Cognitive Processes Underlying Context Effects in Attitude Measurement

Abstract

We begin this article with the assumption that attitudes are best understood as structures in long-term memory, and we look at the implications of this view for the response process in attitude surveys. More specifically, we assert that an answer to an attitude question is the product of a four-stage process. Respondents first interpret the attitude question, determining what attitude the question is about. They then retrieve relevant beliefs and feelings. Next, they apply these beliefs and feelings in rendering the appropriate judgment. Finally, they use this judgment to select a response. All four of the component processes can be affected by prior items. The prior items can provide a framework for interpreting later questions and can also make some responses appear to be redundant with earlier answers. The prior items can prime some beliefs, making them more accessible to the retrieval process. The prior items can suggest a norm or standard of comparison for making the judgment. Finally, the prior items can create consistency pressures or pressures to appear moderate. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

Psychological
Bulletin
1988,
Vol. 103,
No.
3,
299-314
Copyright
1988
by the
American
Psychological
Association, Inc.
0033-2909/88/S00.75
Cognitive Processes Underlying Context
Effects
in
Attitude Measurement
Roger
Tourangeau
and
Kenneth
A.
Rasinski
NORC,
Chicago,
Illinois
We
begin this article
with
the
assumption that attitudes
are
best understood
as
structures
in
long-
term memory,
and we
look
at the
implications
of
this
view
for the
response process
in
attitude
surveys.
More
specifically,
we
assert that
an
answer
to an
attitude question
is the
product
of a
four-
stage
process. Respondents
first
interpret
the
attitude question, determining what attitude
the
ques-
tion
is
about. They then retrieve relevant
beliefs
and
feelings.
Next, they apply these
beliefs
and
feelings
in
rendering
the
appropriate judgment. Finally, they
use
this judgment
to
select
a
response.
All
four
of the
component processes
can be
affected
by
prior
items.
The
prior items
can
provide
a
framework
for
interpreting later questions
and can
also make some responses appear
to be
redundant
with
earlier answers.
The
prior items
can
prime some beliefs, making them more accessible
to the
retrieval
process.
The
prior items
can
suggest
a
norm
or
standard
of
comparison
for
making
the
judgment. Finally,
the
prior
items
can
create consistency pressures
or
pressures
to
appear moderate.
Because
of the
multiple processes involved, context
effects
are
difficult
to
predict
and
sometimes
difficult
to
replicate.
We
attempt
to
sort
out
when
context
is
likely
to
affect
later responses
and
include
a
list
of the
variables that
affect
the
size
and
direction
of the
effects
of
context.
Within social psychology, there
is an
emerging consensus that
attitudes
are
best understood
as
structures that reside
in
long-
term memory (Fazio, Sanbonmatsu, Powell,
&
Kardes,
1986;
Fazio
&
Williams, 1986; Tourangeau, 1984, 1986, 1987; Tour-
angeau
&
Rasinski,
1986;
Wyer&Hartwick,
1984)
and
are
acti-
vated
when
the
issue
or
object
of the
attitude
is
encountered
(Fazio
et
al.,
1986; Fazio
&
Williams, 1986).
The
conventions
that
have
been
found
useful
for
representing other information
in
long-term memory ought
to be
useful
for
representing atti-
tudes
as
well.
In
our own
work,
we
have
found
it
useful
to
repre-
sent
attitudes
as
networks
of
interrelated
beliefs.
Although
we
refer
to the
constituents
of
attitudes
as
beliefs,
we use
this term
loosely
to
encompass memories
of
specific
experiences, general
propositions,
images,
and
feelings.
J.
Anderson
(1983)
and
Bower
(1981)
have
shown
how the
associative-network
formalism
can
also
be
used
to
represent
such
nonproposilional
material
as
images
or
feelings.
Along
these
lines, Sears, Huddy,
and
Schaffer
(1986) proposed
a
struc-
tural model
of
political attitudes that stresses
the
importance
of
such
feelings
in
response
to
political issues. They argued that
responses
to
certain issues
reflect
deep-seated
affect
linked
to
political
symbols. Responses
to an
attitude
item,
thus, depend
An
earlier version
of
this
article
was
presented
at the
41st
Annual
Conference
of the
American Association
for
Public Opinion Research
in
St.
Petersburg, Florida.
We
received support during
the
preparation
of
this article
from
National Science Foundation Grants
SES-8411970
and
SES-8521313,
awarded
to
Roger Tourangeau.
Robert
Abelson,
Norman
Bradburn,
Susan Campbell,
Roy
D'An-
drade,
and
Reid
Hastie
made valuable suggestions
on
this article that
are
incorporated
here.
Correspondence concerning this article should
be
addressed
to
Roger
Tourangeau,
NORC,
1155
East
60th Street, Chicago,
minois60637.
on
the
symbols
it
evokes
and the
affect
attached
to
these sym-
bols.
Other researchers
have
argued
that
attitudes
are
organized
into
schemata (Fiske
&
Dyer,
1985;
Fiske
&
Kinder,
1981;
Has-
tie, 1981)
or
stereotypes (Hamilton, 1981; Linville, 1982; Lin-
ville
&
Jones, 1980;
see
also Cantor
&
Mischel, 1977).
But
whether
attitudes
form
network structures, schemata, stereo-
types,
or
some combination
of
these,
it is
clear
that
the
dimen-
sional
representation
of
attitude structure implicit
in
classical
scaling
techniques, such
as
Likert,
Guttman,
and
Thurstone
scaling,
does
not
fully
capture
the
important structural proper-
ties
of
attitudes.
As we
argue
in
this article,
the
structural
as-
sumptions prevalent
in
current cognitive psychology help
ac-
count
for
important phenomena involving
the
measurement
of
attitudes,
especially
in
survey
settings.
Building
on the
work concerning
the
structure
of
attitudes,
we
propose
a
model
of the
process
by
which attitude questions
are
answered.
We
describe this model
of the
response
process
in the
next section
of the
article. Then,
in the
following
four
sections,
we
describe
how the
context
of a
question—generally,
earlier items
in the
questionnaire—can
affect
each stage
of the
response process
and
offer
hypotheses
about
when
different
types
of
context
effects
are
likely
to
arise.
Our
review
of the
context-effects
literature
in
these sections aims less
for
compre-
hensiveness
than
for
theoretical relevance;
we
stress studies
that
seem
to
illustrate
the
different
processes
that
give
rise
to
context
effects.
In the final
section
of the
article,
we
focus
on the
vari-
ables
that
influence
the
size
and
direction
of
context
effects.
Process
of
Answering
Attitude
Questions
A
recent study
by
Luker
(1984) provides some
indication
of
the
contents
of
attitudes
on
abortion. These attitudes appear
to
have
a
complex
structure,
and
they resemble other long-term
299

300
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSKI
Table
1
Summary
of
Pro-Life
and
Pro-Choice
Views
Topic
Sex
roles
Meaning
of
sex
Motherhood
Premarital
sex
Abortion
Pro-life
way
Men
and
women
are
different.
The
purpose
of sex is
reproduction.
Parenthood
is a
natural
function,
not a
social
role.
Sex
outside
of
marriage
is
wrong.
Abortion
breaks
divine
law.
Pro-choice
view
Men
and
women
are
equal
and
similar.
The
purpose
of
sex
is
to
foster
intimacy.
Parenthood
means
giving
the
child
one's
best
resources.
Teenage
parenthood
is
the
problem,
not
teenage
sex.
Abortion
is a
matter
ofindividual
choice.
memory
structures, such
as
scripts
or
schemata.
Luker
inter-
viewed
activists
on
both sides
of the
abortion issue, people
who
devoted
substantial
amounts
of
time each week
to
working
for
their cause.
She
argued that partisans
on
different
sides
of the
abortion question tend
to
disagree
on a
number
of
related
is-
sues.
Table
1
contains
a
brief summary
of the
major
points
of
contention. Although
Luker
no
doubt sharpened
the
differences
between
the two
sides
in her
presentation
and
although
her
sam-
ple
of
activists probably
had
clearer, more coherent
views
on
the
issue than
a
cross section
of the
general population would,
her
results
are
consistent
with
our own findings
with
a
sample
of
nonactivists
(Tourangeau,
Rasinski,
&
D'Andrade,
1987):
Abortion
beliefs
are not
unidimensional;
they
are
probably
not
even
multidimensional
in any
useful
sense; instead, they seem
to
fall
into
a
small
set of
related topical,
or
thematic, clusters.
If
attitudes
are
structures
in
long-term
memory,
then answer-
ing
an
attitude question
is
likely
to
involve such processes
as
activating
the
relevant
attitude, retrieving
its
contents, synthe-
sizing
an
answer
from
what
has
been retrieved,
and so on. We
have
presented
a
detailed analysis
of the
process
of
answering
attitude questions (Tourangeau, 1984, 1986,
1987;
Tourangeau
&
Rasinski,
1986).
According
to the
model, respondents
first
interpret
the
question.
A key
component
of
this comprehension
process
is
locating
the
relevant attitude structure.
With
well-
formed,
highly
accessible attitudes,
merely
encountering
the
is-
sue
may be
sufficient
to
activate
the
relevant structure; Fazio
et
al.
(1986)
and
Fazio
and
Williams
(1986)
have
argued that
when
the
attitude
is
highly accessible, this initial step
in the
response
process
may be
automatic. Sometimes,
however,
the
process
of
locating
the
relevant attitude
may not be so
easy.
With
unfamil-
iar
issues,
no
specific
attitude
is
readily activated
by the
ques-
tion,
and
respondents must search
for the
relevant attitude.
The
second step
in the
response process
is
retrieval. What
is
retrieved
in a
particular case depends
on the
respondent's
be-
liefs
and on the
demands
of the
question.
An
item
on the use of
abortion
as a
means
of
birth control
may
activate
one set of
beliefs;
an
item,
about abortion
in the
case
of a
threat
to the
mother's
life
may
activate
a
different
set. Respondents with
well-
formed
attitudes
may in
some cases retrieve
a
general evaluation
that
serves
as an
overall
summary
of
their
beliefs
about
the
issue
rather than retrieve
the
underlying
beliefs
themselves
(N. An-
derson
&
Hubert,
1963; Lingle,
Geva,
Ostrom,
Lieppe,
&
Baumgartner,
1979; Lingle
&
Ostrom,
1979).
At the
other
ex-
treme, respondents
who
know
and
care
little
about
an
issue
may
construct
an
attitude
from
superficial
cues present
in the
situa-
tion;
in
persuasion settings,
for
example, such
uninvolved
and
uninformed
respondents
may
base their opinions
on the
attrac-
tiveness
or
credibility
of the
source
of a
persuasive message
(Chaiken,
1980;
Petty
ACatioppo,
1984). Because respondents
are
unlikely
to
retrieve
all
their
beliefs
about
an
issue,
the
re-
trieval
stage
can be
seen
as a
kind
of
sampling process that
over-
represents
the
most accessible
beliefs
or
situational
cues.
In
the
next
stage, respondents must
use the
information
they
have
retrieved
to
render
a
judgment. Sometimes this step
is
triv-
ial. Respondents
who
retrieve
the
belief that abortion
is
murder
do
not
need
to
make
a
complicated judgment
in
deciding
to
disagree with
an
item
that
says,
"Abortion
is not
taking
a
life."
Similarly,
respondents
who
retrieve
a
negative summary evalua-
tion
should
have
little
difficulty
in
endorsing
an
item that
says,
"Personally,
I'm
against
abortion."
In
such
cases,
an
answer
is
not so
much
formulated
as it is
directly retrieved.
In
most cases,
however,
the
question
will
not map so
directly
onto
an
existing
belief
or a
summary evaluation,
and a
more complicated
pro-
cess
will
be
needed
to
generate
a
judgment
from
the set of
re-
trieved
beliefs.
N.
Anderson's
(1974,
1981)
information
inte-
gration
theory
gives
some insight into
the
nature
of
this more
complicated
process.
According
to N.
Anderson (1974,
1981),
the
judgment process involves
the
scaling
of
beliefs
(i.e., placing
them
on
some
underlying
dimension
of
judgment, such
as fa-
vorability);
attaching
a
weight
to
each
one
(i.e., assessing their
relative
credibility
and
importance);
and
combining them into
an
overall judgment, using
an
integration rule, such
as
averag-
ing
or
adding.
The
scaling
of the
beliefs
depends
on the
dimen-
sion
of
judgment
selected
and on the
standard
of
comparison
used
to
anchor
the
dimension.
The
weight
attached
to a
particu-
lar
belief also depends
on the
dimension
of
judgment:
A
belief
that
is
central
to one
dimension (e.g., evaluation)
may be
pe-
ripheral to
another
(e.g.,
likelihood).
If a
number
of
beliefs
or
feelings
about
an
issue
have
been retrieved,
they
may
have
to be
combined
into
an
overall
judgment,
perhaps through
a
process
such
as
averaging
their implications
for the
judgment
(N. An-
derson,
1974,1981).
Sears
etal.(1986)
argued
that
this
integra-
tion
step
is
crucial
with
issues that evoke several political sym-
bols.
In
the final
stage,
respondents
must
report
their
answers.
At
least
two
processes
play
an
important role
in the
reporting stage.
The
format
of
most attitude items requires respondents
to
select
a
response
from
among
a
preestablished
set of
answer catego-
ries.
Thus, respondents must
map
their judgments onto
one of
the
response options.
In
addition
to the
mapping process,
an-
swers
may
undergo
an
editing process
in
which
the
answer
is
checked
for
consistency
with
prior answers
or for
social desir-
ability.
The final
response
given
may be a
compromise between
the
respondent's judgment
and the
dictates
of
consistency
or
social pressure.
In
the
context
of
survey interviews,
in
which
interest
on the
part
of the
respondents
is
relatively
low and
time pressures
are
great,
the
response process
is
likely
to be
carried
out
superfi-
cially.
Respondents
are
unlikely
to
retrieve
all
their
beliefs
on
an
issue; instead,
the
retrieval
process
is
likely
to
yield
a
sample

CONTEXT
EFFECTS
301
of
pertinent
beliefs
that
overrepresents
the
most accessible ones.
In
fact,
a
number
of
persuasion studies
(Chaiken,
1980; Petty
&
Cacioppo,
1984) suggest that
uninvolved
respondents
may
retrieve
virtually
no
issue-relevant
beliefs,
basing their
re-
sponses instead
on
cues
in the
immediate situation.
Similarly,
the
judgment stage
is
unlikely
to
include complex comparisons
involving
a
number
of
dimensions
of
judgment;
instead, salient
dimensions
are
likely
to
receive undue
weight,
and
salient stan-
dards
of
comparison
are
likely
to
serve
as
anchors
for the
judg-
ment.
Putting
these
processes
into
the
foreground
helps emphasize
two
key
points that might otherwise remain obscure. First,
al-
though
attitudes
have
a
static
component—the
component that
resides
in
long-term memory
and
serves
as the
basis
for
answers
to
specific
attitude
questions—they
have
a
dynamic component
as
well.
Answering
any
question requires generating
a
response.
Although
this
process
may be
based
on
existing
structures,
it
takes place on-line
in
real time. (See
Hastie
&
Park's, 1986,
article
on
memory-based judgments.)
This
suggests
a
second
point:
Even
when
the
underlying attitude structure
is
stable,
the
response process need
not be
very
reliable.
To
cite
just
one
source
of
unreliability,
we
note
that
the
retrieval process
may
yield
what
is
most accessible
on a
topic rather than
what
is
most
important.
The
accessibility
of
a
belief
(or
what
Tversky
&
Kah-
neman,
1973,
called
its
availability)
has a
number
of
determi-
nants.
Higgins
and
King
(1981)
cited
six
determinants
of the
accessibility
of a
concept—expectations,
goal relevance,
re-
cency
of
use,
frequency
of
use,
situational
salience,
and
relation
to
other
concepts—and
most
of
these
do not
necessarily relate
to its
long-term strength.
The
unreliability
of the
retrieval pro-
cess (and
the
degree
to
which
the
results
of
retrieval
reflect
short-term
differences
in
accessibility rather than long-term
properties such
as
strength)
is
likely
to be
heightened
in
attitude
measurement settings,
in
which
few
respondents
have
either
the
motive
or the
opportunity
to
reflect
carefully
on
their answers.
As
Tversky
and
Kahneman
demonstrated,
what
is
most
readily
retrieved
from
memory does
not
necessarily
reflect
either real-
ity
or the
contents
of
memory.
Fischoff,
Slovic,
and
Lichtenstein
(1980)
made
a
similar
point
about procedures
for
eliciting
preferences
in
decision-
making
settings.
They
argued that respondents
do not
always
have
well-formed,
coherent opinions
and
that, lacking
such
opinions,
their expressed
preferences
can be
affected
by
subtle
differences
in how the
decision problem
is
posed.
Like
answers
to
attitude questions, decisions
are
often
the
product
of
an
unre-
liable process,
a
process that
can be
influenced
by
apparently
minor changes
in
wording
or
context.
Recent
reviews
of the
survey
literature
(Bradburn,
1982;
Schuman
&
Presser,
1981)
provide
ample evidence that
the
pro-
cess
of
answering attitude questions
in
surveys
can be
affected
by
such
weak
and
momentary
influences
as the
order
in
which
the
items
are
presented. These reviews indicate that
relatively
minor
changes
in
item wording
and
item
context
can
have
dra-
matic
effects
on the
distribution
of the
responses. Item
context
refers
to
earlier material
in the
questionnaire. Generally,
the
context
of
an
item
is
defined
narrowly,
in
terms
of the
preceding
items
in the
questionnaire,
but
other kinds
of
prior material,
such
as
item introductions
or
prior tasks that
the
respondents
have
completed,
can
have
a
similar impact
on
responses
to
later
questions.
In our
examination
of
item-context
effects,
we
focus
primarily
on the
effects
of
earlier questions
but
consider other
forms
of
context
when
they
are
relevant
to the
hypothesized
processes under consideration. Relative
to
authors
of
earlier
re-
views
of
item-context
effects
(e.g., Schuman
&
Presser,
1981),
we
seek
to
impose
a
theoretical structure
on the
range
of
empiri-
cal
findings.
Accordingly,
we
focus
on
studies that appear
to il-
lustrate
different
processes responsible
for
context
effects.
One
particularly well-researched example
of an
item-context
effect
involves
two
items
on
abortion.
One
item
reads,
"Please
tell
me
whether
or not you
think
it
should
be
possible
for
a
preg-
nant
woman
to
obtain
a
legal abortion
if
there
is a
strong chance
of
serious
defect
in the
baby."
The
other item asks about
a
different
situation;
it
concerns legal abortions
when
"the
woman
is
married
and
does
not
want
any
more children." When
these
two
items
are
presented,
it
matters which
one
comes
first.
The
basic
finding
(Schuman
&
Presser,
1981)
is
that
when
the
birth-defect
item comes
first, it
reduces
the
number
of
pro-
choice responses
to the
item about married women.
Answering
the one
question somehow
affects
answers
to the
other.
The
main purpose
of
this article
is to
account
for
some
of
these
well-documented
context
effects,
by
showing
how the
con-
text
of an
item
can
affect
each step
in the
response process.
These context
effects
are
generally regarded
as
measurement
artifacts.
Consistent
with
this
view,
the
mechanisms that
we de-
scribe primarily
involve
the
measurement process rather than
the
underlying attitude structures that
the
questions
were
de-
signed
to
tap. But, although
we
sympathize
with
the
character-
ization
of
context
effects
as
artifacts,
we
argue that
the
processes
that result
in
context
effects
are
interesting substantive phenom-
ena in
their
own right. We
also note that, although context
effects
sometimes produce pseudo-changes
in
attitude, they also
sometimes produce real (i.e., long-lasting)
changes—as
when
the
context induces
one to
apply
a
norm
the
relevance
of
which
had not
been apparent
before.
As
Schuman
(1982)
put it,
"Arti-
facts
are in the
mind
of the
beholder."
In
this article,
we try
to find the
substantive meat
in
what others
may
well
regard
as
methodological poison.
Part
of the
reason that context
effects
are
viewed
with
such
frustration
by
survey researchers
is
that
they
sometimes appear
to be
unreliable.
For
example, Schuman
and
Presser
(19
81)
re-
ported
a
number
of
context
effects
and
also
a
number
of
failures
to
replicate them. (Bishop,
Oldendick,
&
Tuchfarber,
1985,
re-
ported
a
particularly puzzling
set of
inconsistent results.)
It is
not
only
the
magnitude
of the
effects
that
can
vary
but
their
direction
as
well. Prior items sometimes
influence
respondents
to
give
consistent answers later
on, but
prior items sometimes
appear
to
produce inconsistencies. Because
the
term
consis-
tency
carries
with
it a
specific
theoretical connotation,
we
prefer
to use the
terms
carryover
and
backfire
effects
in
describing con-
text
effects.
Thus,
in the
following
sections
of
this
article,
we use
our
model
of the
response
process
to
explain
how
both types
of
effects
can
arise
at
each stage
of the
response process.
Context
and the
Interpretation
Stage
Context
can
affect
the
interpretation
of an
attitude item
in at
least
two
ways.
Prior items
can
provide
an
interpretive
frame-
work
for the
later items,
influencing
the
respondent's
view
of

302
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSK1
what
issue
the
later item
is
supposed
to be
about.
Prior
items
can
also
determine what
is
seen
by the
respondent
as
worth say-
ing
and
what
is
seen
as
redundant. Both
effects
can
influence
responses
to the
later
items.'
Context
as
Interpretive
Framework
In
studies
of
comprehension
and
memory,
it has
been
repeat-
edly
demonstrated
that
people
extract
an
interpretation
from
text,
transfer
this interpretation
(usually
the
gist
of the
passage)
to
memory,
and
subsequently remember this gist rather than
the
verbatim text (see, e.g.,
Bransford
&
Johnson,
1972).
The
gist
frequently
takes
the
form
of
some overall
framework
or
schema with distinguishing particulars
(Bower,
Black,
&
Turner,
1979;
Graesser,
Gordon,
&
Sawyer,
1979;
Reiser, Black,
&
Abel-
son,
1985).
The
overall
framework is
likely
to
consist
of a
mental
model
of
the
situation—that
is, a
script
(Abelson,
1981;
Schank
&
Abelson, 1977)
or a
schema
(Rumelhart,
1975;
Rumelhart
&
Ortony,
1977)—and
the
context
is
important because
it
helps
identify
the
relevant model.
Context
can
influence
interpretation
of
attitude items
in a
survey
questionnaire
in a
similar
way.
In the
survey
setting,
an
attitude structure
is
likely
to
serve
as the
framework
for
inter-
preting
an
item,
and
prior items
may
trigger
the use of a
particu-
lar
attitude.
For
example,
Schuman
and
Presser
(1981)
mea-
sured attitudes
on an
unfamiliar issue,
the
Monetary Control
Bill
(an
obscure bill
before
Congress). Most people
gave
"don't
know" responses
to a
question about their support
for
this bill,
but
among those
who
gave
a
"favor"
or
"oppose"
response,
there
was a
relation between responses
to
this item
and re-
sponses
to an
item
in the
questionnaire
on
inflation.
People
who
were
the
most concerned about
inflation
tended
to
favor
the
Monetary
Control Bill. Schuman
and
Presser argued that this
relation
reflects
the
interpretation some respondents
gave
to the
Monetary Control Bill item;
the
respondents thought that
the
Monetary
Control
Bill
was an
anti-inflation measure
and an-
swered
the
item accordingly.
A
recent study
by
Tourangeau
and
Rasinski
(1986)
provides
more direct evidence supporting Schuman
and
Presser's
(1981)
account.
In our
study,
we
thought
we
could encourage respon-
dents
to
interpret
the
Monetary
Control
Bill
as an
anti-inflation
measure
by
placing
a
series
of
inflation
items immediately
be-
fore
it in one
version
of a
questionnaire.
The
second version
of
the
questionnaire included
the
same
inflation
items placed
before
the
Monetary Control
Bill
question,
but the
items were
scattered throughout
the
questionnaire.
In the
third version,
neutral items with
no
bearing
on the
bill preceded
the
Mone-
tary
Control
Bill
question.
As can be
seen
from
Table
2, the
inflation-context
items increased support
for the
Monetary
Control
Bill
and
decreased
the
"don't
know" rate
but
only
when
the
inflation
items
were
presented
in a
block immediately pre-
ceding
the
Monetary Control Bill item. Evidence
from
a
repli-
cation using
a
different
unfamiliar
issue indicates that when
the
wording
of the
item
clarifies
the
issue,
the
effect
of the
prior
context items presented
in a
block
is
eliminated. This
finding
suggests
that
this
context
effect
arises during
the
interpretation
process
and
can,
therefore,
be
eliminated
by
providing respon-
dents with
a
more
definite
basis
for
comprehending
the
item
than that provided
by the
context.
Table
2
Impact
of
Context
on
Interpretation
of
Unfamiliar
Issues
Context
Favor
Oppose
Dont
know
Item:
Do you
favor
or
oppose
passage
of the
Monetary
Control
Bill?
Block
of
inflation
items
Scattered
inflation
items
Neutral
context
27.5%
8.8%
12.5%
17.5%
20.0%
25.0%
55.0%
71.2%
62.5%
Herr,
Sherman,
and
Fazio
(1984)
also
reported
a
study
in
which judgments
of an
unfamiliar stimulus were assimilated
to
earlier judgments
of
more
familiar
stimuli. They argued
that
the
effect
occurred because
the
target
was
seen
as
belonging
to
the
same category
as the
earlier items.
In
similar results
from
two
studies
of the
attribution
process,
Trope
(1986)
showed
that
situational
cues
affect
the
interpretation
of
behaviors, especially
ambiguous
ones.
Trope argued that
situational
cues
disambigu-
ate
otherwise
uninterpretable
behaviors. Finally, Martin
(1986)
reported
three
additional
studies
that
suggest that context
affects
the
encoding
of
ambiguous stimuli.
In our
terms,
all
these
findings
illustrate carryover
effects
that arise when
respon-
dents interpret
the
target stimulus.
Backfire
at the
Interpretation
Stage
In
some
ways,
survey interviews
are
like conversations.
Two
people—the
interviewer
and the
respondent—take
part,
and the
interview
consists
of
conversational units involving connected
questions
and
answers
on a
given
topic. Although
the
interview
situation
is
clearly
a
very
specialized
form
of
conversation,
it
may
still
follow
many
of the
principles
that
guide more ordinary
and
less structured conversations.
Grice
(1975),
for
example,
argued
that conversations
are
guided
by
such principles
or
max-
ims
as the
maxim
of
quality
(which
demands that
one
should
say
only things that
are
true). Another
Gricean
maxim—that
one
should
be
informative
and
avoid redundancy
(Haviland
&
Clark,
1974)—may
be
especially relevant
to
context
effects
in
attitude measurement. This principle
may
lead respondents
to
give
apparently inconsistent answers
when
in
fact
the
respon-
dents
are
merely interpreting later questions
as
calling
for
new,
nonredundant
information.
Bradbum
(1982)
cited
an
example
that
may
illustrate this
process.
He
reported
that
an
item asking
respondents
to
evalu-
ate
their overall happiness yielded
fewer
"very
happy"
re-
sponses
when
this general question
followed
a
more
specific
one
on
marital happiness than
when
the
general item came
first.
Bradburn
(1982;
see
also Turner
&
Martin, 1984,
pp.
293-294,
1
Context
can,
of
course,
have
other
effects
on the
interpretation
of
attitude
items.
Earlier
items
can,
for
example,
make
later
items
appear
more
extreme
(e.g.,
Higgins
&
Lurie,
1983).
We
consider
some
of
these
other
effects
later,
in
connection
with
the
judgment
process.
In
this
sec-
tion,
we
focus
on
context
effects
that
influence
what
attitude
or
concept
is
seen
as relevant to the
question
or
that
influence
the
perceived
scope
of
the
question.

CONTEXT
EFFECTS
303
on
interpretive contrast) interpreted this result
as
reflecting
the
tendency
for the
respondents
to
exclude their marital happiness
when
the
general item came second because they
felt
that they
had
already covered their marriage
in
their earlier answers.
It
was
as if
respondents interpreted
the
general question
to
mean,
"Aside
from
your marriage,
how
happy
are
you?"
when
the
gen-
eral
item
followed
the
marital item.
The
results
do not
rule
out
other
explanations
for
this context
effect,
but the
exclusion
hy-
pothesis
remains
a
viable account
for
this
context
effect
and
related ones
(Kalton,
Collins,
&
Brook,
1978).
2
Backfire
effects
arise during
the
interpretation stage
when
re-
spondents
interpret related items
as
calling
for new or
different
responses (cf. Martin, 1986).
The
overall direction
of the
con-
text
effect
depends
on the
marginal distribution
of
responses
to
the
prior items.
If respondents had
discounted,
or
subtracted
out, mostly unhappy marriages,
for
example, their overall
hap-
piness would, according
to the
exclusion
hypothesis,
have
been
greater
on the
average
when
the
general item came later.
Carryover
Versus
Backfire
in
Interpretation
Our
discussion
of the
processes leading
to
carryover
and
backfire
effects
at the
interpretation stage
implies
several
hypotheses about
when
each type
of
effect
is
likely
to
occur.
Carryover
effects
reflect
uncertainty about what attitude
is
rele-
vant
to the
item; they
are
likely
to
arise
when
the
attitude issue
is
new
or
unfamiliar
to
many respondents
(as
with
the
Monetary
Control
Bill),
when
the
item wording does
not
make
the
rele-
vant
issue clear,
and
when
the
context items
can be
seen
as
relat-
ing
to the
same general issue (e.g.,
inflation).
Placing
the
context
items
in a
block immediately
before
the
question
on the
unfa-
miliar
issue
may
encourage
the
inference that
the
items
are all
about
the
same topic.
Backfire
effects
at the
interpretation stage
appear
to
reflect
uncertainty
of a
more
specific
kind; these
effects
arise
when
respondents
are
unsure about
the
scope
of an
item, especially
an
item intended
as a
summary item. They
are
likely
to
involve
familiar
issues (such
as
one's
overall happiness)
and
to
occur
when
a
general item
follows
one or
more items
on
particulars that
are
included
in the
general item.
If the
list
of
particulars
is
long enough,
however,
it may
encourage respon-
dents
to
interpret
the
general item
as a
summary
of the
particu-
lars rather than
as a
residual category.
Context
and
Retrieval
Context
can
also
affect
what gets retrieved
or
considered
as
respondents answer
a
particular
question.
In the
memory litera-
ture,
the
influence
of
context
on
retrieval
is
widely
recognized
and is
embedded
in
some
of the key
distinctions
in the field,
such
as
those among
free
recall,
cued recall,
and
recognition.
In
free-recall
tasks,
the
only
cue or
context guiding
the
retrieval
process
is the
fact
that there
is
something
(such
as
items
on a
list)
to be
recalled that
was
learned
in a
particular
time and
place.
In a
cued recall task,
a
more
specific
context
is
provided
for
the
memory search, such
as a
topic heading
or
category
la-
bel. And,
in a
recognition
task,
the
best possible
cue or
retrieval
context
is
provided—the
item itself.
According
to
many
theories
of
long-term
memory,
memory
search
is a
process
in
which items
are
retrieved through
the
spread
of
activation
from
one
item
to
related,
or
linked,
items
in
an
associative network
(J.
Anderson, 1978, 1983; Collins
&
Loftus,
1975).
In
deliberate retrieval situations,
a
node
is
acti-
vated,
and the
activation spreads automatically
from
this node
to
other nodes.
For
example,
in a free-recall
situation,
the
initial
node might represent
the
list learned
in the
course
of
the
experi-
ment
or
specific
situational
features. Once
the
activation pro-
cess
is
initiated,
it
goes
on
automatically
(i.e.,
outside
of
aware-
ness
and
beyond
conscious
control;
Posner,
1978).
Activation
can
spread
to
related ideas that happen
to be
irrelevant
to the
memory
task immediately
at
hand.
The
spread
of
activation
to
the
related ideas makes these ideas more accessible
to the re-
trieval process later.
For
example,
if
subjects
are
asked whether
fruit
was on a
memory list, they
are
quicker later
to
determine
whether
apple
was on the
list. This facilitation
effect
is
often
referred
to
as
priming,
and the
earlier item
or cue
that produces
the
facilitation
effect
is
referred
to as a
prime (Posner,
1978).
Retrieval
and
Attitude
Questions
A
number
of
studies
in the
social judgment literature
have
used priming
as a
means
of
influencing
judgment processes.
Wyer,
Bodenhausen,
and
Gorman (1985),
for
example, showed
that judgments regarding rape cases
were
influenced
by
appar-
ently
unrelated judgments
of
slides that depicted scenes
of
vio-
lence,
sexy
women,
and so on.
Wyer
et
al.
argued that
the
prior
rating task
involving
the
slides activated stereotypes
and
norms
that
were
then
more
likely
to be
applied
to the
judgments
of the
rape cases. (See also
Higgins
&
King,
1981,
and
Higgins,
Rholes,&
Jones,
1977.)
Prior questions
in an
attitude
survey
can
have
this same prim-
ing
effect.
Respondents
who are
asked
a
series
of
questions
oh
women's
rights,
for
example,
may be
more likely
to
retrieve
be-
liefs
consistent
with
the
idea that abortion
is,
say,
a
matter
of
the
woman's
free
choice
when
they
are
asked about abortion
later.
Figure
1
contains
a
more detailed depiction
of
the
hypoth-
esized
process.
The
women's rights items trigger
a
conscious
retrieval
process,
in
which respondents recall their views
on
women's
rights. Activation
can
then spread
from
these
beliefs
to
related
pro-abortion
beliefs, making them more accessible
to
subsequent retrieval
efforts.
The
primed
pro-abortion
beliefs
affect
respondents'
answers
to the
target abortion item
(Tour-
angeau,
1987;
Tourangeau
&
Rasinski,
1986).
This prediction rests
on
several assumptions regarding
the
structure
of
beliefs
about abortion.
One key
assumption,
of
course,
is
that
at
least
some respondents
see
abortion
as
related
2
As
Schuman
and
Presser
(1981,
pp.
42-44)
observed,
the
context
effect
involving
the
happiness
items
does
not
always
replicate,
for
rea-
sons
that
are
not yet
clear.
One
possibility
is
that
when
the
overall
happi-
ness item
comes
after
a
series
of
specific
items,
respondents
correctly
infer
that
the
general item
is
intended
to
summarize
rather
than
exclude
the
specific
domains
in the
earlier
questions.
In any
case,
Kalton, Col-
lins,
and
Brook
(1978)
reported
a
context
effect
consistent with
the
orig-
inal
finding on
marital
and
overall
happiness:
Respondents asked
to
evaluate driving
standards
in
general
became
less
negative when they
had
first
evaluated
the
driving
standards
of
young
people.
Kalton, Col-
lins,
and
Brook
interpreted
this
context
effect
as
reflecting
the
exclusion
or
subtraction
of
the
(primarily negative)
reactions
to
young drivers.

304
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSKJ
Items
on
women's rights
Items
on
traditional
values
Figure
1.
Hypothesized
carryover
effect
during
the
retrieval
stage.
The first
process
(1
in the figure) is
retrieval
of
beliefs
related
to the
context
items;
activation
then
spreads
to
related
beliefs
about
the
target
issue
(2),
rendering
a
response
on one
side
of the
issue
more
likely
(3).
to
women's rights.
The
spread
of
activation
from
one
idea
to
another depends
on the
connection between
the
ideas. With
atti-
tude issues,
beliefs
on one
side
of an
issue tend
to be
more
closely connected than
beliefs
on
opposite sides.
For
this reason,
we
assumed that items
on
women's rights
would
make
pro-
abortion
beliefs
more accessible, without necessarily
affecting
the
accessibility
of
anti-abortion beliefs.
Our
analysis
of the
structure
of
abortion
beliefs
(Tourangeau
et
al.,
1987)
indicates
that this assumption
can be
met, because people
on
different
sides
of an
issue tend
to see the
issue
as
related
to
different
things.
People
who
support legalized abortion
see it as
related
to
women's
rights,
whereas opponents
do not
make
this connec-
tion
(at
least,
not so
directly). Another assumption
is
that
a
large
number
of
respondents could
go
either
way on the
target item.
Partisans
on the
pro-abortion
side
are
likely
to
retrieve pro-
abortion
beliefs
even
in the
absence
of any
priming. Partisans
on
the
other side
have
no
pro-abortion
beliefs
to be
primed.
The
context
effect
should, therefore, occur only
with
respondents
with
mixed
views—those
who can see
abortion
in
pro-choice
terms
but who
would
not
necessarily
do so
without
the
prompt-
ing
of the
context items. Finally,
our
prediction assumes that
in
answering
the
target question, many respondents retrieve spe-
cific
beliefs
rather than some overall evaluation.
We
recently conducted
a
study (Tourangeau
&
Rasinski,
1986)
that provided support
for the
line
of
reasoning depicted
in
Figure
1.
Respondents
who
answered
four
items
on
women's
rights
showed greater support
for
legalized abortion than
did
respondents
who
answered
four
questions concerned
with
tradi-
tional
values.
A
group that received neutral context items
ex-
hibited intermediate levels
of
support.
(In all
these groups,
we
scattered
the
context items among unrelated items
to
deempha-
size
their
relations
to the
target item
on
abortion.)
We
obtained
similar
results
in a
parallel study
with
welfare
spending
as the
target issue.
The
context items seemed
to
prime material that
affected
responses
to the
target items.
There
are a
couple
of
reasons
to
attribute these context
effects
to the
retrieval rather than
the
interpretation stage.
First,
the
issues involved
were
highly
familiar
ones,
and the
items them-
selves
were
relatively straightforward. There
is no
reason
to
sup-
pose that respondents
had any
difficulty
in
comprehending
the
items
or in
identifying
the
relevant attitudes. Second,
the
con-
text
effects
for
abortion
and
welfare
were
most apparent
when
the
context items
were
scattered;
the
context
effects
thought
to
involve
the
interpretation
of
the
item appear
to
be
more marked
when
the
context items
are
presented
in a
block that
is
placed
immediately
before
the
target item. Thus,
different
processes
seem
to be
implicated.
A
number
of
other studies indicate that other contextual vari-
ables
can
activate concepts
or
feelings
that
are
then
carried
over
to a
target item.
Schwarz,
Strack,
Kommer,
and
Wagner
(in

CONTEXT
EFFECTS
305
press) showed that mood,
when
salient,
may be
used
as a
basis
for
judging overall
life
satisfaction
and
that mood
can be
affected
by
such transient contextual variables
as the
weather
or finding a
dime.
According
to
Schwarz
et
al.,
the
context vari-
ables
affect
mood
and
increase
its
accessibility; mood
is
then
used
as a
basis
for
inferring
life
satisfaction. Such mood carry-
over
effects
may
account
for at
least
one
well-known
survey con-
text
effect,
in
which
a
series
of
items designed
to
measure ano-
mie
appeared
to
affect
responses
to
subsequent items assessing
the
level
of
confidence
in
public institutions (Turner
&
Krauss,
1978;
see
also
Johnson
&
Tversky,
1983).
Discounting
Accessible
Concepts
Just because
a
piece
of
information
or a
belief
has
been
primed does
not
guarantee
that
respondents will
use it in
form-
ing
a
judgment
or
answering
a
question.
A
number
of
studies
indicate
that
respondents will discount
or
actively suppress
in-
formation
that they regard
as
suspect
or
irrelevant.
For
exam-
ple, Martin
(1986)
showed that respondents sometimes sup-
press
trait
categories primed
by a
prior task
when
forming
im-
pressions
of a
target person
later;
apparently because they
feel
that
the
later judgment should
be
distinct
from
the
earlier task.
In
a
similar vein, Schwarz
and
Clore
(1983)
argued that respon-
dents
use
mood
as a
basis
for
inferring
their
life
satisfaction
when
mood
is
accessible
but
that respondents discount mood
when
its
connection
to the
weather
is
made salient.
In one
study,
interviewers
called
the
respondents' attention
to the
weather
by
asking about
it;
respondents
in
this group apparently
dis-
counted mood
in
judging
life
satisfaction. According
to
Schwarz
and
Clore, these respondents realized that mood
was
a
poor indicator
of
overall
life
satisfaction because
of its relation
to
such transient factors
as the
weather and, therefore, excluded
mood
in
rating their
life
satisfaction.
In
some
cases,
respondents
appear
to set
aside
the
accessible
concepts
so
that
these concepts
have
no
effect
on the
later judg-
ment (Schwarz
&
Clore,
1983);
in
others,
the
exclusion
of
mate-
rial
that
supports
one
response leads respondents
to
make
the
opposite
response,
and a
backfire
effect
results (Martin, 1986).
Several
other studies
(Higgins
&
King,
1981,
Studies
1 and 5;
Wyer
et
al.,
1985)
illustrate similar
backfire
effects.
For
exam-
ple,
in one
study, Higgins
and
King
(1981,
Study
1)
attempted
to
increase
the
accessibility
of sex
role stereotypes;
in a
subse-
quent recall task, male respondents displayed
a
backfire
effect:
They
incorrectly recalled
a
male target person
as
having more
"female"
than
"male"
characteristics.
One
explanation
for
this
effect
is
that
the
respondents consciously rejected
the
traditional
male
sex
stereotype
for a
more contemporary one.
Discounting
or
rejecting accessible material
may
also
occur
in
attitude surveys,
in
which
the
accessible material
has
been
primed
by
earlier questions
in the
interview.
In our own re-
search
(Tourangeau
&
Rasinski,
1986),
we
have
sometimes
ob-
served
backfire
effects
with
familiar
issues.
For
example,
re-
spondents
who
were
asked
a
series
of
questions about
the
gov-
ernment's responsibility
to
provide certain services, such
as
hospitals
and
roads, subsequently showed reduced support
for
welfare
spending.
The
backfire
effect
was
evident only when
the
context items were presented
in a
block
that
was
placed imme-
diately
before
the
target item
on
welfare.
Carryover
Versus
Backfire
in
Retrieval
Several
variables
may
determine whether material rendered
accessible
by
prior items
is
used
in
formulating
a
response
or
excluded
from
consideration.
The
priming process responsible
for
carryover
effects
at the
retrieval stage
is
thought
to be
auto-
matic
(Posner,
1978) and,
therefore,
operates outside
of
con-
scious awareness.
The
discounting
of
accessible material
ap-
pears
to be a
more controlled process;
it
involves recognizing
that
certain
information
can be
used
in
answering
a
question
and
then deciding
not to use
that information.
This
suggests
two
variables that
may
determine whether
a
carryover
or
back-
fire
effect
occurs—how
much thought respondents give
to
their
answers
and how
obvious
the
context items are. Thoughtful
re-
spondents
are
more
likely
to
assess
the
relevance
or
validity
of
material primed
by
prior items; obvious contexts
are
more
likely
to
arouse
respondents'
suspicions and, thus, trigger
a
con-
scious evaluation
of the
primed
material.
Thus,
in
Schwarz
and
Clore's
(1983)
study, calling
attention
to
context
(i.e.,
the
weather) eliminated
its
effect.
Similarly,
in
our own
work,
when
the
context items
were
presented
in a
block
placed immediately
before
the
target
item,
the
carryover
effect
sometimes disappeared
or was
reversed. Material made accessi-
ble by
such blatant contexts
may be
more
likely
to be
rejected
or
discounted.
Context
Effects
on
Judgments
Answering
an
attitude question generally involves rendering
a
judgment.
In
attitude surveys, judgmental carryover
effects
can
occur when
the
standards
or
dimensions used
in
answering
one
question
are
applied
to
later
questions.
This
is
similar
to
findings
in
studies
of
problem solving that
show
that
one
sticks
with
a
solution strategy
even
when
it
stops being
efficient
(e.g.,
Luchins.
1946).
These
studies
demonstrate
a
kind
of
strategic
carryover
effect,
in
which problem-solving techniques induced
in
solving earlier problems
are
inappropriately applied
to
later
ones.
Many
attitude judgments
are
reasonably
delimited (e.g.,
judgments
of
approval
of a
proposed course
of
action;
Wyer
&
Hartwick,
1984),
but
often
the
relevant dimensions
are
unclear
or
unspecified,
the
criteria
are
ill-defined,
and the
standards
of
comparison
are
vague. Context
can
affect
how the
judgment
is
made, what dimensions
are
judged,
and
what
criteria
or
stan-
dards
of
comparison
are
used.
In the
terms
of
information
inte-
gration theory
(N.
Anderson, 1974,
1981),
context
can
affect
which
dimension
is
used
in
rendering
the
judgment, what
an-
chors
are
used
in the
scaling
of
individual
beliefs,
and
what inte-
gration rule
is
applied.
In
principle, then, context
effects
on
judgment
refer
to the
impact
of
context
on how
beliefs
are
used
in the
judgment pro-
cess; context
effects
on the
retrieval process
refer
to the
impact
of
context
on
what
beliefs
figure
into
the
judgment.
In
practice,
however,
it can be
difficult
to
distinguish
the two
types
of
effects.
Context can,
for
example, suggest
a
standard
to
which subse-
quent items
are
then compared.
In
such cases, context
has
affected
not
only
what
has
been retrieved (i.e.,
the
standard)
but
also
how
other material,
which
would
have
been retrieved
anyway,
is
evaluated.

306
ROGER
TOURANOEAU
AND
KENNETH
A.
RASINSKI
Judgmental
Carryover
A
series
of
studies
by
lyengar,
Kinder, Peters,
and
Krosnick
(1984)
illustrates
this ambiguity.
In
these
studies,
lyengar
et
al.
explored
how
news coverage
can
affect
political attitudes.
In one
study,
subjects watched about
40
min
of
taped news stories with
varying
levels
of
coverage
of the
energy
crisis.
The
subjects then
rated President
Carter's
handling
of
energy
policy
and his
over-
all
performance. These
two
ratings were more highly correlated
when
the
subjects
had
seen more stories
on
energy.
This height-
ened
correlation
may
represent
a
retrieval
effect
(in
which
re-
spondents
were
more
likely
to
consider
Carter's
energy policies
after
they
had
watched
the
news
stories),
a
judgment
effect
(in
which
the
respondents simply
gave
greater weight
to
energy-re-
lated
beliefs
about
Carter),
or
both.
In
other cases,
the
context
may
trigger
the
application
of a
norm, which then provides
the
basis
for a
judgment. Once
again, context probably
affects
both what respondents consider
in
making their judgments
and how
they make
the
judgment.
Schuman
and
Presser
(1981;
see
also
Schuman
&
Ludwig,
1983)
reported
a
context
effect
involving
parallel
items
about
newspaper reporters
from
Communist countries ("Do
you
think
the
United
States
should
let
Communist newspaper
re-
porters
from
other countries come
in
here
and
send back
to
their
papers
the
news
as
they
see
it?")
and
from
the
United
States ("Do
you
think
a
Communist country like Russia should
let
American newspaper
reporters
come
in and
send back
to
America
the
news
as
they
see
it?").
Roughly
half
of the
respon-
dents endorsed
the
item about Communist
reporters
when
it
came
first; the
proportion rose
to
nearly three
fourths
when
the
Communist-reporter item came
after
the
item about American
reporters. According
to
Schuman
and
Presser,
when
the
Ameri-
can-reporter item
is
presented
first, the
item
on
Communist
re-
porters triggers
the
norm
of
fairness
or
evenhandedness,
a
norm
that
is not so
likely
to be
seen
as
relevant
when
the
Communist
item
is
presented
first. By
contrast,
when
the
item
on
Commu-
nist reporters comes
first,
responses
to it are
likely
to be
based
on
attitudes toward communism
or the
Soviet Union. Several
other
well-documented context
effects
(such
as one
involving
items
on
trade restrictions
on
U.S. exports
to
Japan
and on im-
ports
to the
United States
from
Japan) also appear
to
reflect
the
triggering
of a
norm
of
evenhandedness. Schuman
and
Ludwig
gave
a
fuller
discussion
of
these
and
related examples.
Judgmental
Anchors
Context items
can
also suggest reference
points
that
may
serve
as
anchors,
or
standards
of
comparison,
for
later judg-
ments. Judgmental contrast
effects
are a
familiar
idea
in
social
psychology,
dating back
to
Hovland,
Harvey,
and
Sherif
s
(1957)
classic
study.
A
number
of
related
effects
have been
ob-
served, many
of
them
involving
judgments
of the
position
of
an
attitude item
or
persuasive message. Generally,
contrast,
or
backfire,
effects
are
found
(but
see
Schwarz
&
Wyer,
1985).
For
example,
people opposed
to a
message tend
to see the
message
as
more extreme than
do
those
who
agree
with
the
message
(Hovland
etal.,
1957;
Judd
&
Harackiewicz,
1980).
This
is
gen-
erally seen
as a
contrast
effect,
because people
are
thought
to
use
their
own
position
as an
anchor
in
judging other positions
and
to
contrast
opposing opinions with their own.
The
anchor
for
a
judgment need
not be
one's
own
opinion
but
may
be
earlier
items
(Higgins
&
Lurie,
1983; Schwarz
&
Wyer,
1985)
or
even
the
midpoint
of the
response scale (Schwarz
&
Hippler,
1987).
According
to
some accounts
(Ostrom,
1970; Upshaw,
1969),
these contrast
effects
are
not,
strictly
speaking,
judgmental
but
involve
the
process
by
which
the
judgment
is
mapped onto
a
response scale. Some studies (Judd
&
Harackiewicz,
1980),
however,
indicate
that
the
judgment itself
is
affected
by the
stan-
dard
of
comparison.
A
study
by
Strack,
Schwarz,
and
Gschneidinger
(1985)
illus-
trates
how
context
can
affect
the
standard
of
comparison
used
in
making later judgments. They
had
respondents generate
pos-
itive
or
negative personal experiences
from
the
past
and
then
rate their current
life
satisfaction. Respondents
who had
re-
called positive events rated themselves less happy than
did
those
who
had
recalled negative events.
The
past events apparently
served
as an
extreme standard
of
comparison
for
rating current
well-being
and,
thus,
produced judgmental
contrast
effects.
In
a
follow-up
experiment,
Strack
et al.
showed that when
the
past
events were
recalled
vividly
and in
detail,
they
no
longer
func-
tioned
as
judgmental anchors;
instead,
they apparently
affected
the
respondents'
moods and,
thus,
had a
carryover
effect
on
judgments
of
life
satisfaction
(as in
Schwarz
&
Clore,
1983,
which
we
discussed
in the
Retrieval
and
Attitude Questions sec-
tion).
Strack
et al.
(1985) argued
that
the
backfire
effect
obtained
when
pallid events
are
used
as
standards
of
comparison
is
judg-
mental rather than
reflecting
the
mapping
of the
judgment onto
the
response
categories.
Given
the
simple categorical response
format
that
they
used, this interpretation seems quite reason-
able.
A
study
by
Higgins
and
Lurie
(1983)
indicates that
a
judg-
ment based
on a
standard
affects
how the
stimulus
is
remem-
bered, again suggesting that
the
contrast
effect
involves
more
than
the
mapping
process
(cf.
Herr
et
al.,
1984;
see
also
Dillehay
&
Jernigan,
1970,
and
Turner
&
Martin,
1984,
p.
294,
on
per-
ceptual
contrast).
Carryover
Versus
Backfire
in
Judgment
Carryover
effects
at the
judgment stage
for the
most
part
re-
flect
changes
in the
dimension
of
judgment
(as in the
agenda-
setting research)
or
changes
in the
rule used
to
arrive
at the
judg-
ment
(as in the
evenhandedness
findings).
When
the
object
or
issue
being judged
is
highly familiar
and
multifaceted,
such
as
the
President's performance
in
office,
respondents
may not
know
where
to
begin;
they
may, therefore,
be
guided
by
context
in
selecting
a
dimension
for the
evaluation. Carryover
effects
resulting
from
the
application
of the
norm
of
evenhandedness
are
likely
to
occur only under
fairly
circumscribed
conditions;
as
Schuman
and
Ludwig
(1983)
put it,
Context
effects
will
occur
whenever
two
questions
deal
with
differ-
ently
evaluated
competing
parties.
. . .
When
the
comparability
of
the
parties'
positions
is
made
salient,
the
norm
is
evoked
and
prescribes
comparability
of
treatment,
(p.
112)
Backfire
effects
at
this stage reflect
the use of
extreme stan-
dards
of
comparison,
or
judgmental
anchors.
Generally, back-
fire
effects
predominate when
the
prior items
are
extreme
in

CONTEXT
EFFECTS
307
some
way or
are
seen
as
dissimilar
(though
still comparable)
to
the
subsequent item. Carryover
effects
appear
to be the
rule
when
the
anchor
is
seen
as
representing
an
average
or
midpoint.
It is not
always
clear
when
context items will
be
used
as
stan-
dards
of
comparison
for
later
items.
One
obvious prerequisite
would
be
that
the
items
can all be
positioned
on a
single salient
dimension.
If the
question does
not
specify
any
single standard
and
many standards
are
reasonable
(as
with judgments
of
life
satisfaction,
in
which
some respondents
may
compare their cur-
rent
with
their past
lives,
whereas others compare themselves
with
similar others
or
with
the
average
person), then
the
ques-
tion
may be
susceptible
to
judgmental contrast
effects.
Tour-
angeau
and
Rasinski
(1986)
presented results suggesting
an-
other hypothesis. They
found
that
highly
partisan respondents
appear
to
answer
a
series
of
items
on
abortion
by
using
an
abso-
lute standard
but
that respondents
with
mixed views about
abortion appear
to
respond more
flexibly,
perhaps making com-
parisons
among
the
individual items.
Context
and
Response
Selection
The
final
step
in
answering
an
attitude question
is to
select
a
response.
In
both surveys
and
attitude-change
studies,
the
re-
sponse options
are
almost
always
prespecified.
Thus,
one
com-
ponent
of the
response-selection stage
is the
mapping
of the
judgment onto
one
of
the
response categories.
A
second
compo-
nent involves editing
the
chosen response.
The
editing
process
may
reflect
such considerations
as
avoiding inconsistent
or un-
desirable responses.
We
have
already noted that prior items
can
affect
the
map-
ping
process
by
changing
the
anchor
for
the
response scale (Os-
trom,
1970;
Upshaw,
1969).
The
later responses
are
generally
moved
in the
direction
of
contrast
(or
backfire)
away
from
ex-
treme anchors
but not
always.
In an
apparent exception
to the
general
rule,
Schwarz
and
Wyer
(1985)
reported
an
assimilation
(or
carryover)
effect,
in
which
later stimuli
are
seen
as
more
extreme
when
an
extreme anchor
has
been made salient
by a
prior ranking
task.
The
assimilation
to
extreme anchors
ap-
pears
to
involve
the
judgment-to-response mapping process;
the
effect
of the
prior ranking task disappears
when
each
point
on
the
rating scale
is
labeled
(Schwarz
&
Wyer,
1985, Experiment
5).
Sometimes
the
average
of a
group
of
related items
or the
middle alternative among
the
response categories
can
serve
as
an
anchor point
for the
mapping process
(Higgins
&
Lurie,
1983;
Schwarz
&
Hippler,
1987); with such moderate anchors,
assimilation
of the
response
to the
anchor appears
to be the
rule.
Carryover
and
backfire
effects
can
also arise
in the
editing
process. Prior items
can
produce consistency pressures that
affect
answers
to
later questions; prior items
can
also heighten
pressures
to
present oneself
as
moderate
by
giving
seemingly
inconsistent responses
to
related items.
Editing
for
Consistency
Social psychologists
have
assumed
for 3
decades
now
that
people
want
to be or at
least appear
to be
consistent
in
their
beliefs
(Abelson
et
al.,
1968;
Heider,
1958;
Tedeschi,
1981).
Some
of
the
early
work
done under
the
heading
of
cognitive
con-
sistency
is
quite similar
to the
methodological studies carried
out
in
exploring
item-context
effects.
For
example,
McGuire's
(1960)
"Socratic
effect"
research demonstrated that asking
re-
lated questions
on a
topic
can
produce changes
in the
answers.
According
to
McGuire,
asking questions about related
beliefs
can
make
the
relation among them more salient; once
the
rela-
tion
is
made
salient,
people
try to
reduce
the
inconsistencies
among
their beliefs. Although McGuire originally hypothesized
that consistency would
be
achieved only gradually,
over
days
or
weeks,
subsequent results (e.g.,
Wyer
&
Rosen, 1972)
suggest
that
much
of the
inconsistency reduction occurs during
the
ini-
tial session,
when
the
related
beliefs
are first
assessed.
McGuire's
(1960)
Socratic
effect
may be
relevant
to
attitude
measurement
in
surveys (see, e.g.,
Dillehay
&
Jernigan,
1970,
on
response consistency), because
it is
typical
in
surveys
to ask
related
questions
in a
topical block,
and
often
the
introduction
to the
items reinforces
the
connections among them.
The
changes
that result
from
the
juxtaposition
of
previously incon-
sistent
beliefs
are not
necessarily
artifactual
or
short-lived.
In
one
particularly dramatic
demonstration,
Rokeach
(1975)
showed
that
long-term changes
in
attitudes
and
behavior
can be
produced
by
confronting people with
discrepancies
among
their
values.
Consistency
effects
at the
editing stage
can be
distinguished
from
carryover
effects
at
earlier stages
in
several
ways.
Interpre-
tive
carryover
effects
typically
involve
unfamiliar
or
obscure
is-
sues,
whereas consistency pressures
are
likely
to
involve
famil-
iar
issues.
At the
retrieval stage, carryover
effects
are
sometimes
increased
by
scattering
the
context items
(Tourangeau
&
Rasin-
ski,
1986),
whereas consistency pressures should
be
heightened
when
the
related items
are
blocked.
In
addition,
the
carryover
effect
at the
retrieval stage results
from
a
priming process that
is
at
least partially independent
of
responses
to the
context
items;
the
consistency
effect
is, of
course, necessarily dependent
on
responses
to the
prior items.
Editing
for
Purposes
of
Self-Presentation
Respondents
may
report
an
answer that systematically dis-
torts
their underlying judgment.
No one
wants
to
embarrass
him-
or
herself
or to
create
an
uncomfortable situation.
The
interview
is a
social interaction,
and
respondents
may
select
an-
swers
to
present themselves
in a
favorable
light. This desire
to
present oneself
favorably
can
take
some
nonobvious
forms.
Mc-
Guire
and
Millman
(1965) observed that people warned
of an
impending attack
on
their
beliefs
sometimes show anticipatory
belief changes. McGuire
and
Millman attributed this
effect
to
the
receivers' expecting
to be
persuaded
and to
their desire
to
save
face
by
appearing
to
have
agreed with
the
message
all
along.
A
number
of
subsequent
findings
support this self-presentation
account: Respondents snap
back
to
their original position
when
the
anticipated message
is
cancelled
(Cialdini
&
Petty,
1981);
when
the
position
of the
impending message
is
unclear,
the an-
ticipatory change
is in the
direction
of
moderation (Cialdini,
Levy,
Herman,
&
Evenback,
1973).
The
anticipatory-attitude-change
findings
suggest
that people
do
not
want
to
appear gullible
and
that
the
middle ground
ap-
pears
safer
than either extreme. Other things being equal, then,
respondents whose attitudes
are
being measured
may
attempt
to
create
the
impression
of
being moderate
on a
topic. Respon-

308
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSKI
dents
can
create
the
impression
of
moderation
by
using middle
response
categories.
The
response options
are
often
restricted,
however,
to
dichotomous,
pro-con
alternatives.
In
such
cases,
respondents
can
still create
the
appearance
of
moderation
by
balancing
pro and con
answers across
a
series
of
items
on the
same topic.
For
example, respondents with moderate pro-
choice
views
on
abortion
may be
eager
to
give
pro-life
responses
after
they
have
given
a
series
of
pro-choice
answers,
just
to
make
clear
to the
interviewer that they
are not
partisans
or
extremists
on the
issue.
Their responses
to a
given
item may,
in
part,
reflect
their
fears
that
their
earlier answers
may
have created
an
errone-
ously
extreme impression. They
may
edit their later answers
to
offset
this impression. (For
a
related hypothesis,
see
Turner
&
Martin,
1984,
p.
293,
on
response contrast.)
Carryover
Versus
Backfire
in
Editing
When
is the
editing process
likely
to
enhance consistency
as
opposed
to
moderation?
The
consistency
effect
should
be
great-
est
when
a
close, even logical, relation exists among
the
items
(McGuire,
1960),
when
this
relation
has
been made salient
(Wyer
&
Rosen,
1972),
and
when respondents
are
sufficiently
involved
in the
issue
to
care about being consistent.
The
moder-
ation
effect
is
likely
to
occur
when
involvement
in the
issue
is
low
or
when respondents
do, in
fact,
have moderate
views,
which
have
not
been conveyed
by
their responses
to the
prior
items. Thus,
a
series
of
dichotomous items
on a
topic that most
people would agree
(or
disagree) with
may
encourage inconsis-
tent
responses
to a
later item, especially
if the
later item
is
less
consensual
than
those
that
preceded
it.
Variables
Affecting
the
Impact
of
Context
Given
the
multiplicity
of
mechanisms underlying context
effects
and the
empirical
confusions
surrounding them,
it is not
always
possible
to
specify
exactly
when
each type
of
context
effect
will
occur.
In
some cases, there
are
simply
no
relevant
data
or
theories
to
guide predictions.
In
other cases, there
are
data,
but
they give
conflicting
answers. Several investigators
have
reported
attempts
to
replicate context
effects,
only
to
have
the
replications
fail
to
obtain
any
effects
of
context (Bishop
et
al.,
1985; Schuman
&
Presser,
1981,
Appendix
A) or
only
to
obtain
effects
in the
opposite
direction
from
the
original result
(e.g.,
Schuman
&
Presser,
1981,
pp.
42-43).
The
resolving
power
of the
data
is
poor,
and
this places limits
on our
ability
to say
with
any
precision when each type
of
effect
will
occur.
Aside
from
that, context
effects
result
from
complex processes
that
may not
allow simple answers
to
such apparently straight-
forward
questions
as
"When
will
respondents reject accessible
beliefs
rather than being swept along
by
them?" Such questions
are
likely
to
prove
as
difficult
to
answer
as
similar questions
about,
for
example, attitude change
(e.g.,
"When
will
an
expert
be
the
most
effective
persuader?").
For the
time being
at
least,
the
answer must
be
that
it
depends;
and we
have
yet to
identify
all the
variables
that
it
depends
on.
Within these limits,
however,
we
have
identified
in
previous
sections
a
number
of
variables that
affect
the
size
and
direction
of
context
effects.
In
this section
of
the
article,
we
highlight these
variables. Most
of
them
are
related
to the
reliability
of one or
Table
3
Variables
Affecting
Size
and
Direction
of
Context
Effects
Variable
Effect
Variables
affecting
interpretation
Issue familiarity
Attitude accessibility
Unfamiliar
issues more
susceptible
to
interpretive
carryover
effects.
Inaccessible
attitudes
susceptible
to
interpretive carryover
effects.
Variables
affecting
retrieval
Mixed,
or
conflicted,
beliefs
Issue expertise
and
involvement
Question
form
(direct opinion
versus related judgment)
Obviousness
of
context
Depth
of
thought
(mode
and
pace
of
administration)
Mixed
respondents more
susceptible
to
carryover
effects
at
retrieval.
Expert
and
involved respondents
less likely
to
show carryover
effects.
Related judgment items more
susceptible
to
carryover
effects.
Obvious contexts
may be
discounted.
Use of
self-administered
questionnaires
and
conducting
interviews
at a
slow pace
may
reduce context
effects.
Variables
affecting
judgment
Complexity
of
judgment
Similarity
and
extremity
of
context items
Multifaceted
issues susceptible
to
judgmental carryover
effects;
comparative judgments
susceptible
to
backfire
(i.e.,
judgmental
contrast)
effects.
Extreme
or
dissimilar anchors
foster
backfire
(i.e.,
contrast)
effects.
Variables
affecting
response selection
Characteristics
of
anchor (labels,
midpoints versus extremes)
Salience
of the
relation among
items
Labeling
every
option
may
reduce anchoring
effects
during response-mapping
stage; extreme anchors foster
backfire
(contrast)
effects.
Heightening salience
may
increase consistency
effect
for
involved
respondents
and
moderation
effect
for
uninvolved
respondents.
more
of the
processes involved
in
answering
an
attitude
ques-
tion;
it is
when
the
response process
is
unreliable that
it is
most
readily
affected
by
context. These variables
can be
grouped
ac-
cording
to the
stage
of the
attitude-response process that they
primarily
affect.
Table
3
contains
a
summary
list
of
these vari-
ables.
Interpretation
Stage
A
crucial component
of
the
interpretation
of an
attitude item
is
determining which attitude
is
relevant
to the
question.
At
least
two
variables—the
familiarity
of the
issue
and the
accessi-
bility
of
the
attitude—can
influence
whether this determination

CONTEXT
EFFECTS
309
will
be
made reliably
or
will
be
influenced
by the
context
of the
item.
Issue
familiarity.
When
the
issue
is
ambiguous
or
unfamiliar
to
most respondents (e.g.,
the
Monetary Control
Bill),
respon-
dents
may
have
difficulty
in
identifying
a
relevant attitude struc-
ture; they must search
for
one,
and
context
can
bias this search.
By
contrast,
when
the
issue
is a
familiar
one or
when
the
item
wording
clarifies
the
nature
of the
issue,
this
interpretive carry-
over
effect
appears
to be
eliminated.
In
fact,
with highly familiar
issues,
an
attitude
structure
may be
activated automatically
when
the
issue
is
confronted (Fazio
et
al.,
1986).
Context
effects
with
unfamiliar issues generally
run in the
direction
of
carry-
over
(Herr
et
al.,
1984; Tourangeau
&
Rasinski,
1986; Trope,
1986),
with respondents interpreting
the
target item
in
terms
of
the
attitude evoked
by the
context items.
If
the
context items
are to
disambiguate
an
unfamiliar issue,
it may be
important
that they come
in a
block that
is
placed immediately
before
the
target
item (Tourangeau
&
Rasinski,
1986).
When
the
context
items
are
scattered
or
otherwise disguised, respondents
may not
make
the
interpretive connection between
the
target
and
con-
text
items.
Attitude
accessibility.
Issues
so
obscure that
few
respondents
know
which
(if
any)
of
their attitudes
are
relevant
to the
target
item
are an
extreme case,
but
even
with
familiar
issues, individ-
uals
may
differ
markedly
in how
readily they locate
the
relevant
attitude
or
attitudes
when confronted with
an
item. Fazio
et al.
(1986)
and
Fazio
and
Williams
(1986)
have
argued
that
individ-
uals
differ
in the
accessibility
of
their attitudes
on a
given
topic.
It
seems reasonable
to
assume that respondents with
highly
ac-
cessible attitudes
on an
issue
will
identify
(and retrieve) their
attitudes reliably
and
that those
with
less accessible attitudes
will
be
less reliable
in
locating relevant attitude structures
and
more susceptible
to the
effects
of
context. Several variables
are
likely
to
relate
to an
attitude's
long-term accessibility, including
expertise
and
involvement
in the
issue (Fiske
&
Kinder,
1981),
attitude
centrality,
and
direct
experience with
the
issue.
One
line
of
evidence reviewed
by
Converse (1964, 1975)
supports
this reasoning: Expertise
and
involvement lead
to
greater con-
sistency
in
responses
over time. Other variables, such
as
recent
retrieval
of the
attitude
or
related attitudes,
can
have
short-term
effects
on
attitude accessibility.
Retrieval
Stage
We
have
argued
that
retrieval
often
involves
two
steps—re-
trieving
an
overall attitude structure
and
retrieving individual
beliefs
or
feeling
from
within that structure.
The first
step
is
closely linked with interpreting
the
issue;
it may be
difficult
to
identify
the
relevant structure without activating
it
(Fazio
et
al.,
1986).
Thus,
the
reliability
of
retrieving
an
overall structure
de-
pends
on two
variables that
we
have
already
mentioned—the
familiarity
of the
issue
and the
chronic accessibility
of the
atti-
tude. Five additional variables
may
influence
the
retrieval
of
individual
beliefs
and the
susceptibility
of
this stage
of the re-
trieval process
to the
effects
of
context.
Mixed,
or
conflicted,
beliefs.
The
view
that answering
an
atti-
tude question
involves
a
sampling
of
relevant beliefs suggests
that
the
response
process
will
be
unreliable when
the
population
of
beliefs
being sampled
is
heterogeneous,
that
is,
when
the re-
trieved attitude structure includes beliefs supporting
both
sides
of
an
issue.
We
have
found
that
respondents
with mixed views
are
most sensitive
to
differences
in
item wording (Tourangeau
&
Rasinski,
1986);
Smith
(1982)
reviewed evidence that mid-
dle-of-the-road
respondents (who
are
likely
to
have
mixed
views)
are
most vulnerable
to
context
effects
as
well. Respon-
dents
with internally
consistent
views
are
unlikely
to be
affected
by
contextually
induced priming because
the
characteristics
of
samples
of
beliefs
will
not
vary
much when
the
underlying
pop-
ulation
is
homogeneous. When
beliefs
are
mixed,
however,
con-
text
may
shift
the
balance
by
rendering beliefs
on one
side
of
the
issue more accessible
to
retrieval (Tourangeau
&
Rasinski,
1986).
Issue
expertise
and
involvement.
Other
researchers
have
sug-
gested that issue expertise
and
involvement
may
decrease
sus-
ceptibility
to
context
effects
(Fiske
&
Kinder,
1981;
see
also
Rugg
&
Cantril,
1944,
on
attitude
crystallization).
Expertise
about
an
issue
refers
to how
much respondents know about
it;
involvement
refers
to how
much they care about
it. The two
vari-
ables
are
doubtless generally
correlated.
For our
purposes,
the
distinction
is not
crucial, because both experts
and
involved
respondents
are
likely
to
have
tightly interconnected
attitude
structures.
For
such structures,
the
retrieval
process
is
thought
to be
thorough
and
reliable (Smith,
Adams,
&
Schorr,
1978)—
and
unlikely
to be
affected
by
context.
These hypotheses about issue expertise
and
involvement
are
closely
related
to our
earlier hypothesis concerning
the
familiar-
ity of the
issue.
An
unfamiliar issue
is
simply
one for
which
few
respondents
are
expert
or
involved.
We
prefer
to use the
term
familiarity
when characterizing issues
and the
terms
involve-
ment
and
expertise
when characterizing individual respon-
dents.
Still,
we
could restate
our
earlier hypothesis regarding
interpretive
carryover
effects
by
saying
that
expert
and
involved
respondents ought
to be
less
prone
to
such
effects.
Question
form:
Direct
opinions
versus
related
judgments.
Even
when
it is
clear
what
overall
attitude
is
relevant,
a
question
may
still
leave
considerable ambiguity about what
beliefs
are
relevant. Some attitude items
ask for
relatively
clear-cut
agree-
disagree
responses. These
direct
opinion items
may
allow
a
straightforward
readout
of a
belief.
By
contrast,
many
of the
studies that
find
large context
effects
use
questions
that
call
for
a
related judgment, such
as
predicting
the
frequency
of an
event
or
assigning blame
in a
rape case. Respondents
may not be
sure
what
beliefs
are
relevant
to
these
related
judgments
(or how to
use
beliefs
that
they
see as
relevant). When
it is not
clear what
beliefs
are
relevant because
the
item requires
a
complex judg-
ment
rather than
a
direct statement
of
belief,
the
retrieval pro-
cess
will
be
unreliable
and
open
to the
effects
of
context. Under
these
circumstances,
carryover
effects
produced
by
priming
may
be
especially prevalent.
Obviousness
of
context.
Martin (1986,
p.
494) argued
that
most successful priming studies
have
used rather subtle
manip-
ulations
of
context;
for
example,
the
priming task
is
sometimes
presented
as
part
of a
different
study
from
the
target judgment
(Wyer
et
al.,
1985). When
the
prior context
is
made obvious,
respondents
may
discount
or
reject
the
material that
it has
ren-
dered
accessible,
because they judge
it to be
irrelevant
to the
target item
or
they
find it
disagreeable
in
some other way.
For
example,
Schwarz
and
Clore
(1983)
showed that calling
respon-

310
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSKJ
dents' attention
to the
weather eliminated
its
effect
on
judg-
ments
of
life
satisfaction.
The
basic priming
effect
is
thought
to
reflect
an
automatic process (i.e., spreading activation) and,
therefore,
should
not
depend
on
respondents' conscious aware-
ness
of the
connection between
the
context
and
target items.
The
discounting
or
rejection
of
context,
however,
appears
to in-
volve
more controlled
processes.
3
Depth
of
thought.
At first
blush,
it
might seem obvious that
context
effects
are the
products
of
superficial
thinking about
an
issue
and
that
if
respondents could
be
induced
to
think more
deeply
before
they
answered, context
effects
would
disappear
or
at
least
be
reduced. There are,
however,
reasons
to
suppose that
the
opposite
may
sometimes
be the
case: Under some circum-
stances,
increased
thoughtfulness
may
magnify
rather than
di-
minish
the
effects
of
context.
In a
related vein, Wilson
and
Dunn
(1986)
and
Wilson, Dunn,
Bybee,
Hyman,
and
Rotondo
(1984)
have
shown
that increased thoughtfulness
can
decrease
the
correlation between attitudes
and
behaviors
on an
issue.
Our
model
of the
response process assumes that answers
to
attitude
questions
often
reflect
the
sampling
of
individual
beliefs.
If the
sampling
process
is
biased
by
prior items,
additional
thought
may
simply
yield
additional
beliefs
that
are
consistent with
those
sampled earlier (Millar
&
Tesser,
1986; Tesser
&
Leone,
1977).
Of
course, this
effect
must
have
some limit.
As
respon-
dents sample their
beliefs
on a
topic exhaustively, their answers
should
eventually
become stable.
Thus,
the
overall relation
be-
tween
depth
of
thought
and
size
of the
context
effect
may be
curvilinear.
And the
relation
may be
even
more complex
than
that;
as we
have
just noted,
when
the
context
is
obvious, respon-
dents
may
engage
in
more
thoughtful,
controlled processing
and,
in
formulating their
final
answers, they
may
exclude some
of
the
material that they
have
retrieved. Variables that encour-
age
deeper thought
may
make
it
more
likely
that such
controlled
processes
will
be
engaged.
A
number
of
procedural variables
can
encourage
or
discour-
age
thoughtful
processing. Multiple items
on the
target issue
may
encourage repeated attempts
at
retrieval and, thus, more
thorough
sampling
of
relevant
beliefs.
In
addition,
the
pace
of
the
interview (Tourangeau,
Lessler,
&
Salter, 1986)
and the
mode
of
administration
may
affect
how
much thought respon-
dents
give
to
their answers. Recent studies
by
Bishop
and
Hip-
pier
(1986)
indicate that some context
effects
(but
not
all)
are
reduced when
the
questions
are
self-administered rather than
administered
in a
telephone
interview.
Bishop
and
Hippler
at-
tributed
this
mode
difference
to the
slower pace (and greater
thought) typical
of
self-administered questionnaires.
Judgment
Stage
Implicitly
or
explicitly,
different
forms
of
attitude questions
require respondents
to
make
different
types
of
judgments,
rang-
ing
from
relatively
simple judgments
of
agreement
to
extremely
complex
judgments
of
fairness
or
equity.
Even
when respon-
dents retrieve
a
stable
set of
beliefs
or
values relevant
to the
question
at
hand, they
may not
know
exactly
how to use
them
in
making
the
requisite
j
udgment.
As
Fischoff
et
al.
(1980)
pointed
out,
respondents
may
need
practice
in
thinking through
the im-
plications
of
their values
in new
settings.
Complexity
of
the
judgment.
One
variable that
affects
the
reliability
of
the
judgment process
is
the
complexity
of
the
judg-
ment
involved:
The
simpler
the
judgment,
the
less
likely
it is to
be
affected
by
context.
At one
extreme, items
that
map
directly
onto
existing,
highly
accessible beliefs
or
overall evaluations
are
likely
to
produce stable answers;
thus,
an
item asking whether
abortion
is
murder
is, for
most
pro-life
respondents, unlikely
to
require much judgment
at
all. Items
that
map
directly onto
existing
beliefs
are
probably
very
much
the
exception
in
sur-
veys,
with almost
all
items requiring
at
least some judgment.
Many
issues,
such
as the
president's performance
in
office
or
one's
life
satisfaction,
are
inherently multidimensional;
as a
result,
respondents
may
have
a
difficult
time formulating
a co-
herent judgment strategy. With such issues, many
relevant
di-
mensions
of
judgment, each with
its own
criteria
and
standards
of
comparison,
have
to be
integrated
in
some way.
In
such cases,
respondents
may
simplify
their task
by
focusing
on
dimensions
and
standards
of
comparison
that
readily come
to
mind,
and
context
can
determine which dimensions
or
standards come
to
mind.
Even
questions that seem
to be
posed
in
absolute terms
may
involve comparisons
as a key
component
of the
judgment pro-
cess.
For
example, nothing
in the
item about legal abortions
for
women
who are
married
and do not
want
any
more children
appears
to
require comparison
processes.
Yet at
least some
re-
spondents
seem
to
answer this item
by
comparing
the
strength
of
the
reason
given
for
abortion
in
this
case
with
the
stronger
reason mentioned
in the
preceding birth-defect item.
As
Hig-
gins
and
Lurie (1983) argued, many absolute judgments
are in
fact
relative. Thus, when
the
criteria
for an
absolute judgment
are
unclear
or
when
the
judgment
is
inherently comparative
and
an
explicit standard
is
lacking,
the
judgment process
may be
unreliable—and
may
fall
prey
to the
influence
of
contextually
salient criteria
and
anchors.
Characteristics
of
the
context
items.
If a
judgment
is
based
on
a
comparison
with
a
prior item
or
some other anchor sug-
gested
by the
context, then
the
size
and
direction
of the
context
effect
depend
on
such
characteristics
of the
anchor
as its ex-
tremity
and its
similarity
to the
target. Extreme standards
of
comparison
are
likely
to
produce
judgmental
contrast
effects
(Herr
et
al.,
1984;
Herr,
1986;
Judd
&
Harackiewicz,
1980;
Schwarz
&
Strack,
1985),
whereas anchors representing moder-
ate
values (such
as the
average
for a
category
or the
midpoint
of
a
scale) seem
to
produce assimilation
effects
as a
rule (Schwarz
&
Hippler,
1987).
One
reason that extreme context items
may
produce contrast (or,
in our
terms,
backfire)
effects
is
that
the
context
items
may be
seen
as
dissimilar
or
unrelated
to the
tar-
get
item. When
the
standard
suggested
by
context
is
dissimilar
from
the
target, contrast
is the
usual result.
3
The
effects
of
blocking
the
context
items
described
here
depend
on
the
issue's
being
a
familiar
one.
When
the
issue
is
unfamiliar,
the
effects
of
blocking
the
context
items
are
reversed,
with
blocked
items
produc-
ing
carryover
effects.
With
unfamiliar
issues,
context
appears
to
affect
the
interpretation
of
the
item,
and the
context
and
target
items
must
be
presented
in a
block
for an
interpretive
carryover
effect
to
occur
(Tour-
angeau
&
Rasinski,
1986).
Apparently,
with
obscure
issues,
the
blocked
presentation
encourages
respondents
to
infer
that
the
target
item
con-
cerns
the
same
issue
as the
context
items.

CONTEXT
EFFECTS
311
Response-Selection
Stage
The final
stage
of the
response process,
in
which
an
answer
is
selected
from
among
the
options presented,
involves
at
least
two
processes that
can be
affected
by
context.
The
judgment must
be
mapped
onto
one of the
response categories;
in
addition,
the
response
may be
edited
for
consistency
with
prior responses
or
to
create
a
favorable
impression.
The
mapping process
can be
affected
by the
nature
of the
anchors
or
labels
given
in the an-
swer
categories. Context
can
also create pressure
to
appear con-
sistent
or
moderate.
Anchors
for the
response
scale.
Even
when respondents
are
clear that
they
agree
or
disagree with
an
item, they
may
have
difficulty
in
mapping this judgment onto
the
response scale.
Re-
searchers usually provide labels
to
help anchor
the
mapping
process, sometimes labeling
every
option
and
sometimes label-
ing
only
the
extremes. Labeling
all the
options
may
reduce con-
text
effects
(Schwarz
&
Wyer,
1985)
or may
encourage
the use
of the
middle response category
as an
anchor
for the
mapping
process.
Use of the
midpoint
or
middle category
as an
anchor
appears
to
produce
assimilation
effects
(Higgins
&
Lurie,
1983;
Schwarz
&
Hippler,
1987). Higgins
and
Lurie
(1983)
demon-
strated that particularly large context
effects
can
result
when
one
anchor
is
used
in
making
an
initial judgment
but a
different
anchor
is
used
in
mapping
the final
response.
Heightening
the
relations
among
items.
In the
original
So-
cratic
effect
research,
McGuire
(1960)
assumed that,
when
the
connection among logically related items
was
made salient,
people would attempt
to
reduce inconsistencies among their
re-
sponses
to the
items. Subsequent research
(Wyer
&
Rosen,
1972)
shows
that
it is
essential
to the
Socratic
effect
that
the
connections among
the
items
be
salient
to the
respondents.
Some
of the
variables mentioned earlier, such
as the
obvious-
ness
of the
context items
and the
depth
of
thought
given
to the
answers,
may
help make
the
relation among items salient. Edit-
ing
for
consistency
(as
well
as
editing
to
create
an
impression
of
being
a
moderate
on an
issue) presupposes
a
close relation,
even
a
logical one, among
the
items. With less directly related items,
we
would expect
the
editing process
to
reflect
other considera-
tions, such
as
social desirability.
The
pressure
to
appear consistent
or
moderate
is
greater
when
the
connection between
the
context
and
target items
is
obvious. When respondents care about
the
issue, heightening
the
connections among
the
items
is
likely
to
produce
consis-
tency
effects;
when respondents
do not
care about
the
issue,
it
is
likely
to
produce moderation
effects.
Summary
and
Conclusions
Context
can
product either
backfire
or
carryover
effects
that
change
the
apparent distribution
of
opinion
on
related issues.
Context
can
produce these
shifts
by (a)
changing what object
or
issue
is
being judged;
(b)
changing
the
considerations
and
beliefs
that enter into
the
judgment;
(c)
changing
the
dimensions,
stan-
dards,
or
norms that
are
applied
in
making
the
judgment;
and
(d)
changing
how the
judgment
is
reported. Context
effects
are
often
unstable; this instability
may
reflect
the
number
and
com-
plexity
of the
processes that
are
responsible
for the
effects,
as
well
as the
large
number
of
variables
that
can
influence
the
size
and
direction
of
context
effects.
Table
3
offers
a
partial list
of
these variables, along with
a
number
of
specific
embodiments
of
them.
It
should
be
apparent
from
our
account that context
effects
are not
merely
artifactual.
First,
context
effects
are
inextricably
bound
up
with
both
the
structure
of
attitudes
and the
process
of
answering attitude questions.
The
beliefs
and
feelings
that
constitute
an
attitude
are
often
complex
and
mixed.
The
issues
that appear regularly
on
surveys
are
ones that engender endur-
ing
controversy,
and it is
difficult
for
most people
to
take
a
strong
and
clear stand
on
these issues. Given
the
underlying heteroge-
neity
of
people's
beliefs,
it is no
surprise that they
are
suscepti-
ble to the
subtle
effects
of
item wording
or
item context.
And
even
when
beliefs
are
unmixed, judgments
are
still
affected
by
context,
as
long
as the
relevant dimensions
of
judgment
are un-
specified
and the
standards
of
comparison
are
unclear.
A
second reason
for
viewing these
effects
as
more
than
mere
artifacts
is
that
they
are not
necessarily short-lived.
The
attitude
changes
produced
by
consistency pressures
are
known
to be en-
during
on
some
occasions
(e.g.,
Freedman,
1965; Rokeach,
1975).
Most
of the
mechanisms
that
we
distinguish
in
this
arti-
cle can
probably lead
to
enduring changes
in
attitudes.
For ex-
ample, once
a
dimension
is
used
in
making
a
judgment,
it
prob-
ably
tends
to be
reused. More generally,
Lingle
and
Ostrom
(1979)
and
Lingle
et al.
(1979)
have
argued that once
a
judg-
ment
is
made,
it
tends
to be
retrieved
and
applied
in
rendering
related judgments later. Responses
influenced
by
context
may
thus persist
in
memory
and
affect
subsequent responses.
Other
Order
Effects
We
have
not
attempted
to
make
this
article comprehensive
in its
coverage
of
order
effects
in
surveys.
A
number
of
such
effects
do not
involve
the
specific
content
of the
prior items.
For
example,
survey
researchers tend
to put
sensitive items toward
the end of the
questionnaire, partly because they believe that
during
the
course
of the
interview,
rapport gradually builds
be-
tween
the
interviewer
and the
respondent
so
that respondents
are
more
likely
to
admit
to
embarrassing
facts
about themselves
later
in the
interview,
after
a
positive relationship
has
been
es-
tablished. Another
frequently
cited sequence
effect
involves
re-
spondent fatigue.
Tourangeau
et al.
(1986)
showed that under-
reporting
of
dental visits increases
as the
interview
gets
longer.
Aside
from
sheer
fatigue,
the
effects
on
reporting
of
interview
length
or
position within
the
interview
may
reflect
the
respon-
dents'
changing criteria
for
reporting. Criterion
shifts
may be
especially
marked
when
positive responses
to an
item trigger
a
series
of
follow-up
questions,
a
contingency that respondents
are
likely
to
note during
the
course
of a
long
interview.
Tourangeau
et al.
(1986)
showed
how
context
effects
can be
put
to
positive use. They argued
that
warm-up questions
can
trigger
the
recall
of
specific
events,
decreasing
the
amount
of
underreporting. Retrieval cues
can be
valuable
when
they
in-
crease
the
effectiveness
of the
retrieval
process—rather
than
bi-
asing
it.
The
sequence
of the
response options
may
also make
a
differ-
ence
to the
answers. Tourangeau
(1984,
p. 90)
argued that
re-
spondents
may use a
satisficing
rule
in
selecting
a
response,
es-
pecially
when
they
are
confronted with
a
long
series
of
response

312
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSK1
options.
This
hypothesis
implies
that
under
the
right
circum-
stances,
primacy
effects will
be
observed,
a
prediction
recently
confirmed
by
Krosnick
and
Alwin
(1986).
Implications
for
Practice
Although
we
have
not
explicitly
tried
to
spell
out the
implica-
tions
for the
practice
of
attitude
measurement
of the
variables
listed
in
Table
3,
some
of the
variables
clearly
carry
such
im-
plications:
Questions
involving
unfamiliar
issues
should
define
the
issues;
questions
requiring
complex
judgments
should
be
avoided,
or the
relevant
dimensions
and
judgmental
anchors
should
be
clearly
specified;
respondents
should
be
encouraged
to
admit
that
they
do not
have
an
opinion
or
that
their
beliefs
are
conflicted;
and so on. We are
reluctant
to
offer
such
recom-
mendations
until
the
range
of
context
effects—and
the
mecha-
nisms
underlying
them—is
better
understood.
Opinion
re-
searchers
often
bemoan
the
fragmentary
and
noncumulative
nature
of
research
on
context
effects.
We
hope
that
our
model
of
the
process
of
responding
to
attitude
questions
can
help
bring
some
order
to
this
tangled
area.
References
Abelson,
R.
(1981).
Psychological status
of
the
script concept. American
Psychologist,
36,
715-739.
Abelson,
R.,
Aronson,
E.,
McGuire,
W.,
Newcomb,
X,
Rosenberg,
M-,
&
Tannenbaum,
P.
(1968).
Theories
of
cognitive
consistency:
A
sourcebook.
Chicago: Rand
McNally.
Anderson,
J.
(1978).
Language,
memory,
and
thought. Hillsdale,
NJ:
Erlbaum.
Anderson,
J.
(1983).
The
architecture
of
cognition.
Cambridge,
MA:
Harvard
University Press.
Anderson,
N.
(1974). Cognitive algebra.
In L.
Berkowitz
(Ed.),
Ad-
vances
in
experimental
social
psychology
(Vol.
7, pp.
1-101).
New
%rk:
Academic Press.
Anderson,
N.
(1981).
Foundations
of
information
integration
theory.
New
York:
Academic Press.
Anderson,
N.,
&
Hubert,
S.
(1963).
Effects
of
concomitant verbal recall
on
order
effects
in
personality impression formation.
Journal
of
Ver-
bal
Learning
and
Verbal
Behavior,
2,
379-391.
Bishop,
G., &
Hippler,
H. J.
(1986,
May).
A
cross-national
comparison
of
question
form
effects
in
self-administered
vs.
telephone
surveys.
Pa-
per
presented
at the
41
st
Annual Conference
of
the
American Associ-
ation
for
Public Opinion Research,
St.
Petersburg,
FL.
Bishop,
G.,
Oldendick,
R.,
&
Tuchfarber;
A.
(1985).
The
importance
of
replicating
a
failure
to
replicate: Order
effects
on
abortion items. Pub-
lic
Opinion
Quarterly,
49,
105-144.
Bower,
G.
(1981).
Mood
and
memory. American
Psychologist,
36,
129-
148.
Bower,
G.,
Black,
J.,
&
Turner,
T.
(1979).
Scripts
in
comprehension
and
memory.
Cognitive
Psychology,
11,
177-220.
Bradburn,
N.
(1982). Question-wording
eifects
in
surveys.
In R.
Ho-
garth
(Ed.),
Question
framing
and
response
consistency
(pp.
65-76).
San
Francisco: Jossey-Bass.
Bransford,
J.,
&
Johnson,
M.
(1972).
Contextual prerequisites
for un-
derstanding: Some investigations
of
comprehension
and
recall.
Jour-
nal
of
Verbal
Learning
and
Verbal
Behavior,
11,
717-726.
Cantor,
N.,
&
Mischel,
W.
(1977).
Traits
as
prototypes:
Effects
on
recog-
nition memory.
Journal
of
Personality
and
Social
Psychology,
35, 38-
49.
Chaiken,
S.
(1980).
Heuristic versus systematic information processing
and the use of
source versus message cues
in
persuasion.
Journal
of
Personality
and
Social
Psychology,
39,
752-766.
Cialdini,
R.,
Levy,
C,
Herman,
C,
&
Evenback,
S.
(1973).
Attitudinal
politics:
The
strategy
of
moderation.
Journal
of
Personality
and
So-
cial
Psychology,
23,
100-108.
Cialdini,
R.,
&
Petty,
R.
(1981). Anticipatory opinion
effects.
In R.
Petty,
T.
Ostrom,
& T.
Brock
(Eds.),
Cognitive
responses
inpersuasion
(pp.
217-235).
Hillsdale,
NJ:
Erlbaum.
Collins,
A.,
&
Loftus,
E.
(1975).
A
spreading-activation
theory
of
se-
mantic processing.
Psychological
Review,
82,407-428.
Converse,
P.
(1964).
The
nature
of
belief systems
in
mass publics.
In D.
Apter
(Ed.),
Ideology
and
discontent
(pp.
206-261).
New
York:
Free
Press.
Converse,
P.
(1975). Public opinion
and
voting behavior.
In F.
Greenstein
& N.
Polsby
(Eds.),
Handbook
of
political
science
(Vol.
4,
pp.
75-170).
Reading,
MA:
Addison-Wesley.
Dillehay,
R.,
&
Jemigan,
L.
(1970).
The
biased questionnaire
as an in-
strument
of
opinion change.
Journal
of
Personality
and
Social Psy-
chology,
IS,
144-150.
Fazio,
R.,
Sanbonmatsu,
D.,
Powell,
M.,
&
Kardes,
F.
(1986).
On the
automatic
activation
of
attitudes.
Journal
of
Personality
and
Social
Psychology,
37,229-238.
Fazio,
R.,
&
Williams,
C.
(1986). Attitude accessibility
as a
moderator
of
the
attitude-perception
and
attitude-behavior
relations:
An
inves-
tigation
of the
1984 presidential election.
Journal
of
Personality
and
Social
Psychology.
51,
505-514.
Fischoff,
B.,
Slavic,
P., &
Lichtenstein,
S.
(1980). Knowing what
you
want:
Measuring labile values.
In T.
Wallsten
(Ed.),
Cognitive
pro-
cesses
in
choice
and
decision
behavior
(pp.
117-131).
Hillsdale,
NJ:
Erlbaum.
Fiske,
S. X, &
Dyer,
L.
(1985). Structure
and
development
of
social
schemata: Evidence
from
positive
and
negative transfer
effects.
Jour-
nal
of
Personality
and
Social
Psychology,
18,
839-852.
Fiske,
S. X, &
Kinder,
D. R.
(1981).
Involvement, expertise,
and
schema
use:
Evidence
from
political cognition.
In N.
Cantor
& J.
Kihlstrom
(Eds.),
Personality,
cognition,
and
social
interaction
(pp.
171-192).
Hillsdale,
NJ:
Erlbaum.
Freedman,
J.
(1965). Long-term behavioral
effects
of
cognitive disso-
nance.
Journal
of
Experimental
Social
Psychology,
1,
145-155.
Graesser,
A.,
Gordon,
S.,
&
Sawyer,
J.
(1979).
Recognition memory
for
typical
and
atypical actions
in
scripted activities:
Xests
of a
script
pointer
+ tag
hypothesis.
Journal
of
Verbal
Learning
and
Verbal
Be-
havior.18,
319-332.
Grice,
H.
(197
5).
Logic
and
conversation.
In
P.
Cole
&
T.
Morgan
(Eds.),
Syntax
and
semantics:
Vol.
3.
Speech
acts
(pp.
41-58).
New
York:
Seminar
Press.
Hamilton,
D.
(1981).
Cognitive
processes
in
stereotyping
and
intergroup
behavior.
Hillsdale,
NJ:
Erlbaum.
Hastie,
R.
(1981).
Schematic principles
in
person memory.
In E. T.
Hig-
gins,
C. P.
Herman,
&
M.
Zanna
(Eds.),
Social
cognition:
The
Ontario
Symposium
(Vol.
1, pp.
39-88).
Hillsdale,
NJ:
Erlbaum.
Hastie,
R.,
&
Park,
B.
(1986).
The
relationship
between
memory
and
judgment
depends
on
whether
the
judgment
task
is
memory-based
or
on-line.
Psychological
Review,
93,258-268.
Haviland,
S.,
&
Clark,
H.
(1974).
What's new? Acquiring
new
informa-
tion
as a
process
in
comprehension.
Journal
of
Verbal
Learning
and
Verbal
Behavior,
13,512-521.
Heider,
F.
(1958).
The
psychology
of
interpersonal
relations.
New
York:
Wiley.
Herr,
P.,
Sherman,
S.,
&
Fazio,
R.
(1984).
On the
consequences
of
prim-
ing:
Assimilation
and
contrast
effects.
Journal
of
Experimental
Social
Psychology,
19,
323-340.
Herr,
P. M.
(1986).
Consequences
of
priming: Judgments
and
behavior.
Journal
of
Personality
and
Social
Psychology,
51,
1106-1115.

CONTEXT EFFECTS
313
Higgins,
E. T., &
King,
G.
(1981).
Accessibility
of
social
constructs:
Informatidn
processing consequences
of
individual
and
contextual
variability.
In N.
Cantor
& J.
Kihlstrom
(Eds.),
Personality,
cognition,
and
social
interaction
(pp.
69-122).
Hillsdale,
NJ:
Erlbaum.
Higgins,
E. X, &
Lurie,
L.
(1983).
Context,
categorization,
and
recall:
The
"change-of-standard"
effect.
Cognitive
Psychology,
15,
525-547.
Higgins,
E. T,
Rholes,
W.,
&
Jones,
C.
(1977).
Category accessibility
and
impression formation.
Journal
of
Experimental
Social
Psychol-
ogy,
13,141-154.
Hovland,
C.,
Harvey,
O., &
Sberif,
M.
(1957). Assimilation
and
contrast
effects
in
communication
and
attitude change.
Journal
of
Abnormal
and
Social
Psychology,
55,
244-252.
lyengar,
S.,
Kinder,
D.,
Peters,
M.,
&
Krosnick,
J.
(1984).
The
evening
news
and
presidential evaluations.
Journal
of
Personality
and
Social
Psychology,
46.
778-787.
Johnson,
E.,
&
Tversky,
A.
(1983).
Affect,
generalization,
and the
per-
ception
of
risk.
Journal
of
Personality
and
Social
Psychology,
45,
20-
31.
Judd,
C.,
&
Harackiewicz,
J.
(1980). Contrast
effects
in
attitude
judg-
ment:
An
examination
of the
accentuation hypothesis.
Journal
of
Per-
sonality
and
Social
Psychology,
38,
390-398.
Kalton,
G.,
Collins,
M.,
&
Brook,
L.
(1978). Experiments
in
wording
opinion
questions.
Journal
of
the
Royal
Statistical
Society
(Series
C),
27,
149-161.
Krosnick,
J.,
&
Alwin,
D.
(1986, May).
Response
order
effects
in the
measurement
of
values.
Paper presented
at the
41st
Annual
Confer-
ence
of the
American Association
for
Public Opinion Research,
St.
Petersburg,
FL.
Lingle,
J.,
Geva,
N.,
Ostrom,
X,
Lieppe,
M.,
&
Baumgartner,
M.
(1979).
Thematic
effects
of
person judgments
on
impression formation.
Jour-
nal
of
Personality
and
Social
Psychology,
37,
674-687.
Lingle,
J.,
&
Ostrom,
X
(1979). Retrieval selectivity
in
memory-based
impression judgments.
Journal
of
Personality
and
Social
Psychology,
37,
180-194.
Linville,
P.
(1982).
The
complexity-extremity
effect
and
age-based
ste-
reotyping.
Journal
of
Personality
and
Social
Psychology,
42,193-210.
Linville,
P., &
Jones,
E.
(1980). Polarized appraisal
of
out-group
mem-
bers.
Journal
of
Personality
and
Social
Psychology,
28.689-703.
Luchins,
A.
(1946). Classroom experiments
on
mental set. American
Journal
of
Psychology,
59,295-298.
Luker,
K.
(1984). Abortion
and the
politics
of
motherhood.
Berkeley:
University
of
California
Press.
Martin,
L.
(1986).
Set/reset:
Use and
disuse
of
concepts
in
impression
formation.
Journal
of
Personality
and
Social
Psychology,
51,
493-
504.
McGuire,
W.
(1960).
A
syllogistic analysis
of
cognitive
relationships.
In
M.
Rosenberg,
C.
Hovland,
W.
McGuire,
R.
Abelson,
& J.
Brehm
(Eds.),
Attitude
organization
and
change
(pp.
65-111).
New
Haven,
CT:
Yale
University Press.
McGuire,
W.,
&
Millman,
S.
(1965). Anticipatory belief-lowering
fol-
lowing
forewarning
of a
persuasive attack.
Journal
of
Personality
and
Social
Psychology,
2,
471-479.
Millar,
M.,
&
Tesser,
A.
(1986). Thought-induced attitude change:
The
effects
of
schema structure
and
commitment.
Journal
of
Personality
and
Social
Psychology,
51,
259-269.
Ostrom,
X
(1970). Perspective
as a
determinate
of
attitude change.
Journal
of
Experimental
Social
Psychology,
6,
280-292.
Petty,
R.,
&
Cacioppo,
J.
(1984).
The
effects
of
involvement
on re-
sponses
to
argument quantity
and
quality: Central
and
peripheral
routes
to
persuasion.
Journal
of
Personality
and
Social
Psychology.
46,
69-81.
Posner,
M.
(1978).
Chronometric
explorations
of the
human mind.
Hillsdale,
NJ:
Erlbaum.
Reiser,
B.,
Black,
J.,
&
Abelson,
R.
(1985).
Knowledge structures
in the
organization
and
retrieval
of
autobiographical memories.
Cognitive
Psychology,
17,89-137.
Rokeach,
M.
(1975). Long-term value change
initiated
by
computer
feedback.
Journal
of
Personality
and
Social
Psychology,
32,467-476.
Rugg,
D.,
&
Cantril,
H.
(1944).
The
wording
of
questions.
In H.
Cantril
(Ed),
Gauging
public
opinion
(pp.
23-50).
Princeton,
NJ:
Princeton
University
Press.
Rumelhart,
D.
(1975).
Notes
on a
schema
for
stories.
In D.
Bobrow
&
A.
Collins
(Eds.),
Representation
and
understanding:
Studies
in
cognitive
science
(pp.
211-236).
New
York:
Academic
Press.
Rumelhart,
D.,
&
Ortony,
A.
(1977).
The
representation
of
knowledge
in
memory.
In R.
Anderson,
R.
Spiro.
& W.
Montague
(Eds.),
School-
ing
and
the
acquisition
of
knowledge
(pp.
99-135).
Hillsdale,
NJ:
Erl-
baum.
Schank,
R.,
&
Abelson,
R.
(1977).
Scripts,
plans,
goals,
and
understand-
ing.
Hillsdale,
NJ:
Erlbaum.
Schuman,
H.
(1982).
Artifacts
are in the
mind
of the
beholder.
The
American
Sociologist,
17.21-28.
Schuman,
H.,
&
Ludwig,
J.
(1983).
The
norm
of
even-handedness
in
surveys
as
in
life.
American
Sociological
Review,
48.112-120.
Schuman,
J.,
&
Presser,
S.
(1981).
Questions
and
answers
in
attitude
surveys:
Experiments
in
question
form,
wording,
and
context.
New
York:
Academic
Press.
Schwarz,
N.,
&
Clore,
G.
(1983).
Mood,
misattribution,
and
judgments
of
well-being: Informative
and
directive functions
of
affective
states.
Journal
of
Personality
and
Social
Psychology,
45,
513-523.
Schwarz,
N.,
&
Hippler,
H. J.
(1987). What response
scales
may
tell
your
respondents:
Informative functions
of
response alternatives.
In H. J.
Hippler,
N.
Schwarz,
& S.
Sudman
(Eds.),
Social
informa-
tion
processing
and
survey
methodology
(pp. 163-178).
New
\brk:
Springer-
Verlag.
Schwarz,
N.,
&
Strack,
E
(1985).
Cognitive
and
affective
processes
in
judgments
of
subjective well-being:
A
preliminary model.
In H.
Brandstatter
& E.
Kirchler
(Eds.),
Economic
psychology
(pp. 439-
447).
Linz,
Austria:
R.
Trauner.
Schwarz,
N.,
Strack,
E,
Kommer,
D.,
&
Wagner,
D. (in
press).
Soccer,
rooms,
and the
quality
of
your
life:
Further evidence
on
informative
functions
of
affective
states.
European
Journal
of
Social
Psychology.
Schwarz,
N.,
&
Wyer,
R.
(1985).
Effects
of
rank ordering stimuli
on
magnitude ratings
of
these
and
other stimuli.
Journal
of
Experimental
Social
Psychology,
21,30-46.
Sears,
D.,
Huddy,
L.,
&
Schaffer,
L.
(1986).
A
schematic variant
of
sym-
bolic
politics
theory,
as
applied
to
racial
and
gender equality.
In R.
Lau
& D.
Sears
(Eds.),
Political
cognition:
The
19th Annual
Carnegie
Symposium
on
Cognition
(pp.
159-202).
Hillsdale,
NJ:
Erlbaum.
Smith,
E.,
Adams,
N.,
&
Schorr,
D.
(1978). Fact
retrieval and the
para-
dox of
interference.
Cognitive
Psychology,
10,438-464.
Smith,
T.
(1982).
Conditional
order
effects
(GSS Tech. Rep.
No.
33).
Chicago: NORC.
Strack,
E,
Schwarz,
N.,
&
Gschneidinger,
E.
(1985). Happiness
and
reminiscing: The
role
of
time perspective,
affect,
and
mode
of
think-
ing.
Journal
of
Personality
and
Social
Psychology,
47,1460-1469.
Tedeschi,
J.
(1981).
Impression
management
theory
and
social
psycho-
logical
research.
New
York:
Academic
Press.
Tesser,
A.,
&
Leone,
C.
(1977).
Cognitive
schemas
and
thoughts
as
deter-
minants
of
attitude change.
Journal
of
Experimental
Social
Psychol-
ogy,
13,
340-356.
Tourangeau,
R.
(1984). Cognitive
science
and
survey
methods.
In T.
Jabine,
M.
Straf,
J.
Xanur,
& R.
Tourangeau
(Eds.),
Cognitive
aspects
of
survey
methodology:
Building
a
bridge
between
disciplines
(pp.
73-
100).
Washington,
DC:
National Academy
Press.
Tourangeau,
R.
(1986, May).
New
perspectives
from
psychology.
Paper
presented
at the
41st Annual Conference
of the
American Associa-
tion
for
Public Opinion Research,
St.
Petersburg,
FL.

314
ROGER
TOURANGEAU
AND
KENNETH
A.
RASINSKI
Tourangeau,
R.
(1987). Attitude measurement:
A
cognitive
perspective.
In
H.
Hippler,
N.
Schwarz,
& S.
Sudman
(Eds.),
Social
informa-
tion
processing
and
survey
methodology
(pp.
149-162).
New
York:
Springer-
Verlag.
Tourangeau,
R.,
Lessler,
5.,
&
Salter,
W.
(1986).
Cognitive
aspects
of
questionnaire
design:
Pan C
report.
Chicago: NORC.
Tourangeau,
R.,
&
Rasinski,
K.
(1986).
Context
effects
in
attitude
sur-
veys.
Unpublished
manuscript.
Tourangeau,
R.,
Rasinski,
K.,
&
D'Andrade,
R.
(1987).
Beliefs
about
abortion
and
welfare.
Unpublished manuscript.
Trope,
Y.
(1986).
Identification
and
inferential
processes
in
disposi-
tional
attribution.
Psychological
Review,
93,
239-257.
Turner,
C.,
&
Krauss,
E.
(1978).
Fallible indicators
of
the
subjective
state
of
the
nation. American
Psychologist,
23,436-470.
Turner,
C, &
Martin,
E.
(1984).
Surveying
subjective
phenomena
(Vol.
1).
New
York:
Russell Sage Foundation.
Tversky,
A.,
&
Kahneman,
D.
(1973).
Availability:
A
heuristic
for
judg-
ing
frequency
and
probability.
Cognitive
Psychology,
5,
207-233.
Upshaw,
H.
(1969).
The
personal
reference
scale:
An
approach
to
social
judgment.
In L.
Berkowitz
(Ed.),
Advances
in
experimental
social
psychology
(Vol.
4, pp.
315-371).
New
York:
Academic Press.
Wilson,
T., &
Dunn,
D.
(1986).
Effects
of
introspection
on
attitude-
behavior consistency: Analyzing reasons versus
focusing
on
feelings.
Journal
of
Experimental
Social
Psychology,
22,249-263.
Wilson,
T.,
Dunn,
D.,
Bybee,
J.,
Hyman,
D.,
&
Rotondo,
J.
(1984).
Effects
of
analyzing reasons
on
attitude-behavior consistency.
Jour-
nal
of
Personality
and
Social
Psychology,
47,
5-16.
Wyer,
R.,
Bodenhausen,
G., &
Gorman,
T.
(1985). Cognitive mediators
of
reactions
to
rape.
Journal
of
Personality
and
Social
Psychology,
48,
324-338.
Wyer,
R.,
&
Hartwick,
J.
(1984).
The
recall
and use of
belief
statements
as
bases
for
judgments: Some determinants
and
implications.
Journal
of
Experimental
Social
Psychology,
20,65-85.
Wyer,
R.,
&
Rosen,
N.
(1972). Some
further
evidence
for the
"Socratic
effect"
using
a
subjective
probability model
of
cognitive
organization.
Journal
of
Personality
and
Social
Psychology,
24,420-424.
Received
January
28,1987
Revision
received September
8,
1987
Accepted
September
30,1987
Women
and
Minorities:
Reviewers
for
Journal
Manuscripts Wanted
If
you are
interested
in
reviewing
manuscripts
for APA
journals,
the APA
Publications
and
Communications Board would
like
to
invite
your
participation. Manuscript reviewers
are
vital
to the
publication process.
As a
reviewer,
you
would
gain valuable experience
in
publishing.
The
P&C
Board
is
particularly
interested
in
encouraging women
and
ethnic
minority
men and
women
to
participate more
in
this process.
If
you are
interested
in
reviewing
manuscripts,
please write
to
Leslie Cameron
at the
address
below.
Please note
the
following
important points:
• To be
selected
as a
reviewer,
you
must
have
published articles
in
peer-reviewed
journals.
The
experience
of
publication provides
a
reviewer
with
the
basis
for
preparing
a
thorough, objective
evaluative
review.
• To
select
the
appropriate reviewers
for
each manuscript,
the
editor needs detailed informa-
tion. Please include with your letter
your
vita.
In
your
letter, please
identify
which
APA
journal
you
are
interested
in and
describe
your
area
of
expertise.
Be as
specific
as
possible.
For
example,
"social
psychology"
is not
sufficient—you
would need
to
specify
"social
cognition"
or
"attitude
change"
as
well.
•
Reviewing
a
manuscript takes
time.
If
you
are
selected
to
review
a
manuscript,
be
prepared
to
invest
the
necessary time
to
evaluate
the
manuscript thoroughly.
Write
to
Leslie Cameron, Journals
Office,
APA, 1400
N.
Uhle
Street, Arlington, Virginia
22201.