Maybe I Just Got (Un)lucky: One-on-One Conversations and the Malleability of Post-Consumption Product and Service Evaluations

Maybe I Just Got (Un)lucky: One-on-One Conversations and the Malleability of Post-Consumption... Abstract This research focuses on the persuasive impact of a common yet understudied form of word of mouth (WOM): one-on-one conversations in which consumers share and compare past experiences with a product or service. In contrast to prior work on WOM influence, we discover a “positivity effect” in these conversations, such that consumers who share a negative experience form more favorable overall judgments after speaking with someone who had a positive experience, but consumers who share a positive experience are unaffected by learning about another’s negative experience. This effect is mediated by consumers’ dismissal of their own negative experience as a temporary or one-off event in light of the other person’s contrasting positive experience, and is facilitated by positive consumer expectations of product and service performance. We also identify a key boundary condition whereby the positivity effect of one-on-one conversations is moderated by whether consumers have positive or negative expectations of product or service performance. When expectations are negative, the positivity effect is dampened and a negativity effect emerges. word of mouth, experience sharing, consumer expectations, attribution, positivity effect Imagine that you have just purchased the newest iPhone. After one week, the phone’s camera begins to malfunction, leaving you unable to take pictures. You happen to run into a friend and, noticing that she has the same phone, begin to complain about your own experience with it. To your surprise, your friend responds that her phone has been working great and that its camera has been functioning perfectly. How would you react to this conversation? More specifically, how would this casual discussion change your overall views of the iPhone in terms of your attitude, satisfaction, and repurchase intentions? Based on prior word-of-mouth (WOM) research (Chevalier and Mayzlin 2006; Herr, Kardes, and Kim 1991; Schlosser 2005), as well as findings showing that negative (vs. positive) information is more perceptually salient and elicits greater psychological responses (Kahneman and Tversky 1979; Peeters and Czapinski 1990), you might assume that your negative experience would stick, making you unlikely to reconsider your opinion of the phone, even when presented with your friend’s positive experience. However, our findings present evidence to the contrary. Specifically, we show that when consumers share differing accounts of a product’s or service’s performance, those who had a negative experience form more favorable overall evaluations of the product or service, including greater repurchase intentions, after talking to another person who had a contrasting positive experience. On the contrary, those who had a positive experience do not form more unfavorable evaluations after talking to another person who had a contrasting negative experience. This asymmetry suggests a positivity effect1 in these types of conversations, as evaluations formed by a negative (vs. positive) experience appear more malleable and susceptible to the influence of a conversation partner’s contrasting opinion, but not vice versa. To explain this counterintuitive effect, we develop a framework to better understand the persuasive influence of one-on-one (or dyadic) conversations in which consumers share and compare past personal experiences with a product or service. This form of WOM appears to be surprisingly common, as 75% of consumer conversations occur in a face-to-face setting or over the phone (Enda and Mitchell 2013; Keller Fay Group 2014). Additionally, 50–80% of WOM conversations involve consumers recounting direct personal experiences to one another, as opposed to merely passing along firm-generated content such as commercials or celebrity endorsements (Bughin, Doogan, and Vetvik 2010). Prior work has looked at certain aspects of these conversations by studying “storytelling” and its effect on how the sharer evaluates a product or service after talking about an experience with another person (Cowley 2014; Moore 2012). However, little research has investigated more complex interactions in which the consumer shares and then listens to another person’s experience with the same product or service. Thus, our work provides new insight into how consumers in an increasingly homogenized retail landscape react when they share and compare contrasting (vs. similar) personal experiences with the same product or service. We draw on work in attribution theory (Chen and Lurie 2013; Folkes 1984, 1988), disconfirmation bias (Ditto and Lopez 1992; Ditto et al. 1998; Edwards and Smith 1996), and consumer expectations (Folkes and Patrick 2003; Fornell et al. 1996; Johnson, Anderson, and Fornell 1995) to explain why one-on-one WOM conversations are characterized by a positivity effect. We suggest that learning of a conversation partner’s contrasting experience can lead consumers to attribute their own experience to temporary performance heterogeneity, or dismiss it as a one-time event, as opposed to attributing their experience to how the product or service typically performs. We argue that because consumers generally have positive expectations of product and service performance (Fornell et al. 1996; Johnson et al. 1995), they are more likely to dismiss a negative (vs. positive) prior experience as a one-time event upon learning of another person’s contrasting experience. Such dismissal leads consumers to perceive their own negative experience as less diagnostic (Folkes and Patrick 2003), and underlies positive shifts in evaluation. Supporting this proposed process, we show the reverse is also true: when product and service expectations are negative, as opposed to positive, this positivity effect is attenuated and a negativity effect actually occurs. These findings broadly demonstrate that when a product and service experience violates expectations, consumers more readily dismiss it in light of contradictory WOM information. Our research makes several contributions. First, we believe we are the first to provide a comprehensive framework of the persuasive influence of one-on-one conversations where consumers share similar and contrasting direct product and service experiences. Prior work studying consumer responses to contrasting opinions has primarily looked at contexts in which participants communicated within groups or with multiple anonymous others (Herr, Kardes, and Kim 1991; Schlosser 2005). Ours, however, is the first study to examine this phenomenon in a “narrowcasting” WOM context, where consumers are more interested in sharing useful information than making a positive impression on others (Barasch and Berger 2014). Although we focus on private, face-to-face conversations due to their commonality and their significant role in the transmission of WOM, in the general discussion we detail other WOM contexts with similar characteristics where our framework is also applicable. Next, we identify a novel WOM context that elicits a positivity effect, whereby positive information about a product or service is more persuasive than negative information. Although prior work has typically found negativity bias in WOM (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005), we leverage more recent research that has found attenuation of negativity bias in certain contexts (Chen and Lurie 2013), while building on the critical role of positive consumer expectations to make a case for the greater persuasion of positive WOM (Fornell et al. 1996; Johnson et al. 1995). Finally, in identifying dismissal as the attributional process underlying our effect, we believe we are the first to show how and why consumers update initially negative and positive evaluations based on post-consumption conversations with others. THEORETICAL BACKGROUND One-on-One Conversations and the Sharing of Contrasting Experiences Our research focuses on dyadic conversations in which both individuals have had a past personal experience with the same product or service. In our conceptualization of these conversations, consumers both share and compare accounts of satisfactory and unsatisfactory experiences, leading them to make attributions about a product or service’s performance (Folkes and Kotsos 1986). Therefore, we do not focus on the sharing of opinions within more taste-oriented domains, such as movies (He and Bond 2015; Schlosser 2005) or art, but rather on performance contexts where it is less ambiguous whether a product or service did or did not meet expectations. We are particularly interested in one-on-one conversations where consumers share contrasting accounts of a product or service’s performance, and the effect of this disagreement on subsequent consumer evaluations. These types of exchanges are common, as consumers are often eager to share their “true” opinions with one another, especially when products and services are the center of discussion (Chen and Berger 2013; Dubois, Bonezzi, and De Angelis 2016). Supporting this argument, when we presented 65 undergraduates with a scenario in which a classmate told them about a product experience that was the opposite of their own, 85% of participants said they would respond by sharing their own contrasting experience with the classmate (see appendix A). Why Negativity Bias Is Dampened in One-on-One Conversations What happens when these disagreements arise? Past research examining WOM influence suggests that negative shared experiences will be more influential than positive shared experiences (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005), implying a negativity bias. However, recent work has demonstrated that negativity bias in WOM is mitigated under certain circumstances. In particular, Chen and Lurie (2013) argued that negativity bias occurs only because consumers are more likely to attribute negative WOM to the product or service being mentioned, whereas they are more likely to attribute positive WOM to the sharer’s personal motives (e.g., impression management, social positivity norms), which often have little to do with the product’s or service’s actual performance. Supporting this notion, they found that in a context where consumers equally attributed positive and negative online reviews to the product or service itself (vs. the reviewer’s personal motives), they were equally persuaded by both—in other words, negativity bias was attenuated. Building on this finding, we posit that negativity bias is similarly attenuated in one-on-one conversations because consumers, without regard to valence, predominantly attribute the experiences shared in these conversations to the product or service itself (vs. the personal motives of the other person) (Chen and Lurie 2013). Specifically, in dyadic conversations, consumers narrowcast, or recount their experience to just one other person. Prior work suggests that narrowcasting leads consumers to focus more on sharing useful information and less on trying to make a positive impression (Barasch and Berger 2014; Tice et al. 1995). Drawing on these findings, we posit that consumers in dyadic conversations likely share accurate information about their own experience out of a desire to be honest and helpful, which leads them to assume that their conversation partner has the same motive (Epley et al. 2004; Nickerson 1999). As a result, we propose that within dyads, consumers predominantly attribute both positive and negative shared experiences to the product or service being discussed (vs. the personal motives of the other person), thereby attenuating negativity bias (Chen and Lurie 2013). To test this assertion, we asked 80 Amazon Mechanical Turk (MTurk) participants to imagine they were in a one-on-one conversation with a coworker where they shared contrasting experiences at a restaurant (either positive or negative; see appendix B). Following Chen and Lurie (2013), participants first indicated the extent to which they attributed the coworker’s sharing of their experience to personal motives. They then rated the extent to which they attributed the sharing to the restaurant itself.2 We then calculated a causal score by subtracting personal attributions from restaurant experience attributions (MRestaurant = 7.23, SD = 1.40 vs. MPersonal = 6.36, SD = 1.38), where higher scores indicated greater restaurant (vs. personal motive) attributions. A one-sample t-test revealed that the difference score (M = .87, SD = 2.13) was significantly greater than 0 (t(79) = 3.63, p = .001), indicating that participants predominantly attributed the coworker’s WOM to the restaurant. Consistent with our argument, this attribution was equivalent for positive and negative experiences shared by the coworker (MPositive = .76, MNegative = .97; F(1, 78) = .20, NS). These results suggest that consumers in dyadic conversations attribute their conversation partner’s experience to the product or service significantly more than personal motives, regardless of valence. How Sharing Contrasting Experiences Shapes Post-Consumption Evaluations If individuals in dyadic conversations do not rely on simple valence heuristics to determine the accuracy of one another’s experiences, how do they respond when another person’s experience contrasts from their own? Although consumers tend to view their own experiences as highly diagnostic (Herr et al. 1991), they also understand that heterogeneity exists among products and services (He and Bond 2015; Parasuraman, Zeithaml, and Berry 1985). For instance, consumers recognize that some cars in a production line can be dismissed as “lemons” due to manufacturing glitches, while others in the same line perform perfectly fine. Supporting this assertion, Folkes (1984) found that 67% of participants thought an unexpected restaurant experience could be due to a one-time occurrence and was not typical of its chronic performance. This heterogeneity should be particularly salient in exchanges where another’s experience directly contradicts the consumer’s own experience. When this happens, we posit that the consumer will interpret heterogeneity based on her general expectations and beliefs of how products and services typically perform. If a consumer perceives that her conversation partner’s contrasting experience is more consistent with how she would expect the product or service to perform, she should attribute her own experience to temporary performance heterogeneity, or the possibility that her experience was a one-time event (Folkes 1984, 1988). On the contrary, if she perceives that her conversation partner’s experience is less consistent with expectations than her own experience, she should be less likely to attribute her own experience to performance heterogeneity, leaving her evaluations unchanged. Our argument aligns with the idea that people are more critical of information that is inconsistent with their general beliefs and expectations, while uncritically accepting of information that is consistent with these beliefs and expectations (Ditto and Lopez 1992; Ditto et al. 1998). Importantly, past studies have found that consumers are generally satisfied with products and services and tend to have positive expectations regarding their performance (Folkes and Patrick 2003; Fornell et al. 1996; Johnson et al. 1995). Indeed, when we asked 66 undergraduates to rate how they would expect each of 10 popular products and services over $50 (e.g., smartphone, laptop, dinner out) to perform,3 they had very high expectations (MPerformance = 6.08 / 7, SD = .70) (appendix C). Accordingly, we posit that when a consumer shares a negative experience and learns of a conversation partner’s contrasting positive experience, she will perceive her own negative experience as less consistent with positive expectations, causing her to scrutinize her negative experience and attribute it to temporary performance heterogeneity (i.e., dismiss it as a one-time event). This should lead her to see her negative experience as less indicative of the product or service’s chronic performance (Folkes and Patrick 2003), and make her less likely to use this experience in forming a subsequent judgment (Lynch, Marmorstein, and Weigold 1988). As a result, she should adjust her evaluations upward, in line with positive expectations. On the other hand, if she shares a positive experience and her conversation partner shares a contrasting negative experience, her own experience is more consistent with positive expectations, so she will be less likely to scrutinize this positive experience (vs. her partner’s negative experience) and dismiss it as a one-time event. As a result, no shift in her evaluations should occur. Importantly, when the consumer and conversation partner share similar experiences, we do not expect shifts in evaluations since performance heterogeneity should not be salient and consumers have no reason to question the attribution of their own experience or change their evaluation. Accordingly, we consider our control conditions to be those in which the consumer and conversation partner share similar experiences. Thus, our focal contrasts are those that examine the shifts in evaluation when consumers learn of a partner’s contrasting (vs. similar) experience. We present our predictions for the basic positivity effect and its underlying mechanism below: H1:There will be an interactive effect of the consumer’s own experience and her conversation partner’s experience on shifts in the consumer’s general evaluations of a product or service: When a consumer shares a negative experience and learns of a partner’s contrasting positive (vs. similar negative) experience, the consumer will form a more favorable evaluation of the product or service. When the consumer shares a positive experience and learns of a partner’s contrasting negative (vs. similar positive) experience, the consumer will not change her evaluation of the product or service. H2:When the consumer shares a negative experience and learns of a conversation partner’s contrasting positive (vs. similar negative) experience, the consumer’s attribution (or dismissal) of her own experience as a temporary or one-time event will mediate more favorable evaluations. To this point, we have argued that consumers adjust evaluations of their own negative experience upward because they perceive this experience to be less consistent with positive expectations than the conversation partner’s positive experience. We now further investigate this argument by looking at how negative expectations can reverse this effect. The Case of Negative Expectations What happens when consumers have negative expectations? Consistent with our theorizing, we argue that when a consumer with negative product or service expectations has a negative experience and then learns of a conversation partner’s contrasting positive experience, he should perceive his own negative experience as more consistent with expectations. As a result, he should not dismiss his negative experience as a one-time event, leaving initial evaluations unaffected. Conversely, if this consumer has a positive experience and learns of a partner’s contrasting negative experience, he should see his own experience as less consistent with expectations than the partner’s experience, leading him to dismiss his own positive experience as a one-time event, downwardly shifting initial evaluations. In sum: H3:There will be an interactive effect of consumer expectations, the consumer’s own experience, and the conversation partner’s experience on shifts in consumers’ general evaluations of a product or service: When expectations are positive, we will replicate our hypothesis 1 predictions. When expectations are negative, our results will reverse our hypothesis 1 predictions, more specifically: When a consumer shares a negative experience and learns of a partner’s contrasting positive (vs. similar negative) experience, the consumer will not change his evaluation. When a consumer shares a positive experience and learns of a partner’s contrasting negative (vs. similar positive) experience, the consumer will form a more unfavorable evaluation. We summarize our predictions and the studies in which they are tested in table 1. TABLE 1 SUMMARY OF FOCAL PREDICTIONS Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift TABLE 1 SUMMARY OF FOCAL PREDICTIONS Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift We test our predictions in four studies across a diverse array of one-on-one WOM conversations. In studies 1 and 2, we use novel in-person conversation settings to show that when participants share a negative experience with a conversation partner who shares a contrasting positive experience, their initial evaluations shift upward. On the other hand, no shift occurs when participants share a negative experience and learn of the partner’s contrasting positive experience (hypothesis 1). In study 3, we replicate this effect in a vignette-based conversation and find support for dismissal as our proposed process (hypothesis 2). Finally, in study 4, we manipulate performance expectations to moderate the persuasive effect of a conversation partner’s contrasting experience (hypothesis 3). To demonstrate the generalizability and external validity of our findings, we investigate a range of real and simulated dyadic relationships including those with a likeable other (studies 1 and 2), close friend (study 3), and casual acquaintance (study 4). Moreover, we test products and services that participants 1) are obliged to use for a specific task (study 1); 2) are structurally constrained to use within campus life (study 2); 3) chose themselves in real life (study 3); and 4) chose as part of a scenario (study 4). Such variation allows us to demonstrate that this phenomenon applies to products with varying price points and levels of involvement. STUDY 1 In study 1, we provide initial evidence that when consumers share a negative product experience with a conversation partner who shares a contrasting positive (vs. similar negative) experience, they will form more favorable evaluations. We expect no such shift when consumers share a positive product experience and the partner shares a contrasting negative experience (hypothesis 1). To test this prediction, we randomly assign participants to have a negative or positive experience with a university-branded pen and measure time 1 evaluations of it. Next, as an ostensibly separate study on conversation styles, we randomly assign participants to have a conversation with a likeable “student” confederate partner who happens to mention a contrasting or similar experience with the same pen. We then measure time 2 pen evaluations. Method Participants and Design We randomly assigned 139 undergraduates at a large southwestern university to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed experiment. We removed nine participants with whom the confederate could not share his opinion of the pen before the conversation ended, leaving 130 participants. Procedure Upon entering the lab, we asked participants to fill out a sheet of basic biographical information (e.g., year of birth, favorite food) for a study later in the session. We provided all participants with a university-branded ballpoint pen with which to complete this sheet. Pretesting indicated that participants had reasonably positive (though not extreme) expectations of the pen’s performance—that is, significantly above a neutral midpoint of 4 out of 7 (M = 4.92; t(267) = 11.13, p < .001) (see appendix D for further details). The pen thus represented a strong initial test of our theory since it allowed us to test whether participants would dismiss a negative experience based on the moderately positive performance expectations of a mundane product. In the negative experience condition, we had intentionally damaged the ballpoint tip of the pen by briefly holding it up to a flame. Thus, the pen was not visibly damaged yet wrote poorly, resulting in an unambiguously negative product experience. In the positive experience condition, the pen wrote normally. After participants finished the first task, we gave them a new survey stating that we were collecting feedback on the behavioral lab and asked them to evaluate the usability of some lab equipment, such as workstations. As part of this survey, we asked participants to complete a general evaluation of the university-branded pens using a three-item index of attitude, satisfaction, and usage likelihood (α = .93): 1) “How would you rate your attitude toward the [university] pens that we have provided for these studies?” 2) “How satisfied were/are you with the [university] pens that we have provided for these studies?” and 3) “How likely would you be to use the [university] pens that we have provided for these studies?” Here, and in all future studies, across the time 1 and time 2 evaluation questions, we purposefully did not put numbers on the scales (only word labels) to make it more difficult for participants to anchor on initial responses. After a brief delay, we informed participants that they had been selected to participate in an unrelated study on conversation styles and that, as part of the study, we wanted them to talk to another student participant. Unbeknownst to the participant, the conversation partner was an actor paid to pose as a fellow undergraduate. A lab assistant led the participant and conversation partner to a separate room and instructed them to have a short discussion about two things they liked about being a student at the business school. After a short pause, the confederate shared two things he liked about being a business school student and then listened as the participant shared what he or she liked. This conversation was meant to encourage rapport between the participant and the partner and foster likeability (Sinclair et al. 2005) to make the participant more comfortable with the confederate, thereby leading to a more natural interaction. Next, the conversation partner casually asked the participant: “Hey, did they have you take that study where you had to evaluate the lab equipment?” After the participant responded that he or she had done the study, the partner continued, “Did they ask you to rate the pen for some reason? Did yours work okay?” After the participant shared their experience with the pen, the confederate responded by indicating that he had had either a contrasting or similar experience: “That’s interesting, the pen they gave me didn’t write all that well (wrote perfectly fine).” We felt such an exchange would not feel odd or out of place given that participants had just rated the pen and likely viewed it as something common to discuss with the confederate. Finally, at the end of the conversation, the confederate pointed to a bowl of candy and offered some to the participant to further foster likeability (Sinclair et al. 2005). We included this behavior to 1) increase the contextual similarity to a conversation with a likable other, a person with whom these types of discussions occur most frequently (Enda and Mitchell 2013; Keller Fay Group 2014), and 2) decrease impression management pressures, as people are more comfortable being themselves around people they feel close to (Kraus and Chen 2009; Kwang and Swann 2010). At this point, the lab assistant knocked on the door and instructed participants to return to their seats, where they were asked to complete a survey further exploring lab equipment perceptions. We asked participants to reconsider their prior evaluations and respond to the same three-item index for the university-branded pens (α = .91). Finally, although all responses to our questions were completed in private, to account for the possibility that impression management motives were stronger in the negative (vs. positive) own experience condition, we asked participants: “To what extent did you try to make a positive impression on the person you talked to as part of this study? (1 = not at all; 7 = very much so).” Results and Discussion We predicted that participants who shared a negative experience would form more favorable evaluations following exposure to the conversation partner’s contrasting positive (vs. similar negative) experience, while participants who shared a positive experience would not change their evaluations following exposure to the partner’s contrasting negative (vs. similar positive) experience (hypothesis 1). Across all studies, for ease of exposition and increased clarity, we first report the results of our mixed-design ANOVA. We then present within-subjects time 1 to time 2 contrasts, which test whether the magnitude of the evaluation shift is greater than 0. Finally, we follow with our focal between-subjects contrasts of the t2 – t1 difference scores. These complex contrasts test hypothesis 1 by comparing shifts in participant evaluations depending on whether they interacted with a contrasting (vs. similar) partner. Main Results We conducted a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed ANOVA on our evaluation index. There was a main effect of own experience valence (MPositive = 4.91 vs. MNegative = 2.88; F(1, 126) = 83.83, p < .001). Of note, initial evaluations of negative experiences at time 1 fell well below neutral (4) on a seven-point scale, indicating that we successfully manipulated an undoubtedly negative experience in the lab. On the other hand, there was no main effect of the conversation partner’s experience (F(1, 126) = .00, NS) or time (before [t1] vs. after [t2] the conversation) (F(1, 126) = 1.67, p = .20). There were also no significant two-way interactions. Importantly, we found a significant own experience × partner’s experience × timing interaction (F(1, 126) = 7.61, p < .01).4 We next examine our within- and between-subjects contrasts. As described above, because our focal predictions are based on the changes in evaluations in the contrasting and similar conditions, figure 1 and all subsequent figures provide an illustration of shifts in participant evaluations in the form of difference scores before and after the manipulation. Consistent with past research looking at evaluation changes over time (Ramanathan and Williams 2007), we use proc mixed to test both within- and between-subjects planned contrasts. All statistical tests retain the full degrees of freedom. We also report effect sizes for each hypothesized contrast.5 For completeness, we first present within-subjects t2/t1 contrasts across all conditions, and then test hypothesis 1 with our focal complex contrasts. FIGURE 1 View largeDownload slide STUDY 1: SHIFT IN EVALUATIONS (T2 – T1) FIGURE 1 View largeDownload slide STUDY 1: SHIFT IN EVALUATIONS (T2 – T1) Our within-subjects contrasts revealed that among participants sharing a negative experience (or negative sharers), when the conversation partner had a contrasting positive experience, we observed an upward shift in evaluations (t1 = 2.60 vs. t2 = 3.03, Δt2–t1 = .43; t(126) = 3.02, p = .003). On the other hand, when the partner had a similar negative experience, evaluations did not shift (t1 = 3.00 vs. t2 = 2.90, Δt2–t1 = –.10; t(126) = –.74, NS). Among participants who shared a positive experience (or positive sharers), we found no change in evaluations when the partner shared a contrasting negative experience (t1 = 5.03 vs. t2 = 4.92, Δt2–t1 = –.11; t(126) = –.77, NS) or a similar positive experience (t1 = 4.76 vs. t2 = 4.92, Δt2–t1 = .16; t(126) = .99, NS). To test hypothesis 1, we performed between-subjects contrasts of the t2 – t1 difference scores among participants in the contrasting and similar partner conditions. As stated above, we consider the similar partner experience condition our control. Supporting hypothesis 1a, negative sharers whose partner shared a contrasting positive experience increased their evaluations more than those whose partner shared a similar negative experience (ΔContrasting =.43 vs. ΔSimilar = –.10; ηG2= ⁠.055; F(1, 126) = 1.57, p < .01). Supporting hypothesis 1b, positive sharers whose partner shared a contrasting negative experience did not differ from those whose partner shared a similar positive experience (ΔContrasting = –.11 vs. ΔSimilar = .16; ηG2= ⁠.012; F(1, 126) = 1.57, p = .21). Impression Management A 2 (own experience) × 2 (conversation partner’s experience) between-subjects ANOVA on our measure of participants’ desire to please the partner revealed a nonsignificant two-way interaction (F(1, 126) = .97, NS) and no significant contrasts, suggesting no differences in impression management across conditions. Further, a 2 (own experience) × 2 (partner’s experience) × 2 (time) mixed ANOVA controlling for impression management showed that our hypothesized three-way interaction remained significant (F(1, 125) = 7.01, p < .01). Debriefing Our study focal effects and contrasts remained robust to any concerns highlighted in the participant debriefing responses.6 Discussion In study 1, we provide preliminary evidence in support of hypothesis 1. When participants had a negative experience with the pen, they evaluated it more positively when the conversation partner shared a contrasting positive (vs. similar negative) experience. However, when participants had a positive experience, the partner’s contrasting negative experience did not influence evaluations. Further, a measure of impression management provides preliminary evidence that it is unlikely our results are driven by participants wanting to be seen more positively after recounting a negative (vs. positive) experience. While we recognize that the upward shift in evaluations is relatively small, it does represent a nontrivial 17% improvement in initial evaluations, and is consistent with research showing that people typically modify their beliefs gradually and incrementally (Weber and Crocker 1983). Most importantly, in demonstrating this effect, we reverse prior work showing a negativity bias in WOM (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005). Of course, we recognize that because participants only briefly used the pen and it is a lower-involvement product, they may not have given much thought or concern to its failure or success. As a result, they may have formed less concrete initial evaluations, and thus upward shifts in pen evaluations might not generalize to a higher-involvement product or service used on a regular basis. We address this issue in study 2. STUDY 2 In study 2, we seek to replicate our results by asking students to recall and discuss an on-campus service experience, which we posit is more meaningful to them and important to their daily life. We first ask undergraduate participants to think about their experiences with on-campus services. This provides a context where all participants have shared knowledge from which they can reasonably draw similar or contrasting opinions based on an experience. To simulate a dyadic WOM conversation, we randomly assign participants to share a negative or positive on-campus service experience with a confederate posing as another student collecting data for a class project. In addition, we look at whether participants update judgments only of a service experience that is shared (vs. not shared) during the conversation to rule out the possibility that encountering a conversation partner’s positive experience leads to broader positive emotions that influence judgment beyond the specific service discussed. Method Participants and Design We randomly assigned 219 undergraduates from a large southwestern university to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed design experiment. Seven participants who could not recall experiences with a service but completed the survey anyway were removed from the analysis, leaving 212 participants. Procedure Upon entering the lab, participants were randomly assigned to receive one of two versions of an “On-Campus Services Study.” Students were given a list of four on-campus services (dining, health, career, and academic advising) and asked to circle the one they felt most negatively (positively) about, based on an unsatisfactory (satisfactory) experience. Pretesting indicated that participants had positive expectations of all four services (Madvising = 5.58; Mhealth = 5.40; Mdining = 4.94; Mcareer = 5.59; see appendix D for full details). They then wrote two or three sentences describing the experience. Next, participants completed a survey in which they gave an initial evaluation of all four campus services (including the one they selected). We assessed initial evaluations of each service using the same three-item index from study 1 (αdining = .87, αhealth = .92, αcareer = .88, αadvising = .83). Next, we led each participant to a separate room where he or she met the confederate partner. As in study 1, the conversation partner posed as another student and employed a likeability manipulation (Sinclair et al. 2005). As before, this manipulation was done to foster familiarity and make participants feel comfortable. Specifically, when participants entered the room, the partner stood up, shook their hand, and gave them a warm greeting. He then invited them to sit down and introduced himself as a fellow student working on a class project and looking for student feedback regarding on-campus services. Next, he offered participants a piece of candy as a token of appreciation for taking the time to talk with him. He then asked participants if they could briefly share the positive or negative service experience they had chosen to write about. After participants shared their experience, the conversation partner indicated having had either a contrasting or similar experience with the service (e.g., “I actually really like (dislike) [service name]”) and stated that his experience had been either similar or the opposite. He then thanked participants and let them leave. Importantly, to simulate a closer relationship than the last study and thereby increase the generalizability of our findings, before they left, the confederate told participants they would be meeting him later in the session for a different project. This was done to account for the possible limitation that participants may have expected to speak with the confederate only once. Because participants most likely run into their student peers repeatedly and thus expect future interactions, this subsequent conversation expectation was another way to increase the external validity of our lab conversations. Upon returning to the lab, to capture time 2 evaluations, participants completed an additional survey further exploring on-campus services in which we asked them to reconsider their evaluations and respond to the same three-item evaluation index measure for each service (αdining = .89, αhealth = .92, αcareer = .90, αadvising = .87). Lastly, we asked participants to complete final debriefing questions to assess whether they guessed the study hypotheses. As mentioned above, we additionally explored whether participants updated evaluations for the services they did not share with the confederate to see whether other changes that could occur between time 1 and time 2 evaluations, such as more positive emotions, could explain shifts in evaluations that had nothing to do with the conversation. Results and Discussion Main Results We conducted a 2 (own experience) × 2 (conversation partner’s experience) × 2 (time, within) mixed ANOVA on the focal service evaluations. There was a main effect of own experience valence (MPositive = 6.12 vs. MNegative = 3.41; F(1, 208) = 281.43, p < .001) and a main effect of time, whereby evaluations improved after the interaction (MBefore = 4.79 vs. MAfter = 4.87; F(1, 208) = 4.19, p = .04). There was no effect of the conversation partner’s experience (F(1, 208) = 1.31, p = .25). We also found a significant own experience × time interaction (F(1, 208) = 13.46, p < .001). Supporting hypothesis 1, these interactions were qualified by a three-way interaction (F(1, 208) = 5.17, p = .02) (see figure 2). FIGURE 2 View largeDownload slide STUDY 2: SHIFT IN EVALUATIONS (T2 – T1) FIGURE 2 View largeDownload slide STUDY 2: SHIFT IN EVALUATIONS (T2 – T1) Next, our within-subjects contrasts revealed that when negative sharers were exposed to the conversation partner’s contrasting positive experience, their evaluations became more favorable after the interaction (t1 = 3.53 vs. t2 = 3.93, Δt2–t1 = .40; t(208) = 4.75, p < .001). On the contrary, when the partner shared a similar negative experience, evaluations did not shift (t1 = 3.29 vs. t2 = 3.37, Δt2–t1 = .08; t(208) = .94, NS). Next, among positive sharers, there was no change when the partner shared a contrasting negative experience (t1 = 6.15 vs. t2 = 6.05, Δt2–t1 = –.10; t(208) = 1.21, p = .23) or when the partner shared a similar positive experience (t1 = 6.16 vs. t2 = 6.13, Δt2–t1 = –.04; t(208) = .44, NS). Supporting hypothesis 1a, our focal complex contrast on the t2 – t1 difference scores of negative sharers revealed that the shift observed when the partner shared a contrasting positive experience was greater than that observed when the partner shared a similar negative experience (ΔContrasting = .40 vs. ΔSimilar = .08; ηG2= ⁠.032; F(1, 208) = 6.82, p = .01). Supporting hypothesis 1b, the corresponding contrast of the difference scores of positive sharers was not significant (ΔContrasting = –.10 vs. ΔSimilar = –.04; ηG2= ⁠.002; F(1, 208) = .32, NS). Evaluations of Nonselected Services We also tested whether participants’ evaluations of the services they did not share with the conversation partner changed following the conversation to ensure our effects did not emerge as an artifact of the two-time-period study design on a single product or service. It is possible participants felt more positive emotions after the conversation and thus perceived all services more favorably. To investigate this question, we computed participants’ mean evaluations of the three services they did not share with the partner before and after the interaction. For instance, if participants discussed dining services with the partner, we calculated the mean of their evaluations of career, health, and academic advising services. Analysis revealed no significant main effect of the conversation partner’s experience (F(1, 208) = .88, NS) on these nonselected evaluations. There was a significant main effect of own experience valence (MPositive = 4.51 vs. MNegative = 5.20; F(1, 208) = 52.96, p < .001) and timing (MBefore = 4.77 vs. MAfter = 4.87; F(1, 208) = 25.69, p < .001). However, there was no significant three-way interaction (F(1, 208) = .00, NS). Moreover, when breaking out each nonselected service, we found a similarly nonsignificant three-way interaction for health (F(1, 208) = .10, NS), dining (F(1, 208) = .40, NS), advising (F(1, 208) = .10, NS), and career (F(1, 208) = .95, NS) services. Debriefing As in study 1, our results remained robust to suspicion of the confederate.7 Discussion Study 2 replicates our support for hypothesis 1 in a new dyadic conversation setting with a familiar, frequently used service. We also show that the focal effect of consumers shifting their own negative experience upward in the face of another’s contrasting positive experience holds only for shared (vs. unshared) experiences within the dyad, revealing the instrumental role of encountering another’s contrasting opinion in changing opinions, and thereby serving as a type of within-subjects control. We should note that in web appendix B, we also include a between-subjects control condition in a vignette-based study where participants do not imagine sharing their experience with another individual, and show there are no differences between sharing a similar experience and lack of sharing altogether. In study 3, we test our proposed process. STUDY 3 Study 3 shifts to a conversation vignette and seeks to achieve three objectives. First, given that WOM most frequently occurs among close friends and family (Enda and Mitchell 2013), we seek to replicate our focal effect in a context where participants imagined conversing with an actual close friend, with products or services they purchased in the past. Second, using a vignette allows us to better control both transmitter and recipient expression, and is consistent with prior WOM research (Berger and Iyengar 2013; Moore 2012; Schellekens, Verlegh, and Smidts 2010; Zhang, Feick, and Mittal 2014). Third, we look for evidence that more favorable evaluations among negative sharers in the contrasting (vs. similar) partner condition are driven by a process in which participants are more likely to dismiss their negative experience as a one-time event because it is less consistent with positive performance expectations than the conversation partner’s positive experience (hypothesis 2). Method Participants and Design We randomly assigned 238 MTurk participants to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed-design experiment. One participant who could not recall a product or service but completed the survey was removed from the analyses, leaving 237 participants. Procedure We asked participants to think about a product or service costing at least $50 that they had purchased in the previous six months and that they felt performed unsatisfactorily (satisfactorily) (Zhang et al. 2014). Next, we asked them to enter the name of the product or service and write one or two sentences describing their experience. Participants tended to recall commonly used consumer items such as laptops, cell phones, and video games.8 Afterward, we asked participants to evaluate the product or service using the three-item evaluation index from the prior two studies (α = .97). However, because we asked participants to name a product they had previously purchased (and not just used, as with the lab pens and on-campus services), we modified the question, “How likely would you be to use this product or service?” to “How likely would you be to purchase this product or service again?” (1 = extremely unlikely; 7 = extremely likely). Next, we asked participants to “enter the name of a close friend that is important to you and has been for several years” (Kraus and Chen 2009). We asked these participants to imagine they were having a casual conversation with their friend and told him/her about the product or service experience they had written about. Participants in the similar partner condition then read: “To your surprise, [close friend’s name] responds that he/she also really likes (dislikes) that product or service and appears satisfied that the two of you share the same opinion. However, neither of you discuss the matter any further.” In the contrasting partner condition, they read: “To your surprise, [close friend’s name] responds that he/she actually really likes (dislikes) that product or service and appears perplexed that you both do not share the same opinion. However, neither of you discuss the matter any further.” We next captured our mediator, asking participants to complete a four-item measure of the extent to which they attributed their experience to temporary performance heterogeneity—that is, dismissed it as a one-time event following the interaction (α = .88). We based these measures on Folkes and Patrick’s (2003) characterization of performance heterogeneity attribution as the recategorization of an individual experience as an outlier rather than the norm. The measures were: 1) “To what extent do you feel that your experience with this product or service was an isolated event and not representative of the product in general?” 2) “To what extent do you think your experience with this product or service was an accurate indicator of its overall performance?” 3) “To what extent do you think you would have a similar experience upon using this product or service again?” and 4) “To what extent should your experience shape your overall opinion of the product or service?” (all anchored at 1 = not at all, 7 = very much so). Items 2–4 were reverse-coded and all items were combined to form our dismissal index (α = .88). Finally, participants completed the same three-item, time 2 evaluation index (α = .96). Results and Discussion Main Results We performed a 2 (own experience) × 2 (conversation partner’s experience) × 2 (time) mixed ANOVA on the evaluation index. We found main effects of own experience valence (MPositive = 6.11 vs. MNegative = 2.51; F(1, 233) = 723.11, p < .001) and time (MBefore = 4.22 vs. MAfter = 4.40; F(1, 233) = 19.53, p < .001). There was no effect of partner experience (F(1, 233) = 1.29, p = .26). We also found a significant own experience × time (F(1, 233) = 25.92, p < .001) and a marginal partner experience × time interaction (F(1, 233) = 3.68, p = .06). Importantly, these interactions were qualified by a significant three-way interaction (F(1, 233) = 7.39, p < .01) (see figure 3). FIGURE 3 View largeDownload slide STUDY 3: SHIFT IN EVALUATIONS (T2 – T1) FIGURE 3 View largeDownload slide STUDY 3: SHIFT IN EVALUATIONS (T2 – T1) Turning to our within-subjects contrasts, among negative sharers, when the conversation partner had a contrasting positive experience, evaluations shifted upward following the imagined interaction (t1 = 2.24 vs. t2 = 2.83, Δt2–t1 = .59; t(233) = 7.06, p < .001). Interestingly, when the partner had a similar negative experience, evaluations also became more positive (t1 = 2.39 vs. t2 = 2.59, Δt2–t1 = .20; t(233) = 2.37, p = .02). We elaborate on this finding further in the discussion. However, among positive sharers, when the conversation partner had a contrasting negative experience, evaluations did not change (t1 = 6.27 vs. t2 = 6.21, Δt2–t1 = –.06; t(233) = -.74, NS), nor did they shift when the partner had a similar positive experience (t1 = 5.98 vs. t2 = 5.98, Δt2–t1 = .01; t(233) = .07, NS). Supporting hypothesis 1a, a complex contrast on the t2 – t1 difference scores of negative sharers revealed that the shift observed when the partner shared a contrasting positive (vs. similar negative) experience was greater (ΔContrasting = .59 vs. ΔSimilar = .20; ηG2= ⁠.043; F(1, 233) = 10.44, p = .001). Supporting hypothesis 1b, the corresponding complex contrast among positive sharers was nonsignificant (ΔContrasting = –.06 vs. ΔSimilar = .01; ηG2= ⁠.001; F(1, 233) = .033, NS). Mediator (Dismissal) We ran a 2 (own experience: positive vs. negative, between) × 2 (partner’s experience: contrasting vs. similar, between) ANOVA on our dismissal measure. We found a main effect of own experience valence, such that negative (vs. positive) experiences were more likely to be dismissed as temporary in nature (MNegative = 3.22 vs. MPositive = 2.16; F(1, 233) = 47.39, p < .001), indicating that negative (vs. positive) experiences that defy high expectations are more likely to lead consumers to attribute their experience to heterogeneity (Folkes 1984). There was also a main effect whereby dismissal was higher when the partner’s experience was contrasting (vs. similar) (MContrasting = 2.88 vs. MSimilar = 2.49; F(1, 233) = 6.47, p = .01). As expected, there was a significant own experience × partner experience interaction (F(1, 233) = 5.53, p = .02) (see figure 4). FIGURE 4 View largeDownload slide STUDY 3: DISMISSAL OF OWN EXPERIENCE FIGURE 4 View largeDownload slide STUDY 3: DISMISSAL OF OWN EXPERIENCE Supporting hypothesis 2, negative sharers who learned of a partner’s contrasting positive (vs. similar negative) experience were more likely to dismiss their negative experience following the conversation (MContrasting = 3.59 vs. MSimilar = 2.84; ηG2= ⁠.048; F(1, 233) = 11.63, p = .001). Among positive sharers, there was no such difference (MContrasting = 2.17 vs. MSimilar = 2.14; ηG2= ⁠.000; F(1, 233) = .02, NS). Moderated Mediation Analysis We tested moderated mediation using PROCESS model 59 of the bootstrapping process described by Hayes (2013) on the t2 – t1 evaluation index difference scores. This model has been used in other studies to allow the moderator (valence of own experience) to interact with both the independent variable (partner’s experience) and the mediator (dismissal) (Birnbaum et al. 2014; Clark et al. 2017; Goncalo, Vincent, and Krause 2015). This is the most robust model to analyze our effect, given that the valence of the consumer’s own experience also affects the direction of evaluation change following dismissal (see web appendix C for a full explanation of the model, its graphical depiction, and detailed analysis). As predicted, for negative sharers, dismissal mediated the effect of the partner’s contrasting positive (vs. similar negative) experience on upward shifts in evaluations (BIndirect = .24, CI95% [.07, .57]). Specifically, negative sharers dismissed their own experience as a one-time event when they learned of the conversation partner’s contrasting (vs. similar) experience (B = .75, CI95% [.32, 1.19]). As dismissal increased, participants evaluated the product or service more favorably (B = .32, CI95% [.24, .41]). On the contrary, among positive sharers, dismissal did not mediate (BIndirect = –.002, CI95% [–.04, .03]). The index of moderated mediation further confirmed that our moderated mediation analysis was successful (B = –.24, CI95% [–.56, –.07]). Discussion In study 3, we again find support for hypothesis 1 by showing that, when faced with a conversation partner’s contrasting experience, consumers adjust negative judgments upward. Critically, supporting hypothesis 2, when negative (vs. positive) sharers became aware of a partner’s contrasting experience, they were more likely to dismiss their own experience as a one-time event, which mediated the upward shift in evaluations of the product or service. Surprisingly, in this study, participants who shared a negative experience evaluated products and services more positively after learning of a close partner’s similar negative experience. While we did not observe this shift in any other studies, we note that the magnitude of the shift (Δ = .20) is still substantially smaller than the focal contrasting experience (Δ = .59), and, as will be evident next, similar to the time 1 to time 2 shifts in study 4 (Δ = .14, also a vignette). STUDY 4 Thus far, we have argued that consumers dismiss their own negative experiences as one-time events when faced with another’s contrasting positive experience because of their generally high expectations of product and service performance. However, it is theoretically important to test our process in the rarer case where consumers have negative expectations, despite the increased uncertainty and variance that may occur when testing these products and services. Thus, in study 4, we directly manipulate performance expectations. Specifically, in hypothesis 3, we argued that when consumers have negative expectations, those who had a negative experience with a product or service should no longer dismiss their own experience when faced with another’s contrasting positive experience. Analogously, when expectations are negative, one should be more likely to dismiss one’s own positive experience after learning of another’s contrasting negative experience, reversing the positivity effect of prior studies. To test this in study 4, we manipulate expectations of a new restaurant. This study also seeks to further address the alternative explanation of impression management. We found preliminary evidence that impression management motives did not differ among negative (vs. positive) sharers in study 1, but we take two further steps to rule out impression management as a mediator. First, we ask participants to imagine their conversation partner is an acquaintance (vs. a close friend), as increasing social distance heightens impression management concerns and thus provides a more rigorous test for this alternative (Kraus and Chen 2009; Kwang and Swann 2010). Second, we use a four-item measure that explicitly captures whether participants changed their evaluations to satisfy impression management/positivity norms. We acknowledge the possibility that negative (vs. positive) sharers may respond to general impression management concerns by wanting to outwardly appear more agreeable. However, we believe that such concerns are unlikely to mediate our positivity effect since responses for time 2 evaluations are taken privately after the interaction, and research shows that impression management motives influence public, but not private, behavior and opinions (Dubois, Rucker, and Galinsky 2012; Kristofferson, White, and Peloza 2014; Ratner and Kahn 2002). We reasoned that if we indeed did not find mediation by impression management concerns with a more nuanced four-item measure and a more socially distant partner, we could more confidently support our claim that impression management is not the primary driver of the focal effects. Method Participants and Design We randomly assigned 400 MTurk participants to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (restaurant expectations: positive vs. negative, between) × 2 (time: before vs. after conversation, within) mixed-design experiment. We removed eight duplicate participants included due to an error in the filtering of individuals who had previously taken the survey, leaving 392 participants. Of note, our predicted interaction remained significant when these participants were retained (F(1, 384) = 19.54, p < .001), as did our hypothesized contrasts. Procedure We told participants we were conducting a study on how news stories affect product and service evaluations. We first asked participants to imagine they were looking for information on a fictional chain restaurant, The Copper Yard, and to read a top Google search article for this restaurant. In the positive expectations condition, we showed participants a fictional Consumer Reports article rating The Copper Yard as excellent on its food and customer service (see appendix E). In the negative expectations condition, the article rated the restaurant as poor on both these dimensions. Next, we asked participants to imagine they looked up additional information for the Copper Yard franchise in their town. In the positive (negative) expectations condition, we presented them with a fictional local newspaper article describing the restaurant’s high (low) sales and expert opinions of its high- (low-) quality food and service. Both articles highlighted that the restaurant was currently revamping its menu to provide a plausible reason for visiting the restaurant. After reading the articles, participants rated their expectations of The Copper Yard with the same two-item measure used in the product expectations pretest from the literature review (α = .96; see footnote 4 for items). We asked participants to read two articles to add realism to a situation in which consumers might form expectations by learning general information about a national chain, and then looking up specific information about the franchise in their area. We also wanted to ensure that our negative expectations manipulation was sufficiently strong and enduring to set below-neutral expectations. Next, we asked participants to imagine trying The Copper Yard because it was advertising a new happy hour special and was close to their home (see appendix F). Participants in the positive own experience condition imagined having a prompt and pleasant server and eating a tasty dish, while participants in the negative own experience condition imagined having a slow and unpleasant server and eating a mediocre dish. Participants then evaluated The Copper Yard on the same three questions from study 3 (attitude, satisfaction, and purchase likelihood), as well as a fourth item asking “How likely would you be to go to The Copper Yard again in the future?” (α = .96). We added this item to capture intentions to return to the restaurant, as prior work has found repeat loyalty to be an important indicator of consumer restaurant evaluations (Mattila 2001). Next, we asked participants to imagine that, a week after their experience at The Copper Yard, they ran into an acquaintance and struck up a short conversation that included a discussion about local restaurants. As described, we purposefully chose an acquaintance as the conversation partner to support the notion that our effect is robust to varying levels of social distance. Participants then imagined sharing their experience at The Copper Yard with the acquaintance and that the acquaintance then shared a contrasting or similar restaurant experience. To make the conversation seem natural, we asked participants to imagine that they casually shared a few more details about their experiences before exhausting the subject and turning to other topics. Participants then rated the extent to which they would dismiss their own experience as a one-time event following the conversation using the four-item measure from study 3 (α = .88). They also responded to a time 2, four-item evaluation index for The Copper Yard (α = .99). Finally, we asked participants a four-item measure of how much they were influenced to change their opinion of The Copper Yard9 based on impression management concerns. Importantly, while our study 1 impression management measure captured general impression management motives in each condition, our new measure directly asked participants whether they changed their evaluation to satisfy an impression management motive, thereby examining whether impression management motives, as opposed to dismissal, mediated our results. Results and Discussion We predicted that when participants had positive expectations, we would replicate our prior positivity effect: negative sharers would form more positive evaluations upon learning of the conversation partner’s contrasting positive (vs. similar negative) experience, whereas positive sharers would not change their evaluations. On the contrary, when participants had negative expectations, we expected a negativity effect such that negative sharers would no longer form more favorable evaluations following exposure to the partner’s contrasting positive (vs. similar negative) experience. Instead, we predicted that positive sharers would now form more negative evaluations after exposure to the partner’s contrasting negative (vs. similar positive) experience. Of note, because intentionally using products for which consumers have negative expectations is less common, we are less certain of how consumers will behave outside of the explicit contrasts we predict in hypothesis 3b. Manipulation Check A one-way ANOVA confirmed that our expectations manipulation was successful (MPositive = 6.47 vs. MNegative = 2.12; F(1, 384) = 2124.06, p < .001). Main Results We conducted a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (restaurant expectations: positive vs. negative, between) × 2 (time: before vs. after conversation, within) mixed ANOVA on the evaluation index. We found main effects of own experience valence (MPositive = 6.19 vs. MNegative = 2.06; F(1, 384) = 2123.80, p < .001), restaurant expectations (MPositive = 4.57 vs. MNegative = 3.67; F(1, 384) = 101.37, p < .001), and time (MBefore = 4.06 vs. MAfter = 4.19; F(1, 384) = 20.78, p < .001); significant own experience × time (F(1, 384) = 137.66, p < .001), partner experience × time (F(1, 384) = 15.36, p < .001), and expectations × time (F(1, 384) = 39.96, p < .001) interactions; and a marginally significant partner experience × expectations interaction (F(1, 384) = 3.08, p = .08). We also found significant own experience × partner experience × time (F(1, 384) = 44.57, p < .001), own experience × expectations × time (F(1, 384) = 5.06, p < .03), and partner experience × expectations × time (F(1, 384) = 27.26, p < .001) interactions. Most importantly, supporting hypothesis 3, these were qualified by a four-way interaction (F(1, 384) = 19.54, p < .001) (see figure 5). FIGURE 5 View largeDownload slide STUDY 4: SHIFT IN EVALUATIONS (T2 – T1) A. POSITIVE EXPECTATIONS B. NEGATIVE EXPECTATIONS FIGURE 5 View largeDownload slide STUDY 4: SHIFT IN EVALUATIONS (T2 – T1) A. POSITIVE EXPECTATIONS B. NEGATIVE EXPECTATIONS We first examined the positive expectations conditions. Among individuals sharing a negative experience, replicating prior studies, within-subjects evaluations shifted significantly upward in the contrasting partner condition (t1 = 2.11 vs. t2 = 3.43, Δt2–t1 = 1.32; t(384) = 17.44, p < .001). Conversely, these evaluations did not significantly change in the similar partner condition (t1 = 2.31 vs. t2 = 2.45, Δt2–t1 = .14; t(384) = 1.61, p = .11). Next, among individuals sharing a positive experience, evaluations did not change in the contrasting partner condition (t1 = 6.57 vs. t2 = 6.43, Δt2–t1 = –.15; t(384) = –1.58, p = .12), nor did they shift in the similar partner condition (t1 = 6.66 vs. t2 = 6.63, Δt2–t1 = –.03; t(384) = .42, NS). Supporting hypothesis 1 and replicating studies 1–3, a complex contrast on the t2 – t1 difference scores among those sharing a negative experience showed that the shift in the contrasting (vs. similar) partner condition was greater (ΔContrasting = 1.32 vs. ΔSimilar = .14; ηG2= ⁠.222; F(1, 384) = 109.41, p < .001). The corresponding contrast among positive sharers was not significant (ΔContrasting = –.15 vs. ΔSimilar = –.03; ηG2= ⁠.002; F(1, 384) = .87, NS). We next examined the negative expectations conditions. Among individuals who shared a negative experience, within-subjects evaluations shifted significantly upward to relatively the same degree in the contrasting partner condition (t1 = 1.32 vs. t2 = 1.57, Δt2–t1 = .25; t(384) = 3.23, p = .001) and in the similar partner condition (t1 = 1.53 vs. t2 = 1.72, Δt2–t1 = .20; t(384) = 2.16, p = .03). Next, among individuals with negative expectations who shared a positive experience, evaluations shifted significantly downward in the contrasting partner condition (t1 = 5.93 vs. t2 = 5.49, Δt2–t1 = –.43; t(384) = 5.14, p < .001). Interestingly, evaluations also shifted significantly downward in the similar partner condition (t1 = 6.02 vs. t2 = 5.79, Δt2–t1 = –.23; t(384) = 3.10, p = .002), but to a lesser extent. Supporting hypothesis 3, a follow-up contrast on the difference scores of those sharing a negative experience revealed no differences between the two shifts (ΔContrasting = .25 vs. ΔSimilar = .20; ηG2= ⁠.001; F(1, 384) = .21, NS). As we will discuss later, this contrast indicates that the upward shift observed among negative sharers in the contrasting partner condition is unlikely to be related to the partner’s shared experience, which is the focus of our predictions, but rather some other factor. An additional contrast on the difference scores of positive sharers showed that the negative shift in the contrasting negative (vs. similar positive) partner condition was marginally greater (ΔContrasting = –.43 vs. ΔSimilar = –.23; ηG2= ⁠.009; F(1, 384) = 3.51, p = .06). This suggests that although positive sharers became more negative in both conditions, those in the contrasting partner condition were most negative, presumably due to their dismissal of the negative experience as a one-time event, consistent with our theory. Finally, we ran two additional difference score contrasts to test how another’s distinct experience can shape dismissal differently when people have positive versus negative expectations. Consistent with our theory, the upward shift in evaluations among negative sharers in the contrasting partner condition was significantly attenuated when expectations were negative (vs. positive) (ΔPositive exp = 1.32 vs. ΔNegative exp = .25; ηG2= ⁠.199; F(1, 384) = 95.61, p < .001). Next, the downward shift in evaluations among positive sharers in the contrasting partner condition was significantly enhanced when expectations were negative (vs. positive) (ΔPositive exp = –.15 vs. ΔNegative exp = –.43; ηG2= ⁠.013; F(1, 384) = 5.11, p = .02). Mediator (Dismissal) We conducted a 2 (own experience) × 2 (conversation partner’s experience) × 2 (restaurant expectations) between-subjects ANOVA on our dismissal mediator. We found main effects of own experience (F(1, 384) = 10.03, p = .002) and partner experience (F(1, 384) = 76.02, p < .001) and significant own experience × partner experience (F(1, 384) = 5.92, p = .02), own experience × expectations (F(1, 384) = 148.12, p < .001), and partner experience × expectations (F(1, 384) = 7.51, p = .01) two-way interactions. Most importantly, we found a significant own experience × partner experience × expectations interaction (F(1, 384) = 6.59, p = .01). Replicating study 3, planned contrasts revealed that in the positive expectations condition, negative sharers were more likely to dismiss their own experience when the conversation partner shared a contrasting positive (vs. similar negative) experience (MContrasting = 4.30 vs. MSimilar = 2.62; ηG2= ⁠.159; F(1, 384) = 72.38, p < .001). Further, in the negative expectations condition, positive sharers were more likely to dismiss their own experience when the partner shared a contrasting negative (vs. similar positive) experience (MContrasting = 3.36 vs. MSimilar = 2.73; ηG2= ⁠.026; F(1, 384) = 10.35, p < .001). Third, among negative sharers who learned of a partner’s contrasting positive experience, dismissal was significantly higher when general expectations were positive (vs. negative) (MPositive exp = 4.30 vs. MNegative exp = 2.42; ηG2= ⁠.202; F(1, 384) = 97.24, p < .001). Moderated Mediation Analysis We tested moderated mediation using PROCESS model 73 (Hayes 2013) on the t2 – t1 evaluation index difference scores. This model is the three-way analog to PROCESS model 59, used in study 3, and has been used in prior research to allow both moderators to interact with both the independent variable and the mediator10 (Clark et al. 2017) (see web appendix D.1). We first examined participants with positive expectations. For those who had positive expectations and shared a negative experience, dismissal mediated the upward shift in evaluations in the contrasting (vs. similar) partner experience condition (BIndirect = .57, CI95% [.35, .89]). On the contrary, for those who had positive expectations and shared a positive experience, dismissal did not mediate shifts in evaluations (BIndirect = –.02, CI95% [–.12, .05]). Also, as predicted, for those who had negative expectations but shared a positive experience, dismissal mediated the downward shift observed when the partner had a contrasting negative (vs. similar positive) experience (BIndirect = –.14, CI95% [–.25, –.05]). Finally, for participants who shared a negative experience, dismissal mediated the changes in evaluations (BIndirect = .12, CI95% [.02, .30]), though these changes were not different in the contrasting (vs. similar) partner conditions and so should be interpreted in such a light. Alternative Mediator We also tested whether positivity norms/impression management mediated our results. There was a significant three-way interaction on our index (F(1, 382) = 6.53, p = .01). Driving this interaction, under negative expectations, positive sharers had the greatest impression management concerns upon learning of the partner’s contrasting negative experience. Notably, this pattern is inconsistent with our focal evaluation results, where we found the largest shift under positive expectations, when negative sharers learned of the partner’s contrasting positive experience. More importantly, impression management concerns did not mediate in any of the conditions (all confidence intervals included 0). Further, following Zhao, Lynch, and Chen (2010), we tested this variable in parallel with our dismissal mediator (see web appendix D.2 for full details) and again found no evidence that these motives mediated our effect in any conditions. Discussion In study 4, we show that expectations play an important role in how dyadic conversations shape consumer evaluations. When participants had positive expectations, we replicated prior studies and found the same positivity effect and mediation. On the other hand, when participants had negative expectations, we found a different pattern of results. In particular, the upward shift among negative sharers in the contrasting partner condition was dampened, and was no different from the shift in the similar partner condition. Additionally, there was a significant downward shift among positive sharers in the contrasting partner condition, and this shift was marginally greater than the shift in the similar partner condition. Given people are less likely to use products and services for which they have negative expectations, it is less surprising that we found some unpredicted effects in these conditions—most notably increases and drops of ∼.20 from time 1 to time 2 evaluations. We note that though significant, a change of ∼.20 is much smaller than the focal changes documented in prior studies or in the positive expectations of study 4, which are all substantially larger (.42 in study 1, .40 in study 2, .59 in study 3, and 1.32 in the positive expectations condition of study 4). Thus, we expect the significance of these smaller changes is more dependent on the increased sample size of study 4, as opposed to substantial changes in beliefs themselves. Importantly, our focal drop, whereby people holding negative expectations were more likely to lower their own positive evaluations in light of another’s negative evaluations, is more in line with the changes (–.43) across prior studies. GENERAL DISCUSSION Across four studies, we test a framework to illustrate the influence of one-on-one conversations where consumers share and compare experiences with a product or service. We demonstrate a positivity effect in these conversations, whereby consumers who share a negative experience evaluate products and services more favorably after learning about someone else’s contrasting positive experience, but not vice versa. Furthermore, we identify the underlying process behind this positivity effect: positive expectations of product and service performance lead consumers to attribute their own negative experience to temporary performance heterogeneity (i.e., dismiss it as a one-time event). We also delineate a theoretical boundary condition whereby when consumers have negative product and service expectations, they no longer dismiss their own negative experience upon learning of a conversation partner’s contrasting positive experience, and the positivity effect is attenuated. Conversely, negative expectations cause consumers to dismiss their own positive experience after they learn of a conversation partner’s negative experience, leading to a negativity effect. Importantly, our focal pattern of effects remains consistent across a range of real and simulated dyadic interactions with differing levels of closeness (e.g., likable other, close friend, casual acquaintance), a diverse set of products and services (ballpoint pens, campus services, costly [>$50] products bought by participants, a restaurant experience), and oppositely valenced performance expectations. These varying experimental paradigms increase generalizability and help to rule out specific alternative accounts. First, the similar pattern of effects with close others and acquaintances suggests that the malleability of evaluations we observe is unlikely to be driven by spillover from impression management concerns. The range of products used also makes a self-repair account (whereby consumers increase evaluations to make themselves feel better about a bad experience) less plausible. The fundamental attribution error suggests that participants most likely blame the product or service, not themselves, for their negative experience (Bitner, Booms, and Mohr 1994; Folkes and Kotsos 1986), a bias that should be particularly strong with the more mundane pen and campus services used in studies 1 and 2, which were provided to (and not chosen by) participants. Moreover, our study 4 finding—that when expectations are negative, participants who had a positive experience shift their evaluations downward—is inconsistent with a self-repair account, as it is unlikely consumers would want to feel worse about having had a positive experience. In sum, the replication of our focal positivity effect across dyad types and products, its mediation by dismissal, and the subsequent reversal of our positivity effects when expectations are negative provides convergent evidence of expectations, experience, and WOM shaping consumer evaluations. We make several contributions to the literature. First, in recent years, consumer researchers have largely focused on the effect of online WOM, such as consumer reviews. As a result, there is little knowledge regarding the persuasive impact of interpersonal communication that often occurs offline, such as casual conversations where consumers share and compare past product or service experiences. These conversations appear to be surprisingly common and powerful, as consumers in an increasingly homogenized retail landscape often use the same products and services, and share these experiences with one another (Bughin et al. 2010; Enda and Mitchell 2013; Keller Fay Group 2014). Past work has looked at some aspects of these conversations in settings where participants communicate within groups after reading about a product (Herr et al. 1991), or with multiple anonymous online others (Schlosser 2005). Other work has looked at how “storytelling” impacts the sharer (Cowley 2014; Moore 2012), but not how the sharer responds to another person’s experience with the same product or service. Thus, our work provides new insights into the persuasive influence of these more complex interactions. Next, although past WOM research has largely found that negative WOM is more powerful than positive WOM (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005), more recent work has found that this negativity bias is attenuated under certain circumstances (Chen and Lurie 2013). Leveraging this latter finding, we demonstrate a positivity effect in one-on-one conversations, as consumers who have generally positive expectations dismiss their own negative experience as a one-time event after learning of another’s positive experience, but not vice versa. Thus, to our knowledge, ours is the first WOM research to indicate that positive information regarding a product or service experience can be weighed more heavily than negative information. Finally, in identifying the underlying mechanism of our framework, we also integrate work in attribution theory (Chen and Lurie 2013; Folkes 1984, 1988), disconfirmation bias (Ditto and Lopez 1992; Ditto et al. 1998; Edwards and Smith 1996), and consumer expectations (Folkes and Patrick 2003; Fornell et al. 1996; Johnson et al. 1995) to show that dismissal underlies the greater malleability of evaluations formed by negative (vs. positive) product and service experiences, as consumers decrease their subsequent usage of these experiences when they perceive them to be the result of temporary performance heterogeneity. In sum, we believe we are the first to develop a framework to better understand how a common form of dyadic, post-consumption conversation shapes judgments of products and services. Future Research While we tried to increase the external validity of our research by using a variety of dyadic conversation contexts, one limitation of our work is the large breadth of variation in these types of conversations that we do not explore. For instance, we do not look at whether the consumer having one prior experience versus multiple prior experiences moderates how she responds to the conversation partner’s experience. However, our framework would predict that if the consumer has had multiple negative experiences prior to interacting with the partner, she would be less likely to perceive performance heterogeneity solely based on the partner’s contrasting positive experience and, therefore, be less likely to change her evaluation. We also did not explore how consumers might update their evaluations over multiple conversations with the same (or a different) partner. Whereas subsequent conversations with the same partner may not lead to new inferences regarding performance heterogeneity, conversations with a different partner might provide additional cues regarding heterogeneity, and be more persuasive as a result. By investigating these questions in more detail, future research could reveal the full complexity of these conversations. Another question for future research is whether a dyadic conversation context is always necessary for our theorized effect to occur. For instance, two (or more) individuals may also share contrasting and similar experiences within a larger group setting. Prior studies suggest that a negative sharer is more likely to attribute a positive sharer’s experience to personal motives (i.e., impression management) when both are “broadcasting” in front of a group (Barasch and Berger 2014; Chen and Lurie 2013). This should decrease the relative influence of the positive (vs. negative) sharer’s experience and dampen the positivity effect observed in dyadic conversations. This intuition is supported by past studies finding negative (vs. positive) WOM to be more powerful in group contexts (Herr et al. 1991; Schlosser 2005). However, we do acknowledge that certain types of group conversations may have a persuasive effect similar to that of dyadic conversations, such as a consumer sharing a negative experience with two very close friends. If one or both friends respond with a positive experience, the consumer may have little reason to believe that they are trying to show off, and would likely dismiss his own negative experience the same way he might if only talking with one friend. Indeed, he would probably be even more likely to dismiss his negative experience upon learning that both of his friends had a positive experience. Future research could explore under what conditions group conversations have a similar persuasive effect as dyadic conversations, or look at more complex scenarios in which multiple individuals share contrasting or similar experiences that the consumer must interpret and use when forming a new judgment. Future research could also move from face-to-face to other types of interactions. For instance, a person could see a friend’s comment about a product or service on her social media feed or read an online story about another person’s experience. While it is hard to say how people would respond in other contexts, we believe our results would hold in any setting where consumers predominantly attribute another person’s WOM to the product or service itself, as opposed to the reviewer’s personal motives (Chen and Lurie 2013). Thus, it is possible that positivity effects emerge not only in one-on-one contexts, but also in situations where consumers attribute the review directly to the product or service, as signaled by temporal contiguity (Chen and Lurie 2013), trustworthiness of the reviewer, or objectivity of the review source. Studies could also investigate when impression management motives play a larger role in shaping evaluations. In study 4, we did find that these motives were more salient in conversations where an acquaintance shared a contrasting (vs. similar) experience. However, changes in evaluations were not mediated by impression management motives, perhaps because participant responses were made in private. Thus, there are likely situations where impression management influences participant evaluations, perhaps when people must make public evaluations or consume a product or service with a partner they do not know very well. For example, if two people are trying to decide where to go out to eat and their contrasting opinions about a given restaurant emerge as they list possible options, a consumer who wants to be seen as agreeable would probably be more willing to dismiss a positive experience with the restaurant if his partner had a bad experience there and does not want to visit it again. Another interesting question raised by our research is whether WOM being associated with performance versus taste (Spiller and Belogolova 2016) moderates its persuasive impact. Specifically, while our research focused on conversations in which consumers share experiences based on product and service performance, they may also share WOM regarding their tastes, such as opinions about movies or music. One possibility is that expectations are more ambiguous in taste-related domains, as consumers may be uncertain how information about a product or service relates to their own subjective preferences. Additionally, they may not be influenced by another person’s taste-related opinion unless they perceive that person’s tastes to be similar to their own. Future research could look at how consumers construct expectations in taste-related domains, as well as when and why they are persuaded by another’s taste-related opinions. Future research could also further explore how negative expectations impact the way that consumers experience products and services, and how these experiences affect subsequent consumer perceptions. Though less common, consumers are often obligated to use products and services for which they expect performance to be poor (e.g., health insurers, cable companies, airlines). In these cases, products with negative expectations often have some redeeming aspect that leads consumers to try them (extremely low rates, new management taking over, etc.). It would be interesting to examine how these positive attributes are considered by consumers and contribute to evaluation changes. Notably, one limitation of our work is that we measure participant evaluations at only a single level of granularity. In particular, while we capture participants’ overall product and service evaluations following a conversation (including repurchase intentions), we do not measure participants’ evaluations of their own product/service or the conversation partner’s product/service. How might these lower-level evaluations compare to the general evaluations that we capture? We suggest that negative sharers who learn of a partner’s positive experience evaluate their own product or service more negatively following the conversation, as they come to see it as a negative outlier, or “lemon.” On the other hand, their view of the partner’s product or service likely reflects general evaluations, given that they perceive it to be an exemplar of how that product or service typically performs. Future studies could further explore these nuances. Managerial Implications Why is it important to better understand the influence of one-on-one conversations where consumers share past experiences with a product or service? First, in an increasingly franchised retail landscape, there should be greater convergence in the types of products and services consumers try and therefore discuss offline. As a result, it is important to understand how consumer perceptions are shaped when they compare experiences with the same product or service. Second, although there is a broad literature looking at how managers should respond to negative WOM (Blodgett, Hill, and Tax 1997; Schaefers and Schamari 2016), other studies suggest that a greater focus on positive WOM can be equally (or more) beneficial (Ahluwalia 2002; East, Hammond, and Lomax 2008). Our findings give managers a better idea of when and why they should focus on responding to positive versus negative WOM. In particular, in industries where consumers have high expectations, companies may want to prioritize the sharing of positive consumer experiences (via incentives, retweets, etc.), rather than spending all their efforts to mitigate the spread of negative consumer experiences. Conversely, in industries where consumers have low expectations (e.g., airlines, health insurance, specific government services), companies might do better to focus on addressing negative WOM rather than promoting positive WOM, as the latter is more likely to be dismissed. Thus, our findings suggest that WOM strategy can potentially be segmented, depending on consumer expectations of the industry. Further building upon the role of expectations, some managers might assume that they can overcome negative expectations by delivering a positive in-person experience to the customer. While this is somewhat true, our findings suggest that negative expectations continue to loom over positive experiences, as consumers can be persuaded by subsequent negative WOM to lower their evaluations of the product or service. This suggests that if managers are successful at both setting positive expectations and delivering a positive in-person experience, customers will be largely buffered from subsequent negative WOM that they encounter. In providing a framework to evaluate the persuasive impact of these conversations, we have only begun to understand some of the ways in which dyadic, post-consumption WOM may shape product and service evaluations. Interestingly, in a world where performance information regarding products and services is readily available, even when consumers have personal experience to draw from, one-on-one conversations still play a significant role in shaping consumer evaluations and repurchase likelihood. DATA COLLECTION INFORMATION All studies were designed and analyzed by the first author under the guidance of the second author. The data for study 1 was collected at the Arizona State University Behavioral Lab in spring 2016. The data for study 2 was collected at the Arizona State University Behavioral Lab in fall 2014. The data for study 3 was collected on Amazon’s Mechanical Turk in fall 2014. Data for study 4 was collected on MTurk in fall 2017. The data for the first pretest was collected at the Arizona State University Behavioral Lab in fall 2016. The data for the second pretest was collected on MTurk in fall 2017. The data for the third pretest was collected at the Arizona State University Behavioral Lab in fall 2016. The data for the study in web appendix B was collected on MTurk in fall 2014 (data for the added control condition was collected in fall 2016). Research assistants collected the data for studies 1 and 2, as well as all pretests run, at the Arizona State University Behavioral Lab. The first author collected the data for studies 3, 4, pretest 2, and the study in web appendix B. The authors would like to thank Jim Bettman and Kirk Kristofferson for their valuable feedback. The authors are also grateful to the editor, associate editor, and three anonymous reviewers for their insight and guidance. Supplementary materials are included in the web appendix accompanying the online version of this article. APPENDIX A—Pretest 1 We conducted a short pretest in which we asked 65 undergraduate participants to imagine the following scenario: “You are having a conversation with a peer from one of your classes. You run into each other while waiting in line at the grocery store. You don’t know him very well, but he is very friendly and nice. The topic comes around to cleaning and it turns out that you have both purchased the same brand of robot vacuum cleaner.” We next randomly assigned participants to imagine that the peer stated having had either a good or bad experience with the vacuum. We then asked participants to imagine that they had had the opposite experience with the vacuum, and asked, “Would you share your experience with him?” (Yes/No). As expected, 85% of participants indicated that they would share their contrasting opinion of the vacuum in this situation. Moreover, this willingness to share occurred regardless of whether the peer had had a positive or negative experience with the product (χ2 = .53, p = .40). APPENDIX B—Pretest 2 We asked 80 MTurk participants to read the following scenario: “Imagine that you are having a lunchroom conversation with a peer that you know from work. Somehow, the topic comes around to local restaurants, and whether it is worthwhile to go to a restaurant that just opened nearby.” We then randomly assigned participants to imagine that they had a positive (negative) experience at the restaurant, which they shared with the coworker, and that the coworker replied that they had a contrasting negative (positive) experience. We then assessed the extent to which participants attributed the coworker’s experience to a personal motive with the question: “How big of a role do you think personal factors (the other person’s personality, traits, character, personal style, and so on) played in their decision to share their experience?” (1 = minimal role; 9 = maximal role). Next, we assessed the extent to which they attributed the coworker’s experience to the restaurant with the question: “How big of a role do you think the actual restaurant experience (food quality, service, etc.) played in their decision to share their experience?” Both questions came from Chen and Lurie (2013). Further following the procedure from Chen and Lurie (2013), we then subtracted personal motive attribution from restaurant experience attribution, so that positive scores indicated greater restaurant experience (vs. personal motive) attribution. A one-sample t-test revealed the diff score (M = .86, SD = 2.13) to be significantly greater than 0 (t(79) = 3.63, p = .001), indicating that participants predominantly attributed the coworker’s WOM to the restaurant. Consistent with our argument, this attribution was equivalent for positive and negative experiences shared by the coworker (MPositive = .76, MNegative = .97; F(1, 78) = .20, NS). APPENDIX C—Pretest 3 We identified 10 popular consumer products and services (dinner at a nice restaurant, smartphone, headphones, internet service, a night at a hotel, a pair of athletic shoes, car repair service, laptop computer, vacuum cleaner, video game system) that emerged from an open-ended question about a product or service people had purchased in the past six months that was over $50. We then separately asked 66 undergraduate participants to indicate whether they had purchased each product or service in the past two years (1 = yes; 2 = no), as well as how they would expect each product or service to perform, using a two-item satisfaction measure: 1) I would expect to be satisfied with this product and 2) I would expect this product to function well (1 = strongly disagree; 7 = strongly agree) (α = .95). Averaging the satisfaction measure across all 10 products, we found that participants had very positive expectations of product and service performance (MSatisfaction = 6.08 out of 7, SD = .70) (see table 2 for full details). TABLE 2 SATISFACTION EXPECTATIONS FOR COMMONLY PURCHASED PRODUCTS AND SERVICES Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) TABLE 2 SATISFACTION EXPECTATIONS FOR COMMONLY PURCHASED PRODUCTS AND SERVICES Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) APPENDIX D—Expectations Pretest for Study 1 and Study 2 Stimuli In this study, we are interested in students’ perceptions of various products and services. On the following pages, we will list a product or service and ask you to indicate how you perceive that product or service along several dimensions. Study 1 Stimuli Product: A pen from the ASU Bookstore Imagine that you had to use this pen to fill out some information. Two hundred sixty-eight undergraduate participants responded to the following four items (α = .88): “If you were going to use this pen, would you expect it to perform positively?” “If you were going to use this pen, would you expect it to write well?” “If you were going to use this pen, would you expect it not to write well?” (reverse-scored) “Do you think that these pens typically perform well?” (1 = not at all; 7 = very much so). The result of a one sample t-test indicated that mean expectations were above the neutral midpoint (M = 4.92; t(267) = 11.13, p < .001). Study 2 Stimuli Services: Academic Advising, Health Services, Dining Services, and Career Services Two hundred sixty-eight undergraduate participants responded to the following four items for academic advising (α = .82), health services (α = .83), dining services (α = .85), and career services (α = .82): “If you were going to use [service name], would you expect this service to perform positively?” If you were going to use [service name], would you expect a good service experience?” “If you were going to use this service, would you expect a bad service experience?” (reverse-scored) “Do you think that [service name] typically provides a good service experience?” (1 = not at all; 7 = very much so). One sample t-tests indicated that mean expectations of all services were above the neutral midpoint (Madvising = 5.58; t(267) = 22.69, p < .001, Mhealth = 5.40; t(267) = 18.81, p < .001, Mdining = 4.94; t(267) = 12.49, p < .001, Mcareer = 5.59; t(267) = 24.19, p < .001). APPENDIX E—Article Stimuli for Study 4 APPENDIX F— Study 4 Vignettes All participants first read: [Negative expectations condition] The Copper Yard has been advertising a new happy hour special for the past week. Even though you have low expectations of the restaurant, it is close to your house, so you decide to give it a try. [Positive expectations condition] The Copper Yard has been advertising a new happy hour special for the past week. You have high expectations of the restaurant and it is close to your house, so you decide to give it a try. In the negative own experience condition, participants then read: When you enter the restaurant, no one is waiting up front to show you to a table, so you decide to sit at the nearest open table. After what seems like forever waiting, your server, whose name is Mike, finally comes up looking somewhat hurried and bothered. “Can I start you off with something to drink?” he asks briskly. After Mike takes your order, he leaves and doesn’t return with your drink for another 20 minutes. With Mike still there, you take the opportunity to order a happy hour menu item: the Fiesta Nachos. These also take a long time to come out and you find them to be very disappointing—stale tortilla chips, lukewarm bland cheese, and some of the additional toppings you asked for are missing. “I could have made this at home and it would have tasted way better,” you think to yourself. After you finish your food and drink, Mike is nowhere to be found, so you have to get up and find another server to get your bill. After paying, you leave. In the positive own experience condition participants then read: When you enter the restaurant, a greeter warmly welcomes you and shows you to your table. After a short time waiting at your table, your server, whose name is Mike, comes up looking happy and relaxed. “Can I start you off with something to drink?” he asks nicely. After Mike takes your order, he leaves and comes back with your drink in another couple of minutes. With Mike still there, you take the opportunity to order a happy hour menu item: the Fiesta Nachos. These also come out quickly and you find them to be really tasty—fresh tortilla chips, warm zesty cheese, and all of the additional toppings you asked for are there. “I wish I could make this at home,” you think to yourself. After finishing your food and drink, you ask Mike for your bill. After paying, you leave. Next, all participants read: Now, please imagine that the following scenario occurs a week after your experience at The Copper Yard. While out shopping, you run into an acquaintance and strike up a short conversation. In the contrasting partner condition participants read: [Negative own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was pretty quick and my food was good.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. [Positive own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was pretty slow and my food wasn’t that good.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. In the similar partner condition, participants read: [Negative own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was also pretty slow and my food wasn’t that good either.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. [Positive own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was also pretty quick and my food was good too.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. All participants then read: The conversation then turns toward some other topics like mutual friends and how nice the weather has been lately. You look at your watch and realize it’s time to get going. “Good talking to you!” the person says. “Nice to see you!” you respond. Footnotes 1 We do not suggest that this is a bias, similar to negativity bias. Instead, we argue that consumers form more favorable evaluations as the result of a rational process, in which they perceive their own negative experience as a violation of generally positive performance expectations. 2 We measured personal motive attribution with the question: “How big of a role do you think personal factors (the other person’s personality, traits, character, personal style, and so on) played in their decision to share their experience?” We measured restaurant experience attribution with the question: “How big of a role do you think the actual restaurant experience (food quality, service, etc.) played in their decision to share their experience?” (1 = minimal role; 9 = maximal role) (Chen and Lurie 2013). 3 To assess performance, we used the measures: (1) I would expect to be satisfied with this product, and 2) I would expect this product to function well (Completely disagree = 1; Completely agree = 7; α = .95). 4 Of course, the analogous two-way interaction on the t2 – t1 difference scores obtains identical F and p-values (F(1,126) = 7.61, p < .01), so we report interaction effects only for the mixed ANOVA in subsequent studies. 5 Please see web appendix A for graphs showing within-subject evaluations (and their standard deviations) at time 1 and time 2, as well as effect size estimates (⁠ ηG2) of between- and within-subject factors and within-subject pairwise contrasts (Olejnik and Algina 2003). Because between-subjects contrasts represent tests of our hypotheses, we include effect sizes for these analyses in the text itself. 6 Following the study, a debriefing revealed that eight participants had suspicions that could have influenced their evaluations of the pen. Four of these participants indicated that they believed they interacted with the conversation partner to influence their opinion of the pen, three participants felt they were purposely given a faulty pen, and one participant thought the partner was not actually another student participant. However, none of the participants correctly guessed all elements of the study. Importantly, when we removed these participants from our analysis, our predicted three-way interaction remained significant (F(1,118) = 7.33, p < .01), as did all hypothesized contrasts. 7 Examination of the study debriefing questions revealed that two participants felt the conversation partner was trying to change their mind about the service discussed. However, no participant indicated that the partner was part of the study or not being truthful about his position as a student researcher. When we removed these participants, our predicted three-way interaction remained significant (F(1,206) = 4.55, p = .03), as did all hypothesized contrasts. 8 We posited that participants would have positive expectations of the products and services elicited by this prompt, given that it was similar to the recall task we used to generate the list of products and services for the expectations pretest in the theory section (appendix C). 9 To assess social positivity norms and impression management motives, we used the following items: 1) To what extent did your desire to make a positive impression on this person influence your opinion of The Copper Yard? 2) To what extent did you change your opinion of The Copper Yard in order to make a positive impression on this individual? 3) To what extent did you change your opinion of The Copper Yard because you felt pressure to be agreeable? 4) To what extent did you change your opinion of The Copper Yard in order to maintain a positive relationship with this individual? (not at all = 1, very much so = 7; α = .90). 10 Because we used PROCESS model 73, we also include full details of the analysis of the second half of the mediation path, which is the effect of the dismissal × own experience × expectations interaction on evaluations in web appendix D.1. References Ahluwalia Rohini ( 2002 ), “How Prevalent Is the Negativity Effect in Consumer Environments?” Journal of Consumer Research , 29 2 , 270 – 9 . Google Scholar Crossref Search ADS Barasch Alixandra , Berger Jonah ( 2014 ), “Broadcasting and Narrowcasting: How Audience Size Affects What People Share,” Journal of Marketing Research , 51 3 , 286 – 99 . Google Scholar Crossref Search ADS Berger Jonah , Iyengar Raghuram ( 2013 ), “Communication Channels and Word of Mouth: How the Medium Shapes the Message,” Journal of Consumer Research , 40 3 , 567 – 79 . Google Scholar Crossref Search ADS Birnbaum Gurit E. , Ein-Dor Tsachi , Reis Harry T. , Segal Noam ( 2014 ), “Why Do Men Prefer Nice Women? Gender Typicality Mediates the Effect of Responsiveness on Perceived Attractiveness in Initial Acquaintanceships,” Personality and Social Psychology Bulletin , 40 10 , 1341 – 53 . Google Scholar Crossref Search ADS PubMed Bitner Mary Jo , Booms Bernard H. , Mohr Lois A. ( 1994 ), “Critical Service Encounters: The Employee’s Viewpoint,” Journal of Marketing , 58 4 , 95 – 106 . Google Scholar Crossref Search ADS Blodgett Jeffrey G. , Hill Donna J. , Tax Stephen S. ( 1997 ), “The Effects of Distributive, Procedural, and Interactional Justice on Postcomplaint Behavior,” Journal of Retailing , 73 2 , 185 – 210 . Google Scholar Crossref Search ADS Bughin Jacques , Doogan Jonathan , Vetvik Ole Jorgen . ( 2010 ), “A New Way to Measure Word-of-Mouth Marketing,” McKinsey Quarterly, https://www.mckinsey.com/business-functions/marketing-and-sales/our-insights/a-new-way-to-measure-word-of-mouth-marketing. Chen Zoey , Berger Jonah ( 2013 ), “When, Why, and How Controversy Causes Conversation,” Journal of Consumer Research , 40 3 , 580 – 93 . Google Scholar Crossref Search ADS Chen Zoey , Lurie Nicholas H. ( 2013 ), “Temporal Contiguity and Negativity Bias in the Impact of Online Word of Mouth,” Journal of Marketing Research , 50 4 , 463 – 76 . Google Scholar Crossref Search ADS Chevalier Judith A. , Mayzlin Dina ( 2006 ), “The Effect of Word of Mouth on Sales: Online Book Reviews,” Journal of Marketing Research , 43 3 , 345 – 54 . Google Scholar Crossref Search ADS Clark Jason K. , Thiem Kelsey C. , Hoover Ann E. , Habashi Meara M. ( 2017 ), “Gender Stereotypes and Intellectual Performance: Stigma Consciousness as a Buffer against Stereotype Validation,” Journal of Experimental Social Psychology , 68 , 185 – 91 . Google Scholar Crossref Search ADS Cowley Elizabeth ( 2014 ), “Consumers Telling Consumption Stories: Word-of-Mouth and Retrospective Evaluations,” Journal of Business Research , 67 7 , 1522 – 9 . Google Scholar Crossref Search ADS Ditto Peter H. , Lopez David F. ( 1992 ), “Motivated Skepticism: Use of Differential Decision Criteria for Preferred and Nonpreferred Conclusions,” Journal of Personality and Social Psychology , 63 4 , 568 – 84 . Google Scholar Crossref Search ADS Ditto Peter H. , Scepansky James A. , Munro Geoffrey D. , Apanovitch Anne Marie , Lockhart Lisa K. ( 1998 ), “Motivated Sensitivity to Preference-Inconsistent Information,” Journal of Personality and Social Psychology , 75 1 , 53 – 69 . Google Scholar Crossref Search ADS Dubois David , Rucker Derek D. , Galinsky Adam D. ( 2012 ), “Super Size Me: Product Size as a Signal of Status,” Journal of Consumer Research , 38 6 , 1047 – 62 . Google Scholar Crossref Search ADS Dubois David , Bonezzi Andrea , De Angelis Matteo ( 2016 ), “ Sharing with Friends Versus Strangers: How Interpersonal Closeness Influences Word-of-Mouth Valence” , Journal of Marketing Research , 53 5 , 712 – 727 . East Robert , Hammond Kathy , Lomax Wendy ( 2008 ), “Measuring the Impact of Positive and Negative Word of Mouth on Brand Purchase Probability,” International Journal of Research in Marketing , 25 3 , 215 – 24 . Google Scholar Crossref Search ADS Edwards Kari , Smith Edward E. ( 1996 ), “A Disconfirmation Bias in the Evaluation of Arguments,” Journal of Personality and Social Psychology , 71 1 , 5 – 24 . Google Scholar Crossref Search ADS Enda Jodi , Mitchell Amy ( 2013 ), “Friends and Family—Important Drivers of News,” Pew Research Center, http://www.journalism.org/2013/03/17/friends-and-family-important-drivers-of-news/. Epley Nicholas , Keysar Boaz , Van Boven Leaf , Gilovich Thomas ( 2004 ), “Perspective Taking as Egocentric Anchoring and Adjustment,” Journal of Personality and Social Psychology , 87 3 , 327 – 39 . Google Scholar Crossref Search ADS PubMed Folkes Valerie S. ( 1984 ), “Consumer Reactions to Product Failure: An Attributional Approach,” Journal of Consumer Research , 10 4 , 398 – 409 . Google Scholar Crossref Search ADS Folkes Valerie S. ( 1988 ), “Recent Attribution Research in Consumer Behavior: A Review and New Directions,” Journal of Consumer Research , 14 4 , 548 – 65 . Google Scholar Crossref Search ADS Folkes Valerie S. , Kotsos Barbara ( 1986 ), “Buyers’ and Sellers’ Explanations for Product Failure: Who Done It?” Journal of Marketing , 50 2 , 74 – 80 . Google Scholar Crossref Search ADS Folkes Valerie S. , Patrick Vanessa M. ( 2003 ), “The Positivity Effect in Perceptions of Services: Seen One, Seen Them All?” Journal of Consumer Research , 30 1 , 125 – 37 . Crossref Search ADS Fornell Claes , Johnson Michael D. , Anderson Eugene W. , Cha Jaesung , Bryant Barbara Everitt ( 1996 ), “The American Customer Satisfaction Index: Nature, Purpose, and Findings,” Journal of Marketing , 60 4 , 7 – 18 . Google Scholar Crossref Search ADS Goncalo Jack A. , Vincent Lynne C. , Krause Verena ( 2015 ), “The Liberating Consequences of Creative Work: How a Creative Outlet Lifts the Physical Burden of Secrecy,” Journal of Experimental Social Psychology , 59 , 32 – 9 . Google Scholar Crossref Search ADS Hayes Andrew F. ( 2013 ), Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach , New York : Guilford . He Stephen X. , Bond Samuel D. ( 2015 ), “Why Is the Crowd Divided? Attribution for Dispersion in Online Word of Mouth,” Journal of Consumer Research , 41 6 , 1509 – 27 . Google Scholar Crossref Search ADS Herr Paul M. , Kardes Frank R. , Kim John ( 1991 ), “Effects of Word-of-Mouth and Product-Attribute Information on Persuasion: An Accessibility-Diagnosticity Perspective,” Journal of Consumer Research , 17 4 , 454 – 62 . Google Scholar Crossref Search ADS Johnson Michael D. , Anderson Eugene W. , Fornell Claes ( 1995 ), “Rational and Adaptive Performance Expectations in a Customer Satisfaction Framework,” Journal of Consumer Research , 21 4 , 695 – 708 . Google Scholar Crossref Search ADS Kahneman Daniel , Tversky Amos ( 1979 ), “Prospect Theory: An Analysis of Decision under Risk,” Econometrica: Journal of the Econometric Society , 47 2 , 263 – 91 . Google Scholar Crossref Search ADS Keller Fay Group ( 2014 ), “2014 Landmark Study: Return on Word of Mouth,” http://www.womma.org/ReturnOnWOM. Kraus Michael W. , Chen Serena ( 2009 ), “Striving to Be Known by Significant Others: Automatic Activation of Self-Verification Goals in Relationship Contexts,” Journal of Personality and Social Psychology , 97 1 , 58 – 73 . Google Scholar Crossref Search ADS PubMed Kristofferson Kirk , White Katherine , Peloza John ( 2014 ), “The Nature of Slacktivism: How the Social Observability of an Initial Act of Token Support Affects Subsequent Prosocial Action,” Journal of Consumer Research , 40 6 , 1149 – 66 . Google Scholar Crossref Search ADS Kwang Tracy , Swann William B. ( 2010 ), “Do People Embrace Praise Even When They Feel Unworthy? A Review of Critical Tests of Self-Enhancement versus Self-Verification,” Personality and Social Psychology Review , 14 3 , 263 – 80 . Google Scholar Crossref Search ADS PubMed Lynch John G. , Marmorstein Howard , Weigold Michael F. ( 1988 ), “Choices from Sets Including Remembered Brands: Use of Recalled Attributes and Prior Overall Evaluations,” Journal of Consumer Research , 15 2 , 169 – 84 . Google Scholar Crossref Search ADS Mattila Anna S. ( 2001 ), “Emotional Bonding and Restaurant Loyalty,” Cornell Hotel and Restaurant Administration Quarterly , 42 6 , 73 – 9 . Google Scholar Crossref Search ADS Moore Sarah G. ( 2012 ), “Some Things Are Better Left Unsaid: How Word of Mouth Influences the Storyteller,” Journal of Consumer Research , 38 6 , 1140 – 54 . Google Scholar Crossref Search ADS Nickerson Raymond S. ( 1999 ), “How We Know—and Sometimes Misjudge—What Others Know: Imputing One’s Own Knowledge to Others,” Psychological Bulletin , 125 6 , 737 – 59 . Google Scholar Crossref Search ADS Olejnik Stephen , Algina James ( 2003 ), “Generalized Eta and Omega Squared Statistics: Measures of Effect Size for Some Common Research Designs,” Psychological Methods , 8 4 , 434 – 47 . Google Scholar Crossref Search ADS PubMed Parasuraman Anantharanthan , Zeithaml Valarie A. , Berry Leonard L. ( 1985 ), “A Conceptual Model of Service Quality and Its Implications for Future Research,” Journal of Marketing , 49 4 , 41 – 50 . Google Scholar Crossref Search ADS Peeters Guido , Czapinski Janusz ( 1990 ), “Positive-Negative Asymmetry in Evaluations: The Distinction between Affective and Informational Negativity Effects,” European Review of Social Psychology , 1 1 , 33 – 60 . Google Scholar Crossref Search ADS Ramanathan Suresh , Williams Patti ( 2007 ), “Immediate and Delayed Emotional Consequences of Indulgence: The Moderating Influence of Personality Type on Mixed Emotions,” Journal of Consumer Research , 34 2 , 212 – 23 . Google Scholar Crossref Search ADS Ratner Rebecca K. , Kahn Barbara E. ( 2002 ), “The Impact of Private versus Public Consumption on Variety-Seeking Behavior,” Journal of Consumer Research , 29 2 , 246 – 57 . Google Scholar Crossref Search ADS Schaefers Tobias , Schamari Julia ( 2016 ), “Service Recovery via Social Media: The Social Influence Effects of Virtual Presence,” Journal of Service Research , 19 2 , 192 – 208 . Google Scholar Crossref Search ADS Schellekens Gaby A. C. , Verlegh Peeter W. J. , Smidts Ale ( 2010 ), “Language Abstraction in Word of Mouth,” Journal of Consumer Research , 37 2 , 207 – 23 . Google Scholar Crossref Search ADS Schlosser Ann E. ( 2005 ), “Posting versus Lurking: Communicating in a Multiple Audience Context,” Journal of Consumer Research , 32 2 , 260 – 5 . Google Scholar Crossref Search ADS Sinclair Stacey , Lowery Brian S. , Hardin Curtis D. , Colangelo Anna ( 2005 ), “Social Tuning of Automatic Racial Attitudes: The Role of Affiliative Motivation,” Journal of Personality and Social Psychology , 89 4 , 583 – 92 . Google Scholar Crossref Search ADS PubMed Spiller Stephen A. , Belogolova Lena ( 2016 ), “On Consumer Beliefs about Quality and Taste,” Journal of Consumer Research , 43 6 , 970 – 91 . Tice Dianne M. , Butler Jennifer L. , Muraven Mark B. , Stillwell Arlene M. ( 1995 ), “When Modesty Prevails: Differential Favorability of Self-Presentation to Friends and Strangers,” Journal of Personality and Social Psychology , 69 6 , 1120 – 38 . Google Scholar Crossref Search ADS Weber Reneé , Crocker Jennifer ( 1983 ), “Cognitive Processes in the Revision of Stereotypic Beliefs,” Journal of Personality and Social Psychology , 45 5 , 961 – 77 . Google Scholar Crossref Search ADS Zhang Yinlong , Feick Lawrence , Mittal Vikas ( 2014 ), “How Males and Females Differ in Their Likelihood of Transmitting Negative Word of Mouth,” Journal of Consumer Research , 40 6 , 1097 – 108 . Google Scholar Crossref Search ADS Zhao Xinshu , Lynch John G. Jr. , Chen Qimei ( 2010 ), “Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis,” Journal of Consumer Research , 37 2 , 197 – 206 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Consumer Research Oxford University Press

Maybe I Just Got (Un)lucky: One-on-One Conversations and the Malleability of Post-Consumption Product and Service Evaluations

Loading next page...
 
/lp/ou_press/maybe-i-just-got-un-lucky-one-on-one-conversations-and-the-uJFLL21iX9
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0093-5301
eISSN
1537-5277
D.O.I.
10.1093/jcr/ucy028
Publisher site
See Article on Publisher Site

Abstract

Abstract This research focuses on the persuasive impact of a common yet understudied form of word of mouth (WOM): one-on-one conversations in which consumers share and compare past experiences with a product or service. In contrast to prior work on WOM influence, we discover a “positivity effect” in these conversations, such that consumers who share a negative experience form more favorable overall judgments after speaking with someone who had a positive experience, but consumers who share a positive experience are unaffected by learning about another’s negative experience. This effect is mediated by consumers’ dismissal of their own negative experience as a temporary or one-off event in light of the other person’s contrasting positive experience, and is facilitated by positive consumer expectations of product and service performance. We also identify a key boundary condition whereby the positivity effect of one-on-one conversations is moderated by whether consumers have positive or negative expectations of product or service performance. When expectations are negative, the positivity effect is dampened and a negativity effect emerges. word of mouth, experience sharing, consumer expectations, attribution, positivity effect Imagine that you have just purchased the newest iPhone. After one week, the phone’s camera begins to malfunction, leaving you unable to take pictures. You happen to run into a friend and, noticing that she has the same phone, begin to complain about your own experience with it. To your surprise, your friend responds that her phone has been working great and that its camera has been functioning perfectly. How would you react to this conversation? More specifically, how would this casual discussion change your overall views of the iPhone in terms of your attitude, satisfaction, and repurchase intentions? Based on prior word-of-mouth (WOM) research (Chevalier and Mayzlin 2006; Herr, Kardes, and Kim 1991; Schlosser 2005), as well as findings showing that negative (vs. positive) information is more perceptually salient and elicits greater psychological responses (Kahneman and Tversky 1979; Peeters and Czapinski 1990), you might assume that your negative experience would stick, making you unlikely to reconsider your opinion of the phone, even when presented with your friend’s positive experience. However, our findings present evidence to the contrary. Specifically, we show that when consumers share differing accounts of a product’s or service’s performance, those who had a negative experience form more favorable overall evaluations of the product or service, including greater repurchase intentions, after talking to another person who had a contrasting positive experience. On the contrary, those who had a positive experience do not form more unfavorable evaluations after talking to another person who had a contrasting negative experience. This asymmetry suggests a positivity effect1 in these types of conversations, as evaluations formed by a negative (vs. positive) experience appear more malleable and susceptible to the influence of a conversation partner’s contrasting opinion, but not vice versa. To explain this counterintuitive effect, we develop a framework to better understand the persuasive influence of one-on-one (or dyadic) conversations in which consumers share and compare past personal experiences with a product or service. This form of WOM appears to be surprisingly common, as 75% of consumer conversations occur in a face-to-face setting or over the phone (Enda and Mitchell 2013; Keller Fay Group 2014). Additionally, 50–80% of WOM conversations involve consumers recounting direct personal experiences to one another, as opposed to merely passing along firm-generated content such as commercials or celebrity endorsements (Bughin, Doogan, and Vetvik 2010). Prior work has looked at certain aspects of these conversations by studying “storytelling” and its effect on how the sharer evaluates a product or service after talking about an experience with another person (Cowley 2014; Moore 2012). However, little research has investigated more complex interactions in which the consumer shares and then listens to another person’s experience with the same product or service. Thus, our work provides new insight into how consumers in an increasingly homogenized retail landscape react when they share and compare contrasting (vs. similar) personal experiences with the same product or service. We draw on work in attribution theory (Chen and Lurie 2013; Folkes 1984, 1988), disconfirmation bias (Ditto and Lopez 1992; Ditto et al. 1998; Edwards and Smith 1996), and consumer expectations (Folkes and Patrick 2003; Fornell et al. 1996; Johnson, Anderson, and Fornell 1995) to explain why one-on-one WOM conversations are characterized by a positivity effect. We suggest that learning of a conversation partner’s contrasting experience can lead consumers to attribute their own experience to temporary performance heterogeneity, or dismiss it as a one-time event, as opposed to attributing their experience to how the product or service typically performs. We argue that because consumers generally have positive expectations of product and service performance (Fornell et al. 1996; Johnson et al. 1995), they are more likely to dismiss a negative (vs. positive) prior experience as a one-time event upon learning of another person’s contrasting experience. Such dismissal leads consumers to perceive their own negative experience as less diagnostic (Folkes and Patrick 2003), and underlies positive shifts in evaluation. Supporting this proposed process, we show the reverse is also true: when product and service expectations are negative, as opposed to positive, this positivity effect is attenuated and a negativity effect actually occurs. These findings broadly demonstrate that when a product and service experience violates expectations, consumers more readily dismiss it in light of contradictory WOM information. Our research makes several contributions. First, we believe we are the first to provide a comprehensive framework of the persuasive influence of one-on-one conversations where consumers share similar and contrasting direct product and service experiences. Prior work studying consumer responses to contrasting opinions has primarily looked at contexts in which participants communicated within groups or with multiple anonymous others (Herr, Kardes, and Kim 1991; Schlosser 2005). Ours, however, is the first study to examine this phenomenon in a “narrowcasting” WOM context, where consumers are more interested in sharing useful information than making a positive impression on others (Barasch and Berger 2014). Although we focus on private, face-to-face conversations due to their commonality and their significant role in the transmission of WOM, in the general discussion we detail other WOM contexts with similar characteristics where our framework is also applicable. Next, we identify a novel WOM context that elicits a positivity effect, whereby positive information about a product or service is more persuasive than negative information. Although prior work has typically found negativity bias in WOM (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005), we leverage more recent research that has found attenuation of negativity bias in certain contexts (Chen and Lurie 2013), while building on the critical role of positive consumer expectations to make a case for the greater persuasion of positive WOM (Fornell et al. 1996; Johnson et al. 1995). Finally, in identifying dismissal as the attributional process underlying our effect, we believe we are the first to show how and why consumers update initially negative and positive evaluations based on post-consumption conversations with others. THEORETICAL BACKGROUND One-on-One Conversations and the Sharing of Contrasting Experiences Our research focuses on dyadic conversations in which both individuals have had a past personal experience with the same product or service. In our conceptualization of these conversations, consumers both share and compare accounts of satisfactory and unsatisfactory experiences, leading them to make attributions about a product or service’s performance (Folkes and Kotsos 1986). Therefore, we do not focus on the sharing of opinions within more taste-oriented domains, such as movies (He and Bond 2015; Schlosser 2005) or art, but rather on performance contexts where it is less ambiguous whether a product or service did or did not meet expectations. We are particularly interested in one-on-one conversations where consumers share contrasting accounts of a product or service’s performance, and the effect of this disagreement on subsequent consumer evaluations. These types of exchanges are common, as consumers are often eager to share their “true” opinions with one another, especially when products and services are the center of discussion (Chen and Berger 2013; Dubois, Bonezzi, and De Angelis 2016). Supporting this argument, when we presented 65 undergraduates with a scenario in which a classmate told them about a product experience that was the opposite of their own, 85% of participants said they would respond by sharing their own contrasting experience with the classmate (see appendix A). Why Negativity Bias Is Dampened in One-on-One Conversations What happens when these disagreements arise? Past research examining WOM influence suggests that negative shared experiences will be more influential than positive shared experiences (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005), implying a negativity bias. However, recent work has demonstrated that negativity bias in WOM is mitigated under certain circumstances. In particular, Chen and Lurie (2013) argued that negativity bias occurs only because consumers are more likely to attribute negative WOM to the product or service being mentioned, whereas they are more likely to attribute positive WOM to the sharer’s personal motives (e.g., impression management, social positivity norms), which often have little to do with the product’s or service’s actual performance. Supporting this notion, they found that in a context where consumers equally attributed positive and negative online reviews to the product or service itself (vs. the reviewer’s personal motives), they were equally persuaded by both—in other words, negativity bias was attenuated. Building on this finding, we posit that negativity bias is similarly attenuated in one-on-one conversations because consumers, without regard to valence, predominantly attribute the experiences shared in these conversations to the product or service itself (vs. the personal motives of the other person) (Chen and Lurie 2013). Specifically, in dyadic conversations, consumers narrowcast, or recount their experience to just one other person. Prior work suggests that narrowcasting leads consumers to focus more on sharing useful information and less on trying to make a positive impression (Barasch and Berger 2014; Tice et al. 1995). Drawing on these findings, we posit that consumers in dyadic conversations likely share accurate information about their own experience out of a desire to be honest and helpful, which leads them to assume that their conversation partner has the same motive (Epley et al. 2004; Nickerson 1999). As a result, we propose that within dyads, consumers predominantly attribute both positive and negative shared experiences to the product or service being discussed (vs. the personal motives of the other person), thereby attenuating negativity bias (Chen and Lurie 2013). To test this assertion, we asked 80 Amazon Mechanical Turk (MTurk) participants to imagine they were in a one-on-one conversation with a coworker where they shared contrasting experiences at a restaurant (either positive or negative; see appendix B). Following Chen and Lurie (2013), participants first indicated the extent to which they attributed the coworker’s sharing of their experience to personal motives. They then rated the extent to which they attributed the sharing to the restaurant itself.2 We then calculated a causal score by subtracting personal attributions from restaurant experience attributions (MRestaurant = 7.23, SD = 1.40 vs. MPersonal = 6.36, SD = 1.38), where higher scores indicated greater restaurant (vs. personal motive) attributions. A one-sample t-test revealed that the difference score (M = .87, SD = 2.13) was significantly greater than 0 (t(79) = 3.63, p = .001), indicating that participants predominantly attributed the coworker’s WOM to the restaurant. Consistent with our argument, this attribution was equivalent for positive and negative experiences shared by the coworker (MPositive = .76, MNegative = .97; F(1, 78) = .20, NS). These results suggest that consumers in dyadic conversations attribute their conversation partner’s experience to the product or service significantly more than personal motives, regardless of valence. How Sharing Contrasting Experiences Shapes Post-Consumption Evaluations If individuals in dyadic conversations do not rely on simple valence heuristics to determine the accuracy of one another’s experiences, how do they respond when another person’s experience contrasts from their own? Although consumers tend to view their own experiences as highly diagnostic (Herr et al. 1991), they also understand that heterogeneity exists among products and services (He and Bond 2015; Parasuraman, Zeithaml, and Berry 1985). For instance, consumers recognize that some cars in a production line can be dismissed as “lemons” due to manufacturing glitches, while others in the same line perform perfectly fine. Supporting this assertion, Folkes (1984) found that 67% of participants thought an unexpected restaurant experience could be due to a one-time occurrence and was not typical of its chronic performance. This heterogeneity should be particularly salient in exchanges where another’s experience directly contradicts the consumer’s own experience. When this happens, we posit that the consumer will interpret heterogeneity based on her general expectations and beliefs of how products and services typically perform. If a consumer perceives that her conversation partner’s contrasting experience is more consistent with how she would expect the product or service to perform, she should attribute her own experience to temporary performance heterogeneity, or the possibility that her experience was a one-time event (Folkes 1984, 1988). On the contrary, if she perceives that her conversation partner’s experience is less consistent with expectations than her own experience, she should be less likely to attribute her own experience to performance heterogeneity, leaving her evaluations unchanged. Our argument aligns with the idea that people are more critical of information that is inconsistent with their general beliefs and expectations, while uncritically accepting of information that is consistent with these beliefs and expectations (Ditto and Lopez 1992; Ditto et al. 1998). Importantly, past studies have found that consumers are generally satisfied with products and services and tend to have positive expectations regarding their performance (Folkes and Patrick 2003; Fornell et al. 1996; Johnson et al. 1995). Indeed, when we asked 66 undergraduates to rate how they would expect each of 10 popular products and services over $50 (e.g., smartphone, laptop, dinner out) to perform,3 they had very high expectations (MPerformance = 6.08 / 7, SD = .70) (appendix C). Accordingly, we posit that when a consumer shares a negative experience and learns of a conversation partner’s contrasting positive experience, she will perceive her own negative experience as less consistent with positive expectations, causing her to scrutinize her negative experience and attribute it to temporary performance heterogeneity (i.e., dismiss it as a one-time event). This should lead her to see her negative experience as less indicative of the product or service’s chronic performance (Folkes and Patrick 2003), and make her less likely to use this experience in forming a subsequent judgment (Lynch, Marmorstein, and Weigold 1988). As a result, she should adjust her evaluations upward, in line with positive expectations. On the other hand, if she shares a positive experience and her conversation partner shares a contrasting negative experience, her own experience is more consistent with positive expectations, so she will be less likely to scrutinize this positive experience (vs. her partner’s negative experience) and dismiss it as a one-time event. As a result, no shift in her evaluations should occur. Importantly, when the consumer and conversation partner share similar experiences, we do not expect shifts in evaluations since performance heterogeneity should not be salient and consumers have no reason to question the attribution of their own experience or change their evaluation. Accordingly, we consider our control conditions to be those in which the consumer and conversation partner share similar experiences. Thus, our focal contrasts are those that examine the shifts in evaluation when consumers learn of a partner’s contrasting (vs. similar) experience. We present our predictions for the basic positivity effect and its underlying mechanism below: H1:There will be an interactive effect of the consumer’s own experience and her conversation partner’s experience on shifts in the consumer’s general evaluations of a product or service: When a consumer shares a negative experience and learns of a partner’s contrasting positive (vs. similar negative) experience, the consumer will form a more favorable evaluation of the product or service. When the consumer shares a positive experience and learns of a partner’s contrasting negative (vs. similar positive) experience, the consumer will not change her evaluation of the product or service. H2:When the consumer shares a negative experience and learns of a conversation partner’s contrasting positive (vs. similar negative) experience, the consumer’s attribution (or dismissal) of her own experience as a temporary or one-time event will mediate more favorable evaluations. To this point, we have argued that consumers adjust evaluations of their own negative experience upward because they perceive this experience to be less consistent with positive expectations than the conversation partner’s positive experience. We now further investigate this argument by looking at how negative expectations can reverse this effect. The Case of Negative Expectations What happens when consumers have negative expectations? Consistent with our theorizing, we argue that when a consumer with negative product or service expectations has a negative experience and then learns of a conversation partner’s contrasting positive experience, he should perceive his own negative experience as more consistent with expectations. As a result, he should not dismiss his negative experience as a one-time event, leaving initial evaluations unaffected. Conversely, if this consumer has a positive experience and learns of a partner’s contrasting negative experience, he should see his own experience as less consistent with expectations than the partner’s experience, leading him to dismiss his own positive experience as a one-time event, downwardly shifting initial evaluations. In sum: H3:There will be an interactive effect of consumer expectations, the consumer’s own experience, and the conversation partner’s experience on shifts in consumers’ general evaluations of a product or service: When expectations are positive, we will replicate our hypothesis 1 predictions. When expectations are negative, our results will reverse our hypothesis 1 predictions, more specifically: When a consumer shares a negative experience and learns of a partner’s contrasting positive (vs. similar negative) experience, the consumer will not change his evaluation. When a consumer shares a positive experience and learns of a partner’s contrasting negative (vs. similar positive) experience, the consumer will form a more unfavorable evaluation. We summarize our predictions and the studies in which they are tested in table 1. TABLE 1 SUMMARY OF FOCAL PREDICTIONS Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift TABLE 1 SUMMARY OF FOCAL PREDICTIONS Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift Expectations Own experience Partner’s contrasting experience Process Change in own evaluations (relative to similar experience) Hypothesis and study tested Positive Negative Positive Dismiss own experience as one-time event Upward shift Hypothesis 1 (basic effect): studies 1–4 Hypothesis 2 (mediation): studies 3 and 4 Positive Negative Do not dismiss own experience No change Negative Negative Positive Do not dismiss own experience No change Hypothesis 3 (moderation by expectations): study 4 Positive Negative Dismiss own experience as one-time event Downward shift We test our predictions in four studies across a diverse array of one-on-one WOM conversations. In studies 1 and 2, we use novel in-person conversation settings to show that when participants share a negative experience with a conversation partner who shares a contrasting positive experience, their initial evaluations shift upward. On the other hand, no shift occurs when participants share a negative experience and learn of the partner’s contrasting positive experience (hypothesis 1). In study 3, we replicate this effect in a vignette-based conversation and find support for dismissal as our proposed process (hypothesis 2). Finally, in study 4, we manipulate performance expectations to moderate the persuasive effect of a conversation partner’s contrasting experience (hypothesis 3). To demonstrate the generalizability and external validity of our findings, we investigate a range of real and simulated dyadic relationships including those with a likeable other (studies 1 and 2), close friend (study 3), and casual acquaintance (study 4). Moreover, we test products and services that participants 1) are obliged to use for a specific task (study 1); 2) are structurally constrained to use within campus life (study 2); 3) chose themselves in real life (study 3); and 4) chose as part of a scenario (study 4). Such variation allows us to demonstrate that this phenomenon applies to products with varying price points and levels of involvement. STUDY 1 In study 1, we provide initial evidence that when consumers share a negative product experience with a conversation partner who shares a contrasting positive (vs. similar negative) experience, they will form more favorable evaluations. We expect no such shift when consumers share a positive product experience and the partner shares a contrasting negative experience (hypothesis 1). To test this prediction, we randomly assign participants to have a negative or positive experience with a university-branded pen and measure time 1 evaluations of it. Next, as an ostensibly separate study on conversation styles, we randomly assign participants to have a conversation with a likeable “student” confederate partner who happens to mention a contrasting or similar experience with the same pen. We then measure time 2 pen evaluations. Method Participants and Design We randomly assigned 139 undergraduates at a large southwestern university to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed experiment. We removed nine participants with whom the confederate could not share his opinion of the pen before the conversation ended, leaving 130 participants. Procedure Upon entering the lab, we asked participants to fill out a sheet of basic biographical information (e.g., year of birth, favorite food) for a study later in the session. We provided all participants with a university-branded ballpoint pen with which to complete this sheet. Pretesting indicated that participants had reasonably positive (though not extreme) expectations of the pen’s performance—that is, significantly above a neutral midpoint of 4 out of 7 (M = 4.92; t(267) = 11.13, p < .001) (see appendix D for further details). The pen thus represented a strong initial test of our theory since it allowed us to test whether participants would dismiss a negative experience based on the moderately positive performance expectations of a mundane product. In the negative experience condition, we had intentionally damaged the ballpoint tip of the pen by briefly holding it up to a flame. Thus, the pen was not visibly damaged yet wrote poorly, resulting in an unambiguously negative product experience. In the positive experience condition, the pen wrote normally. After participants finished the first task, we gave them a new survey stating that we were collecting feedback on the behavioral lab and asked them to evaluate the usability of some lab equipment, such as workstations. As part of this survey, we asked participants to complete a general evaluation of the university-branded pens using a three-item index of attitude, satisfaction, and usage likelihood (α = .93): 1) “How would you rate your attitude toward the [university] pens that we have provided for these studies?” 2) “How satisfied were/are you with the [university] pens that we have provided for these studies?” and 3) “How likely would you be to use the [university] pens that we have provided for these studies?” Here, and in all future studies, across the time 1 and time 2 evaluation questions, we purposefully did not put numbers on the scales (only word labels) to make it more difficult for participants to anchor on initial responses. After a brief delay, we informed participants that they had been selected to participate in an unrelated study on conversation styles and that, as part of the study, we wanted them to talk to another student participant. Unbeknownst to the participant, the conversation partner was an actor paid to pose as a fellow undergraduate. A lab assistant led the participant and conversation partner to a separate room and instructed them to have a short discussion about two things they liked about being a student at the business school. After a short pause, the confederate shared two things he liked about being a business school student and then listened as the participant shared what he or she liked. This conversation was meant to encourage rapport between the participant and the partner and foster likeability (Sinclair et al. 2005) to make the participant more comfortable with the confederate, thereby leading to a more natural interaction. Next, the conversation partner casually asked the participant: “Hey, did they have you take that study where you had to evaluate the lab equipment?” After the participant responded that he or she had done the study, the partner continued, “Did they ask you to rate the pen for some reason? Did yours work okay?” After the participant shared their experience with the pen, the confederate responded by indicating that he had had either a contrasting or similar experience: “That’s interesting, the pen they gave me didn’t write all that well (wrote perfectly fine).” We felt such an exchange would not feel odd or out of place given that participants had just rated the pen and likely viewed it as something common to discuss with the confederate. Finally, at the end of the conversation, the confederate pointed to a bowl of candy and offered some to the participant to further foster likeability (Sinclair et al. 2005). We included this behavior to 1) increase the contextual similarity to a conversation with a likable other, a person with whom these types of discussions occur most frequently (Enda and Mitchell 2013; Keller Fay Group 2014), and 2) decrease impression management pressures, as people are more comfortable being themselves around people they feel close to (Kraus and Chen 2009; Kwang and Swann 2010). At this point, the lab assistant knocked on the door and instructed participants to return to their seats, where they were asked to complete a survey further exploring lab equipment perceptions. We asked participants to reconsider their prior evaluations and respond to the same three-item index for the university-branded pens (α = .91). Finally, although all responses to our questions were completed in private, to account for the possibility that impression management motives were stronger in the negative (vs. positive) own experience condition, we asked participants: “To what extent did you try to make a positive impression on the person you talked to as part of this study? (1 = not at all; 7 = very much so).” Results and Discussion We predicted that participants who shared a negative experience would form more favorable evaluations following exposure to the conversation partner’s contrasting positive (vs. similar negative) experience, while participants who shared a positive experience would not change their evaluations following exposure to the partner’s contrasting negative (vs. similar positive) experience (hypothesis 1). Across all studies, for ease of exposition and increased clarity, we first report the results of our mixed-design ANOVA. We then present within-subjects time 1 to time 2 contrasts, which test whether the magnitude of the evaluation shift is greater than 0. Finally, we follow with our focal between-subjects contrasts of the t2 – t1 difference scores. These complex contrasts test hypothesis 1 by comparing shifts in participant evaluations depending on whether they interacted with a contrasting (vs. similar) partner. Main Results We conducted a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed ANOVA on our evaluation index. There was a main effect of own experience valence (MPositive = 4.91 vs. MNegative = 2.88; F(1, 126) = 83.83, p < .001). Of note, initial evaluations of negative experiences at time 1 fell well below neutral (4) on a seven-point scale, indicating that we successfully manipulated an undoubtedly negative experience in the lab. On the other hand, there was no main effect of the conversation partner’s experience (F(1, 126) = .00, NS) or time (before [t1] vs. after [t2] the conversation) (F(1, 126) = 1.67, p = .20). There were also no significant two-way interactions. Importantly, we found a significant own experience × partner’s experience × timing interaction (F(1, 126) = 7.61, p < .01).4 We next examine our within- and between-subjects contrasts. As described above, because our focal predictions are based on the changes in evaluations in the contrasting and similar conditions, figure 1 and all subsequent figures provide an illustration of shifts in participant evaluations in the form of difference scores before and after the manipulation. Consistent with past research looking at evaluation changes over time (Ramanathan and Williams 2007), we use proc mixed to test both within- and between-subjects planned contrasts. All statistical tests retain the full degrees of freedom. We also report effect sizes for each hypothesized contrast.5 For completeness, we first present within-subjects t2/t1 contrasts across all conditions, and then test hypothesis 1 with our focal complex contrasts. FIGURE 1 View largeDownload slide STUDY 1: SHIFT IN EVALUATIONS (T2 – T1) FIGURE 1 View largeDownload slide STUDY 1: SHIFT IN EVALUATIONS (T2 – T1) Our within-subjects contrasts revealed that among participants sharing a negative experience (or negative sharers), when the conversation partner had a contrasting positive experience, we observed an upward shift in evaluations (t1 = 2.60 vs. t2 = 3.03, Δt2–t1 = .43; t(126) = 3.02, p = .003). On the other hand, when the partner had a similar negative experience, evaluations did not shift (t1 = 3.00 vs. t2 = 2.90, Δt2–t1 = –.10; t(126) = –.74, NS). Among participants who shared a positive experience (or positive sharers), we found no change in evaluations when the partner shared a contrasting negative experience (t1 = 5.03 vs. t2 = 4.92, Δt2–t1 = –.11; t(126) = –.77, NS) or a similar positive experience (t1 = 4.76 vs. t2 = 4.92, Δt2–t1 = .16; t(126) = .99, NS). To test hypothesis 1, we performed between-subjects contrasts of the t2 – t1 difference scores among participants in the contrasting and similar partner conditions. As stated above, we consider the similar partner experience condition our control. Supporting hypothesis 1a, negative sharers whose partner shared a contrasting positive experience increased their evaluations more than those whose partner shared a similar negative experience (ΔContrasting =.43 vs. ΔSimilar = –.10; ηG2= ⁠.055; F(1, 126) = 1.57, p < .01). Supporting hypothesis 1b, positive sharers whose partner shared a contrasting negative experience did not differ from those whose partner shared a similar positive experience (ΔContrasting = –.11 vs. ΔSimilar = .16; ηG2= ⁠.012; F(1, 126) = 1.57, p = .21). Impression Management A 2 (own experience) × 2 (conversation partner’s experience) between-subjects ANOVA on our measure of participants’ desire to please the partner revealed a nonsignificant two-way interaction (F(1, 126) = .97, NS) and no significant contrasts, suggesting no differences in impression management across conditions. Further, a 2 (own experience) × 2 (partner’s experience) × 2 (time) mixed ANOVA controlling for impression management showed that our hypothesized three-way interaction remained significant (F(1, 125) = 7.01, p < .01). Debriefing Our study focal effects and contrasts remained robust to any concerns highlighted in the participant debriefing responses.6 Discussion In study 1, we provide preliminary evidence in support of hypothesis 1. When participants had a negative experience with the pen, they evaluated it more positively when the conversation partner shared a contrasting positive (vs. similar negative) experience. However, when participants had a positive experience, the partner’s contrasting negative experience did not influence evaluations. Further, a measure of impression management provides preliminary evidence that it is unlikely our results are driven by participants wanting to be seen more positively after recounting a negative (vs. positive) experience. While we recognize that the upward shift in evaluations is relatively small, it does represent a nontrivial 17% improvement in initial evaluations, and is consistent with research showing that people typically modify their beliefs gradually and incrementally (Weber and Crocker 1983). Most importantly, in demonstrating this effect, we reverse prior work showing a negativity bias in WOM (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005). Of course, we recognize that because participants only briefly used the pen and it is a lower-involvement product, they may not have given much thought or concern to its failure or success. As a result, they may have formed less concrete initial evaluations, and thus upward shifts in pen evaluations might not generalize to a higher-involvement product or service used on a regular basis. We address this issue in study 2. STUDY 2 In study 2, we seek to replicate our results by asking students to recall and discuss an on-campus service experience, which we posit is more meaningful to them and important to their daily life. We first ask undergraduate participants to think about their experiences with on-campus services. This provides a context where all participants have shared knowledge from which they can reasonably draw similar or contrasting opinions based on an experience. To simulate a dyadic WOM conversation, we randomly assign participants to share a negative or positive on-campus service experience with a confederate posing as another student collecting data for a class project. In addition, we look at whether participants update judgments only of a service experience that is shared (vs. not shared) during the conversation to rule out the possibility that encountering a conversation partner’s positive experience leads to broader positive emotions that influence judgment beyond the specific service discussed. Method Participants and Design We randomly assigned 219 undergraduates from a large southwestern university to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed design experiment. Seven participants who could not recall experiences with a service but completed the survey anyway were removed from the analysis, leaving 212 participants. Procedure Upon entering the lab, participants were randomly assigned to receive one of two versions of an “On-Campus Services Study.” Students were given a list of four on-campus services (dining, health, career, and academic advising) and asked to circle the one they felt most negatively (positively) about, based on an unsatisfactory (satisfactory) experience. Pretesting indicated that participants had positive expectations of all four services (Madvising = 5.58; Mhealth = 5.40; Mdining = 4.94; Mcareer = 5.59; see appendix D for full details). They then wrote two or three sentences describing the experience. Next, participants completed a survey in which they gave an initial evaluation of all four campus services (including the one they selected). We assessed initial evaluations of each service using the same three-item index from study 1 (αdining = .87, αhealth = .92, αcareer = .88, αadvising = .83). Next, we led each participant to a separate room where he or she met the confederate partner. As in study 1, the conversation partner posed as another student and employed a likeability manipulation (Sinclair et al. 2005). As before, this manipulation was done to foster familiarity and make participants feel comfortable. Specifically, when participants entered the room, the partner stood up, shook their hand, and gave them a warm greeting. He then invited them to sit down and introduced himself as a fellow student working on a class project and looking for student feedback regarding on-campus services. Next, he offered participants a piece of candy as a token of appreciation for taking the time to talk with him. He then asked participants if they could briefly share the positive or negative service experience they had chosen to write about. After participants shared their experience, the conversation partner indicated having had either a contrasting or similar experience with the service (e.g., “I actually really like (dislike) [service name]”) and stated that his experience had been either similar or the opposite. He then thanked participants and let them leave. Importantly, to simulate a closer relationship than the last study and thereby increase the generalizability of our findings, before they left, the confederate told participants they would be meeting him later in the session for a different project. This was done to account for the possible limitation that participants may have expected to speak with the confederate only once. Because participants most likely run into their student peers repeatedly and thus expect future interactions, this subsequent conversation expectation was another way to increase the external validity of our lab conversations. Upon returning to the lab, to capture time 2 evaluations, participants completed an additional survey further exploring on-campus services in which we asked them to reconsider their evaluations and respond to the same three-item evaluation index measure for each service (αdining = .89, αhealth = .92, αcareer = .90, αadvising = .87). Lastly, we asked participants to complete final debriefing questions to assess whether they guessed the study hypotheses. As mentioned above, we additionally explored whether participants updated evaluations for the services they did not share with the confederate to see whether other changes that could occur between time 1 and time 2 evaluations, such as more positive emotions, could explain shifts in evaluations that had nothing to do with the conversation. Results and Discussion Main Results We conducted a 2 (own experience) × 2 (conversation partner’s experience) × 2 (time, within) mixed ANOVA on the focal service evaluations. There was a main effect of own experience valence (MPositive = 6.12 vs. MNegative = 3.41; F(1, 208) = 281.43, p < .001) and a main effect of time, whereby evaluations improved after the interaction (MBefore = 4.79 vs. MAfter = 4.87; F(1, 208) = 4.19, p = .04). There was no effect of the conversation partner’s experience (F(1, 208) = 1.31, p = .25). We also found a significant own experience × time interaction (F(1, 208) = 13.46, p < .001). Supporting hypothesis 1, these interactions were qualified by a three-way interaction (F(1, 208) = 5.17, p = .02) (see figure 2). FIGURE 2 View largeDownload slide STUDY 2: SHIFT IN EVALUATIONS (T2 – T1) FIGURE 2 View largeDownload slide STUDY 2: SHIFT IN EVALUATIONS (T2 – T1) Next, our within-subjects contrasts revealed that when negative sharers were exposed to the conversation partner’s contrasting positive experience, their evaluations became more favorable after the interaction (t1 = 3.53 vs. t2 = 3.93, Δt2–t1 = .40; t(208) = 4.75, p < .001). On the contrary, when the partner shared a similar negative experience, evaluations did not shift (t1 = 3.29 vs. t2 = 3.37, Δt2–t1 = .08; t(208) = .94, NS). Next, among positive sharers, there was no change when the partner shared a contrasting negative experience (t1 = 6.15 vs. t2 = 6.05, Δt2–t1 = –.10; t(208) = 1.21, p = .23) or when the partner shared a similar positive experience (t1 = 6.16 vs. t2 = 6.13, Δt2–t1 = –.04; t(208) = .44, NS). Supporting hypothesis 1a, our focal complex contrast on the t2 – t1 difference scores of negative sharers revealed that the shift observed when the partner shared a contrasting positive experience was greater than that observed when the partner shared a similar negative experience (ΔContrasting = .40 vs. ΔSimilar = .08; ηG2= ⁠.032; F(1, 208) = 6.82, p = .01). Supporting hypothesis 1b, the corresponding contrast of the difference scores of positive sharers was not significant (ΔContrasting = –.10 vs. ΔSimilar = –.04; ηG2= ⁠.002; F(1, 208) = .32, NS). Evaluations of Nonselected Services We also tested whether participants’ evaluations of the services they did not share with the conversation partner changed following the conversation to ensure our effects did not emerge as an artifact of the two-time-period study design on a single product or service. It is possible participants felt more positive emotions after the conversation and thus perceived all services more favorably. To investigate this question, we computed participants’ mean evaluations of the three services they did not share with the partner before and after the interaction. For instance, if participants discussed dining services with the partner, we calculated the mean of their evaluations of career, health, and academic advising services. Analysis revealed no significant main effect of the conversation partner’s experience (F(1, 208) = .88, NS) on these nonselected evaluations. There was a significant main effect of own experience valence (MPositive = 4.51 vs. MNegative = 5.20; F(1, 208) = 52.96, p < .001) and timing (MBefore = 4.77 vs. MAfter = 4.87; F(1, 208) = 25.69, p < .001). However, there was no significant three-way interaction (F(1, 208) = .00, NS). Moreover, when breaking out each nonselected service, we found a similarly nonsignificant three-way interaction for health (F(1, 208) = .10, NS), dining (F(1, 208) = .40, NS), advising (F(1, 208) = .10, NS), and career (F(1, 208) = .95, NS) services. Debriefing As in study 1, our results remained robust to suspicion of the confederate.7 Discussion Study 2 replicates our support for hypothesis 1 in a new dyadic conversation setting with a familiar, frequently used service. We also show that the focal effect of consumers shifting their own negative experience upward in the face of another’s contrasting positive experience holds only for shared (vs. unshared) experiences within the dyad, revealing the instrumental role of encountering another’s contrasting opinion in changing opinions, and thereby serving as a type of within-subjects control. We should note that in web appendix B, we also include a between-subjects control condition in a vignette-based study where participants do not imagine sharing their experience with another individual, and show there are no differences between sharing a similar experience and lack of sharing altogether. In study 3, we test our proposed process. STUDY 3 Study 3 shifts to a conversation vignette and seeks to achieve three objectives. First, given that WOM most frequently occurs among close friends and family (Enda and Mitchell 2013), we seek to replicate our focal effect in a context where participants imagined conversing with an actual close friend, with products or services they purchased in the past. Second, using a vignette allows us to better control both transmitter and recipient expression, and is consistent with prior WOM research (Berger and Iyengar 2013; Moore 2012; Schellekens, Verlegh, and Smidts 2010; Zhang, Feick, and Mittal 2014). Third, we look for evidence that more favorable evaluations among negative sharers in the contrasting (vs. similar) partner condition are driven by a process in which participants are more likely to dismiss their negative experience as a one-time event because it is less consistent with positive performance expectations than the conversation partner’s positive experience (hypothesis 2). Method Participants and Design We randomly assigned 238 MTurk participants to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (time: before vs. after conversation, within) mixed-design experiment. One participant who could not recall a product or service but completed the survey was removed from the analyses, leaving 237 participants. Procedure We asked participants to think about a product or service costing at least $50 that they had purchased in the previous six months and that they felt performed unsatisfactorily (satisfactorily) (Zhang et al. 2014). Next, we asked them to enter the name of the product or service and write one or two sentences describing their experience. Participants tended to recall commonly used consumer items such as laptops, cell phones, and video games.8 Afterward, we asked participants to evaluate the product or service using the three-item evaluation index from the prior two studies (α = .97). However, because we asked participants to name a product they had previously purchased (and not just used, as with the lab pens and on-campus services), we modified the question, “How likely would you be to use this product or service?” to “How likely would you be to purchase this product or service again?” (1 = extremely unlikely; 7 = extremely likely). Next, we asked participants to “enter the name of a close friend that is important to you and has been for several years” (Kraus and Chen 2009). We asked these participants to imagine they were having a casual conversation with their friend and told him/her about the product or service experience they had written about. Participants in the similar partner condition then read: “To your surprise, [close friend’s name] responds that he/she also really likes (dislikes) that product or service and appears satisfied that the two of you share the same opinion. However, neither of you discuss the matter any further.” In the contrasting partner condition, they read: “To your surprise, [close friend’s name] responds that he/she actually really likes (dislikes) that product or service and appears perplexed that you both do not share the same opinion. However, neither of you discuss the matter any further.” We next captured our mediator, asking participants to complete a four-item measure of the extent to which they attributed their experience to temporary performance heterogeneity—that is, dismissed it as a one-time event following the interaction (α = .88). We based these measures on Folkes and Patrick’s (2003) characterization of performance heterogeneity attribution as the recategorization of an individual experience as an outlier rather than the norm. The measures were: 1) “To what extent do you feel that your experience with this product or service was an isolated event and not representative of the product in general?” 2) “To what extent do you think your experience with this product or service was an accurate indicator of its overall performance?” 3) “To what extent do you think you would have a similar experience upon using this product or service again?” and 4) “To what extent should your experience shape your overall opinion of the product or service?” (all anchored at 1 = not at all, 7 = very much so). Items 2–4 were reverse-coded and all items were combined to form our dismissal index (α = .88). Finally, participants completed the same three-item, time 2 evaluation index (α = .96). Results and Discussion Main Results We performed a 2 (own experience) × 2 (conversation partner’s experience) × 2 (time) mixed ANOVA on the evaluation index. We found main effects of own experience valence (MPositive = 6.11 vs. MNegative = 2.51; F(1, 233) = 723.11, p < .001) and time (MBefore = 4.22 vs. MAfter = 4.40; F(1, 233) = 19.53, p < .001). There was no effect of partner experience (F(1, 233) = 1.29, p = .26). We also found a significant own experience × time (F(1, 233) = 25.92, p < .001) and a marginal partner experience × time interaction (F(1, 233) = 3.68, p = .06). Importantly, these interactions were qualified by a significant three-way interaction (F(1, 233) = 7.39, p < .01) (see figure 3). FIGURE 3 View largeDownload slide STUDY 3: SHIFT IN EVALUATIONS (T2 – T1) FIGURE 3 View largeDownload slide STUDY 3: SHIFT IN EVALUATIONS (T2 – T1) Turning to our within-subjects contrasts, among negative sharers, when the conversation partner had a contrasting positive experience, evaluations shifted upward following the imagined interaction (t1 = 2.24 vs. t2 = 2.83, Δt2–t1 = .59; t(233) = 7.06, p < .001). Interestingly, when the partner had a similar negative experience, evaluations also became more positive (t1 = 2.39 vs. t2 = 2.59, Δt2–t1 = .20; t(233) = 2.37, p = .02). We elaborate on this finding further in the discussion. However, among positive sharers, when the conversation partner had a contrasting negative experience, evaluations did not change (t1 = 6.27 vs. t2 = 6.21, Δt2–t1 = –.06; t(233) = -.74, NS), nor did they shift when the partner had a similar positive experience (t1 = 5.98 vs. t2 = 5.98, Δt2–t1 = .01; t(233) = .07, NS). Supporting hypothesis 1a, a complex contrast on the t2 – t1 difference scores of negative sharers revealed that the shift observed when the partner shared a contrasting positive (vs. similar negative) experience was greater (ΔContrasting = .59 vs. ΔSimilar = .20; ηG2= ⁠.043; F(1, 233) = 10.44, p = .001). Supporting hypothesis 1b, the corresponding complex contrast among positive sharers was nonsignificant (ΔContrasting = –.06 vs. ΔSimilar = .01; ηG2= ⁠.001; F(1, 233) = .033, NS). Mediator (Dismissal) We ran a 2 (own experience: positive vs. negative, between) × 2 (partner’s experience: contrasting vs. similar, between) ANOVA on our dismissal measure. We found a main effect of own experience valence, such that negative (vs. positive) experiences were more likely to be dismissed as temporary in nature (MNegative = 3.22 vs. MPositive = 2.16; F(1, 233) = 47.39, p < .001), indicating that negative (vs. positive) experiences that defy high expectations are more likely to lead consumers to attribute their experience to heterogeneity (Folkes 1984). There was also a main effect whereby dismissal was higher when the partner’s experience was contrasting (vs. similar) (MContrasting = 2.88 vs. MSimilar = 2.49; F(1, 233) = 6.47, p = .01). As expected, there was a significant own experience × partner experience interaction (F(1, 233) = 5.53, p = .02) (see figure 4). FIGURE 4 View largeDownload slide STUDY 3: DISMISSAL OF OWN EXPERIENCE FIGURE 4 View largeDownload slide STUDY 3: DISMISSAL OF OWN EXPERIENCE Supporting hypothesis 2, negative sharers who learned of a partner’s contrasting positive (vs. similar negative) experience were more likely to dismiss their negative experience following the conversation (MContrasting = 3.59 vs. MSimilar = 2.84; ηG2= ⁠.048; F(1, 233) = 11.63, p = .001). Among positive sharers, there was no such difference (MContrasting = 2.17 vs. MSimilar = 2.14; ηG2= ⁠.000; F(1, 233) = .02, NS). Moderated Mediation Analysis We tested moderated mediation using PROCESS model 59 of the bootstrapping process described by Hayes (2013) on the t2 – t1 evaluation index difference scores. This model has been used in other studies to allow the moderator (valence of own experience) to interact with both the independent variable (partner’s experience) and the mediator (dismissal) (Birnbaum et al. 2014; Clark et al. 2017; Goncalo, Vincent, and Krause 2015). This is the most robust model to analyze our effect, given that the valence of the consumer’s own experience also affects the direction of evaluation change following dismissal (see web appendix C for a full explanation of the model, its graphical depiction, and detailed analysis). As predicted, for negative sharers, dismissal mediated the effect of the partner’s contrasting positive (vs. similar negative) experience on upward shifts in evaluations (BIndirect = .24, CI95% [.07, .57]). Specifically, negative sharers dismissed their own experience as a one-time event when they learned of the conversation partner’s contrasting (vs. similar) experience (B = .75, CI95% [.32, 1.19]). As dismissal increased, participants evaluated the product or service more favorably (B = .32, CI95% [.24, .41]). On the contrary, among positive sharers, dismissal did not mediate (BIndirect = –.002, CI95% [–.04, .03]). The index of moderated mediation further confirmed that our moderated mediation analysis was successful (B = –.24, CI95% [–.56, –.07]). Discussion In study 3, we again find support for hypothesis 1 by showing that, when faced with a conversation partner’s contrasting experience, consumers adjust negative judgments upward. Critically, supporting hypothesis 2, when negative (vs. positive) sharers became aware of a partner’s contrasting experience, they were more likely to dismiss their own experience as a one-time event, which mediated the upward shift in evaluations of the product or service. Surprisingly, in this study, participants who shared a negative experience evaluated products and services more positively after learning of a close partner’s similar negative experience. While we did not observe this shift in any other studies, we note that the magnitude of the shift (Δ = .20) is still substantially smaller than the focal contrasting experience (Δ = .59), and, as will be evident next, similar to the time 1 to time 2 shifts in study 4 (Δ = .14, also a vignette). STUDY 4 Thus far, we have argued that consumers dismiss their own negative experiences as one-time events when faced with another’s contrasting positive experience because of their generally high expectations of product and service performance. However, it is theoretically important to test our process in the rarer case where consumers have negative expectations, despite the increased uncertainty and variance that may occur when testing these products and services. Thus, in study 4, we directly manipulate performance expectations. Specifically, in hypothesis 3, we argued that when consumers have negative expectations, those who had a negative experience with a product or service should no longer dismiss their own experience when faced with another’s contrasting positive experience. Analogously, when expectations are negative, one should be more likely to dismiss one’s own positive experience after learning of another’s contrasting negative experience, reversing the positivity effect of prior studies. To test this in study 4, we manipulate expectations of a new restaurant. This study also seeks to further address the alternative explanation of impression management. We found preliminary evidence that impression management motives did not differ among negative (vs. positive) sharers in study 1, but we take two further steps to rule out impression management as a mediator. First, we ask participants to imagine their conversation partner is an acquaintance (vs. a close friend), as increasing social distance heightens impression management concerns and thus provides a more rigorous test for this alternative (Kraus and Chen 2009; Kwang and Swann 2010). Second, we use a four-item measure that explicitly captures whether participants changed their evaluations to satisfy impression management/positivity norms. We acknowledge the possibility that negative (vs. positive) sharers may respond to general impression management concerns by wanting to outwardly appear more agreeable. However, we believe that such concerns are unlikely to mediate our positivity effect since responses for time 2 evaluations are taken privately after the interaction, and research shows that impression management motives influence public, but not private, behavior and opinions (Dubois, Rucker, and Galinsky 2012; Kristofferson, White, and Peloza 2014; Ratner and Kahn 2002). We reasoned that if we indeed did not find mediation by impression management concerns with a more nuanced four-item measure and a more socially distant partner, we could more confidently support our claim that impression management is not the primary driver of the focal effects. Method Participants and Design We randomly assigned 400 MTurk participants to a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (restaurant expectations: positive vs. negative, between) × 2 (time: before vs. after conversation, within) mixed-design experiment. We removed eight duplicate participants included due to an error in the filtering of individuals who had previously taken the survey, leaving 392 participants. Of note, our predicted interaction remained significant when these participants were retained (F(1, 384) = 19.54, p < .001), as did our hypothesized contrasts. Procedure We told participants we were conducting a study on how news stories affect product and service evaluations. We first asked participants to imagine they were looking for information on a fictional chain restaurant, The Copper Yard, and to read a top Google search article for this restaurant. In the positive expectations condition, we showed participants a fictional Consumer Reports article rating The Copper Yard as excellent on its food and customer service (see appendix E). In the negative expectations condition, the article rated the restaurant as poor on both these dimensions. Next, we asked participants to imagine they looked up additional information for the Copper Yard franchise in their town. In the positive (negative) expectations condition, we presented them with a fictional local newspaper article describing the restaurant’s high (low) sales and expert opinions of its high- (low-) quality food and service. Both articles highlighted that the restaurant was currently revamping its menu to provide a plausible reason for visiting the restaurant. After reading the articles, participants rated their expectations of The Copper Yard with the same two-item measure used in the product expectations pretest from the literature review (α = .96; see footnote 4 for items). We asked participants to read two articles to add realism to a situation in which consumers might form expectations by learning general information about a national chain, and then looking up specific information about the franchise in their area. We also wanted to ensure that our negative expectations manipulation was sufficiently strong and enduring to set below-neutral expectations. Next, we asked participants to imagine trying The Copper Yard because it was advertising a new happy hour special and was close to their home (see appendix F). Participants in the positive own experience condition imagined having a prompt and pleasant server and eating a tasty dish, while participants in the negative own experience condition imagined having a slow and unpleasant server and eating a mediocre dish. Participants then evaluated The Copper Yard on the same three questions from study 3 (attitude, satisfaction, and purchase likelihood), as well as a fourth item asking “How likely would you be to go to The Copper Yard again in the future?” (α = .96). We added this item to capture intentions to return to the restaurant, as prior work has found repeat loyalty to be an important indicator of consumer restaurant evaluations (Mattila 2001). Next, we asked participants to imagine that, a week after their experience at The Copper Yard, they ran into an acquaintance and struck up a short conversation that included a discussion about local restaurants. As described, we purposefully chose an acquaintance as the conversation partner to support the notion that our effect is robust to varying levels of social distance. Participants then imagined sharing their experience at The Copper Yard with the acquaintance and that the acquaintance then shared a contrasting or similar restaurant experience. To make the conversation seem natural, we asked participants to imagine that they casually shared a few more details about their experiences before exhausting the subject and turning to other topics. Participants then rated the extent to which they would dismiss their own experience as a one-time event following the conversation using the four-item measure from study 3 (α = .88). They also responded to a time 2, four-item evaluation index for The Copper Yard (α = .99). Finally, we asked participants a four-item measure of how much they were influenced to change their opinion of The Copper Yard9 based on impression management concerns. Importantly, while our study 1 impression management measure captured general impression management motives in each condition, our new measure directly asked participants whether they changed their evaluation to satisfy an impression management motive, thereby examining whether impression management motives, as opposed to dismissal, mediated our results. Results and Discussion We predicted that when participants had positive expectations, we would replicate our prior positivity effect: negative sharers would form more positive evaluations upon learning of the conversation partner’s contrasting positive (vs. similar negative) experience, whereas positive sharers would not change their evaluations. On the contrary, when participants had negative expectations, we expected a negativity effect such that negative sharers would no longer form more favorable evaluations following exposure to the partner’s contrasting positive (vs. similar negative) experience. Instead, we predicted that positive sharers would now form more negative evaluations after exposure to the partner’s contrasting negative (vs. similar positive) experience. Of note, because intentionally using products for which consumers have negative expectations is less common, we are less certain of how consumers will behave outside of the explicit contrasts we predict in hypothesis 3b. Manipulation Check A one-way ANOVA confirmed that our expectations manipulation was successful (MPositive = 6.47 vs. MNegative = 2.12; F(1, 384) = 2124.06, p < .001). Main Results We conducted a 2 (own experience: positive vs. negative, between) × 2 (conversation partner’s experience: contrasting vs. similar, between) × 2 (restaurant expectations: positive vs. negative, between) × 2 (time: before vs. after conversation, within) mixed ANOVA on the evaluation index. We found main effects of own experience valence (MPositive = 6.19 vs. MNegative = 2.06; F(1, 384) = 2123.80, p < .001), restaurant expectations (MPositive = 4.57 vs. MNegative = 3.67; F(1, 384) = 101.37, p < .001), and time (MBefore = 4.06 vs. MAfter = 4.19; F(1, 384) = 20.78, p < .001); significant own experience × time (F(1, 384) = 137.66, p < .001), partner experience × time (F(1, 384) = 15.36, p < .001), and expectations × time (F(1, 384) = 39.96, p < .001) interactions; and a marginally significant partner experience × expectations interaction (F(1, 384) = 3.08, p = .08). We also found significant own experience × partner experience × time (F(1, 384) = 44.57, p < .001), own experience × expectations × time (F(1, 384) = 5.06, p < .03), and partner experience × expectations × time (F(1, 384) = 27.26, p < .001) interactions. Most importantly, supporting hypothesis 3, these were qualified by a four-way interaction (F(1, 384) = 19.54, p < .001) (see figure 5). FIGURE 5 View largeDownload slide STUDY 4: SHIFT IN EVALUATIONS (T2 – T1) A. POSITIVE EXPECTATIONS B. NEGATIVE EXPECTATIONS FIGURE 5 View largeDownload slide STUDY 4: SHIFT IN EVALUATIONS (T2 – T1) A. POSITIVE EXPECTATIONS B. NEGATIVE EXPECTATIONS We first examined the positive expectations conditions. Among individuals sharing a negative experience, replicating prior studies, within-subjects evaluations shifted significantly upward in the contrasting partner condition (t1 = 2.11 vs. t2 = 3.43, Δt2–t1 = 1.32; t(384) = 17.44, p < .001). Conversely, these evaluations did not significantly change in the similar partner condition (t1 = 2.31 vs. t2 = 2.45, Δt2–t1 = .14; t(384) = 1.61, p = .11). Next, among individuals sharing a positive experience, evaluations did not change in the contrasting partner condition (t1 = 6.57 vs. t2 = 6.43, Δt2–t1 = –.15; t(384) = –1.58, p = .12), nor did they shift in the similar partner condition (t1 = 6.66 vs. t2 = 6.63, Δt2–t1 = –.03; t(384) = .42, NS). Supporting hypothesis 1 and replicating studies 1–3, a complex contrast on the t2 – t1 difference scores among those sharing a negative experience showed that the shift in the contrasting (vs. similar) partner condition was greater (ΔContrasting = 1.32 vs. ΔSimilar = .14; ηG2= ⁠.222; F(1, 384) = 109.41, p < .001). The corresponding contrast among positive sharers was not significant (ΔContrasting = –.15 vs. ΔSimilar = –.03; ηG2= ⁠.002; F(1, 384) = .87, NS). We next examined the negative expectations conditions. Among individuals who shared a negative experience, within-subjects evaluations shifted significantly upward to relatively the same degree in the contrasting partner condition (t1 = 1.32 vs. t2 = 1.57, Δt2–t1 = .25; t(384) = 3.23, p = .001) and in the similar partner condition (t1 = 1.53 vs. t2 = 1.72, Δt2–t1 = .20; t(384) = 2.16, p = .03). Next, among individuals with negative expectations who shared a positive experience, evaluations shifted significantly downward in the contrasting partner condition (t1 = 5.93 vs. t2 = 5.49, Δt2–t1 = –.43; t(384) = 5.14, p < .001). Interestingly, evaluations also shifted significantly downward in the similar partner condition (t1 = 6.02 vs. t2 = 5.79, Δt2–t1 = –.23; t(384) = 3.10, p = .002), but to a lesser extent. Supporting hypothesis 3, a follow-up contrast on the difference scores of those sharing a negative experience revealed no differences between the two shifts (ΔContrasting = .25 vs. ΔSimilar = .20; ηG2= ⁠.001; F(1, 384) = .21, NS). As we will discuss later, this contrast indicates that the upward shift observed among negative sharers in the contrasting partner condition is unlikely to be related to the partner’s shared experience, which is the focus of our predictions, but rather some other factor. An additional contrast on the difference scores of positive sharers showed that the negative shift in the contrasting negative (vs. similar positive) partner condition was marginally greater (ΔContrasting = –.43 vs. ΔSimilar = –.23; ηG2= ⁠.009; F(1, 384) = 3.51, p = .06). This suggests that although positive sharers became more negative in both conditions, those in the contrasting partner condition were most negative, presumably due to their dismissal of the negative experience as a one-time event, consistent with our theory. Finally, we ran two additional difference score contrasts to test how another’s distinct experience can shape dismissal differently when people have positive versus negative expectations. Consistent with our theory, the upward shift in evaluations among negative sharers in the contrasting partner condition was significantly attenuated when expectations were negative (vs. positive) (ΔPositive exp = 1.32 vs. ΔNegative exp = .25; ηG2= ⁠.199; F(1, 384) = 95.61, p < .001). Next, the downward shift in evaluations among positive sharers in the contrasting partner condition was significantly enhanced when expectations were negative (vs. positive) (ΔPositive exp = –.15 vs. ΔNegative exp = –.43; ηG2= ⁠.013; F(1, 384) = 5.11, p = .02). Mediator (Dismissal) We conducted a 2 (own experience) × 2 (conversation partner’s experience) × 2 (restaurant expectations) between-subjects ANOVA on our dismissal mediator. We found main effects of own experience (F(1, 384) = 10.03, p = .002) and partner experience (F(1, 384) = 76.02, p < .001) and significant own experience × partner experience (F(1, 384) = 5.92, p = .02), own experience × expectations (F(1, 384) = 148.12, p < .001), and partner experience × expectations (F(1, 384) = 7.51, p = .01) two-way interactions. Most importantly, we found a significant own experience × partner experience × expectations interaction (F(1, 384) = 6.59, p = .01). Replicating study 3, planned contrasts revealed that in the positive expectations condition, negative sharers were more likely to dismiss their own experience when the conversation partner shared a contrasting positive (vs. similar negative) experience (MContrasting = 4.30 vs. MSimilar = 2.62; ηG2= ⁠.159; F(1, 384) = 72.38, p < .001). Further, in the negative expectations condition, positive sharers were more likely to dismiss their own experience when the partner shared a contrasting negative (vs. similar positive) experience (MContrasting = 3.36 vs. MSimilar = 2.73; ηG2= ⁠.026; F(1, 384) = 10.35, p < .001). Third, among negative sharers who learned of a partner’s contrasting positive experience, dismissal was significantly higher when general expectations were positive (vs. negative) (MPositive exp = 4.30 vs. MNegative exp = 2.42; ηG2= ⁠.202; F(1, 384) = 97.24, p < .001). Moderated Mediation Analysis We tested moderated mediation using PROCESS model 73 (Hayes 2013) on the t2 – t1 evaluation index difference scores. This model is the three-way analog to PROCESS model 59, used in study 3, and has been used in prior research to allow both moderators to interact with both the independent variable and the mediator10 (Clark et al. 2017) (see web appendix D.1). We first examined participants with positive expectations. For those who had positive expectations and shared a negative experience, dismissal mediated the upward shift in evaluations in the contrasting (vs. similar) partner experience condition (BIndirect = .57, CI95% [.35, .89]). On the contrary, for those who had positive expectations and shared a positive experience, dismissal did not mediate shifts in evaluations (BIndirect = –.02, CI95% [–.12, .05]). Also, as predicted, for those who had negative expectations but shared a positive experience, dismissal mediated the downward shift observed when the partner had a contrasting negative (vs. similar positive) experience (BIndirect = –.14, CI95% [–.25, –.05]). Finally, for participants who shared a negative experience, dismissal mediated the changes in evaluations (BIndirect = .12, CI95% [.02, .30]), though these changes were not different in the contrasting (vs. similar) partner conditions and so should be interpreted in such a light. Alternative Mediator We also tested whether positivity norms/impression management mediated our results. There was a significant three-way interaction on our index (F(1, 382) = 6.53, p = .01). Driving this interaction, under negative expectations, positive sharers had the greatest impression management concerns upon learning of the partner’s contrasting negative experience. Notably, this pattern is inconsistent with our focal evaluation results, where we found the largest shift under positive expectations, when negative sharers learned of the partner’s contrasting positive experience. More importantly, impression management concerns did not mediate in any of the conditions (all confidence intervals included 0). Further, following Zhao, Lynch, and Chen (2010), we tested this variable in parallel with our dismissal mediator (see web appendix D.2 for full details) and again found no evidence that these motives mediated our effect in any conditions. Discussion In study 4, we show that expectations play an important role in how dyadic conversations shape consumer evaluations. When participants had positive expectations, we replicated prior studies and found the same positivity effect and mediation. On the other hand, when participants had negative expectations, we found a different pattern of results. In particular, the upward shift among negative sharers in the contrasting partner condition was dampened, and was no different from the shift in the similar partner condition. Additionally, there was a significant downward shift among positive sharers in the contrasting partner condition, and this shift was marginally greater than the shift in the similar partner condition. Given people are less likely to use products and services for which they have negative expectations, it is less surprising that we found some unpredicted effects in these conditions—most notably increases and drops of ∼.20 from time 1 to time 2 evaluations. We note that though significant, a change of ∼.20 is much smaller than the focal changes documented in prior studies or in the positive expectations of study 4, which are all substantially larger (.42 in study 1, .40 in study 2, .59 in study 3, and 1.32 in the positive expectations condition of study 4). Thus, we expect the significance of these smaller changes is more dependent on the increased sample size of study 4, as opposed to substantial changes in beliefs themselves. Importantly, our focal drop, whereby people holding negative expectations were more likely to lower their own positive evaluations in light of another’s negative evaluations, is more in line with the changes (–.43) across prior studies. GENERAL DISCUSSION Across four studies, we test a framework to illustrate the influence of one-on-one conversations where consumers share and compare experiences with a product or service. We demonstrate a positivity effect in these conversations, whereby consumers who share a negative experience evaluate products and services more favorably after learning about someone else’s contrasting positive experience, but not vice versa. Furthermore, we identify the underlying process behind this positivity effect: positive expectations of product and service performance lead consumers to attribute their own negative experience to temporary performance heterogeneity (i.e., dismiss it as a one-time event). We also delineate a theoretical boundary condition whereby when consumers have negative product and service expectations, they no longer dismiss their own negative experience upon learning of a conversation partner’s contrasting positive experience, and the positivity effect is attenuated. Conversely, negative expectations cause consumers to dismiss their own positive experience after they learn of a conversation partner’s negative experience, leading to a negativity effect. Importantly, our focal pattern of effects remains consistent across a range of real and simulated dyadic interactions with differing levels of closeness (e.g., likable other, close friend, casual acquaintance), a diverse set of products and services (ballpoint pens, campus services, costly [>$50] products bought by participants, a restaurant experience), and oppositely valenced performance expectations. These varying experimental paradigms increase generalizability and help to rule out specific alternative accounts. First, the similar pattern of effects with close others and acquaintances suggests that the malleability of evaluations we observe is unlikely to be driven by spillover from impression management concerns. The range of products used also makes a self-repair account (whereby consumers increase evaluations to make themselves feel better about a bad experience) less plausible. The fundamental attribution error suggests that participants most likely blame the product or service, not themselves, for their negative experience (Bitner, Booms, and Mohr 1994; Folkes and Kotsos 1986), a bias that should be particularly strong with the more mundane pen and campus services used in studies 1 and 2, which were provided to (and not chosen by) participants. Moreover, our study 4 finding—that when expectations are negative, participants who had a positive experience shift their evaluations downward—is inconsistent with a self-repair account, as it is unlikely consumers would want to feel worse about having had a positive experience. In sum, the replication of our focal positivity effect across dyad types and products, its mediation by dismissal, and the subsequent reversal of our positivity effects when expectations are negative provides convergent evidence of expectations, experience, and WOM shaping consumer evaluations. We make several contributions to the literature. First, in recent years, consumer researchers have largely focused on the effect of online WOM, such as consumer reviews. As a result, there is little knowledge regarding the persuasive impact of interpersonal communication that often occurs offline, such as casual conversations where consumers share and compare past product or service experiences. These conversations appear to be surprisingly common and powerful, as consumers in an increasingly homogenized retail landscape often use the same products and services, and share these experiences with one another (Bughin et al. 2010; Enda and Mitchell 2013; Keller Fay Group 2014). Past work has looked at some aspects of these conversations in settings where participants communicate within groups after reading about a product (Herr et al. 1991), or with multiple anonymous online others (Schlosser 2005). Other work has looked at how “storytelling” impacts the sharer (Cowley 2014; Moore 2012), but not how the sharer responds to another person’s experience with the same product or service. Thus, our work provides new insights into the persuasive influence of these more complex interactions. Next, although past WOM research has largely found that negative WOM is more powerful than positive WOM (Chevalier and Mayzlin 2006; Herr et al. 1991; Schlosser 2005), more recent work has found that this negativity bias is attenuated under certain circumstances (Chen and Lurie 2013). Leveraging this latter finding, we demonstrate a positivity effect in one-on-one conversations, as consumers who have generally positive expectations dismiss their own negative experience as a one-time event after learning of another’s positive experience, but not vice versa. Thus, to our knowledge, ours is the first WOM research to indicate that positive information regarding a product or service experience can be weighed more heavily than negative information. Finally, in identifying the underlying mechanism of our framework, we also integrate work in attribution theory (Chen and Lurie 2013; Folkes 1984, 1988), disconfirmation bias (Ditto and Lopez 1992; Ditto et al. 1998; Edwards and Smith 1996), and consumer expectations (Folkes and Patrick 2003; Fornell et al. 1996; Johnson et al. 1995) to show that dismissal underlies the greater malleability of evaluations formed by negative (vs. positive) product and service experiences, as consumers decrease their subsequent usage of these experiences when they perceive them to be the result of temporary performance heterogeneity. In sum, we believe we are the first to develop a framework to better understand how a common form of dyadic, post-consumption conversation shapes judgments of products and services. Future Research While we tried to increase the external validity of our research by using a variety of dyadic conversation contexts, one limitation of our work is the large breadth of variation in these types of conversations that we do not explore. For instance, we do not look at whether the consumer having one prior experience versus multiple prior experiences moderates how she responds to the conversation partner’s experience. However, our framework would predict that if the consumer has had multiple negative experiences prior to interacting with the partner, she would be less likely to perceive performance heterogeneity solely based on the partner’s contrasting positive experience and, therefore, be less likely to change her evaluation. We also did not explore how consumers might update their evaluations over multiple conversations with the same (or a different) partner. Whereas subsequent conversations with the same partner may not lead to new inferences regarding performance heterogeneity, conversations with a different partner might provide additional cues regarding heterogeneity, and be more persuasive as a result. By investigating these questions in more detail, future research could reveal the full complexity of these conversations. Another question for future research is whether a dyadic conversation context is always necessary for our theorized effect to occur. For instance, two (or more) individuals may also share contrasting and similar experiences within a larger group setting. Prior studies suggest that a negative sharer is more likely to attribute a positive sharer’s experience to personal motives (i.e., impression management) when both are “broadcasting” in front of a group (Barasch and Berger 2014; Chen and Lurie 2013). This should decrease the relative influence of the positive (vs. negative) sharer’s experience and dampen the positivity effect observed in dyadic conversations. This intuition is supported by past studies finding negative (vs. positive) WOM to be more powerful in group contexts (Herr et al. 1991; Schlosser 2005). However, we do acknowledge that certain types of group conversations may have a persuasive effect similar to that of dyadic conversations, such as a consumer sharing a negative experience with two very close friends. If one or both friends respond with a positive experience, the consumer may have little reason to believe that they are trying to show off, and would likely dismiss his own negative experience the same way he might if only talking with one friend. Indeed, he would probably be even more likely to dismiss his negative experience upon learning that both of his friends had a positive experience. Future research could explore under what conditions group conversations have a similar persuasive effect as dyadic conversations, or look at more complex scenarios in which multiple individuals share contrasting or similar experiences that the consumer must interpret and use when forming a new judgment. Future research could also move from face-to-face to other types of interactions. For instance, a person could see a friend’s comment about a product or service on her social media feed or read an online story about another person’s experience. While it is hard to say how people would respond in other contexts, we believe our results would hold in any setting where consumers predominantly attribute another person’s WOM to the product or service itself, as opposed to the reviewer’s personal motives (Chen and Lurie 2013). Thus, it is possible that positivity effects emerge not only in one-on-one contexts, but also in situations where consumers attribute the review directly to the product or service, as signaled by temporal contiguity (Chen and Lurie 2013), trustworthiness of the reviewer, or objectivity of the review source. Studies could also investigate when impression management motives play a larger role in shaping evaluations. In study 4, we did find that these motives were more salient in conversations where an acquaintance shared a contrasting (vs. similar) experience. However, changes in evaluations were not mediated by impression management motives, perhaps because participant responses were made in private. Thus, there are likely situations where impression management influences participant evaluations, perhaps when people must make public evaluations or consume a product or service with a partner they do not know very well. For example, if two people are trying to decide where to go out to eat and their contrasting opinions about a given restaurant emerge as they list possible options, a consumer who wants to be seen as agreeable would probably be more willing to dismiss a positive experience with the restaurant if his partner had a bad experience there and does not want to visit it again. Another interesting question raised by our research is whether WOM being associated with performance versus taste (Spiller and Belogolova 2016) moderates its persuasive impact. Specifically, while our research focused on conversations in which consumers share experiences based on product and service performance, they may also share WOM regarding their tastes, such as opinions about movies or music. One possibility is that expectations are more ambiguous in taste-related domains, as consumers may be uncertain how information about a product or service relates to their own subjective preferences. Additionally, they may not be influenced by another person’s taste-related opinion unless they perceive that person’s tastes to be similar to their own. Future research could look at how consumers construct expectations in taste-related domains, as well as when and why they are persuaded by another’s taste-related opinions. Future research could also further explore how negative expectations impact the way that consumers experience products and services, and how these experiences affect subsequent consumer perceptions. Though less common, consumers are often obligated to use products and services for which they expect performance to be poor (e.g., health insurers, cable companies, airlines). In these cases, products with negative expectations often have some redeeming aspect that leads consumers to try them (extremely low rates, new management taking over, etc.). It would be interesting to examine how these positive attributes are considered by consumers and contribute to evaluation changes. Notably, one limitation of our work is that we measure participant evaluations at only a single level of granularity. In particular, while we capture participants’ overall product and service evaluations following a conversation (including repurchase intentions), we do not measure participants’ evaluations of their own product/service or the conversation partner’s product/service. How might these lower-level evaluations compare to the general evaluations that we capture? We suggest that negative sharers who learn of a partner’s positive experience evaluate their own product or service more negatively following the conversation, as they come to see it as a negative outlier, or “lemon.” On the other hand, their view of the partner’s product or service likely reflects general evaluations, given that they perceive it to be an exemplar of how that product or service typically performs. Future studies could further explore these nuances. Managerial Implications Why is it important to better understand the influence of one-on-one conversations where consumers share past experiences with a product or service? First, in an increasingly franchised retail landscape, there should be greater convergence in the types of products and services consumers try and therefore discuss offline. As a result, it is important to understand how consumer perceptions are shaped when they compare experiences with the same product or service. Second, although there is a broad literature looking at how managers should respond to negative WOM (Blodgett, Hill, and Tax 1997; Schaefers and Schamari 2016), other studies suggest that a greater focus on positive WOM can be equally (or more) beneficial (Ahluwalia 2002; East, Hammond, and Lomax 2008). Our findings give managers a better idea of when and why they should focus on responding to positive versus negative WOM. In particular, in industries where consumers have high expectations, companies may want to prioritize the sharing of positive consumer experiences (via incentives, retweets, etc.), rather than spending all their efforts to mitigate the spread of negative consumer experiences. Conversely, in industries where consumers have low expectations (e.g., airlines, health insurance, specific government services), companies might do better to focus on addressing negative WOM rather than promoting positive WOM, as the latter is more likely to be dismissed. Thus, our findings suggest that WOM strategy can potentially be segmented, depending on consumer expectations of the industry. Further building upon the role of expectations, some managers might assume that they can overcome negative expectations by delivering a positive in-person experience to the customer. While this is somewhat true, our findings suggest that negative expectations continue to loom over positive experiences, as consumers can be persuaded by subsequent negative WOM to lower their evaluations of the product or service. This suggests that if managers are successful at both setting positive expectations and delivering a positive in-person experience, customers will be largely buffered from subsequent negative WOM that they encounter. In providing a framework to evaluate the persuasive impact of these conversations, we have only begun to understand some of the ways in which dyadic, post-consumption WOM may shape product and service evaluations. Interestingly, in a world where performance information regarding products and services is readily available, even when consumers have personal experience to draw from, one-on-one conversations still play a significant role in shaping consumer evaluations and repurchase likelihood. DATA COLLECTION INFORMATION All studies were designed and analyzed by the first author under the guidance of the second author. The data for study 1 was collected at the Arizona State University Behavioral Lab in spring 2016. The data for study 2 was collected at the Arizona State University Behavioral Lab in fall 2014. The data for study 3 was collected on Amazon’s Mechanical Turk in fall 2014. Data for study 4 was collected on MTurk in fall 2017. The data for the first pretest was collected at the Arizona State University Behavioral Lab in fall 2016. The data for the second pretest was collected on MTurk in fall 2017. The data for the third pretest was collected at the Arizona State University Behavioral Lab in fall 2016. The data for the study in web appendix B was collected on MTurk in fall 2014 (data for the added control condition was collected in fall 2016). Research assistants collected the data for studies 1 and 2, as well as all pretests run, at the Arizona State University Behavioral Lab. The first author collected the data for studies 3, 4, pretest 2, and the study in web appendix B. The authors would like to thank Jim Bettman and Kirk Kristofferson for their valuable feedback. The authors are also grateful to the editor, associate editor, and three anonymous reviewers for their insight and guidance. Supplementary materials are included in the web appendix accompanying the online version of this article. APPENDIX A—Pretest 1 We conducted a short pretest in which we asked 65 undergraduate participants to imagine the following scenario: “You are having a conversation with a peer from one of your classes. You run into each other while waiting in line at the grocery store. You don’t know him very well, but he is very friendly and nice. The topic comes around to cleaning and it turns out that you have both purchased the same brand of robot vacuum cleaner.” We next randomly assigned participants to imagine that the peer stated having had either a good or bad experience with the vacuum. We then asked participants to imagine that they had had the opposite experience with the vacuum, and asked, “Would you share your experience with him?” (Yes/No). As expected, 85% of participants indicated that they would share their contrasting opinion of the vacuum in this situation. Moreover, this willingness to share occurred regardless of whether the peer had had a positive or negative experience with the product (χ2 = .53, p = .40). APPENDIX B—Pretest 2 We asked 80 MTurk participants to read the following scenario: “Imagine that you are having a lunchroom conversation with a peer that you know from work. Somehow, the topic comes around to local restaurants, and whether it is worthwhile to go to a restaurant that just opened nearby.” We then randomly assigned participants to imagine that they had a positive (negative) experience at the restaurant, which they shared with the coworker, and that the coworker replied that they had a contrasting negative (positive) experience. We then assessed the extent to which participants attributed the coworker’s experience to a personal motive with the question: “How big of a role do you think personal factors (the other person’s personality, traits, character, personal style, and so on) played in their decision to share their experience?” (1 = minimal role; 9 = maximal role). Next, we assessed the extent to which they attributed the coworker’s experience to the restaurant with the question: “How big of a role do you think the actual restaurant experience (food quality, service, etc.) played in their decision to share their experience?” Both questions came from Chen and Lurie (2013). Further following the procedure from Chen and Lurie (2013), we then subtracted personal motive attribution from restaurant experience attribution, so that positive scores indicated greater restaurant experience (vs. personal motive) attribution. A one-sample t-test revealed the diff score (M = .86, SD = 2.13) to be significantly greater than 0 (t(79) = 3.63, p = .001), indicating that participants predominantly attributed the coworker’s WOM to the restaurant. Consistent with our argument, this attribution was equivalent for positive and negative experiences shared by the coworker (MPositive = .76, MNegative = .97; F(1, 78) = .20, NS). APPENDIX C—Pretest 3 We identified 10 popular consumer products and services (dinner at a nice restaurant, smartphone, headphones, internet service, a night at a hotel, a pair of athletic shoes, car repair service, laptop computer, vacuum cleaner, video game system) that emerged from an open-ended question about a product or service people had purchased in the past six months that was over $50. We then separately asked 66 undergraduate participants to indicate whether they had purchased each product or service in the past two years (1 = yes; 2 = no), as well as how they would expect each product or service to perform, using a two-item satisfaction measure: 1) I would expect to be satisfied with this product and 2) I would expect this product to function well (1 = strongly disagree; 7 = strongly agree) (α = .95). Averaging the satisfaction measure across all 10 products, we found that participants had very positive expectations of product and service performance (MSatisfaction = 6.08 out of 7, SD = .70) (see table 2 for full details). TABLE 2 SATISFACTION EXPECTATIONS FOR COMMONLY PURCHASED PRODUCTS AND SERVICES Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) TABLE 2 SATISFACTION EXPECTATIONS FOR COMMONLY PURCHASED PRODUCTS AND SERVICES Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) Product or service % who have purchased in the past two years Mean expected satisfaction (out of 7) Dinner at a nice restaurant 92% 6.27 (SD = .84) Smartphone 91% 6.34 (SD = .80) Headphones 85% 6.03 (SD = .89) Internet service 83% 6.24 (SD = .97) A night at a hotel 78% 5.98 (SD = .91) A pair of athletic shoes 77% 5.87 (SD = 1.00) Car repair service 75% 5.94 (SD = 1.08) Laptop computer 62% 6.37 (SD = .86) Vacuum cleaner 48% 5.84 (SD = .93) Video game system 42% 5.95 (SD = 1.19) APPENDIX D—Expectations Pretest for Study 1 and Study 2 Stimuli In this study, we are interested in students’ perceptions of various products and services. On the following pages, we will list a product or service and ask you to indicate how you perceive that product or service along several dimensions. Study 1 Stimuli Product: A pen from the ASU Bookstore Imagine that you had to use this pen to fill out some information. Two hundred sixty-eight undergraduate participants responded to the following four items (α = .88): “If you were going to use this pen, would you expect it to perform positively?” “If you were going to use this pen, would you expect it to write well?” “If you were going to use this pen, would you expect it not to write well?” (reverse-scored) “Do you think that these pens typically perform well?” (1 = not at all; 7 = very much so). The result of a one sample t-test indicated that mean expectations were above the neutral midpoint (M = 4.92; t(267) = 11.13, p < .001). Study 2 Stimuli Services: Academic Advising, Health Services, Dining Services, and Career Services Two hundred sixty-eight undergraduate participants responded to the following four items for academic advising (α = .82), health services (α = .83), dining services (α = .85), and career services (α = .82): “If you were going to use [service name], would you expect this service to perform positively?” If you were going to use [service name], would you expect a good service experience?” “If you were going to use this service, would you expect a bad service experience?” (reverse-scored) “Do you think that [service name] typically provides a good service experience?” (1 = not at all; 7 = very much so). One sample t-tests indicated that mean expectations of all services were above the neutral midpoint (Madvising = 5.58; t(267) = 22.69, p < .001, Mhealth = 5.40; t(267) = 18.81, p < .001, Mdining = 4.94; t(267) = 12.49, p < .001, Mcareer = 5.59; t(267) = 24.19, p < .001). APPENDIX E—Article Stimuli for Study 4 APPENDIX F— Study 4 Vignettes All participants first read: [Negative expectations condition] The Copper Yard has been advertising a new happy hour special for the past week. Even though you have low expectations of the restaurant, it is close to your house, so you decide to give it a try. [Positive expectations condition] The Copper Yard has been advertising a new happy hour special for the past week. You have high expectations of the restaurant and it is close to your house, so you decide to give it a try. In the negative own experience condition, participants then read: When you enter the restaurant, no one is waiting up front to show you to a table, so you decide to sit at the nearest open table. After what seems like forever waiting, your server, whose name is Mike, finally comes up looking somewhat hurried and bothered. “Can I start you off with something to drink?” he asks briskly. After Mike takes your order, he leaves and doesn’t return with your drink for another 20 minutes. With Mike still there, you take the opportunity to order a happy hour menu item: the Fiesta Nachos. These also take a long time to come out and you find them to be very disappointing—stale tortilla chips, lukewarm bland cheese, and some of the additional toppings you asked for are missing. “I could have made this at home and it would have tasted way better,” you think to yourself. After you finish your food and drink, Mike is nowhere to be found, so you have to get up and find another server to get your bill. After paying, you leave. In the positive own experience condition participants then read: When you enter the restaurant, a greeter warmly welcomes you and shows you to your table. After a short time waiting at your table, your server, whose name is Mike, comes up looking happy and relaxed. “Can I start you off with something to drink?” he asks nicely. After Mike takes your order, he leaves and comes back with your drink in another couple of minutes. With Mike still there, you take the opportunity to order a happy hour menu item: the Fiesta Nachos. These also come out quickly and you find them to be really tasty—fresh tortilla chips, warm zesty cheese, and all of the additional toppings you asked for are there. “I wish I could make this at home,” you think to yourself. After finishing your food and drink, you ask Mike for your bill. After paying, you leave. Next, all participants read: Now, please imagine that the following scenario occurs a week after your experience at The Copper Yard. While out shopping, you run into an acquaintance and strike up a short conversation. In the contrasting partner condition participants read: [Negative own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was pretty quick and my food was good.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. [Positive own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was pretty slow and my food wasn’t that good.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. In the similar partner condition, participants read: [Negative own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was also pretty slow and my food wasn’t that good either.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. [Positive own experience condition] Somehow, the conversation turns to local restaurants and you bring up your recent experience at The Copper Yard. After listening to your experience, the person says, “That’s interesting, I ate there once and my server was also pretty quick and my food was good too.” You each go on to casually share a few more details about your experiences until you’ve exhausted the subject. All participants then read: The conversation then turns toward some other topics like mutual friends and how nice the weather has been lately. You look at your watch and realize it’s time to get going. “Good talking to you!” the person says. “Nice to see you!” you respond. Footnotes 1 We do not suggest that this is a bias, similar to negativity bias. Instead, we argue that consumers form more favorable evaluations as the result of a rational process, in which they perceive their own negative experience as a violation of generally positive performance expectations. 2 We measured personal motive attribution with the question: “How big of a role do you think personal factors (the other person’s personality, traits, character, personal style, and so on) played in their decision to share their experience?” We measured restaurant experience attribution with the question: “How big of a role do you think the actual restaurant experience (food quality, service, etc.) played in their decision to share their experience?” (1 = minimal role; 9 = maximal role) (Chen and Lurie 2013). 3 To assess performance, we used the measures: (1) I would expect to be satisfied with this product, and 2) I would expect this product to function well (Completely disagree = 1; Completely agree = 7; α = .95). 4 Of course, the analogous two-way interaction on the t2 – t1 difference scores obtains identical F and p-values (F(1,126) = 7.61, p < .01), so we report interaction effects only for the mixed ANOVA in subsequent studies. 5 Please see web appendix A for graphs showing within-subject evaluations (and their standard deviations) at time 1 and time 2, as well as effect size estimates (⁠ ηG2) of between- and within-subject factors and within-subject pairwise contrasts (Olejnik and Algina 2003). Because between-subjects contrasts represent tests of our hypotheses, we include effect sizes for these analyses in the text itself. 6 Following the study, a debriefing revealed that eight participants had suspicions that could have influenced their evaluations of the pen. Four of these participants indicated that they believed they interacted with the conversation partner to influence their opinion of the pen, three participants felt they were purposely given a faulty pen, and one participant thought the partner was not actually another student participant. However, none of the participants correctly guessed all elements of the study. Importantly, when we removed these participants from our analysis, our predicted three-way interaction remained significant (F(1,118) = 7.33, p < .01), as did all hypothesized contrasts. 7 Examination of the study debriefing questions revealed that two participants felt the conversation partner was trying to change their mind about the service discussed. However, no participant indicated that the partner was part of the study or not being truthful about his position as a student researcher. When we removed these participants, our predicted three-way interaction remained significant (F(1,206) = 4.55, p = .03), as did all hypothesized contrasts. 8 We posited that participants would have positive expectations of the products and services elicited by this prompt, given that it was similar to the recall task we used to generate the list of products and services for the expectations pretest in the theory section (appendix C). 9 To assess social positivity norms and impression management motives, we used the following items: 1) To what extent did your desire to make a positive impression on this person influence your opinion of The Copper Yard? 2) To what extent did you change your opinion of The Copper Yard in order to make a positive impression on this individual? 3) To what extent did you change your opinion of The Copper Yard because you felt pressure to be agreeable? 4) To what extent did you change your opinion of The Copper Yard in order to maintain a positive relationship with this individual? (not at all = 1, very much so = 7; α = .90). 10 Because we used PROCESS model 73, we also include full details of the analysis of the second half of the mediation path, which is the effect of the dismissal × own experience × expectations interaction on evaluations in web appendix D.1. References Ahluwalia Rohini ( 2002 ), “How Prevalent Is the Negativity Effect in Consumer Environments?” Journal of Consumer Research , 29 2 , 270 – 9 . Google Scholar Crossref Search ADS Barasch Alixandra , Berger Jonah ( 2014 ), “Broadcasting and Narrowcasting: How Audience Size Affects What People Share,” Journal of Marketing Research , 51 3 , 286 – 99 . Google Scholar Crossref Search ADS Berger Jonah , Iyengar Raghuram ( 2013 ), “Communication Channels and Word of Mouth: How the Medium Shapes the Message,” Journal of Consumer Research , 40 3 , 567 – 79 . Google Scholar Crossref Search ADS Birnbaum Gurit E. , Ein-Dor Tsachi , Reis Harry T. , Segal Noam ( 2014 ), “Why Do Men Prefer Nice Women? Gender Typicality Mediates the Effect of Responsiveness on Perceived Attractiveness in Initial Acquaintanceships,” Personality and Social Psychology Bulletin , 40 10 , 1341 – 53 . Google Scholar Crossref Search ADS PubMed Bitner Mary Jo , Booms Bernard H. , Mohr Lois A. ( 1994 ), “Critical Service Encounters: The Employee’s Viewpoint,” Journal of Marketing , 58 4 , 95 – 106 . Google Scholar Crossref Search ADS Blodgett Jeffrey G. , Hill Donna J. , Tax Stephen S. ( 1997 ), “The Effects of Distributive, Procedural, and Interactional Justice on Postcomplaint Behavior,” Journal of Retailing , 73 2 , 185 – 210 . Google Scholar Crossref Search ADS Bughin Jacques , Doogan Jonathan , Vetvik Ole Jorgen . ( 2010 ), “A New Way to Measure Word-of-Mouth Marketing,” McKinsey Quarterly, https://www.mckinsey.com/business-functions/marketing-and-sales/our-insights/a-new-way-to-measure-word-of-mouth-marketing. Chen Zoey , Berger Jonah ( 2013 ), “When, Why, and How Controversy Causes Conversation,” Journal of Consumer Research , 40 3 , 580 – 93 . Google Scholar Crossref Search ADS Chen Zoey , Lurie Nicholas H. ( 2013 ), “Temporal Contiguity and Negativity Bias in the Impact of Online Word of Mouth,” Journal of Marketing Research , 50 4 , 463 – 76 . Google Scholar Crossref Search ADS Chevalier Judith A. , Mayzlin Dina ( 2006 ), “The Effect of Word of Mouth on Sales: Online Book Reviews,” Journal of Marketing Research , 43 3 , 345 – 54 . Google Scholar Crossref Search ADS Clark Jason K. , Thiem Kelsey C. , Hoover Ann E. , Habashi Meara M. ( 2017 ), “Gender Stereotypes and Intellectual Performance: Stigma Consciousness as a Buffer against Stereotype Validation,” Journal of Experimental Social Psychology , 68 , 185 – 91 . Google Scholar Crossref Search ADS Cowley Elizabeth ( 2014 ), “Consumers Telling Consumption Stories: Word-of-Mouth and Retrospective Evaluations,” Journal of Business Research , 67 7 , 1522 – 9 . Google Scholar Crossref Search ADS Ditto Peter H. , Lopez David F. ( 1992 ), “Motivated Skepticism: Use of Differential Decision Criteria for Preferred and Nonpreferred Conclusions,” Journal of Personality and Social Psychology , 63 4 , 568 – 84 . Google Scholar Crossref Search ADS Ditto Peter H. , Scepansky James A. , Munro Geoffrey D. , Apanovitch Anne Marie , Lockhart Lisa K. ( 1998 ), “Motivated Sensitivity to Preference-Inconsistent Information,” Journal of Personality and Social Psychology , 75 1 , 53 – 69 . Google Scholar Crossref Search ADS Dubois David , Rucker Derek D. , Galinsky Adam D. ( 2012 ), “Super Size Me: Product Size as a Signal of Status,” Journal of Consumer Research , 38 6 , 1047 – 62 . Google Scholar Crossref Search ADS Dubois David , Bonezzi Andrea , De Angelis Matteo ( 2016 ), “ Sharing with Friends Versus Strangers: How Interpersonal Closeness Influences Word-of-Mouth Valence” , Journal of Marketing Research , 53 5 , 712 – 727 . East Robert , Hammond Kathy , Lomax Wendy ( 2008 ), “Measuring the Impact of Positive and Negative Word of Mouth on Brand Purchase Probability,” International Journal of Research in Marketing , 25 3 , 215 – 24 . Google Scholar Crossref Search ADS Edwards Kari , Smith Edward E. ( 1996 ), “A Disconfirmation Bias in the Evaluation of Arguments,” Journal of Personality and Social Psychology , 71 1 , 5 – 24 . Google Scholar Crossref Search ADS Enda Jodi , Mitchell Amy ( 2013 ), “Friends and Family—Important Drivers of News,” Pew Research Center, http://www.journalism.org/2013/03/17/friends-and-family-important-drivers-of-news/. Epley Nicholas , Keysar Boaz , Van Boven Leaf , Gilovich Thomas ( 2004 ), “Perspective Taking as Egocentric Anchoring and Adjustment,” Journal of Personality and Social Psychology , 87 3 , 327 – 39 . Google Scholar Crossref Search ADS PubMed Folkes Valerie S. ( 1984 ), “Consumer Reactions to Product Failure: An Attributional Approach,” Journal of Consumer Research , 10 4 , 398 – 409 . Google Scholar Crossref Search ADS Folkes Valerie S. ( 1988 ), “Recent Attribution Research in Consumer Behavior: A Review and New Directions,” Journal of Consumer Research , 14 4 , 548 – 65 . Google Scholar Crossref Search ADS Folkes Valerie S. , Kotsos Barbara ( 1986 ), “Buyers’ and Sellers’ Explanations for Product Failure: Who Done It?” Journal of Marketing , 50 2 , 74 – 80 . Google Scholar Crossref Search ADS Folkes Valerie S. , Patrick Vanessa M. ( 2003 ), “The Positivity Effect in Perceptions of Services: Seen One, Seen Them All?” Journal of Consumer Research , 30 1 , 125 – 37 . Crossref Search ADS Fornell Claes , Johnson Michael D. , Anderson Eugene W. , Cha Jaesung , Bryant Barbara Everitt ( 1996 ), “The American Customer Satisfaction Index: Nature, Purpose, and Findings,” Journal of Marketing , 60 4 , 7 – 18 . Google Scholar Crossref Search ADS Goncalo Jack A. , Vincent Lynne C. , Krause Verena ( 2015 ), “The Liberating Consequences of Creative Work: How a Creative Outlet Lifts the Physical Burden of Secrecy,” Journal of Experimental Social Psychology , 59 , 32 – 9 . Google Scholar Crossref Search ADS Hayes Andrew F. ( 2013 ), Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach , New York : Guilford . He Stephen X. , Bond Samuel D. ( 2015 ), “Why Is the Crowd Divided? Attribution for Dispersion in Online Word of Mouth,” Journal of Consumer Research , 41 6 , 1509 – 27 . Google Scholar Crossref Search ADS Herr Paul M. , Kardes Frank R. , Kim John ( 1991 ), “Effects of Word-of-Mouth and Product-Attribute Information on Persuasion: An Accessibility-Diagnosticity Perspective,” Journal of Consumer Research , 17 4 , 454 – 62 . Google Scholar Crossref Search ADS Johnson Michael D. , Anderson Eugene W. , Fornell Claes ( 1995 ), “Rational and Adaptive Performance Expectations in a Customer Satisfaction Framework,” Journal of Consumer Research , 21 4 , 695 – 708 . Google Scholar Crossref Search ADS Kahneman Daniel , Tversky Amos ( 1979 ), “Prospect Theory: An Analysis of Decision under Risk,” Econometrica: Journal of the Econometric Society , 47 2 , 263 – 91 . Google Scholar Crossref Search ADS Keller Fay Group ( 2014 ), “2014 Landmark Study: Return on Word of Mouth,” http://www.womma.org/ReturnOnWOM. Kraus Michael W. , Chen Serena ( 2009 ), “Striving to Be Known by Significant Others: Automatic Activation of Self-Verification Goals in Relationship Contexts,” Journal of Personality and Social Psychology , 97 1 , 58 – 73 . Google Scholar Crossref Search ADS PubMed Kristofferson Kirk , White Katherine , Peloza John ( 2014 ), “The Nature of Slacktivism: How the Social Observability of an Initial Act of Token Support Affects Subsequent Prosocial Action,” Journal of Consumer Research , 40 6 , 1149 – 66 . Google Scholar Crossref Search ADS Kwang Tracy , Swann William B. ( 2010 ), “Do People Embrace Praise Even When They Feel Unworthy? A Review of Critical Tests of Self-Enhancement versus Self-Verification,” Personality and Social Psychology Review , 14 3 , 263 – 80 . Google Scholar Crossref Search ADS PubMed Lynch John G. , Marmorstein Howard , Weigold Michael F. ( 1988 ), “Choices from Sets Including Remembered Brands: Use of Recalled Attributes and Prior Overall Evaluations,” Journal of Consumer Research , 15 2 , 169 – 84 . Google Scholar Crossref Search ADS Mattila Anna S. ( 2001 ), “Emotional Bonding and Restaurant Loyalty,” Cornell Hotel and Restaurant Administration Quarterly , 42 6 , 73 – 9 . Google Scholar Crossref Search ADS Moore Sarah G. ( 2012 ), “Some Things Are Better Left Unsaid: How Word of Mouth Influences the Storyteller,” Journal of Consumer Research , 38 6 , 1140 – 54 . Google Scholar Crossref Search ADS Nickerson Raymond S. ( 1999 ), “How We Know—and Sometimes Misjudge—What Others Know: Imputing One’s Own Knowledge to Others,” Psychological Bulletin , 125 6 , 737 – 59 . Google Scholar Crossref Search ADS Olejnik Stephen , Algina James ( 2003 ), “Generalized Eta and Omega Squared Statistics: Measures of Effect Size for Some Common Research Designs,” Psychological Methods , 8 4 , 434 – 47 . Google Scholar Crossref Search ADS PubMed Parasuraman Anantharanthan , Zeithaml Valarie A. , Berry Leonard L. ( 1985 ), “A Conceptual Model of Service Quality and Its Implications for Future Research,” Journal of Marketing , 49 4 , 41 – 50 . Google Scholar Crossref Search ADS Peeters Guido , Czapinski Janusz ( 1990 ), “Positive-Negative Asymmetry in Evaluations: The Distinction between Affective and Informational Negativity Effects,” European Review of Social Psychology , 1 1 , 33 – 60 . Google Scholar Crossref Search ADS Ramanathan Suresh , Williams Patti ( 2007 ), “Immediate and Delayed Emotional Consequences of Indulgence: The Moderating Influence of Personality Type on Mixed Emotions,” Journal of Consumer Research , 34 2 , 212 – 23 . Google Scholar Crossref Search ADS Ratner Rebecca K. , Kahn Barbara E. ( 2002 ), “The Impact of Private versus Public Consumption on Variety-Seeking Behavior,” Journal of Consumer Research , 29 2 , 246 – 57 . Google Scholar Crossref Search ADS Schaefers Tobias , Schamari Julia ( 2016 ), “Service Recovery via Social Media: The Social Influence Effects of Virtual Presence,” Journal of Service Research , 19 2 , 192 – 208 . Google Scholar Crossref Search ADS Schellekens Gaby A. C. , Verlegh Peeter W. J. , Smidts Ale ( 2010 ), “Language Abstraction in Word of Mouth,” Journal of Consumer Research , 37 2 , 207 – 23 . Google Scholar Crossref Search ADS Schlosser Ann E. ( 2005 ), “Posting versus Lurking: Communicating in a Multiple Audience Context,” Journal of Consumer Research , 32 2 , 260 – 5 . Google Scholar Crossref Search ADS Sinclair Stacey , Lowery Brian S. , Hardin Curtis D. , Colangelo Anna ( 2005 ), “Social Tuning of Automatic Racial Attitudes: The Role of Affiliative Motivation,” Journal of Personality and Social Psychology , 89 4 , 583 – 92 . Google Scholar Crossref Search ADS PubMed Spiller Stephen A. , Belogolova Lena ( 2016 ), “On Consumer Beliefs about Quality and Taste,” Journal of Consumer Research , 43 6 , 970 – 91 . Tice Dianne M. , Butler Jennifer L. , Muraven Mark B. , Stillwell Arlene M. ( 1995 ), “When Modesty Prevails: Differential Favorability of Self-Presentation to Friends and Strangers,” Journal of Personality and Social Psychology , 69 6 , 1120 – 38 . Google Scholar Crossref Search ADS Weber Reneé , Crocker Jennifer ( 1983 ), “Cognitive Processes in the Revision of Stereotypic Beliefs,” Journal of Personality and Social Psychology , 45 5 , 961 – 77 . Google Scholar Crossref Search ADS Zhang Yinlong , Feick Lawrence , Mittal Vikas ( 2014 ), “How Males and Females Differ in Their Likelihood of Transmitting Negative Word of Mouth,” Journal of Consumer Research , 40 6 , 1097 – 108 . Google Scholar Crossref Search ADS Zhao Xinshu , Lynch John G. Jr. , Chen Qimei ( 2010 ), “Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis,” Journal of Consumer Research , 37 2 , 197 – 206 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Journal of Consumer ResearchOxford University Press

Published: Dec 1, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off