ArticlePDF Available

Self-Esteem and Memory

Authors:

Abstract and Figures

This article describes two potential bases for memory bias associated with global self-esteem. According to the mood-congruence model, activation of either dimension of self-esteem (self-competence or self-liking) produces an affective state that facilitates retrieval of traces that are consistent with that state while hindering retrieval of traces that are inconsistent. According to the relevance model, activation of either dimension results in superior encoding of matching negative content by individuals who are low on the dimension. Three studies were conducted to determine which model best accounts for the pattern of bias across distinct content categories. Results were generally consistent with the relevance model.
Content may be subject to copyright.
Self-Esteem and Memory
Romin W. Tafarodi and Tara C. Marshall
University of Toronto
Alan B. Milne
University of Aberdeen
This article describes two potential bases for memory bias associated with global self-esteem. According
to the mood-congruence model, activation of either dimension of self-esteem (self-competence or
self-liking) produces an affective state that facilitates retrieval of traces that are consistent with that state
while hindering retrieval of traces that are inconsistent. According to the relevance model, activation of
either dimension results in superior encoding of matching negative content by individuals who are low
on the dimension. Three studies were conducted to determine which model best accounts for the pattern
of bias across distinct content categories. Results were generally consistent with the relevance model.
“The devil!...what beastly incidents our memories insist on cher-
ishing!...theugly and disgusting...thebeautiful things we have to
keep diaries to remember.”
So observes Charles Marsden in Eugene O’Neill’s Strange
Interlude (1928), following the shameful recollection of his en-
counter with a prostitute. As with many bleak perceptions in
O’Neill’s plays, the lines betray the author’s own frustration in
trying to transcend the troubling elements of his past. A tortured
soul, O’Neill was as ambivalent about himself as he was about the
family and society he grew up in. Plagued by guilt and self-doubt
until his death from illness in 1953, his literary success owed much
to the “ugly and disgusting” ghosts that haunted him. Through the
alchemy of his craft, dark memories were transformed into tragic
and moving characterizations of conflict, betrayal, and moral
frailty. Perhaps more than he wished, O’Neill saw himself in the
unhappy past, and the past in his unhappy self. Had the playwright
been more at peace with himself, would the shadows of his past
have been so intrusive? More generally, do our generalized atti-
tudes toward ourselves affect what we are able to remember of our
experiences?
The interdependence of memory and personal identity has rarely
been doubted. The reciprocal influences that account for this
relation have received much attention in recent years (Bruner,
1994; Kihlstrom, 1997; Singer & Salovey, 1996). In this vein, we
focus here on the role of self-esteem in producing selective mem-
ory of what is experienced. Two hypothetical models of selectivity
as a function of self-esteem are presented and tested. Toward
describing the key elements of these models, the duality of global
self-esteem is first discussed.
Self-Competence and Self-Liking
Self-esteem is essentially a valuative phenomenon. Value, as
applied to a person, can be understood according to the axiological
distinction between means and ends (Dewey, 1939). That is, indi-
viduals are seen as good for what they can do (instrumental and
technical value) as well as who they are (character, appearance,
social identity, and inherent worth as a person). Informally, this is
often expressed as the distinction between “respect” and “liking.”
The former is founded on observable abilities, skills, and talents;
the latter on moral qualities, attractiveness, membership in valued
groups, and other aspects of social worth. Admittedly, the two
aspects are not independent, for a specific competency may be
celebrated as a virtue that is inherently good, and, likewise, qual-
ities that are judged inherently good can be used instrumentally to
great effect. Despite this overlap, the distinction is worth main-
taining for the purpose of discussing the duality of self-esteem.
Namely, we assign two distinct types of value to ourselves just as
we do to others. Consistent with this, the competence and social-
worth aspects of self-esteem have been distinguished by a range of
theorists over the past half-century (Bandura, 1986; Brissett, 1972;
Brown, 1998; Diggory, 1966; Franks & Marolla, 1976; Gecas,
1971; Silverberg, 1952; White, 1963). The most explicit treatment
was offered by Tafarodi and Swann (1995, 2001), who labeled the
two aspects self-competence (SC) and self-liking (SL).
SC is defined as the valuative experience of oneself as a causal
agent, an intentional being that can bring about desired outcomes.
As a generalized trait, it refers to the overall positive or negative
conception of oneself as a source of power and efficacy. The more
successful one has been in fulfilling the countless intentions that
constitute a lifetime of action, the stronger and more effective one
feels. As an aspect of personal identity, this strength is experienced
as positive value, irrespective of any secondary, moral-aesthetic
significance that overlays it. SC is founded on self-efficacy, de-
fined by Bandura (1989) as “people’s beliefs about their capabil-
ities to exercise control over events that control their lives” (p.
Romin W. Tafarodi and Tara C. Marshall, Department of Psychology,
University of Toronto, Toronto, Ontario, Canada; Alan B. Milne, Depart-
ment of Psychology, University of Aberdeen, Aberdeen, Scotland.
This research was supported by Social Sciences and Humanities Re-
search Council of Canada Grant 410–97–1509 awarded to Romin W.
Tafarodi. We thank Debbie Prentice, Jonathan Freedman, and Caroline Ho
for their comments on an earlier version of this article and Veronica Cheng,
Evelyn Gallego, Daniela Guccione, Mariko Lui, and Salwa Maarouf for
their assistance with data collection. Special thanks goes to Sara Davis, Jo
Franklin, and Kate Redman for their contributions to pilot research.
Correspondence concerning this article should be addressed to Romin
W. Tafarodi, Department of Psychology, University of Toronto, 100 St.
George Street, Toronto, Ontario M5S 3G3, Canada. E-mail: tafarodi@
psych.utoronto.ca
Journal of Personality and Social Psychology Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 84, No. 1, 2945 0022-3514/03/$12.00 DOI: 10.1037/0022-3514.84.1.29
29
1 LINE LONG
1175; Bandura, 1992). In its generalized form, self-efficacy refers
to the overall assurance or faith that individuals have in their
ability to achieve their goals (Sherer et al., 1982; Tipton & Worth-
ington, 1984; Woodruff & Cashman, 1993). As Bandura (1990)
has pointed out, however, self-efficacy is distinguishable from
self-esteem. Specifically, whereas self-efficacy refers to beliefs
and expectations about what one can do, SC refers to the direct
valuative significance (both cognitive and affective) of those be-
liefs and expectations for ones self-identity. In other words, SC is
the hedonic imprint that efficacy beliefs leave on the self-concept.
It is how we experience ourselves as a result of believing we can
do things.
The moral-aesthetic significance of ones characteristics and
actions reflects the social side of value. This is represented in
self-esteem as SL, defined as the valuative experience of oneself as
a social object, a good or bad person. As a generalized trait, it
reduces to ones chronic, overall sense of social worth. Social
worth refers to our value as persons, where value is defined within
the particular moral framework of the society in which we live.
The use of socialhere is not intended to imply that SL is limited
to our conception of how worthy we appear to those around us. It
is certainly true that we often perceive that others judge us (or
would judge us, if they knew enough) just as we judge ourselves.
It is also true that self-judgment is often powerfully affected by the
judgments of others. SL, however, refers to the social value we
assign to ourselves, irrespective of how we think others see us. So,
for example, a pillar of the community who is loved and admired
by all may nonetheless suffer a crushing sense of inadequacy that
rests on privately perceived shortcomings. Self-esteem is ulti-
mately a personal valuation. Even when sitting alone in a dark and
soundless room, we all exist as objects to ourselves, positioned in
physical, temporal, and moral space (Harre´, 1991). We require
neither a real nor imagined audience to stand as audience to
ourselves. Accordingly, we cannot help but judge ourselves ac-
cording to the criteria for goodness that we have internalized,
such as charm, beauty, integrity, humanity, divinity, group mem-
bership, and countless other sources of personal worth. SL, insofar
as it refers to social worth, bears some resemblance to Learys
(1999a) conception of self-esteem as a psychological meter, or
gauge, that monitors the quality of peoples relationships with
others (p. 33; Leary, 1999b) and reflects the persons general
sense that he or she is the sort of person who is valued and
accepted by other people(p. 34). It differs from Learys sociom-
eter account, however, by upholding the long-standing definition
of self-regard as the individuals own sense of personal value,
which need not be consistent with the individuals sense of being
valued by others. Again, it is possible, albeit extraordinary, to
scorn oneself in the face of universal adoration and to adore
oneself in the face of universal scorn.
Whether SC and SL are collapsed into a unitary concept of
global self-esteem or used as distinct and separate constructs
ultimately depends on the practical aims of the researcher. A third
approach is to posit a higher order self-esteem that represents the
intercorrelation of the two dimensions. This approach, however, is
open to question. Measures of general self-esteem such as Rosen-
bergs (1965) Self-Esteem Scale do not produce appreciable vari-
ance beyond that jointly accounted for by SC and SL (Tafarodi &
Milne, 2002). Thus, measures of general self-esteem appear to
be redundant with measures of generalized SC and SL, suggesting
that the same pair of constructs is measured in both cases. More-
over, the overlap of the two dimensions requires no explanation
beyond their developmental interdependence and method factors
that affect measurement. Hence, the correlation of SC and SL does
not require the introduction of an additional hypothetical construct.
The sizeable correlation (r .50..70 across studies) does, how-
ever, require formal evidence of discriminant validity to justify
maintaining the conceptual separation. This has been consistently
supported in confirmatory factor analyses, including one con-
ducted within the multitraitmultimethod framework (Tafarodi &
Milne, 2002; Tafarodi & Swann, 1995, 2001).
In summary, the common approach of collapsing SC and SL
into a gross conception of global self-esteem may be appropriate in
many contexts. This strategy, however, may prove self-limiting in
other contexts. Specifically, explicit separation of the two dimen-
sions is recommended when theory suggests that they diverge in
their unique associations with the variables of interest. Such is the
case, we suggest, for selective memory.
Self-Esteem and Selective Memory
Memory bias or selectivity as a function of stable disposition or
character is defined here as the tendency for those high on some
personality trait to recall or recognize a particular type of previous
experience better or worse than those low on the trait, controlling
for general mnemonic ability. Compelling evidence for such se-
lectivity as a function of self-esteem has been rare, as no studies
have clearly addressed it. Of the relevant research, two recent
articles deserve attention. Story (1998, Study 1) found no differ-
ences in accurate recall of positive and negative personality feed-
back for college students who were low versus high in global
self-esteem. When feedback was not accurately recalled, however,
those high in self-esteem provided more positive guesses than did
those low in self-esteem. As less than 40% of the feedback items
were accurately recalled, on average, the study is more revealing
of the significance of self-esteem for forced reproduction of past
events that are not remembered. Furthermore, SC was not distin-
guished from SL, leaving their unique associations unknown. In a
recognition memory study that did adopt the two-dimensional
distinction, Tafarodi (1998) found that those high in SC but low in
SL remembered more negative than positive feedback, whereas
those high in SL but low in SC remembered more positive than
negative feedback. These findings are suggestive. However, they
involve highly select groups and therefore do not reveal the more
general associations of SC and SL with memory. Nor do they
reveal anything about underlying processes.
In contrast to the paucity of research on self-esteem and mem-
ory, the significance of depression for memory has received much
attention. Overall, depression appears to be associated with nega-
tive memory bias for emotionally toned information (see Blaney,
1986; MacLeod & Mathews, 1991; Mineka & Nugent, 1995, for
reviews). Explanations of this association have focused on the
dysphoric mood that is characteristic of depression (e.g., Bower,
1981) or on negative beliefs about the self (e.g., Beck, 1976) as
operative factors. The latter implicates self-esteem (Bernet, In-
gram, & Johnson, 1993; Metalsky, Joiner, Hardin, & Abramson,
1993; Roberts & Monroe, 1994, 1999), recommending direct ex-
amination of its associations with memory bias.
What is needed is a theoretical and methodological framework
capable of explaining the contextual and semantic specificity of
any selectivity associated with self-esteem. Toward that end, we
30
TAFARODI, MARSHALL, AND MILNE
describe two models for how individual differences in self-esteem
may have consequences for what is remembered.
Two Models of Selectivity:
Mood-Congruence and Relevance
The self-concept, or self as represented in memory, is a complex
cognitive representation consisting of both semantic and episodic
elements (Kihlstrom & Cantor, 1984; Klein & Loftus, 1993). SC
and SL are relevant here as generalized valuative aspects of the
semantic self-concept, situated within its hierarchical structure.
The representation of ones own value is arguably the most con-
sequential element within this structure, and therefore should be
associated with considerable emotion. Whatever its level in an
individual, self-esteem cannot be dispassionate. Thus, the repre-
sentations of SL and SC are at once cognitive and affective. More
than just cold beliefs, they are powerful attitudes with motivational
implications.
The differences between SC and SL suggest that their activation
within the self-concept is heightened by qualitatively different
contextual and stimulus-based factors. SC should be responsive to
performance contexts and external primes that connote success/
failure in the pursuit of goals. SL should be responsive to contexts
and primes that heighten awareness of ones significance as a
moral-aesthetic social object. Many real-world contexts and con-
ceptual stimuli are relevant to both dimensions of self-esteem, but
this natural conjunction takes nothing away from their potential for
differential activation.
How might heightened activation of either SC or SL give rise to
memory bias? Two possibilities are outlined here.
Mood-Congruence
The first possibility is one form of mood-congruence effect (see
Blaney, 1986; Bower, 1981; Sedikides, 1992; Segal, 1988; Singer
& Salovey, 1988), whereby information that is consistent with the
valence of ones emotional state tends to be more readily remem-
bered than information that is inconsistent or unrelated. So, for
example, recalling the past weeks minor triumphs is easier when
one is elated than depressed, independent of how one felt at the
time of those events. Successes are inherently positive and there-
fore consistent with a happy state. According to Bowers (1981)
associative network account of the mnemonic effects of mood,
each specific emotion is represented as a node or unit in memory.
When a particular emotion is experienced, the activation of the
corresponding node increases. The excitation then spreads to as-
sociated memory nodes, including knowledge and episodic traces
that are consistent with the valence of the emotion (J. R. Anderson,
1989). The increased subthreshold activation, or accessibility, of
these associated nodes renders them easier to retrieve from mem-
ory while the emotion persists (Tulving & Pearlstone, 1966).
During this time, the individual will be biased toward remember-
ing valuatively consistent information. Also, because excitation of
one emotion node is assumed to inhibit activation of contrasting
emotion nodes (e.g., sad vs. happy) and their associated memory
traces, memory for valuatively inconsistent information should
suffer.
The application of this form of mood-congruence to self-esteem
is straightforward. Both low SC and low SL feel bad, and both
high SC and high SL feel good. Heightened activation of either
dimension will therefore affect mood in a manner that corresponds
to the individuals standing on the dimension. According to the
foregoing account, this should enhance memory for valuative
information that is consistent with the resulting affective state
while impairing memory for information that is inconsistent.
Those who are high on the activated dimension will enjoy en-
hanced memory for positive information and inhibited memory for
negative information. The opposite holds for those who are low on
the activated dimension. This translates into the prediction that
those high in SC (independent of SL) should be better than those
low in SC at remembering positive content but worse at remem-
bering negative content. The same holds for SL (independent of
SC).
Relevance
The second possibility for how self-esteem gives rise to selec-
tive memory relates to the personal relevance of valuative infor-
mation. Those who are low in SC tend to be preoccupied with their
perceived inability and lack of success. They are therefore quicker
than those high in SC at identifying information suggestive of
failure or inefficacy (Tafarodi & Milne, 2002). Such information,
we assume, is richly encoded because of its relevance to the
abiding personal concerns of these individuals and elaborated on
through connection with the dense network of semantic and epi-
sodic nodes surrounding the representation of low SC. Attempts to
recall the information would be facilitated by the additional re-
trieval cues that this connection provides (Alba & Hasher, 1983).
Thus, frustrated actions and threats to goal fulfillment should be
keenly remembered by those who lack SC. More generally, this
translates into the prediction that those low in SC (independent of
SL) should be better than those high in SC at remembering content
related to weak agency. The opposite, however, cannot be said of
content related to strong agency. Those with negative self-views
often hold stringent self-ideals and experience intense dissatisfac-
tion when falling short of these ideals (Higgins, Klein, & Strau-
man, 1987; Kuiper, Olinger, & MacDonald, 1988). Preoccupation
with ones failings entails a keen awareness with what one has
failed to achieve, embody, or otherwise live up to. This suggests
that conceptual nodes representing imperatives of success,
achievement, and the realization of goals are at least as strongly
associated in memory with the representation of low SC as with
high SC. Consistent with this reasoning, those low in SC have been
found to be as quick as those high in SC at identifying information
related to strong agency (Tafarodi & Milne, 2002). Thus, there is
little reason to expect richer encoding and greater elaboration of
information related to strong agency by those high than by those
low in SC. Such information is equally relevant for both groups
and therefore should be equally memorable.
A parallel argument applies to SL. Those who are low on this
dimension tend to be preoccupied with concerns about social
worth. Dominant themes include guilt over perceived transgres-
sions, concerns about physical appearance, dissatisfaction with
social identity, and fears of rejection or disapproval by others.
Those low in SL are therefore quicker than those high in SL at
identifying information suggestive of badness or unworthiness
(Tafarodi & Milne, 2002). As was the case for SC, the heightened
personal relevance of this information should render it especially
memorable for those who lack SL. This translates into the predic-
tion that those low in SL (independent of SC) should be better than
31
SELF-ESTEEM AND MEMORY
those high in SL at remembering content related to low social
worth. As before, however, content related to high social worth is
expected to be as relevant to those low as those high in SL,
resulting in similar memory for it.
The relevance model suggests a form of schematicity (Markus,
1977; Markus, Smith, & Moreland, 1985) for domain-specific
negative content in those low on either dimension of self-esteem.
It should be noted, however, that a parallel schematicity for pos-
itive content does not apply to those high on either dimension, as
there is little basis for assuming a distinctive focus on ideals in this
group. We suggest that fixation and preoccupation in the context of
self-esteem is bred of perceived inadequacy, not fulfillment, and is
limited to negative rumination. Furthermore, the consequences of
relevance for those low on either dimension should not generalize
to schema-inconsistent positive content (see Markus, 1977, Study
1). Relevance, as described here, is valence-specific.
The mood-congruence and relevance models describe distinct
forms of selectivity. The mood-congruence model posits a
retrieval-based process
1
by which the affective state aroused by
activation of either dimension of self-esteem leads to subthreshold
excitation of trace content that is consistent with that state and
inhibition of trace content that is inconsistent. The resulting in-
creased or decreased accessibility of this content directly affects
the likelihood of retrieving it. The relevance model, however,
posits an encoding-based process by which negative content re-
flecting on either activated dimension is represented with greater
richness, and is therefore better remembered, by those who are low
on that specific dimension. The two models differ in several
important ways. First, selectivity based on the mood-congruence
model is hypothesized to occur during retrieval, whereas selectiv-
ity based on relevance assumes differential encoding. Second,
mood-congruence predicts selectivity for both positive and nega-
tive content, whereas relevance predicts selectivity only for neg-
ative content. Third, selectivity due to mood-congruence hinges
solely on the valence of the content to be remembered, whereas
selectivity due to relevance hinges on the substantive address of
the content (agency vs. social worth). That is, according to the
mood-congruence model, activation of either dimension of self-
esteem biases memory for all emotionally toned content, as con-
gruence here relates only to gross valence. In contrast, the rele-
vance model holds that activation of either dimension of self-
esteem produces selectivity only for negative content that is
specifically related to that dimension. The differences between the
two models, including their divergent predictions, are summarized
in Table 1.
The mood-congruence and relevance models do not represent
strictly competing alternatives. The distinct processes they de-
scribe may coincide. Even so, appropriate tests of the association
of self-esteem with selective memory should be capable of con-
firming or disconfirming both models, given that each generates
distinctive predictions that are falsifiable, even in contexts where
the alternative process is in effect. The three studies reported here
provide such tests. Each assesses the unique or independent asso-
ciations of SC and SL with memory for content representing
distinct categories of valuative information.
Study 1
Overview
College students viewed a series of trait words, including some
representing low and high agency and social worth. They were
then asked to recall the words. Students SC and SL scores were
used to predict the degree of recall within each semantic category.
Selectivity uniquely attributable to each dimension of self-esteem
1
A distinct form of mood-congruence effect, whereby matching affec
-
tive states at encoding and retrieval facilitate memory, is not as relevant.
The consistency of affect across the brief latency periods examined in the
studies that follow is assumed to hold equally across individuals, regardless
of their levels of self-esteem. An encoding-based mood-congruence effect,
however, is a more plausible alternative. Evidence for this form of the
effect is fairly consistent (see Singer & Salovey, 1988), but comes mainly
from studies that did not use controlled presentation of content. This fits
with Bowers (1981) original casting of encoding-based mood-congruence
effects as operating through selective attention, perceptual salience, lower
recognition threshold, and selective learning. One implication of Bowers
(1981) claims is that mood-congruent content is subject to greater attention
and thought because it is more relevant to the mood-dependent motiva-
tional state of the perceiver. This sounds much like our relevance model,
substituting mood associated with self beliefs for the beliefs themselves.
Such a mood-based relevance account, however, does not predict the
asymmetry or specificity of our relevance model. In fact, it generates the
same predictions in relation to self-esteem and memory as does its
retrieval-based cousin. As such, we prefer to revisit the encoding versus
retrieval issue in relation to mood only if the results obtained are at all
consistent with a mood-congruence account.
Table 1
Comparison of Mood-Congruence and Relevance Models of Self-Esteem-Consistent Selectivity
Model feature Mood-congruence Relevance
Locus of effect Retrieval Encoding
Symmetry Symmetrical (applies to both negative
and positive content)
Asymmetrical (applies only to negative content)
Specificity Match defined by valence Match defined by valence and semantic address
Predictions Self-competence Self-competence
3 memory for negative content
3 memory for positive content
3 memory for content suggestive of weak
agency
Self-liking Self-liking
3 memory for negative content
3 memory for positive content
3 memory for content suggestive of low
social worth
Note. Arrows represent independent associations, with direction indicated to the left.
32
TAFARODI, MARSHALL, AND MILNE
was tested for correspondence with the mood-congruence and
relevance models.
Method
Participants
Participants were 114 women enrolled in an introductory psychology
course at the University of Toronto, Toronto, Ontario, Canada. All students
participated in exchange for course credit. The modal age was 19 years.
Materials and Procedure
Participants were individually run through a computer-based task by a
female experimenter. At the outset, they were informed that the task
involved thinking about the meaning of a number of personality trait
words. Care was taken to avoid any clue that memory for the trait words
would be tested, while also emphasizing that each word was to be carefully
attended to.
Mnemonic selectivity as a function of self-esteem requires activation of
the self-concept during encoding (relevance model) or retrieval (mood-
congruence model). As with other forms of self-related bias, selectivity due
to SC and SL should not extend to contexts where the semantic network
containing representations of these attitudes is not activated (e.g., Pi-
etromonaco & Markus, 1985; Pyszczynski, Hamilton, Herring, & Green-
berg, 1989; Schlenker & Britt, 1996). To confirm this contingency, encod-
ing context (EC) was manipulated. One half of the participants were
instructed to consider each trait word presented in relation to their own
personality (self-reference condition), whereas the other half were in-
structed to consider each word in relation to the average female student
(other-reference condition). No selectivity related to self-esteem was ex-
pected in the latter condition.
Sixty trait words were serially presented. Controlled presentation was
used to fix inspection time (item strength). Each word appeared in lower-
case font centered on an otherwise blank monitor screen for 3 s, followed
by1sofblank screen, and then the next word. Participants were instructed
to think carefully about each word for the full duration of its presentation.
The first 6 and last 6 words were presented in the same order for all
participants. The order of the remaining 48 words was random. Five
categories were represented by these words (see the Appendix): High
Competence (C), Low Competence (C), High Social Worth (W),
Low Social Worth (W), and Neutral (N). The 16 words in the last
category (N) were selected from the neutral range (neither positive nor
negative in perceived meaning) in N. H. Andersons (1968) normed list of
trait adjectives and were not clearly indicative of either competence or
social worth. The neutral words were included to permit estimation of
general retrieval ability, an important individual-difference variable to
control for in tests of selectivity (Bors & MacLeod, 1996). The remaining
four categories were represented by 8 words each. These words had been
confirmed through preliminary research to be highly indicative of their
semantic category, as reflected in college students judgments of their
applicability to the experience of low and high SC and SL. There was no
need to match categories on normative memorability (e.g., word frequency,
imagability), as all model predictions are associative rather than cross-
categorical. To reduce structural similarity of words across valence cate-
gories, negated versions of the positive words were avoided in representing
the negative categories (e.g., incompetent was not used with compe-
tent, nor unlikable with likable).
Immediately following the presentation phase, participants were given 5
min to write down as many words as they could recall. The balance of
competence- and social-worth-related words in the memory set was in-
tended to produce similarly heightened activation of SC and SL within the
self-concept during self-referential encoding, as required by the relevance
model to produce selectivity in line with both dimensions at the same time.
Furthermore, because recall was tested without delay, the increased acti-
vation of both dimensions would extend into retrieval, providing the
context needed for selectivity due to mood-congruence.
Finally, participants completed the Self-Liking/Self-Competence Scale
(SLCS; Tafarodi & Swann, 1995), followed by several other self-report
measures that are not relevant here.
2
The SLCS consists of two 10-item
subscales, one designed to measure trait SC and the other trait SL. Re-
spondents indicate degree of agreement with global statements reflecting
low or high SC (e.g., I dont succeed at much; I am a capable person)
or SL (e.g., I feel worthless at times; I like myself). The reliability and
validity of the correlated subscales have been supported elsewhere across
a range of applications (e.g., Aidman, 1999; Bosson & Swann, 1999;
Tafarodi, 1998; Tafarodi, Lang, & Smith, 1999; Tafarodi & Milne, 2002;
Tafarodi & Swann, 1995, 1996, 2001; Tafarodi & Walters, 1999). The
duration of the entire procedure, including full debriefing, was approxi-
mately 45 min.
Results
Preliminary
Two participants were eliminated as clear multivariate outliers,
as reflected in their discontinuously high Mahalanobis D
2
values
for the variables analyzed below (see Khattree & Naik, 1995). This
left a final sample size of 112 (57 self-reference and 55 other-
reference). Total-sample means for SC and SL were 40.13
and 35.48, respectively, with no reliable difference across encod-
ing conditions, t(110) 0.95, p .34 for SC and t(110) 0.79,
p .43 for SL. The SLSC correlation (r .62) was comparable
with values found in past studies and similar across conditions,
t(108) 1.30, p .20.
Selectivity in Recall
The dependent variables were the correct recall proportions for
the four categories of self-esteem-relevant words (C,C,W,
and W). The proportion of N words correctly recalled served
only as a covariate in the analyses that follow, to control for
general retrieval ability. In calculating proportions, close repro-
ductions that preserved the valence (e.g., ignorable for ignored;
liked for likable) were counted as half a word to credit partial
memory. Only .78% of the recalled words qualified as such.
3
Means and standard deviations appear in Table 2. Consistent with
past research (see Symons & Johnson, 1997, for a review), better
overall recall resulted from self-reference than from other-
reference, F(1, 110) 136.50, p .0001. This reflects the
enhanced organization and semantic elaboration of explicitly self-
relevant information.
Simultaneous multiple regression was used to examine selectiv-
ity uniquely attributable to each dimension of self-esteem. This
2
Participants had completed the SLCS once before, during a mass
testing session conducted at the beginning of the academic term. Analysis
confirmed that change in SC and SL scores across the two administrations
was unrelated to encoding condition, as would be expected for a trait
measure. Furthermore, degree of change (holding resulting levels of SC
and SL constant) was unrelated to memory performance. The same inde-
pendence was confirmed in Studies 2 and 3. Thus, SLCS scores on the
second administration did not appear to be influenced by the preceding
memory task.
3
Neither eliminating these partial memories from the recall counts nor
assigning them full credit altered the pattern of results reported here. The
same was true in Study 3.
33
SELF-ESTEEM AND MEMORY
required statistically controlling for each dimension of self-esteem
when testing the association of the other. All noncategorical vari-
ables were standardized prior to model testing to facilitate inter-
pretation of interactions. Dummy coding was used to represent EC.
Recall for each of the four categories was regressed on N recall,
EC, SC, SL, EC SC, and EC SL.
4
The interactions serve as
slope shift coefficients, which, if significant, indicate unequal
partial associations for SC or SL across encoding conditions. Thus,
only SC, SL, EC SC, and EC SL are predictors relevant to the
selectivity hypotheses.
The results for the self-esteem predictors appear in Table 3. The
mood-congruence model but not the relevance model predicts
selectivity for positive words in the self-reference condition. Con-
sistent with the relevance model, none of the self-esteem predictors
were significant in relation to recall of C or W words. Both
models predict selectivity for W words in the self-reference
condition. They differ, however, in that the mood-congruence
model predicts that both SL and SC will be independently asso-
ciated with recall, whereas the relevance model predicts that only
the semantically matching dimension, SL, will be independently
associated. The results supported the relevance model, with only
the EC SL interaction emerging as significant,
⫽⫺.53,
t(105) ⫽⫺2.53, p .01 (see Figure 1). Given the dummy coding
used, the nonsignificant partial association of SL with memory for
W words that appears in Table 3,
.01, t(105) 0.05, p
.96, represents the simple slope for the other-reference condition,
indicating no selectivity in this EC. The same, however, was not
true of the simple slope for SL in the self-reference condition,
.52, t(105) ⫽⫺3.62, p .0005. The direction of this significant
slope reveals that more W words were recalled by those lower in
SL, as predicted by the relevance model. Additional testing re-
vealed that the simple slope for SL was significantly more negative
than that for SC in the self-reference condition, F(1, 105) 7.89,
p .006, but not in the other-reference condition, F(1,
105) 0.02, p .88. Furthermore, multivariate comparisons
using Wilkss lambda confirmed that the simple slope for SL in the
self-reference condition was significantly more negative for W
recall than for C (p .03), W (p .01), and C recall (p
.02). Finally, both models predict selectivity for C words in the
self-reference condition. The mood-congruence model again pre-
dicts that both SL and SC will be independently associated with
recall, whereas the relevance model predicts that only SC will be
independently associated. Contrary to both models, no self-esteem
predictors were significant.
Discussion
The results provide partial support for the relevance model of
selectivity. The mood-congruence model, however, received no
support. With regard to the latter, null result, it may be that
variability in mood as a function of self-esteem is quite modest in
a random sample of college students. The sample distributions are
telling here. For both subscales, the absolute attitudinal midpoint
is 30. Someone with this score would, in the aggregate, be per-
fectly ambivalent or neutral, similarly endorsing items reflecting
low and high self-esteem. The percentage of the sample below this
midpoint was 2% for SC and 23% for SL. This implies only a
modest number of participants with sharply negative feelings
about themselves on SL and virtually none on SC. In contrast,
much of the previous evidence for mood-congruence in memory
comes from studies where intensive mood induction was used to
examine the effects of acutely elated or depressed states, or where
4
Initial models included the SC SL interaction, all quadratic terms,
and their related slope shift terms. As none of these higher order terms were
close to significance, they were eliminated from the model to preserve
degrees of freedom and focus testing, as is appropriate (Darlington, 1990).
The same was true for all other regressions conducted in the three studies
reported in this article. Furthermore, for all regressions conducted, sample
size was sufficient to ensure power .80 for squared partial coefficients
representing smallmedium effect sizes, according to the detailed recom-
mendations of Green (1991). Effects any smaller than this would be of
negligible substantive significance. Examination of variance inflation fac-
tors and eigenvalues (see Freund & Littell, 1992) confirmed that the
SCSL correlation did not result in problems of multicollinearity for any of
the regressions reported. Neither was there evidence of suppression, as
reflected in comparison of direct versus partial SC and SL associations (see
Tzelgov & Henik, 1991). Finally, preliminary analyses confirmed homo-
geneity of covariance across encoding conditions for all covariates used in
the regressions.
Table 2
Proportion Recall as a Function of Semantic Category and
Encoding Context
Category
Encoding
Self-reference Other-reference
High competence .40 (.18) .27 (.19)
Low competence .28 (.12) .17 (.14)
High social worth .50 (.18) .28 (.15)
Low social worth .41 (.17) .26 (.13)
Neutral .53 (.15) .17 (.09)
Note. Standard deviations appear in parentheses.
Table 3
Proportion Recall Predicted by Self-Esteem as a Function of
Semantic Category
Category Predictor
t(105) p
High competence
SC .08 0.53 .60
SL .01 0.08 .94
ECSC .06 0.24 .81
ECSL .13 0.57 .57
Low competence
SC .05 0.32 .75
SL .03 0.16 .87
ECSC .04 0.18 .86
ECSL .17 0.77 .44
High social worth
SC .10 0.68 .50
SL .14 0.94 .35
ECSC .02 0.07 .94
ECSL .21 0.99 .32
Low social worth
SC .03 0.23 .82
SL .01 0.05 .96
ECSC .23 1.07 .29
ECSL .53 2.53 .01
Note. SC self-competence; SL self-liking; EC encoding context
(self-reference vs. other-reference).
34
TAFARODI, MARSHALL, AND MILNE
chronically depressed and nondepressed participants were com-
pared. The affective range of the present study does not parallel the
power of such discrete, manufactured contrasts. The variability in
self-esteem within a nonselect sample of students is likely to
correspond to subtle differences in gross emotional state. The same
variability, however, may correspond to more pronounced differ-
ences in what individuals are differentially concerned with. This is
consistent with the form of selectivity found in relation to SL.
As predicted by the relevance model, SL, independent of SC,
was inversely related to recall of words conveying lack of social
worth. This supports the hypothesis that the greater relevance of
such material for those with doubts about their social value results
in richer and more accessible memory traces. Also as predicted,
there was no association of SL with words conveying high social
worth, which are held to be similarly relevant at all levels of SL.
Support for the relevance model, however, was only partial, as a
parallel associative pattern was not found for SC. This requires
explanation. Nearly 30% of the words in the low SC category were
recalled in the self-reference condition, discounting the possibility
of ceiling or floor effects. The limited sampling of participants
with low SC, mentioned above, is one possible cause of the failure
to detect an association. Before examining other possibilities, the
findings need to be replicated in extension to other forms of
memory. Accordingly, we looked at recognition memory for the
same material in Study 2.
5
Study 2
Method
Participants
Participants were 126 women enrolled in an introductory psychology
course at the University of Toronto. All students participated in exchange
for course credit. The modal age was 19 years.
Materials and Procedure
The materials and procedure were identical to that of Study 1 through to
the end of the presentation of the series of trait words. EC was manipulated
as before. Immediately following the last word, instructions for a digit-
detection task appeared on the monitor screen. This task was a distraction,
aimed at producing enough memory decay to allow for a sufficiently
discriminating recognition test. Pilot testing had revealed that without such
decay, recognition accuracy was generally too high. For the task, partici-
pants viewed strings of digits presented for 500 ms each with a 1-s
interstimulus interval. They were required to press a bar whenever they saw
the digit 5 in a string. The task lasted 3 min.
Recognition memory was then tested by combining the 48 trait words
that had been presented (8 C,8C,8W,8W, and 16 N) with 48
new trait words. One third (16) of the new words were drawn from the
neutral range of N. H. Andersons (1968) list to provide a matched set of
foils for the 16 old N words. Another third was taken from the positive
range of the list (e.g., realistic, enthusiastic) to provide foils for the old C
and W words. The final third was taken from the negative range of the
list (e.g., shallow, hostile) to provide foils for the C and W words.
None of the new positive and negative words were highly indicative of
competence or social worth. This avoided close synonymy with their old
counterparts, a feature that would have caused significant mnemonic con-
fusion (Roediger, McDermott, & Robinson, 1998). The 96 words were
presented in random order and participants made forced old-new judg-
ments about each of them. Afterward, participants completed the SLCS and
several unrelated measures before being fully debriefed. The entire session
lasted approximately 45 min.
Results
Preliminary
Two participants were eliminated, 1 due to confessing that she
had expected a memory test and rehearsed accordingly, and the
other as a clear multivariate outlier on the variables analyzed
below. This left a final sample size of 124 (61 self-reference
5
Both recognition and free recall reflect explicit memory, which is
primarily sensitive to conceptual processing of sensory content (Jacoby,
1983). Recall, however, is somewhat more effortful, requiring self-
generation of cues and candidates for retrieval from episodic memory
(Raaijmakers & Shiffrin, 1981). Recognition, as most often tested, requires
discriminative judgments of whether old and new presented items have
been encountered before in a specific context. Such judgments may be
guided as much by the familiarity of the item as by its retrieval from
episodic memory (Gardiner & Java, 1993; Mandler, 1980). The mood-
congruence and relevance models of selectivity apply to retrieval-based
recognition or true remembering (Tulving, 1985) just as they do to free
recall. Their relation to familiarity-based recognition is less straightfor-
ward. The mood-congruence model assumes greater accessibility of con-
tent that is consistent with the gross valence of the emotion node activated
in memory at the time of recognition judgments. This should similarly
increase the familiarity of old and new valence-congruent items in the
recognition test and therefore have little net effect on measures of discrim-
inative sensitivity (see Bower, 1998). The relevance model implies that
those with lower standing on a dimension of self-esteem will tend toward
higher chronic or base familiarity of negative content that is conceptually
consistent with that dimension. Recognition, however, is affected by situ-
ational familiarity, or the proportional increment in base familiarity pro-
duced by its prior occurrence in a specific context (Engelkamp & Zimmer,
1994; Mandler, 1980). This proportional increment will be lower with
increasing base familiarity, suggesting that relevantcontent may actually
be less contextually familiar to those low in self-esteem and therefore more
difficult to discriminate on this basis alone in a recognition memory test.
The upshot of this analysis is that recognition evidence for either selectivity
model could only be attributed to enhanced episodic retrieval of consistent
content, not the greater subthreshold familiarity of its features.
Figure 1. Free recall of low social worth words as a function of self-
liking and encoding context.
35
SELF-ESTEEM AND MEMORY
and 63 other-reference). Total-sample means for SC and SL
were 39.26 and 33.69, respectively. The SCSL correlation was
r .66. Both means and the correlation were similar for the two
encoding conditions (ps .25).
Selectivity in Recognition
A major advantage of recognition tests it that they allow for the
estimation of subjective decision factors that affect the measure-
ment of memory. A standard index of recognition strength or
sensitivity is d, which is independent of response criterion, the
subjective threshold for concluding that a test item was seen before
on the basis of its familiarity. Because criterion oppositely affects
the proportion of old information correctly recognized (hits) and
the proportion of new information incorrectly recognized (false
alarms), neither proportion represents a satisfactory measure of
retention. d, computed as z(hit rate) z(false-alarm rate) (Mac-
millan & Creelman, 1990), takes both proportions into account so
as to effectively remove the influence of criterion (Banks, 1970;
Murdock, 1982; Zuroff, Colussy, & Wielgus, 1986).
Using the five hit and three false-alarm rates, d was computed
separately for the different categories of trait words. False-alarm
rate for positive foils was used in computing d for high C and
W words, whereas false-alarm rate for negative foils was used in
conjunction with C and W words. False-alarm rate for neutral
foils was used for N words. Hits and false alarms were not
analyzed separately, given their confounding with criterion and the
demonstrated symmetry in their susceptibility to factors affecting
memory strength (Glanzer & Adams, 1985, 1990; Glanzer, Ad-
ams, & Iverson, 1991). Means and standard deviations for
category-specific d are given in Table 4. Higher values of d
reflect greater recognition accuracy. Mirroring the recall findings,
self-reference produced greater overall accuracy than did other-
reference, F(1, 122) 11.83, p .0008.
As in Study 1, simultaneous multiple regression was used to
examine selectivity uniquely attributable to each dimension of
self-esteem. Recognition (d) for each of the four self-esteem-
relevant categories was regressed on N recognition (covariate),
SC, SL, EC, EC SC, and EC SL. As can be seen in Table 5,
the results for the self-esteem predictors paralleled those of
Study 1. No selectivity was evident for C,W,orC words,
and only the EC SL was significant for W words,
⫽⫺.48,
t(117) ⫽⫺2.25, p .03 (see Figure 2). Given the dummy coding
used, the nonsignificant partial association of SL with memory for
W words that appears in Table 5,
.00, t(117) ⫽⫺0.03, p
.98, represents the simple slope for the other-reference condition,
indicating no selectivity there. In contrast, the corresponding sim-
ple slope for SL in the self-reference condition,
⫽⫺.48,
t(117) ⫽⫺3.08, p .003, reveals that W words were better
recognized by those lower in SL in this EC, as predicted by the
relevance model. Additional testing revealed that the simple slope
for SL was significantly more negative than that for SC in the
self-reference condition, F(1, 117) 5.93, p .02, but not in the
other-reference condition, F(1, 117) 0.15, p .70. Furthermore,
multivariate comparisons confirmed that the simple slope for SL in
the self-reference condition was significantly more negative for
W recognition than for C (p .04), W (p .003), and C
recognition (p .001).
6
Discussion
The recognition results converge with the recall results of
Study 1 in revealing SL-consistent selectivity for content convey-
ing low social worth. No selectivity was evident, however, for
content conveying high social worth. This asymmetry fits with the
relevance model but not the mood-congruence model. Further-
more, selectivity was evident only in the context of self-reference,
confirming its dependence on activation of the self-concept, and
specifically SL, during encoding.
The results, however, were also asymmetric in a manner that
does not fit with the relevance model. Namely, there was no
evidence for SC-consistent selectivity in memory for content con-
veying weak agency. This returns us to the puzzle of why personal
relevance should be consequential for memory in relation to SL
but not SC.
As noted earlier, one possibility pertains to the narrow sampling
of SC. Recall that only 2% of the participants in Study 1 could be
assumed to see themselves as incompetent in the absolute sense.
The figure was 4% here. The high SC of the participants reflects
their status as first-year students at a nationally top-ranked univer-
sity. The lack of acutely negative attitudes on this dimension,
however, would not preclude detecting a uniform linear associa-
tion such as that found for SL, as long as there was sufficient
variation in the sample, irrespective of range. In fact, the SC : SL
ratio of sample standard deviations was .69 in Study 1 and .66 in
Study 2. Thus, SC was no less than two thirds as variable as SL,
indicating an appreciable amount of variation in the former. This
suggests that the clearly disconfirmatory results for SC were not
primarily due to the characteristics of the sampling distribution.
Furthermore, there was no indication that the weak agency words
were insensitive as an index of memory bias. There was nearly as
much variance in memory for this category as for low social worth
words in Study 1, and more in Study 2.
6
Neither the mood-congruence nor the selectivity model proposes dif
-
ferences in response criterion as a function of self-esteem. Although the
mood-congruence model suggests greater subjective familiarity of both old
and new valence-congruent content during the recognition test, this implies
nothing about the location of the subjective criterion for judging an item
familiar enough to be judged old. Consistent with this, exploratory analyses
revealed that neither dimension of self-esteem was associated with crite-
rion, or c—computed as .5[z(hit rate) z(false-alarm rate)] (see Ingham,
1970; Macmillan & Creelman, 1990)for any of the self-esteem-relevant
word categories.
Table 4
Recognition Accuracy (d) as a Function of Semantic Category
and Encoding Context
Category
Encoding
Self-reference Other-reference
High competence 2.31 (1.15) 1.88 (0.89)
Low competence 2.78 (1.13) 2.15 (0.90)
High social worth 2.37 (1.05) 1.90 (0.81)
Low social worth 3.15 (1.01) 2.68 (0.94)
Neutral 2.30 (0.83) 2.01 (0.83)
Note. Standard deviations appear in parentheses.
36
TAFARODI, MARSHALL, AND MILNE
A second possibility is that, despite the design of the study, SC
and SL were not similarly activated by the EC and the equal
number of relevant words in the memory set. Related findings
(Tafarodi, Tam, & Milne, 2001) suggest that traits words indica-
tive of low social worth, when subject to explicit self-reference of
the sort used here, produce targeted activation of SL. Therefore,
one might reasonably expect explicit self-reference of words rep-
resenting weak agency to produce targeted activation of SC, the
matching dimension. A potential problem, however, is that the
self-reference manipulation used here may itself activate SL, irre-
spective of the particular word being self-referenced. That is,
deliberately focusing on the extent to which any trait wordeven
one that relates to agencyapplies to ones personality (Am I
the kind of person who can be seen as ____ ?) appears to
require consideration of oneself as a social object. Such consider-
ation would activate the social dimension of self-esteem, SL,
thereby preventing clean, asymmetric activation of SC by an
agency-related word. The results of Studies 1 and 2 are consistent
with this possibility. Unintended joint activation of SC and SL by
weak agency words may have obscured the independent associa-
tions of both dimensions with memory for these words. A better
strategy for targeted activation of SC would defend against this
possibility.
One should recall that SC is rooted in agency, or goal-directed
action. As such, it should be maximally activated by settings that
feature individual ability and success/failure outcomes. We created
such an EC in Study 3, using an explicit performance task that
allowed memory for successes and failures to be examined. More-
over, we manipulated the context to create two distinct encoding
conditions, one intended to heighten activation of SC, and the
other, SL.
Study 3
Overview
Participants solved anagrams that had been selected to produce
an equal number of successes and failures. EC was manipulated
such that some participants saw the task for what it was, a test of
puzzle-solving ability (highlighting SC). Other participants were
led to believe that the task would reveal good and bad aspects
of their personality (highlighting SL). Free recall of the original
words was tested later. SC and SL were used to predict recall for
success and failure trials. Selectivity uniquely attributable to each
dimension of self-esteem was tested in relation to the mood-
congruence and relevance models.
Method
Participants
Participants were 152 women enrolled in an introductory psychology
course at the University of Toronto. All students participated in exchange
for course credit. The modal age was 19 years.
Materials and Procedure
Participants were individually run through a computer-based task by a
female experimenter. At the outset, they were informed that the task
involved anagrams. As in the previous studies, care was taken to avoid any
clue that memory of the material would be tested.
Individual performance is inherently self-relevant for ability-based tasks
where success is desired. Strong manipulation of self-reference was there-
fore not feasible here. Recall that selective memory as a function of
self-esteem requires activation of the self-concept during encoding (rele-
vance model) or retrieval (mood-congruence model). The motivational
engagement of participants with the task was apparent from the high level
of performance (and implied effort) and from comments made during
debriefing. Activation of the self-concept in this context therefore can be
assumed (Alper, 1946, 1957; Coopersmith, 1960).
Participants worked on solving a series of subanagrams.For each trial,
a common five-letter, CVCVC word (C consonant, V vowel; e.g.,
Figure 2. Recognition accuracy (d) for low social worth words as a
function of self-liking and encoding context.
Table 5
Recognition Accuracy (d) Predicted by Self-Esteem as a
Function of Semantic Category
Category Predictor
t(117) p
High competence
SC .14 1.12 .26
SL .16 1.12 .27
ECSC .32 1.38 .17
ECSL .10 0.45 .66
Low competence
SC .12 1.09 .28
SL .07 0.55 .58
ECSC .02 0.11 .91
ECSL .29 1.47 .14
High social worth
SC .04 0.21 .83
SL .12 0.81 .42
ECSC .28 1.17 .25
ECSL .22 0.95 .34
Low social worth
SC .09 0.75 .45
SL .00 0.03 .98
ECSC .21 0.94 .35
ECSL .48 2.25 .03
Note. SC self-competence; SL self-liking; EC encoding context
(self-reference vs. other-reference).
37
SELF-ESTEEM AND MEMORY
PAPER, WIDOW, TOTAL) was presented in uppercase at the center of the
monitor screen for 10 s. The words, taken from Locascio and Ley (1972)
and Kuc˘era and Francis (1967), were not indicative of agency or social
worth. Participants were instructed to focus on each word throughout its
presentation while rearranging its letters to derive from it as many common
four-letter words as possible. The instructions made clear that only one
solution was required to successfully solve the puzzle, but multiple solu-
tions were to be sought if the first solution was found with time remaining.
This provision kept inspection time constant. Once the word disappeared,
a response box appeared in its place, asking participants to type in their
solutions. After typing these, participants clicked on a button labeled done
to advance to the next trial. For trials where no solution was found,
participants clicked on a button labeled none found to advance to the next
trial.
There were 55 trials, of three types. For the 15 success trials, only
CVCVC words known to have at least one common solution, and usually
more, were used (e.g., METAL). If the participant came up with at least one
valid solution, as nearly all did, the trial was taken to represent a fulfilled
goal and, therefore, a subjective success. For the 15 failure trials, only
words that had no solution were used (e.g., VISIT). If the participant
clicked the “none found” key in response to these words, as nearly all did,
the trial was taken to represent an unfulfilled goal, and, therefore, a
subjective failure. Of importance, participants were misled to believe that
both success and failure trials had common solutions. This was accom-
plished by emphasizing at the outset that all the words had common
solutions unless otherwise marked. Neither success nor failure trials were
marked. Participants therefore had every reason to believe that they could
succeed on each of these trials. Finally, the 15 neutral trials involved words
that had no solution and were marked at presentation as “probably no
solution available” because no one had found one yet. If the participant
clicked the “none found” key in response to these words, as nearly all did,
the trial was taken to represent neither subjective success nor failure, as a
solution could not be reasonably expected. The 45 critical trials, presented
in random order, were preceded and followed by 5 success trials to
familiarize participants with the task and guard against primacy and re-
cency effects in memory. These buffer trials were not used in assessing
memory.
Our aim was to avoid the spontaneous priming of SL that we suspected
had rendered explicit self-reference of agency-related trait words problem-
atic in the first two studies. As such, the logic of this study was a deliberate
departure from that of the first two. The details of the success trials were
considered positive content. Similarly, the details of the failure trials were
considered negative content. The focal detail of each trial is the CVCVC
word itself. According to the theory presented at the outset, memory for
this detail should be associated with self-esteem. Specifically, insofar as
the task is experienced as a clear performance situation reflecting ability
and agency, SC should be selectively activated. Each CVCVC word is
encoded during the 10 s it remains on the screen. If, during that time, a
solution is not found where one was expected, the relevance model predicts
that the word will be more deeply encoded, and therefore better remem-
bered, by those lower in SC. In contrast, the relevance model predicts no
association with SC for positive details, words for which solutions are
found. As before, the mood-congruence model predicts both better memory
for negative details (failure trial words) and worse memory for positive
details (success trial words) by those lower in SC.
The predictions for SC, however, are conditional. Specifically, EC was
manipulated as a factor in the study such that participants were assigned to
one of two conditions. The procedure as described so far represents the
performance condition, designed as an EC that activates SC over SL. To
ensure that any SC-consistent selectivity could be attributed to this new
EC, a second, character condition was designed to activate SL over SC.
Here, the implicit focus on ability and performance was overridden by
adding the following paragraph to the initial task instructions:
Please be aware that the words that have valid, common-word solu-
tions are not designed to assess your ability to solve word puzzles. Nor
are they designed to assess some other specific or general ability.
Rather, the particular words for which you manage to find solutions
will give us insight into the GOOD or LIKABLE aspects of your
character, whereas the particular words for which you are unable to
find solutions will give us insight into the BAD or DISLIKABLE
aspects of your character. In other words, you can think of this task as
an indirect personality test, the validity of which has been demon-
strated in past research. Keep this in mind as you work through the
task.
By shifting the emphasis to qualities of character in this second condi-
tion, we sought to highlight the relevance of the trials for social worth.
This, we expected, would increase the activation of SL much more than
SC, allowing the former to drive selectivity. The manipulation of EC to
highlight one or the other dimension of self-esteem allowed the prediction
of memory bias in line with SC when the focus was on ability but memory
bias in line with SL when the focus was on character.
7
Following the task, participants were given 5 min to recall as many of
the CVCVC words as possible. Free recall rather than recognition was
tested because of its more modest procedural demands. The parallel find-
ings of Studies 1 and 2, and their interpretation, suggest that a recognition
test would have yielded similar results. Afterward, participants completed
the SLCS, the Positive and Negative Affect Schedule–Expanded Form
(PANAS-X; Watson & Clark, 1994), the Beck Depression Inventory–II
(BDI-II; Beck, Steer, & Brown, 1996). The PANAS-X and BDI-II were
included to directly examine the association of self-reported affective
differences with selectivity. We suggested earlier that the extent of affec-
tive variation in nonselect samples of students might have been too modest
to produce discernible selectivity due to mood-congruence in the first two
studies. If so, then the belief content of self-esteem, rather than the gross
mood associated with it, is what accounts for the selectivity observed. To
provide support for this interpretation, positive affect (PA), negative affect
(NA), and depressive dysphoria were examined here as alternative predic-
tors of selectivity. Evidence that these dimensions are unrelated to memory
in a context where self-esteem is related would confirm that affect is not
the basis of the phenomenon. The entire session lasted approximately 50
min.
Results
Preliminary
Two participants were eliminated as clear multivariate outliers.
This left a final sample size of 150 (76 in the performance
condition and 74 in the character condition). Total-sample means
for SC and SL were 39.75 and 34.90, respectively. The SC–SL
correlation was r .63. Both means and the correlation were
similar for the two encoding conditions (all ps .38).
Selectivity in Recall
Failure and neutral trials resulted in the expected “none found”
response 95% and 93% of the time, respectively. Success trials
7
We thank one of the editors, Debbie Prentice, for suggesting this
general strategy. We should also note that our predictions contrast with the
possibility that those high in self-esteem will be more inclined to remember
failure because it is inconsistent with their high performance expectations.
This possibility, although intuitive, has received little support in the liter-
ature. Rather, in contexts where failures or incompleted tasks are remem-
bered better than successes or completed tasks, those who can be assumed
to be lower in self-esteem show this tendency to a greater degree than do
their high self-esteem counterparts (Alper, 1946, 1957; Coopersmith, 1960;
Moot, Teevan, & Greenfield, 1988).
38
TAFARODI, MARSHALL, AND MILNE
resulted in at least one (M 1.70) valid solution 92% of the time.
Only trials that result in the expected response (valid solution vs.
none found) can be safely assumed to produce the intended
subjective state (success, failure, or neutral). As such, proportion
recall was computed as the number of reproduced words of each
trial type divided by the number of trials of that type for which the
participant had given the expected response. (Reproduced words
representing trials for which participants had not given the ex-
pected response were not included in the number of words repro-
duced.) As in Study 1, close reproductions counted as half a word.
Such cases accounted for only .40% of the recalled words. Mean
proportion recall was .18 (SD .08) for success trials, .19 (SD
.07) for failure trials, and .18 (SD .09) for neutral trials. These
proportions did not differ significantly across encoding conditions
(all ps .17). The modest amount of recall in general is consistent
with the high degree of mnemonic interference from the self-
generated solution words.
Predicting selectivity from self-esteem. As in the first two
studies, simultaneous multiple regression was used to examine
selectivity uniquely attributable to each dimension of self-esteem.
Recall proportions for success and failure trials were separately
regressed on recall proportion for neutral trials (covariate), SC, SL,
EC (encoding condition: performance, character), EC SC, and
EC SL.
8
The relevant results appear in Table 6. Consistent with
the relevance but not the mood-congruence model, no selectivity
was evident for success trial words. For failure trial words, how-
ever, both interactions were clearly significant: EC SC,
.98,
t(143) 3.55, p .0005; EC SL,
⫽⫺.71, t(143) ⫽⫺2.70,
p .008. Given the dummy coding used, the significant partial
association of SC with memory for failure trial words that appears
in Table 6,
⫽⫺.78, t(143) ⫽⫺3.29, p .001, represents the
simple slope for the performance condition, indicating that failure
trial words were recalled better by those lower in SC, as predicted
by the relevance model. In contrast, the corresponding simple
slope for SC in the character condition was nonsignificant,
.20, t(143) 1.43, p .15, confirming the absence of selectivity
when social worth was emphasized (see Figure 3). The form of the
EC SL for failure trials was also consistent with predictions.
Given the dummy coding used, the nonsignificant partial associ-
ation of SL with memory for failure trial words that appears in
Table 6,
.33, t(143) 1.51, p .13, represents the simple
slope for the performance condition, indicating no selectivity in
line with this dimension when the focus was on ability. Although
this slope is clearly positive (see Figure 4), the high standard error
and resulting probability indicates no reliable association. In con-
trast, the corresponding simple slope for SL in the character
condition,
⫽⫺.39, t(143) ⫽⫺2.63, p .009, confirms that
failure trial words were recalled better by those lower in SL when
social worth was emphasized.
Additional testing revealed that the simple slope for SC was
significantly more negative than that of SL in the performance
condition, F(1, 143) 6.55, p .01, but significantly less neg-
ative than that of SL in the character condition, F(1, 143) 4.96,
p .02. Furthermore, multivariate comparisons using Wilkss
lambda confirmed that the simple slope for SC in the performance
condition was significantly more negative for failure trial words
than for success trial words (p .03). Similarly, the simple slope
for SL in the character condition was significantly more negative
for failure trial words than for success trial words (p .02).
Predicting selectivity from affective measures. Were the affec-
tive measures similarly effective in predicting recall of the success
and failure trial words? To answer this question, we adopted the
recommendation of MacLachlan (1985) to avoid content redun-
dancy by eliminating any items from these alternative measures
that appeared to be indicators of self-esteem. Accordingly, the
items strong and proud were removed from the PANAS-X 10-item
PA scale, and ashamed was removed from the 10-item NA scale.
Similarly, 4 items were removed from the 21-item BDI-II (failure,
self-dislike, self-criticalness, and worthlessness). The resulting
means were 20.70 (SD 5.28) for PA, 12.83 (SD 4.42) for NA,
and 9.23 (SD 6.74) for the BDI-II. The means did not differ
significantly across encoding conditions (all ps .31).
As applied to normal college populations, the BDI-II measures
diffuse maladaptive functioning (Beck, Steer, & Garbin, 1988)
8
Participants who generated many rather than few solutions on success
trials might be expected to experience greater interference in memory. To
control for this, we conducted supplementary regressions where the aver-
age number of solutions generated per success trial was included as a
covariate. This addition did not change the pattern of results reported.
Table 6
Proportion Recall Predicted by Self-Esteem as a Function of
Trial Type
Trial type Predictor
t(143) p
Success
SC .12 0.48 .63
SL .06 0.23 .81
ECSC .28 0.93 .35
ECSL .02 0.08 .94
Failure
SC .78 3.29 .001
SL .33 1.51 .13
ECSC .98 3.55 .0005
ECSL .71 2.70 .008
Note. SC self-competence; SL self-liking; EC encoding context
(performance vs. character).
Figure 3. Free recall of failure trial words as a function of self-
competence and encoding context.
39
SELF-ESTEEM AND MEMORY
or self-reported distress (Coyne, 1994) rather than depression
per se. In the present study, BDI-II score was correlated with PA
and NA at .37 and .66, respectively. The PANA correlation was
.18. Regression models parallel to those estimated using the
SLCS predictors were estimated for the two alternative measures.
First, recall proportions for success and failure trials were re-
gressed on recall proportion for neutral trials, PA, NA, EC, EC
PA, and EC NA. The relevant results appear in Table 7. As
before, no significant selectivity was evident for success trial
words. Of more importance, however, no selectivity was evident
for failure trial words either. Next, to examine the possibility of
distress-consistent selectivity, recall proportions for success and
failure trials were regressed on recall proportion for neutral trials,
BDI-II, EC, and EC BDI-II. Mirroring the results for PA and
NA, neither success nor failure trial words showed significant
selectivity (see Table 7).
9
Discussion
The results confirm that EC is critical for determining which
dimension of self-esteem biases memory.
10
When the task focused
on performance and ability, only SC was independently related to
recall of negative (failure-linked) content. However, when the task
drew attention to participants personality or character, only SL
was independently related to recall. This contextual specificity,
together with the absence of selectivity for positive (success-
linked) content, provides support for the relevance model.
The pattern of selectivity was again inconsistent with the mood-
congruence model. Moreover, individual differences in mood, as
indexed by the PANAS-X and the BDI-II, did not predict recall of
negative or positive content. This null finding does not gainsay the
broader importance of mood-congruence effects for memory, of
which there is ample evidence in the literature. It does, however,
support our claim that the association of self-esteem with selective
memory is due primarily to the differential relevance of content
rather than differences in mood. That is, typical variation in
self-esteem may correspond more clearly to differences in what
individuals are specifically concerned with when experiencing
their own value than it does to how they feel overall.
The emergence of SC as a source of selectivity in a true
performance context lends credence to our suspicion that self-
referential encoding did not selectively activate this dimension of
self-esteem in the first two studies. This invites the generalization
9
Our reliance on reduced affective scales raises the question of compa
-
rability with the original measures. Cronbachs alpha was negligibly af-
fected by item exclusion: BDI-II .88 for 17 items versus .90 for 21 items;
PA .81 for 8 items versus .84 for 10 items; and NA .80 for 9 items
versus .82 for 10 items. Moreover, using the full scales in place of the
reduced scales did not change the pattern of regression results: PA, NA,
and BDI-II scores remained nonsignificantly associated with memory for
both success and failure trial words. Finally, entering the three reduced
affective scales alongside SC and SL in the regressions did not change the
pattern of results for success or failure trial words. Critically, both the SC
and SL interactions remained significant in the case of failure trial words
(p .0004 and p .01, respectively), with the pattern of simple slopes as
before. The same was true when the full scales were used (p .0006 and
p .02, respectively).
10
We claimed at the outset that our two-dimensional approach to self-
esteem affords predictive advantages over the unidimensional alternative.
To confirm this, we combined SC and SL into a single score representing
what is commonly measured using Rosenbergs (1965) Self-Esteem Scale
(see Tafarodi & Milne, 2002, on this redundancy). This new predictor was
substituted in for SC and SL in the regressions reported in Study 1. Results
revealed that the combined score was not significantly associated with
recall of C,C,W,orW words. The same strategy was applied to
Study 2. Here, the combined score was associated with recognition mem-
ory for W words in the same manner as SL had been, but model
prediction was weakened (R
2
reduced from .36 to .32). Finally, in Study 3,
the combined score was not significantly associated with recall of success
or failure trial words. Taken as a whole, this pattern illustrates the benefits
of maintaining the conceptual separation of SC and SL in relation to
memory bias.
Figure 4. Free recall of failure trial words as a function of self-liking and
encoding context.
Table 7
Proportion Recall Predicted by Affective Measures as a
Function of Trial Type
Predictor
tp
Success trials
PA .08 0.75 .46
NA .15 1.30 .20
EC PA .30 1.86 .07
EC NA .11 0.65 .52
BDIII .10 0.88 .38
EC BDIII .20 1.16 .25
Failure trials
PA .04 0.33 .74
NA .19 1.70 .10
EC PA .03 0.18 .86
EC NA .01 0.09 .93
BDIII .14 1.22 .22
EC BDIII .07 0.39 .70
Note. The first four and last two predictors listed under both types of
trials are contained in different regression models, with error df 143 and
145, respectively. PA positive affect; NA negative affect; EC
encoding context (performance vs. character); BDIII Beck Depression
InventoryII.
40
TAFARODI, MARSHALL, AND MILNE
that contexts that focus self-awareness on personality rather than
goal-directed action are more relevant for SL than for SC. Dis-
covering the full import of this distinction for the relative influence
of the two dimensions on thought and behavior will require ex-
tended investigation. Nonetheless, some speculation is appropriate
here.
Consideration of the concrete consequences of ones actions in
relation to intended or desired outcomes should asymmetrically
engage the SC dimension of self-esteem. In contrast, consideration
of the social or moral implications of these same consequences for
ones character and social identity should asymmetrically engage
SL. Individuals caught up in the flow of physically or mentally
taxing activity, especially solitary activity, often lack the capacity
or inclination to reflect on the latter, secondary significance of
their actions and outcomes. Take, for example, the experience of
rushing by car to an important out-of-town meeting. The demands
of fighting with traffic and finding the unfamiliar location may
leave you unable to reflect on your social worth in the situation.
This blindness, however, will not prevent you from experiencing a
profound, even visceral, awareness of your own agency (or lack
thereof) and its direct valuative significance as you discover your-
self either late and lost or approaching the destination with time to
spare. In contexts such as this, SC should dominate as the source
of memory bias. Many other activities tend to place the focus of
reflexive attention on ones character and social significance
ones personhood. Mindful social interaction does so necessarily
because of the reciprocal mirroring it affords (Kohut, 1977).
Even solitary activities, however, can promote such a focus, inso-
far as individuals are both disposed and able to consider the
broader implications of what they are doing or have done for who
they are. Here, SL should dominate as the source of memory bias.
Returning to the studies, we should note that the laboratory
setting does not, in general, give rise to a typical level of self-
consciousness. Rather, research participants in the laboratory are
acutely aware of their social significance, wondering how they are
being perceived, interpreted, compared, evaluated, and measured
by the researchers. This bug-in-a-jar mindset, coupled with
consideration of the self-relevance of both social worth and agency
words, may have contributed to our failure to achieve targeted
activation of SC in the first two studies. In contrast, participants in
Study 3 worked at a brisk pace on a demanding task that, in the
performance condition, did not draw attention to social worth.
Presumably, the task was engaging enough to limit their capacity
to reflect on the laboratory scrutiny they were under.
In everyday life, most of what we do holds implications for our
social significance, suggesting that SL may be subject to greater
activation than SC in general. However, as we rush through our
typical day, we are frequently so taken up with the mechanics of
getting things done that we become oblivious to how we appear to
ourselves and others as social objects. At these times, our
reflexive experience is limited largely to that of the self-as-
agent. For those of us whose lives are replete with nonsocial,
goal-directed activities that are at least moderately demanding,
SC may receive at least as much activation as SL. It is therefore
difficult to draw any firm conclusions about which dimension
of self-esteem is likely to be the greater source of memory bias
on the whole and the implications of this for the architecture of
the self-concept. The answers may depend in large part on the
individual.
General Discussion
In this article, we explored the association of self-esteem with
selective memory. In Studies 1 and 2, SL, independent of SC, was
negatively associated with memory for traits words conveying lack
of social worth. The association held only when these words were
encoded in relation to the self, suggesting selective activation of
SL. In Study 3, SL, independent of SC, was negatively associated
with memory for failure-related content. However, the association
held only when performance outcomes were described as diagnos-
tic of social worth rather than ability, again suggesting selective
activation of SL. In contrast, when performance outcomes were
perceived to reflect ability, SC, independent of SL, was negatively
associated with memory for failure-related content, suggesting
selective activation of SC.
Taken together, these findings are consistent with our claim that
information and experiences related to low social worth or weak
agency are most relevant to those who see themselves as embod-
ying these deficits. Greater personal relevance appears to enhance
the strength and efficiency of encoding such that the resulting
memory traces are more easily retrieved in the future. Though not
examined here, heightened spontaneous memory or intrusions
might also be expected, suggesting, in the extreme case, a predis-
position to obsessive ideation (James & Kendell, 1997). Of im-
portance, there appear be two planes of relevance, corresponding
to the two dimensions of self-esteem and their associated cognitive
structures. The functional relevance of sensory content for either
dimension appears to require activation of the corresponding struc-
ture. This activation is determined by both the wider EC and the
semantic or thematic nature of the content.
The findings also support our claim that valuative content re-
lated to high social worth or strong agency is equally relevant to
those low and high on the corresponding dimension of self-esteem.
The ideals against which we measure ourselves may be similarly
represented and associated with self-esteem across individuals,
such that positive content is neither more accessible nor subject to
superior encoding by those who see themselves as satisfying rather
than falling short of these ideals. Several findings in the literature
are consistent with this conclusion. McGuire and McGuire (1996,
Study 1) found that those low vs. high in self-esteem did not differ
in their ability to generate desirable self-relevant characteristics,
although lows were more inclined to see themselves as lacking
rather than possessing these characteristics. Epstein (1992) found
that those low vs. high in constructive thinking (a dimension that
is highly correlated with self-esteem) did not differ in the extent to
which they elaborated on, in an overgeneralizing manner, in-
stances of personal success and approval. Using a Stroop color-
naming task with positive and negative trait adjectives, Segal,
Gemar, Truchon, Guirguis, and Horowitz (1995) produced results
indicating that negative self-descriptive content is more integrated
within the depressive than the nondepressive self-concept (depres-
sion is highly correlated, inversely, with self-esteem). Their results
also suggested, however, that positive self-descriptive content is
no more integrated in the nondepressive than the depressive self-
concept. Finally, Teasdale, Taylor, Cooper, Hayhurst, and Paykel
(1995) found that depressives were no less (in fact, more) likely
than nondepressives to provide sentence completions suggestive of
success and approval when those completions reflected their con-
tingencies of self-acceptance (ideals).
41
SELF-ESTEEM AND MEMORY
The absence in the present studies of any evidence for selectiv-
ity due to mood-congruence suggests that the acute dysphoria
associated with clinical depression or produced by mood induction
procedures is not characteristic of those who are merely lower than
average in SC or SL. As mentioned, the SLCS score distributions
revealed that few participants held strongly negative attitudes
toward themselves. The affective range of the student samples
therefore can be assumed to be quite limited, precluding a pro-
nounced mood-congruence effect. In contrast, one does not have to
feel acutely despondent to be selectively concerned with negative
information. Selective memory as a general function of self-
esteem, then, appears to be due more to the content of self-
conception than to mood. A more detailed understanding of how
this content comes to bias recollection of the past requires further
investigation. For one thing, our claim that the memory bias
observed across studies was due to differential encoding as a
function of self-esteem needs to be directly tested. Furthermore,
the distinct cognitive and motivational consequences of heightened
relevance need to be clearly identified and their separate and
interactive effects on memory processes better understood.
Finally, the broader significance of the link between trait self-
esteem and memory may be more complex, and tragic, than is
apparent in a single-session study. Implicit in the relevance inter-
pretation of the findings is the assumption that self-esteem influ-
ences memory rather than being influenced by it. This assumption
was supported by the lack of any immediate effects of the memory
tasks on the trait measure of self-esteem. More broadly, however,
the remembered past may have profound consequences for the
persistence of self-esteem. Across time, self-esteem has been
shown to fluctuate as a function of negative life events, with
greater frequency of negative events predicting lower self-esteem
(e.g., Lakey, Tardiff, & Drew, 1994; Miller, Kreitman, Ingham, &
Sashidharan, 1989; Mullis, Youngs, Mullis, & Rathge, 1993; Pear-
lin, Menaghan, Lieberman, & Mullan, 1981; Tafarodi & Milne,
2002; Tafarodi & Walters, 1999). Assuming the impact of negative
events to be distributed over time, greater memory for these events
may exacerbate their injury to self-esteem. This implies a vicious
circle. First, low SC and SL promote memories of deficit-relevant
events. These memories then serve as reminders of inadequacy,
confirming the validity of self-doubts and sustaining the negative
preoccupation they produce. Therapeutic interventions aimed at
breaking this circle might include intensive training in deliberate
encoding, rehearsal, and retrieval strategies. Insofar as we are what
we remember of our lives, willful control of memory may help
create the conditions needed to improve self-esteem.
In summary, we have presented evidence that self-esteem is
associated with selective memory for negative information. The
form of bias is consistent with our claim that those who are low in
SC or SL are especially concerned with perceived inadequacy on
the corresponding valuative dimension (agency or social worth).
This heightened concern appears to render deficit-related experi-
ences especially memorable. That self-doubts can weave them-
selves into the fabric of memory in this way suggests that an
unflinching focus on the pains and failures of the past may be more
tragic than heroic. For Eugene ONeill, the playwright we men-
tioned at the outset, preoccupation with the past was fueled by his
contempt for the illusions of his parents, and a desire to avoid a
world where truth is untrue and life can hide from itself. He
relentlessly sought to confront the ghosts that haunted him. In
doing so, however, he may have created a past that was more a
reflection of his tortured soul than of his actual experiences.
References
Aidman, E. V. (1999). Measuring individual differences in implicit self-
concept: Initial validation of the Self-Apperception Test. Personality
and Individual Differences, 27, 211228.
Alba, J. W., & Hasher, L. (1983). Is memory schematic? Psychological
Bulletin, 93, 203231.
Alper, T. G. (1946). Memory for completed and incompleted tasks as a
function of personality: An analysis of group data. Journal of Abnormal
and Social Psychology, 41, 403420.
Alper, T. G. (1957). Predicting the direction of selective recall: Its relation
to ego strength and N achievement. Journal of Abnormal and Social
Psychology, 55, 149165.
Anderson, J. R. (1989). A rational analysis of memory. In H. L. Roediger,
III & F. I. M. Craik (Eds.), Varieties of memory and consciousness:
Esssays in honour of Endel Tulving (pp. 195210). Hillsdale, NJ:
Erlbaum.
Anderson, N. H. (1968). Likableness ratings of 555 personality-trait words.
Journal of Personality and Social Psychology, 9, 272279.
Bandura, A. (1986). Social foundations of thought and action: A social
cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1989). Human agency in social cognitive theory. American
Psychologist, 44, 11751184.
Bandura, A. (1990). Conclusion: Reflections on nonability determinants of
competence. In R. J. Sternberg & J. Kolligian, Jr. (Eds.), Competence
considered (pp. 315362). New Haven, CT: Yale University Press.
Bandura, A. (1992). Exercise of personal agency through the self-efficacy
mechanism. In R. Schwarzer (Ed.), Self-efficacy: Thought control of
action (pp. 338). Washington, DC: Hemisphere Publishing Corpora-
tion.
Banks, W. P. (1970). Signal detection theory and human memory. Psy-
chological Bulletin, 74, 8199.
Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New
York: International Universities Press.
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck
Depression InventoryII. San Antonio, TX: The Psychological Corpo-
ration.
Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties
of the Beck Depression Inventory: Twenty-five years later. Clinical
Psychology Review, 8, 77100.
Bernet, C. Z., Ingram, R. E., & Johnson, B. R. (1993). Self-esteem. In C. G.
Costello (Ed.), Symptoms of depression (pp. 141159). New York: John
Wiley & Sons.
Blaney, P. H. (1986). Affect and memory: A review. Psychological Bul-
letin, 99, 229246.
Bors, D. A., & MacLeod, C. M. (1996). Individual differences in memory.
In E. L. Bjork & R. A. Bjork (Eds.), Memory (pp. 411441). San Diego,
CA: Academic Press.
Bosson, J. K., & Swann, W. B., Jr. (1999). Self-liking, self-competence,
and the quest for self-verification. Personality and Social Psychology
Bulletin, 25, 12301241.
Bower, G. H. (1981). Mood and memory. American Psychologist, 36,
129148.
Bower, G. H. (1998). An associative theory of implicit and explicit mem-
ory. In M. A. Conway, S. E. Gathercole, & C. Cornoldi (Eds.), Theories
of memory. Vol. 2 (pp. 2560). Hove, United Kingdom: Psychology
Press.
Brissett, D. (1972). Toward a clarification of self-esteem. Psychiatry, 35,
255262.
Brown, J. D. (1998). The self. New York: McGraw-Hill.
Bruner, J. (1994). The remembered self. In U. Neisser & R. Fivush
(Eds.), The remembering self: Construction and accuracy in the self-
narrative (pp. 4154). New York: Cambridge University Press.
42
TAFARODI, MARSHALL, AND MILNE
Coopersmith, S. (1960). Self-esteem and need achievement as determinants
of selective recall and repetition. Journal of Abnormal and Social
Psychology, 60, 310317.
Coyne, J. C. (1994). Self-reported distress: Analog or ersatz depression?
Psychological Bulletin, 116, 2945.
Darlington, R. B. (1990). Regression and linear models. New York:
McGraw-Hill.
Dewey, J. (1939). Theory of valuation. Chicago: University of Chicago
Press.
Diggory, J. C. (1966). Self-evaluation: Concepts and studies. New York:
Wiley.
Engelkamp, J., & Zimmer, H. D. (1994). The human memory: A multi-
modal approach. Seattle, WA: Hogrefe & Huber Publishers.
Epstein, S. (1992). Coping ability, negative self-evaluation, and overgen-
eralization: Experiment and theory. Journal of Personality and Social
Psychology, 62, 826836.
Franks, D. D., & Marolla, J. (1976). Efficacious action and social approval
as interacting dimensions of self-esteem: A tentative formulation
through construct validation. Sociometry, 39, 324341.
Freund, R. J., & Littell, R. C. (1992). SAS System for Regression (2nd ed.).
Cary, NC: SAS Institute.
Gardiner, J. M., & Java, R. I. (1993). Recognizing and remembering. In
A. E. Collins, S. E. Gathercole, M. A. Conway, & P. E. Morris (Eds.),
Theories of memory, Vol. 1 (pp. 163188). Hove, United Kingdom:
Psychology Press.
Gecas, V. (1971). Parental behavior and dimensions of adolescent self-
evaluation. Sociometry, 34, 466482.
Glanzer, M., & Adams, J. K. (1985). The mirror effect in recognition
memory. Memory & Cognition, 13, 820.
Glanzer, M., & Adams, J. K. (1990). The mirror effect in recognition
memory: Data and theory. Journal of Experimental Psychology: Learn-
ing, Memory, and Cognition, 16, 516.
Glanzer, M., Adams, J. K., & Iverson, G. (1991). Forgetting and the mirror
effect in recognition memory: Concentering of underlying distributions.
Journal of Experimental Psychology: Learning, Memory, and Cogni-
tion, 17, 8193.
Green, S. B. (1991). How many subjects does it take to do a regression
analysis? Multivariate Behavioral Research, 26, 499510.
Higgins, E. T., Klein, R. L., & Strauman, T. J. (1987). Self-discrepancies:
Distinguishing among self-states, self-states conflicts, and emotional
vulnerabilities. In K. Yardley & T. Honess (Eds.), Self and identity:
Psychosocial perspectives (pp. 173186). New York: Wiley.
Ingham, J. G. (1970). Individual differences in signal detection. Acta
Psychologica, 34, 3950.
Jacoby, L. L. (1983). Remembering the data: Analyzing interactive pro-
cesses in reading: Journal of Verbal Learning and Verbal Behavior, 22,
485508.
James, I. A., & Kendell, K. (1997). Unfinished processing in the emotional
disorders: The Zeigarnik effect. Behavioural and Cognitive Psychother-
apy, 25, 329337.
Khattree, R., & Naik, D. N. (1995). Applied multivariate statistics with SAS
software. Cary, NC: SAS Institute.
Kihlstrom, J. F. (1997). Consciousness and me-ness. In J. D. Cohen &
J. W. Schooler (Eds.), Scientific approaches to consciousness (pp. 451
468). Mahwah, NJ: Erlbaum.
Kihlstrom, J. F., & Cantor, N. (1984). Mental representations of self.
Advances in Experimental Social Psychology, 17, 147.
Klein, S. B., & Loftus, J. (1993). Behavioral experience and trait judgments
about the self. Personality and Social Psychology Bulletin, 19, 740745.
Kohut, H. (1977). The restoration of the self. New York: International
Universities Press.
Kucˇera, H., & Francis, W. N. (1967). Computational analysis of present-
day American English. Providence, RI: Brown University Press.
Kuiper, N. A., Olinger, L. J., & MacDonald, M. R. (1988). Vulnerability
and episodic cognitions in a self-worth contingency model of depression.
In L. B. Alloy (Ed.), Cognitive processes in depression (pp. 289309).
New York: Guilford Press.
Lakey, B., Tardiff, T. A., & Drew, J. B. (1994). Negative social interac-
tions: Assessment and relations to social support, cognition, and psy-
chological distress. Journal of Social and Clinical Psychology, 13,
4262.
Leary, M. R. (1999a). Making sense of self-esteem. Current Directions in
Psychological Science, 8, 3235.
Leary, M. R. (1999b). The social and psychological importance of self-
esteem. In R. M. Kowalski & M. R. Leary (Eds.), The social psychology
of emotional and behavioral problems: Interfaces of social and clinical
psychology (pp. 197221). Washington, DC: American Psychological
Association.
Locascio, D., & Ley, R. (1972). Scale-rated meaningfulness of 319
CVCVC words and paralogs previously assessed for associative reaction
time. Journal of Verbal Learning and Verbal Behavior, 11, 243250.
MacLachlan, I. M. (1985). Psychometric contamination in correlational
studies of depression of self-esteem. Psychology and Psychiatry, 13,
443444.
MacLeod, C., & Mathews, A. M. (1991). Cognitive-experimental ap-
proaches to the emotional disorders. In P. Martin (Ed.), Handbook of
behavior therapy and psychological science (pp. 116150). New York:
Pergamon Press.
Macmillan, N. A., & Creelman, C. D. (1990). Response bias: Character-
istics of detection theory, threshold theory, and nonparametric in-
dexes. Psychological Bulletin, 107, 401413.
Mandler, G. (1980). Recognizing: The judgment of previous occurrence.
Psychological Review, 87, 252271.
Markus, H. (1977). Self-schemata and processing information about the
self. Journal of Personality and Social Psychology, 35, 6378.
Markus, H., Smith, J., & Moreland, R. L. (1985). The role of the self-
concept in the perception of others. Journal of Personality and Social
Psychology, 49, 14941512.
McGuire, W. J., & McGuire, C.V. (1996). Enhancing self-esteem by
directed-thinking tasks: Cognitive and affective positivity asymmetries.
Journal of Personality and Social Psychology, 70, 11171125.
Metalsky, G. I., Joiner, T. E., Hardin, T. S., & Abramson, L. Y. (1993).
Depressive reactions to failure in a naturalistic setting: A test of the
hopelessness and self-esteem theories of depression. Journal of Abnor-
mal Psychology, 102, 101109.
Miller, P. M., Kreitman, N. B., Ingham, J. G., & Sashidharan, S. P. (1989).
Self-esteem, life stress and psychiatric disorder. Journal of Affective
Disorders, 17, 6575.
Mineka, S., & Nugent, K. (1995). Mood-congruent memory biases in
anxiety and depression. In D. L. Schacter (Ed.), Memory distortions:
How minds, brains, and societies reconstruct the past (pp. 173193).
Cambridge, MA: Harvard University Press.
Moot, S. A., Teevan, R. C., & Greenfield, N. (1988). Fear of failure and the
Zeigarnik effect. Psychological Reports, 63, 459464.
Mullis, R. L., Youngs, G. A., Mullis, A. K., & Rathge, R. W. (1993).
Adolescent stress: Issues of measurement. Adolescence, 28, 267279.
Murdock, B. B., Jr. (1982). Recognition memory. In C. R. Puff (Ed.),
Handbook of research methods in human memory and cognition (pp.
126). New York: Academic Press.
ONeill, E. (1928). Strange interlude. New York: Horace Liveright.
Pearlin, L. I., Menaghan, E. G., Lieberman, M. A., & Mullan, J. T. (1981).
The stress process. Journal of Health and Social Behavior, 22, 337356.
Pietromonaco, P. R., & Markus, H. (1985). The nature of negative thoughts
in depression. Journal of Personality and Social Psychology, 48, 799
807.
Pyszczynski, T., Hamilton, J. C., Herring, F. H., & Greenberg, J. (1989).
Depression, self-focused attention, and the negative memory bias. Jour-
nal of Personality and Social Psychology, 57, 351357.
Raaijmakers, J. G. W., & Shiffrin, R. M. (1981). Search of associative
memory. Psychological Review, 88, 93134.
43
SELF-ESTEEM AND MEMORY
Roberts, J. E., & Monroe, S. M. (1994). A multidimensional model of
self-esteem in depression. Clinical Psychology Review, 14, 161181.
Roberts, J. E., & Monroe, S. M. (1999). Vulnerable self-esteem and social
processes in depression: Toward an interpersonal model of self-esteem
regulation. In T. Joiner & J. C. Coyne (Eds.), The interactional nature of
depression: Advances in interpersonal approaches (pp. 149187).
Washington, DC: American Psychological Association.
Roediger, H. L., McDermott, K. B., & Robinson, K. J. (1998). The role of
associative processes in creating false memories. In M. A. Conway, S. E.
Gathercole, & C. Cornoldi (Eds.), Theories of memory, Vol. 2 (pp.
187245). Hove, United Kingdom: Psychology Press.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton,
NJ: Princeton University Press.
Schlenker, B. R., & Britt, T. W. (1996). Depression and the explanation of
events that happen to self, close others, and strangers. Journal of
Personality and Social Psychology, 71, 180192.
Sedikides, C. (1992). Changes in the valence of the self as a function of
mood. In M. S. Clark (Ed.), Review of personality and social psychol-
ogy: Vol. 14. Emotion and social behavior (pp. 271311). Thousand
Oaks, CA: Sage.
Segal, Z. V. (1988). Appraisal of the self-schema construct in cognitive
models of depression. Psychological Bulletin, 103, 147162.
Segal, Z. V., Gemar, M., Truchon, C., Guirguis, M., & Horowitz, L. M.
(1995). A priming methodology for studying self-representation in major
depressive disorder. Journal of Abnormal Psychology, 104, 205213.
Sherer, M., Maddux, J. E., Mercandante, B., Prentice-Dunn, S., Jacobs, B.,
& Rogers, R. W. (1982). The Self-Efficacy Scale: Construction and
validation. Psychological Reports, 51, 663671.
Silverberg, W. V. (1952). Childhood experience and personal destiny. New
York: Springer.
Singer, J. A., & Salovey, P. (1988). Mood and memory: Evaluating the
network theory of affect. Clinical Psychology Review, 8, 211251.
Singer, J. A., & Salovey, P. (1996). Motivated memory: Self-defining
memories, goals, and affect regulation. In L. L. Martin & A Tesser
(Eds.), Striving and feeling: Interactions among goals, affect, and self-
regulation (pp. 229250). Mahwah, NJ: Erlbaum.
Story, A. L. (1998). Self-esteem and memory for favorable and unfavor-
able personality feedback. Personality and Social Psychology Bulle-
tin, 24, 5164.
Symons, C. S., & Johnson, B. T. (1997). The self-reference effect in
memory: A meta-analysis. Psychological Bulletin, 121, 371394.
Tafarodi, R. W. (1998). Paradoxical self-esteem and selectivity in the
processing of social information. Journal of Personality and Social
Psychology, 74, 11811196.
Tafarodi, R. W., Lang, J. M., & Smith, A. J. (1999). Self-esteem and the
cultural trade-off: Evidence for the role of individualism-collectivism.
Journal of Cross-Cultural Psychology, 30, 638658.
Tafarodi, R. W., & Milne, A. B. (2002). Decomposing global self-esteem.
Journal of Personality, 70, 443483.
Tafarodi, R. W., & Swann, W. B., Jr. (1995). Self-liking and self-
competence as dimensions of global self-esteem: Initial validation of a
measure. Journal of Personality Assessment, 65, 322342.
Tafarodi, R. W., & Swann, W. B., Jr. (1996). Individualism-collectivism
and global self-esteem: Evidence for a cultural trade-off. Journal of
Cross-Cultural Psychology, 27, 651672.
Tafarodi, R. W., & Swann, W. B., Jr. (2001). Two-dimensional self-
esteem: Theory and measurement. Personality and Individual Differ-
ences, 31, 653673.
Tafarodi, R. W., Tam, J., & Milne, A. B. (2001). Selective memory and the
persistence of paradoxical self-esteem. Personality and Social Psychol-
ogy Bulletin, 27, 11791189.
Tafarodi, R. W., & Walters, P. (1999). Individualism-collectivism, life
events, and self-esteem: A test of two trade-offs. European Journal of
Social Psychology, 29, 797814.
Teasdale, J. D., Taylor, M. J., Cooper, Z., Hayhurst, H., & Paykel, E. S.
(1995). Depressive thinking: Shifts in construct accessibility or in sche-
matic mental models? Journal of Abnormal Psychology, 104, 500507.
Tipton, R. M., & Worthington, E. L. (1984). The measurement of gener-
alized self-efficacy: A study of construct validity. Journal of Personality
Assessment, 48, 545548.
Tulving, E. (1985). Memory and consciousness. Canadian Journal of
Psychology, 26, 112.
Tulving, E., & Pearlstone, Z. (1966). Availability versus accessibility of
information in memory for words. Journal of Verbal Learning and
Verbal Behavior, 5, 381391.
Tzelgov, J., & Henik, A. (1991). Suppression situation in psychological
research: Definitions, implications, and applications. Psychological Bul-
letin, 109, 524536.
Watson, D., & Clark, L. A. (1994). Manual for the Positive and Negative
Affect ScheduleExpanded Form. Iowa City: The University of Iowa
Press.
White, R. W. (1963). Ego and reality in psychoanalytic theory: A proposal
regarding independent ego energies. Psychological Issues, 3(3, Mono-
graph 11).
Woodruff, S. L., & Cashman, J. F. (1993). Task, domain, and general
self-efficacy: A re-examination of the self-efficacy scale. Psychological
Reports, 72, 423432.
Zuroff, D. C., Colussy, S. A., & Wielgus, M. S. (1986). Additional
comments on depression, selective memory, and response bias. Cogni-
tive Therapy and Research, 10, 271273.
44
TAFARODI, MARSHALL, AND MILNE
Received November 24, 2001
Revision received July 16, 2002
Accepted July 16, 2002
Appendix
Memory Set Used in Studies 1 and 2
High Competence (C) Low Social Worth (W)
competent ugly
effective rejected
strong isolated
successful neglected
capable inferior
powerful ignored
energetic despised
talented criticized
Low Competence (C) Neutral (N)
weak sentimental
afraid talkative
dependent thrifty
passive religious
failure curious
helpless excited
lacking systematic
defeated serious
High Social Worth (W) definite
likable deliberate
popular meticulous
attractive meditative
appreciated quick
worthy idealistic
accepted precise
interesting subtle
important
45
SELF-ESTEEM AND MEMORY
... The representation, recall, and maintenance of autobiographical memories have shown to be disturbed by individuals with poor psychological well-being, where they are systematically biased for negative information [11]. Memory bias and negative interpretations of events contribute to a vicious cycle, where one's psychological health contributes to the tendency toward negative interpretation of memories, and this bias also maintains low selfesteem, depression, or anxiety [19][20][21]. In fact, there is a consensus that self-esteem, anxiety, and depression are core elements of psychological health [22][23][24]. ...
... Previous studies have shown that the midline cortical structures, such as the medial prefrontal cortex (PFC), ACC, posterior cingulate cortex (PCC), and precuneus that mediate self-referential processing are also closely involved in the expression of self-esteem [29]. Global self-esteem, which includes self-competence and self-liking, is related to selective memory, especially toward a negatively salient stimulus [19]. Individuals with lower self-esteem tend to be more concerned with how others perceive them, and this makes heightened negative events more memorable. ...
... Moreover, the results also revealed that one's level of self-esteem positively influences the perception of one's father, particularly in positive situations. This finding further highlighted the influence of self-esteem on how positively one perceives his or her father for positive memories, rather than rumination on exaggerated negative affects seen in previous patient studies [19,41]. ...
Article
Full-text available
The recollection of childhood memories is affected by the subjects involved, such as father and mother, and by the context. This study aimed to clarify the neural influence of autobiographical memory related to the parent-child relationship on psychological health in adulthood. Twenty-nine healthy volunteers participated in a functional magnetic resonance imaging experiment using a childhood memory recollection task, in which they appraised the emotion a parent would have provided in a given situation. Whole-brain univariate and psychophysiological interaction analyses were performed. Neuroimaging results indicated notable involvement of the caudal anterior cingulate cortex and precuneus in autobiographical memory related to the parent-child relationship, and their activities were closely associated with the level of depression and self-esteem, respectively. The functional connectivity results indicated increased connectivity between the caudal anterior cingulate cortex and fusiform gyrus for the father-positive condition compared to the mother-positive condition and there was a positive correlation between the strength of connectivity between the two regions and the anxiety level. Our findings suggest the processing of negative affect and the personalness of autobiographical memories are distinctly engaged depending on the parent in question and the situational valence. The present study illuminates the impact of autobiographical memory processes on various dimensions of psychological health.
... Secondly, agreeableness may negatively influence the occurrence and development of the main cognitive components of anger rumination: anger memories, anger afterthought, understanding of causes, thoughts of revenge. Agreeableness is a kind of prosocial trait (Costa et al. 1991), according to trait-congruency perspective (Reidy and Richards 1997;Tafarodi et al. 2003). People with high levels of agreeableness attend to take in, remember and recall positive interpersonal information, as opposed to information regarding anger experiences. ...
... This result is consistent with existing research, which has indicated that agreeableness negatively predicts another type of rumination (hostile rumination) over time (Caprara et al. 2013), and suggesting that agreeableness may have a negative effect on ruminating on negative experiences (e.g., anger incidents) regarding other people. The underlying mental mechanism of the positive social function of agreeableness may be due to the positive cognitive preference toward others and interpersonal communication (e.g., Meier et al. 2006), as suggested by the trait-congruency theory (Reidy and Richards 1997;Tafarodi et al. 2003). The present study appears to provide new support for the perspective that personality affects people's cognition (e.g., Kokkinos et al. 2017). ...
Article
Full-text available
Aggression is a type of negative social behavior. Agreeableness and anger-related cognition are thought to be important factors that affect aggression. The longitudinal relations among agreeableness, anger-related cognition and aggression, and the affective cognitive path underlying the relationship between agreeableness and aggression are not clear, however. In this study, 942 college students were investigated twice at an interval of six months, using the Buss-Perry Aggression Questionnaire, Anger Rumination Scale, and agreeableness subscale of the NEO Five-Factor Inventory. The results indicate that: agreeableness negatively predicts anger rumination and aggression after six months; anger rumination positively predicts aggression over time; and anger rumination mediates the longitudinal association between agreeableness and aggression. These results suggest that the prosocial personality may withstand aggression through resisting anger-related cognition. This study deepens our understanding of the relationships between personality and aggression, allowing a development of the General Aggression Model, in terms of recognizing the cognitive pathway for personality to influence aggression, and provides theoretical guidance on reducing the generation of aggression in daily life.
... Although the pairwise relationships between attachment, selfesteem, or rumination and depression have been investigated, there has been no study on the relationship of these four variables within the same model. In particular, considering the close relationship between self-esteem and rumination, those with low explicit self-esteem tend to focus more on their defects and shortcomings and more sufficiently encode negative information, forming more negative emotions manifested with more rumination (Tafarodi et al., 2003). A longitudinal study has shown that rumination can mediate the effect of self-esteem on depression (Kuster et al., 2012), self-esteem and rumination may have a chain mediation effect between attachment style and depression. ...
Article
Full-text available
In this study, we discuss the effects of attachment on depression and the mediating roles of self-esteem and rumination in Chinese seniors. We assessed 431 using the Experiences in Close Relationships Inventory, the Rosenberg Self-Esteem Scale, the Ruminative Responses Scale, and the Short Form of Center for Epidemiologic Studies Depression Scale. Both attachment anxiety and attachment avoidance significantly predicted depression in seniors. Attachment anxiety and attachment avoidance positively predicted rumination but negatively predicted self-esteem. Structural equation models showed that rumination and self-esteem fully mediated the effects of attachment anxiety and attachment avoidance on depression. The attachment of seniors is related to depression, and self-esteem and rumination have chain mediating effects between attachment and depression.
... It has also been shown to reduce problem solving effectiveness (Lyubomirsky & Nolen-Hoeksema, 1995) and to decrease self-esteem (Roberts et al., 2016). Given that individuals with lower self-esteem also tend to be more attentive to negative information and ruminate more (Tafarodi et al., 2003), stopping negative thoughts could potentially be a mechanism to prevent a vicious cycle of negative thinking, lower well-being, and a negative self-concept. ...
Article
Full-text available
Dealing with unwanted thoughts is a recurrent phenomenon in everyday life. The present study focuses on intrusive thoughts in the work context and examines the protective function of thought control for self‐esteem. Possible mediators (negative affect, task focus) and individual differences in the ability to control unwanted thoughts are also considered. We assessed 143 employees' individual ability to suppress thoughts using the think/no‐think paradigm, followed by a five‐day experience sampling study in the work context. Multilevel analyses showed that individuals with lower suppression abilities experienced higher negative affect and lower self‐esteem when they tried to suppress intrusive thoughts, whereas individuals with higher suppression abilities did not. The findings reveal the protective nature of thought suppression abilities, but also highlight possible detrimental aspects of unsuccessfully engaging in thought control. The results provide a basis for recommendations to individuals on dealing with intrusive thoughts. This article is protected by copyright. All rights reserved.
... It is possible that self-liking and self-competence may have meaningfully unique characteristics, but just been underestimated in the current structure of the SLCS-R. For example, previous research showed that this two dimensions have showed divergent patterns of unique relations to memory (Tafarodi, Marshall, Milne, & A. B., 2003) and word recognition (Tafarodi & Milne, 2002). Therefore, further research should explore the dimensionality of the SLCS-R in other cultural context or age group by more advanced statistical methods and modeling methods to clarify whether these two specific factors were theoretically distinct constructs from global selfesteem. ...
Article
Full-text available
Although more and more studies concern about the development of self-esteem in preadolescents, there is a lack of systematic verification of tools for measuring self-esteem for preadolescents. To address that, the present study verified the measurement model, reliability, and dimensionality of the Self-liking/Self-competence Scale-Revised (SLCS-R; Tafarodi & Swann, 2001), a widely used scale to measure self-esteem, in a sample of 616 Chinese preadolescents aged 9–13 years old. The results showed that the bifactor model with a general factor and two specific factors was the best fit model to the data. Multiple index of the bifactor model showed that the common variance of the SLCS-R is mostly account to the general factor, and only the general factor demonstrated satisfactory reliability. These results suggested that the SLCS-R was essentially unidimensional, despite some multidimensionality for Chinese preadolescents.
... Self-esteem-broadly defined as how much individuals value and like themselves-may be a personality factor related to reduced suicide risk (Moran, Macrae, Heatherton, Wyland, & Kelley, 2006;Overholser, Adams, Lehnert, & Brinkman, 1995;Tafarodi, Marshall, & Milne, 2003). Self-esteem is a key component of psychological health and can serve as a protective factor against psychopathology over time (Brown, 2010;Du et al., 2013;Henriksen, Ranøyen, Indredavik, & Stenseng, 2017). ...
Article
Background Childhood‐onset bipolar disorder (BD) has considerable morbidity and mortality, including suicide. Many risk factors have been identified for suicidality, but the potential role of personality traits as assessed by a computer‐assisted self‐report measure remains unclear. Aims To address this gap in knowledge, we tested relations between pathological‐range personality traits and suicidal ideation among young adults whose childhood‐onset BD was prospectively confirmed by enrollment in the Course and Outcome of Bipolar Youth study (COBY) as children (n = 45) and a newly enrolled group of typically developing controls (TDCs; n = 52) both cross‐sectionally and longitudinally after 1.5 years of follow up. Materials & Methods Personality traits were assessed with the computerized Schedule for Nonadaptive and Adaptive Personality‐2 (SNAP‐2). Results Cross‐sectionally, we found that participants with BD had elevated Suicide Proneness and Low Self‐esteem versus TDCs at baseline. Furthermore, longitudinal analyses in the BD participants for whom we had 1.5 years of prospectively collected illness‐course data showed that greater Suicide Proneness and Low Self‐esteem prospectively predicted greater levels, shorter time until occurrence, and greater frequency of suicidal ideation during the follow‐up. Conclusion Our findings suggest the role of specific personality‐related vulnerabilities in the course of BD that, pending replication, could contribute to development of interventions focused on personality traits among individuals with BD.
... Past research has considered the relationship between explicit SE and memory bias (Tafarodi, Marshall, & Milne, 2003). They found that low SE individuals selectively remember schema-congruent information (e.g., information related to failure or inefficacy). ...
Article
In this study, we examined attention and memory biases for aggressive information in two groups of college students. Individuals with fragile high self‐esteem (n = 30) and individuals with secure high self‐esteem (n = 30) first performed a dot‐probe task investigating attention bias, followed by a memory task. Incidental free recall of words presented in the memory task was then completed to assess memory bias. Results revealed that individuals with fragile high self‐esteem exhibited significant attention and memory biases for aggressive words compared with secure high self‐esteem individuals. Attention bias for aggressive words was positively correlated with memory bias in individuals with fragile high self‐esteem, but no correlation was found for individuals with secure high self‐esteem. These findings suggest that individuals with fragile high self‐esteem selectively attend to and remember aggression‐related information. They may process information in ways that are congruent with an aggression‐related schema. This study reveals the aggressive cognitive processes of individuals with fragile high self‐esteem, which may be related to aggression.
Article
Full-text available
Abstract The purpose of this study was to investigate the effectiveness of cognitive-behavioral self-esteem group therapy on assertiveness, Social anxiety, Shyness, and self-efficacy social anxious high school boys in year 2010–2011. The research method was quasi-experiment with two group's pretest-posttest design. The population contains of the third and second grades high school boy students of Ilam city. Social Phobia Inventory (SPIN) was administrated to 240 with of cluster random sampling method, 32 individuals were selected using structured clinical interview for social phobia, based on diagnostic criteria of DSM-IV. The subjects were then randomly assigned into two groups each with 16 members: the control group and the experimental group. The Social Phobia Inventory (SPIN), Revised Cheek And Buss Scale(RCBS), self-efficacy scale, and assertiveness for adolescent scale, were administered at pre test to all participants. The experimental group was trained in self-esteem group training via thirteen group sessions on a weekly. During this period no intervention was given to the control group. Then questionnaires were administered again at post test. The result of multiple covariance analysis indicated that self-esteem group training significantly decreased the amount of social anxiety and Shyness in the experimental group as compared to the control group (P=0.0001).This training caused significant increase in the scores of assertiveness and self-efficacy in the experimental group as compared the control group (P=0.0001). Findings of this study revealed that self-esteem group training can significantly and effectively increase assertiveness and self- efficacy, it also reduced social anxiety and Shyness. Keywords: self-esteem Self-efficacy assertiveness Shyness social anxiety
Article
Memory bias entails preferential recall of a certain kind of information over another. The present research explores the self-verification motive in memory bias to maintain self-esteem. It is hypothesised that people with high self-esteem will remember more positive content and people with low self-esteem will remember more negative content to maintain their self-esteem. Two experiments were conducted to test it. Experiment one (N = 48) uses two groups of high and low self-esteem based on their self-esteem score. In experiment two (N = 44), self-esteem is manipulated in the laboratory by providing positive or negative feedback to participants on a performance task. In both experiments, memory bias is measured by the number of positive or negative adjectives recalled after seeing them on a computer screen. Results of both studies confirm the hypotheses and show that people remember/recall the words that fit into their present self-schema more and forget the words which go against it.
Chapter
In linear regression the mean surface in sample space is a plane. In non-linear regression the mean surface may be an arbitrary curved surface but in other respects the models are similar. In practice the mean surface in most non-linear regression models will be approximately planar in the region(s) of high likelihood allowing good approximations based on linear regression techniques to be used. Non-linear regression models can still present tricky computational and inferential problems. (Indeed, the examples here exceeded the capacity of S-PLUS for Windows 3.1.)
Article
Insofar as people organize information about and evaluations of important topics in connected and coherent systems, attitudes may be changed from within by enhancing the salience of information already present virtually within the person's belief system without communicating new information from outside sources. A cognitive positivity bias is predicted such that stimulus evaluation (e.g., self-esteem) is affected more by characteristics that the stimulus possesses than by ones it lacks. Experiment 1 tested relations between participants' momentary self-esteem and concurrently salient desirable(vs. undesirable) self-characteristics possessed (vs. lacked). Experiments 2 and 3 changed participants' self-esteem by using directed-thinking tasks to manipulate the salience of desirable (vs. undesirable) self-characteristics possessed (and, to a lesser extent, lacked).
Article
Self-efficacy is typically viewed as task-specific. Bandura also discussed the concept at a “domain-linked” level and general level. Sherer and his colleagues developed the Self-efficacy Scale to measure general self-efficacy expectancies in education/vocation and social areas. A reexamination of the Self-efficacy Scale indicated that the scale was more intricate than originally reported. It captured aspects of strength, magnitude, and generality of efficacy. The scale showed appropriate relationships to other personality measures. Criterion validity was established as the scale differentiated performance expectations. Evidence for the concept of domain efficacy was reported. The Self-efficacy Scale was a good measure for domain efficacy in the academic area; the scale has not yet been verified as a general efficacy scale. Research should be directed toward exploring the role of the total concept of efficacy in the cognitive process.