Vital Dimensions in Volume Perception: Can the Eye Fool the Stomach?
Author(s): Priya Raghubir and Aradhna Krishna
Source: Journal of Marketing Research, Vol. 36, No. 3, (Aug., 1999), pp. 313-326
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/3152079
Accessed: 17/04/2008 09:55
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ama.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We enable the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
http://www.jstor.org
PRIYARAGHUBIR
and ARADHNA
KRISHNA*
Given the numberof volume judgments made by consumers, for example, deciding which package is largerand by how much, it is surprising that littleresearch pertainingto volume perceptions has been done in
marketing.In this article, the authors examine the interplayof expectations based on perceptualinputs versus experiences based on sensory
inputin the context of volume perceptions.Specifically,they examine biases in the perceptionof volume due to container shape. The height of
the containeremerges as a vitaldimensionthat consumers appear to use
as a simplifyingvisual heuristicto make a volume judgment. However,
perceived consumption,contraryto perceived volume, is related inversely to height. This lowered perceived consumption is hypothesized and
shown to increase actual consumption.A series of seven laboratoryexperiments programmaticallytest model predictions. Results show that
perceived volume, perceived consumption, and actual consumptionare
relatedsequentially.Furthermore,the authors show that containershape
affects preference,choice, and postconsumptionsatisfaction.The authors
discuss theoretical implicationsfor contrast effects when expectancies
are disconfirmed,specificallyas they relateto biases in visual information
processing, and provide managerial implications of the results for
package design, communication,and pricing.
Dimensions in
Vital
Can
the
Eye
Volume
Fool
e
Perception:
Stomach?
Mavis,l a psychology doctorate,went to the Beach Bar
and had to keep getting refills of red wine. She got tired
of the walk and asked for a largerglass, for which she
paid a higher price. She was surprisedwhen she realized she had paid a higher price for the same volume.
At the same pub on anotherday, Joe, a marketingprofessor, insisted that the Carlsbergpint was larger than
the Fosters pint. Whereasthe Fosters is served in a kegshaped glass, the Carlsberg glass is taller and shaped
like a tankard.They both contain one pint of beer.
'Examplesarebasedon actualincidents.Nameshavebeendisguisedto
maintainanonymity.
David, a supplier of bar equipment, purchased new
teacups for his wife because their old ones seemed
small. After using the new teacups, David's wife complained that they were not as satisfying as their old set.
David was astonishedbecause the new cups looked bigger. It was evidently an illusion; the cups were the same
volume.
*PriyaRaghubiris an assistantprofessor,HaasSchool of Business,
Universityof California,Berkeley(e-mail:raghubir@haas.berkeley.edu).
AradhnaKrishnais an associateprofessor,BusinessSchool,Universityof
Theauthorsacknowledge
theasMichigan(e-mail:aradhna@umich.edu).
sistanceof JasminLee in preparing
the stimulimaterialsfor Study1, Iris
Chowfor helpin conductingthe footnotedstudy,PeterDeMeyerforhelp
in conductingStudies3 and4, andHeatherHoneafor helpin conducting
Studies2, 5, 6, and7. TheythankBobKriderandDonLehmannfortheir
at the BDTCampat the
helpfulsuggestions.Theyalso thankparticipants
University of Colorado at Boulder and the Berkeley/Davis/Santa
Clara/Stanford
Colloquium,ItamarSimonson,the threeJMRreviewers,
andAssociateEditorMichaelHoustonfor theirmanyhelpfulcomments
and suggestions. This research was funded partially by grant
DAG95/96.BM77from the ResearchGrantsCommission,Hong Kong,
throughthe HongKongUniversityof ScienceandTechnologyandawarded to the first author.This articleis dedicatedto the authors'children,
Shikhar,Siddhant,andKamya,whoseresistanceto foodprompted
thisinTointeractwithcolleagueson specificarticlesin thisissue,see
vestigation.
"Feedback"
on theJMRWebsiteat www.ama.org/pubs/jmr.
On a domestic flight, Sandy decided to try a new Lychee fruit drink, packaged in a tetrapackof 200 milliliters (ml). He was surprisedat how full he felt midway
throughdrinkingthe juice. It had seemed such a small
portionwhen he accepted it.
For a business class transatlantic flight, the airline
changed the glasses in which it served champagnefrom
saucers to flutes. Sheila, who was tryingto cut down on
alcohol, asked to get just half a glass. She was surprised
at how quickly it was gone and wished she had asked
for more.
In this article, we examine the effect of container shape
on volume perceptions. Given that actual volume has been
Journalof Marketing
Research
313
Vol. XXXVI (August 1999), 313-326
314
shown to affect actual consumption positively (Wansink
1996), we explore the implicationsof higher perceived volume on both actual and perceived consumption.The interplay among three constructs, perceived volume (volume
perception preconsumption),perceived consumption (volume perceptionpostconsumption),and actual amount consumed, are investigated systematically.Our results suggest
that one's eyes can fool one's stomach.
Accurate volume judgments are complicated processes
that requireestimationof lineardimensions and their aggregation per normatively correct formulas. Heuristic processors of real-world,three-dimensionalinformationare likely
to simplify the volumejudgmenttask in terms of one or two
dimensions, which can lead to systematic biases in volume
perceptions.Our precedingexamples, far from demonstrating anomalousconsumerbehavior,may reflect the standard
way in which consumers make volume judgments. In this
article, we explore which dimensions dominate heuristic
processing of volume judgments;the ensuing perceptualbiases; the mannerin which these perceptionsare amendedin
the face of experientialsensory inputs;and the implication
of these biases for marketersinterestedin package choice,
actual consumption, and postconsumptionsatisfaction.Results of seven studies show that elongation of a container
has a positive effect on volume perceptions, actual consumption, package preference, and package choice but a
negative effect on perceived consumption and postconsumption satisfaction. Although volume perceptions have
been studied extensively in cognitive psychology, albeit
with inconsistent results, they have not received much attention in marketing. Consumption perceptions, to our
knowledge, have not been studied previously. This is puzzling because both volume and consumption perceptions
have many implicationsfor package shape decisions.
Package shape decisions are increasingin importancefor
managers.For example, a manufacturerof papercups is facing competition from a company that makes larger cups.
The manufacturerwants to design a larger paper cup that
maximizes perceived volume for the same amount of raw
material.What shape cups should it make? There does not
appear to be conventional wisdom regarding package
shapes, and a variety of shapes abound in the marketplace.
Although many shapes are now part of the brand image
(e.g., the Coca-Cola bottle), for a new productintroduction,
manufacturersmust decide on the dimensions of the package. To the extent thatconsumersdo not read volume information on the packaging,packagesthatappearlargerwill be
more likely to be purchased,ceteris paribus. Volume estimation is also importantif it affects actual consumption.
Certainpackage shapes might representa double-win situation; that is, they may be more likely to be chosen because
they are perceived to be bigger, and because they are perceived to be bigger, they also may be consumed faster.
Apartfrom the marketingimplicationsfor packaging,this
article also contributesto researchin cognitive psychology.
We propose and investigate the constructof perceived consumption. We also examine for the first time consequences
of volume perceptionand perceived consumptionfor actual
consumption.From a theoreticalstandpoint,we uncover an
interestingillusion, the perceived size-consumption illusion
(PCI). This illusion suggests that perceptionsof volume, as
a function of the elongation of a container,reverse before
JOURNAL OF MARKETINGRESEARCH, AUGUST 1999
and after consumption.We theorize that this occurs because
of the inconsistency between seeing and experiencing; that
is, subjects' relative perceptionsof two objects reverse before and after experiencingthe stimulus.This illusion is related to a highly researchedeffect in the cognitive psychology literature,the size-weight illusion (SWI), which was
documented more than a century ago (Charpentier1891).
The similarity in the two illusions-expectation based on
perceptualinputis contradictedby a sensory experientialinput, which leads to a contrasteffect (a reversal in the perception)-suggests that the nomological constructunderlying the two illusions is the disconfirmedexpectation.
The article is organized into four sections (see Figure 1)
that systematically investigate the effect of package shape
on volume perceptions,preferenceand choice, consumption
(perceived and actual), and postconsumptionsatisfaction.
The first section explores the antecedentsof volume perceptions. We summarize existing research from cognitive
psychology on volume perceptions and test the elongation
hypothesis, which states that taller containersare perceived
to be bigger. Study 1 demonstratesthis bias in volume perceptions for cylindrical shapes (cans, jars, bottles, and so
forth). Study 2 examines whetherthe effect is attenuatedin
conditions of high motivation and strengthenedundercognitive load. The second section introducesthe construct of
perceived consumptionand explores consequences of biases in volume perceptions on perceived consumption. We
summarizeprevious literatureon the SWI and derive implications for the perceived consumption construct. Study 3
tests the effect of container shape on perceptions of consumption, holding both actual and consumed volume constant. The third section explores consequences of biases in
volume perceptions (pre- and postconsumption)on actual
consumption.Studies 4 and 5 measurethe effect of container shape on actual consumptionand test whetherthis effect
is mediated by perceived consumption. The fourth section
(Studies 5-7) examines the consequences of these biases,
that is, whether the perceived volume effect translates to
preference, choice, and postconsumption satisfaction. We
conclude with theoretical implications for the manner in
which consumersuse visual cues to make spatialjudgments
and, more generally,the implicationsof sensory experiential
inputs that contradictperceptuallybased expectancies. We
also offer managerialimplicationsfor packagedesign, communication,and pricing.
ANTECEDENTSOF BIASESIN VOLUMEPERCEPTION
Judgmentsof size, area, and volume are far from trivial,
requiring complicated formulas and calculations. Take a
simple example of everyday occurrence. To compare volumes of two juice containers-one in a cuboid carton and
the other in a cylindrical can (both 240 ml)-a consumer
would need to make five linearjudgments:the heights of the
two containers, the width and depth of the cuboid carton,
and the diameter of the cylindrical can. These linear estimates2 then would have to be combined according to geometric formulasfor the two shapes, which would need to be
2Forcylinders, ratherthan using the linearestimate directly,the estimated diameterwould need to be halved and the result squaredto form the input for the next process. The value of the third parameteralso would need
to be retrievedfrom memory.For many consumers,this numbermay not be
easily accessible.
VolumePerception
315
Figure 1
ANTECEDENTS
ANDCONSEQUENCESOF VOLUMEPERCEPTION
?_
Antecedentsof Biases in
VolumePerception
Effects of Container
Shape on Perceived
Consumption
e
i
i
i
Effects of
Perceived
Volumeon
Actual
Consumption
Consequencesof
VolumePerception
*Motivation
*CognitiveLoad
retrieved from memory. The resulting two numbers then
would have to be comparedto determinewhetherone container carried more juice than the other. Of course, consumers instead might read the label of the container.
However, research consistently has demonstratedthat consumers seldom read details beyond the final price of the
product and, often, not even that (Dickson and Sawyer
1986). Furthermore,decisions of this nature are made by
millions of consumers multiple times a week, if not every
day. In addition, they are made in a short period of time.
Although the preceding example is a caricatureof the normatively correct process involved when making volume
judgments, the purpose is to illustrate that, because of the
level of effort involved in making accuratejudgments, consumers are likely to resortto judgment shortcutsin a tradeoff between effort and accuracy (Payne, Bettman, and
Johnson 1988).
When the level of accuracy desired does not warrantthe
effort required,consumersmay resortto simplifying rules of
thumb, or heuristics, for volume judgments. They may rely
on one or two dimensions and ignore or underweightthe
third to make volume estimation easier. Although most researchersstudying volume judgments would agree with this
proposition,there is no general agreementon whetherthese
effects, which have been shown in children, continue for
adults.
Prior Researchon Biases in VolumeJudgmentDue to Shape
More than 50 years ago, Piaget studiedchildren'sperceptions of volume. In a typical Piagetian experiment,colored
liquid was poured from a tall cylinder into a shorter,wider
cylinder.The height of the liquid in the second cylinder was
lower. Childrenthen were asked whetherthe volume of the
liquid had remained the same or had been reduced. Those
who recognized that the volume had remained the same
were exhibiting "conservationof mass." In a series of studies, Piaget (1968; Piaget, Inhelder, and Szeminska 1960)
found that primaryschool children appearedto use only the
height of the container when making volume judgments;
they believed that the volume had been reduced when the
liquid was poured into a wider glass. The predominantuse
of a single dimension-height-to make three-dimensional
judgments was termedthe "centrationhypothesis."
Using Piaget's experiments as a basis, Holmberg (1975)
proposed the elongation hypothesis, in which height was
conceptualized not in terms of an absolute metric but in
units of width. This hypothesis stated that the greater the
height-to-widthratioof a container,the greaterwas the estimated volume. Holmbergfound supportfor this hypothesis
using both cylindrical and cuboid shapes. Frayman and
Dawson (1981) tested the elongation effect for cylindersand
found weak supportfor the effect. At low volumes (<128 cubic centimeters[cm3]),elongation has a significanteffect on
perceived volume, with short cylinders perceived to be
316
JOURNAL OF MARKETINGRESEARCH, AUGUST 1999
smaller than medium and tall cylinders. However, as the
volume of the cylinders increases in absolute terms, the
elongation effect reduces, with no significantdifferences in
volume estimates (by elongation) for cylinders >1024 cm3.
Support for the elongation effect also is found by Been,
Braunstein,and Piazza (1964) and Pearson (1964); if two
cylinders of equal volume are reduced to new identical volumes, one by reducingthe height and the other by reducing
the width, the cylinder whose height is reduced appears
smaller.
It is not clear,however,thatthe elongationeffect is robust
enough to carry across contexts and experimental procedures. Specifically, it is not clear if these effects translateto
consumer judgments of volume for frequently purchased
products.They may not for several reasons. First, many of
the experiments previously reported were conducted with
grade school age and younger children and indicate that estimation procedures change as the subjects become older
(e.g., Piaget, 1968; Piaget, Inhelder,and Szeminska 1960).
Second, the experimentsused stimuli that were constructed
per the experimentaldesign of the researcher(e.g., styrofoam or white cardboardby Fraymon and Dawson 1981;
gray painted wood by Holmberg 1975) and were not familiar to subjects.Consumers,in contrast,are typically familiar
with the containers of frequentlypurchasedproducts,such
as soda cans and jam jars, and may have their own consumption experiences to guide their volume estimate. This
might make them less susceptible to an elongation effect.
Third, in many of the studies previously reviewed, volume
judgments were elicited using apparatusesthat consumers
would not use in naturalsettings. Holmberg(1975), for ex-
ample, used an apparatusfor which the subject had to turna
knob for raising or lowering a cylinder through a hole in a
plane to matchthe volume of a given object. This leaves the
question of whether adult consumers would make biased
volumejudgmentsof familiarcontainers.In Study 1, we test
the following:
Hi: Holdingactualvolumeconstant,moreelongatedcontainers
areperceivedto havehighervolumes.
Study 1: The Effect of ContainerShape on Perceived
Volume
Method. Subjects were 40 undergraduatebusiness students at the Hong Kong University of Science and
Technology, who completed the experimentaltask for partial course credit. Their average age was 21 years. Thirtyfive (87.5%) were female students,and five were male students. Thirty-seven(92.5%) were right-handed.
Twenty-sevencylindricalboxes, jars, and bottles of commonly used packages (e.g., beer cans, cheese balls, baby
food, and so on) were collected. Packageschosen were commonly purchased products at the university supermarket/
coffee shop (see Table 1). Packageschosen differedin shape
so thatthere was a wide variationin height (from 3.4 to 27.9
cm), maximum width (from 2.15 to 10.2 cm), and actual
volume (from 90 to 2330 ml). Each container was covered
with white paperto disguise its brandname and conceal all
volume information.3
3Note that disguising the boxes also may reduce subjects' use of experiential informationto estimate volume.
Table 1
OF STIMULI
USED INSTUDY1
DESCRIPTION
Description
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Gerberwith a blue lid
Chutneycontainer
Gerber"ThirdFoods"
Robertson'spreserve(340 gm)
Skippy PeanutButter(340 gm)
MarjoramFlakes (.4 oz.)
Spice Island's CinnamonSticks
DairyFarmYogurt(475 gm)
Meadow Lea Margarine(250 gm)
Meadow Lea Margarine(500 gm)
San Miguel Beer (330 ml)
CentrumVitamins
KraftCheez Whiz (250 gm)
Maya Chilli Chutney (237 ml)
Glass box with white lid
Ahmed's tandooripaste
PlantersPeanuts(340 gm)
PlantersCheez Balls (141 gm)
Nestle Coffee Mate
Calistoga MineralWater(296 ml)
Maritinelli'sapple juice (296 ml)
Calistoga juice (296 ml)
Knudsenpapayanectar(236 ml)
Oranginadrink(200 ml)
Coca-Cola (I litre)
PlantersCheez Balls (262 gm)
Pedigree Dog Food (700 gm)
Material
Glass
Glass
Glass
Glass
Plastic
Glass
Glass
Plastic
Plastic
Plastic
Tin
Plastic
Glass
Glass
Glass
Glass
Cardboard
Cardboard
Glass
Glass
Glass
Glass
Glass
Glass
Plastic
Tin
Tin
Height (cm)
7.7
6.2
8.5
10.4
8.0
8.7
10.2
10.5
3.4
6.9
10.9
9.7
10.1
13.0
13.0
11.0
8.7
17.1
16.5
16.5
9.7
17.8
14.0
13.6
27.9
19.5
13.8
Maximum
Width(cm)
Minimum
Width(cm)
Elongation
Volume(ml)
2.5
2.9
2.15
2.9
3.2
2.0
1.6
4.8
5.9
5.8
2.9
2.45
2.6
2.6
2.7
3.1
5.1
10.2
3.5
2.7
3.0
2.7
2.5
2.4
4.4
7.5
4.25
1.8
2.7
2.0
2.6
3.15
1.8
1.5
3.8
5.0
4.75
2.8
2.1
2.2
2.6
2.6
2.9
5.1
10.2
3.5
1.2
1.7
1.75
1.7
1.2
1.2
7.5
4.25
1.54
1.07
1.98
1.79
1.25
2.17
3.19
1.09
.29
.59
1.88
1.98
1.94
2.5
2.41
1.77
.85
.84
2.36
3.06
1.62
3.3
2.8
2.83
3.17
1.3
1.62
125
135
175
285
330
90
120
520
330
570
340
175
240
240
360
310
635
1250
650
305
315
305
245
200
1205
2330
730
Notes: Actual volume (last column) may differ from packagedescription(column 2). Measuresof both maximumand minimumwidth are given to account
for variationin shapes.
VolumePerception
317
Among the cylindrical containers found in the marketplace, there was a high correlation between surface area
viewed head-on (i.e., the shelf facing area, or height x maximum diameter)and volume (R = .90). Height also was correlatedhighly with actual volume (R = .60). However, elongation (height/maximum diameter) was correlated weakly
with actual volume (R = -.19).
Subjects were tested individually.They were told that the
experiment was concerned with how people made judgments under time pressure.The experimenterpresentedthe
different packages one at a time and asked the subject to
arrangethe packages in ascending orderof volume. The order of presentationwas randomizedand different for each
subject.To reduce noise in the estimates, subjectsthen were
shown a can of Diet Coke as a reference and told that its
volume was 355 ml. They were asked to estimate the volume (in ml) of each of the 27 containersin either ascending
or descending order,counterbalancedbetween subjects.
Results and discussion. To test HI, we estimated regression models, with estimated volume as the dependent variable and height, or a variablebased on height (e.g., elongation or surface area), as the predictor variable. Several
alternativemodels were run to identify the most parsimonious model and counterany alternativeexplanationsfor regression results (because containershapes had not been manipulatedsystematically to control for the presence of other
variables that potentially could affect volume estimates,
such as material, surface area, and so forth). The results of
six of these models are reportedin Table 2. Both height and
elongation (Models 1 and 2) have a significanteffect on perceived volume (,Ps= 12.55, 3.91; t = 7.35, 4.58; R2 = .703,
.694, respectively), even when actual volume is included in
the regression equation.4Models 3 and 4 show that height
affects perceived volume beyond surface area. Surface area
is significant if height is not included in the equation (Model 3: i3= 2.49, t = 6.37) but dropsto nonsignificancewith the
inclusion of height (Model 4: P = .88, t = 1.57).5 Surface
area and height have a high correlation(R = .69).
The containerswere of four differentmaterials:glass, tin,
plastic, and cardboard.Model 5 suggests that the material
also may affect volume perceptions;plastic containerswere
perceived as larger than glass containers (P = 37.18, t =
2.06). This is consistent with previous researchin cognitive
psychology that shows that the makeup of a shape (material, color) can affect the perceived size of the shape
(Gundlachand Macoubry 1931). Model 6 indicates that the
shape of the container significantly affects perceived volume beyond actual volume and height.
The results show that height in an absolute or relative
sense (versus elongation), both on its own and along with
the width dimension (e.g., surface area),affects volume perceptions substantially.Taller shapes are perceived as larger
than shorter ones. We investigate this elongation effect in
the remainderof this article. Specifically, in the next study,
we examine the robustness of the elongation effect under
4Derivingthe most appropriatepsychophysical model for estimated volume is not the purposeof this article. Thus, it is not our objective to determine whether height or elongation is a better predictorof estimated volume. Both supportHi.
5We thank reviewers for suggesting that we performregressions incorporatingsurface area, containermaterial,and containershape.
Table2
REGRESSIONMODELSFORSTUDY1
Actual volume
Height
Model 1
Model 2
Model 3
Model 4
Model 5
Model 6
.63
(35.0)
.72
(48.72)
.69
(46.56)
.64
(33.43)
.62
(24.73)
.64
(34.61)
9.77
(3.99)
13.22
(7.14)
9.44
(4.23)
12.55
(7.35)
39.1
(4.58)
Elongationa
Surface areab
2.49
(6.37)
-
Dummy for tinc
-
Dummy for plastic
-
.88
(1.57)
-7.65
(-.27)
-
37.18
(2.06)
Dummy for cardboard
33.22
(1.05)
Shaped
R2e
9.68
(2.14)
.703
.694
.699
.703
.705
.704
aHeight/(maximumdiameter/width).
bHeight(maximumdiameter/width- minimumdiameter/width).This model explained higher variancethan two others that used height(maximumdiameter/width)and height(minimumdiameter/width).
cThe base for the materialdummies is glass.
d(Maximumdiameter/width- minimumdiameter/width)2.
eAs comparingacross the R2s of the differentmodels demonstrates,the parsimoniousheight/elongationmodels (1 and 2) explain a large amountof the variation. Addition of parametersdoes not reduce their parametervalues or substantiallyincrease the model predictiveness.
Notes: Table 2 entries representparameterestimates (t-values) of the tests for parsimonyof parametersand alternativeexplanations for the elongation
heuristic.
318
manipulationsthat shed light on the theoreticalantecedents
of the use of height as a volume heuristic.
Study2: Does ContainerShape AffectPerceivedVolume
Automatically?
There are two possible reasons for the use of the height
heuristic. Consumers may use the heuristic either consciously to reduce the effort involved in making a complex
three-dimensionaljudgment, knowing that this may lead to
a less accuratejudgment, or in an automatic,unintentional
manner(cf. Bargh 1989). If the use of the height heuristicis
a controlleddecision, then increasingmotivationto make an
accuratejudgment should reduce the bias and increasing
cognitive load should increase the bias. Conversely, if it is
automatic, the bias should be robust and less likely to be
moderatedby ability or motivation manipulations,particularly if height forms the initial input for a volume judgment
(e.g., Gilbert 1989). On the basis of prior researchthat has
demonstratedthat the use of salient visual cues to make a
spatial judgment has an automatic aspect (Raghubir and
Krishna1996), we expect the elongation effect to be robust
to motivationand ability manipulations.
Method. Subjects were 20 undergraduatesat the University of Californiaat Berkeley,who participatedin the study
for partialcourse credit. The experimentaltask was to estimate the capacity of two five-ounce glasses, one a tall, fluted champagne glass and the other a round wine glass. All
subjectswere asked to judge the volume of the two glasses.
Cognitive load was manipulatedby asking subjectsto listen
to a tape and count the numberof times a word beginning
with the letter "t" was spoken. Low load subjects heard the
tape but were asked to disregardit. All subjects were given
a 90-second time limit to complete the task. To ensure that
they took the task seriously, all subjects also were given an
incentive for accuracy.There was a $50 rewardfor the subject who made the most accurateestimate dependingon the
task (volume estimate in the low load condition and estimation of words startingwith a "t"in the high load condition).
Subjects were asked to estimate the volumes of the two
glasses in fluid ounces (ozs.), which served as the dependent measure,and to rate the difficulty of the volume perception task on a seven-point semantic differentialscale ("Not
at All"/"VeryDifficult"), which served as the manipulation
check.
Results. The volume estimation task was rated easier by
those in the low load condition (Mean = 3.63) than in the
high load condition (Mean = 2.70; F(l,16) = 2.03, p < .10),6
which suggests that the manipulationworked as intended.
The analysis was a 2 (ability) x 2 (elongation)ANOVA,
with the first factor between subjects and the second within
subjects.The dependentmeasurewas volume perceptionin
fluid ozs. As predictedby HI, the taller glass was perceived
to be larger(Mean = 6.69) thanthe shortglass (Mean= 6.12;
F(l1,14)= 4.90, p < .05), irrespectiveof whethersubjectshad
paid more attentionto the volume estimationtask or the tape
(F < I for all effects involving ability). Thus, ability did not
emerge as a moderatorof the shape effect. The study shows
that the effect of elongation is robust;increasingprocessing
ability and motivation to make an accuratejudgment does
6Note that degrees of freedom are 16 because of partialnonresponseto
this question.
JOURNAL OF MARKETINGRESEARCH, AUGUST 1999
not reduce it. This result adds to literaturedocumentingthat
the use of visual cues to make spatialjudgmentsmay be partially automatic(Raghubirand Krishna1996).
Apart from neither demanding nor consuming cognitive
effort, automaticprocesses are characterizedas uncontrollable, unintentional, and outside conscious awareness
(Bargh 1989). If the use of elongation as a visual cue to
make volume judgments is automaticby these criteria, its
effect should follow when volume judgments incorporate
additionalsources of information,such as actual consumption. In the next section, we examine the implicationsof this
bias in perceived volume when it confronts actual experience, specifically, when the experience contradictsthe prior
expectation.To do this, we must considerthe implicationsof
container shape for consumption. We next introduce the
construct of perceived consumption and examine how the
elongation of a containeraffects it.
EFFECTSOF CONTAINERSHAPE ON PERCEIVED
CONSUMPTION
There is a rich literatureon expectancy disconfirmation
(for a review, see Stangorand McMillan 1992) and illusory
correlation (Chapmanand Chapman 1969) in social psychology and on learning from experience in marketing
(Hoch and Deighton 1989). Much of the expectancy disconfirmationliteraturefocuses on recall and recognition of informationthat is congruentversus incongruentwith expectations, leading to a contrastaway from initial expectations.
Stangor and McMillan performa meta-analysisof 54 such
experiments and show that incongruent informationis recalled betterthan congruentinformationwhen the information involves behaviorsratherthan traits.
Although the expectancy disconfirmationliteraturein social psychology focuses on traitsand behaviors of others as
inputs,some expectancydisconfirmationliteraturein cognitive psychology focuses on self-experiencedsensory inputs.
This latter literatureis more germane to our research, because we want to study perceived consumption,a self-experienced sensory input. A highly researched effect of this
genre is the SWI. Charpentier(1891) first demonstratedthe
SWI, in which bigger objects of the same weight were perceived to be lighter. For example, a pound of cotton wool
seemed lighter than a pound of lead. Since then, many others (e.g., Luczak and Ge 1989; Sarris and Heineken 1976;
Usnadze 1931; for a review, see Jones 1986) have replicated the effect.
Several explanations have been proposed for the SWI.
The most accepted are based on expectancy theory and divided broadlyinto two streams.One suggests thatthe bias is
haptic in nature(i.e., due to touch); the other suggests it is
visual in nature. In the haptic stream of research, Woodworth (1921) proposes that prior experience with objects
leads observersto expect that a large object will be heavier
than a smaller object. This sets up expectations that could
affect the force an observerapplies when lifting an object.A
greater lifting force applied to a larger object causes the
larger object to be judged lighter (see also Davis and
Roberts 1976; Ellis and Lederman 1993; Nakatani 1985;
Pick and Pick 1967; Ross 1969).
In the visual stream,on the basis of his informationintegrationmodel, Anderson(1970) arguesthatsize, or volume,
is an object propertythat affects perceived heaviness along
VolumePerception
with weight. Size affects perceived heaviness throughan expectation of how heavy an object of that size should be.
Specifically, Anderson proposes the Averaging Model, in
which the judged heaviness of the object as seen and lifted
is a function of the weighted average of felt (but not seen)
weight and expected (seen but not felt) weight (see also
Cross and Rotkin 1975). Masin and Crestoni's (1988) results also support the visual hypothesis. They control for
both haptic information, by having the object lifted by
pulling on a string going througha pulley, and volume expectations, by showing the stimulus but manipulatingthe
time at which it was seen. They find that the SWI occurred
when vision was allowed (i.e., the volume expectancycould
be developed) but only when the object was viewed at the
same time it was lifted. If viewing was prior to lifting, the
SWI disappeared. In summary, though there is still some
controversy regarding the basis of the SWI, the prevalent
view appearsto be that people expect the smaller object to
be lighter.However, when they actually lift the small object,
their experience contradictstheir expectation, leading to a
contrasteffect. The opposite is true of the large object. This
results in the smaller object being perceived as heavierthan
the largerobject. There is also evidence to show that this illusion may be partiallyautomatic,in as much as it is uncontrollable, because it does not reduce when subjects are told
that the objects weigh the same (Floumoy 1894).
We propose that perceived consumption is analogous to
perceived heaviness, in as much as it is a function of (1) an
initial perceptualvisual input (volume) and (2) a subsequent
experiential sensory input (actual consumption).We expect
that a similar perception-experience illusion occurs in the
consumption scenario: When subjects see a tall glass, they
perceive it to be largerthana short glass, but when they start
drinking,theirexperience contradictstheirexpectations.Because it is less voluminous than they expected, they believe
that they have consumed more from the containerthey expected would contain less, the shortercontainer.This effect,
the PCI, is stated formally as follows:
is greaterwhenthecontaineris less
H2:Perceivedconsumption
(versusmore)elongated,holdingactualvolumeconsumed
constant.
Note that the PCI requiresthat consumershave expectancies based on elongation of the container.This suggests that
volume perceptions should mediate the effect of elongation
on perceived consumption:
(H2)is
H3:The effect of elongationon perceivedconsumption
mediatedby its effecton volumeperceptions
(H1).
Study 3: Perceived Volume-ConsumptionSwitch
In this study, we test for the PCI, a reversalin perceptions
of volume pre- versus postconsumption. We examine
whether the same person finds the taller glass to contain
more (Hi) and believes he or she has drunkmore from the
shallower glass (H2) and whetherthe effect of elongation on
perceived consumption is explainable by the effect of elongation on perceived volume (H3).
Method. Subjects were students in a graduatemarketing
class at Columbia University.The experimenthas two phases: In week 1, HI was tested, and in week 2, H2 was tested.
There were 18 common data points across the two phases of
319
data collection, which were used to test the perceptionreversal within subjects and examine the mediation hypothesis (H3.)
In week 1, subjects were informedthat the study was being conducted by an established plastic cup manufacturing
company thatwas interestedin getting theiropinions on two
new designs of cups it was thinkingof introducing.Subjects
were given two cups of identical volume (10 fluid ozs.) but
different shapes, one more elongated than the other. The
more elongated cup had an approximateheight of 8.6 cm
and a base diameterof 4.8 cm. The less elongated cup had a
height of 7.8 cm and a base diameter of 5.0 cm. Subjects
were asked to estimate the capacities of both cups (in fluid
ozs.), which served as the dependentmeasureto test Hi. For
a benchmark,they were told that a can of soda has 12 fluid
ozs. The order in which volume was estimated for the two
cups was counterbalanced.Subjects also were asked questions regardingthe aesthetics of the two cups to maintainthe
cover story.Consistentwith the cover story,they were asked
to choose which cup they found more appealing. The two
cups were markedwith letters selected at random.
In week 2, the same cups were filled to half their capacity with water.7Subjects were told the following:
We arean establishedplasticcup manufacturing
companythatwantsto diversifyintomineralandspringwaters.We havejust comeup withtwo new formulations
of springwaterthatwe wouldlike you to try.To get a
good idea of boththe products,we wouldlike you to
finishthe entirespringwaterin bothglassesover the
classsession.Also, we wouldlikeyouto startwithone
glass,finishit completelyandthendrinkfromtheotherglass.PleasestartwithglassL (orP).Turnto thenext
pagewhenyou havefinishedbothglasses.
Again, both cups were markedwith lettersselected at random. Orderof tasting was counterbalancedacross subjects.
Subjects were asked which taste they preferred and how
much they believed they had consumed from each cup. To
make the cover story more believable, they also were asked
how often they bought spring water, which brand they
bought, and when they consumed it (e.g., after sports, with
meals).
Results. A 2 (measures: perceived volume versus perceived consumption)x 2 (shape: tall versus short)ANOVA
showed that, though the two main effects were significant
(F(1,17) = 50.42 and 11.72 for shape and measure, respectively, ps < .01), their interaction was also significant
(F(l,17) = 16.25, p < .001). Although the perceived volume
of the tallercup was greater(Mean = 9.050) than that of the
shallower cup (Mean = 7.750; F(l,17) = 23.15, p < .01, T2 =
.577), as predicted by Hi, the opposite was true for perceived consumption.In supportof H2,consumptionwas perceived as lower for the taller cup (Mean = 4.722) versus the
shallower cup (Mean = 5.056; shape contrastF(1,17) = 2.27,
p < .10, qT2= .118). Thus, both HI and H2 were supported.
To test H3, we conducted a mediation analysis to determine whether the effect of cup shape on perceived consumption was mediated by perceived volume estimates.
This is by far the strongesttest of the hypothesis thatpeople
expect the shortercup to contain less liquid and thatthis ex7Theywerenotfilledto capacityto minimizeconsistencypressuresbe-
tween the two phases of data collection.
JOURNAL OF MARKETINGRESEARCH, AUGUST 1999
320
pectation is what drives their perceptionthat they have consumed more from it. As per Baronand Kenny (1986), to establish mediation, we must demonstratethat (1) the independent variable (shape of cup) affects the dependent
variable (perceived consumption). This test is significant
(H2 results); (2) the independentvariable(shape of cup) affects the proposed mediating variable (perceived volume).
This test is also significant(Hi results);and (3) the effect of
the independent variable (cup) on the dependent variable
(perceived consumption) reduces to nonsignificance (perfect mediation)or in effect size (partialmediation)when the
analysis incorporatesthe mediatingvariable(perceived volume) as a covariate, whereas the effect of the mediating
variable is significant. This demonstratesthat the variance
that was explained previously by the independentvariable
now can be explained by the mediatingvariable,which implies thatthe independentvariableexerts its effect on the dependent variable indirectlythroughthe mediating variable.
This analysis is reportednext.
An ANCOVAon perceivedconsumptionof the two cups,
including perceived volumes of the two cups as covariates,
shows that the effect of shape of cup, which was significant
without the covariate, reducedto nonsignificance when the
covariate was added(F(1,14) = .02, p > .90, T12= .001). The
effect of the covariate was marginally significant, despite
the small sample size (F(1,16) = 2.27, p < .10, 1q2= .124),
and the beta coefficient was in the expected direction([5 =
-.35, t= 1.51).
This result supportsthe propositionthat the switch in relative volumes from the tallerto the shallowercup in the preversus postconsumptiontask is due to a contrast between
what is expected and what is experienced.8In the next section, we explore the implications of this on actual (as opposed to perceived) consumption.
EFFECT OF PERCEIVEDVOLUMEON ACTUAL
CONSUMPTION
Prior researchhas proposed(Assuncao and Meyer 1993;
Blattberg et al. 1978) and demonstrated (Wansink and
Deshpande 1994; Wardand Davis 1978) thatstockpilinghas
a direct effect on consumption.The reasons offered are that
lower unit cost throughstockpilingon a deal stimulatesconsumption (Assuncao and Meyer 1993) and that consumers
want to bring their inventory down to an acceptable level
(Blattberg et al. 1978). This research suggests that consumption is relatedpositively to inventorylevels. This effect
may be due to higher actual volume, higher perceived volume, or both.
8Because the PCI is a novel finding, we replicatedit with some methodological variations to test for robustness.The experiment was conducted
with 37 undergraduatestudentsat a Hong Kong business school. Two cups
of identical volume (8 ozsJ240 ml) but different heights (8.5 versus 5.75
cm) were chosen. Their base diameterswere within a quarter-inchof each
other (elongated = 5.25 cm, shallow = 5.5 cm).
Under the guise of a taste test, both cups were filled with 7-Up or Sprite
(actual volume = approximately222 ml), and subjects were asked to drink
one of the cups completely before drinkingfromthe other.Orderof cup and
brandof soda tasted first was counterbalancedbetween subjects.After disposing of the two cups, subjectsestimatedhow much they had drunkfrom
each. This question was embeddedamong otherquestions to make the cover story realistic.
Results support H2 and replicate Study 3. Subjects estimated that they
had consumed a smaller volume from the tall cup (163.51 ml) versus the
short cup (175.68 ml; F(1,36) = 4.12, p < .05).
Studies holding actual volume constanthave demonstrated thatpackage size positively affects consumption(Folkes,
Martin,and Gupta 1993; Wansink1996; but for null results,
see Moore and Winer 1978). Specifically, Folkes, Martin,
and Gupta(1993) propose that large packages lead to higher consumption because consumers are less worried about
replacementtransactioncosts. Wansink(1996) suggests that
the same effect may be observed if consumers believe that
largerpackages have lower unit costs, and so, even holding
actual volume constant, larger package sizes may lead to
greaterconsumption.Otherreasons suggested for the effect
of stock volume on usage volume are that largerpackages
are more difficult to control and thus lead to overpouring
(Stewart 1994) or that consumers are eager to finish larger
package sizes because of inventory holding costs (Hendon
1986).
These studies have examined the effect of actual differences in volume or differencesin package size on usage. We
now suggest that, even holding actual volume and package
size constant,to the extent that consumersbelieve different
shaped containers have different volumes, container shape
can affect consumptionlevel. Thus, we hypothesize that the
positive effect of actualvolume on usage will translateto the
domain of perceived volume. Based on Studies 1-3, consumers would consume more from more elongated containers, which are perceived to be larger-a direct effect of perceived volume on actual consumption.
An alternativeroute for the same effect is by perceived
consumption. One of the implications of the PCI is that
when subjects see a tall glass, they perceive it to be larger
(Hi), but when they startdrinkingfrom the glass, they realize that it is not as big as they thought.They then overcompensate, as is reflected in lower estimates of perceived
consumption (H2 and H3). This also may result in overcompensation in their actual consumption from the glass.
That is, the overcompensationmight lead to their drinking
more from the elongated glass than from the less elongated
glass. Such a mechanismpoints to an indirecteffect of perceived volume on actual consumption through perceived
consumption.It implies thatconsumptionwill be greaterthe
more elongated the container is. Thus, we expect the
following:
H4:The moreelongatedthe container,the greateris the actual
consumption.
(H4)is meH5:Theeffectof elongationon actualconsumption
diated by perceived consumption(H2).
Study4: Do People Drink Morefrom TallerGlasses?
Method.Subjectswere 16 graduatestudentswho engaged
in the experimentas part of a class at Columbia University.
The study is similar to a typical Piaget mass conservation
experiment (Piaget, Inhelder, and Szeminska 1960). We
used a one-way between-subjectsdesign, with shape of container manipulatedat two levels (shallow glass versus tall,
deep glass). The glasses were the same as those used in
Study 3, that is, of identical capacity and containing the
same volume. At the beginning of class, subjects were told
thatan establishedsoft drinkcompany that had not yet marketed cola wanted their opinions on two formulationsit had
developed. It was giving them a glassful of each so they
could get a true feeling of the formulation. Subjects were
VolumePerception
told to drinkas much or as little as they wantedof either formulationand that they could switch back and forthbetween
formulationsto determinewhich was better.Which glass the
subject drank from first was counterbalancedacross subjects. At the end of the class, subjects were asked which of
the two formulationsthey liked more, which glass was more
appealing to them, how much soda they drank in a week,
and their gender. The amount left in the glasses was measured after the completion of the experiment and used to
compute actual consumption.
Results and discussion. H4 was supported.The analysis
was a repeated measures MANOVA, with consumption as
the dependent variable, shape as the within-subjectsindependentvariable,and the two counterbalancingordersas between-subjectsindependentvariables(which glass's capacity they estimated first and which glass they drank from
first). The MANOVArevealed a significanteffect for shape;
the more (versus less) elongated glass had greaterconsumption (Means = 6.91 versus 6.20 ml; F(1,12) = 23.07, p <
.0001). Neither order of administrationfactor exerted main
or interaction effects. Elongation did not affect any other
measures. Thus, actual consumption is greater from more
elongated glasses. The next two studies examine the route
by which elongation affects actual consumption.
Study 5: Delineating Mediation Paths
Method.Thirty-threeundergraduates,
drawnfromthe same
pool as used in Study 2, participatedfor partialcourse credit.
The glasses from Study 2 were used as stimuli(tall flutedand
roundwine 5-oz. glasses). The studyhadtwo parts.In the first
part, subjects estimated the volumes of the two glasses.
Subjectsthen performedan unrelatedtask for approximately
30 minutes.The second task was assigned between subjects.
Subjects had to choose one of two locations in the experimental room.These locations had been set up with eitherthe
tall or the shortglass. Subjectshad to consume threetypes of
snack foods-one pretzel thin, two corn chips, and threepotato chips, in that order-under the guise of a taste test for
snacks. They were asked to drinkenough waterbetween the
tastings "to remove the taste of the snacks." One experimenterwalked aroundwith a bottleof waterto refill glasses,
takingcare to only refill glasses thatwere completelyempty.
The otherexperimenterunobtrusivelyrecordedconsumption
(by quarterglass). After filling in questionsto keep the cover
story intact(e.g., which snack was saltiest,which made them
thirstiest,which was the easiest taste to remove, and which
taste they liked best), all subjectswere asked to estimate the
amountof waterthey had drunk(in fluid ozs.) to remove the
tasteof each snack.This measurewas used for perceivedconsumption.The elaboratesnack food guise was requiredto reduce suspicionthatthe two partsof the studywere related,reduce demand artifacts, and control for order effects due to
priormeasurementof volume perceptions.Therewas a suspicion check at the end of the study to identify anyone who
guessed thatthe two partsof the study were connected.At the
end of the study,subjectswere askedto help themselvesto the
remainingsnack food and were excused.
Results and discussion. Six subjects indicated suspicion
of a connection between the two partsof the study and were
removed from the sample. Anotherdid not complete all the
measures, leaving a usable sample of 26 subjects to test the
hypotheses.
321
As predicted by HI, the shorter glass was estimated to
contain 5.6 ozs., as compared with the taller glass, which
was estimated to contain 6.0 fluid ozs. (F(1,22) = 4.67, p <
.05).
H2 made a directionalpredictionfor perceived consumption, holding actual consumptionconstant. In this study, as
actual consumption varied, the directional prediction was
inappropriate.Thus, we computeda measurefor which a directionalpredictionis possible:
PerceivedConsumption
Error= (PerceivedConsumption
- ActualConsumption)/Actual
Consumption.
Per H2, the perceived consumption versus actual consumption should be greater for the short glass compared
with the tall glass. The means are in the correct direction
(Error= .48 versus .34 for the short and tall glass, respectively; F(1,24) = 2.39, p < .10, one-tailed).
Subjects in the tall glass condition consumed more water
on average than those in the short glass condition (Means =
8.4 versus 6.9 fl ozs.; F(1,24) = 1.96, p < .10, one-tailed), in
supportof H4.
To examine whethervolume perceptionsdirectly mediated the effect of containershape on actual consumptionor if
this effect is through perceived consumption, as was hypothesized (Hs), we conducted two separateANCOVAs. In
the first, the difference in perceived volume between the tall
and shortglass, as elicited in the first partof the experiment,
was included as a covariate in the ANOVA on actual consumption. The effect of elongation marginally increased
(F(1,23) = 2.15), and the covariate was not significant (F <
1), which is not consistent with a mediation pattern(Baron
and Kenny 1986). The second ANCOVAused the perceived
consumption error as the covariate. This analysis shows a
strong mediation pattern:The effect of the covariate is significant (F(1,19) = 12.72, p < .01) with the sign in the correct direction (P = -.62), whereas the effect of elongation
drops to F < 1, the conventional level of a null effect. This
patternsupportsH5 and indicates that the route to increased
actual consumption is mediated by the effect of elongation
on perceived consumption.
The results of this study replicatethe elongationeffect on
volume perceptions and actual consumption; elongated
glasses are perceived to contain more priorto consumption,
and actual consumptionis greaterfrom these glasses. These
results suggest that managers should construct more elongated containers so that consumers believe they are bigger
and consume them faster. This, however, is contingent on
consumerspreferringa containerthatis perceivedto be larger. We now test for this effect.
CONSEQUENCESOF VOLUMEPERCEPTION
We first reanalyze some of the data collected as part of
Study 5 to test whether consumer perceptions of volume
translateto choice. Next, we reportresults of two additional
studies thatexamine whetherthe elongationeffects translate
to preferenceand postconsumptionsatisfaction.The formal
hypothesis tested is as follows:
H6:Themoreelongatedthecontainer,themoreit is preferred.
322
JOURNAL OF MARKETINGRESEARCH, AUGUST 1999
Study5: An Extensionto Choice
Method. In Study 5, all subjects (n = 33) had to choose
one of two locations in the experimentalroom to complete
the second part of the experiment.This part of the experimental procedurewas run between subjects.There were an
equal numberof locations with a tall or a short glass. Thus,
subjectswere constrainedby the availabilityof a glass when
choosing their seat, and a person who wanted a tall or short
glass may not have gottenone. The orderin which the glasses were picked shows which glasses were chosen more readily by subjects.9
Dependent variables. We measure the difference in preferred choice between the tall and short glasses in the following three ways:
1. The difference in the mean rank, or the mean order in which
the glasses are chosen. If there are four glasses and the choice
orderis tall-short-tall-short,
themeanranksaretwo (i.e., [1 +
3]/2)fortallandthree(i.e., [2 + 4]/2) forshort.Thelowerthe
mean rank,the higher is the preference;
2. The mean availabilityacross all choice occasions. In a situation of choice without replacement,if a glass is preferred,it
will be chosen more readilyand have lower availability;thus,
thehighertheavailability
forthetallor shortglass,thelower
the preference
forit is; and
3. Themeanavailability
conditional
on choice.A tall(short)glass
orin spiteof its
maybe chosenbecauseit hashighavailability
low availability.
of tall(short)glassesacrossoccaAvailability
sionson whichtheywerechosenreflectswhetherglasseswere
or in spiteof low availabilchosenbecauseof highavailability
ity. The higherthe meanavailabilityconditionalon choice, the
lower the preferenceis. Thus, if there are four glasses and the
the availabilityfor the tall
choice orderis tall-short-talk-short,
glass across the four choice occasions is 2/4, 1/3, 1/2, and 0/1,
whichyieldsa meanavailability
of .3333acrossall fourocca-
sions anda meanavailabilityof .50 acrossthe two occasions on
for
whichthe tallglasswas chosen.Similarly,the availability
theshortglasson thefourchoiceoccasionsis 2/4, 2/3, 1/2,and
of .6667acrossall four
1/1,whichresultsin a meanavailability
choiceoccasions(or one meanavailabilityof the tall glass)
with a meanavailabilityof .83 acrossthe two occasions, condi-
tionalon theshortglassbeingchosen.
Results. Study 5 was conducted in two sessions with 13
and 20 students, respectively.10The mean rank for the tall
glass is lower in both session 1 (Means = 3.67 versus 9.86
for tall and short glasses, respectively; p < .001 using the
Wilcoxin-signedrankstestll) and session 2 (Means = 9.2 for
tall and 11.8 for short;p < .001). The mean ranksshow that,
on average, the tall glass was chosen before the short glass.
In session I (n = 13), the mean availability across all
choice occasions was .1563 for the tall glass versus .8437 for
the shortglass. The patternis the same in session 2, in which
the mean availabilitiesare .3688 versus .6362 for the tall and
short,respectively (n = 20). Although these two proportions
are not significantlydifferentfrom .5 because of their small
sample sizes, the directionof the differencesupportsH6.
If we observe only the mean availability conditional on
choice, in session 1, the mean availabilities are .32 for the
9An experimentwith replacementof glasses would have given a moredirect measureof choice for tall versus shortglasses. However, data without
replacementof glasses were requiredto test the consumptionhypotheses.
("Thefirst session had six tall and seven shortglasses.
1We treateda tall and a shortglass as one pair and rank-orderedthe differences in ranksfor the Wilcoxin test.
tall glass when the tall glass was chosen (n = 6) versus .96
for the short glass when the short glass was chosen (n = 7;
proportionsdifferentfrom each other at p < .05). In session
2, the mean availabilities are .44 versus .71 for the tall and
short, respectively (n = 10 each; proportionsdifferent from
each other at p < .05).
Thus, H6 is supported.These analyses demonstratethat
the tall glass was preferredto the short glass. We may conjecture that this occurred because the subjects preferred
glasses that appearedlarger.12
Study 6: An Extensionto Preference
Subjects (n = 53), drawnfrom the same pool as Studies 2
and 5, were shown the two glasses used in Studies 2 and 5
and asked which glass they preferred.An overwhelmingmajority (77%, or 41 of 53 subjects) preferredthe tall glass to
the short one (p < .01). Ten preferredthe short glass, and 3
subjectswere indifferentbetween the two. Thus, the elongation effect translatesfrom thejudgmentto the preferencedomain. Note that the elongation effect should extend to preference only in conditions in which more is preferredto less.
In situationsin which consumers wish to minimize volume
for reasonsof storagecapacity (e.g., yogurt)or consumption
(e.g., candy bars), the managerialimplication of offering a
larger-lookingcontainersize is unclear.
Study 7: An Extensionto PostconsumptionSatisfaction
Postconsumption satisfaction should be based on the
amount people believe they have consumed. Thus, the effect of elongation on postconsumptionsatisfaction should
be in the same direction as its effect on perceived consumption ratherthan on actual consumption.Thus, we predict the following:
H7:The more elongatedthe container,the lower is postconsumptionsatisfaction,given thatactualconsumptionis the
same.
Method.Subjects(n = 40) were drawnfrom the same pool
as Studies 2, 5, and 6 (no subject participatedin more than
one of these studies), and the stimuli used was the sameround wine and tall fluted glasses. The experimenthad two
parts:volume perceptionand consumption,with their order
manipulatedbetween subjects. That is, half completed the
volume perception measures prior to consumption, as in
Study 5, whereas the other half completed these measures
postconsumption.This was to rule out the possibility that
our previous results were due to orderof measurement.The
volume perceptionpartof the experimentwas akin to Study
5, in which subjectsestimatedvolume for both the shortand
the tall glass.
The cover story for the second stage of the experiment
was that it was a taste test for juice. Subjects were given a
(tall or short) glass filled with 4.5 ozs. of a fruitjuice drink
and asked to drinkit as they would normally.After answering an open-endedquestion about its taste, to increasecredibility of the cover story,subjectsratedthe juice using a seven-point scale in terms of how satisfying it was ("Not at
All'/"Very").Because satisfactionis a function of both perceived volume and taste, we also collected closed-ended
12Notethat height is not the only feature that is different between the
taller flutes and the shorter,round wine glasses. Their shape also differs,
and this could potentiallyinfluence subjects'choices.
Volume Perception
measures of taste, including how fruity and refreshing the
juice was. Besides the perceptualmeasure for satisfaction,
we determined satisfaction with a more behavior-based
measure.We asked subjects whetherthey wanted a refill of
the glass ("Definitely Not"/"DefinitelyYes"). To the extent
that subjectsdid not find the drinksatisfying, they should be
more likely to want a refill.
Results. There were no significant ordereffects on any of
the measures, and accordingly,orderis ignored.
In supportof HI, regardingvolume perceptionand replicating previous studies, the taller glass was perceived to
contain more (Mean = 6.91) than the shorterglass (Mean =
6.46; F(1,38) = 4.03, p < .05).
Regarding postconsumptionsatisfaction, an ANOVA on
the satisfaction measure, using type of glass as the independent variableand includingfruitinessand refreshingratings as covariates, showed thatthejuice was perceivedto be
more satisfying when sipped from the short glass compared
with the tall one (Means = 5.85 versus 5.55; F(1,36) = 2.96,
p < .05, one-tailed). Both covariateswere significant.Thus,
the shape of the glass affected postconsumptionsatisfaction.
Consistent with the PCI, this can be explained by consumers' perceived consumption being greater from the
shorterglass.
An ANOVA on the refill requests,though not significant,
provides furthersupportto this finding.Those drinkingfrom
a tall glass wanted a refill to a greater extent than those
drinking from the short glass (Means = 4.90 versus 4.35;
F(1,37) = 1.48, p < .12, one-tailed), and refill requests were
related to how satisfying they found the juice (F(1,37) =
2.82, p <. 10).
In summary, in this section, we investigated the consequences of the effects of shape on volume perceptionsbefore and after consumption. We found that more elongated
containers are preferredand are more likely to be chosen
preconsumptionbut are believed to be less satisfying than
less elongated containers of the same volume. We now discuss the implications of our results for theory and practice.
GENERALDISCUSSION
In this article, we examine the effect of elongation on (1)
perceived volume, (2) perceived consumption, (3) actual
consumption, (4) postconsumption satisfaction, and (5)
choice. As described in Figure 1, our model suggests that
package shape directly affects perceived volume and,
throughthis, indirectly and inversely affects perceived consumption. Perceived consumption,though occurringsubsequent to actual consumption,affects the amountconsumed;
the less people think they are drinking,the more they drink
to compensate. Thus, the net effect of elongation on actual
consumption is positive and by way of the perceived consumption route. Perceived consumption, in turn, affects
postconsumption satisfaction directly, which implies that
the net effect of elongation on satisfaction is negative.
Finally, the positive effect of elongation on volume perception translatesto preferenceand choice.
Specifically, based on the literaturein cognitive psychology, we propose that consumers use the simplifying heuristic of a container's elongation to estimate its volume. An
empirical test shows that, even for frequentlyused and purchased package shapes, this is true.The more elongated the
container,the greaterthe perceived volume of the container
323
is (Study 1). This robust effect may be explainable using
Study 2 results, which are consistent with the proposition
that the use of elongation to judge volume may be partially
automatic.We then proposedthe constructof perceivedconsumption and related it to the robust SWI from cognitive
psychology. We proposed that the SWI is a specific case of
a more general perceptual-experienceillusion and that another manifestationof that illusion is the PCI. The PCI proposes that perceived consumptionis relatedinversely to the
perceived volume of a product. Study 3 demonstratesthis
effect and providesevidence thatthe contrasteffect is mediated by an expectancy disconfirmation.This is a novel finding in both the cognitive psychology and marketingliterature. Next, we examined the implicationof containershape
on actualconsumption.In Studies 4 and 5, we found thatthe
effect of container shape on actual consumption mirrored
the patternof perceived volume (i.e., the more elongated the
container,the greater the consumption from that container
is) and thatthis was due to the effect elongation had on perceived consumption.Studies 5-7 showed that the perceived
volume effect translatedto preference,choice, and postconsumptionsatisfaction.
Ourresearchbuilds on previouswork in both psychology
and marketing and examines the antecedents and consequences of biased volume perceptions.We find that an object's elongation affects consumerjudgments and behavior
in simple though not always intuitive ways.
TheoreticalImplications
The PCI and the manner in which it affects actual consumptionis of theoreticalinterestto informationprocessing
researchers.At a general level, this is a paradigmin which
there is a judgment based on two sources of information
(volume and consumption) that are inconsistent with each
other.The sequential natureof the two informationsources
places the former as a reference against which the latter is
processed. This results in a contrasteffect due to the inconsistency between the two informationsources-a tallercontainer appears larger in volume versus a shortercontainer,
but when consumed from, it does not appearto contain as
much volume as expected. Although contrast effects have
been demonstrated using attitudinal data, this research
shows these effects with less ambiguous sensory inputs. To
our knowledge, this is the first time an effect of this type has
been documentedin marketing.
Our research also introduces volume perceptions as an
area of study. Whereas prior research has focused on the
consumptioneffects of higher actual volume (Wansinkand
Deshpande 1994; Wansinkand Ray 1992, 1996), we focus
on the effects of perceived volume. Because many consumers do not read the labels on packages that declare true
volume (Dickson and Sawyer 1986), the effects of perceived
volume seem importantto study.Furthermore,we introduce
the concept of perceived consumptionto marketingand psychology research. Although smell and taste always have
been accepted as perceived stimuli, consumption as a perceived stimulus has not been investigated.
The set of seven studies shows that consumersare prone
to biases in volume judgment. Consumers may use the
height of the package to anchortheir volume estimates and
then adjust their volume perceptions subsequently to account for width and shape differences.A similar anchorand
324
adjustprocess has been suggested to apply to distance perception (Raghubirand Krishna 1996) and numerosityestimates (Krishnaand Raghubir1997). Similarto Gilbert,Pelham, and Krull (1988), we also propose that the initial
anchor may be an automaticinput with subsequentadjustment a more controlled process (Raghubir and Krishna
1996).
Managerial Implications
Our studies demonstratethatcertainshapes are perceived
to be bigger in volume thanothersof identicalvolume. This
is demonstratedwith the most common shapes used for food
items (e.g., cans and glasses) and frequently purchased
package shapes (e.g., soda cans). In addition, it is demonstratedusing stimuli that are actual package shapes picked
off groceryshelves. Therefore,if people do not readthe volume informationon the package,packages that appearlarger will be more likely to be purchased,ceteris paribus.Thus,
manufacturersmay find it beneficial to sell productsin certain types of package shapes. We also find that perceptions
of larger volumes are associated with larger consumption
(Studies 4 and 5). Therefore,more elongated packages may
present a double-win situation for managers;not only are
they more likely to be purchased,but afterbeing purchased,
they also will be consumed at a faster rate. However, they
may lead to lower postconsumptionsatisfaction,thoughthis
may induce greaterconsumption(Study 7).
For purchase decisions, it has been assumed that consumerssubconsciouslyfocus on purchasingthe packagethat
has lowest unit cost (Isakson and Maurizi 1973). If they do
not read package labels and unit price information,consumers may choose the package that appears largest if the
set of brandshas similar (or equal) prices. This implies that
perceptionsof package volume become important.It has not
been tested if consumers (1) would focus on which package
looks bigger at the time of purchaseor (2) would remember
their feelings of postconsumption(dis)satisfactionfrom the
previous purchase.For "one-shot"purchases,or when consumers have not tried different package shapes, packages
that appear larger would be purchased because postconsumptionsatisfactionwould have no role in the purchasedecision. We also propose that, even if consumers have tried
different packages, they use visual perception ratherthan
their postconsumption satisfaction, because temporal distance makes the formermore salient. In addition,with larger interpurchase cycles, postconsumption dissatisfaction
may be difficult to recall. We have shown (in Study 1) that
consumers perceive more elongated packages to be larger,
even when they are frequently purchased packages. This
would indicate that disconfirmationof package size after
consumptionmay not lead consumersto revise their volume
judgmentssufficiently in the long term.With shortinterpurchase/interconsumptioncycles, postconsumption dissatisfaction may be easy to recall and thereforesalient. However, this suggestion needs furtherresearch.
Ourfindings are importantfor managersbecause package
shape decisions can have a major impact on a company's
sales. For example, in 1984-85, Lipton India elongated the
shape of Lipton tea packages. Unit sales increasedby more
than 10%.13Ourstudies also have implicationsfor shapes of
13Thisinformationwas obtainedthroughpersonal communicationwith
a brandmanager.
JOURNAL OF MARKETINGRESEARCH, AUGUST 1999
differentpackage sizes of the same brand.If a manufacturer
has differentpackage sizes, should it maintainthe same ratio along dimensions, and thereforemaintainthe shape (i.e.,
a cube remains a cube, just a larger one), or should it keep
the width the same and increase the package size along the
height dimension?Ourfindings suggest thatthe latterwould
result in the larger size being perceived as even larger.Another argument is that the manufacturershould have the
most elongatedpackageshape for the packagesize it is keen
on selling overall. This should result in consumersthinking
of this package as a better value for money and, generally,
in largersales.
The package shape decision also has implications for
pricing and communication. The Hong Kong-based distilled bottled water company,Watson's,decided to increase
the volume of its small size bottled water from 700 ml to
800 ml. Competitors sell a 770 ml bottle. Can Watson's
chargea higher price for the new bottle? Should it maintain
the symmetrybetween height and width or elongate the bottle? If the height of the bottle remains unchanged(to make
storage easier), does Watson'sneed to bring the higher volume of the bottle to consumers'attentionor will consumers
easily spot this change? Watson'smust consider if it should
advertise that its new bottle is larger than its old one and
largerthan its competitors'bottles. Our results indicate that
Watson's should market its new bottled water in a more
elongated bottle, particularlyif it wishes to charge a higher
price. If Watson'sdoes not elongate the bottle, it definitely
should advertisethe size change, as consumersmay underestimate the increasedvolume.
Downsizing decisions (reducing volume while keeping
price the same) also must account for volume perceptions.
Several well-known brandsare downsizing ratherthan cutting price, and less than a quarterof them reporta negative
impact (Adams, di Benedetto, and Chandran1991). Adams,
di Benedetto, and Chandran (1991) report that when
Charmindownsized its roll from 500 to 380 sheets in early
1987, and then to 350 sheets, the package price remained
unchanged,the paper sheets were fluffed up to reduce the
visible effect of downsizing, and packagingcommunication
focused on "fluffiness."There was no negative impact of
downsizing on Charmin.It appearsthat consumers did not
realize that the effective price increase was 8%. This example shows that the issue of volume perceptionis important
for reasons of consumer welfare. If companies charge a
higher price for their product because they expect consumersto estimatethem to be bigger thancompetingbrands,
consumers must be made aware of their own biases in volume perception.
StudyLimitationsand Areasfor FurtherResearch
Although we showed that elongation increases perceived
volume, it is importantto determineboundaryconditions for
this effect. For example, does the effect only hold when
height is salient, and not when width is more salient than
height? How well does the model make predictionsfor consumerproductsacross form categories;that is, how do boxes compare with bottles? Are there limits to the elongation
effect at a particularelongation ratio; that is, what happens
to perceptionsof the volume of test tubes?What happensto
volume and consumptionbiases as packages become more
familiar?Do they get reduced,or do they persisteven in sit-
VolumePerception
uations of high familiarity?What happens if new package
shapes are introduced that are counter to expectations for
the category set by competitors? Does this exacerbate the
bias as consumers' volume estimation task becomes more
difficult?
In our experiments on consumption, the liquid was
pouredinto the containersfor the subjects.Anotherinteresting issue for furtherresearch is whether consumers would
pour different amounts of liquid into containersthat appear
to be of different volumes.
Anotherdirection for additionalresearchis to study biases in area perceptions of nearly two-dimensional products,
such as pizzas, cookies, and so forth. What are the implications of the manner in which volume judgments are made
for areajudgments?Is the areajudgmenttask easier thanthe
volume judgment task, thereby resulting in reduced use of
heuristics and less biased size estimates? Furthermore,a
more rigorousspecification of the mannerin which different
dimensions are aggregatedwould lead to greaterspecificity
in predictions. How do consumers integratethe height and
width dimensions to make a size judgment?Is it throughthe
use of the simple elongation parameter,such as height in
terms of width, or is it throughan additive or otherprocess?
Furthermore,though in this article we did not disentangle
the constructs of elongation and height (when height was
manipulated,elongation was manipulated,and vice versa), a
more rigorous model specification would assist in identifying which of these constructshas better predictive validity
in volume estimates.
In this study, we focused on one dimension of packaging
shape-elongation. However,many other aspects of packaging conceivablycould affect perceivedvolume and consumption, for example,aspectsof packageshapeotherthanelongation, color, material,aesthetic appeal, and so forth. Study 2
shows that the use of elongation as a source of information
may be partiallyautomatic.This researchcould be extendedto
test whetherconsumershave an inabilityto control their relianceon the elongationheuristiceven when they areawareof
the elongationbias (Bargh1989). If so, the questionof how to
educateconsumersso they are less prone to these biases becomes an importantpublicpolicy issue. Will consumereducation be useful, or mustthe mannerin whichproductsarepackaged be regulatedso thatvolume informationis highly salient?
REFERENCES
C.
Adams,Anthony, Anthonydi Benedetto,andRajanChandran
(1991),"CanYouReduceYourPackageSizeWithoutDamaging
Sales?" Long Range Planning, 24 (4), 86-96.
Anderson, Norman H. (1970), "AveragingModel Applied to the
Size-Weight Illusion,"Perceptionand Psychophysics,8 (1), 1-4.
Assuncao, Joao and RobertJ. Meyer (1993), "The Rational Effect
of Price Uncertainty on Sales-Price Relationships,"Management Science, 39 (May), 517-35.
Bargh,JohnA. (1989), "ConditionalAutomaticity:Varietiesof Automatic Influence in Social Perceptionand Cognition"in Unintended Thought,James S. Uleman and JohnA. Bargh,eds. New
York:Guilford Press, 3-51.
Baron, Rueben M. and David A. Kenny (1986),"The Moderator-Mediator VariableDistinction in Social Psychological Research: Conceptual, Strategic, and Statistical Consideration,"
Journal of Personalityand Social Psychology, 51 (6), 1173-82.
Been, Richard T., Myron L. Braunstein, and Mildred H. Piazza
(1964), "Judgment of Volume Reduction in Distorted Metal
Containers,"Journal of Engineering Psychology, 3 (16), 23-27.
325
Blattberg, Robert, Thomas Buesing, Peter Peacock, and Subrata
Sen (1978), "Identifyingthe Deal Prone Segment," Journal of
MarketingResearch, 15 (August), 369-77.
Chapman,Loren J. and Jean P. Chapman(1969), "IllusoryCorrelation as an Obstacle to the Use of Valid Psychodiagnostic
Signs," Journal of AbnormalPsychology, 74 (3), 271-80.
Charpentier,A. (1891), "Analyze Experimentalede Quelques Elements de la Sensation de Poids [ExperimentalStudy of Some
Aspects of Weight Perception],"Archives de Physiologie Normales et Pathologiques, 1 (3), 122-35.
Cross, David V. and Laurence Rotkin (1975), "The Relation Between Size and Apparent Heaviness," Perception and Psychophysics, 18 (2), 79-87.
Davis, Christopher M. and William Roberts (1976), "Lifting
Movements in the Size-Weight Illusion," Perception and Psychophysics, 20 (July), 33-36.
Dickson, Peter R. and Alan G. Sawyer (1986), "Point of Purchase
Behavior and Price Perceptions of Supermarket Shoppers,"
Marketing Science Institute Report No. 86-102. Cambridge,
MA: MarketingScience Institute.
Ellis, R.R. and S.J. Lederman(1993), "The Role of Haptic Versus
Visual Volume Cues in the Size-Weight Illusion," Perception
and Psychophysics, 53 (3), 315-24.
Flournoy,T. (1894), "De l'Influence de la PerceptionVisuelle des
Corps sur Leur Poids Apparent[The Influenceof Visual Perception on the Apparent Weight of Objects]," L'Annee Psychologique, 1, 198-208.
Folkes, Valerie S., Ingrid M. Martin, and Kamal Gupta (1993),
"When to Say When: Effects of Supply on Usage," Journal of
ConsumerResearch, 20 (December), 467-77.
Frayman,Bruce J. and William E. Dawson (1981), "The Effect of
Object Shape and Mode of Presentationon Judgments of ApparentVolume,"Perceptionand Psychophysics,29 (1), 56-62
Gilbert, Daniel T. (1989), 'Thinking Lightly About Others:Automatic Components of the Social Inference Process" in Unintended Thought,James S. Uleman and John A. Bargh,eds. New
York:GuilfordPress, 189-211.
, BrettW. Pelham,and Douglas S. Krull(1988), "OnCognitive Busyness:WhenPersonPerceiversMeet PersonsPerceived,"
Journalof Personalityand Social Psychology,54 (May), 733-40.
Gundlach, C. and C. Macoubrey(1931), "The Effect of Color on
ApparentSize," AmericanJournal of Psychology, 43, 109-11.
Hendon, Donald W. (1986), Battlingfor Profits. Jonesboro, AR:
Business ConsultantsInternational.
Hoch, Stephen J. and John Deighton (1989), "Managing What
Consumers Learn from Experience,"Journal of Marketing,53
(2), 1-20.
Holmberg, Lennart (1975), "The Influence of Elongation on the
Perception of Volume of Geometrically Simple Objects," Psychological Research Bulletin, 15 (2), 1-18
Isakson, Hans R. and Alex R. Maurizi (1973), "The Consumer
Economics of Unit Pricing,"Journal of MarketingResearch, 10
(August), 277-85.
Jones, Lynette (1986), "Perceptionof Force and Weight,"Psychological Bulletin, 100 (1), 29-42.
Krishna,Aradhnaand Priya Raghubir(1997), "The Effect of Line
Configurationon Perceived Numerosityof Dotted Lines,"Memory and Cognition, 25 (July), 492-507.
Luczak, Holger and ShuangshengGe (1989), "Fuzzy Modelling of
Relations Between Physical Weight and Perceived Heaviness:
The Effect of Size-Weight Illusion in IndustrialLifting Tasks,"
Ergonomics, 32 (7), 823-37.
Masin, Sergio C. and Loredana Crestoni (1988), "Experimental
Demonstrationof the Sensory Basis of the Size-Weight Illusion," Perceptionand Psychophysics,44 (October),309-12.
Moore,WilliamL. and Russell S. Winer(1978), "An Experimentto
Determinethe Effects of Package Size on Consumption,"working paper,GraduateSchool of Business, ColumbiaUniversity.
326
JOURNALOF MARKETING
RESEARCH,AUGUST1999
Nakatani, Katsuya (1985), "Applicationof the Method of Fixed
Set to the Size-WeightIllusion,"PsychologicalResearch,47
(December), 223-33.
Payne, John W., James R. Bettman, and Eric J. Johnson (1988),
"The Adaptive Decision-Maker: Effort and Accuracy in
Choice,"Journalof Experimental
Psychology:Learning,Memory,and Cognition,14 (3), 534-53.
Pearson, Richard G. (1964), "Judgmentof Volume from Pho-
and MotorSkills,18
tographsof ComplexShapes,"Perceptual
(3), 889-900.
Piaget, Jean (1968), "Quantification,Conservationand Nativism,"
976-79.
Science,162(November),
, B. Inhelder,andA. Szeminska(1960), TheChild'sConNewYork:BasicBooks.
ceptionof Geometry.
Pick, HerbertL. and Anne D. Pick (1967), "A Developmentaland
AnalyticStudyof the Size-WeightIllusion,"Journalof ExperimentalChildPsychology,5 (3), 362-71.
Raghubir,Priya and AradhnaKrishna(1996), "As the Crow Flies:
Bias in Consumers' Map-Based Distance Judgments,"Journal
of ConsumerResearch,23 (June),26-39.
Ross, Helen E. (1969), "WhenIs Weight Not Illusory?"Quarterly
Journalof Experimental
Psychology,21 (4), 346-55.
Sarris,Victor and EdgarHeineken (1976), "An ExperimentalTest
of Two MathematicalModels Applied to the Size-Weight Illu-
sion,"Journalof Experimental
Psychology:HumanPerception
2 (2), 295-98.
and Performance,
Stangor, Charles and David McMillan (1992), "Memory for Expectancy-Congruentand Expectancy-IncongruentInformation:
A Review of the Social and Social Development Literatures,"
PsychologicalBulletin,111(1), 42-61.
Stewart,Brad(1994), CanYouTrusta Tomatoin January?New
York:BantamBooks.
Usnadze, D. (1931), "Concerningthe Size-Weight Illusion and Its
Analogues [Ueber die Gewichtstaeuschungund ihre Analoga],"
Forschung,14, 366-79.
Psychologische
Wansink,Brian (1996), "CanPackage Size Accelerate Usage Vol-
60 (July),1-14.
ume?"Journalof Marketing,
and Rohit Deshpand6(1994), "Outof Sight, Out of Mind:
Pantry Stockpiling and Brand Usage Frequency,"Marketing
Letters,5 (1), 91-100
and Michael L. Ray (1992), "Estimatingan Ad's Impacton
One's Consumptionof a Brand,"Journalof AdvertisingRe9-16.
search,32 (May-June),
and
(1996), "Advertising Strategies to Increase
Usage Frequency,"Journal of Marketing,60 (January),
31-46.
Ward, Ronald W. and James E. Davis (1978), "A Pooled CrossSection Times Series Model of Coupon Promotion," Ameri-
can Journal of AgriculturalEconomics, 60 (November),
393-401.
A Studyof MentalLife.New
R.S.(1921),Psychology:
Woodworth,
York:Holt.