influence of contagious versus noncontagious product

Influence of Contagious versus Noncontagious Product Groupings on Consumer Preferences

ARUL MISHRA *

This article studies the influence of product groupings on consumer preferences. Specifically, it is proposed that when each product in two groups has an equal chance of a gain, consumers prefer to choose from a group that appears more contagious (e.g., products arranged close together, similarly, or symmetrically). However, when each product in two groups has an equal chance of a loss, consumers prefer to choose from a group that appears less contagious (e.g., products arranged apart, dissimilarly, or asymmetrically). Across three experiments, the effect is demonstrated, and contagion theory is used to explicate the underlying process.

C ontagion theory explores the age-old belief that qualities are contagious and transferable. It is believed that

qualities can be transferred from a source to a target and that proximity to the source enhances feelings of contagion (Rozin, Markwith, and Nemeroff 1992; Rozin, Millman, and Nemeroff 1986; Rozin and Nemeroff 2002). The consumer behavior literature has also looked at the role of contagion. Argo, Dahl, and Morales (2006) find that consumers evaluate as less favorable those products previously touched by others. Morales and Fitzsimons (2007) show that consumers believe that disgusting products in a shopping cart are able to contaminate other products in the cart.

Instances of product groupings are abundant in the marketplace (e.g., on retail shelf displays), prompting questions such as, "Are people likely to believe that a good or bad quality in a product is transferable to others in the same group? Does the entire group start reflecting the good or bad qualities inherent in one of its members?" Very little attention has been paid to the contagiousness of different types of groupings on consumer preferences. The focus of

* AmI Mishra is assistant professor at the David Eccles School of Business, 470B BuC 1645 East Campus Center Drive, University of Utah, Salt Lake City, UT 84112 ([email protected]). This work is based on the author's dissertation at the University of Iowa under the supervision of Dhananjay (DJ) Nayakankuppam. The author is indebted to DJ for his invaluable suggestions and feedback. This work was a runner up for the 2006 SCP Sheth Dissertation Proposal Award and was partially supported by it-the author gratefully acknowledges the support of SCPo The article benefited from the helpful feedback received from Irwin Levin, Himanshu Mishra, Peter Riesz, Rick Tubbs, David Sanbonmatsu, and Paul Windschilt. The author would also like to thank the editor, the associate editor, and three reviewers for their very helpful comments and suggestions.

John Deighton served as editor and Ann McGill served as associate editor for this article.

Electronically published November 12, ZOOS

73

this article is on how different product groupings, and the belief that good and bad qualities spread in a group, affect consumer preferences, in the domains of both gains and losses.

Consider an instance of one form of grouping with products arranged close together (and hence perceived as more contagious; Rozin and Nemeroff 2002) as opposed to far apart. It is proposed that when two groups provide an equal chance of a gain (e.g., one product in each group contains a gift coupon), then consumers prefer choosing from the group in which products are kept close together (the "close" contagious group). However, when the two groups provide an equal chance of a loss (e.g., one product in each group has a defect), consumers prefer choosing from the group in which the products are kept far apart (the "apart" noncontagious group). Such a pattern of preference would indicate that in the domain of gain consumers feel there is a higher likelihood of drawing the gain product (the specific product with a gift coupon) from the close group, while in the domain of loss they feel there is a lower likelihood of drawing the loss product from the apart group. Normative theory would predict that consumers should be indifferent between the two groups since both groups offer equal likelihood of drawing the gain (or the loss) product. However, this article demonstrates that consumers use the grouping, a noninformative factor, as a cue in their decision. In the early stages of the present research a preliminary study was conducted to illustrate the proposed effect and provide a better understanding of it before building an underlying theoretical account.

PRELIMINARY STUDY

A simple game of chance was devised in which participants could either win or lose money based on their choices.

© 2008 by JOURNAL OF CONSUMER RESEARCH, Inc .• Vol. 36 • June 2009

All rights reserved. 0093-53011200913601-0003$10.00. DOl: 10.10861595716

by guest on October 17, 2016

http://jcr.oxfordjournals.org/D

ownloaded from

http://jcr.oxfordjournals.org/

74

One hundred and twenty-seven participants were told that they would be playing a game of chance and were randomly assigned to a gain or a loss game. Participants were first shown two identical tables and informed that each table had 16 balls, eight red and eight blue. However, since the balls were wrapped in a silver foil, they were not able to identify which were the blue and which were the red balls. One table had the 16 balls arranged close together (close group), and the other table had the 16 balls arranged apart (apart group). Participants first chose the table from which they would play the game. Once they had chosen a table containing either the close group or the apart group, they were asked to pick up a ball from the chosen group. Participants in the gain game were told that if they picked a blue ball they would win $3, while participants in the loss game were told that if they picked a blue ball they would lose $3. The dependent variable was the table from which participants chose to play the game. After the participants had played the gain or loss game, they were asked to recall the number of blue balls in each group (since picking up a blue ball entailed a win or a loss) to test whether they were able to correctly assess the probability of winning or losing. Finally, participants were asked to guess the real purpose of the game and whether they found anything about the experimental setting unusual.

The domain, gain versus loss, had a significant impact on the choices made by the participants (X 2 (1) = 9.41, p < .002). In the gain game, a significantly greater number of participants chose to play from the table with the close group rather than the table with the apart group (64.52% vs. 35.48%; X2(1) = 5.22, p < .02). A reversal of this was observed for the loss game in which participants chose to play from the apart group rather than from the close group (63.08% vs. 36.92%; X2(l) = 4.45, p < .04). Participants thus behaved as if they had a greater likelihood of winning from the close group and a lower likelihood of losing from the apart group. The results indicated, however, that 86.6% of participants were able to state correctly that there were eight blue balls in each group, indicating that they correctly assessed the probability of winning or losing. Of the remaining participants, 5.5% stated a number greater than eight, and 7.8% stated a number less than eight. Moreover, none of the participants were able to guess the real purpose of the game, and neither did they report that any part of the experimental setting appeared unusual.

The preliminary study provided an initial demonstration of the proposed effect and indicated that participants were successful in correctly assessing the probability inherent in the game. The remaining part of this article is devoted to investigating the underlying process that causes the effect and to demonstrating its implications in consumer decision domains. Since the effect deals with the transmission of good or bad qualities, findings from contagion theory are first discussed and then used to conceptualize a process account. Experiments 1-3 use different contagion cues to replicate the proposed effect and test the viability of the contagion account. The experiments are designed to dem-

JOURNAL OF CONSUMER RESEARCH

on strate the implication for product choice, product packaging, and retail decisions. Finally, implications, limitations, and future research directions are discussed.

THEORETICAL DEVELOPMENT AND CONCEPTUALIZATION

Contagion Theory

Contagion theory suggests that there is a transfer of quality from a source to a target. The quality exchanged could be physical, mental, or moral in nature and positive or negative in valence. In a study, Rozin et al. (1986) demonstrate that when participants themselves had filled sugar in two bottles and randomly labeled one of them sucrose and the other sodium cyanide, they still show a marked disinclination for drinking a sugar solution from the bottle labeled sodium cyanide. The results of this study indicated that even an innocuous label is assumed to transmit the poisonous qualities of sodium cyanide to a sugar solution. Rozin et al. (1986) find that objects that are considered representative of a source object are believed to possess the same qualities as the source object. For instance, people show reluctance to eat chocolates that have disgusting shapes, hit a dart on a dart board depicting John F. Kennedy's face, or tear up duplicate photographs of loved ones (Rozin and Nemeroff 2002).

Research has demonstrated various characteristics of contagion (Rozin et al. 1992; Rozin and Nemeroff 2002). First, physical proximity enhances the sense of contagion (Morales and Fitzsimons 2007). The closer the objects are kept in a group, the greater is the feeling that qualities are transferred among the member objects. Second, objects that appear similar are considered to possess or share similar characteristics. Groups consisting of members with similar likes, backgrounds, or appearance are perceived to be more cohesive and appear to share the same qualities compared to groups with dissimilar group members (Crawford, Sherman, and Hamilton 2002). Third, once the target is in contact with the contaminating source, the quality of the source is considered permanently transferred to the target, even when the source is removed (Rozin and Nemeroff 2002). Fourth, contagion is considered "holographic"-if anyone member of a group is considered to have a specific property, then the complete group, of which the member is one component, is considered to have the property (Rozin and Nemeroff 2002). Work on entitativity also provides insights into the processes people may use in generalizing characteristics over members of a group. Although this work has studied the influence of coherent groups in the social judgment domain, its findings can be extended to understand the spread of qualities in a group. Entitative groups are defined as groups that are based on features of proximity, similarity, symmetry, collective movement, and common fate (Y zerbyt, Rogier, and Fiske 1998). High entitative groups foster a perception of "groupness" and permit generalizability of qualities across members (Crawford et al. 2002). Specifically, when people process information about highly enti-



ownloaded from


CONTAGIOUS PRODUCT GROUPINGS

tative groups, they ascribe qualities of one member to the entire group; however, this transference of qualities is hindered for low entitative groups (Spencer-Rodgers, Hamilton, and Sherman 2007).

Recent investigations in marketing literature have also looked at the role of contagion. Argo et al. (2006) find that consumers evaluate as less favorable those products that they perceive have been previously touched by others. Any cue in the retail environment that signals to the consumer that someone else has touched the product results in increased perceived contamination; however, such a contamination is shown to dissipate over time. Morales and Fitzsimons (2007) find that consumers believe that disgusting products are able to contaminate other products by transferring their offensive qualities. Using a scenario in which consumers are shown products kept in a shopping cart, Morales and Fitzsimons (2007) demonstrate that even without physical contact and with only a perception of nearness consumers experience disgust for products kept near disgusting products (e.g., trash bags, cat litter, feminine napkins, etc.). In sum, contagion theory suggests that people hold beliefs that qualities are transferable and can influence preferences and behavior.

Conceptualization

The literature on contagion suggests that qualities are believed to be transferred across group members. A group that is considered contagious appears to spread the qualities of one of its members to the complete group such that the entire group, and all its members, seems to possess the quality (Rozin and Nemeroff 2002). Using the reviewed findings in contagion theory (Morales and Fitzsimons 2007; Rozin and Nemeroff 2002) and entitativity (Crawford et al. 2002; Spencer-Rodgers et al. 2007) it can be inferred that group features such as similarity, proximity, or symmetry are said to enhance perceived contagion. People are more likely to believe that good or bad qualities will be more pervasive in such groups as opposed to groups in which the products are arranged far apart, dissimilarly, or asymmetrically. For the purposes of this article, proximal, similar, and symmetric groups are called contagious groups, and distal, dissimilar, and asymmetric groups are called noncontagious groups. In sum, contagious groups appear to facilitate the transmission of qualities, while noncontagious groups appear to hinder the transmission of qualities.

Based on these propositions, the following predictions can be made for contagious and noncontagious groups in the gain and loss domain. In the domain of gain, if one product has a gain associated with it (a gift coupon), then all products in the contagious group are perceived to be infused with the gain. Consumers feel that by choosing from such a group they are increasing their chances of picking up the gain product. However, loss is also perceived to spread strongly across the contagious group, reducing consumers' willingness to choose from such a group. The operating belief would be that picking up a product from contagious groups would increase one's chance of choosing the loss product, which is never preferred.

75

Converse predictions can be made for the noncontagious groups. In the domain of gain, the gain seems isolated to one specific product and is not perceived to spread across to the other products in the group. Due to the lack of infusion of the gain quality across the noncontagious group, consumers do not prefer to choose from it since they do not believe they have a good chance of picking up the gain product. In contrast, in the domain of loss, the loss associated with one product appears to spread less strongly in the noncontagious group as compared to the contagious group. By choosing from the noncontagious group, consumers feel that they have reduced their likelihood of picking up the loss product. Therefore, consumers prefer to choose from the contagious group in the gain domain and from the noncontagious group in the loss domain.

Previewing briefly, experiment 1 tests the effect in a product choice domain. Experiment 2 tests the proposed account by demonstrating the mediating role of contagion. Experiment 3 uses a contagion-priming task to moderate the effect. The participants in the experiments were business undergraduate students who took part in the experiments for partial course credit. Gender data were collected for each of the experiments but did not influence any results. Therefore, they are not discussed further.

EXPERIMENT 1 Experiment 1 had two aims. First, to use the feature of

proximity to manipulate contagion, since proximity is said to facilitate the transmission of qualities (Rozin and Nemeroff 2002). Second, to replicate the proposed effect in a consumer choice context. One hundred and fifteen participants took part in the experiment and were taken into a room one at a time. Participants were informed that, similar to a retail environment, they would be shown groups of products and would have to choose one product. Each participant was shown two groups of nine ketchup bottles arranged on two identical tables. One group of ketchup bottles was arranged close together (contagious group), while the other group was arranged apart (noncontagious group). The ketchup bottles were arranged as depicted in figure 1.

Participants were given the questionnaire and left alone to read the instructions and provide their responses. They were randomly assigned to a gain or loss domain. In the loss domain, participants were told that one ketchup bottle in each of the groups had a defective lid, causing the ketchup to splash out. However, it was not known which of the nine bottles had the defective lid because all of the bottles were sealed. Picking the defective bottle would entail a loss of $3, which was the price of the bottle, since the defective bottle would be unusable. Participants in the gain domain were told that one bottle in each of the groups contained a gift coupon for $3 but that it was not known which of the nine bottles had the gift coupon.

To ensure that location of the bottles did not influence choice, the order was counterbalanced by the contagious group appearing on the right for some participants and appearing on the left for others. Additionally, each bottle in both



ownloaded from


76 JOURNAL OF CONSUMER RESEARCH

FIGURE 1

EXPERIMENTAL SETUP FOR EXPERIMENTS 1-3

0 0 0

Ketehup 0 0 0

[j] ------0 0 0 000 000 0 0 0

CLOSE GROUP APART GROUP

WITHIN PARTICIPANT ASSIGNMENT. EXPERIMENT I

• • • • 0 0 • 0

I Toothp.... r---- • • • • e • e

• • • • • 0 0 • • • • • 0 0 • 0

SIMILAR GROUP DISSIMILAR GROUP

BETWEEN PARTICIPANT ASSIGNMENT, EXPERIMENT 2

0 0 Ketchup 0 0 0 0 [j]~0 0

0 0 0 00 0

0 0 0 0

SYMMETRIC GROUP ASYMMETRIC GROUP

WITHIN PARTICIPANT ASSIGNMENT, EXPERIMENT 3

NOTE.-Color version available as an online enhancement.

groups was assigned a number from one to nine, which marked its place in the group. Participants could not see the numbers. The number was used to determine whether there was a specific pattern in which participants chose the bottles, from the middle, the edges, or randomly. Thus, the design of the experiment was a 2 (domain: gain vs. loss) x 2 (location order: contagious group on left vs. on right) between-participants design, with each participant seeing both the contagious and the noncontagious groups.

Participants indicated whether they would choose a bottle from the contagious group or the noncontagious group or whether they were indifferent between the two groups. To simulate realistic behavior, the task had real payoffs. On the questionnaire, they were asked to pick up a bottle from their chosen group and take it with them. Finally, participants were asked to list the reasons for choosing from either of the two groups. Participants who had responded that they were indifferent were still told to pick a bottle from either group and were later asked to explain the reasons they had felt indifferent between the two groups. Participants left the room through another door, and they were thanked and debriefed.

Results and Discussion

In the gain domain, significant differences emerged across participants' choices (x 2 (2) = 31.53, p < .0001). Specifically, 66.07% of participants selected a bottle from the contagious group, as compared to 28.57% who selected from the noncontagious group (X 2 (1) = 8.32, p < .004) and 5.36% who were indifferent between the two groups (X2(1) = 28.90, p < .0001). In the loss domain, significant differences emerged across participants' choices (x2(2) = 24.64, p < .0001). Specifically, a pattern in reverse to that of the gain domain emerged: 61.02% of participants selected a bottle from the noncontagious group, as compared to 30.51 % who selected from the contagious group (X2(1) = 6, p < .02) and 8.47% who were indifferent between the groups (X2(1) = 15.21, P < .0001). The location order did not interact with domain (F < 1, p > .4).

Notably, no specific pattern emerged in the manner in which participants chose the bottles. They were as likely to choose from the middle as from the edges of the groups in both the gain and the loss domains. Thought protocols were



ownloaded from



analyzed to understand the underlying psychological processes behind participants' choices. Two independent judges who were blind to the research hypothesis and the experimental conditions coded participants' thoughts to discover thoughts of contagion or spread of qualities, likelihood of winning or losing being greater in the contagious versus noncontagious group, and one type of grouping appearing better than the other. No differences emerged in participants' thoughts in either the gain or the loss domains. In fact, quite a few participants in both the gain and the loss domains actually acknowledged that the mathematical probability of gaining or losing was the same in both groups, a pattern consistent with the findings of the preliminary study, which suggests that perceived contagion influences subjective probability estimates but not the actual estimates. No thoughts of perceived contagion were reported.

Experiment 1 replicated the effect in a product choice situation in which participants preferred to choose from the close (contagious) group in the gain domain but from the apart (noncontagious) group in the loss domain. Both the preliminary study and experiment 1 used proximity to demonstrate the effect. One aim of experiment 2 was to test the viability of the proposed contagion account using a different feature of contagion, that is, similarity.

EXPERIMENT 2: MEDIATION BY CONTAGION

Experiment 2 used the group feature of similarity to manipulate contagion. That is, the distance between the products was kept the same in both groups, but the package color was either the same (similar, contagious group) or different (dissimilar, noncontagious group). Similar groups are expected to increase perceived transmission of qualities, while dissimilar groups are likely to reduce perceived transmission (Rozin and Nemeroff 2002). A second aim of the experiment was to test whether contagion played a mediating role. In order to test for mediation, similarity was manipulated as a feature of contagion, and the resultant perceived contagion was measured. Considering the limitations of thought protocols in identifying the role of contagion and in line with the suggestions provided by Kahn, Luce, and Nowlis (2006), a more unobtrusive method of a response time measure was used (Luce 1998). The response time to recognize high- and low-contagion words was used to identify the mediating role of contagion.

Third, experiment 2 demonstrates the implications for product packaging. Research has demonstrated that visual cues about the product package, such as size, shape (Folkes and Matta 2004), spokesperson (Garretson and Burton 2005; Priester and Petty 2003), or color (Jacobs et al. 1991), can influence purchase intentions. This experiment was designed to examine the influence of a similar- versus a dissimilarcolored package group on perceived contagion and subsequent willingness to buy. Finally, this experiment tests the robustness of the effect in a between-participant design in which participants saw only one group, the contagious or

77

the noncontagious group, rather than both groups together as in the previous experiments. The design of the experiment was thus a 2 (domain: gain vs. loss) x 2 (package color: same-colored product packages group vs. different-colored product packages group) between-participants design.

Method

One hundred and fifty-nine participants took part in the experiment for partial course credit. They were randomly assigned to one of the four between-participants conditions. Participants were told that, similar to a retail environment, they would be shown a group of products, and their evaluations about the displayed products would be elicited. They were taken into a room one at a time and shown a group of 16 toothpaste packs arranged on a table. The arrangement of the toothpaste packs was identical in all conditions (as depicted in fig. 1). The only difference was that in the samecolored condition all toothpaste packs were of the same red color (contagious group), while in the different-colored condition the toothpaste packs had been colored red, blue, green, and yellow (noncontagious group). Both groups had the toothpaste packs arranged in a square form with four packs in a row and each pack separated by a distance of about 6 inches from the other packs. The packs did not display any brand information. Since package color was a between-participants factor, each participant was exposed to only one group of toothpastes, either the same-colored group or the different-colored group. Participants were then randomly assigned to the gain or the loss condition. Participants in the gain condition were told that one toothpaste pack among the 16 had a gift coupon for $3, while participants in the loss condition were told that one pack of the 16 was filled with a smaller quantity of toothpaste.

Participants were asked to observe the toothpaste packs and to provide their responses about the toothpaste on a laptop computer kept on a nearby table. When participants began the computer-based questionnaire, they were informed that due to a technical glitch the questions about the toothpaste would appear later; in the meantime, they were asked to complete a separate word recognition task. Participants were told that either words or non words would appear on the computer screen. If they saw a word, they should press P on the keyboard, and if they saw a nonword, they should press Q as fast as possible. High-contagion words (transmit, spread, pervade), low-contagion words (confine, isolate, separate), or nonwords (decuyte, rotall, bashiting) appeared on the computer screen. The speed with which participants responded to the words formed a measure of their perceived contagion. It was expected that the concept of contagion would be rendered more accessible for participants who had seen same-colored product packages compared to those who had seen different-colored product packages, for both the gain and the loss conditions. Therefore, the former should be faster at recognizing high-contagion words, while the latter should be faster at recognizing lowcontagion words.

After participants had completed the word recognition



ownloaded from


78

task, they were asked to provide responses about the toothpaste packages that they had seen first. They were asked their willingness to buy a pack from the group in front of them on a 5-point scale, with 1 indicating a low willingness to buy, and 5 indicating a high willingness to buy. Then they were debriefed and left the room.


Willingness to Buy. The analysis yielded a main effect for gain and loss. Participants in the gain domain indicated a greater willingness to buy (M = 3.29) than those in the loss domain (M = 3.01; F(1, 155) = 2.87,p < .09). Further, a significant domain x package-color interaction emerged (F(l, 155) = 38.11, p < .0001). A decomposition of the interaction revealed that in the gain domain participants in the same-colored package condition (contagious group) indicated a higher willingness to buy toothpaste (M = 3.72) compared to participants in the different-colored package condition (noncontagious group; M = 2.86; F(1,78) = 16.95, p < .0001). The loss domain exhibited a reversed effect, with participants in the different-colored package condition indicating a greater willingness to buy (M = 3.58) than participants in the same-colored package condition (M = 2.45; F(l,77) = 21.12, p < .0001). Figure 2a graphs the results for willingness to buy.

Perceived Contagion. Logarithmically transformed response times were averaged across high-contagion words, low-contagion words, and nonwords that were subsequently subjected to a MANOV A. The same pattern of result emerged for the gain and the loss domains, and subsequent analyses used data collapsed across domains. Analysis yielded a significant word x package-color interaction (F(2, 156) = 25.19, p < .0001). A decomposition ofthis interaction across words revealed that (i) for high-contagion words, participants assigned to the same-colored package condition responded faster (M = - .095) than participants assigned to the different-colored package condition (M = .245; F(1, 157) = 34.47, p < .0001); (ii) for low-contagion words, participants assigned to the different -colored package condition responded faster (M = .005) than participants assigned to the same-colored package condition (M = .151; F(1,157) = 8.16, p < .004); and (iii) for nonwords, no significant difference emerged across the same- and the different-colored package condition (F < .05, p> .9). Figure 2b graphs the results for high- and low-contagion words.

Mediation. An overall high perceived-contagion score (Lh) for each participant was calculated by averaging the response times for the high-contagion words. Similarly, an overall low perceived-contagion score (L/) for each participant was calculated by averaging the response times for the low-contagion words. Finally, to obtain a single measure of contagion for each participant, a difference score was created by subtracting the response time to low-contagion words from the response time to high-contagion words (Lh - L/). Lower values of the difference score indicated higher per-


FIGURE 2

WILLINGNESS TO BUY AND MEASURE OF CONTAGION FOR EXPERIMENT 2

Willingness to buy 4.5

3.72

3.5

2.5

1.5

Gain

.same color

A. Willingness to buy

Log response time 0.3

0.2

0.1

o

0.245

3.58

2.45

Loss

C Different color

0.151

0.005

word low contagion word

-0.1 -0.095

-0.2

llsame color C Different color

B. Measure of perceived contagion

ceived contagion since the recognition for high-contagion words is faster. Higher values indicated lower perceived contagion since the recognition for high-contagion words is slower.

The proposed contagion account suggests that similar (contagious) groups engender greater perceived contagion, subsequently influencing consumer preferences. If this is true, then perceived contagion should mediate the influence of group features on consumers' willingness to buy. Considering the same- versus different-colored package a manipulation of contagion, the response times for high- and low-contagion words as an indicator of perceived contagion, and willingness to buy as the outcome variable, separate mediation analyses were run for the gain and the loss domains (Baron and Kenny 1986).



ownloaded from



In the gain domain, the package color had a significant influence on willingness to buy (t(l) = -4.12, p < .0001). The package color also predicted perceived contagion (t(1) = 4.78, p < .0001). The perceived contagion had a significant influence on willingness to buy (t(1) = -16.11, P < .0001). Finally, when a full model was run, examining the joint impact of package color and perceived contagion on willingness to buy, the influence of package color on willingness to buy no longer remained significant (t(l) = -.11, p> .90). However, the influence of perceived contagion on willingness to buy remained significant (t(l) = -14.02, p < .0001). In the loss domain, the package color again had a significant influence on willingness to buy (t(I) = 4.60, p < .0001). The package color significantly predicted perceived contagion (t(1) = 4.95, p < .0001), and perceived contagion had a significant influence on willingness to buy (t(1) = 13.83, p < .0001). However, when a full model was run to examine the joint impact of package color and perceived contagion on willingness to buy, the influence of package color on willingness to buy became nonsignificant (t(1) = .93, p> .35), but the influence of perceived contagion on Willingness to buy remained significant (t(1) = 11.58, p < .0001). A Sobel test also suggested that the influence of contagious versus noncontagious groups on product preferences was mediated through perceived contagion in both the gain (Sobel test statistics = -4.78, p < .001) and the loss domains (Sobel test statistics = 4.81, p < .001).

The results of experiment 2 demonstrate the following findings. First, participants displayed a preference to choose from the similar-colored product package group in the gain domain and the dissimilar-colored product package group in the loss domain. Second, such a preference was mediated by perceived contagion. Thus, this pattern of results demonstrates the role of contagion in the proposed effect.

EXPERIMENT 3: MODERATION BY CONTAGION PRIME

Experiment 3 had a twofold objective. First, experiments 1 and 2 used the group features of proximity and similarity to manipulate transmission of qualities. Therefore, experiment 3 used the group feature of symmetry to manipulate high versus low perceived contagion since symmetric groups facilitate greater transmission of qualities compared to asymmetric groups (Crawford et al. 2002; Spencer-Rodgers et al. 2007). It is proposed that in the gain domain participants would prefer to choose from the symmetric group (contagious group), and in the loss domain they would prefer to choose from the asymmetric group (noncontagious group). Second, in order to gain further support for the proposed contagion account, experiment 3 primed feelings of high versus low contagion and studied their subsequent moderating influence.

Contagion Prime

The prime consisted of four scenarios that included descriptions of an action either spreading to others (increased

79

contagion) or not spreading and remammg isolated (decreased contagion). The actions consisted of a person laughing and other people in the vicinity also starting to laugh (or not), one person yawning and others also beginning to yawn (or not), the spread (or not) of word-of-mouth messages about free ice cream, and a person slipping down an icy slope and others also slipping because of her (or not). Therefore, the same action (e.g., laughing or yawning) was considered either contagious or not contagious, depending on whether it was spread or not. A pretest was conducted to test the efficacy of the prime.

Ninety-one participants took part in the pretest and were randomly assigned to read the four scenarios consisting of either high- or low-contagion primes, after which they did a wordlnonword recognition task on a computer. Words appeared one at a time on the computer screen, and participants were asked to press Q if it was a nonword and P if it was a word, as quickly as possible. The computer recorded the response time between presenting the stimulus and registering a response. Each participant was shown six highcontagion words (spread, pervade, diffuse, mingle, transmit, permeate), six low-contagion words (separate, isolate, detached, confine, segregate, limited), and six nonwords (conudrick, cota, helempki, bashitig, donkangh, va1cuni). It was expected that high contagion would be rendered more accessible for participants who had read scenarios of high contagion, whereas the concept of low contagion would be made more accessible for participants exposed to the lowcontagion prime.

Logarithmically transformed response times were averaged across high-contagion words, low-contagion words, and nonwords, which were subsequently subjected to a MANOV A. Analysis yielded a significant prime x accessibility interaction (F(2, 88) = 16.5I,p < .0001). A decomposition of this interaction across accessibility revealed that (i) for high-contagion words, participants primed with high contagion responded faster (M = -.113) than participants primed with low contagion (M = .054; F(1,89) = 8.25, p < .005); (ii) for low-contagion words, participants primed with low contagion responded faster (M = - .157) than participants primed with high contagion (M = .015; F(I, 89) = 6.11, p < .01); and (iii) for nonwords, no significant difference emerged across participants primed with high and low contagion (F < .51, p > .47). Figure 3 graphs the results.

Method

Two hundred and thirty participants took part in the experiment for partial course credit. They were assigned randomly to one of the following six between-participant conditions: 3 (prime: high contagion vs. low contagion vs. control) x 2 (domain: gain vs. loss). Participants were first primed by being asked to read the scenarios of high or low contagion or assigned to the control condition. Subsequently, each of them was taken to a separate room one at a time and shown two groups of nine ketchup bottles, one group arranged symmetrically (contagious group) and the other group arranged asymmetrically (noncontagious group). The



ownloaded from


80

FIGURE 3

PRETEST OF CONTAGION PRIME, EXPERIMENT 3

Log (Response time) 0.2

0.15

0.1

0.05

0+--,..--

-0.05

-0.1

-0.15 -0.113

-0.2

0.054

DPrime Hi-contagion

0.D15

-0.157

.Prime Low-CIOntagion

symmetric group was arranged in the form of a square with bottles kept about a foot apart from each other (as depicted in fig. 1). The asymmetric group had the bottles arranged in a haphazard, zigzag manner with some bottles near, some far from each other, and none of them in a straight line. l

As in experiment 1, half the participants were told that one ketchup bottle in each group had a gift coupon (gain domain), and the other half were told that one ketchup bottle in each group had a defective lid (loss domain). Participants were told that it was not known which of the nine bottles had the gift coupon or the defective lid. They were asked from which group they would prefer to choose a ketchup bottle. They used a 5-point scale, with 1 indicating a preference to choose from the noncontagious group, 3 indicating indifference, and 5 indicating a preference to choose from the contagious group. As a token of appreciation, participants were also asked to pick a ketchup bottle from either group.


The analysis revealed a prime x domain interaction (F(5,224) = 18.80, p < .0001). A decomposition of the interaction across domains showed that the proposed effect emerged in the control condition. Consistent with the findings of previous experiments, participants in the gain domain preferred to choose from the contagious group (M = 3.48), while those in the loss domain preferred to choose from the

lIn ord~r to ensure that the groups of products appeared symmetric or asymmetrIc, a separate test was conducted with 19 participants. They were shown both groups of ketchup bottles and asked to rate how asymmetric (7) to s.y~metric (1) they found the two groups on a 7-point scale. The results mdlcated that the participants rated the symmetric group lower on the scale (M = 2.21, a value significantly lower than the scale midpoint [4]; t(18) = -3.54, p < .002) and the asymmetric group higher on the scale (M = 4.89, a value significantly higher than the scale midpoint [4]; t(18) = 2.85, p < .01).


noncontagious group (M = 2.47; F(l, 74) = 11.11, p < .001). Those primed with high contagion demonstrated an enhanced effect. They indicated a greater preference to choose from the contagious group in the gain domain (M = 4.11) and a greater preference to choose from the noncontagious group in the loss domain (M = 1.86; F(1,76) = 95.76, p < .001). However, for those primed with low contagion the effect did not emerge since no difference appeared in preferences for the gain or the loss domain (M = 3.07 vs. M = 2.94; F(1,74) = .51, p> .47). Figure 4 graphs the moderation by contagion. The enhanced effect for participants primed with high contagion and the reduced effect for those primed with low contagion demonstrate the moderating role of contagion and provide further support for the proposed contagion account.

GENERAL DISCUSSION

This article demonstrates a new phenomenon that suggests that consumers prefer to choose from contagious groups (proximal, similar, or symmetric) in the gain domain and from noncontagious groups (distal, dissimilar, and asymmetric ~ in the loss domain. The proposed account using contagIOn theory suggests that groups in which the products are arranged proximally, similarly, or symmetrically appear more contagious and seem to facilitate the spread of both gain and loss. If anyone product in the contagious group has a gain or loss quality, then this quality is perceived to spread across the whole group such that all the products in the group start reflecting the gain or loss. However, a distal, dissimilar, or asymmetric group appears less contagious, and if one product in the group has a gain or loss quality, then this quality does not spread as pervasively across the whole group. Experiment 1 demonstrated the effect in a product choice domain. Experiment 2 explored the mediating influence of perceived contagion in a between-participant design. Finally, experiment 3 moderated the effect by priming high and low contagion to provide support for the proposed account.

Contagion theory has found its applications in several

FIGURE 4

MODERATION BY CONTAGION PRIME

pref symmetric 5

4

3

2

1 pref

asymmetric

4.11

control high contagion low contagion

IIIIgain [Jloss



ownloaded from



domains, including marketing. The findings of this article add to research on contagion theory in two ways. First, the experiments show the differential influence of contagiousness of a group (e.g., proximity, similarity, or symmetry) on product preferences in the domain of gains and losses. Perceived contagion causes people to choose an option that would not be predicted by normative probability theory. Second, the article demonstrates that a more contagious group appears to spread the qualities of gain and loss more than a less contagious group. Moreover, product grouping is an irrelevant cue that gets used by people to make product choices since the chance of gain or loss remains the same in both the contagious or the noncontagious group.

For a majority of consumer decisions taking place within the store, information is processed in a more bottom-up manner (Hoch and Deighton 1989), and most purchases are unplanned and spontaneous (Dreze, Hoch, and Purk 1994). In such a situation, displays and product groupings are some of the external cues that can influence consumer decisions. Therefore, findings of this research suggest that products with gains attached to them should be arranged to form contagious groups (e.g., more symmetrically), but products with some loss (e.g., defective items) should be placed in noncontagious (e.g., dissimilar) groups.

If the advertising message needs to increase consumers' perceived contagion, then contagious groups should be depicted. However, if the message aims to convey a feeling that things do not spread but remain isolated, then the advertisement might depict noncontagious groups. There are also implications for the insurance sector. For instance, consumers may buy more travel insurance, accident plans, or preventive plans if messages depict contagious groups, causing them to think that they are more susceptible to the vagaries of nature.

Encouraging people to eat healthful food has been an uphill task since the general perception is that healthful food is not tasty. The article suggests that if one item in a group is said to have a prominent good or bad quality, then it is likely that the entire group will be perceived to have that quality. Thus, if a menu represents a tasty food like ice cream near a group of healthful foods, the quality of tastiness is likely to spread across the whole group. This may increase consumers' likelihood of sampling from the healthful food section. Future studies could look into how the healthiness or tastiness quality can transfer to other food items kept as a group.

The results of the preliminary study and the thought protocols of experiment 1 suggest that participants were able to correctly assess the objective probability of gain or loss in both groups. However, intuitively, they prefer the contagious group in the gain domain and the noncontagious group in the loss domain. This inconsistent pattern alludes to a discrepancy between objective and intuitive probability assessments, a finding that follows from previous research, which suggests that intuitive probability assessments do not necessarily reflect objective probabilities (Windschitl2002). It would be worthwhile for future studies to examine this

81

discrepancy in probability assessments caused due to contagious groups. Another limitation of the studies is that product preferences were not in an actual store environment in which a consumer is exposed to several different types of contextual cues. Future studies should look into how contagious versus noncontagious product grouping in an actual store environment can influence purchase behavior.

REFERENCES

Argo, Jennifer J., Darren W. Dahl, and Andrea C. Morales (2006), "Consumer Contamination: How Consumers React to Products Touched by Others," Journal of Marketing, 70 (April), 81-94.

Baron, Reuben M. and David A. Kenny (1986), "The ModeratorMediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations," Journal of Personality and Social Psychology, 51 (December), 1173-82.

Crawford, Mathew T., Steven J. Sherman, and David L. Hamilton (2002), "Perceived Entitativity, Stereotype Formation, and the Interchangeability of Group Members," Journal of Personality and Social Psychology, 83 (November), 1076-94.

Dreze, Xavier, Stephen J. Hoch, and Mary E. Purk (1994), "Shelf Management and Space Elasticity," Journal of Retailing, 70 (Winter), 301-26.

Folkes, Valerie and Shashi Matta (2004), "The Effect of Package Shape on Consumers' Judgments of Product Volume: Attention as a Mental Contaminant," Journal of Consumer Research, 31 (September), 390-401.

Garretson, Judith A. and Scot Burton (2005), "The Role of Spokescharacters as Advertisement and Package Cues in Integrated Marketing Communications," Journal of Marketing, 69 (October), 118-32.

Hoch, Stephen J. and John Deighton (1989), "Managing What Consumers Learn from Experience," Journal of Marketing, 53 (April), 1-20.

Jacobs, Laurence, Charles Keown, Reginald Worthley, and Kyungn Ghymn (1991j,"Cross-Cultural Colour Comparisons: Global Marketers Beware!" International Marketing Review, 8 (3), 21-30.

Kahn, Barbara E., Mary Frances Luce, and Stephen M. Nowlis (2006), "Debiasing Insights from Process Tests," Journal of Consumer Research, 33 (June), 131-38.

Luce, Mary Frances (1998), "Choosing to Avoid: Coping with Negatively Emotion-Laden Consumer Decisions," Journal of Consumer Research, 24 (March), 409-33.

Morales, Andrea C. and Gavan Fitzsimons (2007), "Product Contagion: Changing Consumer Evaluations through Physical Contact with 'Disgusting' Products," Journal of Marketing Research, 44 (May), 272-83.

Priester, Joseph R. and Richard E. Petty (2003), "The Influence of Spokesperson Trustworthiness on Message Elaboration, Attitude Strength, and Advertising Effectiveness," Journal of Consumer Psychology, 13 (4), 408-21.

Rozin, Paul, Maureen Markwith, and Carol Nemeroff (1992), "Magical Contagion Beliefs and Fear of AIDS," Journal of Applied Social Psychology, 22 (July), 1081-92.

Rozin, Paul, Linda Millman, and Carol Nemeroff (1986), "Operation of the Laws of Sympathetic Magic in Disgust and Other



ownloaded from


82

Domains," Journal of Personality and Social Psychology, 50 (April), 703-12.

Rozin, Paul and Carol Nemeroff (2002), "Sympathetic Magical Thinking: The Contagion and Similarity 'Heuristics,'" in Heuristics and Biases: The Psychology of Intuitive Judgment, ed. Thomas Gilovich, Dale Griffin, and Daniel Kahneman, Cambridge: Cambridge University Press, 201-16.

Spencer-Rodgers, Julie, David L. Hamilton, and Steven J. Sherman (2007), "The Central Role of Entitativity in Stereotypes of


Social Categories and Task Groups," Journal of Personality and Social Psychology, 92 (March), 369-88.

Windschitl, Paul D. (2002), "Judging the Accuracy of a Likelihood Judgment: The Case of Smoking Risk," Journal of Behavioral Decision Making, 15 (January), 19-35.

Yzerbyt, Vincent Y., Anouk Rogier, and Susan T. Fiske (1998), "Group Entitativity and Social Attribution: On Translating Situational Constraints into Stereotypes," Personality and Social Psychology Bulletin, 24 (October), 1090-1104.



ownloaded from


influence of contagious versus noncontagious product

Documents