The Sleeper Effect and Negative Political Advertising
Author(s): Ruth Ann Weaver Lariscy and Spencer F. Tinkham
Source: Journal of Advertising, Vol. 28, No. 4 (Winter, 1999), pp. 13-30
Published by: M.E. Sharpe, Inc.
Stable URL: http://www.jstor.org/stable/4189122
Accessed: 26/08/2010 14:48
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=mes.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
M.E. Sharpe, Inc. is collaborating with JSTOR to digitize, preserve and extend access to Journal of
Advertising.
http://www.jstor.org
The
Ruth
Sleeper
Ann
Weaver
Effect
Lariscy
and
and
Spencer
Negative
Political
Advertising
F. Tinkham
is examined
to explain how the impact of negative political
The sleeper effect phenomenon
advertising
persists
?over
time. Using a true experimental
?and
even increases
design with attack and defensive
message
were asked to report their vote and its certainty
adult subjects
treatments,
after message
immediately
exposure
callback.
Results
indicate
that a defensive
and in a delayed
advertisement
the attack is
telephone
following
over the next few weeks, the impact of the attack ad increases
initially
effective; however,
substantially.
an initial perception
that the assailant
has low credibility
has only a temporary
Similarly,
suppressive
impact
on the effectiveness
and a credibility-driven
of the attack ad. Our documentation
of both an order-driven
who are attacked.
these results offer
for those candidates
Further,
sleeper effect poses strategic
challenges
into the determinants
theoretical
and the pervasiveness
insights
of the sleeper effect in political
campaigns.
Ruth Ann Weaver Lariscy (Ph.D.,
University of Missouri) is an
Associate Professor, College of
Journalism and Mass
Communication, at the University of
Georgia.
Spencer F. Tinkham (Ph.D.,
University of Illinois at Urbana) is a
Professor, College of Journalism and
Mass Communication, at the
University of Georgia.
Acknowledgments: The authors wish
to thank the James M. Cox, Jr.
Institute for partial support of this
research, and to thank the
anonymous reviewers, particularly
Reviewer C, for thorough and
insightful contributions.
Journal of Advertising,
Volume XXVIII, Number 4
Winter 1999
In the 1980's and the late 1970's, when negative
advertising
strategies
were still regarded
as somewhat
the volume
and style of nega"maverick,"
became a mainstay
of election
tivity in campaigns
coverage
(Advertising
Age
Television
such advertising
I Radio Age 1988; Will 1986). While
is
disliked
Hill
and
and
(Ansolabehere
1989)
consistently
Iyengar 1995;
thought
to alienate
of potential
voters (Freedland
large numbers
1994; Rothenberg
1998;
its effectiveness
is endorsed
and
1995),
Teepen
by its continuing
at
of
use
level
(Jamieson
increasing
every
campaigning
1992; Newsweek
and Weaver Lariscy
1997).
1996; Tinkham
The prevalence
and increasing
of negative
attack
ads (West
legitimacy
raises
the
how
of
to
to
and discount
the influence
1992)
of
question
respond
an attack
candidates
who will almost
be
among
particularly
inevitably
attacked.
In the 1980's a popular
was to "take the high road" if
strategy
and
not
the
accusation
with
a response.
Michael
Dukakis'
attacked,
dignify
lack of response
to the "Willie
Horton"
ad in 1988 is cited by political
and commentators
and Campbell
(Jamieson
1992) as well as by
analysts
Dukakis
himself
1997) as a contributing
Journal^
(Washington
C-Span,
factor to the landslide
Bush victory.
the
use
of
unrelated
to the attack
Similarly,
(as atpositive
messages
to
counter
does
not
offset
the specific message
of the
ads)
tempts
negative
attacker.
While these ads may bolster the evaluation
of the attacked
candithan the attack
ad. Labeled
the "negativity
date, they are less powerful
results
from
numerous
information
and judgmental
bias,"
processing
studies
indicate
that individuals
informadecision-making
weigh negative
tion more heavily
than positive
information
(Kellerman
1984; 1989; Lau
it (Ito, Larsen and Smith 1998;
1982; 1985) and are more likely to remember
Dhillon
and
Lang 1991; Lang,
Dong 1995).
Another
that is sometimes
advocated
is for the
strategy
by consultants
candidate
to inoculate
a negative
attack early in the campaign.
This
against
referred to as "stealing
the opponent's
thunder"
approach,
colloquially
(Call
of attack (virtually
1995) is resisted
1996; Nelson
by most potential
targets
all candidates!),
for fear that inoculation
can raise potentially
damaging
issues about themselves.
This paper explores
how defense
attack ads impacts
the
against
negative
of those attacks
effectiveness
over time, as well as what effects, if any, will
last until election
was motivated
day. This study
by strategic
political
1990;
The
14
rather than by the desire to test aldecision-making
We also wanted
theoretical
ternative
interpretations.
in the current
to examine
particularly
empirically,
make
candidates
in which
environment
frequently
the use of negative
advertising,
against
pledges
"works" in the manner
whether
advertising
negative
this study preit does. Further,
believe
consultants
in an applied
context,
sents,
data-gathering
procethe mass
suited for observing
dures that are perfectly
called the sleeper effect
communication
phenomenon
to explore
an opportunity
(SE). As such, it provides
into theoof this phenomenon
and extend discussion
while remaining
retical territory,
by strategrounded
of voter
Thus, we examine
patterns
gic motivations.
over time and within
to attack advertising
response
of the sleeper effect.
context
the theoretical
Effect
Sleeper
and
in its most general
effect is defined,
Sleeper
characterized
sense, as the phenomenon
descriptive
at some point in time after a
impact
by increased
et al. 1978). Thus, in
event occurs (Gruder
stimulus
the sleeper effect
communication
studies,
persuasive
has a greater defound when a message
is typically
attitudes.
impact on receivers'
layed than immediate
and Sheffield
Lumsdaine
by Hovland,
conthe sleeper effect seemed both empirically
counter-intuitive.
and theoretically
Typitradictory
erodes over time
of a message
cally the effectiveness
there is a diminishing
of factors;
due to a variety
of persuasive
impact. Sleeper effect is the
persistence
First
named
(1949),
when "a persuasive
it is observed
message
opposite:
than initial impact on subjects'
(has) greater
delayed
attitudes..."
(Gruder et al. 1978).
have been adtheoretical
Various
explanations
the phenomenon
to explain
(See
vanced
and tested
and classificaAllen and Stiff [1989] for an overview
see also Cook, et al.
tion of many of the SE studies;
Kelman
Journal
of Advertising
and
Hovland
that when a
1953), specifies
is initially
as untrustworthy,
the
regarded
will have reduced
message
impact
upon first expobecause
it is more critisure, yet be better learned
source
(Allen and Stiff 1989; Gruder et al.
cally examined
as time
1981). However,
1978; Petty and Cacioppo
the source is forgotten,
or dissociated
from
passes,
in effect from the
the message,
an increase
yielding
content
itself.
message
Simply
put, while we may
discount
because of who said it, or
initially
something
we encountered
it, over time we will forget the
it from the message,
but rememsource, or dissociate
ber the message.
A more recently
for the
advanced
explanation
effect is the "differential
sleeper
decay" interpretawhere
tion
and Hutchinson
(Moore
1985;
Pratkanis,
Greenwald
and Leippe 1988). This interpretation
posits that a sleeper effect occurs when the impact of the
cue decays faster than does the impact of
discounting
and Leippe 1988).
the message
Greenwald
(Pratkanis,
If the discounting
cue is a source cue, it deteriorates
cue. Thus, over
more rapidly than does the message
more quickly than
time, the source cue is forgotten
In their review of persistence
theories of
the message.
(1993)
however,
change,
Eagly and Chaiken
documents
this interthat not enough research
'The differential
decay theory of the sleeper
pretation:
attitude
argue
of forgetting,
needs
effect, which is based on principles
as havbefore it can be regarded
more substantiation
While the jury is still out on
ing strong
support."
for sleeper
differential decay as a theoretical
explanation
effect, those who base their research on this interpretainterest in the phenomenon.
tion have rekindled
as an explanatory
In its 40-year
phenomhistory
both periods of relaenon, sleeper effect has generated
The
of discredit.
as well as periods
tive excitement
of the effect). The most widely
[1979] for a history
that dominated
used principle,
(and the explanation
is dissociation
until recently)
and discussion
research
routes (Eagly and Chaiken
of memory
[1993] p. 612)
cue (e.g.,
evaluated
where "an unfavorably
persuasion
with
associated
a negative
source) that was initially
from it
dissociated
becomes
the message
presumably
in the mid-1950s
term was widely popular
through
the 1970s (Eagly and Chaiken
1993); yet, numerous
and
were unable
to replicate
researchers
findings
and Watkins
the phenomenon
substantiate
(Johnson
some to
and Worrall
1970) leading
1971; Schulman
claim the time had come to "lay the sleeper effect to
rest" (Gillig and Greenwald
1974). Amidst the confuturned
attention
sion and controversy,
away from
or not a sleeper
whether
that documented
research
under which a sleeper
effect occurred, to the conditions
over time." This negative
source, or cue, is something
with the
that is associated
situation
in the persuasive
it?that
to discount
and causes recipients
is,
message
as less valid, truthful, etc., than
to regard the message
discount.
it would be regarded without the negative
called the
This dissociation
originally
explanation^
effect may or may not be observed.
that examines
For our research,
strategy
political
effect phenomfor the sleeper
as an applied context
is more imof these conditions
enon, documentation
theoretiof the competing
portant than consideration
for the sleeper effect. We empirically
cal explanations
discounting
cue hypothesis
(Hovland
and Weiss
1951;
examine
attack
and defense
political
advertising
mes-
Winter
1d
1999_
and explore
whether
or not such strategic
sages,
and their relative
effectivegive-and-take
messages,
criteria for observation
of
ness, may meet established
an absolute
effect.
sleeper
There are two different,
are often categorized
as
sleeper effect ?a statistidocumentation
of persuasion
between
cally significant
the first message
and the delayed
exposure
impact?
is the dominant
in this
model, and the one examined
For an absolute
research.
sleeper effect to occur, there
must be a delayed
increase
in persuasive
impact. It is
in contrast
understood
with the relative sleeper effect
in which the impact
of the initial
message
may dea discounting
cline without
cue. The relative
sleeper
a loss of persuasive
effect involves
for both
impact
but with no delayed
high and low credible
sources,
in impact
increase
for either
(Cook and Flay 1978;
Cook et al. 1979; Eagly and Chaiken
1993).
For Observing
Conditions
Sleeper
Effect. Regard-
Types of Sleeper
Effects.
over-time
that
phenomena
sleeper effects. The absolute
less
of their theoretical
all of the following
origin,
have been posited as requirements,
conditions
or necfor occurrence
of the sleeper effect.
essary conditions,
some of the conditions
obNotably,
require empirical
servation
or testing
to determine
whether
or not they
have been satisfied.
Other conditions
are not empirion the
cally determined;
rather,
they are dependent
research
or the data-gathering
design
procedures
in the particular
employed
study for their validation.
The first four conditions
for observation
of an absolute sleeper effect are based on the dissociation
explanation (Cook et al. 1979). These include:
itself must be persuasive;
it must
1) the message
have a significant
initial impact on attitudes
determined);
(empirically
cue must be strong enough in
2) the discounting
its immediate
impact to inhibit the attitude
would otherwise
change the original message
have produced
determined);
(empirically
time prior to the
3) there must be sufficient
for the discounting
cue and
delayed measure
the original
to become
dissociated
message
and
(methodologically
determined);
of the discounting
cue and
4) the dissociation
the original
must
occur quickly
message
that forgetting
will not have yet ocenough
curred (empirically
determined).
Petty and Cacioppo
add a fifth condition:
(1981)
assess
these
four
cue and the message
5) the discounting
content
affect attitude change separately, and they must
not affect each other (empirically
determined).
From the differential
of sleeper
decay explanation
conditions
are advanced
effect, three additional
by
Greenwald
and Leippe
Pratkanis,
(1988).
They assert that for an absolute
effect to occur, each
sleeper
of the following
must also be observed:
6) the
to the important
subjects
pay attention
in the message
arguments
(methodologically
and/or empirically
determined);
cue must be received after the
7) the discounting
and
message
determined);
(methodologically
must rate the trustworthiness
8) the subjects
of the message
communicator
immediately
after receiving
the discounting
cue (methodologically
determined).
Conditions
of the design
3, 7 and 8 are functions
and may be satisfied
inclusion
of
meaby
specified
and the ordering
of
surements,
explicit
instructions,
treatments.
Condition
that submessage
6, stating
to the message
jects should pay attention
arguments,
measurement
of atmay be met empirically
through
tention; but it may also be met methodologically
through
instructions
and procedures
that emphasize
"paying
attention."
determination
has been
Methodological
more prevalent
in previous
sleeper effect studies than
has empirical determination
Greenwald
and
(Pratkanis,
that (beLeippe 1988). It should be noted, however,
cause some conditions
are empirical
and others are
and for Condition
6 they may be eimethodological;
need not be mutually
ther) the conditions
exclusive.
For this study we will not argue that a political
meets only one set of the criteria?either
campaign
the original
dissociation
conditions
or the latter differential
we will examine
each
decay ones. Rather,
condition
in the political
context
and asadvertising
sess its relevance.
As different
of
interpretations
effect
have evolved,
conditions
have been
sleeper
context
of this study provides
added; the political
an
ideal setting
for controlled
observation.
From our overview
of the sleeper
effect literature
and requisite
conditions
for its observation,
we offer
the following
research
hypothesis:
HI: All specified
conditions
for observing
an
absolute
in
sleeper effect will be satisfied
this study of attack and defense
political
and
advertising.
all eight of the conditions
is the asUnderlying
that
an inherently
sumption
can
persuasive
message
have its impact
moderated
by its being associated
with a low credible
source. Further,
the presumption
exists that an effective
cue does in fact
discounting
the credibility
of the source.
derogate
Additionally,
the conditions
mandate
that issues
of both the con-
10
tent
The
of the discounting
cue and the order of exposure
cue
be considered.
Thus, beyond
discounting
which
led to HI, three issues
criteria
specific
to the
the
require
additional
consideration
and
will
result
in
H2.
determining
Cue. Over the years a wide
Content of Discounting
of both source and message
attributes
have
variety
served as discounting
cues. In early studies
the discue was often a simple
of the
counting
designation
of the message
source
Lumsdaine
and
(Hovland,
et al. 1953; Hovland
and
Sheffield
1949; Hovland
of
Weiss 1951), flashed on the screen at the beginning
Journal
of Advertising
attack by creating
resistance
to arguments
sequent
in the attack message.
Thus the inoculation
theory
tradition
is based on the premise
that a discounting
cue prior to an attack
interacts
with the
message
attack itself and influences
the impact ofthat
attack.
The extensive
effect literature
that
sleeper
suggests
an absolute
effect is not a reliable
sleeper
phenomenon when the discounting
cue precedes
the message,
even though the earlii.e., the inoculation
condition,
finding was that, at the introducthe high credible
source
message,
a stronger
effect than the low
produced
persuasive
credible
source, but that after a period of time (often
the high credible
source message
3-4 weeks),
experi-
est research
was based on this pre-message
cue. Perthe explanation
for this is that Pratkanis,
haps
Greenwald
and Leippe (1988) viewed the discounting
cue as a separate
the early studies
whereas
message,
the discounting
cue in the original perincorporated
suasive
itself.
message
Source
Effects
of communicator
credCredibility.
have a long, rich hisibility on message
processing
enced a decay
source
credible
in persuasive
while the low
impact,
exhibited
the sleeper effect.
message
initial
studies
and findings,
to these
Subsequent
was accomplished
the discounting
by direct manipuof a communicator
lation of the credibility
delivering
the message
(see Cook et al. 1979 for a complete
The discounting
cue has not
review and discussion).
source or a communicabeen a manipulated
always
In one set of studies
the dishowever.
tor variable,
written
that
statement
cue was a specific
counting
and Miller 1966;
(i.e., Greenberg
tory in persuasion
Hovland
et al. 1953; Littlejohn
and
1971; Petty
is recognized
as a
1984). Source credibility
Cacioppo
on multiple
complex
phenomenon,
dependent
cognitive and affective
factors.
indimensions
Cognitive
clude judgments
about perceived
attributes
such as
and competence/expertise
of the mespower, prestige,
dimension
concerns
sage source. The affective
judgments regarding
such elements
as attractiveness,
dyand trustworthiness
and
namism,
(Batra,
Myers
and told
at the end of the persuasive
message
the information
they had just been exposed
subjects
to was false and inaccurate
(Gruder et al. 1978). In
and Leippe 1988)
Greenwald
later studies (Pratkanis,
cues were expanded
beyond a simple statediscounting
atto include
the source
ment derogating
explicit
of the source of the initial
tacks on the credibility
Aaker
a film.
tion
The
of the
typical
initial
came
it a separate
message
altogether.
making
message,
or a
source
a manipulated
Whether
characteristic,
in all sleeper
of the message
argument,
derogation
cue provides
a
to date the discounting
effect studies
to the source.
direct reference
Cue. One of the condiPlacement
of Discounting
Greenwald
and Leippe
added
tions
by Pratkanis,
of the
that observation
is the specification
(1988)
cue
that the discounting
effect also requires
sleeper
be introduced
Many source cues
after the message.
to and remember
affect how we respond
messages
when the source is identified
(Eagly
prior to a message
and Chaiken
1965). An extensive
1975; Husek
body
cues prior
which examines
discounting
can be found under the rubric of
to attack messages
inoculation
theory (Burgoon et al. 1976; McGuire 1961;
and Papageorgis
McGuire
1961; Pfau and Burgoon
in effect as
cues are designed
1988). Such discounting
to
influence
the subwhich
intended
are
forewarnings
of literature
1996, pp. 402-404).
cannot be viewed as a simple
Yet, source credibility
additive
formula that asks "how much of each ingredient?" does the source contain (Anderson
1971; Eagly
and Chaiken
1996). In fact,
1993, p. 249-50;
Haley
in recent years has emphasized
research
that cred-
so literally;
that it is
ibility need not be interpreted
more important
to maintain
focus on the overall, global evaluation
of the source than on weighted
shares
of individual
and
(Batra,
parts
Myers
component
Aaker 1996; Eagly and Chaiken
1993).
(1993, p. 249) argue that source
Eagly and Chaiken
with other
that interacts
being a variable
credibility,
of both the message
and the communicaattributes
is not best analyzed
in isolation.
tion setting,
Some
this point
as well
research
advertising
supports
(Debevec and Iyer 1986; Gotlieb and Sarei 1991; Haley
1996; Stern 1994). The object of an ad (be it a branded
is the central source
candidate)
product or a political
imof advertising
copy; but it is strongly
component
namely the sponpacted by other source components:
and the vehicle
sor, the endorser,
(Batra, Myers and
Aaker 1996, p. 402). Thus, the global credibility
of the
of the
may impact the perceived
credibility
company
with elements
of
which may interact
spokesperson
the message.
Winter
1999
?
In our study,
as in much of the relatively
recent
on source credibility
literature
and advertising
mesassesssages, we focus on a global or on an overall
ment of the source's
because
of
Further,
credibility.
the competitive
context
of the political
we
campaign,
will be examining
source credibility
of not one, but
two message
sources:
the assailant
and the defendant who is the target of attack.
In this relative
asa global rather than ingredient-specific
sessment
apwell-suited.
From this overview
proach is particularly
of the literature
dealing with the content of discountas well as the role of
ing cues and their placement,
source credibility,
we offer this second hypothesis:
H2: Order of placement
of the discounting
cue
and perceived credibility of the attacker will
be determinants
of an absolute sleeper effect.
In our analysis,
source credibility
of the attacker
after message
will be assessed
(immediately
exposure)
in terms of how it differs from source credibility
of the
to the defensive
ad.
target upon exposure
Time
Elapsed
Subsequent
Much
Between
Initial
and
Testing
has
been
effects
written
about the persistence
of
over
time
&
Chaiken
(see
persuasive
Eagly
review of persistence).
The most
[1993] for a thorough
common persistence
function
is the rate of decay, and
central to this function
is the amount
of elapsed
time
between
first and subsequent
measurements.
opinion
As previously
in most studies of the sleeper
discussed,
cue is placed just prior to, emeffect, the discounting
bedded within,
or just after the message
itself; and
more recently
it has been conceived
as a separate
or following
the base message.
message,
preceding
Across
all sleeper
effect studies,
the first
however,
measurement
has been taken
opinion
immediately
on nuThen, depending
following
message
exposure.
merous
factors (primarily
the research
design itself)
a second opinion measurement
follows?hours,
days,
or weeks
later.
When
Greenwald
and
Pratkanis,
conducted
17
simulations
and
(1988)
Leippe
replications of previous
effect studies,
sleeper
they varied
the amount
of time for the delayed
measure
from
several minutes
to over six weeks
(in four experiments)
(in three experiments).
Earliest
studies (Hovland,
Lumsdaine
and Sheffield
Hovland
and Mandell
Peterson
and
1949;
1957;
Thurstone
observed
a more powerful
1970)
1933;
effect as longer periods
of time elapsed.
Alsleeper
over a
though the sleeper effect has been documented
wide range of varying
times (from minutes
to many
tradiweeks), studies that emerged from the dissociation
tion tend to document
greatest evidence of sleeper effect
over longest
amounts
of time
(rather than shortest)
between
initial and delayed measurement
Although research on this topic is limited, we do predict an increased
power of sleeper effect over time. Specifically:
H3:When
an absolute
effect is obsleeper
it will be stronger
as the elapsed
served,
time between
and demessage
exposure
layed
The
Sleeper
measurement
Effect
increases.
and
the
Political
Campaign
This study extends
examination
of criteria
for observation
of an absolute
effect
to
the
sleeper
political
context.
Most previous
campaign
sleeper effect studies have been conducted
in laboratory
and
settings
have manipulated
either source or communicator
credfor the discounting
cue. We suggest
that a
ibility
with its nearly constant
back and
political
campaign,
forth attack/defense
is
a
conmessages,
prototypical
text for examination
of the sleeper
effect. Other apof sleeper effect principles
to political
camplications
are
but
not
unheard
of.
One
case
rare,
paigns
study
a U.S. Senate campaign
offers the differenexamining
tial decay interpretation
of the longitudinal
source/meswherein
the source cue deteriorates
sage separation,
more rapidly than does a message
cue (Pfau, Parrott
and Lindquist
Since
the
notion
of differing rates
1992).
of decay from message
source and message
content was
first advanced,
it
has
found
little
however,
empirical
Greenwald
and Leippe 1988).
absolute
effect as a theosleeper
retical
for
the
effectiveness
of
explanation
greater
attack
ads
over
the
negative
time,
negative
political
attack ad is considered
the base message
because
it is
more likely (given the negativity
to
have
a
stronbias)
ger initial impact. The attacker
is, thus, seen as initisupport (Pratkanis,
To examine
the
The defensive
natuating attitude
change.
message
the
role
of
the
cue
in
that
it
rally plays
discounting
to
"discount"
either
the
source
of
the
attack
attempts
or the attack message
content
or both. The defensive
is seen as attempting
to suppress
the impact
message
of the attack.
Some defensive
that
messages
merely
held by their sponsor may not work
support positions
as discounting
cues because
to
they don't attempt
the
of
source
the
attack.
These
indiderogate
directly
rect defenses
(i.e., general
positive
supportive
ads,
the power to suppress
mesOnly direct defensive
the
(which
of
the
sages
question
attacker)
credibility
of discounting
cues in tradimay act in the manner
tional sleeper
effect studies.
Unlike
earlier
sleeper
etc.) may
endorsements,
the impact of an attack.
lack
18
effect
The Journal
studies
(where
the
the discounting
or was identified
cue was part of
as the source/
the political
advertis-
itself,
message
of the message),
communicator
context
contains
two separate
(attack
ing
messages
and defense)
which
may often serve as the initial
and discounting
cue. The sleeper effect, with
message
its underpinnings
inherent
variables
of source credibility
in persuasion
theory,
and message
has far-reachfor both the study and the practice of
ing implications
its role
political
negative
advertising.
Understanding
in potential
of both attack and devoters' processing
fense advertising
issues remay illuminate
messages
lated
to both
and mandated
for
guidelines
Andrews
Elliott
(see
1991;
1996;
advertising
political
for increased
1991; for discussions
Lipman
calling
We are
regulation
governing
negative
advertising).
the
of
a natueffect
to
investigation
extending
sleeper
In
ralistic setting?the
the
political
campaign.
political campaign,
the defensive
assumes
the
message
role of the discounting
cue.
voluntary
Method
of
Overview.
This study reports a primary
analysis
data from a true experiment
in which 339 particito a televised
attack ad
negative
pants were exposed
and (depending
on random
to
treatments)
assignment
either (1) no other message
(the control group); (2) a
ad that precedes
the attack ad; or (3) a
ad that follows
the attack
ad. In order to
defensive
the empirical
and methodological
conditions
assess
that previous
research
has deemed
for obnecessary
an
we
absolute
considered
each
sleeper effect,
serving
the extent
and assessed
to which this excondition
or observed
met it through
either
design
periment
our
this
of
assessment,
Following
patterns
response.
us
method
that enabled
focused on the relationships
an
of our predictions
to assess
the validity
involving
absolute
effect. Thus we will be examining
sleeper
defensive
of Advertising
mid or late callback).
Our criterion
for observing
the
effect
was
immediate
and
of
sleeper
delayed certainty
vote for a specific candidate,
the attacker
versus the
of this analysis,
this score takes
target. For purposes
the perspective
of the attacking
and is
candidate,
thus characterized
scores favoring the atby positive
tacker and negative
scores favoring
the target of attack. For all manipulated
in the design,
variables
were
the perceived
For
subjects
randomly
assigned.
credibility factor, subjects were blocked by dividing their
scores at the median. The defensive
ads in both pre-and
in
their
post-attack
exposures
vary
specific content,
such that some are direct appeals
(which explicitly
the attacker
or his message)
and others are
derogate
indirect
bolster the attacked
appeals (which positively
It
should
be
noted
that
candidate).
only the direct defensive
are
consistent
with the content
of
messages
cues in all prior sleeper effect studies.
discounting
Volunteer
in a
Subjects.
participants
residing
of Northeast
area
10-county
Georgia were enrolled in
at Athens Technical
classes
Institute.
day or evening
The recruited sample was purposive
in that we sought
of registered
high representation
voters, over-age-25
and
of racial miadults,
proportional
representation
norities. The resulting
sample contained 221 registered
voters, 65% of whom had voted in the last national
election. Forty-two
percent were over 25 years old and
24% were members
of minority racial groups. Because
were
to treatment
condisubjects
randomly
assigned
in this study
tions, experimental
subgroups
analyzed
are consistent
with overall sample characteristics.
The Attack Ad. The negative
Stimuli.
political commercial that is central to this analysis
was produced
media consultant
Jon Franzen,
under the
by political
the
of
Harriman
for
auspices
Foundation,
general use
in edited form by candidates
in the
for
seats
running
U.S. House of Representatives.
Within each congressional
the relationship
order of exposure
to the
between
attack and defensive
the
relative
credads,
perceived
of
imthe
time
between
and
attacker,
elapsed
ibility
the ad could be edited for a specific
district,
In
our
was edited for
campaign.
study the commercial
a fictitious
in
candidate
the state of
Congressional
We
selected
a
state
different
Kentucky.
intentionally
from the residence
of study participants
to avoid po-
mediate
tential
and delayed
voting responses.
a mixed 2x2x3x2
facThis study employed
Design.
in which
the first three factors
were
torial design
and the fourth factor was
between-subjects
predictors
based
a within-subjecte
on the repeated
predictor
immediate
versus
measure,
delayed voting response.
factors were, respectively:
The three between-subjects
cue (before or
to the discounting
(1) order of exposure
of the atafter the attack);
(2) perceived
credibility
tacker
tween
(low versus
the immediate
time behigh); and (3) elapsed
and the delayed response
(early,
context contamination
from any known candidates in the ongoing campaign
cycle.
This attack
candidate
ad, sponsored
by fictional
Pat Michaels,
is subtly humorous
even though
it is
also a character
attack on his fictional
John
opponent
Boorman.
the commercial,
a lie detector
Throughout
instrument
is in use; this is the ad's primary
visual
content. As each of Boorman's
claims (about his Vietnam military
his position
on tax issues,
and
service,
his concern for the people of Kentucky)
is mentioned
and refuted
in a voice-over,
the lie detector
graph
Winter
1999
that a lie was
wide fluctuations,
indicating
the
lie
detector
mathe
last
In
told.
sequence,
being
when
the
assailant
chine
(Pat
actually
explodes
claims to "care about
notes that Boorman
Michaels)
exhibits
the people of Kentucky."
Each of the discounting
The Defensive
Messages.
ads was likewise
produced by area conprofessionally
cue in many studies
the discounting
Whereas
sultants.
that calls the credibility
consists of a simple statement
source into question (e.g., Cook et al. 1979;
of a message
and Leippe
Greenwald
et al. 1978; Pratkanis,
mescues
are
entire
our
advertising
1988),
discounting
of
the
attack
the
(Boorman)
target
by
sages, sponsored
the same purpose,
to accomplish
i.e., to call into quesThe use
of the assailant
tion the credibility
(Michaels).
the ecoenhances
defensive
of a complete
message
Gruder
of the study in that it characterizes
logical validity
the "give and take" in a political
campaign.
were classified
In this study the defensive
messages
the credibility
as direct (those that explicitly
question
or indirect (those that bolster the credof the attacker)
ads
of attack).
Direct defensive
ibility of the target
claimed
voton
Boorman's
attack
Michaels'
challenged
service, and concern for the people,
ing record, military
on
truthfulness
of Boorman's
evidence
and provided
ads supported Boorman's
these topics. Indirect defensive
about
only positive information
credibility by providing
to the opponent's
him, without making specific references
and
Karrh
claims
[1997] for a similar
(see
Halpern
of
direct
and
indirect
defenses).
conceptualization
of the
Check.
Following
production
Manipulation
was
to
each
attack and defensive
subjected
messages,
check using a separate sample
colfrom the same technical
of
source
was
the
that
experimensubsequently
lege
1994
and Weaver
tal subjects
(see Tinkham
Lariscy
After
of
the
for a thorough
study).
description
pretest
to recogwere instructed
the pre-test,
participants
deand indirect
ads as well as direct
nize attack
in
all
stimulus
materials
were
shown
fenses,
they
attack ad
format. The Michaels
commercial
finished
identified
was correctly
pre-test
by 87.1% of these
had a
Boorman's
and
discounting
messages
subjects;
the
within
score
of
76.8%
mean
pre-test
accuracy
in the text of only one defengroup. Minor revisions
a rigorous manipulation
students
of 70 volunteer
its message
in order to clarify
ad were made
on
this
based
strategy
pre-test.
The instrument
includes
four secInstrumentation.
which
were
in
a
50-minute
all
of
tions,
completed
session.
Section
1, completed
prior to ad expotesting
sive
toward poattitudes
sure, assesses
general
subjects'
of political
measures
litical advertising
and general
The second section,
after exinvolvement.
completed
19
of
posure to the first ad, contains
global evaluations
the ad and the two candidates.
In this section
sub17 different
like
items,
jects also rated
diagnostic
formatted
as
semantic
differen"ethical/
unethical,"
tial scales.
Section
3 followed
to their secexposure
ond ad and duplicated
the items from Section 2. Section 4 asked basic demographic
and also
information;
contained
the candidate
choice (vote) and choice cerwhich
form the primary
measures,
tainty
repeated
in
variables
the
These will be
dependent
experiment.
described
in detail
below.
Measures.
Two composite
measures
were
Composite
in
the
a
relative
source
cred(1)
employed
analysis:
and (2) a relative
vote certainty
score.
ibility measure
score between
the evaluation
The first is a difference
of the attacker
and the evaluation
of the target of
Each evaluation
was assessed
attack.
immediately
to the message
following
exposure
sponsored
by the
and by the target
attacker
(that is at the time of
Based on five evaluative
semanmessage
encoding).
tic differential
internal
items,
consistency/reliabilities
and alpha=.97.
As noted
are, respectively:
alpha=.98
the
relative
source
score
is based
credibility
previously,
of both source
and target
on global evaluations
of
in an assessment
of concurrent
attack. However,
vatwo cognitive
and affective
lidity of this measure,
factor
scores
based
on the 17 dimessage-generated
a
scales
exhibited
correlation
of .61
multiple
agnostic
with the global
measure.
This
indicates
credibility
that the global measure
both the
effectively
represents
and
affective
dimensions
of
source
cognitive
credibility.
The second composite
score (relative
vote certainty)
is the algebraic
of
a
contrast-coded
vote score
product
the
scale
Table
1 for its
(see
certainty
by
multiplied
These
measures
have
definition).
operational
certainty
a long history in attitude
research
and
(see: Fishbein
and
Krosnick
where
1998)
Ajzen 1975; Petty
they are
most often linked with intention
to perform a particuWe measure
lar behavior.
the intention
to vote for a
candidate?Boorman
or
Michaels?and
particular
this choice
scale
by a 7-item
multiply
certainty
al.
et
This
is
re1998).
(Wegener
two-step
process
as
a
standard
behavioral
intention
indicator
garded
and Ajzen
(Fishbein
1975; Fazio and Zanna
1978;
and
Warland
et al. 1998). For
1973; Wegener
Sample
who voted for the same candidate
at each
subjects
a
of
vote
time,
simple
regression
delayed
certainty
immediate
vote certainty
against
yielded a test/retest
coefficient
of R=.91 (p<.001);
for those subreliability
who
exhibited
an
inconsistent
vote
jects
pattern upon
of the certainty
score was also exretest,
reliability
ceptionally
high, R=.84 (p<.001).
Clearly, whether subswitched
their
vote
or
maintained
jects
support for the
The
20
score was
their first vote certainty
of their second, indicating
very little
highly predictive
These
reliabilities
random error in the measure.
high
of any significant
slope inpreclude the interpretation
to the
over time being attributed
creases or decreases
toward the mean."
artifact called "regression
statistical
the sleeper
effect prebecause
This is important
which might be merely
dicts positive
slope changes,
measures.
for unreliable
artifactual
dependent
absolute
levits high test-retest
reliability,
Despite
did
exvote certainty
vs. delayed
els of immediate
same
candidate,
over time, but only
mean increase
a significant
and delayed
whose immediate
among those subjects
This significant
increase
vote did not change (p<.001).
hibit
candidates.
for voters who switched
the same
Despite their switch, these voters maintained
The net efover time nevertheless.
level of certainty
increase among the voters who did
fect of the certainty
of choice was signot switch was that their certainty
nificantly
greater than that for voters who did switch
at the time of the delayed vote (p <.001).
was not observed
Procedure
over a four-week
Data were gathered
period, durat the technical
sessions
college.
ing day and evening
recruits were asked
Within participating
classrooms,
form. Using
a random
consent
to sign a standard
were all given a number that correstart, volunteers
condition
selected
treatment
to a randomly
sponded
to report to an assigned
and instructed
laboratory
In the lab, trained
room for their treatment.
testing
and communicafrom psychology
students
graduate
where
a structured
followed
tion programs
protocol,
could be
that no questions
were instructed
subjects
or the materials
about the questionnaire
answered
they
could occur during the testing
would see; no talking
session; and that there are no right or wrong answers.
1 of the instruSection
When subjects
completed
to read with him/
asked subjects
ment, the moderator
her the following
Setting
scenario:
the Scene for the Pat Michaels
Boorman Campaign
vs. John
All of the information you are about to see or read
has been produced for a fictional political camto actual people or true
paign. Any resemblance
names is entirely coincidence. For the purposes of
this study, please assume that you live in the 10th
district of Kentucky. You intend to
Congressional
vote on November 8 in the General Election.
The two candidates for Congress are John Boorman
and Pat Michaels. Neither of them has served before in the U.S. Congress, making this an "open"
Journal
of Advertising
race. Intentionally, no political party affiliation will be
mentioned in the information you are about to see.
This is typical of most political advertising today.
introduction
of the scenario, subjects were
Following
for a 30-second
to
their
first
message
period.
exposed
#1
In each treatment
condition,
message
exposure
of Section 2 of the instruwas followed by completion
ment. For all subjects
except the control group (who
saw only the televised
to the second
attack), exposure
of
Section
3
of
the instrumessage
plus completion
ment followed. The final section, in addition to asking
asked the quesinformation,
personal
demographic
"For
whom
would
if
vote
the
election
were
tion,
you
held today?" followed by a measure
of the certainty
of
This
is
the
that choice.
measure
of the study,
repeated
also asked during the telephone
call-backs.
Particiwas complete
pants were then told that the session
and that some of them would be called in the near
to their participation.
future
related
Two-thirds
of
for the callback
the subjects
were randomly
selected
interview
(n=225). Of these, 10 subjects were unreachcallback subsamples,
96 received
able. In the remaining
and 119 received the
treatment
the indirect defensive
direct one. In the followup
calls, the rantelephone
domly selected subgroup was asked once again, "If the
were held today between
Pat Michaels
and
election
John Boorman,
how would you vote?"
This was fol"How certain are you
lowed with the second question.
The use of the candidates'
names in
of this decision?"
the use of their names on the
the question
replicates
interviewtest ballot in the original session; additionally,
ers described the 7-unit certainty scale for respondents.
are central to the sleeper
measures
These over-time
in choice over
in this study. Changes
effect analyzed
shifts in
time were examined
by looking at individual
(the
choice, within the called back group of subjects
the excallback
period began three days following
in choice among
those
as well as shifts
periment);
time periods after excalled back at various
subjects
1) is operationalized
posure. "Early callback" (Interval
calls that occurred
as those repeat-measure
during
"Mid
the first seven days of the call-back
procedure.
callbacks"
(Interval
2) were conducted
during the second seven days of the callback period. "Late callbacks"
(Interval
3) began during the fifth week and were completed during the sixth week of the callback period.
Results
Results
research
and discussion
are organized
hypothesis.
1. The first hypothesis
Hypothesis
conditions
for observation
all specified
around
each
advanced
that
of an absolute
Winter
1999
21
study of attack
In
this context,
political
advertising.
of
a
cue
ad
the
role
the defensive
discounting
plays
that initiates
attiand the attack ad is the message
have been established
tude change.
Eight conditions
some of
from prior research.
As previously
discussed,
3
lend
these conditions
and
them(numbers
5)
1, 2,
effect
sleeper
and defensive
will be observed
in this
to being
observed
while
others
empirically
7
and
are
more
8)
(numbers
4, 6,
methodological.
and methodological
conditions
Because
all empirical
in our study, the first hypothesis
are achieved
is fully
selves
supported.
Condition
Specifically:
1 (empirically
the original
determined):
in its
(the attack
ad) must be persuasive
message
own right. The data in this study provide
multiple
indicants
of the independent
impact of the
persuasive
attack ad. First, the effect of the attack ad in isolation
over the target
of attack;
there
favors the attacker
that subgroup
was no backlash
observed.
of
Among
who saw the attack
first, or saw only the
subjects
attack in the control group, initial global attitudes
on
a -3 to +3 scale toward
the attacker,
the target
of
and the ad itself
illustrate
this
(A/atattack,
A/
sd=1.46;
sd=1.22;
tacker=-.33,
A/target=-1.29,
the fact that the ad
sd=1.88).
Thus, despite
ad=-.63,
itself was slightly disliked, as was its sponsor, the negative ad nevertheless
worked to the persuasive
benefit
of its sponsor. The difference
between
sponsor and tarwas statistically
get attitudes
significant
(p<.001).
Further evidence
of the positive persuasive
impact of
is found by examining
the attack ad in isolation
immewithin the control group.
diate vote and vote certainty
The percentage
vote within the control group strongly
in that 80% cast their vote for this
favored the attacker,
The corresponding
candidate!
attacking
certainty score
of the control group votes was 2.40, higher than all like
scores in all treatment
groups (p <.05). Thus in both
attitudinal
and behavioral
terms, the attack ad had
strong persuasive
impact in its own right.
2 (empirically
Condition
the discountdetermined):
the impact of the
ing cue must substantially
suppress
attack.
Table 1 provides
documentation
that the diad is capable of suppressing
rect defensive
the immediate impact
of the attack
the indirect
ad, whereas
in countering
defense
is less powerful
its immediate
whether
a discounting
cue subimpact. Additionally,
diminishes
the impact
of the attack
destantially
but also on its position
pends not only on its content
in relation
to the attack.
Since this study considers
Table
1
of means for both conboth,
reports patterns
tent and order. Note that there is a significant
main
effect observed
for order (p<.001),
as well as a significant interaction
between
order and message
content
of cell means
indicates
that the
(p<.01). The pattern
nature of this interaction
is characterized
by greatest
effect when the defensive
content
inhibiting
message
is a direct derogation
and when that message
follows
the attack
ad. This inhibiting
effect is significantly
than when
indirect
follow the
greater
derogations
attack (p<.05), or when either type of derogation
precedes the attack
that the initial
(p<.01).
Recalling
vote certainty
within
the control group was 2.40, a
mean
vote certainty
of-2.13
(over 4.5 units less than
the control) might be viewed as not only statistically
but also of practical
inhibiting
(p<.001)
significance.
Because
direct derogation
ads preceding
and followthe least and the
had, respectively,
ing the attack
most inhibiting
effect, these two conditions
provide a
framework
for examining
2 and 3.
Hypotheses
Condition
3 (methodologically
there
determined):
must be ample time prior to the delayed
for
measure
the attack
and defense
ads to become
dissociated.
Like previous
effect studies,
we meet this
sleeper
condition
a wide range
by our design, which possesses
of elapsed
times for delayed
measurement,
ranging
from three to 48 days; of these,
fully half occurred
more than two weeks
after the treatments.
This
of delay is greater
amount
than or equal to that provided in other sleeper
effect studies
to test
designed
the dissociation
hypothesis.
4 (empirically
Condition
dissociation
determined):
between
the attack and defense
must occur
messages
that memory
of the attack message
enough
quickly
will not have faded. This condition,
in essence,
rethat the attack have a sustained
Eviimpact.
for this can be found by examining
our control
Within
that portion
of the control
group responses.
a telephone
83.3% initially
group receiving
callback,
voted for the attacker,
Michaels.
At the time of callmaintained
this
back, 75% of this group of subjects
same voting decision,
that the attack mesconfirming
quires
dence
does have a sustained
sage, in isolation,
impact. Mean
of elapsed
number
the immediate
and
days between
delayed votes for the control group was 9.5 days. The
sustained
of the attack
within
the
impact
message
control group demonstrates
that it was not quickly
thus empirically
Condition
4.
forgotten,
supporting
Circumstantial
evidence
for this condition
being
met is also found in our choice of an attack political
advertisement
as stimulus
material.
As previously
information
discussed,
negative
gains more attention
and is better remembered
than is positive
information (Mizerski
1982; Pinkleton
1997; Till and Shimp
bias" provides
1998). This "negativity
a reasonable
basis for assuming
that a negative
will demessage
cay at a slower rate than its less remarkable
positive
The
22
How
Table 1
Immediate
Vote Certainty
(VC) is Influenced
by Content
Order of the Defensive
Advertising
Message
Journal
of Advertising
and
(n=321)
Cell Means:
Preceding
Following
Attack
Attack
(VC)1
_Defensive
Direct Derogation
Row Means
Order of Exposure
Vote Certainty
Message
Content
Indirect Derogation
.20
.46
-.07
(184)
(94)
(90)
-1.55
-2.13
-.38
(137)
(92)
(45)
-.82
-.17
(186)
(135)
Column
Means
Analysis
of Variance
Table
2-tailed
Squares
of
Degrees
Freedom
Mean
Source
Square
?
Order
Content
Interaction2
219.763
10.935
94.522
1
1
1
219.763
10.935
94.522
317
13.779
Sum of
Residual
4367.942
15.949
.794
6.860
.001
.374
.009
'Vote Certainty (VC) is defined as the algebraic product of a contrast-coded vote score (where 1=vote for assailant, 0=abstain, and -1 =vote for
target of attack) and a 7-unit global certainty of vote scale (where 1=not at all certain and 7=completely certain)?so that a larger positive
score indicates greater certainty with respect to a vote for the assailant and a larger negative score indicates greater certainty of a vote for
the target candidate.
2When a direct derogation followed the attack, vote certainty (VC) favoring the attacker was substantially suppressed. The LSD (least significant difference) test indicates that this condition exhibits a mean score significantly lower than when an indirect derogation follows the attack (p <.05) or when either type of derogation precedes the attack (p <.01 ).
the posOur design also accommodates
counterpart.
of the negative
of early decay
by
message
sibility
within
the
first
of
the
half
callbacks
fully
measuring
in the
effect is observed
13 days. Thus, if a sleeper
in the condition where
(3) follows the attack. However,
the direct defense precedes the attack, significant
additive and interactive
effects were observed (p<.05). This
means that when a direct defense precedes the attack
that the attacker
we are confident
first two weeks,
would not have yet been forgotten.
The attack
5 (empirically
Condition
determined):
rather than
ads must act independently
and defensive
an additive
would not preclude
jointly. This condition
of an
an
absence
it
would
but
on
vote,
require
impact
evaluinteraction
effect. In order to test this condition,
was measured
ation of both candidates
immediately
meato each ad. These post-exposure
after exposure
in a hierarwere included
sures and their interaction
the way that attack ad is processed.
ad, it influences
This makes the resulting
vote dependent
not only on
but also on their comeach ad considered
separately,
bined effect. Thus, it might be argued that a direct
which precedes
an attack contribdefensive
message
as one message (Eagly
utes to both ads being processed
and Chaiken
1993). Once again, the two order conditions that are in sharp contrast with one another corread treatments.
Clearly,
spond to the direct defensive
our attack and defensive
acted independently
messages
in all but one treatment
This one exception
condition.
2.
when testing
will be further examined
Hypothesis
Thus, condition 5 is satisfied.
vote certainty
in which immediate
chical regression
If an interaction
term
variable.
was the dependent
this is evidence
to be non-significant,
were observed
In three of
that the impact of each ad is independent.
in Table 1 this
described
conditions
the four treatment
was found to be true: (1) where a direct defense follows
or
the attack; (2) where the indirect defense precedes
Conditions
determined):
6,7 and 8 (methodologically
is satisfied
three conditions
through the
and protocol of our study. Condition
6?
must make note of the imporparticipants
each of these
procedures
that study
Winter
1999
23
_
in the message?was
tant argumente
satisfied
by two
of our data gathering
elements
First, prior
procedure.
to exposure
to any ad, the moderator
instructed
the
close
to
ad
to
attention
the
pay
they
participants
were about to see since they would only see it once.
each advertisement
subSecond, following
exposure,
ad
not
rated
the
the
on
and
candidates
jects
only
global
evaluative
scales,
on several
but also rated
each
advertis-
individual
attributes.
This
ing message
to
think
and
them
about
note
process required
imporof the ad. Our design also clearly
tant content elements
the discounting
7?that
satisfies
condition
cue must
of our
be received
after the message?as
half
fully
the
direct
defensive
treatments
place
advertising
after the attack. Similarly,
8 specicondition
message
must rate the trustworthiness
of
fies that subjects
to the discountafter exposure
the source immediately
comad), which our subjects
ing cue (the defensive
the
after
both
and
the
for
opponent
pleted
sponsor
each message
exposure.
The empirical
evidence
and the procedures
that
in our design
the
conclusion
were followed
support
1 is supported.
That is, all empirical
that Hypothesis
conditions
identified
and methodological
previously
as requisites
are satisfied
for observing
an absolute
effect
sleeper
in our political
advertising
experiment.
2. Perceived
source credibility
and order
Hypothesis
cue will be related
of exposure
to
to the discounting
An
effect.
absolute
observation
of an absolute
sleeper
when the credibility
effect is expected
of the
sleeper
ad follows
is initially
low and the defensive
attacker
2 reports
a repeated
Table
measures
the attack.
in
which
source
and
MANOVA
credibility
to the discounting
cue are considered
order-of-exposure
of
and deimmediate
as between-subjects
predictors
2 are
Of
to
vote
pertinence
Hypothesis
layed
certainty.
main effects of order of exposure (p=.013)
the significant
and source credibility (p=.002) as well as the significant
effects between these two factors and change
interaction
in vote certainty (respectively
p=.006 and p=.011) between
is the
immediate
and delayed measures.
Also of interest
order of
interaction
between
absence
of a significant
that these
and source credibility,
suggesting
exposure
vote certainty
factors influence
two between-subjects
manner.
in a relatively
This is further
independent
a
the
of
interaction
absence
three-way
supported
by
order
and
source
of
involving
exposure
credibility.
The two main effects of order and credibility
can be
lower
vote
is
described
associated
simply:
certainty
in which the discounting
with the treatment
cue follows the attack
and when source credibility
is relalower
attacker
of
for
the
versus
the
attack.
tively
target
Recall that a low score on the vote certainty
scale may
be interpreted
as indicating
a relatively
certain vote for
the target of attack. This means that a defensive
mesfollows
which
the
attack
or
which
calls
into
sage
question the credibility
of the attacker
the imsuppresses
the target.
pact of the attack ad, thus benefitting
1
statistics
that expliFigure
provides
descriptive
cate the interaction
effects.
it
Further,
reports shifts
in actual
of votes for the attacker.
Note that
percent
in
increases
both
and
vote persignificant
certainty
cent within the cells characterized
low
initial
source
by
were
observed.
When
source
was
credibility
credibility
cue followed
the attack, the
low, and the discounting
most
dramatic
increases
were observed.
While Hy2
with
dealt
of
these sleeper
efpothesis
prediction
that when the discounting
fects, it is worth noting
cue preceded
the attack
and source credibility
was
of
for
votes
the
attacker
dehigh, percent
initially
a
in
for
the
attacker
clined, indicating
decay
support
in this one condition.
evidence
of the signifiFigure 2 presents
graphical
cant interaction
between order-of-exposure
and change
in vote certainty,
and the significant
of
interaction
source
and change
in vote certainty.
Note
is
effect
indicated
for
the
sleeper
treatment.
This is evidenced
by the
of
the
significant,
positive slope
within-subjects
change
in vote certainty
when it was measured
immediately
credibility
an absolute
attack"
"following
that
and again after delay. In
message
following
exposure
note
that
when
the
cue precontrast,
discounting
no significant
ceded the attack,
difference
was observed between
the immediate
and delayed
vote certainty scores.
Also demonstrated
in Figure
2 is the pattern
of
of perceived
interaction
source credibility
and change
in vote certainty.
This interaction
effect is similar to
that just described.
of this similarity
Because
one
the
that
think
treatment
and
might
order-of-exposure
the blocking
factor of perceived
source credibility
are
the evidence in Table
Recall, however,
interdependent.
2 indicates
that the treatment
and blocking
factors
have statistically
effects.
Also, a chi
independent
the
treatment
and
square analysis
relating
blocking
factors
indicates
that their frequencies
are statistiof each other, justifying
their sepacally independent
rate consideration.
Note that when relative
credibilof
attacker
the
is
an
absolute
low,
ity
initially
sleeper
effect is observed.
This is evidenced
by the significant
shift in vote certainty
for that group which
positive
had
the
attacker
as having
low
initially
perceived
relative
when
to
the
credibility
compared
target of
attack.
In contrast,
when
initial
source
credibility
favors the attacker,
the shift in vote certainty
between
immediate
and delayed
is minimal.
24
The
Journal
of Advertising
CM
Q fe I
5-s
t- CO t- CD
?- s> h* y
rsl r^"co
I
c ?&
f o aw
<3 ex
SS
5
8?
co io CM3" CM
cm co
O) IO CO t- TCO SCO
? *? i =
?
"9.w
^"<d?~7>?sF
f oiS > c ? P
S
"O
F
lisi
F
?
??g
?
(?
??
C
ce
?
< f
F >
> ?
CO
c
f
CM?
CO
o. ? g-d
s'-s S 8>
5S
SIS
d
.o co
?SS-S
fi
?
5|
O ?)
,2 -^ <o E
"S S ||
ft s=*.^ co
c
XXX
?
?
O o
>f
CL
^13
f
??
o
t- <2 ?
SS*
S
CO5lg ??
S?
?S*
.a f om ?II
F >
JO
o
O
g
F
D>
f-CO
c
f
o
f
0.
f
o
>
f.?
o
Ils
IL COw
?fi
?S S
? "?
e
.?
3 $5
2 *
a*
? e
Ijj O
?) ^ CMrlOt-ON
CD O O O
1
e
S
?
evi
o
F
O CO COT"
CM
" CMCOCO
CD ? CO
c
IS
F
"?
O
I
?
as
f
815
CD CO?- O) ?fr
p in ?? io co
CO COCOCO O?
tt- CO IT)
?
LU
si**
o 5 f *"
m .S COC C
I ? ? -S r
Si
8 5
C H- ^ F
a *- i= <x>f
? "c?? .S ?
f f
lega)
*
3
?- f
c .s
T? x F
f
3 s w
?- ? f
to & fc
? 3 ? ? g
2 S S i I
.LU
o
.3? 8
sa
w
?
?
?i
COCQ
ijf|i
u. f J) f f
?
f
?
?
?^ ??
^ ? <?>
? ? CO?
li
> f
a.
f_ 5 -S ?
? ? "2 42 ??
t- 3 f .F (?
5 *- M 3 (
CO
r
f
o
f
o
>
c/>
?
?
co
o:
' t? >?
o
O)
~
p ?xs. ?
|??5??
00
Q
F g o c5
F
? ?
~a
^
'-3
"S
<u
?
c '-3
CGC
*
*
* * -?
Winter
1999
Order-Driven
25
and
Credibility-Driven
Sleeper
Figure
Effects:
2
An Analysis
of Two
Interactions1
Independent
(n-119)
+3p
+3p
Discounting Cue
Before Attack (n=62)
F
.co
?
s)
i?
Discounting Cue
After Attack (n=57)
+2-
00
>
+1-
+1?
??
o
co
o
(0
c
??
(0
?c
f
O
F
???
Low Relative Source
Credibility (n=56)
+2-
CO
>
f
5
High Relative Source
Credibility (n=63)
0-
5
c
^
co
-1-
.76
.71
0.00
0-
-1-
f
O
-2
F
-1.95
-2'-2.18
I
-3J
Immediate
-3J
Delayed
53.2%
Delayed
54.8%
.52.4%
Dw
47.4%
Shift
in Vote
for
Percentage
Shift
in Vote
Attacker
Order-Driven
Interaction
(oe
50.0%
?
vc)
for
Percentage
Attacker
Interaction
Credibiiity-Driven
(sc
?
vc)
1The order-driven interaction (oe ? vc) is significant at p<.01, and the credibility-driven interaction (sc ? vc) is significant at p<.05. These two
interactions are statistically independent: no significant (oe ? sc) or three-way interaction was observed (see Table 2).
2A simple-effects analysis of the order-driven interaction shows an absolute sleeper effect in both vote certainty (p<.001) and In the attacker's
vote percentage (p<.01) only when the direct discounting cue follows the attack.
3A simple-effects analysis of the credibility-driven interaction shows an absolute sleeper effect in both vote certainty (p<.001) and in the
attacker's vote percentage (p<.001) only when relative credibility of the attacker is initiallylow.
The
26
Journal
of Advertising
in Table 2 and in Figures
1
The evidence
(presented
of an absolute
the presence
and 2) documents
sleeper
that is, when
effect under the specified
conditions;
cue follows the attack and when the
the discounting
is initially
low.
of the attacker
credibility
perceived
in
absolute
effect
the
absence
of
an
Further,
sleeper
has initial strong impact favoring its sponadvertising
for a considerable
sor, but that this impact is sustained
of
time.
Given
the
most
active phase of a
that
period
cue preceding
the contrasting
conditions
(discounting
of the atand high perceived
the attack
credibility
documents
research
that
tacker)
supports
previous
of
of each of these as determinants
the importance
2 is strongly
the sleeper effect. Hypothesis
supported.
3. The sleeper
effect, when observed,
Hypothesis
more powerful
over time. Figure 3 conwill become
suppressed
by a direct defensive
advertising
message,
or when it is suppressed
low credibility
of
by perceived
the attacking
these
defensive
or
candidate,
strategies
source effects are merely temporary.
negative
from this study
also give creStrategic
findings
dence to the unswerving
of
adveradvocacy
negative
consultants.
have
tising by political
They
long argued
that "negative
(Totten
advertising
works...period."
1991). Now they can argue that it also has lasting
of subjects whose response
that subsample
patan absolute
exhibited
effect, either desleeper
or by initially
low pertermined
by order-of-exposure
or by both. For these sleeper
ceived source credibility
and delayed vote certainty,
immediate
effect subjects,
siders
terns
actual percent of vote for the attacker,
into
three periods
of elapsed
time beis partitioned
occurred.
Note
that
the
stronfore the delayed callback
for
the
middle
time
effect occurs
period,
gest sleeper
This
and tapers off for the latest callback
subgroup.
effect grew stronger
means
that the sleeper
during
of elapsed
the first and second intervals
time followand then was observed
exposure,
ing initial message
With
to exhibit
moderate
decay.
respect to the actual
of
this
a
votes,
statistically
signifirepresented
percent
vote for
cant shift such that the highest
percentage
Michaels
was significantly
higher during the second
as well as their
than it was for the first one (73.1% vs. 31.7%).
within callback
a significant
intervals,
Further,
posifor
the
first
two
tive shift was observed
(the early and
This was not the case for the latest
mid callbacks).
a positive but nonwhich exhibited
callback
interval,
interval
3 is partially
shift. Thus, Hypothesis
supsignificant
increase in percent of votes
ported in that a substantial
for the attacker
was observed,
up to three weeks folIn
fact, the vote for Michaels
exposure!
lowing message
the
mid-callback
time period could
(the attacker), during
landslide
have
reached
be said to
proportions.
Discussion
We began this study motivated
advertising
insight into political
by the desire to test alternative
by a desire to gain
rather than
strategy
theoretical
explana-
We recognize,
tions of the sleeper effect phenomenon.
have far-reaching
that our findings
implihowever,
and theorists.
cations for both strategists
Our findings support the use
Strategic
Implications.
in a political campaign.
of negative
attack advertising
that negative
attack
Not only do we have evidence
is of fairly short duration, these suspolitical campaign
tained effects of negative
attack are particupolitical
salient.
Even
the
when
larly
impact of the attack is
of the target of attack,
impact. From the perspective
is
not
it can
negative
advertising
only damaging,
wreak havoc that lasts until election
day.
So what should a candidate
who has been attacked
do to counter the damage? A direct defense,
issued as
after
the attack
as possible,
should
be
quickly
mounted.
It must be realized,
that the suphowever,
effect of this defense
is likely to be shortpressive
lived. This suggests
that reinstatement
of the debe unfense, through
frequently
repeated
messages,
dertaken
election
and
that
these
defenses
through
day
use explicit rebuttal
that call into question
strategies
the credibility
of the attacker.
A savvy political
consultant may also advise the target of attack to use his
or her own attack, recognizing
the inherently
stronof
information.
ger impact
negative
The assailant
should withdraw
from attacking
sufto
so
that
the suppression
election
ficiently
prior
day
that results from low perceived
has time to
credibility
the target of attack should
dissipate.
Concurrently,
be prepared
to respond
to those attackers
who untheir attack into the final hours.
wisely continue
Theoretical
not our original
Implications.
Although
intent, the findings from this study provide substantial
under which a sleeper effect
insight into the conditions
is likely to be observed, the determinants
of the sleeper
effect phenomenon,
and how long it is likely to last.
With respect to the conditions
under which a sleeper
effect is observed,
we organized
research which previconditions
established
ously explored
by classifying
them into two groups:
those that require empirical
observation
and those that are inherent
to the design
or data-gathering
This
procedure.
categorization
should be useful to future researchers
to obseeking
serve a sleeper
effect. We further
that
recognized
these conditions
are not necessarily
exclumutually
sive,
nor
findings
are they necessarily
interdependent.
broaden rather than narrow the possible
Our
situ-
1999
Winter
Absolute
27
Sleeper
Effects
Observed
Figure 3
Time
at Different
Intervals
Following
Message
Exposure1
(n=86)
+3
?
48
days
^^
"?
Interval 1 (n=41)
?^f
'
""
O
O
Interval
Interval 2 (n=26)
Interval 3 (n=19)
2
73.1%
23.1%
to-
Interval
3
31.6%
v47.4%
^o
W
Shift
in Vote
for
Percentage
Attacker3
11nterval1 measurements took place duringthe firstweek of the callback period, median=8.89 days followingmessage exposure. Interval2 callbacks
occurred duringthe second week, median=14.13 days. Interval3 callbacks took place duringthe fifthand sixth weeks, median=34.80 days.
2Differencebetween Interval1 immediate and delayed vote certaintyis significantlydifferent(p<.05); and Interval2 immediate vs. delayed vote certaintyIs
significantlydifferent(p<.001). Difference between Interval3 vote certaintymeasures not significant.Immediatevote certaintyscores do not differ.
3Tne 14.6 percentage point increase in attacker'svote for Interval1 is significantat p<.05 (one-tailed).The 50.0 point increase for Interval2 is significantat
p<.001. The 15.8 point increase for Interval3 is not significant.Delayed vote percentage for Interval2 is significantlyhigher than that for Interval1
(p<.001) and for Interval3 (p<.05, one-tailed). Immediatevote percentages do not differ.
The
28
effect may be observed,
in which a sleeper
by
that each of the
the requirement
calling into question
must be met.
conditions
on
Dual-Determinants
Early literature
Hypothesis.
the role of low source
effect emphasized
the sleeper
as a cause of the sleeper effect, while more
credibility
facsituational
has stressed
recent research
multiple
of which is order of exposure
tors, the most important
we idenour testing
cue. Through
to the discounting
ations
tified
strong
for both of these determinants.
support
a significant
absolute sleeper
we observed
the atinitially
perceived
participants
Specifically,
effect when
at the time of exposure
tacker to have low credibility
with
ad. This is perfectly
to the negative
compatible
studies
that dominated
the dissociation
early sleeper
A second
and independent
effect research.
sleeper
ad folwhen a direct defensive
effect was observed
attack.
This finding is consistent
lowed the negative
of the sleeper
with the differential
decay explanation
at least to the extent that those
effect phenomenon,
decay explanation
posit this
inspired by the differential
condition
for its
order as a requisite
specific message
of
observation.
The fact that these dual determinants
of
the sleeper effect were observed to act independently
that there
each other calls into question the assumption
is a single explanation
for the phenomenon.
Rather,
for the sleeper efboth prior theoretical
explanations
fect may be concurrently
valid; they
with nor contradictory
incompatible
our design and response
Because
that
sistent with all prior conditions
the
for observing
fied as necessary
need
to be neither
of one another.
are conpatterns
have been specieffect, we
sleeper
these conditions
are now able to assess how essential
that
are. Our dual, independent
suggest
explanations
is not an essential
condition
low perceived
credibility
cue
for the sleeper effect to occur, nor is a discounting
a necessary
condithe persuasive
message
following
These
increase
the
tion for its occurrence.
findings
effect phenomenon
of the sleeper
relevance
by exin which it may be obthe possible
settings
panding
narrow
it from some previously
served, and releasing
as a
the order requirement
By dismissing
the
Condition
7)
(see
sleeper
necessary
prerequisite
in more naturaleffect phenomenon
may be observed
strucand is not dependent
on highly
istic settings
tured experimental
protocols.
that only
it was previously
For example,
thought
could
cue followed the message
when the discounting
was invalieffect occur. This assumption
the sleeper
of the sleeper effect independated by our observation
of low
order. Similarly,
dent of message
perceptions
to be
that were previously
source credibility
thought
of the sleeper effect, are not
for observation
essential
confines.
Journal
of Advertising
a sleeper effect was observed as a function of
necessary;
or not the source was iniorder of exposure,
whether
as possessing
low credibility.
Order of
tially evaluated
exposure and source credibility may be additive in their
effect. Our findings support this notion through evidence
that the strongest sleeper effect occurred for subjects who
cue after the attack and
were exposed to the discounting
who also attributed low credibility to the attacker.
Assessment.
methMethodological
Any experimental
as well as
odology raises issues of internal
validity,
issues associated
with transfer
of results beyond the
With our true experimental
deto internal
given priority
validity,
yet
has certain
our design
with rearguably
strengths
and ecological
spect to external
validity.
experimental
sign we have
setting.
Internal
Our successful
random
Validity.
assignment procedure,
evidenced
of
by initial comparability
controls most threats
to internal
subgroups,
validity.
our careful
administration
of the design
Further,
rehearsed
identical
(trained
moderators,
protocols,
our confidence
that internal
etc.) enhances
validity
was preserved
As previously
during implementation.
the high reliability
of the dependent
variable
factor effects as artiprecludes
interpreting
blocking
thus increasing
our confidence
that credibilfactual,
of the sleeper effect.
ity is in fact a true determinant
External and Ecological
Validity. Can we state when
noted,
a sleeper
a reliable
will occur? Previous
documentation
of
on highly strucsleeper effect has depended
tured
and carefully
controlled
laboratory
settings
treatments
Greenwald
and Leippe 1988).
(Pratkanis,
While we also utilized
an experimental
methodology,
our findings
are based on real-world
stimuli
and a
albeit
in which
the ordering
of
scenario,
fictional,
is very likely to occur. The real world
treatments
attack and defense
adver"give and take" of political
tisements
an ideal context
that simulates
provides
the operations
used in experimental
setpreviously
an absolute
tings to document
sleeper effect. Further
the applicability
of our findings
to a
strengthening
naturalistic
of an absosetting is that the observation
effect
lute sleeper effect didn't depend on the strict method/
order manipulation,
but was also achieved
indepenthe relative
credibility
dently through
perceptions.
in the decision
Arguably,
process leading to a vote,
it is likely that the relative
of both candicredibility
dates will be assessed
Those
during the campaign.
who use a negative
attack ad are setting
up conditions in which low levels of credibility
are likely to be
at least initially.
In the real world, a sleeper
observed,
effect
is very likely to occur when a negative
attack is
made. For voters who react with disdain toward candidates who attack (whether
or not a defensive
message
Winter
1999
29
follows), a sleeper effect is likely to occur. That is, the
attack increases.
over time impact of the negative
which examIn some of our own previous
research,
but not over time impact of negative
ined immediate
candidates
of the potenwe have advised
attacks,
but
effects
of
advertising,
negative
tially
powerful
and Weaver Lariscy
also warned of its risks (Tinkham
we con1997). After this study of over-time
impact,
clude,
risky
done
albeit unhappily,
only in the short
to the
that
are
negative
messages
initial
that
term;
any
damage
an effective
deeither through
attacker,
mounted
by the target or by initial low credibilwill
from
voters who do not like attackers,
ity ratings
is but temporal.
The
not last. Damage
to an attacker
candidate
can be damaged
is
primary way an attacking
if he or she launches
a "final hour" attack, immediately
wherein
some voters will
prior to the actual election,
fense
the attacker negatively
and this shortinitially evaluate
whether
or
term effect may not have time to dissipate,
it
the
victim.
not any defense is launched
by
against
In summary,
we offer several
simple statements:
If an attack
stands
alone,
unanswered,
it is
effective.
If an attack
still
mately,
is refuted,
be effective.
the
attack
will,
ulti-
If an attacker
his or her own creddamages
itself will still have
ibility, the attack message
positive
impact over time.
If attacked,
the best defense
for the
therefore,
victim seems to be a strong, swift offense.
Reference
Advertising Age (26 October 1998), "Across the U.S., Politicians
Are On the Attack," 69(43), 3; 53.
Allen, Mike and James B. Stiff (1989), "Testing Three Models for
the Sleeper Effect," Western Journal of Speech Communication, 53 (fall), 411-426.
Anderson, Norman (1971), "Integration Theory and Attitude
Change," Psychological Review, 78, 171-206.
Andrews, Edmund (1991), "F.C.C. Seeks to Strengthen Rules on
Negative Political Advertising," New York Times, June 14,
Section D, 18.
Ansolabehere, Stephen and Shanto Iyengar (1995), Going Negative: How Attack Ads Shrink and Polarize the Electorate, New
York: Free Press.
Batra, Rajeev, John G. Myers and David Aaker (1996), Advertising
Management (5th Ed.) Upper Saddle River, New Jersey:
Prentice Hall.
Burgoon, Michael, J. D. Riese, M. Butler, J. Montgomery, C. L.
Stinnett, W. D. Miller, M. K. Long, M. Vaughn and B. Caine
(1976), "Propensity of Persuasive Attack and Intensity of Pretreatment Messages as Predictors of Resistance Persuasion,"
Journal of Psychology, 92, 123-129.
Call, John A. (1996), "Making the Research Work for You," Trial
(Jury Persuasion Issue), 32(April), 20-27.
Cook, Thomas D., Charles L. Gruder, Karen M. Hennigan and
Brian R. Flay (1979), "History of the Sleeper Effect: Some
Logical Pitfalls in Accepting the Null Hypothesis," Psychological Bulletin, 86(4), 662-679.
Cook, Thomas P. and Brian R. Flay (1978), "The Temporal Persistence of Experimentally Induced Attitude Change," In L.
Berkowitz, ed., Advances in Experimental Social Psychology,
(Vol. 11), New York: Academic Press.
Debevec, Kathleen and Easwar Iyer (1986), "The Influences of
Spokespersons in Altering a Product's Gender Image," Journal of Advertising, 15(4), 12-20.
Eagly, Alice H. and S. Chaiken (1975), "An Attribution Analysis of
the Effect of Communicator Characteristics on Opinion
Change: The Case of Communicator Attractiveness," Journal
of Personality and Social Psychology, 32, 136-144.
The Psychology of Attitudes,
and-(1993),
Fort Worth, TX: Harcourt Brace Javanovich.
Elliott, Stuart (1996), "Advertising Agencies Make Pitch to Politicians and Consultants, Urging Them to Clean Up Their Act,"
New York Times, April 29, Section D, 27.
Fazio, R.H. and M.P. Zanna (1978), "On the Predictive Validity of
Attitudes: The Role of Direct Experience and Confidence,"
Journal of Personality, 46, 228-243.
Fishbein, Martin and leek Ajzen (1975), Belief, Attitudes, Intention
and Behavior, Reading, MA: Addison-Wesley Publishing Co.
Freedland, Jonathon (1994), "The Great Democratic Turnoff,"
Guardian, October 29, Section 2, 26.
Gillig, P.M. and A.G. Greenwald (1974), "Is It Time to Lay the
Sleeper Effect to Rest?" Journal of Personality and Social
Psychology, 29, 132-139.
Gotlieb, Jerry and Dan Sarei (1991), "Comparative Advertising
Effectiveness: The Role of Involvement and Source Credibility," Journal of Advertising, 20(1), 38-45.
Greenberg, Bradley and Gerald Miller (1966), "The Effect of LowCredible Sources on Message Acceptance," Speech Monographs,
33, 127-136.
Gruder, Charles L., Thomas D. Cook, Karen M. Hennigan, Brian
R. Flay, Cynthia Alessis and Jerome Halamaj (1978), "Empirical Tests of the Absolute Sleeper Effect Predicted From
the Discounting Cue Hypothesis," Journal of Personality and
Social Psychology, 36(10), 1061-1075.
Haley, Eric (1996), "Exploring the Construct of Organization as
Source: Consumers' Understanding of Organizational Sponsorship of Advocacy Advertising," Journal of Advertising, (Summer) 25(2), 19-35.
Hill, Ronald P. (1989), "An Exploration of Voter Responses to Political Advertisements," Journal of Advertising, 18(4), 14-22.
Hovland, Carl I. and Wallace Mandell (1957), "Is There a ILaw of Primacy' in Persuasion?" in The Order of Presentation in Persuasion,
Carl I. Hovland, ed., New Haven, CT: Yale University Press, 1-22.
, Irving Janis, Harold Kelly and Walter Weiss (1953),
"The Influence of Source Credibility on Communication Effectiveness," Public Opinion Quarterly, 15(4), 635-650.
and W. Weiss (1951), The Influence of Source Credibility on Communication Effectiveness," Public Opinion Quarterly, 15, 635-650.
, AA. Lumsdaine and F.D. Sheffield (1949), Experiments on Mass Communication, Princeton, NJ: Princeton University Press.
Husek, T.R. (1965), "Persuasive Impacts on Early, Late, or No
Mention of a Negative Source," Journal of Personality and
Social Psychology, 2, 125-128.
30
Ito, Tiffany ?., Jeff T. Larsen and ?. Kyle Smith (1998), "Negative
Information Weighs More Heavily on the Brain: The Negativity Bias in Evaluative Categorizations," Journal of Personality and Social Psychology, 75(4), 887-900.
Jamieson, Kathleen H. (1992), Dirty Politics: Deception, Distraction, Democracy, New York: Oxford University Press.
and Karlyn K. Campbell (1992), The Interplay of
Influence: News, Advertising, Politics and Mass Media,
Belmont, CA: Wadsworth, Inc.
Johnson, H.H. and T. Watkins (1971), "Trends in Dissonance Research and the Chapanis Criticisms," Personality and Social
Psychology Bulletin, 3, 244-247.
Karrh, James A. and David Halpern (1997), "Nothing to Lose?
Assessing the Impact of Competitive Position on Responses to
Negative Political Advertising," in M. Carole Macklin (ed.),
Proceedings of the 1997 Conference of the American Academy
of Advertising, 23-36.
Kellerman, Kathy (1984), The Negativity Effect and Its Implications
for Initial Interaction," Communication Monographs, 51, 37-55.
(1989), The Negativity Effect in Interaction: It's
All in Your Point of View," Human Communication Research,
16(2), 147-183.
Kelman, H.C. and Carl Hovland (1953), "Reinstatement of the
Communicator in Delayed Measurement of Opinion Change,"
Journal of Abnormal and Social Psychology, 48, 327-335.
Lang, A. (1991), "Emotion, Formal Features and Memory for Televised Political Advertisements," in F. Biocca, ed., Television
and Political Advertising Vol. 1: Psychological Processes,
Hillsdale, NJ: Lawrence Erlbaum Associates, 221-243.
Lang, Annie, Kulijinder Dhillon and Quingwen Dong (1995), The
Effects of Emotional Arousal and Valence on Television Viewers' Cognitive Capacity and Memory," Journal of Broadcasting & Electronic Media, 39(3), 313-327.
Lau, Richard R. (1985), Two Explanations for Negativity Effects
in Political Behavior," American Journal of Political Science,
29(1), 119-138.
(1982), "Negativity in Political Perceptions," Political Behavior, 4, 353-357.
Lipman, Joanne (1991), "Agencies to Attack Negative Political
Ads," Wall Street Journal, May 17, Section B, 4.
Littlejohn, Stephen (1971), "A Bibliography of Studies Related to
Variables of Source Credibility," Bibliographic Annual in
Speech Communication, 2, 1-40.
McGuire, William J. (1961), The Effectiveness of Supportive and
Refutational Defenses in Immunizing and Restoring Beliefs
Against Persuasion," Sociometry, 24, 184-197.
and D. Papageorgis (1961), The Relative Efficacy
of Various Types of Prior Belief-Defense in Producing Immunity Against Persuasion," Journal of Abnormal and Social
Psychology, 62, 327-337.
Mizerski, Richard W. (1982), "An Attribution Explanation of the
Disproportionate Influence of Unfavorable Information," Journal of Consumer Research, 9 (December) 301-310.
Moore, D.L. and J.W. Hutchinson (1985), The Influence of Affective
Reactions to Advertising: Direct and Indirect Mechanism of Attitude Change," in L. F. Alwitt & A. A. Mitchell, eds., Psychological
Processes and Advertising Effect: Theory, Research, and Applications, Hillsdale, NJ: Lawrence Erlbaum Associates.
Nelson, No?lle (1995), "Jury Inoculation: Acknowledging the Weaknesses in Your Case Can Take the Sting Out of an Opponent's
Attack," Los Angeles Daily Journal, Sept. 29, Vol. 108(186),
SI, col. 2.
Newsweek (1996), The Limits of Negativity," September 23, p. 42.
Peterson, R.C. and L. Thurstone (1933; 1970), Motion Pictures and
the Social Attitudes of Children. New York: Arno Press.
The
Journal
of Advertising
Petty, Richard E. and Jon A. Krosnick (1998), eds., Attitude
Strength: Antecedents and Consequences, Mahwah, NJ:
Lawrence Erlbaum Associates.
and John T. Cacioppo (1984), "Source Factors and
the Elaboration Likelihood Model of Persuasion," in Advances
in Consumer Research, 11, T.C. Kinnear, ed., Ann Arbor, MI:
Association for Consumer Research, 668-672.
Attitudes and Persuasion:
and-(1981),
Classic and Contemporary Approaches, Dubuque, LA: Wm. C,
Brown Company.
Pfau, Michael and Michael Burgoon (1988), "Inoculation in Political Campaign Communication," Human Communication Research, 15(1), 91-111.
, Roxanne Parrott and Bridget Lindquist (1992),
"An Expectancy Theory Explanation of the Effectiveness of
Political Attack Television Spots: A Case Study," Journal of
Applied Communication Research, 20(3), 235-253.
Pinkleton, Bruce (1997), The Effects of Negative Comparative Political
Advertising on Candidate Evaluations and Advertisement Evaluations: An Explanation," Journal of Advertising, 26,19-29.
Pratkanis, Anthony R., Anthony Greenwald and Michael R. Leippe
(1988), "In Search of Reliable Persuasion Effects: The Sleeper
Effect is Dead; Long Live the Sleeper Effect," Journal of Personality and Social Psychology, 54, 203-218.
Priester, J.R. and Richard E. Petty (1995), "Source Attribution and
Persuasion: Perceived Honesty as a Determinant of Message
Scrutiny,"Personality / Social Psychology Bulletin, 21, 637-54.
Rothenberg, Randall (1990), "Voters Complain Negative Ads Are Driving Them Away," New York Times, November 6, Section B, 9.
Sample, J. and R. Warland (1973), "Attitudes and the Prediction of
Behavior," Social Forces, 51, 292-304.
Schulman and Worrall (1970), "Salience Patterns, Source Credibility
and the Sleeper Effect,"Public Opinion Quarterly, 34, 371-382.
Stern, Barbara (1994), "A Revised Communication Model for Advertising: Multiple Dimensions of the Source, the Message,
and the Recipient," Journal of Advertising, 23(2), 5-16.
Teepen, Tom (1995), Toxic Rhetoric Poisons America," Atlanta
Journal / Constitution, 3 (January), Section A, 9.
Television / Radio Age (1988), "Gauging TVs Influence On Voters:
Debates, Negative Ads, and Personality," June 13, p. 34.
Till, Brian D. and Terence Shimp (1998), "Endorsers in Advertising: The Case of Negative Celebrity Information," Journal of
Advertising, 27(Spring).
Tinkham, Spencer F. and Ruth Ann Weaver Lariscy (1994), "Ethical
Judgments of Political Television Commercials as Predictors of
Attitude Toward the Ad," Journal of Advertising 23(3), 1-15.
The Role of Advertising Atand-(1997),
tention and Political Involvement in Assessing the Cognitive
and Affective Impact of Negative Political Advertising," in M.
Carole Macklin, ed., Proceedings of the 1997 Conference of the
American Academy of Advertising, 22.
Totten, Greg (1991), Personal Interview, March 19, Strubble-Totten Political Consultants, Washington, D.C. Washington Journal , 1997, Broadcast interview, C-Span.
Wegener, Duane T., John Downing, Jon Krosnick and Richard
Petty (1998), "Measures and Manipulations of Strength-Related Properties of Attitudes: Current Practice and Future
Directions," in Richard Petty & Jon Krosnick, eds., Attitude
Mahwah, NJ:
Strength: Antecedents and Consequences,
Lawrence Erlbaum Associates.
West, Darrell M. (1992), "Reforming Campaign Ads," PS 25, March,
74-77.
Will George (1986), "In the Hot Stove League," Newsweek, January 20.