between age and the MSCEIT ............................ 66
3-7 Pearson product-moment correlations between academic indicators and the MSCEIT....66
3-8 Pearson product-moment correlations between the Big Five personality dimensions
and the M SC E IT ............................................................................67
3-9 Pearson product-moment correlations between psychological well-being (PWB) and
the M SC E IT ................. .......... ......... ...........................................67
3-10 Pearson product-moment correlations between peer attachment and the MSCEIT..........68
3-11 Pearson product-moment correlations between alcohol and cigarette use and the
M S C E IT ............................................................... ................ 6 8
3-12 Factorial analysis of variance results for Overall EIl......... ................................... 69
3-13 Factorial analysis of variance results for Strategic EI ..................................................69
3-14 Factorial analysis of variance results for Experiential EI ........................................70
3-15 Observed score intercorrelations of the MSCEIT subtests (N = 150) .............................70
3-16 Observed score intercorrelations of the MSCEIT subtests and first-, second-, and
third-order factors (N = 150). .................................................................... ..................71
3-17 First-order inter-factor correlations of the MSCEIT (N = 150)............... ...................72
3-18 MSCEIT parameter estimates for one-, two-, three-, four-, nested, and four-branch
factor m models (N = 150) ........................... ...... ..................................... .. .... .. 72
3-19 Goodness of fit indices for one-, two-, three-, four-, nested, and four-branch factor
m o d e ls (N = 1 5 0 ) ...............................................................................................................7 3
3-20 Hierarchical regression predicting grade point average (N = 148).................................74
3-21 Hierarchical regression predicting SAT (N = 143)................................. ............... 75
3-22 Hierarchical regression predicting psychological well-being (N = 150)...........................76
3-23 Hierarchical regression predicting Positive Relations with Others (N = 150)...................77
3-24 Hierarchical regression predicting peer attachment (N = 150)........................................78
3-25 Hierarchical regression predicting cigarette use (N = 150) ............................................79
3-26 Hierarchical regression predicting alcohol use (N = 150) ...........................................80
LIST OF FIGURES
1-1 Carroll's Three-Stratum Theory of intelligence ........................ .............43
1-2 Factor structure of the four-branch ability model of E..................................................43
3-1 General factor m odel................................. .. .......... ............... 80
3-2 O blique tw o-factor m odel ......................................................................... ...................8 1
3-3 O blique four-factor m odel ........................................................................ ...................8 1
3-4 O blique three-factor m odel ....................................................................... ................... 82
3-5 N ested m odel ..............8.............................82
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
THE VALIDITY OF EMOTIONAL INTELLIGENCE AND ITS ABILITY TO PREDICT
Eric Allan Rossen
Chair: John H. Kranzler
Major: School Psychology
Emotional intelligence (El) has gained considerable attention within the last decade. This
surge in interest is due, in part, to claims about EI's ability to predict important outcomes such as
life satisfaction, peer attachment, substance abuse, depression, and loneliness. This popularity
also has been attributed to attempts at expanding our understanding of human abilities beyond
general intelligence, or psychometric g. Although psychometric g is the most predictive factor of
many important educational and occupational outcomes, at least 50% of the variance can be
attributed to other variables. Thus, researchers have looked for alternative constructs such as El
to supplement or replace intelligence tests to improve prediction.
Although a number of conceptualizations of El have emerged within the last decade,
Mayer and Salovey's (1997) theory of El has the most empirical support. They postulate that El
consists of the ability to (1) accurately perceive emotions, (2) use emotions to facilitate thinking,
(3) understand emotions, and (4) manage emotions. The Mayer-Salovey-Caruso Emotional
Intelligence Test (MSCEIT) was developed to measure these constructs. Although the test's
authors provide some evidence to substantiate the MSCEIT's validity, questions remain.
The purpose of this study was to examine the construct of El as measured by the MSCEIT.
Specifically, this study investigated the validity of the MSCEIT by examining its internal
structure, relationship to external criteria, and incremental validity. Results of this study revealed
that internal consistency estimates were acceptable for higher-order factors, although
unacceptable for some subtests. Overall El and scaled scores on the MSCEIT correlated with
personality and g in the low to moderate range, demonstrating convergent and divergent
evidence of validity. Overall El also correlated significantly with self-reported SAT scores,
psychological well-being, peer attachment, and alcohol use. However, after controlling for the
effects ofg and personality, El contributed little to no additional explained variance in a number
of real-life outcomes. Alcohol use was the only criterion on which El explained additional
variance. Further, results indicated that the proposed theoretical structure of the MSCEIT is an
improper model. Thus, the present study raises concerns about the construct of El and its
measurement with the MSCEIT.
REVIEW OF THE LITERATURE
Emotional intelligence (EI) is generally defined as the ability to use emotions to reason,
problem-solve, manage relationships, enhance thought, and succeed. Over the past 15 years, El
has received considerable attention in the social sciences as well as in the popular media. In fact,
the American Dialect Society listed El as one of the most useful new phrases of 1995 (American
Dialect Society, 1996). The construct of El ignited such public interest that El appeared as the
cover story in TIME magazine (Gibbs, 1995), in popular newspapers, internet websites, and in
trade texts dealing with self-help, management practices, and assessment (Matthews, Zeidner, &
Although El has recently received an increased amount of attention, the relationship
between cognition and affect has been discussed for centuries. Dating as far back as ancient
Greece and Rome, philosophers have debated the relationship between cognition and emotion.
Some associated affect and strong emotion with weakness and irrational thought (Grewal &
Salovey, 2005). Stoics, for example, believed that one must avoid extreme emotions in order to
think reasonably and rationally (Still & Dryden, 1999). Conversely, philosophers such as Plato
and Aristotle believed that thought plays a significant role in the expression of human emotion.
They believed that emotion results from one's thoughts and beliefs about the world
(Fortenbaugh, 1975). For example, an individual may experience fear only if they appraise the
situation as dangerous. Similar discussions addressing the relationship between cognition and
emotion continue in modern thought and psychological research, exploring new perspectives on
the interaction between emotion and thought, and seeking rules to describe when and why
emotions arise (Mayer, Salovey, & Caruso, 2002).
Precursors of El in the Intelligence Literature
Although no consensus definition exists, there is general agreement as to what intelligence
is. An American Psychological Association task force defined intelligence as the "ability to
understand complex ideas, to adapt effectively to the environment, to learn from experience, to
engage in various forms of reasoning, to overcome obstacles by taking thought" (Neisser et al.,
1996, p. 77). Another definition originating from a treatise signed by 52 well-established
intelligence researchers conceptualized intelligence as "a very general mental capability that,
among other things, involves the ability to reason, plan, solve problems, think abstractly,
comprehend complex ideas, learn quickly and learn from experience ... it reflects a broader and
deeper capability for comprehending our surroundings" (Gottfredson, 1997a, p. 13). Despite
some differences among existing definitions, many agree that the domain of human intelligence
is multidimensional, including the ability to reason, learn, problem solve, think about complex
ideas, and adapt to our environment and experiences.
In the late 1800s, Binet and Simon devised the first intelligence tests, which contained a
variety of cognitive tasks used to distinguish children with mental retardation from those with
behavior problems. Spearman (1904) discovered that scores on all measures of cognitive ability
were positively correlated, suggesting that a common underlying trait may influence
performance. He claimed that this common latent trait, which he referred to as "g," is central to
performance on all cognitive tasks (Kamphaus, 2001). Although Spearman recognized the
influence of additional specific factors in determining test performance, g appeared to be the
most important. Thus, Spearman explained intellectual abilities as consisting of psychometric g
representing common variance shared across all measures of cognitive ability, and specific
factors representing test or subtest variance unique to individual measures.
At present, a great deal of data substantiate g as the best available predictor of a range of
important outcomes, including both academic and occupational success (e.g., Glutting, Watkins,
& Youngstrom, 2003; Gottfredson, 2002; Hermstein & Murray, 1994; Jensen, 1998; Johnson,
Bouchard, Krueger, McGue, & Gottesman, 2004; McGrew, 2005; Sattler, 1993). As Gottfredson
(1997b) noted, no other single measurable predictor has such consistently high predictive
validity for job performance, including job complexity and trainability. In addition, evidence
suggests that g is associated with success in training programs, occupational status, rates of crime
and delinquency, mate selection, health risk behavior, and artistic/scientific creativity (Jensen,
1998; Lubinski, 2004; Wasserman & Tulsky, 2005).
Overall intelligence quotient (IQ) scores derived from standardized intelligence tests are
good estimates of psychometric g. IQ scores, at best, explain 50% of the variance in predicting
occupational and academic outcomes, and vary by the specific criterion. This relationship seems
to be extremely robust, and confirmed by decades of research (Neisser et al. 1996).
Criticisms of Psychometric g
Although g is the most active ingredient in test prediction, the effects of intelligence
simply provide an advantage by improving the odds of success (Gottfredson, 1997b).
Approximately 50 to 75% of the variance associated with positive outcomes can be attributed to
other variables and traits. Further, given the multidimensional nature of intelligence,
psychometric g does not fully represent the array of cognitive abilities that constitute human
intelligence (Daniel, 1997; Horn & Blankson, 2005; Sternberg, 1997). Thus, researchers have
examined the influence of additional abilities and variables that may contribute to our
understanding of individual differences and prediction of human behavior (Van Rooy, Alonso, &
Cattell-Horn Gf-Gc Theory
Cattell (1943, 1963) proposed two types of intelligence based on a series of factor analytic
studies: fluid (Gf) and crystallized (Gc). Fluid intelligence refers to reasoning and problem
solving abilities. More specifically, fluid intelligence refers to the ability to solve novel problems
or tasks that cannot be performed automatically, or both. On the other hand, crystallized
intelligence refers to learned information obtained through experience, and can be described as a
person's breadth of acquired knowledge in language, cultural concepts, and the application of
this knowledge (Kamphaus, 2001; McGrew, 2005).
Horn (1979) expanded upon Cattell's model by positing additional abilities beyond Gf and
Gc. Based on an extensive program of factor analytic research, Horn (1991) had identified 10
correlated second-order abilities, including fluid intelligence, crystallized intelligence, short-term
acquisition and retrieval, visual intelligence, auditory intelligence, long-term storage and
retrieval, cognitive processing speed, correct decision speed, quantitative knowledge, and
reading and writing skills (McGrew, 2005). This eventually became known as the Cattell-Horn
Gf-Gc theory. Notably, the Cattell-Horn Gf-Gc theory did not incorporate g as a common
Carroll (1993) proposed a new structure of intelligence resulting from a structural analysis
of over 460 different datasets. This model consisted of three hierarchical strata, with each
subsequent stratum increasing in breadth. Within this framework, Stratum I reflects 70 narrow
cognitive abilities. Stratum II consists of eight broad cognitive abilities that are similar to those
proposed by the Cattell-Horn Gf-Gc theory, including fluid intelligence, crystallized intelligence,
general memory and learning, broad visual perception, broad auditory perception, broad retrieval
ability, broad cognitive speediness, and processing/decision speed. Stratum III is characterized
by psychometric g, the general factor that accounts for variance in all cognitive abilities
(Flanagan & Ortiz, 2001) (see Figure 1-1). At present, a hierarchical three-stratum theory of
cognitive abilities is arguably the most widely accepted theory of intelligence, and implies the
multidimensionality of human intelligence.
Although the three-stratum theory appears to be the best description of cognitive abilities,
considerable variance remains unexplained when all cognitive factors are controlled for. Thus,
psychometric theories of intelligence may be incomplete (Sternberg, 2003), which suggests that
other abilities and noncognitive factors likely are important in prediction (Raven, 2002). For
example, many individuals with high levels of intelligence sometimes demonstrate difficulty
taking into account their own and other people's feelings, make poor decisions, and have
difficulty achieving success (Lopes & Salovey, 2004). On the other hand, those who behave in
socially and emotionally intelligent ways may experience success in relationships, academics,
and occupational settings (Kang, Day, & Meara, 2005). Wechsler (1950), known for his
pioneering work in intelligence testing, noted the need to measure non-intellective aspects of
intelligence to explain the variance in academic and job success unaccounted for by traditional
E. L. Thomdike (1920) defined social intelligence as "the ability to understand and manage
men and women, boys and girls-to act wisely in human relations" (p. 228). Generally, this
refers to how individuals understand, interact, and deal with people, and is similar to existing
definitions of El (Austin & Saklofske, 2005).
Following decades of research, however, social intelligence failed to emerge as a valid
construct (Cronbach, 1960; Davies, Stankov, & Roberts, 1998). This failure is in part due to
difficulties operationalizing and objectively measuring social intelligence. For example, Weis
and SUi (2005) argued that assessing social intelligence in real-life contexts while controlling for
the environment, moods, social norms, and values remains a significant and perhaps
insurmountable obstacle. The measurement of social intelligence requires capturing how
individuals respond to social and environmental stimuli, including how they might adapt their
responses, facial expressions, gestures, and voice intonations in a given situation. Thus, formal
standardized testing processes are unable to measure the construct of social intelligence
(Thorndike, 1920). Therefore, it appears that social intelligence remains an elusive construct.
A marked increase in research exploring the interaction of emotion and intellectual thought
occurred in the 1980s, in part due to Howard Gardner. Gardner (1993) defined intelligence as the
ability to solve problems or create products that have societal or cultural value. In his book,
Frames ofMind (1983), Gardner postulated seven intelligence, including linguistic (related to
reading, writing, and story-telling), logical-mathematical (related to arithmetic, patterns, and
strategy), bodily-kinetheti/i (knowledge through sensation), spatial (thinking in images and
pictures), musical (related to auditory awareness), intrapersonal (related to awareness of one's
own feelings), and interpersonal (related to positive peer relationships).
Gardner argued that his seven proposed intelligence function independently. Further, he
claimed the intelligence are not dominated by g or other broad cognitive abilities. These claims,
however, have not been substantiated by empirical research, in part due to the difficulty of
creating tests that can objectively measure each individual intelligence independently. Criticism
also has focused on the specific intelligence themselves; some have argued that musical
intelligence, for example, can be considered a talent or cognitive style rather than a stand-alone
construct (e.g., Morgan, 1996).
Among Gardner's distinct intelligence, intrapersonal intelligence and interpersonal
intelligence are most closely related to EI. Gardner (1983) broadly defined intrapersonal
intelligence as the ability to access one's emotions, discriminate among different feelings, label
them, and then use them to facilitate behavior. According to his theory, intrapersonal intelligence
helps us understand who we are and what drives our behavior. Gardner (1983) characterized
interpersonal intelligence as the ability to identify and make distinctions among the emotions of
others. This ability relates to our understanding of other people's emotions, motives, and
subsequent behavior. Gardner's conceptualization of personal intelligence is therefore similar to
current conceptualizations of El (Grewal & Salovey, 2005).
Triarchic Theory of Intelligence
Sternberg's (1985) triarchic theory of intelligence is largely based on the notion that,
although traditional methods of intellectual assessment provide important information about an
individual, they are incomplete. Thus, individuals who score high on traditional psychometric IQ
tests may not always function well in everyday life. His theory consists of three kinds of
intelligence: analytical (or componential), creative (or experiential), and practical (or contextual).
Analytical intelligence is similar to the traditional psychometric definition of intelligence,
emphasizing the importance of academic problem solving, information processing, and fluid
reasoning ability. Creative intelligence refers to an individual's ability to synthesize and react to
novel stimuli, emphasizing the importance of automaticity. Practical intelligence refers to the
ability to learn from past experiences and deal with everyday real-world tasks and problems, and
can be characterized as common sense.
Of the three components of Sternberg's theory, practical intelligence has been most
frequently linked to El (e.g., Austin & Saklofske, 2005; Matthews et al., 2002). Sternberg,
Wagner, and Okagoski (1993) have drawn particular attention to the importance of "tacit
knowledge" in practical intelligence, which refers to an implicit understanding of how to deal
with a given situation without formal instruction.
Sternberg and his colleagues have argued that measures of tacit knowledge have
demonstrated the capacity to predict academic and occupational outcomes uniquely from
traditional IQ and personality tests (Hedlund & Sternberg, 2000; Wagner, 1987; Wagner &
Sternberg, 1990). However, critics argue that measures of tacit knowledge lack empirical
support, and merely describe knowledge acquired from experience rather than an intellectual
ability (Gottfredson, 2001; Jensen, 1993; Ree & Earles, 1993).
Emergence of Contemporary Research on El
Although several theories have attempted to explain the variance unaccounted for by
traditional theories of intelligence, most have encountered significant criticism regarding their
measurement, or conceptual framework, or both. Recently, the construct of El has emerged in an
attempt to explain additional variance in human behavior.
Goleman and the popularization of EI. Goleman (1995) created a model of El
characterized by five broad areas: knowing one's emotions (recognizing and monitoring
feelings), managing emotions (emotional regulation), motivating oneself (goal-oriented),
recognizing emotions in others (empathic awareness), and handling relationships (managing the
emotions of others). Goleman considered El to encompass "a set of abilities which include self-
control, zeal and persistence, and the ability to motivate oneself" (p. xii). He further specified
that El consists of the ability to "motivate oneself and persist in the face of frustrations; to
control impulse and delay gratification; to regulate one's moods and keep distress from
swamping the ability to think; to empathize and to hope" (p. 34). Goleman (1998) later described
El as including up to 25 skills and characteristics that promote success, such as initiative,
teamwork, and self-awareness, and likened El to individual character.
Goleman (1995) made strong claims about the ability of El to predict important real-life
outcomes. In his book, Emotional Intelligence: Why It Can Matter More Than IQ, he contended
that El accounts for up to 80% of the variance in academic and occupational success. Goleman
asserted that El provides "an advantage in any domain in life, whether in romance and intimate
relationships or picking up the unspoken rules that govern success in organizational politics" (p.
The influence of Goleman's book on popularizing El became immediately apparent. It
stimulated a great deal of research on EI, as evidenced by the competing theories of El that
emerged shortly after its publication. In addition, intervention and training programs were
developed to provide parents with strategies to improve their children's EI, and some businesses
hired El coaches to enhance worker productivity (Grewal & Salovey, 2005). By 1997, over 20
formal programs of social-emotional learning were incorporated into school curricula, many of
which emphasized the teaching of El (Elias et al., 1997).
Goleman, however, was criticized for asserting that El is a more important predictor of
success than IQ without providing empirical support for these claims (Landy, 2005; Mayer &
Cobb, 2000; Mayer, Salovey, & Caruso, 2004a). For example, Goleman (1998) claimed El has
greater predictive validity for occupational performance than IQ. However, as Matthews et al.
(2002) argued, no published studies actually confirm this relationship, and the unpublished
investigation that Goleman cites does not actually include any measure of EI.
Critics of Goleman also contended that his definition of El is overinclusive and unclear,
incorporating aspects of cognition, personality, motivation, emotions, neurobiology, and
intelligence (Locke, 2005; Matthews et al., 2002). In response to Goleman's claims, Mayer,
Salovey, and Caruso (2000b) stated "the unexplained 80% of success appears to be in large part
the consequence of complex, possibly chaotic interactions among hundreds of variables playing
out over time" (p. 412).
At present, many existing models of El are fairly distinct from one another despite most
using the same El label (Ashkanasy & Dasborough, 2003). The next sections summarize these
main theories of EI. Following this, the most widely researched theory of El with the greatest
empirical support, the four-branch ability model of El (Mayer & Salovey, 1997), is reviewed.
El Theory and Research
At present, the theoretical framework of El is a topic of debate among researchers (Van
Rooy & Viswesvaran, 2004). Mayer and colleagues made a distinction between two main
competing groups of El theories-mixed (or sometimes referred to as trait El) and ability models
(Mayer, Salovey, & Caruso, 2000a).
Mixed Models and Their Measurement
In mixed models, El does not exclusively consist of emotion or intelligence (Neubauer &
Freduenthaler, 2005). Instead, El is seen as a mixture of cognitive abilities and personality traits
that may predict success in various domains (Mayer et al., 2000b). Although most mixed model
El theories make reference to cognitive abilities utilized in the processing of emotional
information, these theories focus more on personality traits and attributes such as optimism and
motivation (Goldenberg, Matheson, & Mantler, 2006; Livingstone & Day, 2005).
Mixed models generally are measured by self-report questionnaires, which assess an
individual's beliefs about his or her competencies in areas of EI. Typically, respondents are
given a series of statements regarding their emotional understanding, awareness, and control.
They are then asked to indicate on a Likert scale the extent to which the statements describe how
they feel, think, or behave in most situations. Some items from popular self-report El measures
include, "I am aware of my emotions as I experience them" (Schutte et al., 1998), "It is a
problem controlling my anger" (Bar-On, 1997), and "Sometimes I can't tell what my feelings
are" (Salovey, Mayer, Goldman, Turvey & Palfai, 1995). Authors of mixed model El measures
claim that they predict success and other important outcomes fairly well (Bar-On, 1997;
Goleman, 1998; Schutte et al., 1998).
The most popular and widely cited mixed model El theories and their corresponding self-
report El measures are presented below.
Bar-On's model of Social-Emotional Intelligence (S-EI). In Bar-On's S-EI theory,
several interrelated emotional and social components that impact intelligent behavior and the
ability to cope with the demands and pressures of daily life are combined. Bar-On (2005) noted
the influence of related constructs in the development of his theory, including social intelligence
and Gardner's (1983) intrapersonal and interpersonal intelligence.
S-EI consists of 15 competencies that measure five higher-order factors. The first,
Intrapersonal .\kil/ (comprised of Self-Regard, Emotional Self-Awareness, Assertiveness,
Independence, and Self-Actualization), refers to the ability to recognize, understand, and express
emotions. The second, Interpersonal .kill/ (comprised of Empathy, Social Responsibility, and
Interpersonal Relationship), refers to the ability to understand the emotions of others. The third
factor, Adaptability (comprised of Reality-Testing, Flexibility, and Problem-Solving), refers to
the ability to handle change and solve problems. The fourth, Stress Management (comprised of
Stress Tolerance and Impulse Control), refers to the ability to manage emotions. The fifth factor,
General Mood (comprised of Optimism and Happiness), refers to the ability to generate positive
affect and be self-motivated (Bar-On, 2005).
To measure these constructs, Bar-On constructed the Emotional Quotient Inventory (EQ-i;
Bar-On, 1997), a 133 item self-report measure. Of the existing El measures, the EQ-i is among
the most widely used (Van Rooy & Viswesvaran, 2004).
According to Bar-On (1997), the EQ-i is predictive of success in academics and success
among US Air Force recruiters. In addition, the EQ-i was found to moderately correlate with
measures of psychological well-being, physical health, self-actualization, and social interaction
(Bar-On, 1997, 2001, 2004, 2005; Bar-On & Fund, 2004). In an independent study using the
Emotional Quotient Inventory: Youth Version (EQ-i:YV), Overall El correlated significantly
with academic success among a large sample of high school students (Parker et al., 2004).
Schutte et al. El theory. The El theory proposed by Schutte et al. (1998) is largely based
on Salovey and Mayer's (1990) original model of El, consisting of a general second-order El
factor as well as three first-order factors, namely, Appraisal and Expression of Emotion,
Regulation of Emotion, and Utilization of Emotion. The Appraisal and Expression of Emotion
factor is comprised of perception of emotion and empathy. Regulation of Emotion includes
regulating emotions in the self as well as in others. The third factor, Utilization of Emotion,
includes flexible planning, creativity, attention, and motivation (Schutte et al., 1998).
To measure these components, Schutte and colleagues developed the Schutte Self-Report
Inventory (SSRI; Schutte et al., 1998). The SSRI correlated moderately with alexithymia (i.e.,
difficulty describing and recognizing one's own emotions), depression, and academic success.
Saklofske, Austin and Minski (2003) also found moderate correlations between the SSRI and
alexithymia, self-reported happiness, satisfaction with life, depression-proneness, social
loneliness, family loneliness, and romantic loneliness. These findings support the validity of the
SSRI. As predicted by El theory, lower levels of El and higher levels of academic success, life
satisfaction, and psychological well-being are associated with depression and loneliness.
Trait Meta-Mood Theory. Salovey et al. (1995) developed a theory to explain individual
differences in meta-mood, or the process by which one responds to their emotional states
(Fitness & Curtis, 2005). According to this theory, El consists of three factors, including
Attention to Emotion, Emotional Clarity, and Emotion Repair. Attention to Emotion refers to the
awareness of inner feelings and emotions. Emotional Clarity refers to the ability to distinguish
among different feelings and emotional states. Emotion Repair refers to the ability to regulate
emotions and repair negative emotional experiences.
The Trait Meta-Mood Scale (TMMS; Salovey et al., 1995) was developed to measure these
three factors. Unlike most El measures, the TMMS does not yield an overall El score. Martinez-
Pons (1997) found evidence for the three-factor model of the TMMS. However, in a study using
exploratory factor analysis, Palmer, Gignac, Bates, and Stough (2003) found that oblique three-
and four-factor solutions of the TMMS were proper models. Palmer et al. (2003) asserted that
these differences result from cultural differences between samples, as the participants in
Martinez-Pons' study were from the USA and theirs were from Australia. Thus, the theory is
Salovey et al. (1995) found that the TMMS correlated with criteria such as depression,
optimism, locus of control, mood recovery, and goal orientation. Among a sample of
undergraduate students, Fitness and Curtis (2005) found that Attention to Emotion was
moderately correlated with self-reported measures of empathy and the ability to use complex
reasoning when thinking about other's behaviors. A self-report measure of self-control correlated
positively with Emotional Clarity and Emotion Repair. They also found that destructive,
maladaptive responses to interpersonal conflict were negatively associated with Emotional
Clarity and Emotion Repair. However, constructive responses were not significantly associated
with the TMMS.
Emotional Competence Theory. Goleman, Boyatzis, and McKee (2002), extending upon
Goleman's (1998) earlier theoretical model of EI, presented a model consisting of 18
competencies that comprise four clusters: (1) Self-Awareness, (2) Self-Management, (3) Social
Awareness, and (4) Social Skills. According to their theory, El is a set of competencies that
promotes understanding of emotional information and leads to improved performance (Boyatzis
& Sala, 2004). To measure these competencies, Sala (2002) developed the Emotional
Competence Inventory-Version 2 (ECI-2).
The ECI-2 is 72-item self-report El measure which incorporates self-report and manager
and peer ratings to measure emotional competencies related to EI. A number of studies have
found that the ECI-2 is predictive of job performance and success, salary increases, college
student retention, managerial styles and organizational climate (e.g., Boyatzis & Sala, 2004;
Cavalo & Brienza, 2002; Humphrey, Kellett, & Sleeth, 2001; Lloyd, 2001; Nel, 2001; Sevinc,
2001; Stagg & Gunter, 2002; Williams, 2003). However, most of the cited empirical studies
evaluating the ECI-2 are only available in technical reports, unpublished manuscripts, and
working papers (Matthews et al., 2002). Conte (2005) argued that, because most of these studies
have not appeared in published scientific journals or been subjected to blind peer-review, the
ECI-2 does not deserve serious consideration.
Criticism of Mixed Model Theories of El
Critics of mixed model El theories argue that incorporating qualities such as motivation,
optimism, empathy, and other traits into a single psychological entity called El is problematic
(Mayer et al., 2000). They maintain that, although studying these variables together may provide
These results led to questions about the appropriateness of labeling these tests as measures of El
(Neubauer & Freduenthaler, 2005). According to Mayer, Salovey, and Caruso (2004a), mixed
models "often have little or nothing specifically to do with emotion or intelligence and,
consequently, fail to map onto the term emotional intelligence" (p. 197).
Several empirical studies have further investigated the association between several widely
used self-report El and personality measures (Dawda & Hart, 2000; Hedlund & Sternberg, 2000;
McCrae, 2000; Newsome et al., 2000; Petrides & Furnham, 2000, 2001; Saklofske et al., 2003;
Van Rooy & Viswesvaran, 2004; Warwick & Nettelbeck, 2004). Each of these studies found
substantial overlap between self-report El and personality measures. This is somewhat expected
considering the inclusion of such variables as empathy and motivation in several mixed model
theories, which coincide with components of the personality domain.
Given these criticisms and apparent shortcomings, Conte (2005) argued that mixed models
are not as viable as ability models. As a result, the predominant theories are ability models of EI.
Among them, Mayer and Salovey's (1997) model is generating the most research.
Ability Models of El
Salovey and Mayer (1990) proposed a hierarchical model in which El consists of Appraisal
and Expression of Emotion, Regulation of Emotion, and Utilization of Emotion. Their
conceptualization of El was strongly influenced by the movement to broaden the construct of
intelligence (Mayer, Salovey, & Caruso, 2004a). They did not necessarily assume that El is
fundamentally independent from g; rather, they described El as another way to measure
intelligent behavior (Landy, 2005).
Mayer and Salovey (1997) later revised their model and developed a four-branch
hierarchical model to isolate El as a mental ability and separate it from well-known personality
traits (Neubauer & Freudenthaler, 2005). They asserted that people think intelligently about
emotions and that those emotions can facilitate intelligent thought. In particular, their interest lies
in understanding individual differences in the processing of affective information. They defined
El as a collection of abilities that combine to form four oblique first-order factors, or branches:
(1) Perceiving, Appraising, and Expressing Emotions; (2) Using Emotions to Facilitate Thought;
(3) Understanding Emotions; and (4) Managing Emotions.
The first branch, Perceiving, Alpi ,aiiingi. and Expressing Emotions, refers to how well
individuals identify emotions and emotional content. In other words, an emotionally intelligent
individual should be able to distinguish among facial and postural expressions of emotion,
identify their own bodily sensations, and monitor internal feelings. Additional emphasis is placed
on recognizing emotions and feelings in other people as well as oneself. Finally, this branch
refers to the ability to express feelings accurately through the face, voice, and related
communication channels (Mayer et al., 2004a).
The second branch, Using Emotions to Facilitate Thought, refers to the ability to
effectively use feelings and emotions to assist thinking. An emotionally intelligent individual
generates specific emotions to support problem solving. It further includes the ability to use
emotions to bring attention to important events to harness emotions for more effective and
rational decision making (Salovey & Pizarro, 2002).
The third branch, Understanding Emotions, involves the ability to understand complex
emotions and the similarities and differences between them (e.g., liking vs. loving, annoyance vs.
anger, etc.). An emotionally intelligent individual can comprehend emotion within the context of
relationships (e.g., sadness often accompanies a loss; anger often accompanies an argument)
(Caruso, Mayer & Salovey, 2002). In addition, emotionally intelligent individuals can
understand complex feelings, such as simultaneous feelings of love and hate, or recognizing the
combination of fear and surprise as awe (Neubauer & Freudenthaler, 2005).
The fourth branch, Managing Emotions, refers to the ability to be aware of one's emotions.
Emotionally intelligent individuals can recognize negative emotions without repressing them.
Perhaps more adaptive is the ability to regulate emotions, and rather than act hastily on them,
harness the emotional experience as motivation in the future.
The four branches further combine into pairs to form two oblique second-order factors;
Experiential El and Strategic El. Experiential El represents the ability to accurately perceive,
respond to, and manipulate emotional information without necessarily understanding it (Mayer et
al., 2002). This factor is derived from the Perceiving, Appraising, and Expressing Emotions and
Using Emotions branches. Strategic El represents the ability to understand emotions and use
them strategically for planning and self-management (Mayer et al., 2002) and is derived from the
Understanding Emotions and Managing Emotions branches. These second-order factors combine
to yield a unitary third-order factor, Overall El (see Figure 1-2).
Measurement of Ability Models of El
In light of the psychometric and theoretical issues surrounding the use of self-report
measures, many researchers have focused their attention on measures predicated on the ability
model of EI, or ability measures (e.g., Caruso, Mayer, & Salovey, 2002; Mayer, Caruso, &
Salovey, 1999, 2000; Mayer & Salovey, 1997; Mayer, Salovey, & Caruso, 2000c; Mayer,
Salovey, Caruso, & Sitarenios, 2003). Mayer (2001) contends that, because g is typically
measured by ability measures, so too should the construct of El for it to fall within the domain of
intelligence. Mayer and colleagues assert that ability measures are better indicators of one's
maximal El performance, and that they most directly operationalize El as a cognitive ability.
At present, few ability measures exist. Among them, fewer still have evidence
substantiating their reliability or validity. For example, the Emotional Intelligence Scale for
Children (EISC; Sullivan, 1999) is not commercially available, in part due to difficulties
establishing the validity and reliability of the scale. The Emotional Accuracy Research Scale
(EARS; Mayer & Geher, 1996), which was developed to measure the identification of other's
emotions, demonstrated poor reliability as well as small and unstable relationships with external
criteria. The ability measure with the most empirical support is the Mayer-Salovey-Caruso
Emotional Intelligence Test (MSCEIT; Mayer, Salovey, & Caruso, 2002).
Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT)
The MSCEIT1 consists of eight subtests that are divided into four first-order factors, two
second-order factors, and Overall EI, and is designed to measure Mayer and Salovey's (1997)
four-branch ability model of EI. Mayer, Salovey, Caruso, and Sitarenios (2003) report findings
supporting the structure of the MSCEIT through confirmatory factor analysis with a sample of
1,985 adults. Although both the one- and two-factor solutions demonstrated good fit based on fit
indices, a constrained four-factor solution, in which the covariances between Perceiving and
Using Emotions and between Understanding and Managing Emotions were constrained to
equality, demonstrated the best fit.
Mayer et al. (2002) also found support for one-, two-, and four-factor solutions as well as
an integrated hierarchical model reflecting the structure of their four-branch ability model, with
1The original version of the MSCEIT, the Multi-Factor Emotional Intelligence Scale
(MEIS), underwent two revisions to provide a more psychometrically sound instrument that
takes less time to administer: the MSCEIT Version 1.1 (V1.1) and, most recently, the MSCEIT
Version 2.0, referred to throughout as the MSCEIT.
eight subtests associated with four first-order factors; first-order factors associated with two
second-order factors; and second-order factors associated with an overall third-order factor.
Gender, Racial/Ethnic, and Cross-Cultural Differences
Within the normative sample, women, on average, scored slightly higher than men on all
first-order factors of the MSCEIT, and those differences were all statistically significant (Mayer
et al., 2002). However, Livingstone and Day (2005) found that gender was only significantly
associated with the Perceiving Emotions scale. Small racial/ethnic group differences were also
noted. Mean scores obtained by White/non-Hispanics and Hispanic Americans were statistically
significantly higher than those obtained by African Americans and Asian Americans on Overall
EI, as well as on the Using Emotions, Understanding Emotions, and Managing Emotions factors.
These differences, however, were modest, with the largest difference explaining only 5% of the
variance in scores on the Managing Emotions scale (Mayer et al., 2002). In one study examining
cultural differences, no significant differences were found when comparing scores between
Americans and Germans, indicating that cultural differences did not significantly impact
performance (Lopes et al., 2004).
Relationship with Self-Report El Measures
In general, ability model measures of El are unrelated to most mixed model measures of El
(Barchard & Hakstian, 2004; Zeidner, Shani-Zinovich, Matthews, & Roberts, 2005). For
example, correlations between the MSCEIT and the Trait Meta-Mood Scale (TMMS; Salovey et
al., 1995) are nonsignificant (Lopes, Salovey, & Strauss, 2003; Warwick & Nettlebeck, 2004).
Brackett and Mayer (2003) reported low but significant correlations between the MSCEIT and
the EQ-i and SSRI. They further reported nonsignificant partial correlations between the
MSCEIT and the EQ-i and SSRI when controlling for the Big Five personality dimensions. In
general, self-report measures appear to tap different constructs than those measured by the
MSCEIT (Barchard & Hakstian, 2004). This distinction is important given the aforementioned
criticisms of self-report measures.
Relationship with Age and Experience
Intellectual capacities improve with age (Fancher, 1985). Therefore, Mayer et al. (1999)
claim that, if El is a cognitive ability, it should vary with age, improving from childhood into
adulthood. Mayer and colleagues found support for this hypothesis. Individuals ranging in age
from 18 to 24 years scored significantly lower, on average, than older groups (> 25 years) on
Overall El and three of the four first-order factors on the MSCEIT (Mayer et al., 2002).
Goldenberg et al. (2006) also found a positive and significant relationship between age and
Overall EI, Managing Emotions, and Understanding Emotions on the MSCEIT.
According to their theory, El is expected to increase with experience as well (Roberts,
Zeidner, & Matthews, 2001). For example, college students' critical thinking ability has been
found to change significantly with each year of college (Berger, 2005; King & Kitchener, 1994;
Pascarella, 1999; Perry, 1981). Students in higher education typically develop the ability to
identify and evaluate sources of information, think critically, problem solve, and effectively
communicate their ideas (Miller & Winston, 1990). For example, in a group of college and non-
college students matched on age, those attending college scored higher in critical thinking after
one year (Pascarella, 1999). Thus, some have argued that the college experience itself is largely
responsible for these changes in cognitive, moral, and psychosocial development (Pascarella &
Terenzini, 2005; Perry, 1981). In other words, certain experiences, such as years in formal
education, may lead to increases in EI. Day and Carroll (2004) found that years in college
correlated significantly and positively with Managing Emotions, Using Emotions, and Perceiving
Emotions on the MSCEIT. Goldenberg, Matheson, and Mantler (2006) also found that years of
education correlated significantly and positively with Overall EI, Managing Emotions, and
Convergent evidence of validity. According to theory, El should be positively related to
psychometric g based on the assumption that all cognitive abilities correlate to some extent (Van
Der Zee, Thij s, & Schakel, 2002). If, however, El and g correlate too highly, they would not
appear to be distinct constructs. Thus, their association should be, at most, moderate.
Mayer et al. (2002) claim that the relationship between the El and g is positive and low to
moderate. Results of research support this claim (r = .27 to .45). For example, Bastian, Bums,
and Nettelbeck (2005) found a significant moderate correlation between the MSCEIT and the
Raven's Advanced Progressive Matrices (RAPM; Raven, Court, & Raven, 1993). Brackett and
Mayer (2003) found a significant moderate correlation between the MSCEIT and Verbal SAT,
which is considered to be indicator of crystallized intelligence. In another study, the Wonderlic
Personnel Test (WPT; Wonderlic, 1992) correlated significantly with the MSCEIT (Schulte, Ree,
& Carretta, 2004). Zeidner et al. (2005) also found a significant correlation between the MSCEIT
and the Wechsler Intelligence Scale for Children-Revised (WISC-R; Wechsler, 1974)
Vocabulary subtest, which is highly correlated with total WISC-R scores and used as an
indicator for crystallized intelligence. Thus, El appears related to, yet distinct from, g.
Divergent evidence of validity. Schaie (2001) noted the importance of demonstrating
divergent evidence of validity among El measures (i.e., distinction from existing personality
measures), stating that research must "show that a new set of constructs is not simply ... an
alternative way of describing already established personality dimensions" (p. 245). Thus, to be
considered as a separate construct, correlations between El and personality should not be too
high (Van der Zee, Thij s, & Schakel, 2002).
Many personality instruments use forced-choice item formats to measure five basic
clusters of personality traits, known as the Big Five (McCrae & Costa, 2003). Extensive research
has supported the existence of the Big Five factors, which almost completely account for the
domain of personality descriptors (Goldberg, 1993). The Big Five are generally identified as
extraversion (outgoing and assertive), agreeableness (kind, easygoing, and generous),
conscientiousness (organized, conforming, and self-disciplined), emotional stability (relaxed and
at ease), and openness (imaginative, curious, and open to new experiences) (Goldberg, 1992).
Research indicates that El, as defined by the ability model, generally is distinguishable
from the factors of the Big Five. For example, Zeng and Miller (2003) reported nonsignificant
correlations between overall scores on the MSCEIT and the Big Five personality dimensions.
Some studies have demonstrated a low to moderate relationship between the MSCEIT and the
dimensions of Agreeableness and Openness to Experience (e.g., Brackett & Mayer, 2003;
Brackett, Mayer, & Warner, 2004; Day & Carroll, 2004); however, weighted mean correlations
across five studies also suggest positive and significant relationships with Emotional Stability,
Extraversion, and Conscientiousness (range = .06 to .11), although these relationships were weak
(Mayer et al., 2004a). Despite data suggesting significant correlations between El and some of
the Big Five personality dimensions, these relationships generally are low.
Predicting Important Outcomes
In addition to its moderate relationship with measures ofg and personality, some measures
of El have demonstrated associations with important real-life outcomes such as life satisfaction,
quality of social relationships, alcohol abuse, cigarette smoking, and academic success (Bar-On,
1997; Goleman, 1995; Mayer et al., 2002; Schutte et al., 1998; Van Der Zee et al., 2002). To
substantiate validity, if El is indeed within the domain of intelligence, high levels of El should
provide real-life adaptive benefits to an individual based on the demonstrated relationship
between cognitive ability and positive real-life outcomes (Matthews, Zeidner, & Roberts, 2005).
Surprisingly few studies have been conducted examining the predictive validity of some of the
most widely used and cited El measures (Gohm, 2004). In addition, studies that support their
predictive validity are rarely replicated (Brody, 2005).
A critical stipulation when determining the predictive power of these new measures is
determining their incremental validity, or the increased variance explained above and beyond
related constructs (e.g., g and personality). Few studies have examined this criterion (Brody,
2004); these studies are discussed below.
Academic success. One important criterion of interest is academic success. Goleman
(1995) claimed that EI-related skills are critical for academic achievement. Lopes and Salovey
(2004) agreed, suggesting that understanding emotional vocabulary may help children develop
language skills and interpret literature, and the ability to manage emotions can facilitate control
of attention, intrinsic motivation, and ameliorate anxiety associated with exams or difficult
assignments. Subsequently, American schools have seen an increase in programs claiming to
teach El and improve academic outcomes. In order to defend the inclusion of these programs,
one must demonstrate a significant relationship between El and academic outcomes.
The Understanding Emotions factor of the MSCEIT predicted math performance in both
males and females after controlling for general mental ability (Lyons & Schneider, 2005). Mean
scores on Overall EI, Understanding Emotions, and Managing Emotions were also significantly
higher for a sample of high-school gifted students than their non-gifted peers (Zeidner et al.,
2005). This is consistent with the proposed relationships of general intelligence and academic
achievement with giftedness and EI.
Psychological well-being. Ryff (1989a) derived a multidimensional model of
psychological well-being comprised of six distinct components of well-being, including
Autonomy (a sense of self-determination and independence), Environmental Mastery (the ability
to manage one's life and surroundings), Personal Growth (a feeling of continued growth and
development as a person), Positive Relations with Others (sustaining quality relationships),
Purpose in Life (the feeling that one's life is meaningful), and Self-Acceptance (positive
evaluations of oneself). Several factor analytic studies have supported the six-factor model of
psychological well-being as well as a single higher order factor (Clarke, Marshall, Ryff, &
Wheaton, 2001; Ryff & Keyes, 1995; Van Dierendonck, 2004).
Brackett and Mayer (2003) found a positive and significant relationship between the
overall composite scores on the Scales of Psychological Well-Being (SPWB; Ryff, 1989a) and
the MSCEIT. In another study, the Positive Relations with Others scale correlated significantly
with Managing Emotions on the MSCEIT (Lopes, Salovey & Strauss, 2003). Thus, El may
contribute to psychological health. However, the authors did not report the relationships between
the MSCEIT and the individual scales of the SPWB. Further, they did not examine this
relationship while controlling for g or personality, leaving the possibility for the significant
relationships due to covariance with related constructs. At present, no published data examining
the relationship between the six individual factors of psychological well-being and the MSCEIT
Peer attachment. Interpersonal relationships are important in the development of positive
self-image and personal identity (Koon, 1997). In addition, among adolescents and college
students, quality of peer attachment moderates important outcomes, including educational and
occupational achievement, school adjustment, college satisfaction, self-esteem, locus of control,
optimism, and competence (Cotterell, 1992; Fass & Tubman, 2002).
Lopes and colleagues (2003, 2004) found that higher mean scores on ability measures of
El are associated with more pro-social behavior and higher quality of relationships with peers
after controlling for the Big Five personality dimensions, g, and gender. Izard et al. (2001) found
that the ability to recognize and interpret emotional cues in facial expressions, similar to the
faces subtest of the MSCEIT, have long-term effects on social behavior and academic success
after controlling for verbal ability and temperament. Some argue that emotional knowledge
provides a foundation for effective communication and social relationships (Izard et al., 2001).
People must understand and manage their own emotions as well as perceive and understand
others emotions to navigate the social world (Lopes et al., 2004). Despite existing studies finding
generally consistent results, only a limited number of studies have addressed this relationship.
Further research is needed to confirm and clarify the relationship between emotional competency
on social relationships and peer attachment (Lopes et al., 2003; Lopes et al. 2004).
Cigarette and alcohol use. Smoking and drinking habits are associated with personality,
alexithymia, and lower than average scores on IQ measures (Kauhanen, Julkunen, & Salonen,
1992; Trinidad & Johnson, 2002; Von Knorring & Oreland, 1985). Considering the relationships
of El with those same variables, it seems reasonable to suggest that those with low levels of El
may be more prone to cigarette or alcohol abuse. Some researchers have hypothesized that those
who are more emotionally intelligent can cope with emotions and deal with social pressures to
help them resist substance abuse (Goleman, 1995; Trinidad & Johnson, 2002).
In one study with middle school students, Trinidad and Johnson (2002) found a significant
negative relationship between an earlier version of the MSCEIT and cigarette and alcohol abuse.
Although El accounted for only 2% to 4% of the variance associated with the use of cigarettes
and alcohol, these relationships remained significant after controlling for the effects of age,
gender, and grades. Further, Bracket, Mayer, and Warner (2004) found the MSCEIT to
incrementally predict drug use and alcohol use among males.
Criticism of Ability Models
Despite the research supporting ability models, several criticisms have been raised as well.
Structure of EI. The analysis of the factor structure of the MSCEIT conducted by Mayer
et al. (2003) indicated that a four-factor solution was the best fitting model. Gignac (2005),
however, argued that the one-, two-, and four-factor models were not appropriate models based
on an examination of goodness-of-fit indices. Generally, better fitting models are associated with
fit indices such as the Tucker-Lewis index (TLI) and normed fit index (NFI) that approach 1.0,
with values between .90 and .95 as indicative of only marginal fit (Hu & Bentler, 1999). Root
Mean Square Error of Approximation (RMSEA) values lower than .05 indicate good fit (Browne
& Cudeck, 1993). Gignac (2005) re-analyzed the Mayer et al. (2003) data and found the TLI was
below .90 for both the one- and two-factor solutions, and the RMSEA was .12 and .09 for the
one- and two-factor solutions, respectively, indicating poor fit.
Gignac (2005) also analyzed a supplementary four-factor solution by removing the
equality constraints from the Mayer et al. (2003) analysis, and found the four-factor solution to
exhibit good fit (X2(14, N= 1985) = 50.91, TLI = .96). However, this model yielded a non-
positive definite matrix, suggesting that the four-factor model was unacceptable. Gignac (2005)
further noted that the Perceiving and Using factors correlated .97, suggesting that they do not
measure separate constructs. As a result, Gignac (2005) analyzed a three-factor model with the
Perceiving and Using factors combined into one factor and found that this model was no worse
fitting than the unconstrained four-factor solution (X2(15, N= 1985) = 51.46; AX2(1, N= 1985)
=.55, p > .75). The three-factor model proposed was therefore the most parsimonious, whereas
the data did not support the proposed one-, two-, and four-factor solutions.
Palmer, Gignac, Manocha, and Stough (2005) examined the factor structure of the
MSCEIT with a sample of 450 adults. Similar to previous findings, both the one- and two-factor
solutions did not exhibit good fit, although the factor loadings were positive and statistically
significant on the general factor and second-order factors. Palmer et al. found the four-factor
model to exhibit good fit, although they reported a high correlation between the Using Emotions
and Managing Emotions factors (r = .90). A supplementary analysis with the four subtests
contributing to these two factors combined into one factor exhibited excellent fit and was not
worse fitting than the oblique four-factor solution (Ay2(3) = 4.95, p= .175). As a result, the four-
factor solution was not considered an appropriate model. Although this three-factor model differs
from Gignac's (2005) proposed three-factor model, both merged the Using Emotions factor with
another factor. Thus, results of this research indicate that the Using Emotions factor does not
appear to measure unique variance (Palmer et al., 2005).
Palmer et al. (2005) conducted additional CFAs using nested models. They found the best
fitting model to consist of a general factor model on which all eight subtests load saliently, a
nested orthogonal Perceiving Emotions factor derived from two subtests, nested oblique
Understanding Emotions and Managing Emotions factors derived from two subtests each, and
the removal of the Using Emotions factor based on consistently strong correlations with other
Finally, despite the authors reporting good fit for the hierarchical structure of the test,
which represents "the most direct test of the MSCEIT .. ." (Mayer et al., 2002, p. 63), these
analyses were not reported, or replicated, anywhere other than in the MSCEIT manual. In
addition, the authors only report goodness-of-fit indices without reporting chi-square statistics.
Gignac (2005) raised additional concerns, noting that in order to claim the existence of
four first-order factors, more tasks or subtests would need to be added. Although latent variables
consisting of only two observed variables can be analyzed using structural equation modeling,
they are more likely to yield indefinite positive matrices than latent variables with three or more
observed variables (Wothke, 1993). Mulaik and Millsap (2000) further argued that latent
variables with two or three positively correlated indicators will always yield excellent model fit
based on fit indices. Mayer et al. (2005) acknowledged that until these concerns are addressed,
selecting the most appropriate structural model of the MSCEIT should be postponed.
Given the incompatible findings among existing studies and concerns over too few
subtests, the structure of the MSCEIT remains unclear. Further, no existing studies have
replicated the analysis of the four-branch hierarchical structure of the test. Thus, additional
hierarchical confirmatory factor analyses are needed to guide the interpretation of scores.
Convergent evidence of validity. Despite reported low to moderate relationships between
El and g, the results of available research are somewhat conflicting. For example, although the
MSCEIT and WPT were moderately associated in one study (Schulte, Ree, & Carretta, 2004),
they demonstrated a non-significant relationship in another (Zeng & Miller, 2003). Salovey,
Mayer, Caruso, and Lopes (2003) also found non-significant correlations with the MSCEIT and
the Vocabulary subtest of the Wechsler Adult Intelligence Scale-Third Edition (WAIS-III;
Wechsler, 1997) and with self-reported SAT scores. Further, Warwick and Nettelbeck (2004)
found that only the Managing Emotions subtest correlated significantly with abstract reasoning
ability as measured by the Differential Aptitude Tests (DAP; Bennett, Seashore, & Wesman,
1989). Given the inconsistency and paucity of available data on the convergent validity of EI,
additional research on the EI's relationship with g is needed (Conte, 2005).
Divergent evidence of validity. Variability within the existing data prevents definitive
conclusions about EI's relationship with personality. For example, Openness to Experience has
shown inconsistent correlations with EI, demonstrating positive significant relationships in some
studies (e.g., Brackett & Mayer, 2003), and negative significant relationships in others (e.g.,
Lopes et al., 2003). Varied results such as these preclude definitive statements about the
divergent evidence of validity of EI. Thus, additional research is needed to examine the
relationship between the El and personality.
Development of EL Some research has found only minimal support for claims that El
increases with age and experience. For example, Day and Carroll (2004) found Perceiving
Emotions to be the only scale significantly related to age in years (r = -.14), with younger
participants surprisingly scoring higher. In their study, participants consisted of undergraduate
students only, ranging in age from 17 to 54. In addition, Gohm and Clore (2002) found no
increase in El with years in college. Generally, the investigations examining experience and age
differences in El have been both meager and contradictory (Matthews et al., 2002). Thus,
additional research is needed to examine this relationship between age, experience (notably years
in higher education), and El (Day & Carroll, 2004).
Predicting academic success. Of the few studies examining the relationship between El
and academic success, most have used self-report measures, and of those, most have neglected to
control for cognitive ability or personality, or both (Barchard, 2003; Bar-On, 1997; Matthews et
al., 2002; Schutte et al., 1998; Van Der Zee et al., 2002). One study found that scores on the
MSCEIT predicted academic success (Brackett & Mayer, 2003). However, the MSCEIT
demonstrated negligible incremental validity for academic success after controlling for
personality and cognitive ability. Further, Bastian et al. (2005) found a nonsignificant
relationship between the MSCEIT and academic success without controlling for cognitive ability
and personality. More research is needed to examine the relationship between El and academic
Predicting cigarette and alcohol use. Brackett and Mayer (2003) examined the ability of
the MSCEIT, EQ-i, and the SSRI to predict drug, cigarette, and alcohol use. Among these three
measures and three dependent variables, the only significant relationship when controlling for
the Big Five personality dimensions and g was between the EQ-i and alcohol use. Given the
inconsistency of results, further research is needed to examine these relationships.
Though research on El is still in its infancy (Daus & Ashkanasy, 2005), a rapid increase in
research on El has taken place over the past decade, perhaps due to claims of EI's predictive
validity as well as the desire to expand our understanding of human abilities. Consequently,
several distinct theories of El have emerged. Among them, the ability model of El (Mayer &
Salovey, 1997), as measured by the MSCEIT, has received the most empirical support.
However, questions remain regarding its validity and internal structure. As a result, the ability
model of El requires more research before it can be considered a valid construct.
Purpose of the Current Study
The aim of this study is to further examine the validity of El with a sample of university
students. The specific research questions that will be addressed are (1) To what degree does El
correlate with intelligence, or g, and the Big Five personality characteristics? (2) Does El
correlate with age and years in college? (3) Do different groups (racial/ethnic group, gender, and
year of study) perform differently on an ability measure of EI? (4) Does El incrementally predict
real-life outcomes beyond that predicted by intelligence and personality? (5) Is the proposed
factorial structure of the four-branch ability model of El supported by hierarchical confirmatory
70 narrow abilities
Figure 1-1. Carroll's Three-Stratum Theory of intelligence.
Figure 1-2. Factor structure of the four-branch ability model of EI.
Participants consisted of 150 undergraduate students (62 freshmen, 43 sophomores, 24
juniors, and 21 seniors; 40 men and 110 women) from over 30 fields of study. The age of
participants ranged from 17-36 years (M = 19.7, SD = 2.1). The majority of respondents were
White/non-Hispanic (68.7%), followed by African Americans (14.7%), Hispanic Americans
(9.3%), Asian Americans (5.3%), Other (1.3%), and Native Americans (0.7%). All participants
were enrolled in a four-year public university in North Central Florida. Students had the option
to participate in the study to receive research credit or extra credit for an undergraduate course.
All participants were treated in accordance with the "Ethical Principles of Psychologists and
Code of Conduct" (American Psychological Association, 2002).
Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT)
Emotional intelligence (EI) was measured with the MSCEIT. Normative data was collected
from 5,000 respondents from 50 research sites. Most of the data came from U.S. sites, although
the United Kingdom, Canada, Malta, South Africa, Australia, Switzerland, Scotland, the
Philippines, India, Slovenia, and Sri Lanka also contributed to normative data collection with
English speaking respondents (Mayer et al., 2002). This measure takes approximately 30-40
minutes to complete and is administered via the internet to individuals age 17 and above, where
their responses are computer scored by the test publisher.
The MSCEIT includes 141 items measuring an individual's ability to solve emotional
problems on eight separate tasks, divided into four factors (i.e., Perceiving Emotions, Using
Emotions, Understanding Emotions, and Managing Emotions). The Perceiving Emotions factor
is comprised of the Faces task and the Pictures task. These subtests examine how emotions are
perceived and appraised in faces and pictures/designs. Examinees are asked to indicate the level
of emotion expressed in several images of faces and designs using the following emotions:
anger, sadness, happiness, disgust, fear, surprise, and excitement.
The Using Emotions factor is made up of the Sensations and Facilitation tasks. These
subtests require examinees to describe emotions using non-emotional vocabulary and identify
specific feelings or emotions that may improve or interfere with one's performance on cognitive
or behavior tasks. Specifically, in the Sensations task, participants must generate an emotion and
match sensations to them. In the Facilitation task, participants are asked to match the moods that
typically accompany specific cognitive tasks or behaviors.
The Understanding Emotions factor is made up of the Blends and Changes tasks. These
subtests ask examinees about their reasoning and understanding of emotions. Specifically, in the
Blends task, participants identify emotions that, when combined, form other emotions. The
Changes task asks respondents to identify an emotion that is the result of the intensification of
Finally the Managing Emotions factor is measured by the Emotion Management and
Social Management tasks. These tasks provide examinees with situations and ask for the best
social response for managing the feelings of the situation (Mayer, 2001). Specifically, the
Emotion Management task provides participants with a story, and asks them to select the action
that would be most effective in reaching a specified emotional outcome. The Social Management
task requires participants to select the most effective actions for an individual to use in the
management of another's feelings.
Reliability. Mayer et al. (2002) reported split-half reliabilities of .79 to .91 for the four
first-order factors and .91 for Overall El. Brackett and Mayer (2003) reported test-retest
reliability after a period of three weeks at .86 for the total MSCEIT. Mayer et al. (2003),
however, found low internal consistency estimates for the scales, ranging from .55 to .88.
Validity. Results of research generally support the validity of the MSCEIT. Several studies
have found significant and moderate correlations between the MSCEIT and g-loaded tests,
demonstrating convergent evidence of validity and suggesting that El is a unique construct from
g (Bastian, Burns, & Nettelbeck, 2005; Brackett & Mayer, 2003; Schulte, Ree, & Carretta, 2004;
Zeidner et al., 2005). In addition, several studies have found that the MSCEIT generally is
distinguishable from personality (Brackett and Mayer, 2003; Brackett, Mayer, & Warner, 2004;
Day & Carroll, 2004; Zeng & Miller, 2003). The MSCEIT has also been found to correlate
significantly with several important outcomes, including academic success (Lyons & Schneider,
2005; Zeidner et al., 2005), psychological well-being (Brackett & Mayer, 2003; Lopes, Salovey
& Strauss, 2003), peer attachment (Lopes et al., 2003; Lopes et al., 2004), and cigarette and
alcohol use (Bracket, Mayer, & Warner, 2004; Trinidad & Johnson, 2002). Perhaps more
importantly, the MSCEIT uniquely contributed to the prediction of academic success, peer
attachment, and cigarette and alcohol use after controlling for g, personality, or both (Bracket et
al., 2004; Lopes et al., 2003; Lopes et al., 2004; Lyons & Schneider, 2005; Trinidad & Johnson,
International Personality Item Pool (IPIP)
Personality was assessed using the IPIP (Goldberg, 1999) personality scale. The IPIP is a
100-item scale which measures the Big Five personality dimensions (Extraversion,
Agreeableness, Conscientiousness, Emotional Stability, and Openness to Experience). The IPIP
has 20 items for each of the five personality traits. The IPIP was created to serve the same
empirical and applied functions of the NEO-FFI (Costa & McCrae, 1992), which is among the
most commonly used and cited measures of the Big Five (Cucina, Goldenberg, & Vasilopoulos,
2005). The internal consistencies of the scales reported by Goldberg are high: Extraversion (a =
.91), Agreeableness (a = .88), Conscientiousness (a = .88), Emotional Stability (a = .91), and
Openness (a=.90). When corrected for attenuation, correlations with the respective scales on the
NEO-FFI range from .88 to .93. This strong relationship was supported by Lim and Ployhart
(2002), who also found support for the five-factor model of personality through a confirmatory
factor analysis of the IPIP.
Wonderlic Personnel Test (WPT)
The WPT (Wonderlic, 1992) is a widely-used, objective measure of general cognitive
ability (g). It can be administered to adults individually or in groups. The WPT is a timed test (12
minutes) consisting of 50 items that increase in difficulty as the test progresses, yielding one total
score. Test item subject areas include verbal, quantitative, and spatial ability.
The reliability and validity of the WPT is well established (Geisinger, 2001). The WPT
manual reports alternate-forms reliability coefficients ranging from .73 to .95, split-half
reliability estimates from .88 to .94, and test-retest reliability estimates from .84 to .94 after up to
a five-year period (Dodrill, 1983; Schoenfeldt, 1985; Wonderlic, 1992). Evidence of concurrent
validity has been established with various measures of intelligence. For example, the WPT
correlates between .91 and .93 with the Wechsler Adult Intelligence Scale (WAIS; Wechsler,
1955), and between .56 and .80 with aptitude G (general learning ability) of the General Aptitude
Test Battery (Dodrill, 1981; McKelvie, 1992; Wonderlic, 1992; Mullane & McKelvie, 2001). In
addition, the WPT scores correlate .86 with the Verbal IQ scale of the Wechsler Adult
Intelligence Scale Revised (WAIS-R; Wechsler, 1981) and .84 with the Performance IQ scale
(Hawkins, Faraone, Pepple, Seidman, & Tsuang, 1990). Furthermore, the WPT was found to
correlate .66 with the Composite IQ score of the Kaufman Adult Intelligence Test (KAIT;
Kaufman & Kaufman, 1993), .62 with the Crystallized IQ score, and .54 with the Fluid IQ score
(Bell, Matthews, Lassiter, & Leverett, 2002).
Academic success. Participants were asked to provide self-reports of college grade point
average (GPA) and total SAT/ACT score as an indicator of academic success. When both SAT
and ACT scores are reported, only SAT scores were used for analysis. For students reporting
only ACT scores, their score was converted into an approximate equivalent SAT score based on
concordance tables provided by the College Entrance Examination Board (Dorans, 1999).
Dorans reported a strong relationship between the ACT and SAT (r = 0.92). In a study
investigating the accuracy of self-reported estimates of GPA and SAT scores among
undergraduate students, reported GPA and SAT scores correlated significantly with official
records (r = .97 and .88, respectively) (Cassady, 2001).
Psychological well-being. To assess psychological well-being, participants completed the
Scales of Psychological Well-Being-Short Form (SPWB-SF; Ryff, 1989a). This survey consists
of 43 items designed to measure six dimensions of well-being: Autonomy, Environmental
Mastery, Personal Growth, Positive Relations with Others, Purpose in Life, and Self-Acceptance.
Participants were asked to respond to various statements and indicate on a six-point Likert scale
how true each statement is of them. Higher scores indicate greater well-being. The internal
consistencies of the scales for a sample of 202 adults are moderate to good: Autonomy (a = .77),
Environmental Mastery (a = .80), Personal Growth (a = .81), Positive Relations with Others (a =
.79), Purpose in Life (a = .86), and Self-Acceptance (a = .84) (C.D. Ryff, personal
communication, January 18, 2006). In a longitudinal study consisting of 6,875 adults, the internal
consistencies of the scales reported were moderate: Autonomy (a = .70), Environmental Mastery
(a = .72), Personal Growth (a = .76), Positive Relations with Others (a = .78), Purpose in Life (a
=.79), and Self-Acceptance (a = .79) (C.D. Ryff, personal communication, January 18, 2006).
Test-retest reliability estimates are available only for the 20-item parent scales, and range
from .81 to .88 (Ryff, 1989b). Correlations between the 20-item scales and related measures of
positive functioning (i.e., life-satisfaction, self-esteem, affect balance, internal control, and
morale) were all positive and significant; whereas correlations with measures of negative
functioning (i.e., external locus of control, depression) were negative and significant (Ryff,
Peer attachment. To assess quality of social relationships, participants completed Section
III of the Inventory of Parent and Peer Attachment (IPPA; Armsden & Greenberg, 1987). This
section contains 25 items on a five-point Likert scale focusing on peer relationships. Section III
of the IPPA is a self-report measure with questions regarding relationships that fall under three
emerging factors: Trust (10 items; a = .91), Communication (8 items; a = .87), and Alienation (7
items; a = .72). These factors yield an overall score, referred to as Peer Attachment (a = .92). To
obtain an overall Peer Attachment score, negatively worded items were reverse-scored and then
the response values in each section were summed together. Three-week test-retest reliability for
the IPPA Peer Attachment scale was .86.
Evidence of criterion-related validity reported by the authors indicates that the IPPA Peer
Attachment scale correlates with related measures of social relationships. Scores correlated .57
with the Tennessee Self-Concept Scale (TSCS; Fitts, 1965) and .32 with the Peer Utilization
factor from the Inventory of Adolescent Attachment (Greenberg et al., 1984). In addition, the
IPPA was found to correlate with several affective states as measures by Bachman's Affective
States Index (Bachman, 1970). Specifically, scores correlated positively and significantly with
Self-Esteem and Life-Satisfaction; scores correlated negatively and significantly with Alienation,
Irritability/Anger, Depression/Anxiety, and Guilt (Armsden & Greenberg, 1987).
Cigarette and alcohol use. To measure alcohol use, participants completed the Alcohol
Use Disorders Identification Test (AUDIT; Babor, Higgins-Biddle, Saunders, & Monteiro,
2001). This measure was developed to identify harmful drinking habits and contains 10 items
pertaining to alcohol use and dependency. Response format is on a five-point Likert scale. The
AUDIT has demonstrated concurrent validity with other measures of alcohol use/dependency,
and has also successfully differentiated among hazardous and non-hazardous drinkers in college
populations (Bohn, Babor, & Kranzler, 1995; Kokotailo et al., 2004). In a study comparing
scores on the AUDIT with DSM diagnostic criteria among college students, the AUDIT showed
high sensitivity, correctly identifying 94% of the sample who had met diagnostic criteria as high-
risk alcohol users, and moderate specificity, correctly classifying approximately two-thirds of the
individuals who did not meet diagnostic high-risk alcohol use (Fleming, Barry, & MacDonald,
To measure cigarette/tobacco use, participants were asked to provide information on the
number of packs of cigarettes smoked in one week, the number of cigarettes smoked per day,
total years of smoking, and how soon after waking from sleep is the first cigarette smoked.
Responses were rated on a five-point Likert scale to produce a total raw score ranging from 0 to
Demographic variables. Information was collected on age, gender, and racial/ethnic
group. In addition, students were asked about their major field of study, their current year of full-
time study, and their current membership/leadership in campus groups or clubs.
Participants were first asked to complete a demographic questionnaire, including
information relevant to academic success (i.e., GPA and SAT/ACT scores). Following the
demographic questionnaire, participants were administered the WPT, SPWB-SF, AUDIT and
cigarette use questionnaire, Section II of the IPPA, the 100-item IPIP personality scale, and the
MSCEIT in randomized order using a balanced Latin-square to control for order effect.
To assure full effort and truthfulness, all questionnaires, surveys, and protocols were
completed anonymously. In order to preserve anonymity, students were assigned an individual
identification numbers for the remainder of the battery. Testing was group administered by the
author in one session lasting approximately one hour.
Descriptive statistics were calculated for all demographic, outcome, and predictor
variables. Internal consistency (a) was calculated for all eight subtests, and first-, second- and
third-order factor scores of the MSCEIT.
Pearson product-moment correlation coefficients were calculated to examine relationships
between all variables and scores on the MSCEIT.
A factorial analysis of variance (ANOVA) was conducted with a = .05 to examine group
differences (racial/ethnic group, gender, and year of study) on second- and third-order factors.
Confirmatory factor analyses (CFA) of the MSCEIT were conducted. The MSCEIT
consists of 141 items designed to measure four first-order factors, each measured by two
subtests. Mayer et al. (2002) theorized that the four first-order factors would correlate positively,
and yield a general El factor as well as a two-factor model divided into Experiential El and
Strategic EI. Thus, one-, two-, and four-factor oblique models were compared to a three-factor
model that reflects alternative interpretations of the MSCEIT (Gignac, 2005). The one-, two-,
and four-factor models also were compared to each other. A CFA was also conducted for a
general factor model with a nested orthogonal Perceiving Emotions factor, and oblique
Understanding Emotions and Managing Emotions factors to address additional interpretations
(Palmer et al., 2005). Finally, a general factor model with nested second- and third-order factors
was analyzed, which reflects the Mayer et al. (2002) four-branch theoretical model and the
MSCEIT. Goodness of fit indices used for this analysis include change in chi-squared (AX2)
along with degrees of freedom, normed fit index (NFI), Bentler's comparative fit index (CFI),
and the root mean square error of approximation (RMSEA). Absolute close-fit values < .06 and
incremental close-fit values > .95 were considered satisfactory.
A series of multiple regression analyses also were conducted to examine the incremental
validity of the Overall El score over the overall WPT score and the five IPIP personality
dimensions in predicting the criterion variables (academic success, psychological well-being,
Positive Relations with Others scale of the SPWB, substance use, and peer attachment). The
WPT was entered in the first step, followed by the Big Five personality dimensions in the second
step, and the MSCEIT in the third step. The order of predictor variables in the model was
selected to determine whether the addition of El after g and personality will add significantly to
the prediction of the criterion variables. Analyses will be conducted at the (p < .05) level of
significance. Because up to three measures will be used in the regression analysis, we can expect
increments in R to be small, although this does not necessarily indicate lack of contribution to
incremental prediction of the criterion (Hunsley & Meyer, 2003).
The results of this study are presented in five sections. The first section presents
descriptive statistics for all of the variables investigated in this study. The second section
includes results of correlation analyses used to investigate significant relationships between
emotional intelligence (EI) and age, outcome variables, and predictor variables. The following
sections present results for factorial ANOVA, confirmatory factor analysis, and hierarchical
Table 3-1 displays descriptive statistics for the outcome variables measured in the study.
The mean scores for self-reported SAT (M 1186) and GPA (M 3.25) are slightly higher than
national averages for undergraduate students (SAT M = 1026; GPA M 3.18) (National Center
for Education Statistics). The mean scores reported for the AUDIT and Survey of Cigarette Use
are raw scores. A maximum of 40 points is possible for the AUDIT, and 16 for the Survey of
Cigarette Use. High scores indicate a higher level of substance use. For the Survey of Cigarette
Use, the sample was restricted in range.
The mean scores for the IPPA are raw scores, with higher scores indicating a greater
relationship/attachment to peers. The means and SDs of the raw scores on the SPWB are
reported for the individual scales and total PWB. A maximum of 48 points is possible for the
Purpose in Life scale, and 42 for the remaining five scales. Scores were highest for the Positive
Relations with Others scale (M 36.3) and lowest in Autonomy (M 30.4).
Descriptive statistics of predictor variables are presented in Table 3-2. For this sample, the
mean and standard deviation of the WPT (M 23.4, SD = 4.9) were comparable to those from
the standardization sample (M= 21, SD = 7) as well as values obtained in previous studies (see
Furnham & Petrides, 2003; Zeng & Miller, 2003). Scores on the WPT were calculated by
summing correct responses given within the allotted time. A maximum score of 50 points is
possible for the WPT; however, few individuals (less than 1%) generally attempt more than 40
items on the test (Geisinger, 2001).
Each of the five dimensions of personality were measured on a 100-point scale. Scores
generally were highest for Agreeableness (M 82.6, SD = 9.2), followed by Openness to
Experience (M 74.9, SD = 10.7), Conscientiousness (M 74.2, SD = 13.7), Extraversion (M=
71.9, SD = 13.6), and Emotional Stability (M 68.2, SD = 13.9).
Standard scores reported for the MSCEIT were uncorrected for age, gender, or education
level. As can be seen from Table 3-2, significant scatter exists among subtests. Specifically, the
mean score and standard deviation on the Faces subtest (M 113, SD = 24.2) was over 0.5
standards deviations higher than the average subtest score of 100 obtained from the
standardization sample. All other scores were within 0.3 standard deviations of the mean score of
100. Internal consistency for the instrument was calculated using Cronbach's alpha. Coefficients
for the eight subtests ranged from .45 to .90, indicating that the subtests had moderate to high
internal consistency. Coefficients for the four scale-level scores (a = .72 to .89), two area-level
scores (a = .81 to .88), and Overall El score (a = .89) indicated acceptable internal consistency.
These findings are consistent with those of Mayer et al. (2002).
Mean Score Differences by Race/Ethnicity, Gender, and Year of Study
Mean scores on the MSCEIT by race/ethnicity are presented in Table 3-3. Mean scores
among participants who identified themselves as American Indian (n = 1) or Other (n = 2) are
not included in the description below due to their low representation within the sample.
Mean scores among Hispanic Americans were higher on Perceiving Emotions, Managing
Emotions, Strategic EI, Experiential EI, and Overall El than any other identifiable groups. Mean
scores among African Americans were higher on Using Emotions and Experiential EI, and were
lower on Understanding Emotions, Managing Emotions, and Strategic El than other racial/ethnic
group's mean scores. Mean scores among Asian Americans were lower on Perceiving Emotions,
Using Emotions, and Experiential Emotions than the other group's mean scores. However, none
of these differences reached statistical significance. The effects of race/ethnicity are further
addressed in Tables 3-12 through 3-14.
Mean scores on the MSCEIT by gender are presented in Table 3-4. Women generally
scored the same or higher than men. These differences did not reach statistical significance. The
effects of race/ethnicity are further addressed in Tables 3-12 through 3-14.
Mean scores on the MSCEIT by year of study are presented in Table 3-5. Mean scores
among students in their first year of study were higher on Perceiving Emotions, Understanding
Emotions, Experiential Emotions, and Overall EI. Mean scores among students in their fourth
year of study were higher on Managing Emotions and Strategic EI. However, none of these
differences reached statistical significance. The effects of race/ethnicity are further addressed in
Tables 3-12 through 3-14.
Table 3-6 presents correlations between age in years and MSCEIT raw scores. As can be
seen from the table, age did not significantly correlate with performance on the MSCEIT.
Although age ranged from 17 to 36 years, the sample was restricted in range compared to the
normative sample and the age range for which this measure was designed (M= 19.7, SD = 2.1).
Table 3-7 presents correlations between academic indicators (self-reported SAT and GPA,
and WPT) and the MSCEIT. As can be seen, Understanding Emotions and Strategic El
correlated significantly with all three academic indicators. In addition, Overall El correlated
moderately and significantly with self-reported SAT (r = .31, p < .01) which often is used as an
indicator of Gc, and the WPT (r = .24, p < .01).
GPA generally correlated lower with the MSCEIT than SAT or the WPT. Notably, self-
reported GPA was not associated with Overall El. This may be due to grade inflation among
undergraduate students and subsequent decrease in variability (SD = 0.47). However, the
correlations between El and the objective measures of academic performance, SAT and the
WPT, were consistent with one another.
Table 3-8 presents correlations between the Big Five personality dimensions and the
MSCEIT. As can be seen in this table, Agreeableness correlated significantly with first-order
factor, second-order factor, and Overall El scores except the Perceiving Emotions factor.
Openness to Experience correlated significantly, although low-to-moderately, with
Understanding Emotions, Managing Emotions, Strategic El, and Overall El. These findings
generally are consistent with previous studies suggesting low-to-moderate associations between
the MSCEIT and Agreeableness and Openness to Experience (e.g., Brackett & Mayer, 2003;
Brackett, Mayer, & Warner, 2004; Janovics & Christiansen, 2001; Mayer et al., 2004a). In
addition, Emotional Stability correlated significantly, although moderately, with Using
Emotions, Experiential El, and Overall El. Conscientiousness correlated significantly with
Table 3-9 presents correlations between the SPWB and performance on the MSCEIT.
Positive Relations with Others correlated significantly with the MSCEIT. These findings are
consistent with previous studies that measured the Positive Relations with Others scale as an
indicator of peer attachment (e.g., Lopes et al., 2003). Self-Acceptance correlated significantly
with Using Emotions, Managing Emotions, Experiential EI, and Overall EI. Environmental
Mastery correlated significantly with Using Emotions, Managing Emotions, and Experiential EI.
The Autonomy scale correlated significantly, although negatively, with Understanding Emotions.
Purpose in Life correlated significantly with Managing Emotions. Total PWB correlated
significantly with Using Emotions, Managing Emotions, Experiential EI, and Overall EI.
Table 3-10 presents correlations between peer attachment and performance on the
MSCEIT. As can be seen from the table, Managing Emotions, Strategic EI, and Overall El
correlated significantly, although low to moderately, with peer attachment.
As can be seen in table 3-11, reported alcohol use significantly correlated with
performance on the MSCEIT. Specifically, higher scores on the MSCEIT are associated less
alcohol use. Cigarette use did not significantly correlate with performance on the MSCEIT.
Factorial Analysis of Variance (ANOVA)
Tables 3-12 to 3-14 display results of factorial ANOVAs conducted to examine group
differences (gender, race/ethnicity, and year of study) in performance on the MSCEIT. Initial
analyses were conducted for Strategic EI, Experiential EI, and Overall El scores. As can be seen
from the tables, the analysis resulted in no significant main or interaction effects for the
Levene's test was used to examine the assumption of homogeneity of variances. This
assumption was not met. When race/ethnicity was removed from the ANOVA, evidence for
violation of this assumption was not significant. However, results of the analysis still indicated
no main or interaction effects for gender or year of study.
Confirmatory Factor Analysis
To examine the latent structure of the MSCEIT, the data were analyzed using latent
variable structural equation modeling with the LISREL 8.7 program. Standard scores from all
eight subtests were entered into the analysis. Six individual models were tested: (a) a general
factor model (Figure 3-1), (b) an oblique two-factor model consisting of Strategic El and
Experiential El (Figure 3-2), (c) an oblique four-factor model consisting of the Perceiving,
Using, Understanding, and Managing Emotions factors (Figure 3-3), (d) an oblique three-factor
model combining the Perceiving Emotions and Using Emotions factors into one factor (Figure 3-
4), (e) a general factor model with a nested orthogonal Perceiving Emotions factor, and oblique
Understanding Emotions and Managing Emotions factors (Figure 3-5), and (f) a general factor
model with two nested second-order factors and four first-order factors (Figure 1-2).
Tables 3-15 and 3-16 display the intercorrelations between MSCEIT subtests, first-,
second-, and third-order factor observed scores. Table 3-17 displays first-order inter-factor
Tables 3-18 and 3-19 display standardized parameter estimates and goodness-of-fit indices
for the one-, two-, four-, three-, and nested factor models. As can be seen in Table 3-18, the one-
factor model (Figure 3-1) yielded a 2 (20) = 76.79, as well as non-satisfactory levels of fit (NFI
=.70, CFI = .75, RMSEA = .14). However, as can be seen in Table 3-17, all factor loadings on
the general El factor were positive and significant, ranging from .34 to .58.
The oblique two-factor model (Figure 3-2) yielded a X (19) = 46.68. This model was
significantly better fitting than the one-factor model (AX2 (1) = 30.11, p < .001). In addition, all
factor loadings were positive and significant. However, the chi-square was significant (p < .001),
indicating that this model was not well fitting. Further, this model yielded non-satisfactory
goodness-of-fit indices (NFI = .81, CFI = .88, RMSEA = .09).
Gignac (2005) supported an oblique three-factor model (Figure 3-4) that combined the
Perceiving and Using Emotions factors into one factor. This was based on a high correlation
between the two factors (.97), and resulted in an excellent fitting model. For this sample, the
inter-factor correlation was not as high (.78), although this relationship was the strongest among
the factors and justified exploring the three-factor model proposed by Gignac (2005). The
oblique three-factor model yielded a X2 (17) = 17.78, which was not statistically significant
(p=.40), indicating that the model was excellent fitting (NFI = 93, CFI = 1.00, RMSEA = .01).
All factor loadings were positive and significant. In addition, this model was not worse fitting
than the oblique four-factor model (AX2 (3) = 3.69).
The general factor model with a nested orthogonal Perceiving Emotions factor and oblique
Understanding and Managing Emotions factors (Figure 3-5) was analyzed based on Palmer et al.
(2005). This model yielded a X2 (16) = 10.77. This was not statistically significant (p = .80),
indicating that the model was excellent fitting (NFI= .96, CFI = 1.0, RMSEA = 0.0).
Finally, the four-branch model (Figure 1-2) consisting of a general factor, two nested
second-order factors, and four nested first-order factors was analyzed to assess the
appropriateness of the theoretical structure of the MSCEIT. This model yielded a X2 (16) = 16.07.
This was not statistically significant (p = .45), indicating that the model was excellent fitting
(NFI= .94, CFI = 1.0, RMSEA = 0.0). However, as can be seen in Table 3-18, two factor
loadings were greater than 1.0, indicating negative residual variance. Thus, the four-branch
theoretical model of the MSCEIT four is an improper model.
Several hierarchical multiple regression analyses were conducted to examine whether El
predicts a significant portion of the variance accounted for, beyond IQ and personality, in the
prediction of the criterion variables (academic success, PWB, the Positive Relations with Others
scale of the SPWB, peer attachment, alcohol use, and cigarette use). General intelligence as
measured by the WPT was entered in the first block. The Big Five personality dimensions were
entered into the second block. Overall El from the MSCEIT was entered into the third block.
Each regression took the same form.
Tables 3-20 and 3-21 display results of the multiple hierarchical regressions for academic
success (i.e., GPA and SAT, respectively). As can be seen from Table 3-20, IQ was a minor,
although significant predictor of GPA (R2 = .03, p < .05). When the Big Five personality
dimensions were added in the second step, they predicted an additional 13 percent (p < .01) of
variance in GPA. Overall El did not significantly predict additional variance when added in the
third step (p > .05). The full model was responsible for only 12% of the explained variance in
self-reported GPA. These findings are consistent with those found in previous studies (e.g.,
Table 3-21 indicates that IQ was a significant predictor of SAT (R2 = .49, p < .01). When
the Big Five personality dimensions were added in the second step, they only predicted an
additional 7% of the variance in SAT (p < .01), with Openness to Experience as the only
significant dimension (p < .01). Overall El did not significantly predict additional variance when
added in the third step (p > .05). The full model was responsible for 54% of the variance in self-
Table 3-22 displays results of the multiple hierarchical regression for the total PWB score.
IQ was a minor, although significant predictor ofPWB (R2 = .03, p < .05). Interestingly, betas
for WPT were negative, suggesting that those with lower IQ experience higher levels of PWB.
When the Big Five personality dimensions were added in the second step, they predicted an
additional 58% (p < .01) of variance in PWB, with Conscientiousness, Extraversion, and
Emotional Stability emerging as significant predictors in the model. Overall El did not
significantly predict additional variance when added in the third step (p > .05). In the full
regression equation, IQ was no longer a significant predictor of PWB. The full model was
responsible for 60% of the explained variance in PWB.
Positive Relations with Others
Table 3-23 displays results of the multiple hierarchical regression for the Positive
Relations with Others scale of the SPWB. Although the Positive Relations with Others scale is
only an individual scale within the SPWB, it has been used in other studies as an indicator of
peer relationships (see Lopes et al., 2003). IQ was a minor, although significant predictor of
PWB (R2 = .03, p < .05). Betas for WPT were negative, suggesting that those with lower IQ
have more positive relations with others. When the Big Five personality dimensions were added
in the second step, they predicted a significant amount of additional variance (R2 = .46, p < .01)
in Positive Relations with Others. Overall El significantly predicted only 1 percent of additional
variance when added in the third step (p > .05). In the full regression equation, IQ was no longer
a significant predictor of Positive Relations with Others. The full model was responsible for 49%
of the variance in predicting Positive Relations with Others.
Table 3-24 displays regression results for peer attachment. As can be seen in the table, IQ
was not a significant predictor of peer attachment (p > .05). When the Big Five personality
dimensions were added in the second step, they predicted an additional 17% (p < .01) of the
variance. Following these gains, Overall El did not significantly account for additional variance
in the prediction of peer attachment. Extraversion and Conscientiousness were the only
significant predictors in the full regression equation. The overall model was responsible for 16%
of the variance in predicting peer attachment.
Table 3-25 displays regression results for cigarette use. None of the predictor variables
accounted for a significant amount of explained variance in cigarette use. However, the beta
value for IQ was negative and significant (p < .05) in the second step of the regression,
indicating that those with lower levels of IQ may be more prone to cigarette use. These findings
should be viewed with caution, however, due to the restriction of range within the sample.
As can be seen in Table 3-26, IQ was not a significant predictor of alcohol use (p > .05).
When the Big Five personality dimensions were added in the second step, they predicted an
additional 13% (p < .01) of the variance. Following these gains, Overall El significantly
accounted for an additional 4% of unique variance (p < .01) in alcohol use. Agreeableness and
Overall El were the only significant predictors in the full regression equation. These findings are
consistent with previous studies (e.g., Brackett et al., 2004; Trinidad & Johnson, 2002). Betas
were negative for both Agreeableness and Overall EI, and suggest that individuals who use
alcohol more frequently are less agreeable and have lower levels of EI. The full model was
responsible for 14% of the variance in predicting alcohol use.
Table 3-1. Descriptive statistics for outcome variables.
Autonomy 150 30.4
Environmental 150 31.7
Personal 150 35.7
Positive 150 36.3
Purpose in 150 40.3
Self- 150 34.0
Total PWBf 150 208.4
Note: aMaximum score = 40; bMaximum score
dMaximum score = 42; eMaximum score = 48;
=16; cMaximum score =
fMaximum score =258.
Table 3-2. Descriptive statistics for predictor variables.
Variable M SD Range
WPTa 23.4 4.9 11-37
Extraversionb 71.9 13.6 40-99
Social Mgmt.C 96.4
Using Emotions' 100.5
Experiential EIC 102.7
Strategic EIC 97.4
Overall EIC 100.1
Note: N= 150; norm groupM 21, SD
7; bMaximum score =
groupM= 100, SD
Table 3-3. Mean scores on the MSCEIT by race/ethnicity.
American Asian African Hispanic White/non
Indian American American American -Hispanic Other
(n = 1) (n = 8) (n = 22) (n = 14) (n = 103) (n = 2)
Emotions 95.6 102.4 108.9 109.5 102.7 108.5
Using Emotions 73.7 99.5 105.0 100.7 99.8 101.1
Understanding 107.1 99.7 91.7 99.4 98.3 103.2
Managing 110.4 100.5 93.7 100.8 96.3 95.8
Experiential El 83.6 100.6 107.7 105.4 101.5 106.9
Strategic EI 113.2 100.6 92.9 100.7 97.5 99.2
Overall EI 94.9 101.4 98.9 104.2 99.7 104.6
Table 3-4. Mean
Scores on the MSCEIT
(n = 40)
(n = 110)
Table 3-5. Mean scores on the MSCEIT by year of study.
Table 3-6. Pearson product-moment correlations between age and the MSCEIT.
Note: *p < .05; **p
Table 3-7. Pearson product-moment correlations between academic indicators
SAT Self-reported GPA
Perceiving Emotions 0.07 0.03
Using Emotions 0.09 -0.04
Note: *p < .05; **p < .01.
and the MSCEIT.
(n = 62)
(n = 43)
(n = 24)
(n = 21)
Table 3-8. Pearson product-moment correlations between the Big Five personality dimensions
and the MSCEIT.
Strategic El 0.06
Overall El 0.05
Note: *p= .05; **p = .01.
Table 3-9. Pearson product-moment correlations between psychological well-being (PWB) and
Note: *p = .05;
0.16* 0.22** 0.31** 0.17*
Table 3-10. Pearson product-moment correlations between peer attachment and the MSCEIT.
Strategic El 0.18*
Overall El 0.16*
Note: *p= .05; **p =.01.
Table 3-11. Pearson product-moment correlations between alcohol and cigarette use and the
Note: *p = .05; **p
AUDIT Cigarette Use
Table 3-12. Factorial analysis of variance results for Overall EI.
Sum of Mean
Source of variation squares df square F Sig. ofF
Gender 16.16 1 16.16 0.10 0.75
Race/ethnicity 298.73 5 59.75 0.38 0.86
Year of study 92.62 3 30.87 0.19 0.90
Gender* Race 92.72 4 23.18 0.15 0.96
Gender Year of
study 82.42 3 27.47 0.17 0.91
Race Year of study 1235.33 8 154.42 0.97 0.46
Gender Race Year
of study 345.86 4 86.47 0.54 0.70
Error 19213.88 121 158.79
Corrected total 21913.29 149
Note: *p= .05; **p = .01.
Table 3-13. Factorial analysis of variance results for Strategic EI.
Sum of Mean Sig.
Source of variation squares df square F ofF
Year of study
Gender Year of
Race Year of study
Gender Race Year
Note: *p= .05; **p = .01.
12.86 0.14 0.71
118.09 3 39.36 0.44 0.73
761.94 5 152.39 1.70 0.14
4.20 0.05 0.99
211.64 4 52.91 0.59 0.67
682.46 8 85.31 0.95 0.48
164.95 4 41.24 0.46 0.77
Table 3-14. Factorial analysis of variance results for Experiential EI.
Source of variation Sum of squares df square F ofF
Gender 99.21 1 99.21 0.45 0.50
Race/ethnicity 398.24 3 132.75 0.60 0.62
Year of study 1413.76 5 282.75 1.27 0.28
Gender Race 210.39 3 70.13 0.32 0.81
Gender Year of
study 160.45 4 40.11 0.18 0.95
Race Year of study 1429.02 8 178.63 0.81 0.60
Gender Race Year
of study 918.79 4 229.70 1.04 0.39
Error 26836.42 121 221.79
Corrected total 31617.09 149
Note: *p= .05; **p = .01.
Table 3-15. Observed score intercorrelations of the MSCEIT subtests (N= 150).
Faces Pictures Sensations station Changes Blends Ment
Sensations 0.20* 0.30**
Facilitation 0.21** 0.36** 0.37**
Changes 0.20* 0.13 0.26** 0.22**
Blends 0.10 0.03 0.02 0.05 0.48**
Management 0.09 0.19* 0.12 0.22** 0.23** 0.23**
Management 0.15 0.11 0.23** 0.27** 0.29** 0.17* 0.51**
Note: *p= .05; **p = .01.
Table 3-16. Observed score intercorrelations of the MSCEIT subtests and first-, second-, and
third-order factors (N = 150).
0.85** 0.23** 0.68**
0.30** 0.80** 0.33** 0.70**
0.33** 0.26** 0.81** 0.62**
0.29** 0.81** 0.64**
Note: *p= .05; **p = .01.
Table 3-17. First-order inter-factor correlations of the MSCEIT (N
Perceiving Using Understand Managing
Emotions Emotions Emotions Emotions
Note: *p= .05; **p = .01.
Table 3-18. MSCEIT parameter estimates for one-, two-, three-, four-, nested, and four-branch
factor models (N
(Mayer et al., 2002)
I I I
0.36 0.41 0.46 0.41 0.35 0.26 0.52
0.42 0.57 0.69 0.55 0.50 0.56 0.65
0.46 0.54 0.57 0.56 0.59
0.52 0.64 0.64 0.64 0.65
II III II
II III II
0.53 0.50 1.03 1.00 0.37 0.61 1.01
0.34 0.39 0.46
0.53 0.65 0.62 0.64 0.29 0.63 0.63
0.58 0.68 0.81 0.79 0.37 0.63 0.80
Table 3-19. Goodness of fit indices for one-, two-, three-, four-, nested, and four-branch factor
models (N = 150).
Nested Four-branch model
One- Two- Four- Three- (Palmer et al., (Mayer et al., 2002)
Model fit factor factor factor factor 2005)
x2 76.79 46.68 14.09 17.8 10.77 16.07
df 20 19 14 17 16 16
NFI 0.70 0.81 0.94 0.93 0.96 0.94
CFI 0.75 0.88 1.00 1.00 1.00 1.00
RMSEA 0.14 0.09 0.01 0.01 0.00 0.00
Note: The one-, two-, three- and four-factor models were specified such that error terms were
uncorrelated. In the nested (Palmer et al., 2005) model, Overall El as a general factor was
restricted to be uncorrelated with the three scale-level factors. In the four-branch model (Mayer
et al., 2002), factor loadings and residual variances for the second-order factors were set equal to
each other. Reported factor loadings for both the nested and four-branch models are completely
standardized. Boldfaced roman numerals indicate the factors specified in each model and below
each roman numeral are the parameter estimates for each of the subtests associated with that
factor. All other loadings were fixed at zero.
Table 3-20. Hierarchical regression predicting grade point average (N = 148).
Step 1 WPT
Step 2 Personality
Step 3 MSCEIT
Note: Total Adjusted R2= .12. F(7, 148)
B SE B
8 Block AR2
Table 3-21. Hierarchical regression predicting SAT (N
Predictor variables B SE B 1 AR2
Step 1 WPT 0.49**
WPT 19.98 1.71 .70**
Step 2 Personality 0.07**
WPT 18.13 1.74 .64**
Extraversion -0.35 0.68 -.03
Agreeableness 0.49 0.98 .03
Conscientiousness -0.76 0.63 -.07
Emotional Stability 0.40 0.60 .04
Openness 3.44 0.86 .27**
Step 3 MSCEIT 0.00
WPT 17.50 1.82 .61**
Extraversion -0.23 0.69 -.02
Agreeableness 0.06 1.05 .00
Conscientiousness -0.79 0.63 -.08
Emotional Stability 0.25 0.61 .03
Openness 3.38 0.86 .26**
MSCEIT 0.89 0.76 .08
Note: Total Adjusted R2= .54. F(7, 143) = 25.22. *p < .05, **p < .01.
Table 3-22. Hierarchical regression predicting psychological well-being (N
Step 1 WPT
Step 2 Personality
B SE B 13
-0.91 0.41 -.18
-0.23 0.28 -.05
0.71 0.11 .39**
0.02 0.16 .01
Emotional Stability 0.60
Step 3 MSCEIT
Emotional Stability 0.58
Note: Total Adjusted R2= .60. F(7, 150) = 32.52. *p
Table 3-23. Hierarchical regression predicting Positive Relations with Others (N = 150).
Predictor variables B SE B 8 AR2
Step 1 WPT 0.03*
WPT -0.18 0.08 -.18*
Step 2 Personality 0.46**
WPT -0.04 0.06 -.04
Extraversion 0.11 0.02 .30**
Agreeableness 0.19 0.04 .36**
Conscientiousness 0.06 0.02 .18**
Emotional Stability 0.09 0.02 .27**
Openness -0.05 0.03 -.12
Step 3 MSCEIT 0.01*
WPT -0.07 0.06 -.07
Extraversion 0.11 0.02 .32**
Agreeableness 0.16 0.04 .32**
Conscientiousness 0.06 0.02 .17**
Emotional Stability 0.09 0.02 .25**
Openness -0.06 0.03 -.13
MSCEIT 0.05 0.03 .13*
Note: Total Adjusted R2= .49. F(7, 150) =21.114. *p < .05, **p < .01.
Table 3-24. Hierarchical regression predicting peer attachment (N
Predictor variables B SE B 8 AR2
Step 1 WPT 0.02
WPT -0.32 0.18 -.15
Step 2 Personality 0.17**
WPT -0.13 0.18 -.06
Extraversion 0.16 0.07 .20*
Agreeableness 0.14 0.10 .12
Conscientiousness 0.15 0.07 .19*
Emotional Stability 0.12 0.06 .16*
Openness 0.00 0.09 .00
Step 3 MSCEIT 0.01
WPT -0.18 0.18 -.08
Extraversion 0.17 0.07 .21*
Agreeableness 0.11 0.11 .09
Conscientiousness 0.15 0.07 .18*
Emotional Stability 0.11 0.06 .14
Openness -0.01 0.09 -.01
MSCEIT 0.08 0.08 .08
Note: Total Adjusted R2 .16. F(7, 150) = 4.97. *p < .05, **p < .01.
Table 3-25. Hierarchical regression predicting cigarette use (N
Predictor variables B SE B 13 AR2
Step 1 WPT 0.02
WPT -0.03 0.02 -.13
Step 2 Personality 0.04
WPT -0.04 0.02 -.18*
Extraversion 0.00 0.01 -.03
Agreeableness -0.02 0.01 -.15
Conscientiousness 0.00 0.01 -.04
Emotional Stability 0.01 0.01 .09
Openness 0.02 0.01 .16
Step 3 MSCEIT 0.01
WPT -0.04 0.02 -.16
Extraversion 0.00 0.01 -.04
Agreeableness -0.01 0.01 -.12
Conscientiousness 0.00 0.01 -.04
Emotional Stability 0.01 0.01 .11
Openness 0.02 0.01 .16
MSCEIT -0.01 0.01 -.08
Note: Total Adjusted R2 .02. F(7, 150) = 1.4. *p < .05, **p < .01.
Table 3-26. Hierarchical regression predicting alcohol use (N:
Step 1 WPT
B SE B
Step 2 Personality
-0.16 0.05 -.29**
Emotional Stability -0.01
Step 3 MSCEIT
Emotional Stability 0.01
Note: Total Adjusted R2= .14. F(7, 150)= 4.43.*p <
Figure 3-1. General factor model.
Figure 3-2. Oblique two-factor model.
Figure 3-3. Oblique four-factor model.
Figure 3-4. Oblique three-factor model.
Figure 3-5. Nested model.
The aim of this study was to examine the validity of emotional intelligence (El).
Specifically, this study investigated whether an ability measure of EI, the MSCEIT, measures a
psychological construct distinct from other related constructs. The second aim was to examine
whether El is greater in older individuals and/or those with more years in college. The third aim
was to examine group differences (racial/ethnic group, gender, and year of study) in performance
on an ability measure of EI. The fourth aim was to examine the degree to which the El
incrementally predicts real-life outcomes beyond IQ and personality. The fifth aim was to
examine the latent structure of the ability model of EI. A discussion of the most relevant findings
of each research question and their implications for future research will be discussed in the
Research Question 1: Evidence of Validity
Convergent Evidence of Validity
A primary goal of this study was to investigate the validity of El by examining the
convergent evidence of validity of the MSCEIT. To claim that El measures a distinct cognitive
ability, the MSCEIT should demonstrate positive and low to moderate correlations with closely
related constructs such as g. If the relationship between the MSCEIT and IQ is too high, for
example, it may be seen as another measure of intellectual ability. If the relationship is too low,
El may not necessarily be considered a cognitive ability. The findings of this study support
claims that the MSCEIT measures a related, yet distinct ability from g.
In a correlational analysis, IQ and Overall El correlated .24 (p < .01). These results are
similar to the modest relationships reported in several previous studies (Brackett & Mayer, 2003;
Bastian, Burns, & Nettelbeck, 2005; Zeidner et al., 2005). Further, Overall El and self-reported
SAT correlated moderately and significantly (r = .31, p < .01). Although SAT was not used as
an indicator of intelligence in this study, several studies have described positive associations
between the SAT and cognitive ability. Thus, as hypothesized, El appears to measure a cognitive
ability that is related to, but distinct from g.
To date, the most frequently interpreted score of the MSCEIT is Overall EI. In the present
investigations, analyses of the area-level and scale-level scores were also conducted.
Interestingly, in this sample both IQ and self-reported SAT scores were significantly related to
Overall EI, Strategic EI, and Understanding Emotions, but not to Experiential EI, Perceiving
Emotions, Using Emotions, or Managing Emotions. This is somewhat consistent with claims that
the Understanding Emotions scale is most strongly associated with general cognitive ability
(Mayer, Salovey, Caruso, & Sitarenios, 2001). However, based on the theory that all cognitive
abilities should correlate to some extent (Van der Zee et al., 2002), it might be expected that all
factors of the MSCEIT would be significantly related to IQ. The absence of positive manifold
may be due in part to the association of some factors with personality and others more so with
cognitive ability. These findings may also indicate problems associated with the latent structure
of the MSCEIT.
Divergent Evidence of Validity
To claim that the MSCEIT measures a construct distinct from personality, it must also
demonstrate divergent evidence of validity by correlating moderately, at most, with personality.
Generally, the MSCEIT met this criterion. In this study, Overall El correlated significantly with
Openness to Experience (r = .22, p < .01), Emotional Stability (r = .24, p < .01), and
Agreeableness (r = .32,p < .01). The strongest association was found between Managing
Emotions and Agreeableness (r = .42, p < .01). These findings are somewhat comparable to
those found in previous studies (e.g., Brackett & Mayer, 2003; Brackett, Mayer, & Warner,
2004; Mayer et al., 2004a). Thus, El is most closely related to personal qualities such as
kindness, generosity, calmness, curiosity, and willingness to try new things.
Perceiving Emotions did not correlate significantly with any of the personality dimensions
or academic indicators listed above. Although this does not preclude its inclusion in the factorial
structure, this may suggest that some aspects of El consist of abilities outside the realm of
intelligence and personality.
The strength of the relationships between the MSCEIT and some of the Big Five
personality dimensions in this study (range .02 to .42) was slightly higher than those reported in
previous research (e.g., Mayer et al., 2004a; Zeng & Miller, 2003). Still, they remain lower than
those found for self-report measures (Davies et al., 1998). Consequently, the ability model of El
has more substantiation than mixed models.
Correlations in the low to moderate range suggest that El is distinct from personality,
although it may overlap somewhat across several personality traits or domains. Thus, personality
overlaps somewhat with an individual's EI.
Research Question 2: EI, Age, and Experience
The purpose of this question was to examine whether El is greater among older individuals
and/or those with more years in college. Mayer et al. (1999) claimed that this criterion is critical
for El to be considered a cognitive ability. In the present study, this criterion was not met. Years
of education was used as an indicator of experience. Correlations between raw scores on the
MSCEIT and age in years and experience were near zero and not significant. In fact, participants
in their first year of study scored higher on several factors than those with more years of study.
These findings may be due in part to the restriction in range of the sample. All participants were
students at a large southeastern university, and the vast majority were between 18 and 20 years
of age. In addition, the demographic questionnaire did not specify whether participants should
indicate the number of years they had attended college or their class status based on number of
Research Question 3: Group Differences
The purpose of this question was to examine whether different groups demonstrate
differences in performance on EI. Results of this study revealed that women generally scored
slightly higher than men, on average, although these differences were not significant. Mean
differences based on years of education also were not significant. Finally, no mean differences
were found between different racial/ethnic groups. Generally, Hispanic Americans scored
slightly higher than other groups in this sample. However, the sample was relatively
homogeneous with regards to demographic characteristics; thus these findings should be viewed
Research Question 4: Predicting Real-Life Outcomes
The purpose of this question was to explore the relationship between El and important
outcomes when controlling for g and personality. El is of little use in applied settings if it does
not improve our prediction of human behavior above and beyond related constructs. For
interpretative purposes, the Overall El score was used for regression analyses. There were
several important findings.
In the correlation analyses, self-reported GPA demonstrated weak but significant
correlations with Strategic El (r = .21, p < .01) and Understanding Emotions (r = .19, p < .05),
and was unrelated to Overall El (r = .14, p > .05). Based on a series of hierarchical multiple
regression analyses, El accounted for no additional unique variance in self-reported GPA after
controlling for g and personality. Further, these findings are consistent with those found from
previous studies (Bastian et al., 2005).
Although self-reported SAT correlated significantly with Overall El (r = .31, p < .01), El
did not account for additional variance in self-reported SAT scores after controlling for g and
personality. Given the significant reduction of the contribution of El when g and personality are
controlled, these findings suggests that EI, g, and personality share considerable amount of
overlapping variance with regards to the prediction of self-reported SAT scores.
As expected, g emerged as a significant predictor for both SAT and GPA. Interestingly,
Conscientiousness emerged as the most powerful predictor of self-reported GPA in the full
regression equation. Those high in conscientiousness typically are self-disciplined, organized,
and conforming; qualities that are critical to success among undergraduate students. Thus,
although the relationship between Conscientiousness and academic success could be expected,
its strength as a predictor over g is somewhat surprising.
Although several scales were derived from the SPWB, only the total PWB score was used
for the hierarchical multiple regression analysis. In the correlation analyses, PWB correlated
significantly with Overall El (r = .19,p < .05), Experiential El (r = .20, p < .05), Managing
Emotions (r = .24, p < .01), and Using Emotions (r = .17, p < .05). Based on a series of
hierarchical multiple regression analyses, El did not account for additional variance in total PWB
after controlling for g and personality.
Personality emerged as a significant predictor of total PWB in the full regression equation.
Notably, Extraversion, Conscientiousness, and Emotional Stability were significant predictors of
total PWB. Thus, in this study, those who were outgoing and assertive, organized, self-
disciplined, relaxed, and at ease demonstrated the highest levels of PWB.
In the correlation analyses, peer attachment correlated significantly with Overall El (r =
.16, p < .05), Strategic El (r = .18, p < .05), and Managing Emotions (r=.22, p < .01). Based on
a series of hierarchical multiple regression analyses, El did not account for additional variance in
peer attachment after controlling for g and personality.
Extraversion and Conscientiousness emerged as the only significant predictors of peer
attachment in the full regression equation. Interestingly, these same personality dimensions
contributed significantly to the prediction of psychological well-being, which may indicate a
relationship between an individual's psychological well-being and the quality of peer attachment.
A regression analysis also was conducted with the Positive Relations with Others scale of
the SPWB as another indicator of peer attachment. This scale has been used in previous studies
as a proxy for the quality of peer relationships. Further, this scale correlated significantly with El
in the present sample. In this study, El emerged as a significant predictor of the Positive
Relations with Others scale after controlling for g and personality. The contribution of El was
low, however, accounting for only 1% of the variance. The strongest predictors in the model
were Extraversion, Agreeableness, Conscientiousness, and Emotional Stability.
Lopes et al. (2003) also found a significant contribution of El in predicting Positive
Relations with Others when controlling for g and personality, with El accounting for an
additional 5% of variance. The difference in magnitude of unique variance explained by El
between the present investigation and that of Lopes et al. (2003) may be due to differences in
how cognitive ability was measured. Lopes et al. (2003) used scores on a Vocabulary subtest
from a popular IQ test as an estimate of intelligence, whereas the present study used a measure of
psychometric g. Further, although the sample described by Lopes et al. (2003) also consisted of
undergraduate students, participants were all from Yale University, and thus likely to be highly
restricted in range on cognitive ability.
Correlations between all scores on the MSCEIT and a measure of cigarette use were not
significant. Based on a series of hierarchical multiple regression analyses, El predicted no
additional unique variance in total cigarette use after controlling for g and personality. In fact,
neither g nor personality contributed significantly to the prediction of cigarette use in the full
regression equation. This may be due in part to the limited number of cigarette smokers in the
sample. Only 8% of participants indicated use of cigarettes to any degree. Among those who
reported cigarette use, all used cigarettes rather infrequently.
In the correlation analyses, alcohol use correlated negatively and significantly with Overall
EI, Strategic EI, Experiential EI, and all first-order factor scores except Understanding Emotions.
Based on a series of hierarchical multiple regression analyses, El emerged as a significant
predictor of alcohol use after controlling for g and personality, accounting for an additional 4%
of the variance. Interestingly, Agreeableness also contributed significantly to the prediction of
alcohol use in the full regression equation. Negative betas for both the MSCEIT and
Agreeableness indicated that those with higher levels of El and who are kind and easy-going
report lower levels of alcohol use.
Research Question 5: Latent Structure of the MSCEIT
The aim of this question was to examine the factor structure of the MSCEIT. The results of
the confirmatory factor analyses determined that the general factor model with three nested
scale-level factors (Palmer et al., 2005) emerged as the most appropriate model. Additionally, the
three-factor model (Gignac, 2005) also indicated good fit.
Both the one- and two-factor models were not well-fitting models, despite positive and
significant factor loadings. Although the four-factor and hierarchical four-branch models
emerged as excellent fitting models based on goodness-of-fit indices, both models contained
factor loadings above 1.0, resulting in a negative residual variance. Thus, these models also were
considered improper models.
The Using Emotions scale emerged as problematic in the four-branch model, yielding a
factor loading greater than 1.0 on Experiential El. This is consistent with findings from both
Gignac (2005) and Palmer et al. (2005), who removed this scale in their structural equation
models due to high correlations with other scales. In fact, the only well-fitting models within this
study were those with the Using Emotions scale removed. In addition, the Changes subtest
yielded a factor loading greater than 1.0 in the four-branch model, and a loading of 1.0 in the
three-factor model, indicating that this subtest also may be problematic. Thus, the Using
Emotions scale and the Changes subtest should be revised or removed in subsequent revisions of
Although the residual variances of the oblique four-factor and hierarchical four-branch
models could have been constrained to be positive, this solution does not account for the
insufficient number of subtests contributing to each first-order factor. Fit indices will always
yield excellent fit for models consisting of factors with only two indicators (Mulaik & Millsap,
2000). Generally, at least three observed variables should contribute to a factor (Bollen, 1989).
These results raise serious concerns regarding the interpretation of the MSCEIT. Although
the authors claim the appropriateness of one-, two-, four-factor, and a hierarchical nested four-
branch structural model for the MSCEIT, these findings have yet to be supported or replicated by
additional research. Mayer et al. (2005) acknowledged these concerns, and stated that the factor
structure of the MSCEIT can not be interpreted at this time. As a result, until additional analyses
are conducted or changes to the test are made, scores from the MSCEIT should be interpreted
with caution, if at all.
Summary and Implications for Future Research
The results of the current study indicate that EI, as defined by the ability model, is distinct
from both g and personality. In addition, the first-, second-, and third-order factor scores on the
MSCEIT demonstrate adequate levels of internal consistency. However, the internal consistency
of several subtests were low. In addition, with the exception of alcohol use, the results also
indicate that after controlling for the effects ofg and personality, El contributes little to no
additional explained variance in a number of real-life outcomes. These findings generally are
consistent with previous research (Brody, 2004). Results suggest that among undergraduate
students, although El correlates with academic success, alcohol use, psychological well-being,
and peer attachment, the variance explained is redundant with the variance explained by
personality and g. In addition, the data did not support the proposed factor structure of the
MSCEIT, precluding the interpretation of scores. Thus, at this point, the nature of El remains
Still, El should not be discarded. As several studies have demonstrated evidence of the
validity and reliability of the MSCEIT, future work should prospectively examine predictive
validity with other related criteria among varied samples. For example, research is needed to
examine relationships between El and behavior (e.g., behavior referrals, diagnoses of
emotional/behavioral disorders, school suspensions), social-emotional functioning, achievement,
cognitive ability, personality, and temperament among children and adolescents. Further, the
relationship between El and goodness-of-fit between children and their teachers and parents also
should be explored. Before schools can continue implementing intervention programs to improve
EI, research needs to determine the relationship between El and student outcomes in school
settings (Rossen, 2006).
Future studies should examine the development of El through longitudinal studies. Such
research could answer several questions about the nature of EI, such as whether El is a stable
ability or malleable as a function of experience, age, and/or intervention. Also, longitudinal data
could help determine how well El predicts future behaviors and attitudes.
Future research on El should incorporate some other theories of contemporary measures of
intelligence to better understand El (Bastian et al., 2005; Kaufman & Kaufman, 2001). For
example, examining how stratum II cognitive abilities such as Gf and Gc relate to specific first-
order factors of the MSCEIT may guide the interpretation of scores.
Finally, developing more subtests or parceling existing subtests is necessary to determine
the most appropriate theoretical model of the MSCEIT (Gignac, 2005; Palmer et al., 2005). At
least three subtests should uniquely contribute to each scale. Thus, a total of 12 subtests should
be considered as a minimum in subsequent revisions of the MSCEIT. Further, given concerns
noted in the factor analysis, the Using Emotions scale should be revised or removed.
As noted previously, this sample was restricted to undergraduate college students at a
highly selective institution. Although this is an important population for psychological research,
the generalizability of these findings to other populations is limited. Further, including
undergraduate students from other local colleges with a wider range of age, race/ethnicity,
cigarette use, and general college experience may have produced a different pattern of results.
Further, collecting age data in years may have resulted in restricted variance. Findings related to
the relationship between El and age may have been different had information been collected on
age in months.
The tests selected to measure g and the Big Five personality traits were selected due to
their ease of administration and the evidence supporting their validity, but they are not the most
widely used measures in research. Subsequently, these measures do not have as strong empirical
support as other measures such as the Wechsler scales or the NEO-FFI. Although the measures
used in the present study demonstrated relationships with the other variables as expected, results
may have been different had more commonly used measures such as the NEO-Five-Factor
Inventory (NEO-FFI; Costa & McRae, 1992) or Wechsler scales been included.
Finally, this study was based largely on participants' self-reports about their behaviors and
perceptions of themselves. Despite evidence of validity and reliability for nearly all measures
included in the present study, self-reports are susceptible to inaccurate or distorted responses to
create favorable impressions (Barrick & Mount, 1996). Although efforts were made to reduce the
likelihood of deceptive response styles by preserving participants' anonymity, research is needed
on the relationship between El and actual behavior rather than self-reported behavior (Caruso et
Despite noted limitations, the present study suggests that ability measures of El may
provide an additional tool for psychologists to study and understand human behavior and
development. The construct of El represents a promising avenue for future research and potential
This and all other information is being used for data collection purposes only. Participation in
this study is voluntary.
Please answer the following questions. Remember your name is not being used, you are
anonymous please be truthful.
1. How old are you?
2. Gender (circle one): Male Female
3. Race/Ethnicity (circle one):
American Indian Hispanic American
Asian American White/non-Hispanic
African American Other
4. Please circle the number next to your current year of study.
1. Freshman 3. Junior
2. Sophomore 4. Senior
5. Please write your major field of study at the university?
6. Please write your SAT score: or ACT score:
7. Please write your current undergraduate GPA:
8. Please list any campus groups/clubs for which you are currently a member:
9. Please list any campus groups/clubs for which you hold a leadership role:
INTERNATIONAL PERSONALITY ITEM POOL (IPIP) PERSONALITY SCALE
On the following pages, there are phrases describing people's behaviors. Please use the rating
scale below to describe how accurately each statement describes you. Describe yourself as you
generally are now, not as you wish to be in the future. Describe yourself as you honestly see
yourself, in relation to other people you know of the same sex as you are, and roughly your same
age. So that you can describe yourself in an honest manner, your responses will be kept in
absolute confidence. Please read each statement carefully, and then circle the corresponding
1: Very Inaccurate
2: Moderately Inaccurate
3: Neither Inaccurate nor Accurate
4: Moderately Accurate
5: Very Accurate
1. I am the life of the party. 1 2 3 4 5
2. I insult people. 1 2 3 4 5
3. I am always prepared. 1 2 3 4 5
4. I get stressed out easily. 1 2 3 4 5
5. I have a rich vocabulary. 1 2 3 4 5
6. I often feel uncomfortable around others. 1 2 3 4 5
7. I am interested in people. 1 2 3 4 5
8. I leave my belongings around. 1 2 3 4 5
9. I am relaxed most of the time. 1 2 3 4 5
10. I have difficulty understanding abstract ideas. 1 2 3 4 5
11. I feel comfortable around people. 1 2 3 4 5
12. I am not interested in other people's problems. 1 2 3 4 5
13. I pay attention to details. 1 2 3 4 5
14. I worry about things. 1 2 3 4 5
15. I have a vivid imagination. 1 2 3 4 5
16. I keep in the background. 1 2 3 4 5
17. I sympathize with others' feelings. 1 2 3 4 5
18. I make a mess of things. 1 2 3 4 5
19. I seldom feel blue. 1 2 3 4 5
20. I am not interested in abstract ideas. 1 2 3 4 5
21. I start conversations. 1 2 3 4 5
22. I feel little concern for others. 1 2 3 4 5
23. I get chores done right away. 1 2 3 4 5
24. I am easily disturbed 1 2 3 4 5
25. I have excellent ideas. 1 2 3 4 5
26. I have little to say. 1 2 3 4 5
27. I have a soft heart. 1 2 3 4 5
28. I often forget to put things back in their proper place. 1 2 3 4 5
29. I am not easily bothered by things. 1 2 3 4 5
30. I do not have a good imagination. 1 2 3 4 5
31. I talk to a lot of different people at parties. 1 2 3 4 5
32. I am not really interested in others. 1 2 3 4 5
33. I like order. 1 2 3 4 5
34. I get upset easily. 1 2 3 4 5
35. I am quick to understand things. 1 2 3 4 5
36. I don't like to draw attention to myself. 1 2 3 4 5
37. I take time out for others. 1 2 3 4 5
38. I shirk my duties. 1 2 3 4 5
39. I rarely get irritated. 1 2 3 4 5
40. I try to avoid complex people. 1 2 3 4 5
41. I don't mind being the center of attention. 1 2 3 4 5
42. I am hard to get to know. 1 2 3 4 5
43. I follow a schedule. 1 2 3 4 5
44. I change my mood a lot. 1 2 3 4 5
45. I use difficult words. 1 2 3 4 5
46. I am quiet around strangers. 1 2 3 4 5
47. I feel others' emotions. 1 2 3 4 5
48. I neglect my duties. 1 2 3 4 5
49. I seldom get mad. 1 2 3 4 5
50. I have difficulty imagining things. 1 2 3 4 5
51. I make friends easily. 1 2 3 4 5
52. I am indifferent to the feelings of others. 1 2 3 4 5
53. I am exacting in my work. 1 2 3 4 5
54. I have frequent mood swings. 1 2 3 4 5
55. I spend time reflecting on things. 1 2 3 4 5
56. I find it difficult to approach others. 1 2 3 4 5
57. I make people feel at ease. 1 2 3 4 5
58. I waste my time. 1 2 3 4 5
59. I get irritated easily. 1 2 3 4 5
60. I avoid difficult reading material. 1 2 3 4 5
61. I take charge. 1 2 3 4 5
62. I inquire about others' well-being. 1 2 3 4 5
63. I do things according to a plan. 1 2 3 4 5
64. I often feel blue. 1 2 3 4 5
65. I am full of ideas. 1 2 3 4 5
66. I don't talk a lot. 1 2 3 4 5
67. I know how to comfort others. 1 2 3 4 5
68. I do things in a half-way manner. 1 2 3 4 5
69. I get angry easily. 1 2 3 4 5
70. I will not probe deeply into a subject. 1 2 3 4 5
71. I know how to captivate people. 1 2 3 4 5
72. I love children. 1 2 3 4 5
73. I continue until everything is perfect. 1 2 3 4 5
74. I panic easily. 1 2 3 4 5
75. I carry the conversation to a higher level. 1 2 3 4 5
76. I bottle up my feelings. 1 2 3 4 5
77. I am on good terms with nearly everyone. 1 2 3 4 5
78. I find it difficult to get down to work. 1 2 3 4 5
79. I feel threatened easily. 1 2 3 4 5
80. I catch on to things quickly. 1 2 3 4 5
81. I feel at ease with people. 1 2 3 4 5
82. I have a good word for everyone. 1 2 3 4 5
83. I make plans and stick to them. 1 2 3 4 5
84. I get overwhelmed by emotions. 1 2 3 4 5
85. I can handle a lot of information. 1 2 3 4 5
86. I am a very private person. 1 2 3 4 5
87. I show my gratitude. 1 2 3 4 5
88. I leave a mess in my room. 1 2 3 4 5
89. I take offense easily. 1 2 3 4 5
90. I am good at many things. 1 2 3 4 5
91. I wait for others to lead the way. 1 2 3 4 5
92. I think of others first. 1 2 3 4 5
93. I love order and regularity. 1 2 3 4 5
94. I get caught up in my problems. 1 2 3 4 5
95. I love to read challenging material. 1 2 3 4 5
96. I am skilled in handling social situations. 1 2 3 4 5
97. I love to help others. 1 2 3 4 5
98. I like to tidy up. 1 2 3 4 5
99. I grumble about things. 1 2 3 4 5
100. I love to think up new ways of doing things. 1 2 3 4 5
INVENTORY OF PARENT AND PEER ATTACHMENT (IPPA) SECTION III
Directions: Indicate whether the following items are never or almost never true, seldom true,
sometimes true, often true, or always or almost always true.
1 2 3 4 5
I -------------------I -------------------I I I
Never, Seldom Sometimes Often Always,
or Almost never or Almost always
1. I like to get my friends' point of view on things I'm concerned about.
2. My friends sense when I'm upset about something.
3. When we discuss things, my friends consider my point of view.
4. Talking over my problems with my friends makes me feel ashamed or foolish.
1 2 3 4 5
5. I wish I had different friends.
6. My friends understand me.
7. My friends encourage me to talk about my difficulties.
1 2 3 4
8. My friends accept me as I am.