Available online at www.sciencedirect.com Journal of CONSUMER PSYCHOLOGY Journal of Consumer Psychology 21 (2011) 226 – 239 Can including pros and cons increase the helpfulness and persuasiveness of online reviews? The interactive effects of ratings and arguments☆ Ann E. Schlosser Marketing and Evert McCabe Faculty Fellow, Foster School of Business, University of Washington, Box 353200, Seattle, WA 98195-3200, USA Received 12 November 2010; revised 2 March 2011; accepted 12 April 2011 Available online 17 May 2011 Abstract One guideline given to online reviewers is to acknowledge a product's pros and cons. Yet, I argue that presenting two sides is not always more helpful and can even be less persuasive than presenting one side. Specifically, the effects of two- versus one-sided arguments depend on the perceived consistency between a reviewer's arguments and rating. Across a content analysis and three experiments that vary the information provided in the online review and whether the ratings are positive or negative, the results support these predictions. Furthermore, beliefs that the reviewer is able (vs. willing) to tell the truth mediated the effects. © 2011 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved. Keywords: Word-of-mouth communication; Online peer reviews; Two-sided arguments; Internet; Credibility; Persuasion Communication between consumers—or word-of-mouth (WOM) communication—is a powerful force in shaping consumers' product attitudes (e.g., Brown and Reingen, 1987; Herr, Kardes, and Kim, 1991; Sultan, Parley, and Lehmann, 1990). One popular vehicle for WOM communication is the Internet, making the inclusion of online peer reviews—that is, reviews written by consumers on the Internet—a necessary feature for websites wishing to attract and retain consumers. For instance, over half (58%) of consumers prefer sites with peer reviews (Decker, 2007) and nearly all (98%) online shoppers reported reading peer reviews before making a purchase (Freedman, 2008). However, simply offering online peer reviews is likely insufficient to attract and retain consumers. Web site owners need visitors to write reviews that consumers find helpful and persuades them to buy. Consequently, it is important to understand what makes some online peer reviews more helpful and persuasive than others. ☆ The author thanks Wayne Hoyer, Tiffany White, Richard Yalch, and the University of Illinois Urbana-Champaign Marketing Department for their helpful comments on earlier versions of this paper, and Dante Batingan, Celeste Chen and HeeSun Choi for their research assistance. E-mail address: aschloss@uw.edu. One guideline offered by experts and websites to presumably make reviews more helpful and persuasive is for reviewers to present both a product's pros and cons in their reviews (Burns, 2008; JupiterResearch, 2008). Even some websites have separate fields that directly ask reviewers to identify pros and cons (e.g., epinions.com). Such a guideline has intuitive appeal for several reasons. First, in the advertising literature, providing two sides rather than one side can make an ad more credible and thus persuasive (Crowley and Hoyer, 1994). Second, even without prompting, speakers often present multiple sides in an attempt to appeal to everyone when an audience's views are unknown (Tetlock, Skitka, and Boettger, 1989). Indeed, those posting reviews online tend to mention a product's pros and cons more than those reviewing a product privately do (Schlosser, 2005). Yet, despite evidence that consumers read the text of online reviews rather than rely solely upon such summary statistics as the average star rating (Chevalier and Mayzlin, 2006), most research on online peer reviews has ignored the written content of reviews due to the cost and effort incurred in measuring it (Godes and Mayzlin, 2004), instead focusing on more easily quantifiable measures such as product ratings (e.g., Forman, Ghose, and Wiesenfeld, 2008; Gershoff, Mukherjee, and Mukhopadhyay, 2003; Godes and Mayzlin, 2004; Sen and Lerman, 2007). Even though this research has 1057-7408/$ - see front matter © 2011 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jcps.2011.04.002 yielded important insights, by ignoring the written content of reviews, an important piece to understanding the persuasiveness and helpfulness of online reviews—such as the effectiveness of presenting two sides versus one side—may be missed. I propose that reviews that include a product's pros and cons will not always be more helpful, credible and persuasive—it will depend on the reviewer's product rating. Consequently, despite reviewers' attempts to publicly state the “right” arguments, under certain conditions, such attempts may backfire by triggering concerns about the reviewers' qualifications, thereby reducing how helpful and persuasive they are. In fact, I argue that presenting two sides can cause a review with an extremely favorable rating to produce product judgments that are as low if not lower than a review with a moderately favorable rating. Although online peer reviews involve consumer-to-consumer communication, I use the term “consumer” to refer to the person who reads a review to form a product judgment, and the term “reviewer” to refer to the person who wrote a review after a product experience. Furthermore, with the exception of study 3, the studies focus on positive reviews because online peer reviews tend to be positive (Chevalier and Mayzlin, 2006; Freedman, 2008). For instance, one study found that 80% of product ratings are either 4 or 5 stars (Bazaarvoice, 2008). I begin with an overview of the literature on message sidedness, WOM communication and credibility, followed by the theoretical development of the hypotheses. Four studies follow that test the robustness of these predictions by varying the method of data collection (i.e., content analysis vs. online experiment), the information provided along with the online peer reviews, the product evaluated, and whether ratings are positive or negative. In addition, different types of credibility are tested as mediators. Taken together, the results suggest that depending on the reviewer's rating, presenting two- (vs. one-) sided arguments can backfire by reducing credibility perceptions, helpfulness and persuasiveness. Furthermore, five-star ratings do not always produce greater product attitudes and purchase intentions than four-star ratings. Literature review In consumer research, the credibility and persuasiveness of two- (vs. one-) sided arguments have been studied most extensively in the context of advertising (for a review, see Crowley and Hoyer, 1994). Mixed results have emerged regarding the effects of two-sided advertisements on persuasion. Whereas some studies have shown positive effects (Etgar and Goodwin, 1982; Kamins, Brand, Hoeke, and Moe, 1989), others have obtained nonsignificant or mixed results (Golden and Alpert, 1987; Kamins and Assael, 1987; Kamins and Marks, 1988; Settle and Golden, 1974). Nonsignificant effects are to be expected given the tradeoff inherent in providing negative product information in an advertisement; although providing some negative information increases advertiser credibility, there is the cost of divulging negative product information (Crowley and Hoyer, 1994). Such costs are offset under certain conditions (for a meta-analysis, see Eisend, 2006). For instance, two-sided advertisements yield positive effects 227 when the negative and positive attributes are negatively correlated (Bohner, Einwiller, Erb, and Siebler, 2003; Pechmann, 1992), when individuals are promotion- (rather than prevention-) focused (Florack, Ineichen, and Bieri, 2009), when consumer suspicion of an ulterior motive is reduced (DeCarlo, 2005), and when negative information was voluntarily disclosed and consumers are under cognitive load (Eisend, 2010). One of the more consistent findings is that including some negative product information in an ad enhances advertiser credibility (cf., Bohner et al., 2003; Kamins and Assael, 1987; Settle and Golden, 1974; Swinyard, 1981). This is because acknowledging some negative aspects of a product acts against the advertiser's self-interest to sell the product. Consequently, consumers deem the advertiser to be more trustworthy than if the ad contained only positive (i.e., one-sided) arguments. Yet, with advertising, the source has a clear incentive to present only one side. With WOM communication, however, the source's motivations are less clear, and as a result, the findings from advertising contexts may not generalize to WOM contexts. Traditionally, WOM communication differed in multiple ways from advertising; it was personally rather than financially motivated, synchronous, ephemeral, oral and informal between parties who “interact with each other in real time and space” (Stern, 1994, p. 7). However, with computer-mediated communication, many of these differences are eliminated. Although it is often still personally rather than financially motivated, it can be asynchronous, stored, textual, and perhaps most importantly, is not bounded by temporal or physical space. As a result, it often occurs between people in one's “virtual life” rather than “experiential reality life”—conditions that require recipients to imagine communicators and their communication goals (Stern, 1994). Like advertising effectiveness, the persuasiveness of offline WOM communication has been attributed to communicator characteristics, such as communicator credibility (Bearden and Etzel, 1982; Laczniak, DeCarlo, and Ramaswami, 2001; Mizerski, 1982). Credibility is often known when communicating with friends or family, or inferred from such social context cues as tone of voice or body language. Yet, unlike offline WOM communication, anyone can easily broadcast their product opinions to strangers via the Internet (Hoffman and Novak, 1996), regardless of their qualifications. As a result, individuals read product experiences from those whose motivations and abilities are unknown to them. Furthermore, because most online contexts lack social context cues (Dubrovsky, Kiesler, and Sethna, 1991), individuals must use other information—such as the content of a review—to determine a reviewer's credibility. The content of most online peer reviews consist of (1) an overall product evaluation (i.e., the rating) and (2) a written explanation for this evaluation (i.e., the arguments; Burns, 2008). Oftentimes, this is the only information consumers have from which to determine whether to trust reviewers and thus accept their arguments. I propose that individuals use reviewers' arguments in conjunction with their ratings to judge their credibility, which in turn affects how persuasive the reviews are. According to attribution theory (e.g., Jones and Davis, 1965; Kelley, 1973), communicator perceptions play a critical role in 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 determining whether a message is persuasive. Specifically, a message is persuasive to the extent that it is considered valid rather than due to irrelevant causes (Kelley, 1973), such as reporting or knowledge biases. Credibility is the extent to which individuals perceive the communicator to be willing and able to convey the truth (Eagly and Chaiken, 1993; Hovland, Janis and Kelley, 1953; Wood and Eagly, 1981). For instance, reviewers may be willing to tell the truth (i.e., their reporting is unbiased), but lack credibility because they do not have the skills to properly evaluate the product (i.e., their knowledge is biased). When the communicator is deemed both willing and able to tell the truth (i.e., is credible), the message is accepted. When the communicator lacks credibility, however, individuals discount the message as invalid and thus the message is unpersuasive. Past research has shown that acting in a manner that disconfirms expectations can bolster credibility. For instance, one reason why providing two- (vs. one-) sided arguments in an ad bolsters credibility is that identifying both pros and cons is inconsistent with advertisers' presumably extremely favorable opinion of the product (Crowley and Hoyer, 1994). Consequently, advertisers using two-sided arguments are deemed trustworthy. Likewise, relative to confirming expectations, disconfirming expectations reduces attributions that the communicator's arguments are due to his/her background (e.g., political affiliation or pressures to sell a product) and increases attributions that the communicator's arguments are based on factual evidence (e.g., research; Wood and Eagly, 1981). Yet, in these instances, the audience has information about the communicator's background—information that calls into question the communicator's trustworthiness. These findings are unlikely to generalize to situations in which background information about the communicator is absent (Wood and Eagly, 1981)—a situation most often experienced when reading online peer reviews. In fact, disconfirming expectations may backfire. I propose that unlike advertising, because peer reviewers do not have a clear incentive to lie, they will generally be deemed trustworthy. However, because peer reviewers are often strangers, their ability to accurately evaluate the product is largely unknown. To determine if a reviewer is able to convey the truth, consumers will likely compare the rating and arguments. In this case, disconfirming expectations should reduce (not enhance) credibility. For instance, reviewers who strongly endorse a product will likely be expected to present only pros. If reviewers instead disconfirm expectations by presenting two sides, their ability to accurately review the product will likely be questioned more than if they confirmed expectations by presenting one side. This prediction is based upon evaluative–cognitive consistency theory. First proposed by Rosenburg (1956), evaluative–cognitive consistency is “the consistency between one's abstract evaluation of an attitude object and the evaluative content of one's beliefs about the object” (Eagly and Chaiken, 1993, p. 114) and exists when “objects are liked to the extent that they are seen as possessing desirable attributes” (Scott, 1969, p. 263). For instance, holding only favorable beliefs about a product should make it appear entirely good and thus lead to extremely favorable evaluations. In contrast, holding both favorable and unfavorable beliefs should make it appear neither entirely good nor bad and thus should lead to moderate evaluations. Indeed, when individuals derive their attitudes from their beliefs, considering both the pros and cons produces moderate judgments, whereas considering only the pros (vs. cons) leads to extreme judgments (Brauer, Chambres, Niedenthal, and Chatard-Pannetier, 2004; Linville, 1982; Tetlock, 1983). Consequently, considering two sides (vs. a single side) is evaluatively consistent with holding moderate (vs. extreme) judgments. Importantly, those with highly consistent explanations for their attitudes are as aware of inconsistent information as those with less consistent explanations for their attitudes (Chaiken and Yates, 1985). Consequently, individuals can (and often do) justify their attitudes to others by referring to inconsistent information (Sengupta and Johar, 2002; Simonson and Nowlis, 2000; Tetlock, 1983). For instance, individuals often express arguments that seem acceptable regardless of whether these arguments are evaluatively consistent with their publicly stated attitudes (Schlosser and Shavitt, 1999, 2002). Likewise, online reviewers will sometimes modify their product ratings to convey a desirable impression, such as to lower their ratings to appear discerning independent of whether they modified their arguments (Schlosser, 2005). Yet, such attempts to express acceptable and persuasive arguments and attitudes while ignoring the consistency between these expressed arguments and attitudes may backfire by making reviewers appear less capable, and in turn, less helpful and persuasive. Although evaluative–cognitive consistency was initially developed and later questioned for its ability to always explain attitude formation (Eagly and Chaiken, 1993), I propose that this theory does capture how individuals expect others to explain their attitudes. Consequently, this theory may be more relevant for understanding how consumers judge WOM communication. Hypotheses Helpfulness When judging whether a review is helpful, individuals likely consider both (1) the diversity of information presented, and (2) the evaluative consistency of the review. In general, because two-sided reviews provide more diverse information than onesided reviews, they should be deemed more informative and helpful. For instance, a review in which the reviewer likes everything about a product will likely be deemed less helpful than one in which the reviewer identifies what was liked and disliked. Yet, the magnitude of this difference likely also depends upon whether the reviewer's rating is evaluatively consistent with a two-sided review. Inconsistencies between a reviewer's rating and review will likely dampen the perceived helpfulness of a two-sided review by raising concerns about the reviewer's ability. For example, presenting multiple sides when one's rating is extreme might convey that the reviewer is less qualified to review the product. Hence, the perceived 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 228 helpfulness should be higher for two- than one-sided reviews, but only when the corresponding rating is moderate (vs. extreme). In other words, an additive effects model is proposed whereby helpfulness is enhanced by two-sided arguments because more diverse information is presented (two-sided = 1; one-sided = 0) as well as by the consistency between reviewers' arguments and ratings (consistency = 1; inconsistency = 0). As a result, two-sided reviews should be more helpful than one-sided reviews only when the rating is moderate because this would be two-sided (+ 1) and internally consistent (+ 1) rather than onesided (+ 0) and inconsistent (+ 0). In contrast, when the rating is extreme, the difference between two- and one-sided reviews should be reduced because the reviews would be either twosided (+ 1) but internally inconsistent (+ 0) or one-sided (+ 0) but consistent (+ 1). H1. Rating extremity and arguments should interact such that individuals deem two-sided reviews to be more helpful than one-sided reviews only when the reviewers' product ratings are moderate. Product judgments Although two-sided arguments should enhance perceived helpfulness, they are unlikely to have the same effect on persuasion. Indeed, as mentioned earlier, two-sided ads are not always more persuasive than one-sided ads—although identifying cons can enhance credibility, it can also reduce consumers' expectations of product benefits (for discussion, see Bohner et al., 2003; Crowley and Hoyer, 1994). Thus, I propose a different additive effects model for persuasion than helpfulness, in which persuasion is enhanced by one-sided arguments because only information supporting a position is presented (one-sided = 1; two-sided = 0) as well as the consistency between arguments and evaluation (consistency = 1; inconsistency = 0). Consequently, one-sided reviews should be more persuasive than two-sided reviews when the rating is extreme because this would be one-sided (+ 1) and internally consistent (+ 1) rather than two-sided (+ 0) and inconsistent (+ 0). However, this advantage of one- (vs. two-) sided reviews should be reduced when the rating is moderate because the reviews would be either one-sided (+ 1) but internally inconsistent (+ 0) or two-sided (+ 0) but consistent (+ 1). H2. Rating extremity and arguments should interact such that reviews with one-sided arguments are more persuasive than reviews with two-sided arguments only when the rating is extreme. Four studies were conducted to test these predictions. Study 1 is a content analysis of online movie reviews to test H1 in a naturalistic setting while measuring the effects of other relevant variables. The remaining studies are online experiments that directly test the interactive effects of reviewers' ratings and arguments on perceived helpfulness and product judgments (H1–H2) when the rating is positive (studies 2–4) and negative (study 3), and to identify the mechanism underlying these interactive effects (study 4). Because consumers often consult 229 others' opinions for experiential products (West and Broniarczyk, 1998), the reviewed products were books (studies 2 and 4) and movies (studies 1 and 3). Study 1: content analysis Method The first study involved a content analysis of positive peer reviews posted at Yahoo! Movies, where users post online movie reviews as well as indicate whether they found a review to be helpful. At this site, users can rate a movie A, B, C, D or F (the basis for the rating variable) and write a review (the basis for the arguments variable). A posted review is followed by the question “Was this review helpful?” and two response options: “Yes” and “No.” Yahoo! indicates how many people found each review helpful as well as how many people “voted.” The dependent variable (helpfulness) was calculated by dividing the number of people who voted that they found the review helpful by the total number who submitted votes for that review. If H1 is supported, then more visitors should deem two-sided reviews to be helpful than one-sided reviews only when these reviews have moderate (vs. extreme) ratings (i.e., a B rather than A). The decision to focus on positive peer reviews was based on prior research (e.g., Bazaarvoice, 2008; Chevalier and Mayzlin, 2006), and a content analysis of the target movies and two other movies that revealed that relatively few reviews were negative (6–20% of reviews had a rating of D or F), of which few (4–17%) had at least five individuals submit helpfulness votes. Two undergraduate students who were unaware of the experimental hypotheses began by considering alphabetically movies from 2004 that were reviewed at Yahoo! Movies. From these, the first two that had diverse ratings and a high degree of user activity were chosen. In addition, to asses the generalizability of the effects, the movies were from different genres (the genre variable): science fiction (“Alien vs. Predator”) and drama (“Friday Night Lights”). The same two students then independently selected peer reviews for either “Friday Night Lights” (n = 70) and “Alien vs. Predator” (n = 77) that met the following criteria: the review (1) was posted within a week after the movie was released, (2) had either an extremely or moderately favorable rating (A or B; the rating variable), and (3) had at least five individuals who submitted helpfulness “votes” (the average number of voters across reviews was 47.5). From these 147 reviews, 114 were analyzed. Reviews were omitted due to the author admitting to not seeing the film (n = 2) and being too short (i.e., had fewer than 2 sentences; n = 13). A similar number of reviews were deleted across conditions (all χ2s b 1). In addition, because two-sided messages are less effective when over half of the content is negative (Crowley and Hoyer, 1994; Eisend, 2006), 18 reviews were omitted that were largely (over 50%) negative. This variable was used to screen reviews rather than as a covariate in the reported analyses because it is highly correlated with the arguments variable (r = .74, or 55% related). However, the results do not meaningfully differ between deleting these reviews from the analyses and controlling for this variable in ANCOVAs. 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 Results The proportion of helpfulness votes was analyzed with a 2 (rating) × 2 (arguments) × 2 (genre) ANCOVA while controlling for variables likely to affect how believable the reviewer is (e.g., anonymity), the effectiveness of two-sided messages (e.g., title valence), and additional movie information provided by the reviewer (e.g., whether the reviewer rated all attributes given by Yahoo!). These covariates had less than 33% common variation with the independent variables. Moreover, none of the covariates were significant (all Fs(1, 106) b 1). As expected, the two-sided reviews received a higher proportion of helpfulness votes than the one-sided reviews did (F(1, 94) = 4.08, p b .05), which as predicted in H1, was qualified by a rating × arguments interaction (F(1, 94) = 5.93, p b .05; see Fig. 1A). Planned contrasts revealed that more helpfulness votes were given to two- than one-sided reviews when the rating was moderately favorable (Ms = .73 vs. .49, F(1, 94) = 8.69, p b .01; all reported means from ANCOVAs are covariateadjusted means), but not when it was extremely favorable (Ms = .59 vs. .59, F(1, 94) b 1). Furthermore, two-sided reviews were deemed helpful by more voters when the rating was Proportion of Helpfulness Votes A 0.8 0.75 Extreme (A) Moderate (B) 0.7 0.65 0.6 0.55 0.5 0.45 0.4 One-sided Two-sided Written Arguments B Proportion of Helpfulness Votes The two students then served as judges who independently coded the arguments of each review as one- or two-sided (the arguments variable) using scoring procedures developed to assess whether written documents reflect more than a single side (Baker-Brown et al., 1992). According to these procedures, a review is one-sided if only the pros are mentioned (e.g., “AVP is a great, fun summer flick, a long awaited reward for the fans of the Sci-Fi Horror genre”), whereas a review is two-sided if both pros and cons are mentioned (e.g., “The action sequences are outstanding, although the way they were shot can make your eyes spin a bit”). The judges had 90% agreement and resolved all disagreements through discussion. In addition, to control for other variables likely to affect consumers' perceptions of a reviewer's credibility and thus helpfulness, the judges also recorded (1) whether the reviewer was anonymous in terms of location or gender (yes/no for each), and (2) whether there were any grammatical or spelling errors (yes/no). Furthermore, to control for message structure variables shown in prior research to reduce the effectiveness of twosided messages (Crowley and Hoyer, 1994), and as a result, may cause two-sided messages to be ineffective even with moderate ratings, the judges also coded (1) whether the title was positive, negative or neutral, (2) whether the first sentence was positive, negative or neutral, (3) whether they refuted a negative claim, and (4) whether the reviewer criticized the plot (an important attribute of a film). Finally, because in addition to providing a written review and overall movie rating, reviewers can rate each of four attributes (story, acting, direction, visuals) with a letter grade from A to F, the judges coded whether the reviewer rated all attributes (yes/no), and whether they rated all of these attributes the same (yes/no). The judges had 78–88% agreement in coding these categories and resolved all disagreements through discussion. 0.80 0.75 Extreme (A) Moderate (B) 0.70 0.65 0.60 0.55 0.50 0.45 0.40 Same Varied Rated Attributes Fig. 1. Study 1: effects of ratings and arguments on the percent of helpful votes. moderately than extremely favorable (Ms = .73 vs. .59, F(1, 94) = 4.06, p b .05). Because reviewers' ratings of the four attributes (story, acting, directing, visuals) could also constitute their arguments for their overall rating, thereby affecting the number of helpfulness votes a review receives independent of what is written, the proportion of helpfulness votes was analyzed using the rated attributes as the independent variable instead of the written arguments. Specifically, a 2 (rating) × 2 (rated attributes: same or varied) × 2 (genre) ANCOVA was conducted with the same covariates used previously except that the rated attributes variable was replaced by the arguments variable. As before, none of the covariates were significant (all Fs(1, 94) b 1.39, NS). The only significant effect was a rating × rated attribute interaction (F(1, 94) = 5.57, p b .05; see Fig. 1B). Planned contrasts revealed that a review with varied (i.e., diverse) attribute ratings received more helpfulness votes than a review with the same rating across all attributes did when the overall product rating was moderately favorable (Ms = .76 vs. .52, F(1, 94) = 6.51, p = .01), but not when it was extremely favorable (Ms = .57 vs. .60, F(1, 94) b 1). In fact, reviews with varied attribute ratings received more helpfulness votes when the overall rating was moderately than extremely favorable (Ms = .76 vs. .52, F(1, 94) = 5.04, p b .05). These findings support the argument that when judging helpfulness, consumers consider both (1) the diversity of information provided, and (2) the consistency between the arguments provided and the overall rating. In other words, consistency is a necessary but not sufficient condition for perceived helpfulness. Although rating every attribute as “good” (a “B”) is consistent with an overall rating 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 230 of “B,” these findings reveal that when explaining a moderate attitude, it is less helpful to say that every attribute was “good” (i.e., mention only pros) than to rate some attributes as “outstanding” and others as “mediocre” (i.e., to include pros and cons). Conclusions Many websites encourage reviewers to discuss a product's pros and cons within their reviews. This is likely because presenting more than one side would seem to be more informative and helpful than presenting only one side. Yet, the results of study 1 indicate that this is not always the case. A content analysis of peer reviews posted at a popular website supports H1: reviews with two- (vs. one-) sided arguments received a significantly higher proportion of helpful votes only when the reviewer's rating was moderately favorable. When the reviewer's rating was extremely favorable, the advantage of presenting two- (vs. one-) sides disappeared. Whereas the results of study 1 shed light on how the text and rating of a review interact to affect helpfulness perceptions (H1) and qualify the assumption that presenting two sides will be more helpful than presenting one, it remains unclear whether these variables interact to influence product judgments (H2). I propose that encouraging reviewers to present multiple sides (vs. a single side) will backfire if the reviewer has an extremely favorable rating. Specifically, because two-sided arguments are less consistent with an extremely favorable rating than onesided arguments, product judgments should be lower when reviews are two- (vs. one-) sided and the rating is extremely favorable. In fact, two-sided reviews with extremely favorable ratings (e.g., 5 stars) should cause lower product judgments than two-sided reviews with less favorable but more consistent ratings (e.g., 4 stars). This is one objective of study 2. The others are to test H1 and H2 in a controlled environment (an online experiment) with a different product (books). Study 2: online experimental analysis of positive peer reviews Method Design and sample The sample consisted of 201 undergraduates who participated in exchange for partial completion of a course requirement. Participants were randomly assigned to one of four conditions in a 2 (rating: extreme vs. moderate) × 2 (arguments: one-sided vs. two-sided) experimental design. For the rating manipulation, the reviewer's rating was either 5 stars (extreme) or 4 stars (moderate). For the arguments manipulation, the written content of the review was entirely positive (one-sided) or positive with one negative claim (two-sided). Materials and procedure At the start of the study, participants were instructed to imagine that they are looking for a book on business practices, are considering the book by Jim Collins entitled “Good to Great: Why Some Companies Make the Leap…and Other's 231 Don't” and go to www.epinions.com. Participants then viewed a screen capture from epinions.com, which provided an image of the book, key background information (e.g., the author, book category) as well as “More Information” describing the content of the book. Below this was a peer review, which was a modification of an actual peer review for this book at epinions. com. Across conditions, the reviews began with the same title (“A great book for MBAs”), author (“by danielm”) and date. Alongside this information was the peer's rating of 4 or 5 stars (the rating manipulation). Below this was a list of pros and cons followed by a written review, which varied across arguments conditions. In the one-sided condition, the pros were concise, data-driven and well-crafted, and the cons were “none.” The same pros were identified in the two-sided condition, but the con was “need more company examples, even from the loser side.” Those in the one-sided condition then read “If you aim to become a real leader in a big institution, then this is the book you should spend time on. It is about leadership and how to build a great organization and last long. At 300 pages, Good-toGreat is a comprehensive, well-researched project that is well written and entertaining. There are many examples to illustrate the author's key points too.” The arguments were the same for those in the two-sided condition except the last sentence was replaced with “There should be more examples to illustrate the author's key points, however.” This manipulation is consistent with scoring procedures for recognizing one or more sides (Baker-Brown et al., 1992). Furthermore, several characteristics of the two-sided review were held constant to maximize its effectiveness. Specifically, two-sided messages are most effective when fewer than 50% of the attributes are attacked, and when the attacked attributes are no more important than the positive attributes (Crowley and Hoyer, 1994). The latter condition and the perceived internal consistency were pretested (as discussed later). After reading the review, participants completed a survey, which began by assessing individuals' attitudes toward the product on three semantic differential scales ranging from −3 to +3 and anchored with bad/good, uninteresting/interesting and dislike/like (α = .87), and their purchase intentions on three semantic differential scales ranging from −3 to +3 and anchored with unlikely/likely, impossible/possible, and improbable/probable (α = .89). To control for individual differences in attitudes toward business books in general, this variable was assessed next using the same three semantic differential scales used to measure product attitudes (α = .93). Afterwards, for perceived helpfulness, amidst seven other distractor variables, participants reported how helpful and informative they found the review to be on a 1-to-5 scale where higher numbers reflect greater helpfulness (r = .63, p b .001). To also control for variation in participants' knowledge and interest in business books in general, toward the very end of the study, participants reported how knowledgeable they are about and how interesting they find business books to be on a 1to-5 scale. Finally, to examine how online peer reviews affect attitude formation (not change), participants were asked at the end of the study whether they had heard of the target book prior to the study. Because 23 participants had heard of it and 5 were unsure if they had, their data were deleted, leaving a sample of 173. There 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 were no significant differences across conditions in whether they had heard of the book (χ2s b 2.62, NS), and the results do not meaningfully change when including these participants; the rating × arguments interaction was significant for helpfulness, product attitudes and intentions (Fs N 4.32, ps b .05). than two-sided review (Ms = 3.89 vs. 3.02, F(1, 102) = 14.98, p b .001), whereas the moderately favorable rating was perceived to be more consistent with the two- than one-sided review (Ms = 3.91 vs. 3.25, F(1, 102) = 8.96, p b .005). Results Pretests Two pretests were conducted. The first was designed to test whether the attacked attribute (number of examples) is no more important than the positive attributes. Fifty participants reported how important the criteria (comprehensive, well-written, wellresearched, entertaining and number of examples) are to evaluating business books on a scale from 1 to 5 where higher numbers reflect greater importance. As predicted, the number of examples was rated as less important than how well-written and well-researched the book is (Ms = 4.14 vs. 4.48 and 4.48, respectively, ts(49) N 2.27, ps b .05), and as important as how comprehensive and entertaining the book is (Ms = 4.14 vs. 4.00 and 4.16, respectively, ts(49) b 1). The internal consistency of the reviews was pretested with 108 undergraduates who were randomly assigned to one of the reviews. After reading the review, to assess internal consistency, participants indicated on a scale ranging from 1 (not at all) to 5 (very much) the extent to which (1) the written content of the review supported the reviewer's rating, (2) what the reviewer said was consistent with the rating, and (3) they expected the reviewer's rating given what the reviewer said about the book (α = .75). Participants also reported how typical they found the online review to be and how often they read online reviews (neither of which were significantly affected by rating and/or arguments, Fs(1, 104) b 1). These were covariates in the 2 (rating) × 2 (arguments) ANCOVA. How often they read online reviews was a significant covariate (F(1, 102) = 5.51, p b .05). More importantly, the rating × arguments interaction was significant (F(1, 102) = 22.58, p b .001). As expected, the extremely favorable rating was perceived to be more consistent with the one- Perceived helpfulness, product attitudes and intentions were analyzed with a 2 (rating)× 2 (arguments) ANCOVA while controlling for attitudes toward, knowledge of and interest in business books. Not surprisingly, attitudes toward and interest in business books in general were significant covariates (Fs(1, 163)N 4.30, ps b .05). Supporting H1 and replicating the results of study 1, a rating× arguments interaction was significant (F(1, 163)= 4.66, p b .05). Planned contrasts revealed that reviews were deemed significantly more helpful when they had two- than onesided arguments and the rating was moderately favorable (Ms = 3.40 vs. 2.66, F(1, 163) = 14.68, p b .001), rather than extremely favorable (Ms = 3.10 vs. 2.94, F(1, 163)b 1, see Table 1). Consistent with H2, the rating × argument interaction was also significant for both product attitudes and intentions (Fs(1, 169) N 8.02, ps b .01, see Table 1). Again, not surprisingly, attitudes toward business books in general was a significant covariate for both product attitudes and intentions (both Fs(1, 169) N 26.29, ps b .001), and knowledge of and interest in business books were significant covariates for intentions (Fs(1, 169) N 4.07, ps b .05). More importantly, the benefit of mentioning two sides backfired with extreme ratings. Specifically, when the rating was extreme, two-sided arguments caused lower product attitudes than onesided arguments (Ms = 1.04 vs. 1.48, F(1, 169) = 4.62, p b .05), although this difference was nonsignificant for intentions (Ms = .46 vs. .80, Fs(1, 169) = 1.71, NS). It was only when the reviewer's rating was moderate that two-sided arguments were more persuasive than one-sided arguments (Ms = 1.44 vs. .87, F(1, 169) = 6.87, p b .01 for attitudes; Ms = 1.06 vs. .33, Fs(1, 169) = 7.36, p b .01 for intentions). According to the Table 1 Studies 2–4: pairwise comparisons of means across rating extremity and arguments. Two-sided arguments One-sided arguments F-value Study Rating Dependent variable Extreme rating Moderate rating Extreme rating Moderate rating Extremity × arguments 2 Positive 3 Positive or negative 4 Positive Helpfulness Product attitudes a Intentions a Helpfulness Product attitudes b Intentions b Helpfulness Product attitudes a Intentions a Ability Truthfulness 3.10a 1.04a .46ac 2.62a 4.64a 3.83a 3.41a .91a − .44a 3.14a 3.72a 3.40b 1.44b 1.06b 3.18b 4.80a 4.55bc 3.69b 1.60b .15bc 3.46b 3.66a 2.94ab 1.48b .80bc 2.84ab 5.30b 4.79b 3.43a 1.54b .61b 3.45b 3.26b 2.66a .87a .33a 2.60a 4.72a 4.20ac 2.77c 1.32ab − .30ac 2.91a 3.17b 4.66* 11.56** 8.03** 8.69** 4.61** 8.80** 11.03** 5.32* 7.57** 11.70** .01 Note. Different subscripts reflect significantly different directional tests at p ≤ .05. All reported means from ANCOVAs are covariate-adjusted means. *p b .05 **p ≤ .01. a Product attitudes and intentions are reported on a scale from − 3 to +3 where higher numbers reflect more favorable product attitudes and higher intentions. b Product attitudes and intentions are reported on a 1-to-7 scale where higher numbers reflect greater persuasion toward the reviewer's positive or negative position. 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 232 additive effects model for persuasion, when the ratings are moderate, one- and two-sided reviews should have similar persuasive effects. However, these findings suggest that the consistency between the content of the review and rating may have had greater influence on participants' product judgments than the valence of the content itself did. In further support that the internal consistency of a review matters, extremely (vs. moderately) favorable ratings did not always produce higher product judgments. When the arguments given were two-sided, product judgments were lower with reviews that had extremely than moderately favorable ratings (Ms = 1.44 vs. 1.04, F(1, 169) = 3.91, p = .05 for attitudes; Ms = 1.06 vs. .46, F(1, 169) = 5.42, p b .05 for intentions). Taken together, these findings suggest that in addition to affecting perceived helpfulness, the rating of a review interacts with the arguments provided to influence its persuasiveness. Consequently, neither two-sided arguments nor extremely favorable ratings always led to higher product judgments. Conclusions The results of an online experiment replicate and extend those from a content analysis. Overall, two- (vs. one-) sided messages and extreme (vs. moderate) ratings did not always lead to greater helpfulness and persuasiveness. Specifically, replicating the results of study 1, two-sided reviews were deemed more helpful than one-sided reviews only when the rating was moderately rather than extremely favorable. Moreover, providing further support for an additive effects model that includes internal consistency, two- (vs. one-) sided reviews were not always more persuasive; when the reviewer's rating was extreme, one-sided arguments were as persuasive as or more persuasive than two-sided arguments. Furthermore, an extremely favorable rating did not guarantee higher product judgments. When the accompanying arguments were twosided, a moderately favorable rating produced higher product judgments. Thus, although reviewers may cover multiple sides in an effort to appear helpful, balanced and fair, or to follow a website's advice to include both sides, and in turn, be more persuasive than if they presented a single side, such a strategy can backfire when the reviewer's rating is extremely favorable. Although prior research (e.g., Bazaarvoice, 2008; Chevalier and Mayzlin, 2006) and the results from study 1 indicate that positive reviews are more prevalent than negative reviews, study 3 was designed to test the generalizability of H1 and H2 to negative reviews. It is possible that because it is less socially appropriate to be disparaging than positive (Mizerski, 1982) and/or because negative information is more diagnostic than positive information (Herr et al., 1991), reviewers who provide negative ratings may be deemed especially credible and thus more helpful and persuasive regardless of their written arguments. If either of these accounts are supported, then perceived helpfulness and persuasiveness should be unaffected by whether one or more sides are given in a review and whether this is consistent with the reviewer's rating. That is, the valence of the review would moderate the rating × arguments effects on helpfulness and persuasion. However, if the results generalize 233 to negative reviews, then H1and H2 should be supported regardless of whether the review is positive or negative. Study 3: online experimental analysis of positive vs. negative peer reviews Method Design and sample Two-hundred and thirty-one undergraduates participated in exchange for partial completion of a course requirement. They were randomly assigned to one of eight reviews, which varied across three manipulated variables: rating extremity (moderate vs. extreme), rating valence (positive vs. negative), and arguments (one- vs. two-sided). Materials and procedure The experimental method was the same as study 2 with the following exceptions. First, participants were instructed to imagine that they planned to see a movie and in deciding what to see, visited a website where consumers post their reviews. At the site, they come across a review of Xiao Xiao, an animated martial arts film. To ensure that everyone understood the rating system, participants were told in the negative (positive) condition that the reviewer's rating was on a scale from 1 to 5 stars, where fewer (more) stars reflect less (more) favorable evaluations. The moderate (vs. extreme) rating was 2 (vs. 1) in the negative condition and 4 (vs. 5) in the positive condition. The arguments for the ratings were adaptations of an actual review posted at Yahoo! Movies. For the one-sided/negative review, only negative information was presented: “This flash video was a poor medium for showing all those moves that Jet Li displays in his movies. There's the usual mix of attacks where the hero is invincible, even as he gets double-teamed. The action sequence is completely impossible and unentertaining.” For the one-sided/positive review, only positive information was presented: “This flash video was a great medium for showing all those fun moves that Jet Li displays in his movies. A great mix of attacks, even some real tension as our hero gets double-teamed. The action sequence is completely believable and entertaining.” For the two-sided/positive and two-sided/ negative conditions, the last sentence was replaced with: “Although the action sequence is sort of impossible, this can be accepted for entertainment purposes.” As before, this manipulation is consistent with scoring procedures for message sidedness (Baker-Brown et al., 1992). After reading the review, participants completed a survey, which began by assessing product attitudes, intentions to watch the film and perceived helpfulness using the scales from study 2. Because positive and negative reviews are being compared, the product attitude and intention scores were converted from − 3-to-+3 scores to 1-to-7 persuasiveness scores such that higher scores reflect greater persuasion. Specifically, a persuasiveness score of 7 corresponds to the highest score in the direction consistent with the reviewer's position (i.e., +3 and − 3 for positive and negative reviews, respectively), and a score of 1 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 reflects the score furthest from the reviewer's position (i.e., − 3 and +3 for positive and negative reviews, respectively). Pretest Using the same procedure and questions used in the pretest for study 2, 145 undergraduates were randomly assigned to one of the reviews. A 2 (rating) × 2 (arguments) ANCOVA using how typical they found the review and how often they read reviews yielded a significant rating × arguments interaction (F(1, 139) = 27.03, p b .001; neither covariate was significant, F(1, 139) b 1). As expected, an extreme rating was perceived to be more consistent with one- than two-sided arguments (Ms = 4.14 vs. 3.26, F(1, 139) = 20.84, p b .001), whereas a moderate rating was perceived to be more consistent with two- than one-sided arguments (Ms = 3.76 vs. 3.12, F(1, 139) = 8.63, p b .005). Results Perceived helpfulness, product attitudes and intentions were analyzed with a 2 (extremity) × 2 (valence) × 2 (arguments) ANOVA. Consistent with H1, an extremity × arguments interaction was significant (F(1, 223) = 8.69, p b .005, see Table 1). Replicating the results of the first two studies that examined positive reviews, reviews with two-sided arguments were deemed significantly more helpful than those with one-sided arguments when the rating was moderate (Ms = 3.18 vs. 2.60, F(1, 223) = 6.48, p = .01), but not when the rating was extreme (Ms = 2.62 vs. 2.84, F(1, 223) = 1.16, NS). Moreover, valence did not significantly moderate this interaction (F(1, 223) = 1.48, NS). Thus, regardless of whether the review had a positive or negative rating, two-sided reviews were deemed more helpful than one-sided reviews only when the reviewer's corresponding rating was moderate. Supporting H2 and replicating the results of study 2, an extremity × arguments interaction was significant for both attitudes and intentions (Fs(1, 223) ≥ 4.61, ps ≤ .05, see Table 1). Moreover, supporting the generalizability of H2 to positive and negative reviews, this interaction was not moderated by valence for either dependent variable (Fs(1, 223) ≤ 1.69, NS). In other words, regardless of whether the rating was positive or negative, the persuasive effects of providing one or two sides depended on the extremity of the reviewer's corresponding rating. Specifically, reviewers with an extreme rating were less persuasive when they provided two- than one-sided arguments (Ms = 4.64 vs. 5.30, F(1, 223) = 6.98, p b .01 for attitudes; Ms = 3.83 vs. 4.79 , F(1, 223) = 8.84, p b .005 for intentions). Furthermore, consistent with the proposed additive effects model, when the reviewer's rating was moderate, there was little difference between providing two- and one-sided arguments (Ms = 4.80 vs. 4.72, F(1, 223) b 1 for attitudes; Ms = 4.55 vs. 4.20, F(1, 223) = 1.53, NS for intentions). In fact, mentioning two sides reduced the persuasiveness of an extreme rating; a moderate rating led to higher intentions than an extreme rating when the arguments were two-sided (Ms = 4.55 vs. 3.83, F(1, 223) = 5.67, p b .05 for intentions) and produced similar product attitudes (Ms = 4.80 vs. 4.64, F(1, 223) b 1). Conclusions Study 3 provides evidence that the interactive effects observed in studies 1 and 2 emerge regardless of whether reviews have positive or negative ratings. As predicted, for both positive and negative reviews, perceived helpfulness was higher for two- than one-sided reviews only when the rating was moderate. Furthermore, the results indicate that two-sided arguments are not necessarily more persuasive than one-sided arguments in the context of peer reviews. Specifically, when the reviewers' ratings were extreme, participants were persuaded more by one- than two- sided arguments. These findings suggest that including pros and cons in a review will not always be more helpful or persuasive than presenting only one side (all pros or cons). Likewise, extreme (vs. moderate) ratings will not always lead to more extreme judgments. Study 4 was designed to test credibility as the underlying mechanism of H1 and H2. Theory and research suggest that credibility mediates the relationship between communication and persuasion (Eagly, Chaiken, and Wood, 1981; Hilton, 1995; Laczniak et al., 2001), although whether credibility mediates the relationship between two-sided ads and persuasion has been inconclusive (Crowley and Hoyer, 1994). One reason for this may be that credibility is often measured as a single, unidimensional construct, whereas credibility has been argued to be a multidimensional construct reflecting beliefs about the communicators' ability and their willingness to tell the truth (Eagly and Chaiken, 1993; Hovland et al., 1953). By measuring only one dimension, the effect of the other—and its implications for source credibility and message acceptance—is overlooked. For instance, although consumers might trust that a source is willing to tell the truth, they may discount a message because they question the source's abilities. I propose that it is beliefs about a reviewer's ability rather than truthfulness that mediate the predicted effects. Specifically, unlike advertisers who include some negative information in their ads to convey truthfulness (Crowley and Hoyer, 1994), such an effect is unlikely to generalize to peer reviewers who have less incentive to lie. Instead, consumers likely infer that reviewers are willing to tell the truth (although they may differ in the extent to which they believe this) and instead question the reviewer's expertise. One way to ascertain reviewers' ability to convey the truth is to compare their arguments with their ratings. By definition, experts know more about their area of expertise than nonexperts, and consequently, are more aware of the pros and cons than nonexperts (Brauer et al., 2004; Linville, 1982), although sometimes they provide only a single side rather than multiple sides when giving advice (Brauer et al., 2004). I propose that consumers will infer from two- (vs. one-) sided reviews that reviewers are knowledgeable, but only if the reviewers' corresponding ratings are consistent with these arguments (i.e., moderate rather than extreme). When the corresponding rating is extreme, providing one side should convey that reviewers are knowledgeable more than presenting two sides. If consumers believe that the reviewer is able to tell the truth, then they should accept (rather than discard) the message. Thus, beliefs about an online peer reviewer's ability to tell the truth 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 234 should influence whether they perceive the review to be helpful and are persuaded by it. H3. Beliefs about reviewers' abilities will be higher with a twothan one-sided review when the reviewers' corresponding rating is moderate. In contrast, beliefs about reviewers' abilities will be higher with a one- than two-sided review when the reviewers' corresponding rating is extreme. H4. Beliefs about reviewers' abilities will mediate the interactive effect between ratings and arguments on helpfulness and product judgments. In addition to examining credibility, the robustness of the rating × arguments effects was explored further by comparing the internal consistency of a review with a different type of consistency: consistency with others. For example, even if a review is evaluatively consistent, the review may be inconsistent with others by having a rating that differs from the average product rating. To assess the effects of such a scenario, across conditions in study 4, the reviewer's and average rating were inconsistent; when the reviewer gave the product 4 (vs. 5) stars, the average rating was 5 (vs. 4) stars. If participants are persuaded more by the average star rating than the content of a single review, then their product attitudes and intentions should be higher when the average rating is 5 than 4 stars (even though the reviewer's rating is 4 than 5 stars). Alternatively, participants might compare the arguments provided in a review with the average rating, thereby finding consistency (inconsistency) with others in an internally inconsistent (consistent) review. For instance, although a two-sided review is relatively inconsistent with a reviewer's 5-star rating, it is consistent with an average 4-star rating. As a result, participants may assume that the reviewer covered both the pros and cons to appeal to an audience with less favorable attitudes, and thus beliefs about their abilities are not adversely affected. Likewise, although a one-sided review is relatively consistent with a reviewer's 5-star rating, it is inconsistent with an average 4-star rating. By ignoring the audience's less favorable position, reviewers may seem unaware of other positions, and thus beliefs about their abilities are adversely affected. However, if participants are persuaded more by the text of a peer review than summary statistics (Chevalier and Mayzlin, 2006)—and in particular, are persuaded more by the internal consistency of a review than its consistency with others—then their product attitudes and intentions should be higher only when the reviewer's arguments and ratings are consistent, regardless of the average star rating. Study 4: perceived ability (vs. willingness) to tell the truth as mediators Method One hundred nineteen undergraduates participated in exchange for partial completion of a course requirement. They were randomly assigned to one of four conditions in a 2 (rating: moderate vs. extreme) × 2 (arguments: one- vs. twosided) design. The procedure and materials were the same as 235 those used in study 2 with some exceptions. First, in addition to the book description and one peer review, participants received the average rating of the book from nine reviewers. This rating contradicted the one reviewer's rating such that when the reviewer gave the book 4 (vs. 5) stars, the average rating was 5 (vs. 4) stars. Conditions in which the reviewer's rating was consistent with the average star rating were not run because these conditions would not tease apart the effects of internal consistency from the effects of consistency with others. Second, to measure reviewer credibility, participants were given six statements—three each for the reviewer's ability and truthfulness—and rated their agreement on a 1-to-5 scale where higher numbers reflect greater credibility. An exploratory factor analysis yielded a two- (rather than single-) factor solution with eigenvalues greater than one. A sample item from the ability factor is that the reviewer “is an expert of movies,” and for the truthfulness factor, “is telling the truth” (αs N .6). To reduce demand effects on mediation tests, the variables were measured in reverse order to their predicted effects (Pechmann, 1992); the dependent variables (helpfulness, product attitudes and intentions) were measured before the possible mediators (ability and truthfulness). Third, participants recorded how many people they thought reviewed the film in order to control for this. Results Helpfulness All variables were analyzed with a 2 (rating) × 2 (arguments) ANCOVA while controlling for variation in the number of people participants thought reviewed the film as well as the alternate mechanism (truthfulness). Although truthfulness was a significant covariate (F(1, 114) = 39.75, p b .001), the hypothesized rating × arguments interaction emerged. Specifically, above and beyond the effects of truthfulness, an argument's main effect (F(1, 114) = 11.47, p = .001) was qualified by a rating × arguments interaction (F(1, 114) = 12.61, p = .001). Consistent with H1, the two-sided review was only perceived to be significantly more helpful than the one-sided review when it was more consistent with the reviewer's rating—that is, four stars (Ms = 3.69 vs. 2.77, F(1, 114) = 25.14, p b .001) rather than five stars (Ms = 3.41 vs. 3.43, F(1, 114) b 1, see Table 1). Indeed, a two-sided review was deemed more helpful when the corresponding rating was four than five stars (Ms = 3.69 vs. 3.41, F(1, 114) = 13.72, p b .001). Product judgments For product attitudes and intentions, consistent with H2, the rating × arguments interactions were significant (Fs(1, 114) = 5.29, p b .05 for attitudes; F(1, 112) = 7.59, p b .01 for intentions), despite the truthfulness covariate being significant for attitudes (F(1, 114) = 21.64, p b .001) and intentions (F(1, 112) = 26.03, p b .001). Specifically, consistent with H2, two-sided arguments were less persuasive than one-sided arguments when the rating was extreme (Ms = .91 vs. 1.54, F(1, 114) = 5.32, p b .05 for attitudes; Ms = −.44 vs. .61, F(1, 114) = 6.93, p b .01 for intentions), whereas there were no differences in persuasion 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 between the two- and one-sided reviews when the rating was moderate (Ms = 1.60 vs. 1.32, F(1, 114) b 1 for attitudes; Ms = .15 vs. −.30, F(1, 112) = 1.41, NS for intentions). Providing additional support for a model including internal consistency, when the arguments were two-sided, a four-star rating led to higher attitudes and intentions than a five-star rating (Ms = 1.60 vs. .91, F(1, 114) = 6.67, p = .01 for attitudes; Ms = .15 vs. − .44, F(1, 112) = 3.08, p b .05 one-tailed for intentions). Credibility Consistent with H3, the rating × arguments interaction was significant for beliefs about the reviewer's ability (F(1, 115) = 11.70, p b .001), but not for truthfulness (F(1, 115) b 1, see Table 1). Although participants found reviewers of two- (vs. one-) sided reviews to be more able when their rating was moderate (Ms = 3.46 vs. 2.91, F(1, 113) = 9.92, p b .005), they found reviewers to be less able when their rating was extreme (Ms = 3.14 vs. 3.45, F(1, 113) = 2.60, p = .05 one-tailed). Furthermore, participants found reviewers of two-sided reviews to be more able when the rating was four than five stars (Ms = 3.46 vs. 3.14, F(1, 113) = 3.25, p b .05 one-tailed), and they found reviewers of a one-sided review to be more able when the rating was five than four stars (Ms = 3.45 vs. 2.91, F(1, 113) = 9.33, p b .005). Because bootstrapping techniques are recommended over Baron and Kenny (1986) and Sobel techniques for assessing mediation (Preacher, Rucker, and Hayes, 2007), H4 was assessed with 1,000 bootstrap resamples and a bias-corrected and -accelerated 95% confidence interval (CI). Confidence intervals that do not contain zero are significant at p b .05 (Preacher et al., 2007). As predicted, ability significantly mediated the rating × arguments effects on helpfulness (95% CI: .18 to .74), product attitudes (95% CI: .03 to .66), and purchase intentions (95% CI: .10 to 1.12), whereas truthfulness did not for helpfulness (95% CI: −.11 to .14), product attitudes (95% CI: −.10 to .11), or purchase intentions (95% CI: −.34 to .30). Sobel tests for mediation were consistent with H4 and these conclusions (all three Sobel test statisticsN 2.04, ps b .05 for ability; all three Sobel test statistics ≤ .60, NS, for truthfulness). General discussion Online peer reviews are important from both the customer and merchant perspective. Industry statistics reveal that the majority of consumers research products and services online (Bazaarvoice, 2008; Freedman, 2008) and prefer sites with peer reviews (Bazaarvoice, 2008; Decker, 2007; Freedman, 2008). Moreover, online peer reviews can influence such important marketing outcomes as customer satisfaction (Freedman, 2008) and sales (Chevalier and Mayzlin, 2006; Forman et al., 2008; Godes and Mayzlin, 2004, 2009). Consequently, understanding how consumers use online peer reviews—in particular, which reviews are deemed the most helpful, produce higher product judgments and why—is a timely and important topic. This research yields at least four important contributions: (1) it contributes to the message sidedness literature by showing in a WOM context when and why one-sided arguments are deemed as helpful as and more persuasive than two-sided arguments; (2) it contributes to the online peer review literature by demonstrating the importance of considering both the qualitative and quantitative aspects of online peer reviews when explaining helpfulness and persuasiveness; (3) it contributes to the online peer review and message sidedness literatures by identifying the underlying mechanism through which two- (vs. one-) sided reviews are helpful and persuasive; and (4) it contributes to the attitudinal and social influence literatures by demonstrating that individuals are influenced by the evaluative consistency of WOM communication. First, this research sheds light on the effective use of twosided messages in WOM communication. In the advertising literature, presenting multiple sides can be more persuasive than an entirely favorable (i.e., one-sided) ad because mentioning some negative information is contrary to an advertiser's sales motive (Crowley and Hoyer, 1994). Consequently, two-sided ad claims are attributed to the advertiser's truthfulness rather than expertise. Yet, the current research suggests that different effects emerge in WOM communication contexts, where the speaker has less (if any) incentive to lie. Specifically, instead of affecting perceptions of an online peer reviewers' truthfulness, I argue and find that depending on the reviewers' ratings, two(vs. one-) sided arguments affect perceptions of their ability to tell the truth. As a result, two-sided reviews are not always more helpful and persuasive than one-sided reviews. In fact, when the reviewer's rating was extreme, two-sided reviews were no more helpful and were less persuasive than one-sided reviews. The implications of two (vs. one-) sided arguments for helpfulness and persuasiveness of online reviews has likely been ignored due to the time and cost incurred in analyzing written content (Godes and Mayzlin, 2004). Thus, the second contribution is demonstrating the importance of considering both the qualitative (i.e., written) and the quantitative content of a review (i.e., ratings) to better understand what makes reviews helpful and persuasive. Specifically, the results of four studies—varying in methodology, products evaluated, and whether the reviewers' ratings were positive or negative—collectively support the predictions that whether consumers found a review helpful and were influenced by it depended upon both the extremity of the reviewers' ratings and whether their corresponding arguments contained one or two sides. Consequently, although reviewers are often encouraged to mention a product's pros and cons in their reviews (Burns, 2008; JupiterResearch, 2008) or may present multiple sides to appeal to an audience with unknown (Tetlock et al., 1989) or diverse views (Schlosser, 2005), such reviews were not always deemed the most helpful (studies 1–4) or persuasive (studies 2–4). Such reviews were only more helpful and persuasive when the reviewers' ratings were more evaluatively consistent with a two-sided review (i.e., moderate rather than extreme; see Table 1). Furthermore, extreme ratings did not always produce more extreme product judgments than moderate ratings. When a review contained a product's pros and cons, consumers' product judgments were higher when the reviewer's corresponding rating was moderately than extremely favorable (studies 2–4). It was only when the review contained only one side (the pros) 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 236 that consumers product judgments were higher when the reviewer's rating was extremely than moderately favorable. Moreover, these results generalized across positive and negative reviews; consumers were persuaded more by a review with an extreme than moderate rating only when the corresponding arguments were one- (vs. two-) sided (study 3). Thus, encouraging reviewers to cover multiple sides in their reviews weakened the effects of an extreme rating. Third, study 4 sheds light on the mediating role of credibility on the perceived helpfulness and persuasiveness of online peer reviews. Because most research examining the helpfulness and persuasiveness of online peer reviews have involved content analyses (e.g., Chevalier and Mayzlin, 2006; Forman et al., 2008; Godes and Mayzlin, 2004), the underlying mechanisms have been ignored. In the advertising literature, because of advertisers' financial incentives to lie, the effectiveness of two-sided ads has been largely attributed to increasing beliefs in the advertiser's willingness (rather than ability) to tell the truth. Yet, I propose and find that participants believed more in reviewers' ability to tell the truth when their arguments were more evaluatively consistent with their posted ratings. However, when the reviewers' arguments were less evaluatively consistent with their ratings, participants questioned the reviewers' ability to tell the truth. This interactive effect was only significant for participants' beliefs in the reviewers' ability (vs. willingness) to tell the truth. In fact, the rating × arguments interaction was significantly mediated by ability but not truthfulness. Thus, whether reviewers gave two(vs. one-) sided arguments interacted with rating to affect participants' beliefs about their ability (vs. willingness) to tell the truth, which in turn affected perceived helpfulness and persuasiveness. Finally, this research contributes to the attitudinal and social influence literatures. Specifically, evaluative–cognitive consistency was initially developed to explain how individuals arrive at their own attitudes (Rosenburg, 1956). However, later research suggests many other routes to attitude formation (Eagly and Chaiken, 1993) and that people may be less consistent than previously thought (Cialdini, Trost, and Newsom, 1995). I argue that even if individuals are not always consistent, they likely expect others to be consistent. Indeed, the current research suggests that evaluative–cognitive consistency theory is useful in explaining how people expect others to explain their attitudes. The more consistent speakers' arguments are with their attitudes, the more capable they seem, and the more helpful and persuasive they are likely to be. Managerial implications Taken together, the results suggest that although marketers may wish to attract reviewers who give their products extremely favorable ratings in order to increase purchases, such evangelists do not always produce the highest product attitudes and intentions. Reviewers who mentioned a product's pros and cons were less persuasive when their corresponding ratings were extremely than moderately favorable. In fact, what would seem to be the least favorable online peer review (i.e., a moderately favorable rating with both positive and negative claims) 237 produced product judgments that were as favorable as what would seem to be the most favorable review (i.e., an extremely favorable rating with only positive claims). How four-star/twosided reviews can produce product judgments as high as fivestar/one-sided reviews (and higher than five-star/two-sided reviews) can be explained in terms of the internal consistency of a review and the implications of this for consumers' beliefs about the reviewers' ability. Industry experts recommend educating reviewers on how to write helpful and persuasive online reviews. One recommendation is to present both the pros and cons of a product (Burns, 2008; JupiterResearch, 2008). Indeed, some websites such as epinions. com have fields in the online peer review forms for identifying the pros and cons. Consequently, even a well-intentioned reviewer with an extremely favorable rating may follow instructions to mention both sides in their review and thus be less credible and persuasive because of it. The present results add a qualification to the recommendation to mention multiple sides; reviewers should mention cons only if these cons influenced their product evaluation. If not, and they endorse the product with the highest rating possible, then a more persuasive and credible tactic is to mention only the pros. Thus, to avoid encouraging reviewers to generate cons that they might not otherwise consider or mention, a better strategy may be to instruct reviewers to be consistent: to explain what they liked if they give a product the highest rating possible, and to include what they disliked only if they have a moderately favorable rating. Limitations and future research directions Because the objective of this research was to identify whether and how two- (vs. one-) sided arguments in a review affects how helpful and persuasive a review is, participants in studies 2–4 received a single peer review. Although these studies are not as externally valid as study 1, this tradeoff was made to increase the internal validity of the findings. An interesting extension of this research would be to examine whether the observed effects might vary depending on characteristics of having other reviews available (e.g., its order of appearance among all reviews), and how the reviewer has judged other products. For instance, applying Kelley's (1967) covariation theory, evaluatively consistent reviews should be more persuasive when there is high than low distinctiveness (e.g., the reviewer gives only one product an extremely favorable rating rather than all products extremely favorable ratings). Still, the results of studies 1 and 4 attest to the robustness of the findings. For instance, the results of study 1 were consistent with H1 and replicated in studies 2–4 even though multiple factors (e.g., where the review appeared relative to other reviews, how many other reviews were available, how the reviewer rated other products, etc.) were allowed to vary. Furthermore, the results of study 4 replicated the results of studies 1–3 even though there were summary statistics that contradicted the reviewer's rating. Although it may seem surprising that information about a reviewer (i.e., whether reviewers disclosed their location or gender) was unrelated to perceived helpfulness in study 1, such effects may 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239 depend on the community creating a culture of self-identification (Forman et al., 2008), which may not be present at Yahoo! Movies. Indeed, viewing reviewers' profiles required clicking on a link to access these pages, which consumers may have had little motivation to do. Finally, although this research focused on online peer reviews, the results likely generalize to offline WOM contexts in which consumers are unaware of the speakers' background, such as when discussing a product with another customer in a store. Whereas the results of study 4 indicate that the mediator of the predicted effects is perceptions of reviewer's abilities rather than truthfulness, a worthwhile extension would be to examine whether the mediator might vary depending on cues undermining beliefs that the reviewer is willing to be truthful, such as the presence of incentives or posting on an e-commerce or manufacturer's website. Under such circumstances, consumers may judge reviewers more on the basis of truthfulness than ability, thereby moderating the interactive effects of ratings and message sidedness on credibility, helpfulness and persuasion. Another limitation and thus opportunity for extension is that the products evaluated here (books and movies) were both experience goods. Prior research has found differences in the perceived helpfulness of reviews across product categories (e.g., Sen and Lerman, 2007). For instance, it may be possible that by focusing on experience goods, taste differences could introduce idiosyncratic noise. Yet, such noise would provide a conservative test of the hypotheses. It is possible, however, that because consumers often consult others' opinions more for experiential than search goods (West and Broniarczyk, 1998), they may be less inclined to consult or fully process online peer reviews of search products. Another characteristic of the products used in these studies is that they involve relatively little social and financial risk. On the one hand, because consumers will likely process reviews more carefully when risk is high, the studies presented here would be a conservative test of the predicted effects. On the other hand, when risk— especially social risk—is high, consumers may attend more to the summary statistics than be influenced by any one review. These would be valuable extensions of this research. References Baker-Brown, G., Ballard, E. J., Bluck, S., de Vries, B., Suedfeld, P., & Tetlock, P. E. (1992). The conceptual/integrative complexity scoring manual. In C. P. Smith, & J. W. Atkinson (Eds.), Motivation and personality: handbook of thematic content analysis (pp. 401–418). New York: Cambridge University Press. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51 (6), 117382. Bazaarvoice (2008). Industry Statistics. http://www.bazaarvoice.com/industryStats. html#ConsumerRating Bearden, W., & Etzel, M. J. (1982, September). Reference group influence on product and brand purchase decisions. Journal of Consumer Research, 9, 183–194. Bohner, G., Einwiller, S., Erb, H., & Siebler, F. (2003). When small means comfortable: Relations between product attributes in two-sided advertising. Journal of Consumer Psychology, 13(4), 454–463. Brauer, M., Chambres, P., Niedenthal, P. M., & Chatard-Pannetier, A. (2004, January). The relationship between expertise and evaluative extremity: The moderating role of experts' task characteristics. Journal of Personality and Social Psychology, 86, 5–18. Brown, J., & Reingen, P. H. (1987, December). Social ties and word-of-mouth referral behavior. Journal of Consumer Research, 14, 350–362. Burns, M. (2008). The design guide for online customer reviews. http://www. forrester.com/Research/PDF/0,5110,44373,00.pdf Chaiken, S., & Yates, S. (1985, December). Affective-cognitive consistency and thought-induced attitude polarization. Journal of Personality and Social Psychology, 49, 1470–1481. Chevalier, J. A., & Mayzlin, D. (2006, August). The effect of word of mouth on sales: Online book reviews. Journal of Marketing Research, 345–354. Cialdini, R. B., Trost, M. R., & Newsom, J. T. (1995, August). Preference for consistency: The development of a valid measure and the discovery of surprising behavioral implications. Journal of Personality and Social Psychology, 69, 318–328. Crowley, A. W., & Hoyer, W. D. (1994, March). An integrative framework for understanding two-sided persuasion. Journal of Consumer Research, 20, 561–574. DeCarlo, T. E. (2005). The effects of sales message and suspicion of ulterior motives on salesperson evaluation. Journal of Consumer Psychology, 15(3), 238–249. Decker, S. (2007). 84% Trust user reviews over a critic. Online at. http://www. bazaarblog.com/2007/07/03/84-trust-user-reviews-over-a-criticmarketingsherpa/ Dubrovsky, V. J., Kiesler, S., & Sethna, B. N. (1991). The equalization phenomenon: Status effects in computer-mediated and face-to-face decision making groups. Human–Computer Interaction, 6, 119–146. Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX: Harcourt Brace Jovanovich. Eagly, A. H., Chaiken, S., & Wood, W. (1981). An attribution analysis of persuasion. In John H. Harvey, William Ickes, & Robert F. Kidd (Eds.), New directions in attribution research, vol. 3 (pp. 37–62). Hillsdale, NJ: Erlbaum. Eisend, M. (2006). Two-sided advertising: A meta-analysis. International Journal of Research in Marketing, 23(2), 187–198. Eisend, M. (2010). Explaining the joint effect of source credibility and negativity of information in two-sided messages. Psychology & Marketing, 27(11), 1032–1049. Etgar, M., & Goodwin, S. A. (1982). One-sided versus two-sided comparative message appeals for new brand introductions. Journal of Consumer Research, 8(4), 460–465. Florack, A., Ineichen, S., & Bieri, R. (2009). The impact of regulatory focus on the effects of two-sided advertising. Social Cognition, 27(1), 37–56. Forman, C., Ghose, A., & Wiesenfeld, B. (2008, September). Examining the relationship between reviews and sales: The role of reviewer identity disclosure in electronic markets. Information Systems Research, 19, 291–313. Freedman, L. (2008). Merchant and customer perspectives on customer reviews and user-generated content. Online at. www.powerreviews.com/socialshopping/solutions/whitepaper/2008_WhitePaper_0204_4.pdf Gershoff, A., Mukherjee, A., & Mukhopadhyay, A. (2003). Consumer acceptance of online agent advice: Extremity and positivity effects. Journal of Consumer Psychology, 13, 161–170. Godes, D., & Mayzlin, D. (2004, July-August). Using online conversation to study word-of-mouth communication. Marketing Science, 23, 545–560. Godes, D., & Mayzlin, D. (2009, July–August). Firm-created word-of-mouth communication: Evidence from a field test. Marketing Science, 28, 721–739. Golden, L., & Alpert, M. I. (1987). Comparative analysis of the relative effectiveness of one- and two-sided communication for contrasting products. Journal of Advertising, 16(1), 18–25. Herr, P. M., Kardes, F. R., & Kim, J. (1991, March). Effects of word-of-mouth and product-attribute information on persuasion: An accessibility–diagnosticity perspective. Journal of Consumer Research, 17, 454–462. Hilton, D. J. (1995, September). The social context of reasoning: Conversational inferences and rational judgment. Psychological Bulletin, 118, 248–271. Hoffman, D. L., & Novak, T. P. (1996, July). Marketing in hypermedia computer-mediated environments: Conceptual foundations. Journal of Marketing, 60, 50–68. 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 238 Hovland, C. I., Janis, I. L., & Kelly, H. H. (1953). Communication and persuasion: Psychological studies of opinion change. New Haven, CT: Yale University Press. Jones, E., & Davis, K. (1965). From acts to dispositions: The attribution process in person perception. In Leonard Berkowitz (Ed.), Advances in experimental social psychology, vol. 2 (pp. 219–266). New York: Academic Press. JupiterResearch (2008). Online consumer opinions/use of user generated content in the UK. http://www.bazaarvoice.com/whitePapers.html Kamins, M. A., & Assael, H. (1987, February). Two-sided versus one-sided appeals: A cognitive perspective on argumentation, source derogation, and the effect of disconfirming trial on belief change. Journal of Marketing Research, 24, 29–39. Kamins, M. A., Brand, M. J., Hoeke, S. A., & Moe, J. C. (1989). Two-sided versus one-sided celebrity endorsements: The impact on advertising effectiveness and credibility. Journal of Advertising, 18(2), 4–10. Kamins, M. A., & Marks, L. J. (1988). An examination into the effectiveness of two-sided comparative price appeals. Journal of the Academy of Marketing Science, 16(2), 64–71. Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska symposium of motivation, vol. 15 (pp. 192238). Lincoln: University of Nebraska Press. Kelley, H. H. (1973). The processes of causal attribution. The American Psychologist, 28, 107–128. Laczniak, R. N., DeCarlo, T. E., & Ramaswami, S. N. (2001). Consumers' responses to negative word-of-mouth communication: An attribution theory perspective. Journal of Consumer Psychology, 11(1), 57–73. Linville, P. W. (1982, February). The complexity-extremity effect and age-based stereotyping. Journal of Personality and Social Psychology, 42, 193–211. Mizerski, R. (1982, December). An attribution explanation of the disproportionate influence of unfavorable information. Journal of Consumer Research, 9, 301–310. Pechmann, C. (1992, November). Predicting when two-sided ads will be more effective than one-sided ads: The role of correlational and correspondent inferences. Journal of Marketing Research, 29, 441–453. Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated mediation hypotheses: Theory, methods, and prescriptions. Multivariate Behavioral Research, 42, 185–227. Rosenburg, M. J. (1956). Cognitive structure and attitudinal affect. Journal of Abnormal and Social Psychology, 53, 367–372. 239 Schlosser, A. E. (2005, September). Posting versus lurking: Communicating in a multiple audience context. Journal of Consumer Research, 32, 260–265. Schlosser, A. E., & Shavitt, S. (1999). Effects of an approaching group discussion on product responses. Journal of Consumer Psychology, 8(4), 377–406. Schlosser, A. E., & Shavitt, S. (2002, June). Anticipated discussion and product judgments: Rehearsing what to say can affect your judgments. Journal of Consumer Research, 29, 101–115. Scott, W. A. (1969, August). Structure of natural cognitions. Journal of Personality and Social Psychology, 12, 261–278. Sen, S., & Lerman, D. (2007). Why are you telling me this? An examination into negative consumer reviews on the Web. Journal of Interactive Marketing, 21(4), 76–94. Sengupta, J., & Johar, G. V. (2002). Effects of inconsistent attribute information on the predictive value of product attitudes: Toward a resolution of opposing perspectives. Journal of Consumer Research, 29(1), 39–56. Settle, R. B., & Golden, L. L. (1974). Attribution theory and advertiser credibility. Journal of Marketing Research, 11(4), 181–185. Simonson, I., & Nowlis, S. M. (2000, June). The role of explanations and need for uniqueness in consumer decision making: Unconventional choices based on reasons. Journal of Consumer Research, 27, 49–68. Stern, B. (1994). A revised communication model for advertising: Multiple dimensions of the source, the message and the recipient. Journal of Advertising, 23(2), 6–15. Sultan, F., Parley, J. U., & Lehmann, D. R. (1990, February). A meta-analysis of applications of diffusion models. Journal of Marketing Research, 27, 70–77. Swinyard, W. R. (1981). The interaction between comparative advertising and copy claim variation. Journal of Marketing Research, 18(2), 175–186. Tetlock, P. E. (1983, July). Accountability and complexity of thought. Journal of Personality and Social Psychology, 45, 74–83. Tetlock, P. E., Skitka, L., & Boettger, R. (1989, October). Social and cognitive strategies for coping with accountability: Conformity, complexity, and bolstering. Journal of Personality and Social Psychology, 57, 632–640. West, P. M., & Broniarczyk, S. M. (1998, June). Integrating multiple opinions: The role of aspiration level on consumer response to critic consensus. Journal of Consumer Research, 25, 38–51. Wood, W., & Eagly, A. (1981). Stages in the analysis of persuasive messages: The role of causal attributions and message comprehension. Journal of Personality and Social Psychology, 40(2), 246–259. 15327663, 2011, 3, Downloaded from https://myscp.onlinelibrary.wiley.com/doi/10.1016/j.jcps.2011.04.002 by <shibboleth>-member@uni-bremen.de, Wiley Online Library on [22/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License A.E. Schlosser / Journal of Consumer Psychology 21 (2011) 226–239