ArticlePDF Available

Abstract and Figures

The objective of this systematic review of studies using self-affirmation manipulations was to identify research gaps and provide information to guide future research. We describe study characteristics, categories of manipulations, and report effects on various dependent variables. Our search strategies yielded 47 eligible articles (69 studies). Manipulations varied by affirmation domain (values or personal characteristics), attainment (participant- or investigator-identified), and procedure (scale, essay, feedback, etc.). Most dependent variables were cognitive. Strong effects of self-affirmation were found for attitudes and persuasion/bias, but future work is needed for variables with mixed results including risk cognitions, intentions, and behavior. Suggestions and considerations for future research involving self-affirmation manipulations are discussed.
No caption available
… 
No caption available
… 
No caption available
… 
Content may be subject to copyright.
Sezf
and
Identity,
5:
289—354,
2006
Psychology
Press
Copyright
©
2006
Psychology
Press
Taylor
&
Francis
Group
ISSN:
1529-8868
print/1529-8876
online
DOT:
lO.1080/15298860600805325
Experimental
Manipulations
of
Self-Affirmation:
A
Systematic
Review
AMY
McQUEEN
University
of Texas-Houston,
School
of
Public
Health,
Houston,
Texas,
USA
WILLIAM
M.
P.
KLEIN
University
of
Pittsburgh, Pittsburgh,
Pennsylvania,
USA
The
objective
of
this
systematic
review
of
studies using
self-affirmation
manipulations
was
to
identify
research
gaps and
provide
information
to
guide
future
research.
We
describe
study
characteristics,
categories
of
manpulations,
and
report
effrcts
on
various
dependent variables.
Our
search
strategies
yielded
47
eligible
articles (69
studies).
Manipulations
varied
by
affirtnation
domain (values or
personal
characteris
tics),
attainment
(participant-
or
investigator-identified),
and
procedure
(scale,
essay,feedback,
etc.). Most
dependent variables
were
cognitive.
Strong
efjcts
of
self-
affirmation
were
foundfor
attitudes
and
persuasion/bias,
but
future
work
is
neededfor
variables
with
mixed results
including
risk
cognitions,
intentions,
and
behavior.
Suggestions
and
considerations
fur
flaure
research
involving
sef
affirmation
manipulations
are
discussed.
Self-enhancement
theorists
(Steele,
1988;
Tesser,
1988)
suggest
that
individuals
have
a
fundamental
need
or
motive
to
maintain
a
positive,
global
self-evaluation.
When
the
self-concept
is
threatened,
people
experience
psychological
discomfort
that
they
are
motivated
to
reduce.
Several
strategies for
reducing
such
discomfort
have
been
proposed.
Tesser’s
(1988)
self-evaluation
maintenance
model posits
that
threatened
individuals
may
minimize
the
importance of
the
threatened
domain
or
distance
themselves
from
a
superior-performing
individual
or
close
other. General
self-
enhancement
involves
seeking
out
or
interpreting
situations
in
order
to
attain
a
positive
self-view.
Self-enhancement
may
be
achieved
from
within
(i.e.,
intraperso
nal)
or
by
making
downward
social
comparisons
(i.e.,
interpersonal)
(Aspinwall
&
Taylor,
1993;
Wood,
Giordano-Beech,
&
Ducharme,
1999).
In
contrast,
Steele’s
(1988)
self-affirmation
theory
outlined
one
method
for
maintaining
or restoring
one’s
global positive
self-image
from
threats,
which
required
affirming
some
important
aspect
of
the
self
that
is
unrelated
to
the
threatened
domain.
Thus,
self-affirmation
can
serve
as
a
buffer
or
coping
resource
when
the
self
is
threatened
(Steele,
Spencer,
&
Lynch,
1993)
and
is
not
simply
an ephemeral
positive
feeling
or
activation
of
the
self-concept
(Steele
&
Liu,
1983).
Received
3
August
2005;
accepted
16
May
2006
Correspondence
should
be
addressed
to:
Amy
McQueen.
University
of
Texas-Houston,
School
of
Public
Health,
Center
for
Health
Promotion
&
Prevention
Research,
7000
Fannin,
Suite
2568,
Houston
Texas
77030,
USA.
E-mail:
Amy.Mcqueen@uth.tmc.edu
http://www.psypress.com/sai
289
290
A.
Mc
Queen
&
W.
M.
P.
Klein
Self-affirmations
are expected
to
reduce
the
psychological
discomfort
associated
with
cognitive
dissonance
(Festinger,
1957),
even
though
they
may
not
resolve
or
dismiss
the
actual
problem
or
cause
of
the
inconsistency.
People
may
not
attempt
to
dismiss
every
threat,
but
instead
accept
or
acknowledge
some,
as
long
as
their
global
self-evaluation
remains
positive. Thus,
the
motivation
for
self-affirmation
may
depend
on
the
importance
or
magnitude
of
the
threat
and
current
evalu-
ations
of
self-worth.
Although
countless
strategies
for
self-affirmation may
exist,
Steele (1988)
posited
that
after
a
threat
to
their
self-image,
people would
affirm
their
self-adequacy
through
whatever
means
were
most
readily available
or
salient.
The
outcome
of
interest
in
early self-affirmation
studies
was
attitude
change
following
a
dissonant
act
(Simon,
Greenberg,
&
Brehm,
1995;
Steele
&
Liu,
1983).
The same
induced-compliance
paradigm that
is
common
in
studies
inducing
cognitive
dissonance
was
used
in
self-affirmation
studies.
Cognitive
dissonance
was
increased
when
participants
were
led
to
believe
that
they
freely
chose to
write
a
counter-attitudinal
essay
on
a
topic
of
personal
importance,
such
as
a
large
tuition
increase
at
their
university.
Self-affirmation
was
manipulated
in
the
laboratory
by
having
participants
complete
a
scale
that
affirmed
a
strongly-held
value. Such
manipulations
were
introduced
after
dissonance
was
aroused,
but
before
attitudes
were
reassessed.
The
degree
of
attitude
change
in
the
direction
of
the
position
taken
in
the
essay
was
a
measure
of
cognitive
dissonance.
The
attenuation
of
such
change,
when
given
the
opportunity
to
affirm
an
important
value
or
characteristic,
represented
the
effect
of
self-affirmation.
Later
empirical work
identified
additional
boundary
conditions
for
the
effect
of
self-affirmation
on
the
reduction
of
dissonance.
Some
dissonant
behaviors
(e.g.,
hypocrisy)
may
only
be
resolved
by
attempting
to
directly
change
the
behavior,
despite
the
attractiveness
and
importance
of
alternative
affirmation
opportunities
(Stone,
Wiegand,
Cooper,
&
Aronson,
1997).
Situations
also may
arise
in
which
one’s
chosen
affirmation
fails.
Failed
self-affirmation
attempts
restore
dissonance
and
produce
attitude
change
(Galinsky,
Stone,
&
Cooper,
2000).
Similarly, evidence
suggests
that
affirming
self-attributes
(i.e.,
compassion)
that
are
related
to
the
self-
threat
(i.e.,
essay
advocating
decreased funding
for
handicapped
students)
may
increase
dissonance
(Blanton, Cooper,
Skurnik,
&
Aronson,
1997;
Stone
&
Cooper,
2003).
Individual
differences
may also
influence
the
effect
of
self-affirmation.
Self-
affirmation may
produce
greater
attitude
change
(Steele
et
al.,
1993,
study
2;
Stone
&
Cooper,
2003)
and
greater
perceived
stress (Creswell,
Welch,
Taylor, Sherman,
Gruenewald,
&
Mann,
2005)
among
individuals
with
low
self-esteem
compared
to
those with high
self-esteem.
However,
other
evidence
suggests
that
the
effects
of
self-
affirmation
manipulations
are similar
for
individuals
with
high
and
low
self-esteem
(Spencer,
Fein,
&
Lomore,
2001,
study
2).
Recent
work
suggests
that
self-affirmation
of
important
values
may
decrease
biased
information
processing
and
influence
behavioral intentions
to
reduce
risks
when
presented
with
negative
(i.e.,
self-threatening)
health
information
(Sherman,
Nelson,
&
Steele,
2000).
Such
positive
findings
may
promote
the
adoption
of
similar
self-affirmation
manipulations
by
a
diverse
audience
of
researchers
attempting
to
influence
behavior
change
processes
through
large-scale
interventions.
However,
there
are
multiple
manipulations
in
use,
no
indication
of
which
technique
is
“best”,
and
numerous
outcome
variables being
investigated.
Consequently,
a
review
of
studies
involving
self-affirmation
manipulations
seems
necessary.
Others
have
reviewed
self-affirmation
theory (Aronson,
Cohen,
&
Nail,
1999;
Sherman
&
Cohen,
Self-Affirmation
Manipulations
291
2002,
2006;
Steele,
1988);
however,
our
goal was
to
review
the
manipulations
of
self-
affirmation
that
are
employed
to
test
the
theory.
Of
note,
most
of
the
research
examining self-affirmation
has been
conducted
to
support
or
refine the
theory;
however,
no
published
study
has
closely
examined
the
specific
methods
by
which
self-affirmation
is
manipulated
in
the
laboratory.
Using
the
methods
for
a
systematic
review
of
the
literature
(The
Campbell
Collaboration,
2001;
Weed,
1997),
we
specified
our
goals,
search
methods, inclusion/exclusion
criteria,
and
methods
for
summarizing
data
from
primary
studies
in
order
for
others
to
be
able
to
critically
evaluate
and
be
able
to
replicate
the
work.
The
goal
of
this
review
was to:
(I)
describe
the
studies
that
use
self-affirmation
manipulations;
(2)
categorize
and
describe
self-affirmation
manipulations
being
used
in
the
literature;
and
(3)
examine
the
effects
of
self-affirmation
manipulations
on
dependent
measures
using
standard
effect
size
estimates. An
expected
contribution
of
this
review
is
to
provide investigators interested
in
using
a
self-affirmation
manipulation
ample
references
and considerations for conducting
their
experiments,
and
to
contribute
to
the refinement,
measurement,
and
application
of
self-affirmation
theory
by
describing
gaps
in
the
experimental
research.
Method
Inclusion
Criteria
and
Search
Strategy
Although
many
strategies
or
techniques
could
be
self-affirming,
we
sought
to
focus
our
review
on studies
employing
experimental
manipulations
designed
specifically
to
invoke
self-affirmation.
In
addition,
eligible
studies
had
to
examine
the
effect
of
self-
affirmation
on
dependent
variables
such
as
attitudes,
mood,
and
behavior. Following
Steele
and
Liu’s
(1983)
suggestions,
we
defined
self-affirmation
as
any
affirmation
of
some
important
aspect
of
the
self
(i.e.,
personal
values,
characteristics,
or
positive
qualities).
We
excluded studies
that:
(1)
did
not
include
an equivalent
comparison
condition
not
exposed
to
the
self-affirmation
manipulation
(Boney-McCoy,
Gibbons,
&
Gerrard,
1999;
Kaplan
&
Krueger,
1999;
McGregor,
Zanna,
Holmes,
&
Spencer,
2001);
(2)
did
not
involve
a
self-threat,
but
used
strategies
believed
to
induce
positive
self-evaluations
(Harber,
2005;
McGuire
&
McGuire,
1996);
(3)
used
an experimental
manipulation
eliciting
self-enhancement
strategies
other
than
self-
affirmation
(such
as
derogation
of
others;
Beauregard
&
Dunning,
1998);
or
(4)
used
measures
to
assess
self-affirmation
as
a
dependent
variable—either
survey-based
(Lockwood,
Dolderman,
Sadler,
&
Gerchak,
2004;
Murray,
Bellavia,
Feeney,
Holmes,
&
Rose,
2001;
Murray,
Holmes,
MacDondald,
&
Ellsworth,
1998)
or
behavioral
(Stone
et
al.,
1997).
Behaviors
such
as
donating
to
charity,
volunteering
one’s
time,
or
helping someone
could
be
self-affirming.
However,
to
be
included
in
our
review
as
a
self-affirmation
manipulation,
the
presence
or absence
of
the
behavior
would
have
to
be
associated
with
some
dependent
variable.
Although
we
included
several
studies
identified
in
our literature
search
that
used
the
completion
of
a
self-esteem
scale
as
a
self-affirmation
manipulation
(Kimble,
Kimble,
&
Croy,
1998), we
did
not
perform
additional
searches
for
all
studies
using
self-esteem
scales,
because
we
felt
the
vast
self-esteem
literature
goes
beyond
the
scope
of
this
review
(and,
as
noted
earlier,
many
such
studies
do
not
involve
an
explicit
self-threat).
Similarly,
we
included
studies
identified
in
our literature
search
that
used
positive
feedback
as a
self-affirmation
manipulation, but
did
not
conduct
an
additional
292
A.
Mc
Queen
&
W.
M.
P.
Klein
search
of
all
studies employing
bogus
feedback.
Such studies
go
well
beyond the
scope
of
our
review
because
they
were
not
specifically
designed
to
test
self-
affirmation
theory
and
many
such studies do
not
employ
a
self-threat
unrelated
to
the ostensibly
affirming
positive
feedback.
We
did
not
restrict
eligible
studies
to
a
particular
age
group,
study
setting,
population,
outcome variable,
or
year
of
publication.
However,
for
practical
purposes,
the search
was
limited
to
English
language published
articles
in
peer-reviewed
journals
involving
human
participants.
The
search
for
published
articles
began
with
a
keyword
search
using
the
terms
“self-affirm$
or
self
affirm$”
in
the
PsycINFO
database
(1967-current)
on
10
April
2005.
The
keyword
“affirmation”
did
not
yield
any
unique,
relevant
articles.
The
authors
independently
reviewed
the
database
search results
and
agreed
on
study
eligibility.
To
find
eligible
articles
in
addition
to
those
identified
through
PsycINFO,
the
first
author
scanned
the reference
lists
from
eligible
articles,
as
well
as
tables
of
contents
for
the
two
most
recently
published
issues
and any
in-
press articles
available
online
from
the
journals
from
which
we
already
identified
at
least
two eligible articles:
Journal
of
Experimental
Social
Psychology,
Journal
of
Personality
and
Social
Psychology,
Motivation
and
Emotion,
and
Personality
and
Social
Psychology
Bulletin.
A
direct
request
to
authors
for
recent
or
in press
studies
involving
a
self-affirmation
manipulation
was
posted
to
the
Society
for
Personality
and
Social
Psychology
(SPSP)
electronic
Internet
listserv,
2
February
2005.
A
final
search
was
conducted
on
6
March
2006
using
PsycINFO
and
a
hand
search
of
the
journals
listed.
Data
Extraction
Procedure
and
Effect
Size
Calculation
The
following
information
was
extracted from
all
eligible
studies
within
each
relevant
article:
(a)
the
study
paradigm
or
procedure;
(b)
study
setting,
population,
and
any
selection
criteria;
(c)
the
threat
to
the
self;
(d)
specifics
of
the
self-affirmation
manipulation;
(e)
the
measures
used
as
a
manipulation
check
for
the
self-affirmation
manipulation;
(1)
the
type
of
control
or
comparison
condition;
(g)
the
primary
independent
and
dependent
variables:
(h)
the
results
relevant
to
the
self-affirmation
manipulation; and
(i)
data
needed
to
calculate
effect
sizes,
including
exact
cell
sizes
of
the
number
of participants
included
in
analyses.
All
data
extraction
and
effect-size
calculations
were
performed
by
the
first
author.
The
second
author
recalculated
10%
of
the
effect
sizes
and
found
no errors.
We
concluded
that
no
systematic errors
were
made.
We
attempted
to
summarize
all
analytical
findings
relevant
for
our
review
regardless
of
statistical
significance,
as
reflected
most
often
by
p-values.
Given
that
p-values
are
influenced
by
both
the
actual
effect
size
and
the
sample
size,
the
effect
size
is
needed
to
interpret
whether
the
results
from
significance
tests are
meaningful.
Additionally,
because
effect
size
measures
are
based on
standard
metrics,
they
are
useful
for
making
comparisons
across studies.
Because
most
analyses examined
mean
differences
across experimental
conditions,
we
calculated
a
standardized
effect
size
statistic
(Cohen’s
a1
using
an
Excels
program
(Wilson,
1996)
and
other formulas
as
needed
(Lipsey
&
Wilson,
2001;
Rosenthal,
1994)
to
divide
the
group
mean
difference
by
the
pooled
standard
deviation.
Whenever
possible,
interactions
were
broken
down into
specific
contrast
effects
as
recommended
by
Rosnow
and
Rosenthal
(1996,
2003).
Cohen’s
dis
a
directional
coefficient
and
ranges
from
to
+.
As
a
rule
of
thumb,
these
effect
sizes
can
be
interpreted
as
small
(d=
0.20),
medium
(d=0.50)
and
large
(d=0.80;
Cohen,
1988).
Effect
sizes
can
be
combined
SeiiAffirmation
Manipulations
293
using
meta-analytic
techniques
when
quantitative
results
are
conceptually
and
statistically
comparable
(Lipsey
&
Wilson,
2001).
Results
Search
Results
In ‘total,
the
keyword
search
of
the
PsycINFO
database
produced
275
hits,
which
were
reduced
to
238
by
limiting
inclusion
to
publications
in
English
with
human
participants.
The
number
of
publications
was
reduced
after
we
limited
inclusion
to
peer
reviewed
journals
(n
122)
and
deleted
duplicate
references
(ii
=
120).
After
careful
evaluation
of abstracts
and
full
text
when
further
details
were
needed,
we
found
that
92
articles did
not
report
results
from
a
study
employing
a
self-affirmation
manipulation
and
were
excluded.
Twenty-eight
articles
were eligible
for
review.
Additional
eligible
articles
were
found
through
hand
searches
of
recent
or
in-press
articles
available
online
(n
=
1),
direct solicitation
using
the
SPSP
listserv
(n
=
7),
and
from
references
cited
in
selected
articles
(n
5).
In
the
final
search on
6
March
2006,
additional
eligible
articles
were
obtained
from
PsycINFO
(n
=4)
and
a
hand
search
(n 2).
Therefore,
47
eligible
articles
were
found
in
total;
32
from
PsycINFO
and
15
through
alternative
search
methods,
which
contained
a
total
of
69
eligible
studies.
Systematic
Review
Results
Data
extracted
from
eligible
studies
are
summarized
in
text
and
table
form
and
include
descriptions
of
the
studies
using
self-affirmation
manipulations,
categoriza
tions
of
the types
of
manipulations
used,
and
their
effects
on
various
dependent
variables.
Additionally,
we
discuss
the
role
of
constructs
that
emerged
in
our
review
of
the
literature
as
important
variables
for consideration.
Study
Characteristics
Appendix
Table
1
summarizes
the
final
sample
size,
any
specific
selection
criteria
used,
the
study
location,
the
threat
to
self,
and
whether
a
self-affirmation
manipulation
check
was
used
in
each
study.
The
studies
are
presented
in
the
same
order
as
Appendix
Table
2
for
ease
of
comparison
across
tables.
Samples.
With
the
exception
of
two
studies,
the
majority
of
participants
exposed
to
eligible
self-affirmation
manipulations
were
undergraduate
college
and
university
students.
Additionally,
the
majority
of
the
studies
(51
of
69)
were
conducted
in
the
United
States.
Although
the
study
populations
appear
similar,
many
studies
involved selection
criteria
that
restricted
their
sample based
on
attitudes,
self-esteem,
sex,
race/ethnicity,
and other
characteristics
relevant
to
the
study
topic
(see
Appendix
Table
1).
Most
studies
failed to
report
the
exact
cell sizes,
but
the
17
studies
that
did
so
are
noted
(*)
in
Appendix
Table
1.
Many
studies
indicated
that
“random
assignment”
was
used;
therefore,
for
all
studies
that
did
not
specify
exact
cell
sizes,
equal
cell
sizes
were
assumed and
fractions
were
used
in
22
cases
as
noted
(#)
in
Appendix Table
1.
We
report
these
details because
cell
sizes
were
often needed
to
calculate the
reported
effect
sizes
in
Appendix
Table
2.
294
A.
Mc
Queen
&
W.
M. P.
Klein
Threat.
The
forced-choice
or
induced-compliance
paradigm
common
in
studies
of
cognitive
dissonance
was
used
in
15
studies.
Of
these
15,
11
presented
the
cognitive
dissonance
and
self-affirmation
manipulations
as
independent
studies,
sometimes
employing
the
use
of separate
experimenters
and
separate
rooms.
The
variety
of
self-threats,
other
than
dissonance,
that
were
employed
made
categorizing
across
studies
difficult
(see
Appendix Table
1).
While some
threats
were
more
overt
such
as
negative
feedback
on
an
important
attribute
and
losing
an
athletic
match,
others
were
more
implicit to the
task
such
as
evaluating
outgroup
members,
assessing
job
security
after
recent
layoffs,
evaluating
risk
information, and
responding
to
laboratory
stress-inducing
tasks.
Additionally,
the
order
of
manipulating
self-affirmation varied
(Appendix
Table
1).
Most
studies
manipulated
self-affirmation
before
(n
=
27)
or
after
(n
=
38)
the self
had
been
threatened;
however,
three
studies
introduced
the
self-threat
then
manipulated
self-affirmation
prior
to
completing
the
self-threat.
One
study
presented
a
self-affirmation
prior
to
the
threat
in
one
condition,
and
in
conjunction
with
the
threat
in
another
condition.
Control
conditions.
Various
control
conditions
were
used.
Most
studies
asked
non-affirmed
participants
to
complete
a
value
scale
of
low
importance
or write
an
essay
about
why
a
value
of
low
importance
to
them
might
be
important
to
someone
else
or
to
another
student
at
their
university
(see
Appendix
Table
1).
Alternatively,
Cohen,
Aronson,
and
Steele
(2000)
suggested
that participants
may
use
any
self-
reflective
writing
opportunity
to
self-affirm
(p.
1154);
therefore,
they
instructed
participants
in
the
non-affirmation
condition
to
write
everything
that
they
had
eaten
or
drunk
in
the
past
48
hours
(see
also
Harvey
&
Oswald,
2000,
study
2).
Several
studies
used
similar
non-self-focused tasks,
whereas
other
studies
did
not
employ
filler
tasks
for
non-affirmed
participants
(see
Appendix
Table
1).
Manipulation
checks.
Thirteen
of
the
20
studies
employing
the
Allport-Vernon
values
scales
(discussed
below)
examined
whether
participants
with
a
greater
orientation
to
the
specific
value
addressed
in
the
scale
endorsed
more
items
than
participants
with
lower
orientations
(Appendix
Table
1).
Endorsing
more
items
was
described
as
an
indication
of
appropriate
selection
of
high-
and
low-oriented
participants,
as well as
greater
self-affirmation.
However,
in
two
studies
reporting
significant
effects
of
self-affirmation
manipulations,
one
found
no
differences
between
groups
(Liu
&
Steele,
1986,
studies
I
&
2),
and
one
excluded
from
analyses
the
five
participants
who
did
not
endorse
more
value
items
(Creswell
et
al.,
2005).
Few
studies
that
did
not
use
the
value
scales
included
manipulation
checks.
Reed
and
Aspinwall
(1998)
cited
Steele
et
al.’s
(1993)
use
of
a
self-esteem
inventory
as
self-
affirming
to
individuals
with
high
self-esteem
as
a
reason
for
their
inability
to
perform
a
traditional
manipulation
check
without
risking
contamination
among
control
participants
(p.
105;
see
also
Blanton,
Pelham,
DeHart,
&
Carvallo,
2001,
p.
376).
Other
manipulation
checks
varied
by
study.
Two
studies used
survey
items
to
assess
whether
participants
believed
the
feedback
(Cohen
et
al.,
2000;
Heine
&
Lehman,
1997).
Two
studies
assessed
post-manipulation
self-evaluations
with
a
single
question,
“In
general,
how
do
you
feel
about
yourself?”
(Cohen
et
al.,
2000,
study
3;
Sherman
et
al.,
2000,
study
1).
Three
studies
used
judges’
ratings
of
participant’s
essays
(Dillard,
McCaul,
&
Magnan,
2005;
Harris
&
Napper,
2005;
Schimel,
Arndt,
Banko,
&
Cook,
2004).
One
study
asked
participants.
“How
meaningful
did
you
find
the
essay?”
to
assess
participants’
reactions
to their
assigned
Sef-Affirmarion
Manipulations
295
study
condition
(Siegel,
Scillitoe,
&
Parks-Yancy,
2005),
and
another
study asked,
“How
important
was
it
for
you
to
do
well?”
(Kimble
et
al.,
1998).
Categorization
of
Se(f-Affirination
Manipulations
Various
methods
were
used
to
manipulate
self-affirmation,
but
most
consistently
varied
by
affirmation
domain,
attainment,
and
procedure.
The
affirmation
domain
usually
focused
on a
specific
value or
personal
characteristic.
Attainment
of
a
value
or
positive
characteristic
was
either
provided
(usually
in
a
short’
list)
to
the
participant
by
the
study
investigator
or
participants
were
asked
to
identify
a
positive
personal
characteristic
or
experience
on
their
own.
The procedure
or
method
of
manipulating
self-affirmation
also
varied;
some
participants
were
asked
to
respond
to
specific
scales
or
a
short
list
of
questions,
some were
asked
to
write
a
list
or
an
essay,
and
others
were
asked
to
use
imagery
techniques
or think
about
their positive
qualities.
Other
methods
used
to
manipulate
self-affirmation included
the
provision
of
self-esteem
scales,
positive
feedback,
indirect
methods,
and
an
expected
opportunity
to
perform
a
positive
behavior.
The
value
scale
self-affirmation
manipulation
used
in
Steele
and
Liu’s
(1983)
first
study
was
the
manipulation
type
used
most
often
in
subsequent
studies
(21
of
69
studies).
All
value
scales
employed
some
version
of
the
list
of
values
from
Allport,
Vernon,
and
Lindzey
(1960)
including:
theoretical,
economic, aesthetic,
social,
political,
and
religious.
What
defines this
type
of
self-affirmation
manipulation
is
the
list
of
values
provided
to
participants
from
which
they
chose
their
most
important.
However,
the way
in
which
the
value
scales
were
administered
varied
across
studies.
Steele
and
Liu
(1983;
Liu
&
Steele,
1986)
screened
college
students
using
Schorr’s
80-item
Test
of
Value
Activities,
which
measured
the
six
general value
orientations
contained
in
the
Ailport—Vernon
Study
of
Values
Scale.
Participants
were
specifically
selected
based
on
their
high
(top
third)
or
low
(bottom
third)
scores
on
Schorr’s
economic
and
political
subscale.
Steele
and
Liu
(1983)
reported
that
participants
who
completed
the Ailport—Vernon
economic—political
subscale
after
a
dissonance
manipulation
were self-affirmed
only
when
they
had
high
economic
and
political
values.
Lower
attitude-change
scores
among
self-affirmed
compared
with
non-affirmed
participants
was
reported
as
evidence
of
a successful
self-affirmation
manipulation.
Liu
and
Steele
(1986)
noted
that
identifying
important
aspects
of
participants’
self-concepts
ahead
of
time
allowed
for
more
personalized
manipula
tions
of
the
self-affirmation
process.
In
later
studies,
participants
were
not
selected
based
on
specific
orientations.
Instead
of
selecting
participants
who
reported
specific
values,
Tesser
and
Cornell
(1991,
study
1)
were
the
first
to
match
participants’
value
ratings
with
the
appropriate
value
scale
to
be
completed
(i.e.,
affirmation
par
ticipants:
highest
value;
control
participants:
low
value).
Alternative
scales with
individualized
written
responses
were
also
used.
Two
studies
asked
participants
to
choose
a
self-definition
from
a
list
provided,
insert
the
term
into
six
sentence
stems,
and
complete
each
sentence;
sentences
were
varied
according
to
the
experimental
condition
(Schimel
et
al.,
2004,
studies
1
&
2).
This
list
of
“self-definitions”
varied from
the
list
of
general
values because
it
included
12
professions:
engineer,
entrepreneur,
mathematician,
comedian,
doctor,
nurse,
lawyer,
musician,
artist,
athlete,
scientist,
and
student.
In
another
study,
all
parti
cipants
in
the
self-affirmation
condition
were
given
a
single
value
scale
that
asked
participants
if
they
had
ever
performed
10
kindness
behaviors,
and,
if
so,
to
write
a
description
of
the
event
(Reed
&
Aspinwall,
1998).
296
A.
Mc
Queen
&
W.
M.
P.
Klein
A
value
essay
self-affirmation
manipulation
was
used
in
19
of
the
69
studies.
Self-affirmation
participants
were
asked
to
circle
their
most
important
value
from
a
list
provided
(similar
to
those
included
in
Ailport
et
al.,
1960)
and
write
a few
paragraphs
about
why
it
was
important
to them.
Similarly,
another
value essay
technique
required
participants
to
rank
a
list
of
11
values
and
personal
charac
teristics
developed
by
Harber
(1995)
in
order
of
personal
importance.
Then
participants
wrote
an
essay
describing
why
their
highest-ranked
value
was
important
to
them and
a
time when
it,
had
been
particularly
important
(Sherman
et
al.,
2000,
study
2),
how
it
had
made
them
feel
good
about
themselves
(Cohen
et
al.,
2000,
study
1),
or
a
time
in
their
lives
when
it
had
proved
meaningful
(Cohen
et
al.,
2000,
study
3).
Other participants
were
asked
to
write
why
the
selected
value
was
important
to
them
and
how
they
had acted
consistently
with
it
in
the
past and
planned
to
act
consistently
with
it
in
the
future (Shrira
&
Martin,
2005)
or
why
it
was
important
and how
they
used
it
in
their
everyday
life,
if
possible
describing
specific
occasions on
which
the
value
determined
what
they
did
(Harris
&
Napper,
2005).
To
create
an
interdependent
self-affirmation,
Hoshino-Browne, Zanna,
Spencer,
Zanna,
Kitayama,
and
Lackenbauer
(2005)
asked
participants
to,
“Select
the
most
important
value
for
themselves
and
their
families
and
explain
the
reasons
why
they
and
their
families
share
that
particular
value”
(p.
303).
Thirteen
self-affirmation
manipulations
involved
writing
tasks
whereby
partici
pants
were
asked
to
identify
and
write
about
a
positive
characteristic
or
experience.
Example
instructions
included
writing
about
a life
event
that
made
participants
feel
proud
(Klein,
Blier,
&
Janze,
2001,
study
1)
or
four
examples
of
times
when
participants
had
demonstrated
their
most
cherished
personal
characteristic
(Matz
&
Wood,
2005,
study
2).
Another
study
asked
participants
to,
“Write
a
short
description
of
an
area
of
your
life
that
is
both
important
to
you
and
makes
you
feel
proud.
It
can
be
any
aspect
of
your
identity,
a
talent,
a
relationship,
or
a
basic
value”
(Blanton
et
al.,
2001,
p.
375).
Three
studies
asked
participants
to
list
three
positive aspects
of
themselves
(Van
den
Bos,
2001,
study
2)
and
one
asked
them
to
think
about
these
positive
characteristics
for
a
few
minutes
(De
Cremer
&
Sedikides,
2005,
study
6).
Two
studies
asked
participants
to
list
positive
things
about
themselves
for
five
minutes
(Harvey
&
Oswald,
2000,
studies
I
&
2).
Wiesenfeld,
Brockner,
Petzall,
Wolf,
and
Bailey (2001)
asked
affirmation
participants
to
write
an
essay
about
an
incident
at
work
or
an
aspect
of
their
job
that
made
them
feel
good
about
themselves
or
understand
themselves
better.
This
is
one
example
of
an
affirmation
focus
within
the same
domain
as
the
outcome
variables
being
examined.
Two
studies
employed
two
different
self-affirmation
manipulations.
One
study
included
two
self-affirmation
manipulations:
one
to
reflect
intrinsic
values, one
to
reflect
extrinsic
values
(Schimel
et
al., 2004,
study
3).
For
the
intrinsic
affirmation,
participants
were
asked
to,
“Describe
in
detail
two
qualities
that
most
reveal
who
you
are
as
a
person
(e.g.,
values,
hobbies,
or
personality
traits)..
.
and
to
describe
how
each
of
these
qualities
reflects
your
true
self”
(p.
90).
For
the
extrinsic
affirmation,
participants
were
asked to,
“Think
of at
least
two
achievements
that
reveal
how
competent
and
talented
you
are
(e.g.,
good
grades,
winning
an
award,
or
getting
promoted
at
work)
and
to
describe
what you
have
achieved
and
how
each
accomplishment
reflects
your
competence
and
success
as
a
person”
(p.
90).
Another
study
asked
participants
in
one
condition
to,
“Think
of
a
time
you
felt this
way
and
write
it
down,”
for
eight
affirmations
(intelligent, kind,
honest, strong-willed
and persistent,
friendly,
good-hearted
and
caring,
a
good
significant
other,
and
Self-Affirmation
Manipulations
297
compassionate)
prior
to
reading
eight
tobacco-product
warning
messages
(Dillard
et
al., 2005).
In
another
condition,
participants
viewed
one
affirmation
along
with
each
warning
message.
Thirteen
studies used
self-affirmation
manipulations
that
substantively
differed
from
those
previously
discussed
(Appendix
Table
2).
Two studies
had
participants
complete
self-esteem
measures
to
make
self-resources
salient.
Six
studies
affirmed
participants
by
providing
bogus
feedback
from
tests
of
personality
(n
=
4),
social
perceptiveness
(n
=
1)
or
social
skills
and
ambition
(n
=
1).
One
study
affirmed a
valued
skill
by
verbally
saying,
“Good!
You
are
driving
well.
Please
continue
for
another
two
trials” (Ben-An,
Flonian,
&
Mikulincer,
1999,
p.
41).
One
study
had
participants
visualize
someone
who either:
(a)
liked
them
contingent
on
certain
standards
(comparison
condition);
(b)
liked
them
uncontingently
(self-affirmed
condition);
or
(c)
they
knew
but
did
not
interact
with
socially
(control
condition).
One
study
had
participants
complete
sentence
scrambles
that
created
affirming
sentences
either
related
or
unrelated
to
the
self-threat
(Stone
&
Cooper,
2003),
one
study asked
participants
about
their
likelihood
of
seeing
a
movie
about
an
issue
they
had
previously
rated
for
general
and
personal
importance
(Simon
et
al.,
1995),
and
one
study
had
participants
complete
a
scale
of
Motivation
to
Control
Prejudiced
Reactions
to
reduce
stereotype
threat
before
completing
a
measure
of
implicit
attitudes
(Frantz,
Cuddy,
Burnett,
Ray,
&
Hart,
2004).
In
an early
study,
before
participants
wrote
counter-attitudinal
essays
against
increased
funding
for
treatment
and
facilities
for
the
handicapped,
they
were
told
that
they
would
have
the
opportunity
to
record
exams
onto
audio-cassettes
for
blind students during
a
break
in
the
study
(Steele
&
Liu,
1981).
This
is
the only
study
that
examined
the
effects
of
an
anticipated
self-affirmation
strategy.
Effects
of
S4f-Affirmation Manipulations
on
Dependent
Variables
The
use
of
multiple
dependent
measures
violates
the
assumption
of
stochastic
independence
of
effect
size
estimates
(Hedges
&
01km,
1985);
a
requirement
for
meta-analysis.
Therefore,
instead
of
combining
studies
that
measured
their
primary
dependent
variable
in
similar
ways,
we
present
effect
sizes
for
multiple
dependent
variables
per
study
in
Appendix
Table
2
for
a
more
thorough
comparison
across
studies.
To
examine
whether
the
effect
sizes
were
consistently
larger
for
certain
types
of
manipulations
compared
with
others,
we
organized
categories
of
dependent
variables
by
manipulation
type
(Appendix
Table
2).
The range
of
absolute
effect
sizes
for
specific
contrasts
between
self-affirmation
and
control
conditions
varied
by
dependent
variable (Appendix
Table
2).
However,
some
categories
of
self-affirmation
manipulations
contain
only
a
few
studies,
statistical
contrasts
vary
by
study,
and
not
all
categories
of
dependent
variables
are
examined within
each
category
of
the
manipulations.
Therefore,
comparisons
of
results
and
interpretations
should
be
made with
caution.
For
these
reasons,
meta-analyses
were
deemed
inappropriate
for
many
dependent
variables
and
uninformative
for
others.
For
example,
for
the
10
studies
that
assessed
attitude
and
reported
self-affirmed
vs.
not-affirmed
contrasts,
an
insufficient
number
(<3)
of
different
manipulation
types
were
available
to
make
statistical
comparisons.
Although
a
sufficient
number
(>3)
of
studies
were
available
to
compare
presentation order
(before
or
after
self-threat)
on
weighted
effect
size,
no
significant
difference
was
detected
(z
=
0.36,
p
>
.05).
Similar
results
were
found
for
the
six
studies
that
assessed
persuasion
similarly,
but
other
dependent
variables
had
even
fewer
directly
comparable
studies.
298
A.
Mc
Queen
&
W.
M.
P.
Klein
Self-affirmation
consistently
reduced
attitude
change and
psychological
discom
fort
following
a
dissonant
act,
as
long
as
the
self-affirmation
manipulation
was
not
challenged
or
negated
(e.g.,
Galinsky
et
al.,
2000).
However,
self-affirmation
may
only
decrease
bias
when
the
issue
is
of
high
personal
importance
(Cohen
et
al.,
2000;
Correll,
Spencer,
&
Zanna,
2004).
Compared
with
non-affirmed
participants,
self-
affirmed
participants
were
more
accepting
of
counter-attitudinal
arguments
(Cohen
et
a!.,
2000;
Correll
et
al.,
2004),
demonstrated
less
prejudice
(Fein
&
Spencer,
1997;
Zarate
&
Garza,
2002,
study
1),
but
viewed
a
non-stigmatized
other
more
stereotypically
when
the
left
hemisphere
was
activated
(Shrira
&
Martin,
2005),
chose
upward
social
comparisons
or
fewer
downward comparisons
(Spencer
et
al.,
2001,
study
3;
Wood
et
al.,
1999,
study
3),
and
made
fewer
external
attributions
for
other’s
behaviors
(Liu
&
Steele,
1986)
and
team
performance
(Sherman
&
Kim,
2005).
Self-affirmation
also
may
influence
the
associations
among
cognitions.
In
Klein
et
al.
(2001),
self-affirmation
in
the
presence
of
social-comparison
threat
dis
associated
risk
perceptions
from
risk
factors.
Additionally,
Sherman
and
Kim
(2005)
concluded
that
self-affirmation
eliminated
the
link
between
self-judgments
and
judgments
of
one’s
group
(athletic team).
As
illustrated
in
Appendix Table
2,
the
majority
of
dependent
variables
examined
have
been
cognitive
in
nature.
However,
eight
studies
measured task
performance;
four
of
which
found
no
effect
of
self-affirmation.
In
two studies,
intrinsic
self-affirmations
improved
performance
on a
subtraction
task,
but
extrinsic
affirmations
did
not
(Schimel et
al.,
2004,
studies
1
&
2).
Three
studies
assessed
changes
in
behavior
following
the
presentation
of
personally-relevant
health-risk
information
(Harris
&
Napper,
2005;
Reed
&
Aspinwall,
1998;
Sherman
et
al.,
2000,
study
2),
but
only one
reported
a
significant
positive,
immediate
influence
of
self-affirmation
on
health
behaviors
(Sherman
et
al.,
2000,
study
2).
Another
study
involved
repeated
measures
for
heart
rate,
blood
pressure,
and cortisol
levels
before,
during,
and
after
a
laboratory
stress-inducing
task,
but
only
found
significant
differences
in
cortisol
levels
(Creswell
et
al.,
2005).
Specifically,
non-affirmed
participants
exhibited
a
significant
increase
in
saliva
cortisol
levels
from
pre-
to
20
minutes
post-stress
onset,
whereas
the
change
was
not
significant
among
self-affirmed
participants.
A
review
of
the
mixed
effects
of
self-affirmation
on
cognitions
and
behaviors
related
to
health
can
also
be
found
in
Harris
and
Napper
(2005).
Although
Sherman
et
al.
(2000)
found
only
positive
support
for
self-affirmation
on
perceived
risk
and
intention
to
change
one’s
risky
behavior,
other
studies
showed
mixed
effects
of
self-
affirmation
on risk
cognitions
(Klein
et
al.,
2001,
study
1;
Reed
&
Aspinwall,
1998).
Harris
and
Napper
(2005)
reported
increased
perceived
risk
and
intention
to
reduce
heavy
alcohol
consumption,
but
only
among
self-affirmed
participants
who
were
heavy
drinkers.
However,
they
found
no
differential
effect
of
self-affirmation
on
drinking behavior
at
follow-up.
Another
study
that
used
two
self-affirmation
manipulations
found
no
effect
of
either
method
on smokers’
recall
of
risk
information,
evaluation
of
risk
messages,
mood,
or
motivation
to
quit
smoking
(Dillard
et
al.,
2005).
Re-Evaluation
of
the
Effects
of
Self-Affirmation
Manipulations
on
Dependent
Variables
When
sufficient
data
were
presented,
effect
sizes
were
calculated
for
mean
differences
reported
as
non-significant
by
authors.
In
some
cases,
our
estimates
produced
some
SeUAffirmarion
Manipu1ations
299
meaningful
results
(d>
0.30)
that
were
previously
overlooked.
For
example,
Matz
and
Wood
(2005,
study
2)
reported
no
significant
differences
for
positive
and
negative
emotional
reactions,
but
we
suggest
that
self-affirmation increased
positive
emotions
and
to
some
degree
decreased
negative
emotions.
Harris and
Napper
(2005)
examined
effects
of
self-affirmation
on
drinking
risk
over
time,
which
produced
fewer
significant
findings
than
our
estimated
effect
sizes
for
Time
1
only.
Specifically,
we
found
that
compared
with
non-affirmed
heavy
drinkers,
self-affirmed
heavy
drinkers
had
lower
self-esteem,
higher
negative
affect,
and
reported greater
perceived
evidence
strength.
Among
lighter
drinkers,
being
self-affirmed
decreased
perceived
evidence
strength,
negative
mood,
and
slightly
increased
self-esteem.
The
differences
in
results
illustrate
that
when
accounting
for
repeated
measures
in
a
factorial
design,
the
effects
of
self-affirmation
on
specific
dependent
measures
are
smaller
than
cross-sectional
approaches
to
data
analysis.
Lastly,
Sherman
and
Kim
(2005)
reported
a
non-significant
interaction
between game
outcome
and
self-
affirmation
status.
Our
estimate
of
effect
size
(d=
0.37)
suggested
a
small,
but
meaningful
difference; the
direction
could
not
be
determined
with
the
information
provided.
Role
of
Mood
Steele
and
Liu
(1983)
argued
that
the
effects
of
self-affirmation
on
attitude
change
were
not
due
to
changes
in
positive
mood.
In
fact,
self-affirmed
individuals
reported
less
positive
mood.
Later
studies
suggested
that
self-affirmation
buffered
the
negative
effects
of
helplessness
training
on
mood
(Liu
&
Steele,
1986).
However,
some
researchers
have
reported
a
positive
effect
of
self-affirmation
on
mood
relative
to
comparison
conditions
(Galinsky
et
al.,
2000,
study
1;
Koole,
Smeets,
van
Knippenberg,
&
Dijksterhuis,
1999,
study
3;
Van
den
Bos,
2001,
study
2)
and
others reported
no
significant
effect
(Cohen
et
al.,
2000,
study
3;
Klein
et
al.,
2001,
study
1;
Sherman
et
al.,
2000,
study
1,
2005,
study
1;
Shrira
&
Martin,
2005;
Wiesenfeld,
Brockner,
&
Martin,
1999,
study
I).
One
study found
no
significant
differences
in
self-justifying
attitude
change
as
a
function
of
positive
or
negative
induced
mood
(Steele
et
al.,
1993,
study
3);
therefore,
the
authors
concluded
that
mood induction
does
not
serve
as
a
proxy
for
self-affirmation.
However,
according
to
mood-manipulation
checks,
participants
did
not
maintain
their induced
mood
state
throughout
the
experiment.
Role
of
SelEsteem
The
role
of
self-esteem
varied across
studies
involving
self-affirmation
manipulations
and
was
examined
as
both
an
independent
and dependent
variable.
Three
studies
selected
participants
based
on
self-esteem
scores
(Schimel
et
al.,
2004,
study
2;
Spencer
et
al.,
2001,
study
2;
Wood
et
a!.,
1999,
study
3),
and
one
study
measured
self-esteem
and categorized
participants
based
on
scores
(Gramzow
&
Gaertner,
2005,
study
3).
Various
measures
were
used
to
identify
participants
with
high
and
low
self-esteem
including
Rosenberg,
Janis
Field,
and
the
Self-Rating
Scale
(Flemming
&
Courtney,
1984).
In
Creswell
et
a!.
(2005)
their
index
of
self
resources
for
personal
growth
included
self-esteem
(Rosenberg,
1965),
tendency
to
self-enhance
(Taylor
&
Gollwitzer,
1995),
and
optimism
(Scheier,
Carver,
&
Bridges.
1994)
scales.
The reliability
of
median
self-esteem
scores
across
samples
is
unknown.
300
A.
Mc
Queen
&
W.
M.
P.
Klein
Five
studies
measured
some
form
of
self-esteem
as
a
dependent
variable.
One
study
that
used
an
implicit
measure
of
self-esteem
was
the
only
one
to
find
a
positive
effect
of
self-affirmation
(Koole
et
al.,
1999,
study
3).
Another
study
found
higher
collective
self-esteem
in
athletic
game
winners
compared
with
losers,
but
only
among
those
not
affirmed
(Sherman
&
Kim,
2005,
study
2).
Discussion
This
is
the
first
attempt
to
closely
examine
the
specific
methods
by
which
self-
affirmation
has
been
manipulated
in
the
laboratory.
This
review
provided
support
for
self-affirmation
theory,
but
also
raised
questions
to
be
examined
in
future
research.
We
review
our
understanding
of
the
self-affirmation
construct,
the
effects
of
self-affirmation
manipulations,
and
possible causal
mechanisms,
and
offer
suggestions for
future
studies
involving self-affirmation
manipulations.
Our
review
describes
several
different
approaches that
have been
used
in
the
literature
to
induce
self-affirmation, which
consistently
vary
by
affirmation
domain
(value or
positive
personal
characteristics),
attainment
(participant-
or
investigator-identified),
and
procedure
(scale,
essay,
imagery, feedback, priming).
What
is
Se(f
Affirmation?
Self-esteem
is
defined
as
a
global
self-evaluation
(on
a
positive
negative
continuum),
whereas self-affirmation
is
the
active
affirmation
of
some
important
aspect
of
one’s
self-concept
that
is
unrelated
to
a
self-threat.
Only
one
study
that
examined
the
effect
of
self-affirmation on
self-esteem
found
a
positive
association,
so
it
is
unclear whether
affirming
an
important
aspect
of
the
self
will
increase
one’s
global
self-evaluation.
In
fact,
according
to
Steele
(1988),
self-affirmation
strategies
are
used
to
reduce
the
self-
threat
by
maintaining
or
restoring
an
overall
positive
sense
of
self
so
perhaps
no
substantive
change
in
global self-esteem
should
be
evident.
Future
studies
using
more
implicit
measures
of
self-esteem
(as
used
in
Koole
et
al.,
1999)
may
provide
either
more
support
for
this
argument
or
demonstrate
differences
in
implicit
and
explicit
self-esteem
after
self-affirmation
or positive
self-evaluation.
A
type
of
self-enhancement
that
occurs
after
a
public
threat
to
self-esteem
is
compensatory
self-inflation
(Greenberg
&
Pyszczynski,
1985).
Although
both
self-
affirmation
and
compensatory
self-inflation are
believed to
restore
one’s
self-regard
after
a
threat,
compensatory
self-inflation
may
generate
a
“lofty
self-image”
that
requires
additional
defensive
strategies
to
be
maintained
(p.
279).
Self-affirmation
posits
that
individuals
draw
on
unthreatened,
but
important
aspects
of
their
self-
concept
to
maintain
rather
than
maximize
positive self-worth.
Self-affirmation
may
serve
as
a
“buffer”
or
resource
to
cope
with
a
self-threat
(Liu
&
Steele,
1986);
therefore,
self-affirmation
without
a
threat
to
self
may
be
more
appropriately
referred
to
as
a
positive
self-verification
or
self-evaluation
strategy.
We
excluded
several
studies
from
our
review
because
their
manipulation
of
the
self
concept
was
not
in
conjunction
with
a
self-threat
(Arndt,
Schimel,
Greenberg,
&
Pyszczynski,
2002;
Harber,
2005;
McGuire
&
McGuire,
1996).
Similar
to the
buffering
model’s
relationship
between
perceived
social
support
and
health
outcomes
during
times
of
stress,
self-affirmation,
by
definition,
may
be
most
applicable
in
times
of
self-threat
(Cohen
&
Wills,
1985).
Positive
self-evaluation
in
general,
like
low
social
isolation,
is
expected
to
promote
positive
outcomes
through
different
causal
mechanisms.
SelJAffirination Manipulations
301
Defining
an esoteric concept
like
self-affirmation
is
difficult
and
being
able
to
control
its
presence
or
absence
in
an experiment
hinges
on such
definitions.
Our
inclusion
of
studies
using
self-affirmation
manipulations
of
different
methods
(i.e.,
positive
feedback
compared
with
value
scales
or
essays)
obscures
the
notion
of
what
constitutes
a
self-affirmation.
For
example,
can
an
individual
be
self-affirmed
by
positive
comments
or
evaluations
from
someone
else?
Results
from
Steele
et
al.
(1993)
support
the
assertion
that
positive
feedback
after
dissonance
is
aroused has
effects
on
attitude
change
similar
to
other
self-affirmation
manipulations.
However,
the
effects
of
feedback
may
be
moderated
by
culture
(Heine
&
Lehman,
1997).
Future
studies
should compare
positive
feedback
against
other
types
of
self-
affirmation
manipulations
within
the
same
study
to
determine whether
their
effects
on
the
same
dependent
variables
are
similar
or
different.
Further
consideration
should
also
be
given
to
the
nature
of
the
positive
feedback,
because
affirmation
of
the
individual’s
success
or
status
(e.g.,
academic
competence)
may
be
more
external
to
the
self
and
feel
more
conditional.
Specifically,
self-affirmations
of
intrinsic
or
self-
determined
values
(i.e.,
learning
for
self-betterment)
may
have
a
greater
positive
effect
on
behavior
compared
with
affirmations
of
extrinsic
or
socially
imposed
values
(i.e.,
academic
achievement;
Schimel
et
al.,
2004).
Similarly,
affirmations
on
characteristics
that
are
amenable
to
change may
be
different
from
those
perceived
to
be
more
innate or
trait-like.
Studies
have
shown
that
when self-affirmations
are
in
a
domain
related
to
the
self-threat,
dissonance
or
defensiveness
is
increased,
which
supports
Steele’s
(1988)
theory
that
self-affirmation
must
occur
in
an
unrelated
domain
(Blanton
et
al.,
1997;
Stone
&
Cooper,
2003).
Unambiguous
self-threat
may require
self-
affirmation
in
an
explicitly
unrelated
domain. Alternatively,
it
has
been
hypothesized
that
self-affirmations
compatible
with
a persuasive
message
will
reduce
resistance
to
persuasion
(Jacks
&
O’Brien,
2004).
The
majority
of
self-
affirmation
manipulations
reviewed
here
were
in
domains
unaffected
by the
self-
threat.
However,
one
study
in
our
review
used
a
writing task related
to
the
threatened domain
and
found
that
self-affirmation
eliminated
the
effect
of
job
insecurity
on
mood
(Wiesenfeld
et
al.,
2001).
Future
studies
may
seek to
further
explore
whether
self-affirmations
in
a
related
or
compatible
domain
can
be
equally
effective
as
those
in
an unrelated
domain.
Related domains
may
only
be
effective
if
the
self-threat
is
more
ambiguous
or
less
injurious.
It
may
be
that
the
domains
within
an
aspect
of
the self
are
more
discrete
than
first
envisioned
and
affirming
oneself
on
one’s
broad
job
skills
and
applicability/marketability
may
buffer
against
the
threat
of
a
specific
job
layoff.
What
Are the
Effects
of
SeFAffirmation
Manipulations?
Self-affirmation
manipulations
of
various
types
appear
to have
more
consistently
positive
effects
on
attitude
change
after dissonance
arousal,
acceptance of
counter-
attitudinal
arguments, and reductions
in
prejudice,
downward
comparisons,
and
external
attributions of
others’
behavior
compared
with
other
dependent
variables
reviewed.
Evidence
is
mixed
concerning
the
effects
of
self-affirmation
manipulations
on
risk
cognitions,
intention,
behavior,
task performance,
physiological
stress
responses,
mood,
and
self-esteem.
Insufficient
numbers
of
studies
employing
different
manipulation
types
and
assessing
similar
dependent
variables
for
similar
self
affirmed
and
non-affirmed groups
were
available
to
determine whether
any
one
manipulation
type
significantly
affected
observed
effect
sizes.
302
A.
Mc
Queen
&
W.
M.
P.
Klein
What
Are
the
Causal
Mechanisms
That Explain
the
Effects
of
Seif-Affirmation
Manipulations?
Self-affirmation
is
posited
to
reduce
individuals’ defensiveness
to
a
self-threat,
but
the
specific
causal mechanisms
remain
unknown.
Mediators.
Trivialization
and
mood
have
been
examined
as
potential
mediators
of
the
effect
of
self-affirmation,
but
results
are
inconsistent
across
studies. Simon
et
al.
(1995)
found
support
for
trivialization
of
counter-attitudinal
behavior
in
conjunction
with
self-affirmation, whereas
Koole
et
al.
(1999)
found
no
effect
of
self-
affirmation
on
measures
of
trivialization
across
three
studies.
Despite
multiple
reports
that
self-affirmation
manipulations
did
not
affect
mood,
Koole
et
al.
(1999,
study
3)
reported
that
the
effect
of
self-affirmation
on
recognition
accuracy
was
mediated
by
positive
mood,
but
not
self-esteem.
Further,
self-esteem
was
not
related
to
positive
mood,
supporting
the
independence
of
the
two
constructs.
However,
implicit measures
were
used
for
both
positive
mood
and
self-esteem,
and
such
measures
are
not
yet
well
established. Results
reported
by
Koole
et
al.
(1999)
are
consistent
with
ideas
that
self-defensive
processes
(i.e.,
dissonance
reduction,
self-evaluation
maintenance,
self-affirmation)
are
mediated
by
affect
of
which
people
are
not
explicitly
aware
(Tesser,
Martin,
&
Cornell,
1996).
Moderators.
Self-affirmation
manipulations
are
effective
when
they
are
not
negated
(Galinsky
et
al., 2000)
and
when
they
are
related
to
an
issue
of
high
personal
importance
(Cohen
et
al.,
2000;
Correll
et
al., 2004).
Self-affirmation
manipulations
may
be
less
effective
for
more
chronically
activated
dissonance-arousing
behaviors
such
as
smoking
or
risky
alcohol
use
for
which
individuals
have
already
developed
defensive
strategies.
Additionally,
cultural
factors
may
moderate
the
effect
of
self-
affirmation.
Individuals
from
collectivistic
cultures
may
require dissonance
and
self-
affirmation
manipulations
that
respectively
threaten
and
affirm
important
aspects
of
the self-concept
that
is
relevant
to
a
collective
or
shared identity
(Heine
&
Lehman,
1997;
Hoshino-Browne
et
al.,
2005).
Order.
A
methodological
factor
that
may
influence
how
self-affirmation
affects
cognitions
and
behaviors
is
whether
the
self-affirmation
manipulation
is
presented
before
or
after
the
self-threat.
Multiple
studies
showed
positive
effects
of
self-
affirmations
presented
either
before
or after
the
self-threat;
therefore,
one
may
conclude
that
the
order
of
presentation
is
not
important
and
that
the
effect
is
robust
(Sherman
et
al.,
2000).
Information
processing
and
evaluation
may
be
equally
affected
by
either
presentation
order
if sufficient
time
is
provided
before
dependent
measures
are
assessed.
Similarly,
given
the
short
time
intervals
between
delivery
of
the
self-threat
and
self-affirmation
manipulation
in
most
laboratory
experiments
immediately
prior
to
measurement
of
dependent
variables,
little
difference
in
order
may
be
expected.
Alternatively,
different
mediators
may
be
responsible
for
the
effect
of
self-affirmation
manipulations
depending
on
whether
the
effect
is
one
of
prevention
or
repair
of
harms
to
the self-concept.
Self-Esteem.
As
noted
earlier,
little
support
exists
for
the
effect
of
self
affirmation
on
dependent
measures
of
self-esteem.
However,
self-esteem
appears
to
moderate
the
effect
of
self-affirmation,
which
also
supports
the
independence
of
Setf-Affirmation
Manipulations
303
these
constructs.
Several
studies
have
shown
that
self-affirmations
in
a
domain
unrelated
to the
self-threat
produce
greater
attitude
change
(Steele
et