MEIA Technical Manual - SIGMA Assessment Systems

Transcription

MEIA Technical Manual - SIGMA Assessment Systems
Multidimensional Emotional
Intelligence Assessment
Robert P. Tett, Ph.D., Alvin Wang, Ph.D., & Kevin E. Fox, Ph.D.
Technical Manual
Advancing the Science of Human Assessment since 1967.
Multidimensional Emotional Intelligence Assessment
MEIA
Acknowledgments
We gratefully acknowledge the assistance of Margaret H. Thomas, Randy D. Fisher,
Arelis Martinez, and Joseph Griebler in the early phases of scale development. Portions of this manual are based on articles by Tett, Fox, and Wang (2005) and Tett
and Fox (in press).
Copyright © 2006 by SIGMA Assessment Systems, Inc. All rights reserved. May
not be reproduced in whole or in part in any form or by any means without written
permission of SIGMA Assessment Systems, Inc.
Version 1.03.
SIGMA Assessment Systems, Inc.
P.O. Box 610757
Port Huron, MI 48061-0757
Published simultaneously in Canada by:
Research Psychologist Press, Inc.
P.O. Box 3292, Station B
London, ON N6A 4K3.
Printed in Canada
ii
Contents
Chapter 1
Introduction................................................................................. 1
Overview.......................................................................................................1
The Emotional Intelligence Construct..........................................................1
The EI Trait Domain.....................................................................................5
The Challenge of Social Desirability in Self-Description............................7
Table 1.1: Multidimensional Emotional Intelligence Assessment (MEIA)
scales, definitions, and sample items........................................... 8
Administration and Scoring..........................................................................9
Norms............................................................................................................9
Table 1.2: College student norms for the Multidimensional Emotional Intelligence Assessment (MEIA).......................................................... 9
Reliabilities and Interscale Correlations.....................................................10
Table 1.3: Test-retest reliabilities for the 10 MEIA primary scales over a 4to 6- week interval (N = 68 college students)............................ 11
Table 1.4: Correlations among 10 MEIA primary scales and alpha coefficients based on college student norms (N = 332).................... 12
Table 1.5: Correlations among 10 MEIA primary scales and alpha coefficients based on college student norms for men (upper right, N =
127) and women (lower left, N = 188)...................................... 12
Summary.....................................................................................................13
Chapter 2
Construction of the MEIA......................................................... 15
Study 1: Development of Preliminary EI Scales......................................15
Preparation of Item Sets........................................................................15
Confirmation of Content Coverage.......................................................16
Data Collection.....................................................................................16
Item Analyses........................................................................................16
Evaluation of Convergent and Discriminant Validity...........................17
Relations with PRF-Desirability...........................................................18
Summary...............................................................................................18
Table 2.1: Correlations between the 10 EI and 15 JPI-R and PRF-Desirability (DES) subscales in Study 1 (N = 138)............................ 19
Table 2.2: Correlations between the 10 EI and 15 JPI-R subscales controlling for PRF-Desirability in Study 1 (N = 138)......................... 19
Study 2: Assessing the Uniqueness of EI as a Trait Domain....................20
Item Preparation....................................................................................20
Data Collection.....................................................................................20
Item Analyses........................................................................................21
iii
Evaluation of Convergent and Discriminant Validity...........................21
Table 2.3: Numbers of items, means, standard deviations, alphas, and correlations among the 10 EI subscales in Study 2a (N = 183)...... 22
Table 2.4: Correlations between the 10 EI and 15 JPI-R and PRF-Desirability (DES) subscales in Study 2 (N = 163)............................ 22
Table 2.5: Correlations between the 10 EI and 15 JPI-R subscales, controlling for PRF-Desirability in Study 2 (N = 163)......................... 24
Summary...............................................................................................25
Table 2.6: EI subscale variance accounted for by PRF-Desirability (DES),
JPI-R subscales, and remaining EI subscales (N = 163).......... 25
Chapter 3
Validation of the MEIA.............................................................. 27
Content Validity..........................................................................................27
Criterion Validity.........................................................................................27
Incremental Validity....................................................................................28
Table 3.1: Correlations between MEIA subscales and distal self-report
criteria: satisfaction and adaptability (N = 152)...................... 29
Construct Validity.......................................................................................30
Table 3.2: Regression results showing incremental validity of MEIA primary scales in predicting satisfaction and cross-cultural adaptability (N = 152)........................................................................ 31
Social Desirability.......................................................................................33
Structural Validity.......................................................................................33
Table 3.3: Correlations between MEIA subscales and the Big Five (minimarkers), affectivity, self-monitoring, and Crowne-Marlowe
social desirability (N = 152)...................................................... 34
Table 3.4: Correlations between MEIA subscales and the Hogan Personality Inventory (N = 119).............................................................. 34
Table 3.5: Rotated loadings and eigenvalues of 10 orthogonal components
derived from 29 MEIA item bundles (N = 152)......................... 36
Table 3.6: Rotated loadings, eigenvalues*, and communalities** (h2) of 10
MEIA subscales (N = 152)........................................................ 37
Table 3.7: Results of model testing (N = 184)............................................ 39
Table 3.8: Correlations between MEIA core composites and proximal outcomes (N = 184)........................................................................ 41
Summary.....................................................................................................42
Chapter 4
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)........................................................................ 43
Emotional Intelligence at Work..................................................................43
Norms..........................................................................................................44
Table 4.1: Norms for the Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W) in the total sample and several
subgroups................................................................................... 45
Table 4.2: Norms for the Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W) in men and women....................... 45
iv
Table 4.3: Alphas and interscale correlations for the Multidimensional
Emotional Intelligence Assessment-Workplace (MEIA-W) (N =
653)............................................................................................ 46
Reliability & Interscale Correlations..........................................................46
Content Validity..........................................................................................46
Construct Validity.......................................................................................47
Table 4.4: Correlations between the MEIA-W and the Hogan Personality
Inventory - Short Form............................................................... 48
Criterion Validity.........................................................................................48
Job Performance Ratings............................................................................49
Criterion Validity Results............................................................................50
Table 4.5: Final item analysis results for task and contextual performance
behaviors forming composites for use in criterion validation of
the MEIA-W............................................................................... 51
Table 4.6: Correlations between the MEIA-W and supervisor ratings of
task and contextual performance (N = 116).............................. 52
Incremental Validity....................................................................................52
Table 4.7: Incremental validity of ability (PAF) over personality (HPI-SF)
and EI (MEIA-W) (N = 50)........................................................ 53
Table 4.8: Incremental validity of personality (HPI-SF) over ability (PAF)
and EI (MEIA-W) (N = 50)........................................................ 54
Structural Validity.......................................................................................54
Table 4.9: Incremental validity of EI (MEIA-W) over ability (PAF) and
personality (HPI-SF) (N = 50).................................................. 55
Table 4.10: Correlations among subscales for the MEIA and MEIA-W..... 56
Table 4.11: Fit results for the MEIA-W (N = 225)...................................... 56
Table 4.12: Correlations between MEIA-W core composites and proximal
outcomes (N = 225)................................................................... 58
Summary.....................................................................................................58
References...................................................................................................... 59
v
Chapter 1
Introduction
Overview
The Multidimensional Emotional Intelligence Assessment (MEIA) is a 150-item, selfreport measure of emotional intelligence (EI) based on Salovey and Mayer’s (1990)
multidimensional conceptualization of EI, and takes about 20 minutes to complete.
The MEIA includes 10 primary scales and an Infrequency scale. The 10 primary
scales are Recognition of Emotion in the Self (RecSlf), Regulation of Emotion in the
Self (RegSlf), Recognition of Emotion in Others (RecOth), Regulation of Emotion
in Others (RegOth), Nonverbal Emotional Expression (NvExp), Empathy (Emp),
Intuition vs. Reason (IvR), Creative Thinking (CrTh), Mood Redirected Attention
(MRA), and Motivating Emotions (MotEm). The first six scales (RecSlf, RegSlf,
RecOth, RegOth, NvExp, Emp) capture core EI facets, and the latter four (IvR, CrTh,
MRA, MotEm) assess proximal EI outcomes. The Infrequency scale detects cases
of non-purposeful responding, removal of which reduces unwanted random variance
from MEIA data. Raw scores on the MEIA are converted to standard scores based
on norms from 332 American college students and the normed scores are presented
as a MEIA profile.
The MEIA is intended for research on emotional intelligence conceptualized as a
dispositional trait (see below) and may also be used as a counseling aid, for example,
in career centers. Research questions relevant to the MEIA include but are not limited
to (a) understanding and predicting success in school, work, and other life domains,
(b) the personological bases of life satisfaction and psychological well-being, (c)
linkages between EI and psychopathology (e.g., alexithymia, depression), (d) bases
of health-related behavior (e.g., smoking, exercise), (e) emotional development, (f)
emotional labor, (g) emotional contagion, (h) interpersonal and group processes
(e.g., friendship, spousal relations, parenting, team building), (i) the heritability of
emotional functioning, and (j) the ability/trait distinction. A parallel measure, the
Multidimensional Emotional Intelligence Assessment – Workplace (MEIA-W), is
available for use in work settings.
The Emotional Intelligence Construct
The nature of EI and its content are matters of ongoing interest. Two dominant perspectives that have emerged in this area are those of ability and personality (Davies,
1
Chapter 1
Stankov, & Roberts, 1998; Mayer, Salovey, & Caruso, 2000; Roberts, Zeidner, &
Matthews, 2001). Each approach carries important theoretical and measurement
implications. As an ability or skill, EI is a capacity to engage in valued behavior,
entails a degree of mutability (e.g., through training), and calls for measurement in
the context of correctness (i.e., right and wrong answers). As personality, on the
other hand, EI is a relatively stable motivational construct and encourages assessment by way of self-description. Research questions diverge accordingly. From an
ability perspective, it is reasonable to examine agreement in what counts as right
and wrong answers (Mayer et al., 2000), the uniqueness of EI in relation to more
traditional conceptualizations of intelligence (i.e., g; Roberts et al., 2001), the role
of EI in decision processes, and conditions affecting the accuracy of emotional
perception, appraisal, and related processes. From a personality perspective, it is
appropriate to consider the effects of response biases (Roberts et al., 2001), relations
with extant trait typologies (e.g., the Big Five; Davies et al., 1998), and the possibility
of bi-directional linkages with important outcomes (i.e., positive relations in some
situations, negative in others; Tett, Jackson, Rothstein, & Reddon, 1999).
The ability approach to EI has been advocated most strongly by Mayer & Salovey
(1997) and Mayer et al. (2000) in developing the Mayer Salovey Caruso Emotional
Intelligence Test (MSCEIT). Mayer and Salovey (1997) define EI as “the ability to
perceive and express emotion, assimilate emotion in thought, understand and reason
with emotion, and regulate emotion in the self and others” (p. 10). Whereas Mayer
et al. (2000) explicitly caution against a personality approach to EI, others have promoted such an approach entailing self-report. Bar-On (1997), for example, defines
EI as “an array of noncognitive capabilities, competencies, and skills that influence
one’s ability to succeed in coping with environmental demands and pressures” (p. 14).
He offers 15 specific dimensions (e.g., emotional self-awareness, empathy) within
5 broader categories of intrapersonal skills, interpersonal skills, adaptability, stressmanagement, and general mood. Similarly, Petrides and Furnham (2003) define EI
as “a constellation of behavioral dispositions and self-perceptions concerning one’s
ability to recognize, process, and utilize emotion-laden information” (p. 278; italics
in original). Included in this “constellation” are emotionally relevant personality
traits, such as empathy and perseverance, and self-perceived social abilities allowing
conceptualization as “emotional self-efficacy.”
It is important to note that, although the ability and personality perspectives on EI
carry distinct measurement and research implications, it does not follow that they
are mutually exclusive. Rather, we suggest that EI captures a unique amalgam of
cognitive and emotional functioning calling explicitly for assessment as both ability
and personality. EI is distinct from g by its direct involvement of emotion, which
is closely tied to motivation. We suggest that EI, unlike g, includes an inherently
motivational component and that personality captures this component by its conceptualization in terms of psychological needs. In short, ability captures the “can do”
of EI, whereas personality captures the intrinsic “will do.” This conceptualization
2
Introduction
of EI as the combination of ability and personality is depicted in Figure 1.1. The
ability/personality distinction is elaborated briefly, below, as it applies to EI.
Figure 1.1: Emotional intelligence as a combination of ability and personality domains
Personality traits have long been considered as needs (e.g., Murray, 1938), satisfaction of which is intrinsically motivating (Tett & Burnett, 2003). Cognitive ability,
in contrast, is a capacity, and motivation to engage that capacity derives either from
extrinsic rewards (e.g., salary) or from companion need-based traits like self-esteem,
self-efficacy, or achievement striving (Tett & Burnett, 2003). EI resembles both
ability and personality with respect to motivation. People engage in EI as an ability
because it brings extrinsic rewards and helps satisfy self- and achievement-related
needs. As a personality trait, EI is engaged because it is intrinsically rewarding.
Regulating one’s own emotions by reducing unpleasant affect or increasing pleasurable affect, for example, is inherently desirable. Recognizing one’s own emotions is
intrinsically rewarding for those high in the need for self-awareness, and regulating
how others feel can be directly motivating to those high in the need for control. In
short, we suggest that EI dimensions warrant treatment as personality traits because
they bear conceptualization as needs in addition to capacities.
A second feature of personality traits that distinguishes them from abilities and
makes them relevant to consideration of EI is that the value of a given personality
3
Chapter 1
trait is situationally dependent, whereas the value of a given ability is attached to the
construct itself (Tett & Burnett, 2003). People are rarely, if ever, judged too high
on an ability or skill. Correspondingly, ability measures rely on right and wrong
answers (i.e., value is built-in). In personality terms, on the other hand, the level
of a given trait is distinct from its value. With respect to EI, for example, empathy
is generally a desirable human quality, yet there are many situations where being
empathic can be counterproductive (e.g., health care workers who empathize too
strongly with those in need of urgent care; security personnel whose primary duty is
to protect a nation or organization, sometimes at the expense of individuals and their
emotional states). Recognizing others’ feelings is readily conceived as an ability,
but its value too depends on situational demands (Elfenbein & Ambady, 2002). As
a third example, recognizing one’s own emotions can be beneficial or detrimental,
depending on other factors (Davis & Franzoi, 1999). In each case, the value of EI
is context dependent, in keeping with a general personality perspective.
A third reason for examining EI as a personality domain stems from the observation that personality-oriented measures of EI (i.e., based on self-report) have been
found to relate meaningfully to important life outcomes. As noted by Mayer et al.
(2000), self-reported ability is of questionable validity due to its susceptibility to bias.
If EI is (purely) an ability, then self-report EI measures should be weakly related to
conceptually relevant outcomes. To the contrary, research involving such measures
has shown notable linkages with life satisfaction (Ciarrochi, Chan, & Caputi, 2000;
Palmer, Donaldson, & Stough, 2002; Saklofske, Austin, & Minski, 2003; Slaski
& Cartwright, 2002), self-esteem (Ciarrochi et al., 2000), job satisfaction (Wong
& Law, 2002), job performance (Slaski & Cartwright, 2002; Wong & Law, 2002),
and academic performance (Schutte, Malouff, Hall, Haggerty, Cooper, Golden, &
Dornheim, 1998). These findings support assessment of EI by self-report and, accordingly, a personality-based perspective.
In light of the forgoing analysis, we define EI as intraindividual consistency and
interindividual uniqueness in the capacity and willingness to perceive, understand,
regulate, and express emotions in the self and others. Our definition warrants
brief discussion along several lines. First, by “intraindividual consistency” and
“interindividual uniqueness,” we mean to articulate the trait-like quality of EI, the
former term denoting prediction of future behavior from past behavior and the latter,
differentiation among individuals. Second, as noted above, we see EI as a natural
combination of ability (i.e., “capacity”) and personality (i.e., “willingness”) that
follows directly from the joining of two distinct but interrelated psychological domains: intelligence and emotion. Third, the perception, understanding, regulation,
and expression of emotions are conceived to be distinct but interrelated processes
(e.g., one can understand only that which is perceived; regulation of internal states
affects emotional expression) and all more specific EI content facets (e.g., regulation
of emotion in the self, empathy) can be considered in terms of the noted processes.
Finally, EI deals with emotions in both the self and others, as noted elsewhere (cf.
4
Introduction
Mayer et al., 2000).
Notably, the MEIA is not designed to capture the entirety of our definition of
EI, but rather only that portion dealing with EI as a personality domain, that is,
the “willingness” component. We offer the MEIA as a multidimensional measure
that is expected to be compatible with corresponding ability-based tests. Use of
the MEIA and ability-based EI tests can be expected to provide a more complete
understanding of EI than is possible with the use of either type of measure alone.
Hence, those seeking to assess EI as a broader construct (i.e., beyond the domain of
personality) are urged to consider use of ability-based EI measures (e.g., MSCEIT)
in addition to the MEIA.
The EI Trait Domain
Several models of EI facets have been proposed. The first to appear in the literature
and the most influential to date is that of Salovey and Mayer (1990). Mayer and
Salovey’s (1997) revision targets EI exclusively as an ability. The original model offers a more suitable basis for the construction of self-report measures, as in the case
of the MEIA, owing to its broader conceptualization of EI. Empathy, for example,
is included in the 1990 model but not the revision as it does not conform to Mayer
and Salovey’s (1997) ability orientation. Salovey and Mayer’s (1990) initial model
is depicted in Figure 1.2 and each of its 10 facets is described briefly, below. Some
MEIA scale names are different from Salovey and Mayer’s labels (in parentheses)
to clarify targeted content. Sample true-keyed items are also provided.
Recognition of Emotion in the Self (Emotion in the Self: Verbal). The first of
Salovey and Mayer’s (1990) EI facets denotes the degree to which one is in touch
with his or her feelings and can describe those feelings in words. An example would
be recognizing when one is afraid and finding the words to describe that feeling.
This aspect of EI is expected to facilitate interpersonal relationships by contributing
to self-awareness and the effective communication of one’s emotional states. (“If I
am upset, I know the cause of it.”)
Nonverbal Emotional Expression (Emotion in the Self: Nonverbal). The second
dimension refers to the communication of one’s feelings to others through bodily
(i.e., nonverbal) expression. An example is the willingness to show affection toward
a friend or loved one by way of a tender touch or hug. The outward expression of
one’s true feelings is expected to provide a basis for trust in interpersonal relationships. (“Emotionally, I am very easy to read.”)
Recognition of Emotion in Others (Emotion in Others: Nonverbal). What people
tell us they are feeling often provides an incomplete message. Richer information can
be obtained by subtler, nonverbal cues, such as tone of voice and facial expression.
Attending to such cues is important for maintaining good interpersonal relations
because the appropriateness of one’s own behavior in social situations depends in
part on what others are experiencing emotionally. (“I can tell how people are feeling
5
Chapter 1
even if they never tell me.”)
Empathy (Emotion in Others: Empathy). This dimension also deals with others’
feelings; but, more than being concerned with reading those feelings, it captures
willingness to understand them by relating them to one’s own experiences. Empathy
may be considered a major contributor to EI in that it not only shows willingness
to understand others’ feelings but also indicates accessibility to one’s own. (“I am
sensitive to the feelings of other people.”)
Regulation of Emotion in the Self. Two components in Salovey and Mayer’s (1990)
model pertain to regulating emotions. With respect to the self, strong emotions can
overpower rational thinking, leading to critical errors in judgment (i.e., what Goleman, 1995, calls “emotional highjackings”). Such errors can put the individual in
physical danger or jeopardize interpersonal relations and goal attainment. Selfregulation can be achieved unconsciously, as in anticipation of pleasant outcomes,
as well as consciously, as in the deliberate selection or avoidance of mood inducing
situations. (“I can keep myself calm even in highly stressful situations.”)
Figure 1.2: Salovey & Mayer’s (1990) Model of Emotional Intelligence
Regulation of Emotion in Others. Managing interpersonal relationships can
entail influencing others’ affective states. Particular cases include (1) impression
management, as in the employment setting when trying to be hired or promoted, (2)
leadership, as in motivating others to reach a shared goal, and (3) friendship, as in
cheering up a companion when he or she is feeling down. (“Usually, I know what
it takes to turn someone else’s boredom into excitement.”)
Intuition versus Reason (Flexible Planning). The last 4 components of Salovey
and Mayer’s model pertain to “applications” of EI. In the context of the MEIA,
6
Introduction
they are also referred to as “proximal outcomes.” The first of these applications is
the use of emotions in pursuing important life goals. Mood swings can provide a
broader array of emotional considerations in decision-making. In present efforts,
this dimension was conceived as a preference to base decisions on feelings over
logic. (“I often use my intuition in planning for the future.”)
Creative Thinking. A second application of EI is its contribution to creativity.
Salovey and Mayer suggest that emotions can facilitate creative thinking by altering
the organization and use of information in memory. Creativity, in turn, expands one’s
options in making important life decisions. (“People think my ideas are daring.”)
Mood Redirected Attention. A third utilization of EI described by Salovey and
Mayer is the propensity to attend to information about the self when powerful — usually negative — emotions occur. Such emotions can improve one’s self-awareness
by directing attention to their source, which, in turn, can reveal latent priorities. An
example would be the revelation of a personal weakness in dealing with children
brought on by frustration in taking care of a friend’s toddler. (“Having strong emotions forces me to understand myself.”)
Motivating Emotions. The last application of EI refers to drive and perseverance
in pursuing one’s goals. People high in EI are more ambitious and optimistic, traits
that can serve those individuals themselves, as well as motivate those around them.
(“I believe I can do almost anything I set out to do.”)
Table 1.1 summarizes the 10 primary MEIA scales, with definitions and sample
items, and their corresponding sources in Salovey and Mayer’s (1990) model. The
scales have been reordered in light of research findings supporting a structure different from that proposed by Salovey and Mayer (1990). (See Chapter 3.)
The Challenge of Social Desirability in Self-Description
Social desirability has long been a focus of research (e.g., Crowne & Marlowe, 1964;
Edwards, 1957; Jackson & Messick, 1958) and the subject of controversy as a source
of bias (e.g., Paulhus, 1984) versus valid trait variance (e.g., Ones, Viswesvaran, &
Reiss, 1996) in self-report measures. As a bias, it warrants control in scale construction and/or score interpretation (Bar-On, 1997). As a source of valid trait variance,
on the other hand, removing or limiting desirability undermines validity. Helmes
(2000) offers the following compromise: In constructing self-report measures, “a
reasonable criterion in selecting items would be that items be associated more strongly
with the construct of interest than with a measure of social desirability” (p. 36).
Explicit consideration of social desirability in the assessment of EI using selfreport has been notably lacking in the literature (Roberts et al., 2001). As an ability
(Mayer & Salovey, 1997), EI is generally recognized as desirable. Accordingly, selfdescription on EI facets can be expected to reveal individual differences not only in
the targeted construct (e.g., empathy), but also the tendency to describe oneself in a
favorable light. In developing the MEIA, we took Helmes’ (2000) position, noted
7
Chapter 1
Table 1.1: Multidimensional Emotional Intelligence Assessment (MEIA) scales, definitions, and sample
items
S & M (1990)
label*
Core EI
MEIA label
Abbrev.
Definition
Sample item
Emotion in the Self:
Verbal
Recognition of
Emotion in the Self
RecSlf
being in touch with one’s If I am upset, I know the
feelings and describing
cause of it.
those feelings in words
Regulation of
Emotion in the Self
Regulation of
Emotion in the Self
RegSlf
controlling one’s
own emotional
states, particularly in
emotionally arousing
situations
I can keep myself calm
even in highly stressful
situations.
Emotion in Others:
Nonverbal
Recognition of
Emotion in Others
RecOth
attending to others’
nonverbal emotional
cues, such as facial
expressions and tone of
voice
I can tell how people
are feeling even if they
never tell me.
Regulation of
Emotion in Others
Regulation of
Emotion in Others
RegOth
managing others’
emotional states,
particularly in
emotionally arousing
situations
Usually, I know what it
takes to turn someone
else’s boredom into
excitement.
Emotion in the Self:
Nonverbal
Nonverbal
Emotional
Expression
NvExp
communicating one’s
feelings to others
through bodily (i.e.,
nonverbal) expression
I like to hug those who
are emotionally close to
me.
Emotion in Others:
Empathy
Empathy
Emp
understanding others’
emotions by relating
them to one’s own
experiences
I am sensitive to the
feelings of other people.
Proximal Outcomes
Flexible Planning
Intuition versus
Reason
IvR
using emotions in the
pursuit of life goals;
basing decisions on
feelings over logic
I often use my intuition
in planning for the
future.
Creative Thinking
Creative Thinking
CrTh
using emotions to
facilitate divergent
thinking
People think my ideas
are daring.
Mood Redirected
Attention
Mood Redirected
Attention
MRA
interpreting strong
–usually negative–
emotions in a positive
light
Having strong emotions
forces me to understand
myself.
Motivating
Emotions
Motivating
Emotions
MotEm
pursuing one’s
goals with drive,
perseverance, and
optimism
I believe I can do almost
anything I set out to do.
* Salovey & Mayer (1990)
8
Introduction
above, seeking to ensure that every item correlates stronger with its own scale than
it does with any other scale, including one measuring social desirability. Success in
meeting this objective would mean that the MEIA is less susceptible to desirability
response bias, and hence to unwanted variance, than self-report EI measures lacking
that criterion in item selection.
Administration and Scoring
The MEIA and MEIA-W can be administered and scored online at www.sigmatesting.
com. The test-taker is directed to a URL (link) where the test is located and asked
to indicate his or her agreement with each item on a 6-point scale. Once the test is
completed, the administrator logs in to score the responses and generate a report.
Norms
Means and standard deviations for the MEIA’s 10 primary scales are provided in Table
1.2, based on the responses of 332 American college students of whom approximately
60% are women. The students were undergraduates, mostly in their freshman year,
representing a wide range of academic and career interests. They were enrolled at
Table 1.2: College student norms for the Multidimensional Emotional Intelligence Assessment (MEIA)
All
N = 332
Men
N = 127
Women
N = 188
Men vs. Women
(M = 1, F = 2)
2-tailed
SEI Scale
Mean SD
Mean SD
Mean SD
t
p
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
Proximal Outcomes
IvR
CrTh
MRA
MotEm
4.13
3.84
4.40
4.27
4.04
3.99
.77
.91
.75
.71
.76
.77
4.23
4.07
4.30
4.26
3.90
3.85
.71
.85
.73
.70
.78
.77
4.05
3.66
4.49
4.29
4.10
4.07
.79
.91
.75
.72
.74
.76
2.03
4.03
-2.20
-.31
-2.25
-2.50
.04
.00
.03
.76
.03
.01
3.52
3.55
4.46
4.63
.77
.67
.78
.74
3.45
3.57
4.49
4.66
.79
.66
.78
.75
3.58
3.52
4.46
4.59
.75
.65
.79
.74
-1.52
.73
.41
.76
.13
.46
.68
.45
Inf
Infa
1.48
1.57
.39
.56
1.46 .39
1.59 .59
1.49
1.57
.39
.55
-.65
.27
.51
.79
20.20 3.78
20.25 3.39
20.19 4.02
.14
.89
Age (years)
Cases with Inf > 2.5 (N = 20) have been dropped (1-6 scale).
a
Inf for all cases, including > 2.5; N(all) = 352, N(men) = 138, N(women) = 197.
9
Chapter 1
one of two Oklahoma colleges, and responses were gathered between fall of 2001
and fall of 2004. Cases with unusually high Infrequency (Inf) scores (i.e., >2.5
on a 1-6 scale; N = 20) were dropped prior to calculating the norms to remove the
unwanted effects of non-purposeful responding. Norms are provided for MEIA Inf
based on all 332 respondents to allow normative comparisons on Inf per se, should
the need arise. Norms are also provided for men (N = 127) and women (N = 188)
separately, in the event that uses of the MEIA warrant normative comparisons within
sex. Notably, as shown at the right side of Table 1.2, men scored significantly higher
on RecSlf and RegSlf, whereas women scored higher on RecOth, Emp, and NvExp.
Reliabilities and Interscale Correlations
In general, reliability in psychological measurement denotes consistency in an individual’s responses to a test or survey. Consistency is important because it conveys
that a given score on a test is credible, that it is not simply the result of random
fluctuations in an individual’s responding. It is important to realize that consistency
(i.e., reliability) does not imply accuracy (i.e., validity): a test may be reliable (i.e.,
show consistency in measuring something) without being valid (i.e., measuring a
particular targeted trait). Reliability, however, is required to show validity: a test
that yields inconsistent scores within itself (i.e., one with low reliability) cannot be
expected to yield predictable or otherwise meaningful relations with other measures
(i.e., show validity).
There are different types of score consistency and, hence, different types of reliability. Two common forms are test-retest and internal consistency reliability. The
first captures the degree to which a test yields the same scores for the same test
takers at two points in time, typically separated by two or more weeks. Most traits,
including those targeted by the MEIA, are defined as relatively stable dispositions.
Corresponding measures, then, should show relatively high stability over time (i.e.,
test-retest reliability). Internal consistency reliability captures the degree to which
all the items on a given scale (e.g., the individual items on a given MEIA scale) elicit
similar responses (i.e., agree/ disagree) from a given test taker. Thus, it represents
the similarity or “homogeneity” of a scale’s items. Cronbach’s alpha, the most common index of internal consistency reliability, ranges from 0 (no reliability) to 1.00
(perfect reliability). Values in the range of .60 to .90 are common for trait measures
similar to those of the MEIA.
Test-retest reliability coefficients (correlations between scores from two administrations of the same test) for the MEIA primary scales are provided in Table 1.3,
based on responses from 68 college students who completed the MEIA twice with
a 4- to 6-week interval separating test administrations. Ranging from .67 to .88,
the test-retest coefficients for the MEIA are very much in line with those reported
for scales targeting similar psychological dispositions (e.g., personality traits) and
suggest sufficient stability in test scores over time to allow meaningful score interpretations.
10
Introduction
Table 1.3: Test-retest reliabilities for the 10 MEIA primary scales over a 4- to 6- week interval (N = 68
college students)
SEI scale
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
Proximal Outcomes
IvR
CrTh
MRA
MotEm
Mean
Median
Test-retest
reliability
.72
.88
.67
.83
.80
.84
.76
.76
.74
.79
.78
.78
Correlations among the subscales of a multidimensional inventory (such as the
MEIA) are informative by revealing the latent structure of the subscale set, that is,
how the subscales cluster together. Knowing which subscales are interrelated can
facilitate score interpretations. For example, if Scale A is correlated positively with
Scale B, then people who score highly on one can be expected, in general, to score
highly on the other. By extension, those who score at opposite ends (i.e., high on
one, low on the other) can be identified as relatively rare cases.
Correlations among the 10 primary MEIA scales, along with alpha reliability
coefficients, are reported in Table 1.4 based on the total normative sample (N = 332
college students). Correlations and alphas based on men and women are offered
separately in Table 1.5. Alpha coefficients in both tables are moderately strong,
overall (range = .75 to .85, based on the combined data and for women; .75 to .82
for men), suggesting good internal consistency reliability. The MEIA interscale
correlations are generally weak to moderate in strength (range = -.23 to .57, based
on the combined data; -.26 to .61 for men; -.15 to .54 for women), suggesting that
each scale captures a relatively unique aspect of EI. Cases of greater overlap (e.g.,
RecOth and RegOth correlate .57 in the combined data, Table 1.4) suggest opportunity to consider EI in slightly broader terms, but the overlap is not so great as to
render scores on overlapping scales completely redundant. Correlations among the
MEIA scales and redundancies with other measures are reviewed in greater detail
in Chapter 3 with respect to the MEIA’s validity.
Comparing correlations between men and women (Table 1.5) suggests overall
similar patterns. The largest exceptions (e.g., RecSlf and NvExp correlate .24 in
men vs. .03 in women) may suggest uniqueness in the nature of EI within sex, and
11
Chapter 1
Table 1.4: Correlations among 10 MEIA primary scales and alpha coefficients based on college student
norms (N = 332)
Core EI
RecSlf
Core EI
RecSlf
.75
RegSlf
.37
RecOth .30
RegOth .34
NvExp
.09
Emp
.32
Proximal Outcomes
IvR
-.05
CrTh
.12
MRA
.09
MotEm .47
RegSlf
RecOth RegOth
Proximal Outcomes
NvExp
Emp
.85
.09
.14
-.14
-.03
.83
.57
.20
.28
.81
.31
.40
.83
.42
.77
-.23
.13
-.14
.42
.13
.26
.19
.31
.16
.27
.26
.45
.24
.07
.22
.32
.11
.16
.15
.10
IvR
CrTh
MRA
MotEm
.82
.19
.26
-.02
.77
.07
.16
.75
.23
.81
Values in bold are alpha coefficients.
p < .05, two-tailed, for | r | = .11; p < .01, two-tailed, for | r | = .14.
Table 1.5: Correlations among 10 MEIA primary scales and alpha coefficients based on college student
norms for men (upper right, N = 127) and women (lower left, N = 188)
Core EI
RecSlf
RegSlf
Core EI
RecSlf .75\.75
.27
RegSlf
.40
.85\.82
RecOth .33
.14
RegOth .31
.12
NvExp
.03
-.15
Emp
.28
-.07
Proximal Outcomes
IvR
-.05
-.15
CrTh
.12
.19
MRA
.01
-.14
MotEm .44
.37
Proximal Outcomes
RecOth RegOth
NvExp
Emp
IvR
CrTh
MRA
MotEm
.38
.06
.83\.82
.54
.11
.26
.47
.25
.61
.81\.82
.27
.44
.24
-.14
.31
.41
.85\.79
.47
.43
.03
.29
.40
.38
.75\.78
.02
-.26
.10
.03
.04
.11
.19
.06
.28
.26
.21
.07
.24
-.13
.31
.28
.06
.16
.51
.47
.24
.44
.08
.32
.14
.29
.16
.37
.24
.28
.29
.48
.15
.12
.27
.12
.39
.17
.29
.31
.81\.82
.21
.25
.04
.16
.78\.74
.13
.15
Values in bold are alpha coefficients (women\men).
For men, p < .05, two-tailed, for | r | = .18; p < .01, two-tailed, for | r | = .23.
For women, p < .05, two-tailed, for | r | = .15; p < .01, two-tailed, for | r | = .19.
12
.29
-.06
.02
.21
.77\.74
.31
.20
.81\.82
Introduction
correspondingly, the way that MEIA scores might be interpreted per sex. On the
other hand, fluctuations between samples of the current size (127 men, 188 women)
may reflect random aberrations due to sampling error (i.e., gathering data from 127
different college student men might yield differences from the current set of 127
men just as large as those seen here between men and women). Further research on
sex differences in EI structure may help explain the noted patterns of correlations.
Until evidence accrues warranting clear separation of men and women in consideration of EI structure, MEIA users are urged to rely on correlations based on both
sexes combined (Table 1.4).
Summary
The MEIA is a brief self-report measure of that part of emotional intelligence conceived as a set of relatively stable and distinct personality-type traits bearing on the
perception, control, and use of emotions in the self and others. It is not intended
to assess the entirety of EI, which also includes emotional ability components, but
rather more specifically just the 10 dispositional facets articulated by Salovey and
Mayer (1990). The remainder of this manual describes how the MEIA was developed
(Chapter 2), evidence bearing on the MEIA’s validity (Chapter 3), and the workplace
version, the MEIA-W (Chapter 4).
13
Chapter 2
Construction of the MEIA
The MEIA was constructed following in key respects the construct-oriented approach developed by Douglas N. Jackson, Ph.D. (Jackson, 1970) in creating the
Personality Research Form (PRF; Jackson, 1989) and Jackson Personality Inventory–Revised (JPI–R; Jackson; 1994). In general, this approach combines rational
and empirical methods to yield reliable and construct-valid self-report measures.
The method begins with clear definitions of targeted constructs. Items are written
to capture each construct, as defined, and to minimize susceptibility to acquiescence
and social desirability biases. Responses to the items are collected from a sample
representative of targeted populations and statistical analyses are undertaken to
identify items meeting several quantitative criteria. In light of those results, current
items are modified or new items are written, then further data are collected, new
analyses undertaken, and so on in a cycle designed to yield scales demonstrating
acceptable reliability and validity. The specific steps undertaken in developing the
MEIA are described below in the context of 2 initial studies undertaken from 1996
to 1998 at the University of Central Florida.
Study 1: Development of Preliminary EI Scales
Our first objective was to develop item sets representing each of the 10 components of Salovey and Mayer’s (1990) model. Efforts were directed, in particular, to
creating scales having (a) broad yet unique domain coverage in support of content
validity, (b) control of acquiescence response bias by inclusion of equal numbers of
true- and false-keyed items per scale, (c) adequate internal consistency reliability,
and (d) evidence supporting convergent and discriminant validity in relations with
selected personality traits, including social desirability.
Preparation of Item Sets
Items representing each of the 10 EI constructs were written independently by 1 female and 3 male psychology professors with collective expertise spanning personality,
experimental, cognitive, clinical, social, industrial/organizational, and measurement
areas. Item writers were asked to create 6 positively and 6 negatively keyed selfdescriptive statements per dimension. Items were to be (a) clearly representative
of a specific dimension as defined by Salovey and Mayer (1990) (see Table 1.1 and
15
Chapter 2
corresponding discussion); (b) readable at the grade 8 level, (c) as brief as possible,
(d) void of double negatives and other complications, and (e) void of overly desirable or undesirable content. Redundant items from multiple writers were merged or
dropped. The 307 surviving items were pooled randomly to form a preliminary list.
Confirmation of Content Coverage
To help ensure that items were uniquely representative of their targeted scales, the
item writers independently recategorized each item with respect to both dimensions
and keying. An item was retained if at least 2 of the 3 judges not contributing that
item placed it on its targeted scale. The 122 items surviving this step were combined
randomly with the 16 items from the PRF Desirability scale and 6 newly written
infrequency items (e.g., “I have lived north of the arctic circle,” 3 keyed in each
direction) to form a single questionnaire. The infrequency items were developed
to detect cases of nonpurposeful responding, removal of which would make results
more interpretable. A Likert scale was used ranging from strongly disagree (1) to
strongly agree (6).
Data Collection
The preliminary item set was administered to a total of 174 American undergraduate
university students enrolled in a personality class. Sixty-six percent ranged in age
from 20 to 24 years, and 70% were female. Respondents were encouraged to answer
all statements even if they were unsure in particular cases. Six participants had
mean infrequency scores of 3 or higher and 10 additional cases included a number
of missing responses, leaving N = 158.
Item Analyses
After reflecting responses to false-keyed items (i.e., such that 6’s became 1’s, 5’s
became 2’s, etc.), item analysis proceeded in 2 stages. First, reliability output for
each of the 10 item sets was reviewed with respect to both Cronbach’s alpha and
each item’s corrected item-total correlation (CITr; i.e., the correlation between an
item and the unit-weighted sum of the remaining items on the same scale). Items
with a low (or negative) CITr were removed over several iterations to yield relatively
homogeneous scales. In the second stage, every item was correlated with the sum
of the items retained on each of the other scales as well as the PRF-Desirability total
score. Any item that correlated more strongly with a non-targeted scale, including Desirability, than it did with its own scale (as per its CITr) was dropped. This
procedure was repeated over several iterations to yield a homogenous and unique
item set per dimension.
Results supported scale construction efforts in several respects. Most importantly,
a single set of relatively homogeneous items was retained for each of the 10 targeted
dimensions. The means, ranging from 3.6 to 4.8 (median = 4.3) on a 6-point scale,
suggested lack of ceiling and floor effects, and the standard deviations, ranging
16
Construction of the MEIA
from .68 to 1.02 (median = .76), suggested individual differences sufficient to allow meaningful correlations involving the EI scales. Alphas for these preliminary
scales ranged from .52 for NvExp to .84 for RegSlf (median alpha = .74). These
values, although modest in some cases, were encouraging because scales with the
lowest alphas were those with the fewest items (4 items on each of 2 scales, alphas
= .52 and .58), and alpha would be expected to increase with the addition of new
items in Study 2 (see below). Also, following Helmes (2000), each item on each
scale was retained if it correlated with the other items on its own scale higher than
it did with any other scale, including PRF-Desirability. Thus, internal consistency
for each scale reflected homogeneity that is maximally unique with respect to the
targeted content of that scale. Uniqueness in content coverage reduces unwanted
interscale dependencies (i.e., due to shared bias), a property that would benefit by
the addition of new, similarly developed items.
Evaluation of Convergent and Discriminant Validity
Responses to items surviving the item analysis were averaged to form 10 EI scale
scores. The same students had been given the JPI-R earlier in the term. The number
of cases usable for both measures was 138. Convergent and discriminant validity was
assessed by correlating each scale with each of the JPI-R’s 15 scales (scale names are
provided below in tables of results). Prior to data collection, predictions of positive,
negative, or neutral (i.e., r = 0) relations were made independently by each of the 4
item writers, as well as a graduate student and a senior undergraduate student, based
on construct definitions. Inter-judge consistency in predicting EI–JPI-R relations
was far above chance levels. Unanimous agreement was obtained for 5 directional
predictions (all positive except as noted): EI Emp with JPI-R Empathy, EI RegSlf
with JPI-R Anxiety (negative), EI RegOth with JPI-R Social Astuteness, EI CrTh
with JPI-R Innovation, and EI MotEm with JPI-R Energy Level. Agreement of 83%
(i.e., 5 out of 6 judges) was obtained for EI RecOth and JPI-R Empathy, EI Emp and
JPI-R Tolerance, and EI CrTh and JPI-R Traditional Values (negative). Agreement
of 66% was obtained in 3 cases, and unanimous agreement in predicting a zero correlation was achieved in 51 of the remaining 139 linkages.
Correlations between the preliminary EI scales and the JPI-R and PRF Desirability
are shown in Table 2.1. Several points are noteworthy. First, convergent validity for
some scales is reflected in predicted directional correlations. As expected, EI Emp
correlated with JPI-R Empathy (.58), EI RegSlf correlated negatively with JPI-R
Anxiety (-.61), EI CrTh correlated with JPI-R Innovation (.66), and EI MotEm correlated with JPI-R Energy Level (.57). These correlations are the strongest in the set
and are 4 of the 5 predicted with 100% consensus. The other unanimous prediction,
EI RegOth with JPI-R Social Astuteness, yielded a weaker-than-expected correlation
of .12. Correlations between each of the two contributing scales and Desirability
(i.e., positive in the case of RegOth and negative for Social Astuteness) suggest a
suppressor effect. Partialing Desirability from the noted correlation gives a value
17
Chapter 2
of .31, more in keeping with expectations. Two of the 3 relations predicted with
83% agreement are significant though modest: EI RegOth correlated .21 with JPI-R
Empathy, and EI Emp correlated .18 with JPI-R Tolerance. The correlation between
EI CrTh and JPI-R Traditional Values, predicted with 83% agreement, was weaker
than expected (-.09). Overall, with few exceptions, results supported the convergent
validity of the preliminary scales and their continued development.
Of the 51 unanimous predictions of r = 0, 16 were significant (two-tailed). The
resulting 31% is higher than chance levels, suggesting that either the scales are
measuring non-targeted constructs or the predictions are incomplete. Most of the
unpredicted relations permit interpretation in keeping with the intended purposes
of the scales. For example, that EI Emp correlated .34 with JPI-R Anxiety suggests
emotionality, a general dimension to which both empathy and anxiety contribute
(Jackson, 1994). Similarly compelling post hoc rationales can account for most of
the remaining stronger relations. Study 2 allowed replication using refined scales.
Relations with PRF-Desirability
The far right column of Table 2.1 contains correlations with PRF-Desirability. The
highest values involve MotEm (.58), RegSlf (.49), RegOth (.44), and RecSlf (.42).
Rating the self highly on these scales could reflect true individual differences on
what are arguably the most desirable qualities in the set. To assess the role of social
desirability in current validation efforts, Desirability scores were partialed out of
relations between the proposed and JPI-R scales. Results are reported in Table 2.2.
Given the .58 correlation between EI MotEm and Desirability, it is not surprising
that validities involving the former are substantially reduced in several cases (e.g.,
.22 to .06 for JPI-R Breadth of Interest). Similar drops are evident for other scales.
Validities increase in some cases (e.g., EI Emp in relations with several JPI-R scales),
suggesting a suppressor role for Desirability like that noted above regarding EI
RegOth in relation to JPI-R Social Astuteness. Overall, validity interpretations are
clarified by the removal of Desirability, but results for targeted relations are largely
unchanged.
Summary
Results from Study 1 encouraged further development of self-report measures of
the 10 components of Salovey and Mayer’s (1990) model. In particular, the scales
showed (a) good domain representation as per high consensus among independent judges of individual items, (b) adequate internal consistency reliability while
maximizing uniqueness of domain coverage, (c) adequate convergent validity with
respect to selected JPI-R scales, and (d) adequate discriminant validity in light of
predicted r = 0 relationships and plausible interpretations of the stronger exceptions.
All told, results supported further efforts to develop self-report measures of EI as a
multidimensional disposition.
18
Construction of the MEIA
CPX
.03
.30**
.06
.22*
.11
.13
.23**
.12
.08
.09
BDI
.20*
.66**
.16
.13
.00
.20*
.08
.07
.30**
.09
INV
.15
.13
-.02
.19
.11
.01
.09
.15
-.19*
.18*
TOL
.27**
.10
.34**
-.14
-.46**
.10
-.08
.26**
.21*
.58**
EMP
.15
-.05
.26**
-.38**
-.61**
-.08
-.21*
.00
.13
.34**
AXY
.08
-.05
.13
-.30**
-.34**
-.16
-.26**
-.14
.03
.19**
CPR
.08
-.08
.00
-.03
-.21*
.19*
.00
.17*
-.01
.32**
SOC
.06
.15
-.05
.41**
.25**
.51**
.23**
.34**
.07
-.07
SCF
-.03
.13
-.17*
.57**
.44**
.32**
.21*
.24**
-.02
.05
ENL
-.10
.20*
.13
-.21*
-.11
.12
-.17*
-.12
.25**
-.03
SAS
.06
.32**
.12
.14
.06
.11
-.13
-.18*
.04
-.16
RKT
-.09
-.16
-.02
.26**
.26**
.12
.21*
.09
-.07
-.05
ORG
-.06
-.09
.01
.27**
.09
.20*
.07
.24**
.04
.23**
TRV
.10
.09*
-.03
.33**
.11
.19*
.11
.22**
.07
.45**
RSY
.10
.05
-.10
.58**
.49**
.44**
.42**
.34**
.12
.23**
DESa
JPI-R subscale
Table 2.1: Correlations between the 10 EI and 15 JPI-R and PRF-Desirability (DES) subscales in Study 1 (N = 138)
MEIA subscale
Appraisal & Expression
RecSlf
.05
NvExp
-.02
RecOth
.09
Emp
.07
Regulation
RegSlf
-.11
RegOth
.01
Applications
IvR
.09
CrTh
.50**
MRA
.21*
MotEm
.01
.09
.50**
.22*
-.01
-.15
.00
.04
-.03
.09
.07
CPX
.00
.30**
.09
.06
-.04
-.01
.12
.02
.05
.03
BDI
.20*
.66**
.16
.16
.00
.22**
.09
.08
.30**
.10
INV
.14
.12
.00
.10
.02
-.10
.00
.10
-.22**
.15
TOL
.28**
.10
.33**
-.11
-.48**
.17*
-.04
.30**
.23**
.61**
EMP
.18*
-.05
.25**
-.29**
-.57**
.06
-.10
.10
.17*
.42**
AXY
.11
-.04
.11
-.19*
-.25**
-.04
-.17*
-.05
.07
.25**
CPR
.07
-.08
.01
-.12
-.31**
.16
-.06
.14
-.03
.30**
SOC
.03
.14
-.02
.28**
.10
.42**
.09
.26**
.03
-.15
SCF
-.10
.13
-.15
.37**
.23**
.09
-.04
.08
-.11
-.07
ENL
-.07
.22**
.11
-.05
.05
.31**
-.05
-.02
.30**
.03
SAS
.07
.32**
.11
.23**
.12
.17*
-.10
-.16
.05
-.15
RKT
-.13
-.18*
.01
.11
.14
-.02
.09
.00
-.11
-.12
ORG
-.09
-.10
.04
.15
-.05
.09
-.06
.17*
.01
.19*
TRV
.07
.09
.01
.17*
-.08
.03
-.05
.12
.03
.41**
RSY
JPI-R subscale
Table 2.2: Correlations between the 10 EI and 15 JPI-R subscales controlling for PRF-Desirability in Study 1 (N = 138)
*p < .05, **p < .01, two-tailed tests.
CPX = Complexity; BDI = Breadth of Interest; INV = Innovation; TOL = Tolerance; EMP = Empathy; AXY = Anxiety; CPR = Cooperativeness; SOC = Sociability;
SCF = Social Confidence; ENL = Energy Level; SAS = Social Astuteness; RKT = Risk Taking; ORG = Organization; TRV = Traditional Values; RSY =
Responsibility; DES = PRF-Desirability.
N = 158 for these correlations
a
MEIA subscale
Appraisal & Expression
RecSlf
NvExp
RecOth
Emp
Regulation
RegSlf
RegOth
Applications
IvR
CrTh
MRA
MotEm
*p < .05, **p < .01, two-tailed tests.
CPX = Complexity; BDI = Breadth of Interest; INV = Innovation; TOL = Tolerance; EMP = Empathy; AXY = Anxiety; CPR = Cooperativeness; SOC = Sociability;
SCF = Social Confidence; ENL = Energy Level; SAS = Social Astuteness; RKT = Risk Taking; ORG = Organization; TRV = Traditional Values; RSY = Responsibility.
19
Chapter 2
Study 2: Assessing the Uniqueness of EI as a Trait Domain
The prominence of emotional intelligence as a research-worthy construct is a relatively recent phenomenon. Given the conceptualization of EI as a trait, it is important
to establish its independence from more established trait constructs. A similar point
arises in considering the uniqueness of each of the 10 facets of EI proposed by Salovey and Mayer (1990). Principles of parsimony and practicality led us to assess the
degree to which each scale captures variance independent of that on the remaining
scales. Failure to establish such uniqueness in coverage could prompt a reduction
in the number of scales and corresponding reconsideration of the EI construct.
Specific goals in Study 2 were to (a) improve each subscale’s internal consistency
reliability by the addition of domain-specific items, (b) balance positively and negatively keyed items to control acquiescence response bias, (c) reassess convergent and
discriminant validity in light of relations with specific JPI-R subscales, (d) evaluate
the test-retest stability of the EI subscales, and (e) examine the overall uniqueness
of the EI subscales as distinct from social desirability, mainstream personality traits,
and the remaining EI facets.
Item Preparation
New items were prepared by the 4 original item writers and a female undergraduate
student according to the same criteria used in Study 1 (e.g., avoidance of double
negatives). Results from Study 1 suggested that reliability of at least .80 could
be achieved using a total of 12 items per scale, 6 keyed in each direction. The
number of new items written for each scale was determined under the assumption
that half would survive item analysis. As in Study 1, new items were assessed for
their fidelity to targeted constructs by subjective categorization by dimension and
keying. A graduate student was added as a sixth judge. An item was retained for
further consideration if it was categorized correctly by at least 3 of the 5 judges not
contributing that item. The new items were classified with at least 60% consensus
in sufficient numbers to warrant data collection and item analysis.
Data Collection
All surviving new items were combined randomly with the scale items surviving
Study 1, the 6 infrequency items, and the 16 PRF Desirability items for a total of
189. The 6-point response format used in Study 1 was used again here. The revised questionnaire was administered to a total of 191 American undergraduates in
a developmental psychology class. Ages ranged from 18 to 49 years with a mean
of 21.8, and 74% were female. The JPI-R and a brief demographic data sheet were
given on a different day. Removal of cases with mean infrequency scores of at least
3 or a preponderance of missing responses left 183 usable cases for the EI scales and
163 for both the EI and JPI-R measures. Sixty-eight participants completed the EI
20
Construction of the MEIA
items 4 to 6 weeks after the first administration, permitting evaluation of test-retest
reliability (see Table 1.3).
Item Analyses
The procedure used in Study 1 was repeated here to derive maximally homogeneous
scales. Specifically, each scale retained only those items correlating higher with the
sum of the remaining items on the targeted scale than with total scores on any other
scale, including Desirability. Results supported efforts to improve the preliminary
scales. Means, standard deviations, alpha coefficients, and inter-scale correlations
are presented in Table 2.3. As in Study 1, variability on each scale was deemed
adequate to allow correlations among scales and with other measures. For 6 scales,
item analysis led to removal of between 1 and 4 items that survived Study 1 (total
= 13 items dropped). Resulting alphas range from .74 for Emp to .87 for MotEm
(median = .81). These values are higher than those derived in Study 1 due to the
addition of new items. The number of items per scale averages 12.1 in Study 2
(range = 8 to 15) compared to 6.6 in Study 1 (range = 4 to 11). Of the 3 scales with
alphas under .80, 2 have fewer than the desired 12 items (8 and 10) and the other
has an alpha of .79 for 13 items.
Attempts to balance item keying were successful on the whole. An exception was
Mood Redirected Attention (MRA) for which no false-keyed items survived. The
reason for this is not clear. Negatively keyed items were dropped disproportionately
during both item sorting and item analysis, suggesting that judges and respondents
treated the negatively keyed items on that scale differently. In any case, MRA may
be more susceptible to acquiescence response bias than are the other scales.
Evaluation of Convergent and Discriminant Validity
As in Study 1, scale validities were assessed using correlations with the JPI-R and
with PRF Desirability. The same predictions forwarded in Study 1 were used here
to guide evaluation of convergent and discriminant validity for particular scales.
Unpredicted but interpretable relations observed in Study 1 were also considered
as prospective replications. Finally, to assess the uniqueness of the EI scales, we
hierarchically regressed each one (individually) onto (a) Desirability (forced entry),
(b) the 15 JPI-R scales (stepwise), and (c) the remaining 9 EI scales (stepwise).
Retention of a given EI scale as representative of a distinct trait would be supported
to the degree that reliable variance remains unaccounted for by the other scales in
the analysis.
Consistent with Study 1, correlations with the JPI-R scales and Desirability were
generally supportive of the EI scales’ convergent and discriminant validity. Results
are shown in Table 2.4. Four of the 5 EI–JPI-R linkages predicted with unanimous
agreement are .45, .55, -.56, and .60, all in the expected direction paralleling Study
1 results. The fifth relation, between EI RegOth and JPI-R Social Astuteness, was
only .13, but it rose to .25 after Desirability was partialed out. This too mirrors
21
22
4.20
3.98
4.48
4.56
3.79
4.49
3.77
3.88
4.55
4.60
13
12
13
12
8
15
Mean
12
13
13
10
N of
items
.66
.75
.69
.71
.86
.70
.77
.75
.72
.68
SD
-.11
.14
.28**
.34**
.38**
.34**
.82
.44**
.32**
.06
RecSlf
.13
.03
.12
.36**
.23**
.30**
.25**
.80
.19**
.28**
NvExp
.02
.47**
.34**
.42**
.23**
.55**
.20**
.04
.83
.24**
RecOth
Appraisal & Expression
.30**
.10
.36**
.15*
-.13
.41**
-.09
.18*
.17*
.74
Emp
.11
.37**
.33**
.46**
.03
.83
.85
.24**
-.10
.18*
-.04
.37**
.14
.08
.48**
.34**
RegOth
.20**
.00
.10
-.29**
RegSlf
Regulation
.15*
.36**
.04
.81
.16*
.22**
.79
.04
.25**
.07
.10
-.03
.46**
.08
CrT
-.13
.11
-.14
.14
.01
.30**
IvR
-.03
-.05
.16*
-.04
.12
.11
.08
.60**
.01
.14
.17*
.22**
.07
.47**
.10
.11
INV
.10
.06
.19*
.13
BDI
.06
.19*
-.15*
-.01
.05
.03
-.03
.03
-.09
.01
TOL
.27**
.08
.16*
-.03
-.32**
.25**
-.11
.22**
.05
.55**
EMP
.08
-.20**
-.06
-.21**
-.56**
-.07
-.23**
-.03
-.21**
.19*
AXY
.06
-.31**
.12
-.28**
-.29**
-.18*
-.29**
-.09*
-.27**
.13
CPR
.12
-.01
-.03
.13
.05
.17*
-.03
.22**
.04
.13
SOC
.03
.37**
-.02
.33**
.27**
.40**
.24**
.28**
.30**
.05
SCF
JPI-R subscale
.00
.19*
-.17*
.45**
.28**
.20**
.13
.22**
.11*
-.06
ENL
.13
.03
.06
.00
-.04
.13
-.14
-.12
.11
.10
SAS
.09
.38**
.11
.19*
.08
.18*
-.02
-.09
.20**
-.08
RKT
-.01
-.18*
.04
.47**
.14
.14
.20**
.16*
.08
-.04
ORG
-.02
-.36**
.09
.23**
.09
.07
.07
.12
-.07
.16*
TRV
*p < .05, **p < .01, two-tailed tests.
CPX = Complexity; BDI = Breadth of Interest; INV = Innovation; TOL = Tolerance; EMP = Empathy; AXY = Anxiety; CPR = Cooperativeness;
SOC = Sociability; SCF = Social Confidence; ENL = Energy Level; SAS = Social Astuteness; RKT = Risk Taking; ORG = Organization;
TRV = Traditional Values; RSY = Responsibility; DES = PRF-Desirability.
Appraisal & Expression
RecSlf
-.03
NvExp
-.06
RecOth
.14
Emp
.11
Regulation
RegSlf
-.19*
RegOth
.11
Applications
IvR
.01
CrTh
.46**
MRA
.10
MotEm
-.01
CPX
.07
-.05
.07
.09
.08
.10
.02
.11
.25**
.62**
.45**
.47**
.49**
.52**
.32**
.27**
DES
.07
.19**
.07
.87
.13
.24**
.06
.06
.30**
-.01
MotEm
.10
.18**
.03
.31**
RSY
Table 2.4: Correlations between the 10 EI and 15 JPI-R and PRF-Desirability (DES) subscales in Study 2 (N = 163)
MEIA subscale
a
.25**
.13
.76
.21**
-.17*
.26**
.19*
-.01
.28**
.32**
MRA
Applications
*p < .05, **p < .01, two-tailed tests; alphas are shown in boldface.
Values below the diagonal are zero-order correlations; those above the diagonal are partial correlations with PRF-Desirability removed.
Appraisal
& Expression
RecSlf
NvExp
RecOth
Emp
Regulation
RegSlf
RegOth
Applications
IvR
CrTh
MRA
MotEm
MEIA subscale
Table 2.3: Numbers of items, means, standard deviations, alphas, and correlations among the 10 EI subscales in Study 2a (N = 183)
Chapter 2
Construction of the MEIA
findings from Study 1. Of the 3 predictions based on 83% consensus, only 1 proved
significant here. Specifically, EI CrTh and JPI-R Traditional Values correlated -.36,
which compares to a value of -.09 from Study 1. Overall, the predicted directional
relations between the proposed scales and the JPI-R were supported.
For relations expected to yield r = 0, the strongest exceptions from Study 1 replicated in Study 2. Several additional cases proved significant in both studies. For
example, EI RecSlf correlated with JPI-R Organization .21 in Study 1 and .20 in
Study 2. This may reflect shared measurement of detailed thinking, consistent with
the Big Five dimension of Conscientiousness. Similarly plausible post hoc rationales can be offered in the remaining cases. Overall, significant EI–JPI-R relations
in both studies are consistent with the intended purposes of the proposed scales, in
support of their validity.
Table 2.5 contains correlations between the EI and JPI-R scales with Desirability
removed. Similar to corresponding results from Study 1, the partial correlations in
most cases are weaker yet many retain their significance. Some relations strengthen,
suggesting a suppressor effect on the part of Desirability (e.g., .13 to .25 for JPI-R
Social Astuteness and EI RegOth).
Unlike in Study 1, correlations among the EI scales were reviewed in Study 2
for evidence bearing on validity. Interscale correlations are reported in Table 2.3,
both with and without controlling for Desirability (above and below the diagonal,
respectively). The strongest zero-order correlation was observed between RegOth
and RecOth (r = .55). Correcting for unreliability using the respective alphas yields
.66, representing 44% overlap in reliable variance. This supports efforts to develop
scales that capture unique portions of the EI domain. Partialing out Desirability
weakens many relationships, especially those involving MotEm. Nonetheless, 21
of the 45 partial correlations are significant (compared to 31 zero-order correlations), suggesting that observed relations are due to more than shared overlap with
Desirability.
Finally, Table 2.6 reports regression results addressing the uniqueness of the
proposed EI scales. The 3 columns under R2 show for each scale the incremental
contributions of Desirability, the JPI-R scales, and the remaining EI scales, respectively. The right column summarizes these findings as the percentages of reliable
variance attributable to the combination of Desirability and significant other scales.
As shown, the 10 EI scales retained between 30% (CrTh) and 76% (Intuition vs.
Reason; IvR) of their reliable variance after accounting for the other variables. This
largely supports the distinctiveness of the proposed scales as self-report measures.
On average, Desirability accounted for 20% of the reliable variance, (range = 0%
for IvR to 45% for MotEm). The JPI-R scales accounted for a further average 24%
(range = 4% for RegSlf to 56% for CrTh), and the remaining EI scales accounted
for another average 12% (range = 0% for RegSlf to 19% for RecOth).
23
24
.07
.02
.18*
.11
.15*
.20**
.06
.47**
.08
.08
-.17*
.17*
.01
.47**
.12
.04
BDI
.01
-.03
.16*
.14
CPX
.08
.61**
.02
.21**
.15*
.14
-.01
-.04
.18*
-.03
INV
.06
.18*
-.17*
-.07
.03
-.01
-.07
.00
-.11
-.01
TOL
.27**
.08
.16*
-.06
-.37**
.27**
-.14
.24**
.05
.57**
EMP
.09
-.18*
.01
-.04
-.51**
.08
-.10
.15*
-.15*
.30**
AXY
.06
-.30**
.07
-.19
-.22**
-.08
-.22**
.03
-.22**
.20**
CPR
.12
-.03
-.06
.06
-.01
.13
-.10
.18*
.01
.12
SOC
JPI-R subscale
.02
.36**
-.06
.28**
.21**
.36**
.17
.22**
.26**
.00
SCF
-.01
.17*
-.26**
.37**
.18*
.07
-.02
.09
.03
-.15*
ENL
.14
.05
.10
.13
.04
.25**
-.07
-.04
.17*
.15*
SAS
.09
.39**
.12
.28**
.11
.23**
.01
-.08
.23**
-.07
RKT
-.03
-.23**
-.05
.32**
-.03
-.05
.02
-.05
-.02
-.16*
ORG
-.03
-.40**
.03
.08
-.03
-.08
-.07
-.02
-.16*
.09
TRV
.06
-.09
-.01
-.16*
-.09
-.09
-.08
.00
-.07
.24**
RSY
*p < .05, **p < .01, two-tailed tests.
CPX = Complexity; BDI = Breadth of Interest; INV = Innovation; TOL = Tolerance; EMP = Empathy; AXY = Anxiety; CPR = Cooperativeness; SOC = Sociability;
SCF = Social Confidence; ENL = Energy Level; SAS = Social Astuteness; RKT = Risk Taking; ORG = Organization; TRV = Traditional Values;
RSY = Responsibility.
Appraisal & Expression
RecSlf
NvExp
RecOth
Emp
Regulation
RegSlf
RegOth
Applications
IvR
CrTh
MRA
MotEm
MEIA subscale
Table 2.5: Correlations between the 10 EI and 15 JPI-R subscales, controlling for PRF-Desirability in Study 2 (N = 163)
Chapter 2
Construction of the MEIA
Table 2.6: EI subscale variance accounted for by PRF-Desirability (DES), JPI-R subscales, and
remaining EI subscales (N = 163)
R2
MEIA subscale
DES
JPI-R
(forced)
(stepwise)
Appraisal & Expression
Recognition of Emotion in Self
Nonverbal Emotional Expression
Recognition of Emotion in Others
Empathy
Regulation
Regulation of Emotion in Self
Regulation of Emotion in Others
Applications
Intuition vs. Reason
Creative Thinking
Mood Redirected Attention
Motivating Emotions
Average
Remaining EI
Alpha
(stepwise)
reliability
% reliable
variance
accounted for*
.25
.28
.08
.08
.28
.36
.25
.40
.41
.41
.44
.48
.82
.80
.83
.74
50.0
51.3
53.0
64.9
.19
.23
.47
.38
.47
.53
.85
.83
55.3
63.9
.00
.00
.05
.39
.08
.46
.19
.58
.19
.57
.35
.60
.79
.81
.76
.87
24.1
70.4
46.1
69.0
.16
.35
.45
.81
54.8
*100(R2/alpha)
Summary
With few exceptions, the goals of scale construction were met: (a) each item was
reliably judged by independent experts as representative of its targeted EI facet,
in support of the scales’ content validity, (b) each EI scale demonstrated adequate
internal consistency reliability, (c) each EI scale demonstrated adequate test-retest
reliability in keeping with the measurement of EI as a relatively stable dispositional
trait, (d) 9 of the 10 scales (all but MRA) controlled acquiescence response bias by
inclusion of equal numbers of true- and false-keyed items, (e) every item on each
scale correlated stronger with the remaining items on the targeted scale than with
total scores on any other scale, including Desirability, (f) EI scale scores correlated
meaningfully and, in the majority of cases, predictably with thematically linked
personality scale scores, in support of convergent validity, (g) EI scale scores, for the
most part, were uncorrelated with conceptually unrelated personality scale scores,
in support of discriminant validity, (h) EI scale scores showed low to moderate correlations with PRF Desirability in keeping with the conceptualization of EI as an
overall desirable quality but without undermining the interpretability of scale scores
based on self-report, and (i) each EI scale showed relatively little overlap with Desirability, mainstream personality traits, and the remaining EI facets, in support of EI
as a distinct trait domain composed of multiple, relatively distinct facets.
Minor refinements in the 10 primary EI scales would be undertaken in subsequent
studies to yield 12 items per scale, 6 keyed in each direction (except MRA, which
25
Chapter 2
contains 8 true-keyed items). In addition, efforts are underway to develop 2 new
scales, targeting Delay of Gratification and Emotional Appropriateness using the
same methods described above regarding development of the 10 primary scales.
26
Chapter 3
Validation of the MEIA
Test validation is the systematic assessment of how well scores on a test reflect true
differences among test takers on the characteristic the test is intended to measure.
Three primary classes of validity evidence are widely recognized as appropriate
in test validation. Content validity is the degree to which test items are judged by
experts as capturing a targeted construct. Criterion validity denotes the degree to
which test scores show meaningful relations with variables judged worthy of prediction (i.e., criteria). Construct validity encompasses all evidence bearing on whether
or not a test measures what it is intended to measure, including content-related and
criterion-related evidence. Two broad subclasses of construct validity evidence are
convergent validity: how well test scores correlate meaningfully (either positively
or negatively) with scores on other measures of the same or conceptually related
constructs; and discriminant validity: how well test scores are uncorrelated with
scores on measures of conceptually unrelated constructs. Evidence regarding the
validity of the MEIA is reviewed below within the noted categories.
Content Validity
As described in Chapter 2, the MEIA was developed from the earliest stages to
possess content validity. To begin, items were written by multiple psychological
experts with diverse backgrounds and interests to capture each of the 10 facets of
EI proposed by Salovey and Mayer (1990). Most test authors rely solely on such
construct-targeted item-writing as evidence for content validity. Indeed, this approach is generally recognized as offering an adequate basis for inferring content
validity (Anastasi & Urbina, 1997). In developing the MEIA, we took the added
step of confirming that each item written for a particular scale is judged reliably
by experts not contributing that item to be representative of the targeted construct
(i.e., 1 of the 10) and not any of the remaining 9 constructs, in light of the entire set
of construct definitions. It is notable that many items written for a given scale did
not pass independent confirmation in the second step (avg. = 60% dropped). The
confirmatory procedure employed here offers added assurance that the MEIA scales
are valid with respect to the interpretability of the items they contain.
Criterion Validity
Trait-EI measures developed by others have been linked to both self- and other-rated
outcomes. Self-report criteria have included biographical outcomes (e.g., drug use),
27
Chapter 3
chronic emotions (e.g., depression, anxiety), attitudes (e.g., life satisfaction), and
self-rated performance. Other-rated criteria have included actual or expected performance (e.g., GPA), experimental outcomes (i.e., positivity of generated stories),
and others’ attitudes. The MEIA has been linked to several self-report criteria, as
described below.
The MEIA was administered to 152 undergraduate students enrolled in a psychology course at a community college or university, both located in a southwestern
American city (mean age = 22.1 years, 66% female). Measures of life satisfaction
and cross-cultural adaptability were also administered. Life Satisfaction was assessed using Diener, Emmons, Larsen, and Griffin’s (1985) 5-item Satisfaction With
Life Scale (SWLS; e.g., “I am satisfied with my life”). We also used 2 subscales of
Roberts and Clifton’s (1992) Quality of Student Life Questionnaire, targeting quality
of relationships with other students (5 items) and with faculty (9 items). Sample
items, respectively, include “I get on well with the other students in my class(es)”
and “Professors listen to what I say.” Positive correlations with the EI subscales were
expected in all 3 cases. The 50-item Cross-Cultural Adaptability Inventory (Kelley
& Meyer, 1992) includes 4 subscales. Emotional Resilience assesses recovery from
negative emotions and positive reaction to new experiences. Perceptual Acuity assesses sensitivity to environmental information including others’ nonverbal cues.
Personal Autonomy assesses preservation of one’s own values while simultaneously
respecting others’ values. Flexibility-Openness assesses enjoyment in thinking and
behaving in new ways.
Correlations between the MEIA and the 8 self-report criteria are reported in Table
3.1. Criterion validity is evident by notable positive relations. Regarding satisfaction, MEIA MotEm correlates .47 with life satisfaction and .49 with satisfaction
with faculty; MEIA RegOth correlates .53 with satisfaction with other students. The
median r for all 30 validity coefficients involving satisfaction is .30. This compares
favorably to criterion validities for an adjective-based Big 5 personality measure
administered to the same sample, which had a median correlation of .25. Regarding
cross-cultural adaptability, 9 of the 10 MEIA scales (all but IvR) yielded 3 or more
significant positive relations, the strongest obtaining for MotEm (range = .36 to .57),
RecOth (.31 to .51), and RegOth (.31 to .45). These results suggest, in keeping with
scale definitions, that individuals more willing to adapt to other cultures tend to be
more optimistic, persevering, and other-oriented.
Incremental Validity
As an extension of criterion validity, it is relevant also to consider the degree to which
a scale contributes uniquely to explaining criterion variance, after controlling for the
contributions of more established measures. As reported in Chapter 2, regression
results showed that the 10 MEIA scales retained from 30% (CrTh) to 76% (IvR)
of their reliable variance after accounting for PRF Desirability, the 15 scales on the
JPI-R, and the remaining MEIA scales (see Table 2.6). These results suggest that the
28
Validation of the MEIA
Table 3.1: Correlations between MEIA subscales and distal self-report criteria: satisfaction and
adaptability (N = 152)
Satisfaction
MEIA scale
Life
Satisfaction
Satisfaction with
Other Students
Satisfaction
with Faculty
Core EI
Recognition of Emotion in the Self
Regulation of Emotion in the Self
Recognition of Emotion in Others
Regulation of Emotion in Others
Nonverbal Emotional Expression
Empathy
.39**
.31**
.38**
.39**
.40**
.14
.27**
.11
.41**
.53**
.41**
.27**
.29**
.20**
.27**
.42**
.35**
.26**
Proximal EI Outcomes
Intuition vs. Reason
Creative Thinking
Mood Redirected Attention
Motivating Emotions
-.08
-.05
.23**
.47**
.04
.13
.31**
.36**
.02
.04
.31**
.49**
Adaptability
MEIA scale
Socio-cultural Emotional
Adaptation
Resilience Flexibility
Perceptual
Acuity
Personal
Autonomy
Core EI
Recognition of Emotion in the Self
Regulation of Emotion in the Self
Recognition of Emotion in Others
Regulation of Emotion in Others
Nonverbal Emotional Expression
Empathy
.28**
.30**
.27**
.25**
.06
.27**
.40**
.42**
.43**
.39**
.11
.28**
.22**
.15
.31**
.31**
.39**
.31**
.30**
.16*
.51**
.45**
.38**
.38**
.39**
.26**
.31**
.38**
.27**
.25**
Proximal EI Outcomes
Intuition vs. Reason
Creative Thinking
Mood Redirected Attention
Motivating Emotions
-.13
.13
.16*
.39**
-.05
.29**
.26**
.57**
-.01
.20*
.31**
.36**
-.02
.25**
.35**
.36**
-.03
.12
.35**
.44**
*p < .05 two-tailed, **p < .01 two-tailed.
MEIA captures individual differences that are sufficiently independent of other, more
common sources to warrant separate consideration in studies of human behavior.
We assessed the MEIA’s incremental validity by adding the primary scales to
regressions of the satisfaction and cultural adaptability criteria, after accounting
for the contributions of a variety of personality and related measures (N = 152).
For each of the 7 criteria, 4 sets of variables were entered in consecutive steps: (1)
Crowne-Marlowe social desirability, (2) the Big 5 personality dimensions, (3) Positive Affect (PA), Negative Affect (NA), and self-monitoring, and (4) the 10 MEIA
primary scales (forced entry in step 1, stepwise in steps 2 to 4). (Measurement details
29
Chapter 3
are offered below in the Construct Validity section.) Incremental validity would be
evident with a significant increase in adjusted R2 at step 4.
Results, presented in Table 3.2, show that the MEIA scales contributed independently to all 7 criteria: from 4% (Satisfaction with Faculty) to 19% (Perceptual Acuity) of the total variance (mean = 12.3%), and from 11% (Satisfaction with Faculty)
to 46% (Personal Autonomy) of the explained variance (mean = 29.1%), based on
adjusted R2s. (The adjusted values account for upwardly biased estimates owing
to capitalization on chance.) All MEIA primary scales except RegOth contributed
uniquely to at least 1 criterion, Emp contributing in 5 cases, RecOth and MRA each
in 4 cases, and RegSlf in 2. Beta weights and partial correlations generally support
the expectation that EI contributes positively to satisfaction and cross-cultural adaptability. The only exception is that IvR contributes negatively to Perceptual Acuity:
acuity is lower in those relying on intuition over logic. We judge the acuity-logic
connection to be interpretable, in support of the criterion validity of IvR.
Construct Validity
Evidence bearing on the construct validity of the MEIA has accrued from several
sources, including the two preliminary investigations described in Chapter 2. Results
reported in Tables 2.1 and 2.2 (from Study 1) and Tables 2.4 and 2.5 (from Study
2) reveal interpretable and, in many cases, predictable linkages between the initial
MEIA scales and conceptually related personality measures on the JPI-R (Jackson,
1994), in support of convergent validity. Overall weaker relations between the MEIA
scales and conceptually dissimilar personality variables, shown in the same tables,
offer support for discriminant validity.
A third study, noted above in the criterion validity section (N = 152), also yielded
evidence bearing on the MEIA’s construct validity. Instead of the JPI-R, as used in
the earlier studies, Saucier’s (1994) mini-markers were administered to assess the
Big Five personality dimensions: Agreeableness, Conscientiousness, Extroversion,
Emotional Stability, and Openness to Experience. Positive Affect (PA) and Negative
Affect (NA) were measured as traits (not states) using the 20-item Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988). PA items include
“interested,” “excited,” and “enthusiastic,” and NA items include “jittery,” “hostile,”
and “irritable.” Self-monitoring was assessed using Gangestad and Snyder’s (1985)
true/false measure, containing 9 true-keyed items (e.g., “I would probably make a
good actor”) and 9 false-keyed items (e.g., “I feel a bit awkward in public and do not
show up quite as well as I should”). Finally, social desirability was assessed using
Crowne and Marlowe’s (1964) 33-item scale. One of the 18 true-keyed items is “I
always practice what I preach.”
Correlations between the MEIA scales and the 9 other measures are reported in
Table 3.3. Convergent validity is largely supported in relations involving personality, affect, and self-monitoring. For example, MEIA RegSlf correlates .61 with
Emotional Stability and -.60 with NA, MEIA Emp correlates .57 with Agreeable-
30
Validation of the MEIA
Table 3.2: Regression results showing incremental validity of MEIA primary scales in predicting
satisfaction and cross-cultural adaptability (N = 152)
Step Dependent variable/unique predictor
Life satisfaction
Satisfaction with other students
1 a
2b
3b
4b
Social Desirability
Extraversion
Agreeableness
Positive Affect
Recognition of Emotion in Others
Empathy
Satisfaction with faculty
1 a
2b
3b
4b
Social Desirability
Extraversion
Conscientiousness
Positive Affect
Empathy
Emotional resilience
1 a
2b
3b
4b
Social Desirability
Extraversion
Emotional Stability
Openness
Positive Affect
Negative Affect
Motivating Emotions
Recognition of Emotions in Others
Creative Thinking
Mood Redirected Attention
Flexibility
1 a
2b
4b
Social Desirability
Openness
Agreeableness
Extraversion
Emotional Stability
Empathy
Mood Redirected Attention
1 a
2b
3b
4b
Social Desirability
Conscientiousness
Extraversion
Positive Affect
Nonverbal Emotional Expression
Recognition of Emotion in Others
Regulation of Emotion in the Self
R
R2
Adj R2
R2c
Betad
r
Partial rd
.34
.47
.50
.56
.61
.64
.66
.12
.22
.25
.32
.37
.41
.44
.11
.21
.23
.30
.35
.39
.41
.11**
.10**
.02*
.07**
.05**
.04**
.02**
.06
.16*
-.13
.26**
.26**
.20**
.18**
.34
.44
.16
.49
.40
.38
.32
.07
.18
-.14
.25
.29
.24
.21
.19
.47
.51
.55
.60
.63
.04
.22
.26
.30
.36
.40
.03
.21
.25
.28
.34
.38
.03*
.18**
.04**
.03**
.06**
.04**
.12
.30**
.02
.16*
.26**
.23**
.19
.42
.25
.44
.44
.27
.13
.32
.02
.17
.29
.23
.31
.44
.51
.57
.61
.10
.19
.26
.33
.37
.09
.18
.24
.31
.35
.09**
.09**
.06**
.07**
.04**
.15*
.18*
.13
.33**
.22**
.31
.29
.38
.52
.27
.16
.19
.13
.30
.25
.30
.47
.55
.57
.61
.64
.68
.71
.73
.74
.09
.22
.30
.33
.37
.41
.46
.50
.53
.55
.08
.21
.28
.31
.35
.38
.44
.47
.50
.52
.08**
.13**
.07**
.03*
.04**
.03**
.06**
.03**
.03**
.02**
.09
.09
.17*
-.09
.01
-.17*
.30**
.21**
.23**
.17**
.30
.34
.39
.29
.48
-.43
.57
.43
.29
.26
.12
.11
.17
-.10
.01
-.17
.30
.26
.25
.22
.33
.48
.55
.57
.59
.66
.68
.11
.23
.30
.33
.35
.43
.47
.11
.22
.28
.31
.32
.41
.44
.11**
.11**
.06**
.03*
.01*
.09**
.03**
.13
.19**
.03
.18**
.32**
.35**
.21**
.33
.33
.43
.25
.32
.39
.31
.15
.23
.03
.21
.33
.34
.25
continued...
31
Chapter 3
Table 3.2 (cont.)
Step Dependent variable/unique predictor
Perceptual Acuity
1 a
2b
4b
Social Desirability
Openness
Agreeableness
Extraversion
Recognition of Emotion in Others
Empathy
Mood Redirected Attention
Intuition vs. Reason
Personal autonomy
1 a
2b
3b
4b
Social Desirability
Openness
Conscientiousness
Positive Affect
Mood Redirected Attention
Recognition of Emotion in the Self
Empathy
Regulation of Emotion in the Self
R
R2
Adj R2
R2c
Betad
r
Partial rd
.20
.45
.52
.55
.64
.68
.69
.71
.04
.20
.27
.30
.41
.46
.48
.51
.03
.19
.26
.29
.39
.43
.46
.48
.03**
.16**
.07**
.03**
.10**
.04**
.03**
.02**
.09
.14*
.14
.12
.34**
.27**
.19**
-.16**
.20
.39
.38
.28
.52
.38
.35
-.02
.10
.16
.14
.15
.39
.28
.24
-.21
.23
.39
.43
.46
.52
.58
.60
.62
.05
.15
.18
.21
.27
.33
.36
.38
.05
.14
.17
.19
.25
.31
.33
.35
.05**
.09**
.03*
.02*
.06**
.06**
.02*
.02*
.05
.18**
.09
.05
.25**
.19**
.21**
.19*
.23
.30
.30
.36
.35
.39
.27
.26
.06
.21
.09
.05
.28
.20
.24
.19
forced entry; bstepwise entry; cbased on adjusted R2 values; dfinal model
*p < .05, **p < .01, two-tailed.
a
ness, MEIA RegOth correlates .51 with Extroversion and .50 with PA, MEIA CrTh
correlates .66 with Openness, and MEIA MotEm correlates .66 with Conscientiousness and .64 with PA. These relations, the strongest in the set, are consistent with
construct definitions. Linkages in other cases are also informative. People higher
on Conscientiousness, for example, recognize and regulate both their own and others’ emotions (.20 to .29), favor logic over intuition (-.24), and are lower on creative
thinking (-.22). Consistent with prior research (Fox & Spector, 2000; Palmer et al.,
2002), the MEIA scales tend to correlate positively with PA and negatively with
NA (median rs = .32 and -.12, respectively). An exception is IvR, which correlated
-.02 with PA and .23 with NA: those relying on intuition in decision making tend to
experience more negative emotions. Finally, self-monitoring correlates with CrTh
(.41), RegOth (.36), RecOth (.31) and Emp (-.28), a pattern that fits self-monitoring
as “the extent to which individuals strategically cultivate public appearances” (Gangestad & Snyder, 2000; p. 530).
Further evidence bearing on the construct validity of the MEIA derives from
correlations with the Hogan Personality Inventory (HPI; Hogan & Hogan, 1995),
based on data from the normative sample. After dropping cases with high scores on
MEIA Inf (>2.5) or low scores on the HPI Validity scale (<10), final N = 119. The
correlations, reported in Table 3.4, are largely interpretable. For example, MEIA
RegSlf correlates strongest with HPI Adjustment (.70), MEIA MotEm correlates
strongest with HPI Ambition (.52), and MEIA CrTh correlates strongest with HPI
32
Validation of the MEIA
Intellectance (.59). The pattern of relations is consistent, on the whole, with those
reported earlier involving other personality measures (Tables 2.1 and 2.4 for the
JPI-R; Table 3.3 for Saucier’s Big Five mini-markers). Correlations with the HPI
occupational scales, shown in the right half of Table 3.4, are further illuminating.
MEIA RegSlf, for instance, shows a dominant correlation with HPI Stress Tolerance
(.66) and moderate relations with HPI Service Orientation (.45), Clerical (.46), Reliability (.32), and Managerial (.35). HPI Managerial also correlates meaningfully with
MEIA MotEm (.67), RecSlf (.37), RegOth (.37), NvExp (.26), MRA (.21), and CrTh
(.18). All told, correlations with the HPI support the construct validity of the MEIA.
Social Desirability
Social desirability offers a special case of discriminant validity, in that self-report
measures should be expected to yield relatively weak relations with desirability
scales (Helmes, 2000). Previous studies have shown that self-report trait-EI may
be vulnerable to socially desirable responding (e.g., Bar-On, Brown, Kircaldy, &
Thorne, 2000; Hemmati, Mills, & Kroner, 2004). In developing the MEIA, we addressed this issue by having item writers avoid desirable and undesirable language,
eliminating items correlating more strongly with a desirability scale than with the
remaining items on their targeted scales, and controlling for desirability in relations
involving the proposed scales. Our findings suggest that (a) desirability accounts
for relatively modest proportions of reliable scale variance (mean = 20% in Study
2 using PRF-Des, mean = 6% in Study 3 using the Marlowe-Crowne), (b) MEIA
facets differ in their susceptibility to desirable responding (IvR: .02, MotEm: .62 in
Study 2; MRA: .00, RegSlf: .38 in Study 3), and (c) relations among EI facets and
between the facets and other variables showed similar overall patterns after controlling for desirability (all 3 studies).
We suggest that, as a generally desirable trait, EI elicits higher self-descriptions
in those who see themselves in a desirable light. Self-reported EI need not pose
serious threats to validity, however, when efforts are made to limit desirability saturation in scale construction, as was the case here. The degree to which self-report EI
measures are vulnerable to deliberate faking (e.g., as in personnel selection settings)
is a matter for further research.
Structural Validity
Construct validity bears further consideration in terms of how well conceptually
distinct subscales of a test interrelated to form more general clusters. Results of
studies designed to investigate the structure of EI shed light on the structural validity of the MEIA. The MEIA’s structure has been assessed in four ways, as follows.
PCA of item bundles. Total scores on item subsets within each of the 10 primary
scales (i.e., 3 “item bundles” per scale) were subjected to a clustering technique
called principal components analysis (PCA). The MEIA’s structure would be sup-
33
34
-.24**
-.21**
.09
.67**
.18*
-.05
.12
.15
Adjust-
ment
.06
-.12
.12
.27**
.05
-.14
.12
.12
.13
.02
.18*
.30**
.20*
.52**
Soci-
ability
.30**
.30**
.23**
.36
.21*
-.02
Ambi-
tion
.14
.22*
.13
.22*
.22*
.26**
.20*
.42**
.28**
.34**
Like
ability
-.17
.07
-.16
.26**
.24
.62
.04
.07
.02
-.09
-.23**
.00
.05
.33**
.21*
.20*
.14
.21*
.32**
.47**
.11
.59**
.13
.21*
.09
.23**
.18*
.23**
.03
.21*
-.07
.35**
.15
.28**
.27**
.31**
.13
.19*
.12
.15
Pru- Intellec- School
dence tance Success
HPI Primary Scales
Cases with low HPI Validity (<10) or high MEIA Inf (>2.5) dropped.
Core EI
RecSlf
.38**
RegSlf
.70**
RecOth
.07
RegOth
.04
NvExp
.14
Emp
.07
Proximal Outcomes
IvR
-.05
CrTh
.25**
MRA
-.15
MotEm
.34**
MEIA Scales
-.06
.35**
.22**
.28**
.22**
.07
.31**
.50**
.33**
-.13
.05
.66**
.20**
.20**
.17*
.00
.44**
.47**
.05
.03
Openness
to Expression
-.02
.12
.33**
.62**
.33**
.22**
.35**
.50**
.31**
.07
.24**
.02
.10
-.35**
-.29**
-.59**
-.16*
-.17*
-.07
.12
.07
.26**
-.06
.20*
.17
.45**
.15
.25**
.21*
.25**
Service
Orient’n
-.06
.25**
-.13
.37**
.39**
.66**
.11
.02
.09
-.07
-.14
.01
.00
.23**
.25**
.32**
.03
.08
.28**
.39**
Stress Reli-
To.
ability
General Occ. Scales
-.08
-.14
.02
.38**
.15
.39**
-.03
.18*
.28**
.22**
.12
.21*
.09
.45**
.36**
.46**
.18*
.31**
.22*
-.01
.24**
.39**
.19*
.23**
.18*
.14
.20*
.37**
.15
-.03
.04
.18*
.21*
.67**
.37**
.35**
.17
.37**
.26**
.17
ManaClerical Sales gerial
Specific Occ. Scales
.01
.41**
.04
.02
.11
.03
.32**
.35**
.00
-.28**
Positive Negative
Self
Social
Affectivity Affectivity Monitoring Desirability
Affectivity
Table 3.4: Correlations between MEIA subscales and the Hogan Personality Inventory (N = 119)
*p < .05 two-tailed, **p < .01 two-tailed.
.30**
.23**
.23**
.23**
.15
.14
Emotional
Stability
Big Five Personality
Conscien-
tiousness Extroversion
.12
.10
.12
.24**
.40**
.57**
Agreeable-
ness
MEIA Scales
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
Proximal EI Outcomes
IvR
CrTh
MRA
MotEm
Table 3.3: Correlations between MEIA subscales and the Big Five (mini-markers), affectivity, self-monitoring, and Crowne-Marlowe
social desirability (N = 152)
Chapter 3
Validation of the MEIA
ported to the degree that the 3 “bundles” from each scale combine to form 10 distinct
clusters or components, one per MEIA facet. Each of the 10 MEIA primary scales
was divided randomly into three sets (“bundles”) of four items, two keyed in each
direction. MRA, consisting of just 8 true-keyed items, was divided into two sets of
four items. Total scores on the resulting 29 item bundles were entered into a PCA
using the Study 3 data (N = 152). The components were extracted orthogonally (i.e.,
independently of each other) and rotated orthogonally to facilitate interpretations.
The results are presented in Table 3.5. As shown, item bundles from the same MEIA
scale load dominantly and exclusively on (i.e., contribute to) a single component.
Specifically, item bundles from each subscale load highly (.64 to .88, mean = .80)
and uniquely (60% of cross-loadings <.10; max = .30) on 1 of the 10 independent
components. This strongly supports the intended dimensionality of the EI items as
representing 10 distinct EI facets and confirms the internal consistency reliability
of each scale.
PCA of primary scale total scores. In the second analysis of structural validity,
total scores on the 10 MEIA primary scales (i.e., based on all items per scale, not
item bundles) were subjected to PCA. Validity would be supported to the degree
that the resulting clusters bear unique interpretations, that is, that they “make sense”
conceptually at a general level. Using the Study 3 data (N = 152), three components
were extracted and rotated to maintain their independence, facilitating interpretations. Results are shown in Table 3.6. The three components accounted for 60% of
the total variance and 75% of the reliable variance (based on alphas).
Variables loading dominantly on Factor I include MotEm, RecSlf, RegSlf, and
IvR (negative). These EI facets share a focus on the self, suggesting the label Self
Orientation for Factor I. Facets loading strongest on Factor II are NvExp, Emp, and
MRA. Together, the first 2 constructs suggest exchange of emotional information
between the self and others. MRA may contribute to this exchange by its positive
valuation of emotional experience. We labeled Factor II as Emotional Sharing. Factor III shows prominent loadings for CrTh and both RecOth and RegOth. Creative
thinking denotes distinctiveness of one’s own ideas relative to those of others. As
a set, these 3 facets share an outward focus, which we labeled Other Orientation.
The self/other distinction, reflected in Factors I and III, is similar to that reported
elsewhere on EI (e.g., the EQi includes Intrapersonal and Interpersonal subscales;
the WEIP separates Dealing With Own Emotions from Dealing With Others’ Emotions). Salovey and Mayer’s (1990) model posits primary separation of appraisal
and regulation of emotion and secondary separation of self and other within each
of those broader categories. Our results suggest the reverse hierarchical order: self/
other as the primary distinction, with recognition/regulation embedded within each
of self and other. Emotional Sharing appears to be a uniquely emergent construct.
Other self-report EI measures (SREI, WEIP, TEIQue) include empathy as a facet
of EI, but the configuration of empathy with both nonverbal emotional expression
and willingness to grow from one’s emotional experiences (i.e., MRA) has not been
identified in prior research.
35
Chapter 3
Table 3.5: Rotated loadings and eigenvalues of 10 orthogonal components derived from 29 MEIA item
bundles (N = 152)
Parcel
Component
1
2
3
4
5
6
7
8
9
10
RecOth_3
RecOth_1
RecOth_2
.85
.82
.81
.09
.04
-.06
.10
.15
.05
.13
.00
-.04
.10
.03
.23
.11
.16
.09
.07
-.14
.11
.18
.10
-.05
.15
.23
.19
.14
.03
-.05
RegSlf_3
RegSlf_1
RegSlf_2
.03
.02
.01
.88
.87
.77
.14
.08
.28
-.10
-.10
-.02
.10
.17
.26
.12
-.01
-.06
-.08
-.18
.01
-.07
-.05
.07
.11
.01
-.01
-.03
.06
-.19
MotEm_1
MotEm_3
MotEm_2
-.01
.19
.18
.20
.21
.07
.83
.82
.77
.00
.03
.06
.14
.12
.30
.10
.01
-.07
-.11
-.09
-.12
.02
.14
.07
.07
.17
.12
.11
.05
.12
Emp_1
Emp_3
Emp_2
.03
.00
.04
-.11
-.02
-.08
.05
.02
.00
.86
.84
.81
.01
.04
.00
.00
-.11
.04
.15
-.01
.17
.05
.17
.17
.01
.01
.19
.01
.09
.06
RecSlf_3
RecSlf_1
RecSlf_2
.09
.18
.11
.22
.24
.09
.25
.15
.13
.03
.05
-.02
.82
.78
.74
-.01
.10
-.01
-.08
-.17
.03
.03
.08
.15
.07
-.01
.23
-.03
.19
-.06
CrtTh_1
CrtTh_2
CrtTh_3
.10
.07
.11
-.01
-.02
.08
.03
.06
-.06
-.05
-.05
.03
.09
.03
-.08
.86
.81
.76
.05
.11
.08
-.03
-.06
.01
.10
.10
.07
-.12
.03
.15
IvR_2
IvR_3
IvR_1
-.12
.14
.06
-.05
.00
-.21
-.03
-.14
-.14
-.01
.25
.14
-.14
-.12
.06
.08
.17
.05
.82
.77
.74
-.02
.24
-.02
.16
-.06
-.05
.07
-.02
.13
NvExp_1
NvExp_2
NvExp_3
.00
.25
-.02
-.09
-.07
.15
.03
.10
.14
.12
.18
.23
.16
-.06
.27
-.03
.04
-.17
.18
-.06
.01
.84
.79
.64
.05
.09
.30
.12
.13
-.02
RegOth_2
RegOth_3
RegOth_1
.29
.30
.20
-.04
.17
.04
.23
.18
.03
.07
-.01
.23
.22
.10
.05
-.04
.31
.27
.08
-.07
.07
.13
.05
.22
.76
.70
.70
.03
.10
.17
MRA_2
MRA_1
.09
.01
.03
-.14
.14
.09
.08
.08
.11
-.07
.04
.03
.03
.15
.03
.19
.19
.01
.87
.83
2.51
2.50
2.41
2.41
2.35
2.31
2.10
2.07
2.02
1.71
post-rotation
eigenvalues
Confirmatory SEM of 6 core MEIA scales. Building on the latter PCA results,
the structure of just the 6 core EI facets was assessed using confirmatory structural
equation modeling (SEM) applied to new data. This approach allows direct comparisons among competing structures based on unique a priori conceptualizations,
and entails assessing how well each alternative structure fits observed correlations
among actual scale scores. As depicted in Figure 1.2, Salovey and Mayer’s (1990)
model of EI posits two core EI factors, one, Regulation of Emotion, consisting of
Regulation in the Self and in Others, and the other, Appraisal & Expression of Emotion, comprising the remaining 4 core dimensions. This configuration was compared
36
Validation of the MEIA
Table 3.6: Rotated loadings, eigenvalues*, and communalities** (h2) of 10 MEIA subscales (N = 152)
Component / MEIA subscale
I
II
III
h2
I Self Orientation
Motivating Emotions
Recognition of Emotion in the Self
Regulation of Emotion in the Self
Intuition versus Reason
.74
.74
.72
-.57
.24
.20
-.22
.38
.16
.18
.07
.31
.64
.62
.56
.57
II Emotional sharing
Nonverbal Emotional Expression
Empathy
Mood Redirected Attention
.23
-.15
.00
.77
.75
.54
.01
-.03
.20
.64
.58
.34
III Other Orientation
Creative Thinking
Recognition of Emotion in Others
Regulation of Emotion in Others
-.12
.30
.34
-.16
.24
.41
.84
.64
.64
.75
.56
.69
post-rotation eigenvalues
post-rotation variance explained
2.23
22
1.98
20
1.73
17
5.94
59
*An eigenvalue is the number of variables (of the 10 MEIA scales in this case) that a given component represents. PCA is
useful, in general, to the degree that each component captures more variance than that contributed by 1 variable (i.e., has
an eigenvalue > 1). **A communality (h2) is the proportion of variance in the given MEIA scale that is accounted for by all
three components combined. Variables with low communalities share less in common with the remaining variables and,
accordingly, are more unique.
to two alternative configurations, one a second 2-factor structure, and the other, a
unique 3-factor structure. Both alternatives are described below.
In addition to Salovey and Mayer’s 2-factor configuration, depicted in Figure 1.2,
the 6 core EI facets permit conceptualization under 2 other broader constructs, each
with 3 facets. Combining RecSlf, RegSlf, and NvExp suggests Internal Orientation.
Inclusion of the first 2 aspects is intuitively obvious. The third (NvExp) is internally focused, as judging one’s emotional expressivity assumes awareness of one’s
inner states. Combining RecOth, RegOth, and Emp suggests External Orientation.
Similar to Internal Orientation, the first 2 dimensions are clearly relevant. Empathy
is included here for its focus on the emotions of others (Jordan, Ashkanasy, Hartel,
& Hooper, 2002).
The 6 core EI facets also permit a 3-factor configuration, the first 2 factors corresponding to the 2 general dimensions in the second 2-factor model. Specifically,
Self Orientation contains RecSlf and RegSlf, and Other Orientation contains RecOth
and RegOth. The 2 remaining core facets, NvExp and Emp, can be combined to
form a third factor, Emotional Sharing. The reason is that both NvExp and Emp
denote a willingness to communicate with others in emotional terms, Emp denoting
receptiveness to how others are feeling, and NvExp, openness to sharing one’s feelings with others. This 3-factor structure is similar to that which emerged from the
PCA reported in the preceding section. The difference is that all 10 MEIA scales
37
Chapter 3
were involved in the PCA, whereas only the 6 core MEIA scales are involved in the
current confirmatory analysis.
In all three multi-factor EI models offered above, the factors were presumed to
be correlated and, hence, were allowed to covary in model testing. A fourth model,
positing a single general factor, was included to assess the possibility that core EI
is unidimensional. Lack of support for this model would encourage assessment of
the three alternative, multi-factor models. The 4 targeted models are depicted in
Figure 3.1.
The MEIA was administered to 196 incoming freshmen from a private American
university, the large majority (95%) ranging in age from 18 to 20 years, and 54%
female. All initial cases were those reporting complete data. Removal of multivariate
outliers, critical in undertaking confirmatory structural analyses, yielded N = 184.
Figure 3.1: One-, two-, and three-factor models based on Salovey & Mayer’s (1990) core EI dimensions
38
Validation of the MEIA
Results of model testing (AMOS 4.0) are reported in Table 3.7. Improvements
in fit are evident, moving from single-factor to multi-factor solutions. The 3-factor
solution yielded inadmissible parameter estimates. Modification indices revealed
correlated error terms for NvExp and RecSlf. Ullman, (2001), Millsap (2002), and
Byrne (1998) argue that such cases are admissible in SEM when supported on conceptual grounds. The noted correlation is consistent with the Internal Orientation
factor in our 2-factor model. The modified model showed good overall fit. Final
path coefficients are presented in Figure 3.2.
Table 3.7: Results of model testing (N = 184)
Model
df
Chi sq.
GFI
AGFI
TLI
NFI
RMR
1-factor
9
78.50
.88
.72
.58
S&M 2-factor
8
78.44
.88
.69
.52
2-factor
8
63.07
.90
.73
.62
3-factorinadmissable parameter estimates
modified 3-factora 5
9.99
.98
.93
.95
Chi sq.
diff.
df
.73
.73
.78
.047
.045
.048
.06
15.43*
1
1
.97
.017
68.51*
4
*includes correlated error terms for RecSlf & NvExp
GFI = Goodness of Fit Index
AGFI = Adjusted Goodness of Fit Index
TLI = Tucker-Lewis Index
NFI = Normed Fit Index
RMR = Root Mean Residual
GFI to NFI: values range from .00 (poor fit) to 1.00 (perfect fit). Values above .90 are generally considered to indicate acceptable fit.
RMR: .00 = no erro and perfect fit; higher values indicate poorer fit. Values below .05 are generally considered to indicate
acceptable fit.
Salovey and Mayer’s (1990) primary aim was to articulate the major components
and utilizations of EI; how the various facets might combine to form more general
composites was of lesser concern. Perhaps not surprisingly, then, current results failed
to support their 2-factor configuration of the 6 core EI dimensions. Consistent with
previous research (e.g., Jordan et al., 2002; Petrides & Furnham, 2000), results also
did not support a unidimensional interpretation of EI. Rather, a complex structure
of 3 correlated factors was identified.
Self Orientation, comprising RecSlf and RegSlf, is that part of EI that is especially
inwardly focused. Those high on this composite describe themselves as in touch
with and in control of their own emotions. Other Orientation, including RecOth
and RegOth, is the outwardly focused counterpart to Self Orientation, high scorers
describing themselves as both perceptive of others’ feelings and willing to alter
them. Emotional Sharing captures a uniquely communal aspect of EI, denoting the
dual tendency to be receptive to how others are feeling (Emp) and to be authentic
in expressing one’s own emotions (NvExp).
39
Chapter 3
Figure 3.2: Final path coefficients for the modified 3-factor model of emotional intelligence based on
student sample (N = 184).
The self-other distinction is well recognized in the emotion and personality literatures and related domains (e.g., social cognition, motivation, leadership). Indeed,
Salovey and Mayer (1990) clearly articulate between self and other within both
recognition and regulation of emotion. Wong and Law (2002) found support for the
distinction with respect to emotional appraisal, but not emotional regulation. Current
results support the distinction in both respects. In fact, consistent with Jordan et al.
(2002), our results suggest that the self-other distinction is more prominent than the
appraisal-regulation distinction, as factors were derived here for self versus other,
each including both appraisal (i.e., recognition) and regulation facets. Accordingly,
EI models proposing primary separation of appraisal and regulation of emotion (e.g.,
Mayer & Salovey, 1997; Salovey & Mayer, 1990) warrant reconfiguration in terms
of the primacy of the self-other distinction.
Unique to the current study was identification of Emotional Sharing as a core
trait-EI component, consisting of NvExp and Emp. We had proposed this possibility
based on understanding these specific traits as similarly communicative in nature.
This feature of EI has been discussed by many; yet, to our knowledge, current findings are the first to suggest Emotional Sharing as a distinct EI component.
It is noteworthy that the factors/composites are moderately positively interrelated
(mean = .32). Whether use of an overall EI score or a more articulated set of facet
scores is warranted is both a practical and conceptual issue. Dealing with a single,
overall score is simpler, but it implies that facets can compensate for one another in
determining the overall score. Thus, 2 individuals might score moderately high on
40
Validation of the MEIA
a general EI index despite having markedly different facet scores. Such possibilities
would undermine efforts to understand why EI is related (or unrelated) to relevant
outcomes. We recommend, accordingly, that researchers assess EI as a multidimensional construct and that use of a single overall score be evaluated against the
explanatory benefits of more specific measures.
Relations between core MEIA and proximal outcome scales. The fourth assessment of the MEIA’s structural validity entailed correlating the three core EI composite
scores, derived from the previous analysis (i.e., Self Orientation, Other Orientation, Emotional Sharing), with the four proximal outcomes (i.e., IvR, CrTh, MRA,
MotEm). As in the PCA applied to total scores on all 10 primary scales, relations
between the core scales and proximal outcomes should “make sense” conceptually,
in light of the constructs those scales were designed to assess. Results are reported
in Table 3.8.
Generally, relations are positive, in keeping with an overall unified conceptualization of EI. Not surprisingly, MotEm shows the strongest correlation with Self
Orientation (.55), and intuitive thinking (IvR) is linked strongest to Emotional Sharing (.27; i.e., relying on intuition and sharing emotions each suggest emotionality).
Notably, being intuitive is associated with lower self-regulation and self-recognition
of emotion (-.16). This suggests compensatory value in EI facets; e.g., reliance
on intuition may be desirable but at the expense of lowered self orientation. Such
complexities undermine popular simplistic descriptions of EI (e.g., Goleman, 1998)
and support the use of multi-faceted EI measures like the MEIA.
Summary of structural validity. Results of the 4 analyses described above offer
overall support for the structural validity of the MEIA. Specifically, (a) PCA of 29
item bundles (2 for MRA, 3 per remaining primary scale) yielded a separate, welldefined component for each of the 10 MEIA primary scales consistent with scale
development objectives; (b) PCA of the 10 intact primary scales yielded 3 interpretable components, Self Orientation, Other Orientation, and Emotional Sharing,
replicating and extending findings from previous EI research; (c) confirmatory SEM
Table 3.8: Correlations between MEIA core composites and proximal outcomes (N = 184)
MEIA core composites
Self Orientation
Other Orientation
Emotional Sharing
MEIA Proximal Outcomes
IvR
CrTh
MRA
MotEm
-.16*
.13
.27*
.21*
.35*
.22*
.09
.32*
.36*
.55*
.45*
.30*
*p < .05
Self Orientation = (RecSlf + RegSlf) / 2
Other Orientation = (RecOth + RegOth) / 2
Emotional Sharing = (NvExp + Emp) / 2
IvR = Intuition vs. Reason
CrTh = Creative Thinking
MRA = Mood Redirected Attention
MotEm = Motivating Emotions
41
Chapter 3
provided independent support (i.e., based on new data) for the 3 noted components,
based on just the 6 core MEIA scales rather than all 10 primary scales; and (d) correlations between the core EI composites and the 4 proximal outcomes yielded an
overall interpretable pattern consistent with scale definitions.
Summary
Analyses conducted to date offer substantial support for the validity of the MEIA
as a self-report measure of emotional intelligence conceived as, in part, a multidimensional disposition. Specifically, (a) in support of content validity, the MEIA’s
items are reliably interpretable as representative of EI facets proposed by Salovey
and Mayer (1990); (b) in support of criterion validity, the MEIA scales have been
found to correlate meaningfully with a number of self-report criteria, including life
satisfaction and cultural adaptability; and (c) in support of construct validity, studies have yielded largely interpretable patterns of relations between the MEIA and
assorted other trait measures, including different measures of social desirability, and
among the MEIA scales themselves. Validation is an ongoing process. Currently
available results suggest that the MEIA may prove useful in further studies of emotional intelligence toward clarifying its dispositional nature, scope, and measurement.
42
Chapter 4
Multidimensional Emotional Intelligence
Assessment - Workplace (MEIA-W)
The relevance of emotional intelligence in work settings has been frequently discussed
(e.g., Abraham, 1999; Ashkanasy & Daus, 2002; George, 2000), particularly with
respect to the prediction of job performance and corresponding use in personnel
selection. A workplace version of the MEIA has been created for such purposes.
The MEIA-W is a close duplicate of the original MEIA, targeting the same 10 dimensions proposed by Salovey and Mayer (1990) and including the same number
of items per scale with half keyed in each direction (except MRA, which, as in the
MEIA, includes 8 true-keyed items). The MEIA-W items are simple work-related
versions of the original items. For example, the MEIA RegSlf item, “I can keep
myself calm even in highly stressful situations” was converted to “I can keep myself
calm even in highly stressful work situations.” Similarly, the word “people” in some
MEIA items was replaced by “co-workers” for the MEIA-W. Use of such “at work”
items has been reported to improve validity in personality tests over more general
versions administered in work settings (Bing, Whanger, Davison, & VanHook, 2004;
Robie, Schmit, Ryan, & Zickar, 2000; Schmit, Ryan, Stierwalt, & Powell, 1995).
The validity of the MEIA-W and its relations with job performance is the focus of
ongoing research. Below, we offer preliminary norms for the MEIA-W, based on
responses from over 600 working adults, as well as results of initial investigations
into the psychometric properties of the MEIA-W, including its reliability and validity. We begin with a brief review of the academic literature bearing on the role of
EI in work settings.
Emotional Intelligence at Work
How EI might contribute to job performance and to understanding workplace behavior is a matter of increasing interest. George (2000) offered several bases for
EI-performance relationships in leadership. For example, empathy may dispose an
individual to offer greater support to co-workers, thereby contributing to productive
interpersonal relationships. People with positive moods are likely to evaluate their
opportunities more favorably and, accordingly, consider a broader range of options
when making decisions. Being in touch with one’s emotions can provide internal
signals guiding selection of appropriate courses of action and facilitate anticipation
of reactions in the self and others. George identified 5 more specific work-related
43
Chapter 4
outcomes of EI: (1) EI can help one to be goal-focused and create a common vision
of important work objectives and how to achieve them. (2) EI can contribute to
instilling in others an appreciation for why valued objectives are important. (3) EI
can be expressed in creating and maintaining enthusiasm and optimism in co-workers
regarding goal pursuit, resulting in cooperation and trust among organizational members. (4) EI allows greater flexibility in decision making and dealing with change.
Finally, (5) EI can assist in establishing a meaningful identity for an organization
by management of people’s emotional reactions to the values and norms that make
up an organization’s culture.
Ashkanasy and Daus (2002) offer several related mechanisms though which EI may
be linked to job performance. For example, individuals high in EI are more likely
to inspire and motivate others in team efforts, offer assistance to co-workers when
they need it, communicate more effectively through self-expression and recognition
of other’s emotions, and manage others’ feelings toward increasing job satisfaction
and commitment. Similarly, Abraham (1999) argues that EI can help one adapt in
dealing with conflict and stress, manage customers and co-workers, and encourage
organizational citizenship and helping behavior.
All told, EI is expected to contribute to a wide range of work outcomes relating to
the appraisal, management, and utilization of emotions in the self and others. The
MEIA-W promises to facilitate corresponding research in applied settings in terms
of Salovey and Mayer’s multidimensional conceptualization of EI. In the following
sections, preliminary norms are offered followed by empirical evidence bearing on
key psychometric properties of the MEIA-W.
Norms
Preliminary norms for the MEIA-W are available based on a total sample of 653
working adults. Of the 459 reporting age, the mean is 35.8 (s = 9.6). The large
majority of the entire sample (83%) are from a publishing company in Tulsa, OK,
and the remainder are from an internationally diverse group of MBA students in
Australia (12%), and blue collar workers at a Tulsa manufacturing plant (5%). Despite
heavy reliance on a single organization, the total sample offers sufficient numbers
of workers in each of several job categories to permit separate norms. Of the 549
participants reporting job titles, 86% are classifiable as holding white collar positions,
42% are white-collar supervisors, 44% are white-collar non-supervisors, 14% are in
sales, and 10% are in customer service. (Note that some jobs are classifiable under
multiple categories). Norms for all these groups are presented in Table 4.1. The
means vary little across jobs. Exceptions include slightly higher means on RegSlf,
RecOth, RegOth, CrTh, and MotEm, and a slightly lower mean on Emp in sales,
and slightly higher means on RegOth and MotEm in customer service.
Table 4.2 presents means and standard deviations separately for men and women.
Means are significantly different in several cases, men scoring higher than women
on RegSlf and RegOth, and lower than women on NvExp and Emp. The observed
44
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
Table 4.1: Norms for the Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
in the total sample and several subgroups
All
N = 653
White Collar
N = 471
(86%)
WC Supervisory WC Non-Sup.
N = 232
N = 239
(42%)
(44%)
Sales
N = 76
(14%)
Cust. Service
N = 54
(10%)
MEIA-W scale
Mean SD
Mean SD
Mean SD
Mean SD
Mean SD
Mean SD
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
4.81
4.86
4.50
4.51
4.24
3.90
.65
.80
.62
.63
.52
.58
4.90
4.96
4.53
4.53
4.27
3.89
.60
.74
.61
.64
.51
.59
4.89
4.93
4.55
4.59
4.29
3.92
.60
.76
.59
.62
.52
.57
4.90
4.98
4.51
4.47
4.25
3.85
.60
.73
.63
.66
.50
.61
5.03
5.16
4.75
4.73
4.36
3.68
.64
.65
.55
.54
.56
.64
4.92
5.03
4.66
4.72
4.33
3.89
.52
.68
.60
.57
.38
.60
Proximal Outcomes
IvR
CrTh
MRA
MotEm
2.93
4.21
3.74
5.05
.67
.51
.86
.59
2.87
4.23
3.68
5.15
.65
.50
.85
.53
2.90
4.22
3.73
5.15
.63
.49
.85
.51
2.84
4.24
3.63
5.13
.67
.52
.85
.55
2.86
4.40
3.61
5.22
.72
.51
.91
.53
2.85
4.33
3.71
5.19
.66
.56
.78
.50
For each of the 5 subsamples, percentages are based on available N = 549.
Table 4.2: Norms for the Multidimensional Emotional Intelligence Assessment Workplace (MEIA-W) in men and women
Men
N = 321
(51%)
Women
N = 310
(49%)
Sex
differences
MEIA-W scale
Mean
SD
Mean
SD
t
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
4.78
4.96
4.50
4.56
4.17
3.82
.69
.79
.65
.63
.53
.60
4.86
4.79
4.51
4.45
4.32
3.98
.61
.79
.58
.63
.50
.53
-1.59
2.69**
-.11
2.23*
-3.75**
-3.58**
Proximal Outcomes
IvR
CrTh
MRA
MotEm
2.88
4.24
3.76
5.08
.67
.52
.90
.60
2.96
4.20
3.70
5.03
.67
.50
.80
.57
-1.57
.98
.89
1.20
Percentages are based on available N = 631.
differences are relatively modest. In any event, sex-based norms should not be used
in personnel selection applications. They are offered here for normative comparisons
not involved in selection.
45
Chapter 4
Use of the current norms for the MEIA-W is encouraged with caution owing to
limited representativeness of the broader working population. Future norming of the
MEIA-W will target a more diverse array of occupations, industries, and cultures.
Reliability & Interscale Correlations
Alpha coefficients for the MEIA-W and correlations among its 10 subscales are
reported in Table 4.3. The reliabilities in most cases are in the acceptable range
(median = 78.5). Internal consistency is lower than desired, however, in both NvExp
(.62) and CrTh (.60). The reason for these lower values is unclear. Further research
is expected to offer improvements in reliability through item refinement and, possibly,
item replacement. The pattern of interscale correlations is largely consistent with
that revealed in Table 1.4 for the MEIA based on college students. The correlations
are stronger, overall, in the MEIA-W. One possible explanation for this is that scores
on the MEIA-W are more uniformly affected by socially desirable responding, owing to perceived greater value of EI in work settings. Future research addressing
the susceptibility of the MEIA-W to self-deception and impression management
is expected to clarify current findings. In the meantime, it should be noted that
correlations among the MEIA-W scales, even after accounting for limitations due
to unreliability, are not so high as to suggest that any 2 subscales are assessing the
same facet of workplace EI. Further discussion of interscale correlations is offered
below under structural validity.
Content Validity
As discussed in Chapter 3, content validity in the context of a self-report measure
is concerned with how well a set of items captures a targeted construct domain.
Table 4.3: Alphas and interscale correlations for the Multidimensional Emotional Intelligence Assessment-Workplace (MEIA-W) (N = 653)
MEIA-W scale
RecSlf RegSlf RecOth RegOth NvExp Emp
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
.82
.64**
.51**
.44**
.45**
.06
.89
.36**
.44**
.20**
-.03
.81
.59**
.41**
.12**
.82
.46**
.23**
.62
.34**
Proximal Outcomes
IvR
CrTh
MRA
MotEm
-.25**
.29**
-.21**
.64**
-.35**
.32**
-.24**
.68**
-.05
.35**
.10**
.39**
-.04
.42**
.06
.47**
.01
.07*
.15** -.02
.02
.08*
.26** -.02
*p < .05, **p < .01, two-tailed tests.
46
IvR
CrTh
MRA
MotEm
.69
.78
.10**
.30**
-.35**
.60
.05
.79
.32** -.17**
.78
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
Content validity may be inferred from the writing of items targeting a specifically
defined construct (Anastasi & Urbina, 1997) and may be further documented by
demonstrating that items are reliably and independently classified by experts into
their targeted clusters. A unique feature of the MEIA, relative to other self-report EI
measures developed to date, is that such additional documentation was undertaken
during scale development. Specifically, only those items written for a given scale
were retained for further analysis if sorted into that scale by a majority of experts,
excluding the item writer, based on item review in light of construct definitions.
This type of evidence has not been gathered with respect to the MEIA-W. However,
because each MEIA-W item is a simple work-related copy of a content-validated
MEIA item, we believe an inference of content validity in the case of the MEIA-W
is well grounded. In any event, the extension of content validity to the MEIA-W,
conferred by simple word changes, is unique relative to other available work-oriented
EI measures.
Construct Validity
Correlations between the MEIA-W and both the Hogan Personality Inventory – Short
Form (HPI-SF; Hogan, Brinkmeyer, & Hogan, 1996) and the Personnel Assessment
Form (PAF; Jackson, 1999), a measure of cognitive ability, allow Inferences of
construct validity. The HPI-SF is a 115-item self-report measure of 7 normal-range
traits derived from the five-factor model of personality. Scales and brief definitions
are as follows: Adjustment denotes calmness and self-acceptance; Ambition reflects
social self-confidence, energy, and leaderlike qualities; Sociability represents the need
to interact with other people; Likeability refers to tact and interpersonal sensitivity;
Prudence denotes conscientiousness, conformity, and dependability; Intellectance
refers to creativity and interest in intellectual matters; and School Success reflects
enjoyment of academic activities and pursuit of educational achievement (Hogan,
Brinkmeyer, & Hogan, 1996). The PAF is a 14-minute test providing separate scores
for verbal and quantitative ability as well as overall intelligence. Developed explicitly for use in employment settings, the PAF total score correlates .80 with more
established measures of intelligence (Gellatly, Paunonen, Meyer, & Jackson, 1991).
The MEIA-W, HPI-SF and PAF were completed by 40 employees at a manufacturing company and 10 at a publishing company, both located in Tulsa, OK (total N
= 50). Of this sample, 30% were female, 74% were White, 12% Black, and 14%
Native American. Mean age was 35 years and mean company tenure was 49 months.
Correlations between the MEIA-W and both the HPI-SF and PAF are reported
in Table 4.4. Relations between the MEIA-W and the HPI-SF are largely interpretable. For example, RecSlf and RegSlf correlate .33 and .48, respectively, with
HPI-Adjustment, CrTh correlates .43 with HPI-Intellectance, and MotEm correlates
.26 and .34 with HPI-Ambition and School Success, respectively. As to cognitive
ability, CrTh correlates .36 with verbal ability, and IvR correlates -.27 with math
ability, the latter signifying overlap between quantitative skill and preference for
47
Chapter 4
Table 4.4: Correlations between the MEIA-W and the Hogan Personality Inventory - Short Form
Hogan Personality Inventory - Short Form
Cognitive ability
MEIA-W subscale
ADJ
AMB
SOC
LIK
PRU
INT
SCH
VA
MA
Total
Core EI
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
.33*
.48*
.12
.38*
.00
.17
.29*
.22
.26*
.39*
.01
-.12
.17
.06
.30*
.21
.13
-.10
.01
.10
.16
.23*
.06
.29*
-.09
-.22
-.07
.01
.15
.42*
.03
.12
.08
.14
-.12
-.29*
.25*
.16
.35*
.38*
.08
.07
.04
-.27*
.26*
.22
.16
.11
.16
.06
.12
.12
.08
-.13
.09
-.19
.24*
.21
.15
.04
Proximal Outcomes
IvR
CrTh
MRA
MotEm
-.02
-.16
-.18
.20
-.17
.22
.04
.26*
-.19
.14
.05
.11
.11
.08
-.13
-.06
.19
-.24*
.07
-.07
-.07
.43*
-.15
.23*
-.04
.20
.20
.34*
.04
.36*
.18
.19
-.27*
.18
.05
.31*
-.06
.33*
.16
.25*
*r >= .23, corresponding to p < .10 two-tailed or p < .05 one-tailed.
ADJ = Adjustment, AMB = Ambition, SOC = Sociability, LIK = Likeability, PRU = Prudence,
INT = Intellectance, SCH = School Success, VA = Verbal Ability, MA = Math Ability and the
Personnel Assessment Form (cognitive ability) (N = 50).
rational thinking. All told, the noted relationships are consistent with the construct
definitions targeted for assessment in developing the MEIA-W, in support of construct
validity. Furthermore, the pattern of relationships between the MEIA-W and the
HPI-SF largely mirrors that observed for the MEIA in relation to the JPI-R, reported
in Chapter 3 (see Table 3.4).
Criterion Validity
Research linking self-reported EI and job performance is sparse. Slaski and Cartwright (2002) report that EI, assessed with Bar-On’s (1997) Emotional Quotient
Inventory (EQ-i), correlated .22 with overall managerial performance. Wong and Law
(2002) found that a 16-item self-report measure of EI correlated .21 with supervisor
ratings of government workers’ performance. Subscales targeting 4 more specific
facets of EI showed validities ranging from .08 to .27. They also reported correlations of .40 with job satisfaction, and .14 (n.s.) with organizational commitment.
In a related effort, Abraham (2000) observed that EI contributed to organizational
commitment but not job satisfaction in customer service reps. Finally, Dulewicz and
Higgs (2000) report that a personality-based measure of EI was related to managerial advancement.
Although contributing uniquely to the growing literature on the role of EI in the
workplace, the noted studies offer limited evidence regarding linkages between EI
and job performance. In particular, few studies have examined EI as an articulated
construct with multiple facets, and fewer with respect to multiple dimensions of job
performance. In this section, results are reported showing relations between the
MEIA-W and job performance within a multidimensional framework.
48
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
A total of 116 employees of a publishing company (N = 76) and box manufacturing company (N = 40), both located in Tulsa, OK, completed the MEIA-W and were
rated on performance by their immediate supervisors. Jobs varied in each location,
sample positions including editor, IT programmer, plant manager, receptionist, and
crate assembler. Women constituted 46% of the sample, and of the 96 people reporting race, 81% were White, 6% Black, 2% Asian, 9% Native American, and 1%
Hispanic. Mean age was 35 years and mean company tenure, 49 months.
Job Performance Ratings
Performance ratings were obtained by interviewing each participant’s direct supervisor regarding 32 task and 18 contextual performance behaviors derived from 3
sources: performance appraisal instruments in use at the two companies, the academic
literature on task and contextual performance (e.g., Borman & Motowidlo, 1993),
and a detailed taxonomy of performance competencies developed by Tett, Guterman, Bleier, and Murphy (2000). All told, 8 task and 6 contextual dimensions (3 to
5 behaviors per dimension) were targeted. The task dimensions were Productivity,
Dependability, Job Knowledge, Problem Solving, Leadership, Interpersonal Relations, Flexibility, and Management, and the contextual dimensions were Helping,
Persistence, Volunteering, Rule Following, Organizational Support, and General
Contextual Performance. Supervisors judged their subordinates on each behavior
using a 0-to-8 scale with anchors ranging from “never / not at all” to “always / all
the time.” Three general performance measures were formed by averaging scores
across task components, across contextual components, and across all components.
Performance rating interviews lasting 1 to 3 hours were conducted in person or
by phone. Eighteen supervisors rated between 2 and 19 subordinates (mean = 6.1).
Behaviors judged by the supervisor as irrelevant for a ratee’s job were left unrated,
resulting in missing data for 19 of the 50 behaviors. Sample size was at least 102
for all but 6 behaviors (N range = 25 to 90). Judges were informed that their ratings
would be used only for research purposes and would remain confidential. Participants completed the MEIA-W within a month of having their performance rated.
Differences among supervisors in leniency/severity bias led us to adjust the criterion data as follows. First, we calculated the mean rating per supervisor, averaging
across all dimensions and all of the given supervisor’s ratees (range = 5.3 to 7.0).
Second, we averaged these supervisor means across supervisors to yield an overall
mean. Third, we took the difference between each supervisor’s mean and the overall
mean, positive differences indicating leniency and negative differences indicating
severity. Finally, we subtracted that difference, per supervisor, from every rating by
that supervisor, such that lenient supervisors’ ratings were adjusted downward and
severe supervisors’ ratings were adjusted upward. This resulted in all raters having
the same adjusted mean (averaging across all dimensions and all of his or her ratees),
but leaving individual differences among ratees within raters completely intact.
49
Chapter 4
Item analysis of the adjusted behavior ratings (i.e., treating behaviors as items)
guided improvements in composite reliability and interpretability. Specifically, any
behavior was removed from its composite if doing so would increase alpha (increases
were substantial in several cases). This was repeated one behavior at a time until
alpha for each composite was maximized based on retained behaviors. We then
added a behavior (from the set of rejects) to an intact composite if doing so (a) made
sense conceptually in light of retained behaviors, and (b) would improve reliability.
This was done separately within the sets of task and contextual behaviors. Some of
the original composites “fell apart,” yielding 4 task and 5 contextual performance
composites, with alphas ranging from .59 to .91 (median = .73). Several composites
were relabeled to reflect their modified content. Twelve of the 50 behaviors, including
all those with N < 102, remained unassigned owing to their uniqueness in relation
to the composites. These behaviors were excluded from subsequent analyses. Final
item analysis results are offered in Table 4.5.
Criterion Validity Results
Correlations between the MEIA-W and performance variables are reported in Table
4.6. Results are noteworthy in several respects. First, all 10 MEIA-W scales are
significantly related to at least 1 performance composite, and half are related to at
least 2. Of the 90 main linkages, 20 are significant (22%), exceeding chance levels.
Second, in keeping with overall expectations, 18 of the 20 noted significant linkages
(90%) are positive, including all 13 of those involving the 6 core MEIA-W scales.
Third, regarding the 2 exceptions to the positive correlations, MRA correlates -.16
and -.17 with Positive Thinking and Initiative, respectively, in keeping with this
EI dimension denoting a tendency to focus on negative past events. Fourth, the
performance composite Flexibility accounts for 4 of the 7 significant correlations
involving task performance, and Loyalty, for 6 of the 13 significant relations involving contextual performance. Fifth, of the 10 MEIA-W scales, RegOth and NvExp
are the two strongest and most consistent predictors, the former showing the only
significant relation with Productivity and the largest relationships with all three overall
performance scores (range = .23 to .26). Sixth, the MEIA-W scales are proportionally more frequently related to the contextual composites (26%) than to the task
composites (17.5%). Correspondingly, the 6 core MEIA-W scales show a higher
proportion of significant relations (24%) relative to the 4 proximal outcomes (19%).
These findings offer preliminary support for the criterion validity of the MEIAW within a multidimensional framework. The two most consistent predictors of
supervisory ratings in this sample are RegOth and NvExp, and the performance
dimensions most successfully predicted by the MEIA-W are Flexibility and Loyalty.
Further research on the criterion validity of the MEIA-W is warranted, particularly
with respect to (a) the differential relevance of EI to performance in selected jobs
(e.g., high relevance: customer service vs. low relevance: manual labor), and (b)
the effects of deliberate response distortion in job applicants versus incumbents.
50
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
Table 4.5: Final item analysis results for task and contextual performance behaviors forming composites
for use in criterion validation of the MEIA-W
Task performance
Productivity
alpha = .91
N = 98
n items = 7
Project Management
alpha = .82
N = 88
n items = 6
Professionalism
alpha = .73
N = 116
n items = 4
Flexibility
alpha = .84
N = 104
n items = 6
Positive thinking
alpha = .62
N = 116
n items = 2
Initiative
alpha = .63
N = 89
n items = 2
Normative support
alpha = .60
N = 116
n items = 4
Loyalty
alpha = .59
N = 102
n items = 3
Extra effort
alpha = .78
N = 104
n items = 4
prd1
prd2
prd3
prd4
prd5
prd6
prd7
tsk1
tsk2
tsk3
tsk4
tsk5
tsk6
pro1
pro2
pro3
pro4
flx1
flx2
flx3
flx4
flx5
flx6
.80
.77
.76
.75
.68
.68
.62
.67
.66
.58
.58
.56
.52
.59
.55
.51
.44
.71
.69
.68
.66
.61
.44
Maximizes productivity and achieves work-related goals.
Completes work assignments thoroughly and accurately.
Meets deadlines; adheres to schedules; is punctual.
Works well with little supervision.
Uses good judgment when making decisions.
Ensures customer satisfaction and product quality.
Shows organized work habits; utilizes time efficiently.
Inspires others to strive to achieve shared goals.
Monitors current work unit progress in light of standards and goals.
Identifies work-related problems.
Actively seeks, integrates, and shares job-related information.
Understands how the company operates as a whole.
Plans and coordinates the steps needed to complete tasks before action is taken.
Accepts responsibility for own actions, decisions, and directions to co-workers.
Works well with others.
Works well under pressure; stays calm in stressful situations.
Communicates clearly and effectively.
Readily grasps new ideas and concepts.
Balances competing work demands (e.g., co-workers, the bottom line).
Shows imagination, creativity, and resourcefulness in solving problems.
Shows decisiveness; does not hesitate in making tough decisions.
Adapts to new situations and immediate work demands.
Encourages sharing of ideas; seeks others’ work-related opinions.
Contextual performance
pos1
pos2
ini1
ini2
org1
org2
org3
org4
loy1
loy2
loy3
ext1
ext2
ext3
ext4
.46
.46
.46
.46
.45
.42
.38
.28
.57
.36
.32
.74
.59
.54
.52
Undertakes routine duties with energy and optimism.
Demonstrates positive attitudes toward the job.
Takes the initiative in solving job-related problems.
Offers ideas for improving the work unit or organization.
Volunteers for extra assignments, tasks, or committees.
Encourages others in dealing with difficult work situations.
Avoids taking shortcuts that bend the rules.
Upholds company policies, even if unpopular.
Displays loyalty to the organization; shows commitment to stay with the company.
Supports co-workers in helping them deal with their personal problems.
Openly demonstrates public support for the company.
Goes beyond the call of duty in getting things done.
Spends extra time and stays late to complete a job.
Persists in overcoming obstacles to complete a task.
Pitches in to help others when they need it.
Values in the second column are corrected item-total correlations (i.e., between an item and the unit-weighted sum of the
remaining items in the composite.)
51
Chapter 4
Table 4.6: Correlations between the MEIA-W and supervisor ratings of task and contextual performance (N = 116)
Core EI
Performance criterion
Task performance
Productivity
Project management
Professionalism
Flexibility
Contextual performance
Positive thinking
Initiative
Normative support
Loyalty
Extra effort
Mean performance
Task
Contextual
Overall
Proximal Outcomes
RecSlf RegSlf RecOth RegOth NvExp Emp
IvR
CrTh MRA MotEm
.08
-.02
.04
.13
.06
-.03
.19*
.09
.08
.02
-.06
.13
.19*
.12
.11
.31*
.10
.17*
.10
.17*
.08
.07
.02
.08
-.06
.05
-.11
.13
.04 .03 .09
-.02 -.07 .05
-.12 -.11 -.07
.18* .01 .17*
.13
.06
-.04
.22*
.05
.23*
.11
.03
.08
.13
-.02
.07
-.08
.16*
.02
.01
.22*
.03
.29*
.22*
.04
.14
.01
.29*
.04
-.09
.03
.01
.24*
.07
-.05
-.11
.26*
.05
.06
-.02
.00
.06
-.02
-.01
-.16*
-.17*
.00
.10
-.07
.17*
.16*
.05
.01
.14
.07
.14
.11
.09
.19*
.15*
.06
.05
.06
.23*
.25*
.26*
.17*
.18*
.19*
.08
.08
.09
.01
.05
.03
.03
.00
.02
-.05
-.11
-.09
.08
.17*
.13
*p < .10, two-tailed and < .05, one-tailed.
Incremental Validity
The value of any new measure rests on how much it contributes beyond previously
established measures. The incremental validity of the MEIA-W was assessed in
predicting and understanding job performance beyond personality and ability tests,
based on the same sample of 50 workers reported above under construct validity.
Groups of 2 to 4 were administered the HPI-SF and PAF approximately 5 months
after completing the MEIA-W and performance assessments. Each set of variables
was correlated with the 9 job performance composites described in Table 4.5, as
well as means for task, contextual, and overall performance. Of the 120 correlations
involving the MEIA-W, 19 (16%) were significant; of the 84 correlations involving
the HPI-SF, 17 (20%) were significant; and, of the 24 correlations involving the PAF,
7 (29%) were significant.
The incremental validity of the MEIA-W was assessed by trying to account for
each significant relationship between a given MEIA-W scale and a given performance
composite by entering the HPI-SF and PAF scales into hierarchical stepwise regression in earlier steps. For example, in assessing the unique contribution of MEIA-W
Emp to Loyalty (r = .35; N = 50), PAF Verbal and Math were allowed to enter in block
1 (p to enter = .05, p to remove = .10), then all 7 HPI-SF scales were allowed to enter
in block 2 before forcing Emp to enter in block 3. Incremental validity would be
evident if R2 increases significantly (p < .05) at block 3. For comparison purposes,
the same procedure was used in assessing the incremental validity of ability over
personality and EI, and personality over ability and EI. Unique effects were further
articulated by comparing initial zero-order correlations to corresponding partial
52
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
correlations for the given predictor and performance measures. Incremental validity
in this case would be reflected in a smaller drop from zero-order to partial values.
Tables 4.7, 4.8, and 4.9 present results of the regressions for incremental validity of the PAF, HPI-SF, and MEIA-W, respectively, each controlling for the other 2
measures. As shown in Table 4.7, PAF Verbal contributes uniquely to relations with
2 performance composites (Productivity and Initiative) out of a total of 7 significant
zero-order relationships involving either ability type (29%). The HPI-SF, particularly
School Success, offered the most consistent challenges to ability. Table 4.8 shows
that, of the 17 significant zero-order relations involving the HPI-SF, 7 survived after
controlling for the PAF and MEIA-W (41%). School Success, for example, survived
2 of 7 linkages, with PAF Verbal accounting for most of the attrition. Table 4.9 reveals that 12 of the 19 significant validities involving the MEIA-W (63%) remained
significant after controlling for the HPI-SF and PAF. RegOth survived in 4 of 7
cases (57%), most notably, its relation with Flexibility, where R2 at step 3 increased
Table 4.7: Incremental validity of ability (PAF) over personality (HPI-SF) and EI (MEIA-W) (N = 50)
Step 1
Step 2
Step 3
Performance
criterion
HPI-SF
MEIA-W
PAF
R2 ch.
zero
order r
partial
r
Productivity
SCH
Verbal
Math
.09**
.02
.40
.29
.32
.16
Project Management
PRU
RecSlf
IvR
Verbal
.04
.24
.23
Flexibility
SCH
RegOth
IvR
Verbal
.02
.30
.18
Initiative
INT
Verbal
.10**
.34
.34
Overall Task Perf.
SCH
Verbal
.05
.32
.23
Overall Performance
SCH
Verbal
.03
.27
.17
SCH = School Success, PRU = Prudence, INT = Intellectance
*p < .05, **p < .01
by .12. RecOth survived in 1 of 3 cases (Flexibility), Emp survived in each of 2
cases, and 4 other MEIA-W scales each survived single significant relations with
performance. Comparisons between mean absolute value zero-order correlations and
corresponding mean partial correlations, derived from the right side of each table,
show drops from .31 to .23 (25%) for the PAF, .28 to .22 (21%) for the HPI-SF, and
.29 to .24 (16%) for the MEIA-W.
Our results suggest that EI stands to contribute uniquely to predicting job performance beyond personality and ability and support the use of the MEIA-W for such
purposes. A full 63% of the 19 significant relations involving the MEIA-W survived
53
Chapter 4
Table 4.8: Incremental validity of personality (HPI-SF) over ability (PAF) and EI (MEIA-W) (N = 50)
Step 1
Step 2
Step 3
Performance
criterion
PAF MEIA-W
HPI-SF
R2 ch.
Productivity
Verbal
zero
order r
partial
r
SCH
.02
.30
.17
RecSlf
Emp
PRU
.00
.24
.01
Professionalism
RegSlf
MotEm
AMB
SOC
SCH
.06*
.09**
.08**
.25
.28
.23
.27
.32
.32
Flexibility
Verbal
SCH
.01
.32
.14
Positive Thinking
AMB
SCH
.09**
.07*
.30
.27
.30
.27
Initiative
Verbal
RecSlf
RegSlf
INT
.05*
.29
.25
Loyalty
Emp
INT
.08**
-.38
-.31
Extra Effort
RegOth
INT
SCH
.04
.04
.25
.31
.22
.21
Overall Task Perf.
SCH
.04
.30
.20
Overall Context Perf.
RegOth
AMB
SCH
.04
.03
.27
.25
.19
.18
Overall Performance
Verbal
AMB
.05
.27
.22
SCH
.05
.31
.23
Project Management
Verbal
RegOth
IvR
Verbal
SCH = School Success, PRU = Prudence, AMB = Ambition, SOC = Sociability, INT = Intellectance
*p < .05, **p < .01
competition with 7 established and varied personality variables and math and verbal
abilities. Of the 10 MEIA-W scales, 8 contributed uniquely to performance, with
RegOth showing the strongest unique effect (12%, p < .01) in its contribution to
Flexibility, compared to all predictor-criterion pairings. Overall, the evidence suggests that, although demonstrating meaningful overlap with personality traits and
cognitive ability, trait-EI represents a unique domain with potential to expand our
understanding and prediction of workplace behavior. Interestingly, the MEIA-W
showed the strongest unique contributions to predicting job performance, after controlling for the remaining variables. Further research on the incremental validity of
the MEIA-W is warranted involving more specific personality variables.
Structural Validity
In Chapter 3, evidence was presented bearing on the structure of the six core-EI dimensions proposed by Salovey and Mayer (1990) as assessed by the MEIA (RecSlf,
54
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
Table 4.9: Incremental validity of EI (MEIA-W) over ability (PAF) and personality (HPI-SF) (N = 50)
Step 1
Step 2
Step 3
Performance
criterion
PAF HPI-SF
MEIA-W
R2 ch.
Productivity
Verbal
zero
order r
partial
r
RegOth
.02
.23
.16
Project Management
Verbal
RecSlf
Emp
.07*
.07*
-.26
.29
-.28
.27
Professionalism
RecSlf
.08*
.30
.30
Flexibility
Verbal
RecOth
RegOth
MotEm
.09*
.12**
.06*
.37
.41
.30
.32
.37
.26
Initiative
Verbal
MotEm
.01
.23
.10
Normative Support
IvR
.08*
.28
.28
Loyalty
INT
AMB
NvExp
RecOth
Emp
RegOth
MRA
.06*
.04
.06*
.05*
.02
.31
.24
.35
.27
.25
.27
.23
.29
.26
.18
Extra Effort
SCH
RecOth
RegOth
.02
.06*
.24
.34
.15
.25
Overall Task Perf.
RegOth
.03
.23
.18
RegOth
.03
.25
.18
RegOth
.05*
.27
.22
SOC
INT
SOC
ADJ
Verbal
Overall Context Perf.
AMB
SOC
Overall Performance
Verbal
SOC = Sociability, INT = Intellectance, ADJ = Adjustment, AMB = Ambition, SCH = School Success
*p < .05, **p < .01
RegSlf, RecOth, RegOth, NvExp, and Emp). Here, we offer corresponding results
for the MEIA-W. The initial sample consisted of 237 employees at 3 organizations:
114 from an American publishing company, 50 from an American manufacturing
company, and 73 from an Australian MBA program. Gender was closely balanced
(50.5% female; age unavailable). Removal of multivariate outliers, using Mahalanobis distance, yielded a final sample of 225. Manufacturing workers completed
the MEIA-W at the worksite; publishing company workers and Australian MBA
students completed a web-based version on their own time.
Correlations among the 10 MEIA-W scales are reported in Table 4.10 along with
corresponding results for the MEIA, based on responses from 184 college students,
to allow direct comparisons. Results of model testing (AMOS 4.0) are reported in
Table 4.11. As in the case of the MEIA, significant improvements in fit are evident,
moving from single-factor to multi-factor solutions. The modified 3-factor solution
55
Chapter 4
Table 4.10: Correlations among subscales for the MEIA and MEIA-W
Core EI facets
EI scale
Proximal Outcomes
RecSlf RegSlf RecOth RegOth Emp NvExp
IvR
CrTh MRA MotEm
Core EI facets
RecSlf
RegSlf
RecOth
RegOth
NvExp
Emp
--
.47*
.29*
.30*
.36*
.15*
.61*
--
.16*
.19*
.08
-.06
.48*
.34*
--
.61*
.31*
.38*
.35*
.32*
.60*
--
.42*
.46*
.14*
.09
.15*
.26*
.51*
--
.41*
.15*
.44*
.46*
--
.37*
-.16* .33*
-.23* .30*
.02 .41*
.13 .34*
.10 .20*
.04 -.07
Proximal outcomes
IvR
CrTh
MRA
MotEm
-.04
.19*
.22*
.47*
-.24*
.17*
-.06
.47*
.10
.36*
.24*
.33*
.14
.26*
.33*
.47*
.18*
.17*
.36*
.16*
.29*
.22*
.26*
.37*
--
.34*
.23*
.03
-.32*
-.25*
-.03
.08
-.07
.10
.57*
.57*
.35*
.39*
.22*
.09
.13* .29* -.23*
--
-.03 .27*
.19* --
-.12
.23* .31* --
*p < .05
Correlations for the MEIA (student sample; N = 184) are below the diagonal and values for the MEIA-W (worker sample; N
= 225) are above the diagonal.
Table 4.11: Fit results for the MEIA-W (N = 225)
Model
df
Chi sq.
GFI
AGFI
TLI
NFI
RMR Chi sq. diff df
1-factor
S&M 2-factor
2-factor
3-factor
modified 3-factora
9
8
8
6
5
98.70
96.87
75.35
23.94
12.92
.88
.89
.91
.97
.98
.72
.70
.75
.89
.93
.61
.56
.67
.88
.94
.75
.76
.81
.94
.97
.046
.044
.040
.019
.012
1.83
23.35*
74.76*
85.78*
1
1
3
4
includes correlated error terms for RecSlf & NvExp; Chi sq. difference = 11.02 (df = 1, p < .001) in comparison to the original
3-factor model
*p < .001 in comparison to the 1-factor model
a
reported in Chapter 3 for the MEIA, including correlated error terms for NvExp
and RecSlf (see Table 3.7), replicated here, fit indices ranging from .93 to .98. Final
path coefficients are presented in Figure 4.1 (compare with Figure 3.2 for college
student data on the MEIA).
The high degree of similarity in the factor structures of the 6 core-EI dimensions
assessed with the MEIA and MEIA-W supports use of the MEIA-W as a workplace
surrogate of the original MEIA. As a further comparison between the MEIA and
MEIA-W, correlations between the 3 core EI composites and 4 proximal outcomes
based on the MEIA-W are reported in Table 4.12 (see Table 3.8 for corresponding
results based on the MEIA).
Unlike in the case of the 6 core EI dimensions, correlations involving the 4 proximal outcomes vary between the student and worker samples. Whereas MRA in
workers reflects low Self-Orientation, MRA in students (see Table 3.8) reflects high
56
Multidimensional Emotional Intelligence Assessment - Workplace (MEIA-W)
Other Orientation. Also, Emotional Sharing in students is related to all 4 proximal
outcomes, whereas, in workers, it is related to only MotEm. Reasons for these discrepancies are unclear. In the case of MRA, the sorts of past problems considered
in responding to the MRA items may vary between groups (i.e., workplace vs. less
context-specific). The groups might also vary in their approach to judging past difficulties, perhaps relating to maturity. Regarding Emotional Sharing, students’ use
of IvR, CrTh, MRA, and MotEm suggests greater reliance on communication of
emotions among friends and family, relative to workers, whose EI applications may
be more job-focused. Further research is needed to explore differences between
populations in how core EI dimensions are used.
Figure 4.1: Final path coefficients for the modified 3-factor model of emotional intelligence based on
worker sample (N =225)
In sum, the structure of EI core facets assessed with the MEIA-W replicated well
the structure observed using the MEIA. Correlations involving the applications,
however, varied notably between measures and samples. This suggests that core EI
structure is more universal than how core dimensions are applied. Accordingly, EI
models that include proximal outcomes as central to EI may not replicate as well
across populations as EI models excluding such outcomes. We urge MEIA-W users
to treat core facets separately from their applications.
57
Chapter 4
Table 4.12: Correlations between MEIA-W core composites and proximal outcomes (N = 225)
MEIA-W Proximal Outcomes
MEA-IW core composites
Self Orientation
Other Orientation
Emotional Sharing
IvR
CrTh
MRA
MotEm
-.22*
.08
.08
.35*
.42*
.08
-.32*
-.03
.02
.63*
.41*
.19*
*p < .05
Self Orientation = (RecSlf + RegSlf) / 2
Other Orientation = (RecOth + RegOth) / 2
Emotional Sharing = (NvExp + Emp) / 2
IvR = Intuition vs. Reason
CrTh = Creative Thinking
MRA = Mood Redirected Attention
MotEm = Motivating Emotions
Summary
The MEIA-W was created through simple word additions (e.g., “at work”) and substitutions (“co-worker” in place of “person”) to be a close surrogate of the MEIA
for use in worker populations. Empirical evidence on the MEIA-W collected to
date suggests that its measurement properties, like those of its parent MEIA, are
favorable, on the whole. Internal consistency reliabilities (alphas) are, for the
most part, in the acceptable range (.60 to 89; mean = .76, median = .79). Content
validity can be judged as having carried over from that documented for the MEIA.
Construct validity is largely supported by interpretable relations with the HPI-SF
(Big Five personality) and PAF (verbal and math abilities). Criterion validity is
established by multiple significant relationships with supervisors’ ratings of task
and contextual performance. Incremental validity of the MEIA-W is evident in
predicting job performance after controlling for Big Five personality and cognitive
ability and, moreover, compares favorably to corresponding results for the HPI-SF
(personality) and PAF (ability). The structure of the 6 core MEIA-W scales is very
similar to the structure of the 6 core MEIA scales, in support of a stable 3-factor
solution. Relations between the 3 core EI composites and the 4 proximal outcomes
vary between the MEIA-W and MEIA. Such differences may reflect differences in
the populations under investigation (i.e., students versus workers).
In sum, the MEIA-W has yielded promising results with respect to its reliability
and validity for use with worker populations. Further research is needed to examine
the conditions under which it is likely to offer better predictions of job performance
(e.g., in jobs for which EI is judged to be most relevant, e.g., leadership, sales, customer service, health care). Researchers examining such possibilities are urged to
treat each facet of trait-EI targeted by the MEIA-W as a potentially separate construct and to weigh the practical advantages of a single, overall EI score against the
diagnostic advantages of facet-level scores within a multidimensional framework.
58
References
Abraham, R. (1999). Emotional intelligence in Organizations: A conceptualization.
Genetic, Social & General Psychology Monographs, 125, 209-226.
Abraham, R. (2000). The role of job control as a moderator of emotional dissonance
and emotional intelligence-outcome relationships. The Journal of Psychology, 134, 169-184.
Anastasi, A., & Urbina, S. (1997). Psychological Testing (7th ed.) Upper Saddle
River, NJ: Prentice-Hall.
Ashkanasy, N. M., & Daus, C. S. (2002) Emotion in the workplace: The new challenge for managers. Academy of Management Executive, 16, 76-86.
Bar-On, R. (1997). The Bar-On emotional quotient inventory (EQ-i): Technical
manual. Toronto, Canada: Multi-Health Systems.
Bar-On, R., Brown, J. M., Kirkcaldy, B. D., & Thome, E. P. (2000). Emotional expression and implications for occupational stress; An application of the
Emotional Quotient Inventory (EQ-i). Personality and Individual Differences, 28, 1107-1118.
Bing, M. N., Whanger, J. C., Davison, H. K., & VanHook, J. B. (2004). Incremental
validity of the frame-of-reference effect in personality scale scores: A
replication and extension. Journal of Applied Psychology, 89, 150-157.
Borman, W. C., & Motowidlo, S. J. (1993) Expanding the Criterion Domain to Include Elements of Contextual Performance. In Schmitt, N., & Borman,
W. C., Personnel Selection in Organizations (pp. 71-98). San Fransico:
Jossey-Bass.
Byrne, B. M. (1998). Structural equation modeling with AMOS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum.
Ciarrocchi, J., Chan, A. Y., & Caputi, P. (2000). A critical evaluation of the emotional
intelligence construct. Personality and Individual Differences, 28, 539561.
Crowne, D. P., & Marlowe, D. (1964). The approval motive. Studies in evaluative
dependence. New York: John Wiley & Sons.
Davies, M., Stankov, L., & Roberts, R. D. (1998) Emotional intelligence: In search of
an elusive construct. Journal of Personality and Social Psychology, 75,
989-1015.
Davis, M. H., & Franzoi, S. L. (1999). Self-awareness and self-consciousness. In V.
J. Derlega, B. A. Winstead & W. H. Jones (Eds.), Personality: Contemporary theory and research (2nd ed.) (pp. 307-338). Chicago: Nelson-Hall.
59
References
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with
life scale. Journal of Personality Assessment, 49, 71-75.
Dulewicz, V., & Higgs, M. (2000). Emotional Intelligence. A review and evaluation
study. Journal of Managerial Psychology, 15 ,341-372.
Edwards, A. L. (1957). The social desirability variable in personality assessment and
research. New York: Dryden.
Elfenbein, H. A., & Ambady, N. (2002). Predicting workplace outcomes from the
ability to eavesdrop on feelings. Journal of Applied Psychology, 87, 963971.
Fox, S., & Spector, P. E. (2000). Relations of emotional intelligence, practical intelligence, general intelligence and trait affectivity with interview outcomes:
It’s not all just ‘g.’ Journal of Organizational Behavior, 21, 203-220.
Gangestad, S. W., & Snyder, M. (1985). “To carve nature at its joints”; On the
existence of discrete classes in personality. Psychological Bulletin, 92,
317-349.
Gangestad, S. W., & Snyder, M. (2000). Self-monitoring: Appraisal and reappraisal.
Psychological Bulletin, 126, 530-555.
Gellatly. I. R., Paunonen, S. V., Meyer, J. P., & Jackson, D. N. (1991) Personality, vocational interest, and cognitive predictors of managerial job performance
and satisfaction. Personality and Individual Differences, 12, 221-231.
George. J. M. (2000). Emotions and Leadership: The role of emotional Intelligence.
Human Relations, 53, 1027-1055.
Goleman, D. (1995). Emotional intelligence. New York: Bantam Books.
Goleman, D. (1998). Working with emotional intelligence. New York: Bantam Books.
Helmes, E. (2000). The role of social desirability in the assessment of personality
constructs. In R. D., Goffin & E. Helmes (Eds.), Problems and solutions
in human assessment: Honoring Douglas N. Jackson at 70 (pp. 21-40).
Norwell, MA: Kluwer Academic Publishers.
Hemmati, T., Mills, J. F., & Kroner, D. G. (2004). The validity of the Bar-On emotional intelligence quotient in an offender population. Personality and
Individual Differences, 37, 695-706.
Hogan, J., Brinkmeyer, K., & Hogan, R. (1996). The Hogan Personality Inventory
Form-S Manual Tulsa, OK: Hogan Assessment Systems.
Hogan, R., & Hogan, J. (1995). Hogan Personality Inventory Manual. Tulsa, OK:
Hogan Assessment Systems.
60
References
Jackson, D. N. (1970). A sequential system for personality scale development. In C.
D. Spielberger (Ed.), Current topics in clinical and community psychology (pp. 61-96). New York: Academic Press.
Jackson, D. N. (1989). Personality Research Form Manual (3rd ed.). Port Huron, MI:
Research Psychologists Press.
Jackson, D. N. (1994). Manual for the Jackson Personality Inventory – Revised. Port
Huron, MI: Sigma Assessment Systems.
Jackson, D. N. (1999). Personnel Assessment Form Manual. Port Huron, MI: Sigma
Assessment Systems, Inc.
Jackson, D. N., & Messick, S. (1958). Content and style in personality assessment.
Psychological Bulletin, 55, 243-252.
Jordan, P. J., Ashkanasy, N, M., Hartel, C. E., & Hooper, G. S. (2002). Workgroup
emotional intelligence: Scale development and relationship to team process efectiveness and goal focus. Human Resource Management Review,
12, 195-214.
Kelley, C., & Meyer, J., (1992). CCAI (Cross-Cultural Adaptability Inventory)
Manual. Minneapolis, MN: National Computer Systems.
Mayer, J. D., & Salovey, P. (1997). What is emotional intelligence? In P. Salovey &
D. Sluyter (Eds.), Emotional development and emotional intelligence
education Implications for educators (pp. 3-31). New York: Basic Books.
Mayer, J. D., & Salovey, P., Caruso, D. R. (2000). Competing models of emotional intelligence. In R.J. Sternberg (Ed.) Handbook of human intelligence (2nd
ed.), (pp. 396-420). New York: Cambridge University Press.
Millsap, R. E. (2002). Structural equation modeling: A users-guide. In F. Drasgow &
N. Schmitt (Eds.), Measuring and analyzing behavior in organizations:
Advances in measurement and data analysis (pp. 257-301). San Francisco, CA: Jossey-Bass.
Murray, H. (1938). Explorations in personality. New York: Oxford University Press.
Ones, D. S., Viswesvaran, C., & Reiss, A. D. (1996). Role of social desirability in
personality testing for personnel selection: The red herring. Journal of
Applied Psycholgy, 81, 660-679.
Palmer, B. R., Donaldson, C., & Stough, C. (2002). Emotional intelligence and life
satisfaction. Personality and Individual Differences, 33, 1091-1100.
Paulhus, D. L. (1984). Two-component models of socially desirable responding.
Journal of Personality and Social Psychology, 46, 598-609.
61
References
Petrides, K. V., & Furnham, A. (2000). On the dimensional structure of emotional
intelligence. Personality and Individual Differences, 29, 313-320.
Petrides, K. V., & Furnham, A. (2003). Trait emotional intelligence: Behavioral
validation in two studies of emotion recognition and reactivity to mood
induction. European Journal of Personality, 17, 29-57.
Roberts, L. W., & Clifton, R. A. (1992). Measuring the affective quality of life of university students: The validation of an instrument. Social Science Indicators, 27, 113-137.
Roberts, R. D., Zeidner, M., & Matthews, G. (2001). Does Emotional Intelligence
Meet Traditional Standards for an Intelligence? Some New Data and
Conclusions. Emotion, 1, 196-231.
Robie, C., Schmit, M. J., Ryan, A. M., & Zickar, M. J. (2000). Effects of item context specificity on the measurement equivalence of a personality inventory. Organizational Research Methods, 3, 348-365.
Saklofske, D. H., Austin, E. J., & Minski, P. S. (2003). Factor structure and validity
of a trait emotional intelligence measure. Personality and Individual Differences, 34, 707-721.
Salovey, P., & Mayer, J. D. (1990). Emotional intelligence. Imagination, Cognition
and Personality, 9, 185-211.
Saucier, G. (1994). Mini-markers: A brief version of Goldberg’s unipolar Big-Five
markers. Journal of Personality Assessment, 63, 506-516.
Schmit, M. J., Ryan, A. M., Stierwalt, S. L., & Powell, A. B. (1995). Frame-of-reference effects on personality scale scores and criterion-related validity.
Journal of Applied Psychology, 80, 607-620.
Schutte, N. S., Malouff, J. M., Hall, L. E., Haggerty, D. J., Cooper J. T., Golden, C.
J., & Dornheim, L. (1998). Development and validation of a measure of
emotional intelligence. Personality and Individual Differences, 25, 167177.
Slaski, M., & Cartwright, S. (2002). Health, performance and emotional intelligence:
An exploratory study of retail managers. Stress and Health, 18, 63-68.
Tett, R. P., & Burnett, D. B. (2003). A personality trait-based interactionist model of
job performance. Journal of Applied Psychology. 88, 500-517.
Tett, R. P., & Fox, K. E. (in press). Confirmatory factor structure of emotional intelligence in student and worker samples. Personality and Individual Differences.
62
References
Tett, R. P., Fox, K. E., & Wang, A. (2005). Development and Validation of a SelfReport Measure of Emotional Intelligence as a Multidimensional Trait
Domain, Personality and Social Psychology Bulletin.
Tett, R. P., Guterman, H. A., Bleier, A., & Murphy, P. J. (2000). Development and
content validation of a “hyperdimensional” taxonomy of managerial
competence. Human Performance, 13, 205-251.
Tett, R. P., Jackson, D. N., Rothstein, M., & Reddon, J. R. (1999). Meta-analysis of
bi-directional relations in personality-job performance research, Human
Performance, 12, 1-29.
Ullman, J. B. (2001). Structural equation modeling. In B. G. Tabachnick & L. S. Fidell (Eds.), Using multivariate statistics (4th ed.), (pp. 653-771). Boston,
MA: Allyn and Bacon.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief
measures of positive and negative affect: The PANAS scales. Journal of
Personality and Social Psychology, 54, 1063-1070.
Wong, C., & Law, K. S. (2002). The effects of leader and follower emotional intelligence on performance and attitude: An exploratory study. The Leadership Quarterly, 13, 243-274.
63
SIGMA Assessment Systems, Inc.
P.O. Box 610757, Port Huron, MI 48061-0757
Research Psychologists Press, Inc.
P.O. Box 3292, Station B, London, ON N6A 4K3