Introduction

Word of mouth communication (WOMC), or interpersonal communications about products, services, firms or brands in which the participants are not commercial sources (Bone, 1995), is a vital marketplace phenomenon that influences various categories of consumer behavior. The effects of WOMC are generally considered even more powerful than company-initiated messages for the reasons that: consumers perceive the sources through which it occurs to be credible (Richins, 1983); information is perceived to be custom-tailored (Wirtz & Chew 2002); and consumers do not question the motives of people who produce WOMC (Silverman, 1997). Because of its ubiquity in the marketplace and powerful effects on consumer behavior, WOMC has drawn attention from researchers since the 1960s (Wirtz & Chew, 2002), whose concerns can be largely categorized into the antecedents and consequences of WOMC.

Regarding the consequences of WOMC, Dichter (1966) reveals that positive WOMC can increase consumers’ intentions to purchase innovative products, as it reduces perceived risks of the products. Conversely, Holmes and Lett (1977) reveal that negative WOMC can deter consumers from considering a brand, as it damages the company’s reputation. In general, researchers (Arndt, 1967; Mizerski, 1982; Wright 1974) suggest that negative WOMC have stronger influences on the individual’s brand evaluation than positive WOMC. Recently, researchers pay increasing attention to the factors that moderate the effects of WOMC, including the manner in which information is presented (e.g., pallid versus vivid presentation), the availability of prior impressions of the brand (Herr, Kardes, & Kim 1991), consumers’ causal attributions regarding the underlying motives for WOMC they encounter (Laczniak, DeCarlo, & Ramaswami, 2001), and consumers’ perception of justice (Blodgett, Granbois, & Walters, 1993).

On the other hand, given the status of WOMC as “a dominant force in the marketplace” (Mangold, Miller, & Brockway, 1999, p.73), marketers and researchers take interest in the factors that can influence WOMC, reflecting their need to have more control over the process in a proactive manner. For example, a group of researchers (Heckman & Guskey, 1998) reveal that satisfaction/dissatisfaction with purchase experiences have influences on WOMC. Wirtz and Chew (2002) find that the effects of consumer satisfaction on WOMC are moderated by incentives, customer-company tie strength and consumers’ deal proneness, implying that a company can proactively manage WOMC by providing incentives and strengthening its tie with customers. Consumers’ identification with a company is also suggested as an antecedent of positive WOMC, with its effects moderated by consumers’ commitment to their relationship with the company (Brown, et. al., 2005). Engel and others (1990) and Rogers (1995) also reveal that consumers’ dispositional characteristics can also be an antecedent of WOMC, by showing that opinion leaders tend to generate more WOMC then non-leaders.

The current literature on WOMC has a number of limitations. First, the literature has largely failed to incorporate the perspective that a receiver of messages in one WOMC instance could become a sender in an ensuing WOMC situation, with information rippling out through chain-reactions of multiple WOMC situations embedded in the web of social networks. The dynamic, “chain-reaction” view of WOMC is distinct from the way advertising is typically modeled, in which an advertiser is pinpointed as a sender of messages and viewers as receivers. The perspective is a logical corollary and a group communication extension of the early “interactive model” of communication (e.g., Schramm, 1954), which conceptualizes communication as an interactive, on-going process, with participants being receivers and senders of messages alternately (e.g., Schramm, 1954).

Second, the current literature does not explore the potential role of the dynamics of small group communication, whose major constructs include group norms, group identity, peer pressure, need for social approval etc., in WOMC processes. This deletion is notable, because WOMC often takes place in a group communication setting, and perceptions of group norms are known to have influences on both consumer behavior (Batra, Homer, & Kahle, 2001; Meyer & Anderson, 2000; Venkatesan, 1966) and communicative patterns in small group settings (Neuwirth & Frederick, 2004).

The current study aims to address the above two limitations. First, employing the dynamic, chain-reaction view of WOMC and theories of group norm, we suggest that consumers’ exposure to WOMC regarding a consumption entity will affect their perceived group norm about the entity as well as their own attitudes towards it, which will further influence their communicative patterns in ensuing WOMC situations. In particular, drawing on the literature on the antecedents and consequences of WOMC and group norms, we predict that college students’ exposure to their peers’ conversation about one specific area of consumption, the idea of going to Cancun, Mexico to spend an imminent spring break, will influence their perceptions of the average student’s attitudes towards the idea, as well as their own attitudes to it. We further expect that, because of these two kinds of influence, students’ exposure to the initial WOMC situation will affect their own WOMC patterns regarding the same issue in a projective role playing task resembling a group communication setting.

The current project focuses on the issue of choosing a spring break destination, more specifically Cancun’s desirability as a spring break destination, as a research setting for a number of reasons. First, we assume that choosing a spring break destination is more susceptible to group norm influences than other consumption categories, because decision making in choosing a spring break destination is to a large part a consensual process, where students need to seek approval from their friends. Fisher (1993) reports that students consider choosing where to go for a spring break, together with changing hairstyle and choosing a brand of athletic shoes, perfume and music, to be a consumption decision greatly influenced by others. Second, because the data were collected just a week before a spring break, we assume that the topic was relevant to the majority of students, thus increasing their motivation for participation and reduce maturation effects. Third, we choose Cancun among many spring break destinations under the assumption that it is generally known as a possible spring break place, but not many students have actually been to the place for spring break, enabling the stimulus to function effectively as a source attitude formation and group norm perception.

Literature Review

Norms are indispensable to understanding various aspects of human behavior (Campbell, 1954). Although most Americans think of themselves as independent, consciously or unconsciously, they are subject to influences from group patterns and expectations in a variety of contexts, including consumer behavior (Batra, Homer, & Kahle, 2001; Meyer & Anderson, 2000; Venkatesan, 1966). Potential reasons for these influences are many, including affiliation needs (Festinger, 1954) and pressures toward group conformity (Asch, 1951, 1952). Norms, and the mechanisms through which they influence human behavior, can be either descriptive or injunctive. Descriptive norms can influence people by describing what is typical and or normal, and functions as evidence regarding effective and adaptive action for a particular situation, whereas injunctive norms signify what the morally approved/disapproved conduct is in a given situation (Cialdini, Kallgren, & Reno, 1991; Cialdini, Reno, & Kallgren, 1990).

Where do people catch group norms? Cialdini, Reno and Kallgren (1990), through a field experimental setting, show that observing a confederate dropping a bag from a fast food restaurant into an otherwise clean environment can make people more aware of the descriptive norm that people do not litter in the given area. Reno, Cialdini and Kallgren (1993), through a similar research setting, show that observing a confederate pick up garbage from a dirty environment can turn on people’s awareness of the injunctive norm that littering is socially disapproved. Social contact with various constituents, including parents, faculty and peers can all be a source through which perceptions of group norms emerge (Perkins, 2002, p.164). We expect that college students’ exposure to conversation among their peers about the desirability of Cancun as a spring break destination will influence their perception of the descriptive norm among their peers, particularly the average student’s attitudes towards the idea of going to Cancun to spend the upcoming spring break.

H1a: Students exposed to largely unfavorable comments from their peers about Cancun’s desirability as a spring break destination will perceive the average college student’s attitudes towards Cancun to be more negative.

H1b: Students’ exposed to largely unfavorable comments from their peers about Cancun’s desirability as a spring break destination will have more negative attitudes towards Cancun.

The literature from spiral of silence theory indicate that individuals form their perception of the majority opinion through multiple sources, typically mass media, and the perceived majority opinion discourages them from expressing their opinion when it deviates from the majority opinion (Noelle-Neumann, 1993). Can a similar phenomenon occur in a WOMC situation, particularly students’ conversation in a classroom setting regarding the issue of choosing spring break destination? We assume it is, because there is no reason to believe that fear of social isolation, which is the psychological rationale for predicting that people with minority opinions tend not to publicly express their opinion, is confined to political issues. The Asch (1951; 1956) experiments show that experimental manipulation, namely having confederates present incorrect answers in the task of comparing line lengths, can affect people’s willingness to express their correct opinion. Neuwirth and Frederick (2004) show that perceptions of majority attitudes predict people’s willingness to ask their acquaintance to let them drive when the acquaintance has been drinking. Neither of the above two settings is related to political issues. Therefore, we assume that the power of perceived norms to influence people’s opinion expression applies to WOMC, especially for consumption categories and situations for which the need to achieve social and peer approval is greater. As Fisher (1993) reveals that college students consider choosing a spring break place as one such area of consumption decision making, we make the following predictions.

H2a: Students exposed to largely negative comments from their peers about Cancun’s desirability as a spring break destination will mention a smaller number of favorable comments about Cancun in a projective role playing task.

H2b: Students exposed to largely negative comments from their peers about Cancun’s desirability as a spring break destination will mention a greater number of unfavorable comments about Cancun in a projective role playing task.

If H2a and H2b are supported, the alternative explanation for the association between exposure and number of favorable/unfavorable comments will be that exposure to unfavorable comments about Cancun negatively influences the students’ own attitudes, which result in a greater number of unfavorable comments and smaller number of favorable comments in the role playing task. Therefore, to further test the internal validity of the predictions made in H2a and H2b, we test the hypotheses that the perceived attitudes of the average student will be significantly associated with the number of positive and negative words even after controlling for the students’ own attitudes towards the idea of going to Cancun for spring break.

H3a: After controlling for their own attitudes, students’ perception of the average student’s attitudes towards Cancun as a spring break destination will be positively associated with the number of favorable comments of Cancun they mention in a hypothetical role playing task.

H3b: After controlling for their own attitudes, students’ perception of the average student’s attitudes towards Cancun as a spring break destination will be negatively associated with the number of unfavorable aspects mentioned of Cancun in a hypothetical role playing task.

Method

Design

To test H1a, and H1b, one-way ANOVA design was used, with valence of the conversation (positive versus split versus no conversation) as the independent variable and the subjects’ perception of the average undergraduate’s attitudes towards Cancun as a spring break destination as the dependent variable. To test H2a and H2b, a one-way ANOVA design was used, with the valence of conversation (positive versus split versus no conversation) as the independent variable. The dependent variable varied according to each hypothesis. To test H3a and H3b, multiple regression models were built, with gender, age, ethnicity and the subjects’ own attitudes towards Cancun as a spring break destination as control variables and perception of the average undergraduate’s attitudes as the independent variable. Dependent variables varied along the predictions made in the three hypotheses. Two versions of audio-taped conversation about Cancun were designed as experimental stimuli to manipulate the valence of conversation. The audio-tapes of the two conversation versions and their transcripts were used as stimuli concurrently. Other variables were measured in a survey session following the experiment.

Subjects & Procedure

From two elementary advertising courses at a large Southern state university, 109 female and 29 male undergraduate students participated in the experiment in return for minor extra credit. The experiment was conducted a week before a spring break season. The class curriculum did not include topics that might have sensitized the students to the purpose of the current research. Students signed up for participation in class, and they were invited to a laboratory where the experiment was conducted. The informed consent process was secured first. Then, an experimenter delivered general instructions about the study. The subjects were directed to read the transcript and listen to the conversation carefully. The experimenter stated:

PLEASE LISTEN TO THIS CONVERSATION VERY CAREFULLY. After you listen to this tape, we will ask you a few questions about the place the students are talking about and the people who have this conversation. We have provided a transcript of this conversation to help you better follow the conversation.

The subjects were randomly assigned to one of the three conversation valence conditions, that is, negative valence, split valence, and no conversation that served as a control group. Following exposure, they were first familiarized with, and then completed a questionnaire containing, among others, their own attitudes towards Cancun as a spring break destination, their perceptions of the average undergraduate’s attitudes, and a projective role playing task. The projective task was designed to tap into their WOMC in a hypothetical situation of the kind that students would typically encounter in class. Upon completion of the questionnaire, the subjects were debriefed and thanked.

The Stimulus Materials

Two hypothetical versions of audio-taped conversation regarding the issue of spring break destinations were created by having a number of undergraduates enact the provided transcripts. To increase experimental reality, the materials were developed based on ten graduate students’ responses to a survey containing open-ended questions such as “What are the things that make a place a good spring break place?” and “What are the issues you care about most when you choose a spring break place?” Further, regarding the origin of the materials, the experimenter stated:

A travel agency has been conducting focus groups with students to understand where they want to spend the coming spring break. Focus groups are routinely tape-recorded to be used as research data. An interesting conversation took place among participants during a break in a session. Because the conversation was taped during a break, the tape had a poor quality. So you will listen to a reenacted version of the audio-taped conversation. To help your understanding, we also provide you with a transcript. You may read it while you listen to the conversation.

The two versions were created to manipulate the overall valence of the participants’ views about Cancun as a spring break destination. In one version (See Appendix 1), named “negative valence,” the students expressed overwhelmingly unfavorable views about Cancun as a spring break place. In the other version (See Appendix 2), named “split valence,” the first student expressed exactly the same unfavorable view, but the other students responded with moderately to strongly favorable views. The two versions had similar word counts, with 286 words in the negative valence version and 285 in the split version.

Measures

Demographics

The subjects self-reported their age, gender, socio-economic status (SES) and ethnic backgrounds. Self-reported SES was measured by asking the subjects to check which of the five categories, including working class, lower middle class, middle class, upper middle class and upper class, they considered their family to be. Age was asked with an open-ended question.

Three Attitude Measures

The subjects reported three categories of attitudes towards the idea of going to Cancun for the coming spring break, namely, their own attitudes, their perceived attitudes of the students in the tape-recorded conversation, and their perceived attitudes of the average college student. They reported each of the three attitude categories by checking a six-item semantic differential scale that ranged between one (negative, unfavorable, unlikable, undesirable, unenjoyable, and unappealing) to seven (positive, favorable, desirable, enjoyable, and appealing). Instructions varied according to the attitude category. For example, the subjects reported their perceived attitudes of the average student by completing the incomplete sentence, “In your thinking, the average _________ (the college name) student’s attitudes towards the idea of going to Cancun for the coming spring break would be…” by checking the six-item scale.

Word of Mouth Variables

Number of favorable/unfavorable comments generated was measured through a projective role-playing technique. Rooted in psychodynamics, marketing and communication researchers use projective techniques when respondents are expected to be reluctant or unwilling to reveal their ideas because of the sensitive nature of the topic or when they need to measure the motives the respondents themselves may not be aware of (Kassarjian, 1974; Will, Eadie, & MacAskill, 1996). More relevant to the current project, Wirtz and Chew (2002) used a role playing technique to tap into various aspects of WOMC, suggesting that using the technique “discourages social desirability effects ... and reduces problems involving individual differences in reactions and in personal circumstances to the research contexts” (p. 146).

Considering that American culture places emphasis on independence and self-determination (Hofstede, 1980), and therefore assuming that most people do not enjoy the idea that they may be influenced by others’ opinions in their consumption decision making, an approach similar to Wirtz and Chew (2002) was developed for the current project to bypass the subjects’ social desirability bias and lower their motives for defending their self-image. The subjects were instructed to imagine the following situation. The underlying assumption is that people tend to project and reveal their inner thoughts and motives in interpreting and responding to a projective stimulus.

Jennifer is a student at _________ (the college name) and the spring break is several weeks away. She has been thinking about some places, but hasn’t yet decided where to go. One day she comes to class and finds her acquaintances in the middle of conversation about spring break places. Soon one of them says, “Does anybody go to Cancun? I know it could be a place to go, but really don’t know if it’s good for spring break.” The person continues to ask, “Hey Jennifer, what do you think?

The subjects were then asked to write what Jennifer would say in response. Their responses were coded by the two authors for the number of favorable and unfavorable comments.

Results

Preliminary Findings

Demographics

Of the 138 subjects, the majority (79 percent) were females. Overall, they ranged in age from 18 to 25 (M = 20.30, SD = 1.25). Their ethnic backgrounds were mostly whites (76 percent), followed by Hispanics (14 percent), Asian Americans (4 percent), and African-Americans (3 percent). Most of the participants were from middle class (25 percent) or upper middle class (60 percent) families. Only 2.9 percent reported that they were from working class families.

Attitude Measures

On the whole, the subjects’ own attitudes toward Cancun as a spring break destination were above the mid-point, with a mean of 5.01 on a seven-point scale (SD = 1.70). The six-items were internally consistent (α = .98). The perceived average students’ attitudes toward Cancun were 5.05 on the average (SD = 1.51). The items also exhibited high internal consistency (α = .98). Perceived attitudes of the conversation participants were 3.63 (SD = 1.48), slightly below the mid-point on the average. The scale was internally consistent (α = .97).

Word of Mouth Measures

The two authors content-analyzed the subjects’ open-ended responses to look for references to favorable or unfavorable aspects about Cancun as a spring break destination.

After getting familiarized with the coding guidelines developed by the first author, the two authors independently coded 35 of the 138 total responses. Examples of favorable comments included “nice beach,” “warm weather,” and “my parents’ favorite place.” Examples of unfavorable comments included “too far away,” “too expensive,” and “only people who cannot legally drink in the U.S would go there.” Both of the measures were internally consistent, evidenced by high correlations between the two coders for the number of unfavorable comments (r = .83, p < .01), and for the number of favorable comments (r = .94, p < .05). The subjects made .90 (SD = 1.24) favorable comments and 1.07 (SD = 1.15) favorable comments on the average.

Manipulation Check

Predictions made in H1, H2a and H2b were based on the assumption that the subjects would infer the average student’s attitudes from information presented in the two versions of conversation, especially the varying degrees of their peers’ favorable/unfavorable remarks about Cancun’s desirability as a spring break destination. The assumption could be valid only when the three groups differ in their perceived attitudes of the conversation participants. An independent sample t-test was run, with the valance of conversation as independent variable and perceived attitudes of the participants in the conversation as the dependent. Results indicate that manipulation was successful, as the mean of the negative valence group was 2.20 lower than the split valence group on a seven-point scale. The difference was statistically significant (p < .001).

Testing the Hypotheses

Hypothesis 1

H1 predicted that exposure to largely unfavorable comments from their peers about Cancun’s desirability as a spring break destination will result in more negative perceived attitudes of the average college student towards Cancun. A one-way ANOVA test supported this prediction [F(2, 129) = 7.96, p < .001] (See Table 1 & 2).

The means of the negative, split, and control groups were 4.42, 5.46 and 5.43, respectively. Bonferroni post hoc multiple comparisons showed that the negative valence group significantly differed from the other two, but the difference between the split and control groups was minimal. 

Hypotheses 2a and 2b 

H2a and H2b predicted that exposure to largely unfavorable comments from their peers about Cancun’s desirability as a spring break destination would result in their mentioning a smaller number of favorable aspects, and a greater number of negative aspects of Cancun as a spring break destination. Two one-way ANOVA tests revealed that both H2a [F(2, 133) = 3.61, p < .05] and H2b [F(2, 135) = p < .001] were supported (See Table 2).

Concerning the number of favorable aspects mentioned, the negative, split and control valence groups on the average exhibited .77, 1.25 and 1.40 respectively. However, Bonferroni post hoc multiple mean comparisons indicated that control group did not differ from neither of the two treatment groups, although the difference between the negative valence and control groups appeared large. This might have occurred because the control group had a smaller sample size (n = 20) and therefore had an inflated standard error, compared to the other two.

Concerning the number of unfavorable aspects mentioned, the negative, split control valence groups on the average recorded 1.36, .66 and .35 respectively. Bonferroni post hoc multiple comparisons indicated the negative valence groups had statistically significant mean differences from the other two groups. In sync with the subjects’ overall positive attitudes towards Cancun, the control group mentioned the least negative aspects on the average (M = 5.01, SD = 1.70).

Hypotheses 3a and 3b

It was predicted that the perception of the average college student’s opinion would be negatively associated with the number of favorable aspects referred to in the role playing (H3a) and positively associated with the number of unfavorable aspects mentioned. Two hierarchical multiple regression models were built to test the hypotheses. In both of the models, demographics (gender, age, SES and ethnic background) were entered as the first step, followed by the subjects’ own attitudes towards Cancun as the second step. Perceived average student’s attitudes were entered last. In the complete model, the perceived average student’s attitudes significantly predicted the number of favorable aspects mentioned (β = .18, p < .05), supporting H3a (See Table 3).

However, H3b was not supported (See Table 3), as perceived attitudes did not significantly predict the number of unfavorable aspects (β = .16, p = .077), although the relation was negative as H3b predicted (See Table 4).

The subjects’ own attitudes were significant in the model for H3a (r = .21, p < .05) and H3b (r = -.32, p < .001). If the subjects’ own attitudes had not been controlled for, the perceived average student’s attitudes would have been significant for both H3a (r = .21, p < .05) and H3b (r = -.20, p < .05).

Discussion

Based on the literature on the influences of perceived group norms on human behavior, the current study reveals a mechanism through which individuals’ exposure to WOMC could affect their own communication patterns in a later WOMC situation. Particularly, by means of an experiment with 138 undergraduate students, this study reveals that students’ exposure to largely unfavorable versus mixed comments from their peers about the desirability of Cancun as a spring break destination results in differences in their perceptions of the average student’s attitudes towards Cancun, and the amount of unfavorable aspects they mention in a later hypothetical, projective WOMC situation. This study confirmed from a multiple regression test that even when the students’ own attitudes, together with demographic variables, are controlled for, the perceived average student’s attitudes still significantly predict the amount of favorable aspects mentioned, but not unfavorable aspects mentioned.

The findings have a number of practical implications for the practice of city marketing, considering that the effects of WOMC matter more for services marketing (Mangold, Miller, & Brockway, 1999). Fisher (1993) suggests that choosing a spring break place, together with wearing athletic shoes, choosing a music genre and wearing perfume, is a consumption decision making area where the approval of peers is very important. The situation in which peer approval is perceived and achieved can most aptly be conceptualized as one in which people get involved in small group communication and sense the majority opinion and approval based on the favorable and/or unfavorable comments from others about a given object. By showing that individuals’ former WOMC experience can affect their own future WOMC behavior through its effects on the individuals’ perceived group norms, the findings of the current project imply that it is imperative for city marketing practitioners to develop ways to influence potential visitors’ perceptions of the group norms.

Theoretically, the findings of this study imply that a process similar to “group think,” “pluralistic ignorance,” or “spiral of silence,” which have been mostly explored in the domain of political communication or public opinion research, can occur in at least some consumption decision making areas where the approval from peers and/or other reference groups matter. We interpret this as meaning that the need for social approval, or fear of disapproval and isolation, is not confined to opinion expression about political issues.

Limitations & Future Research

This study has a number of limitations. First, the sample size of the control group was smaller than the other two groups. This seems to be the reason that the mean difference between control group and the other two groups in the number of positive aspects mentioned is fairly large but statistically insignificant. However, a more serious limitation of this study is that results of the multiple regression test implied that the influence of the subjects’ own attitudes was greater than that of the perceived average student’s attitudes, which resulted in failure to support H3b. However, this is a limitation as well as an opportunity for further research, because it suggests that exposure to WOMC may affect individuals’ later WOMC behavior through affecting both their own attitudes and perceptions of group norm. Currently, the literature does not exist on the effects of exposure to WOMC on consumers’ attitudes towards a product or service that they have never used.

In addition, it will be theoretically and practically meaningful to further test if the findings observed in this project carry over to different consumption decision making categories, particularly the ones stipulated by Fisher (1993) as areas where perceived social approval matters. If the carry-over occurs, it will imply that it is imperative for marketing communications practitioners to influence WOMC in their favor at an early stage, because the favorable/unfavorable impression expressed in early WOMC situations can carry over to ensuing WOMC situations. Especially when it comes to a new brand or product category, harnessing early adopter’s WOMC seems to be particularly important, because their stories can ripple out through multiple social networks and WOMC situations. It will be also meaningful to see how the effects differ along consumers’ dispositional characteristics, such as opinion leadership and seeking.