The Effect of Anonymity and Norms on Language Intensity in Consumer Reviews
Elizabeth Tompkins, 10426388 Master’s Thesis
University of Amsterdam Graduate School of Communication Master’s Programme Communication Science
Supervisor: Peeter Verlegh January 31st, 2014
Abstract
Online consumer reviews are an important form of word-of-mouth (WOM). However, online reviews often include intense and aggressive language which can make it unpleasant for individuals to participate in WOM and can affect the persuasiveness of reviews. This research explores the effect of anonymity and social norms (operationalized as other consumer
reviews) on the intensity of the language used in online consumer reviews. A 2 x 2
experimental design was employed to test these effects. Two hundred eighty-four participants completed partially written reviews with high or low intensity words which were pre-tested with an independent sample. The participants were randomly assigned to one of the four conditions: intense normative cue, identified-intense normative cue, anonymous-moderate normative cue, identified-anonymous-moderate normative cue. Results of an ANCOVA showed no main effect of anonymity or norms on review intensity; a significant interaction effect of anonymity and norms on review intensity was found. More research is needed to fully understand the effect of anonymity and norms on language intensity in consumer reviews.
The Effect of Anonymity and Norms on Language Intensity in Consumer Reviews In today’s connected world, many people rely on online consumer reviews (referred to as electronic word-of-mouth (eWOM)), for information on products and services. Online reviews can be found across the internet on social media, branded websites, blogs, and consumer opinion platforms such as epinions.com or yelp.com (Hennig-Thurau, Walsh, & Walsh, 2003). Online consumer reviews have the potential to reach an immense amount of people and like traditional word-of-mouth; they can greatly influence consumer decisions (Hennig-Thurau, Gwinner, Walsh, & Gremler, 2004).
However, the use of aggressive, intense and socially disagreeable language is rampant in online communication, including eWOM. Such language can make it unpleasant for users to participate in online forums, which may cause some consumer voices to go unheard. Also, intense language – which reflects the degree by which a speaker’s attitude deviates from neutrality (Bradac, Bowers, & Courtright, 1979) – may affect the persuasiveness of consumer reviews (Bradac et al., 1979; Burgoon, Jones, & Stewart, 1975). This may make it hard for consumers to make unbiased decisions about brands and creates an added challenge for brand managers who strive to maintain a consistent brand image across all brand communication, including WOM.
In recent years, researchers have advanced several explanations for the socially unacceptable language that is often seen across the internet. One widely cited cause is anonymity, which has been linked to flaming – an extreme form of intense online
communication (Alonzo & Aiken, 2004; Christopherson, 2007; Reinig, Briggs, Brandt, & Nunamaker, 1997; Suler, 2004). Another possibility for the prolific use of intense language online is the influence of social norms (Lea & Spears, 1991). In computer mediated
communication (CMC) language often acts as a social cue and may influence the language an individual uses in their own messages (Walther, 2007).
Although electronic word-of-mouth has received a lot of attention among academics and practitioners alike, the effect of anonymity and social norms on language intensity in online consumer reviews has yet to be directly explored by the scientific community. Therefore, the purpose of this research is to understand to what extent being anonymous or identified when posting a negative online review influences the intensity of the language used in the review and to what extent normative cues in the form of other reviews influence the intensity of the language used in an individual’s review and how this varies for anonymous and identified individuals.
Theoretical Background
The Effect of Anonymity on Language Intensity in Online Reviews
On an online consumer review platform, individuals may consciously adapt the language they use in their reviews in order to manage their impressions. Goffman’s theories of identity and social performance suggest that through self-presentation people continually adjust their self-expressive behavior in order to maintain a desired impression in social situations, a process referred to as impression management (Schau & Gilly, 2003). Some people are more inclined to manage their impression and are better at self-management than others. Self-monitoring theory asserts that individuals must have both the desire and the skill to adjust their behaviour to maintain their desired social impressions (Snyder, 1974). Offline, individuals manage their impressions through their facial expressions, body language and physical appearance, whereas in CMC, message content and language are commonly used for impression management (Walther, 2007).
Anonymity – which can be defined as a state in which a person is not identifiable (Marx, 1999) – takes away the need for impression management and allows individuals to express their true mind (Joinson, 2001). It has been argued that anonymity leads to increased self-disclosure online (Joinson, 2001). A study of communication on weblogs found that
anonymity is positively related to self-disclosure (Qian & Scott, 2007). Other research, described in detail below, also links anonymity to deindividuation and decreased inhibition (Postmes & Spears, 1998). This suggests that when people feel anonymous online they are more likely to write what they really think and feel and are less likely to monitor their behavior.
Language intensity has often been described as a measure of affect; an individual’s emotions are shown in the intensity of the language used in his/her messages (Rogan, 1995). Research has shown that attitude influences language intensity (Burgoon & Miller, 1971). When people feel strongly about something (e.g. a negative experience at a restaurant) their emotions manifest themselves in the language used in their messages, making the language in the messages more intense. Language intensity is not only a factor that pertains to the sender of a message; the intensity of the language used in a message can also make messages more or less persuasive to receivers (Bradac et al., 1979; Burgoon et al., 1975).
Theory and research indicates that online anonymity leads to reduced impression management. This can in turn lead to increased self-disclosure of one’s opinions or emotions. Thus if an individual has a negative experience with a brand and naturally feels angry, upset, or aggressive, the language used in the review should reflect those sentiments when the individual posts the review anonymously. Alternatively, when the individual is identified as the writer of the review he/she will be more likely to adjust the language used in the review to gain/maintain social approval. With this in mind, this research proposes the following
hypothesis:
H1: The language in reviews by anonymous individuals will be more intense than in reviews by identified individuals.
The influence of social norms on people’s behavior can be described as a desire to conform to the expectations of others (Deutsch & Gerard, 1955). To determine what behavior is desired in specific social situations, people use normative cues which provide insight into a group’s social norms. Social norms can be broken into two categories: injunctive norms and descriptive norms. Injunctive norms refer to generally accepted social values and morals, whereas descriptive norms describe what the majority of individuals actually do in certain situations (Cialdini, Reno, & Kallgren, 1990).
Some researchers have suggested that the nature of CMC reduces normative influence due to a reduction in social cues (Kiesler, Siegel, & McGuire, 1984). However, other research has also shown very pro-social behaviour in CMC (Joinson, 2001). While it is true that CMC lacks the extensive normative cues that are found in face-to-face interaction, CMC still provides normative cues in the form of message content and language (Walther, 2007). Furthermore, it has even been suggested that the lack of social cues found in CMC may increase the impact of social norms (Lea & Spears, 1991).
Online and offline the influence of norms can vary greatly by situation and person. Social desirability theory argues that some people have a high need for social approval whereas other individuals have a low need for social approval (Snyder, 1974). Individuals who have a high need for social approval are more motivated to conform to group norms than low need for approval individuals. Norms are also situation specific. A study on social norms and littering found that normative influence varied by norm type (injunctive or descriptive) and the situation in which the normative cue was observed (i.e. whether the normative cue was made salient) (Cialdini et al., 1990).
Latane’s dynamic social impact theory asserts that in a given social space (such as an online consumer review platform) individuals will be more likely to conform to the attitudes, beliefs and behaviors of the majority within that social space than to attitudes, beliefs and
behaviors of the minority or less proximate individuals (Cialdini & Goldstein, 2004). If on a given review platform, the majority of reviews contain highly intense language, the use of intense language may be the norm for that platform. Given Latane’s theory, people will be more likely to conform to the platform’s group norm of using intense (or moderate) language than follow a general social norm of using moderate language in communication. With this in mind the following is hypothesized:
H2: Normative cues (i.e. other reviews) will influence the intensity of the language used in reviews such that intense (moderate) reviews by others will result in more intense (moderate) individual reviews.
The Effect of Norms and Anonymity on Language Intensity in Online Reviews A widely studied effect of anonymity is deindividuation, which refers to a psychological state of decreased self-evaluation and decreased evaluation apprehension (Postmes & Spears, 1998). Deindividuation theory suggests that being unidentified and therefore unaccountable, releases people from inner restraints and allows them behave in ways that would normally be inhibited by established social norms (Postmes & Spears, 1998). Some argue that deindividuation results in aggressive, emotional and intense anti-social behavior because people naturally repress these feelings and anonymity allows people to show their true colors (Reicher, Spears, & Postmes, 1995). Online and in CMC,
deindividuation has been linked to flaming in various situations. A study of anonymity in online classrooms found that deindividuation, operationalized by anonymity, increased instances of aggressive communication (Reinig et al., 1997).
The majority of research on anonymity and deindividuation asserts that
deindividuation results in anti-normative behavior because deindividuation frees individuals from the influence of social norms. However, empirical support for these effects remains mixed. One explanation for this may be that most deindividuation studies have
conceptualized anti-normative behavior in terms of general social norms, while ignoring possible context or situation specific norms (Postmes & Spears, 1998). A meta-analysis of 60 deindividuation studies found strong support for the influence of situational norms on
anonymous and deindividuated behavior (Postmes & Spears, 1998). Since social norms are highly situational (Cialdini et al., 1990), it is possible that deindividuation actually results in normative behavior that is in line with the situational norms but out of line with broadly accepted social norms.
Some academics have tried to determine whether deindividuation causes anti-normative behavior or whether anonymity can actually result in conformity to situation specific norms. An experiment by Johnson and Downing (1979) looked at how anonymity and normative cues in the form of costumes interact to promote anti-social (i.e. negative) or pro-social (i.e. positive) behavior. The results showed that for both identified and anonymous individuals, positive normative cues (nurses’ uniforms) resulted in more pro-social behavior while negative normative cues (Klu Klux Klan robes) resulted in more anti-social behavior (Johnson & Downing, 1979). Furthermore, the results showed that anonymous participants were actually more responsive to normative cues than identified participants (Johnson & Downing, 1979).
The Social Identity explanation of Deindividuation Effects (SIDE) was devised to understand the effects of norms on deindividuated individuals. According to the SIDE, when individuals are anonymous and no group or social identity is available, individuals will act in accordance with their internal standards. When social identity rather than personal identity is salient, the model argues that anonymity leads to a greater influence of situational group norms (Joinson, 2001). The SIDE also stipulates that social identity can be salient without actual group interaction and with only the potential for group formation (Reicher et al., 1995).
On online consumer review platforms where individuals interact with other people and form a group of consumer reviewers, social identity may be more salient than personal identity. Thus, anonymous individuals should be more sensitive to the norms on the online platform than identified individuals. Based on the SIDE and prior research on anonymity and norms, the following hypothesis is proposed:
H3: The effect of normative cues on the intensity of the language used in reviews will be stronger for anonymous individuals than for identified individuals.
Conceptual model:
Method Design
This research was conducted to understand to what extent anonymity and norms affect the intensity of the language used in online consumer reviews. The study also considered a possible interaction effect of anonymity and norms on the intensity of the language used in reviews. A 2 (anonymity: anonymous vs. identified) x 2 (normative cue: intense other reviews vs. moderate other reviews) between subjects experimental design was conducted to test these effects.
Sample
Due to time and budget constraints a non-probability sample using a snowball
technique was used in this research. As a part of this research, 362 participants were recruited
Normative cue: Intense/Moderate other reviews Review Intensity (DV) Anonymity (IV): Anonymous/Identified
to take part in the experiment. Participants were invited to take part in the study by a link sent via a social networking site (i.e. Facebook). Data collection online via Facebook was used to ensure participants had a Facebook profile (which was part of the anonymity manipulation) and because it was assumed that people using Facebook would be internet savvy and thus be familiar with online reviews. Of the 362 participants who started the survey, 78 failed to give their informed to consent to take part in the survey. These results were removed from the main analysis. In total, 284 results were retained for this research (64.7% female; age: M = 31.8, SD = 10.9, range = 54). Participation was completely voluntary and without incentive. All participants’ results remained anonymous.
Procedure
The website Qualtrics was used to generate the survey for this research. By clicking the survey link participants were directed to the survey website and randomly assigned to one of the four following conditions: anonymous-intense normative cue, identified-intense
normative cue, anonymous-moderate normative cue, identified-moderate normative cue. The survey started with an introduction which thanked the participants for their time and
explained the study. Participants were told that the study was about online reviews. In order to proceed to the main part of the survey, participants were required to give their informed consent.
Following the introduction, the participants were asked to imagine themselves in a scenario involving a bad experience with a brand. In order to reduce the influence of previous attitudes towards specific brands and to avoid the possibility of a brand being unknown by participants, a specific brand name was never given. Instead, the scenario described a bad experience with an unnamed restaurant. (For complete scenario see appendix.) The participants were then asked to write an online review about the restaurant. Following the method used in prior research on intense language (Burgoon & Miller, 1971) and prior
research on consumer reviews (Moore, 2012); the participants were provided with a
completed review with several blank spaces. Participants were instructed to choose a word to fill in each blank space. While completing this task the participants were exposed to two other negative reviews which contained either intense or moderate language. To manipulate anonymity the participants were either told to imagine their review would be posted with their Facebook profile name and photo or that their review would be posted anonymously. Participants were asked to imagine their review would be posted with a Facebook profile name and photo in the identified condition because social networking sites connect users to ‘real life’ family, friends, acquaintances, and institutions which reduce users’ anonymity online (Zhao, Grasmuck, & Martin, 2008).
Once the review was completed the participants were asked a number of demographic questions and other questions such as how negative the restaurant scenario was perceived to be, how much the participants noticed the normative cues and how much they perceived them to be moderate or intense. Missing answers were requested before participants could continue the survey. At the end of the survey participants were thanked for their time, debriefed and provided with a space to provide feedback or questions about the experiment.
Stimuli Material
Four negative online reviews were written from different fictional reviewers to be used as normative cues. Two reviews used intense language while the other two used moderate language. The pairs of moderate reviews and intense reviews were identical with the exception of the intense or moderate words. All the reviews were pre-tested with an independent sample. (See appendix for the reviews.)
Pretest
A pretest among 21 participants (65% female, age: M = 25.35, SD = 6.76) was conducted to ensure that the experimental stimuli had the anticipated effect. To measure the
intensity of the other consumer reviews, the participants were asked to read each review and rate the intensity of the review on a 5-point Likert scale (1-no intensity to 5-extermely high
intensity)(Rogan, 1995). A one-way within subjects ANOVA showed a significant difference
between reviews, Wilks’ Lambda = .133, F (3, 18) = 39.28, p < .001. A Bonferroni post hoc test was used to make comparisons between the four reviews. Pairwise comparisons were not statistically significant between the two intense reviews (p = .81) and the two moderate reviews (p = 1.00). However, comparisons did show a significant difference between the intense and moderate reviews (p < .001) and indicate that the intense reviews (MIntenseReview1
= 4.29, SE = .12, MIntenseReview2 = 4 .52, SE = .13) are significantly more intense than the moderate reviews (MModerateReivew1 = 2.62, SE = .15, MModerateReivew2 = 2.81, SE = .20). Therefore, the manipulation of the normative cues was deemed effective.
Rogan’s (1995) intensity scale ranging from (1-no intensity to 5-extermely high
intensity) was also used to determine an intensity level (low/high) of each of the words that
would be used in the main experiment. A one-way within subjects ANOVA showed a significant difference between words, Wilks’ Lambda = .02, F (9, 12) = 51.06, p < .001. Bonferroni comparisons further indicated significant differences (p < .001) between all high and low intensity word pairings; bad (M=2.57, SE=.13) and horrible (M=3.91, SE=.17), poor (M=2.48, SE=.11) and terrible (M=3.71, SE=.12), not good (M=1.81, SE=.11) and awful (M=3.38, SE=.16), not tasty (M=2.05, SE=.11) and absolutely disgusting (M=4.38, SE=.12), don’t care for (M=1.91, SE=.12) and really hate (M=3.86, SE=.25).
Measures
Review intensity. A list of ten pre-rated words was used in order to determine review intensity (the dependent variable). The participants were given two words, a high intensity word and a low intensity word, for each of the five blanks in the review and asked to choose one word to fill in each of the blanks. The lower intensity options included the following
words: bad, poor, not good, not tasty, and don’t care for. The higher intensity options included: horrible, terrible, awful, absolutely disgusting, and really hate. For each participant’s review an intensity score was calculated based on the average of the total number of low and high intensity words used in the review. The total intensity score per review ranged from 5-10 before being averaged.
Other variables. Several additional variables were measured in the survey. Demographic variables included age, gender and English fluency. To measure English fluency, participants were asked if they were native English speakers. If the participants indicated that they were not native speakers they were then asked to rate their understanding of the English language on a 5-point Likert scale ranging from 1-not well at all to 5- very
well.
Other variables were included to check the experimental manipulations. First the participants were asked how well they read the other online reviews. This was measured using a 5-point Likert scale ranging from 1-not well at all to 5-very well. Then participants were asked to rate the intensity of the reviews using the same 5-point Likert scale that was used in the pre-test. How negative the participants’ viewed the restaurant scenario was also measured. The participants were asked to rate how negative the restaurant experience was on a 5-point Likert scale (1- not negative at all to 5-very negative).
The participants’ frequency writing online reviews and self-monitoring were also measured. How often participants write online reviews was measured with a 5-point Likert scale ranging from 1- very rarely to 5- very often. Self-monitoring was measured using Synder’s (1974) 25-point true or false scale which included statements such as “In order to
get along and be liked, I tend to be what people expect me to be rather than anything else”
and “I would not change my opinions (or the way I do things) in order to please someone else
Results
A number of descriptive analyses were conducted on the sample. Of the 205
participants that gave their level of English fluency, 49% were native English speakers and of the non-native speakers, 95.4% answered that they understand English well (23.1%) or very well (72.3%). Results showed that very few of the participants frequently write online reviews, as nearly 90% of participants indicated that they only write reviews very rarely (30%), rarely (30%) or sometimes (29%). Further descriptive analyses were conducted to understand how the participants responded to the experimental manipulations. The majority of participants (75%) indicated that they read the other reviews (i.e. the normative cue) at least fairly well and more than 95% of participants found the restaurant scenario to be negative (36%) or very negative (60%).
To check for alternative explanations, a Spearman correlation analysis was conducted to see if a relationship exists between the dependent variable (review intensity) and various control variables. The control variables included the following: age, gender, English fluency, and self-monitoring. Spearman correlations were used because the data included rank-ordered variables. (The following variables were scored dichotomously: review intensity, gender, English fluency and self-monitoring.) Results showed significant correlations between review intensity and the participants’ English fluency (rs (201) = -.15, p = .04) and the level of self-monitoring (rs (202) = -.14, p = .04). Therefore these variables were controlled for in further analyses.
An analysis of covariance (ANCOVA) was used to test the hypotheses. The ANCOVA was performed with review intensity as the dependent variable and anonymity, normative cues, an anonymity-normative cues interaction, English fluency, and self-monitoring as fixed factors. The results of the ANCOVA showed no significant effect of English fluency (p = .15) or self-monitoring (p = .47) on review intensity.
Hypothesis one stated that reviews written anonymously will be more intense than reviews written by identified individuals. The ANCOVA showed no main effect of
anonymity on review intensity F (1, 185) = .04, p = .84. Therefore the first hypothesis was not supported. The second hypothesis specified that moderate other reviews will result in more moderate reviews and intense other reviews will result in more intense reviews. The ANCOVA showed no main effect of the normative cue on review intensity F (1, 185) = .03,
p = .87. Thus the second hypothesis was also not supported by the results of the ANCOVA.
The final hypothesis stated the effect of the normative cue on review intensity will be stronger for anonymous individuals than for identified individuals. The ANCOVA showed a significant interaction effect of anonymity and normative cues on review intensity F (1,185) = 4.59, p = .03. However, the third hypothesis was not supported. See table 1 below for the estimated marginal means.
Table 1: Mean Review Intensity Scores for Anonymity and Normative Cue Interaction Moderate Normative Cue Intense Normative Cue Anonymous M = 1.47, SE = .05 M = 1.37, SE = .05
Identified M = 1.37, SE = .05 M = 1.45, SE = .05
A simple effects test was conducted to determine if significant differences exist between the mean scores for review intensity of the four conditions (anonymous-intense normative cue, identified-intense normative cue, anonymous-moderate normative cue, identified-moderate normative cue). Pairwise comparisons indicated there is not a significant difference between review intensity for anonymous individuals exposed to the intense
normative cue compared to the moderate normative cue (p = .11) or between identified individuals exposed to the intense normative cue compared to the moderate normative cue (p = .16). There is also not a significant difference between review intensity for individuals exposed to the intense normative cue who were anonymous compared to those who were identified (p = .19) or between individuals exposed to the moderate normative cue who were anonymous compared to those who were identified (p = .09).
Conclusion and Discussion
This research was conducted to understand the potential effect of anonymity and norms on the intensity of the language used in online consumer reviews. It was posited that individuals would use more intense language in their reviews if they were anonymous rather than identified. This research also proposed that normative cues in the form of other online reviews would play a role in the language used in online reviews. It was suggested that being exposed to more intense reviews would influence individuals to use more intense language in their reviews and being exposed to more moderate reviews would influence individuals to use more moderate language in their reviews. Furthermore, an interaction effect between norms
and anonymity, whereby anonymous individuals would be more sensitive to the normative cues than identified individuals, was suggested. To test these hypotheses, an experiment was conducted in which anonymity and normative cues were manipulated. The results of the experiment did not support the hypothesized effects.
The results of this research seem out of in line with previous research on anonymity and self-disclosure in blogs (Qian & Scott, 2007) and did not provide support for arguments that anonymity leads to increased self-disclosure online (Joinson, 2001). Based on identity theory and deindividuation theory, it was expected that anonymity would release individuals from the need to manage their impressions and make participants less inhibited. This would allow them to express their true feelings in their online reviews which would be reflected in the intensity of the language used in the reviews, since language intensity is a common measure of a speaker’s emotion (Rogan, 1995). The results of this research indicated that participants found the restaurant experience negative; however these negative feelings were not reflected as expected in the intensity of the language used in their reviews.
Theoretical details in deindividuation theory may account for the unexpected results of this study. Much of deindividuation research is based on the assumption that aggressive behavior, like the use of intense language which was expected in this study, is an outcome deindividuation. However, classic deindividuation theory does not unambiguously state that aggressive behavior is an outcome of deindividuation, rather that deindividuation causes inhibited behavior, such as aggression, to become disinhibited (Johnson & Downing, 1979). It is possible that aggressive behavior may (or may not) be inhibited among certain
individuals, whereas other behaviors may (also) be inhibited. Gergen, Gergen, and Barton’s 1973 study found deindividuation resulted not in aggressive behavior but instead in intimate behavior among participants, suggesting that among their participants the salient inhibited behavior was intimacy rather than aggression (Johnson & Downing, 1979).
Although this study tried to trigger feelings of aggression in the participants through the negative restaurant scenario, it is possible that the participants did not experience inhibited aggression. If this was the case, anonymity could have resulted in disinhibited behavior but the behavior which was disinhibited was simply not aggressive in nature.
Therefore, the intensity of the language used in the reviews was not a reflective of aggressive emotions.
Alternatively, the surprising results of this study may be due to the manipulation of anonymity, which may not have been as effective as planned. In other research on the effects of anonymity and deindividuation, anonymity has been manipulated more realistically than in this research. For example, in the realm of CMC, anonymity has been manipulated by
broadcasting individuals in the identified condition in real time via a webcam (Joinson, 2001) or by having participants in the identified condition sit in plain sight of others while they type their messages on computers (Lea & Spears, 1991). Whereas in offline studies of
deindividuation, anonymity is often manipulated through the use of concealing costumes while name tags and no concealing costumes are used for identified individuals (Postmes & Spears, 1998).
Due to ethical concerns, in this research anonymity was only manipulated by asking the participants to imagine themselves anonymous or identified. The participants in the identified condition were asked to imagine their review would be posted with their Facebook name and profile but the participants knew the results of the survey, including the reviews, would remain anonymous. Therefore, there is good reason to believe that individuals in the identified condition did not really feel identified. Unfortunately, there was not a question included in the survey to check what extent the anonymity manipulation worked.
Like anonymity, the results of this research did not show an effect of norms on language intensity. It was expected that the normative cues would influence reviewer
behavior since research has shown people to be highly susceptible to social influence
(Cialdini et al., 1990; Deutsch & Gerard, 1955). Some researchers have suggested that online, where other cues are unavailable, language and message content are used as social cues (Walther, 2007). Therefore it was anticipated that the language used in other people’s reviews would act as a normative cue and would influence the language used in individuals’ reviews. However, the results of this research showed no main effect of the normative cues on review intensity and thus did not support this theoretical reasoning.
It is possible that the manipulation of the normative cues was not as effective as expected, which may account for the unanticipated results. Although the results indicated that the majority of participants read the other reviews, it is possible the content of the reviews and/or the number of reviews included may not have been sufficient to serve as a normative cue for using moderate or intense language in online reviews. If the language intensity norm was not made salient through the normative cue, the normative influence would likely not have been effective, as other research on norms has shown that making the norm salient is important for the norm to influence behavior (Cialdini et al., 1990).
The results of this research showed an interaction effect between anonymity and the normative cues on review intensity. However the hypothesized interaction effect was not confirmed. The saliency of group identity in the survey may account for this unexpected result. The hypothesized effect was largely based on the SIDE which argues that
deindividuation can lead to a greater compliance with norms when group identity is made salient (Reicher et al., 1995). In this research it was assumed that group identity is likely to emerge on an online consumer review platform. However, it is possible that within the survey, the concept of an online reviewer community was not prominent enough, leading to lack of group identity. This potential lack of group identity could cause the anonymity-norm interaction to behave unexpectedly.
Although this research does not offer further insight into the direction of the
anonymity-norm interaction, the results do offer further support for deindividuation theories which describe a strong relationship between anonymity and social norms on deindividuated behavior (Lea & Spears, 1991; Postmes & Spears, 1998; Reicher et al., 1995). How
anonymity and norms influence language intensity in consumer reviews remains unclear. However, the results of this research seem to be in line with the results of a meta-analysis of deindividuation studies which found a firm link between norms and deindividuated behavior (Postmes & Spears, 1998).
Limitations and Future Research
There are several limitations to this study. However, a main limitation of this
research lies in the operationalization of review intensity. Review intensity was measured in a very structured way (i.e. filling in the blanks of a pre-written review) which likely allowed for little self-expression among participants. This is a major weakness since in essence this research was interested in human expression, reflected in the intensity of the language used in the reviews under various conditions (i.e. anonymous/identified) and normative influences. If participants were unable to express themselves fully, it is not surprising that the study was unable to find variations in the way participants expressed themselves in their reviews under the different conditions.
Although much of language intensity research has been conducted using the same method used in this research (Burgoon & Miller, 1971; Burgoon et al., 1975), a content analysis of anonymous and identified online consumer reviews may have been more suitable for answering this research question. A content analysis would allow for naturally written reviews to be studied which would probably reveal a much wider range of self-expression and variations in language intensity. Several researchers have developed coding metrics for coding the language intensity of natural discourse which could be used to code language
intensity in online consumer reviews (Rogan, 1995). Unfortunately, due to limited resources, a content analysis was not possible for this research.
Future research on anonymity, norms and language intensity in online consumer reviews should consider a research design which would allow participants to write their reviews more naturally without using partially completed reviews and pre-rated words. More realistic anonymous or identified conditions, like those discussed above, are also advised. An ideal research design would involve a review platform created specifically for the research, which allows reviewers to write their own reviews in free form. The platform could require users to be identified with their posts or allow them to post anonymously and could also expose reviewers to other consumer reviews. Such a method would likely allow researchers to more effectively measure language intensity, provide a more realistic manipulation of anonymity and norms and would of course, increase the ecological validity of the research. Implications for Theory and Practice
The results of this research indicated that anonymity does not influence language intensity in online consumer reviews. This raises some theoretical questions regarding anonymity and electronic communication, such as whether or not research on flaming and other aggressive language in CMC can be applied to eWOM. Some researchers have argued that individuals use the internet to post inflammatory remarks as a way to satisfy certain desires, such as entertainment, relaxation, escapism and passing time (Alonzo & Aiken, 2004). These motivations to flame are very different from motivations to participate in eWOM which include desires for social interaction, economic reasons, and helping other consumers (Hennig-Thurau et al., 2004). Thus eWOM may be inherently different from other types of CMC where aggressive language is often found and therefore new theories, specific to eWOM, may need to be developed to better understand language intensity in online consumer reviews.
The results of this research also indicated that normative cues in the form of other reviews do not influence review intensity. More research should be undertaken to understand to what extent the content of online consumer reviews and other electronic messages
influence computer mediated interactions. Theories of reduced normative influence in CMC, such as those described by Kiesler, Siegel and McGuire (1984), should be explored more deeply as the results of this research may provide support for such theoretical reasoning.
The results of this research leave many questions unanswered regarding intense language in online consumer reviews. It was anticipated that this research would provide some practical insights for consumer review platforms and consumer reviewers. Of particular interest was whether online consumer review platforms could help curb the use of intense and socially unacceptable language in online reviews by requiring individuals to identify
themselves when posting reviews or by changing users’ exposure to other reviews. However, further research is required before such recommendations can be given to online consumer review platforms which would help them improve the way they collect consumer reviews.
References
Alonzo, M., & Aiken, M. (2004). Flaming in electronic communication. Decision Support
Systems, 36(3), 205-213.
Bradac, J. J., Bowers, J. W., & Courtright, J. A. (1979). Three language variables in
communication research: Intensity, immediacy, and diversity. Human Communication
Research, 5(3), 257-269.
Burgoon, M., Jones, S. B., & Stewart, D. (1975). Toward a message-centered theory of persuasion: three empirical investigations of language intensity. Human Communication
Research, 1(3), 240-256.
Burgoon, M., & Miller, G. R. (1971). Prior attitude and language intensity as predictors of message style and attitude change following counter attitudinal advocacy. Journal of
Personality and Social Psychology, 20(2), 246.
Christopherson, K. M. (2007). The positive and negative implications of anonymity in internet social interactions:“On the internet, nobody knows you’re a dog”. Computers in
Human Behavior, 23(6), 3038-3056.
Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity.
Annu.Rev.Psychol., 55, 591-621.
Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal of
Deutsch, M., & Gerard, H. B. (1955). A study of normative and informational social
influences upon individual judgment. The Journal of Abnormal and Social Psychology,
51(3), 629.
Hennig-Thurau, T., Gwinner, K. P., Walsh, G., & Gremler, D. D. (2004). Electronic word-of-mouth via consumer-opinion platforms: What motivates consumers to articulate
themselves on the internet? Journal of Interactive Marketing, 18(1), 38-52.
Hennig-Thurau, T., Walsh, G., & Walsh, G. (2003). Electronic word-of-mouth: Motives for and consequences of reading customer articulations on the internet. International
Journal of Electronic Commerce, 8(2), 51-74.
Johnson, R. D., & Downing, L. L. (1979). Deindividuation and valence of cues: Effects on prosocial and antisocial behavior. Journal of Personality and Social Psychology, 37(9), 1532.
Joinson, A. N. (2001). Self‐disclosure in computer‐mediated communication: The role of self‐awareness and visual anonymity. European Journal of Social Psychology, 31(2), 177-192.
Kiesler, S., Siegel, J., & McGuire, T. W. (1984). Social psychological aspects of computer-mediated communication. American Psychologist, 39(10), 1123.
Lea, M., & Spears, R. (1991). Computer-mediated communication, de-individuation and group decision-making. International Journal of Man-Machine Studies, 34(2), 283-301.
Marx, G. T. (1999). What's in a name? some reflections on the sociology of anonymity. The
Postmes, T., & Spears, R. (1998). Deindividuation and antinormative behavior: A meta-analysis. Psychological Bulletin, 123(3), 238.
Qian, H., & Scott, C. R. (2007). Anonymity and Self‐Disclosure on weblogs. Journal of
Computer‐Mediated Communication, 12(4), 1428-1451.
Reicher, S. D., Spears, R., & Postmes, T. (1995). A social identity model of deindividuation phenomena. European Review of Social Psychology, 6(1), 161-198.
Reinig, B. A., Briggs, R. O., Brandt, S. A., & Nunamaker, J. (1997). The electronic classroom on fire: Why it happens, and how to put out the flames. System Sciences,
1997, Proceedings of the Thirtieth Hawaii International Conference On, , 2 639-647.
Rogan, R. G. (1995). Language intensity: Testing a content‐based metric. Communication
Reports, 8(2), 128-135.
Schau, H. J., & Gilly, M. C. (2003). We are what we post? self‐presentation in personal web space. Journal of Consumer Research, 30(3), 385-404.
Snyder, M. (1974). Self-monitoring of expressive behavior. Journal of Personality and Social
Psychology, 30(4), 526.
Suler, J. (2004). The online disinhibition effect. Cyberpsychology & Behavior, 7(3), 321-326.
Walther, J. B. (2007). Selective self-presentation in computer-mediated communication: Hyperpersonal dimensions of technology, language, and cognition. Computers in Human
Zhao, S., Grasmuck, S., & Martin, J. (2008). Identity construction on Facebook: Digital empowerment in anchored relationships. Computers in Human Behaviour, 24(5), 1816-1836.
Appendix Stimuli Material Intense other reviews:
1) This restaurant is horrendous! The service is atrocious and the food is gross! 2) I can't stand this restaurant! The food is revolting and the waiters are dreadful.
Everything about this restaurant is horrific! Moderate other reviews:
1) This restaurant is unexceptional. The service is inadequate and the food is so-so. 2) I don’t prefer this restaurant. The food is second-rate and the waiters are mediocre.
Everything about this restaurant is unsatisfactory. Restaurant scenario:
Imagine you make reservations to go to an expensive restaurant for a special occasion. When you arrive at the restaurant, it takes a long time before you are helped. When the waiters finally greet you, they say they have lost your reservation and you will have to wait for an available table. You wait almost half an hour for your table. When you are finally seated, you are in an uncomfortable spot. When the food arrives it is not good and some of it is even cold. You try to politely mention this to your waiter but he just ignores you. Not once does anyone ask you how the food is or if you would like anything else. After this experience, you are ready to leave. But again, it takes a long time to get the bill and when you mention this, no one apologizes – they hardly even say goodbye when you leave.
