UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Double you? Function and form of implicit and explicit self-esteem
Albers, L.W.A.
Link to publication
Citation for published version (APA):Albers, L. W. A. (2010). Double you? Function and form of implicit and explicit self-esteem. Amsterdam.
General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.
Download date: 12 Oct 2020
DOUBLE YOU?
DOUBLE YOU?
Function and form of implicit and explicit
self-esteem
ACADEMISCH PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus
prof. dr. D.C. van den Boom ten overstaan van een door het college voor promoties
ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel
op vrijdag 15 januari 2010, te 10:00 uur
door
Lucas Wilhelmus Antonius Albers
geboren te Cuijk en Sint Agatha
Promotor: Prof. dr. A. Dijksterhuis
Copromotor: Dr. M. Rotteveel
Overige leden: Prof. dr. H. Aarts
Prof. dr. A.H. Fischer
Prof. dr. J.A. Forster
Dr. R.W. Holland
Prof. dr. S.L. Koole
Prof. dr. J. van der Pligt
Faculteit Faculteit der Maatschappij- en Gedragswetenschappen
Cover Design: Janneke Albers
5
Preface
In this dissertation I examine function and form of implicit self-esteem, which
reflects the relatively less accessible part of self-esteem. Specifically, this
dissertation is about the additional value implicit self-esteem has in
understanding self-esteem, one of the most researched concepts in
Psychology. I will argue that in order to place the concept of implicit self-
esteem within the realms on current research on self-esteem, understanding of
both measuring implicit self-esteem and its relation with explicit self-esteem
is necessary.
Though several measures exist to capture implicit self-esteem, in this
dissertation I will focus on optimizing and addressing one of the most widely
used measures, the name letter test (Kitayama & Karasawa, 1997). The name
letter test consists of judging the different letters of the alphabet on their
attractiveness. A name letter effect is then calculated based on attractiveness
ratings of someone’s name letters while correcting for specific and general
letter liking, in order to avoid confounds. This name letter effect is
subsequently used as a measure of implicit self-esteem based on the
assumption that by assessing someone’s evaluation of well established
attributes of the self (e.g., such as someone’s initial letters) in an unobtrusive
Preface
6
way (e.g., under the guise of simply evaluating all alphabet letters),
someone’s implicit evaluation of the self is measured.
In Chapter 1, I will give an overview of the existing literature on the
relation between implicit and explicit self-esteem and then integrate the
findings from the research presented in the chapters 2, 3, and 4 of this
dissertation. Therefore, I did not include a general conclusion as a last chapter
in my dissertation. It should also be noted that because all chapters were
written separately, they can be read on their own.
In Chapter 2, I will focus on optimizing the current calculation method
of the name letter test in order to improve its validity as a measure of implicit
self-esteem. I will argue that the traditional calculation method results in a
name letter score that is confounded, because it cannot separate the influence
of name letter liking from overall letter liking. In three experiments I will
show that this (theoretical) confound of overall letter liking indeed is
problematic in empirical findings when using the traditional name letter score.
I will propose an alternative calculation method for the name letter score to
remedy these issues and to arrive at a more valid measure for implicit self-
esteem.
In Chapter 3, I will discuss function and form of implicit self-esteem in
relation to affect-regulation processes. More specifically I will first show that
implicit self-esteem relates over and above explicit self-esteem and across
time to the personality dimensions Emotional Stability and Neuroticism, the
two personality dimensions that are most strongly related to affective
regulation. In a second and third experiment I will further show that implicit
self-esteem can differentiate between the kinds of affect regulation required:
It is of special importance in regulating affect when the self is threatened. On
the other hand, our explicit measures of self-esteem did not relate to self-
regulation and rather seems to predict a more general state of affect.
In Chapter 4, I will, in search for “the real self”, further dig into the
relation between implicit and explicit self-esteem. I will propose, based on
both theoretical and empirical evidence, that though measures of implicit self-
Preface
7
esteem may not be without problems, in essence implicit self-esteem is closer
to “the real self” than explicit self-esteem. In an additional experiment I
further substantiate this proposition by showing that explicit self-esteem more
resembles implicit self-esteem when an automatic goal to be honest is
activated.
Contents
Preface 5
Chapter 1: Identifying Implicit and Explicit self-esteem 11
Chapter 2: Towards Optimizing the Name Letter Test 35
as a Measure of Implicit Self-Esteem
Chapter 3: Implicit Self-Esteem Goes a Long Way: Common Grounds 59
of Personality, Self-Esteem and Affect Regulation
Chapter 4: Digging for the real attitude: Lessons from research on 81
implicit and explicit self-esteem.
References 105
Summary 117
Summary in Dutch (Nederlandse Samenvatting) 123
Acknowledgements (Dankwoord) 129
KLI Dissertation Series 131
11
Chapter 1
Identifying Implicit and Explicit self-esteem
Chapter 1
12
“One's own self is well hidden from one's own self; of all mines of treasure,
one's own is the last to be dug up.”
Friedrich Wilhelm Nietzsche, 1883
In the 20th century most psychological measures resembled each other with
regards to their explicit or direct nature. That is, psychological constructs
were measured by asking individuals to indicate their attitudes, feelings,
cognitions or behavior. For example, when measuring self-esteem someone
had to indicate personal agreement with items such as “I have a positive
attitude towards myself”. A collection of agreements and disagreements with
similar statements is then taken as an individual score on the measurement of
that psychological construct, in this example the score on the Rosenberg Self-
Esteem Scale (Rosenberg, 1965). The sheer prevalence of these self-report
methods of psychological constructs suggests their usefulness and established
value within the field of psychology. However, they rely on two basic
assumptions when applying those measures as indicators of who we are and
how we function: Individuals will and can introspectively indicate what they
think, feel and do. Challenging both assumptions has led in the late 20th
century into a surge of attention in the use and development of implicit, or
indirect, measuring instruments (Fazio & Olson, 2003). However,
falsification of each of these assumptions poses different problems for
psychological research.
The possibility that someone may not be willing to honestly answer
self-report measures is generally acknowledged to pose a validity problem for
the measurement of psychological constructs although the importance of the
problem is subject of ongoing debate (Holtgraves 2004; Alliger & Dwight,
2000). Indirect or implicit measures circumvent this problem since the subject
simply does not know what is being assessed so that the impact of social
Implicit and Explicit Self-Esteem
13
desirability factors greatly diminishes.
The idea that individuals actually may not be able to have full
introspective access to themselves can already be found in the works of 18th
and 19th century influential philosophers such as Leibniz, Schopenhauer and
Nietzsche, but has only been substantiated by empirical evidence conducted
in the last few decades (for reviews see for example Dijksterhuis, & Bargh,
2001; Epley, Savitsky & Kachelski, 1999; Greenwald & Banaji, 1995;
Lewicki, Hill & Czyzewska, 1992). Because traditional self-report measures
cannot capture someone’s introspectively inaccessible parts, our
understanding of psychological functioning might benefit from indirect or
implicit measures that do try the grasp our nonconscious attitudes and
feelings.
So besides circumventing validity problems in self-report measures,
the development of indirect or implicit measures may first and foremost be
useful in exploring the role that both nonconscious content and processes play
in our psychological functioning. In this dissertation I will focus on the latter
and try to demonstrate that implicit measures have added value over and
above explicit measures in understanding the self.
Specifically, the focus will be on the development of the psychological
construct of self-esteem, one of the most researched concepts in psychology,
and its division into two different concepts, namely explicit and implicit self-
esteem. First, a discussion on the emergency of the concept of implicit self-
esteem is presented that focuses on the implicit attitude towards the self and
that resulted especially from research on implicit attitudes and their influence
on our judgments and behavior. This is followed by an analysis and overview
of several instruments that measure implicit self-esteem. Since these measures
are scarcely out of the egg, optimizing them can effectively contribute to our
understanding of implicit self-esteem. In this regard the focus will be on the
development of the name letter test and methods to optimize it as a measure
of implicit self-esteem. The relation between implicit and explicit self-esteem,
which is like any other implicit-explicit relation subject of ongoing debate,
Chapter 1
14
will first be discussed and then connected to other psychological constructs.
Specifically, results will be presented showing that implicit self-esteem relates
over and above explicit self-esteem to important affect-related personality
traits and furthermore serves as a useful predictor of affect-regulation. I will
furthermore attest that these empirical results form the core in the theoretical
development of understanding function and form of implicit self-esteem.
Though I agree with Nietzsche that our own ‘real’ self is well hidden from us,
implicit self-esteem measures may serve as effective tools in digging for it.
This is why this state-of-the-art-overview is ended with an attempt to mine
the self.
The development of implicit self-esteem
Within social psychological research self-esteem has a long-standing
tradition, like any attitude, to be the result of conscious evaluative processes.
Indeed, empirical account of self-esteem resided almost solely in explicit self-
evaluative measures such as the Rosenberg Self-Esteem Scale (Rosenberg,
1965). Psychological research using this explicit conceptualization of self-
esteem showed that people in general have an unrealistically positive explicit
self-esteem, or explicit self-evaluation, that actually promotes psychological
well-being (Taylor & Brown, 1988). People with high explicit self-esteem
feel better in the face of threat or after failure (e.g., Brown & Marshall, 2001;
DiPaula & Campbell,2002; Lane, Jones & Stevens, 2002; Pyszczynski,
Greenberg, Solomon, Arndt & Schimel, 2004) though there is somewhat
conflicting evidence regarding the specificity of this effect: there are also
studies showing that the ability of high-self-esteem individuals to deal better
with threats to the self rather reflects a general disposition in feeling happier
independent of the circumstances (see Baumeister, Campell, Krueger, & Vos,
2003, for a review on the relation between explicit self-esteem and
happiness).
The contention that self-attitudes comprise only conscious evaluations
radically changed when a growing body of empirical research on attitudes
Implicit and Explicit Self-Esteem
15
showed proof for the existence of implicit attitudes that can influence both
behavior and explicit judgments in a profound manner. Attitudes or
stereotypes can be activated outside of awareness and subsequently influence
our judgments of other people (e.g., Bargh and Pietromonaco,1982) of
ourselves (e.g., Baldwin, Carrel, and Lopez,1991) and even induce behavior
in line or contrary to those attitudes or stereotypes (e.g., Bargh, Chen &
Burrows, 1996; Chen & Bargh, 1997; Dijksterhuis & Bargh, 2001).
Furthermore, research has shown that subliminal exposure to a stimulus
influences our attitude toward that stimulus outside of our awareness
(Bornstein & D'Agostino, 1992; Kunst-Wilson & Zajonc, 1980, Zajonc,
2001). This attention to implicit attitudes prompted Greenwald and Banaji
(1995) to introduce the concept of implicit self-esteem that is generally
defined as the introspectively inaccessible part of our self-evaluations (for
reviews see Greenwald & Banaji, 1995; Koole & DeHart, 2007; Pelham &
Hetts, 1999). In contemporary self-esteem research a division has thus been
made between explicit self-esteem, a form of self-evaluation that occurs under
conscious self-reflection, and implicit self-esteem that is regarded as an
automatic and unconscious evaluation of the self.
Though the concept of implicit self-esteem itself is generally not
contested, both the validity and reliability of contemporary implicit measures
are under debate. Bosson, Swann and Pennebaker (2000) compared seven
different measures of implicit self-esteem on both their validity and
reliability. A meager two measures had acceptable test-retest reliabilities, the
different implicit measures did not correlate among each other and
correlations between the implicit self-esteem measures and several measures
of explicit self-esteem failed to reach significance in most instances.
However, the two implicit measures with acceptable test-rest reliabilities each
did predict criterion variables known to be related to measures of explicit self-
esteem, though both implicit measures each predicted different criterion
variables. These results presented by Bosson, Swann & Pennebaker (2000)
might not seem that promising in measuring implicit self-esteem, there are,
Chapter 1
16
however, several rays of hope.
First of all, since the observation of Bosson et al. (2000) that implicit
self-esteem measures are still in their infancy, several improvements have
been made. In the next section “Measuring Implicit Self-Esteem” these recent
developments and the resulting improved psychometric qualities of the two
most used measures of implicit self-esteem, the self-esteem implicit
association test (self-esteem IAT), originally developed by Greenwald and
Farnham (2000) and the name letter test developed by Kitayama & Karasawa
(1997) will be discussed.
Secondly, ever since the observation by Bosson, Swann & Pennebaker
(2000), additional evidence shows that measures of implicit and explicit self-
esteem do seem to correlate with each other, albeit low to moderate. In the
section “The Relation between Implicit and Explicit Self-Esteem” this
evidence is presented, suggesting that previous accounts of the absence of
correlations between implicit and explicit self-esteem measures are mainly the
result of several situational factors and psychometric qualities suppressing
this correlation.
Thirdly, there is extensive empirical evidence showing that implicit
self-esteem measures predict important criterion variables and frequently do
so even better than explicit self-esteem measures. Especially the ability of
implicit self-esteem measures to predict the emotion regulation of a person in
self-threatening circumstances is promising. In the section “Function and
Form of Implicit Self-Esteem” these findings will be discussed in more detail.
Measuring Implicit Self-Esteem
Since they seem most promising in measuring implicit self-esteem, in this
section the psychometric qualities are discussed of both the self-esteem IAT,
originally developed by Greenwald and Farnham (2000) and the name letter
test, developed by Kitayama and Karasawa (1997).
The self-esteem IAT is based on the general IAT developed by
Greenwald, McGhee, and Schwarz (1998) and measures the strength of
Implicit and Explicit Self-Esteem
17
automatic associations between concepts. Along these lines, the assumption
behind the self-esteem IAT is that someone with a more positive implicit self-
concept possesses stronger (automatic) associations between the self-concept
and positive stimuli as opposed to the self and negative stimuli, and thus
should be relatively faster in associating self-items with positive stimuli.
Specifically, the self-esteem IAT measures the relative speed with which a
participant can associate a self-related stimulus with a positive stimulus
compared to a negative stimulus. The self-esteem IAT of Greenwald and
Farnham (2000) consists of two critical blocks of categorization tasks
whereby in each block one response to two sets of paired stimuli is assessed.
In each block two sets of paired stimuli are projected on a screen throughout
that block. In one critical block the first set always consists of a combination
of a self-related stimulus and a positively valenced stimulus (e.g., “me/
pleasant”), while the second set always comprises a not-self related stimulus
and a negatively valenced stimulus (e.g., “not-me/ unpleasant”). A target
stimulus that can be of either four kinds of stimuli (i.e., self-related, not self-
related, positive or negative) appears and has to be correctly classified as fast
as possible to one of the two sets of paired stimuli. In the other critical block
the combination of the two sets of paired stimuli is reversed so that now the
self-related stimulus is combined with a negative stimulus (e.g., “me/
unpleasant”) and the not self-related stimulus with a positive stimulus (e.g.,
“not me/ pleasant”). The difference in response times is subsequently used as
a measure of implicit self-esteem. Since the introduction of the IAT the
scoring algorithm used to assess the IAT effect has been improved resulting
in increased correlations between IAT measures and their parallel explicit
measures as well as improved psychometric qualities of the IAT itself
(Greenwald, Nosek, & Banaji, 2003). Contrary to these improvements, the
psychometric properties of the IAT have been challenged by several
researchers (see e.g., Blanton & Jaccard, 2006; Rothermund & Wentura,
2004). Specific to the self-esteem IAT, Karpinski (2004) challenged the
notion that the IAT could serve as an adequate measure of implicit self-
Chapter 1
18
esteem. He argued that since the self-esteem IAT is a relative measure that
contrasts positive and negative associations with the self-concept always with
positive and negative associations with the “not-self concept” it is not clear
whether it actually measures the evaluation of the self-attitude or rather the
evaluation someone has of the specific not-self concept used in the IAT. For
example, using a nonspecific other category would lead to different results
than using best friend as category of the not-self concept, since the two
concepts would differ, expectedly, in the positive and negative associations
attached to them.
Though Pinter and Greenwald (2005) partly met this assertion by
showing that a nonspecific other category or a not-me category are relative
neutral in valence, I agree that the inherent relative nature of the (self-esteem)
IAT can be theoretically problematic as a specific measure of implicit self-
esteem, since someone with predominant positive associations with the self-
concept and neutral associations with the not-self concept will show the same
level of implicit self-esteem opposed to someone with neutral associations
with the self-concept and negative association with the not-self concept. To
counter these problems Karpinski and Steinman (2006) recently developed
the single category IAT, modeled after the single target IAT developed by
Wigboldus, Holland, and Van Knippenberg (2005), that assesses the strength
of the evaluative associations of a single category such as the self-concept.
Karpinski and Steinman (2006) show in their research good psychometric
properties of the self-esteem single category IAT, such as a higher correlation
with explicit measures of self-esteem compared to the traditional self-esteem
IAT, though future research is needed to substantiate these preliminary
findings.
Another common measure for implicit self-esteem is the name letter
test, which assesses the liking people have for the letters of their own name.
The idea that the liking people have for their own name letters could serve as
a measure for implicit self-esteem was first put forward by Greenwald and
Banaji (1995). They based their idea on the demonstration by Nuttin (1985)
Implicit and Explicit Self-Esteem
19
that people tend to prefer the letters in their own name to letters that are not
part of their name. Greenwald and Banaji (1995) argued that since this name
letter effect occurred without the need to think consciously about the self and
without conscious knowledge of the purpose of evaluating letters, it could
serve as a measure of implicit self-esteem. This idea was implemented by
Kitayama and Karasawa (1997) who had participants rate, in random order,
all the letters of the alphabet for their attractiveness. To control for the
baseline attractiveness of the different alphabet of the alphabet, they then
calculated the name letter score as the attractiveness of the name letters minus
the average attractiveness of those same letters rated by participants not
possessing those name letters. This name letter test and its calculation method
of the name letter score has become an important tool to measure implicit
self-esteem, whereby most research using the name letter test focuses only on
the initials of participants, because people prefer their initials to the other
name letters (Koole & DeHart, 2007). The basic assumption underlying this
name letter test is that someone’s name letters are attached to the implicit self-
concept so that evaluations of these name letters represent that persons’
implicit self-esteem and research from Koole, Dijksterhuis, and van
Knippenberg (2001) has shown that the name letter effect is indeed especially
apt in assessing self-evaluations in the absence of conscious self-reflection.
To fully appreciate the potential importance of the name letter test as a
measure of implicit self-esteem, we need to insure that the way we calculate a
name letter score indeed reflects the evaluation of the self-associated name
letters. As mentioned before the traditional name letter score is calculated as
the rated attractiveness of name letters minus the average attractiveness of
those same letters rated by participants not possessing those name letters.
Though this calculation method controls for the baseline attractiveness of the
different letters of the alphabet it neglects the possible influence of general
letter liking of a specific person. Specifically, the traditional calculation
method of the name letter score treats the not-name letter evaluations of a
person as irrelevant evaluations of that person, so these evaluations only
Chapter 1
20
function as a baseline score for calculating the name letter score of other
people. However, not addressing the influence of not-name letter evaluations
introduces a confound. It means that a response bias in letter liking or other
influences on general letter liking can affect the traditional name letter score.
If we want to use the traditional name letter score as a measure of implicit
self-esteem this can be problematic, because someone who happens to dislike
letters in general (e.g., because he just failed a language test) will have a
lower name letter score than someone who happens to like letters (e.g.,
because she just won a game of scrabble).
In the second chapter of this dissertation the assertion that the
traditional calculation method of the name letter effect is confounded is
discussed into more detail and substantiated with empirical evidence showing
that the traditional calculation method of the name letter score cannot
adequately distinguish the evaluation of name letters from not-name letters
and it may therefore mistake response bias or general letter liking for implicit
self-esteem. Because it is crucial for the name letter score as a measure
implicit self-esteem to distinguish the evaluation of the self-associated name
letters from the not-name letters a new scoring algorithm that circumvents the
confound in the traditional scoring algorithm is proposed. The main
difference with the traditional scoring algorithm is that the new scoring
algorithm assesses the unique contribution of the name or initial letter
evaluation by simultaneously controlling for the influence of not-name letter
liking and liking of the specific alphabet letters. Instead of subtracting the
specific evaluation of the different alphabet letters, the new scoring algorithm
removes its influence on the name letter evaluation. More importantly, to
avoid confound of general letter liking, the new scoring algorithm also
removes the influence that not-name letter evaluation has on name letter
evaluation. Statistically this means determining the size of the relation
between name-letter evaluation with both not-name letter evaluation and the
specific alphabet letter evaluation, and then remove their influence using a
regression approach (Cohen, Cohen, West & Aiken, 2003). Because this
Implicit and Explicit Self-Esteem
21
alternative scoring algorithm of the name letter score is by definition
unrelated to not-name letter evaluation it represents a more valid measure of
implicit self-esteem compared to the traditional scoring algorithm.
The relation between implicit and explicit self-esteem
The different accounts on the relation between implicit and explicit self-
esteem show a somewhat mixed picture. Some researchers advocate that the
two forms of self-esteem are independent constructs because they did not find
significant correlations between implicit and explicit self-esteem (e.g.,
Baccus, Baldwin & Packer, 2004; Jordan, Spencer, Zanna, Hoshino-Browne
& Correll, 2003; Robinson & Meier, 2005; Shimizu & Pelham, 2004;
Spalding & Hardin, 1999). Others, however, did find significant correlations
at least in some experimental conditions or in some samples (e.g., Albers,
Dijksterhuis & Rotteveel, 2009; Bosson, Swann, & Pennebaker, 2000;
Dijksterhuis, Albers & Bongers, 2008; Greenwald & Farnham, 2000; Jones,
Pelham, Mirenberg, & Hetts, 2002; Jordan, Whitfield, & Zeigler-Hill, 2007;
Koole, Dijksterhuis & van Knippenberg, 2001; Olson, Fazio, & Hermann,
2007; Pelham, Koole, Hardin, Hetts, Seah & DeHart, 2005) suggesting that an
independency account of both forms of self-esteem is unwarranted. This
contention is further substantiated by two recent meta-analyses showing a
consistent and reliable, albeit weakly, positive relation between explicit self-
esteem and implicit self-esteem measured with both the name letter test
(Krizan & Suls, 2008) and the self-esteem IAT (Hofman, Gawronski,
Gschwendner, Le, & Schmitt, 2005). Based on the presented empirical
evidence I believe it is safe to conclude that implicit and explicit self-esteem
are in fact related. The fact that implicit and explicit self-esteem do not relate
in a strong and straightforward way, may be attributable to at least two main
categories of factors that attenuate the relation between implicit and explicit
self-esteem.
First of all, there is growing evidence that the relation between implicit
and explicit self-esteem can be enhanced by psychological variables that
Chapter 1
22
reduce (the need for) an active construction process during the measurement
of explicit self-esteem. Koole, Dijksterhuis and Van Knippenberg (2001)
showed, for example, that preventing participants from actively constructing
explicit self-esteem (i.e., under high cognitive load) increased the relation
between implicit and explicit self-esteem, while Krizan and Suls (2008)
demonstrated similar results under circumstances when the self-concept is
already activated. Furthermore, Olson, Fazio, and Hermann (2007)
demonstrated an enhanced correlation between implicit and explicit self-
esteem after asking participants to be honest, an effect that was also obtained
by Dijksterhuis et al. (2008) who showed that priming participants with a goal
to be honest resulted in higher correlations between implicit and explicit self-
esteem.
Besides preventing self-presentational strategies, individual differences
in the extent to which self-esteem is deliberately expressed also seem to
influence the relation between implicit and explicit self-esteem. Along these
lines, Pelham et al. (2005) showed that women, who are considered to be
more in touch with their inner feelings and thus rely less on construction
processes, show more overlap between measures of implicit and explicit self-
esteem than men. In a similar fashion, Jordan, Whitfield, and Zeigler-Hill
(2007) demonstrated that people with a higher faith in their own intuition
showed increased correspondence between their implicit and explicit self-
esteem.
Secondly, it may be noted that the generally low correlations between
measures of implicit and explicit self-esteem are also at least in part caused by
the fact that implicit measures are still in a developing stage. Their reliability
is often low (Bosson, Swann & Pennebaker, 2000), and it is not fully
understood yet what exactly drives the effects of some of the implicit
measures. Recently, various researchers have proposed improvements to
various measures of implicit self-esteem. Both Karpinski (2004) and Albers,
Dijksterhuis and Rotteveel (2009) suggested improvements to implicit
measures of self-esteem that will likely result in more meaningful correlations
Implicit and Explicit Self-Esteem
23
between implicit and explicit measures of self-esteem.
The evidence presented before support the notion that implicit and
explicit self-esteem do relate in a predictable fashion and as such might reflect
the two parts of a dual process, whereby implicit self-esteem reflects well-
learned associations of the self that are effortlessly activated, as opposed to
the need of an active construction process to express explicit self-esteem.
Both initiatives to strengthen implicit measures of self-esteem as well as
research addressing psychological variables influencing the relation between
implicit and explicit self-esteem give rise to optimism for the usability of self-
esteem measures and their combined contribution in understanding self-
esteem processes in the future.
Function and form of implicit self-esteem
In the previous sections I have tried to show that though measures of implicit
self-esteem are relatively young and still in a developing stage they are
promising in contributing to our understanding of self-esteem, especially
because these measures are, unlike measures of explicit self-esteem, less
likely to be subject to both self-presentational strategies and inaccessibility of
one’s implicit attitudes. Research on the relation between measures of
someone’s implicit and explicit self-esteem shows a small but reliable relation
that is enhanced in an intuitive mindset, when the self is activated, or when an
active construction process of our self-attitude is absent.
So there is convincing empirical support that both forms of self-esteem have
something in common, but to determine what that something may look like
we need to take our exploration a step further. In this respect it might be
insightful to investigate both the unique and shared validity of both implicit
and explicit self-esteem in predicting other relevant psychological constructs
and processes. Not only could it specify the relevance of implicit self-esteem
over and above those other psychological constructs, but it could also clarify
on both the similarities and the differences between implicit and explicit self-
esteem themselves.
Chapter 1
24
The empirical findings on the relation between both implicit and
explicit self-esteem and other psychological traits are sparse but do
consistently demonstrate a relation with those personality traits that have a
strong affective component. Both Connor and Barrett (2005) and Robinson
and Meier (2005) found a positive relation between implicit self-esteem and
Emotional Stability. Robinson and Meier (2005) furthermore showed that
implicit self-esteem still uniquely predicted Emotional Stability when
controlling for explicit self-esteem, however in both experiments implicit and
explicit self-esteem were uncorrelated, so it remains unclear whether implicit
and explicit self-esteem would also share variance with Emotional Stability
when they would in fact be related. This is why in our own research,
presented in Chapter 3 of this dissertation, we investigated whether implicit
and explicit self-esteem would also share variance with Emotional Stability
and three additional personality dimensions (i.e., Extraversion, Conformism
and Egoism) under circumstances when implicit and explicit self-esteem are
in fact related. Taken together the results from our own research demonstrated
that implicit self-esteem, explicit self-esteem, and the two personality
dimensions Extraversion and Emotional Stability were all intrinsically related.
Implicit and explicit self-esteem correlated positively and participants with
high implicit and explicit self-esteem were emotionally more stable and
extraverted. Though the correlations between implicit self-esteem and the
explicit measures were only moderate to small in nature, the variance they
shared was stable across an extended period of time. Because implicit and
explicit related to each other, we could also address the nature of their shared
relation with explicit personality traits. Though explicit self-esteem mediated
the relation between implicit self-esteem and the personality traits
Extraversion and Emotional Stability, in general we did not found full
mediation, suggesting that part of the relation between implicit self-esteem
and (explicit) personality variables is independent from explicit self-esteem,
and thus demonstrates the additional value of implicit self-esteem. Unlike
explicit self-esteem, implicit self-esteem was not related to the personality
Implicit and Explicit Self-Esteem
25
dimensions Egoism and Conformism. These results suggest that the common
grounds between implicit and explicit self-esteem reside predominately in
personality traits that are most strongly connected to affective experience
(DeNeve, & Cooper, 1998).
In order to determine the nature of this unique and shared variance it is
interesting to bridge these findings with research investigating the relation
between both implicit and explicit self-esteem and affect-regulation itself.
Both Robinson and Meier (2005) and Conner and Barrett (2005) showed a
relation between both implicit and explicit self-esteem and affective
experience, with both higher implicit and explicit self-esteem resulting in less
negative affect in daily live. Interestingly, in the study of Connor and Barrett
(2005) implicit self-esteem, as opposed to explicit self-esteem, did not predict
a measure of positive affect in daily life. Since implicit self-esteem did predict
negative affect and not positive affect this might mean that implicit self-
esteem may be specifically related to coping with self-threatening
experiences. Because their study was not set out to test what kind of daily
experiences resulted in these effects it is, however, not clear whether high
implicit self-esteem is indicative of less experienced negative affect in general
or whether it actually specifically pertains to self-threatening and not to self-
affirming experiences.
In another study, Robinson and Meier (2005) did demonstrate that
implicit self-esteem predicted experimenter-reported negative affect after a
self-threat manipulation, but because they did not include a control or self-
affirming condition, and did not measure self-reported affect, it still remains
unclear whether the relation between implicit self-esteem and (negative)
affect is unique to the self threatening circumstances, and if so, whether this
relation would also apply to the actual affective experience of an individual.
This is why we conducted an experiment (see Chapter 3 of this dissertation)
where we elaborated on the role that both implicit and explicit self-esteem
play in affect regulation. Five months prior to the actual experiment
participants completed measures of their implicit and explicit self-esteem and
Chapter 1
26
the personality measures Extraversion and Emotional Stability. At the actual
experiment, participants first completed the same measures for both implicit
and explicit self-esteem and then were subjected to either a self-threatening or
self-affirming manipulation upon which they indicated how they felt. Based
on previous findings of Robinson and Meier (2005) and Meager and Aidan
(2004) and the contention that implicit self-esteem functions as a buffer to
negative affect (Koole & DeHart, 2007) we expected the function of implicit
self-esteem to reside especially in self-threatening circumstances. This is
indeed what we found. Though explicit self-esteem and the personality
dimensions Extraversion and Emotional Stability were predictive of general
experienced affect, implicit self-esteem was specifically related to
experienced affect under self-threatening circumstances. High implicit self-
esteem individuals experienced less negative feelings opposed to low implicit
self-esteem individuals when the self was threatened, but not when the self
was affirmed, an effect that we even could demonstrate with our measure of
implicit self-esteem that we collected 5 months prior to the actual self-threat.
This specific ability for implicit self-esteem to do so was even further
supported by our findings that explicit self-esteem, Extraversion and
Emotional Stability could not do so.
These findings suggest that that the common grounds of implicit and
explicit self-esteem reside mainly in the unaffected self, but diverge in
function when the self is affected. Regardless of the way the self is affected,
explicit self-esteem remains indicative of an individual’s general level of
positive or negative affect. The opposite is true for implicit self-esteem: high
implicit self-esteem individuals feel less negatively affected by threats to their
selves. In this respect our studies support the contention that implicit self-
esteem functions as buffer that only comes into play when it is needed.
Will the real self-esteem please stand up
In the previous paragraphs I have tried to demonstrate that the combined
effort of the explicit and implicit self-esteem constructs has additional value
Implicit and Explicit Self-Esteem
27
in our attempts to understand function and form of the self, or at least that of
our self-esteem. However, by introducing the concept of implicit self-esteem
we have also complicated matters. Instead of one, we now have two concepts
aimed at measuring self-esteem. So how should we address this complication?
Which concept better reflects our ‘real’ self-esteem? As a first step it makes
sense to consider the empirical evidence presented in the previous paragraphs
demonstrating that implicit and explicit self-esteem are in fact related (in
Chapter 4 of this dissertation additional empirical evidence is provided to
support this notion). From a theoretical point of view such a relation may be
apparent, but these results make it also unlikely that implicit and explicit self-
esteem may each have their own independent concept counterpart from an
empirical point of view. This suggests that the constructs of implicit and
explicit self-esteem are at least indicative of one underlying system that I will
denote from now on as our core self-esteem. This core self-esteem, or core
self-attitude, reflects our self-esteem before it is actually measured,
verbalized, or acted upon. Both implicit and explicit self-esteem try to grasp
this core self-esteem, the question is, however, whether they do equally well.
To answer to what extend both implicit and explicit self-esteem reflect
the core self-esteem, basically there are three areas of empirical data that are
important to consider. The first area consists of the previously mentioned
findings showing that the relation between implicit and explicit self-esteem is
enhanced by factors that reduce (the need for) an active construction process
of responding to the explicit self-esteem measures. In itself this is not that
problematic, because core relations between different psychological processes
do support our understanding of human functioning, but because explicit self-
esteem could need more construction, in general there is a bigger chance that
explicit self-esteem measures will be contaminated with other psychological
processes than measures of implicit self-esteem. This makes explicit self-
esteem a less pure measure that is more dissociated from the core self-esteem
than implicit self-esteem.
The second area of findings actually demonstrates that measures of
Chapter 1
28
explicit self-esteem not only have a bigger change of being distorted, but that
empirical findings show that these distortions occur and actually result in the
inability of explicit self-esteem measures to differentiate in important self-
opinions. After a thorough review of the existing literature Baumeister et al.
(2003) concluded that high explicit self-esteem individuals not only comprise
a heterogeneous group of individuals with important differences in their self-
opinions but also that in order to denote these differences and its behavioral
consequences other additional measures such as self-deception, impression
management and narcissism are necessary. If explicit self-esteem itself can
not differentiate between important differences in self-opinions, obviously, it
is impossible to accurately reflect core self-self-esteem with only someone’s
explicit self-esteem.
Binding the findings from the previous paragraphs together provide
further support for the contention that explicit self-esteem is further
dissociated from core self-esteem than implicit self-esteem. Because high
explicit self-esteem individuals comprise a heterogeneous group that is
influenced by other psychological processes and measures of explicit self-
esteem require more construction than measures of implicit self-esteem,
theoretically, implicit self-esteem should be less susceptible to the influence
of those other psychological forces. The third area of findings actually
pertains to empirical support for this notion. Several studies that measured
both implicit and explicit self-esteem demonstrated a positive relation
between explicit self-esteem, but not implicit self-esteem, and measures that
represent self-presentational strategies, such as self-deception, defensive self-
presentation bias, self-enhancement and unrealistic optimism, (e.g., Bosson,
Brown, Ziegler-Hill & Swann, 2003; Jordan, Spencer, Zanna, Hoshino-
Browne, & Correll, 2003; Olson, Fazio, & Hermann, 2007; Riketta, 2005). If
we consider core self-esteem to be a stable self-attitude that really reflects
how we feel about ourselves, this also implies that this self-attitude has no
‘reason’ to misrepresent that attitude. It is ‘just’ the attitude of the self, and
only other internal and external psychological forces can instigate a
Implicit and Explicit Self-Esteem
29
motivation to try to misrepresent this core self-esteem. Because explicit self-
esteem, as opposed to implicit self-esteem, positively relates to
misrepresenting core self-esteem, this further demonstrates that implicit self-
esteem is closer to core self-esteem than explicit self-esteem.
Moreover, an interesting pattern emerges among those studies that
regressed both implicit and explicit self-esteem and their interaction onto the
different dependent measures of self-presentation. While high implicit self-
esteem individuals in general comprises a homogenous group, high explicit
self-esteem individuals were heterogeneous: Those with high explicit self-
esteem and low implicit self-esteem showed more defensive self-presentation
bias, self-enhancement and unrealistic optimism than individuals with both
high explicit and high implicit self-esteem (e.g., Bosson, Brown, Ziegler-Hill
& Swann, 2003; Jordan, Spencer, Zanna, Hoshino-Browne, & Correll, 2003;
Olson, Fazio, & Hermann, 2007). I believe this further supports the idea that
implicit self-esteem is a more pure form of core self-esteem than explicit self-
esteem.
Conclusion
In this chapter I focused on the additional value of the construct of implicit
self-esteem in understanding the self. Though the current measures of implicit
self-esteem are still in their infancy, they do predict self-relevant behavior,
affect-regulation and affective personality traits, and they do so over and
above measures of explicit self-esteem. Empirical findings regarding its
function show that implicit self-esteem, if well developed, effectively buffers
against threats to the self. Because implicit self-esteem is closer to core self-
esteem than explicit self-esteem, this also suggests that this buffering function
might be representing an important role our core self-esteem plays in
supporting the self.
Besides the additional value in understanding the self, measures of
implicit self-esteem have also contributed in understanding the desirability of
having high-self-esteem by supporting the notion that high scores on
Chapter 1
30
traditional measures of explicit self-esteem could be either the result of
actually having high self-esteem or the result of other, less desirable self-
presentational strategies. This means that the numerous studies that have
demonstrated less desirable outcomes of having high explicit self-esteem do
not necessarily reflect the consequences of having high self-esteem but may
as well reflect the consequences of not wanting to acknowledge not having
high self-esteem.
Because a score on a measure of explicit self-esteem can be indicative
of two completely different self-opinions, it may no longer be warranted to
use those traditional measures of explicit self-esteem on their own. That is,
every measure of explicit self-esteem should be accompanied by another
measure to separate the bold from the boisterous. In this respect it is wise, but
not sufficient, to include explicit measures that are indicative of these self-
presentational strategies, because these explicit measures may also be subject
to other unwanted factors influencing its construction. Furthermore, explicit
psychological measures require at least some awareness of (the outcome of)
these psychological processes, and thus exclude the possibility that less
conscious processes could also help our understanding of the self. Because
measures of implicit self-esteem need less construction and do not require
awareness of processes related to the self, they represent the best candidates
in furthering our understanding of the self.
We started this state-of-the-art-overview with a statement from
Nietzsche protesting that of all treasures, it is our own that we dig up last. In
‘Thus spoke Zarathustra’ Nietzsche states that this is the result of a lifetime
experience wherein the societal norms are both imposed on us but which we
also impose on ourselves, thus increasing the distances between who we
really are and who we believe or say we are. Because measures of implicit
self-esteem are important empirical tools to diminish precisely these
distances, the self is no longer the last treasure we need to mine for.
Implicit and Explicit Self-Esteem
31
Limitations and Future Directions
In this chapter I may have come across as being rather harsh on the usability
of explicit self-esteem measures and rather lenient towards measures of
implicit self-esteem. Admittedly, measures of explicit self-esteem have been
used in literally thousands of psychological studies for over more than four
decades, so if such a mature concept as explicit self-esteem still raises serious
concerns regarding its functionality and usability, some exhortation is
justified. In addition, even though recently successful developments have
been made in improving the psychometric qualities of measures of implicit
self-esteem such as the name letter test and the implicit self-esteem IAT, a
critical note is also in place regarding the psychometric qualities of implicit
self-esteem measures.
Although implicit self-esteem as a construct is less prone to biases than
explicit self-esteem and thus closer to the core self-esteem, one may argue
that the relatively lower reliability of the current implicit self-esteem
measures causes distortions of a similar magnitude as when measuring
explicit self-esteem, so in this respect there is no intrinsic evidence to favor
the current implicit self-esteem measures over explicit self-esteem measures.
Furthermore, although it is easier for someone to fake on a measure of explicit
self-esteem it also seems possible to fake on measures of implicit self-esteem,
such as the IAT (Fiedler & Bluemke, 2005). Empirical evidence concerning
the ability to fake on the name letter test is absent, but it seems very plausible
that if someone is aware of the specifics of the current name letter test and its
scoring algorithm, she will be able, at least to some extent, to influence the
outcomes on the name letter score. This means that effort should not only be
put in identifying and reducing distortions on explicit self-esteem measures
but also on measures of implicit self-esteem.
Another area that warrants additional investigation concerns the special
status that is attributed to individuals with low implicit and high explicit self-
esteem. As mentioned before, these individuals seem to show the most
defensive behaviors, it remains unclear however, whether these findings
Chapter 1
32
really reflect defense behaviors of individuals with low implicit and true high
explicit self-esteem or rather of individuals with low implicit and faked high
explicit self-esteem (or apply to both groups). Because scores on explicit self-
esteem measures can be easily faked, individuals that score high on an
explicit self-esteem measure could really believe to have high explicit self-
esteem but might as well be reluctant in disclosing their knowingly low
explicit self-esteem. This means that defensive behaviors could arise as well
as a consequence of low implicit and explicit self-esteem combined with
reluctance in disclosing the self to others than as a consequence of the
combination between true high explicit and low implicit self-esteem. Because
these two possible profiles clearly differ from a psychological point of view, I
believe further investigation is necessary. One way to shed some light on the
origins of these defensive behaviors would be to test whether similar effects
would still occur when explicit self-esteem is measured under circumstances
when (the need for) active construction is reduced, so that it becomes harder
for individuals to misrepresent their true explicit self-esteem.
Although studies in this area are limited, it seems that the different
measures of implicit self-esteem do not seem to correlate with each other.
Because, in principle, the different measures of implicit self-esteem try to
grasp the same construct, they should be related to at least some extend. The
fact that they do not warrants empirical and theoretical investigation. In this
respect both the reliability and validity of the current implicit self-esteem
measures are in need of further improvement. For those implicit self-esteem
measures that do seem to have acceptable psychometric qualities, such as the
name letter test and the (single category) IAT, it may be wise to look both at
the conceptual differences in the measures themselves and at psychological
factors influencing the relation between them. Since the IAT uses different
stimuli and requires a different response than the name letter test, they may
measure different aspects of implicit self-esteem. In this respect it would be
interesting to investigate whether aggregating the scores on both existing
measures adds more to the prediction of important criterion variables (e.g.,
Implicit and Explicit Self-Esteem
33
affect-regulation) than adapting both measures in such a way that they
resemble each other more on a conceptual level (e.g., time response window
in the name letter test). In a similar way, it would be interesting to look at the
influence of different psychological contexts (e.g., self-relevance) on the
relation between the different implicit self-esteem measures.
The last critical note to address concerns the stereotype that is
(automatically) activated when confronted with the word “implicit”
throughout this dissertation. It may imply that in the current state of affairs it
is already possible to investigate function and form of that part of the self we
are completely unaware of. I believe, however, that with the current implicit
measurers this is not really possible, because these measures are still subject
to awareness of what is being measured and require, at least to some extent,
awareness on responding to those measures. So, in this respect it is important
to regard the implicitness of the current self-esteem measures from a relative
perspective, whereby the added value of those relatively implicit measures in
understanding our selves is paramount to the discussion how implicit they
really are.
35
Chapter 2
Towards Optimizing the Name Letter Test as a
Measure of Implicit Self-Esteem*
A common measure for implicit self-esteem is the name letter effect,
traditionally calculated as the rated attractiveness of someone’s initials or
name letters minus the average attractiveness of those same letters rated by
people not having those initial or name letters. We present evidence showing
this calculation method is confounded with general letter liking. As such, it
cannot adequately distinguish the evaluation of name letters from not-name
letters, and may therefore mistake response bias or general letter liking for
implicit self-esteem. We propose an alternative calculation method that is
without confounds and assesses the unique contribution of name letter
evaluation by simultaneously controlling for the influence of not-name letter
liking and general liking of the specific letters.
*This chapter is based on: Albers, L., Dijksterhuis, A., & Rotteveel, M. (2009). Towards
Optimizing the Name Letter Test as a Measure of Implicit Self-Esteem. Self and Identity, 8,
63-77.
Chapter 2
36
The notion that most of our behavior is evoked consciously has been
successfully challenged by psychological research using measures and
manipulations that either prevent conscious access and deliberation, or
conceal its real intention (Bargh & Chartrand, 1999). Within the field of self-
esteem, this research led to the development of the concept of implicit self-
esteem that is generally defined as the introspectively inaccessible part of our
self-evaluations (for reviews see Greenwald & Banaji, 1995; Koole &
DeHart, 2007; Pelham & Hetts, 1999). A common measure for implicit self-
esteem is the name letter test, which assesses the liking people have for the
letters of their own name (i.e., self-associated evaluations) compared to the
liking of these same letters among people who do not have those letters in
their names. In this article, we argue that the way the score of the name letter
test is traditionally calculated is not adequate to distinguish name letter liking
from not-name letter liking and as such is sensitive to response bias or overall
letter liking. We propose an alternative calculation method that remedies
these issues and serves as a more valid measure for implicit self-esteem.
Origins of the Name Letter Test
The phenomenon that people tend to prefer the letters of their own name to
not-name letters was originally demonstrated by Nuttin (1985) and coined the
name letter effect. Because the effect occurred without the need to think
consciously about the self and without conscious knowledge of the purpose of
evaluating letters, Greenwald and Banaji (1995) suggested the name letter
effect could serve as a measure of implicit self-esteem.
This idea was implemented by Kitayama and Karasawa (1997) who
had participants rate, in random order, all the letters of the alphabet for their
attractiveness. To control for the baseline attractiveness of the different
letters, they calculated the name letter score as the rated attractiveness of the
name letters minus the average attractiveness of those same letters as rated by
participants not possessing those name letters (i.e., they subtracted the
“normative” liking score of the specific alphabet letters). The name letter test
Towards Optimizing the Name Letter Test
37
has become an important tool to measure implicit self-esteem (Koole &
DeHart, 2007) and the method described above is the predominant method to
calculate the name letter score. In all research cited in this paper the method
was used to calculate a name letter score, although Jones, Pelham, Mirenberg,
and Hetts (2002) used the absolute liking scores of name letters, without
subtracting the normative liking score of the specific alphabet letters, in
addition to the traditional method.
In recent years, the name letter effect has contributed considerably to
our understanding of implicit self-esteem. Research using the name letter test
has shown that the name letter effect is indeed especially apt in assessing self-
evaluations in the absence of conscious self-reflection (Koole, Dijksterhuis, &
van Knippenberg, 2001). Furthermore, implicit self-esteem has been linked to
childhood experiences and parenting styles (DeHart, Pelham, & Tennen,
2006), and it has been found to moderate unrealistic optimism and self-
enhancing behaviors (Bosson, Brown, Zeigler-Hill, & Swann, 2003). High
implicit self-esteem functions as a buffer helping people to deal with threats
to the self (Jones, Pelham, Mirenberg, & Hetts, 2002) or failure (Dijksterhuis,
2004). Finally, it has been found that people gravitate more to other persons,
places and professions that share their name letters (Pelham, Carvallo, &
Jones 2005; Pelham, Mirenberg, & Jones, 2002), demonstrating that name
letter liking can affect major life decisions.
The Importance of Distinguishing Name Letters from Not-Name Letters
To fully appreciate the potential importance of the concept of implicit self-
esteem, we need to insure that the way we measure implicit self-esteem
indeed reflects implicit self-evaluative processes. As mentioned before, the
traditional name letter score is calculated as the rated attractiveness of name
letters minus the average attractiveness of those same letters rated by people
not possessing those name letters. This means that not-name letter evaluations
of a person are treated as irrelevant evaluations in calculating the name letter
score of that person, and only function as a baseline score for calculating the
Chapter 2
38
name letter score of other people.
However, not addressing the influence of not-name letter evaluations
introduces a confound. It means that a response bias in letter liking or other
influences on general letter liking can affect the traditional name letter score.
If we want to use the traditional name letter score as a measure of implicit
self-esteem this can be problematic, because someone who happens to dislike
letters in general (e.g., because he just failed a language test) will have a
lower name letter score than someone who happens to like letters (e.g.,
because she just won a game of scrabble).
A hypothetical example can illuminate this problem. In Table 2.1, we
show a hypothetical dataset with letter-evaluations and the traditional name
letter score. To simplify matters we only show evaluations of three different
letters, but the same logic would apply if we would use all letters of the
alphabet. What happens here? Although Suzy evaluates the S, her own name
letter, as less attractive than other letters, according to the traditional
calculation method she clearly shows a positive name letter score. This is not
caused by the fact that she likes her name letter, but by the fact that she likes
letters in general. Jack and Frank prefer their name letter to other letters, but
in fact, Jack has a negative name letter score, and the name letter score of
Frank suggests no preference for his name letter over the not-name letters.
Table 2.1 Hypothetical Letter-Evaluations and Traditional Name Letter Scores
F S J NLStrada
Frank 6 4 5 0
Suzy 9 7 9 3
Jack 3 4 6 -1
Note. NLStrad represents the name letter score using the traditional calculation method.
a Calculating for example the NLStrad for the initial F of Frank = 6 – [(9 + 3)/2] = 0.
Because the traditional name letter score is sensitive to general letter
liking, any factor, manipulation or response bias (not just winning a game of
Towards Optimizing the Name Letter Test
39
scrabble) that influences general letter liking will, by definition, influence the
traditional name letter score and thus appear to have influenced implicit self-
esteem. Suppose we would show participants either a funny cartoon or, in a
control condition, something more neutral. Subsequently, they rate letters, and
because the traditional name letter score differs between conditions, we
conclude that watching cartoons enhances implicit self-esteem. Although it is
not impossible that watching cartoons could positively affect implicit self-
esteem, in this case it is much more likely that differences in mood caused
participants to evaluate letters in general as more or less attractive (Mayer,
Gaschke, Braverman, & Evans, 1992). Because the traditional name letter
score is not equipped to tell one from the other, it mistakes affect for implicit
self-esteem.
An Alternative Calculation Method for the Name Letter Score
If we want to use the name letter score as a measure of implicit self-esteem, it
is crucial that we are able to distinguish between the evaluation of the name
letters and the not-name letters. In order to achieve this we believe it is
necessary that the evaluation of not-name letters is incorporated in calculating
a name letter score. One way to do this is to extend the traditional calculation
method by also subtracting the not-name letter evaluation from the name
letter evaluation. However, simply subtracting the not-name letter evaluation
could theoretically lead to the same problem as mentioned before, as the
actual size of the impact of not-name letter evaluation is not known. It could
lead to the same confound as in the traditional name letter score and it could
also lead to overcorrection resulting in a similar confound but now in opposite
direction: Someone who happens to like letters will have a lower name letter
score then someone who happens to dislike letters. We believe this possibility
makes this calculation method a less suitable option. In the results section we
will present empirical evidence that supports this notion.
Another alternative method for incorporating the influence of not-name
letter liking is to standardize the evaluation of the alphabet letters of each
Chapter 2
40
individual separately, prior to calculating the traditional name letter score.
Though this method would remove at least part of the confound of general
letter liking, this method will also partly remove the influence of the ‘true’
name letter effect. This is because in standardizing each letter evaluation, the
mean of someone’s overall letter evaluation is subtracted from the specific
letter evaluation, and then followed by dividing it by the standard deviation of
someone’s overall letter evaluation. If we compare two individuals, each with
the same not-name letter evaluations, but with one person showing higher
evaluation on the name letters, both the standard deviation and mean letter
evaluation of the latter person will be higher so that the standardized scores of
the name letter evaluations of this person will be relatively lower.
Furthermore, because the standardized letter evaluations of the not-name
letters partly determines the standardized name letter evaluations, and vice
versa, a confounding correlation between the two could still ensue. In the
results section we will present empirical evidence that supports these possible
problems with this calculation method.
Hence, in order to eliminate a possible confound of general letter liking
it is not sufficient to subtract someone’s not-name letter evaluation from the
traditional name letter score. In addition, standardizing each individual’s letter
evaluation prior to calculating the traditional name letter effect could remove
part of the name letter effect itself.
If we want to use this name letter evaluation as a measure of implicit
self-esteem both the influence of not-name letter liking and specific letter
evaluation need to be controlled for. We can accomplish this statistically
using the following two steps (Cohen, Cohen, West & Aiken, 2003): First,
both the average not-name letter evaluation and the average normative score
of the name letters are regressed onto the average name letter evaluation. The
standardized regression coefficients that follow from this analysis are a
measure of the influence both potential confounds have on the name letter
evaluation. Secondly, to remove the influence of both potential confounds we
need to multiply both the not-name letter evaluation and the normative score
Towards Optimizing the Name Letter Test
41
by their unstandardized regression coefficients, before subtracting them from
the name letter evaluation. This results in the calculation method for the
alternative name letter score that is depicted in Equation 1:
NLS = NLE – (B1 * NNL + B2 * SLE) (1)
In this equation, B1 and B2 are the unstandardized regression
coefficients when simultaneously predicting name letter evaluation (NLE)
from both the not-name letter evaluation (NNL) and the normative score of
the name letters (SLE). This method results in a name letter score (NLS) that,
by definition, is unrelated to both not-name letter evaluation and the specific
letter evaluation of the name letters.
To recap, our main point is that we should insure that the name letter
score is not confounded with response bias or any factor that influences
general letter liking. The alternative name letter score we propose indeed does
this, but remains sensitive to factors that do influence the evaluation of the
self-associated name letters. We believe that it makes the alternative
calculation method of the name letter score a better measure of implicit self-
esteem than the traditional method. In the remainder of this article, we will
present data from three different studies that empirically support this notion.
Advantages of the Alternative Name Letter Score: Empirical Evidence
Overview of the Studies
We collected name letter evaluations from three different studies to compare
the traditional calculation method of the name letter score with the alternative
calculation method we propose. We will only discuss the results of every
study to the extent they pertain to the calculation method itself. For more
information on the results of Study 1 and 2, we refer to Albers, Dijksterhuis,
and Rotteveel (2009), and for Study 3 to Albers, Dijksterhuis, and Rotteveel
(2004). In every study, participants completed the experiment on a PC-
computer. We administered the name letter test as follows: Participants were
Chapter 2
42
informed that they were about to evaluate simple stimuli, in this case
evaluating the attractiveness of the letters. They were asked to give their first
impression on each letter and not to think too long but to respond intuitively.
Letters were always presented in random order. In every study participants
indicated their liking of every letter by clicking on the corresponding area of a
9-point scale (Study 1) or a continuous scale (Study 2 and 3), ranging from
ugly to beautiful. In Study 2 and 3 the scores on the letter evaluation could
range from -10, representing the left end of the scale (i.e., ugly), to 10,
representing the right end of the scale (i.e., beautiful) and a score of zero
representing the middle point of the scales (i.e., neutral).
At the end of Study 1 and 2, participants provided their full names on
an informed consent form. In Study 3, participants provided their initials. All
participants were undergraduates from the University of Amsterdam and
received course credits for participation. We will first provide some additional
information about each individual study.
Study 1. A total of 384 participants (269 female) completed the name
letter test and Heatherton and Polivy’s (1991) State Self-Esteem Scale
(SSES). The order of the name letter test and SSES was counterbalanced.
After completing these measures, 184 out of these 384 participants also
completed the Dutch Narcissism Scale (Ettema & Zondag, 2002). This scale
measures two distinct aspects of narcissism. The first aspect addresses
internally directed narcissism, and measures someone’s belief of being
important to other people (e.g., “I am disappointed that people do not always
see me for who I really am”). The second aspect is externally directed
narcissism, and measures someone’s believe of influencing other people (e.g.,
“I have no problems in persuading others to do what I deem necessary”).
Study 2. All 128 participants (82 female) completed the name letter test
prior to the Rosenberg Self-Esteem Scale (Rosenberg, 1965). After a filler
task they answered several questions concerning their participation in the
Towards Optimizing the Name Letter Test
43
experiment, and the perceived stability of their self-esteem.
Study 3. Fifty-eight participants (40 female) were randomly assigned to
write either about a recent positive experience or a recent negative experience
and then indicated how they felt after describing the event on two continuous
scales ranging from respectively very negative to very positive, and very
angry to very cheerful. Scores on both scales were aggregated to form one
measure of experienced affect (α = .82), with higher scores reflecting more
positive affect. They then completed the name letter test, followed by the
Rosenberg Self-Esteem Scale (Rosenberg, 1965).
Calculating the traditional and alternative name letter scores.
Most research using the name letter test focuses only on the initials of
participants, because people prefer their initials to the other name letters
(Koole & DeHart, 2007). This is why we will mainly focus on comparing the
traditional and alternative name letter score using as name letters the
evaluation of the initials of participants and not the evaluation of all name
letters. We will however, also address the relation between the evaluation of
all the name letters and not-name letter evaluations, to show that both initial
evaluation and evaluation of all name letters give rise to the same confound
when using the traditional calculation method of the name letter score.
We calculated the traditional name letter score (NLStrad) as the
average evaluation of someone’s initials (NLE) minus the normative score of
these initials¹ (SLE) which we calculated as the average evaluation of the
corresponding letters rated by participants not possessing those name or initial
letters² (SLE), yielding the following equation (see also Kitayama &
Karasawa, 1997, for a similar approach):
NLStrad = NLE – SLE (2)
We calculated the alternative name letter score (NLSalt) as the average
Chapter 2
44
evaluation of someone’s initials (NLE) minus the influence of both the
average evaluation of not-name letters³ (NNL) and the evaluation of those
same letters rated by participants not possessing those name or initial letters
(SLE) yielding Equation 3. In the Appendix we present a detailed description
of the different steps involved in calculating this alternative name letter score.
NLSalt = NLE – (B1 * NNL + B2 * SLE) (3)
In Table 2.2, we present the regression coefficients for the three different
studies.
Because the calculation method we propose determines whether in any
given sample both not-name letter evaluation and the normative score of the
name letters have a confounding influence on the name letter score, a name
letter score does not have to be corrected if its corresponding confounds are
absent (i.e., non-significant) in that sample.
Table 2.2 Summary of Regression Analyses for NNL and SLE predicting NLE
B SE B β p
Study 1 (N = 384)
NNL 0.61 0.09 .34 .000
SLE 0.67 0.25 .13 .010
Study 2 (N = 128)
NNL 0.48 0.15 .28 .001
SLE 0.37 0.29 .11 .207
Study 3 (N = 58)
NNL 0.98 0.23 .51 .000
SLE -0.26 0.45 -.07 .569
Note. R² = .14 for Study 1; R² = .09 for Study 2; R² = .26 for Study 3; NLE = average evaluation
of initials; in Study 1 and 2, NNL = not-name letter evaluation; in Study 3, NNL = not-initial
letter evaluation; SLE is the average evaluation of the same initials evaluated by participants for
whom it is not a name-letter (Study1 and 2) or not an initial letter (Study3).
Towards Optimizing the Name Letter Test
45
This is why we excluded NNL or SLE from Equation 3, when its
corresponding unstandardized regression coefficient (i.e., B1 or B2) showed
no significant relation with NLE. However, in all our studies we did find a
confounding influence of not-name letter liking, and hence, B1 was always
included. Furthermore, in- or excluding B2 when it was not significant did not
result in different results than presented in this article.
In our previous section, we mentioned two other possible calculation
methods of the name letter score that, in theory, should be less effective than
the method we propose in Equation 2. To empirically test whether these other
calculation methods indeed show poorer psychometric qualities, we also
included them in our analyses. The first calculation method is an extension of
the traditional calculation method by also subtracting the average evaluation
of not-name letters (NNL). This means that both unstandardized regression
coefficients B1 and B2 from Equation 3 are now fixed to a value of 1,
resulting in the following equation we used to calculate this possible name
letter score:
NLSext = NLE – SLE – NNL (4)
In the second possible method, the letter evaluations of an individual are
first standardized before calculating the traditional name letter score. This
means that first the mean letter evaluation of an individual is subtracted from
each letter evaluation of that individual. These corrected alphabet letter
evaluations are then divided by the standard deviation accompanying the
mean letter evaluation of that individual. With these standardized letter
evaluations we calculate a name letter score in the same way as the traditional
name letter score is calculated.
Results
We first address the problematic (i.e., confounded) relation between the
traditional calculation method of the name letter score and the evaluation of
Chapter 2
46
not-name letters. Secondly, we look at the impact general letter liking can
have on the name letter score depending on which calculation method of the
name letter score is used, followed by evidence concerning the relation
between not-name letter evaluation and several relevant explicit measures.
Finally, we address the relation between explicit self-esteem and implicit self-
esteem, as measured with the different calculation methods.
The confound of not-name letter evaluation on the traditional name letter
score.
We mentioned before that the traditional name letter score is confounded with
general letter liking. We tested this by correlating the traditional name letter
score of participants with their average evaluation of the not-name letters. In
every study, the traditional name letter score correlated positively with not-
name letter evaluation. The correlation between traditional name letter score,
based on the initials, and the not-name letter evaluation was r = .34, p < .000,
in Study 1; in Study 2, r = .29, p < .001; and in Study 3, r =.48, p < .000. The
correlation between the traditional name letter score, based on all name
letters, and the not-name letter evaluation was even more pronounced in the
two studies where we obtained all name letters from participants, r = .60 , p <
.000, in Study 1; and in Study 2, r = .44, p < .000. These positive correlations
do not make sense theoretically. If we conceive of not-name letters as merely
neutral, no correlation should exist between not-name letter evaluation and
the traditional name letter score. Our data however, suggest that between 8
and 36 % of the variance in the traditional name letter score is actually caused
by not-name letter liking.
Extending the traditional name letter score by also subtracting the not-
name letter evaluation, as depicted in Equation 4, resulted in both Study 1 and
2 in a name letter score that was still confounded with not-name letter
evaluation, but now in the opposite direction. The name letter score and the
not-name letter evaluation correlated negatively, in Study 1, r = -.23, p < .000;
in Study 2, r = -.29, p < .001. This means that the extended calculation
Towards Optimizing the Name Letter Test
47
method of the traditional name letter score is not suitable as a specific
indicator of name letter liking.
Standardizing each letter evaluation of an individual before calculating
the traditional name letter score resulted in a name letter score that still
correlated negatively with not-name letter evaluation in Study 1, r = -.11, p <
.04. In Study 2 this confound was only marginal significant, r = -.14, p < .11,
while in Study 3 the confound was absent, r = .08, p = .55. Though the
standardizing method did show some effect of the confound of general letter
liking, the impact of this confound is smaller compared to the traditional
calculation method or the extended calculation method.
The alternative name letter score we proposed in Equation 4 is by
definition uncorrelated with not-name letter evaluation. Because this method
gets rid of any confound with general letter liking, we believe that in the light
of this confound it is the best candidate to calculate a name letter score.
The influence of general letter liking on the traditional name letter score.
The positive correlations between the traditional name letter score and the
evaluation of not-name letters may have other undesirable consequences. If
some variable affects general letter liking, the results from the traditional
name letter score will lead to the unwarranted conclusion that this variable is
related to implicit self-esteem. Although we did not specifically design our
studies to test this in practice, we did find such an influence of general letter
liking on the traditional name letter score when we used gender of the
participant as a factor. In all three studies, men showed significantly higher
ratings of both name and not-name letter evaluation than women did. This
difference in general letter liking actually resulted, in a higher traditional
name letter score for men than for women: in Study 1, F(1, 382) = 6.0, p <
.02; in Study 2, F(1, 126) = 4.7 , p < .03; in Study 3, F(1, 56) = 6.2, p < .05.
Standardizing each letter evaluation before calculating the traditional
name letter score led the differences between men and women in Study 1 and
2 ( both ps > .22) to disappear, except for Study 3, where we still found
Chapter 2
48
marginally significant difference F(1,56) = 3.6, p < .06. In contrast, both our
alternative calculation method as well as the extended calculation method led
the differences between men and women to disappear in all studies4 (all ps >
.26).
The results from this section show that significant differences in implicit
self-esteem between certain groups can in fact be caused by general letter
liking when we use the traditional name letter score. This means that in these
instances general letter liking is mistaken for implicit self-esteem. We believe
this is an unwanted aspect of the traditional name letter score that is resolved
in our alternative name letter score, because the influence of general letter
liking is removed.
The relation between not-name letter evaluation and relevant explicit
measures.
The findings discussed above clearly demonstrate that correlations between
the traditional name letter score and other variables should be treated with
caution as these correlations may be caused by general letter liking. In Table
2.3, we present data that show that in both Study 1 and 3, we did find
additional evidence for such an influence of general letter liking on the
traditional name letter score.
In Study 1, we found a significant relation between the traditional
name letter score and both internally directed narcissism, r = .20, p < .007,
and externally directed narcissism, r =.18, p < .02. However, also not-name
letter evaluation correlated with externally directed narcissism, r = .27, p <
.0003. On the other hand, the alternative name letter score, did not correlate
significantly with externally directed narcissism, r = .08, p = .29, but did
correlate with internally directed narcissism, r = .20, p < .006. This suggests
that not-name letter evaluation mediates the relation between the traditional
name letter score and externally directed narcissism. We confirmed this in a
mediation analysis using the SPSS-Macro from Preacher & Hayes (2005),
with the traditional name letter score as independent, external narcissism as
Towards Optimizing the Name Letter Test
49
dependent, and not-name letter evaluation as mediator, showing full
mediation of not-name letter evaluation. The bootstrapped estimate of the
indirect effect was 0.016, p < 0.01, while the direct effect of the traditional
name letter score on external narcissism after controlling for not-name letter
evaluation was no longer significant, β = .09, t(181) = 1.13 , p = .26. Though
both the alternative and traditional name letter score show that implicit self-
esteem relates to internally directed narcissism, the traditional name letter
score leads to the unwarranted conclusion that implicit self-esteem also relates
to external narcissism.
Table 2.3 Correlations Between the Traditional and Alternative Name Letter
Scores, Not-name Letter Evaluation, and Relevant Explicit Measures.
N NLSalt NLStrad NLSext ZNLS NNL
Study 1
Narcissism internal 184 .20** .20** .20** .19* .03
Narcissism external 184 .08 .18* .02 -.01 .27***
SSES 384 .19*** .23*** .16** .15** .15**
Study 2
RSES 128 .29*** .27** .24** .17 .05
Study 3
Affect PE 30 .10 .36* .14 .12 .43*
Affect NE 28 .58*** .61*** .61*** .55** .21
RSES 58 .32* .47*** .29* .16 .44***
Note. SSES = State Self-Esteem Scale; RSES = Rosenberg Self-Esteem Scale; Affect PE = affect positive
event; Affect NE = affect negative event; NLStrad = traditional name letters score; NLSalt = alternative
name letter score; NLSext = extended name letter score; ZNLS = traditional name letter score based on the
standardized letter evaluations; in Study 1 and 2, NNL = evaluation of not-name letters; in Study3, NNL =
evaluation of not-initial letters.
* p < .05. ** p < .01. *** p < .001.
In Study 3, we found another unwanted influence of general letter liking.
Chapter 2
50
In Table 2.3, it is shown that the traditional name letter score correlated
significantly with reported affect, both after describing a negative and after a
positive personal experience, demonstrating that in both conditions a higher
traditional name letter score was associated with more positive affect.
Furthermore, these correlations did not differ between the two conditions, z =
1.2, p = .23. However, not-name letter evaluation also correlated positively
with the reported affect after describing a positive personal experience,
suggesting that the relation between the traditional name letter score and
affect could in fact be due to general letter liking. Indeed, the alternative name
letter score that controls for not-name letter evaluation did not correlate with
the reported affect after describing a positive personal experience, but did
correlate after describing a negative personal experience, resulting in a
significant difference between the two correlations, z = 2.0, p < .04. Again,
using the traditional or alternative name letter score results in different
conclusions on the theory of implicit self-esteem. Should we have used the
traditional name letter score we would have concluded – erroneously- that
after self-threat and after self-enhancement, implicit self-esteem relates
similarly to the way we feel. The results from the alternative name letter score
actually show that this relation can be specified to situations where the self is
threatened.
The results from both the extended calculation method and standardizing
letter evaluations before calculating a traditional name letter score were
similar to our proposed alternative method, showing that on these aspects
these methods had better psychometric qualities then the traditional
calculation method.
The relation between implicit and explicit self-esteem. Though measures of
implicit and explicit self-esteem both refer to the same construct (i.e., self-
esteem) they tend to correlate only weakly (Pelham et al., 2005). We have
shown that when we use the traditional name letter score, not-name letter
evaluation can obscure the relation between implicit self-esteem and other
Towards Optimizing the Name Letter Test
51
relevant constructs. Because measures of explicit self-esteem could be equally
sensitive to such a response bias, it is important to rule out the possibility that
the relation between implicit and explicit self-esteem is in fact (partly) driven
by similarities in response bias on its measures. In Table 2.3, we present the
correlations between two measures of explicit self-esteem, not-name letter
evaluation, and the different name letter scores. Though in both Study 1 and 3
we found a significant correlation between the measure of explicit self-esteem
and not-name letter evaluation, in all studies the correlation between the
alternative name letter score (i.e., controlling for not-name letter evaluation)
and the measure of explicit self-esteem remained significant, in Study 1, r =
.19, p < .0002; in Study 2, r = .29, p = < .001; in Study 3, r = .32 , p < .02.
Mediation analysis with the traditional name letter score as independent,
explicit self-esteem as dependent, and not-name letter evaluation as mediator,
confirmed the absence of complete mediation in Study 1, 2 and 3, but did
show partial mediation of not-name letter evaluation in Study 3: bootstrapping
resulted in a significant indirect effect of 0.39, p < .05, while the direct effect
of the traditional name letter score on the measure of explicit self-esteem
remained significant after controlling for not-name letter evaluation, β = .28,
t(55) = 2.5, p < .01. These results show that though both the alternative and
the traditional name letter score correlate reliable with explicit self-esteem,
the alternative name letter score we propose provides a more valid estimate of
this relation.
The extended name letter score did correlate significant with explicit
self-esteem in all three studies though in study 1 this correlation appeared to
be less strong compared to the traditional name letter score, z = 1.6 , p < .06
(one-tailed).
Standardizing letter evaluations prior to calculating the traditional
name letter score resulted in a significantly lower correlation in Study 1
compared to the traditional name letter score, z = 1.7 , p < .05 (one-tailed). In
Study 2, this difference was marginally significant compared to both the
traditional calculation method and our proposed alternative calculation
Chapter 2
52
method, respectively z = 1.3, p < .10 (one-tailed); z = 1.5, p < .07 (one-tailed).
Furthermore, in Study 3 the correlation between the standardized name letter
score and explicit self-esteem was no longer significant, r = .16, p = .23, and
significant lower compared to the traditional name letter score, z = 2.5 , p <
.005, and marginally significant lower compared to our proposed alternative
name letter score, z = 1.3, p < .10 (one-tailed).
When we compare the correlations between explicit self-esteem and
the different name letter scores, both our proposed alternative name letter
score and the traditional name letter score do equally well, and better than
both the extended and standardized name letter score. However, the relation
between the traditional name letter score and explicit self-esteem was
mediated, and thus confounded, by not-name letter evaluation in Study 3. We
believe that these findings again rule in favor of our proposed alternative
name letter score as a measure of implicit self-esteem.
Discussion
In this article, we presented an alternative method for calculating the name
letter score and demonstrated that, compared to the traditional calculation of
the name letter score, the alternative name letter score we propose has better
psychometric qualities and serves as a more adequate measure of implicit
self-esteem. Both on a theoretical and empirical level we presented
converging evidence showing that the traditional name letter score does not
adequately deal with the evaluation of not-name letters, potentially leading to
unwarranted conclusions about implicit self-esteem. In all the three studies we
presented, the traditional name letter score correlated positive with not-name
letter evaluation. This confound could have led to the unwarranted conclusion
that men had higher implicit self-esteem than women, whereas our proposed
alternative method did not give rise to this problem. Furthermore, the
traditional name letter score suggested that implicit self-esteem is related to
both internally and externally directed narcissism and is similarly related to
experience of both self-threat and self-enhancement. The alternative name
Towards Optimizing the Name Letter Test
53
letter score however, only showed a relation between implicit self-esteem and
internally directed narcissism, and further specified that implicit self-esteem is
only related to affective experiences after self-threat (incidentally, this result
corroborates data by Conner and Barrett (2005) who showed that implicit
self-esteem predicts negative but not positive feelings in daily life).
The two other possible name letter scores we presented besides our
alternative name letter score did not fully resolve these issues. Simply
subtracting not-name letter evaluation still resulted in the confound of general
letter liking, making it impossible to guaranty that implicit self-esteem effects
based on such name letter scores are not the result of factors influencing
general letter liking. Standardizing every letter evaluation before calculating
the traditional name letter score did show better psychometric qualities
compared to the traditional name letter score, but resulted in a less sensitive
measure of implicit self-esteem, because part of the ‘true’ name letter effect is
also removed from the name letter score.
In sum, when using the traditional name letter score, it remains unclear
what part of the relation between the traditional name letter score and other
relevant variables or factors actually reflect the specific influence of implicit
self-esteem. Our alternative calculation method of name letter effect is
calculated in such a way that precludes the possibility in any given sample
that the resulting name letter scores are confounded with either general letter
liking or the normative rating of the name letters. We believe this makes our
alternative calculation method of the name letter score a more valid measure
of implicit self-esteem with better psychometric qualities.
Chapter 2
54
Endnotes
¹ The normative score of someone’s initials (SLE) is based on letter
evaluations of participants for whom it is not a name letter. This means that
the normative score of a prevalent name letter will be based on fewer
participants than the normative score of a rare name letter, and the former
could potentially be less reliable. However, in all our studies, the normative
ratings of prevalent name letters actually showed smaller standard deviations
than those of rare letters. This means that prevalent letters are judged more
similar and need fewer normative ratings than less frequent letters to become
reliable. In addition to this finding we also correlated the average evaluation
of each letter from Study 1 with its corresponding average evaluation of that
letter in Study 2, resulting in a correlation of r = .78, p < .000 (i.e., showing
an adequate overall reliability of the normative scores). If the prevalence of a
letter influences the reliability, than controlling for this influence should
lower the correlation mentioned before. In fact, the partial correlation of the
evaluations of letters between Study 1 and 2, while controlling for prevalence,
actually resulted in a slightly higher correlation of r = .81, p < .000,
suggesting that the normative ratings of subjects with common name letters
do not suffer from diminished reliability.
²In Study 3, we used the average evaluation of participants not
possessing the initial letters. In Study 1 and 2, we used the average evaluation
of the letters rated by participants not possessing those name letters. However,
using the average evaluation of participants not possessing the initial letters in
Study 1 and 2 resulted in almost identical name letter scores and did not
influence any other reported relations.
³In Study 1 and 2, we calculated someone’s average not-name letter
evaluation excluding all name letters of that person. In Study 3, where we
only obtained the initials of the participants, we calculated someone’s average
letter evaluation excluding the initials of that person. 4Though these findings clearly demonstrate the impact a confound of
not-name letter evaluation can have on the name letter score, the gender
Towards Optimizing the Name Letter Test
55
effects themselves might not generalize outside the University of Amsterdam.
In their research, Pelham et al. (2005) did not find any differences between
men and women on the traditional name letter score among six different
samples, one of the samples also consisting of Dutch students, but at a
different university.
Chapter 2
56
APPENDIX
Calculating the Alternative Name Letter Score
Below we have described the steps necessary to calculate the alternative name
letter score. A full SPSS-syntax of these steps can be obtained from the first
author.
To calculate the alternative name letter score, we assume that all name
letters of a participant can be identified. Furthermore, we assume that a name
letter score based on someone’s initials is calculated (the same approach
described below can be used to calculate the name letter score based on all
name letters but now by substituting the evaluation of the initials with the
evaluation of all name letters). First, it is necessary to create a (SPSS)-data
file with the following variables:
1. Someone’s initials
2. The evaluation of these initials
For example if someone’s name is Suzy Jones, the scores on the letter ‘S’ and
‘J’ represent the evaluations of the initials.
3. The 26 evaluations of all letters, whereby the evaluations of the letters
that are in fact name-letters for that individual are set to missing values
For Suzy Jones this means that for the 26 alphabet letter evaluations, the
evaluations of the letters ‘S’, ‘U’, ‘Z’, ‘Y’, ‘J’,‘O’, ‘N’, ‘E’ are set to missing
values
4. The mean of someone’s not-missing evaluations of the alphabet letters
from step 3 is calculated (NNL). This mean represents the average
evaluation of the not-name letters for each individual.
5. The mean of each alphabet letter from the not-missing evaluations of
the alphabet letters from step 3 is calculated. These means represent
the normative ratings of the alphabet letters based on all not-name
letter evaluations.
6. The two variables representing the normative ratings of someone’s
initials are created.
Towards Optimizing the Name Letter Test
57
For Suzy Jones the two new variables that are created in step 6 are the
normative scores on the letters ‘S’ and ‘J’ from step 5.
7. Two new variables are created: the mean of the two variables from step
6 (SLE) and the mean evaluation of the initials (NLE)
8. The variables NNL and SLE are regressed onto NLE, and the
corresponding unstandardized regression coefficients B1 and B2 can
now be used to calculate the name letter score using the following
equation:
NLS = NLE – (B1 * NNL + B2 * SLE)
59
Chapter 3
Implicit Self-Esteem Goes a Long Way: Common Grounds of
Personality, Self-Esteem and Affect Regulation*
*This chapter is based on: Albers, L., Dijksterhuis, A., & Rotteveel, M. (2009). Implicit self-
esteem goes a long way: Common grounds of Personality, Self-Esteem and Affective
Regulation. Manuscript submitted for publication.
Chapter 3
60
Self-esteem research has a long and rich history in several areas of
psychology, but until recently self-esteem theory was almost entirely based on
conscious self-reports. Inspired by social cognition research demonstrating
the value of assessing evaluations in a more unobtrusive way, the last decade
self-esteem researchers have shown an increasing interest in the more
intuitive components of the self. Consequently, several implicit self-esteem
measures were developed such as the name letter test (Kitayama & Karasawa,
1997), the self-esteem IAT (Greenwald & Farnham, 2000) and the implicit
self-evaluation survey (Hetts, Sakuma, & Pelham, 1999).
Though the different measures of implicit self-esteem do not seem to
correlate (e.g., Riketta, 2005; Bosson, Swann, & Pennebaker, 2000), two
recent meta-analyses do show a consistent, albeit weak, positive relations
between explicit self-esteem and implicit self-esteem measured with both the
name letter test (Krizan & Suls, 2008) and the self-esteem IAT (Hofman,
Gawronski, Gschwendner, Le, & Schmitt, 2005). Furthermore, most implicit
self-esteem researchers do not assume that implicit and explicit self-esteem
are theoretically completely independent (Chapter 4 of this thesis), suggesting
that some kind of relation between implicit and explicit self-esteem should
exist. Consequently, research on the relation between implicit and explicit
self-esteem diverged in into two different directions.
First, a considerable amount of research focused on the differences
between someone’s implicit and explicit self-esteem, showing that especially
having high explicit and low implicit self-esteem can result in maladaptive
self-regulatory strategies. For example, people with high explicit and low
implicit self-esteem display more unrealistic optimism (Bosson, Brown,
Zeigler-Hill, & Swann, 2003), are prone to prejudice (Jordan, Spencer, Zanna,
Hoshino-Browne, & Correll, 2005), and are more narcissistic and display less
stable self-esteem (Zeigler-Hill, 2006).
A second approach to implicit self-esteem research focuses on the
possibility that implicit and explicit self-esteem reflect the two parts of a dual
process, whereby implicit self-esteem reflects well-learned associations of the
Implicit Self-Esteem, Personality, and Affect Regulation
61
self that are effortlessly activated, as opposed to the need of an active
construction process to express explicit self-esteem (Chapter 4 of this thesis).
Though research in this area is limited, its results support the idea that
circumstances that reduce (the need for) an active construction process causes
explicit self-esteem to correlate stronger with implicit self-esteem. Koole,
Dijksterhuis and Van Knippenberg (2001) showed, for example, that
preventing participants from actively constructing explicit self-esteem (i.e.,
under high cognitive load) increased the relation between implicit and explicit
self-esteem, while Olson, Fazio, and Hermann (2007) demonstrated an
enhanced correlation between implicit and explicit self-esteem after asking
participants to be honest 9see also Chapter 4) .
Besides preventing self-presentational strategies, individual differences in
the extent to which self-esteem is deliberately expressed also seem to
influence the relation between implicit and explicit self-esteem. Along these
lines, Pelham et al. (2005) showed that women, who are considered to be
more in touch with their inner feelings and thus rely less on construction
processes, show more overlap between measures of implicit and explicit self-
esteem than men. In a similar fashion, Jordan, Whitfield, and Zeigler-Hill
(2007) demonstrated that people with a higher faith in their own intuition
showed increased correspondence between their implicit and explicit self-
esteem.
The first approach described before resulted in a better understanding of
what happens when the two forms of self-esteem conflict, whereas the latter
sheds light on the factors influencing the relation between the two forms of
self-esteem. However, both approaches are, de facto, less effective in
specifying the nature of the common grounds between implicit and explicit
self-esteem. Put differently, they specify when implicit and explicit self-
esteem overlap or what happens when the two conflict, but do not show us
what these common grounds might be and where they diverge.
In order to clarify what the common grounds between implicit and
explicit self-esteem are, and where they would diverge, a different approach is
Chapter 3
62
needed. In this respect we think it is insightful to look at possible relations
that both implicit and explicit self-esteem have with other relevant
psychological constructs. First of all, it would place implicit and explicit self-
esteem within a broader context of complex human functioning. Secondly, it
could specify the relevance of implicit self-esteem over and above those other
psychological constructs.
More importantly, the nature of the relation between other psychological
construct and both implicit and explicit self-esteem, especially under
circumstances when the latter two are correlated, can provide insight into both
the similarities and the differences between implicit and explicit self-esteem
themselves. To our knowledge, there is only one study that actually related
both implicit and explicit self-esteem to a third psychological construct.
Robinson and Meier (2005) showed in two experiments that both implicit and
explicit self-esteem were independently and both positively related to the
personality dimension Emotional Stability. Because in both experiments
implicit and explicit self-esteem were uncorrelated, obviously these results
can only clarify on the differences between implicit and explicit self-esteem
in predicting Emotional Stability. This is why in our first study we wanted to
examine whether implicit and explicit self-esteem will also share variance
with Emotional Stability under circumstances when implicit and explicit self-
esteem are in fact related. Furthermore, we also decided to include additional
personality dimensions to extend on the findings of Robinson & Meier
(2005). Specifically we wanted to address whether implicit self-esteem is
especially related to those personality dimensions that have a strong affective
component. Reasons for this were threefold. Affective experience is
positively related to both explicit self-esteem (Baumeister, Campell, Krueger,
& Vos, 2003) and the personality traits Extraversion and Emotional Stability
(DeNeve, & Cooper, 1998). Secondly, both Connor and Barrett (2005) and
Robinson and Meier (2005) showed a positive relation between implicit self-
esteem and both Emotional Stability and affective experience in daily live.
Thirdly, explicit self-esteem seems to be most strongly related to the
Implicit Self-Esteem, Personality, and Affect Regulation
63
personality traits Extraversion and Emotional Stability (Robins, Tracy, &
Trzesniewksi, Potter, & Gosling, 2001). These findings suggest that implicit
and explicit self-esteem will similarly relate to both Emotional Stability and
Extraversion.
Another interesting question that arises from the previous paragraph taps
into the specific role both implicit and explicit self-esteem play in their
relation to affective experience. As mentioned before, both Robinson and
Meier (2005) and Conner and Barrett (2005) showed a relation between both
implicit and explicit self-esteem and affective experience, with both higher
implicit and explicit self-esteem resulting in less negative affect in daily live.
Interestingly, in the study of Connor and Barrett (2005) implicit self-esteem,
as opposed to explicit self-esteem, did not predict a measure of positive affect
in daily live, suggesting that especially implicit self-esteem might be
specifically related to coping with self-threatening experiences. Because their
study was not set out to test what kind of daily experiences resulted in these
effects it is, however, not clear whether high implicit self-esteem is indicative
of less experienced negative affect in general or whether it actually
specifically pertains to self-threatening and not to self-affirming experiences.
In another study, Robinson and Meier (2005) did demonstrate that implicit
self-esteem predicted experimenter-reported negative affect after a self-threat
manipulation, but because they did not include a control or self-affirming
condition, and did not measure self-reported affect, it still remains unclear
whether the relation between implicit self-esteem and (negative) affect is
unique to the self threatening circumstances, and if so, whether this relation
would also apply to the actual affective experience of an individual.
As mentioned before, the purpose of our present studies was set out to
gain both a better insight into the common grounds of implicit and explicit
self and to further our understanding in identifying the circumstances when
both forms of self-esteem might differ in relative importance. Based on the
evidence presented before we expected that general affect is an essential
aspect of both implicit and explicit self-esteem processes, while the kind of
Chapter 3
64
affect-regulation required might differ between the two. This is why in Study
1 we first examined the combined predictive validity of implicit and explicit
self-esteem for those personality dimensions that have a strong affective
component. In Study 2 and 3 we specifically focused on the relation between
both implicit and explicit self-esteem and affect regulation under self-
threatening versus self-affirming circumstances.
Study 1
In Study 1, we set out to show that implicit self-esteem has incremental
validity over and above explicit self-esteem, but only in predicting those
personality dimensions that have a strong affective component. We used the
Dutch Personality Inventory (DPQ; Luteijn, Starren & Van Dijk, 2000) to
measure personality, so we could assess the predictive validity of implicit and
explicit self-esteem on the ‘affective’ personality measures Emotional
Stability and Extraversion and compare it with the predictive validity on two
additional personality measures, Conformity and Egoism, that do not have a
strong affective component.
Furthermore, we wanted to validate our contention that implicit and
explicit self-esteem should share common variance in predicting Emotional
Stability and Extraversion under circumstances when implicit and explicit
self-esteem are related. This is why we chose to use the name letter test as a
measure of implicit self-esteem and conduct our study among a large enough
sample to insure adequate power to detect small to medium correlations.
Method
A time 1, 384 participants (269 female) completed the name letter test and the
Heatherton and Polivy’s (1991) State Self-Esteem Scale (SSES) on a PC-
computer. All participants were undergraduates from the University of
Amsterdam and received course credits for participation. The order of the
name letter test and SSES was counterbalanced. At time 2, approximately one
week later, 362 of these participants (257 female) completed the Dutch
Implicit Self-Esteem, Personality, and Affect Regulation
65
Personality Questionnaire (DPQ). In the Netherlands, the DPQ (Luteijn,
Starren & Van Dijk, 2000) is the most used personality questionnaire that
measures seven personality traits, which can be reduced to four personality
dimensions, Emotional Stability, Extraversion, Conformity and Egoism. The
dimensions Emotional Stability and Extraversion are equal to their Big Five
counterparts, the dimension Conformity resembles the Big Five-factor
Conscientiousness and the dimension Egoism is only measured with the DPQ
(Barelds & Luteijn, 2002). At time 3, approximately 5 months later, 54 of the
384 participants completed the name letter test and the SSES as premeasures
of another experiment (in the section of Study 3 full details are presented).
We included these name letter scores and SSES from these participants to test
the stability of the relations between implicit self-esteem, explicit self-esteem
and personality. At the end of the experiment participants provided their full
names on an informed consent form.
The name letter test. We administered the name letter test as follows:
Participants were informed that they were about to evaluate simple stimuli, in
this case evaluating the attractiveness of the alphabet letters. We asked them
to give their first impression on each letter and not to think too long but to
respond intuitively. Letters were always presented in random order. At both
times participants indicated their liking of every letter by clicking on the
corresponding area of a 9-point scale (time 1) or a continuous scale (time 3),
ranging from ugly to beautiful. A marker appeared at the area a participant
clicked on and by clicking the ‘next’ button, they could continue with the next
evaluation. On the continuous scale a score of 10 represented the right end of
the scale (i.e., beautiful), a score of -10 the left end of the scale (i.e., ugly),
and a score of zero the middle of the scale (i.e., neutral).
We calculated the name letter score according to the proposed method by
Albers, Dijksterhuis Rotteveel (2009, Chapter 2 of this thesis). First, we
calculated the average evaluation of someone’s initials (NLE), the average
evaluation of those specific letters rated by participants for whom it is not a
Chapter 3
66
name letter (SLE) and someone’s average evaluation of the alphabet letters
excluding all not-name letters (NNL). We then regressed NNL and SLE onto
NLE, and calculated the name letter score (NLS) as depicted in Equation 1:
NLS = NLE – (B1 * NNL + B2 * SLE) (1)
In this equation, B1 and B2 are the unstandardized regression coefficients
when simultaneously predicting NLE from both NNL and SLE. We excluded
NNL or SLE from Equation 1, when its corresponding unstandardized
regression coefficient (i.e., B1 or B2) showed no significant relation with NLE
(i.e., p > 0.1).The resulting name letter score (NLS) is, by definition,
unrelated to both not-name letter evaluation and the specific alphabet letter
evaluation. This insures that the name letter effect is not confounded with
factors that influence general letter liking instead of specific name letter
liking.
Results
In none of the analyses we conducted did we find any gender differences or
order effects. We calculated the name letter score according to Equation 1. To
calculate the name letter score at time 1, we included both NNL and SLE in
the equation because the regression analysis showed that both the
corresponding unstandardized regression coefficients where significant,
B1 = .61, t(381) = 7.1, p < .000; B2 = .67, t(381) = 2.6, p < .009, and
calculated the name letter score (NLS) as depicted in Equation 2.
NLS = NLE – (.61 * NNL + .67 * SLE) (2)
At time 3, we calculated the name letter score as depicted in Equation 4
(presented in the result section from Study 3).
Correlations between implicit and explicit self-esteem. To explore the
Implicit Self-Esteem, Personality, and Affect Regulation
67
relation between implicit and explicit self-esteem we correlated the name
letter score and SSES, at both time 1 and 3. The results are presented in Table
3.1, showing that, albeit moderately to small, at both time 1 and time 2,
implicit and explicit self-esteem correlated significantly with each other.
Furthermore, the relation between implicit and explicit self-esteem measured
at different times (i.e., correlating self-esteem at time 1 with time 3) remained
significant, suggesting that the relation between implicit and explicit self-
esteem is relative stable over time.
Table 3.1 Correlations Between Personality Traits, Implicit and Explicit Self-esteem
N Implicit Explicit N Implicit Explicit
SE1 SE1 SE2 SE2
Explicit SE1 384 .19***
Implicit SE2 53 .54*** .28*
Explicit SE2 53 .36** .82*** 115 .34***
Extraversion 362 24*** .42*** 49 .43** .60***
Emotional Stability 362 .22*** .58*** 49 .41** .72***
Conformity 362 .04 -.24*** 49 -.01 -.25†
Egoism 362 .01 -.23*** 49 -.11 -.39**
Note. SE1 and SE2 are self-esteem scores at time 1 and 2 respectively.
† p < .10. * p < .05. ** p < .01. *** p < .001.
Correlations between self-esteem and personality traits. To explore the
relation between personality and both implicit and explicit self-esteem we
correlated the name letter score and SSES, measured at both time 1 and time
3, with the personality traits administered at time 2. We present these
correlations in Table 3.1. Explicit self-esteem correlated highest with
Emotional Stability and Extraversion, at both time 1 and time 3, supporting
previous findings (Robins et al., 2001) that high self-esteem individuals tend
to be emotionally stable and extraverted. Conformity and Egoism both
correlated negatively with explicit self-esteem, suggesting that high self-
Chapter 3
68
esteem individuals are less rigid and egotistic. Implicit self-esteem also
correlated significantly at both time 1 and time 3 with the personality
dimensions Emotional Stability and Extraversion, suggesting that similar to
high explicit self-esteem individuals, high implicit self-esteem individuals are
also emotional stable and extraverted. As expected, implicit self-esteem did
not relate to the personality dimensions Egoism and Conformity, showing that
implicit self-esteem is indeed specifically related to those personality traits
that are most strongly related to affective experience.
The extra predictive validity of implicit self-esteem. The previous
results show that implicit self-esteem shares variance with both explicit self-
esteem and the personality dimensions Emotional Stability and Extraversion.
However, it remains unclear whether implicit and explicit self-esteem are
independently related to these personality dimensions, or that the relation
between implicit self-esteem and personality is in fact mediated by explicit
self-esteem. To determine this, we first conducted two mediation analyses
using the SPSS-Macro from Preacher and Hayes (2005), with implicit self-
esteem and explicit self-esteem, measured at time 1, and the specific
personality dimension measured at time 2 as variables. Mediation analysis
with Emotional Stability as dependent and implicit self-esteem as independent
showed only partial mediation of explicit self-esteem: bootstrapping resulted
in a significant indirect effect of 0.14, p < .01, while the direct effect of
implicit self-esteem on Emotional Stability remained significant after
controlling for explicit self-esteem, β = .13, t(359) = 2.9, p < .004. A
mediation analysis with Extraversion as dependent and implicit self-esteem as
independent also showed partial mediation of explicit self-esteem:
bootstrapping resulted in a significant indirect effect of 0.10, p < .01, while
the direct effect of implicit self-esteem on Extraversion remained significant
after controlling for explicit self-esteem, β = .18, t(359) = 3.8, p < .000.
We then tested whether these partial mediation effects of implicit self-
esteem would sustain using the implicit and explicit self-esteem measures
Implicit Self-Esteem, Personality, and Affect Regulation
69
administered at time 3. Though sample size was considerable smaller (N =
49), a mediation analysis with Extraversion as dependent variable and implicit
self-esteem as independent variable again showed only partial mediation of
explicit self-esteem: bootstrapping resulted in a significant indirect effect of
.27, p < .05, while the direct effect of implicit self-esteem on Extraversion
remained significant after controlling for explicit self-esteem, β = .30, t(46) =
2.7, p < .01. We obtained similar results for Emotional Stability. A mediation
analyses only showed partial mediation of explicit self-esteem: bootstrapping
resulted in a marginal significant indirect effect of 0.32, p < .07, while the
direct effect of implicit self-esteem on Extraversion remained significant after
controlling for explicit self-esteem, β = .26, t(46) = 2.6, p < .01.
Discussion
Taken together these results show that implicit and explicit self-esteem are
related to the personality traits Emotional Stability and Extraversion over an
extended period of time, but also that each have their independent relation
with these personality traits. Since explicit self-esteem related to all
personality traits and implicit self-esteem related only to Emotional Stability
and Extraversion and not to Conformity and Egoism, these results further
show that implicit and explicit self-esteem share variance in predicting those
personality traits that have a strong affective component, suggesting that the
common grounds of implicit and explicit self-esteem reside in affect-
regulation.
Study 2
In our second study we wanted to investigate the differences between implicit
and explicit self-esteem by assessing the relative importance of the two forms
of self-esteem as a function of different forms of affect regulation. We
expected implicit self-esteem to be especially important when affect
regulation due to self-threat is required and explicit self-esteem when the self
is affirmed. First of all, Robinson and Meier (2005) showed that under ego
Chapter 3
70
threat only implicit self-esteem and not explicit self-esteem was a significant
predictor of experimenter reported negative affect. Secondly, Meager and
Aidan (2004) reported similar results, demonstrating that only implicit self-
esteem was predictive of a participants’ implicit mood after negative
feedback, while explicit self-esteem only related to the evaluation of the
confederate who provided the negative feedback.
This is why, in our second study, we tested whether implicit self-
esteem would relate differently to affective experience after self-threat
compared to a self-reinforcing condition. Because the research mentioned
before suggests that implicit self-esteem is especially related to negative
affect we expected implicit self-esteem to be more strongly related to
experienced affect after the self is threatened than after the self is boosted.
Method
Fifty-eight participants (40 female) were randomly assigned to a self-
enhancing or self-threat condition. In the self-enhancement condition
participants described a recent positive self-relevant experience, in the self-
threat a recent negative self-relevant experience. After describing the personal
event they then indicated how they felt on two continuous scales ranging from
respectively very negative to very positive, and very angry to very cheerful.
Scores on both scales were aggregated to form one measure of experienced
affect (α = .82), scores ranging from -20 to 20, with positive scores reflecting
positive affect and negative scores reflecting negative affect. After this they
completed the name letter test as described in Study 1, with the exception that
we only obtained the initials of the participants and not their full names at the
end of the experiment. Upon completion of the name letter test participants
filled out the Rosenberg Self-Esteem Scale (Rosenberg, 1965). We calculated
the name letter score according the method depicted in Equation 1. Because in
this study we only obtained the initials of the participants, we calculated
someone’s average letter evaluation excluding the initials of that person.
Implicit Self-Esteem, Personality, and Affect Regulation
71
Results
We calculated the name letter score according to Equation 1. Because the
regression analysis showed that only the unstandardized regression coefficient
of NNL was significant, B1 = .98, t(56) = 4.4, p < .000, we only included
NNL and not SLE¹, resulting in Equation 3:
NLS = NLE – .98 * NNL (3)
Affect. Experienced affect in both conditions differed significantly, F(1,
56) = 266.9, p < .0001. In the self-threat condition participants experienced
negative affect, M = -3.7, SD = 4.5, while in the self-enhancing condition
participants experienced positive affect, M = 14.0, SD = 3.7, suggesting that
the manipulation was successful. Neither implicit nor explicit self-esteem
differed between both conditions, respectively F(1, 56 ) = 1.58, p = .21 and
F(1,56) = .02, p = .90. Though implicit and explicit self-esteem only
correlated in the self-threat condition, r = .45, p < .02 and not after the self
was enhanced, r = .15, p = .42, these correlations did not differ significantly
from each other, z = 1.2, p = .24.
To determine whether implicit self-esteem and experienced affect were
more strongly related in the self-threat condition compared to the self-
enhancing condition we conducted a regression analyses with condition, the
centered measure of implicit self-esteem and the corresponding two-way
interaction as predictors and experienced affect as dependent. This resulted in
a significant interaction between implicit self-esteem and condition, β = .16,
t(54) = 2.0, p < .05. As expected, separate regression analysis for both the
self-enhancement and self-threat condition with implicit self-esteem as
predictor and experienced affect as dependent variable, showed that in the
self-threat condition implicit self-esteem significantly predicted experienced
affect, β = .58, t(26) = 3.6, p < .001 while in the self-enhancement condition
implicit self-esteem failed to reach significance as a predictor of experienced
affect, β = .1, t(28) = .51, p = .61, showing that only after self-threat higher
Chapter 3
72
implicit self-esteem was predictive of less negative experienced affect.
Similar regression analysis with the centered measure of explicit self-
esteem, condition, and the corresponding two-way interaction as predictors,
and experienced affect as dependent, did not result in a significant interaction
between explicit self-esteem and condition, β = .02, t(54) = .3, p = .77. As
expected, separate regression analysis for both the self-enhancement and self-
threat condition with explicit self-esteem as predictor and experienced affect
as dependent variable showed that in both conditions explicit self-esteem
significantly predicted experienced affect; in the self-threat condition, β = .41,
t(26) = 2.3, p < .03; in the self-enhancement condition, β = .38, t(28) = 2.2, p
< .04, showing higher explicit self-esteem was predictive of less negative
experienced affect, regardless whether the self was threatened or enhanced. In
the self-threat condition explicit self-esteem no longer predicted experienced
affect after controlling for implicit self-esteem, β = .18, t(25) = 1.0, p = .32,
suggesting that implicit self-esteem fully mediated the relation between
explicit self-esteem and experienced affect after the self is threatened. We
confirmed this in a mediation analysis with experienced affect as dependent
and explicit self-esteem as independent showing full mediation of implicit
self-esteem: bootstrapping resulted in a significant indirect effect of .11, p <
.02.
These results show that while explicit self-esteem is similarly related to
experienced affect regard after both self-enhancement and self-threat, implicit
self-esteem is specifically related to experienced affect only after the self is
threatened and under these circumstances even fully mediates the relation
between explicit self-esteem and experienced affect.
Discussion
The results from this study support previous research suggesting that implicit
self-esteem is related to affective processes especially when the self is
threatened. Besides the specific influence implicit esteem has under these
circumstances on someone’s implicit mood (Meager & Aidan, 2004) and
Implicit Self-Esteem, Personality, and Affect Regulation
73
someone’s observed negative affect (Robinson & Meier, 2005) we
demonstrated this also to be true for someone’s (explicitly) experienced
affect. Because in our study we included a self-enhancement condition we
could further demonstrate that this relation between implicit self-esteem and
affect is indeed specific to the self-threatening circumstances and not to any
circumstances where the self is activated.
But moreover, our results also extend previous research by showing that
implicit and not explicit self-esteem is at the core of coping with negative
affect. Because explicit self-esteem could not differentiate between
someone’s experienced affect after self-threat compared to self-enhancement
it suggests that the ability of explicit self-esteem to predict experienced affect
rather reflects subjective well being than a specific process of coping with
negative affect. On the other hand, implicit self-esteem did differentiate in
affective experience when affective coping was required, showing that high
implicit self-esteem individuals are more resilient against threats to the self.
Because in this study we measured both implicit and explicit implicit
after self-threat we should be careful in interpreting these results when
denoting implicit and explicit self-esteem as predictors of affect regulation.
Though in our study we did not find clear influences of condition on implicit
and explicit self-esteem, we only found the expected significant correlation
between implicit and explicit self-esteem after the self was threatened, and not
after the self was affirmed (though both correlations did not differ
significantly from each other). This means that experienced affect itself may
have influenced both forms of self-esteem, so it remains unclear whether
implicit and explicit esteem are similar predictive of someone’s affective
experience when measured before the experimental manipulation.
Study 3
In order to investigate the specific roles both implicit and explicit self-esteem
play in predicting experienced affect, in Study 3 we made two changes
compared to Study 2. First of all, both measures of self-esteem were
Chapter 3
74
administered before participants were submitted to a self-enhancing or self-
threatening manipulation. Secondly, participants were submitted to a different
experimental manipulation in order to increase the external validity of our
research.
Method
At the start of the experiment, 117 participants (93 female) completed the
name letter test
followed by the Heatherton and Polivy’s (1991) State Self-Esteem Scale
(SSES). We assigned them randomly to either a self-threat or a self-
reinforcing condition. In both conditions participants first rated the
attractiveness of same-sex faces compared to their own attractiveness by
clicking on a continuous scale, ranging from ugly to beautiful, with a score of
25 represented the right end of the scale (i.e., beautiful), a score of -25 the left
end of the scale (i.e., ugly), and a score of zero the middle of the scale (i.e.,
neutral). We selected the faces from a set of faces that we pre-tested on
attractiveness, such that in the self-threat condition, participants rated eight
attractive faces and in the self-reinforcing condition eight unattractive faces.
In both conditions, participants first rated four neutral faces. After rating the
faces participants indicated how they felt by clicking on two continuous
scales ranging from respectively very negative to very positive, and from not
in the least cheerful to very cheerful, with a score of 25 represented the right
end of the scale (i.e., very positive or very cheerful), a score of -25 the left end
of the scale (i.e., very negative or not in the least cheerful), and a score of zero
the middle of the scale (i.e., neutral). Scores on both scales were averaged to
form one measure of experienced affect (α = .94), with positive scores
reflecting positive affect and negative scores reflecting negative affect.
Results
In none of the analyses we conducted did we find any gender differences. We
calculated the name letter score according to Equation 1. To calculate the
Implicit Self-Esteem, Personality, and Affect Regulation
75
name letter score, we only included SLE in the equation because the
regression analysis including NNL showed no significant corresponding
unstandardized regression coefficient for NNL, B1 = .23, t(114) = 1.3, p =
.21, resulting in a significant corresponding unstandardized regression
coefficient of SLE, B2 = .82, t(115) = 2.9, p < .004, and calculated the name
letter score (NLS) as depicted in Equation 4.
NLS = NLE – .82 * SLE (4)
As a manipulation check, we compared the rating of the faces in both
conditions., The average rating of the faces in the self-threat condition
differed significantly from the average rating in the self-enhancing condition,
F(1,115) = 290, p < .0001. Participants rated the attractive faces as being
more attractive than themselves, M = 9.2, SD = 6.1, and the unattractive faces
as less attractive than themselves, M = -13.0, SD = 7.9, suggesting that the
manipulation was successful.
To test whether implicit and explicit self-esteem were differently
related to experienced affect depending on condition (i.e., self-enhancement
versus self-threat), we conducted a regression analysis with the dummy coded
condition, the centered measures of implicit and explicit self-esteem and the
corresponding two-way interactions between condition and both implicit and
explicit self-esteem as predictors, and experienced affect as dependent,
resulted in the expected significant interaction between implicit self-esteem
and condition, β = .23, t(110) = 2.1, p < .03, while both the interaction
between explicit self-esteem and condition and between explicit and implicit
self-esteem were not significant (both ps > .58). Separate regression analysis
for both the self-enhancement and self-threat condition with implicit self-
esteem as predictor and experienced affect as dependent variable, showed that
though in the self-enhancement condition implicit self-esteem failed to reach
significance as a predictor, β = .16, t(57) = .51, p = .61, in the self-threat
condition implicit self-esteem indeed did significantly predict experienced
Chapter 3
76
affect, β = .40, t(56) = 3.3, p < .002, and remained significant after controlling
for explicit self-esteem, β = .25, t(55) = 2.5, p < .02, suggesting that explicit
self-esteem only partially mediated the relation between implicit self-esteem
and experienced affect. We confirmed this in a mediation analysis with
experienced affect as dependent variable and implicit self-esteem as
independent variable showing partial mediation of explicit self-esteem:
bootstrapping resulted in a significant indirect effect of .43, p < .05.
In order to try to replicate the findings described in the previous
paragraph and test whether the reported effects of self-esteem on affective
experience would sustain over an extended period of time we conducted
similar analyses but now with the measures of both implicit and explicit self-
esteem and personality that we assessed approximately 5 months before the
current study (for a detailed description see Study 1). The results support our
previous findings. Though sample size was considerable smaller, implicit
self-esteem still significantly predicted, over an extended period of
approximately 5 months, experienced affect when the self was threatened, β =
.50, t(24) = 2.8, p < .01, but not when the self was enhanced, β = .06, t(25) =
.30, p = .77. Explicit self-esteem and both affective personality dimensions
(i.e., Extraversion and Emotional Stability) did not differentiate between
conditions and all predicted experienced affect both after the self was
threatened or enhanced (all ps < .09), while both ‘non-affective’ personality
dimensions Conformity and Egoism did not predict experienced affect in
either condition (all ps > .27).
Discussion
The results from this study support the idea that implicit self-esteem and not
explicit self-esteem functions as a buffer when the self is threatened (Koole &
DeHart, 2007). In our study, high implicit self-esteem individuals experienced
less negative feelings opposed to low implicit self-esteem individuals when
the self was threatened, an effect that we even could demonstrate with our
measure of implicit self-esteem that we collected 5 months prior to the actual
Implicit Self-Esteem, Personality, and Affect Regulation
77
self-threat. This specific ability for implicit self-esteem to do so was even
further supported by our findings that explicit self-esteem, Extraversion and
Emotional Stability could not do so.
General Discussion
The results from our studies support and extend on the findings of Robinson
and Meier (2005) and Meager and Aidan (2004), showing that implicit and
explicit self-esteem share common grounds with those personality traits that
deal with affective experience and expression. We could furthermore
demonstrate that implicit self-esteem is especially important in dealing with
threats to the self. In fact, it was the only psychological construct we
measured that was specifically related to negative affect. Though our
measures of explicit self-esteem, Emotional Stability and Extraversion each
related to affective experience, they did not differentiate between
circumstances when the self was threatened or boosted.
In our first study we demonstrated that implicit self-esteem, explicit
self-esteem, and the two personality dimensions Extraversion and Emotional
Stability are all intrinsically related. Implicit and explicit self-esteem
correlated positively and participants with high implicit and explicit self-
esteem were emotionally more stable and extraverted. Though these
correlations between implicit self-esteem and the explicit measures were only
moderate to small in nature, the variance they shared stayed stable across
time. Because implicit and explicit related to each other, we could also
address the nature of their shared relation with explicit personality traits.
Though explicit self-esteem mediated the relation between implicit self-
esteem and the personality traits Extraversion and Emotional Stability, in
general we did not found full mediation, suggesting that part of the relation
between implicit self-esteem and (explicit) personality occurs independent
from explicit self-esteem, and thus demonstrates the additional value of
measuring implicit self-esteem. Unlike explicit self-esteem, implicit self-
esteem was not related to the personality dimensions Egoism and
Chapter 3
78
Conformism. These results suggest that the common grounds between
implicit and explicit self-esteem reside predominately in personality traits that
are most strongly connected to affective experience (DeNeve, & Cooper,
1998), and support the idea that affective regulation is a core function of self-
esteem.
In our second and third study we elaborated on the role of self-esteem
in affective regulation. Based on previous findings of Robinson and Meier
(2005) and Meager and Aidan (2004) and the contention that implicit self-
esteem functions as a buffer to negative affect (Koole & DeHart, 2007) we
expected the function of implicit self-esteem to reside especially in self-
threatening circumstances. This is also what we found. Though explicit self-
esteem and the personality dimensions Extraversion and Emotional Stability
were predictive of a general experienced affect, implicit self-esteem was
specifically related to experienced affect under self-threatening
circumstances.
These results from our studies suggest that that the common grounds of
implicit and explicit self-esteem reside mainly in the unaffected self, but
diverge in function when the self is affected. Regardless of the way the self is
affected, explicit self-esteem remains indicative of an individual’s general
level of positive or negative affect. The opposite is true for implicit self-
esteem: high implicit self-esteem individuals feel less negatively affected by
threats to their selves. In this respect our studies support the contention that
implicit self-esteem functions as buffer that only comes into play when it is
needed.
Implicit Self-Esteem, Personality, and Affect Regulation
79
Endnotes
¹Including both NNL and SLE into the regression when predicting
NLE showed that the corresponding unstandardized regression coefficient of
NNL was significant, B1 = .98, t(55) = 4.4, p < .000, while the corresponding
unstandardized regression coefficient of SLE was not significant, B2 = -.26,
t(55) = -.57, p = .57.
81
Chapter 4
Digging for the real attitude: Lessons from research on implicit
and explicit self-esteem*
*This chapter is based on: Dijksterhuis, A., Albers, L., & Bongers, K. (2008). Digging for
the real attitude: Lessons from research on implicit and explicit self-esteem. In R. E. Petty,
R. H. Fazio, & P. Briñol (Eds.), Attitudes: Insights from the new wave of implicit measures
(pp. 229-250). New York: Psychology Press.
Chapter 4
82
It has once been argued that attitude formation and attitude expression are
more reminiscent of architecture than of archeology. Rather than uncovering
true, deeper beliefs and values, people’s attitudinal expressions are the result
of often distorted, temporary constructions created on the spot (Bettman, Luce
& Payne, 1998; see also Dijksterhuis & Nordgren, 2006). Researchers
administer questionnaires aimed at measuring how people think about
themselves, about George W. Bush, or about Chocolate Chip cookies, and
people construct what we call attitudes. Just as there is nothing wrong with
architecture, there is nothing inherently wrong with measuring attitudes with
questions, that is, explicitly. Usually, the goal behind measuring attitudes is to
predict behavior, and indeed, explicitly measured attitudes often do: We are
fairly positive about ourselves helping us to navigate life reasonably well, we
vote against George W. Bush, and we eat way too many Chocolate Chip
cookies.
Still, there is something about this practice that makes it somewhat
unsatisfactory. If an old manuscript suggests an undiscovered tomb in the
Egyptian desert, we send archeologists to find it, rather than ask architects to
recreate the tomb on the basis of some vague descriptions. Have you ever
been in a museum, staring in awe at some beautiful piece of old art, only then
to discover (by reading the brochure that was handed to you at the entrance)
that you are looking at a replica? The real statue created by Michelangelo is in
an area inaccessible to the public, and you are looking at a copy made in 1987
in some Florentine factory. Even in such cases, when a replica looks exactly
like the original, looking at it just does not feel quite right. It is nothing more
than a minor nuisance, and it certainly does not spoil your entire day, but you
had preferred to see the real thing.
One way to look at explicitly measured attitudes is to assume that what
one measures is all there is. Explicitly measured attitudes are what they are,
and there are no such things as underlying, “real” attitudes. This is
unsatisfactory of course, as we know enough about unconscious affective and
cognitive processes to assume that a “7” on a 9-point Likert-scale about
Digging for the Real Attitude
83
Chocolate Chip cookies is not just something hovering in the air. It must
come from somewhere, somehow. It may be the result of a construction
process, but we hope it is, at least to some extent, influenced by deeper
psychological forces. Hence, an alternative viewpoint that is more realistic
(and much more exciting) is to assume that there are such things as “real”
attitudes and that what we assess with attitude measures is at least in part
based on this real thing. Sure, due to poor construction we routinely end up
with very poor replica’s, indeed much more reminiscent of the work of
architects working on the basis of vague descriptions than of contemporary
Florentine artists who can make a detailed copy of the Michelangelo statue.
After all, when we answer a questionnaire we often have not much to work
with other than perhaps some vague hints, such as subtle affective reactions or
old memories of past behavior. Still, somewhere in that Egyptian desert is the
real thing we are looking for. Some process sparked by our millions of brain
cells represents that real attitude.
The observation that attitudes are more the result of architecture rather
than of archeology was made before psychologists started to develop implicit
attitude measures. In our view, the creation and development of implicit
measures is of paramount importance, because it fundamentally changes the
way we think (at least it should) about attitudes. In the present chapter, we
will argue that implicit attitude measurement is not just another style of
architecture. Instead, we will review evidence (and present some new
evidence) strongly suggesting that implicit measurement reflects archeology.
Amateur archeology with limited equipment perhaps, but archeology.
In this chapter, we focus on arguably the most important attitude we
have: Self-esteem, or the attitude towards the self. However, our hope is that
our thinking is generalizable to attitudes in general. Most of this chapter deals
with research on the relation between three protagonists: Explicitly measured
self-esteem, implicitly measured self-esteem, and that what we until now
called “the real thing”. We start out by defining what we conceive of as this
hypothesized “real” attitude and by proposing three hypotheses concerning
Chapter 4
84
the relationship between explicit and implicit self-esteem. We then discuss
relevant research with the aim to differentiate between these hypotheses and
to decide which of the three is the most plausible. Before we end with some
conclusions, we present an experiment that we recently conducted.
Three alternative hypotheses on the relation between explicit and implicit
self-esteem.
A chapter that features the term “the real thing” too often runs the risk of
alienating a scientific audience (but perhaps attracting people from the music
industry!), so a definition is in order. The “real” attitude, we propose, is the
evaluative “tone” that is automatically activated upon the perception of the
attitude object (Bargh, Chaiken, Govender & Pratto, 1992; Fazio,
Sanbonmatsu, Powell & Kardes, 1986). With more multifaceted and/or
important attitude objects, it is perhaps more appropriate (see e.g., Cacioppo,
Crites, Berntson & Coles, 1993) to define the attitude as the sum of the
various evaluative tones that are automatically activated. It is a proposed
underlying construct constituting the core of the attitude, sitting there waiting
to be excavated. It is itself undisturbed by biasing processes that occur when
the attitude is measured or verbalized, but at the same time it feeds such
processes. From now on, we call it the core attitude or core self-esteem (see
also Dijksterhuis, 2004).
What is the relation between this core attitude and implicitly and
explicitly measured attitudes? Or, to turn to self-esteem, is there one core self-
esteem that is related to both measures of implicit and explicit self-esteem?
And if so, how are they related? Let us briefly discuss three possible
hypotheses pertaining to this relation.
1. The independence hypothesis. According to this hypothesis, implicit
and explicit self-esteem are independent constructs. They happen to partly
share their name, they happen to be about the same object, but they are
Digging for the Real Attitude
85
unrelated. Both implicitly measured self-esteem and explicitly measured
esteem are based on their own underlying core construct. Implicit self-esteem
could be based on the automatically activated evaluative tone, whereas
explicit self-esteem could be based on the evaluative tone that becomes
apparent only when one explicitly or consciously reflects on the self.
2. The equal relationship hypothesis. This hypothesis states that
implicit and explicit self-esteem are related because they are both related to
the same core attitude, the one defined above. They are simply different
manifestations of this core. In addition, the two manifestations do not differ as
to how well they represent that core. They measure a different aspect, but
generally do equally well.
3. The hierarchy hypothesis. This hypothesis also assumes that implicit
and explicit self-esteem are related because they are both related to the same
core. However, here implicit measures of self-esteem better represent core
self-esteem than explicit measures of self-esteem. That is, implicit self-esteem
digs deeper and more closely approaches the hidden Egyptian tomb. This
hypothesis also implies that what we measure explicitly is partly based (and
can be partly predicted by) what we find when we measure implicitly1.
In what follows, we will make a (stepwise) comparison between the
plausibility of the hypotheses by reviewing evidence. We will start with
comparing the independence hypothesis with the remaining two, the equal
relationship hypothesis and the hierarchy hypothesis, whereby no distinction
will be made between latter two yet.
Before we move on, it should be noted that although we only review
evidence on self-esteem, the three hypotheses encompass possible relations
between implicit and explicit measures of attitudes in general. Indeed,
versions of both the independence hypothesis and the equal relationship
hypothesis shine through in work on racial attitudes (i.e., prejudice, Dovidio,
Kawakami, Johnson, Johnson & Howard, 1997). Likewise, the hierarchy
Chapter 4
86
hypothesis is in part based on, and fully in line with, the work by Fazio and
colleagues on the MODE-model (see e.g., Fazio, 1990; Olson & Fazio, 2008).
The MODE-model also views an attitude as a “core” that can be automatically
activated, whereby implicit measures are more proximal indicators of these
automatically activated attitudes than more downstream explicit measures.
Are implicit and explicit self-esteem related?
If we find evidence for the notion that implicit and explicit self-esteem are
related, this implies that both the equal relationship hypothesis and hierarchy
hypothesis are more plausible than the independence hypothesis. In our view,
there are currently three relevant sets of research findings. First, quite a
number of researchers have directly investigated the relation between implicit
and explicit self-esteem by assessing correlations between the two. A second
fruitful avenue is to investigate whether the same specific levels of implicit
and explicit self-esteem have the same or comparable consequences for other
psychological processes. A third way to shed light on the relationship between
implicit and explicit self-esteem it to see if there are experimental
manipulations that affect both implicit and explicit self-esteem in comparable
ways.
Are implicit and explicit self-esteem correlated? The answer is “sort
of”. Some researchers did not find correlations (Baccus, Baldwin & Packer,
2004; Bosson, Swann & Pennebaker, 2000; Jordan, Spencer, Zanna, Hoshino-
Browne & Correll, 2003; Spalding & Hardin, 1999), others did find
significant correlations (DeHart, Pelham & Tennen, 2006; Greenwald &
Farnham, 2000), yet others found significant correlations in some
experimental conditions or in some samples and not in others (Jones, Pelham,
Mirenberg, & Hetts, 2002; Koole, Dijksterhuis & van Knippenberg, 2001;
Pelham, et al., 2005). Various people have concluded that implicit and
explicit self-esteem correlate “weakly at best”. There is no arguing with that
conclusion, and on the basis of the current state of affairs we cannot say much
Digging for the Real Attitude
87
about the plausibility of the independence hypothesis. Rejecting the
independence hypothesis would have required more consistent correlations
between explicit and implicit self-esteem. On the other hand, given that some
researchers did find significant correlations, the correlational data cannot be
interpreted as support for the independence hypothesis either.
It may be noted that the generally low correlations between measures
of implicit and explicit self-esteem are at least in part caused by the fact that
implicit measures are still in a developing stage. Their reliability is often low
(Bosson, Swann & Pennebaker, 2000), and it is not fully understood yet what
exactly drives the effects of some of the implicit measures. Recently, various
researchers have proposed improvements to various measures of implicit self-
esteem. Both Karpinski (2004) and Albers, Dijksterhuis and Rotteveel (2009)
suggested improvements to implicit measures of self-esteem that will likely
result in more meaningful correlations between implicit and explicit measures
of self-esteem. Wentura, Kulfanek and Greve (2005) even proposed an
interesting new measure that alleviates some problems of other measures.
Such initiatives to strengthen implicit measures of self-esteem give rise to
optimism and it is likely that researchers will obtain higher and more
consistent correlations between implicit and explicit self-esteem in the future.
Do implicit and explicit self-esteem have comparable consequences?
Explicit self-esteem is known to be predictive of many things, but arguably
the best known fact is that it is related to how people cope with negative
experiences: High levels of explicit self-esteem help people cope with
negative feedback or negative experiences in general. High explicit self-
esteem forms a “buffer” against stress and experiences of failure (see e.g.,
Dodgson & Wood, 1998; Shrauger & Rosenberg, 1970; Steele, 1988). For
instance, it has been observed that people with low explicit self-esteem
exhibit stronger emotional reactions after failure than people with high
explicit self-esteem (Brown & Dutton, 1995) and that people with low explicit
self-esteem demonstrate impaired motivation after failure whereas individuals
Chapter 4
88
with high self-esteem generally do not (e.g., DiPaula & Campbell, 2002;
Shrauger & Rosenberg, 1970).
After only a few years of research on implicit self-esteem, we can
safely conclude that implicit self-esteem has indeed comparable
consequences. Spalding and Hardin (1999) demonstrated that low implicit
self-esteem individuals show more anxiety during a confronting interview
than high implicit self-esteem individuals. Greenwald and Farnham (2000)
showed that implicit self-esteem is negatively related to motivation after
failure such that people with low self-esteem show a stronger decrease in
motivation than people with high self-esteem. Baccus, Baldwin and Packer
(2004) demonstrated that people with high implicit self-esteem show less
aggression after an insult than people with lower implicit self-esteem. Finally,
Dijksterhuis (2004) showed that people with high implicit self-esteem show
no changes in mood after negative feedback, whereas people with lower
implicit self-esteem reported a more negative mood after negative feedback.
Indeed, high implicit self-esteem is a buffer against negative experiences, just
as high explicit self-esteem is. These findings strongly suggest that implicit
and explicit self-esteem are to some extent related, rendering the
independence hypothesis less plausible.
Do (some) experimental manipulations have the same effects on
implicit and explicit self-esteem? Currently, there are two areas of research
that indeed suggest this to be the case. First, threats to the self have been
known to decrease explicit self-esteem. For instance, both Dutton and Brown
(1997), and Heatherton and Polivy (1991) found that people report lower
explicit self-esteem after negative intelligence feedback. In recent years,
various researchers have reported comparable consequences of threats to the
self on implicit self-esteem. Jones, Pelham, Mirenberg and Hetts (2002) asked
participants to write about a negative aspect of their personality and
demonstrated that this lowered implicit self-esteem. In addition, Dijksterhuis
(2004) gave participants (bogus) negative intelligence feedback and found
Digging for the Real Attitude
89
that it lowered participants’ score on an implicit measure of self-esteem. It is
also known that people engage in self-affirming behavior in order to repair
“dents” in their self-esteem. And again, engaging in self-affirmation after
threat has been shown to both restore explicit (Steele, Spencer & Lynch,
1993), as well as implicit self-esteem (Koole, Smeets, van Knippenberg &
Dijksterhuis, 1999).
Secondly, it has been demonstrated that both explicit self-esteem and
implicit self-esteem can be changed by evaluative conditioning (Baccus,
Baldwin & Packer, 2004; Dijksterhuis, 2004; Riketta & Dauenheimer, 2003).
Evaluative conditioning (see De Houwer, Thomas & Baeyens, 2001, for a
review) is a technique in which is an attitude object (the Conditioned Stimulus
or CS [plural CSi]) is repeatedly paired with either a positive or a negative
stimulus (the Unconditioned Stimulus or US [plural USi]). After a number of
pairings, the CS takes on the valence of the USi. In our view, evaluative
conditioning is fascinating because it changes an attitude at its core. Earlier,
we defined the core attitude as the evaluative tone (or tones) that become
automatically activated upon the perception of the attitude object. And it is
this evaluative tone evaluative conditioning directly works on.
The procedures used by the different researchers differed only subtly.
Baccus et al., (2004) presented participants with self-relevant words (such as
their own names) on a computer screen, and in the experimental condition the
words were followed by smiling faces. In a control condition, self-relevant
words were randomly paired with smiling, frowning and neutral faces.
Dijksterhuis (2004) presented participants repeatedly with the word “I”, in the
experimental condition followed by positive adjectives. In the control
condition, neutral adjectives followed the word “I”. In some of the
experiments, all this information was presented subliminally. Riketta and
Dauenheimer (2003) followed almost exactly the same procedure, except that
whereas in the experiments by Dijksterhuis the adjectives immediately
followed the word “I”, in the Riketta and Dauenheimer experiments the word
“I” and the positive adjectives were presented simultaneously. Both Baccus et
Chapter 4
90
al., (2004) and Dijksterhuis (2004) assessed implicit self-esteem after the
evaluative conditioning procedure, whereas Riketta and Dauenheimer (2004)
measured self-esteem explicitly. Crucially, in all sets of studies it was found
that evaluative conditioning increased self-esteem2.
Where does this leave things? Although the findings on correlations
between implicit and explicit self-esteem are inconclusive, other evidence is
not. First, high (and low) implicit and explicit self-esteem have comparable
consequences for how people deal with negative experiences. Second, various
experimental manipulations (threat to the self, evaluative conditioning) have
the same effect on implicit as on explicit self-esteem. In our view, this makes
the independence hypothesis untenable. There is some sort of relation
between implicit and explicit self-esteem. Put differently, they must at least to
some extent represent the same underlying core.
Is implicit self-esteem closer to the core than explicit self-esteem?
Now that we have rejected the independence hypothesis, we can begin to
analyze which of the two remaining hypotheses is the most plausible. Is the
equal relationship hypothesis, whereby (measures of) explicit and implicit
self-esteem represent the underlying core attitude equally well (or equally
poorly) the best descriptor of the current state of affairs? Or are the relevant
findings better described by hierarchy hypothesis, stating that implicit
measures of self-esteem represents core self-esteem better?
In order for the hierarchy hypothesis to trump over the equal
relationship hypothesis, it has to be proven that explicit self-esteem is more
dissociated from the core attitude than implicit self-esteem. If this is true, it
should be possible to demonstrate why this dissociation is indeed more
pronounced. As our opening lines suggested, explicit attitudes are often active
constructions, more reminiscent of architecture than of archeology. A
prediction one can derive from the conceptualization of explicit self-esteem as
a construction process is that, since constructive processes are easier to
change than underlying representations, explicit self-esteem must be easier to
Digging for the Real Attitude
91
change than implicit self-esteem. Another way to differentiate between the
two hypotheses is to examine the construction process itself. Is there evidence
for the architectural aspect of explicit self-esteem? Can we find evidence for
biasing psychological forces that leads explicit self-esteem away from its
underlying core? In what follows, we first look at the changeability of explicit
and implicit self-esteem. Later, we examine the evidence for the notion that
explicit self-esteem is a construction partly based on biasing processes that
are not related to the core attitude.
Is explicit self-esteem easier to change than implicit self-esteem? We
already discussed evidence that shows that both implicit and explicit self-
esteem can be changed, at least for a brief period of time, by various
experimental manipulations. However, what can we say about more enduring
changes as a result of major life events?
There is indeed some evidence for a greater flexibility of explicit self-
esteem. Hetts and Pelham (2003) found people whose birthday was
overlooked (!) reported low implicit self-esteem, whereas their explicit self-
esteem was on a normal level. One could assume that when one’s birthday is
overlooked, this initially has negative consequences for both implicit and
explicit self-esteem. However, due to the assumed nature of explicit self-
esteem as more of an active construction process, explicit self-esteem can be
easier brought to more normal levels than implicit self-esteem. Although we
concede that this interpretation of the findings of Hetts and Pelham (2003) is
somewhat speculative, other research from Hetts, Pelham and colleagues
(1999) more firmly support the hierarchy hypothesis. Hetts, Sakuma and
Pelham (1999) assessed implicit and explicit self-esteem among Asian-
Americans who immigrated relatively recently. They reasoned that such
major life events would affect both explicit and implicit self-esteem, but that
it is more likely, due to the nature of explicit self-esteem measures, that
explicit self-esteem changes more quickly than implicit self-esteem. This is
exactly what they found. Whereas recent immigrants still demonstrated low
Chapter 4
92
implicit self-esteem, their explicit self-esteem soon appeared to be back to
normal levels. Fully in line with the hierarchy hypothesis, they concluded that
conscious constructions are more malleable than “deeper” unconscious
representations (see also DeHart, Pelham & Tennen, 2006).
Is explicit self-esteem a construction process? One could argue that
explicit self-esteem must largely be a construction process, simply because
people do not have much conscious access to deeper, unconscious processes.
Following Nisbett and Wilson (1977) one could reasonably assume that
explicit self-esteem (or explicit measures in general) relies on introspective
processes to an extent that is unwarranted and perhaps even unrealistic. We
simply do not know how we truly feel about ourselves, so we have no choice
but to engage in construction. We are architects working with poor and vague
instructions.
Pelham et al., (2005) recently reported evidence supporting this idea.
They reasoned that some people may have better access to how they truly feel
about themselves than others. Now the better people have access to core self-
esteem, the less need there is for construction. That means that, assuming
implicit self-esteem reflects core self-esteem better than explicit self-esteem,
implicit self-esteem and explicit self-esteem should correlate higher among
people who have better access to their core self-esteem. Pelham et al., (2005)
reasoned this could well mean that gender moderates the correlation between
implicit and explicit self-esteem. After all, aren’t women generally better at
accessing their deeper feelings than men? Socialization processes make
women trust their feelings and intuitions more (Pacini & Epstein, 1999) and
we know that women are generally better than men in expressing their
emotions (e.g., Lakoff, 1990). Pelham et al., (2005) compared six samples
from three different countries and indeed confirmed their prediction. Among
men, explicit and implicit self-esteem did not correlate in any of the samples,
whereas significant correlations were found in all samples for women
(ranging in size from .11 to .51).
Digging for the Real Attitude
93
Other evidence for explicit self-esteem as a construction comes from
studies suggesting that explicit self-esteem assesses factors other than the core
attitude towards the self. For example, various researchers have found that
explicit self-esteem correlates significantly with style of self-presentation,
impression management, and self-deception (Greenwald & Farnham, 2000;
Jordan, Spencer, Zanna, Hoshino-Browne & Correll, 2003; Raskin, Novacek,
& Hogan, 1991). These findings support the hierarchy hypothesis. Explicit
self-esteem may be assessing a mixture of core self-esteem, and various
essentially unrelated motives. Especially the finding that explicit self-esteem
is correlated with self-deception is interesting. The higher one’s explicit self-
esteem, the greater the possibility that people’s construction work reflects an
attempt to fool oneself.
If one is willing to assume that motivated construction takes effort, one
can derive a straightforward prediction from the notion that explicit self-
esteem correlates with various motives. Obviously, the people whose explicit
self-esteem does not reflect core self-esteem are the ones who have to engage
in effortful strategies to maintain this inconsistency. Specifically, people with
high explicit self-esteem but low implicit self-esteem are the true construction
workers. They engage in self-presentation and self-deception, thus, they
expend most effort. Conversely, given that implicit self-esteem represents this
core attitude quite well, people with comparable explicit self-esteem and
implicit self-esteem (both high or both low) do not engage in much
construction and hence, do not expend much effort.
There is indeed some evidence that maintaining high self-esteem in the
face of negative experiences requires work. People have to “explain things
away,” for instance by changing the way they interpret experiences or by
making self-serving attributions (e.g., Crocker & Major, 1989; Pelham,
DeHart & Carvallo, 2003). Importantly however, there is evidence that this is
especially true for people who want to maintain high explicit self-esteem in
the face of low implicit self-esteem (Bosson, Brown, Ziegler-Hill and Swann,
2003; Jordan, et al., 2003; McGregor & Marigold, 2003; see also Jordan,
Chapter 4
94
Logel, Spencer, Zanna & Whitfield, 2008).
Bosson et al., (2003) investigated two groups of participants. They
compared people with both high explicit and implicit self-esteem and people
with high explicit but low implicit self-esteem (often called fragile or
defensive self-esteem, see e.g., Kernis, 2003). They found that people with
low implicit self-esteem engaged more in unrealistic optimism. In addition,
they found evidence that supports the notion that high explicit self-esteem can
be related to self-deception. Participants were presented with four personality
profiles about themselves ranging from highly unflattering to highly
flattering. They were then asked to rate the accuracy of each of the profiles
(that were allegedly written by clinical psychology students) and as it turned
out, participants with low implicit and high explicit self-esteem rated the very
flattering profile as more descriptive of themselves than participants with both
high explicit and implicit self-esteem. The different profiles are given in the
Appendix to the Bosson et al., (2003) article and this makes the data even
more interesting, as the flattering profile is indeed rather extreme, including
the phrase “knows that affection and admiration from others are well-
deserved” (p. 183).
Jordan et al., (2003) distinguished between the same two groups:
People whose explicit and implicit self-esteem are high versus people whose
explicit self-esteem is high, while their implicit self-esteem is low. They first
established that people with low implicit self-esteem showed more narcissistic
behavior. In later experiments, they obtained more direct evidence for the idea
that maintaining high explicit self-esteem based on low implicit self-esteem
takes effort. They demonstrated that individuals with high explicit but low
implicit self-esteem showed much more defensive behavior. They engaged
more in in-group bias, they demonstrated more prejudice when threatened,
and they put more effort in dissonance reduction.
McGregor and Marigold (2003) investigated effects of personal
uncertainty on “compensatory conviction”. Conviction refers to the extremity
and certainty of important personal attitudes and compensatory conviction is
Digging for the Real Attitude
95
the tendency to increase the extremity of such attitudes and the commitment
with which such attitudes are held. People under uncertainty generally show
compensatory conviction, but McGregor and Marigold (2003) showed that
this is especially true for people with low implicit and high explicit self-
esteem. That is, relative to people with both high implicit and explicit self-
esteem, people with low implicit and high explicit self-esteem engaged in
more compensatory conviction regarding such moral topics as the death
penalty or abortion.
The work by Jordan and colleagues (2008) on people with low implicit
and high explicit self-esteem is consistent with our reasoning. Moreover, their
analysis sheds some more light on why people with low implicit and high
explicit self-esteem have to engage in defensive effort. Jordan et al., reason
that implicit self-esteem is not so much unconscious as it is preconscious.
Sometimes, especially in the face of threats, people become aware of their
(low level of) implicit self-esteem. This fleeting awareness is assumed to be
aversive among people with low implicit and high explicit self-esteem,
leading to what they call “nagging doubts”. These nagging doubts, in turn,
will motivate defensive effort. Jordan et al. (2008), present some interesting
first evidence for their reasoning.
To conclude, the research on people with high explicit and low implicit
self-esteem clearly shows that maintaining high explicit self-esteem when
implicit self-esteem is low is, at least sometimes, hard work – construction
work. In general, the evidence for explicit self-esteem as a construction
process with many inherent biases is strong, rendering the hierarchy
hypothesis more plausible than the equal relation hypothesis.
Finding support for the hierarchy hypothesis
With the hierarchy hypothesis coming out as the most plausible, in the last
part of this chapter we try to corroborate the hierarchy hypothesis by
discussing (and to some extent testing) the support for a few hypotheses
derived from the hierarchy hypothesis. The first hypothesis following from
Chapter 4
96
the hierarchy hypothesis is that there should be an asymmetry in frequency of
occurrence of different combination of implicit and explicit self-esteem. That
is, we can predict that the combination high explicit/low implicit self-esteem
is more common than the combination low explicit/high implicit self-esteem.
The second hypothesis pertains to the fact that, if we assume explicit self-
esteem is a construction, variations in the degree to which people engage in
active construction should affect the relation between implicit and explicit
self-esteem. That is, we expect implicit and explicit self-esteem to correlate
higher when there is less construction. Both hypotheses will be further
discussed, starting with the asymmetry hypothesis.
Is there an asymmetry? First, let us make the rather safe assumption
that the construction process underlying explicit self-esteem biases explicit
self-esteem more often in a positive rather than in a negative fashion. After
all, people are known to be motivated to see themselves (and have others see
them) in a positive light. This was already suggested earlier by the finding
that explicit self-esteem is correlated with self-presentation style, self-
deception and impression management (Greenwald & Farnham, 2000; Jordan
et al., 2003; Raskin, et al., 1991). It may certainly be the case that people
strategically report low explicit self-esteem (perhaps because they want to
come across as modest), but this is likely to be relatively rare.
If this reasoning is correct we should be able to witness the following
asymmetry: For people with incongruent implicit and explicit self-esteem
(i.e., one is high, the other low), the combination low implicit/high explicit
self-esteem should occur much more often, or among many more people, than
the combination high implicit/low explicit self-esteem. People with high
implicit self-esteem are seldom motivated to report low explicit self-esteem,
whereas low implicit self-esteem individuals may often report relatively high
explicit self-esteem (even if, as we have seen, it is often hard work). Now is
there such an asymmetry?
One problem is that whether one finds support for this asymmetry or
Digging for the Real Attitude
97
not depends on where one draws the line. When do we categorize implicit
self-esteem and explicit self-esteem as truly low or truly high? The evidence
for the hypothesized asymmetry is, up to this point, only suggestive. First,
various people have made the same prediction (Epstein, 1983; O’Brien &
Epstein, 1988; Kernis, 2003). Others have argued that the combination of high
explicit/low implicit self-esteem is indeed uncommon in Western cultures but
not in Asian cultures (see Kitayama & Uchida, 2003, for a brief review).
Kitayama and Uchida (2003) reported that the combination of high
explicit/low implicit self-esteem can be found among Western participants but
only (or at least mostly) under highly specified measuring circumstances.
Concretely, Western participants only showed the high explicit/low implicit
self-esteem combination in the context of close, interdependent relations.
Perhaps also telling is the fact that the combination low implicit/high explicit
has been named – as defensive or fragile self-esteem – and its consequences
have been investigated by an increasing number of research groups (e.g.,
Bosson et al., 2003; Jordan et al., 2003; Kernis, 2003; McGregor & Marigold,
2003), whereas the combination of high implicit/low explicit self-esteem has
received relatively little attention (for exceptions, see Jordan et al., 2008;
Kitayama and Uchida, 2003). Still, we concede that research is needed to
more strongly corroborate this hypothesized asymmetry.
Variations in construction and the relation between implicit and
explicit self-esteem. The second hypothesis derived from the hierarchy
hypothesis is much easier to test. If explicit self-esteem is partly a
construction process guiding people away from core self-esteem and therefore
also from implicit self-esteem, it means that the less construction there is, the
more explicit self-esteem should correlate with implicit self-esteem. After all,
the less construction there is, the better explicit self-esteem should represent
core self-esteem.
Koole, Dijksterhuis and van Knippenberg (2001) have reported
supportive evidence for this hypothesis. In one experiment, they first assessed
Chapter 4
98
people’s implicit self-esteem. Later they measured explicit self-esteem and
they measured the time it took participants to complete the explicit self-
esteem items. They hypothesized that the longer people would take to
complete the measure of explicit self-esteem, the more they engaged in active
construction. By measuring response times they assessed natural variations in
people’s degree of construction. They then divided the participants in two
groups: Fast responders and slow responders. In support of the hierarchy
hypothesis, for fast responders the correlation between explicit and implicit
self-esteem was high (.51), whereas for slow responders there was no
correlation at all (-.06). In sum, the less people engaged in active construction
during assessment of explicit self-esteem, the more explicit self-esteem
correlated with implicit self-esteem.
In another experiment, Koole et al., (2001) manipulated rather than
measured construction. Again, they first measured participants’ implicit self-
esteem. Subsequently, explicit self-esteem was assessed and this was either
done under cognitive load or not. Obviously, cognitive load prevents people
from engaging in too much active construction and the experimenters
predicted that explicit self-esteem would correlate with implicit self-esteem
under load, but not necessarily under normal conditions. Indeed, this is what
they found. The correlation between explicit and implicit self-esteem was
high under load (.48) and absent under normal conditions (-.15). This fully
supports the hierarchy hypothesis.
Further support for the hierarchy hypothesis: An experiment
The experiment we report here extends the experiments reported by Koole et
al., (2001). Again, we tried to manipulate the extent to which participants
would engage in active construction processes biasing explicit self-esteem
away from core self-esteem. Before participants’ completed measures of
implicit and explicit self-esteem, we subliminally primed half of our
participants with the goal to be honest. This was done under the guise of a
lexical decision task whereby experimental participants were subliminally
Digging for the Real Attitude
99
presented with words such as honest, sincere, and true. Control participants
were not presented with words related to honesty. We then measured implicit
self-esteem by name-letter preferences and explicit self-esteem with
Heatherton and Polivy’s (1991) State Self-Esteem Scale3. We tested three
hypotheses.
Hypothesis 1. The first hypothesis is the most important and the most
straightforward. Assuming that explicit self-esteem is in part the result of a
construction process biasing the view of the self in a positive way, the goal to
be honest should lead people to engage in this biased construction to a lesser
extent. The goal to be honest should lead reported explicit self-esteem to be
more strongly related to core self-esteem and therefore to implicit self-esteem.
This means that the correlation between the measures of implicit self-esteem
and explicit self-esteem should be higher for people with a primed honesty
goal than for control participants.
Hypothesis 2. The advantage of the SSES scale is that explicit self-
esteem is divided into three subscales, Appearance self-esteem, Performance
self-esteem, and Social self-esteem. Hypothesis 2 pertains to these subscales.
One could argue that the extent to which people can positively construct self-
esteem differs for the different subscales. After all, reality constraints differ
between the different domains they represent. Appearance self-esteem is
probably the hardest to strategically bias in a positive way. We can maintain
that our attractiveness is on par with that of Brad Pitt or Jennifer Lopez, but it
does not make sense. It’s a form of absurd self-deception and we know it. A
mild form of self-deception is likely to be easier in the domain of
Performance self-esteem. The truth is still to some extent objective, but at
least one can easily switch between different domains (“Yes, I lost a game of
Trivial Pursuit against friends, but I had an A+ for Intro Social Psychology!”).
Social self-esteem is, with the exception of extreme cases perhaps, likely the
easiest one to steer towards rosiness. One can think about many different
Chapter 4
100
relationships, and it is often possible to flexibly interpret social behavior (“His
insult was not personal, he must have been in an awful mood”). If this
reasoning is valid, this would mean that, for people without an honesty goal,
appearance self-esteem correlates highest with implicit self-esteem, whereas
social self-esteem correlates lowest with implicit self-esteem. For participants
with an honesty goal, these differences should disappear.
Hypothesis 3. Assuming that explicit self-esteem, although partly
constructed, does to some extent represent core self-esteem, this should also
be true for the three subscales. If participants primed with honesty engage less
in construction, their self-esteem should better reflect core self-esteem (as
reflected in Hypothesis 1). As this should be true for all subscales, it follows
that the different subscales should “converge” towards core self-esteem and
therefore also to each other. Hence, the correlations between the subscales
should be higher among people primed with honesty.
In total, seventy-one undergraduate students participated in the
experiment, 37 in the control condition, 34 in the honesty-prime condition.
The correlations pertaining to the hypotheses are listed in the Table 4.1. As
can be seen, hypothesis 1 was supported, although the difference just failed to
reach significance. As predicted, we found a high correlation between explicit
and implicit self-esteem after honesty priming, and no such correlation for
control participants. Hypothesis 2 also received support in that, for control
participants, the correlation between implicit self-esteem and appearance self-
esteem was highest, whereas the correlation between implicit self-esteem and
social self-esteem was lowest. The honesty prime significantly increased the
correlation between implicit self-esteem and social self-esteem and between
implicit self-esteem and performance self-esteem (although this latter effect
was marginally significant), whereas the honesty prime did not affect the
correlation between implicit self-esteem and appearance self-esteem. Finally,
hypothesis 3 also received support. Correlations between subscales were
generally higher under honesty conditions than under control conditions, with
Digging for the Real Attitude
101
one of them reaching conventional levels of significance. As predicted, the
different subscales converged because the honesty prime decreased the
amount of construction work.
In sum, although some of the evidence was statistically somewhat
weak, the results support the hierarchy hypothesis. The effects of priming of
the honesty goal were exactly as predicted.
Table 4.1 Correlations between implicit and explicit self-esteem (and its
subscales) and between the subscales as a function of condition.
Honesty Control Difference
Correlations with implicit SE (p – one-tailed)
Overall explicit SE .54* (.51*) .21 (.17) < .06 (< .06)
Appearance .41* (.43*) .33* (.29†) ns (ns)
Performance .41* (.45*) .14 (.16) < .09 (< .12)
Social .48* (.42*) .05 (.04) < .03 (< .05)
Correlations among subscales
Appearance-Performance .63* .51* ns
Appearance-Social .70* .42* < .02
Performance-Social SLE .77* .68* ns
Note. Correlations between parentheses are based on the new scoring algorithm for the name
letter test proposed by Albers, Dijksterhuis, & Rotteveel (2009)
* p < .05. † p < .10.
Conclusions
To conclude, the hierarchy hypothesis best describes the relation between
implicit and explicit self-esteem. Both explicit and implicit self-esteem are in
part based on the same underlying construct, that what we called core self-
esteem. However, due to the fact that explicit self-esteem is often the
consequence of active and biased construction processes, it represents core
self-esteem less well than implicit self-esteem. Furthermore, as implicit self-
Chapter 4
102
esteem represents core self-esteem better, explicit self-esteem can be partly
predicted by implicit self-esteem. In addition, it was shown that (in line with
Koole et al., 2001) the correlation between explicit and implicit self-esteem
can be increased by interfering with the active construction process
explicit self-esteem is partly based on.
Before ending we would like to remark that we do not see implicit self-
esteem as an infinitely better construct than explicit self-esteem. In addition,
we certainly do not argue that we should stop using the latter. Such a claim
would clearly be unwarranted. The relation between explicit and implicit self-
esteem is often so weak that it clearly pays off to investigate both and to
scrutinize combinations of consistent and inconsistent combinations, as
interesting recent research clearly shows. Another reason for not solely
relying on implicit self-esteem is the fact that measures of implicit self-esteem
are in a sense still in a developing stage. For some measures, the underlying
processes driving its effects are not fully understood. Although some
important improvements have been proposed recently (Albers, Dijksterhuis &
Rotteveel, 2009; Karpinski, 2004; Wentura, Kulfanek & Greve, 2005), there
is still quite some work to be done to optimize implicit measurement.
However, we do want to maintain that explicit self-esteem is a less
pure form of self-esteem. In addition to being less pure though, it is also more
rich and multifaceted. It is in part construction rather than excavation work. It
only weakly reflects core self-esteem and it is affected by self-deception,
impression management, and self-presentation style. However, this inherent
richness is not in itself problematic, after all, explicit self-esteem predicts
quite a number of psychological processes very well.
To recapitulate, both explicit and implicit self-esteem clearly have their
value, also in an Egyptian desert. Measuring explicit self-esteem may be
architecture, but it is pretty good architecture with means we are familiar
with. Measuring implicit self-esteem, on the other hand, is sincere archeology,
but with equipment that still leaves things to be desired.
Digging for the Real Attitude
103
Endnotes
¹One could raise the reverse hierarchy hypothesis, namely that explicit
self-esteem is closer to the core than implicit self-esteem. However, such a
hypothesis is at odds with so much psychological knowledge that it cannot be
seriously defended. Conscious processes are by necessity preceded by
unconscious processes (at least when one maintains that consciousness resides
in the brain). Hence, one cannot be conscious of an attitude (“I really like
Chocolate Chip cookies”) without preceding unconscious attitudinal
processes (such as positive affective reactions upon the perception of
Chocolate Chip cookies). One way out would be to say that attitudes are only
attitudes when they are conscious and that the core is to be found in
consciousness. Such a conceptualization is possible, but it would have some
undesirable consequences, the least problematic being that the current chapter
would be superfluous (as implicit attitudes would not exist). However, it
would also render the attitude concept rather limp as we are not that often
consciously aware of our attitudes, except perhaps during communication. Of
course, we are very often aware of attitude objects of course (“Ah, cookies”)
but not of the attitude. In addition, the reverse hierarchy hypothesis would
severely constrain the number of cases where attitudes can predict behavior,
because even if we are consciously aware of an attitude, this very often
happens only after we act (such as when one mindlessly reaches for Chocolate
Chip cookies, and only then thinks “I’m fond of them!”).
²It should be noted that we take the liberty here to interpret the Riketta
and Dauenheimer findings in terms of evaluative conditioning. The authors
themselves favor a different explanation for their findings.
³The order in which implicit self-esteem and explicit self-esteem were
administered was counterbalanced. Order did not affect the results.
105
References
Albers, L., Dijksterhuis, A., & Rotteveel, M. (2004). [The relation between
implicit self-esteem and affective experience after self-threat and self-
enhancement]. Unpublished raw data.
Albers, L., Dijksterhuis, A., & Rotteveel, M. (2009). Implicit self-esteem goes
a long way: Common grounds of Personality, Self-Esteem and
Affective Regulation. Manuscript submitted for publication.
Albers, L., Dijksterhuis, A., & Rotteveel, M. (2009). Towards Optimizing the
Name Letter Test as a Measure of Implicit Self-Esteem. Self and
Identity, 8, 63-77.
Alliger, G. M., & Dwight, S. A. (2000). A meta-analytic investigation of the
susceptibility of integrity tests to response distortion. Educational and
Psychological Measurement, 60, 59-72.
Baccus, J. R., Baldwin, M. W., & Packer, D. J. (2004). Increasing implicit
self-esteem through classical conditioning. Psychological Science, 15,
498-502.
Baldwin, M. W., Carell, S. E. & Lopez, D. F. (1991). Priming relationship
schemas: My advisor and the pope are watching me from the back of
References
106
my mind. Journal of Experimental Psychology, 26, 435-454.
Barelds, D. P. H., & Luteijn, F. (2002). Measuring personality: A comparison
of three personality questionnaires in The Netherlands. Personality and
Individual Differences, 33, 499–510.
Bargh, J. A., Chaiken, S., Govender, R., & Pratto, F. (1992). The generality of
the automatic evaluation effect. Journal of Personality and Social
Psychology, 62, 893-912.
Bargh, J. A., & Chartrand, T. L. (1999). The unbearable automaticity of
being. American Psychologist, 54, 462-479.
Bargh, J. A., Chen, M., & Burrows, L. (1996). The automaticity of
social behavior: Direct effects of trait concept and stereotype
activation on action. Journal of Personality and Social
Psychology, 71, 230-244.
Bargh, J. A., & Pietromonaco, P. (1982). Automatic information processing
and social perception: The influence of trait information presented
outside of conscious awareness on impression formation. Journal of
Personality and Social Psychology, 43, 437-449.
Baumeister, R. F., Campbell, J. D., Krueger, J. I. & Vohs, K. D. (2003). Does
high self-esteem cause better performance interpersonal success,
happiness, or healthier lifestyles? Psychological Science in the Public
Interest, 4, 1-44.
Bettman, J. R., Luce, M. F., & Payne, J. W. (1998). Constructive consumer
choice processes. Journal of Consumer Research, 25, 187-217.
Blanton, H., & Jaccard, J. (2006). Tests of multiplicative models in
psychology: A case study using the unified theory of implicit attitudes,
stereotypes, self-esteem, and self-concept. Psychological Review, 113,
155–169.
Bornstein, R. F., & D’Agostino, P. R. (1992). Stimulus recognition and the
mere exposure effect. Journal of Personality and Social Psychology,
63, 545-552.
Bosson, J. K., Brown, R. P., Zeigler-Hill, V., & Swann, W. B., Jr. (2003).
References
107
Self-enhancement tendencies among people with high explicit self-
esteem: The moderating role of implicit self-esteem. Self and Identity,
2, 169-187.
Bosson, J. K., Swann, W. B., & Pennebaker, J. W. (2000). Stalking the
perfect measure of implicit self-esteem: The blind men and the
elephant revisited. Journal of Personality and Social Psychology, 79,
631-643.
Brown, J. D., & Dutton, K. A. (1995). The thrill of victory, the complexity of
defeat: Self-esteem and people's emotional reactions to success and
failure. Journal of Personality and Social Psychology, 68, 712-722.
Brown, J. D., & Marshall, M. A. (2001). Self-esteem and emotion: Some
thoughts about feelings. Personality and Social Psychology Bulletin,
27, 575-584.
Brown, J. D., & McGill, K.L. (1989). The cost of good fortune: When
positive life events produce negative health consequences. Journal of
Personality and Social Psychology, 57, 1103-1110.
Cacioppo, J. T., Crites, S. L. Jr., Berntson, G. G., & Coles, M. G. H. (1993). If
attitudes affect how stimuli are processed, should they not affect the
event-related brain potential? Psychological Science, 4, 108-112.
Chen, M., & Bargh, J.A.(1997). Nonconscious behavioral
confirmation processes: The self-fulfilling consequences of
automatic stereotype activation. Journal of Experimental Social
Psychology, 33, 541-560.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple
regression/correlation analysis for the behavioral sciences (3rd ed.).
Mahwah, NJ: Lawrence Erlbaum Associates.
Conner, T, & Barrett, L. F. (2005). Implicit self-attitudes predict spontaneous
affect in daily life. Emotion, 5, 476-488.
Crocker, J., & Major, B. (1989). Social stigma and self-esteem: The self-
protective properties of stigma. Psychological Review, 96, 608-630.
References
108
De Houwer, J., Thomas, S., & Baeyens, F. (2001). Associative learning of
likes and dislikes: A review of 25 years of research on human
evaluative conditioning. Psychological Bulletin, 127, 853-869.
DeHart, T., Pelham, B. W., & Murray, S. L. (2004). Implicit dependency
regulation: Self-esteem, relationship closeness, and implicit evaluation
of close others. Social Cognition, 22, 126-146.
DeHart, T., Pelham, B. W., & Tennen, H. (2006). What lies beneath:
Parenting style and implicit self-esteem. Journal of Experimental
Social Psychology, 42, 1-17.
DeNeve, K. M., & Cooper, H. (1998). The happy personality: A meta-
analysis of 137 personality traits and subjective well-being.
Psychological Bulletin, 124, 197-229.
Di Paula, A., & Campbell, J.D. (2002). Self-esteem and persistence in the
face of failure. Journal of Personality and Social Psychology, 83, 711-
723.
Dijksterhuis, A. (2004). I like myself but I don’t know why: Enhancing
implicit self-esteem by subliminal evaluative conditioning. Journal of
Personality and Social Psychology, 86, 345-355.
Dijksterhuis, A., Albers, L., & Bongers, K. (2008). Digging for the real
attitude: Lessons from research on implicit and explicit self-esteem. In
R. E. Petty, R. H. Fazio, & P. Briñol (Eds.), Attitudes: Insights from
the new wave of implicit measures (pp. 229-250). New York:
Psychology Press.
Dijksterhuis, A., & Bargh, J. A. (2001). The perception-behavior expressway:
The automatic effects of social perception on social behavior. In M.P.
Zanna (Ed.), Advances in Experimental Social Psychology, (Vol.33,
pp. 1-40). San Diego: Academic Press.
Dijksterhuis, A., & Nordgren, L. F. (2006). A theory of unconscious thought.
Perspectives on Psychological Science, 1, 95-109.
Dodgson, P. G., & Wood, J. V. (1998). Self-esteem and the cognitive
accessibility of strength and weaknesses after failure. Journal of
References
109
Personality and Social Psychology, 75, 178-197.
Dovidio, J. F., Kawakami, K., Johnson, C., Johnson, B., & Howard, A.
(1997). On the nature of prejudice: Automatic and controlled
processes. Journal of Experimental Social Psychology, 33, 510-540.
Dutton, K. A., & Brown, J. D. (1997). Global self-esteem and specific self-
views as determinants of people's reactions to success. Journal of
Personality and Social Psychology, 73, 139-148.
Epley, N., Savitsky, K., & Kachelski, R.A.(1999). What every skeptic should
know about subliminal persuasion. Skeptical Inquirer, 23, 40-45+.
Epstein, S. (1983). The unconscious, the preconscious, and the self-concept.
In J. Suls & A. Greenwald (Eds.), Psychological perspectives on the
self (Vol. 2, pp. 219-247). Hillsdale, NJ. Lawrence Erlbaum.
Ettema, J. H. M., & Zondag, H. J. (2002). De Nederlandse Narcisme Schaal
(NNS): Psychodiagnostisch gereedschap [The Dutch Scale of
Narcissism]. Psycholoog, 37, 250-255.
Fancher, R. E. (1996). Pioneers of psychology (3nd ed.). New York: Norton.
Fazio, R.H. (1990). Multiple processes by which attitudes guide behavior:
The MODE model as an integrative framework. In M.P. Zanna (Ed.),
Advances in Experimental Social Psychology (Vol. 23, pp. 75-109).
New York: Academic Press.
Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition
research: Their meaning and uses. Annual Review of Psychology, 54,
297-327.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On
the automatic
activation of attitudes. Journal of Personality and Social Psychology, 50, 229-
238.
Fiedler, K., & Bluemke, M. (2005). Faking the IAT: Aided and unaided
response control on the implicit association tests. Basic and Applied
Social Psychology, 27, 307-316.
References
110
Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition:
Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4-
27.
Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association
Test to measure self-esteem and self-concept. Journal of Personality
and Social Psychology, 79, 1022-1038.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. K. L. (1998). Measuring
individual differences in implicit cognition: The Implicit Association
Test. Journal of Personality and Social Psychology, 74, 1464-1480.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and
using the Implicit Association Test: I. An improved scoring algorithm.
Journal of Personality and Social Psychology, 85, 197-216.
Heatherton, T. F., & Polivy, J. (1991). Development and validation of a scale
for measuring state self-esteem. Journal of Personality and Social
Psychology, 60, 895-910.
Heimpel, S. A., Wood, J. V., Marshall, M. A., & Brown, J. D. (2002). Do
people with low self-esteem really want to feel better? Self-esteem
differences in motivation to repair negative moods. Journal of
Personality and Social Psychology, 82, 128-147.
Hetts, J. J., Pelham, B. W. (2001). A case for the nonconscious self-concept.
In G.B. Moskowitz (Ed.), Cognitive Social psychology: The Princeton
symposium on the legacy and future of social cognition (pp. 105-124).
Mahwah, NJ: Erlbaum.
Hetts, J. J., & Pelham, B. W. (2003). The ghosts of Christmas past: Reflected
appraisals and the perils of near Christmas birthdays. Manuscript in
preparation.
Hetts, J. J., Sakuma, M., & Pelham, B. W. (1999). Two roads to positive
regard: Implicit and
explicit self-evaluations and culture. Journal of Experimental Social
Psychology, 35, 512-559.
Hofmann, W., Gawronski, B., Gschwender, T., Le, H., & Schmitt, M. (2005).
References
111
A meta-analysis on the correlation between the Implicit Association
Test and explicit self-report measures. Personality and Social
Psychology Bulletin, 31, 1369-1385.
Holtgraves, T. M. (2004). Social desirability and self-reports: Testing models
of socially desirable responding. Personality and Social Psychology
Bulletin. 30, 161-172
Jones, J. T., Pelham, B. W., Mirenberg, M. C., & Hetts, J. J. (2002). Name-
letter preferences are not merely mere exposure: Implicit egotism as
self-regulation. Journal of Experimental Social Psychology, 38, 170-
177.
Jordan, C. H., Spencer, S. J., Zanna, M. P., Hoshino-Browne, E., & Correll, J.
(2003). Secure and defensive high self-esteem. Journal of Personality
and Social Psychology, 85, 969-978.
Jordan, C. H., Whitfield, M., & Zeigler-Hill, V. (2007). Intuition and the
correspondence between implicit and explicit self-esteem. Journal of
Personality and Social Psychology, 93, 1067-1079.
Jordan, C. H., Logel, C., Spencer, S. J., Zanna, M. P., & Whitfield, M. L.
(2008). The Heterogeneity of self-esteem: Exploring the interplay
between implicit and explicit self-esteem. In R. E. Petty, R. H. Fazio,
& P. Briñol (Eds.), Attitudes: Insights from the new wave of implicit
measures (pp. 251-284). New York: Psychology Press.
Karpinski, A. (2004). Measuring self-esteem using the implicit self-
association test: The role of the other. Personality and Social
Psychology Bulletin, 30, 22-34.
Karpinski, A., & Steinman, R. B. (2006). The Single Category Implicit
Association Test as a Measure of Implicit Social Cognition. Journal of
Personality and Social Psychology, 91, 16-32.
Kernis, M. H. (2003). Towards a conceptualization of optimal self-esteem.
Psychological Inquiry, 14, 1-26.
Kitayama, S., & Karasawa, M. (1997). Implicit self-esteem in Japan: Name
References
112
letters and birthday numbers. Personality and Social Psychology
Bulletin, 23, 736-742.
Kitayama, S., & Uchida, Y. (2003). Explicit self-criticism and implicit self-
regard: Evaluating self and friend in two cultures. Journal of
Experimental Social Psychology, 39, 476-482.
Koole, S. L., & Dehart, T. (2007). Self-Affection without Self-Reflection:
Origins, Models, and Consequences of Implicit Self-Esteem. In C.
Sedikides & S. Spencer (Eds.), The Self in Social Psychology (pp. 36 -
86). New York: Psychology Press.
Koole, S. L., Dijksterhuis, A., & van Knippenberg, A. (2001). What's in a
name: Implicit self-esteem and the automatic self. Journal of
Personality and Social Psychology, 80, 669-685.
Koole, S.L., & Pelham, B.W. (2003). On the nature of implicit self-esteem:
The case of the Name-letter effect. In. S. Spencer & M.P. Zanna (Eds.),
Ontario Symposium on Personality and Social Psychology (Vol. 7).
Koole, S. L., Smeets, K., van Knippenberg, A., & Dijksterhuis, A. (1999).
The cessation of rumination through self-affirmation. Journal of
Personality and Social Psychology, 77, 111-125.
Krizan, Z., & Suls, J. (2008). Are implicit and explicit measures of self-
esteem related? A meta-analysis for the Name-Letter Test. Personality
and Individual Differences, 44, 521-531.
Kunst-Wilson, W. R., & Zajonc, R. B. (1980). Affective discrimination of
stimuli that cannot be recognized. Science, 207, 557–558.
Lakoff, R. T. (1990). Talking power: The politics of language. New York:
Basic Books.
Lane, A. M., Jones, L., & Stevens, M. (2002). Coping with failure: The
effects of self-esteem and coping on changes in self-efficacy. Journal
of Sport Behavior, 25, 331-345.
Lewicki, P., Hill, T., & Czyzewska, M. (1992). Nonconscious acquisition of
information. American Psychologist, 47, 796-801.
Luteijn, F., Starren, J., & Van Dijk, H. (2000). NPV; Nederlandse
References
113
Persoonlijkheidsvragenlijst, handleiding (herziene uitgave). [DPQ;
Dutch Personality Questionnaire, manual (revised edition)]. Lisse, The
Netherlands: Swets & Zeitlinger.
Mayer, J. D., Gaschke, Y. N., Braverman, D. L., Evans, T. W. (1992). Mood-
congruent judgment is a general effect. Journal of Personality and
Social Psychology, 63, 119-132.
McFarlin, D. B., Baumeister, R. F., & Blascovich, J. (1984). On knowing
when to quit: Task failure, self-esteem, advice, and nonproductive
persistence. Journal of Personality, 52, 138-155.
McGregor, I., & Marigold, D. C. (2003). Defensive zeal and the uncertain
self: What makes you so sure? Journal of Personality and Social
Psychology, 85, 838-852.
Meagher, B. E., & Aidman, E. (2004) Individual Differences in Implicit and
Declared Self-Esteem as Predictors of Response to Negative
Performance Evaluation: Validating Implicit Association Test as a
Measure of Self-Attitudes . International Journal of Testing , 4, 19-42.
Nisbett, R. E., & Wilson, T. D. (1997). Telling more than we can know:
verbal reports on mental processes. Psychological Review, 84, 231-
259.
Nuttin, J. M. (Jr.) (1985). Narcissism beyond Gestalt and awareness: The
name letter effect. European Journal of Social Psychology, 15, 353-
361.
O’ Brien, E. J., & Epstein, S. (1988). The multidimensional self-esteem
inventory. Odessa, FL: Houghton Mifflin.
Olson, M. A., Fazio, R. H., & Hermann, A. D. (2007). Reporting tendencies
underlie discrepancies between implicit and explicit measures of self-
esteem. Psychological Science, 18, 287-291.
Olson, M. A., & Fazio, R. H. (2008). Implicit and explicit measures of
attitudes: The perspective of the MODE model. In R. E. Petty, R. H.
Fazio, & P. Briñol (Eds.), Attitudes: Insights from the new wave of
implicit measures (pp. 19-65). New York: Psychology Press.
References
114
Pacini, R., & Epstein, S. (1999). The relation of rational and experiential
information processing styles to personality, basic beliefs, and the
ration-bias phenomenon. Journal of Personality and Social
Psychology, 76, 972-987.
Pelham, B. W., Carvallo, M., & Jones, J. T. (2005). Implicit egotism. Current
Directions in Psychological Science, 14, 106-110.
Pelham, B.W., DeHart, T., & Carvallo (2003). Implicit effects of stigmatizing
names. State University of New York at Buffalo, unpublished
manuscript.
Pelham, B. W., & Hetts, J. J. (1999). Implicit and explicit personal and social
identity: Toward a more complete understanding of the social self. In
T.R. Tyler, R.M. Kramer, & O.P. John (Eds.), The psychology of the
social self (pp. 115–143). Hills- dale, NJ: Erlbaum.
Pelham, B. W., Koole, S. L., Hardin, C. D., Hetts, J. J., Seah, E., & De Hart,
T. (2005). Gender moderates the relation between implicit and explicit
self-esteem. Journal of Experimental Social Psychology, 41, 84-89.
Pelham, B. W., Mirenberg, M.C. & Jones, J. T. (2002). Why Susie sells
seashells by the seashore: Implicit egotism and major life decisions.
Journal of Personality and Social Psychology, 82, 469-487.
Pinter, B., & Greenwald, A. G. (2005). Clarifying the role of the "other"
category in the self-esteem IAT. Experimental Psychology, 52, 74-79.
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for
estimating indirect effects in simple mediation models. Behavior
Research Methods, Instruments and Computers, 36, 717-731.
Pyszczynski, T., Greenberg, J., Solomon, S., Arndt, J., & Schimel, J. (2004).
Why do people need self-esteem? A theoretical and empirical review.
Psychological Bulletin, 130, 435-468.
Raskin, R., Novacek, J., & Hogan, R. (1991). Narcissism, self-esteem, and
defensive self-enhancement. Journal of Personality, 59, 19-38.
Riketta, M. (2005). Gender and socially desirable responding as moderators
of the correlation between implicit and explicit self-esteem. Current
References
115
Research in Social Psychology, 11, 14-28.
Riketta, M., & Dauenheimer, D. (2003). Manipulating self-esteem with
subliminally presented words. European Journal of Social Psychology,
33, 679-699.
Robins, R. W., Tracy, J. L., Trzesniewksi, K., Potter, J., & Gosling, S. D.
(2001). Personality correlates of self-esteem. Journal of Research in
Personality, 35, 463-482.
Robinson, M. D., & Meier, B. P. (2005). Rotten to the core: Neuroticism and
implicit evaluations of the self. Self and Identity, 4, 361-372.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ:
Princeton University Press.
Rothermund, K., & Wentura, D. (2004). Underlying processes in the Implicit
Association Test (IAT): Dissociating salience from associations.
Journal of Experimental Psychology: General, 133, 139-165.
Shimizu, M., & Pelham, B. W. (2004). The unconscious cost of good fortune:
implicit and explicit self-esteem, positive life events, and health.
Health Psychology, 23, 101-105.
Shrauger, J. S., & Rosenberg, S. E. (1970). Self-esteem and the effects of
success and failure on performance. Journal of Personality, 38, 404-
417.
Spalding, L. R., & Hardin, C. D. (1999). Unconscious unease and self-
handicapping: Behavioral consequences of individual differences in
implicit and explicit self-esteem. Psychological Science, 10, 535-539.
Stake, J. E., Huff, L., & Zand, D. (1995). Trait self-esteem, positive and
negative events, and event-specific shifts in self-evaluation and affect.
Journal of Research in Personality, 29, 223-241.
Steele, C. M. (1988). The psychology of self-affirmation: Sustaining the
integrity of the self. In L. Berkowitz (Ed.), Advances in Experimental
Social Psychology (Vol. 21, pp. 261-302). New York: Academic Press.
Steele, C. M., Spencer, S. J., & Lynch, M.(1993) Self-image resilience and
dissonance: The role of affirmational resources. Journal of
References
116
Personality and Social Psychology, 64, 885-896.
Taylor, S. E. & Brown, J. D. (1988). Illusion and well-being: A social
psychological perspective on mental health. Psychological Bulletin,
103, 193—210.
Wentura, D., Kulfanek, M., & Greve, W. (2005). Masked affective priming
by name letters: Evidence for a correspondence of explicit and implicit
self-esteem. Journal of Experimental Social Psychology, 41, 654-663.
Wood, J. V., Heimpel, S. A., Newby-Clark, I. R., Ross, M.(2005). Snatching
Defeat from the Jaws of Victory: Self-Esteem Differences in the
Experience and Anticipation of Success. Journal of Personality and
Social Psychology, 89, 764-780.
Zeigler-Hill, V. (2006). Discrepancies between implicit and explicit self-
esteem: Implications for narcissism and self-esteem instability. Journal
of Personality, 74, 119-143.
117
Summary
It is hard to find a person that does not have a deep routed feeling of being a
self. A self that is distinguishable from its surroundings and other selves, who
in their own way, distinguish their selves similarly. Moreover, all these selves
seem to attach existential meaning to the affective condition their selves are
in, so that feeling good about oneself is often the ultimate goal. That this
contention is not without support becomes clear when consulting the Web.
The term “Self-help” results in almost one billion hits on the internet,
approaching the far broader concept of health and outnumbering concepts
such as god, religion, war, peace, success, medicine, money, and even food.
The preoccupation with valuing ourselves is also apparent in
psychological research with self-esteem as one of its most studied concepts.
In the 20th century this concept of self-esteem resembled most other
psychological concepts in that the explicit, or direct, approach of defining,
and measuring it was the dominant approach. That is, psychological
constructs were measured by asking individuals to indicate their attitudes,
feelings, cognitions or behavior. For example, when measuring self-esteem
someone had to indicate personal agreement with items such as “I have a
positive attitude towards myself”. A collection of agreements and
Double You?
118
disagreements with similar statements is then taken as an individual score on
self-esteem, in this example the score on the Rosenberg Self-Esteem Scale
(Rosenberg, 1965). The prevalence of these self-report methods of
psychological constructs suggests their usefulness and established value
within the field of psychology. However, they rely on two basic assumptions
about who we are and how we function: Individuals will and can
introspectively indicate what they think, feel and do. A challenge of both
assumptions has led to a surge of attention in the use and development of
implicit, or indirect, measuring instruments in the late 20th century (Fazio &
Olson, 2003).
Falsification of each of these two basic assumptions poses different
problems for psychological research. The possibility that someone may not be
willing to honestly answer self-report measures is generally acknowledged to
pose a validity problem for the measurement of psychological constructs
although the importance of the problem is subject of ongoing debate
(Holtgraves 2004; Alliger & Dwight, 2000). Indirect or implicit measures
circumvent this problem since the subject simply does not know what is being
assessed so that the impact of social desirability greatly diminishes. The idea
that individuals actually may not be able to have full introspective access to
themselves can already be found in the works of 18th and 19th century
influential philosophers such as Leibniz, Schopenhauer and Nietzsche, but has
only been substantiated by empirical evidence conducted in the last few
decades (for reviews see for example Dijksterhuis, & Bargh, 2001; Epley,
Savitsky & Kachelski, 1999; Greenwald & Banaji, 1995; Lewicki, Hill &
Czyzewska, 1992; Nisbett & Wilson, 1977). Because traditional self-report
measures cannot capture someone’s introspectively inaccessible parts, our
understanding of psychological functioning benefits from indirect or implicit
measures that do try the grasp these attitudes and feelings. Hence, besides
circumventing validity problems in self-report measures, the development of
indirect or implicit measures may first and foremost be useful in exploring the
role that both nonconscious content and processes play in our psychological
Summary
119
functioning.
In this dissertation I examine function and form of implicit self-esteem,
which reflects the relatively less accessible part of self-esteem. Specifically,
this dissertation is about the additional value implicit self-esteem has in
understanding self-esteem. I will argue that in order to place the concept of
implicit self-esteem within the realms on current research on self-esteem,
understanding of both the measurement of implicit self-esteem and its relation
with explicit self-esteem is necessary.
In Chapter 1 an overview of the existing literature on implicit and
explicit self-esteem is integrated with the findings from the research presented
in the chapters 2, 3, and 4 of this dissertation. First, a discussion on the
emergence of the concept of implicit self-esteem is presented, followed by an
analysis and overview of several instruments that measure implicit self-
esteem. The relation between implicit and explicit self-esteem, which, like
any other implicit-explicit relation, is subject to ongoing debate, will then be
discussed and connected to other psychological constructs. Specifically,
results will be presented showing that implicit self-esteem relates over and
above explicit self-esteem to important affect-related personality traits and
furthermore serves as a useful predictor of affect-regulation. I will attest that
these empirical results form the core in the theoretical development of
understanding function and form of implicit self-esteem, and conclude that
although the current measures of implicit self-esteem are still in their infancy,
they already serve as important empirical tools in diminishing the distance
between our real self and the self we believe or say we are.
In Chapter 2, I will focus on optimizing one of the most widely used
implicit self-esteem measures, the name letter test (Kitayama & Karasawa,
1997). The name letter test consists of judging the different letters of the
alphabet on their attractiveness. A name letter effect is then traditionally
calculated as the rated attractiveness of someone’s initials or name letters
minus the average attractiveness of those same letters rated by people not
having those initial or name letters. This name letter effect is subsequently
Double You?
120
used as a measure of implicit self-esteem based on the assumption that by
assessing someone’s evaluation of well established attributes of the self (such
as someone’s initials) in an unobtrusive way (e.g., under the guise of simply
evaluating all alphabet letters), someone’s implicit evaluation of the self is
measured. It is argued that the traditional calculation method results in a name
letter score that is confounded, because this method cannot adequately
distinguish the evaluation of name letters from not-name letters, and may
therefore mistake response bias or general letter liking for implicit self-
esteem. This is why an alternative scoring algorithm for the name letter score
is proposed that is without confounds and assesses the unique contribution of
name letter evaluation by simultaneously controlling for the influence of not-
name letter liking and general liking of the specific letters. In three
experiments it is shown that the confound of overall letter liking is indeed
problematic in empirical findings on implicit self-esteem when using the
traditional name letter score while the alternative scoring algorithm remedies
this issue and as such results in a more valid measure for implicit self-esteem.
In Chapter 3, I will discuss function and form of implicit self-esteem in
relation to affect-regulation processes. Using the new scoring algorithm for
the name letter test presented in Chapter 2 it is first empirically demonstrated
that implicit self-esteem relates over and above explicit self-esteem and across
time to the personality dimensions Emotional Stability and Neuroticism, the
two personality dimensions that are most strongly related to affect regulation.
In a second and third experiment I will further show that implicit self-esteem
can differentiate between the kind of affect regulation required: It is of special
importance in regulating affect when the self is threatened. On the other hand,
the explicit measures of self-esteem did not relate to this kind of self-
regulation but rather seemed to predict a more general state of affect. These
results suggest that measuring implicit self-esteem has additional value in our
understanding of both function and form of self-esteem
In Chapter 4, I will, in search for “the real self”, further dig into the
relation between implicit and explicit self-esteem. Based on both theoretical
Summary
121
and empirical evidence it is proposed, that though measures of implicit self-
esteem may not be without problems, in essence implicit self-esteem is closer
to our core self-esteem than explicit self-esteem. First of all evidence is
provided that the relation between implicit and explicit self-esteem is
enhanced by factors that reduce (the need for) an active construction process
of responding to the explicit self-esteem measures. Secondly, empirical
evidence is presented demonstrating that high scores on traditional measures
of explicit self-esteem could be either the result of actually having high self-
esteem or the result of other, less desirable self-presentational strategies.
Conversely, measures of implicit self-esteem are unrelated to these
psychological processes. Because measures of explicit self-esteem do not
have the inherent capacity to distinguish between these possibilities, this
provides further support for the proposition that explicit self-esteem is a less
pure measure that is more dissociated from core self-esteem than implicit self-
esteem. In an additional experiment this proposition is further substantiated
by showing that explicit self-esteem more resembles implicit self-esteem
when an automatic goal to be honest is activated.
123
Summary in Dutch
(Nederlandse Samenvatting)
Het is lastig iemand te vinden die geen diepgeworteld gevoel heeft een zelf te
zijn. Een zelf dat zich duidelijk onderscheidt van zijn omgeving en andere
personen die, op hun beurt, zichzelf op dezelfde wijze onderscheiden.
Bovendien lijken al deze ‘zelven’ een existentiële betekenis te verbinden aan
de toestand waarin zij zich bevinden, waardoor je goed voelen over jezelf
vaak een levensdoel op zich is. Dat deze observatie niet zomaar uit de lucht
gegrepen is wordt duidelijk wanneer we kijken naar het overweldigende
aanbod op het gebied van zelfhulp. De Engelse vertaling van zelfhulp (i.e.,
Self-help) geeft bijna 1 miljard hits op het internet, waarmee het veel bredere
Engelstalige begrip ‘health’ (i.e., gezondheid) in aantal benaderd wordt en de
Engelse vertalingen van begrippen zoals god, geloof, oorlog, vrede, succes,
geneeskunde, geld en voedsel zelfs overstegen worden.
De grote behoefte om ons goed over onszelf te voelen komt ook sterk
naar voren in wetenschappelijk psychologisch onderzoek, waarbij
zelfwaardering één van de meest onderzochte onderwerpen is. Net zoals bij
andere psychologische constructen werd ook zelfwaardering in de twintigste
Double You?
124
eeuw vooral als een expliciet en bewust construct benaderd. Bij het meten van
zelfwaardering betekende dit dat iemand de mate van instemming diende aan
te geven op beweringen zoals “Ik heb een positieve houding naar mijzelf”. De
mate waarin die persoon vervolgens instemde met verschillende gerelateerde
beweringen werd vervolgens als maat genomen voor iemands zelfwaardering.
Hoewel deze expliciete manier van meten ook nog tegenwoordig de gangbare
methode is om psychologische constructen te meten, is het wel zo dat de
laatste decennia steeds meer aandacht is gekomen voor het ontwikkelen van
meer impliciete of indirecte meetmethodes. Dit komt vooral omdat de
aannames, die ten grondslag liggen aan de expliciete meetmethodes en ervan
uitgaan dat personen zowel in staat als bereid zijn aan te geven wat zij over
zichzelf denken en voelen, steeds vaker in twijfel worden getrokken (Fazio &
Olson, 2003).
De falsificatie van deze twee basisaannames zorgen elk voor hun eigen
problemen binnen psychologisch onderzoek. De mogelijkheid dat iemand niet
bereid is eerlijk zijn wordt in het algemeen gezien als een probleem bij het
meten van het betreffende psychologische construct (Holtgraves 2004; Alliger
& Dwight, 2000). Indirecte of impliciete meetmethodes kunnen dit probleem
omzeilen omdat de ondervraagde persoon zich niet bewust is van datgene dat
gemeten wordt zodat de impact van sociaal wenselijk gedrag sterk wordt
verminderd. De mogelijkheid dat iemand niet in staat is om volledige
introspectieve toegang te hebben tot zichzelf wordt al benoemd in het
gedachtegoed van invloedrijke filosofen zoals Leibniz, Schopenhauer en
Nietzsche, maar wordt pas de laatste paar decennia ondersteund door
empirisch bewijs (voor een overzicht zie bijvoorbeeld Dijksterhuis, & Bargh,
2001; Epley, Savitsky & Kachelski, 1999; Greenwald & Banaji, 1995;
Lewicki, Hill & Czyzewska, 1992; Nisbett & Wilson, 1977). Omdat de
traditionele zelfrapportage methodes niet in staat zijn de ontoegankelijke
aspecten van onze psyche bloot te leggen kan ons begrip over ons
psychologisch functioneren en de rol die onbewuste processen hier in spelen
dus baat hebben bij indirecte of impliciete meetmethodes die erop gericht zijn
Summary in Dutch
125
dat wel te doen.
In dit proefschrift onderzoek ik vooral de impliciete zelfwaardering,
die betrekking heeft op het relatief minder toegankelijke deel van onze
zelfwaardering. De nadruk zal hierbij liggen op de toegevoegde waarde die
deze impliciete benadering heeft voor het beter begrijpen van de functie en
vorm van zelfwaardering.
In Hoofdstuk1 wordt de bestaande literatuur over impliciete en
expliciete zelfwaardering geïntegreerd met de bevindingen van het onderzoek
dat gepresenteerd wordt in de hoofdstukken 2, 3, en 4 van dit proefschrift.
Allereerst wordt de opkomst van het concept impliciete zelfwaardering
besproken, gevolgd door een overzicht en analyse van de verschillende
methoden om impliciete zelfwaardering te meten. De relatie tussen impliciete
en expliciete zelfwaardering zal dan onder de loep genomen worden en
vervolgens worden gerelateerd aan andere psychologische constructen.
Specifiek zal empirisch aangetoond worden dat impliciete zelfwaardering
vooral gerelateerd is aan persoonlijkheidstrekken die een sterke affectieve
component bezitten en verder kan fungeren als een bruikbare voorspeller voor
affectregulatie. Ik zal verder aangeven dat deze empirische bevindingen de
kern vormen in de theoretische ontwikkeling en het begrip van de functie en
vorm van impliciete zelfwaardering en concluderen dat ondanks dat de
huidige meetmethodes voor impliciete zelfwaardering nog in hun
kinderschoenen staan, ze zelfs nu al fungeren als belangrijk empirisch
gereedschap om de afstand te verkleinen tussen ons echte zelf en het zelf dat
we denken of zeggen te zijn.
In Hoofdstuk 2 zal de nadruk liggen op het op het optimaliseren van de
naamletter test (Kitayama & Karasawa, 1997), één van de meest gebruikte
methodes om impliciete zelfwaardering te meten. De test is gebaseerd op de
aantrekkelijkheid die iemand toedicht aan de verschillende letters van het
alfabet. De traditionele rekenmethode om hier vervolgens een naamletter
effect voor te berekenen bestaat uit de aantrekkelijkheid die iemand dan blijkt
te geven aan zijn initialen of naamletters te verminderen met de gemiddelde
Double You?
126
aantrekkelijkheid die andere personen, die de betreffende initialen of
naamletters niet in hun eigen naam hebben, aan die letters geven. Dit
naamletter effect wordt vervolgens gebruikt als maat voor de impliciete
zelfwaardering gebaseerd op de aanname dat door het op onopvallende wijze
vaststellen van iemands waardering voor kenmerken die sterk geassocieerd
zijn met het zelf (zoals onder het mom van het simpelweg waarderen van alle
letters van het alfabet daarmee onopvallend de waardering bepalen die iemand
geeft aan zijn initialen), de impliciete waardering die een persoon over
zichzelf heeft gemeten kan worden. Er zal zowel theoretisch als empirisch
worden aangetoond worden dat de traditionele rekenmethode niet in staat is
een onderscheid te maken in de waardering van iemands naamletters en de
overige letters van het alfabet, zodat het geen goede methode is om iemands
impliciete zelfwaardering vast te stellen. Om deze reden zal er een nieuw
algoritme worden voorgesteld dat wel in staat is dit onderscheid te maken en
resulteert in een betere methode om impliciete zelfwaardering mee vast te
stellen.
In Hoofdstuk 3 zal ik de functie en vorm van impliciete zelfwaardering
in relatie tot affectregulatie bespreken. Met gebruik van het nieuwe algoritme
voor het naamletter effect, besproken in Hoofdstuk 2, zal er eerst empirisch
worden aangetoond dat impliciete zelfwaardering bovenop expliciete
zelfwaardering gerelateerd is aan de persoonlijkheidseigenschappen
Emotionele Stabiliteit en Neuroticisme, de twee persoonlijkheidsdimensies
die het sterkst gerelateerd zijn aan affectregulatie. In een tweede en derde
experiment zal verder aangetoond worden dat impliciete zelfwaardering
vooral een belangrijke rol speelt bij affectregulatie wanneer het zelf bedreigd
wordt, terwijl de expliciete zelfwaardering niet gerelateerd lijkt te zijn aan
deze vorm van zelfregulatie maar eerder indicatief is voor de algemene
gemoedstoestand van een persoon. Deze resultaten duiden op de toegevoegde
waarde die het meten van impliciete zelfwaardering heeft in ons begrip van
zowel functie als vorm van zelfwaardering.
In Hoofdstuk 4 zal ik in de zoektocht naar het ‘werkelijke’ zelf verder
Summary in Dutch
127
ingaan op de relatie tussen impliciete en expliciete zelfwaardering, waarbij
zowel op grond van theoretisch als empirisch bewijs geconcludeerd wordt dat
in essentie de impliciete zelfwaardering dichter bij onze werkelijke
zelfwaardering ligt dan de expliciete zelfwaardering. Allereerst zal er bewijs
gegeven worden dat laat zien dat de relatie tussen impliciete en expliciete
zelfwaardering versterkt wordt door factoren die de (behoefte tot) actieve
constructie verminderen bij het meten van iemands expliciete zelfwaardering.
Daarnaast wordt er empirisch bewijs gepresenteerd dat laat zien dat hoge
scores bij het meten van expliciete zelfwaardering zowel het gevolg kunnen
zijn van een werkelijke hoge zelfwaardering maar ook het gevolg van een
minder wenselijke zelfpresentatiestrategieën die een eigenlijk lage expliciete
zelfwaardering maskeren, terwijl het meten van impliciete zelfwaardering niet
beïnvloed wordt door deze strategieën. Dit geeft verdere ondersteuning aan de
het idee dat expliciete zelfwaardering een minder zuivere maat is die verder
verwijderd is van onze werkelijke zelfwaardering dan impliciete
zelfwaardering. In een extra experiment wordt dit idee verder ondersteund
door te laten zien dat expliciete zelfwaardering meer op impliciete
zelfwaardering gaat lijken wanneer een onbewust doel om eerlijk te zijn
wordt geactiveerd.
129
Acknowledgements
(Dankwoord)
Zelfs een mars van duizend mijl begint met de eerste stap. Nu het einde van
mijn proefschriftmars in zicht is wil ik deze woorden van Lao Tse graag
aangrijpen om iedereen die deel is geweest van mijn reis met heel mijn hart te
bedanken.
Ik ben deze reis ooit begonnen met de eerste stap, eigenlijk zonder doel
of verwachting over de afloop. Promoveren was ondergeschikt aan
onderzoeken. Ondergeschikt aan de uitdagingen die de wetenschap mij te
bieden had. Dat beviel uitstekend, maar veranderde toen na een paar jaar
duidelijk werd dat mijn omzwervingen mij niet wezenlijk dichter bij het
promoveren zelf hadden gebracht. Het doel om te promoveren werd steeds
saillanter en drukte mijn onderzoekspassie soms naar de tweede plaats. Een
op intelligentie geënte bergbeklimming werd daardoor steeds meer een proeve
van doorzettingsvermogen. Hoewel mij dat niet altijd even goed beviel, bleek
ik over meer doorzettingsvermogen te beschikken dan ik aanvankelijk had
gedacht. Een soort van onbewuste vastberadenheid die nieuw voor mij was.
Alleen bleek die wel een beetje te botsen met mijn eigenwijze neiging het op
mijn eigen voorwaarden te doen. Iets te perfectionistisch. Iets te idealistisch.
Iets te moralistisch. En die voorwaarden bleken soms moeilijk de kop in te
drukken. Daarmee heb ik mijzelf vaak door te krappe kloven proberen te
Acknowledgements
130
duwen en mezelf op wel erg kronkelende paadjes en te steile hellingen doen
belanden. En dus kwam ik wel eens vast te zitten, verloor ik perspectief of
gleed ik uit. Maar evenzo vaak hebben jullie mij weer vlot getrokken, het
juiste pad gewezen, mij verzorgd en rechtop gezet. Zelfs wanneer ik tijdens
het uitglijden jullie wel eens heb laten uitglijden. Ik kan niet vaak genoeg
zeggen hoeveel mij dat waard is.
Ap en Mark, bedankt voor jullie vriendschap, ondersteuning, geduld en
wetenschappelijk inzicht. Ik heb er veel van geleerd en veel aan jullie gehad.
Dit geldt ook voor alle andere collega’s van de verschillende vakgroepen
Sociale Psychologie, de vakgroep Arbeids- en Organisatie Psychologie aan de
UvA en mijn tijd bij het Social Skills Lab. Door jullie heb ik veel mooie
onuitwisbare herinneringen. Met weemoed denk ik vaak aan die tijd terug,
niet in het minst door Yael, Catherina, Helma, Maaike, Kirsten, Sjoerd,
Maaike, Karin, Clemens, Daphne, Job, Michael, Sjoerd, Gerben, Sanne, Titia,
Petra, Tessa, Carlijn, Marielle, Pam, Lotte, Lotte, Anja, Karin en Ho Jung.
Wat een mooie tijd met mooie mensen was het.
En dan mijn leven buiten de UvA. Al mijn vrienden, bedankt voor jullie
warmte en levenswijsheden en dan zeker Stefan, Edo, Mike, Chris, Petra,
Maria, Janneke, Jeroen, Jeroen, Lobke, Daphne, Carlijn, Sjoerd, Sumit en
Yann. Marije, Roos en Julie, dank voor de mooie momenten die ik met jullie
heb mogen delen.
Sjoerd dank je voor je ongekunsteldheid, je vriendschap en Etta James.
Waar zouden we zijn zonder muziek.
Sumit, dank voor je steun en de gepassioneerde discussies over het leven
in het algemeen en de wetenschap in het bijzonder. We hebben veel gemeen.
Yann, merci de votre amitié et chanter le Blues du Lac de St.Croix avec
moi.
Chris, dank voor je recht-voor-z’n-raap-zijn, onuitputbare energie en
hernieuwde vriendschap. Ik zal oefenen om je wat beter bij te kunnen houden.
Lobke, dank voor je ondersteuning, eigenzinnigheid en goede
gesprekken. Het doet me goed.
Dankwoord
131
Edo, dank voor je grote humor en behulpzaamheid. Je bent een mooi
mens.
Carlijn, dank voor je mensenkennis, je interesse en oprechtheid. Ik leer
er veel van.
Marjolein, dank voor je lieve warmte en dat je me naar Cambodja
stuurde. Het was zo nodig en zo’n goed idee.
Stefan, dank voor jouw empathisch vermogen en diepgaand inzicht in
mijn menselijk zijn. Het is mij veel waard.
Jeroen dank voor je vriendschap, je levenshouding, goede gesprekken en
backgammon. Daarbij bestaat er geen betere zwager voor mijn zus.
Daphne, dank voor al je vriendschap, humor in mooie tijden en grote
steun op moeilijke dagen. Ik prijs me heel gelukkig dat ik je ken.
Jeroen, kerel. Dank voor al meer dan zeventien jaar magie.
Onbetaalbaar. Onvoorwaardelijk. Ongelooflijk.
Janneke, Janneke, Janneke. Wat een fantastische zus en wat een
ontzettend mooi mens ben jij. Wat zou ik zonder je moeten.
Liefste Pap en Mam. Jullie stonden, staan en zullen altijd aan mijn basis
staan. Zonder enig voorbehoud was jullie liefde er en zal die er altijd zijn.
Jullie zijn wie ik hoop te zijn.
133
The “Kurt Lewin Institute Dissertation Series” started in 1997. Since 2008 the following dissertations have been published: 2008-1: Marijke van Putten: Dealing with missed opportunities. The causes and boundary conditions of inaction inertia 2008-2: Marjolein Maas: Experiential Social Justice Judgment Processes 2008-3: Lonneke de Meijer: Ethnicity effects in police officer selection: Applicant, assessor, and selection-method factors 2008-4: Frederike Zwenk: Voice by Representation 2008-5: Margreet Reitsma: The Impact of Linguistically Biased Messages on Involved Receivers 2008-6: Marcus Maringer: Feeling one thing, seeing another: Emotion comparison effects in person judgments 2008-7: Hanneke Heinsman: The competency concept revealed: Its nature, relevance, and practice 2008-8: Joris Lammers: Toward a more social social psychology of power 2008-9: Daniël Fockenberg: Between Good and Evil: Affective Priming in Dynamic Context 2008-10: Arne van den Bos: Why we stereotype influences how we stereotype: self- enhancement and comprehension effects on social perception 2008-11: Lidewij Niezink: Considering Others in Need: On Altruism, Empathy and Perspective Taking 2008-12: Aad Oosterhof: Better together: Antecedents and consequences of perceived expertise dissimilarity and perceived expertise complementarity in teams 2008-13: Femke ten Velden: Negotiation in dyads and groups: The effects of social and epistemic motives 2008-14: Maike Wehrens: How did YOU do? Social comparison in secondary education 2008-15: Kyra Luijters: Making Diversity Bloom: Coping Effectively with Cultural Differences at Work 2008-16: Ilona de Hooge: Moral emotions in decision making: Towards a better understanding of shame and guilt 2008-17: Lindred L. Greer: Team Composition and Conflict: The Role of Individual Differences 2008-18: Sezgin Cihangir: The Dark Side of Subtle Discrimination: How targets respond to different forms of discrimination 2008-19: Giel Dik: On the contagiousness of others’ goals: The role of perceiving effort 2008-20: Lotte van Dillen: Dealing with negative feelings: The role of working memory in emotion regulation 2008-21: Marijn Poortvliet: Information exchange examined: An interpersonal account of achievement goals 2008-22: Sjoerd Pennekamp: Dynamics of disadvantage: Uncovering the role of group- based anger 2008-23: Chris Reinders Folmer: Cooperation and communication: Plastic goals and social roles 2009-1: Marijke Leliveld: Ethics in Economic Decision-Making 2009-2: Monique Pollmann: Accuracy and Bias in Person Perception 2009-3: Krispijn Faddegon: Regulatory Focus in Group Contexts
134
2009-4: Lieven Brebels: Mirror, mirror on the wall… Procedural fairness as an evaluative and regulatory looking-glass self 2009-5: Daphne Wiersema: Taking it personally: Self-esteem and the protection of self-related attitudes 2009-6: Judith D.M. Grob: Dial E for Emotion: Context and Consequences of Emotion Regulation 2009-7: Katherine Stroebe: Is this about me? Responding to subtle discrimination – beyond an individual versus group perspective 2009-8: Menno Vos: Identity patterns in diverse work groups: Improving social integration outcomes through relational identities 2009-9: Lennart Renkema: Facing Death Together: Understanding The Consequences of Mortality Threats 2009-10: Michael Vliek: Group-based social comparison processes: An intragroup level of analysis 2009-11: Karlijn Massar: Unconscious rivals: The automatic evaluation of rivals in jealousy-evoking situations 2009-12: Bart Terwel: Origins and consequences of public trust: Towards an understanding of public acceptance of carbon dioxide capture and storage 2009-13: Emma ter Mors: Dealing with information about complex issues: The role of source perceptions 2009-14: Martijn Veltkamp: On the Instigation of Implicit Motivation: How Deprivation and Positive Affect Cause Motivated Behavior 2009-15: Marret K. Noordewier: Consistency and the unexpected 2009-16: Sytske van der Velde: Imitation of Emotion: How meaning affects the link between imitation and liking 2009-17: Jacomijn Hofstra: Attaching Cultures: The role of attachment styles in explaining majority members' acculturation attitudes 2009-18: Jacqueline Tanghe: Affect in Groups: Convergence, Conditions and Consequences 2009-19: Anne Marike Lokhorst: Using Commitment to Improve Environmental Quality 2009-20: Jonathan van ‘t Riet: Framing Health Communication Messages 2009-21: Suzanne Pietersma: Persuasive Health Communication: A Self-Perspective 2009-22: Remco Wijn: A functional perspective on the justice judgment process and its consequences 2009-23: Niels van de Ven: The bright side of a deadly sin: The psychology of envy 2009-24: Anthon Klapwijk: The Power of Interpersonal Generosity 2010-1: Maarten Wubben: Social Functions of Emotions in Social Dilemmas 2010-2: Joyce Rupert: Diversity faultiness and team learning 2010-3: Daniel Lakens: Abstract Concepts in Grounded Cognition 2010-4: Luuk Albers: Double You? Function and Form of Implicit and Explicit Self-Esteem