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.