Archive ouverte UNIGE

https://archive-ouverte.unige.ch

Chapitre de livre 2013 Published version Open Access

This is the published version of the publication, made available in accordance with the publisher’s policy.

The GRID meets the wheel: assessing emotional feeling via self-report

Scherer, Klaus R.; Shuman, Vera; Fontaine, Johnny; Soriano, Cristina

How to cite

SCHERER, Klaus R. et al. The GRID meets the wheel: assessing emotional feeling via self-report. In:
Components of emotional meaning: A sourcebook. Oxford : Oxford University Press, 2013. doi:
10.1093/acprof:oso/9780199592746.003.0019

This publication URL: https://archive-ouverte.unige.ch//unige:97384

Publication DOI: 10.1093/acprof:oso/9780199592746.003.0019

© This document is protected by copyright. Please refer to copyright holder(s) for terms of use.
 Chapter 18

The GRID meets the Wheel: Assessing

emotional feeling via self-report
Klaus R. Scherer,1 Vera Shuman, Johnny J. R. Fontaine,
and Cristina Soriano

The GRID study has provided a wealth of new data of high relevance to understand the semantics
of emotion terms. This data can be profitably applied to create new tools for emotion research, or

13
to further develop the existing ones. Here, we illustrate one such application by describing how the

20
GRID paradigm was used to improve and further validate a popular tool for emotion assessment,
namely the Geneva Emotion Wheel (GEW), a self-report measure of feelings.
,
ss
Componential theories define emotion as a process during which several components such as
e
physiological responses, motor expression, and cognitive representations (of both eliciting events
Pr

and self-perceived response patterns) become synchronized over a limited period of time (see
y

Chapter 1). One of these components is Subjective Feeling, a holistic cognitive representation
it

that integrates the temporarily coordinated changes of the other components into a succinct, well
rs
ve

formed Gestalt, allowing the individual to reach awareness of his/her state and label it—stating that
he/she “has” or “feels” a particular emotion. In order to study the feeling component of emotion,
i
Un

psychologists need to rely on self-report. There is no other means but to ask the individual to report
on the nature of his/her experience, since feeling is defined as a subjective cognitive representa-
d
or

tion of the emotional state which reflects a unique integration of mental and bodily changes in the
xf

context of a particular event (see Chapter 1). Emotion researchers currently use various paradigms
)O

for self-report, including the more recently developed GEW, to be reviewed below.
(c

18.1 Classic self-report emotion assessment methods

Psychologists sometimes ask the participants in a study to describe their feelings in their own
words. While this procedure may yield interesting information, it is fraught with problems. For
example, people differ with respect to their verbal ability and richness of vocabulary (e.g., Gohm &
Clore, 2000), which makes it difficult to compare reports across individuals or to rely on their
accuracy. This is a major problem in controlled experimental research where fine-grained scalar
measurement for a circumscribed number of specific feeling states is required. In consequence,
psychologists generally use forced-choice self-reports of emotional experience. There are two
major approaches: (1) the discrete emotion labels approach, and (2) the dimensional rating ap-
proach. The following section, based on Scherer (2005a), reviews these two traditions.

1 Corresponding author: Klaus R. Scherer. Swiss Center for Affective Sciences—University of Geneva. 7, Rue
des Battoirs, CH-1205 Geneva, Switzerland. [email protected]

19-Fontaine-Chap18.indd 281 7/17/13 3:45 PM

 282 COMPONENTS OF EMOTIONAL MEANING

The discrete emotion labels approach is used by scholars and laymen alike to categorize the stream
of emotional experience into separate states profiting from the existence of specific emotion words
and expressions in language (the type of words used in the GRID study). As shown in Chapter 3,
while there are differences between languages with respect to the richness of the emotion lexicon
and the meanings of related words, there is also a high degree of overlap. Darwin (1872) has used
this convergence to postulate the evolutionary continuity of a set of fundamental emotions and
the observable physiological and expressive symptoms that accompany them. This approach has
been revived by Tomkins (1962), and has been popularized by Izard (1971; differential emotions
theory) and Ekman (1972; basic emotion theory). In this tradition, categorizing emotional experi-
ences according to the emotion words available in natural languages, it is typically assumed that
the language-based categories reflect unique appraisal and response patterns (facial, vocal, and
physiological) driven by typical event appraisals.
Researchers adopting the discrete emotion approach to assess emotional experience use scales
with nominal, ordinal, or interval characteristics. Typically, respondents are presented with a list
of emotion terms and are asked (1) to check the terms that best describe the emotion experienced
(nominal scale); (2) to indicate on a scale (generally with 3 to 7 points) whether the emotion was

13
experienced “a little,” “somewhat,” or “strongly” (ordinal scale); or (3) to use an analog scale (e.g.,

20
an underlying dimension from 0 to 100) to indicate exactly how much or how intensely the emo-
tion was experienced (interval scale). Methods vary on whether respondents are to respond choos-
,
ss
ing only the most pertinent emotion category, two or more categories to indicate possible blends,
e
or all categories in a list (replying with “no” or “0” for the emotions that are not at all appropriate
Pr

to describe the experience). Some standardized instruments of this kind have been proposed in
y

the literature (e.g., the Differential Emotion Scale; Izard, 1991). However, most researchers create
it
rs

ad hoc lists of emotion categories that seem pertinent for a specific research aim, without worry-
ve

ing too much about the representativeness of the chosen list or how well results obtained with the
i

specific list may compare to results obtained with other emotion lists.
Un

The results obtained with the emotion label approach are generally highly plausible and easily
d

interpretable, given that widely shared and frequently used natural language labels tend to be em-
or

ployed. However, it is often difficult or even impossible to compare results across different studies
xf

in which widely different sets of emotion labels have been used. In addition, the statistical analysis
)O

of these data is hampered by the abundance of missing data and the difficulty to analyze and inter-
pret the frequently reported emotion blends (Larsen et al., 2009; Scherer, 1998; Scherer & Ceschi,
(c

2000). Often, problems are encountered of differential familiarity of respondents with particular
emotion words, as well as differential interpretation of the meaning of the terms provided by the
researcher. In addition, there are other problems with a discrete emotion response format, such as
confusions (e.g., in the case of very extensive word lists), order effects, and other types of artifacts
such as demand characteristics (e.g., the choice of specific emotion words may give away the re-
search aim).
The dimensional approach in the self-report assessment of emotional experience was pioneered
by the German psychologist Wilhelm Wundt (1896), who used introspection to develop a struc-
tural description of subjective feelings consisting of their position in a three-dimensional space
formed by the dimensions of valence (positive–negative), arousal (calm–excited), and ten-
sion (tense–relaxed). This proposal has had a strong impact, both on the measurement of feeling
(e.g., Schlosberg, 1954) and on the assessment of emotional connotations of language concepts in
general (e.g., Osgood, Suci, & Tannenbaum, 1957). The theoretical foundations of this approach
and its recent research development are discussed in detail in Chapter 2 of this book, showing that
this domain of inquiry is currently dominated by a two-dimensional valence × arousal model.

19-Fontaine-Chap18.indd 282 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 283

To obtain a self-report of feeling with this approach, respondents are typically asked to indicate
how positive or negative and how excited or aroused they feel (either in two separate steps, or
by providing a two-dimensional surface and asking the respondent to determine the appropriate
position in it; Larsen et al., 2009). As a result, the emotional feeling of the person is described by
a point in this valence × arousal space. In some cases, respondents are also asked to separately
evaluate the positive and negative parts of the valence scale (see Chapter 2). In other cases, three
dimensions (valence, arousal, and dominance) are assessed (e.g., Self Assessment Manikin
Test, Bradley & Lang, 1994). This simple and reliable method lends itself to advanced statistical
processing due to the general use of interval scaling. However, the information obtained is limited
to the degree of positive or negative feeling and the degree of felt bodily excitation. Furthermore,
the dimensional approach does not allow differentiating intensity of feeling from bodily excitation,
which are clearly different constructs. For example, while intense anger is likely to be characterized
by high arousal, intense sadness is often characterized by very low arousal. Thus, mild sadness
and intense sadness could not be differentiated based on a valence × arousal space alone, as the
arousal level is low in both cases. This is a problem for researchers who are interested in clearly
differentiating emotions like sadness and depression or grief. Obviously, any attempt to reduce

13
positions in a multidimensional space to a two-dimensional representation will face this problem.

20
Another disadvantage is that, while most lay persons have little problem evaluating the positivity
or negativity of a feeling (or event) and the approximate degree of their felt arousal, the resulting
,
ss
point in two-dimensional space has no specific meaning for them and cannot be communicated
e
to others. It would seem very strange to tell someone that I feel 2.3 positive and 1.6 aroused. Emo-
Pr

tional sharing (Rimé, 2009) is an important social phenomenon, directly linked to the adaptive
y

function of emotion communication through expression, and this function would not be well
it
rs

served by a two-dimensional metric.

ve

Most importantly, however, in the two-dimensional valence × arousal space several rather
i

different emotions are close neighbors. Figure 18.1 reproduces the mapping of the terms Rus-
Un

sell (1983) used as markers for his emotion circumplex in two-dimensional valence-arousal
d

space (original terms used by Russell in capital letters). In this figure, emotions as different as
or

anger and fear, and other related terms, appear located in the immediate vicinity, as they all share
xf

high negative arousal. In consequence, dimensional ratings on the valence × arousal space
)O

by themselves do not seem advisable if a researcher is interested in diagnosing the quality of a

particular feeling. Reisenzein (1994) reviews this problem and suggests using appraisal theory to
(c

disambiguate the quality of neighboring states in valence × arousal space (Gehm & Scherer,
1988). As shown in this volume, the use of a discrete emotion label provides at least probabilistic
information on the prototypical appraisal patterns differentiating the respective emotions. While
appraisal profiles obviously provide the means for a more fine-grained differentiation (Scherer,
Schorr, & Johnstone, 2001; and Chapters 1 and 12), this alone does not solve the problem of the
lack of an appropriate self-report instrument that is precise, valid, and economical.
Clearly, the two classic self-report approaches reported above have both advantages and disad-
vantages. Given the central role of emotion self-report in emotion research, it is surprising that
relatively few attempts have been made so far to develop new instruments to avoid some of their
shortcomings. In particular, it might be worth investing in the development of an instrument ca-
pable of combining the advantages of the precise differentiation provided by natural language la-
bels with the simple organizational structure afforded by a two-dimensional space. One possibility
would be to use discrete emotion labels and arrange them graphically in a two-dimensional affect
space, allowing the user to rapidly orient in this space. Such an instrument would be useful for a
wide range of research goals.

19-Fontaine-Chap18.indd 283 7/17/13 3:45 PM

 284 COMPONENTS OF EMOTIONAL MEANING

Active
+ bellicose
adventurous + AROUSED
ASTONISHED ALARMED+ hostile
+ lusting TENSE + hateful
+ envious
High Power/Control + trimphant ANGRY AFRAID Obstructive
EXCITED + enraged + defiant
+ self-confident + ambitious + conceited ANNOYED + contemptueus
+ courageous + angry
+ feeling superior + jealous
DISTRESSED
+ indignant + disgusted
+ convinced + loathing
DELIGHTED FRUSTRATED
enthusiastic + + disoontented
+ elated + light-hearted + impatient
+ determined + suspicious + bitter
+ amused excited + + insulted
HAPPY + passionate
+ joyous
+ distrustful
+ expectant
Positive + interested + bored startled + Negative
+ feel well + impressed + disappointed
PLEASED MISERABLE
+ amorous astonished + + apathetic
GLAD + dissatisfied
+ confident
+ taken aback
+ content + hopeful + despondent

13
+ worried + uncomfortable
+ relaxed + despondent SAD

20
+ solemn + longing feel guilt +
+ attentive DEPRESSED GLOOMY
SERENE languid + + desperate
+ ashamed
CONTENT + friendly
pensive +
,
ss + embarrassed
AT EASE + contemplative
SATISFIED polite + + serious + melancholic
RELAXED + wavening + lonely
e
CALM + hesitant
Pr

Conducive peaceful + + consceintious BORED anxious Low Power/Control

sad + + dejected + insecure
y

empathic + DROOPY
it

reverent + + doubtful
rs

SLEEPY TIRED
Passive
ve

Figure 18.1 Alternative dimensional structures for the semantic space of emotions.
i
Un

Reproduced from Social Science Information, 44(4), What are emotions? And how can they be measured?
Scherer, K. R., pp. 695–729, Figure 1 © 2005, Sage Publications, with permission.
d
or
xf

18.2 The Geneva Emotion Wheel—design features

)O

Based on these considerations, the Geneva Emotion Wheel (GEW) was designed to combine a dis-
(c

crete and a dimensional approach in the self-report assessment of emotion. It consists of a theory-
based circular arrangement of discrete emotion terms in two-dimensional space according to the
following design criteria (Scherer, 2005):
1 Two dimensions form the underlying structure of the instrument— valence and
control/power
2 The instrument uses standard emotion labels from natural languages to capitalize on respond-
ents’ intuitive understanding of the semantic field of emotion terms
3 Emotion terms are displayed in a systematic fashion by aligning them with respect to the
underlying dimensional structure
4 The instrument allows the systematic assessment of the intensity of feelings
5 The instrument’s user-friendly graphical design allows the respondent to rapidly understand
the principle and use the instrument in a reliable fashion.
Below we describe how each of these design criteria was addressed in the development of the GEW:
1 Choice of dimension. There can be little doubt that valence (positive–negative or pleasant–
unpleasant), separating positive and negative emotions, constitutes the most important

19-Fontaine-Chap18.indd 284 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 285

dimension of affective space. The decision on the second dimension is less obvious. Despite
the fact that, as shown above, the advocates of activation or arousal as the second major
dimension have dominated the scene for the last few decades, there are a number of drawbacks
to this choice. As shown above, an arousal dimension is very limited in its usefulness to dif-
ferentiate between emotions due to the fact that most emotion families have several members
that differ with respect to their degree of arousal. Furthermore, it is not always clear exactly
what is meant by the terms “arousal” and “activation.” While arousal generally refers to
sympathetic arousal in the sense of the dominance of the sympathetic branch of the autono-
mous nervous system, activation is often used in connection with motivation or action ten-
dencies, not necessarily requiring a high degree of sympathetic arousal.
Beyond arousal/activation, there are of course alternative dimensions that can be chosen as
a second dimension after valence. For example, Wundt (1905) proposed “tension–relaxation”
and Schlosberg (1954) “attention–rejection.” Most importantly, Osgood and collaborators (1957)
highlighted the importance of a potency (or power, dominance) dimension in their seminal work
on the semantic differential, placing this dimension second after evaluation (valence), and be-

13
fore activity (activation, arousal). This fits very well with an appraisal account of emotion. If
emotions are indeed elicited and differentiated by appraisal patterns (see Chapter 1), the structure

20
of the emotional space should be largely determined by the major appraisal dimensions. The close
,
link between the appraisal checks “intrinsic (un)pleasantness” and “goal (in)consistency” or “goal
ss
conduciveness/obstruction” on the one hand, and the valence dimension on the other, is obvi-
e
Pr

ous. The same is true for the link between the power/potency dimension and the coping potential
check (which determines the degree of control available to the individual in a situation, as well
y
it

as the power available to exercise that control; see Scherer, 1984b). As shown by numerous stud-
rs

ies (see reviews in Ellsworth & Scherer, 2003; Scherer, Schorr, & Johnstone, 2001), the appraisal
ve

dimensions that seem to have the strongest impact on emotion differentiation are indeed goal
i
Un

conduciveness (representing valence) and coping potential (control/power), corresponding

to Lazarus’ (1968) pioneering distinction between primary and secondary appraisal. Obviously,
d

differences in arousal/activation and intensity are also important determinants of feeling but
or

they are more likely to define differences within an emotion family rather than between emo-
xf

tion families. In consequence, it was decided to use a two-dimensional affect space including the
)O

dimensions of valence (based on pleasantness and goal conduciveness appraisals) and power/
(c

potency/control (based on coping potential appraisals) to organize the discrete emotion labels to
be measured by the GEW.
Scherer (2005) further justified this theory-driven decision by arguing that this two-
dimensional structure fits the organization of emotion terms in the two-dimensional space
obtained, for example, through the analysis of similarity ratings. He provides an example, repro-
duced in Figure 18.1, in which a two-dimensional structure (conducive/obstructive × high/low
control–power) found for 80 German terms (reported in Scherer, 1984b, pp. 47–55) is super-
imposed on the item distribution of English terms obtained by Russell (1983). Th e figure shows
that both a classic valence by arousal structure (as postulated by Russell) and a valence
(conducive/obstructive) by control/power structure (as based on an appraisal model) can be
justified, as the respective axes are just rotated by 45°. The latter structure provides a theoreti-
cally more homogeneous solution, as both factors represent the two major appraisal criteria.

2 Choice of emotion families. It is difficult to decide a priori, on purely theoretical grounds,

which words, labels, or expressions should be chosen to represent the discrete states within the
different regions of the continuous two-dimensional space spanned by valence and power/
potency. In the interest of replicability of results across studies, it was attempted to choose a

19-Fontaine-Chap18.indd 285 7/17/13 3:45 PM

 286 COMPONENTS OF EMOTIONAL MEANING

standard set of emotion families that met three conditions: (1) having been frequently used
in past research (to ensure their utility in the assessment instrument), (2) covering most of
the regions or segments of the two-dimensional space (to be able to map most positions in
affective space), and (3) affording an arrangement of the terms around the rim of the wheel
in approximately equal distances. As one might imagine, it is virtually impossible to find a
set of emotion terms that equally satisfies all three conditions. In consequence, a pragmatic
approach was adopted in the design of the GEW, giving greater weight to the first criterion to
achieve a compromise that was satisfactory from the standpoint of potential users. Note that
to accommodate users with specific needs, it is accepted that they replace part of the standard
terms with terms of their choice, provided that the terms used are differentiated by valence
and control/power and can be reasonably represented on a circle in this space.
3 Circular arrangement of emotion terms. In the GEW, emotion terms representative
of major emotion families are arranged in a circle (see Figure 18.3). The two underlying
dimensions, valence and control/power, separate the wheel into four quadrants: Negative
valence–low control/power, negative valence–high control/power, positive valence–

13
low control/power, and positive valence–high control/power. The position of the emotion
terms in these clusters should correspond to their values on the valence and control/power

20
dimensions.
,
ss
4 Intensity ratings. Members of any one specific emotion family can be expected to vary among
each other with respect to intensity (e.g., irritation–anger–rage), which, as argued above, may
e
Pr

correlate with, but is not the same as, physiological arousal. It was, therefore, decided to map
intensity on the distance between the rim and the hub of the wheel, representing the intensity
y
it

of a specific emotional experience as the distance of its position from the central point in the
rs

valence–control/power space (see also Reisenzein, 1994; Russell, 1980, p. 1170).

ve

5 Ease of use. The wheel interface is easy to understand. Participants are asked to identify an
i
Un

experienced or imagined emotion among the various options provided. They are also asked to
rate its intensity on the basis of the distance from the hub of the wheel, which implies choos-
d
or

ing one of the answer circles increasing in size from the hub to the rim (the larger the circle,
xf

the more intense the emotion is reported to be). Thus, the meaning of the response options is
)O

quite intuitive. Also, in the interest of reading ease, the number of emotion families is limited.
Finally, the alignment of the emotion terms based on the underlying dimensions should facili-
(c

tate the usability of the GEW.

Note that the resulting instrument, although conceived in a very different fashion and for a rather
different purpose, corresponds in several aspects to the various proposals of personality assess-
ment instruments based on the notion of an “interpersonal circumplex”, with the dimensions of
“warmth vs hostility” or “love vs hate ‘on the one hand and’ dominance vs submission” on the other
(Leary, 1957; Wiggins & Trobst, 1997). The two dimensions of nurturance/hostility (or warmth/
coldness) and dominance/submission are highly comparable to valence and control/power
(which has often been called dominance in the literature). The arrangement of words in the GEW
turns out to be also very similar to the emotion distribution in Plutchik’s (1980b) emotion cir-
cumplex color wheel (see Figure 18.2), even though this theorist started from the notion of adap-
tation-oriented basic emotions. One might almost surmise that he arranged the emotions around
the circumplex with an implicit valence by power structure in mind.
In what follows, two stages of the development and of the investigation of the structural validity
of the GEW will be described.

19-Fontaine-Chap18.indd 286 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 287

opt
serenity

e
imi

lov
sm
interest acceptance
joy

agg anticipation trust

ress ecstasy ion
iven miss
ess sub
vigilance admiration

annoyance anger rage terror fear apprehension

13
20
loathing amazement
t
emp grief
, aw
ss
t e
con
surprise
e
disgust
Pr

boredom sadness
y

distraction
it
dis
rs
e
ors

app
ve
rem

pensiveness
rov
i

al
Un
d
or
xf

Figure 18.2 Plutchik’s emotion circumplex.

)O

Reproduced from American Scientist, 89, Plutchik, R., The nature of emotions: Human emotions have deep
evolutionary roots, a fact that may explain their complexity and provide tools for clinical practice, pp. 344–350
(c

© 2001, The Scientific Research Society. Reprinted by permission of American Scientist, magazine of Sigma Xi,
The Scientific Research Society.

18.3 Stage 1: Development and structural validation of prototype

versions of the GEW
The first prototype (Version 1.0) of the GEW (see Figure 18.3) was developed as a tool for the ver-
bal report of emotions in a study of email communication (financed by the Gottlieb Daimler and
Karl Benz Foundation). In this version, four emotion families were presented per quadrant, yield-
ing a total of 16 emotions in the wheel (which seemed reasonable considering that the number of
“basic emotions” is often considered to be somewhere between 6 and 14). In this version, a separate
word (adjective) was proposed for each level of intensity response option within one emotion fam-
ily (e.g., vexed, irritated, angry, enraged for the anger family; these are not visible in Figure 18.3,
they appeared when the mouse cursor passed over the circles). The choice of the concrete families

19-Fontaine-Chap18.indd 287 7/17/13 3:45 PM

 288 COMPONENTS OF EMOTIONAL MEANING

How did you feel when YOU WROTE this message?

high control
anger pride

hostiltiy elation

contempt happiness

satisfcation
disgust

positive
negative

relief
shame/guilt

13
boredom hope

,20
ss
sadness interest
e
anxiety surprise
Pr

low control
Figure 18.3 Version 1.0 of the Geneva Emotion Wheel. Reproduction of a VisualBasic screen dump;
ity

intermediate labels occurred when the mouse cursor passed over the circles.
rs

Reproduced from Social Science Information, 44 (4), What are emotions? And how can they be measured?
ve

Scherer, K. R., pp. 695–729 © 2005, Sage Publications, with permission.

i
Un
d
or

was largely determined by what are generally considered to be basic or fundamental emotions,
xf

frequently studied in the field.

)O

Based on pilot studies with Version 1.0 of the GEW, a second version (Version 2.0, see Figure 18.4)
was developed with two words (relatively close synonyms) referring to each of the 20 emotion
(c

families; the goal was to emphasize that each response option represented an emotion fam-
ily rather than individual emotions. Furthermore, as the gradation of intensity levels by four
different adjectives from the semantic field of the emotion family in the prototype Version
1.0 proved problematic (in terms of reliability and translatability of the gradation differences;
Bänziger, Tran, & Scherer, 2005), in Version 2.0 the different intensity response options within
one emotion family were therefore represented only with unlabeled circles of different sizes.
Also, some emotions were placed in slightly different positions based on the results of the
initial studies (e.g., “interest” passed to a somewhat higher position on the control/power
dimension).
In several studies, the structural validity of placing the emotion terms in the GEW Versions 1.0
and 2.0 was assessed (see Bänziger, Tran, & Scherer, 2003, 2005; Sacharin, Schlegel, & Scherer,
2012). The results of these studies for the respective GEW Versions address similar issues and
are, therefore, presented together below. Additionally, studies performed in our lab and other labs
examined how well the GEW Versions 1.0 and 2.0 fared compared to other measurement tools
(Caceido & van Beuzekom, 2006; Tran, 2004).

19-Fontaine-Chap18.indd 288 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 289

Irritation Involvement
Anger Interest

Contempt Amusement
Scorn Laughter

Disgust Pride
Repulsion Elation

Envy Happiness
Jealousy Joy

Disappointment No emotion Enjoyment

Regret felt Pleasure

Guilt Other emotion Tenderness

Remorse felt Feeling love

Embarrassment Wonderment
Shame Feeling awe

13
Worry Feeling disburdened

20
Fear Relief

,
ss
Sadness Astonishment
Despair Surprise
e
Pr

Pity Longing
y

Compassion Nostalgia
it
rs

Figure 18.4 Template of Version 2.0 of the GEW (distributed until March 2013 via the website
ve

http://www.affective-sciences.org/researchmaterial and replaced by the new version 3.0 described

i

in this chapter).
Un
d

18.4 Methods
or
xf

Two paradigms were used to assess the validity of the structure of the original version of the GEW:
)O

(1) similarity ratings of emotion words, and (2) direct ratings of the position of these words on the
valence and control/power dimensions. For Version 1.0, Bänziger et al. (2005) performed several
(c

validation tasks with a sample of 28 native English and 31 native French speakers. In a first task, re-
spondents performed pair-wise similarity ratings of the 16 emotion family nouns positioned around
the circumference of the wheel. In a second task, the 64 adjectives (16 × 4) representing the intensity
gradations within families were sorted into the 16 family categories. In a third task, a subsample of 14
English and 15 French speakers rated the 16 emotion family nouns and 64 adjectives by using a dedi-
cated graphic interface (enlarging a circle on the screen to rate intensity and moving markers within
a two-dimensional space to rate valence and control/power). Another subsample (14 English and
16 French speakers) rated the intensity, valence, and control/power associated with each of the 16
emotion nouns and 64 adjectives on continuous rating scales using a mouse-operated slider. For Ver-
sion 2.0, Sacharin et al. (2012) examined ratings of valence and control/power on 11-point scales
in an online study with 40 native English speakers. In this study, valence was defined as follows: “the
situation is experienced as (un)pleasant and enjoyable (disagreeable) and/or is likely to have positive
and desired (negative and undesired) consequences for the person.” In turn, control/power was de-
scribed as “the person believes that he/she can (cannot) influence the situation to maintain or improve
it (if desired).”

19-Fontaine-Chap18.indd 289 7/17/13 3:45 PM

 290 COMPONENTS OF EMOTIONAL MEANING

To examine the usability of the GEW, Tran (2004) compared the emotion assessment—first by
using the GEW with ratings of emotions presented as word lists in a sample of 80 business school
students, and later by administering the GEW in a simulation study with 106 managers attending
executive development seminars in four to seven person teams. Caceido and van Beuzekom (2006)
directly compared the utility of the GEW with the utility of the PrEmo (Desmet, 2003), a graphical
measure for the assessment of discrete emotions.

18.5 Results
With regard to the structural validity of the GEW Version 1.0 (Bänziger et al., 2005), the categori-
zation task showed that the adjectives representing intensity differences within an emotion family
were almost all correctly classified. However, the agreement on the intensity gradation was less
than perfect. With regard to the placement of the main emotion family labels in the valence and
power space, the ratings of similarity were submitted, for each language sample separately, to mul-
tidimensional scaling (MDS) analyses, using an ordinal model and Euclidian distances. The follow-

13
ing fit measures for a two-dimensional solution were computed: English sample, stress = 0.34, RSQ

20
= 0.52; French sample, stress = 0.34, RSQ = 0.39. The arrangement of the 16 emotion families in
the two dimensions largely confirmed the theoretical prediction with respect to the overall cluster-
,
ss
ing of the emotions in high/low power and high/low valence. Exceptions were “interest” in the
e
English sample and “relief ” (soulagement) in the French sample, which were empirically found to
Pr

be placed in the high control/power quadrant rather than the predicted low control/power quad-
y

rant. Similar to the MDS analyses, the data from the direct ratings of valence and control/power
it
rs

supported the differentiated alignment of the emotions along the wheel very well for the valence
ve

dimension. However, the empirical alignment of the emotions on the power/control dimension
i

corresponded less well to the theoretical predictions. This was particularly true for the negative
Un

emotions. These emotions were also rated less reliably on the control/power dimension as indi-
d

cated by large standard deviations.

or

The results of the Sacharin et al. (2012) study can be summarized as follows: for the valence
xf

ratings, 19 out of 20 negative emotions and 15 out of 20 positive emotions were rated as predicted.
)O

Among the negative, only “compassion” was rated as more positive than expected. Among the posi-
tive, “nostalgia,” “longing,” “feeling disburdened,” “astonishment,” and “involvement” were not sig-
(c

nificantly different from the expected position. The results for the control/power ratings are more
problematic—only 8 out of 20 high control/power emotions (all positive emotions) and only 1 out
of 20 low control/power emotions (“sadness”) were rated as predicted. After computing the mean
valence and control/power ratings for each word across raters, control/power and valence
were positively associated as reflected in a positive correlation of valence and control/power rat-
ings across the 40 word ratings, r (40) = 0.718, p <0.001. Furthermore, control/power ratings had
greater standard deviations than valence ratings (2.89, 1.51), t (39) = 11.68, p <0.001. This is in
line with the findings reported by Bänziger et al. (2005) with respect to the difficulty of finding the
predicted arrangement for the negative, low control/power emotions.
Studies examining the usability of the GEW for emotion assessment compared with other
measures showed that the same feelings are associated with different vignette scenarios when
using the nouns in GEW Version 1.0 as word lists (Tran, 2004). For the use with managers in
a quasi-naturalistic environment, very high response rates were obtained in daily assessments
for 8 to 10 days, providing an indicator of the managers’ strong involvement in the emotion as-
sessment task. The GEW was judged to be a particularly useful measurement instrument under

19-Fontaine-Chap18.indd 290 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 291

time pressure and with repeated measurements (Tran, 2004). Furthermore, in Caceido and van
Beuzekom’s (2006) study, respondents overall preferred the GEW over the PrEmo, and judged
the GEW as clear to understand, useful to differentiate between emotions, and appealing in its
visual design.

18.6 Discussion
The studies examining the structural validity of the emotion terms in the valence–control/power
space showed that for GEW Versions 1.0 and 2.0 the placement of the emotion terms along the
valence dimension generally corresponded to prediction. To improve the representation, “com-
passion,” which had been rated as a positive emotion, was to be moved from the negative to the
positive side of the GEW.
In contrast to the findings for the valence dimension, the alignment of emotion terms on
the control/power dimension differs depending on the response paradigm. In the similarity
study with MDS analyses, though the model fit was not very good and the sample size was small,

13
the placement of the emotion terms corresponded well to the predicted alignment. In contrast,

20
the empirical data derived from rating studies tend to deviate from the predictions, especially
in the negative—low control/power quadrant. Inspection of means and variance suggest that,
,
ss
to some extent, the rating results could reflect a response bias in the use of the control/power
e
dimension by some participants, resulting in a large variance (Bänziger et al., 2005; Sacharin
Pr

et al., 2012).
y

Using a larger sample size or re-wording the description of what control/power means might
it
rs

ameliorate this problem. The discussion of the results for the appraisal component in the GRID
ve

study (see Chapter 12) also showed that the wording of the control/power features in the ques-
i

tionnaire may not have been optimal, and changes are proposed for the CoreGRID and further
Un

studies with the full instrument (see Chapter 44). However, it remains to be seen if the changes
d

in wording produce the desired effect. It may well be that the abstract notions of “control” and
or

“power” in connection to emotions are not easy to grasp for non-psychologists, and that it is thus
xf

difficult to obtain explicit and reliable ratings for these concepts.

)O

An alternative explanation for the observed difficulty in obtaining the theoretically predicted
alignment on the power dimension (especially for negative high-power and positive low-power
(c

emotions) is provided by the strong association of valence with control/power ratings (Sacharin
et al., 2012). Indeed, it has recently been suggested that control/power appraisals are valenced
(Shuman, Sander, & Scherer, 2013; Scherer, 2010b). High power is associated with positive affect
and low power with negative affect (e.g., Keltner, Gruenfeld, & Anderson, 2003). A similar cor-
relation was found in the results of the GRID study, as shown in Chapters 8 and 12, suggesting a
strong ecological correlation between negative valence and low control/power.
Thus, even with a larger sample size and revised wording, it may not be possible to obtain inde-
pendent and fine-grained ratings of emotions on the control/power appraisal criterion alone. To
empirically grasp the notion that negative and positive emotional experiences can be associated
with more or less control/power, other methods may be needed that measure not only control and
power, in general, or specific appraisals related to it, but additional components of the emotional
experience, such as action tendencies.
A further substantiation of the structure of the GEW is timely because the GEW is increasingly
used due to its user-friendliness in comparison with other instruments (Caceido & van Beuze-
kom, 2006). The GEW has been applied in a variety of contexts, such as consumer attitudes to

19-Fontaine-Chap18.indd 291 7/17/13 3:45 PM

 292 COMPONENTS OF EMOTIONAL MEANING

internet videos and industrial design products (e.g., Bardzell, Bardzell, & Pace, 2009; Caceido &
van Beuzekom, 2006), the affective evaluation of body movements and vocalizations (e.g., Beck,
Stevens, & Bard, 2009; Pammi & Schröder, 2009), emotions during learning in virtual environ-
ments and in virtual environments with different illumination (e.g., Longhi, Pereira, Brecht, &
Behar, 2009; Santos, 2008), and experience sampling studies of emotions in everyday life (e.g.,
Tschan, Semmer, Messerli, & Janevski, 2010).
Furthermore, the GEW has been applied to assess emotions at different levels of analysis rang-
ing from the individual and the team level emotions of managers (Tran, 2004) to the emotional
climate in a hospital (Wittgenstein, 2008). Specifically, Tran (2004) found that the emotions re-
corded in the different quadrants of the GEW differentially influence team decision making and
cohesion. For example, negative–low control/power emotions were positively associated with
team cohesion, and negative–high control/power emotions were negatively associated with team
cohesion. Furthermore, differentiating between the different intensity levels within an emotion
family in the GEW contributed important information. For example, moderate levels of positive–
low control/power emotions were positively associated with alternative evaluation (a key com-
ponent of decision making), whereas high intensity positive–low control/power emotions were

13
negatively associated with alternative evaluation. Finally, Tran’s work showed that the GEW can be

20
used as a means to help develop team processes. Over the course of her study, it was observed that
participants often used the GEW ratings as a basis to discuss their emotions with their colleagues,
,
ss
yielding self-awareness and awareness at the group level. In addition, by mapping their emotions
e
on the GEW on a regular basis, all group members can see the evolution of the emotional climate
Pr

and can proactively manage it (Tran, Páez, & Sánchez, 2012).

y

To conclude, the existing studies on the GEW Versions 1.0 and 2.0 underscore the utility of the
it
rs

instrument. The placement of the emotion terms in each quadrant of the GEW, however, could not
ve

be satisfactorily justified to date. MDS analyses of similarity ratings indicate that the placement
i

of the emotions predicted in each GEW quadrant are valid, but direct rating studies of the va-
Un

lence and control/power of the same terms were unable to yield the expected placement on the
d

control/power dimension. Ratings of control/power, it seems, have to be worded in drastically

or

different ways to obtain results that reflect the MDS results.

xf
)O

18.7 Stage 2: Development of a standard version of the GEW

(c

based on structural validation with the GRID instrument

After the first results of the GRID study had confirmed that the two major dimensions of the se-
mantic emotion domain are indeed valence and control/power (Fontaine et al., 2007), it became
obvious that the GRID paradigm could constitute a royal road to finalize the validation of the GEW
for a set of major emotion terms. As demonstrated in the chapters of Part III of this volume, the
GRID results clearly establish the existence of a four factor structure for the emotion space, with
activation/arousal and novelty being essential additional factors for a satisfactory mapping of
major emotion terms with respect to their discriminability in low-dimensional space. However, we
decided to stick to a two-dimensional representation of the emotion terms in a wheel structure for
the assessment instrument, as a three-dimensional representation on a two-dimensional paper or
screen surface is confusing, and a four-dimensional representation would require a series of inde-
pendent two-dimensional graphs. Such formats are inacceptable for a self-report instrument that
needs to be immediately obvious to use and economical in terms of time investment. As the two
first dimensions emerging in PCAs of the GRID data, valence and power also explain the largest
percentage of the variance between emotions, this solution seems well justified.

19-Fontaine-Chap18.indd 292 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 293

The GRID study described in this volume provides an ideal framework to obtain similarity met-
rics for emotion words. The GRID provides a very comprehensive feature profile consisting of 142
features covering all components of emotion. In consequence, the assessment of similarity is based
on a very rich set of criteria. Furthermore, the information is obtained for a very large set of lan-
guages and cultures using sizable groups of native speakers. In consequence, all requirements to
obtain a definitive validation of the placement of the emotion terms around the circumference of
the wheel are fulfilled. Most importantly, the power dimension emerges as the second strongest
dimension from the PCA and is clearly identified even if the appraisal component is not included
in the analysis. This consistent with the original choice of control/power as the second factor in
the two-dimensional structure of the GEW (in contrast to arousal in dimensional theories). It
is also consistent with the assumption that the appraisals of control and power strongly affect the
other components producing clear changes, such as dominant action tendencies and loud voice that
are sufficient to determine a clear, overall power factor. In consequence, the use of valence and
power coordinates derived from the GRID data seem to be an ideal solution to solve the issue of
validating the predicted arrangement of the GEW emotion terms in two-dimensional valence by
control/power space.

13
To this end, it was necessary to obtain additional ratings for words used in the GEW that were

20
not rated in the basic GRID study. The list of 24 GRID emotion words (see Chapter 5 for the
criteria of choice) already contained 16 of the 20 items that had been regularly used with the
,
ss
previous versions of the GEW. Four words were missing: “amusement,” “admiration,” “relief,”
e
and “regret.” Therefore we contacted the different collaborators in the GRID study and asked
Pr

them whether they would be willing and able to have these words judged on the 142 features in
y

the same way and using similar groups of participants as for the regular GRID questionnaire.
it
rs

Groups in ten different countries—Switzerland (French), United Kingdom (English), Belgium

ve

(Flemish), China (Mandarin Chinese), Germany (German), Estonia (Estonian), Finland

i

(Finnish), Italy (Italian), Japan (Japanese), Poland (Polish)—agreed to participate and collected
Un

the data using exactly the same procedures as described in Chapters 5 and 6 (see Table 18.1 for
d

sample characteristics).
or
xf
)O

Table 18.1 Sample characteristics

(c

Language Country Region N Age Format Compensation

Total F M Range Mean SD
French Switzerland Geneva 20 12 2 20–45 29.21 8.285 online course credit

English UK Belfast & York 19 10 9 18–22 19.32 1.416 online course credit
Dutch Belgium Gent 15 9 6 21–23 21.87 0.640 online course credit
Chinese China Hong Kong 19 6 11 19–25 21.24 1.602 online course credit
German Germany Berlin 20 19 1 19–37 22.10 4.241 online course credit
Estonian Estonia Tartu 15 8 7 22–29 25.67 1.759 online course credit
Finnish Finland various 18 17 1 18–37 25.33 5.087 online course credit
Italian Italy Bologna 19 12 2 19–26 23.36 1.946 online course credit
Japanese Japan Sapporo 15 9 6 18–21 18.87 1.125 online course credit
Polish Poland Lodz 14 7 7 19–38 23.36 4.749 online course credit

19-Fontaine-Chap18.indd 293 7/17/13 3:45 PM

 294 COMPONENTS OF EMOTIONAL MEANING

In consequence, the new standard version of the GEW presented here contains the following 20
emotion words (the asterisks indicating the four words rated specifically for the GEW validation):
• admiration*
• amusement*
• anger
• compassion
• contempt
• contentment
• disappointment
• disgust
• fear
• guilt
• hate

13
• interest

20
• joy
• love ,
e ss
• pleasure
Pr

• pride
y

• regret*
it
rs

• relief*
ve

• sadness
i
Un

• shame
We first analyzed the reliability of the rating data for the four new terms for each country (fol-
d
or

lowing the procedure described in Chapter 6). These data were then combined with the data for
xf

the 16 GEW terms that were part of the basic GRID study and for which the reliability had been
)O

assessed previously. Then, dissimilarity matrices were produced for each of the country samples by
computing the distances between the feature profiles of the different words. These matrices were
(c

then combined and submitted to MDS (using Proxscal) across countries. Proxscal computes solu-
tions for different dimensionalities. As we are interested in a two-dimensional arrangement of the
emotion terms for the GEW instrument, here we report only the results for the two-dimensional
solution. The stress and fit measures for the two-dimensional solution are as follows: Normal-
ized Raw Stress 0.00481, Stress-I 0.06939, Stress-II 0.13588, S-Stress 0.00503, Dispersion Ac-
counted For (D.A.F.) 0.99519, Tucker’s Coefficient of Congruence 0.99759. These fit indices can
be judged as quite satisfactory in the light of the levels expected according to the MDS literature.
Table 18.2 shows the respective coordinates for the 20 words on the two dimensions (shown as a
two-dimensional plot in Figure 18.5).
The two dimensions underlying the plotted coordinates of the 20 GEW words can be readily inter-
preted—the horizontal dimension corresponding to a general valence dimension (highly compa-
rable to the valence dimension in the overall GRID analysis—see Chapter 7), separating the group
of positive emotions from that of the negative ones. The vertical dimension is not as immediately
obvious, but comes very close to the power dimension in the overall GRID analysis (see Chap-
ter 7). Both of these dimensions correspond directly to the dimensions that have been theoretically

19-Fontaine-Chap18.indd 294 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 295

Table 18.2 Coordinates of the 20 GEW emotion words in

two-dimensional VALENCE by POWER space

Emotion words VALENCE POWER

admiration 0.66 −0.09

amusement 0.67 0.19

anger −0.37 0.47
compassion −0.05 −0.55
contempt −0.55 0.43
contentment 0.77 −0.03
disappointment −0.77 −0.12
disgust −0.68 0.20
fear −0.61 0.07
guilt −0.57 −0.27

13
hate −0.45 0.43

20
interest 0.61 0.25
,
ss
joy 0.68 0.07
e
love 0.58 −0.16
Pr

pleasure 0.71 0.02

y
it

pride 0.72 0.15

rs

regret −0.70 −0.19

ve

relief 0.66 −0.36

i
Un

sadness −0.68 −0.35

d

shame −0.61 −0.16

or
xf
)O

postulated for the GEW. They seem to be closely related to two types of appraisal: on the one hand,
(c

appraisals of unpleasantness and obstructiveness vs pleasantness and conduciveness (underlying

the valence dimension), and on the other hand, appraisals of high vs low control/power (for the
power dimension) (see Chapter 12). We also computed a PCA of these data to compare the MDS
plots to factor score plots. Essentially, the PCA extracted four factors, the first two of which again
correspond to valence and power. The two-dimensional factor score plot yielded coordinates for
the 20 emotion words that were very similar to the MDS solution. A comparison of the positions of
the 20 emotion words in the two-dimensional MDS space shows that the patterning is very compa-
rable to the placement of the same words in the previous Version 2.0 of the GEW shown in Figure
18.4, thus empirically confirming the earlier arrangement, which was based exclusively on theoreti-
cal considerations.
Based on these results, a new version (3.0) of the GEW was constructed. Concretely, the two-
dimensional coordinates shown in Table 18.2 and Figure 18.5 were projected onto the rim of the
circle that represents the wheel structure and arranged equidistantly around the circumference of
the wheel. The result of this empirical GRID-based circular ordering of the 20 terms, GEW Version
3.0, is shown in Figure 18.6.

19-Fontaine-Chap18.indd 295 7/17/13 3:45 PM

 296 COMPONENTS OF EMOTIONAL MEANING

0.6

anger
hate
0.4 contempt

interest
0.2 disgust amusement
pride

fear joy
POWER

pleasure
0
contentment
admiration
disappointment
shame love
–0.2 regret
guilt

sadness relief
–0.4

13
compassion

20
–0.6
–0.8 –0.6 –0.4 –0.2 0 0.2 0.4 0.6 0.8
VALENCE
,
ss
Figure 18.5 Representation of the 20 GEW emotion words in a two-dimensional VALENCE by POWER
e
Pr

space.
yit
rs

Anger Interest
ive

Hate Amusement
Un
d

Contempt Pride
or
xf

Digust Joy
)O
(c

Pleasure
Fear None

Other Contentment
Disappointment

Admiration
Shame

Regret Love

Guilt Relief

Sadness Compassion

Figure 18.6 Template of Version 3.0 of the GEW as based on the GRID validation described
in this chapter.

19-Fontaine-Chap18.indd 296 7/17/13 3:45 PM

 THE GRID MEETS THE WHEEL: ASSESSING EMOTIONAL FEELING VIA SELF-REPORT 297

18.8 Availability and analysis procedures

The most recent version of the GEW (Version 3.0) shown in Figure 18.6 is now available for
non-commercial use by academic researchers interested in self-report assessment of emotion.
The recommended instructions, the final version of the wheel interface, and further information
(about adaptation and translation into other languages and licensing for commercial use of the
instrument) are available on the GEW web page (http://www.affective-sciences.org/gew). There,
the wheel can be downloaded in Word format that can be easily adapted to the needs of different
researchers. The web page also shows information about the currently available versions for differ-
ent languages and computer applications of the wheel.
The 20 items currently used in the wheel, like the 24 items in the GRID study, have been chosen
on the basis of an extensive selection process based on theoretical and empirical considerations.
They represent a fair sampling of the more frequently used emotion terms in different languages
(see Chapter 5 in this volume for the GRID choices). However, they may not constitute an optimal
choice for all possible applications in different fields of research. For example, an event sampling of
emotions occurring in families may require a different selection of emotion categories than an ex-

13
perimental study with a limited set of emotion manipulations. Given the huge diversity of research

20
interests and needs in the field, it would be illusory to propose the use of the current standard ver-
sion for all kinds of application.
,
ss
It should be noted that the specific choice of emotion words is not constitutive for the GEW
e
and the advantages its use confers to researchers. Rather, the fact that the arrangement of the
Pr

terms in two-dimensional space is theory-based and empirically confirmed makes the instru-
ment much easier to use than the usual lists (especially in repeated applications, as participants
ity

will find the appropriate terms much more rapidly and precisely). The use of the instrument is
rs

also facilitated by the anchoring of the meaning of the chosen emotion terms with respect to their
ve

position in the underlying, theoretically determined, affective space. In consequence, users are
i
Un

invited to construct their own wheel if the choice of terms in the standard version is not optimal
for the respective research aim (see instructions for adaptation on the GEW web page provided
d

above). However, we would like to stress that, whenever possible, it is advantageous to use the
or

standard version shown in Figure 18.6 to allow for replication by other researchers and to build
xf
)O

up cumulative databases.
Finally, we provide a brief overview of the analysis procedures. The GEW can be analyzed in two
(c

different ways:
1 Using the classic discrete emotion approach, the ratings on different emotion families
in the GEW structure can be analyzed in a very similar fashion to the procedures used
with standard questionnaires in which emotion terms are listed one below the other and
participants are asked to rate the intensity with which they have experienced each of the
emotions on a five-point ordinal scale. Depending on the purpose of the study, participants
can be asked to (a) select only one emotion family (the strongest they experienced), (b)
choose several emotions they may have experienced simultaneously or in close succession
(producing mixed emotions), or (c) give a response to each emotion scale, with a special
category for “did NOT experience this emotion” (a format that has desirable psychometric
characteristics).
2 Using a dimensional approach allows the researcher to obtain continuous dimensional values by
profiting from the explicit arrangement of the emotion terms in two-dimensional valence by
control/power space. Concretely, each emotion term can be represented by its two coordinates
in the space (provided in Table 18.2) with the respective intensity as a third variable. These

19-Fontaine-Chap18.indd 297 7/17/13 3:45 PM

 298 COMPONENTS OF EMOTIONAL MEANING

three values can then be analyzed separately or in the form of a composite scale. In the case of
mixed emotions, with several emotions rated, the respective coordinates and intensities can be
combined using statistical measures of central tendency. Alternatively, all ratings within one or
more of the four quadrants of the wheel can be combined with the help of statistical measures
of central tendency.

18.9 Conclusion
The GEW is a theory-based instrument for the assessment of emotional experience through self-
report. It combines a dimensional orientation with ratings of intensity for a number of major
emotion families, rated in a categorical fashion. The instrument has been used for many years by
emotion researchers in many different applications. Earlier studies with multidimensional analy-
ses of similarity ratings of emotion words (Bänziger et al., 2005), designed to validate the theoreti-
cal structure that underlies the arrangement of the emotion families on a wheel-like circle, have
essentially confirmed the predicted placement of the emotion families in a two-dimensional space
formed by valence and control/power. However, attempts to use direct ratings of the appraisals

13
that are thought to determine the differentiation of emotions on the control/power dimension

20
have met with mixed success. Two of those studies (Bänziger et al., 2005; Sacharin et al., 2012) have
been reviewed and it is suggested that it might be useful to use a different approach to validate the
,
ss
position of emotion words in a valence × power structure, abandoning the exclusive reliance on
e
coping potential appraisals in favor of a more comprehensive representation of power.
Pr

The GRID paradigm has been found to be ideally suited for this purpose, given that it anchors
y

the four fundamental dimensions of emotional space in feature profiles based on all emotion com-
it
rs

ponents, and given the remarkable stability of these factors over many different languages. The four
ve

dimensions reliably emerge even when individual components (also appraisal) are removed from
the data set (see data reports in Part III of this volume). The analysis of the 20 GEW words with the
i
Un

GRID instrument in 10 countries with 10 different languages has allowed us to firmly validate the
theoretical structure used in the development of the wheel. The precise data provided by these
d
or

GRID results have allowed recalibrating the positioning of the emotion categories on the rim of
xf

the emotion wheel. Based on this recalibration, a new version of the GEW, 3.0, is presented in this
)O

chapter and is made available to interested researchers. While the theoretical structure of the GEW
has been validated in this research, an empirical construct validation remains to be done.
(c

Apart from providing a combination of a dimensional and a categorical approach to emotion

assessment through self-report, the GEW provides an alternative to the dominant valence ×
arousal model in dimensional approaches. We hope that the new Version 3.0 of the GEW will see
widespread use in the future and will help standardize the self-report assessment procedures in the
psychological investigation of emotion and in the interdisciplinary domain of the affective sciences
more generally. We believe that systematic use of a standard version of the GEW would consti-
tute important progress with respect to replicating results across studies, and, most importantly,
allow increased collaboration in designing empirical studies and sharing the data in the emotion
domain. These activities, in turn, would greatly contribute to our effort to better understand the
semantics of emotion terms and their role in the categorization and communication of emotional
experience.

19-Fontaine-Chap18.indd 298 7/17/13 3:45 PM