Content uploaded by Barbara E. Kahn
Author content
All content in this area was uploaded by Barbara E. Kahn on Dec 30, 2014
Content may be subject to copyright.
Journal of Marketing Research
Vol. XLVI (December 2009), 725–738725
© 2009, American Marketing Association
ISSN: 0022-2437 (print), 1547-7193 (electronic)
*Xiaoyan Deng is Assistant Professor of Marketing, Fisher College of
Business, Ohio State University (e-mail: deng_84@fisher.osu.edu).
Barbara E. Kahn is Dean, Schein Family Chair, and Professor of Market-
ing, School of Business Administration, University of Miami (e-mail:
bkahn@miami.edu). This paper is based on “Essay One” of the first
author’s dissertation, written when she was a doctoral student at the Whar-
ton School, University of Pennsylvania, and supervised by the second
author when she was Dorothy Silberberg Professor of Marketing at the
same institution. The authors thank Robert Meyer, Eric Bradlow, and Wes
Hutchinson for their comments on a previous version of this article and the
Wharton Behavioral Laboratory for support in data collection. The authors
also thank the anonymous JMR reviewers for their constructive sugges-
tions. Dilip Soman served as associate editor for this article.
XIAOYAN DENG and BARBARA E. KAHN*
The authors show that location of the product image on a package
f
acade influences consumers’ perceptions of the visual heaviness of the
product and evaluations of the package. The “heavier” (“lighter”) locations
are on the bottom (top), right (left), and bottom-right (top-left) of the
package. For products for which heaviness is considered a positive
attribute, packages with the product image placed at heavy locations
a
re preferred, whereas for products for which heaviness is considered
a negative attribute, packages using light locations are preferred.
Furthermore, in the former category (e.g., snacks), a salient health goal,
as opposed to a neutral goal, weakens the preference for packages
using heavy locations, though this moderating effect of goal is weaker for
healthful snacks than for regular snacks. Store shelf context is a
boundary condition, such that the location effects on perceived product
heaviness and package evaluation appear in a contrasting context but
disappear in an assimilating context. Moreover, perceived product
heaviness mediates (1) the location effect on package evaluation and (2)
the moderating role of store shelf context (i.e., mediated moderation).
Keywords: visual packaging design, product image location, visual
perceived heaviness, package evaluation
Is Your Product on the Right Side? The
“Location Effect” on Perceived Product
Heaviness and Package Evaluation
Packaging is integral to the marketing and distribution of
products. At the point of sale, the product package can play
a pivotal role in a consumer’s purchase decision. Even after
the purchase, the packaging can continue to influence the
consumption experience. Thus, understanding how packag-
ing variables, such as shape, color, and graphics, affect con-
sumer perception, evaluation, and behavior is of theoretical
and managerial importance. Recently, marketing researchers
have focused on the effects of package shape on volume
perception, package preference, choice, and consumption
(e.g., Folkes and Matta 2004; Krishna 2006; Raghubir and
Greenleaf 2006; Raghubir and Krishna 1999; Wansink and
Van Ittersum 2003; Yang and Raghubir 2005).
In this research, we focus on the relative location of prod-
uct images on different package facades. Drawing on the art
and visual perception literature, we hypothesize and demon-
strate a “location effect,” in which product images placed at
certain locations on a package facade appear to be heavier
than the same images placed at other locations. Specifically,
we first show that the “heavier” locations are on the bottom,
right, and bottom-right of the package facade and that the
“lighter” locations are on the top, left, and top-left. We then
show that different locations of the product image can influ-
ence the shopper’s perception of the visual heaviness of the
product and that these differing perceptions can affect pack-
age evaluation. We further hypothesize and demonstrate that
product attribute valence, consumer goal, and store shelf
context can moderate this effect. Before we discuss the
theoretical justification for our hypotheses, however, we
report the findings of a field study in which we observed
some systematic patterns regarding how managers are cur-
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 725
726 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
rently using the location of the product image on a package
facade in the actual market place.
FIELD STUDY: THE LOCATION EFFECT OBSERVED
IN THE SNACK CATEGORY
We conducted a field study in a metropolitan supermar-
ket in the snack category. We wanted to observe (1) how fre-
quently packaging in this category featured an image of the
product; (2) if there was an image, whether the location of
the image varied; and (3) whether the pattern of this varia-
tion differed in any systematic way.
W
e investigated two types of snacks: cookies and crack-
ers. We report summary statistics in Table 1, Panel A, which
shows that 100% of the snack packages displayed a product
image (photography or illustration) on the facade (for exam-
ples, see the Web Appendix at http://www.marketingpower.
com/jmrdec09). In addition, 61% of the cracker packages
featured a health claim (e.g., no trans fat, no or low satu-
rated fat, fat free, no hydrogenated oils, sugar free, no or
low cholesterol), whereas only 28% of the cookie packages
did (t(277) = 6.0, p < .0001), suggesting that the cracker
category is being marketed as a more healthful category.
Consistent with this categorization, 41% of the cracker
packages placed the product image at light locations,
whereas for cookies (the less healthful option), only 30% of
the packages did (t(277) = 2.0, p < .05). We further exam-
ined the relationship between the presence of a health claim
on the package and the presence of a product image placed
at light locations separately for each category. We report the
results in Table 1, Panel B, which shows that for cookies,
this relationship was significant (χ
2
(1) = 31.0, p < .0001),
and the presence of a product image placed at light locations
was 66% for packages with a health claim and only 17% for
those without such a claim (t(136) = 6.3, p < .0001). For
crackers, this relationship showed the opposite pattern,
though it was not significant: 38% of the packages with a
health claim placed the product image at light locations,
whereas 47% of those without a health claim did.
The field study suggests that there are some systematic
patterns in the location of the product image on package
facades. It appears that the location of the product image is
related to product categories and consumers’ shopping
goals. This is intriguing because our conversations with
brand managers and packages designers suggested that the
location effect was not an effect of which they were con-
scious. In addition, we found no discussion of this effect in
our literature search in packaging trade journals and related
articles. However, we know that pictures are more easily
recognized and recalled than text (Childers and Houston
1984) and, furthermore, that pictorial information has a pro-
found impact on consumer perception and judgment (Hol-
brook and Moore 1981). Thus, we are not surprised to
observe in our field study that product image was so preva-
lently used in package design. Given the attention drawn to
images, it would also make sense that, at least at an intuitive
level, managers might be using product images and their
locations to provide product information. Our experiments
are designed to provide a roadmap that will help managers
predict how the location of the product image on a package
facade can be used in a deliberate way to communicate per-
ceptions about the product.
In Study 1, we demonstrate that pictorial objects placed
at heavy locations of a visual field appear to be heavier than
the same objects placed at light locations. Treating a pack-
age facade as the visual field and the product image as a pic-
torial object within that field, in Studies 2–4, we apply this
finding to product packaging and examine the effect of loca-
tion on package evaluation. In Study 2, we show that there
is no universal preference for heavy or light locations. For
products for which heaviness is considered a positive attrib-
ute, packages with the product image placed at heavy loca-
tions are preferred, whereas for products for which heavi-
ness is considered a negative attribute, packages using light
locations are preferred. In Studies 3–4, we focus on the
snack category, for which heaviness is good, and test mod-
erating conditions for the preference for packages using
heavy locations. In Study 3, we identify consumer goal as a
moderator, such that a salient health goal, as opposed to a
neutral goal, weakens the preference for packages using
heavy locations. We also provide evidence that this moder-
Table 1
PRODUCT IMAGE LOCATION OBSERVED IN THE SNACK CATEGORY
A: Summary Statistics
Snack Type Cookie Cracker
Number of packages
a
138% 141%
Product image on the package facade 100% 100%
A single image (cluster) 88% 84%
Package facade shape 1: vertical rectangle 41% 67%
Package facade shape 2: horizontal rectangle 47% 26%
Package facade shape 3: square 12% 8%
Health claim on the package facade 28% 61%
Product image placed at the light locations 30% 41%
B: Product Image Location as a Function of Snack Type and Health Claim
Cookie (n = 138) Cracker (n = 141)
Health Claim: Health Claim: Health Claim: Health Claim:
No (n = 100) Yes (n = 38) No (n = 55) Yes (n = 86)
Product image location placed at light locations 17 (17%) 25 (66%) 26 (47%) 33 (38%)
a
If a package facade contained a transparent window, we excluded that package from our analysis to focus on product graphics only. We excluded 24
cookie and cracker packages.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 726
Is Your Product on the Right Side? 727
ating effect of goal is weaker for healthful snacks than for
regular snacks, echoing the results from our field study. In
Study 4, we identify store shelf context as another modera-
tor. Specifically, the location effects on perceived product
heaviness and package evaluation appear in a contrasting
context but disappear in an assimilating context. Moreover,
we establish that perceived product heaviness mediates (1)
the location effect on package evaluation and (2) the moder-
ating role of store shelf context (i.e., mediated moderation).
STUDY 1: THE LOCATION EFFECT ON VISUAL
HEAVINESS PERCEPTION
Perceptions about the lifted weight of physical objects as
a function of seeing and feeling the objects have been exten-
sively studied in experimental psychology. For example, a
heavily researched effect is the size–weight illusion in
which bigger objects of the same weight feel lighter (Char-
pentier 1891). Conversely, perceived visual weight of picto-
rial objects has been examined in the art and design litera-
ture from a Gestalt psychology perspective (e.g., Arnheim
1974). Here, it is shown that artists can skillfully distribute
perceived visual weight by assigning pictorial objects to dif-
ferent locations of a visual field to achieve a perceptual bal-
ance in the artwork and to convey certain meanings to the
viewer.
On the basis of the Gestalt whole–part relationship, Arn-
heim (1974) argues that no object is perceived as isolated;
that is, seeing an object involves assigning it a place in the
whole, for example, a location in space. A visual field’s
“structural skeleton” (i.e., the boundaries, center, and axes
of symmetry) serves as a frame of reference for assigning
location to an object within the field. This location further
determines the object’s visual weight. The relationship
between location and visual weight can be described as
“bottom-heavy” (i.e., as a pictorial object moves from the
top to the bottom along the vertical axis of a visual field, it
appears heavier) and “right-heavy” (i.e., as a pictorial object
moves from the left to the right along the horizontal axis of
a visual field, it also appears heavier). According to Arn-
heim, both bottom-heavy and right-heavy are “perceptual
forces.” That is, they are the counterpart of physical forces
in visual perception. He argues that bottom-heaviness
occurs because of gravitational pull and right-heaviness
occurs because of the lever effect.
In a world dominated by gravitational pull, heavy things
are anchored on the ground, while things of little weight
(e.g., balloons) float upward. Therefore, people’s visual
weight judgments are in line with these bottom-heavy or
top-light observations. This visual perception is so powerful
that designers tend to make the bottom part of a visual
object appear heavier. As Greenough (1947, p. 24) observes,
“That buildings, in rising from the earth, be broad and sim-
ple at their bases, that they grow lighter not only in fact but
in expression as they ascend, is a principle established. The
laws of gravitation are the root of this axiom. The spire
obeys it. The obelisk is its simplest expression.” This visual
tradition is even observed in situations in which the force of
gravity is not directly relevant. Typographers design the
lower part of a letter or number to be slightly bigger than the
upper part. For example, 3, 8, S, and B look comfortably
poised in their upright position but “macrocephalic” (the
condition of having a head that is excessively large) when
t
urned upside down (Kanizsa and Tampieri 1968).
One explanation for right-heaviness is based on one of
the principles of the lever. This principle states that the
g
reater the distance on the lever from the fulcrum position
that an object is placed, the heavier is the weight that is
needed on the other side to balance the object. On the basis
of this mechanism, Arnheim (1974) proposes a two-step
explanation for the right-heavy perception. First, pictures
are “read” from left to right, and this is why the diagonal
that runs from bottom-left to top-right is perceived as
ascending and the other diagonal (i.e., top-right to bottom-
left) as descending (i.e., if “read” from right to left, these
two diagonals will then be perceived as descending and
ascending, respectively [Wolfflin 1950]). Second, because
the eyes enter a visual field from the left, the left naturally
becomes the anchor point or “visual fulcrum.” Thus, the fur-
ther an object is placed away from the left side (or the ful-
crum), the heavier is the perceived weight. Although this
“visual level effect” was Arnheim’s theory, he had no direct
empirical evidence for it.
However, subsequent research in the field of ocular domi-
nance (i.e., eye dominance or eyedness) has provided sup-
port for Arnheim’s (1974) theory of right-heaviness, albeit
not necessarily directly through his lever principle. Ocular
dominance is “the visual phenomenon where a functional
ocular unilaterality exists in binocular vision—some sort of
physiological preferential activity of one eye over that of the
other when both are used together” (Ogle 1950).
1
Porac and
Coren (1976) report that approximately 65% of the popula-
tion is right-eye dominant, 32% is left-eye dominant, and
3% is ambiocular. (Bourassa, McManus, and Bryden [1996]
report similar statistics based on a meta-analysis covering
research from 1925 to 1992.) Porac and Coren also describe
the influence of ocular dominance on perceptual processing.
For example, relative to the nondominant eye, the dominant
eye has greater voluntary muscle movement strength
(Schoen and Schofield 1935), is more likely to provide the
primary information for the computation of visual direction
(Walls 1951), and has a higher input intensity (Francis and
Harwood 1959). These observations point to a “visual ego”
(the center point for egocentric localization) that resides in
the dominant eye (Porac and Coren 1976).
Visual input from the dominant eye is accentuated during
binocular viewing, suggesting that objects on the same side
as the dominant eye are often overestimated. First, Coren
and Porac (1976) found that objects presented to the domi-
nant eye are perceived as bigger than the same objects pre-
sented to the nondominant eye. Scott and Sumner (1949)
found that objects placed in front of the dominant eye are
perceived as being closer to the observer. Mefferd and
Wieland (1969) had participants bisect a horizontal line, and
they found that the locus of the midpoint shifted toward the
side of the dominant eye. That is, for right-eyed people, the
right segment of the line seemed to “count” more. Finally,
1
Ocular dominance was first documented in the sixteenth century (see
Porta 1593). It is a striking phenomenon that can be demonstrated easily:
First, hold a pencil directly in front of you. Then, keeping both eyes open,
align its tip with a point on a distant wall. Now cover one of your eyes and
see if the pencil remains in good alignment with the target. Now cover the
other eye and see if the alignment maintains. If the pencil shifts out of the
alignment when you cover the left (right) eye, you are categorized as a left-
eyed (right-eyed) person.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 727
728 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
3
We use one-tailed tests for all planned contrasts.
as a indirect test of Arnheim’s (1974) right-heavy principle,
Hirata (1968) had right-eyed participants assign a visual
b
alance point anywhere along the straight line anchored by
two identical objects or two objects of different sizes. Dur-
ing binocular viewing, there was a clear tendency to set the
balance point to the right of the physically expected balance
point, proving that for right-eyed people, objects in the right
visual field had greater visual weight. However, Hirata’s
study did not include left-eyed participants, who might
e
xhibit the same, reduced, or reversed right-heavy percep-
tual tendency.
In light of Arnheim’s (1974) bottom-heavy and right-
heavy visual weight principles and the subsequent support
for right–left visual weight asymmetry among right-eyed
people (who make up 65% of the population), we predict
general bottom-heavy and right-heavy effects across people,
though because of individual differences in eyedness, the
latter effect might be weaker than the former one.
H
1
: A pictorial object placed at the bottom, right, and bottom-
right of a visual field will be perceived as heavier than the
same object placed at the top, left, and top-left, respectively.
Method
All experiments we report herein were computer based
and conducted in the same behavioral laboratory. Partici-
pants were primarily students at a northeastern U.S. univer-
sity who participated in experimental sessions containing
several studies (of which our study was one) and were com-
pensated $10 per hour. We conducted Study 1 (N = 139) to
test H
1
using a one-factor (product image location: top, bot-
tom, left, right, top-left, bottom-right; see the labels used in
Figure 1) within-subject design. We prepared seven stimuli:
one practice stimulus and six study stimuli. The six study
stimuli were the labels used in Figure 1. In the practice stim-
ulus, the disk was placed at the center of the square. We
used a 6 × 6 Latin square to counterbalance the order of the
six study stimuli, such that within a sequence, each stimulus
appeared only once in each of the six positions, and across
the six sequences, each stimulus followed each of the others
only once. This design also enabled us to test H
1
as both a
between-subjects study (using only the first stimulus each
participant rated) and a within-subject study (using all stim-
uli). Participants rated the practice stimulus and then the six
study stimuli, one at a time, on three nine-point scales (–4 =
“unsubstantial/weightless/feathery,” and 4 = “substantial/
weighty/hefty”).
Results
“Location” as a within-subject factor. We averaged the
three scales to form a perceived heaviness index (α = .87).
A repeated measures analysis of variance (ANOVA) with
location as a within-subject factor conducted on this index
revealed a significant main effect of location (F(5, 665) =
48.5, p < .0001, see Figure 1, Panel A).
2
Planned contrasts
indicated that the object located at the bottom was perceived
as visually heavier than the object located at the top (M
top
=
–.72, M
bottom
= .81; t(133) = 6.9, p < .0001), the object at
the right heavier than the object at the left (M
left
= –.30,
M
right
= .65; t(133) = 6.5, p < .0001), and the object at the
bottom-right heavier than the object at the top-left (M
top-
left
= –1.12, M
bottom-right
= .93; t(133) = 10.1, p < .0001).
3
Thus, H
1
was supported. We also calculated a difference
score based on the first two contrasts (i.e., difference
bottom
vs. top
– difference
right vs. left
) for each participant. We found
that this difference was significantly greater than zero
(t(133) = 2.3, p < .05), suggesting that the bottom-heavy
result is a stronger phenomenon than the right-heavy result.
“Location” as a between-subjects factor. A one-way
ANOVA with location as a between-subjects factor con-
ducted on the rating on the first stimulus revealed that the
main effect of location was also significant (F(5, 133) = 7.0,
p < .0001, see Figure 1, Panel B). Again, planned contrasts
indicated that the object located at the bottom was perceived
as visually heavier than the object located at the top (M
top
=
–1.16, M
bottom
= 1.00; t(133) = 4.3, p < .0001), the object at
the right heavier than the object at the left (M
left
= –.31,
M
right
= .54; t(133) = 1.7, p < .05), and the object at the
.93
–1.12
.65
–.30
.81
–.72
–2
–1
0
1
2
Visual Perceived Heaviness
–1.16
1.00
–.31
.54
–1.41
.14
–2
–1
0
1
2
Visual Perceived Heaviness
1 2 3 4 5 6
Top- Bottom-
Top Bottom Left Right Left Right
F
igure 1
A PICTORIAL OBJECT’S LOCATION WITHIN A VISUAL FIELD
DETERMINES ITS VISUAL WEIGHT
A: Within-Subject Effect
B: Between-Subjects Effect
2
For the sake of brevity, in this article, we do not report results that are
not statistically significant.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 728
Is Your Product on the Right Side? 729
bottom-right heavier than the object at the top-left (M
t
op-
l
eft
= –1.41, M
b
ottom-right
= .14; t(133) = 3.0, p < .005). The
planned contrast between bottom-heavy and right-heavy
also indicated that the former was a stronger effect than the
latter (t(133) = 1.9, p < .05).
The underlying mechanism for the right-heavy effect.
Hirata’s (1968) study reveals a right-heavy effect among
right-eyed people. In Study 1, including both right-eyed and
l
eft-eyed participants, we found a right-heavy effect, but it
w
as weaker than the bottom-heavy effect. From the findings
that ocular dominance influences visual processing such
that objects on the same side as the dominant eye are often
overestimated in size (Coren and Porac 1976) and length
(Mefferd and Wieland 1969), it might be conjectured that
the reason the right-heavy effect was smaller than the
bottom-heavy effect is related to heterogeneity in eyedness
among the participants. To test this conjecture, we con-
ducted Study 1a (N = 117).
Participants were shown (1) an object in the left visual
field (see the “3” label in Figure 1); (2) an object in the right
visual field (see the “4” label); (3) side-by-side presentation
of 1 and 2, with 1 on the left side; and (4) side-by-side pres-
entation of 1 and 2, with 1 on the right side, one at a time.
We used four between-subjects conditions to counterbal-
ance the order. The paired presentation in 3 and 4 was con-
sistent with the experimental paradigm used in the ocular
dominance studies.
4
We then assessed participants’ eyed-
ness using an unconscious sighting test, the “alignment test”
(Crovitz and Zener 1962; Hirata 1968), described in n. 1.
5
Participants were instructed to perform the alignment test
eight times with alternating hands and to report each time
which eye led to an alignment failure. Those who reported
stronger alignment for the right eye four times or more were
categorized as right-eyed people; otherwise, they were cate-
gorized as left-eyed people. The results showed that 77%
and 23% of the participants were right-eyed and left-eyed,
respectively.
A 2 × 2 repeated measures ANOVA with location as a
within-subject factor and eyedness as a between-subjects
factor conducted on the perceived heaviness index (α = .82)
revealed a significant main effect of location (F(1, 115) =
23.6, p < .0001), a significant main effect of eyedness (F(1,
115) = 5.2, p < .05), and a significant interaction between
the two (F(1, 115) = 7.8, p < .01). Planned contrasts indi-
cated that the object located at the right side was perceived
as visually heavier than the object located at the left side for
right-eyed participants (M
left
= –.58, M
right
= .55; t(115) =
8.0, p < .0001) but not for left-eyed participants (M
left
= .29,
M
right
= .59; t(115) = 1.2, p > .1). In other words, left-eyed
participants perceived the object in the left visual field as
significantly heavier than right-eyed participants (t(115) =
3.5, p < .001), though they perceived the object in the right
visual field as equally heavy. Therefore, our conjecture that
eyedness moderates the right-heavy perception was
confirmed.
Discussion
In Study 1, using a square visual field, we established that
a pictorial object’s location within a visual field determines
its visual weight. In our subsequent studies, we aim to add
more external validity to our experimental stimuli, so we
u
se six different shaped package facades, including those
we identified in the field study and a circle-shaped package
f
acade that has been observed in gourmet shops (e.g., Dan-
i
sh cookie tins), that display real product images as our
stimuli (illustrated in an abstract form in Figure 2).
A
lthough we gain generalizability using these more real-
istic package shapes, we also introduce a confound because
the different shapes are conducive to different (heavy versus
light) locations. For example, the bottom-heavy principle is
not as relevant to the two horizontal rectangle shapes,
because for these shapes there is not enough space for a pic-
torial object to vary significantly along the height dimen-
sion; similarly, the right-heavy principle is not as relevant to
the two vertical rectangle shapes. Thus, we apply the
bottom-heavy (right-heavy) principle only to the two verti-
cal (horizontal) rectangle shapes. However, we can apply
both principles to square and circle shapes. Therefore, we
redefine heavy (light) as the bottom (top) of vertical rectan-
gle shapes, the right (left) of horizontal rectangle shapes,
and the bottom-right (top-left) of square and circle shapes.
In our subsequent studies, the location effect on perceived
product heaviness refers to the mean difference in perceived
heaviness between products placed at heavy locations (Fig-
ure 2, bottom row) and products placed at light locations
(top row). Similarly, the location effect on package evalua-
tion refers to the mean difference in evaluation between
packages using heavy locations (Figure 2, bottom row) and
packages using light locations (top row). Finally, because in
Study 1 we established the location effect in both a within-
subject and a between-subjects experimental design, in the
following studies, we manipulate location as a within-
subject factor so that we have the power to test the location
effect using a large number of different shapes.
Raghubir and Krishna (1999) demonstrate that con-
sumers’ volume perceptions determine their package pref-
erences, such that the more voluminous the packages are
perceived to be, the more they are preferred. Analogously,
we argue that consumers’ heaviness perceptions translate to
their package preferences as well, such that the heavier the
product appears to be from the package illustration, the
more that package is preferred. On the basis of this argu-
ment, in Studies 2 and 3, we test the location effect on pack-
age evaluation directly without measuring perceived prod-
uct heaviness to control for any potential carryover or
demand effect that measuring the mediator might evoke. In
Study 4, however, we explicitly test the hypothesized medi-
ating role of perceived product heaviness.
STUDY 2: OVERALL VALENCE OF PRODUCT
HEAVINESS MODERATES THE LOCATION EFFECT ON
PACKAGE EVALUATION
As an attribute of a product, heaviness is not universally
preferred. For some products, heaviness is considered posi-
tive (e.g., when weight is positively correlated with durabil-
ity or richness in taste); however, for other products, heavi-
ness is considered negative (e.g., when weight is negatively
correlated with portability). Thus, higher levels of perceived
4
Our analyses found no effect of stimulus presentation format (single
versus pair), stimulus (left-right) position within a pair, and stimulus order.
Thus, in the subsequent analyses, we collapsed data over these factors.
5
For different tests for eyedness and the advantages of unconscious tests,
see Porac and Coren (1976).
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 729
730 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
heaviness may increase or decrease consumer utility as a
function of the valence (positive or negative) assigned to
that attribute. We note that the valence assigned to a product
attribute may further be influenced by the consumer’s cur-
rent affect (Adaval 2001) or whether that attribute is aligned
with a salient goal (Huffman and Houston 1993; Markman
and Brendl 2000), which we address subsequently. If heavi-
ness is considered positive (negative), product image loca-
tions that increase the perception of heaviness will be more
(less) preferred. This is an important hypothesis because it
suggests that there is no universally preferred product image
location (e.g., the bottom or right side of the package
facade), but preferred location is a function of the valence
assigned to heaviness (as a product attribute) with respect to
a specific product category.
Specifically, we investigate two groups of products in
which the valence of heaviness differs. For snack products
(e.g., ChocoStix wafer stick, Chips Ahoy chocolate chip
cookie, and Oreo sandwich cookie), heaviness is good,
because it suggests (1) “getting more” or (2) a richness in
taste. In contrast, for nonfood items (e.g., Logitech Quick-
Cam, Panasonic fluorescent bulb, and Ty Beanie Babies),
heaviness is not good, because it suggests disadvantages
regarding (1) safety (e.g., it is not safe when a heavy light
bulb or computer-mounted video cameras falls out of place
or when kids throw a heavy toy around or at each other) or
(2) portability.
H
2
: For products for which heaviness is considered positive,
packages with the product image placed at heavy locations
will be preferred. In contrast, for products for which heavi-
ness is considered negative, packages with the product
image placed at light locations will be preferred.
Method
We conducted Study 2 (N = 140) to test H
2
using a 2
(product image location: heavy versus light) × 2 (attribute
valence: heaviness is positive versus heaviness is negative)
within-subject design. Manipulation checks (conducted
among a separate group of 124 participants) confirmed that
heaviness was considered a positive attribute for the three
snack products relative to the three nonfood products
(M
difference
= 1.0 on a nine-point scale; t(122) = 5.2, p <
.0001); whereas lightness was considered a positive attrib-
ute for the three nonfood products relative to the three snack
products (M
difference
= 2.1; t(122) = 11.3, p < .0001).
For each of the six products, we prepared 12 schematic
package facades similar to those shown in Figure 2, except
that the axes were removed and the disk was replaced by the
real product image. We organized the 72 stimuli into six
groups of 12 stimuli, such that within each group, (1) heavy
and light locations each occurred six times, (2) the six prod-
ucts each occurred twice, and (3) the six shapes each
occurred twice. We used six sequences to counterbalance
the stimulus order within each group. We randomly
assigned participants to one of the 36 between-subjects con-
ditions (i.e., 6 groups × 6 sequences).
6
In the introduction to the study, participants were shown
a sample schematic package facade along with the corre-
sponding realistic package. This side-by-side comparison
facilitated their understanding of how the schematic draw-
ing translated to an actual package. Then, they viewed the
12 schematic packages assigned to them in pairs and evalu-
6
Our analyses found no effect of the replicate Factors 2 and 3 and of
stimulus order. We also found no effect of stimulus group and sequence.
Thus, in the subsequent analyses, we collapsed data over these factors.
Heavy
Light
Location
Shape
CircleHorizontal RectangleVertical Rectangle Square
H/W: 2 H/W: 1.618 H/W: 1/2 H/W: 1/1.618 H/W: 1 H/W: 1
F
igure 2
MANIPULATIONS OF PRODUCT IMAGE LOCATION AND PACKAGE FACADE SHAPE IN STUDIES 2–4
Notes: H/W: height/width ratio. H/W of 1.618 is the “golden” ratio, or golden section, that has been widely used in product design and package design
(Bloch 1995; Raghubir and Greenleaf 2006). Studies 2 and 3 used all six shapes. Study 4 used the first four shapes.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 730
Is Your Product on the Right Side? 731
ated each package on three nine-point scales (–4 = “dislike/
bad/unsatisfactory,” and 4 = “like/good/satisfactory”). Pre-
senting stimuli in pairs enabled us to reduce the number of
Web pages participants needed to go through from 12 to 6,
thus reducing participant fatigue. The results from Study 1a
also indicated that stimulus presentation format (i.e., single
v
ersus pair) had no effect on visual weight.
Results
We averaged the three scales to form a package evalua-
tion index (α = .96). A 2 × 2 repeated measures ANOVA
conducted on this index revealed a significant location ×
valence interaction (F(1, 139) = 136.0, p < .0001; see Fig-
ure 3). Planned contrasts showed that for products for which
heaviness is good, packages with the product image placed
at heavy locations were evaluated more favorably than
packages with the product image placed at light locations
(M
heavy
= 1.21, M
light
= .19; t(139) = 9.0, p < .0001). For
products for which lightness is good, packages using heavy
locations were evaluated less favorably than packages using
light locations (M
h
eavy
= .23, M
l
ight
= 1.08; t(139) = –7.5,
p < .0001). Thus, H
2
was supported.
Discussion
In Study 2, we confirmed that there is no universally pre-
ferred product image location for package design and that
preferred location is a function of the valence of heaviness
in a specific product category. However, because we could
not control for other attributes that might differ between the
two groups of products, we cannot determine conclusively
that the preferences for different product image locations
were strictly a function of the different valences assigned to
heaviness across the two groups. We remove this confound
in Study 3 by keeping product category constant (i.e., focus-
ing on the snack category for which heaviness is considered
a positive attribute) and manipulate goal to induce differ-
ences in the valence assigned to heaviness. Study 3 will also
h
elp us understand the results of the field study by examin-
ing package evaluations for regular and healthful snacks and
for snacks with and without a health claim.
STUDY 3: CONSUMER GOAL MODERATES THE
LOCATION EFFECT ON PACKAGE EVALUATION
Goals affect how consumers evaluate attribute levels
(
Huffman and Houston 1993; Markman and Brendl 2000)
and assign valence (positive or negative) to a product attrib-
ute. In Study 2, we showed that when no goal was primed,
consumers considered heaviness a positive attribute for
regular snacks, which led to a preference for perceptual
heaviness on the package. However, consumer goals, such
as eating healthfully or being on a diet, could mitigate the
positive valence of heaviness in the snack category. That is,
a less positive valence might be assigned to heaviness
because of its incompatibility with the consumer’s health-
related goal, and as a result, the preference for heavy prod-
uct image locations on the package should become weaker.
On the basis of this argument, we predict a two-way inter-
action between consumer goal and product image location,
such that goal should moderate the preference for packages
using heavy locations (see H
3a
).
In addition to consumer goal, we can compare healthful
snacks (e.g., crackers) with regular snacks (e.g., cookies).
For consumers with a health goal, healthful snacks allow
them to indulge themselves in snacking without compromis-
ing their heath goal. As a result, they should prefer “getting
more” of healthful snacks and thus prefer perceptual heavi-
ness on the package, resembling the tendencies consumers
in the no-goal state exhibited for regular snacks in Study 2.
Although a health goal will weaken the preference for pack-
ages using heavy locations (H
3a
), this weakening effect
should diminish for healthful snacks compared with regular
snacks. In other words, we hypothesize a three-way inter-
action among snack type, consumer goal, and product
image location, such that the two-way interaction described
previously will be further moderated by snack type (see
H
3b
).
H
3a
: There is a preference for packages with the product image
placed at heavy locations under a neutral goal, but this
effect will be weakened under a salient health goal.
H
3b
: A salient health goal weakens the preference for packages
with the product image placed at heavy locations for regu-
lar snacks, but this weakening effect will diminish for
healthful snacks.
Method
We conducted Study 3 (N = 69) to test H
3
using a 2 (prod-
uct image location: heavy versus light) × 2 (consumer goal:
neutral versus health) × 2 (snack type: regular versus health-
ful) mixed design. Consumer goal was the only between-
subjects factor, and we manipulated it using a scrambled
sentence task (Srull and Wyer 1979). Regular snacks were
represented by the three snack products used in Study 2.
Healthful snacks were represented by reduced fat Nilla
.23
1.21
.19
1.08
0
1
2
Heaviness Is Good* Lightness Is Good*
Attribute Valence
Package Evaluation
Heavy location
Light location
Figure 3
PRODUCT ATTRIBUTE VALENCE MODERATES THE EFFECT
OF PRODUCT IMAGE LOCATION ON PACKAGE EVALUATION
*p < .0001.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 731
732 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
Wafers, reduced fat Triscuit crackers, and reduced fat Wheat
Thins.
As in Study 2, we prepared 72 package stimuli, 12 for
each of the 6 snack products.
7
The stimulus construction
and organization also followed Study 2. At the beginning of
the study, participants were asked to complete 20 scrambled
sentences, 10 of which contained a goal-priming word
either for a health goal (e.g., low fat, salad, and workout) or
f
or a neutral goal. Other aspects of this study were similar
to Study 2, except that we used a single-scale measure of
package evaluation (–4 = “dislike,” and 4 = “like”).
Results
A 2 × 2 × 2 repeated measures ANOVA conducted on
package evaluation revealed a significant main effect of
location (F(1, 67) = 216.8, p < .0001), a significant two-way
interaction effect of location × goal (F(1, 67) = 37.1, p <
.0001), and a marginally significant three-way interaction
effect of location × goal × snack type (F(1, 67) = 2.8, p <
.1). We report means and planned contrasts results in Table
2. Planned simple effect contrasts indicated a significant
preference for heavy locations over light locations, which
we refer to as “positive location effect” (PLE = M
h
eavy
location
– M
light location
), in all conditions: 2.69 (t(67) = 11.7,
p < .0001) under neutral goal and .87 (t(67) = 3.6, p <
.0005) under health goal for regular snacks, and 2.19
(t(67) = 9.5, p < .0001) under neutral goal and 1.15 (t(67) =
4.8, p < .0001) under health goal for healthful snacks. How-
ever, consistent with H
3a
, we found evidence that this pref-
erence for heavy locations was weakened by a health goal.
Planned two-way interaction contrasts indicated a signifi-
cantly positive “moderating effect of goal” (MEG =
PLE
neutral goal
– PLE
health goal
) across both types of snacks:
1.82 (t(67) = 5.5, p < .0001) for regular snacks and 1.03
(t(67) = 3.1, p < .005) for healthful snacks. Finally, consis-
tent with H
3b
, we found evidence that this weakening effect
of a health goal further diminished for healthful snacks.
Planned three-way interaction contrasts indicated a signifi-
cantly positive “moderating effect of (snack) type” (MET =
MEG
r
egular snack
– MEG
h
ealthful snack
): .78 (t(67) = 1.7, p <
.05).
Discussion
In Study 3, we held product category constant and
focused on the snack category, for which we demonstrated
in Study 2 that heaviness was considered positive and thus
perceptual heaviness was preferred over perceptual light-
ness on the package. We showed that compared with a neu-
tral goal, a health goal weakened this preference (i.e., a
goal × location two-way interaction). We also showed that
this weakening effect of a health goal further diminished for
healthful snacks compared with regular snacks (i.e., a snack
type × goal × location three-way interaction).
The location effect manifested in the marketplace. In
Study 3, we also provided some systematic evidence to sup-
port our field study findings. If we assume that a health
claim on a package is promoting a health goal for con-
sumers, the results in the field study parallel the results in
Study 3. Specifically, the field study indicated that for regu-
lar snacks (i.e., cookies), packages with a health claim used
light locations significantly more frequently than packages
without a health claim (i.e., 66% versus 17%). This is con-
sistent with the finding in Study 3 that a health goal signifi-
cantly weakened the preference for heavy locations for
regular snacks (i.e., MEG
R
egular
= 1.82). In addition, the
field study indicated that for healthful snacks (i.e., crack-
ers), packages with a health claim used light locations less
frequently than packages without a health claim (i.e., 38%
versus 47%). This is consistent with the finding in Study 3
that the weakening effect of health goal significantly dimin-
ished for healthful snacks (i.e., MET = .78). Overall, we
observe considerable convergence between findings from
the field study and lab Studies 1–3.
The location effect manifested in consumer designs. Stud-
ies 1–3 demonstrate that consumers respond in a predictive
way to different product image locations constructed to test
our theory. However, it would be worthwhile to test whether
consumers actually have a mental model that is consistent
with our theory. To explore this, we conducted Study 3a
(N = 81).
Participants’ task was to design two snack packages by
placing several design elements in a package facade (design
element and package facade were explained to them in sim-
ple terms). On the design Web page, four design elements
7
As in Study 2, our analyses found no effect of the replicate factors,
product instance and shape, and stimulus order. We also found no effect of
stimulus group and sequence. Thus, in the subsequent analyses, we col-
lapsed data over these factors.
Table 2
CONSUMER GOAL AND SNACK TYPE MODERATE THE EFFECT OF PRODUCT IMAGE LOCATION ON PACKAGE EVALUATION
Regular Snack Healthful Snack
Product Image Location
(H: heavy; L: light)
Neutral Goal Health Goal Neutral Goal Health Goal
H L H L H L H L
Package evaluation 1.98*** –.70** 1.12*** .25 1.66 –.53 1.19 .04 Mean
Positive location effect
(PLE: M
heavy
– M
light
)
2.69*** .87** 2.19*** 1.15***
Contrast [1]
Simple effect
Moderating effect of goal
(MEG: PLE
neutral
– PLE
health
)
1.82*** 1.03**
Contrast [2]
Two-way interaction
Moderating effect of (snack) type
(MET: MEG
regular
– MEG
healthful
)
.78***
Contrast [3]
Three-way interaction
*p < .05.
**p < .005.
***p < .0001.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 732
Is Your Product on the Right Side? 733
(digital images of the snack, logo, snack name, and weight
label) were scattered outside a rectangle. Participants were
instructed to drag all four elements into the frame to form a
package design. They could move elements around or click
a “reset” button to try different compositions. When they
w
ere satisfied with the design, they could click a “finished”
button to leave the page, at which point the location of each
design element (x and y coordinates with the top-left corner
o
f the frame as the point of origin) was recorded. After a
practice design, all participants designed one cookie and
one cracker package. Across the two snacks, the logo and
t
he weight label were the same, but the snack image and
product name were different. We used four between-
subjects experimental conditions, such that participants in
Conditions 1 and 2 designed on a horizontal rectangle pack-
age facade (height-to-width ratio =
1
⁄2) and participants in
Conditions 3 and 4 drew on a vertical rectangle frame
(height-to-width ratio = 2). Orthogonal to the package
facade shape, participants in Conditions 1 and 3 designed
for the cookie first and then for the cracker, and in Condi-
tions 2 and 4, this order was reversed.
8
If consumers’ lay theory agreed with our previous find-
ings, we should observe that they use heavy locations more
frequently than light locations for cookie images and do the
opposite for cracker images. The results confirmed this
predication. The relationship between snack image location
and snack type was statistically significant (χ
2
(2) = 10.8,
p < .005). Furthermore, 38% of the participants placed the
cookie image at heavy locations, and 20% placed it at light
locations of the package facade (t(80) = 2.2, p < .05); in
contrast, 17% and 37% placed the cracker image at heavy
and light locations, respectively (t(80) = –2.5, p < .05).
STUDY 4: PERCEIVED PRODUCT HEAVINESS
MEDIATES THE LOCATION EFFECT ON PACKAGE
EVALUATION AND THE MODERATING ROLE OF
STORE SHELF CONTEXT
Although consumers may encounter a product package in
isolation when they shop online, in bricks-and-mortar
stores, a product package is almost always viewed in the
context of other packages on a store shelf. Thus, in an actual
retail store display, the perceived heaviness a consumer may
experience from the graphics of a single package is likely to
be moderated by the graphics of other nearby packages. If
other packages on the shelf are similarly using product
graphics, the effects of a single package are likely to be
mitigated. Conversely, if other packages on the shelf are
using product image locations in light positions, a single
package with the product image in a heavy location is likely
to have a more dramatic effect. Thus, the retailer’s place-
ment of other manufacturers’ packages on the shelf can nul-
lify or reinforce a single manufacturer’s strategy for its own
brand.
Research on attention allocation proposes that consumers
attend to store shelf display through a two-process model.
First, they allocate attentional resources evenly across the
entire display with parallel processing of multiple stimuli.
Second, they concentrate attentional resources on one dis-
play element, with serial processing of different stimuli
(Jonides 1983). Behavioral decision research using eye-
tracking data indicates that these two processes underlie a
three-stage choice model of in-store purchase: The parallel
processing underlies the first stage of choice called “orien-
tation,” in which shoppers overview the entire product dis-
play; the serial processing underlies the second and the third
stages of choice called “evaluation” and “verification,”
respectively, in which shoppers compare between two or
three alternatives and verify the tentatively chosen brand
size (Russo and LeClerc 1994).
T
his literature suggests that store shelf context plays an
i
mportant role in consumer perception and evaluation at the
point of purchase. It also suggests that the effect of context
on evaluation is mediated by perception. In this study, we
test for this mediation mechanism.
The context effect literature distinguishes between two
types of effects: contrast and assimilation. Contrast refers to
the displacement of judgments away from the values of con-
textual stimuli, and assimilation refers to the displacement
of judgments toward the contextual standard (Wedell 1994).
Accordingly, we simulate two types of store shelf context.
A contrasting context is formed when the target package is
different from the contextual packages in terms of product
image location (i.e., a heavy [light] target package is sur-
rounded by light [heavy] packages). An assimilating context
is formed when the target package is the same as the con-
textual packages in terms of product image location (i.e., a
heavy [light] target package is surrounded by heavy [light]
packages).
We propose that for packages with the product image
placed at heavy (light) locations, a contrasting store shelf
context will strengthen the perceptual heaviness (lightness)
of the product, whereas an assimilating context will weaken
this perception. That is, store shelf context is hypothesized
to moderate the location effect on perceived product heavi-
ness, such that contrasting contexts will accentuate and
assimilating contexts will attenuate this effect. We illustrate
this moderation mechanism in Figure 4 and outline it in H
4a
.
We then propose that consumers’ perceptions of product
heaviness will translate to their package evaluations. On the
basis of this mediation hypothesis, we further propose that
store shelf context will also moderate the location effect on
package evaluation, such that contrasting contexts will
accentuate and assimilating contexts will attenuate this
effect. This is a “mediated moderation” because the hypoth-
esized mediation process is responsible for this moderating
effect (Muller, Judd, and Yzerbyt 2005).
9
In other words, if
the perceived product heaviness does not mediate package
evaluation, the moderating effect of store shelf context will
stop at the perception stage without extending to the evalua-
tion stage. We also illustrate this mediated moderation
mechanism in Figure 4 and outline it in H
4b
.
8
Our analyses found no effect of package facade shape and snack order.
Thus, in the subsequent analyses, we collapsed data over these factors.
9
Based on the mediation literature, in a mediated moderation, the path
from the independent variable to the mediator (i.e., X → ME) depends on
the level of the moderator, whereas the effect of the mediator on the
dependent variable (i.e., ME → Y) is constant. In contrast, in a moderated
mediation, the path from the independent variable to the mediator (i.e.,
X → ME) is constant, whereas the effect of the mediator on the dependent
variable (i.e., ME → Y) depends on the level of the moderator. Here, we
propose a mediated moderation rather than a moderated mediation because,
in our case, the moderator is hypothesized to act on the X → ME path
rather than the ME → Y path.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 733
734 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
H
4a
: Store shelf context moderates the location effect on per-
ceived product heaviness.
H
4b
: Perceived product heaviness mediates (i) the location
effect on package evaluation and (ii) the moderating role
of store shelf context in the location effect on package
evaluation.
Method
We conducted Study 4 (N = 42) to test H
4
using a 2 (prod-
uct image location: heavy versus light) × 2 (store shelf con-
text: contrasting versus assimilating) within-subject design.
We used the two regular snack products as in Studies 2 and
3—Chips Ahoy chocolate chip cookie and Oreo sandwich
cookie—to construct the package stimuli.
To prepare the stimuli, for each of the two cookies, we
designed 8 target packages using four different shapes. For
each of the 8 target packages, we created a store shelf simu-
lation by surrounding it with 31 contextual packages that
were of the same four different shapes and various snack
products. The target package was colored, while all contex-
tual packages were black-and-white. In total, we created 16
store shelf simulations (see the Web Appendix at http://
www.marketingpower.com/jmrdec09). They were organized
into four groups of four stimuli, such that within each
group, (1) target packages using heavy and light locations
10
Our analyses found no effect of the replicate factors 3, 4, and 5 on both
perceived heaviness and package evaluation. We also found no effect of
stimulus group and sequence. Thus, in the subsequent analyses, we col-
lapsed data over these factors. However, we found that the effect of stimu-
lus order was significant on perceived heaviness (p < .01, the stimuli in the
second and third positions were rated higher than the stimulus in the first
position) but not on package preference. This is the only significant order
effect we observed across all studies. Because we always controlled for
order by counterbalancing, this result does not compromise our
conclusions.
F
igure 4
THE MODERATING ROLE OF STORE SHELF CONTEXT AND THE MEDIATING ROLE OF PERCEIVED PRODUCT HEAVINESS
A: The Mediated Moderation Model
Perceived
product
heaviness
Package evaluation Product image location
Store shelf context
S
Strengthened
perceived
heaviness/lightness
Weakened
perceived
heaviness/lightness
Increased
package
liking/disliking
Decreased
package
liking/disliking
Heavy/light
product image location
Contrasting
Context
Assimilating
Context
B: Specific Effects of Contrasting and Assimilating Store Shelf Contexts
each occurred twice, (2) contrasting and assimilating con-
texts each occurred twice, (3) the two cookies each occurred
twice, (4) the four shapes of target package each occurred
once, and (5) the four locations of the target package on the
shelf each occurred once. Moreover, we used four
sequences to counterbalance the stimulus order within each
group.
10
Participants were randomly assigned to 1 of the 16
between-subjects conditions (i.e., 4 stimulus groups × 4
sequences).
In the introduction to the study, participants were shown
a sample store shelf simulation composed of schematic
package designs and the corresponding store shelf snapshot
composed of realistic packages (see the Web Appendix at
http://www.marketingpower.com/jmrdec09). Then, they
viewed the four stimuli assigned to them one by one and, for
each stimulus, rated the cookie in the target package on the
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 734
Is Your Product on the Right Side? 735
Models 4–6 tested the mediated moderation hypothesis
(H
4b(ii)
):
(4) Package evaluation = β
40
+ β
41
location + β
42
context
+ β
43
location × context + ε
4
,
three scales used to measure perceived heaviness in Study
1. Finally, they viewed the four stimuli again, one at a time,
and rated the target package on the three scales used to
measure package evaluation in Study 2.
R
esults
Perceived product heaviness. We averaged the three
scales used to measure perceived product heaviness to form
a perceived heaviness index (α = .83). A 2 × 2 repeated
measures ANOVA conducted on this index revealed a sig-
nificant main effect of location (F(1, 41) = 10.6, p < .005)
and a significant interaction effect of location × context
(F(1, 41) = 8.7, p < .01; see Figure 5, Panel A). Planned con-
trasts showed that in the contrasting context, cookies placed
at heavy locations were perceived as visually heavier than
cookies placed at light locations (M
h
eavy
= .43, M
l
ight
= –.72;
t(41) = 4.4, p < .0001); however, this effect disappeared in
the assimilating context. Thus, H
4
a
was supported.
Package evaluation. The three scales used to measure
package evaluation were averaged to form a package
evaluation index (α = .96). The same ANOVA conducted on
this index revealed a significant main effect of location (F(1,
41) = 7.5, p < .01) and a significant interaction effect of
location × context (F(1, 41) = 7.9, p < .01; see Figure 5,
Panel B). Planned contrasts showed that in the contrasting
context, packages with the product image placed at heavy
locations were evaluated more favorably than packages with
the product image placed at light locations (M
heavy
= .88,
M
light
= –.54; t(41) = 3.9, p < .0005); however, this effect
again disappeared in the assimilating context. Thus, we pro-
vide evidence that the moderation of store shelf context in
the location effect on perceived heaviness carries over to
package evaluation.
Mediation and mediated moderation analyses. To test
H
4b
, we estimated the six regression models that Muller,
Judd, and Yzerbyt (2005, Equations 1–6) suggest. Models
1–3 tested the mediation hypothesis (H
4b(i)
):
(1) Package evaluation = β
10
+ β
11
location + ε
1
,
(2) Perceived heaviness = β
20
+ β
21
location + ε
2
, and
(3) Package evaluation = β
30
+ β
31
location
+ β
32
perceived heaviness + ε
3
.
We report the results in Table 3. For Model 1, the overall
effect of location on package evaluation, b
11
, was signifi-
cant; for Model 2, the effect of location on the mediator, b
21
,
was also significant; and for Model 3, the effect of the
mediator on package evaluation, b
32
, was significant, but the
residual direct effect of location on package evaluation
(controlling for the mediator), b
31
, was not, indicating that
perceived heaviness fully mediates the effect of location on
package evaluation. The Sobel (1982) test confirms that the
indirect effect of location on package evaluation through the
mediator was significant (z = 2.4, p < .05). Moreover, there
was an equality relationship among the estimated parame-
ters of these models, such that the difference between the
overall effect and the residual direct effect of location on
package evaluation (b
11
– b
31
) was equal to its indirect
effect through the mediator (b
21
× b
32
), demonstrating
excellent model fit.
F
igure 5
STORE SHELF CONTEXT MODERATES THE EFFECS OF
PRODUCT IMAGE LOCATION ON PERCEIVED PRODUCT
HEAVINESS AND PACKAGE EVALUATION
A: Perceived Product Heaviness
–.15
.43
–.21
–.72
–1
0
1
Contrasting Context** Assimilating Context
Store Shelf Context
Perceived Heaviness
Heavy location
Light location
.36
.88
.37
–.54
–1
0
1
Contrasting Context* Assimilating Context
Store Shelf Context
Package Evaluation
Heavy location
Light location
B: Package Evaluation
*p < .0005.
**p < .0001.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 735
(5) Perceived heaviness = β
50
+ β
51
location + β
52
context
+ β
53
location × context + ε
5
, and
(6) Package evaluation = β
60
+ β
61
location + β
62
context
+ β
63
location × context
+ β
6
4
perceived heaviness
+ β
65
perceived heaviness × context
+ ε
6
.
To establish mediated moderation, we should first observe
an overall moderation of the location effect (i.e., β
43
should
be nonzero). The question then is whether the mediating
process partially or fully accounts for this moderation. If it
does, the moderation of the residual direct effect of the loca-
tion should be reduced or eliminated compared with the
moderation of the overall location effect; that is, β
63
should
be smaller in magnitude than β
43
, or close to zero in the case
of full mediation. The results supported this mechanism (see
Table 3). For Model 4, the overall effect of location on pack-
age evaluation (X → Y) depended on the level of the mod-
erator (i.e., b
43
was significant); for Model 5, the effect of
location on the mediator (X → ME) also depended on the
level of the moderator (i.e., b
53
was significant); and for
Model 6, the effect of the mediator on package evaluation
(ME →Y) was significant (i.e., b
64
was significant) but con-
stant (i.e., b
65
was not significant), and the residual direct
effect of location on package evaluation (controlling for the
mediator) did not depend on the moderator (i.e., b
63
was not
significant). This pattern indicated that perceived heaviness
fully mediated the moderating role of context in the location
effect on package evaluation (Muller, Judd, and Yzerbyt
2005, p. 856). The Sobel test confirmed that the (moder-
ated) indirect effect of location on package evaluation
through the mediator was significant (z = 2.1, p < .05).
Moreover, an equality relationship existed among the esti-
mated parameters of these models, such that the difference
736 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
between the overall (moderated) effect and the residual
(moderated) direct effect of location on package evaluation
(b
4
3
– b
6
3
) was equal to its (moderated) indirect effect
through the mediator (b
64
× b
53
+ b
65
× b
51
), demonstrating
excellent fit.
Discussion
In Study 4, we provided evidence that store shelf context
moderated the location effect on perceived product heavi-
ness, such that this effect appeared in a contrasting context
but disappeared in an assimilating context. In addition, store
shelf context moderated the location effect on package
evaluation, such that this effect appeared in a contrasting
context but disappeared in an assimilating context. Finally,
the latter moderation was due to perceived product heavi-
ness fully mediating the location effect on package
evaluation.
GENERAL DISCUSSION
Summary
In this research, we examine the effects of a package
design element—namely, the location of the product image
within a package facade—on consumers’ perceived product
heaviness and package evaluations. A field study and a
series of laboratory studies provide convergent evidence
that this design element can be used strategically to create
favorable perceptions about the product and evaluations
toward the package.
Specifically, in a field study conducted in a metropolitan
supermarket in the snack category, we find that the location
of the snack image on a package facade varies systemati-
cally as a function of snack type (e.g., cookie versus
cracker) and feature (e.g., containing healthful ingredients
or not), suggesting that the location of the product image is
used to communicate product information. In the art and
visual perception literature, location is among the factors
that determine a pictorial object’s visual weight (e.g., Arn-
heim 1974). Drawing from this literature and the research
Mediation Model Intercept
Product Image
Location (X)
Perceived
Heaviness
(ME)
1 Package
evaluation (Y)
.27
(.15)
.35*
(.15)
2 Perceived
heaviness (ME)
–.16**
(.12)
.30*
(.12)
3 Package
evaluation(Y)
.38**
(.13)
.15
(.13)
.67***
(.08)
Mediated
Moderation Model Intercept
Product Image
Location (X)
Store Shelf
Context (MO)
Location ×
Context
(X × MO)
Perceived
Heaviness
(ME)
Heaviness ×
Context
(ME × MO)
4 Package
evaluation (Y)
.27
(.15)
.35*
(.15)
–.10
(.15)
.36*
(.15)
5 Perceived
heaviness (ME)
–.16
(.12)
.30*
(.12)
.02
(.12)
.27*
(.12)
6 Package
evaluation (Y)
.37**
(.13)
.15
(.14)
–.10
(.13)
.17
(.14)
.65***
(.09)
.02
(.09)
*p < .05.
**p < .005.
***p < .0001.
T
able 3
REGRESSION COEFFICENTS (STANDARD ERRORS) FOR THE MEDIATION AND MEDIATED MODERATION ANALYSES IN STUDY 4
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 736
Is Your Product on the Right Side? 737
on ocular dominance (e.g., Porac and Coren 1976), in Study
1, we experimentally establish that heavy (light) locations
are the bottom (top), right (left), and bottom-right (top-left)
of a visual field because pictorial objects placed at these
l
ocations are perceived as heavier (lighter). We then apply
this “location effect” to package graphic design. In Study 2,
we demonstrate that for products for which heaviness is
c
onsidered a positive attribute, packages with the product
image placed at heavy locations are preferred, whereas for
products for which heaviness is considered negative, pack-
ages using light locations are preferred. This suggests that
there is no universal preference for heavy or light locations
but that the placement of the product image should depend
on the valence assigned to heaviness. Focusing on a product
category for which heaviness is considered positive (i.e., the
snack category), in Study 3, we show that consumers’
salient health goal weakens the preference for packages
using heavy locations. However, the influence of goal
becomes weaker if the snack category is considered health-
ful. This suggests that in the snack category, the preferred
product image location should be a function of snack type
(e.g., regular versus healthful snacks) and consumer goal
(e.g., dieters versus nondieters). These results are consistent
with the field study findings. In Study 4, we explore another
moderating condition for the location effect on package
evaluation as well as the underlying mechanism. We dis-
cover that store shelf context first moderates the location
effect on perceived product heaviness, such that this effect
appears in a contrasting context but disappears in an assimi-
lating context. Then, because perceived product heaviness
mediates the location effect on package evaluation, the
moderating effect of store shelf context further translates to
package evaluation (i.e., mediated moderation), such that
packages using heavy locations are evaluated favorably in a
contrasting context but not in a assimilating context. Our
analyses confirm both the mediation and the mediated mod-
eration processes.
Contributions to the Literatures and Further Research
Visual perception. Bottom-heavy and right-heavy are two
principles of visual weight in the literature of art and visual
perception (e.g., Arnheim 1974). However, bottom-heavy
has never been experimentally tested before, and right-
heavy was tested only among right-eyed people (Hirata
1968). We experimentally test these two principles and con-
firm that eyedness moderates the right-heavy perception.
We also apply these perceptual effects to the context of
packaging and identify moderating conditions. In doing so,
we contribute to the art and visual perception literature.
Research on consumer perceptual processing of visual
cues has shown that consumers frequently make perceptual
judgments (e.g., how big, how long, how many) based on
visual cues (e.g., Folkes and Matta 2004; Krider, Raghubir,
and Krishna 2001; Krishna 2006; Raghubir and Krishna
1999), and these judgments are often biased (for a review,
see Krishna 2007). We also contribute to this line of more
applied research by examining another perceptual judg-
ment—namely, how heavy a product is perceived to be—
and similarly illustrating systematic biases in this visual
perception.
Packaging. Researchers in this area have examined how
package shape affects evaluation through perceived volume
(e.g., Krishna 2006; Raghubir and Krishna 1999; Wansink
and Van Ittersum 2003; Yang and Raghubir 2005). We con-
tribute by examining how the location of the product image
on a package facade affects evaluation through perceived
product heaviness. Perceived heaviness is a particularly
notable construct, and further research should explore its
antecedents and consequences in greater depth. According
to Arnheim (1974), factors such as color, shape, size, direc-
tion, depth, isolation, and intrinsic interest can also deter-
mine visual weight. Among these factors, color and shape
are particularly relevant to package graphic design. We con-
ducted a preliminary study to test whether different colors
induce differential visual heaviness perceptions and found
support for this idea. We pretested the visual weight of ten
colors (the three primary colors, two secondary colors, and
five tertiary colors) that are commonly used in graphic
design. We found that when saturation was held constant,
these ten hues differed considerably from one another in
visual weight. Furthermore, for each hue, the high-saturated
color chip was perceived as significantly heavier than the
low-saturated color chip. These preliminary findings sug-
gest that color has a strong potential to influence package
evaluation through perceived heaviness, and thus this is a
fruitful area for further research. Container shape has been
found to affect volume perception, such that the greater the
height-to-weight ratio of a container, the greater is the esti-
mated volume (i.e., the elongation effect; e.g., Raghubir and
Krishna 1999). Shape can also affect visual weight percep-
tion, and we leave this topic for further research as well.
REFERENCES
Adaval, Rashmi (2001), “Sometimes It Just Feels Right: The Dif-
ferential Weighting of Affect-Consistent and Affect-Inconsistent
Product Information,” Journal of Consumer Research, 28
(June), 1–17.
Arnheim, Rudolf (1974), Art and Visual Perception: A Psychology
of the Creative Eye. Berkeley: University of California Press.
Bloch, Peter H. (1995), “Seeking the Ideal Form: Product Design
and Consumer Response,” Journal of Marketing, 59 (July), 16–30.
Bourassa, D.C., I.C. McManus, and M.P. Bryden (1996), “Hand-
edness and Eye-Dominance: A Meta-Analysis of Their Rela-
tionship,” Laterality, 1 (1), 5–34.
Charpentier, Augustin (1891), “Analyse Experimentale de
Quelques Elements de la Sensation de Poids” [“Experimental
Study of Some Aspects of Weight Perception”], Archives de
Physiologie Normales et Pathologiques, 3, 122–35.
Childers, Terry L. and Michael J. Houston (1984), “Conditions for
a Picture-Superiority Effect on Consumer Memory,” Journal of
Consumer Research, 11 (September), 643–54.
Coren, Stanley and Clare Porac (1976), “Size Accentuation in the
Dominant Eye,” Nature, 260 (April), 527–28.
Crovitz, H.F. and K.A. Zener, (1962), “A Group-Test for Assessing
Hand- and Eye-Dominance,” American Journal of Psychology,
75 (June), 271–76.
Folkes, Valerie and Shashi Matta (2004), “The Effect of Package
Shape on Consumers’ Judgments of Product Volume: Attention
as a Mental Contaminant,” Journal of Consumer Research,
31(September), 390–401.
Francis, J.L. and K.A. Harwood (1959), “The Variation of the Pro-
jection Center with Differential Stimulus and Its Relation to
Ocular Dominance,” Transactions of the International Optomet-
ric Congress, 21, 75–88.
Greenough, Horatio (1947), Form and Function: Remarks on Art,
Design, and Architecture. Berkeley: University of California
Press.
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 737
Hirata, Ken-ichi (1968), “Experimental Study on Right-Left Prob-
lems in Visual Balance,” Psychologia: An International Journal
of Psychology in the Orient, 11 (3–4), 139–42.
Holbrook, Morris B. and William L. Moore (1981), “Feature Inter-
a
ctions in Consumer Judgments of Verbal Versus Pictured Pre-
sentations,” Journal of Consumer Research, 8 (June), 103–113.
Huffman, Cynthia and Michael J. Houston (1993), “Goal-Oriented
Experiences and the Development of Knowledge,” Journal of
Consumer Research, 20 (September), 190–207.
Jonides, John (1983), “Further Toward a Model of the Mind’s
Eye’s Movement,” Bulletin of the Psychonomic Society, 21 (4),
247–50.
Kanizsa, Gaetano and Giorgio Tampieri (1968), “Nuove Osser-
v
azioni Sull’orientamento Retinico ed Ambientale” [“New
Observations on Retinal and Environmental Orientation”], in
Ricerche Sperimentali Sulla Percezione [Experimental Studies
on Perception], G. Kanizsa and G. B. Vicario, eds. Trieste, Italy:
Universita degli Studi di Trieste, 51–68.
Krider, Robert E., Priya Raghubir, and Aradhna Krishna (2001),
“Pizzas: π or Square? Psychophysical Biases in Area Compari-
sons,” Marketing Science, 20 (4), 405–425.
Krishna, Aradhna (2006), “Interaction of Senses: The Effect of
Vision Versus Touch on the Elongation Bias,” Journal of Con-
sumer Research, 32 (March), 557–66.
——— (2007), “Spatial Perception Research: A Integrative
Review of Length, Area, Volume, and Number Perception,” in
Visual Marketing: From Attention to Action, Michel Wedel and
Rik Pieters, eds. Mahwah, NJ: Lawrence Erlbaum Associates,
167–92.
Markman, Arthur B. and C. Miguel Brendl (2000), “The Influence
of Goals on Value and Choice,” in The Psychology of Learning
and Motivation: Advances in Research and Theory, Vol. 39,
Douglas L. Medin, ed. San Diego: Academic Press, 97–128.
Mefferd, R.B., Jr., and B.A. Wieland (1969), “Influence of Eye
Dominance on the Apparent Centers of Simple Horizontal
Lines,” Perceptual and Motor Skills, 28 (3), 847–50.
Muller, Dominique, Charles M. Judd, and Vincent Y. Yzerbyt
(2005), “When Moderation Is Mediated and Mediation Is Mod-
erated,” Journal of Personality and Social Psychology, 89 (6),
852–63.
Ogle, K.N. (1950), Researches in Binocular Vision. New York:
Hafner.
Porac, Clare and Stanley Coren (1976), “The Dominant Eye,” Psy-
chological Bulletin, 83 (5), 880–97.
Porta, J.B. (1593), De Refractione Optices Parte: Libri Novem.
Naples: Carlinum & Pacem.
R
aghubir, Priya and Eric A. Greenleaf (2006), “Ratios in Propor-
tion: What Should the Shape of the Package Be?” Journal of
Marketing, 70 (April), 95–107.
——— and Aradhna Krishna (1999), “Vital Dimensions in Vol-
ume Perception: Can the Eye Fool the Stomach?” Journal of
Marketing Research, 36 (August), 313–26.
Russo, J. Edward and France LeClerc (1994), “An Eye-Fixation
Analysis of Choice Processes for Consumer Nondurables,”
Journal of Consumer Research, 12 (September), 274–90.
S
choen, Z.J. and C.F. Scofield (1935), “A Study of the Relative
Neuromuscular Efficiency of the Dominant and Non-Dominant
Eye in Binocular Vision,” Journal of General Psychology, 12
(10), 156–81.
Scott, R.B. and F.C. Summer (1949), “Eyedness as Affecting
Results Obtained with the Howard and Dolman Depth Percep-
tion Apparatus,” Journal of Psychology, 27 (April), 479–82.
Sobel, Michael E. (1982), “Asymptotic Confidence Intervals for
Indirect Effects in Structural Equation Models,” in Sociological
Methodology, S. Leinhart, ed. San Francisco: Jossey-Bass,
290–312.
Srull, Thomas K. and Robert S. Wyer (1979), “The Role of Cate-
gory Accessibility in the Interpretation of Information About
Persons: Some Determinants and Implications,” Journal of Per-
sonality and Social Psychology, 37 (10) 1660–72.
Walls, G.L. (1951), “A Theory of Ocular Dominance,” A.M.A.
Archives of Ophthalmology, 45 (4), 387–412.
Wansink, Brian and Koert Van Ittersum (2003), “Bottom Up! The
Influence of Elongation on Pouring and Consumption Volume,”
Journal of Consumer Research, 30 (December), 455–63.
Wedell, Douglas H. (1994), “Contextual Contrast in Evaluative
Judgments: A Test of Pre- Versus Postintegration Models of
Contrast,” Journal of Personality and Social Psychology, 66 (6),
1007–1019.
Wolfflin, Heinrich (1950), Principles of Art History: The Problem
of the Development of Style in Later Art. M.D. Hottinger, trans.
New York: Dover.
Yang, Sha and Priya Raghubir (2005), “Can Bottles Speak Vol-
umes? The Effect of Package Shape on How Much to Buy,”
Journal of Retailing, 81 (4), 269–81.
738 JOURNAL OF MARKETING RESEARCH, DECEMBER 2009
jmkr.46.6.725:JMR6B 11/18/09 9:21 AM Page 738
Copyright of Journal of Marketing Research (JMR) is the property of American Marketing Association and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.
