Learning from supervisor negative gossip: The reflective learning process and performance outcome of employee receivers

Workplace gossip is generally viewed as a deviant behavior that negatively affects the work outcomes of employees. However, we argue that this negative view is incomplete. Drawing on the cultural learning perspective of gossip and social learning theory, we examine how the job performance of employee receivers benefits from supervisor negative gossip through reflective learning. On the basis of multi-source, cross-sectional designs, Studies 1 and 2 consistently find that supervisor negative gossip facilitates employee receiver reflective learning and subsequent job performance when controlling for two sets of theory-relevant variables. Study 3, which has a multi-source, cross-lagged panel design, provides further evidence of the directional relationship from supervisor negative gossip to employee receiver job performance through reflective learning. The findings of the three separate field studies support the positive effect of supervisor negative gossip on employee receivers from a learning perspective. We discuss the theoretical and practical implications of these findings in terms of how employee receiver job performance can benefit from workplace negative gossip.


Introduction
Gossip, or evaluative talk between two or more people in which judgments about an absent third party are made (Ellwardt et al., 2012b;Kurland and Pelled, 2000), is ubiquitous in social life. For decades, gossip has received considerable research attention in diverse fields, such as anthropology, social psychology, linguistics, communication (Van Iterson and Clegg, 2008), and organizational behavior that focuses on workplace gossip (e.g. Noon and Delbridge, 1993;Wittek and Wielers, 1998). Early organizational research has mainly viewed workplace gossip as a deviant behavior that violates organizational norms and threatens the well-being of the organization and its members (Robinson and Bennett, 1995). However, this negative view is incomplete, as gossip can serve important social functions, such as helping people gather and learn from such information (Baumeister et al., 2004;Foster, 2004;Grosser et al., 2012). Indeed, the recent literature has adopted a neutral view on workplace gossip, defining it as 'informal and evaluative (i.e. positive or negative) talk from one member of an organization to one or more members of the same organization about another member of the organization who is not present to hear what is said' (Brady et al., 2017: 3).
This definition suggests that workplace gossip occurs between two parties: the gossiper and the gossip receiver. Research on workplace gossip has predominantly examined its impact on the work outcomes of gossipers. Workplace gossip, particularly negative gossip, incurs costs for gossipers, such as high job anxiety, job insecurity, turnover, and low affective well-being at work and performance ratings (Brady et al., 2017;Grosser et al., 2010). However, the negative influence of workplace gossip on gossipers only offers a partial understanding of the effects of workplace gossip. The other important party in gossip, namely gossip receivers (Martinescu et al., 2014), has received scant attention in the workplace gossip literature. In contrast to the negative influence of gossip on gossipers, gossip receivers may be positively affected by gossip according to the cultural learning perspective of gossip (Baumeister et al., 2004). This perspective suggests that gossip receivers can obtain valuable information and vicarious learning experiences from gossip, which helps them be adaptive and effective (Baumeister et al., 2004). Furthermore, we expect this learning process in the workplace to subsequently enhance gossip receiver job performance (i.e. overall job performance on one's basic and core tasks; Tsui et al., 1997). Shifting the research focus to how gossip receivers are affected by gossip allows us to introduce a learning perspective for investigating the positive effect of gossip on their learning process and job performance, thereby extending the nomological network of workplace gossip.
To explore this issue, we examine the effects of negative top-down gossip from supervisors to employee receivers. We focus on gossip from supervisor gossipers to employee receivers, as gossip information from supervisors with higher hierarchical positions than employees is powerful in influencing employees (Cantor et al., 2015;Eisenberger and Stinglhamber, 2011) and can effectively prompt employee receivers' learning (Houmanfar and Johnson, 2004). We also focus on supervisor negative gossip, as the literature has implied that the valence of gossip from people with high social status in organizations is usually negative (Ellwardt et al., 2012a). Moreover, the cultural learning perspective of gossip underscores the learning value of negative gossip (Baumeister et al., 2004). To examine the specific learning process, we draw on the cultural learning perspective of gossip and social learning theory to propose reflective learning, which is a cognitive process of increasing awareness and making sense of personal experiences and developing a new understanding that guides future actions (Anseel et al., 2009;Peltier et al., 2005), as an underlying learning process of employee receivers through which supervisor negative gossip positively affects their job performance.
We contribute to the literature in three ways. First, by orienting research to the other important party of gossip, namely gossip receivers, we add nuanced knowledge to the positive effects of workplace gossip on gossip receivers. By examining the learning process of gossip receivers through which gossip positively affects their job performance, we extend the nomological network of workplace gossip. Second, we investigate reflective learning as a novel and valuable mechanism in the workplace gossip literature, thereby responding to the call of Brady et al. (2017) to delineate the processes underlying gossip and employee work outcomes. In addition, we enrich the cultural learning perspective of gossip by portraying a specific learning process. Third, we examine the positive effects of top-down negative gossip from supervisors to employee receivers to shed light on gossip as a possible managerial approach for prompting employee receivers' learning process and enhancing their job performance. We conduct three field studies to empirically test our hypotheses.

Top-down negative gossip from supervisors to employee receivers
In the workplace, gossip facilitates receivers' learning of organizational rules, norms, and regulations and of appropriate and inappropriate work behaviors (Baumeister et al., 2004;Brady et al., 2017). The gossip literature has highlighted two dimensions of workplace gossip: actors who are involved in gossip, and gossip valence (Brady et al., 2017;Wu et al., 2018). Regarding gossip actors, the workplace gossip literature has mainly focused on horizontal gossip among coworkers about absent others (e.g. Brady et al., 2017;Ellwardt et al., 2012b;Kuo et al., 2015). However, it has ignored top-down gossip from supervisor gossipers to employee receivers. This oversight is unfortunate. Supervisor negative gossip is a common type of supervisory behavior, and supervisors often gossip with employees about other absent employees (Goff and Goff, 1988). Supervisors can use gossip as an informal channel to deliver messages and expectations more efficiently than other formal channels, such as written documents (Michelson et al., 2010;Mishra, 1990;Rogers and Agarwala-Rogers, 1976;Su et al., 2009).
Gossip valence is categorized as either positive or negative (Foster, 2004). Positive gossip refers to positively evaluating a gossip subject, such as praising his/her appropriate and norm-strengthening behaviors, offering social/political support, and defending the subject in his/her absence (Ellwardt et al., 2012a). In contrast, negative gossip refers to negatively evaluating a gossip subject, such as communicating disapproval about his/her poor job performance and underlying inappropriate, norm-violating behaviors (Ellwardt et al., 2012a;Kuo et al., 2018;McDonald et al., 2007). As mentioned earlier, the valence of gossip from people with high social status is usually negative, as they mainly use negative gossip to sanction norm violations and reinforce organizational norms (Ellwardt et al., 2012a). Moreover, whereas the cultural learning perspective of gossip endorses the learning values of both positive and negative gossip, it highlights that negative gossip is more common, informative, and diagnostic than positive gossip in promoting the learning of gossip receivers (Baumeister et al., 2004;Wert and Salovey, 2004). This principle is known as 'bad is stronger than good' (Baumeister et al., 2004: 113). Researchers have argued that rules, norms, and guidelines in a social system are better conveyed and learned through negative, normviolating gossip stories than positive, norm-strengthening gossip stories (Baumeister et al., 2004;Skowronski and Carlston, 1987;Wert and Salovey, 2004). In support of this view, empirical evidence has revealed that people learn more effectively from others' negative, failed experiences than positive, successful experiences (e.g. Bledow et al., 2017;Diwas et al., 2013). Therefore, our focus on supervisor negative gossip is valuable for exploring its effects on the learning process and job performance of employee receivers, as elaborated below.

Supervisor negative gossip and employee receiver reflective learning
Supervisor negative gossip occurs during the daily interactions between a supervisor gossiper and an employee receiver. Supervisors vary the frequency with which they informally and privately convey negative evaluations of the inappropriate work behaviors and unsatisfactory job performance of absent employees to employee receivers (Kuo et al., 2018). We contend that the frequency with which a supervisor negatively gossips with an employee facilitates the employee receiver's reflective learning at work.
The cultural learning perspective of gossip highlights that 'gossip is a learning mechanism' (Baumeister et al., 2004: 116) for gossip receivers. Gossip is an extension of observational/vicarious learning that allows gossip receivers to learn from the heard stories of gossip subjects; such learning goes beyond one's personal experiences and direct observations (Baumeister et al., 2004). In the workplace, when a supervisor frequently shares negative gossip stories with an employee receiver, the employee receiver has many opportunities to vicariously learn from the gossip stories concerning the gossip subject's poor job performance and the underlying inappropriate and norm-violating behaviors. Thus, employee receivers can learn about the rules, norms, and regulations of their organizations, the expectations and standards of their supervisors, and the reasons for unsatisfactory job performance without personally experiencing such inappropriate behaviors and poor performance (Brady et al., 2017;Kuo et al., 2018;Martinescu et al., 2014).
Though the cultural learning perspective of gossip provides a general framework for understanding gossip as an extension of vicarious learning, social learning theory, which introduced the concept of vicarious learning, has been further developed to highlight self-reflection as an important learning process (Bandura, 1991;Davis and Luthans, 1980;Stajkovic and Luthans, 1998). This theory posits that vicarious punishment, or the observed negative consequences of certain behaviors of others, reduces people's tendency to behave in a similar way. Supervisor negative gossip exposes employee receivers to the negative consequences of inappropriate behaviors and unsatisfactory job performance-that is, being the subject of negative gossip and receiving negative evaluations from supervisors. The subject of supervisor negative gossip suffers not only reputation loss (Brady et al., 2017), but also possible unfavorable career consequences, such as low performance ratings and few promotion opportunities (Ellwardt et al., 2012a). Thus, employee receivers are motivated to reduce similar inappropriate behaviors described in the gossip stories and adjust their behaviors to avoid unsatisfactory performance. To achieve this goal, self-reflection is an indispensable learning process (Bandura, 1991;Zimmerman, 2000). After receiving evaluations and comments (especially negative ones) about others, people seek to understand their own behaviors and performance to guide their future behaviors (Zimmerman, 2000(Zimmerman, , 2002a(Zimmerman, , 2002b. In our context, vicarious punishment from supervisor negative gossip increases employee receivers' awareness of their own behaviors and performance and drives them to make sense of whether their behaviors are appropriate and their job performance is satisfactory. This reflective process helps employee receivers leverage the vicarious learning opportunities provided by supervisor negative gossip to develop a new understanding of their experiences that guides them to meet expectations, reduce wrongs and violations, and perform their jobs satisfactorily in the future. Some studies indirectly support our theorizing. For example, Diwas et al. (2013) argued and found that people can learn from the failures of others and that they can learn from their own failures only when they are exposed to a great number of others' failures. They explained that learning from others' failures involves reflection, through which people make sense of what went wrong and analyze the problems of their own behaviors to avoid similar failures in the future. Bledow et al. (2017) also found that failed stories effectively grab people's attention and elicit reflection. In summary, we hypothesize the following: Hypothesis 1: Supervisor negative gossip is positively related to employee receiver reflective learning.

Employee receiver reflective learning and job performance
We contend that employee receiver reflective learning is conducive to subsequent job performance. As discussed earlier, reflective learning is concerned with the cognitive process of strengthening personal awareness, making sense of experiences, and developing new understandings that guide future actions. Reflective learning involves employees' cognitive attempts to become aware of and evaluate their past behaviors and experiences (Ellis et al., 2006;Nilsen and Ellström, 2012). Furthermore, it triggers individuals to detect and diagnose inadequacies or shortcomings in their own behaviors and performance. By doing so, individuals learn how to achieve high performance (Campbell and Lee, 1988) and operate effectively in the future (Daudelin, 1996;Kolb, 1984). Campbell and Lee (1988) argued that reflection on one's own past weakness is conducive to job performance. In addition, reflection guides future actions. Through reflective learning, individuals can detect and identify the areas in which they must improve (Pee et al., 2000), determine alternative solutions to problems, and improve the effectiveness of future actions (Peltier et al., 2005). In summary, reflective learning not only helps learners understand what happened in their experiences (Kelley, 1973), but also helps them develop a new understanding of their own behaviors and performance to carry out future job tasks effectively (Loughran, 2002;Nilsen and Ellström, 2012;Smyth, 1992). Empirical evidence also supports our reasoning (e.g. Markman et al., 2008). Therefore, we hypothesize the following: Hypothesis 2: Employee receiver reflective learning is positively related to job performance.

Mediating effect of employee receiver reflective learning
Integrating Hypotheses 1 and 2, we contend that employee receiver reflective learning mediates the positive relationship between supervisor negative gossip and job performance. We theorize that the frequent supervisor negative gossip behaviors prompt employee receivers to reflect on their own experiences and past behaviors and to learn how to execute their tasks in the future effectively, which ultimately enhance their subsequent job performance. Thus, we hypothesize the following: Hypothesis 3: Employee receiver reflective learning mediates the positive effect of supervisor negative gossip on job performance.

Study overview
We conducted three field studies to test our hypotheses with different, independent samples and thereby cross-validate our findings. In Studies 1 and 2, we used multi-source, cross-sectional designs and two separate samples from different industries. We examined the mediating effect of employee receiver reflective learning on the positive relationship between supervisor negative gossip and employee receiver job performance, while controlling for two sets of theory-relevant variables. In Study 3, 1 we adopted a multi-source, cross-lagged panel design to replicate the findings of Studies 1 and 2 and provide directional evidence of the effect of supervisor negative gossip on job performance through employee receiver reflective learning. Collecting cross-lagged panel data in organizations requires the surveyed organization and its employees to be highly motivated and has a high attrition risk. 2 Therefore, Studies 1 and 2 were designed as pilot studies. After identifying the cross-sectional associations, we conducted Study 3 with a cross-lagged panel design to test the directional relationships.

Study 1
Method Participants and procedures. The data were collected from an onsite, paper-and-pencil survey in a large energy company in eastern China. The surveyed company engaged in the exploration and production, refinement, transportation, distribution, and marketing of oil, gas, and coal and in power generation and trading. With the assistance of the company's Human Resource Management Department, we invited all 250 middle-level managers, and randomly selected one of their direct subordinates to participate. Our sample thus consisted of one-on-one employee-supervisor dyads. Confidentiality was ensured and participation was voluntary. Two separate questionnaires were distributed to the supervisor and employee participants, respectively. The employee participants were asked to rate supervisor negative gossip, reflective learning, and the control variables of supervisor positive gossip, supervisor positive and negative feedback, norm acceptance, and their demographics. The supervisor participants were asked to evaluate the employee participants' job performance. Ultimately, we obtained 212 matched employee-supervisor questionnaire pairs (a response rate of 84.8%).
Of the 212 employee participants, 70.3% were male. Their average age was 28.5 years (SD = 4.9) and their average organizational tenure was 5.1 years (SD = 4.6). In terms of education, 2.4% received secondary school education, 25.5% received junior college education, and 72.1% received college or higher education. Of the matched supervisor participants, 72.2% were male. Their average age was 34.9 years (SD = 7.0). In terms of education, .5% received secondary school education, 20.8% received junior college education, and 78.8% received college or higher education.
Measures. All of the items were originally developed in English and translated into Chinese following a standard back-translation procedure (Brislin, 1980).

Supervisor negative gossip.
We modified a four-item scale used by Ellwardt et al. (2012b) to measure supervisor negative gossip by changing the referent of gossiper from 'I' to 'my supervisor' and the referent of gossip subject from 'managers' to 'other colleagues'. For example, we modified the item, 'At work, I sometimes complain about managers while they are absent' to 'At work, my supervisor sometimes complains about other colleagues while they are absent.' The three other modified items were 'My supervisor sometimes makes negative comments on the behavior of other colleagues while they are absent,' 'If my supervisor feels treated badly by other colleagues, he/she talks about this to me,' and 'My supervisor sometimes criticizes other colleagues for a negative characteristic while they are absent.' Following previous research (e.g. Wu et al., 2018), the employee participants were asked to rate how often their supervisors exhibited the behaviors described in these four items in the past six months on a seven-point Likert scale (1 = never, 7 = always). Cronbach's alpha was .94.
Reflective learning. We measured reflective learning using an eight-item scale developed by Peltier et al. (2005). Based on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree), the employee participants rated the extent of their agreement with the items describing their experiences at work. The sample items were 'I often reflected on my actions to see whether I could improve them' and 'I often tried to think about how I could do something better next time' (α = .83).
Job performance. We used an 11-item scale developed by Tsui et al. (1997) to measure job performance (a = .96). Of the 11 items, six assessed employees' basic task performance in terms of task quantity, quality, and efficiency. Each supervisor rated the extent to which he/she agreed with the items describing the focal employee's performance as better than that of the peers on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree). The sample items were 'This employee's quantity of work is higher than average' and 'This employee strives for higher quality work than required.' The other five items measured core task performance in terms of employees' overall ability, judgement, accuracy, job knowledge, and creativity. The supervisors rated the extent to which the employee had met the performance standards as described in the items on a seven-point Likert scale (1 = unsatisfactory, 7 = excellent). The sample items were 'This employee's ability to perform core job tasks' and 'This employee's creativity when performing core tasks.' Control variables. Consistent with previous studies (e.g. Ellwardt et al., 2012b), we controlled for the employee participants' age, sex, education, organizational tenure, and supervisor positive gossip. We modified the three-item scale of positive gossip used by Ellwardt et al. (2012b) to measure supervisor positive gossip, following the same approach we used to adapt the measurement of supervisor negative gossip mentioned above. One sample item was 'My supervisor sometimes makes a positive comment about other colleagues while they are absent' (α = .89). We also included supervisor positive and negative feedback as the control variables to partial out their plausible effects on employee receiver reflective learning. It has been argued that feedback is a common way to inform employees about any desired/undesired performance and behavior that need persistence or correction in the workplace (Ashford and Cummings, 1983). We used a seven-item scale of performance feedback developed by George and Zhou (2001) to measure supervisor positive and negative feedback. The sample items were 'My supervisor often tells me that my performance is excellent' (positive feedback, α = .87) and 'My supervisor often tells me that my performance is not up to the standard' (negative feedback, α = .91).
The cultural learning perspective of gossip suggests that by disseminating value-laden information, negative gossip can serve as a norm-enforcing mechanism (Grosser et al., 2010) and prompt gossip receivers to understand norm-violating behaviors and accept the norms (Beersma and Van Kleef, 2012). Therefore, norm acceptance may be another plausible mediator in addition to reflective learning that connects supervisor negative gossip to employee receiver job performance. Thus, we controlled for the possible confounded mediating effect of norm acceptance. We measured norm acceptance using a three-item scale from Jackson et al. (2006). One sample item was 'I followed the norms of this department' (α = .88).
Analytical strategies. To test our hypotheses, we estimated a path model with the composite scores of our studied variables using Mplus 7.11 with maximum-likelihood estimation (Muthén andMuthén, 1998-2017). The direct effect of supervisor negative gossip on job performance and the indirect effect of supervisor negative gossip on job performance through reflective learning were specified in the path model. Regarding the control variables, we specified the direct effects of the employee participants' demographics (i.e. age, sex, education, and organizational tenure) and the theory-relevant control variables (i.e. supervisor positive gossip and supervisor positive and negative feedback) on reflective learning and job performance. We also controlled for the potential mediating effect of norm acceptance on the relationship between supervisor negative gossip and job performance.
To test the significance of the mediation effect, we used the MODEL INDIRECT command to test the indirect and direct effects and their standard errors simultaneously (Muthén andMuthén, 1998-2017), in conjunction with the BOOTSTRAP option of the ANALYSIS command with 2000 resampling (Preacher and Hayes, 2008;Preacher et al., 2007). We interpreted the indirect effects with bootstrapped 95% bias-corrected confidence intervals (BC CIs). To assess the effect size of the indirect effect, we retrieved the completely standardized indirect effect (ab cs ) using the STDYX command. We also followed Preacher and Kelley (2011) in reporting the standardized maximum possible indirect effect (κ 2 ). We calculated κ 2 by dividing the indirect effect (ab) by the maximum possible indirect effect M (ab), which was obtained using the MBESS package for R. The effect size was discussed on the basis of Cohen's (1988) recommended cutoff values of .01, .09, and .25, representing small, medium, and large effect sizes, respectively. Furthermore, we followed Hu and Bentler (1999) in reporting the comparative fit index (CFI) and standardized root-mean-square residual (SRMR) to assess model fit. Table 1 presents the means, standard deviations, correlations, and reliabilities for all of the studied variables. We conducted a series of confirmatory factor analyses to test the distinctiveness of our studied variables using individual measurement items as indicators. The six-factor model including employee-rated variables (supervisor negative and positive gossip, supervisor negative and positive feedback, reflective learning, and norm acceptance) demonstrated an acceptable fit (χ 2 (260) = 529.62, CFI = .92, SRMR = .06) and was better than the alternative models ( Table 2).

Results and discussion (Study 1)
As shown in Table 3, supervisor negative gossip was positively and significantly related to reflective learning (B = .12, SE = .03, p < .01), which in turn was positively and significantly related to job performance (B = .48, SE = .18, p < .01). Thus, Hypotheses 1 and 2 are supported. In support of Hypothesis 3, the indirect effect of supervisor negative gossip on job performance through reflective learning was positive and significant (indirect effect = .06, SE = .03; bootstrapped 95% BC CI = [.01, .12]). The indices (ab cs = .07 and κ 2 = .08) indicate a small to medium indirect effect size   (Cohen, 1988). We repeated the analyses without the control variables and obtained similar results.
The results of Study 1 support our hypotheses. Specifically, supervisor negative gossip prompted employee receiver reflective learning, which in turn improved job performance. These effects remained significant even after controlling for the three theory-relevant independent variables of supervisor positive gossip, supervisor positive and negative feedback, and the plausible mediating effect of norm acceptance. Despite the promising findings obtained, three issues must be resolved. First, although the learning value of positive gossip has been well documented in the literature (Baumeister et al., 2004;Martinescu et al., 2014), we found a non-significant relationship between supervisor positive gossip and employee receiver reflective learning. Thus, further studies are needed to replicate this finding. Second, another alternative theoretical explanation for the positive relationship between supervisor negative gossip and job performance may be that employee receivers simply hide the behaviors 3 that they expect their supervisors to disapprove. Third, the positive effects of supervisor negative gossip on reflective learning and job performance may be confounded with trust in supervisor gossiper-employee receiver dyads, as negative gossip is often shared between people who trust each other (Ellwardt et al., 2012b). To address these issues, we conducted Study 2 to replicate the findings obtained in Study 1. Study 2 incorporated trust in supervisor as an additional controlled independent variable and hiding behavior as another plausible mediator.

Study 2
Method Participants and procedures. The participants were employees and their immediate supervisors from a regional subsidiary of a nationwide company operating in the dairy production and distribution industry in northwest China. All of the 375 frontline employees and their immediate supervisors were invited to participate in an online questionnaire survey. Confidentiality was ensured and participation was voluntary. The company's Human Resource Management Department assisted us in distributing two separate questionnaires to the employee and supervisor participants, respectively. The employee participants were asked to rate supervisor negative gossip and the control variables of supervisor positive gossip, supervisor positive and negative feedback, trust in supervisor, hiding behavior, and their demographics. The supervisor participants were asked to evaluate the employee participants' job performance. We obtained 243 matched responses, yielding a response rate of 64.8%. Of the 243 employee participants, 74.9% were male. A total of 53.5% were between 20 and 29 years old, 44.0% were between 30 and 39 years old, and 2.5% were between 40 and 49 years old. Per organizational tenure, 27.2% had worked in the company for less than 1 year, 29.6% had worked in the company for 1 to 3 years, 18.5% had worked in the company for 4 to 6 years, 13.6% had worked in the company for 7 to 9 years, and 11.1% had worked in the company for at least 10 years. In terms of education, 3.3% received secondary school education, 46.5% received junior college education, and 50.2% received college or higher education.

Measures
Supervisor negative gossip. As discussed earlier, supervisor negative gossip gives employee receivers the opportunity to vicariously learn from the gossip subject's unsatisfactory job performance. 4 In keeping with this reasoning, we further modified the gossip content measurement used in Study 1 to be job performance based. Specifically, we modified the items used in Study 1 to: 'At work, my supervisor sometimes complains about other colleagues' job performance while they are absent,' 'My supervisor sometimes makes a negative comment on the job performance of other colleagues while they are absent,' 'If my supervisor feels other colleagues perform their jobs badly, he or she talks about this to me,' and 'My supervisor sometimes criticizes other colleagues for negative job performance while they are absent.' The employees were asked to rate, on a seven-point Likert scale (1 = never, 7 = always), how often their supervisors exhibited the behaviors described in these four items in the past 6 months (α = .93). We adopted the same scales of reflective learning (α = .88) and job performance (α = .97) used in Study 1.
Control variables. As in Study 1, we controlled for the employee participants' demographics of age, sex, education, and organizational tenure. Moreover, we controlled for supervisor positive gossip. Following the same approach of modifying supervisor negative gossip in Study 2, we modified the measurement of supervisor positive gossip used in Study 1 to ensure that gossip content was job performance based. One sample item was 'My supervisor sometimes makes a positive comment about other colleagues' job performance while they are absent' (α = .85). We also controlled for supervisor positive feedback (α = .88) and supervisor negative feedback (α = .87). To remain consistent with the adaptation of the measurements of gossip in this study, we adapted the feedback items used in Study 1 to be job performance based. The sample items were 'My supervisor often tells me that my job performance is excellent' and 'My supervisor often tells me that my job performance is not up to the standard.' As discussed previously, we controlled for trust in supervisor, which was measured using Yang and Mossholder's (2010) five-item scale (α = .95). We also controlled for hiding behavior to partial out its potential mediating effect on the relationship between supervisor negative gossip and employee receiver job performance. To measure hiding behavior, we adapted two relevant items from surface acting (Grandey, 2003)-one type of hiding behavior and developed one additional item based on the conceptualization of hiding behavior that captures the extent to which employees hide behaviors disapproved by their supervisors. The three items were 'I just pretend to perform behaviors I need to display for my job,' 'I put on an act to deal with my supervisor in an appropriate way he/she approves,' and 'I hide the behavior of which my supervisor shows his/her disapproval' (α = .87).
Analytical strategies. We adopted similar procedures to those used in Study 1 by estimating a path model with the composite scores of our studied variables using Mplus 7.11 (Muthén andMuthén, 1998-2017). Given that the subordinates were nested within the supervisors in Study 2, we used a design-based modeling (TYPE = COM-PLEX, ESTIMATOR = MLR) approach to deal with the non-independence owing to the nested data structure. This approach is commonly used to analyze single-level models with non-independent data structures (e.g. Schaubroeck et al., 2017;Wu and Kwok, 2012;Wu et al., 2016). In this path model, the direct effect of supervisor negative gossip on job performance and the indirect effect of supervisor negative gossip on job performance through reflective learning were specified. In addition, we specified the direct effects of employee participants' demographic variables (i.e. age, sex, education, and organizational tenure), supervisor positive gossip, supervisor positive and negative feedback, and trust in supervisor on reflective learning and job performance. We also controlled for the potential mediating effect of hiding behavior on the relationship between supervisor negative gossip and job performance.
To test the significance of the mediation effect, we used the MODEL INDIRECT command in conjunction with 95% confidence intervals (CIs) using the Monte Carlo resampling method with 20,000 repetitions. We also reported the effect size of the indirect effect using ab cs and κ 2 (Preacher and Kelley, 2011). Table 4 presents the means, standard deviations, correlations, and reliabilities for all the studied variables. We examined the discriminant validity of the variables using individual measurement items as indicators. The seven-factor model including employee-rated variables (supervisor negative and positive gossip, supervisor negative and positive feedback, trust in supervisor, reflective learning, and hiding behavior) yielded an acceptable fit (χ 2 (384) = 799.10, CFI = .92, SRMR = .07) and was better than the alternative models ( Table 2).

Results and discussion (Study 2)
As presented in Table 3, supervisor negative gossip was positively and significantly related to reflective learning (B = .10, SE = .03, p < .01), which in turn was positively and significantly associated with job performance (B = .30, SE = .13, p < .05). These results support Hypotheses 1 and 2. In support of Hypothesis 3, the indirect effect of supervisor negative gossip on job performance through reflective learning was positive and significant (indirect effect = .03, SE = .01; 95% CI = [.01, .05]). The indices (ab cs = .04 and κ 2 = .04) indicate a small to medium indirect effect size (Cohen, 1988). We repeated the analyses without the control variables and obtained similar results.
The results of Study 2 constructively replicated the findings obtained in Study 1 with another independent sample and a different set of theory-related control variables. Although the results of Studies 1 and 2 are encouraging, their cross-sectional designs could not provide a cogent examination of the directional association from supervisor negative gossip to job performance through reflective learning, which is the major focus of our hypotheses. Supervisors may select their favorite employees, who may also receive the highest performance rating to gossip with. This implies the possibility of reversed causality. Therefore, we conducted Study 3 with a multi-source, cross-lagged panel design, which is viewed as an effective way to examine the directional association implied in our theorizing (Finkel, 1995).

Study 3
Method Participants and procedures. We collected data from three companies (a sales company, a restaurant, and a property agency) located in northwest China. With the assistance of a liaison person from each company, we approached and invited 214 employees and their immediate supervisors to participate in an online questionnaire survey. Confidentiality was ensured and participation was voluntary. The liaison person of each company assisted us in distributing two separate questionnaires to the employee and supervisor participants, respectively. We collected data at three waves with one-month time intervals. During Wave 1, the employee participants were asked to rate supervisor negative gossip and their demographics. The supervisor participants were asked to evaluate the employee participants' job performance. A total of 204 matched responses were obtained. During Wave 2, the employee participants rated their reflective learning at work. A total of 189 responses were received. During Wave 3, the employee participants rated supervisor negative gossip and the supervisor participants rated the employee participants' job performance. The final sample included 132 matched cases (an overall response rate of 61.7%).
Of the 132 employee participants, 82.6% were male. A total of 7.6% were under 20 years old, 61.4% were between 20 and 29 years old, 26.5% were between 30 and 39 years old, and 4.5% were between 40 and 49 years old. Per organizational tenure, 25.8% had worked in the company for less than 1 year, 34.1% had worked in the company for 1 to 3 years, 22.7% had worked in the company for 4 to 6 years, 7.6% had worked in the company for 7 to 9 years, and 9.8% had worked in the company for at least 10 years. In terms of education, 2.3% received primary school education, 23.5% received secondary school education, 31.1% received junior college education, and 43.2% received college or higher education.
Measures. We adopted the same scales used in Study 2 to measure supervisor negative gossip, reflective learning, and job performance. The reliability for reflective learning (Wave 2) was .93. The reliabilities for supervisor negative gossip were .93 (Wave 1) and .96 (Wave 3). The reliabilities for job performance were .95 (Wave 1) and .98 (Wave 3). Consistent with Studies 1 and 2, we controlled for the employee participants' demographics in terms of age, sex, education, and organizational tenure. Considering the focus on testing the hypothesized directional association and the need for management to simplify the administration of data collection for a cross-lagged survey, Study 3 did not include any theory-relevant control variables.
Analytical strategies. Consistent with Study 2, we estimated a path model with the composite scores of our studied variables using Mplus 7.11 (Muthén andMuthén, 1998-2017). We adopted the same approach used in Study 2 to deal with the non-independence owing to the nested data structure. In this path model, the direct effect of supervisor negative gossip (Wave 1) on employee receiver job performance (Wave 3) and the indirect effect of supervisor negative gossip on employee receiver job performance through reflective learning (Wave 2) were specified. To examine the directional association between supervisor negative gossip and job performance, we specified the reversed direct effect of employee receiver job performance (Wave 1) on supervisor negative gossip (Wave 3) and the reversed indirect effect of employee receiver job performance (Wave 1) on supervisor negative gossip (Wave 3) through reflective learning (Wave 2). To test the significance of the mediation effect, we used the MODEL INDIRECT command in conjunction with 95% CIs using the Monte Carlo resampling method with 20,000 repetitions. We also reported the effect size of the indirect effect using ab cs and κ 2 (Preacher and Kelley, 2011). Table 5 presents the means, standard deviations, correlations, and reliabilities for all of the studied variables. We first examined the discriminant validity of the focal variables (i.e. supervisor negative gossip [Wave 1], reflective learning [Wave 2], and job performance [Wave 3]), with individual measurement items as indicators. The three-factor model yielded an acceptable fit (χ 2 (227) = 525.61, CFI = .91, SRMR = .05) and was better than the alternative models (Table 6). These results support the distinctiveness of our key variables. We then tested the measurement equivalence of supervisor negative gossip and job performance across the two measurement waves (Waves 1 and 3). As shown in Table 6, the configural equivalence (i.e. constraining the factor structure equivalent) and metric equivalence (i.e. constraining the factor structure and factor loadings equivalent) of supervisor negative gossip and job performance across the two waves demonstrated satisfactory fit. The two types of equivalent constraints of both variables did not show significant differences in the model fit indices, ΔCFI < .01 and ΔSRMR < .03 (Chen, 2007;Cheung and Rensvold, 2002). These findings show sufficient measurement equivalence for the two measures across the two waves.

Results and discussion (Study 3)
As presented in Table 7, supervisor negative gossip (Wave 1) was positively and significantly related to reflective learning (Wave 2; B = .06, SE = .02, p < .01), which in turn was positively and significantly associated with job performance (Wave 3; B = .28, SE = .13, p < .05). These results support Hypotheses 1 and 2. In support of Hypothesis 3, the indirect effect of supervisor negative gossip (Wave 1) on job performance (Wave 3) through reflective learning (Wave 2) was positive and significant (indirect effect = .02, SE = .01; 95% CI = [.001, .037]). The indices (ab cs = .03 and κ 2 = .04) indicate a small to medium indirect effect size (Cohen, 1988). We repeated the analyses without the control variables and obtained similar results.
We also tested the reversed direct and indirect effects. The direct relationship between job performance (Wave 1) and supervisor negative gossip (Wave 3) was nonsignificant (B = .07, SE = .24, ns). To test the reversed indirect effect, we first examined the relationship between job performance (Wave 1) and reflective learning (Wave 2), which was non-significant (B = .04, SE = .08, ns). We then tested the relationship between reflective learning (Wave 2) and supervisor negative gossip (Wave 3), which was also non-significant (B = −.07, SE = .23, ns). Thus, the reversed indirect effect was non-significant, which indicated that reflective learning could not mediate the reversed directional association from job performance to supervisor negative gossip. In summary, these findings provide empirical evidence for the directional association from supervisor negative gossip to job performance through reflective learning, which strengthens and validates our theorizing.

General discussion
Drawing on the cultural learning perspective and social learning theory, we provide insight into the underlying learning process of gossip tapped by reflective learning. Our three field studies consistently support that supervisor negative gossip enhances employee receiver job performance through reflective learning. We offer a number of important theoretical implications for the research on gossip and reflective learning.

Theoretical implications
First, by orienting the research focus to gossip receivers, we shed light on the positive effects of workplace negative gossip on the other important party in gossip (Martinescu et al., 2014). Studies have mainly revealed the negative effects of negative gossip on gossipers' job performance (e.g. Brady et al., 2017;Grosser et al., 2010). Recently, researchers have started to propose the various positive functions of negative gossip in groups, such as delivering group norms and sanctioning norm violators (Grosser et al., 2010), imposing social influence (Burt and Panzarasa, 2008;Sommerfeld et al., 2007), and establishing social bonds (Bosson et al., 2006;Dunbar, 2004). However, limited empirical studies have investigated the effects of workplace gossip on gossip receivers. From the perspective of gossip receivers, we find that supervisor negative gossip facilitates employee receiver job performance by triggering the reflective learning process. Thus, we add to the understanding of the positive effects of negative gossip from a learning lens, which extends the nomological network of workplace negative gossip. Second, although the learning value of negative gossip has been documented in the cultural learning perspective of gossip (e.g. Baumeister et al., 2004;Stirling, 1956;Suls, 1977), researchers have not yet empirically examined the learning processes postulated. Moreover, burgeoning studies have called for future research to examine the mediating processes through which gossip influences employee work outcomes (Brady et al., 2017). Drawing on the cultural learning perspective of gossip and social learning theory, we theorize reflective learning as a novel and valuable mechanism between supervisor negative gossip and employee receiver job performance. Moreover, our findings of the mediating role of reflective learning substantiate a specific learning process of negative gossip, thereby enriching the cultural learning perspective of gossip.
Third, by focusing on the effects of negative gossip from supervisors, we introduce a novel direction for examining a prevalent yet under-investigated supervisory behavior in organizations. Apart from gossip between employees and their peers, supervisors may also often gossip with their subordinates about other absent employees (Goff and Goff, 1988). Given that supervisors are influential to employees (Ellwardt et al., 2012b), their negative gossip about other employees' failures can effectively attract employee receivers' attention and prompt their learning. By focusing on top-down gossip, we provide empirical evidence that supervisor negative gossip stimulates reflective learning and enhances the subsequent job performance of employee receivers. We thereby add new knowledge to the research on supervisory behaviors and suggest negative gossip as a possible effective managerial approach to communicate negative information with employees, promote their learning, and enhance their job performance.
Interestingly, from the gossipers' perspective, Grosser et al. (2010) found that employee negative gossip reduces supervisor-rated employee performance. Brady et al. (2017) also revealed a negative correlation between employee negative gossip (targeted at supervisors or coworkers) and peer-rated employee in-role performance. However, we reveal that gossip from supervisors positively affects employee gossip receivers. Our findings across the three field studies provide consistent empirical evidence on the propositions about the positive roles of negative gossip in triggering the learning of gossip receivers (e.g. Martinescu et al., 2014). Our results also indicate that when simultaneously including both positive and negative supervisor gossip in the analytical model, supervisor negative gossip demonstrates a positive and significant relationship with employee receiver reflective learning. However, this does not occur for supervisor positive gossip, despite the positive correlations between supervisor positive gossip and reflective learning (r = 34, p < .01, in Study 1; r = .18, p < .01, in Study 2). The results suggest that supervisor positive gossip has a positive relationship with reflective learning, but that the relationship is not as strong as that of supervisor negative gossip. This finding supports the cultural learning perspective of gossip that negative gossip has a higher learning value than positive gossip (Baumeister et al., 2004). Similarly, Bledow et al. (2017) revealed that vicarious learning through failure stories is more beneficial to learning processes and outcomes than vicarious learning through successful stories. Based on 10 years of data from 71 cardiothoracic surgeons, Diwas et al. (2013) also found that people learn more from others' failures than successes. This emerging line of studies lends empirical support to the 'bad is stronger than good' (Baumeister et al., 2004: 113) principle.
Our finding on the higher learning value of negative gossip than positive gossip is inconsistent with that of Martinescu et al. (2014), who found that positive gossip has a stronger positive effect on self-improvement value than negative gossip. However, Martinescu et al. (2014) examined horizontal gossip among peers, which is different from our focus of top-down gossip from supervisors to employee receivers. Moreover, Martinescu et al. (2014) acknowledged that their findings are inconsistent with the cultural learning perspective of gossip, which may be owing to the content of gossip, as they explained. The content of positive gossip in Martinescu et al. (2014) is competence related, which can facilitate the learning of how to improve one's competence. However, the cultural learning perspective of gossip focuses on the learning of rules, norms, and regulations. As argued, positive gossip containing norm-strengthening stories should be less instructive than negative gossip containing norm-violating stories. Martinescu et al. (2014) further suggested that the content of gossip may have a moderating effect on the relative effectiveness of positive and negative gossip in promoting learning. Therefore, future research should scrutinize the roles of the relationship between gossipers and gossip receivers (horizontal or top-down) and gossip content in the learning processes triggered by gossip.

Practical implications
Although workplace gossip is generally perceived as problematic (Beersma and Van Kleef, 2012), it is omnipresent and reflects how people informally communicate in organizations (Kniffin and Wilson, 2010). In some cases, people gossip (i.e. prosocial gossip) for the purpose of organizational development (e.g. Feinberg et al., 2012;Kniffin and Wilson, 2010). Thus, managers should understand the functional role of workplace gossip. We extend the previous research by offering explanations on the benefits of gossip to employee receivers, thereby offering some important practical implications.
Our results show that supervisor negative gossip enhances job performance through employee receiver reflective learning. Given that information transmits more rapidly through the grapevine (i.e. via gossip) than in formal channels (Beersma and Van Kleef, 2011), our findings reveal that organizational gossip is conducive to managers' effective information dissemination (Grosser et al., 2010). Through frequent negative gossip with employee receivers, managers prompt subordinates to understand the rules, appropriate behaviors, and performance standards in the workplace, which promotes 'management by gossip' (Houmanfar and Johnson, 2004: 129). Notably, we do not advocate malicious speech, but focus on the evaluative nature of negative gossip that involves supervisors' comments on the inappropriate behaviors or performance of target employees. Such value-laden information can help employee receivers learn from undesirable behaviors and effectively guide their future behaviors.
Second, we offer the practical implication regarding the learning mechanism of employee receivers through which supervisor negative gossip promotes their job performance. Learning from one's own failed experience is painful and time consuming (Anderson et al., 2011), but supervisor gossip provides employees with opportunities to learn vicariously from their colleagues' experiences. We show that upon receiving supervisor negative gossip, employee receivers engage in reflective learning and learn lessons from the absent colleagues, about whom their supervisors negatively gossip. Accordingly, the employee receivers can understand the reasons for unsatisfactory job performance, learn how to live up to their workplace's standards and expectations, and guide their own future work behaviors, thereby promoting their own job performance. We recommend that managers incorporate reflective learning into communication and training programs to train employees to engage in self-reflection when receiving value-laden information from supervisors.

Limitations and future research directions
This study is not without limitations. First, we used a cross-lagged panel design in Study 3 to test the directional association from supervisor negative gossip to job performance through employee receiver reflective learning. However, as mentioned, the collection of cross-lagged panel data in organizations is exposed to a high attrition risk. In fact, the total attrition rate of Study 3 was 35.3%. This raised the concern of nonresponse bias. To assess this issue, we tested the potential non-random sampling effects following Goodman and Blum's (1996) four-step procedure, which has been widely adopted in previous studies (e.g. Füllemann et al., 2015;Holman et al., 2010). We found that the non-random sampling issue might have existed, such that the respondents who reported higher frequent supervisor negative gossip were more likely to remain in the subsequent surveys. In addition, the non-random sampling might have affected the means of supervisor negative gossip and employee receiver job performance. However, it did not affect the variances and relationships among the studied variables. The results also indicate that there were no differences between the stayers and leavers in terms of the demographics of age, sex, education, and organizational tenure. Alternative hypotheses may be considered in the future (Goodman and Blum, 1996). As we obtained consistent results across the three studies, the alternative hypotheses may not be a serious issue here. Nevertheless, we encourage future studies with more rigorous longitudinal or experimental designs to replicate our findings.
Second, the sample sizes are relatively small in the three field studies (i.e. N = 212 for Study 1, N = 243 for Study 2, and N = 132 for Study 3). This poses the issue of low statistical power (Cohen, 1988). However, the potential problem of insufficient statistical power (i.e. likelihood of falsely concluding given effects or Type II errors) may not be a major issue in this study, as the findings across the three studies are highly replicated via different samples from various industries.
Third, our three field studies were conducted in the Chinese context. Given that our theorizing is not tied to any cultural dynamics, we expect our results to be generalizable to other cultures. On the basis of multicultural samples, Brady et al. (2017) revealed that the effects of workplace gossip on various organizational outcomes (i.e. uncertainty, emotion validation, self-esteem, norm enforcement, networking, influence, organizational justice, performance, deviance, and turnover) are invariant among cultures. Nevertheless, we encourage future research to replicate our findings in other cultural contexts.
Finally, we focus on the positive effects of supervisor negative gossip on employee receivers. We encourage future studies, as extensions of our study, to integrate the pros and cons and explore the possible boundary conditions involved to provide a full understanding of the effects of supervisor negative gossip on employee receivers.

Conclusion
We extend the scholarly knowledge about the effects of supervisor negative gossip on employee receivers. On the basis of a learning perspective, we illustrate that employee receiver reflective learning is an important mediating process that links supervisor negative gossip to job performance. Thus, we paint a nuanced picture of the positive effects of workplace negative gossip. National Natural Science Foundation of China (grant number 71902092) and the Outstanding Researcher Support System, Nottingham University Business School China, awarded to Jie Wang.