Content uploaded by Irwin P Levin
Author content
All content in this area was uploaded by Irwin P Levin
Content may be subject to copyright.
How Consumers Are Affected by the Framing
of Attribute Information Before and After
Consuming the Product
IRWIN P. LEVIN
GARYJ.GAETH*
Consunners rated several qualitative attributes of
ground
beef that framed the beef
as either "75% lean " or "25% fat." The consumers' evaluations were more favor-
able toward ttie beef labeled "75%
lean"
than that labeled "25% fat." More impor-
tantly, ttie magnitude of this information framing effect lessened when consumers
actually tasted the meat. We discuss these results
in
terms of an averaging model,
which suggests that a diagnostic product experience dilutes the impact of infonna-
tion framing.
J
udgment and decision making research identifies
various contexts or "framing" effects that have
important implications for consumer behavior the-
ory development and application. Our present re-
search explores an information framing effect by
which consumers' product judgments vary as a func-
tion ofthe verbal labels used to define specific product
attributes (Johnson and Levin 1985; Levin et al.
1985).
For example, the judged likelihood of purchas-
ing ground beef was found to be higher when the
ground beef was described (framed) in terms of its
percent-lean rather than its percent-fat (Levin et al.
1985).
We used a more general operational definition of
framing than used in earlier works based on Kahne-
man and Tversky's Prospect Theory (Kahneman and
Tversky 1979; see also Puto 1987; Thaler
1985;
Tver-
sky and Kahneman 1981). These researchers defined
framing effect in the context of choice under uncer-
tainty, where the choice between two alternative ac-
tions was shown to reverse, depending on whether at-
tention was focused on the potential gain or the po-
tential loss associated with each alternative. This
definition applies to areas that require discrete
choices between opposing courses of action that are
typically assessed by probabilities of gains and losses
(e.g. Neale and Bazerman 1985). In contrast, the pres-
ent study examines the effect of deterministic product
•Irwin P. Levin is Professor of Psychology, College of Liberal
Arts,
and
Gary
J.
Gaeth
is
Assistant Professor ofMarketing. College
of Business, both at the University of Iowa, Iowa City, IA 52242.
The
authors wish
to thank three anonymous reviewers
for
their
use-
ful suggestions, Ross Dickerson and Elizabeth Vera for their help
in conducting the research, and Ginny Parrish for her editorial as-
sistance.
374
attribute framing on consumers' overall product
judgments.
To better understand this effect. Levin (1987)
showed that favorable or unfavorable associations
with positively or negatively phrased attribute labels
mediate the evaluation of consumer goods. Different
groups of subjects in that study were asked to evaluate
a hypothetical purchase of ground beef that was alter-
natively described as "75% lean" or "25% fat." Sub-
jects'
evaluations were made on several scales, such as
greasy/greaseless, good tasting/bad tasting, and high
quality/low quality. More favorable associations
were produced on each scale when the beef was de-
scribed in terms of percent-lean rather than per-
cent-fat.
Exposure to externally generated product frames is
but one part of the consumer information process.
Another critical part ofthe process that has been in-
cluded in many of the recent descriptions of con-
sumer behavior is personal product experience (see
Assael 1987; Bettman 1979; Bettman and Park 1980).
Our present study examines the joint effects that
framed product attribute information and personal
product experience have on consumer judgments.
(See Deighton and Schindler 1988 for an assessment
ofthe interaction of advertising and experience with
a service.) To manipulate experience with the prod-
uct, we gave subjects a taste of ground beef that was
also described verbally to them as either percent-lean
or percent-fat. This procedure allowed us to assess
whether the information frame affects consumer
judgments in addition to the effects of personal prod-
uct experience.
We varied the order of the framing and sampling
stages; some subjects tasted the meat before it was la-
© JOURNAL OF CONSUMER RESEARCH • Vol.
15
• December 1988
FRAME
AND
PRODUCT EXPERIENCE375
beled, whereas other subjects tasted
the
meat after the
verbal label.
We
manipulated
the
order
to
reveal
how
the framing
of
attribute information would affect
consumers' decisions
in two
typical sequences
of
events (i.e., when consumers
are
exposed
to
product
information before they have
any
personal experi-
ence with
the
product
and
when personal experience
with
the
product precedes consumers' exposure
to
product information.) Considerable evidence
in the
marketing literature suggests that product labeling
can have
an
impact
on
consumers' decisions prior
to
firsthand product experience.
For
example,
in
their
classic study, Allison
and Uhl
(1964) show that
con-
sumers perceive beers differently depending
on
whether
the
consumers
are
aware
or
unaware
of the
brand
of the
beer. Similarly, Bettman
and
Sujan
(1987) show that "priming" customers
to use
differ-
ent decision criteria affected
the
consumers' differen-
tial
use
of decision attributes.
Indeed,
an
argument
can be
made that
a
major role
of advertising
is to
frame
the
subsequent product
ex-
perience. Deighton suggests that "advertising arouses
an expectation"
and "the
subject tends
to
confirm
the
expectation upon exposure
to
more objective infor-
mation (such
as
evidence
or
product experience)"
(1984,
p. 765).
Deighton (1984;
see
also Darley
and
Gross
1983)
also shows that exposure
to
advertising
influenced consumer inferences drawn from objec-
tive information provided later
by an
"unbiased"
source (e.g.. Consumer Reports). Hoch
and Ha
(1986,
Experiment
1)
extend Deighton's work
by
demon-
strating that advertising
has its
greatest influence
when
it
precedes
an
ambiguous product experience
yet
has
very little impact when
it
succeeds
an
unam-
biguous product experience. They also showed that
an
ad
that precedes product experience
has a
signifi-
cant influence
on
product evaluation when compared
to
a no ad
control condition, whereas
an ad
that
fol-
lowed product experience
was not
significantly
different from
the no ad
control condition.
In our
context,
it
seems reasonable
to
treat
an
advertisement
as
a
frame. Hoch
and Ha's
(1986) research suggests
that
the
framing effect will
be
strongest when
the
product experience
is
nondiagnostic (ambiguous)
and
it
will
be
weakest
(or
overwhelmed) when
the
product experience is diagnostic (unambiguous).
To model the joint effects
of
frame
and
experience,
we view the judgment process
as a
classic example
of"
integrating information from different sources. Infor-
mation Integration Theory (Anderson
1981, 1986)
provides methods
for
testing alternative algebraic
models that describe
the
information integration
pro-
cess.
Tests
of
adding versus averaging models have
proved particularly useful
in
understanding
con-
sumer behavior (Shanteau
1988;
Troutman
and
Shanteau
1976) and are
especially relevant
to the
present study.
An
adding model predicts that
the
effect
of a
given source
of
information will
be
inde-
pendent
of the
number
and
nature
of the
other
sources with which
it is
combined.
An
averaging
model,
in
contrast, predicts that
the
effect
of any
given source will
be
reduced with each piece of added
information. Thus,
an
averaging model, rather than
an adding model, predicts that
the
effect
of
an infor-
mation frame will
be
reduced when
the
consumer
has
firsthand experience with
the
product.
METHOD
Design
and
Procedure
We asked subjects
to
rate ground beef
on
several
qualitative dimensions based
on a
sample taste ofthe
meat
and on a
verbal description
of a key
attribute.
The basic design consisted
of the
between-subjects
factorial manipulation of two levels
of
label (positive
and negative)
and two
temporal orders (taste
the
beef
after receiving
the
label
and
taste
the
beef before
re-
ceiving
the
label).
Half
of the
subjects
in the
"taste after labeling"
condition were told that they would
be
given
a
taste
of
"75%
lean ground
beef"
while
the
other half were
told that they would
be
given
a
taste
of
"25%
fat
ground
beef."
After tasting
the
meat, they were given
a response sheet
for
expressing their reactions.
Sub-
jects
in the
"taste before labeling" condition were
not
told
the
"7o-lean/%-fat" information until after they
tasted
the
ground
beef.
Half
of
these subjects were
then told that
the
meat they
had
just sampled
was
"75%
lean" whereas
the
others were told that
it was
"25%
fat."
They then were given
a
response sheet.
We took several precautions during
the
tasting
stage
to
minimize confounding effects.
The
study
was
conducted
on two
separate evenings using
the
same
four consecutive half-hour periods
per
evening—one
for each cell ofthe experimental design.
One
random
order
of
conditions
was
employed
the
first evening
and
the
reverse order
was
employed
the
second
eve-
ning.
On
each evening,
all
subjects received
a l'/2
ounce sample from
the
same skillet
of
freshly cooked
ground
beef.
(Actually,
the
meat
was
slightly more
lean than
the
reported "75% lean/25% fat.")
The four rating scales that were
on the
response
sheets follow. Subjects were instructed
to
place
an X
in
one
ofthe seven boxes
on
each scale. Note that
the
left-right positions ofthe positive
and
negative poles
vary across scales. These
are the
same scales used
in
the Levin (1987) study.
D
good
D
D
tasting
D
greasy
D
high
D
fat
D
quality
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
bad tasting
D
greaseless
D
low quality
D
lean
376
THE JOURNAL OF CONSUMER RESEARCH
TABLE
1
MEAN RATING SCORES ACROSS TASTE
AND
FRAMING CONDITIONS
Rating scale
Fat/lean
Low quality/high quality
Greasy/greaseless
Bad taste/gcxx) taste
Label-oniy condition*
Positive
5.15
5.33
4.49
5.69
Negative
2.83
3.66
2.96
4.43
Difference"
2.32*
i.6r
1.53*
1.26*
Positive
4.67
4.71
4.13
5.00
Taste after labeling
Negative
3.57
3.95
3.43
4.71
Difference
1.10*
.76*
.70*
.29
Taste before labeling
Positive
4.05
4.43
3.67
5.00
Negative
3.45
4.09
3.05
5.09
Difference
.60=
.34
.62=
-.09
•
Data taken
from
Levin
(1987).
'
Difference between
mean rating score In positive and
negative
framing
conditions.
'p<0.10.
'p<0.05.
•p<0.01.
Subjects
Ninety-six students from introductory psychology
classes
at
the University
of
Iowa signed
up to
partici-
pate in the study the day before the actual session was
to be held. Twelve males and 12 females volunteered
for each ofthe four evening sessions, though
not all
of those who signed
up
actually showed
up for
their
designated session.
The
final sample size
in
each
ex-
perimental condition was
21
in
"positive frame taste
before labeling,"
24 in
"negative frame taste before
labeling," 22
in
"positive frame taste after labeling,"
21
in
"negative frame taste after labeling."
RESULTS
Ratings
on
each scale ranged from
one to
seven,
with higher numbers representing more favorable re-
sponses. Table
1
gives the mean rating score
on
each
scale
for
each experimental condition ofthe present
study
and for the
Levin (1987) study, which
did not
include a tasting phase. We found ratings to be higher
in the positive framing condition than
in
the negative
framing condition, with
one
minor reversal. Differ-
ences between framing conditions were tested
for
sig-
nificance using
a
one-tailed /-test
of the
hypothesis
that the difference is greater than zero.
Although most,
but not
all, individual tests ofthe
framing effect were significant,
we
detected
a
clear
pattern
in the
magnitude
of
the difference scores
in
Table
1.
The largest framing effect was on the fat/lean
scale, which
is the
scale most related
to the
framing
manipulation. The smallest effect was on the bad tast-
ing/good tasting scale, which is the scale most related
to the tasting experience. More importantly, compar-
ing the values ofthe difference scores across columns
in Table 1 shows that
the
magnitude ofthe framing
effect within each scale
was
related
to
whether
sub-
jects actually tasted the ground beef and,
to
some ex-
tent, when they tasted
it.
For each scale,
the
framing
effect tended to be largest when subjects did not actu-
ally taste the meat, less large when subjects tasted
the
meat after being given
the
label,
and
smallest when
subjects tasted the meat before being given the label.
To test
the
significance
of
these observed trends,
ANOVA tests were performed
on the
data from
the
four rating scales.' These results
are
summarized
in
Table 2. Framing condition accounted
for
the largest
percentage ofthe variance on each scale. Tasting con-
dition—whether
the
subjects tasted
the
meat before
receiving the label, after receiving the label, or did not
taste
the
meat
at all
(from Levin 1987)—was
not a
significant source
of
variance. That
is,
whether
and
when subjects tasted the meat did not affect the over-
all favorableness
of
their rating responses. However,
the framing condition
by
tasting condition interac-
tion
was
significant
for
three
of the
four scales,
affirming that the magnitude ofthe framing effect var-
ied across tasting conditions.
The two components ofthis interaction
can be de-
scribed
as (1) the
extent
to
which
the
framing effect
differed between
the
"label-only" condition
of the
Levin (1987) study and the "taste plus labeling" con-
ditions
of the
present study,
and (2) the
extent
to
which
the
framing effect differed between
the
"taste
after labeling" and "taste before labeling" conditions
ofthe present study. Appearing
at the
bottom
of
Ta-
ble
2
as subcomponents ofthe framing by taste inter-
action, these
two
tests (1 df each) show that
the
first
component was significant
for
all scales. This signifi-
cance confirms that
the
framing effect,
as
predicted
by
the
averaging model,
was
reliably reduced when
subjects tasted the meat as compared to when they did
not taste the meat. The second component, however,
was
not
significant
for any
scale. Thus,
the
order
of"
the tasting
and
labeling phases
did not
significantly
affect the magnitude ofthe framing effect.
'Note that the assumption or random assignment of subjects to
conditions clearly has been violated in this between-experiment
test. However, the subjects who participated in both experiments
did come from the same population.
FRAME AND PRODUCT EXPERIENCE377
TABLE 2
ANOVA RESULTS FOR EACH RATING SCALE
Source
Framing condition (positive, negative)
Taste condition (label-only.* taste after
lat)eling.
taste
before labeling)
Framing X taste"
Taste condition compared to no taste condition
Taste after label condition compared to taste before
label condition
df
1,184
2.184
2.184
1.184
1.184
Fat/lean
MS
128.10
2.23
12.51
24.30
.72
F
54.51*
.95
5.32*
10.34*
.31
Low quality/
high quality
MS
63.37
1.61
7.37
13.87
.87
F
30.25*
.77
3.52^
6.62*
.42
Greasy/
greaseless
MS F
61.68 27.28*
3.23 1.43
3.75 1.66
7.42
3.28«
.08 .04
Bad taste/
good taste
MS
24.86
.91
8.36
15.91
.81
F
13.48*
.49
4.53"
8.63*
.44
• Data taken
from
Levin (1987).
" Tbe 2 di
framing
x taste interaction was divided into two separate
1
iH
tests,
which are described
tn
the text.
'p<0.10.
"p<0.05.
•p<0.01.
DISCUSSION
Because our research deals with information fram-
ing effects, it seems only fitting that the present results
can be stated in two different ways. On the one hand,
the alternate labeling of a product attribute in posi-
tive or negative terms did affect consumers' evalua-
tions even when they actually consumed the product.
On the other hand, the labeling or framing effect was
reduced when consumers sampled the product as
compared to when they did not.
The framing influence of attribute labels, even
when consumers can rely on their own experiences,
serves to validate the information framing effect
found in previous studies in which subjects evaluated
hypothetical purchases (Bellizzi and Martin 1982;
Levin et al. 1985). The general reduction ofthe fram-
ing effect that occurs when consumers actually sam-
pled the product is consistent with the averaging
model of information integration (see Shanteau 1988
for a similar model developed for combining tempo-
ral sequences of visual and verbal information). Ac-
cording to this model, the effect of any one source of
information is decreased when it is combined with
another source of information, because the two
sources of information available to the judge are bal-
anced in arriving at an integrated impression. The
present results (Tables
1
and 2) clearly show this pat-
tern. Verbal labels and experiential information thus
appear to be combined through an averaging pro-
cess."
'For the present situation, an averaging model can be formulated
as follows: response = (My/+ w^)l(wf+ w,), where/and fare the
subjective scale values ofthe framed information and the personal
experience, respectively, and My and w, are the weights associated
with
these
different sources of information. Two important features
Our results are also consistent with those of Hoch
and Ha (1986), who found that ads (frames) have an
effect on subsequent product evaluations only when
the product experience is ambiguous. The question is
whether tasting hamburger is an ambiguous experi-
ence.
Examination of differences between rating
scales reveals post hoc support for the relationship be-
tween ambiguity of experience and the frame (ad)
effect proposed by Hoch and Ha (1986). One would
expect that the "taste experience" would be most rele-
vant to the bad tasting/good tasting scale and hence
maximally reduce the framing effect, whereas the
taste experience would be less relevant to the other
scales. The results in Table
1
support this interpreta-
tion by showing the taste-related scale to be least (and
nonsignificantly) affected by the frame when subjects
actually tasted the meat.
The result from the Hoch and Ha (1986) study can
be viewed within the averaging model framework.
Product experience, whether it precedes or follows
product information, will have the greatest weight
when it is diagnostic (unambiguous). According to
the balancing principle of the averaging model, we
would then expect to find stronger framing (advertis-
ing) effects when the product experience is less diag-
nostic, such as sampling perfume or cologne.
The averaging model provides a context for exam-
ining a variety of factors (e.g., temporal effects) that
ofthis model should be noted. First, because the sum ofthe weights
is in the denominator, the effective weight of information frame is
wjl(wf
+ w,). which is less than one if
H',
is not equal to zero. This
means that the effect of information frame is less when personal
experience is provided than when it is not. Second, the weight pa-
rameters are allowed to vary with serial position (for example,
which source of information is processed firist) and by the salience
orambiguity ofthe attribute (as in Hoch and Ha 1986).
378THE JOURNAL
OF
CONSUMER RESEARCH
would extend the current research. When various
sources of information are integrated serially, the rel-
ative importance or weight of each source may de-
pend upon its position within the sequence. The pres-
ent trends suggest a primacy effect whereby the rela-
tive importance of attribute labeling is greater when
it occurs before product consumption than when it
occurs afterwards. This effect is consistent with the
data reported by Hoch and Ha (1986), indicating that
the impact of an ad presented before the product ex-
perience tends to be greater than the impact of one
presented after. (As here, these differences supported
a primacy effect, but were not significant.) The reli-
ability of temporal effects needs to be afiirmed in an
expanded experimental design. The averaging model
predicts that the weight of a given piece of informa-
tion decreases as the number of pieces of information
with which it is combined increases. Thus, the greater
the number of prior personal experiences with a prod-
uct, the less should be the effect of subsequent attri-
bute labeling.
Because information weighting has been shown to
vary directly with the polarity ofthe information pre-
sented (Levin et al. 1973), the quality ofthe personal
experience should also moderate the effect of the
product information frame. Thus, if the ground beef
tastes terrible, for example, it is unlikely that even a
powerful positive frame will lead to a favorable evalu-
ation. Future research designed to test these predic-
tions and quantify the effects ofthe variables of inter-
est would involve manipulating the pleasantness of
personal experiences with the product, as well as the
relative number and temporal sequence of personal
experiences and framed descriptions. Such research
should serve to enhance our understanding of how
consumers respond to advertisements and product
experiences in the marketplace (Goering 1985).
[Received March 1988. Revised July 1988.]
REFERENCES
Allison, Ralph I. and Kenneth P. Uhl (1964), "Influence
of
Beer Brand Identification
on
Taste Perception," Jour-
nal of Marketing
Research,
37 (April), 32-37.
Anderson, Norman H. (1981), Foundations of Information
Integration Theory, New York: Academic Press.
(1986),
"A Cognitive Theory of Judgment and Deci-
sion,"
in
New
Directions
in Research on Decision
Mak-
ing,
eds. B.
Brehmer,
et al.,
Amsterdam,
The
Nether-
lands:
North-Holland, 63-108.
Assael, Henry (1987), Consumer Behavior and Marketing
Action, Boston, MA: Kent.
Bellizzi, Joseph
A.
and Warren
S.
Martin (1982), "The
In-
fluence of National Versus Generic Branding on Taste
Perception," Journal
of
Business Research,
10 (Sep-
tember), 385-396.
Bettman, James
R.
(1979), Information Processing Theory
of Consumer
Choice,
Reading,
MA:
Addison-Wesley.
—••— and C. Whan Park (1980), "Effects
of
Prior Knowl-
edge and Experience and Phase
of
the Choice Process
on Consumer Decision Processes:
A
Protocol Analy-
sis,"
Journal
of
Consumer Research,
7
(December),
234-248.
and Mita Sujan (1987), "Effects of Framing on Eval-
uation
of
Comparable
and
Noncomparable Alterna-
tives
by
Expert
and
Novice Consumers," Journal
of
Consumer
Research,
14
(September), 141-154.
Darley, John M. and Paget H. Gross (1983), "A Hypothesis-
Confirming Bias
in
Labelling Effects," Journal of Per-
sonality and Social
Psychology,
44 (January), 20-33.
Deighton, John (1984),
"The
Interaction
of
Advertising
&nd
Evidence,"
Journal
of Consumer Research, 11 (De-
cember), 763-770.
and Robert
M.
Schindler (1988), "Can Advertising
Influence Experience?" Psychology
&
Marketing,
5
(Summer), 103-115.
Goering, Patricia
A.
(1985), "The Effects
of
Product Trial
on Consumer Expectations, Demand,
and
Prices,"
Journal
of
Consumer
Research,
12
(June), 74-82.
Hoch, Stephen
J.
and Young-Won Ha (1986), "Consumer
Learning: Advertising
and the
Ambiguity
of
Product
Experience," Journal of Consumer Research, 13 (Sep-
tember), 221-233.
Johnson, Richard
D. and
Irwin
P.
Levin (1985), "More
Than"Meets the
Eye:
The Effect ofMissing Information
on Purchase Evaluations," Journal
of
Consumer
Re-
search,
12 (September), 169-177.
Kahneman, Daniel
and
Amos Tversky (1979), "Prospect
Theory: An Analysis
of
Decision Under Risk," Econo-
metrica, 47 (March), 263-291.
Levin, Irwin P. (1987), "Associative Effects
of
Information
Framing," Bulletin
of
the Psychonomics Society,
25
(March), 85-86.
, Richard D. Johnson, Craig
P.
Russo, and Patricia J.
Deldin (1985), "Framing Effects
in
Judgment Tasks
with Varying Amounts
of
Information," Organiza-
tional Behavior
and
Human Decision Processes,
36
(December), 362-377.
-, Linda
L.
Wall, Jeanette
M.
Dolezal,
and
Kent
L.
Norman (1973), "Differential Weighting
of
Positive
and Negative Traits
in
Impression Formation
as a
Function
of
Prior Experience," Journal of Experimen-
tal
Psychology,
97
(January), 114-115.
Neale, M. and
M.
Bazerman (1985), "Perspectives
for
Un-
derstanding Negotiation: Viewing Negotiation
as a
Judgmental Process," Journal
of
Conflict
Resolution,
29 (March), 33-55.
Puto,
Christopher (1987),
"The
Framing
of
Buying
Decx-
%\ons"
Journal
of
Consumer
Research,
14
(December),
301-315.
Shanteau, James (1988), "Consumer Impression Forma-
tion:
The
Integration
of
Visual
and
Verbal Informa-
tion,"
in
Nonverbal Communication
in
Advertising.
eds.
Sidney Hecker and David W. Stewart, Lexington,
MA: D.C. Heath, 42-57.
Thaler, Richard (1985), "Mental Accounting
and Con-
sumer Choice," Marketing
Science,
4
(Summer),
199-
214.
Troutman,
C.
Michael
and
James Shanteau (1976),
"Do
Consumers Evaluate Products by Adding or Averaging
Attribute Information?" Journal
of
Consumer
Re-
search,
3 (September), 101-106.
Tversky, Amos and Daniel Kahneman (1981), "The Fram-
ing
of
Decisions and
the
Psychology
of
Choice,"
5c/-
ence.lW (September),453-458.
