The Influence of Incidental Similarity on Observers’ Causal Attributions and Reactions to a Service Failure

The Influence of Incidental Similarity on Observers’ Causal Attributions and Reactions to a... Abstract Observers’ reactions to a service failure and their attributions of responsibility for its occurrence can depend on fortuitous characteristics of the protagonists that happen to draw their attention. Four field and laboratory experiments show that when observers have an incidental similarity to one of the persons involved in a service failure (the customer or the service provider), their attention is drawn to this protagonist, often leading them to construe the situation from this person’s perspective and consequently to blame the protagonist less for the negative event they observe. However, when an incidentally similar protagonist is rude or has an undesirable personal characteristic (i.e., obesity), observers’ greater attention to that person increases their attributions of responsibility to him or her rather than decreasing it. These opposing effects of incidental similarity on attributions influence not only observers’ evaluations of the persons involved in the situation they observe, but also their willingness to patronize the establishment. These effects occur both when observers actually witness a conflict offline and when they consider it online on the basis of reviews. incidental similarity, attribution, service failures, perspective taking, observer reactions Consumer complaints about a service failure are commonplace. For example, a restaurant visitor might protest that his steak is too well done, or a customer might complain that a product she bought doesn’t perform as it should. People who observe such interactions might attribute responsibility for the incident to either the customer or the service provider, and this attribution could influence not only their evaluation of the protagonists but also their inclination to patronize the provider’s establishment. Consumers’ offline purchase decisions are based on online reviews (Zhu and Zhang 2010), and negative reviews influence 80% of these decisions (Dimensional Research 2013). Thus, customer complaints on online social platforms, which increased 800% between 2014 and 2015 (Causon 2015), capture many viewers’ attention. When people observe or read about someone’s complaint to a service provider, what factors influence their attributions of responsibility, and what are the effects of these attributions? Our research provides some answers to these questions. We found that when observers are not personally involved in a service failure, their inclinations to blame the provider or the customer can depend on quite fortuitous events that have nothing objectively to do with the service failure that occurred. For example, observers’ incidental similarity to one of the protagonists (e.g., a common birthday or first name) can increase the attention they pay to this protagonist, and this attention, in turn, can influence their disposition to blame the protagonist for the failure’s occurrence. Two different processes can underlie the influence of this increased attention, however, and these processes have opposite effects. Our conceptualization of these processes is summarized in figure 1. In the conditions we investigated, participants overhear a customer and a service provider discussing a problem that the customer has encountered. We assumed that if observers are unacquainted with the protagonists, they are unlikely to have much interest in the discussion and consequently do not think very deeply about the conditions that gave rise to the service failure being discussed. If they happen to notice an incidental similarity between themselves and one of the protagonists (protagonist A), however, their attention is drawn to him or her and to the incident in which the person is involved. Then, if neither protagonist has a salient characteristic that could explain the incident, observers construe it from the perspective of protagonist A. Consequently, for reasons suggested both by Jones and Nisbett (1971) and by defensive attribution theory (Shaver 1985), they are less likely to blame protagonist A (see figure 1A), and therefore, by default, are more likely to attribute responsibility for the incident to protagonist B. This attribution, in turn, will affect observers’ willingness to patronize the establishment in which the incident occurred. FIGURE 1 View largeDownload slide Conceptual model FIGURE 1 View largeDownload slide Conceptual model There are contingencies in this prediction, however. For example, if observers are unable to devote the cognitive resources necessary to construe the implications of the situation they observe, the effect of incidental similarity on their perspective and its consequent influence on responsibility attributions may not occur. Moreover, suppose protagonist A has a distinctive feature that observers consider to be undesirable (figure 1B). People are unlikely to empathize with someone they dislike or evaluate unfavorably, and their negative reactions can lead them to blame the person for negative events in which he or she is involved (Regan, Straus, and Fazio 1974). Therefore, observers are unlikely to take the perspective of a protagonist whose behavior or personal characteristics elicit negative affective reactions, so they may blame protagonist A for the service failure regardless of other considerations. Moreover, if their incidental similarity to protagonist A increases their attention to him or her, it could increase their cognitive elaboration of the service failure and this person’s involvement in it, and this elaboration might lead them to make more extreme estimates of his or her responsibility for the incident than they otherwise would (Petty, Haugtvedt, and Smith 1995; Tesser 1978). Our research provides the first evidence that consumers’ incidental similarity to one of the protagonists in a situation in which they are not personally involved can influence their reactions to the situation and attributions of responsibility for its occurrence. Moreover, it is the first to find that totally irrelevant features of a service failure situation can have directionally different effects on observers’ reactions to the situation and their willingness to patronize the establishment in which the failure occurred. CONCEPTUAL BACKGROUND To convey the processes that underlie the effects we report, we first discuss the manner in which incidental similarity to a protagonist can influence the observer’s attention to him or her. We then consider the consequences of this attention for observers’ attributions of responsibility for a service failure and the reactions to the protagonist that result. Incidental Similarity, Attention, and Causal Attribution Consumers’ similarity to a salesperson can influence the quality of their interaction with this person (Crosby, Evans, and Cowles 1990). This influence could occur for several reasons. For example, because people typically regard their own attributes favorably, they are inclined to like others who also possess these attributes (Byrne 1971; for other evidence that similarity increases attraction, see Berscheid and Reis 1998; Folkes 1982). Moreover, information that one’s attributes are shared by another can increase one’s self-esteem (Stalling 1970). People can often infer their similarity to one another from their behavior or from demographic factors (e.g., gender, race, or age) that have stereotype-based implications for their personality, attitudes, and values. However, even an incidental similarity to someone (e.g., a common first name, hometown, or birthday) can have a significant impact on their reactions to this person (Burger et al. 2004; Guéguen, Pichot, and Le Dreff 2005; Jiang et al. 2010; Martin and Guéguen 2013). For example, if individuals’ similarity to one another distinguishes them from others in the immediate situation, it can induce a sense of social connectedness. (For example, two Chicagoans are likely to feel more connected to each other if they meet in Paris than if they meet on a Chicago subway.) Thus, a person’s incidental similarity to a stranger can increase the likelihood of complying with the stranger’s request (Burger et al. 2004; Guéguen et al. 2005), of mimicking his or her nonverbal behaviors (Guéguen and Martin 2009), and of cooperating with the person (Miller, Downs, and Prentice 1998). In the consumer domain, a customer’s incidental similarity to a service provider can increase the customer’s liking for both the provider and the service received (Jiang et al. 2010). The aforementioned effects, however, occur when individuals either actually interact with the person to whom they are incidentally similar or imagine doing so. In contrast, the observers in our research were passive witnesses to an interchange between two persons with whom they were unacquainted and did not expect to interact. Consequently, they had little a priori interest in either party. Therefore, social connectedness is unlikely to account for the effect of incidental similarity in the situations we investigated. Rather, this effect seemed more likely to result from its influence on the relative attention that observers pay to the protagonists. People who witness an interchange between persons they do not know may often not be motivated to construe its implications. If they notice that a feature of one of the protagonists is similar to their own, however, their attention may be drawn to this person and they may be stimulated to think more carefully about the person’s role in the situation at hand. Incidental self-relevant information about a person (the person’s name, hometown, etc.) can elicit selective attention to this person (Bargh 1982; Moray 1959; Tacikowski and Nowicka 2010). In the situation we investigated, therefore, we expected that observers’ attention to a protagonist to whom they are incidentally similar would affect their interpretation of the service failure and their attributions of responsibility for its occurrence. The determinants of these attributions could be both cognitive and motivational. Cognitive Factors: Effects of Perspective Events are often attributed to features of the situation that are most likely to capture attention (Taylor and Fiske 1978). Thus, observers are likely to attribute an actor’s behavior to characteristics of the actor. In contrast, because actors’ attention is focused outward, they are more likely to attribute their behavior to characteristics of the situation or to other persons with whom they interact. These actor-observer differences are well established (Jones and Nisbett 1971). There are some constraints on their occurrence (Malle 2006), including the normative distinctiveness of the actor’s behavior (Hansen and Stonner 1978) and observers’ prior knowledge of the actor (Taylor, Peplau, and Sears 2006). In the conditions we considered, however, observers were unacquainted with the protagonists and their complaint behavior in the situation was not out of the ordinary. Similar considerations suggest that actors’ and observers’ attributional dispositions can be influenced by the perspective from which the actors’ behavior is viewed or imagined. Thus, showing actors a videotape of themselves interacting with another can increase their disposition to attribute their behavior to characteristics of themselves (Storms 1973). By the same token, observers who imagine an interaction from the perspective of one of the protagonists typically attribute responsibility for the events that occur to the other protagonist (Regan and Totten 1975). In the present context, this suggests that if observers imagine a service failure from the perspective of one of the protagonists, they are likely to attribute responsibility for the failure to the other protagonist. Motivational Factors Actor-observer differences in attribution can also depend on the favorableness of the behavior observed (Malle 2006). Individuals often exhibit a self-serving bias (Kunda 1987; Shaver 1985). That is, they typically take responsibility for positive outcomes of their behavior, but accept responsibility for negative outcomes only when no other plausible explanation exists (Arkin, Gleason, and Johnston 1976; Oishi, Wyer, and Colcombe 2000). This tendency, like that postulated by Jones and Nisbett (1971), suggests that observers who construe the implications of a service failure from the perspective of one of the protagonists are unlikely to attribute responsibility for the event to that protagonist; rather, they are inclined to blame the other protagonist. Effects of Protagonist Characteristics The preceding analyses, however, assume that the protagonists in the situation being observed have no distinguishing features that might be used as a basis for attributions. If one protagonist has an attribute or behavioral disposition that observers consider to be undesirable, an additional consideration comes into play. People not only dislike someone whose behavior or attributes elicit negative affective reactions, but also attribute other negative characteristics to the individual. Thus, people are more inclined to blame a negative event on someone they dislike than on someone they like (Regan et al. 1974; see also Alicke 2000). This suggests that if one of the parties involved in a service failure situation has an undesirable characteristic that elicits negative affective reactions, observers will attribute unfavorable characteristics and behavior to this individual and consequently will blame this person for the service failure that occurred. Moreover, the amount of attention they pay to this person is likely to increase their cognitive elaboration of the person’s involvement in the incident and to increase the extremity of their attribution of responsibility (Petty et al. 1995; Tesser 1978). A Two-Process Model The preceding discussion suggests that although observers’ incidental similarity to one of the protagonists in a service failure situation is likely to increase the attention they pay to that person, this attention could influence their attributions of responsibility in the different ways summarized in figure 1. If the protagonists have no distinguishing characteristics that could plausibly account for the service failure (figure 1A), observers’ relatively greater attention to the incidentally similar protagonist is likely to influence the perspective they take when they construe implications of the situation. That is, they construe the situation from the perspective of the protagonist to whom their attention is drawn, and attribute responsibility for the service failure to factors that are salient to them from this imagined perspective (e.g., the other protagonist). Thus: H1: When the customer and service provider involved in a service failure are strangers and have no distinctive judgment-relevant characteristics, observers’ incidental similarity to one of the protagonists will decrease their likelihood of blaming this protagonist for the failure and will increase their likelihood of blaming the other. There are qualifications on this hypothesis. That is, the effects of incidental similarity implied by hypothesis 1 are mediated by the effect of attention on the perspective that observers take when they construe the implications of the service failure (see figure 1A). If this is so, preventing them from thinking about the service failure from this perspective should eliminate these effects. Furthermore, different considerations arise when one of the protagonists has a characteristic that observers dislike or consider to be undesirable (figure 1B). In this case, observers are likely to attribute other undesirable characteristics to the protagonist as well and, therefore, to blame the service failure on this person. This may occur regardless of their incidental similarity to the protagonist. If they happen to be incidentally similar to the protagonist they consider to be responsible for the service failure, however, their relatively greater attention to this protagonist is likely to increase their cognitive elaboration of the service failure and the protagonist’s responsibility for it, and this increased elaboration is likely to increase the extremity of the observers’ attribution (Petty et al. 1995; Tesser 1978). H2: When observers perceive the behavior or personal characteristics of a protagonist in a service failure situation as undesirable, they will attribute greater responsibility for the failure to this protagonist than to the other, and this tendency will be greater when they are incidentally similar to the protagonist than when they are not. Note that the effects of incidental similarity implied by hypotheses 1 and 2 are both mediated by its impact on the attention paid to the similar protagonist. The effect of this increased attention differs, however, depending on whether the protagonist has undesirable features or not. Specifically, H3: The effect of observers’ incidental similarity on attributions of responsibility for a service failure implied by hypotheses 1 and 2 is mediated by its effect on the attention they pay to the protagonist to whom they are similar. To evaluate the implications of hypothesis 2, we considered two characteristics that seemed likely to influence attributions of a protagonist’s responsibility. One characteristic, rudeness, pertained to the protagonist’s behavior during the service failure situation. This factor was suggested by Jiang et al.’s (2010) finding that when participants have viewed a video of a personal trainer who was rude to other staff members, they report a more negative attitude toward the trainer when they are incidentally similar to him than when they are not. (For further evidence that individuals have greater dislike of persons who engage in undesirable behavior when they are similar to these persons, see Cooper and Jones 1969.) These effects could also occur when a protagonist’s undesirable attribute is irrelevant to the service failure situation and has little if any implications for the protagonist’s responsibility for the incident. To this end, we speculated that observers would be more inclined to blame a protagonist if he or she is obese. This characteristic is obviously irrelevant to the service failure situation itself. However, negative reactions to obese individuals are pervasive (Crandall 1994). They are often considered to be lazy, unhealthy, lacking in self-discipline, dishonest, unsuccessful, and stupid (Crandall 1994; Greenleaf et al. 2004; Puhl and Brownell 2001). Consequently, overweight employees are relatively less likely to get a pay raise and to be promoted (Pingitore et al. 1994), are less likely to receive appropriate treatment from physicians (Price et al. 1987), and are more likely to receive poor academic evaluations (Solovay 2000). In the consumer domain, customers have lower satisfaction and purchase intentions when they interact with an overweight service provider (Cowart and Brady 2014) and tend to avoid making the same choices that obese customers have made (McFerran et al. 2010). We therefore expected that observers would have greater inclination to blame an obese protagonist for a service failure than a protagonist of normal weight. Moreover, this disposition should be greater when the observers are incidentally similar to the protagonist than when they are not. Finally, the differences in attributions implied by the first three hypotheses have obvious implications not only for observers’ liking of the protagonists they observe but also for their evaluations of the service and their likelihood of patronizing the provider in the future. Specifically, H4: The effects of observers’ incidental similarity to the protagonists involved in a service failure on their attributions of responsibility for the failure influences their evaluations of the service and their willingness to patronize the provider in the future. Other Considerations Two other factors that might affect attributions in the conditions we investigated are worth noting. Threat to Self-Esteem If observers of a service failure take the perspective of a protagonist to whom they are incidentally similar, they might find negative attributes of this protagonist to be threatening to their self-esteem and might be motivated to dissociate themselves from this person, leading to effects of the sort implied by hypothesis 2. However, we expected the effects implied by hypothesis 2 to be independent of such threats. That is, observers might feel threatened by the undesirable behavior of a protagonist to whom they feel incidentally similar, but this threat does not necessarily mediate the effects of incidental similarity on attributions of responsibility that we identified. Empathy The disposition to take the perspective of a person who is incidentally similar could also be mediated by empathy. That is, individuals who take the perspective of another person often empathize with that person, vicariously experiencing the feelings that he or she is likely to have in the situation at hand (Batson et al. 1981). However, this is not always true. In competitive situations, for example, players often find it useful to imagine the situation from their opponent’s point of view in developing effective game-playing strategies without thinking about the other’s feelings. In short, empathy for another requires taking the other’s perspective, but taking another’s perspective does not necessarily imply empathy (Davis 1983). In several of the studies to be reported, we used a measure of empathy as a proxy for perspective taking in order to evaluate our assumptions about the effect of incidental similarity on perspective taking. However, we assumed that empathy per se was not itself a central factor in the effects we observed. Overview Four experiments examined the validity of our hypotheses and the processes that underlie them. Table 1 summarizes the causal relationships investigated in each experiment. Experiment 1, a field study, found that when observers of a service failure had the same surname as the waitress, they judged her to be less responsible for the failure and evaluated the service more favorably. When they had the same last name as the customer, however, the reverse was true. Experiment 2 corroborated these effects in a different situational context and confirmed that these effects are eliminated when participants are distracted from expending the cognitive resources necessary to think carefully about the service failure. Experiments 3 and 4 confirmed the mediating effect of attention on the impact of incidental similarity and showed that when one of the protagonists in the service failure situation had a negative attribute, incidental similarity of this protagonist increased attributions of blame to this protagonist rather than decreasing it. As table 1 indicates, no single study examined all of the direct and mediating influences of incidental similarity on attributions of responsibility and patronage intentions that we hypothesized. In combination, however, they provided a fairly complete picture of the phenomena we investigated. TABLE 1 SUMMARY OF CAUSAL RELATIONS EVALUATED IN EXPERIMENTS 1–4 Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions TABLE 1 SUMMARY OF CAUSAL RELATIONS EVALUATED IN EXPERIMENTS 1–4 Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions Pretests of the different scenarios we used in experiments 2 through 4 confirmed that participants assigned similar levels of responsibility for the failure to both the service provider and the customer. In addition, a supplementary study ruled out the possibility that the effects of incidental similarity in the conditions we investigated were mediated by feelings of social connectedness. These studies are described in the web appendix. EXPERIMENT 1 This experiment evaluated hypothesis 1 in a field setting. Restaurant customers observed a service failure involving another customer and a waitress, and their incidental similarity to one of the protagonists was unobtrusively manipulated. We expected that participants would blame the waitress less for the failure when their name was similar to that of the waitress than when it was similar to that of the customer. Method Participants were asked to apply for a “mystery shopper” experiment three weeks before the study. Their last names were identified through their applications. Persons whose last names were particularly uncommon were not recruited, leaving 70 participants in the actual experiment. These participants (41.7% males between 20 and 35 years of age) were paid HK$100 (approximately US$13) to take part in a mystery shopper study. They were randomly assigned to cells of a 3 (incidental similarity: provider similarity vs. customer similarity vs. control) between-subjects design. The experiment was conducted at a small local restaurant in Hong Kong. The restaurant was newly opened and the experiment (with the owner’s permission) took place on 15 weekdays between 2:00 p.m. and 6:30 p.m., when the restaurant traffic was minimal. The size of the restaurant was small (around 900 square feet, with 15 tables) and customers could easily see and overhear the interactions between the waiter and the other customer at the next table. Each participant, run individually, arrived at the restaurant in a particular time slot. No participant had been to the restaurant before. They were invited to be mystery shoppers and to evaluate the restaurant’s afternoon tea menu and services. Each participant was given a coupon of HK$100 to spend on food and beverages. When participants arrived, they were asked to sit at a table in a corner of the restaurant to minimize the disturbance from other customers and waiters. Two female confederates (age = 22 years) were recruited to be a waitress and a target customer. In provider similarity conditions, the waitress wore a name tag with a last name that either matched or did not match the participant’s. Furthermore, the waitress mentioned her name to the participants when she greeted them and directed them to a table. In customer similarity conditions, a “reserved” sign with a last name that either matched or did not match the participant’s was put on the table next to the participant’s. The female confederate pretending to be a customer entered the restaurant after the participant was seated. To ensure that each participant noticed the name of the confederate, the waitress referred to “Ms. ____” when she greeted the customer and directed her to the table next to the participant’s. In control conditions, neither the last name on the waitress’s name tag nor the reserved table stand matched the name of the participant. Participants received their order within 10 minutes and the confederate received hers 5 minutes later. Upon receiving it, however, she complained to the waitress that her food was mixed with onions. (Although this might not be universally considered to be a “service failure,” we assumed that observers in the situation we constructed would interpret the customer’s complaint as either an unjustified response that reflected the customer’s idiosyncratic taste or a justified criticism of the way the restaurant prepared this particular dish.) The waitress volunteered to replace the dish and said it would take approximately 15 minutes. The confederate looked unhappy and told the waitress to cancel the order, as she had to leave. Upon leaving the restaurant, participants completed an online questionnaire on their mobile phone. First, they evaluated the overall quality of the service along three scales from 1 (bad/undesirable/unfavorable) to 10 (good/desirable/favorable). Responses to these items were averaged (α =.93). They then evaluated the quality of the food and the physical environment along similar scales (α =.92 and .95, respectively). Finally, participants were asked whether they had personally experienced a service failure at the restaurant (all participants responded negatively) and whether they had observed any service failure that involved another customer. Two participants did not notice that a service failure had occurred and were excluded from the subsequent data analysis. The remaining participants indicated whether the waitress was responsible by responding to three items (“You blame the waitress for the problem you observed,” “You think the waitress was responsible for the problem you observed,” and “You think that the problem described was not in the control of the waitress” (reverse-scored) along a scale from 1 (strongly disagree) to 10 (strongly agree) (α =. 92). Results and Discussion We expected that incidental similarity would affect participants’ overall service evaluations through its mediating impact on attributions of responsibility. These expectations were confirmed. Participants’ evaluations of the restaurant in terms of food, physical environment, and service are summarized in table 2. The effect of incidental similarity on overall service evaluations was significant (F(2, 65) = 18.37, p < .001, ηp2 = .36). Participants rated the service more favorably when they were incidentally similar to the service provider than they did in control conditions (7.17 vs. 6.06, respectively; F(1, 65) = 5.17, p < .01, d = .98) but rated it less favorably when they were incidentally similar to the customer than in control conditions (5.10 vs. 6.06, respectively; F(1, 65) = 3.84, p < .05, d = .86). In contrast, participants’ ratings of food quality or the physical environment did not differ across three conditions (t < 1). TABLE 2 THE IMPACT OF INCIDENTAL SIMILARITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 1 Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) NOTE.—Standard deviations are in parentheses. Means in each row with unlike subscripts significantly differ at p < .05. TABLE 2 THE IMPACT OF INCIDENTAL SIMILARITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 1 Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) NOTE.—Standard deviations are in parentheses. Means in each row with unlike subscripts significantly differ at p < .05. The effect of incidental similarity on attributions of responsibility for the service failure was also significant (F(2, 65) = 16.68, p < .001, ηp2 = .34) and paralleled its effect on service evaluations. Participants were less likely to blame the waitress when they were incidentally similar to her than they were in control conditions (5.38 vs. 6.55, respectively; F(1, 65) = 5.44, p < .01, d = .98) but were marginally more likely to blame the waitress when they were similar to the customer than in control conditions (7.39 vs. 6.55, respectively; F(1, 65) = 2.85, p = .065, d = .74). Responsibility attributions significantly mediated the effects of incidental similarity on service evaluations. Based on a bootstrapping analysis (model 4, Hayes 2018) with 5,000 samples, the 95% CI ranged from –.7683 to –.0512. In summary, experiment 1 provided evidence that observers’ incidental similarity to the protagonists in a service failure situation affected both their attributions of responsibility for the failure and evaluations of the overall quality of the service in general. When observers were incidentally similar to the service provider, they attributed less responsibility to the provider and evaluated the service quality more favorably. When they were incidentally similar to the customer, however, they attributed more responsibility to the provider and evaluated the service less favorably. EXPERIMENT 2 Participants in experiment 1 observed a failure that occurred in the same consumption situation in which they were present. Experiment 2 examined whether a similar effect would occur when observers were not involved in the situation at all. We used a trip advisor website as a context and manipulated the incidental similarity of observers’ last name to that of either a hotel manager or a customer. We also assessed attributions of blame to the customer as well as to the service provider and confirmed these attributions on the basis of thought listing data. We also showed that observers did not think much about the service failure situation when an incidental similarity was not present. In addition, to ensure that the effect of an incidental similarity was not mediated by its impact on the positive affect that participants experienced, participants’ moods were measured. Furthermore, we obtained support for the assumption that the effect of incidental similarity was mediated by the attention paid to the similar protagonist. If this assumption is valid, situational factors that distract observers from differentially attending to the protagonists should decrease their ability to engage in this processing and should consequently decrease or eliminate differences in their attributions of responsibility for the failure that otherwise occurs. This was in fact the case. Method To collect participants’ last name information, we had 288 Hong Kong participants complete an unrelated survey on the university network two weeks before the main study. A section of the survey requested demographic information, including their last name. After eliminating participants whose last names were unlikely to be shared, we selected 192 participants (42.7% males) to participate in a 15 minute study for HK$20 (approximately US$3). They were randomly assigned to cells of a 3 (incidental similarity: provider similarity vs. customer similarity vs. control) × 2 (cognitive load: low vs. high) between-subjects design. Participants performed two ostensibly unrelated tasks. First, on the pretense that we were interested in people’s ability to perform two tasks at once, we asked them to memorize either a 10-digit (high load) or 2-digit (low load) number that they would be asked to recall after they performed a second task in the experiment (Shiv and Fedorikhin 1999). In this task, they viewed a trip advisor online review of a hotel service (see the web appendix). The review contained a dialogue that described a customer’s complaint about the service and the hotel manager’s response. The customer complained that he had made a reservation for a nonsmoking sea view room on the hotel’s website two weeks earlier but that when he arrived at the hotel, he found that the room was unavailable. The hotel manager responded that he had checked the computer system and did not see the customer’s reservation and that only nonsmoking rooms without a sea view were available, as it was during the peak season. The dialogue conveyed that participants had the same last name as the manager in provider similarity conditions, and the same last name as the customer in customer similarity conditions. In control conditions, all names differed. Participants reported the likelihood that they would visit the hotel along three scales from 1 (very unlikely/inclined not to/definitely will not) to 10 (very likely/inclined to/definitely will). They also indicated their inclination to blame both the manager and the customer (i.e., “You blame the manager/customer for the problem you observed,” “You think the manager/customer was responsible for the problem you observed,” and “You think that the problem described was not in the control of the manager/customer” (reverse-scored) along a scale from 1 (strongly disagree) to 10 (strongly agree). Responses to each set of items were averaged (α > .90). Participants reported their mood (α = .91) along three scales from 1 (sad/bad/depressed) to 10 (happy/good/cheerful). After that, they listed the thoughts that went through their minds while they were reading the hotel scenario. Finally, to access the validity of cognitive load manipulation, we had them respond to two items (r > .68): “I found it challenging to read the scenario while trying to remember the number” and “Remembering the number was easy” (reverse-scored) along a scale from 1 (strongly disagree) to 10 (strongly agree). Results The effects of incidental similarity on judgments are summarized in table 3. We expected that incidental similarity would influence inclinations to patronize the establishment under low-cognitive-load conditions and that this influence would be mediated by attributions of responsibility for the incident’s occurrence. Under high-load conditions, however, we expected this influence to be minimal. TABLE 3 THE IMPACT OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 2 Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. TABLE 3 THE IMPACT OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 2 Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. Manipulation Check As expected, participants perceived the situation as more difficult to process under high load than under low load (7.02 vs. 2.78, respectively; F(1, 186) = 348.50, p < .001, ηp2 = .65). Patronage Intentions The main effect of incidental similarity and its interaction with cognitive load were significant (F(2, 186) = 12.34, p < .001, ηp2 = .12 and F(2, 186) = 5.86, p < .01, ηp2 = .06, respectively). Between-cell comparisons shown in table 3 indicate that in low-load conditions, incidental similarity to the provider increased participants’ inclination to visit the hotel (from 5.13 to 6.39; F(1, 186) = 27.95, p < .001, ηp2 = .13), but similarity to the customer decreased it (from 5.13 to 4.41; F(1, 186) = 9.64, p < .01, ηp2 = .05). In the high-load conditions, however, no differences were found between provider similarity conditions (M = 5.48) and either customer similarity conditions (M = 5.11) or control conditions (M = 5.25). Responsibility Attributions Inclinations to blame the service provider and inclinations to blame the customer were analyzed separately. Each analysis yielded a main effect of incidental similarity (F(2, 186) = 15.18, p < .001, ηp2 = .14 and F(2, 186) = 16.28, p < .001, ηp2 = .15, respectively) and an interaction of similarity and cognitive load (F(2, 186) = 8.23, p < .001, ηp2 = .08 and F(2, 186) = 8.85, p < .001, ηp2 = .09, respectively). Between-cell comparisons shown in table 3 indicate that in low-load conditions, incidental similarity to the provider decreased participants’ disposition to blame the provider relative to control conditions (from 5.49 to 4.65; F(1, 186) = 16.18, p < .001, ηp2 = .08) and increased their disposition to blame the customer (from 5.27 to 6.42; F(1, 186) = 25.50, p < .001, ηp2 = .12). In contrast, incidental similarity to the customer increased attributions of responsibility to the provider (from 5.49 to 6.68; F(1, 186) = 31.79, p < .001, ηp2 = .15) and decreased attributions of responsibility to the customer (from 5.27 to 4.14; F(1, 186) = 24.64, p < .001, ηp2 = .12). In contrast, attributions of blame in high-load conditions did not significantly vary over the three similarity conditions. Thought Listings The different perspectives that observers took in incidental similarity conditions, and their consequent attributions of responsibility for the service failure, are reflected in their thought listings. Two coders who were blind to experimental conditions coded each participant’s thoughts in terms of whether they predominantly conveyed the perspective of the provider (e.g., “The manager has no control over the technical problem of the booking system. He has tried his best to satisfy the customer”), the customer (e.g., “I can feel how unhappy the customer is. It is very clear that the manager should send a confirmation”), both (e.g., “Both of them have responsibilities . . .”) or neither. Interjudge agreement was 91.1% and differences were resolved through discussion. The number and proportions of participants whose thought reflected each perspective are summarized in the top section of table 4. A log linear analysis under low-load conditions yielded a significant interaction of incidental similarity and thought type χ2(6) = 77.56, p < .001. Under low-load conditions, 27 of 32 participants in provider-similarity conditions (84%) took the provider’s perspective (χ2(3)= 60.25, p < .001) and 24 of 32 participants in customer-similarity conditions (75%) took the customer’s perspective (χ2(3) =42.25, p < .001). In control conditions, these proportions were low and did not appreciably differ (χ2(3) =3.75, p >.10). In contrast, a similar analysis under high-load conditions yielded only a main effect of thought type χ2(5) = 43.42, p < .001; the majority of participants’ thoughts in these conditions did not reflect either perspective. TABLE 4 PROPORTION OF PERSPECTIVE-RELATED THOUGHTS AND ATTRIBUTION-RELATED THOUGHTS AS A FUNCTION OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD—EXPERIMENT 2 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 NOTE.—Numbers in parentheses indicate the percentage of the thoughts. TABLE 4 PROPORTION OF PERSPECTIVE-RELATED THOUGHTS AND ATTRIBUTION-RELATED THOUGHTS AS A FUNCTION OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD—EXPERIMENT 2 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 NOTE.—Numbers in parentheses indicate the percentage of the thoughts. Observers who took one protagonist’s perspective were expected to attribute responsibility for the service failure to the other protagonist. To confirm this assumption, judges recoded participants’ thoughts in terms of whether they reflected an attribution of responsibility to the provider (e.g., “It is very clear that the manager should send a confirmation. This is his responsibility”), the customer (e.g., “I think the customer should call the manager to reconfirm the booking”), both (e.g., “Both of them have responsibility for the incident”), the situation (e.g., “No one should be blamed. This is the problem of the system”), or neither. Interjudge agreement was 93.2% and differences were resolved through discussion. Results are summarized in the second half of table 4. A log linear analysis under low-load conditions yielded a significant interaction of incidental similarity and thought type χ2(8) = 85.42, p < .001. When participants were not under cognitive load, 25 of 32 participants in provider-similarity conditions (78%; χ2(4) = 67.69, p < .001) attributed responsibility to the customer, whereas 24 of 32 participants in customer-similarity conditions (75%; χ2(4) = 61.13, p < .001) attributed responsibility to the provider. Responsibility attributions in control conditions did not differ. Moreover, a similar analysis under high-load conditions yielded only a main effect of thought type χ2(6) = 61.33, p < .001. Most participants in these conditions did not attribute beliefs to either protagonist. Mediation Bootstrapping analyses (model 8, Hayes 2018) were consistent with the assumption that under low-load conditions, the effect of incidental similarity on attributions of blame to the provider was mediated by perspective-related thoughts about the provider (based on 5,000 samples, 95% CI ranged from –3.2671 to –.5665) and that its effect on attributions of blame to the customer was mediated by perspective-related thoughts about the customer (95% CI ranged from –1.5700 to –.3929). Under high-load conditions, however, the corresponding effects of perspective-related thoughts were not significant; the 95% CI ranged from –.3134 to .7379 in the first case and from –.1255 to .3315 in the second. Mood Participants reported being happier when they were either similar to the provider (M = 6.16) or similar to the customer (M = 6.33) than they were in control conditions (M = 5.76); F(2, 186) = 3.58, p < .05, ηp2 = .04. However, mood was not a significant mediator of the effect of similarity on responsibility attributions (95% CI, from –.0261 to .1095). Discussion Experiment 2 provided evidence that attention mediated the effect of incidental similarity on observers’ attributions of responsibility. Observers under low-load conditions were less inclined to blame a service provider for a service failure when they were incidentally similar to him, but were more inclined to do so when they were incidentally similar to the customer. In this case, their attributions of responsibility mediated their willingness to patronize the hotel. The effect of incidental similarity was not evident, however, when participants were under cognitive load and were less able to think carefully about the situation they observed. Thought listing data confirmed these processes. In control conditions, thoughts about the interaction were generally low, indicating that disinterested observers did not think much about the service failure situation. When incidental similarity was manipulated, participants’ thoughts largely pertained to the protagonist to whom they were incidentally similar and correspondingly reflected a disposition to blame the other protagonist for the service failure. When participants were under cognitive load, however, their attention was apparently diverted from the service failure situation and their incidental similarity to the protagonists had little effect on their thoughts about it. (Note that the lack of effect in high-load conditions provides indirect evidence that the processes underlying the attributions of responsibility were conscious and deliberate. Had they been automatic, cognitive load would have had little effect.) Finally, although participants reported being in a better mood when they discovered that they were incidentally similar to one of the protagonists, mood did not mediate the effect of similarity on responsibility attributions. EXPERIMENT 3 Results of the first two experiments are consistent with our assumption that when the protagonists in a service failure situation have no distinctive features that have implications for judgment, an incidental similarity to one of the protagonists increases their attention to the protagonist and leads them to take this protagonist’s perspective when they interpret the situation. Consequently, they decrease attributions of responsibility for the service failure to this protagonist for reasons similar to those suggested by Jones and Nisbett (1971). An additional experiment we conducted (see web appendix) confirmed our assumption that this effect was mediated by an increase in the attention that observers paid to the incidentally similar provider. Experiments 3 and 4 provided further support for this assumption. Moreover, they showed that if one of the protagonists has a salient personal attribute or engages in behavior that observers consider to be undesirable, the observers are likely to attribute responsibility for a negative event to this protagonist regardless of other considerations (Alicke 2000; Regan et al. 1974). In this case, an incidental similarity that draws observers’ attention to this protagonist may strengthen this attribution rather than decreasing it. Experiment 3 determined the effects of incidental similarity to a protagonist who engaged in rude behavior during the service failure situation. Two parallel studies with different contexts were conducted, one in which the service provider was rude (experiment 3a) and the other in which the customer was rude (experiment 3b). Because both studies yielded similar conclusions, we report experiment 3a here and relegate the details of the other subexperiment to the web appendix. In addition, participants’ perceptions of similarity were measured. We expected that control participants would perceive themselves as more similar to the protagonist when they were incidentally similar with the protagonist than when they were not. When a protagonist has an undesirable attribute, however, we expected that observers’ increased attention to the protagonist that was stimulated by their incidental similarity would not have these effects. Experiment 3a Method In this experiment, we used first names to manipulate the similarity between observers and the service provider. Two weeks before the main study, 220 participants completed a survey identical to that used in experiment 2. Of these, 148 participants (40.5% males) whose first names were moderately uncommon participated in a 15 minute study with HK$20 (approximately US$3). They were randomly assigned to cells of a 2 (service provider similarity: present vs. absent) × 2 (rudeness: present vs. absent) between-subjects design. Participants read a passage in which they were told to imagine that they visited a sporting goods store to buy some sportswear and, while looking around, they overheard a conversation between a customer and a salesperson who was referred to by name (either the same as the participant’s or not (see the web appendix). In the story, the customer asks the salesperson to bring him a pair of running shoes that he had ordered online the day before. However, the salesperson tells him that the shoes have not yet arrived. The customer complains that he had not been told that it would take more than one day for delivery, but the salesperson responds that a note on the store website indicates that the delivery will take one to two days. In the rudeness condition, the salesperson was described as looking impatient and “barking” at the customer when he mentioned the website information about the delivery date. In the control condition, the salesperson’s manner was not mentioned. Then, participants reported their intention to buy the product at the store along scales from 1 (very unlikely/inclined not to/definitely will not) to 10 (very likely/inclined to/definitely will). They also responded to six items concerning their inclination to blame the salesperson and the customer along a scale from 1 (strongly disagree) to 10 (strongly agree). These items were identical to those used in experiment 2 except that the provider was referred to as a “salesperson” rather than the “manager” (α > .90 in each case). In addition, participants reported the extent to which they empathized with both the salesperson’s situation and the customer’s situation along scales from 1 (not at all/definitely no) to 10 (very much/definitely yes). Responses to each set of items were averaged (α > .92 in each case). Furthermore, they responded to three items concerning the extent to which they had paid attention to the salesperson (“To what extent did you pay close attention to the salesperson?” “To what extent did you focus your attention on the salesperson?” and “To what extent did the salesperson grab your attention?” along scales from 1 (little) to 10 (a lot); α = .91. They also indicated their similarity to the salesperson along a scale from 1 (very dissimilar) to 10 (very similar) and their perception of the salesperson’s rudeness along a scale from 1 (not rude at all) to 10 (very rude). Experiment 3a Results The results of this experiment are summarized in table 5. TABLE 5 THE IMPACT OF INCIDENTAL SIMILARITY AND SERVICE PROVIDER’S RUDENESS ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 3A Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. TABLE 5 THE IMPACT OF INCIDENTAL SIMILARITY AND SERVICE PROVIDER’S RUDENESS ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 3A Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. Manipulation Check Participants perceived the salesperson to be more rude in rudeness conditions than in control conditions (7.37 vs. 3.55, respectively; F(1, 144) = 324.59, p < .001, ηp2 = .69). Attention The effects of incidental similarity and rudeness conditions on attention and other dependent variables are shown in table 5. Participants’ attention to the salesperson increased with incidental similarity (from 5.28 to 7.04; F(1, 144) = 92.58, p < .001, ηp2 = .39). No other effects were significant. Perceived Similarity Both rudeness and incidental similarity affected participants’ perceptions of similarity to the provider (F(1, 144) = 144.70, p < .001, ηp2 = .50 and F(1, 144) = 36.65, p < .001, ηp2 = .20, respectively). However, the interaction of these variables was also significant (F(1, 144) = 70.77, p < .001, ηp2 = .33). As shown in table 5, incidental similarity increased perceptions of similarity to the provider in control conditions (from 3.30 to 5.78; F(1, 144) = 209.69, p < .001, ηp2 = .59), but it decreased these perceptions when the provider was rude (from 2.68 to 2.27; F(1, 144) = 6.62, p < .05, ηp2 = .04). Patronage Intentions Unlike previous experiments in which the protagonists were nondescript, we expected that incidental similarity to a rude protagonist would increase attributions of responsibility to this protagonist and that these attributions would be reflected in patronage inclinations. Analyses of participants’ patronage inclinations yielded a main effect of service provider rudeness and an interaction of provider similarity and rudeness (F(1, 144) = 129.59, p < .001, ηp2 = .47 and F(1, 144) = 24.21, p < .001, ηp2 = .14, respectively). In control conditions, participants’ incidental similarity to the salesperson increased the inclination to patronize the store (from 4.86 to 5.97; F(1, 144) = 37.18, p < .001, ηp2 = .21). When the salesperson was rude, however, incidental similarity decreased patronage inclinations (from 3.68 to 2.97; F(1, 144) = 15.69, p < .001, ηp2 = .10). Attributions of Responsibility The effects of the provider’s incidental similarity and rudeness on intentions to blame the provider paralleled their effects on patronage intentions. The main effect of rudeness and its interaction with incidental similarity were significant (F(1, 144) = 83.24, p < .001, ηp2 = .37 and F(1, 144) = 33.13, p < .001, ηp2 = .19, respectively). Incidental similarity to the salesperson decreased perceptions that she or he was to blame in control conditions (from 5.80 to 4.65; F(1, 144) = 41.99, p < .001, ηp2 = .23) but increased these perceptions in rudeness conditions (from 6.41 to 7.32; F(1, 144) = 26.61, p < .001, ηp2 = .16). Analyses of attributions of responsibility to the customer also yielded a main effect of provider rudeness and an interaction of rudeness with incidental similarity (F(1, 144) = 188.16, p < .001, ηp2 = .57 and F(1, 144) = 5.70, p < .05, ηp2 = .04, respectively). Incidental similarity to the provider increased attributions of responsibility to the customer in control conditions (from 5.53 to 6.15; F(1, 144) = 12.31, p < .01, ηp2 = .08) but not when the provider was rude (3.46 vs. 3.21, respectively; F(1, 144) = 2.68, p > .10, ηp2 = .02). Empathy (Perspective) We inferred the perspective that observers took when they construed the implications of the service failure from their estimates of their empathy. We expected that in control conditions, their empathy for a protagonist would increase with their incidental similarity to him or her but that in rudeness conditions, this would not be the case. Observers’ empathy for each protagonist was analyzed separately and yielded analogous results. The provider’s rudeness decreased empathy for the provider (from 5.87 to 3.57; F(1, 144) = 142.20, p < .001, ηp2 = .50) and increased empathy for the customer (from 5.53 to 6.54; F(1, 144) = 31.98, p <.001, ηp2 = .18). In addition, incidental similarity and the provider’s rudeness had interactive effects on empathy for both the provider (F(1, 144) = 21.23, p < .001, ηp2 = .13) and the customer (F(1, 144) = 4.78, p <.05, ηp2 = .03). Table 5 shows that in control conditions, incidental similarity to the provider increased empathy for the provider (from 5.32 to 6.41; F(1, 144) = 32.94, p <.001, ηp2 = .19) and decreased empathy for the customer (from 6.01 to 5.04; F(1, 144) = 30.10, p <.001, ηp2 = .17), consistent with the assumption that it increased the disposition to take the perspective of the provider in this condition. When the provider was rude, however, incidental similarity to him decreased empathy for him (from 3.92 to 3.23; F(1, 144) = 13.64, p <.001, ηp2 = .09) and did not appreciably affect empathy for the customer (6.64 vs. 6.45, respectively; F(1, 144) = 1.66, p >.10, ηp2 = .01), suggesting that observers avoided taking the provider’s perspective in this case. Mediation Bootstrapping analyses (model 91, Hayes 2018) confirmed the mediating effects of attention and responsibility attributions on the interactive effects of the provider’s rudeness and incidental similarity. Specifically, the causal model: incidental similarity → attention × rudeness → attributions of blame to provider → patronage inclinations was reliable (based on 5,000 bootstrapping samples, 95% CI from –2.0736 to –1.0381). However, an alternative model in which the causal sequence of blame and attention was reversed was not supported (95% CI from –.0636 to .0997). In addition, a serial mediation analysis (model 6, Hayes 2018) that included empathy as a mediator was confirmed in the control condition: incidental similarity → attention → empathy for provider → attributions of blame to provider → patronage inclinations (95% CI from .0071 to .2318) but not in the rudeness condition (95% CI from –.0936 to .0392). Experiment 3b A similar experiment confirmed analogous effects of rudeness when the customer in a service failure situation was rude. Birth month was used to manipulate incidental similarity. One hundred twenty participants (49.2% males) imagined that they viewed a video on YouTube depicting a customer’s interaction with a restaurant manager. The customer indicated that he had made a reservation and requested a sea-view table one week ago through the restaurant’s website. However, the manager replied he checked the reservation system and did not see the customer’s special request. In the rudeness condition, the customer screamed and shouted loudly at the manager when he learned that he could not get the sea-view table. In the control condition, the customer’s behavior was not mentioned. In the customer similarity conditions, the dialogue showed that the observers and the customer had the same birth month, whereas in no-similarity condition, this was not the case. In this experiment, incidental similarity and the customer’s rudeness had interactive effects on perceived similarity to the customer (F(1, 116) = 58.46, p < .001, ηp2 = .34), attributions of responsibility (F(1, 116) = 6.75, p < .05, ηp2 = .06), and patronage intentions (F(1, 116) = 23.48, p < .001, ηp2 = .17). Specifically, in control conditions, an incidental similarity to the customer increased observers’ perceptions of their similarity to him (from 3.37 to 5.67; F(1, 116) = 151.51, p < .001, ηp2 = .57), increased their attributions of responsibility to the provider (from 5.68 to 6.52; F(1, 116) = 19.97, p <.001, ηp2 = .15), and decreased their inclination to patronize the establishment (from 4.64 to 2.99; F(1, 116) = 59.82, p <.001, ηp2 = .34). When the customer was rude, however, incidental similarity to him decreased their perceptions of their similarity to him (from 2.27 to 1.70; F(1, 116) = 10.20, p < .01, ηp2 = .08), increased their disposition to blame him (from 5.89 to 6.40; F(1, 116) = 5.87, p < .05, ηp2 = .05), and increased their patronage intentions (from 5.07 to 5.49; F(1, 116) = 4.55, p < .05, ηp2 = .04). (For more details, see the web appendix.) Discussion Experiment 3 confirmed the hypothesis that when a protagonist in a service failure situation engages in undesirable behavior, an incidental similarity to the protagonist increases rather than decreases attributions of responsibility to this protagonist. Although control participants perceived themselves as more similar to the protagonist when they were incidentally similar to the protagonist than when they were not, this difference was reversed when the protagonist was rude. The different effects of incidental similarity on empathy that occurred under rudeness and control conditions confirm our assumption that the processes underlying the effect of incidental similarity do not operate independently. Rather, the effects of incidental similarity on the disposition to take the perspective of an incidentally similar protagonist do not occur if the protagonist has a salient distinguishing characteristic. If this had occurred, incidental similarity would have had similar effects in both the rudeness and control conditions. In fact, although an incidental similarity to the provider increased empathy for him in control conditions, it significantly decreased empathy for him when he was rude. Moreover, similarity to the provider decreased empathy for the customer under control conditions but had no effect when the provider was rude. The effect of empathy in control conditions was presumably influenced by participants’ incidental similarity to the protagonist and was a determinant of their attention to him. When the protagonist was rude, however, empathy was not a determinant of their attention to the protagonist but rather was a consequence of participants’ negative reactions to this behavior. Finally, note that the influence of incidental similarity to a protagonist on observers’ judgments of their general similarity to him depended on whether protagonist was rude or not. As figure 1 indicates, however, these perceptions do not enter into the conceptualization we propose. That is, observers’ incidental similarity to the protagonist directs their attention to him, and this attention leads them either to take the protagonist’s perspective or, if the protagonist is rude, to think more extensively about the protagonist’s responsibility for the service failure. Although perceived similarity may be a correlate of these effects, it does not directly influence the phenomena we predict. EXPERIMENT 4 Experiment 4 determined whether an undesirable feature of a protagonist that is irrelevant to the situation observed (specifically, the protagonist’s obesity) could have similar effects. Two substudies with different contexts were again conducted, one (experiment 4a) in which the customer’s obesity was manipulated and the other (experiment 4b) in which the service provider’s obesity was manipulated. However, because the results of the two studies have virtually identical conclusions, the details of experiment 4b are reported in the web appendix. Experiment 4a Method In this experiment, the obesity of the customer rather than of the provider was varied. Moreover, we obtained a measure of self-threat to determine whether this factor played a role in the effects we observed. We used first names to manipulate incidental similarity. We invited 210 female participants to perform a survey identical to experiment 2 two weeks before the main study. Of these, 160 whose first names were not extremely uncommon participated in a 15 minute study for HK$20 (approximately US$3). They were randomly assigned to cells of a 2 (customer similarity: present vs. absent) × 2 (customer obesity: present vs. absent) between-subjects design. Participants imagined themselves going to a restaurant for dinner. On this pretense, their attention was directed to a computer screen showing a picture of a male waiter and a female customer. Participants saw only the back of a female customer, who sat at a table with the male waiter facing her. The customer was portrayed by either an overweight female model or an average weight model (age = 26). An average-looking male confederate (age = 28) portrayed the waiter. A pretest using a between-subject design (n = 60) confirmed that the female confederate was perceived to be more obese than in control conditions (8.10 vs. 4.67, respectively, along a scale from 1 (not fat at all) to 10 (very fat); F(1, 58) = 226.06, p < .001, ηp2 = .80). The pretest also confirmed the male confederate was average-looking, along a scale from 1 (not handsome at all) to 10 (very handsome); (5.23 vs. 5.37, respectively along a scale from 1 to 10; F < 1). Finally, the customer was referred to by a name that was either the same as or different from the participant’s (see the web appendix). Participants were asked to imagine overhearing a conversation between the two protagonists. In each case, the customer complained to the waiter about waiting a long time for a dessert special. The waiter explained that it took time to prepare the dessert because it was freshly made. The customer asked why she had not been informed of this earlier, and the waiter replied that the menu indicated that customers should order early if they wanted the dessert special. The customer looked unhappy and paid her bill without waiting for the dessert. Then, participants were given a questionnaire similar to that used in experiment 3a, with an exception that measures of the inclination to blame the provider referred to “waiter” rather than “manager” (α > .92 in each case). In addition, they responded to three items used in experiment 3a concerning the extent to which they had paid attention to the customer (α = .90) along scales from 1 (little) to 10 (a lot). They were also asked to write down any features of the customer that happened to draw their attention. Finally, we speculated that self-threat might mediate the effects of incidental similarity on observers’ attributions of blame and their subsequent reactions. To this end, we asked participants to report the extent to which the protagonist’s obesity threatened their self-image (i.e., “To what extent did the customer’s characteristics negatively affect your self-image?”) along a scale from 1 (little) to 10 (a lot) and indicated their perception of the customer’s obesity (i.e., “Do you think the customer is fat?”) along a scale from 1 (not fat at all) to 10 (very fat). Experiment 4a Results The results of this experiment are summarized in table 6. TABLE 6 THE IMPACT OF INCIDENTAL SIMILARITY AND CUSTOMER OBESITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—STUDY 4A Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at the p < .05. TABLE 6 THE IMPACT OF INCIDENTAL SIMILARITY AND CUSTOMER OBESITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—STUDY 4A Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at the p < .05. Manipulation Check Participants perceived the customer to be more obese in the obesity condition than in the control condition (7.73 vs. 2.95, respectively; F(1, 156) = 378.65, p < .001, ηp2 = .71). Attention Participants’ attention to the customer increased with both incidental similarity (from 5.80 to 7.54; F(1, 156) = 95.44, p < .001, ηp2 = .38) and the customer’s obesity (from 6.50 to 6.84; F(1, 156) = 3.60, p = .06, ηp2 = .02). No interaction occurred. Thought listing data were consistent with these attentional differences. Participants reported more features of the customer that happened to draw their attention in the similarity condition than in the control condition (2.18 vs. 1.01, respectively; F(1, 156) = 113.00, p < .001, ηp2 = .42) and more features in the obese condition than in the control condition (1.75 vs. 1.44, respectively; F(1, 156) = 8.17, p < .01, ηp2 = .05). The interaction was not significant. These differences were also evident when features pertaining to the incidental similarity were not considered. That is, participants reported more characteristics of the obese customer (e.g., she is fat, she has long hair, she wears a blue jacket) in the similarity condition than in the control condition (1.78 vs. 1.18, respectively; F(1, 156) = 15.86, p < .001, ηp2 = .91). No difference was evident when the customer was not obese (1.03 vs. .85, respectively; F(1, 156) = 1.35, p > .10, ηp2 = .01). Perceived Similarity Analyses of perceived similarity yielded main effects of customer similarity and obesity (F(1, 156) = 35.57, p < .001, ηp2 = .19 and F(1, 156) = 265.05, p < .001, ηp2 = .63, respectively) and an interaction of these variables (F(1, 156) = 109.78, p < .001, ηp2 = .41). Between-cell comparisons (see table 6) indicated that in control conditions, participants perceived the customer to be more similar to them when they had an incidental similarity to the customer than when they did not (5.78 vs. 3.23, respectively; F(1, 156) = 271.41, p <.001, ηp2 = .64). When the customer was obese, however, this difference was reversed (1.63 vs. 2.33, respectively; F(1, 156) = 21.41, p <.001, ηp2 = .12). Patronage Intentions Analyses of patronage intensions yielded main effects of customer similarity and obesity (F(1, 156) = 7.39, p < .01, ηp2 = .05; and F(1, 156) = 73.23, p < .001, ηp2 = .32, respectively) and an interaction of obesity and incidental similarity (F(1, 156) = 37.29, p < .001, ηp2 = .19). Between-cell comparisons indicated that incidental similarity decreased patronage intentions in control conditions (from 4.73 to 3.08; F(1, 156) = 78.56, p <.001, ηp2 = .33) but increased patronage intentions when the customer was obese (from 5.19 to 5.83; F(1, 156) = 12.43, p <.001, ηp2 = .07). Responsibility Attributions Analyses of inclinations to blame the service provider yielded main effects of both incidental similarity (F(1, 156) = 8.29, p < .01, ηp2 = .05), and the customer’s obesity (F(1, 156) = 77.05, p < .001, ηp2 = .33) and an interaction of these variables (F(1, 156) = 15.10, p < .001, ηp2 = .09). Incidental similarity to the customer increased attributions of responsibility to the provider in control conditions (from 5.48 to 6.65; F(1, 156) = 46.19, p <.001, ηp2 = .23), consistent with the results of earlier studies. However, this difference was not evident when the customer was obese (4.45 vs. 4.63, respectively; F(1, 156) = 1.90, p >.10, ηp2 = .01). Analyses of inclinations to blame to customer also yielded a main effect of obesity and an interaction (F(1, 156) = 69.78, p < .001, ηp2 = .31 and F(1, 156) = 10.31, p < .01, ηp2 = .06, respectively). Incidental similarity decreased attributions of responsibility to the customer in control conditions (from 4.87 to 4.23; F(1, 156) = 10.60, p <.01, ηp2 = .06) but increased these attributions when she was obese (from 5.92 to 6.58; F(1, 156) = 11.23, p <.01, ηp2 = .07). Empathy (Perspective) Separate analyses of empathy for the provider and empathy for the customer yielded an interaction of customer obesity and incidental similarity in each case (F(1, 156) = 6.29, p < .05, ηp2 = .04 and F(1, 156) = 8.78, p < .01, ηp2 = .05, respectively). Table 6 shows that in control conditions, observers’ incidental similarity to the customer increased their empathy for her (from 5.79 to 6.37; F(1, 156) = 10.73, p <.01, ηp2 = .06) and decreased their empathy for the provider (from 4.89 to 4.23; F(1, 156) = 18.10, p <.001, ηp2 = .10), consistent with assumptions that observers took the customer’s perspective in these conditions. When the customer was obese, however, incidental similarity to her decreased observers’ empathy for her (from 4.54 to 4.03; F(1, 156) = 8.47, p <.01, ηp2 = .05) and had little effect on their empathy for the provider (5.65 vs. 5.80, respectively; F(1, 156) = 1.87, p >.10, ηp2 = .01). Thus, although incidental similarity to the obese customer increased observers’ attention to her, it led them to avoid taking her perspective when they construed the implications of the service failure situation. These findings, like those of experiment 3, suggest that when a protagonist has a distinguishing characteristic that is potentially relevant to observers’ attributions of responsibility, the effects of incidental similarity on the perspective they take are not evident. Mediation We assumed that participants’ attention to the customer and their responsibility attributions would mediate the effects of the customer’s obesity and incidental similarity on observers’ patronage intentions; that is, incidental similarity → attention × obesity → attributions of blame to customer → patronage inclinations. Bootstrapping analyses (Hayes 2018, model 91) confirmed this model (based on 5,000 samples, 95% CI from .5279 to 1.4197). However, an alternative model in which the causal sequence of blame and attention was reversed was not supported (based on 5,000 bootstrapping samples, 95% CI from –.0417 to .0587). In addition, a serial mediation analysis (model 6, Hayes 2018) that included empathy as a mediator was confirmed in the control condition: incidental similarity → attention → empathy for customer → attributions of blame to customer → patronage inclinations (95% CI from –.3359 to –.0487). In the obesity condition, however, this model was not supported (95% CI from –.1017 to .3369). Self-Threat The main effect of obesity on self-threat was significant (F(1, 156) = 12.16, p < .01, ηp2 = .07) and the interaction of similarity and obesity was marginally significant (F(1, 156) = 3.68, p = .057, ηp2 = .02). Incidental similarity to the customer increased feelings of self-threat to the customer (from 1.83 to 2.35; F(1, 156) =14.37, p < .001, ηp2 = .08) when the customer was obese, and no effect was found in the control conditions (1.60 vs.1.58; F(1, 156) = 1.24, p > .10, ηp2 = .01). However, a model that assumed that self-threat mediated the effect of incidental similarity on attributions of blame to the customer was not supported under either obesity conditions (95% CI: from –.0323 to .1179) or control conditions (95% CI: from –.0419 to .0156). Experiment 4b A second experiment (described in the web appendix) showed that analogous effects of obesity occur when the provider in a service failure situation is obese. One hundred twenty participants (43% male) imagined going to a bookstore and overhearing a conversation between a female salesperson and a male customer. A picture of the interaction was presented and indicated that the salesperson was either of average weight or excessive weight. (In the picture, the salesperson had her back to the camera, so participants could not see her face.) In the conversation, the customer found that a book he had ordered had not been delivered on time. In similar provider conditions, the dialogue indicated that participants had the same last name as the salesperson. In control conditions, the names did not match. Incidental similarity and the provider’s obesity had interactive effects on perceived similarity to the provider (F(1, 116) = 56.47, p < .001, ηp2 = .33), attributions of responsibility to the provider (F(1, 116) = 21.16, p < .001, ηp2 = .15), and patronage intentions (F(1, 116) = 15.43, p < .001, ηp2 = .12). In control conditions, incidental similarity to the provider increased observers’ perceptions of their similarity to her (from 3.37 to 5.97; F(1, 116) = 182.45, p < .001, ηp2 = .61), decreased their disposition to blame the provider for the incident (from 5.58 to 4.68; F(1, 116) = 19.92, p < .01, ηp2 = .15), and increased patronage intentions (from 4.89 to 6.06; F (1, 116) = 31.99, p < .001, ηp2 = .22). When the provider was obese, however, observers’ incidental similarity to her decreased their perceptions of their similarity to her (from 2.83 to 2.53; F(1, 116) = 3.31, p = .07, ηp2 = .03), increased their inclination to blame the provider for the service failure (from 6.26 to 7.25; F(1, 116) = 24.03, p < .001, ηp2 = .17), and decreased their patronage intentions (from 4.03 to 3.56; F(1, 116) = 5.80, p < .05, ηp2 = .05). Discussion Experiment 4 confirmed the hypothesis that an undesirable feature of a protagonist that is irrelevant to the situation observed (i.e., the protagonist’s obesity) would reverse the effect of incidental similarity on the observer’s responsibility attributions and willingness to patronize the establishment in the future. Mediation analyses also confirmed that attention underlies this reversed effect and ruled out an explanation in terms of self-threat. These findings are consistent with our expectation that incidental similarity increases attention to the provider/customer in an interaction and consequently increases the effect of other characteristics of the provider/customer on judgments of him and the situation. Although control participants perceived themselves as more similar to the protagonist when they were incidentally similar to the protagonist than when they were not, this difference was reversed when the protagonist was obese. Thought listing data also showed that participants reported more thoughts about other characteristics of the obese customer in the similarity condition than in the control condition. In other words, incidental similarity apparently induced participants to think about other differences between themselves and the obese customer and, therefore, lowered their perceived similarity to her. GENERAL DISCUSSION Four field and laboratory experiments show that an incidental similarity to one of the protagonists in a service failure situation can increase observers’ attention to this protagonist and consequently can affect their attributions of responsibility for the failure’s occurrence. The effects of similarity on social interactions (Berscheid and Reis 1998; Byrne 1971) and the determinants of attributions of responsibility have been extensively investigated in previous research (Jones and Nisbett 1971; Weiner 1980,1985). However, our conceptualization of the way these factors influence observers’ reactions to a service failure has both theoretical and empirical implications that have not previously been identified. Theoretical Implications In contrast to assumptions underlying previous research on the effects of perceived similarity (Byrne 1971), we found that the effect of incidental similarity to a protagonist on responses to information about a service failure was mediated by the relative attention that is spontaneously drawn to this protagonist and not by similarity per se. This attention can have the opposite effects on observers’ attributions of responsibility for the incident, depending on whether the protagonist has a distinguishing feature that could affect these attributions. When neither protagonist in the situation has a distinguishing characteristic that is likely to be considered undesirable, observers whose attention is drawn to one protagonist spontaneously construe its implications from this protagonist’s perspective. In these conditions, therefore, an incidental similarity to one protagonist decreases the disposition to blame this protagonist and consequently increases the disposition to blame the other. When one of the protagonists in a service failure situation has an undesirable characteristic or has engaged in undesirable behavior, however, observers generally evaluate this protagonist negatively and blame him or her for the service failure. In this case, their incidental similarity to the protagonist has little, if any, effect on the perspective from which they construe its implications. However, their relatively greater attention to the incidentally similar protagonist increases the thought they devote to their judgment and increases the extremity of their attribution of responsibility. In principle, these two effects of incidental similarity could occur independently. If this were so, however, the opposing effects of increased attention to a protagonist who has an undesirable attribute would offset one another. But, as the results of experiments 3 and 4 indicate, there was no evidence that incidental similarity affected observers’ perspective in these conditions (as inferred from the effect of empathy). Thus, it seems reasonable to conclude that when an incidentally similar protagonist has an undesirable attribute, observers’ disposition to take the protagonist’s perspective is eliminated. Other interpretations of these effects should be considered. For example, people generally avoid taking responsibility for negative events (Arkin et al. 1976; Shaver 1985), and this tendency could contribute to the disposition for observers who took the perspective of a protagonist to attribute responsibility for the service failure to the other. As experiment 4 indicated, however, the motivation to avoid threats to self-esteem did not appear to be a major factor in the conditions we investigated. Moreover, if observers were personally threatened by a negative event for which a similar person was responsible, they would be relatively disinclined to attribute a service failure to such a person regardless of other considerations. In contrast, experiments 3 and 4 showed that incidental similarity increased observers’ disposition to attribute the service failure to a protagonist who had an undesirable characteristic. Finally, experiment 4 provided no evidence that self-threat mediated the effects we observed. Two aspects of our results are worth reiterating. First, previous research on the effects of incidental similarity has typically been restricted to conditions in which (a) individuals interact directly with the similar other, and (b) the effects of the interaction are positive (cf. Burger et al. 2004; Guéguen et al. 2005), to respond to his or her questions on intimate topics (Martin and Guéguen 2013; Miller et al. 1998). In contrast, we found that people’s incidental similarity can affect reactions to a person even when they have no direct contact with the person but only observe a situation in which the person is involved. Second, research on perspective taking indicates that the closeness of people’s relationship to someone has a positive effect on their likelihood of taking the person’s perspective and empathizing with this person (Batson and Shaw 1991; Cialdini et al. 1997; Johnson et al. 2002). In this research, however, similarity has typically been defined in terms of personality, values, or shared experiences. We show that an individual’s trivial similarity that has no objective implications for personality can stimulate observers to empathize with him or her. Moreover, this type of similarity can sometimes decrease empathy rather than increasing it. Other Implications Numerous studies have examined consumers’ attributions and reactions to service failures (Bitner 1990; Chan, Wan, and Sin 2009; Folkes, Koletsky, and Graham 1987; Wan 2013; Wan, Hui, and Wyer 2011). However, these studies largely focused on reactions of a customer who was directly involved in the service failure. Although a few studies have investigated the role of observers (Laufer and Gillespie 2004; Wan, Chan, and Su 2011), they examined only factors that influence observers’ attributions of blame to the service provider. Our research shows that incidental similarity can affect both online and offline observers’ attribution of blame to both the service provider and the customer and that this effect occurs even though the failure is personally irrelevant. The effect of incidental similarity to a protagonist on observers’ responsibility attributions has implications for consumer behavior outside the laboratory. In an actual shopping situation, idiosyncratic characteristics of a service provider are usually conveyed in the provider’s name tag. Our results suggest that this information may fortuitously stimulate customers to identify an incidentally similarity to the provider and consequently to hold the person less responsible not only for problems that involve the customer personally but also for problems that involve others. These effects might only be evident, however, if the provider is not antagonistic or does not have distinguishing physical characteristics that customers consider undesirable. Our results could also have implications for consumers’ reactions to the information conveyed online. Many customers base their offline purchase decisions on online reviews. Online reviewer profiles (with reviewer’s name, picture, the number of reviews posted, etc.) are one of the most important ways that customers assess the reviews’ credibility (Lim and Van Der Heide 2015). Many social media websites convey consumers’ descriptions of a shopping experience they have had at a particular store or restaurant. The description may be either favorable or unfavorable. Communicators might be more effective if they accompany their message with their name or other demographic features that are fortuitously similar to those of the recipients. Some qualifications on our findings should be considered. For example, our conclusions assume that observers not only are incidentally similar to the persons they observe but also evaluate the similarity favorably. Individuals with a strong need for uniqueness may react negatively to indications they are similar to another (Snyder and Fromkin 1977). Moreover, both individual and cultural differences might influence one’s general disposition to think about oneself in relation to others (Markus and Kitayama 1991). Participants in the present research were Asians, who may be particularly sensitive to the similarities between themselves and others and inclined to value social connectedness. Although the effects of incidental similarity have been identified in research on Western cultural samples as well (Burger et al. 2004), this difference could be a consideration in evaluating the generalizability of our findings. DATA COLLECTION INFORMATION The first author designed experiment 1 and supervised the collection of data for it by research assistants at Hong Kong Green Fresh Vegan Restaurant in December 2015. The first author designed experiments 2, 3, 4, and 5 with feedback from the second author. The first author supervised the collection of data for experiments 2, 3, 4, and 5 by research assistants at the Chinese University of Hong Kong. Experiment 4b was run in September 2016, experiment 5b in November 2016, experiment 3 in July 2017, experiment 5a in August 2017, experiment 2 in November 2017, and experiment 4a in March 2018. Data from all experiments were analyzed by the first author and discussed by both authors. The authors gratefully acknowledge the helpful input and guidance from the editor, associate editor, and three reviewers. The authors also gratefully acknowledge a group of research assistants and postgraduate students from the School of Hotel and Tourism Management, The Chinese University of Hong Kong, for their assistance in data collection. In addition, the authors would like to thank Hong Kong Green Fresh Vegan Restaurant for their help in data collection for the field experiment. This research was supported by Grants GRF 14502514 from the Research Grants Council, Hong Kong. Supplemental materials are included in the web appendix accompanying the online version of this article. References Alicke Mark D. ( 2000 ), “ Culpable Control and the Psychology of Blame ,” Psychological Bulletin , 126 4 , 556 – 74 . Google Scholar CrossRef Search ADS PubMed Arkin Robert. M. , Gleason James M. , Johnston Shawn ( 1976 ), “ Effect of Perceived Choice, Expected Outcome, and Observed Outcome of an Action on the Causal Attributions of Actors ,” Journal of Experimental Social Psychology , 12 2 , 151 – 8 . Google Scholar CrossRef Search ADS Bargh John A. ( 1982 ), “ Attention and Automaticity in the Processing of Self-Relevant Information ,” Journal of Personality and Social Psychology , 43 3 , 425 – 36 . Google Scholar CrossRef Search ADS Batson C. Daniel , Duncan Bruce D. , Ackerman Paula , Buckley Terese , Birch Kimberly S. ( 1981 ), “ Is Empathic Emotion a Source of Altruistic Motivation? ” Journal of Personality and Social Psychology , 40 2 , 290 – 302 . Google Scholar CrossRef Search ADS Batson C. Daniel , Shaw Laura L. ( 1991 ), “ Evidence for Altruism: Toward a Pluralism of Prosocial Motives ,” Psychological Inquiry , 2 2 , 107 – 22 . Google Scholar CrossRef Search ADS Berscheid Ellen , Reis Harry T. ( 1998 ), “Attraction and Close Relationships,” in Handbook of Social Psychology , Vol. 2 , ed. Gilbert Daniel T. , Fiske Susan T. , Lindzey Gardner , Boston : McGraw-Hill , 193 – 281 . Bitner Mary Jo ( 1990 ), “ Evaluating Service Encounters: The Effects of Physical Surroundings and Employee Responses ,” Journal of Marketing , 54 2 , 69 – 82 . Google Scholar CrossRef Search ADS Burger Jerry M. , Messian Nicole , Patel Shebani , del Prado Alicia , Anderson Carmen ( 2004 ), “ What a Coincidence! The Effects of Incidental Similarity on Compliance ,” Personality and Social Psychology Bulletin , 30 1 , 35 – 43 . Google Scholar CrossRef Search ADS PubMed Byrne E. Donn ( 1971 ), The Attraction Paradigm , New York : Academic Press . Causon Jo ( 2015 ), “Customer Complaints Made via Social Media on the Rise,” https://www.theguardian.com/media-network/2015/may/21/customer-complaints-social-media-rise/. Chan Haksin , Wan Lisa C. , Sin Leo Y. M. ( 2009 ), “ The Contrasting Effects of Culture on Consumer Tolerance: Interpersonal Face and Impersonal Fate ,” Journal of Consumer Research , 36 2 , 292 – 304 . Google Scholar CrossRef Search ADS Cialdini Robert B. , Brown Stephanie L. , Lewis Brian P. , Luce Carol , Neuberg Steven L. ( 1997 ), “ Reinterpreting the Empathy-Altruism Relationship: When One into One Equals Oneness ,” Journal of Personality and Social Psychology , 73 3 , 481 – 94 . Google Scholar CrossRef Search ADS PubMed Cooper Joel , Jones Edward E. ( 1969 ), “ Opinion Divergence as a Strategy to Avoid Being Miscast ,” Journal of Personality and Social Psychology , 13 1 , 23 – 30 . Google Scholar CrossRef Search ADS Cowart Kelly O. , Brady Michael K. ( 2014 ), “ Pleasantly Plump: Offsetting Negative Obesity Stereotypes for Frontline Service Employees ,” Journal of Retailing , 90 3 , 365 – 78 . Google Scholar CrossRef Search ADS Crandall Christian S. ( 1994 ), “ Prejudice against Fat People: Ideology and Self-Interest ,” Journal of Personality and Social Psychology , 66 5 , 882 – 94 . Google Scholar CrossRef Search ADS PubMed Crosby Lawrence A. , Evans Kenneth R. , Cowles Deborah ( 1990 ), “ Relationship Quality in Services Selling: An Interpersonal Influence Perspective ,” Journal of Marketing , 54 3 , 68 – 81 . Google Scholar CrossRef Search ADS Davis Mark H. ( 1983 ), “ Measuring Individual Differences in Empathy: Evidence for a Multidimensional Approach ,” Journal of Personality and Social Psychology , 44 1 , 113 – 26 . Google Scholar CrossRef Search ADS Dimensional Research ( 2013 ), “Customer Service and Business Results: A Survey of Customer Service from Mid-Size Companies,” https://d16cvnquvjw7pr.cloudfront.net/resources/whitepapers/Zendesk_WP_Customer_Service_and_Business_Results.pdf. Folkes Valerie S. ( 1982 ), “ Forming Relationships and the Matching Hypothesis ,” Personality and Social Psychology Bulletin , 8 4 , 631 – 6 . Google Scholar CrossRef Search ADS Folkes Valerie S. , Koletsky Susan , Graham John L. ( 1987 ), “ A Field Study of Causal Inferences and Consumer Reaction: The View from the Airport ,” Journal of Consumer Research , 13 4 , 534 – 9 . Google Scholar CrossRef Search ADS Greenleaf Christy , Starks Misty , Gomez Laura , Chambliss Heather , Martin Scott ( 2004 ), “ Weight-Related Words Associated with Figure Silhouettes ,” Body Image , 1 4 , 373 – 84 . Google Scholar CrossRef Search ADS PubMed Guéguen Nicolas , Martin Angélique ( 2009 ), “ Incidental Similarity Facilitates Behavioral Mimicry ,” Social Psychology , 40 2 , 88 – 92 . Google Scholar CrossRef Search ADS Guéguen Nicolas , Pichot Nathalie , Le Dreff Gwénaëlle ( 2005 ), “ Similarity and Helping Behavior on the Web: The Impact of the Convergence of Surnames between a Solicitor and a Subject in a Request Made by E-Mail ,” Journal of Applied Social Psychology , 35 2 , 423 – 9 . Google Scholar CrossRef Search ADS Hansen Ranald D. , Stonner David M. ( 1978 ), “ Attributes and Attributions: Inferring Stimulus Properties, Actors’ Dispositions, and Causes ,” Journal of Personality and Social Psychology , 36 6 , 657 – 67 . Google Scholar CrossRef Search ADS Hayes Andrew F. ( 2018 ), Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach , 2nd ed. , New York : Guildford Press . Jiang Lan , Hoegg JoAndrea , Dahl Darren W. , Chattopadhyay Amitava ( 2010 ), “ The Persuasive Role of Incidental Similarity on Attitudes and Purchase Intentions in a Sales Context ,” Journal of Consumer Research , 36 5 , 778 – 91 . Google Scholar CrossRef Search ADS Johnson James D. , Simmons Carolyn H. , Jordav Amanda , Maclean Leslie , Taddei Jeffrey , Thomas Duane , Dovidio John F. , Reed William ( 2002 ), “ Rodney King and O. J. Revisited: The Impact of Race and Defendant Empathy Induction on Judicial Decisions ,” Journal of Applied Social Psychology , 32 6 , 1208 – 23 . Google Scholar CrossRef Search ADS Jones Edward E. , Nisbett Richard E. ( 1971 ), The Actor and the Observer: Divergent Perceptions of the Causes of Behavior , Morristown, NJ : General Learning Press . Kunda Ziva ( 1987 ), “ Motivated Inference: Self-Serving Generation and Evaluation of Causal Theories ,” Journal of Personality and Social Psychology , 53 4 , 636 – 47 . Google Scholar CrossRef Search ADS Laufer Daniel , Gillespie Kate ( 2004 ), “ Differences in Consumer Attributions of Blame between Men and Women: The Role of Perceived Vulnerability and Empathic Concern ,” Psychology and Marketing , 21 2 , 141 – 57 . Google Scholar CrossRef Search ADS Lim Young-shin , Van Der Heide Brandon ( 2015 ), “ Evaluating the Wisdom of Strangers: The Perceived Credibility of Online Consumer Reviews on Yelp ,” Journal of Computer-Mediated Communication , 20 1 , 67 – 82 . Google Scholar CrossRef Search ADS Malle Bertram F. ( 2006 ), “ The Actor–Observer Asymmetry in Attribution: A (Surprising) Meta-Analysis ,” Psychological Bulletin , 132 6 , 895 – 919 . Google Scholar CrossRef Search ADS PubMed Markus Hazel R. , Kitayama Shinobu ( 1991 ), “ Culture and the Self: Implications for Cognition, Emotion, and Motivation ,” Psychological Review , 98 2 , 224 – 53 . Google Scholar CrossRef Search ADS Martin Angélique , Guéguen Nicolas ( 2013 ), “ The Influence of Incidental Similarity on Self-Revelation in Response to an Intimate Survey ,” Social Behavior and Personality: An International Journal , 41 3 , 353 – 6 . Google Scholar CrossRef Search ADS McFerran Brent , Dahl Darren , Fitzsimons Gavan , Morales Andrea ( 2010 ), “ I’ll Have What She’s Having: Effects of Social Influence and Body Type on the Food Choices of Others ,” Journal of Consumer Research , 36 6 , 915 – 29 . Google Scholar CrossRef Search ADS Miller Dale T. , Downs Julie S. , Prentice Deborah A. ( 1998 ), “ Minimal Conditions for the Creation of a Unit Relationship: The Social Bond between Birthdaymates ,” European Journal of Social Psychology , 28 3 , 475 – 81 . Google Scholar CrossRef Search ADS Moray Neville ( 1959 ), “ Attention in Dichotic Listening: Affective Cues and the Influence of Instructions ,” Quarterly Journal of Experimental Psychology , 11 1 , 56 – 60 . Google Scholar CrossRef Search ADS Oishi Shigehiro , Wyer Robert S. Jr. , Colcombe Stanley J. ( 2000 ), “ Cultural Variation in the Use of Current Life Satisfaction to Predict the Future ,” Journal of Personality and Social Psychology , 78 3 , 434 – 45 . Google Scholar CrossRef Search ADS PubMed Petty Richard E. , Haugtvedt Curtis , Smith S. M. ( 1995 ), “Elaboration as a Determinant of Attitude Strength: Creating Attitudes that Are Persistent, Resistant, and Predictive of Behavior,” in Attitude Strength: Antecedents and Consequences , ed. Petty Richard E. , Krosnick Jon A. , Mahwah, NJ : Erlbaum , 93 – 130 . Pingitore Regina , Dugoni Bernard L. , Tindale R. Scott , Spring Bonnie ( 1994 ), “ Bias Against Overweight Job Applicants in a Simulated Employment Interview ,” Journal of Applied Psychology , 79 6 , 909 – 17 . Google Scholar CrossRef Search ADS PubMed Price James H. , Desmond Sharon M. , Krol Ronald A. , Snyder Frank F. , O’Connell Janelle K. ( 1987 ), “ Family Practice Physicians’ Beliefs, Attitudes, and Practices Regarding Obesity ,” American Journal of Preventive Medicine , 3 6 , 339 – 45 . Google Scholar CrossRef Search ADS PubMed Puhl Rebecca , Brownell Kelly D. ( 2001 ), “ Bias, Discrimination, and Obesity, ” Obesity Research , 9 12 , 788 – 805 . Google Scholar CrossRef Search ADS PubMed Regan Dennis T. , Straus Ellen , Fazio Russell ( 1974 ), “ Liking and the Attribution Process ,” Journal of Experimental Social Psychology , 10 4 , 385 – 97 . Google Scholar CrossRef Search ADS Regan Dennis T. , Totten Judith ( 1975 ), “ Empathy and Attribution: Turning Observers into Actors ,” Journal of Personality and Social Psychology , 32 5 , 850 – 6 . Google Scholar CrossRef Search ADS PubMed Shaver Kelly G. ( 1985 ), The Attribution of Blame: Causality, Responsibility, and Blameworthiness , New York : Springer . Shiv Baba , Fedorikhin Alexander ( 1999 ), “ Heart and Mind in Conflict: The Interplay of Affect and Cognition in Consumer Decision Making ,” Journal of Consumer Research , 26 3 , 278 – 92 . Google Scholar CrossRef Search ADS Snyder C. R. , Fromkin Howard L. ( 1977 ), “ Abnormality as a Positive Characteristic: The Development and Validation of a Scale Measuring Need for Uniqueness ,” Journal of Abnormal Psychology , 86 5 , 518 – 27 . Google Scholar CrossRef Search ADS Solovay Sondra ( 2000 ), Tipping the Scales of Justice: Fighting Weight-Based Discrimination , Amherst, NY : Prometheus . Stalling Richard B. ( 1970 ), “ Personality Similarity and Evaluative Meaning as Conditioners of Attraction ,” Journal of Personality and Social Psychology , 14 1 , 77 – 82 . Google Scholar CrossRef Search ADS PubMed Storms Michael D. ( 1973 ), “ Videotape and the Attribution Process: Reversing Actors’ and Observers’ Points of View ,” Journal of Personality and Social Psychology , 27 2 , 165 – 75 . Google Scholar CrossRef Search ADS PubMed Tacikowski Pawel , Nowicka Anna ( 2010 ), “ Allocation of Attention to Self-Name and Self-Face: An ERP Study ,” Biological Psychology , 84 2 , 318 – 24 . Google Scholar CrossRef Search ADS PubMed Taylor Shelley E. , Fiske Susan T. ( 1978 ), “ Salience, Attention, and Attribution: Top of the Head Phenomena ,” Advances in Experimental Social Psychology , 11 , 249 – 88 . Google Scholar CrossRef Search ADS Taylor Shelley E. , Peplau Letitia A. , Sears David O. ( 2006 ), Social Psychology , 12th ed. , Upper Saddle River, NJ : Pearson . Tesser Abraham ( 1978 ), “ Self-Generated Attitude Change ,” Advances in Experimental Social Psychology , 11 , 289 – 338 . Google Scholar CrossRef Search ADS Wan Lisa C. ( 2013 ), “ Culture’s Impact on Consumer Complaining Responses to Embarrassing Service Failure ,” Journal of Business Research , 66 3 , 298 – 305 . Google Scholar CrossRef Search ADS Wan Lisa C. , Chan Elisa K. Y. , Su Lei ( 2011 ), “ When Will Customers Care About Service Failures That Happened to Strangers? The Role of Personal Similarity and Regulatory Focus and Its Implication on Service Evaluation ,” International Journal of Hospitality Management , 30 1 , 213 – 20 . Google Scholar CrossRef Search ADS Wan Lisa C. , Hui Michael K. , Wyer Robert S. ( 2011 ), “ The Role of Relationship Norms in Responses to Service Failures ,” Journal of Consumer Research , 38 2 , 260 – 77 . Google Scholar CrossRef Search ADS Weiner Bernard ( 1980 ), Human Motivation , New York : Holt-Rinehart, & Winston . Weiner Bernard ( 1985 ), “ An Attributional Theory of Achievement Motivation and Emotion ,” Psychological Review , 92 4 , 548 – 73 . Google Scholar CrossRef Search ADS PubMed Zhu Feng , Zhang Xiaoquan (Michael) ( 2010 ), “ Impact of Online Consumer Reviews on Sales: The Moderating Role of Product and Consumer Characteristics ,” Journal of Marketing , 74 2 , 133 – 48 . 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/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Consumer Research Oxford University Press

The Influence of Incidental Similarity on Observers’ Causal Attributions and Reactions to a Service Failure

Loading next page...
 
/lp/ou_press/the-influence-of-incidental-similarity-on-observers-causal-WqHI8uOQHd
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/ucy050
Publisher site
See Article on Publisher Site

Abstract

Abstract Observers’ reactions to a service failure and their attributions of responsibility for its occurrence can depend on fortuitous characteristics of the protagonists that happen to draw their attention. Four field and laboratory experiments show that when observers have an incidental similarity to one of the persons involved in a service failure (the customer or the service provider), their attention is drawn to this protagonist, often leading them to construe the situation from this person’s perspective and consequently to blame the protagonist less for the negative event they observe. However, when an incidentally similar protagonist is rude or has an undesirable personal characteristic (i.e., obesity), observers’ greater attention to that person increases their attributions of responsibility to him or her rather than decreasing it. These opposing effects of incidental similarity on attributions influence not only observers’ evaluations of the persons involved in the situation they observe, but also their willingness to patronize the establishment. These effects occur both when observers actually witness a conflict offline and when they consider it online on the basis of reviews. incidental similarity, attribution, service failures, perspective taking, observer reactions Consumer complaints about a service failure are commonplace. For example, a restaurant visitor might protest that his steak is too well done, or a customer might complain that a product she bought doesn’t perform as it should. People who observe such interactions might attribute responsibility for the incident to either the customer or the service provider, and this attribution could influence not only their evaluation of the protagonists but also their inclination to patronize the provider’s establishment. Consumers’ offline purchase decisions are based on online reviews (Zhu and Zhang 2010), and negative reviews influence 80% of these decisions (Dimensional Research 2013). Thus, customer complaints on online social platforms, which increased 800% between 2014 and 2015 (Causon 2015), capture many viewers’ attention. When people observe or read about someone’s complaint to a service provider, what factors influence their attributions of responsibility, and what are the effects of these attributions? Our research provides some answers to these questions. We found that when observers are not personally involved in a service failure, their inclinations to blame the provider or the customer can depend on quite fortuitous events that have nothing objectively to do with the service failure that occurred. For example, observers’ incidental similarity to one of the protagonists (e.g., a common birthday or first name) can increase the attention they pay to this protagonist, and this attention, in turn, can influence their disposition to blame the protagonist for the failure’s occurrence. Two different processes can underlie the influence of this increased attention, however, and these processes have opposite effects. Our conceptualization of these processes is summarized in figure 1. In the conditions we investigated, participants overhear a customer and a service provider discussing a problem that the customer has encountered. We assumed that if observers are unacquainted with the protagonists, they are unlikely to have much interest in the discussion and consequently do not think very deeply about the conditions that gave rise to the service failure being discussed. If they happen to notice an incidental similarity between themselves and one of the protagonists (protagonist A), however, their attention is drawn to him or her and to the incident in which the person is involved. Then, if neither protagonist has a salient characteristic that could explain the incident, observers construe it from the perspective of protagonist A. Consequently, for reasons suggested both by Jones and Nisbett (1971) and by defensive attribution theory (Shaver 1985), they are less likely to blame protagonist A (see figure 1A), and therefore, by default, are more likely to attribute responsibility for the incident to protagonist B. This attribution, in turn, will affect observers’ willingness to patronize the establishment in which the incident occurred. FIGURE 1 View largeDownload slide Conceptual model FIGURE 1 View largeDownload slide Conceptual model There are contingencies in this prediction, however. For example, if observers are unable to devote the cognitive resources necessary to construe the implications of the situation they observe, the effect of incidental similarity on their perspective and its consequent influence on responsibility attributions may not occur. Moreover, suppose protagonist A has a distinctive feature that observers consider to be undesirable (figure 1B). People are unlikely to empathize with someone they dislike or evaluate unfavorably, and their negative reactions can lead them to blame the person for negative events in which he or she is involved (Regan, Straus, and Fazio 1974). Therefore, observers are unlikely to take the perspective of a protagonist whose behavior or personal characteristics elicit negative affective reactions, so they may blame protagonist A for the service failure regardless of other considerations. Moreover, if their incidental similarity to protagonist A increases their attention to him or her, it could increase their cognitive elaboration of the service failure and this person’s involvement in it, and this elaboration might lead them to make more extreme estimates of his or her responsibility for the incident than they otherwise would (Petty, Haugtvedt, and Smith 1995; Tesser 1978). Our research provides the first evidence that consumers’ incidental similarity to one of the protagonists in a situation in which they are not personally involved can influence their reactions to the situation and attributions of responsibility for its occurrence. Moreover, it is the first to find that totally irrelevant features of a service failure situation can have directionally different effects on observers’ reactions to the situation and their willingness to patronize the establishment in which the failure occurred. CONCEPTUAL BACKGROUND To convey the processes that underlie the effects we report, we first discuss the manner in which incidental similarity to a protagonist can influence the observer’s attention to him or her. We then consider the consequences of this attention for observers’ attributions of responsibility for a service failure and the reactions to the protagonist that result. Incidental Similarity, Attention, and Causal Attribution Consumers’ similarity to a salesperson can influence the quality of their interaction with this person (Crosby, Evans, and Cowles 1990). This influence could occur for several reasons. For example, because people typically regard their own attributes favorably, they are inclined to like others who also possess these attributes (Byrne 1971; for other evidence that similarity increases attraction, see Berscheid and Reis 1998; Folkes 1982). Moreover, information that one’s attributes are shared by another can increase one’s self-esteem (Stalling 1970). People can often infer their similarity to one another from their behavior or from demographic factors (e.g., gender, race, or age) that have stereotype-based implications for their personality, attitudes, and values. However, even an incidental similarity to someone (e.g., a common first name, hometown, or birthday) can have a significant impact on their reactions to this person (Burger et al. 2004; Guéguen, Pichot, and Le Dreff 2005; Jiang et al. 2010; Martin and Guéguen 2013). For example, if individuals’ similarity to one another distinguishes them from others in the immediate situation, it can induce a sense of social connectedness. (For example, two Chicagoans are likely to feel more connected to each other if they meet in Paris than if they meet on a Chicago subway.) Thus, a person’s incidental similarity to a stranger can increase the likelihood of complying with the stranger’s request (Burger et al. 2004; Guéguen et al. 2005), of mimicking his or her nonverbal behaviors (Guéguen and Martin 2009), and of cooperating with the person (Miller, Downs, and Prentice 1998). In the consumer domain, a customer’s incidental similarity to a service provider can increase the customer’s liking for both the provider and the service received (Jiang et al. 2010). The aforementioned effects, however, occur when individuals either actually interact with the person to whom they are incidentally similar or imagine doing so. In contrast, the observers in our research were passive witnesses to an interchange between two persons with whom they were unacquainted and did not expect to interact. Consequently, they had little a priori interest in either party. Therefore, social connectedness is unlikely to account for the effect of incidental similarity in the situations we investigated. Rather, this effect seemed more likely to result from its influence on the relative attention that observers pay to the protagonists. People who witness an interchange between persons they do not know may often not be motivated to construe its implications. If they notice that a feature of one of the protagonists is similar to their own, however, their attention may be drawn to this person and they may be stimulated to think more carefully about the person’s role in the situation at hand. Incidental self-relevant information about a person (the person’s name, hometown, etc.) can elicit selective attention to this person (Bargh 1982; Moray 1959; Tacikowski and Nowicka 2010). In the situation we investigated, therefore, we expected that observers’ attention to a protagonist to whom they are incidentally similar would affect their interpretation of the service failure and their attributions of responsibility for its occurrence. The determinants of these attributions could be both cognitive and motivational. Cognitive Factors: Effects of Perspective Events are often attributed to features of the situation that are most likely to capture attention (Taylor and Fiske 1978). Thus, observers are likely to attribute an actor’s behavior to characteristics of the actor. In contrast, because actors’ attention is focused outward, they are more likely to attribute their behavior to characteristics of the situation or to other persons with whom they interact. These actor-observer differences are well established (Jones and Nisbett 1971). There are some constraints on their occurrence (Malle 2006), including the normative distinctiveness of the actor’s behavior (Hansen and Stonner 1978) and observers’ prior knowledge of the actor (Taylor, Peplau, and Sears 2006). In the conditions we considered, however, observers were unacquainted with the protagonists and their complaint behavior in the situation was not out of the ordinary. Similar considerations suggest that actors’ and observers’ attributional dispositions can be influenced by the perspective from which the actors’ behavior is viewed or imagined. Thus, showing actors a videotape of themselves interacting with another can increase their disposition to attribute their behavior to characteristics of themselves (Storms 1973). By the same token, observers who imagine an interaction from the perspective of one of the protagonists typically attribute responsibility for the events that occur to the other protagonist (Regan and Totten 1975). In the present context, this suggests that if observers imagine a service failure from the perspective of one of the protagonists, they are likely to attribute responsibility for the failure to the other protagonist. Motivational Factors Actor-observer differences in attribution can also depend on the favorableness of the behavior observed (Malle 2006). Individuals often exhibit a self-serving bias (Kunda 1987; Shaver 1985). That is, they typically take responsibility for positive outcomes of their behavior, but accept responsibility for negative outcomes only when no other plausible explanation exists (Arkin, Gleason, and Johnston 1976; Oishi, Wyer, and Colcombe 2000). This tendency, like that postulated by Jones and Nisbett (1971), suggests that observers who construe the implications of a service failure from the perspective of one of the protagonists are unlikely to attribute responsibility for the event to that protagonist; rather, they are inclined to blame the other protagonist. Effects of Protagonist Characteristics The preceding analyses, however, assume that the protagonists in the situation being observed have no distinguishing features that might be used as a basis for attributions. If one protagonist has an attribute or behavioral disposition that observers consider to be undesirable, an additional consideration comes into play. People not only dislike someone whose behavior or attributes elicit negative affective reactions, but also attribute other negative characteristics to the individual. Thus, people are more inclined to blame a negative event on someone they dislike than on someone they like (Regan et al. 1974; see also Alicke 2000). This suggests that if one of the parties involved in a service failure situation has an undesirable characteristic that elicits negative affective reactions, observers will attribute unfavorable characteristics and behavior to this individual and consequently will blame this person for the service failure that occurred. Moreover, the amount of attention they pay to this person is likely to increase their cognitive elaboration of the person’s involvement in the incident and to increase the extremity of their attribution of responsibility (Petty et al. 1995; Tesser 1978). A Two-Process Model The preceding discussion suggests that although observers’ incidental similarity to one of the protagonists in a service failure situation is likely to increase the attention they pay to that person, this attention could influence their attributions of responsibility in the different ways summarized in figure 1. If the protagonists have no distinguishing characteristics that could plausibly account for the service failure (figure 1A), observers’ relatively greater attention to the incidentally similar protagonist is likely to influence the perspective they take when they construe implications of the situation. That is, they construe the situation from the perspective of the protagonist to whom their attention is drawn, and attribute responsibility for the service failure to factors that are salient to them from this imagined perspective (e.g., the other protagonist). Thus: H1: When the customer and service provider involved in a service failure are strangers and have no distinctive judgment-relevant characteristics, observers’ incidental similarity to one of the protagonists will decrease their likelihood of blaming this protagonist for the failure and will increase their likelihood of blaming the other. There are qualifications on this hypothesis. That is, the effects of incidental similarity implied by hypothesis 1 are mediated by the effect of attention on the perspective that observers take when they construe the implications of the service failure (see figure 1A). If this is so, preventing them from thinking about the service failure from this perspective should eliminate these effects. Furthermore, different considerations arise when one of the protagonists has a characteristic that observers dislike or consider to be undesirable (figure 1B). In this case, observers are likely to attribute other undesirable characteristics to the protagonist as well and, therefore, to blame the service failure on this person. This may occur regardless of their incidental similarity to the protagonist. If they happen to be incidentally similar to the protagonist they consider to be responsible for the service failure, however, their relatively greater attention to this protagonist is likely to increase their cognitive elaboration of the service failure and the protagonist’s responsibility for it, and this increased elaboration is likely to increase the extremity of the observers’ attribution (Petty et al. 1995; Tesser 1978). H2: When observers perceive the behavior or personal characteristics of a protagonist in a service failure situation as undesirable, they will attribute greater responsibility for the failure to this protagonist than to the other, and this tendency will be greater when they are incidentally similar to the protagonist than when they are not. Note that the effects of incidental similarity implied by hypotheses 1 and 2 are both mediated by its impact on the attention paid to the similar protagonist. The effect of this increased attention differs, however, depending on whether the protagonist has undesirable features or not. Specifically, H3: The effect of observers’ incidental similarity on attributions of responsibility for a service failure implied by hypotheses 1 and 2 is mediated by its effect on the attention they pay to the protagonist to whom they are similar. To evaluate the implications of hypothesis 2, we considered two characteristics that seemed likely to influence attributions of a protagonist’s responsibility. One characteristic, rudeness, pertained to the protagonist’s behavior during the service failure situation. This factor was suggested by Jiang et al.’s (2010) finding that when participants have viewed a video of a personal trainer who was rude to other staff members, they report a more negative attitude toward the trainer when they are incidentally similar to him than when they are not. (For further evidence that individuals have greater dislike of persons who engage in undesirable behavior when they are similar to these persons, see Cooper and Jones 1969.) These effects could also occur when a protagonist’s undesirable attribute is irrelevant to the service failure situation and has little if any implications for the protagonist’s responsibility for the incident. To this end, we speculated that observers would be more inclined to blame a protagonist if he or she is obese. This characteristic is obviously irrelevant to the service failure situation itself. However, negative reactions to obese individuals are pervasive (Crandall 1994). They are often considered to be lazy, unhealthy, lacking in self-discipline, dishonest, unsuccessful, and stupid (Crandall 1994; Greenleaf et al. 2004; Puhl and Brownell 2001). Consequently, overweight employees are relatively less likely to get a pay raise and to be promoted (Pingitore et al. 1994), are less likely to receive appropriate treatment from physicians (Price et al. 1987), and are more likely to receive poor academic evaluations (Solovay 2000). In the consumer domain, customers have lower satisfaction and purchase intentions when they interact with an overweight service provider (Cowart and Brady 2014) and tend to avoid making the same choices that obese customers have made (McFerran et al. 2010). We therefore expected that observers would have greater inclination to blame an obese protagonist for a service failure than a protagonist of normal weight. Moreover, this disposition should be greater when the observers are incidentally similar to the protagonist than when they are not. Finally, the differences in attributions implied by the first three hypotheses have obvious implications not only for observers’ liking of the protagonists they observe but also for their evaluations of the service and their likelihood of patronizing the provider in the future. Specifically, H4: The effects of observers’ incidental similarity to the protagonists involved in a service failure on their attributions of responsibility for the failure influences their evaluations of the service and their willingness to patronize the provider in the future. Other Considerations Two other factors that might affect attributions in the conditions we investigated are worth noting. Threat to Self-Esteem If observers of a service failure take the perspective of a protagonist to whom they are incidentally similar, they might find negative attributes of this protagonist to be threatening to their self-esteem and might be motivated to dissociate themselves from this person, leading to effects of the sort implied by hypothesis 2. However, we expected the effects implied by hypothesis 2 to be independent of such threats. That is, observers might feel threatened by the undesirable behavior of a protagonist to whom they feel incidentally similar, but this threat does not necessarily mediate the effects of incidental similarity on attributions of responsibility that we identified. Empathy The disposition to take the perspective of a person who is incidentally similar could also be mediated by empathy. That is, individuals who take the perspective of another person often empathize with that person, vicariously experiencing the feelings that he or she is likely to have in the situation at hand (Batson et al. 1981). However, this is not always true. In competitive situations, for example, players often find it useful to imagine the situation from their opponent’s point of view in developing effective game-playing strategies without thinking about the other’s feelings. In short, empathy for another requires taking the other’s perspective, but taking another’s perspective does not necessarily imply empathy (Davis 1983). In several of the studies to be reported, we used a measure of empathy as a proxy for perspective taking in order to evaluate our assumptions about the effect of incidental similarity on perspective taking. However, we assumed that empathy per se was not itself a central factor in the effects we observed. Overview Four experiments examined the validity of our hypotheses and the processes that underlie them. Table 1 summarizes the causal relationships investigated in each experiment. Experiment 1, a field study, found that when observers of a service failure had the same surname as the waitress, they judged her to be less responsible for the failure and evaluated the service more favorably. When they had the same last name as the customer, however, the reverse was true. Experiment 2 corroborated these effects in a different situational context and confirmed that these effects are eliminated when participants are distracted from expending the cognitive resources necessary to think carefully about the service failure. Experiments 3 and 4 confirmed the mediating effect of attention on the impact of incidental similarity and showed that when one of the protagonists in the service failure situation had a negative attribute, incidental similarity of this protagonist increased attributions of blame to this protagonist rather than decreasing it. As table 1 indicates, no single study examined all of the direct and mediating influences of incidental similarity on attributions of responsibility and patronage intentions that we hypothesized. In combination, however, they provided a fairly complete picture of the phenomena we investigated. TABLE 1 SUMMARY OF CAUSAL RELATIONS EVALUATED IN EXPERIMENTS 1–4 Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions TABLE 1 SUMMARY OF CAUSAL RELATIONS EVALUATED IN EXPERIMENTS 1–4 Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions Experiment Conclusion Experiment 1 Incidental similarity → attribution of blame → patronage intentions Experiment 2 Incidental similarity × cognitive load → attribution of blame → patronage intentions Experiment 3 Incidental similarity → rudeness × attention→ attributions of blame→ patronage intentions Experiment 4 Incidental similarity → obesity × attention→ attributions of blame → patronage intentions Pretests of the different scenarios we used in experiments 2 through 4 confirmed that participants assigned similar levels of responsibility for the failure to both the service provider and the customer. In addition, a supplementary study ruled out the possibility that the effects of incidental similarity in the conditions we investigated were mediated by feelings of social connectedness. These studies are described in the web appendix. EXPERIMENT 1 This experiment evaluated hypothesis 1 in a field setting. Restaurant customers observed a service failure involving another customer and a waitress, and their incidental similarity to one of the protagonists was unobtrusively manipulated. We expected that participants would blame the waitress less for the failure when their name was similar to that of the waitress than when it was similar to that of the customer. Method Participants were asked to apply for a “mystery shopper” experiment three weeks before the study. Their last names were identified through their applications. Persons whose last names were particularly uncommon were not recruited, leaving 70 participants in the actual experiment. These participants (41.7% males between 20 and 35 years of age) were paid HK$100 (approximately US$13) to take part in a mystery shopper study. They were randomly assigned to cells of a 3 (incidental similarity: provider similarity vs. customer similarity vs. control) between-subjects design. The experiment was conducted at a small local restaurant in Hong Kong. The restaurant was newly opened and the experiment (with the owner’s permission) took place on 15 weekdays between 2:00 p.m. and 6:30 p.m., when the restaurant traffic was minimal. The size of the restaurant was small (around 900 square feet, with 15 tables) and customers could easily see and overhear the interactions between the waiter and the other customer at the next table. Each participant, run individually, arrived at the restaurant in a particular time slot. No participant had been to the restaurant before. They were invited to be mystery shoppers and to evaluate the restaurant’s afternoon tea menu and services. Each participant was given a coupon of HK$100 to spend on food and beverages. When participants arrived, they were asked to sit at a table in a corner of the restaurant to minimize the disturbance from other customers and waiters. Two female confederates (age = 22 years) were recruited to be a waitress and a target customer. In provider similarity conditions, the waitress wore a name tag with a last name that either matched or did not match the participant’s. Furthermore, the waitress mentioned her name to the participants when she greeted them and directed them to a table. In customer similarity conditions, a “reserved” sign with a last name that either matched or did not match the participant’s was put on the table next to the participant’s. The female confederate pretending to be a customer entered the restaurant after the participant was seated. To ensure that each participant noticed the name of the confederate, the waitress referred to “Ms. ____” when she greeted the customer and directed her to the table next to the participant’s. In control conditions, neither the last name on the waitress’s name tag nor the reserved table stand matched the name of the participant. Participants received their order within 10 minutes and the confederate received hers 5 minutes later. Upon receiving it, however, she complained to the waitress that her food was mixed with onions. (Although this might not be universally considered to be a “service failure,” we assumed that observers in the situation we constructed would interpret the customer’s complaint as either an unjustified response that reflected the customer’s idiosyncratic taste or a justified criticism of the way the restaurant prepared this particular dish.) The waitress volunteered to replace the dish and said it would take approximately 15 minutes. The confederate looked unhappy and told the waitress to cancel the order, as she had to leave. Upon leaving the restaurant, participants completed an online questionnaire on their mobile phone. First, they evaluated the overall quality of the service along three scales from 1 (bad/undesirable/unfavorable) to 10 (good/desirable/favorable). Responses to these items were averaged (α =.93). They then evaluated the quality of the food and the physical environment along similar scales (α =.92 and .95, respectively). Finally, participants were asked whether they had personally experienced a service failure at the restaurant (all participants responded negatively) and whether they had observed any service failure that involved another customer. Two participants did not notice that a service failure had occurred and were excluded from the subsequent data analysis. The remaining participants indicated whether the waitress was responsible by responding to three items (“You blame the waitress for the problem you observed,” “You think the waitress was responsible for the problem you observed,” and “You think that the problem described was not in the control of the waitress” (reverse-scored) along a scale from 1 (strongly disagree) to 10 (strongly agree) (α =. 92). Results and Discussion We expected that incidental similarity would affect participants’ overall service evaluations through its mediating impact on attributions of responsibility. These expectations were confirmed. Participants’ evaluations of the restaurant in terms of food, physical environment, and service are summarized in table 2. The effect of incidental similarity on overall service evaluations was significant (F(2, 65) = 18.37, p < .001, ηp2 = .36). Participants rated the service more favorably when they were incidentally similar to the service provider than they did in control conditions (7.17 vs. 6.06, respectively; F(1, 65) = 5.17, p < .01, d = .98) but rated it less favorably when they were incidentally similar to the customer than in control conditions (5.10 vs. 6.06, respectively; F(1, 65) = 3.84, p < .05, d = .86). In contrast, participants’ ratings of food quality or the physical environment did not differ across three conditions (t < 1). TABLE 2 THE IMPACT OF INCIDENTAL SIMILARITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 1 Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) NOTE.—Standard deviations are in parentheses. Means in each row with unlike subscripts significantly differ at p < .05. TABLE 2 THE IMPACT OF INCIDENTAL SIMILARITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 1 Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) Similar to provider Similar to customer Control Service evaluation 7.17a (1.21) 5.10b (1.15) 6.06c (1.11) Food evaluation 6.94a (1.14) 7.09a (1.22) 7.11a (1.12) Physical environment 6.26a (1.20) 6.16a (1.26) 6.32a (1.09) Attributions of blame to service provider 5.38a (1.22) 7.39b (1.10) 6.55b (1.23) NOTE.—Standard deviations are in parentheses. Means in each row with unlike subscripts significantly differ at p < .05. The effect of incidental similarity on attributions of responsibility for the service failure was also significant (F(2, 65) = 16.68, p < .001, ηp2 = .34) and paralleled its effect on service evaluations. Participants were less likely to blame the waitress when they were incidentally similar to her than they were in control conditions (5.38 vs. 6.55, respectively; F(1, 65) = 5.44, p < .01, d = .98) but were marginally more likely to blame the waitress when they were similar to the customer than in control conditions (7.39 vs. 6.55, respectively; F(1, 65) = 2.85, p = .065, d = .74). Responsibility attributions significantly mediated the effects of incidental similarity on service evaluations. Based on a bootstrapping analysis (model 4, Hayes 2018) with 5,000 samples, the 95% CI ranged from –.7683 to –.0512. In summary, experiment 1 provided evidence that observers’ incidental similarity to the protagonists in a service failure situation affected both their attributions of responsibility for the failure and evaluations of the overall quality of the service in general. When observers were incidentally similar to the service provider, they attributed less responsibility to the provider and evaluated the service quality more favorably. When they were incidentally similar to the customer, however, they attributed more responsibility to the provider and evaluated the service less favorably. EXPERIMENT 2 Participants in experiment 1 observed a failure that occurred in the same consumption situation in which they were present. Experiment 2 examined whether a similar effect would occur when observers were not involved in the situation at all. We used a trip advisor website as a context and manipulated the incidental similarity of observers’ last name to that of either a hotel manager or a customer. We also assessed attributions of blame to the customer as well as to the service provider and confirmed these attributions on the basis of thought listing data. We also showed that observers did not think much about the service failure situation when an incidental similarity was not present. In addition, to ensure that the effect of an incidental similarity was not mediated by its impact on the positive affect that participants experienced, participants’ moods were measured. Furthermore, we obtained support for the assumption that the effect of incidental similarity was mediated by the attention paid to the similar protagonist. If this assumption is valid, situational factors that distract observers from differentially attending to the protagonists should decrease their ability to engage in this processing and should consequently decrease or eliminate differences in their attributions of responsibility for the failure that otherwise occurs. This was in fact the case. Method To collect participants’ last name information, we had 288 Hong Kong participants complete an unrelated survey on the university network two weeks before the main study. A section of the survey requested demographic information, including their last name. After eliminating participants whose last names were unlikely to be shared, we selected 192 participants (42.7% males) to participate in a 15 minute study for HK$20 (approximately US$3). They were randomly assigned to cells of a 3 (incidental similarity: provider similarity vs. customer similarity vs. control) × 2 (cognitive load: low vs. high) between-subjects design. Participants performed two ostensibly unrelated tasks. First, on the pretense that we were interested in people’s ability to perform two tasks at once, we asked them to memorize either a 10-digit (high load) or 2-digit (low load) number that they would be asked to recall after they performed a second task in the experiment (Shiv and Fedorikhin 1999). In this task, they viewed a trip advisor online review of a hotel service (see the web appendix). The review contained a dialogue that described a customer’s complaint about the service and the hotel manager’s response. The customer complained that he had made a reservation for a nonsmoking sea view room on the hotel’s website two weeks earlier but that when he arrived at the hotel, he found that the room was unavailable. The hotel manager responded that he had checked the computer system and did not see the customer’s reservation and that only nonsmoking rooms without a sea view were available, as it was during the peak season. The dialogue conveyed that participants had the same last name as the manager in provider similarity conditions, and the same last name as the customer in customer similarity conditions. In control conditions, all names differed. Participants reported the likelihood that they would visit the hotel along three scales from 1 (very unlikely/inclined not to/definitely will not) to 10 (very likely/inclined to/definitely will). They also indicated their inclination to blame both the manager and the customer (i.e., “You blame the manager/customer for the problem you observed,” “You think the manager/customer was responsible for the problem you observed,” and “You think that the problem described was not in the control of the manager/customer” (reverse-scored) along a scale from 1 (strongly disagree) to 10 (strongly agree). Responses to each set of items were averaged (α > .90). Participants reported their mood (α = .91) along three scales from 1 (sad/bad/depressed) to 10 (happy/good/cheerful). After that, they listed the thoughts that went through their minds while they were reading the hotel scenario. Finally, to access the validity of cognitive load manipulation, we had them respond to two items (r > .68): “I found it challenging to read the scenario while trying to remember the number” and “Remembering the number was easy” (reverse-scored) along a scale from 1 (strongly disagree) to 10 (strongly agree). Results The effects of incidental similarity on judgments are summarized in table 3. We expected that incidental similarity would influence inclinations to patronize the establishment under low-cognitive-load conditions and that this influence would be mediated by attributions of responsibility for the incident’s occurrence. Under high-load conditions, however, we expected this influence to be minimal. TABLE 3 THE IMPACT OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 2 Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. TABLE 3 THE IMPACT OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 2 Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) Similar to provider Similar to customer Control Patronage intention  High cognitive load 5.48a (1.37) 5.11a (1.41) 5.25a (1.26)  Low cognitive load 6.39b (1.45) 4.41c (1.40) 5.13a (10.20) Intention to blame service provider  High cognitive load 5.44a (.94) 5.75a (.95) 5.60a (1.27)  Low cognitive load 4.65b(1.26) 6.68c (1.38) 5.49a (1.35) Intention to blame customer  High cognitive load 5.51a (1.21) 5.17a (1.26) 5.21a (1.29)  Low cognitive load 6.42b (1.47) 4.14c (1.29) 5.27a (1.28) Mood  High cognitive load 5.83a (1.38) 6.01a (1.37) 5.71a (1.14)  Low cognitive load 6.49b (1.13) 6.65b (1.24) 5.81a (1.13) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. Manipulation Check As expected, participants perceived the situation as more difficult to process under high load than under low load (7.02 vs. 2.78, respectively; F(1, 186) = 348.50, p < .001, ηp2 = .65). Patronage Intentions The main effect of incidental similarity and its interaction with cognitive load were significant (F(2, 186) = 12.34, p < .001, ηp2 = .12 and F(2, 186) = 5.86, p < .01, ηp2 = .06, respectively). Between-cell comparisons shown in table 3 indicate that in low-load conditions, incidental similarity to the provider increased participants’ inclination to visit the hotel (from 5.13 to 6.39; F(1, 186) = 27.95, p < .001, ηp2 = .13), but similarity to the customer decreased it (from 5.13 to 4.41; F(1, 186) = 9.64, p < .01, ηp2 = .05). In the high-load conditions, however, no differences were found between provider similarity conditions (M = 5.48) and either customer similarity conditions (M = 5.11) or control conditions (M = 5.25). Responsibility Attributions Inclinations to blame the service provider and inclinations to blame the customer were analyzed separately. Each analysis yielded a main effect of incidental similarity (F(2, 186) = 15.18, p < .001, ηp2 = .14 and F(2, 186) = 16.28, p < .001, ηp2 = .15, respectively) and an interaction of similarity and cognitive load (F(2, 186) = 8.23, p < .001, ηp2 = .08 and F(2, 186) = 8.85, p < .001, ηp2 = .09, respectively). Between-cell comparisons shown in table 3 indicate that in low-load conditions, incidental similarity to the provider decreased participants’ disposition to blame the provider relative to control conditions (from 5.49 to 4.65; F(1, 186) = 16.18, p < .001, ηp2 = .08) and increased their disposition to blame the customer (from 5.27 to 6.42; F(1, 186) = 25.50, p < .001, ηp2 = .12). In contrast, incidental similarity to the customer increased attributions of responsibility to the provider (from 5.49 to 6.68; F(1, 186) = 31.79, p < .001, ηp2 = .15) and decreased attributions of responsibility to the customer (from 5.27 to 4.14; F(1, 186) = 24.64, p < .001, ηp2 = .12). In contrast, attributions of blame in high-load conditions did not significantly vary over the three similarity conditions. Thought Listings The different perspectives that observers took in incidental similarity conditions, and their consequent attributions of responsibility for the service failure, are reflected in their thought listings. Two coders who were blind to experimental conditions coded each participant’s thoughts in terms of whether they predominantly conveyed the perspective of the provider (e.g., “The manager has no control over the technical problem of the booking system. He has tried his best to satisfy the customer”), the customer (e.g., “I can feel how unhappy the customer is. It is very clear that the manager should send a confirmation”), both (e.g., “Both of them have responsibilities . . .”) or neither. Interjudge agreement was 91.1% and differences were resolved through discussion. The number and proportions of participants whose thought reflected each perspective are summarized in the top section of table 4. A log linear analysis under low-load conditions yielded a significant interaction of incidental similarity and thought type χ2(6) = 77.56, p < .001. Under low-load conditions, 27 of 32 participants in provider-similarity conditions (84%) took the provider’s perspective (χ2(3)= 60.25, p < .001) and 24 of 32 participants in customer-similarity conditions (75%) took the customer’s perspective (χ2(3) =42.25, p < .001). In control conditions, these proportions were low and did not appreciably differ (χ2(3) =3.75, p >.10). In contrast, a similar analysis under high-load conditions yielded only a main effect of thought type χ2(5) = 43.42, p < .001; the majority of participants’ thoughts in these conditions did not reflect either perspective. TABLE 4 PROPORTION OF PERSPECTIVE-RELATED THOUGHTS AND ATTRIBUTION-RELATED THOUGHTS AS A FUNCTION OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD—EXPERIMENT 2 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 NOTE.—Numbers in parentheses indicate the percentage of the thoughts. TABLE 4 PROPORTION OF PERSPECTIVE-RELATED THOUGHTS AND ATTRIBUTION-RELATED THOUGHTS AS A FUNCTION OF INCIDENTAL SIMILARITY AND COGNITIVE LOAD—EXPERIMENT 2 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 A. Perspective-related thoughts about Provider Customer Both Neither χ2 High-load conditions  Similar to provider 8 (.25) 3 (.09) 4 (.13) 17 (.53) 15.25, p <.01  Similar to customer 5 (.16) 9 (.28) 3 (.09) 15 (.47) 10.50, p <.05  Control 3 (.09) 5 (.16) 4 (.13) 20 (.62) 24.25, p <.001 Low-load conditions  Similar to provider 27 (.84) 1 (.03) 2 (.06) 2 (.06) 60.25, p <.001  Similar to customer 3 (.09) 24 (.75) 4 (.13) 1 (.03) 42.25, p <.001  Control 5 (.16) 6 (.19) 12 (.37) 9 (.28) 3.75, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 B. Attributions of responsibility To provider To customer To both To situation To neither χ2 High-load conditions  Similar to provider 3 (.09) 4 (.13) 4 (.13) 5 (.16) 16 (.50) 18.31, p <.01  Similar to customer 9 (.28) 4 (.13) 3 (.09) 1 (.03) 15 (.47) 10.50, p <.05  Control 4 (.13) 3 (.09) 1 (.03) 5 (.16) 19 (.59) 32.38, p <.001 Low-load conditions  Similar to provider 1 (.03) 25 (.78) 2 (.06) 2 (.06) 2 (.06) 67.69, p <.001  Similar to customer 24 (.75) 1 (.03) 3 (.09) 3 (.09) 1 (.03) 61.13, p <.001  Control 6 (.19) 3 (.09) 10 (.31) 6 (.19) 7 (.22) 3.94, p >.10 NOTE.—Numbers in parentheses indicate the percentage of the thoughts. Observers who took one protagonist’s perspective were expected to attribute responsibility for the service failure to the other protagonist. To confirm this assumption, judges recoded participants’ thoughts in terms of whether they reflected an attribution of responsibility to the provider (e.g., “It is very clear that the manager should send a confirmation. This is his responsibility”), the customer (e.g., “I think the customer should call the manager to reconfirm the booking”), both (e.g., “Both of them have responsibility for the incident”), the situation (e.g., “No one should be blamed. This is the problem of the system”), or neither. Interjudge agreement was 93.2% and differences were resolved through discussion. Results are summarized in the second half of table 4. A log linear analysis under low-load conditions yielded a significant interaction of incidental similarity and thought type χ2(8) = 85.42, p < .001. When participants were not under cognitive load, 25 of 32 participants in provider-similarity conditions (78%; χ2(4) = 67.69, p < .001) attributed responsibility to the customer, whereas 24 of 32 participants in customer-similarity conditions (75%; χ2(4) = 61.13, p < .001) attributed responsibility to the provider. Responsibility attributions in control conditions did not differ. Moreover, a similar analysis under high-load conditions yielded only a main effect of thought type χ2(6) = 61.33, p < .001. Most participants in these conditions did not attribute beliefs to either protagonist. Mediation Bootstrapping analyses (model 8, Hayes 2018) were consistent with the assumption that under low-load conditions, the effect of incidental similarity on attributions of blame to the provider was mediated by perspective-related thoughts about the provider (based on 5,000 samples, 95% CI ranged from –3.2671 to –.5665) and that its effect on attributions of blame to the customer was mediated by perspective-related thoughts about the customer (95% CI ranged from –1.5700 to –.3929). Under high-load conditions, however, the corresponding effects of perspective-related thoughts were not significant; the 95% CI ranged from –.3134 to .7379 in the first case and from –.1255 to .3315 in the second. Mood Participants reported being happier when they were either similar to the provider (M = 6.16) or similar to the customer (M = 6.33) than they were in control conditions (M = 5.76); F(2, 186) = 3.58, p < .05, ηp2 = .04. However, mood was not a significant mediator of the effect of similarity on responsibility attributions (95% CI, from –.0261 to .1095). Discussion Experiment 2 provided evidence that attention mediated the effect of incidental similarity on observers’ attributions of responsibility. Observers under low-load conditions were less inclined to blame a service provider for a service failure when they were incidentally similar to him, but were more inclined to do so when they were incidentally similar to the customer. In this case, their attributions of responsibility mediated their willingness to patronize the hotel. The effect of incidental similarity was not evident, however, when participants were under cognitive load and were less able to think carefully about the situation they observed. Thought listing data confirmed these processes. In control conditions, thoughts about the interaction were generally low, indicating that disinterested observers did not think much about the service failure situation. When incidental similarity was manipulated, participants’ thoughts largely pertained to the protagonist to whom they were incidentally similar and correspondingly reflected a disposition to blame the other protagonist for the service failure. When participants were under cognitive load, however, their attention was apparently diverted from the service failure situation and their incidental similarity to the protagonists had little effect on their thoughts about it. (Note that the lack of effect in high-load conditions provides indirect evidence that the processes underlying the attributions of responsibility were conscious and deliberate. Had they been automatic, cognitive load would have had little effect.) Finally, although participants reported being in a better mood when they discovered that they were incidentally similar to one of the protagonists, mood did not mediate the effect of similarity on responsibility attributions. EXPERIMENT 3 Results of the first two experiments are consistent with our assumption that when the protagonists in a service failure situation have no distinctive features that have implications for judgment, an incidental similarity to one of the protagonists increases their attention to the protagonist and leads them to take this protagonist’s perspective when they interpret the situation. Consequently, they decrease attributions of responsibility for the service failure to this protagonist for reasons similar to those suggested by Jones and Nisbett (1971). An additional experiment we conducted (see web appendix) confirmed our assumption that this effect was mediated by an increase in the attention that observers paid to the incidentally similar provider. Experiments 3 and 4 provided further support for this assumption. Moreover, they showed that if one of the protagonists has a salient personal attribute or engages in behavior that observers consider to be undesirable, the observers are likely to attribute responsibility for a negative event to this protagonist regardless of other considerations (Alicke 2000; Regan et al. 1974). In this case, an incidental similarity that draws observers’ attention to this protagonist may strengthen this attribution rather than decreasing it. Experiment 3 determined the effects of incidental similarity to a protagonist who engaged in rude behavior during the service failure situation. Two parallel studies with different contexts were conducted, one in which the service provider was rude (experiment 3a) and the other in which the customer was rude (experiment 3b). Because both studies yielded similar conclusions, we report experiment 3a here and relegate the details of the other subexperiment to the web appendix. In addition, participants’ perceptions of similarity were measured. We expected that control participants would perceive themselves as more similar to the protagonist when they were incidentally similar with the protagonist than when they were not. When a protagonist has an undesirable attribute, however, we expected that observers’ increased attention to the protagonist that was stimulated by their incidental similarity would not have these effects. Experiment 3a Method In this experiment, we used first names to manipulate the similarity between observers and the service provider. Two weeks before the main study, 220 participants completed a survey identical to that used in experiment 2. Of these, 148 participants (40.5% males) whose first names were moderately uncommon participated in a 15 minute study with HK$20 (approximately US$3). They were randomly assigned to cells of a 2 (service provider similarity: present vs. absent) × 2 (rudeness: present vs. absent) between-subjects design. Participants read a passage in which they were told to imagine that they visited a sporting goods store to buy some sportswear and, while looking around, they overheard a conversation between a customer and a salesperson who was referred to by name (either the same as the participant’s or not (see the web appendix). In the story, the customer asks the salesperson to bring him a pair of running shoes that he had ordered online the day before. However, the salesperson tells him that the shoes have not yet arrived. The customer complains that he had not been told that it would take more than one day for delivery, but the salesperson responds that a note on the store website indicates that the delivery will take one to two days. In the rudeness condition, the salesperson was described as looking impatient and “barking” at the customer when he mentioned the website information about the delivery date. In the control condition, the salesperson’s manner was not mentioned. Then, participants reported their intention to buy the product at the store along scales from 1 (very unlikely/inclined not to/definitely will not) to 10 (very likely/inclined to/definitely will). They also responded to six items concerning their inclination to blame the salesperson and the customer along a scale from 1 (strongly disagree) to 10 (strongly agree). These items were identical to those used in experiment 2 except that the provider was referred to as a “salesperson” rather than the “manager” (α > .90 in each case). In addition, participants reported the extent to which they empathized with both the salesperson’s situation and the customer’s situation along scales from 1 (not at all/definitely no) to 10 (very much/definitely yes). Responses to each set of items were averaged (α > .92 in each case). Furthermore, they responded to three items concerning the extent to which they had paid attention to the salesperson (“To what extent did you pay close attention to the salesperson?” “To what extent did you focus your attention on the salesperson?” and “To what extent did the salesperson grab your attention?” along scales from 1 (little) to 10 (a lot); α = .91. They also indicated their similarity to the salesperson along a scale from 1 (very dissimilar) to 10 (very similar) and their perception of the salesperson’s rudeness along a scale from 1 (not rude at all) to 10 (very rude). Experiment 3a Results The results of this experiment are summarized in table 5. TABLE 5 THE IMPACT OF INCIDENTAL SIMILARITY AND SERVICE PROVIDER’S RUDENESS ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 3A Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. TABLE 5 THE IMPACT OF INCIDENTAL SIMILARITY AND SERVICE PROVIDER’S RUDENESS ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—EXPERIMENT 3A Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) Similar provider Nonsimilar provider Attention  Rudeness 7.14a (1.16) 5.43b (1.04)  Control 6.95a (1.27) 5.14b (.95) Perceived similarity  Rudeness 2.27a (1.07) 2.68b (.99)  Control 5.78c (1.16) 3.30d (.97) Patronage intention  Rudeness 2.97a (1.19) 3.68b (1.03)  Control 5.97c (1.17) 4.86d (1.08) Inclination to blame provider  Rudeness 7.32a (1.28) 6.41b (.82)  Control 4.65c (1.21) 5.80d (.99) Inclination to blame customer  Rudeness 3.21a (.94) 3.46a (1.12)  Control 6.15b (1.07) 5.53c (1.29) Empathy for provider  Rudeness 3.23a (1.01) 3.92b (1.07)  Control 6.41c (1.43) 5.32d (1.11) Empathy for customer  Rudeness 6.45a (1.08) 6.64a (.96)  Control 5.04b (1.15) 6.01c (1.15) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at p < .05. Manipulation Check Participants perceived the salesperson to be more rude in rudeness conditions than in control conditions (7.37 vs. 3.55, respectively; F(1, 144) = 324.59, p < .001, ηp2 = .69). Attention The effects of incidental similarity and rudeness conditions on attention and other dependent variables are shown in table 5. Participants’ attention to the salesperson increased with incidental similarity (from 5.28 to 7.04; F(1, 144) = 92.58, p < .001, ηp2 = .39). No other effects were significant. Perceived Similarity Both rudeness and incidental similarity affected participants’ perceptions of similarity to the provider (F(1, 144) = 144.70, p < .001, ηp2 = .50 and F(1, 144) = 36.65, p < .001, ηp2 = .20, respectively). However, the interaction of these variables was also significant (F(1, 144) = 70.77, p < .001, ηp2 = .33). As shown in table 5, incidental similarity increased perceptions of similarity to the provider in control conditions (from 3.30 to 5.78; F(1, 144) = 209.69, p < .001, ηp2 = .59), but it decreased these perceptions when the provider was rude (from 2.68 to 2.27; F(1, 144) = 6.62, p < .05, ηp2 = .04). Patronage Intentions Unlike previous experiments in which the protagonists were nondescript, we expected that incidental similarity to a rude protagonist would increase attributions of responsibility to this protagonist and that these attributions would be reflected in patronage inclinations. Analyses of participants’ patronage inclinations yielded a main effect of service provider rudeness and an interaction of provider similarity and rudeness (F(1, 144) = 129.59, p < .001, ηp2 = .47 and F(1, 144) = 24.21, p < .001, ηp2 = .14, respectively). In control conditions, participants’ incidental similarity to the salesperson increased the inclination to patronize the store (from 4.86 to 5.97; F(1, 144) = 37.18, p < .001, ηp2 = .21). When the salesperson was rude, however, incidental similarity decreased patronage inclinations (from 3.68 to 2.97; F(1, 144) = 15.69, p < .001, ηp2 = .10). Attributions of Responsibility The effects of the provider’s incidental similarity and rudeness on intentions to blame the provider paralleled their effects on patronage intentions. The main effect of rudeness and its interaction with incidental similarity were significant (F(1, 144) = 83.24, p < .001, ηp2 = .37 and F(1, 144) = 33.13, p < .001, ηp2 = .19, respectively). Incidental similarity to the salesperson decreased perceptions that she or he was to blame in control conditions (from 5.80 to 4.65; F(1, 144) = 41.99, p < .001, ηp2 = .23) but increased these perceptions in rudeness conditions (from 6.41 to 7.32; F(1, 144) = 26.61, p < .001, ηp2 = .16). Analyses of attributions of responsibility to the customer also yielded a main effect of provider rudeness and an interaction of rudeness with incidental similarity (F(1, 144) = 188.16, p < .001, ηp2 = .57 and F(1, 144) = 5.70, p < .05, ηp2 = .04, respectively). Incidental similarity to the provider increased attributions of responsibility to the customer in control conditions (from 5.53 to 6.15; F(1, 144) = 12.31, p < .01, ηp2 = .08) but not when the provider was rude (3.46 vs. 3.21, respectively; F(1, 144) = 2.68, p > .10, ηp2 = .02). Empathy (Perspective) We inferred the perspective that observers took when they construed the implications of the service failure from their estimates of their empathy. We expected that in control conditions, their empathy for a protagonist would increase with their incidental similarity to him or her but that in rudeness conditions, this would not be the case. Observers’ empathy for each protagonist was analyzed separately and yielded analogous results. The provider’s rudeness decreased empathy for the provider (from 5.87 to 3.57; F(1, 144) = 142.20, p < .001, ηp2 = .50) and increased empathy for the customer (from 5.53 to 6.54; F(1, 144) = 31.98, p <.001, ηp2 = .18). In addition, incidental similarity and the provider’s rudeness had interactive effects on empathy for both the provider (F(1, 144) = 21.23, p < .001, ηp2 = .13) and the customer (F(1, 144) = 4.78, p <.05, ηp2 = .03). Table 5 shows that in control conditions, incidental similarity to the provider increased empathy for the provider (from 5.32 to 6.41; F(1, 144) = 32.94, p <.001, ηp2 = .19) and decreased empathy for the customer (from 6.01 to 5.04; F(1, 144) = 30.10, p <.001, ηp2 = .17), consistent with the assumption that it increased the disposition to take the perspective of the provider in this condition. When the provider was rude, however, incidental similarity to him decreased empathy for him (from 3.92 to 3.23; F(1, 144) = 13.64, p <.001, ηp2 = .09) and did not appreciably affect empathy for the customer (6.64 vs. 6.45, respectively; F(1, 144) = 1.66, p >.10, ηp2 = .01), suggesting that observers avoided taking the provider’s perspective in this case. Mediation Bootstrapping analyses (model 91, Hayes 2018) confirmed the mediating effects of attention and responsibility attributions on the interactive effects of the provider’s rudeness and incidental similarity. Specifically, the causal model: incidental similarity → attention × rudeness → attributions of blame to provider → patronage inclinations was reliable (based on 5,000 bootstrapping samples, 95% CI from –2.0736 to –1.0381). However, an alternative model in which the causal sequence of blame and attention was reversed was not supported (95% CI from –.0636 to .0997). In addition, a serial mediation analysis (model 6, Hayes 2018) that included empathy as a mediator was confirmed in the control condition: incidental similarity → attention → empathy for provider → attributions of blame to provider → patronage inclinations (95% CI from .0071 to .2318) but not in the rudeness condition (95% CI from –.0936 to .0392). Experiment 3b A similar experiment confirmed analogous effects of rudeness when the customer in a service failure situation was rude. Birth month was used to manipulate incidental similarity. One hundred twenty participants (49.2% males) imagined that they viewed a video on YouTube depicting a customer’s interaction with a restaurant manager. The customer indicated that he had made a reservation and requested a sea-view table one week ago through the restaurant’s website. However, the manager replied he checked the reservation system and did not see the customer’s special request. In the rudeness condition, the customer screamed and shouted loudly at the manager when he learned that he could not get the sea-view table. In the control condition, the customer’s behavior was not mentioned. In the customer similarity conditions, the dialogue showed that the observers and the customer had the same birth month, whereas in no-similarity condition, this was not the case. In this experiment, incidental similarity and the customer’s rudeness had interactive effects on perceived similarity to the customer (F(1, 116) = 58.46, p < .001, ηp2 = .34), attributions of responsibility (F(1, 116) = 6.75, p < .05, ηp2 = .06), and patronage intentions (F(1, 116) = 23.48, p < .001, ηp2 = .17). Specifically, in control conditions, an incidental similarity to the customer increased observers’ perceptions of their similarity to him (from 3.37 to 5.67; F(1, 116) = 151.51, p < .001, ηp2 = .57), increased their attributions of responsibility to the provider (from 5.68 to 6.52; F(1, 116) = 19.97, p <.001, ηp2 = .15), and decreased their inclination to patronize the establishment (from 4.64 to 2.99; F(1, 116) = 59.82, p <.001, ηp2 = .34). When the customer was rude, however, incidental similarity to him decreased their perceptions of their similarity to him (from 2.27 to 1.70; F(1, 116) = 10.20, p < .01, ηp2 = .08), increased their disposition to blame him (from 5.89 to 6.40; F(1, 116) = 5.87, p < .05, ηp2 = .05), and increased their patronage intentions (from 5.07 to 5.49; F(1, 116) = 4.55, p < .05, ηp2 = .04). (For more details, see the web appendix.) Discussion Experiment 3 confirmed the hypothesis that when a protagonist in a service failure situation engages in undesirable behavior, an incidental similarity to the protagonist increases rather than decreases attributions of responsibility to this protagonist. Although control participants perceived themselves as more similar to the protagonist when they were incidentally similar to the protagonist than when they were not, this difference was reversed when the protagonist was rude. The different effects of incidental similarity on empathy that occurred under rudeness and control conditions confirm our assumption that the processes underlying the effect of incidental similarity do not operate independently. Rather, the effects of incidental similarity on the disposition to take the perspective of an incidentally similar protagonist do not occur if the protagonist has a salient distinguishing characteristic. If this had occurred, incidental similarity would have had similar effects in both the rudeness and control conditions. In fact, although an incidental similarity to the provider increased empathy for him in control conditions, it significantly decreased empathy for him when he was rude. Moreover, similarity to the provider decreased empathy for the customer under control conditions but had no effect when the provider was rude. The effect of empathy in control conditions was presumably influenced by participants’ incidental similarity to the protagonist and was a determinant of their attention to him. When the protagonist was rude, however, empathy was not a determinant of their attention to the protagonist but rather was a consequence of participants’ negative reactions to this behavior. Finally, note that the influence of incidental similarity to a protagonist on observers’ judgments of their general similarity to him depended on whether protagonist was rude or not. As figure 1 indicates, however, these perceptions do not enter into the conceptualization we propose. That is, observers’ incidental similarity to the protagonist directs their attention to him, and this attention leads them either to take the protagonist’s perspective or, if the protagonist is rude, to think more extensively about the protagonist’s responsibility for the service failure. Although perceived similarity may be a correlate of these effects, it does not directly influence the phenomena we predict. EXPERIMENT 4 Experiment 4 determined whether an undesirable feature of a protagonist that is irrelevant to the situation observed (specifically, the protagonist’s obesity) could have similar effects. Two substudies with different contexts were again conducted, one (experiment 4a) in which the customer’s obesity was manipulated and the other (experiment 4b) in which the service provider’s obesity was manipulated. However, because the results of the two studies have virtually identical conclusions, the details of experiment 4b are reported in the web appendix. Experiment 4a Method In this experiment, the obesity of the customer rather than of the provider was varied. Moreover, we obtained a measure of self-threat to determine whether this factor played a role in the effects we observed. We used first names to manipulate incidental similarity. We invited 210 female participants to perform a survey identical to experiment 2 two weeks before the main study. Of these, 160 whose first names were not extremely uncommon participated in a 15 minute study for HK$20 (approximately US$3). They were randomly assigned to cells of a 2 (customer similarity: present vs. absent) × 2 (customer obesity: present vs. absent) between-subjects design. Participants imagined themselves going to a restaurant for dinner. On this pretense, their attention was directed to a computer screen showing a picture of a male waiter and a female customer. Participants saw only the back of a female customer, who sat at a table with the male waiter facing her. The customer was portrayed by either an overweight female model or an average weight model (age = 26). An average-looking male confederate (age = 28) portrayed the waiter. A pretest using a between-subject design (n = 60) confirmed that the female confederate was perceived to be more obese than in control conditions (8.10 vs. 4.67, respectively, along a scale from 1 (not fat at all) to 10 (very fat); F(1, 58) = 226.06, p < .001, ηp2 = .80). The pretest also confirmed the male confederate was average-looking, along a scale from 1 (not handsome at all) to 10 (very handsome); (5.23 vs. 5.37, respectively along a scale from 1 to 10; F < 1). Finally, the customer was referred to by a name that was either the same as or different from the participant’s (see the web appendix). Participants were asked to imagine overhearing a conversation between the two protagonists. In each case, the customer complained to the waiter about waiting a long time for a dessert special. The waiter explained that it took time to prepare the dessert because it was freshly made. The customer asked why she had not been informed of this earlier, and the waiter replied that the menu indicated that customers should order early if they wanted the dessert special. The customer looked unhappy and paid her bill without waiting for the dessert. Then, participants were given a questionnaire similar to that used in experiment 3a, with an exception that measures of the inclination to blame the provider referred to “waiter” rather than “manager” (α > .92 in each case). In addition, they responded to three items used in experiment 3a concerning the extent to which they had paid attention to the customer (α = .90) along scales from 1 (little) to 10 (a lot). They were also asked to write down any features of the customer that happened to draw their attention. Finally, we speculated that self-threat might mediate the effects of incidental similarity on observers’ attributions of blame and their subsequent reactions. To this end, we asked participants to report the extent to which the protagonist’s obesity threatened their self-image (i.e., “To what extent did the customer’s characteristics negatively affect your self-image?”) along a scale from 1 (little) to 10 (a lot) and indicated their perception of the customer’s obesity (i.e., “Do you think the customer is fat?”) along a scale from 1 (not fat at all) to 10 (very fat). Experiment 4a Results The results of this experiment are summarized in table 6. TABLE 6 THE IMPACT OF INCIDENTAL SIMILARITY AND CUSTOMER OBESITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—STUDY 4A Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at the p < .05. TABLE 6 THE IMPACT OF INCIDENTAL SIMILARITY AND CUSTOMER OBESITY ON OBSERVERS’ REACTIONS TO A SERVICE FAILURE—STUDY 4A Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) Similar customer Nonsimilar customer Attention  Obese 7.78a (.97) 5.90b (1.06)  Average weight 7.30a (1.24) 5.70b (1.20) Perceived similarity  Obese 1.63a (.70) 2.33b (.97)  Average weight 5.78c (1.19) 3.23d (1.00) Patronage intention  Obese 5.83a (1.38) 5.19b (1.05)  Average weight 3.08c (1.08) 4.73d (1.19) Inclination to blame provider  Obese 4.45a (1.08) 4.63a (1.13)  Average weight 6.65b (1.14) 5.48c (1.01) Inclination to blame customer  Obese 6.58a (1.34) 5.92b (1.35)  Average weight 4.23c (1.29) 4.87d (1.10) Empathy for provider  Obese 5.80a (1.02) 5.65a (1.05)  Average weight 4.23b (.89) 4.89c (1.07) Empathy for customer  Obese 4.03a (1.14) 4.54b (1.05)  Average weight 6.37c (1.12) 5.79d (1.32) NOTE.—Numbers in parentheses indicate standard deviations. Means with differing subscripts in each subtable are significantly different at the p < .05. Manipulation Check Participants perceived the customer to be more obese in the obesity condition than in the control condition (7.73 vs. 2.95, respectively; F(1, 156) = 378.65, p < .001, ηp2 = .71). Attention Participants’ attention to the customer increased with both incidental similarity (from 5.80 to 7.54; F(1, 156) = 95.44, p < .001, ηp2 = .38) and the customer’s obesity (from 6.50 to 6.84; F(1, 156) = 3.60, p = .06, ηp2 = .02). No interaction occurred. Thought listing data were consistent with these attentional differences. Participants reported more features of the customer that happened to draw their attention in the similarity condition than in the control condition (2.18 vs. 1.01, respectively; F(1, 156) = 113.00, p < .001, ηp2 = .42) and more features in the obese condition than in the control condition (1.75 vs. 1.44, respectively; F(1, 156) = 8.17, p < .01, ηp2 = .05). The interaction was not significant. These differences were also evident when features pertaining to the incidental similarity were not considered. That is, participants reported more characteristics of the obese customer (e.g., she is fat, she has long hair, she wears a blue jacket) in the similarity condition than in the control condition (1.78 vs. 1.18, respectively; F(1, 156) = 15.86, p < .001, ηp2 = .91). No difference was evident when the customer was not obese (1.03 vs. .85, respectively; F(1, 156) = 1.35, p > .10, ηp2 = .01). Perceived Similarity Analyses of perceived similarity yielded main effects of customer similarity and obesity (F(1, 156) = 35.57, p < .001, ηp2 = .19 and F(1, 156) = 265.05, p < .001, ηp2 = .63, respectively) and an interaction of these variables (F(1, 156) = 109.78, p < .001, ηp2 = .41). Between-cell comparisons (see table 6) indicated that in control conditions, participants perceived the customer to be more similar to them when they had an incidental similarity to the customer than when they did not (5.78 vs. 3.23, respectively; F(1, 156) = 271.41, p <.001, ηp2 = .64). When the customer was obese, however, this difference was reversed (1.63 vs. 2.33, respectively; F(1, 156) = 21.41, p <.001, ηp2 = .12). Patronage Intentions Analyses of patronage intensions yielded main effects of customer similarity and obesity (F(1, 156) = 7.39, p < .01, ηp2 = .05; and F(1, 156) = 73.23, p < .001, ηp2 = .32, respectively) and an interaction of obesity and incidental similarity (F(1, 156) = 37.29, p < .001, ηp2 = .19). Between-cell comparisons indicated that incidental similarity decreased patronage intentions in control conditions (from 4.73 to 3.08; F(1, 156) = 78.56, p <.001, ηp2 = .33) but increased patronage intentions when the customer was obese (from 5.19 to 5.83; F(1, 156) = 12.43, p <.001, ηp2 = .07). Responsibility Attributions Analyses of inclinations to blame the service provider yielded main effects of both incidental similarity (F(1, 156) = 8.29, p < .01, ηp2 = .05), and the customer’s obesity (F(1, 156) = 77.05, p < .001, ηp2 = .33) and an interaction of these variables (F(1, 156) = 15.10, p < .001, ηp2 = .09). Incidental similarity to the customer increased attributions of responsibility to the provider in control conditions (from 5.48 to 6.65; F(1, 156) = 46.19, p <.001, ηp2 = .23), consistent with the results of earlier studies. However, this difference was not evident when the customer was obese (4.45 vs. 4.63, respectively; F(1, 156) = 1.90, p >.10, ηp2 = .01). Analyses of inclinations to blame to customer also yielded a main effect of obesity and an interaction (F(1, 156) = 69.78, p < .001, ηp2 = .31 and F(1, 156) = 10.31, p < .01, ηp2 = .06, respectively). Incidental similarity decreased attributions of responsibility to the customer in control conditions (from 4.87 to 4.23; F(1, 156) = 10.60, p <.01, ηp2 = .06) but increased these attributions when she was obese (from 5.92 to 6.58; F(1, 156) = 11.23, p <.01, ηp2 = .07). Empathy (Perspective) Separate analyses of empathy for the provider and empathy for the customer yielded an interaction of customer obesity and incidental similarity in each case (F(1, 156) = 6.29, p < .05, ηp2 = .04 and F(1, 156) = 8.78, p < .01, ηp2 = .05, respectively). Table 6 shows that in control conditions, observers’ incidental similarity to the customer increased their empathy for her (from 5.79 to 6.37; F(1, 156) = 10.73, p <.01, ηp2 = .06) and decreased their empathy for the provider (from 4.89 to 4.23; F(1, 156) = 18.10, p <.001, ηp2 = .10), consistent with assumptions that observers took the customer’s perspective in these conditions. When the customer was obese, however, incidental similarity to her decreased observers’ empathy for her (from 4.54 to 4.03; F(1, 156) = 8.47, p <.01, ηp2 = .05) and had little effect on their empathy for the provider (5.65 vs. 5.80, respectively; F(1, 156) = 1.87, p >.10, ηp2 = .01). Thus, although incidental similarity to the obese customer increased observers’ attention to her, it led them to avoid taking her perspective when they construed the implications of the service failure situation. These findings, like those of experiment 3, suggest that when a protagonist has a distinguishing characteristic that is potentially relevant to observers’ attributions of responsibility, the effects of incidental similarity on the perspective they take are not evident. Mediation We assumed that participants’ attention to the customer and their responsibility attributions would mediate the effects of the customer’s obesity and incidental similarity on observers’ patronage intentions; that is, incidental similarity → attention × obesity → attributions of blame to customer → patronage inclinations. Bootstrapping analyses (Hayes 2018, model 91) confirmed this model (based on 5,000 samples, 95% CI from .5279 to 1.4197). However, an alternative model in which the causal sequence of blame and attention was reversed was not supported (based on 5,000 bootstrapping samples, 95% CI from –.0417 to .0587). In addition, a serial mediation analysis (model 6, Hayes 2018) that included empathy as a mediator was confirmed in the control condition: incidental similarity → attention → empathy for customer → attributions of blame to customer → patronage inclinations (95% CI from –.3359 to –.0487). In the obesity condition, however, this model was not supported (95% CI from –.1017 to .3369). Self-Threat The main effect of obesity on self-threat was significant (F(1, 156) = 12.16, p < .01, ηp2 = .07) and the interaction of similarity and obesity was marginally significant (F(1, 156) = 3.68, p = .057, ηp2 = .02). Incidental similarity to the customer increased feelings of self-threat to the customer (from 1.83 to 2.35; F(1, 156) =14.37, p < .001, ηp2 = .08) when the customer was obese, and no effect was found in the control conditions (1.60 vs.1.58; F(1, 156) = 1.24, p > .10, ηp2 = .01). However, a model that assumed that self-threat mediated the effect of incidental similarity on attributions of blame to the customer was not supported under either obesity conditions (95% CI: from –.0323 to .1179) or control conditions (95% CI: from –.0419 to .0156). Experiment 4b A second experiment (described in the web appendix) showed that analogous effects of obesity occur when the provider in a service failure situation is obese. One hundred twenty participants (43% male) imagined going to a bookstore and overhearing a conversation between a female salesperson and a male customer. A picture of the interaction was presented and indicated that the salesperson was either of average weight or excessive weight. (In the picture, the salesperson had her back to the camera, so participants could not see her face.) In the conversation, the customer found that a book he had ordered had not been delivered on time. In similar provider conditions, the dialogue indicated that participants had the same last name as the salesperson. In control conditions, the names did not match. Incidental similarity and the provider’s obesity had interactive effects on perceived similarity to the provider (F(1, 116) = 56.47, p < .001, ηp2 = .33), attributions of responsibility to the provider (F(1, 116) = 21.16, p < .001, ηp2 = .15), and patronage intentions (F(1, 116) = 15.43, p < .001, ηp2 = .12). In control conditions, incidental similarity to the provider increased observers’ perceptions of their similarity to her (from 3.37 to 5.97; F(1, 116) = 182.45, p < .001, ηp2 = .61), decreased their disposition to blame the provider for the incident (from 5.58 to 4.68; F(1, 116) = 19.92, p < .01, ηp2 = .15), and increased patronage intentions (from 4.89 to 6.06; F (1, 116) = 31.99, p < .001, ηp2 = .22). When the provider was obese, however, observers’ incidental similarity to her decreased their perceptions of their similarity to her (from 2.83 to 2.53; F(1, 116) = 3.31, p = .07, ηp2 = .03), increased their inclination to blame the provider for the service failure (from 6.26 to 7.25; F(1, 116) = 24.03, p < .001, ηp2 = .17), and decreased their patronage intentions (from 4.03 to 3.56; F(1, 116) = 5.80, p < .05, ηp2 = .05). Discussion Experiment 4 confirmed the hypothesis that an undesirable feature of a protagonist that is irrelevant to the situation observed (i.e., the protagonist’s obesity) would reverse the effect of incidental similarity on the observer’s responsibility attributions and willingness to patronize the establishment in the future. Mediation analyses also confirmed that attention underlies this reversed effect and ruled out an explanation in terms of self-threat. These findings are consistent with our expectation that incidental similarity increases attention to the provider/customer in an interaction and consequently increases the effect of other characteristics of the provider/customer on judgments of him and the situation. Although control participants perceived themselves as more similar to the protagonist when they were incidentally similar to the protagonist than when they were not, this difference was reversed when the protagonist was obese. Thought listing data also showed that participants reported more thoughts about other characteristics of the obese customer in the similarity condition than in the control condition. In other words, incidental similarity apparently induced participants to think about other differences between themselves and the obese customer and, therefore, lowered their perceived similarity to her. GENERAL DISCUSSION Four field and laboratory experiments show that an incidental similarity to one of the protagonists in a service failure situation can increase observers’ attention to this protagonist and consequently can affect their attributions of responsibility for the failure’s occurrence. The effects of similarity on social interactions (Berscheid and Reis 1998; Byrne 1971) and the determinants of attributions of responsibility have been extensively investigated in previous research (Jones and Nisbett 1971; Weiner 1980,1985). However, our conceptualization of the way these factors influence observers’ reactions to a service failure has both theoretical and empirical implications that have not previously been identified. Theoretical Implications In contrast to assumptions underlying previous research on the effects of perceived similarity (Byrne 1971), we found that the effect of incidental similarity to a protagonist on responses to information about a service failure was mediated by the relative attention that is spontaneously drawn to this protagonist and not by similarity per se. This attention can have the opposite effects on observers’ attributions of responsibility for the incident, depending on whether the protagonist has a distinguishing feature that could affect these attributions. When neither protagonist in the situation has a distinguishing characteristic that is likely to be considered undesirable, observers whose attention is drawn to one protagonist spontaneously construe its implications from this protagonist’s perspective. In these conditions, therefore, an incidental similarity to one protagonist decreases the disposition to blame this protagonist and consequently increases the disposition to blame the other. When one of the protagonists in a service failure situation has an undesirable characteristic or has engaged in undesirable behavior, however, observers generally evaluate this protagonist negatively and blame him or her for the service failure. In this case, their incidental similarity to the protagonist has little, if any, effect on the perspective from which they construe its implications. However, their relatively greater attention to the incidentally similar protagonist increases the thought they devote to their judgment and increases the extremity of their attribution of responsibility. In principle, these two effects of incidental similarity could occur independently. If this were so, however, the opposing effects of increased attention to a protagonist who has an undesirable attribute would offset one another. But, as the results of experiments 3 and 4 indicate, there was no evidence that incidental similarity affected observers’ perspective in these conditions (as inferred from the effect of empathy). Thus, it seems reasonable to conclude that when an incidentally similar protagonist has an undesirable attribute, observers’ disposition to take the protagonist’s perspective is eliminated. Other interpretations of these effects should be considered. For example, people generally avoid taking responsibility for negative events (Arkin et al. 1976; Shaver 1985), and this tendency could contribute to the disposition for observers who took the perspective of a protagonist to attribute responsibility for the service failure to the other. As experiment 4 indicated, however, the motivation to avoid threats to self-esteem did not appear to be a major factor in the conditions we investigated. Moreover, if observers were personally threatened by a negative event for which a similar person was responsible, they would be relatively disinclined to attribute a service failure to such a person regardless of other considerations. In contrast, experiments 3 and 4 showed that incidental similarity increased observers’ disposition to attribute the service failure to a protagonist who had an undesirable characteristic. Finally, experiment 4 provided no evidence that self-threat mediated the effects we observed. Two aspects of our results are worth reiterating. First, previous research on the effects of incidental similarity has typically been restricted to conditions in which (a) individuals interact directly with the similar other, and (b) the effects of the interaction are positive (cf. Burger et al. 2004; Guéguen et al. 2005), to respond to his or her questions on intimate topics (Martin and Guéguen 2013; Miller et al. 1998). In contrast, we found that people’s incidental similarity can affect reactions to a person even when they have no direct contact with the person but only observe a situation in which the person is involved. Second, research on perspective taking indicates that the closeness of people’s relationship to someone has a positive effect on their likelihood of taking the person’s perspective and empathizing with this person (Batson and Shaw 1991; Cialdini et al. 1997; Johnson et al. 2002). In this research, however, similarity has typically been defined in terms of personality, values, or shared experiences. We show that an individual’s trivial similarity that has no objective implications for personality can stimulate observers to empathize with him or her. Moreover, this type of similarity can sometimes decrease empathy rather than increasing it. Other Implications Numerous studies have examined consumers’ attributions and reactions to service failures (Bitner 1990; Chan, Wan, and Sin 2009; Folkes, Koletsky, and Graham 1987; Wan 2013; Wan, Hui, and Wyer 2011). However, these studies largely focused on reactions of a customer who was directly involved in the service failure. Although a few studies have investigated the role of observers (Laufer and Gillespie 2004; Wan, Chan, and Su 2011), they examined only factors that influence observers’ attributions of blame to the service provider. Our research shows that incidental similarity can affect both online and offline observers’ attribution of blame to both the service provider and the customer and that this effect occurs even though the failure is personally irrelevant. The effect of incidental similarity to a protagonist on observers’ responsibility attributions has implications for consumer behavior outside the laboratory. In an actual shopping situation, idiosyncratic characteristics of a service provider are usually conveyed in the provider’s name tag. Our results suggest that this information may fortuitously stimulate customers to identify an incidentally similarity to the provider and consequently to hold the person less responsible not only for problems that involve the customer personally but also for problems that involve others. These effects might only be evident, however, if the provider is not antagonistic or does not have distinguishing physical characteristics that customers consider undesirable. Our results could also have implications for consumers’ reactions to the information conveyed online. Many customers base their offline purchase decisions on online reviews. Online reviewer profiles (with reviewer’s name, picture, the number of reviews posted, etc.) are one of the most important ways that customers assess the reviews’ credibility (Lim and Van Der Heide 2015). Many social media websites convey consumers’ descriptions of a shopping experience they have had at a particular store or restaurant. The description may be either favorable or unfavorable. Communicators might be more effective if they accompany their message with their name or other demographic features that are fortuitously similar to those of the recipients. Some qualifications on our findings should be considered. For example, our conclusions assume that observers not only are incidentally similar to the persons they observe but also evaluate the similarity favorably. Individuals with a strong need for uniqueness may react negatively to indications they are similar to another (Snyder and Fromkin 1977). Moreover, both individual and cultural differences might influence one’s general disposition to think about oneself in relation to others (Markus and Kitayama 1991). Participants in the present research were Asians, who may be particularly sensitive to the similarities between themselves and others and inclined to value social connectedness. Although the effects of incidental similarity have been identified in research on Western cultural samples as well (Burger et al. 2004), this difference could be a consideration in evaluating the generalizability of our findings. DATA COLLECTION INFORMATION The first author designed experiment 1 and supervised the collection of data for it by research assistants at Hong Kong Green Fresh Vegan Restaurant in December 2015. The first author designed experiments 2, 3, 4, and 5 with feedback from the second author. The first author supervised the collection of data for experiments 2, 3, 4, and 5 by research assistants at the Chinese University of Hong Kong. Experiment 4b was run in September 2016, experiment 5b in November 2016, experiment 3 in July 2017, experiment 5a in August 2017, experiment 2 in November 2017, and experiment 4a in March 2018. Data from all experiments were analyzed by the first author and discussed by both authors. The authors gratefully acknowledge the helpful input and guidance from the editor, associate editor, and three reviewers. The authors also gratefully acknowledge a group of research assistants and postgraduate students from the School of Hotel and Tourism Management, The Chinese University of Hong Kong, for their assistance in data collection. In addition, the authors would like to thank Hong Kong Green Fresh Vegan Restaurant for their help in data collection for the field experiment. This research was supported by Grants GRF 14502514 from the Research Grants Council, Hong Kong. Supplemental materials are included in the web appendix accompanying the online version of this article. References Alicke Mark D. ( 2000 ), “ Culpable Control and the Psychology of Blame ,” Psychological Bulletin , 126 4 , 556 – 74 . Google Scholar CrossRef Search ADS PubMed Arkin Robert. M. , Gleason James M. , Johnston Shawn ( 1976 ), “ Effect of Perceived Choice, Expected Outcome, and Observed Outcome of an Action on the Causal Attributions of Actors ,” Journal of Experimental Social Psychology , 12 2 , 151 – 8 . Google Scholar CrossRef Search ADS Bargh John A. ( 1982 ), “ Attention and Automaticity in the Processing of Self-Relevant Information ,” Journal of Personality and Social Psychology , 43 3 , 425 – 36 . Google Scholar CrossRef Search ADS Batson C. Daniel , Duncan Bruce D. , Ackerman Paula , Buckley Terese , Birch Kimberly S. ( 1981 ), “ Is Empathic Emotion a Source of Altruistic Motivation? ” Journal of Personality and Social Psychology , 40 2 , 290 – 302 . Google Scholar CrossRef Search ADS Batson C. Daniel , Shaw Laura L. ( 1991 ), “ Evidence for Altruism: Toward a Pluralism of Prosocial Motives ,” Psychological Inquiry , 2 2 , 107 – 22 . Google Scholar CrossRef Search ADS Berscheid Ellen , Reis Harry T. ( 1998 ), “Attraction and Close Relationships,” in Handbook of Social Psychology , Vol. 2 , ed. Gilbert Daniel T. , Fiske Susan T. , Lindzey Gardner , Boston : McGraw-Hill , 193 – 281 . Bitner Mary Jo ( 1990 ), “ Evaluating Service Encounters: The Effects of Physical Surroundings and Employee Responses ,” Journal of Marketing , 54 2 , 69 – 82 . Google Scholar CrossRef Search ADS Burger Jerry M. , Messian Nicole , Patel Shebani , del Prado Alicia , Anderson Carmen ( 2004 ), “ What a Coincidence! The Effects of Incidental Similarity on Compliance ,” Personality and Social Psychology Bulletin , 30 1 , 35 – 43 . Google Scholar CrossRef Search ADS PubMed Byrne E. Donn ( 1971 ), The Attraction Paradigm , New York : Academic Press . Causon Jo ( 2015 ), “Customer Complaints Made via Social Media on the Rise,” https://www.theguardian.com/media-network/2015/may/21/customer-complaints-social-media-rise/. Chan Haksin , Wan Lisa C. , Sin Leo Y. M. ( 2009 ), “ The Contrasting Effects of Culture on Consumer Tolerance: Interpersonal Face and Impersonal Fate ,” Journal of Consumer Research , 36 2 , 292 – 304 . Google Scholar CrossRef Search ADS Cialdini Robert B. , Brown Stephanie L. , Lewis Brian P. , Luce Carol , Neuberg Steven L. ( 1997 ), “ Reinterpreting the Empathy-Altruism Relationship: When One into One Equals Oneness ,” Journal of Personality and Social Psychology , 73 3 , 481 – 94 . Google Scholar CrossRef Search ADS PubMed Cooper Joel , Jones Edward E. ( 1969 ), “ Opinion Divergence as a Strategy to Avoid Being Miscast ,” Journal of Personality and Social Psychology , 13 1 , 23 – 30 . Google Scholar CrossRef Search ADS Cowart Kelly O. , Brady Michael K. ( 2014 ), “ Pleasantly Plump: Offsetting Negative Obesity Stereotypes for Frontline Service Employees ,” Journal of Retailing , 90 3 , 365 – 78 . Google Scholar CrossRef Search ADS Crandall Christian S. ( 1994 ), “ Prejudice against Fat People: Ideology and Self-Interest ,” Journal of Personality and Social Psychology , 66 5 , 882 – 94 . Google Scholar CrossRef Search ADS PubMed Crosby Lawrence A. , Evans Kenneth R. , Cowles Deborah ( 1990 ), “ Relationship Quality in Services Selling: An Interpersonal Influence Perspective ,” Journal of Marketing , 54 3 , 68 – 81 . Google Scholar CrossRef Search ADS Davis Mark H. ( 1983 ), “ Measuring Individual Differences in Empathy: Evidence for a Multidimensional Approach ,” Journal of Personality and Social Psychology , 44 1 , 113 – 26 . Google Scholar CrossRef Search ADS Dimensional Research ( 2013 ), “Customer Service and Business Results: A Survey of Customer Service from Mid-Size Companies,” https://d16cvnquvjw7pr.cloudfront.net/resources/whitepapers/Zendesk_WP_Customer_Service_and_Business_Results.pdf. Folkes Valerie S. ( 1982 ), “ Forming Relationships and the Matching Hypothesis ,” Personality and Social Psychology Bulletin , 8 4 , 631 – 6 . Google Scholar CrossRef Search ADS Folkes Valerie S. , Koletsky Susan , Graham John L. ( 1987 ), “ A Field Study of Causal Inferences and Consumer Reaction: The View from the Airport ,” Journal of Consumer Research , 13 4 , 534 – 9 . Google Scholar CrossRef Search ADS Greenleaf Christy , Starks Misty , Gomez Laura , Chambliss Heather , Martin Scott ( 2004 ), “ Weight-Related Words Associated with Figure Silhouettes ,” Body Image , 1 4 , 373 – 84 . Google Scholar CrossRef Search ADS PubMed Guéguen Nicolas , Martin Angélique ( 2009 ), “ Incidental Similarity Facilitates Behavioral Mimicry ,” Social Psychology , 40 2 , 88 – 92 . Google Scholar CrossRef Search ADS Guéguen Nicolas , Pichot Nathalie , Le Dreff Gwénaëlle ( 2005 ), “ Similarity and Helping Behavior on the Web: The Impact of the Convergence of Surnames between a Solicitor and a Subject in a Request Made by E-Mail ,” Journal of Applied Social Psychology , 35 2 , 423 – 9 . Google Scholar CrossRef Search ADS Hansen Ranald D. , Stonner David M. ( 1978 ), “ Attributes and Attributions: Inferring Stimulus Properties, Actors’ Dispositions, and Causes ,” Journal of Personality and Social Psychology , 36 6 , 657 – 67 . Google Scholar CrossRef Search ADS Hayes Andrew F. ( 2018 ), Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach , 2nd ed. , New York : Guildford Press . Jiang Lan , Hoegg JoAndrea , Dahl Darren W. , Chattopadhyay Amitava ( 2010 ), “ The Persuasive Role of Incidental Similarity on Attitudes and Purchase Intentions in a Sales Context ,” Journal of Consumer Research , 36 5 , 778 – 91 . Google Scholar CrossRef Search ADS Johnson James D. , Simmons Carolyn H. , Jordav Amanda , Maclean Leslie , Taddei Jeffrey , Thomas Duane , Dovidio John F. , Reed William ( 2002 ), “ Rodney King and O. J. Revisited: The Impact of Race and Defendant Empathy Induction on Judicial Decisions ,” Journal of Applied Social Psychology , 32 6 , 1208 – 23 . Google Scholar CrossRef Search ADS Jones Edward E. , Nisbett Richard E. ( 1971 ), The Actor and the Observer: Divergent Perceptions of the Causes of Behavior , Morristown, NJ : General Learning Press . Kunda Ziva ( 1987 ), “ Motivated Inference: Self-Serving Generation and Evaluation of Causal Theories ,” Journal of Personality and Social Psychology , 53 4 , 636 – 47 . Google Scholar CrossRef Search ADS Laufer Daniel , Gillespie Kate ( 2004 ), “ Differences in Consumer Attributions of Blame between Men and Women: The Role of Perceived Vulnerability and Empathic Concern ,” Psychology and Marketing , 21 2 , 141 – 57 . Google Scholar CrossRef Search ADS Lim Young-shin , Van Der Heide Brandon ( 2015 ), “ Evaluating the Wisdom of Strangers: The Perceived Credibility of Online Consumer Reviews on Yelp ,” Journal of Computer-Mediated Communication , 20 1 , 67 – 82 . Google Scholar CrossRef Search ADS Malle Bertram F. ( 2006 ), “ The Actor–Observer Asymmetry in Attribution: A (Surprising) Meta-Analysis ,” Psychological Bulletin , 132 6 , 895 – 919 . Google Scholar CrossRef Search ADS PubMed Markus Hazel R. , Kitayama Shinobu ( 1991 ), “ Culture and the Self: Implications for Cognition, Emotion, and Motivation ,” Psychological Review , 98 2 , 224 – 53 . Google Scholar CrossRef Search ADS Martin Angélique , Guéguen Nicolas ( 2013 ), “ The Influence of Incidental Similarity on Self-Revelation in Response to an Intimate Survey ,” Social Behavior and Personality: An International Journal , 41 3 , 353 – 6 . Google Scholar CrossRef Search ADS McFerran Brent , Dahl Darren , Fitzsimons Gavan , Morales Andrea ( 2010 ), “ I’ll Have What She’s Having: Effects of Social Influence and Body Type on the Food Choices of Others ,” Journal of Consumer Research , 36 6 , 915 – 29 . Google Scholar CrossRef Search ADS Miller Dale T. , Downs Julie S. , Prentice Deborah A. ( 1998 ), “ Minimal Conditions for the Creation of a Unit Relationship: The Social Bond between Birthdaymates ,” European Journal of Social Psychology , 28 3 , 475 – 81 . Google Scholar CrossRef Search ADS Moray Neville ( 1959 ), “ Attention in Dichotic Listening: Affective Cues and the Influence of Instructions ,” Quarterly Journal of Experimental Psychology , 11 1 , 56 – 60 . Google Scholar CrossRef Search ADS Oishi Shigehiro , Wyer Robert S. Jr. , Colcombe Stanley J. ( 2000 ), “ Cultural Variation in the Use of Current Life Satisfaction to Predict the Future ,” Journal of Personality and Social Psychology , 78 3 , 434 – 45 . Google Scholar CrossRef Search ADS PubMed Petty Richard E. , Haugtvedt Curtis , Smith S. M. ( 1995 ), “Elaboration as a Determinant of Attitude Strength: Creating Attitudes that Are Persistent, Resistant, and Predictive of Behavior,” in Attitude Strength: Antecedents and Consequences , ed. Petty Richard E. , Krosnick Jon A. , Mahwah, NJ : Erlbaum , 93 – 130 . Pingitore Regina , Dugoni Bernard L. , Tindale R. Scott , Spring Bonnie ( 1994 ), “ Bias Against Overweight Job Applicants in a Simulated Employment Interview ,” Journal of Applied Psychology , 79 6 , 909 – 17 . Google Scholar CrossRef Search ADS PubMed Price James H. , Desmond Sharon M. , Krol Ronald A. , Snyder Frank F. , O’Connell Janelle K. ( 1987 ), “ Family Practice Physicians’ Beliefs, Attitudes, and Practices Regarding Obesity ,” American Journal of Preventive Medicine , 3 6 , 339 – 45 . Google Scholar CrossRef Search ADS PubMed Puhl Rebecca , Brownell Kelly D. ( 2001 ), “ Bias, Discrimination, and Obesity, ” Obesity Research , 9 12 , 788 – 805 . Google Scholar CrossRef Search ADS PubMed Regan Dennis T. , Straus Ellen , Fazio Russell ( 1974 ), “ Liking and the Attribution Process ,” Journal of Experimental Social Psychology , 10 4 , 385 – 97 . Google Scholar CrossRef Search ADS Regan Dennis T. , Totten Judith ( 1975 ), “ Empathy and Attribution: Turning Observers into Actors ,” Journal of Personality and Social Psychology , 32 5 , 850 – 6 . Google Scholar CrossRef Search ADS PubMed Shaver Kelly G. ( 1985 ), The Attribution of Blame: Causality, Responsibility, and Blameworthiness , New York : Springer . Shiv Baba , Fedorikhin Alexander ( 1999 ), “ Heart and Mind in Conflict: The Interplay of Affect and Cognition in Consumer Decision Making ,” Journal of Consumer Research , 26 3 , 278 – 92 . Google Scholar CrossRef Search ADS Snyder C. R. , Fromkin Howard L. ( 1977 ), “ Abnormality as a Positive Characteristic: The Development and Validation of a Scale Measuring Need for Uniqueness ,” Journal of Abnormal Psychology , 86 5 , 518 – 27 . Google Scholar CrossRef Search ADS Solovay Sondra ( 2000 ), Tipping the Scales of Justice: Fighting Weight-Based Discrimination , Amherst, NY : Prometheus . Stalling Richard B. ( 1970 ), “ Personality Similarity and Evaluative Meaning as Conditioners of Attraction ,” Journal of Personality and Social Psychology , 14 1 , 77 – 82 . Google Scholar CrossRef Search ADS PubMed Storms Michael D. ( 1973 ), “ Videotape and the Attribution Process: Reversing Actors’ and Observers’ Points of View ,” Journal of Personality and Social Psychology , 27 2 , 165 – 75 . Google Scholar CrossRef Search ADS PubMed Tacikowski Pawel , Nowicka Anna ( 2010 ), “ Allocation of Attention to Self-Name and Self-Face: An ERP Study ,” Biological Psychology , 84 2 , 318 – 24 . Google Scholar CrossRef Search ADS PubMed Taylor Shelley E. , Fiske Susan T. ( 1978 ), “ Salience, Attention, and Attribution: Top of the Head Phenomena ,” Advances in Experimental Social Psychology , 11 , 249 – 88 . Google Scholar CrossRef Search ADS Taylor Shelley E. , Peplau Letitia A. , Sears David O. ( 2006 ), Social Psychology , 12th ed. , Upper Saddle River, NJ : Pearson . Tesser Abraham ( 1978 ), “ Self-Generated Attitude Change ,” Advances in Experimental Social Psychology , 11 , 289 – 338 . Google Scholar CrossRef Search ADS Wan Lisa C. ( 2013 ), “ Culture’s Impact on Consumer Complaining Responses to Embarrassing Service Failure ,” Journal of Business Research , 66 3 , 298 – 305 . Google Scholar CrossRef Search ADS Wan Lisa C. , Chan Elisa K. Y. , Su Lei ( 2011 ), “ When Will Customers Care About Service Failures That Happened to Strangers? The Role of Personal Similarity and Regulatory Focus and Its Implication on Service Evaluation ,” International Journal of Hospitality Management , 30 1 , 213 – 20 . Google Scholar CrossRef Search ADS Wan Lisa C. , Hui Michael K. , Wyer Robert S. ( 2011 ), “ The Role of Relationship Norms in Responses to Service Failures ,” Journal of Consumer Research , 38 2 , 260 – 77 . Google Scholar CrossRef Search ADS Weiner Bernard ( 1980 ), Human Motivation , New York : Holt-Rinehart, & Winston . Weiner Bernard ( 1985 ), “ An Attributional Theory of Achievement Motivation and Emotion ,” Psychological Review , 92 4 , 548 – 73 . Google Scholar CrossRef Search ADS PubMed Zhu Feng , Zhang Xiaoquan (Michael) ( 2010 ), “ Impact of Online Consumer Reviews on Sales: The Moderating Role of Product and Consumer Characteristics ,” Journal of Marketing , 74 2 , 133 – 48 . 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/about_us/legal/notices)

Journal

Journal of Consumer ResearchOxford University Press

Published: Jun 6, 2018

There are no references for this article.

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