The Pain and Gain of Gossip:
Different effects of praise and blame gossip on prosocials’ and proselfs’ contributions to the group
Daisy van der Veeken University of Amsterdam
Master thesis
Work- & Organizational Psychology Student number: 0580503
mevr. dr. B. Beersma mevr. drs. J. Koen MSc
E-mail: daisyvanderveeken@gmail.com August 24, 2014
Abstract
Gossip is an unavoidable group process, which organizations struggle to manage. I investigated the disparate research findings on gossip and propose that gossip can have a feedback function in work groups, which has positive or negative effects on group members’ contributions to group goals depending on their social value orientation and the identifiability of their contributions. To test this theory, 194 participants took part in a laboratory experiment with a 2 x 2 x 2 factorial design. Participants were led to believe that they received praise or blame gossip about a fellow group member and that their own contributions to group goals were identifiable to the other group members or not, before being confronted with a social dilemma. Furthermore, their social value orientation was measured. As expected, proself participants contributed more in response to blame gossip, especially when their contributions were identifiable, while prosocial participants contributed more in response to praise gossip regardless of their identifiability. These positive effects were mediated by the internalized norm of social responsibility. In contrast, blame gossip amotivated prosocial participants and made them feel more like a sucker, which, in turn, lowered their contributions to avoid being exploited by a norm violator, especially when their contributions were unidentifiable. Praise gossip taught proself participants that their contributions were dispensable for the team project to succeed. This type of gossip amotivated them, when their contributions were unidentifiable, and was even able to make them feel more like a sucker, which made them free-ride on the efforts of others. With this study I aim to provide organizations with valuable information about the effects of gossip in work groups to minimize the pain and optimize the gain of gossip.
Contents
The Pain and Gain of Gossip... 5
The Feedback Function of Gossip ... 7
Differences in Gossip Processing ... 8
Group Opinion Concerns ... 9
Social Responsibility ... 11
Reciprocity ... 13
Sucker Role ... 15
The Orientation of Work Motivation ... 16
Autonomous versus Controlled Motivation ... 17
Amotivation ... 18
Method ... 21
Participants & Design ... 21
Procedure ... 21
Measures ... 25
Results ... 30
Descriptive Statistics and Treatment of the Data ... 30
Manipulation Checks ... 32
Contributions to the Group ... 32
Group Opinion Concerns ... 35
Social Responsibility ... 36
Reciprocity ... 39
Sucker Role ... 40
Autonomous Motivation ... 41
Amotivation ... 43
Conditional indirect effects ... 44
Discussion ... 47
Limitations and future directions ... 52
Conclusion ... 56
References ... 57
Footnotes ... 72
Appendix ... 81
Exploratory Research: Measures ... 81
Exploratory Research: Results Motives to Gossip ... 82
The Pain and Gain of Gossip
Although we may not like to admit it, we all know it is true: people like to gossip (see e.g., Dunbar, Duncan, & Mariott, 1997; Emler, 1994; Fine, 1977; Goodman & Ben-Ze'ev, 1994; Wilson, Wilczynski, Wells, & Weiser, 2000). This informal exchange of evaluative
information about absent third parties (see for a review Eder & Enke, 1991; Foster, 2004; Turner, Mazur, Wendel, & Winslow, 2003; Michelson, Van Iterson, & Waddington, 2010) seems to be an unavoidable product of socially interconnected people (e.g., Bergman, 1993; Dunbar, 2004); making gossip part of just another day at the office.
Organizations today, however, are confronted with contradictory implications of gossip. This group interaction process can be critical for group performance; it can either have detrimental or advantageous effects on employees’ contributions to their team and
organization, which calls for the question: when and why does gossip have positive and/or negative effects on individuals’ contributions to group goals? For example, on the one hand research shows that gossip can lower employee morale and work productivity (e.g. Baker & Jones, 1996). On the other hand, recent studies focus on the positive effects of gossip and demonstrate that it can serve important functions, for example, heighten group members’ cooperation on collective tasks (Beersma & Van Kleef, 2011a; Feinberg, Willer, Stellar, & Keltner, 2012). Due to a lack of knowledge on how to use this unavoidable group process to their advantage, many organizations today struggle with managing gossip (see e.g. Michelson, et al., 2010).
The purpose of the present study is threefold: First, I reconcile these disparate research findings on gossip by examining both positive and negative effects of gossip on individuals’ contributions to group goals and the individual and situational factors that influence the strength of these effects (i.e. moderators). Gossip occurs in situations where three parties are intertwined: the gossiper, the gossip target, and the gossip recipient (gossip triad; Kurland & Pelled, 2000). In this study I will focus on the gossip recipient. Second, I extend previous
empirical studies on gossip in work groups by examining not only the influence of blame gossip but also praise gossip on contributions to group goals. The valence of the reputational information people exchange can be either negative (i.e. blame gossip) or positive (i.e. praise gossip) (see for a review Eder & Enke, 1991; Foster, 2004; Michelson et al., 2010; Turner et al., 2003). 1 Recent empirical studies show that blame and praise gossip can effectively influence the perceptions and behaviors of people; for example, they show that both can have a strong influence on recipients’ visual perception of the gossip target (Anderson, Siegel, Bliss-Moreau, & Barrett, 2011), impact their target impression formation (Brandt, Vonk, & Van Knippenberg, 2009; 2011) and augment people’s opinion of their colleagues (Burt & Knez, 1995). Furthermore, Sommerfeld, Krambeck, Semman and Milinski (2007)
demonstrated that people use information provided by praise and blame gossip to decide how much cooperation they will provide to a cooperative or non-cooperative partner (Sommerfeld, Krambeck, Semman, & Milinski, 2007). Based on these research findings, it is probable that gossip valence can also influence how people adjust their level of cooperation in work groups, whereby individual contribution behaviors do not only influence one interdependent other (i.e. the gossip target), as in the study of Sommerfeld et al., but also has an impact on other
interdependent group members. However, studies on gossip in work groups mainly focus on negative reputational information sharing. For instance, Beersma and Van Kleef (2011a) demonstrated that the mere threat of blame gossip can effectively deter selfish behavior and promote cooperation. As an exception, Kniffin and Wilson (2005, 2010) provide case study evidence that people can use both blame and praise gossip for the good of the group: they report that a team of rowers instigated gossip blaming a slacking group member and praising the hardworking group members to uphold group norms, especially when confronted with a norm violator. Thus, blame gossip is not the only way for people to uphold group norms, there are two ways: they can spend time maintaining group-serving norms by criticizing norm violating behavior using blame gossip, but also enforce the same norms by praising proper
group behavior using praise gossip Therefore, it is striking that empirical research on gossip in work groups has ignored praise gossip as an effective way to influence group members’ contributions to group goals (also see Wilson et al., 2000).Third, given the scarcity of empirical research providing an explanation for the effect of gossip on employees’
contributions, I examine a variety of possible mediators to explain how an external event as gossip can take on an internal psychological significance and answer the question why blame and praise gossip can both have positive and negative effects on employees’ contributions to group goals. My aim is to provide organizations with valuable information about the effects of gossip in work groups; especially since informal communication, collaboration and teamwork has become fundamental for organizational effectiveness (Baker, 2002). Knowing when and why gossip affects group performance should help create organizational policies minimizing the pain and optimizing the gain of gossip.
The Feedback Function of Gossip
Taking a cultural perspective, Baumeister, Zhang and Vohs (2004) argued that gossip serves as a means of cultural learning. By hearing about the triumphs and misadventures of others, people gain valuable information about their environment and learn which rules, norms and guidelines they have to follow to function effectively within the complex structures of human social life (also see Foster, 2004). As mentioned earlier, recent studies show that people indeed use information provided by gossip to guide their personal perceptions and behaviors. I expect this function of gossip to be no different in an organizational context. At the office, people hear through the grapevine about the individual performance of their fellow group members (i.e. individual feedback in groups; Nadler 1979). This type of feedback teaches them how to regulate their own behaviors and make their performance and team projects into a success (see Kluger & De Nisi, 1996; Nadler, 1979). For example, when people hear that a current group member tarnished his or her reputation in the past by free riding – not putting
much effort into team projects whilst still reaping the benefits of these projects – this may motivate them to put in some extra effort in their current team project to prevent making the same mistake and endure the same costs to their own reputation. However, feedback is a double-edged sword (Kluger & De Nisi, 1996; Nadler, 1979); thus the same blame gossip may also have an adverse effect on recipients’ contribution behaviors and, in turn, on group performance. In response to free-riding, people may also withdraw effort to avoid being taken advantage of by another group member (i.e. sucker effect; see Kerr, 1983). The same goes for praise gossip: on the one hand, hearing about the triumphs of a hardworking group member can encourage others to work harder, because they do not want the group to fail due to their lack of contribution (i.e. Köhler effect, see Köhler, 1926). On the other hand, they may perceive their contributions to be dispensable for the team project to succeed (i.e. free-riding; see Kerr & Bruun, 1983). Therefore, I agree with Baumeister et al. (2004) that gossip can be understood as an extension of observational learning and complement their theory by
proposing that gossip has a feedback function in work groups, which can have positive or negative effects on group members’ contributions to group goals.
Differences in Gossip Processing
The influence of feedback on performance depends on the situational and/or individual characteristics of its recipients (Kluger & De Nisi, 1996, Nadler, 1979). Earlier studies suggest that recipients’ reactions to gossip, specifically their contribution behaviors, depend on the identifiability of their actions (e.g. Beersma & Van Kleef, 2011a) and their social value orientation (Feinberg et al., 2012) – group members’ preference for a distribution of valued outcomes for self and others (see e.g. Beersma, Conlon, & Hollenbeck, 2007; Druckman, 1994; Messinck & McClinktock, 1968; McClintock, 1976; Van Lange, Otten, De Bruin & Joireman, 1997; Weingart, Bennett, & Brett, 1993). When the contributions individuals make to the group are identifiable and their group members are prone to gossip, they will have
concerns about their reputation and group expulsion, which lead them to contribute more resources to the group (e.g . Beersma & Van Kleef, 2011a). However, when individuals’ contributions are unidentifiable, they are able to hide in the crowd leaving their reputations unharmed. Furthermore, Feinberg et al. (2012) demonstrated that this positive effect of blame gossip and identifiability on contributions to group goals is probably driven by group
members with a proself orientation, who attend to the utility of their own outcomes, whether or not relative to others. Their participants were placed in a type of social dilemma that puts their individual interests against the interest of the group (also see Dawes, 1980; Frank, 1988; Komorita & Parks, 1996; Willer, 2009). Those who scored lower on prosociality contributed significantly more when they knew that an observer could easily comment on their behavior to future interaction partners. The same threat of blame gossip, however, had no effect on the contributions of those who scored higher on prosociality (i.e. prosocials), who attend to the utility of joint outcomes. To have a better understanding of these behavioral differences, I will discuss a variety of possible mediators in the relation between gossip and contributions to group goals and the way identifiability and social value orientation (i.e. moderators) create the appropriate conditions for the indirect effect of gossip on group members’ contributions (see Figure 1).
Group Opinion Concerns
Most researchers suggest that group opinion concerns are the main reason why people increase their contributions level in response to blame gossip. These concerns refer to both social belonging as a group member as well as his or her reputation (Beersma & Van Kleef, 2011a), which are both very important to people’s social self-concepts (Sedikides, 2002). Gossip can pose a threat to the image others generally have about someone and even lead to exclusion from their group (see De Cremer & Sedikides, 2008; De Cremer & Tyler, 2005a, 2005b; Hardy & Van Vugt, 2006; Van Vugt & Hardy, 2010). Therefore, when people expect
to be gossiped about, it raises their group opinion concerns and leads them to contribute more to the group. Alternatively, receiving blame gossip about another group member may also make individuals more aware of the effect such a gossip could have on their own reputation, and lead them to increase their contributions to comply with the group norm.
Gossip is often instigated to share and validate reputational information (see Dunbar 1996; Wilson et al., 2000) or, among other motives, it can be used to protect the group from norm violators (see Beersma & Van Kleef, 2012; Dunbar, 2004; Gluckman, 1963; Keltner, Van Kleef, Chen, & Kraus, 2008). Receivers of gossip rely on this useful information as a guide to avoid interacting with selfish and manipulative individuals and instead develop relationships with team players (Sommerfeld et al., 2007; Wedekind & Milinski, 2000). In addition, when individual contribution behaviors are identifiable for interdependent others (e.g. fellow group members), those who would otherwise take advantage of group resources restrain their self-serving behavior and contribute considerably more to the common good to procure or even enhance their reputation and avoid group exclusion(e.g., Barclay, 2004; Barclay & Willer, 2007; Hardy & Van Vugt, 2006; Piazza & Bering, 2008; Wedekind & Milinski, 2000; Willer, 2009). Furthermore, Beersma & Van Kleef (2011a) demonstrated that group opinion concerns actually mediate the interactive effect of blame gossip and
identifiability on people’s contributions to group goals. Based on these previous studies, I expect that receiving negative information about the past performance of a fellow group member may trigger the same effect: heighten receivers’ group opinion concerns and motivate them to put in some extra effort to prevent harm to their own reputation. This effect is likely to be driven by receivers with a proself orientation, because they score higher on impression management questionnaires indicating that proselfs care more about their reputations than prosocials (Willer, Feinberg, Flynn, & Simpson, 2011), and contribute more when their reputation is at jeopardy (Simpson & Willer, 2008). Therefore, I firstly predict a three-way interaction, implying that blame gossip increases contributions to group goals of proself
recipients, especially when their contributions are identifiable for their group members, relative to prosocial recipients (Hypothesis 1a), relative to praise gossip (Hypothesis 1b), and relative to unidentifiable contributions (Hypothesis 1c). Secondly, I predict that the interactive effect of gossip, social value orientation and identifiability on contributions to the group is positively mediated by group opinion concerns, such that blame gossip heightens proselfs’ group opinion concerns and, in turn, their contributions to the group, especially when their contributions are identifiable for their group members, relative to prosocials (Hypothesis 2a), relative to praise gossip (Hypothesis 2b), and relative to unidentifiable contributions
(Hypothesis 2c).
Social Responsibility
In addition, I suggest an alternative reason why people increase their contribution level in response to blame gossip: social responsibility, which has been found to be an important internalized social norm for the functioning of groups and societies (Fleishman, 1988; Ostrom, 1998). This tendency to care about both self and others (e.g. Fiske, Kitayama, Markus, & Nisbett, 1998) seems to be an important antecedent of benevolence and elevates levels of generosity (see Van Lange, 2000). Feelings of social responsibility are “most likely to be activated when decision-makers interpret their decisions in terms of the consequences of their behavior for others (see e.g. Schwartz, 1977), and not only when others are perceived to be in need (Berkowitz & Daniels, 1963; 1964)” (De Cremer & Van Lange, 2001, p. S14). Weber and Hertel (2007) also argue that people who are mainly concerned with social welfare perceive their contributions to group goals to be less dispensable and therefore contribute more to the group than people who are mainly concerned with individual welfare. Therefore, it is not surprising, that De Cremer and Van Lange (2001) found preliminary evidence showing that social responsibility represents a psychologicaldimension underlying, at least partly, the behavioral differences between prosocials and proselfs, and, when confronted with
a social dilemma, prosocials generally feel more socially responsible for the group and, in turn, engage in cooperative behavior and enhance joint outcomes (also see Fleishman, 1988).
Blame gossip, however, can heighten the social responsibility of proself group
members. Receiving negative feedback about a slacking group member may signal to proself recipients that the success of their group project, and thereby their own success, is in danger. To obtain these valued outcomes, proselfs may feel more socially responsible and contribute more to the group. This act to ensure the group’s interest, when people expect others to perform poorly is called social compensation (Karau & Williams, 1997; Lount, Park, Kerr, Messé, & Seok, 2008; Todd, Seok, Kerr, & Messé, 2006; Williams & Karau, 1991). When people do so to ultimately serve their own self-interest it is also referred to as enlightened self-interest (see e.g. Frimer, Walker, Dunlop, Lee & Riches, 2011). Research suggest that the social compensation effect is based on individualistic motives (i.e. competition and status), which supports my idea that proselfs can socially compensate for other group members.
In contrast, the Köhler effect (or indispensability effect) – whereby people contribute more to the group to prevent group failure due to their lack of contribution (see Köhler, 1926) – is based on social motives (i.e. collective welfare; see Kerr, Messé, Seok, Sambolec, Lount, & Park, 2007; Nijstad, Van Vianen, Stroebe, & Lodewijkx, 2004; Weber & Hertel, 2007). These social motives may inspire prosocials to contribute more in response to praise gossip. Receiving positive feedback about a dedicated group member may signal to prosocial recipients that they need to contribute more to comply with the high group norm in order for the group to succeed. To prevent the group from failing because of them, prosocials may feel more socially responsible and, in turn, heighten their contributions to group goals. Based on these research findings, I predict a two-way interaction effect of gossip and social value orientation on contributions to group goals, such that praise gossip heightens the contributions of prosocial recipients, relative to proself recipients (Hypothesis 1d), and relative to blame gossip (Hypothesis 1e). Furthermore, I predict that the interactive effect of gossip and social
value orientation on recipients’ contributions to group goals is positively mediated by social responsibility; such that blame gossip heightens the social responsibility of proself recipients, and, in turn, their contributions to the group, relative to prosocial recipients (Hypothesis 3a), and relative to praise gossip (Hypothesis 3b); while praise gossip heightens the social
responsibility of prosocials and, in turn, their contributions, relative to proselfs (Hypothesis 3c), and relative to blame gossip (Hypothesis 3d).
Reciprocity
Reciprocity is defined as the tendency to reward or punish an interdependent other in
accordance with their level of contribution (see Bowles & Gintis, 2004; Fehr & Fischbacher, 2003; Fehr & Gachter, 2002; Gintis, 2000; Gouldner, 1960). Like social responsibility,
reciprocity is considered an internalized social norm important for group functioning, because it discourages norm-violating behavior and stimulates desired cooperative teamwork
(Fleishman, 1988; Ostrom, 1998). For example, when group members are confronted with free-riding, they tend to react negatively towards these selfish and unfair actions that violate the norm of equality (i.e. similar amounts of efforts and/or contributions; see Haidt, 2001; Horberg & Keltner, 2007). These violations can cause a negative emotional response (e.g. feelings of anger) and disciplinary behavior to punish the free-rider and ultimately restore equality in the group members’ outcomes (see Horberg & Keltner, 2007; Lerner, 1980; Miller, 2001; Stouten, De Cremer, & Van Dijk, 2005), even when such behavior may be costly to their own self-interest (Fehr & Fischbacher, 2004; Fehr & Gachter, 2002;
Fetchenhauer & Huang, 2004; Frank, 1988; Miller, 2001). When people have the tendency to punish a fellow group member (or interdependent other) for violating social norms it is referred to as negative reciprocity, while the tendency to reward other group members for proper group behavior (e.g. high contribution levels) is referred to as positive reciprocity (see Blau, 1964; Perugini, Gallucci, Presaghi, & Ercolani, 2003). Sommerfeld et al. (2007)
demonstrated that people do not only adjust their contribution levels in response to first-hand information, but also in response to gossip: when their participants received positive
comments about an interdependent other’s past contribution behaviors, they were more likely to cooperate over a series of trials than when they received negative comments, even when they had access to contradictory first-hand information. Therefore, I expect that positive and negative reciprocity represent psychological dimensions underlying behavioral assimilation in response to gossip.
Although people generally prefer cooperation and fairplay (Henrich et al., 2001), numerous studies demonstrate that, relative to proselfs, people with a prosocial orientation are more likely to experience feelings of reciprocity and engage in the same level of cooperation (e.g., De Cremer & Van Lange, 2001; Kelley & Stahelski, 1970; Liebrand, 1986; Messick & McClintock, 1968; Pruitt & Kimmel, 1977; Simpson & Willer, 2008; Van Lange, 2000) or expected level of cooperation as an interdependent other (e.g., De Dreu & McCusker, 1997; Van Lange & Liebrand, 1989); when exposed to acts of unfairness, prosocials experience greater affective reactions (Feinberg et al., 2012), and respond with increased amygdale activity (Haruno & Frith, 2010), suggesting that, relative to proselfs, prosocials intuitively divert from inequality (Adolphs, Tranel & Damasio, 1998; Anderson & Phelps, 2002). In addition, after engaging in blame gossip to punish the free-rider, prosocials felt a greater sense of relief (Feinberg et al., 2012). Praise gossip is likely to have the opposite affect: when confronted with a prisoner’s dilemma (i.e. a type of social dilemma) in which an individual’s choice of contribution could be influenced merely by a single previous choice of their partner or by a series of trails (e.g. Kelly & Stahelski, 1970), prosocials accordingly engaged in behavioral assimilation. When both individuals simultaneously needed to make a single choice (e.g. De Dreu & McCusker, 1997; Van Lange & Liebrand, 1989), prosocials expected greater cooperation from the interdependent other and anticipated on this expectation by contributing significantly more relative to proselfs. This also corresponds with the integrative
model of social value orientation, stating that prosocials do not only assign positive weight to the outcomes for themselves, like proselfs, but also to the outcomes for others and to equality in outcomes in interdependent situations (Van Lange, 1999). Based on these research
findings, I expect that receiving gossip about a group member’s past cooperative or non-cooperative choices may also bring about expectations about their future cooperation level, and make prosocials adjust their own cooperation level accordingly. Therefore, I predict an interactive effect of gossip and social value orientation on recipients’ contributions to group goals that is mediated by reciprocity; such that praise gossip heightens prosocials’ positive reciprocity and, in turn, heightens their contributions to the group, relative to proself recipients (Hypothesis 4a), and relative to blame gossip (Hypothesis 4b). In contrast, blame gossip heightens prosocials’ negative reciprocity, and, in turn, lowers their contributions to the group, relative to proself recipients (Hypothesis 4c), and relative to praise gossip (Hypothesis 4d).
Sucker Role
Prosocial group members may also lower their contributions in response to blame gossip, not to punish a norm violator, but to avoid feeling like a sucker. According to Orbell and Dawes (1981), group members feel taken advantage of when they notice a free-rider in their midst, especially when this other group member is fully capable of contributing (also see Kerr, 1983). They, who do behave in accordance with the norms, do all the work and invest their resources while the free-rider exploits them. These group members suffer from motivation loss. To prevent or dissolve being manipulated into the so-called sucker role, in which they are being exploited by a free-riding fellow group member, they reduce their own efforts and contributions (i.e. sucker effect; Kerr, 1983).Having established that gossip is a powerful source of information that can influence recipients’ contribution behaviors; I expect blame gossip about a free-riding group member to have the same sucker effect (i.e. reduction of
group members’ contribution) as first-hand information in work groups. This seems to apply especially to prosocial recipients who, as mentioned earlier, react more negatively towards selfish and unfair acts. Therefore, I predict that the interactive effect of gossip and social value orientation on recipients’ contributions to group goals is negatively mediated by sucker role; such that blame gossip heightens prosocials’ feelings of being a sucker and, in turn, lowers their contributions to group goals, relative to proselfs (Hypothesis 5a), and relative to praise gossip (Hypothesis 5b).
The Orientation of Work Motivation
I propose that the differences between prosocials and proselfs in response to gossip in terms of their underlying psychological dimensions and their resulting contribution behaviors can also be explained by the Self-Determination Theory (SDT) (see e.g. Deci & Ryan, 1985; Gagné & Deci, 2005; Ryan & Deci, 2000a; 2000b). I will first provide a short description of the SDT, and in the next paragraphs apply this theory to the interactive effect of gossip, social value orientation and identifiability on contributions to group goals. In accordance with the SDT, a macro theory of work motivation, individual differences can be attributed to the process of internalization – whereby people take in an external group norm (i.e. values, attitudes, regulations) and transform it into their own, so that it will emanate from their sense of self. Depending on the degree to which such an external regulation has been internalized, people’s motivation to contribute to group goals may vary on a controlled-to-autonomous continuum (see Gagné & Deci, 2005; George, 1995). “SDT postulates that autonomous motivation and controlled motivations differ in terms of both their underlying regulatory processes and their accompanying experiences, and it further suggests that behaviors can be characterized in terms of the degree to which they are autonomous versus controlled” (Gagné & Deci, 2005, p. 334).Whereas some people put effort in their job to get approval and respect from others (e.g. supervisor, colleagues, family, clients) or to avoid being criticized (i.e.
controlled motivation), others personally want to put effort in their job, because they consider it to be important, it aligns with their personal values or because it has personal significance to them (i.e. autonomous motivation).2 “Earlier research suggests that the latter case entails personal endorsement and a feeling of choice, whereas the former involves mere compliance with an external control” (Ryan & Deci, 2000a, p. 60). This type of work motivation
corresponds with the classic type of extrinsic motivation, whereby people’s behavior is initiated and maintained by contingencies external to themselves (e.g., I contribute to group goals when other group members are watching) (Gagné & Deci, 2005). Although SDT differentiates orientations (or quality) of work motivation, it also emphasizes that the amount (or quantity) of motivation matters (Deci & Ryan, 2000), suggesting that all types of
motivation can also be added, representing a person’s total amount of motivation (also see Van den Broeck, Lens, De Witte, & Van Coillie, 2013).
Autonomous versus Controlled Motivation
According to George (1995), negative feedback is “very unlikely to enhance feelings of competence” (p. 331); it makes “the controlling nature of the punishment most salient” (p. 331), and “may discourage social loafing to the extent that it increases the value of individual effort expenditure (e.g. as a means to avoid punishment)” (p. 330). Thus, negative feedback may heighten people’s controlled motivation to contribute to group goals. As mentioned earlier, this positive effect of negative feedback is likely to be especially pronounced in proselfs, who are more likely to engage in selfish behaviors and reputational management. Therefore, I expect blame gossip to heighten recipients’ controlled motivation – i.e. feel like they have to conform to the communicated group norm – when their contributions are identifiable for their fellow group members, especially when recipients have a proself
orientation. One could also argue that proselfs actually want to contribute to group goals (i.e. autonomous motivation) in response to blame gossip, because this type of feedback informs
proselfs that the group’s outcomes, more importantly their own, are in danger. To ensure these desired outcomes, proselfs are likely to put in some extra effort.
In contrast, prosocials are more likely to become autonomously motivated in response to positive feedback about another group member. Positive feedback concerning one’s effectiveness and/or capabilities does “not threaten self-determination and may enhance feelings of competence leading to increased motivation” (George, 1995, p. 331). Praise gossip may have a similar effect; hearing that fellow group members did a good job, may give prosocials a positive image of their group’s performance and make them want to comply with the communicated high group norm (i.e. autonomous motivation).
Based on the SDT, I predict three interaction effects: the first two are two-way interactions between gossip and social value orientation of gossip recipients on contributions to group goals that are mediated by autonomous motivation, implying that praise gossip increases prosocials’ autonomous motivation and, in turn, their contributions to group goals, relative to proselfs (Hypothesis 6a), and relative to blame gossip (Hypothesis 6b). Blame gossip will increase proselfs’ autonomous motivation and, in turn their contributions to group goals, relative to prosocials (Hypothesis 6c), and relative to praise gossip (Hypothesis 6d). The third is a three-way interaction between gossip, social value orientation and identifiability on contributions to group goals that is mediated by controlled motivation, implying that blame gossip increases proselfs’ controlled motivation and, in turn, their contributions, but only when their contributions are identifiable for their group members, relative to prosocials (Hypothesis 7a), relative to praise gossip (Hypothesis 7b), and relative to unidentifiable contributions (Hypothesis 7c).
Amotivation
In sharp contrast to the other orientations of work motivation is amotivation, which is defined as a complete lack of self-determination, intention and motivation to exert any effort (see
Gagné & Deci, 2005). This lack of motivation can be consequential to prosocials receiving blame gossip. As mentioned earlier, they experience greater aversion towards wrong doings of others than proselfs, and thus it is likely that hearing about the selfish acts of a fellow group member may trigger a greater level of amotivation in prosocial group members to contribute to their group goals. Therefore, I predict an interactive effect of gossip and social value orientation on recipients’ contributions to group goals that is negatively mediated by amotivation; such that blame gossip heightens prosocials’ amotivation and, in turn, lowers their contributions to the group, relative to proselfs (Hypothesis 8a), and relative to praise gossip (Hypothesis 8b).
Amotivation can also help explain the negative effect of praise gossip on the
contribution behaviors of group members with a proself orientation. Praise gossip can be used as an opportunity to free ride. This type of selfish behavior occurs when group members perceive their contribution to be dispensable to the group project (Kerr & Bruun, 1983; Olson, 1965). Hardworking group members can signal to other group members that their
contributions are not needed for the group to succeed, and consequently make them reduce their efforts (see Olson, 1965; Orbell & Dawes, 1981; Shepperd, 1993; Stroebe & de Frey, 1982). Group members with a proself orientation are more likely to engage in this type of behavior, because, unlike prosocials, they only assign positive weight to their own
self-interests (Van Lange, 1999). Furthermore, prosocials construe social dilemmas differently and use different decision making heuristics, making it highly unlikely that they will engage in free-riding.Prosocials use and recall cooperative heuristics (e.g. ‘take a problem-solving approach’; ‘play fair’; ‘share and share alike’; De Dreu & Boles, 1998) and construe social dilemmas as a moral issue, whereby cooperation is seen as good behavior and
non-cooperation (or competition) is seen as bad behavior(Liebrand, Janssen, Rijke, & Shure, 1986; Liebrand & Van Run, 1985; McClintock & Liebrand, 1988; Sattler & Kerr, 1991; Van Lange & Liebrand, 1989). It is therefore, unlikely for prosocials to engage in a wrongful
behavior as free-riding in response to praise gossip that depletes joint outcomes of the group and goes against their belief in fair play as a resolution to social dilemmas. In contrast, proselfs use and recall competitive heuristics (‘your gain is my loss’ and vice versa; ‘the winner takes all’; ‘never trust your opponent’) and construe social dilemmas in terms of power, whereby cooperation is seen as a sign of weakness and competition as a sign of strength. Free-riding for them seems like a smart way to manipulate the other group members into getting what to want, which gives them a sense of power. Of course, when their lack of contribution is identifiable for their fellow group members, their potential power and
reputation are in jeopardy. Therefore, I predict a three-way interaction between gossip, social value orientation and identifiability on contributions to group goals, implying that praise gossip lowers proselfs’ contributions, especially when their contributions are unidentifiable for their group members, relative to prosocials (Hypothesis 1f), relative to blame gossip (Hypothesis 1g), and relative to identifiable contributions (Hypothesis 1h). Furthermore, I predict that this three-way interaction is negatively mediated by amotivation, such that praise gossip heightens proselfs’ amotivation and, in turn, lowers their contributions when their contributions are unidentifiable, relative to prosocials (Hypothesis 8c), relative to blame gossip (Hypothesis 8d), and relative to identifiable contributions (Hypothesis 8e). Thus, praise gossip may not only have a positive effect but also an adverse effect on group performance.
In summary, I expect that prosocial and proself individuals differ in their motivational and behavioral responses to praise and blame gossip, and that proselfs are more sensitive to the identifiability of their actions. Specifically, I expect that blame gossip heightens proselfs’ contributions, especially when their contributions are identifiable, and that this effect is positively mediated by group opinion concerns, social responsibility, and the orientation of their work motivation (i.e. autonomous motivation, controlled motivation). Furthermore, I expect that praise gossip heightens prosocials’ contributions to the group via social
blame gossip lowers prosocials’ contributions to the group and that this effect is negatively mediated by negative reciprocity, amotivation and sucker role, while praise gossip lowers proselfs’ contributions, via amotivation, when their contributions are unidentifiable. Figure 1 is a process model that depicts these main proposed effects of gossip. To test this model, participants took part in a laboratory experiment, in which they were led to believe that they received praise or blame gossip about a fellow group member and that their own contributions to group goals were identifiable to the other group members or not, before being confronted with a social dilemma. They also filled in a questionnaire to assess their social value
orientation. This method will be further elaborated upon and my results will be discussed in the following sections.
Method Participants & Design
A total of 194 participants (66 males and 128 females, mean age 22,07 years)took part in this study at the laboratory of the University of Amsterdam (UvA) for 0.5 course credit or
monetary compensation (5 euros). The participants were randomly assigned to four
experimental conditions of a 2 (gossip valence: praise vs. blame) by 2 (identifiability: absent vs. present) factorial design using a double-blind procedure. Furthermore, participants’ social value orientation was measured, resulting in a 2 (gossip valence: praise vs. blame) by 2 (identifiability: absent vs. present) by 2 (social value orientation: proself vs. prosocial) factorial design.
Procedure
Upon arrival at the laboratory, participants were welcomed by the experimenter and seated in individual cubicles in front of a computer. Here they were asked to fill out an informed consent form and told that from that point on all instructions were presented on the computer
screen. As a cover story, the participants read that the purpose of the study was to examine the effects of flexible working (i.e. a policy that allows employees to determine when and where they work) and information channels by comparing a computer-mediated interaction and a face-to-face interaction in work groups (adapted from Beersma & Van Kleef, 2011a; Van Kleef, De Dreu, & Manstead, 2006). The participants would therefore interact in a group with three other participants whom were seated in separate cubicles during the first part of the experiment. During the second part of the experiment they would interact face-to-face with their group. In reality, participants did not interact with anybody else and the second part of the experiment would not actually take place.
Before introducing the participants to their alleged group members, they had to fill out a nine item Decomposed Games Measure (cf. De Cremer & Van Lange, 2001; Messinck & McClintock, 1968; Van Lange, 1999; Van Lange & Kuhlman, 1994) to assess their social value orientation. Thereafter, I used an established procedure to enhance the salience of the group for participants (adapted from Beersma & Van Kleef, 2011a; Homan, Greer, Jehn, & Koning, 2010). To make participants feel that they belonged to a group, they were instructed to fill out their (nick)name and gender and take their own photos with a webcam. Their photo and name would then be displayed along with those of three other people who were
introduced as their group members. The photos and corresponding names included two male students and one female student. In reality, the photos and personal information of the other group members were always the same and the responses that allegedly came from the other group members were preprogrammed.
After participants were digitally introduced to their group members, they received instructions about the group task: a common good dilemma (adapted from Mulder, Van Dijk, De Cremer, & Wilke, 2006). Participants read their group would participate in a decision-making task in which every member needed to make a decision. Each group member received 100 chips as endowment at the start of the session and each chip provided a chance at a price
of 75 euros. They also read that they were free to contribute as many chips as they wanted into two accounts: a personal account (i.e. defect) and a group account (i.e. cooperate in order to establish the common good). The total number of chips given to the group by the four individual group members would be multiplied by two and divided equally among all members, regardless of their contribution. However, the total number of chips given to the group would only be multiplied by two if the group members together contributed at least 200 chips to the group account. The number of chips they decided to keep would accrue
completely to themselves and every chip would give them a chance of winning a price of 75 euros. To enhance the understanding of this situation, some outcome possibilities were explained to the participants (e.g. the issue of free-riding).
To manipulate identifiability, the instructions included information about whether the participants’ group members would or would not be informed about how many chips the participants contributed to the group. To manipulate gossip, participants were led to believe that they could send and receive a private chat-message to and from every group member before engaging in the decision-making task. In all conditions participants would first get the opportunity to send their fellow group members a chat-message. They could also decline to send any messages by pressing ‘x’. Thereafter, participants would receive a chat-message of their two male group members. One of the chat-messages was of a general nature and did not include any type of gossip (i.e. “Here we go.”). The second message contained a gossip comment about the female group member and her previous contribution behaviors. The valence of the gossip chat-message differed across conditions; in the blame gossip condition the gossip had a negative valence and in the praise gossip condition the gossip had a positive valence. These preset gossip comments were adapted from Sommerfeld et al. (2007) and based on actual gossip comments typed by their participants: “It is not relaxed to work with Natasha. Not a team player!” (i.e. blame gossip comment) and “Natasha is a relaxed chick. A
real team player!” (i.e. praise gossip comment).3 The participants also saw that their female group member (i.e. the gossip target) did not send them a chat message.
Next, participants engaged in the common good dilemma. After making a decision about how many chips to contribute to the group account (and thereby how many to keep for themselves), they filled out several questionnaires: group opinion concerns questionnaire (cf. Beersma & Van Kleef, 2011a), sucker role test (Beersma & Van Kleef, 2011b), the social responsibility norm, (adapted from Berkowitz & Lutterman, 1968; De Cremer & Van Lange, 2001), the personal norm of reciprocity (adapted from Perugini et al., 2003), the altruism facet of the NEO Personality Inventory – Revised (NEO PI – R; cf. Feinberg et al., 2012) as a proxy to measure participants’ social value orientation, the Multidimensional Work Motivation Scale (MWMS; adapted from Gagné et al., 2012) to assess participants’ work motivation, and a questionnaire containing manipulation checks (i.e. identifiability, gossip valence). For explorative reasons, the experiment also contained questionnaires to assess measures of motivation to be accepted by the group and group norms in the game. Furthermore, participants filled out the Motives to Gossip Questionnaire (adapted from
Beersma & Van Kleef, 2012; Grosser, Lopez-Kidwell, LaBianca, & Ellwardt, 2012) and some questions that assessed their thoughts on the contributions (i.e. amount of chips) of the other group members to the group account. These explorative measures and corresponding results are further discussed in the appendix. The final questions assessed participants’ thoughts on the purpose of the study and whether they had any suggestions. Thereafter, participants were informed that the interactive part of the experiment would not actually take place. They were thanked and asked to report to the experimenter to collect their credit points or money and, if they wanted, leave they email address to receive a full debriefing and still have a chance to win the 75 euro price.
Measures
To measure contributions to group goals, I recorded the amount of chips participants
contributed to the group project, which could range from 0 to 100 (cf. Beersma & Van Kleef, 2011a; Mulder et al., 2006).
To access participants’ social value orientation, they completed a Decomposed Games Measure, an established task to measure this construct (see e.g. Messinck & McClintock, 1968; Van Lange, 1999; Van Lange & Kuhlman, 1994). According to De Cremer and Van Lange (2001), this measure is “internally consistent, reliable over substantial time periods, not related to measures of social desirability or indices of mood, and there is evidence for its ecological validity in various domains” (p. 10). In accordance with earlier research, the reliability was excellent (Cronbach’s α = .98).The task consists of nine-items, each
containing three alternative outcome distributions with valuable points for themselves and an anonymous (fictional) interaction partner. An example is the choice between alternative (A) 500 points for themselves and 500 points for the other (cooperative choose), (B) 560 points for themselves and 300 for the other (individualistic choice; i.e. maximum amount of points for oneself regardless of the other), or (C) 500 for themselves and 100 for the other
(competitive choice; i.e. maximizing the difference between the outcomes). Both B and C are proself choices. Participants are classified as prosocial, individualistic or competitive when at least six choices are consistent with one of the orientations. Out of the total number of 194 participants, 66 were identified as prosocials, 104 as individuals and 6 as competitors. On the basis of the above criterion, 18 participants could not be classified and where therefore
excluded from further analyses, leaving 176 participants (64 males and 112 females, mean age of 22.18 years). Participants classified as individualists and competitors were combined to form one group of proselfs (also see e.g. De Cremer & Van Lange, 2001).4
I measured group opinion concerns with four items (adapted from Beersma & Van Kleef, 2011a) on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly
agree) (e.g., “During the decision-making task, I thought about how my group members would think about me.”; “During the decision-making task, I was worried about my
reputation.”). In accordance with Beersma and Van Kleef, the reliability of the scale was high (α = .81).
Sucker role was measured with 11 Likert-type items (f.e. “I think that my group members respect me.”) ranging from 1 (strongly disagree) to 5 (strongly agree), to assess how much people feel like a sucker in their group. Initial analysis showed that the alpha reliability of the sucker role scale was high (α = .80). However, I refrained from using two of the items due to corrected item-total correlations below .30 (i.e. “I think that my group members find me naïve.”, corrected item-total correlation of .02; “I think that my group members find me stupid.”, corrected item-total correlation of .20). This resulted in an even higher alpha reliability of .85.
I measured the internalized norm of social responsibility with a five-item scale adapted from De Cremer & Van Lange (2001; Berkowitz & Lutterman, 1968), asking the participants to rate how responsible they feel for the group’s interest on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree) (e.g., “Every group member should contribute to the common good.”; “It is the duty of every group member to contribute to the group the very best he/she can.”). I removed three items from the original questionnaire, because these items were not suitable to adapt to the working in groups domain (i.e. “Our country would be a lot better off if we didn’t have so many elections and people didn’t have to vote so often.”, “People would be a lot better off if they could live far away from other people and never have to do anything for them.”, “At school I usually volunteered for special projects.”). Initial analysis showed that the alpha reliability of the social responsibility scale was very low (α = .46). After sequentially removing three additional items, due to low
corrected item-total correlations, the reliability of the scale was an acceptable .73 (“I feel very bad when I breach a promise.”, corrected item-total correlation of .07; “It makes no sense to
worry about the acts of other group members; you cannot change it anyways.”, corrected item-total correlation of .21; “It is o.k. to disappoint other group members, because you cannot always please everyone.”, corrected item-total correlation of .15). In search of a possible explanation for the initial low reliability of the respective scale, I also performed a factor analysis. This showed 2 extracted principal components with an Eigenvalue greater than 1, which combined explained 57.49% of the variance. After deleting one item (“I feel very bad when I breach a promise.”) due to crossloadings above .4, the 2 components explained 69.33% of the variance. However, further analysis showed that one of the
components deemed unreliable (α = .32); both items had a corrected item-total correlation of .19. Therefore, I was forced to refrain from using this construct in further analyses. The remaining social responsibility construct explained 78.89% of the variance.
The internalized norm of reciprocity was measured using the Personal Norm of Reciprocity (PNR) questionnaire adapted from Perugini et al. (2003)5, which measures three aspects of reciprocity through 18 items with a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree): Reciprocity beliefs (6 items; “I do not behave badly with other group members so as to avoid them behaving badly to me.”), Positive reciprocity (5 items; “If a group member does me a favor, I am ready to return it.”), Negative reciprocity (7 items; “If a group member puts me in a difficult position, I will do the same to him/her.”). I removed the items from the original questionnaire that were not suitable to adapt to the working in groups domain: 3 reciprocity belief items (“When I pay someone compliments, I expect that s/he in turn will reciprocate.”, “If I help tourists, I expect that they will thank me nicely.”, “If I don’t leave a good tip in a restaurant, I expect that in the future I will not get good service.”), 4 positive reciprocity items (“I am ready to do a boring job to return someone’s previous help.”, “If someone asks me politely for information, I’m really happy to help him/her.”, “If someone lends me money as a favor, I feel I should give him/her back something more than what is strictly due.”, “If somebody suggests to me the name of the winning horse at the race, I would
certainly give him/her part of my winnings.”) and 2 negative reciprocity items (“If somebody offends me, I will offend him/her back.”, “If somebody is impolite to me, I become
impolite.”). Initial analysis showed that the alpha reliability of the reciprocity belief scale was .64. After deleting one item (i.e. “It is obvious that if I treat someone badly s/he will look for revenge.”), due to a low corrected item-total correlation of .18, the total scale resulted in a slightly higher alpha reliability of .67. Initial analysis of the positive reciprocity scale showed that the alpha reliability was .63. After sequentially removing two items, due to relative low corrected item-total correlations, the reliability of the positive reciprocity scale was an
acceptable .71 (i.e. “When someone does a favor for me, I feel committed to repay him/her.”, corrected item-total correlation of .29; “I am ready to undergo personal costs to help
somebody who helped me before, corrected item-total correlation of .32). The alpha reliability of the negative reciprocity scale was good (Cronbach’s α = .74).
The orientation of work motivation to contribute to the group was measured using the Multidimensional Work Motivation Scale (MWMS; adapted from Gagné, et al., 2012). The different motivational constructs as described in the SDT (see e.g. Gagné & Deci, 2005) were measured through 31 items, asking the participants to rate the reasons for contribution to their group on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree): External regulation (nine items, e.g. “I contribute to my group, because I feel the pressure of my fellow group members to contribute.”), Introjection (six items, e.g. “I contribute to my group, because otherwise I feel ashamed.”), Identification (four items, e.g. “I contribute to my group, because I find that important.”), Integration (two items, e.g. “I contribute to my group, because it is my duty to contribute to the group.”), Intrinsic motivation (five items, e.g. “I contribute to my group, because the task itself is fun.”)6, Amotivation (five items, e.g. “My contribution to the group is pointless.”). I had to remove three items from the original
MWMS, because these items were not suitable to adapt to the working in groups domain (i.e. one external regulation item, “I work, because I want to avoid poverty.”; one introjection
item, “I work, because I have to prove to myself that I can.”; one integration item, “I work, because I was made for my kind of job.”). Initial analysis showed that the alpha reliability of the respective scales, except for Integration (α = .72), were high (external regulation, α = .82; introjection α = .88; Identification .87; Intrinsic motivation = .93; Amotivation = .85).
However, one external regulation item (“I contribute to my group, because it earns me chips (and therefore money.”) had a very low corrected item-total correlation (-.04), giving me reason to delete this item from the respective scale. This resulted in an even higher alpha reliability of .88.To verify the different motivational constructs of the MWMS, I performed a factor analysis. After removing the cross-loadings above the appropriate cut-off point of .4, which represent substantive values (see Field, 2005; Stevens, 1992), there were 5 principal components extracted with an Eigenvalue greater than 1: External regulation (α = .89), Introjection (α = .91), Autonomous motivation (i.e. Identification and Integration items; α = .88), Intrinsic motivation (α = .93) and Amotivation (α = .85).7
These components combined explained 71.78% of the common variance. The items I deleted from the respective scale due to cross-loadings above .4 were two External regulation items (“I contribute to my group, because its offers me security”; “I contribute to my group, because I want to avoid criticism from my group members”), three Introjection items (“I contribute to my group, because it is my duty.”; “I contribute to my group, because it makes me feel good.”; “I contribute to my group, because I can be proud of myself when I contribute.”) and two Autonomous
motivation items (“I contribute to my group, because it is my duty to contribute.”; “I contribute to my group, because it fits well with my goals). In accordance with earlier research, the External regulation and the Introjection scales were combined to form one Controlled motivation scale (α = .91).
I also included three Likert-type items to check the adequacy of the identifiability manipulation (e.g., “My group members could see how I distributed the chips between myself and the group.”). The reliability of this manipulation check was very high (α = .97). Finally, I
included three items to check the manipulation of the gossip valence (e.g., “My group member(s) shared negative information about another group member.”). After removing one item (“My group member(s) shared information about another group member.”) due to a corrected item-total correlation of .12, the reliability of this manipulation check was also very high (α = .93). In hindsight, this item was indeed unfit to check if the gossip valence
manipulation was successful, because it does not discriminate between blame and praise gossip.
Results
Descriptive Statistics and Treatment of the Data
Table 1 shows the correlations between variables. As can be seen, the manipulation checks correlated with the manipulations in the expected ways. The independent variable gossip valence positively correlated with the internalized norm of social responsibility (r = .18, p = .02), and the dependent variable contributions to the group (r = .18, p = .02). This indicates that when participants received blame gossip, their feelings of social responsibility as well as their contributions to the group increased. Contributions to the group also positively
correlated with identifiability (r = .17, p = .03), social responsibility (r = .39, p = .00), positive reciprocity (r = .18, p = .02), and autonomous motivation (r = .31, p = .00), and negatively correlated with sucker role (r = -.44, p = .00), and amotivation (r = -.47, p = .00). Thus, participants’ contributions also increased, when their contributions were identifiable for their fellow group members, when their feelings of social responsibility increased, when their positive reciprocity tendencies increased, and when their autonomous motivation increased. In contrast, when participants felt more like a sucker and their amotivation increased, their contributions to the group decreased. Furthermore, identifiability positively correlated with group opinion concerns (r = .25, p = .00) and positive reciprocity (r = .18, p = .02), indicating
that when participants’ contributions were identifiable, their group opinion concerns as well as their positive reciprocity tendencies increased. Also, social value orientation positively correlated with social responsibility (r = .37, p = .00), indicating that when participants had a prosocial orientation, they also had increased feelings of social responsibility.
I analyzed the data with 2 (gossip: praise vs. blame) x 2 (identifiability: absent vs. present) x 2 (social value orientation: proself vs. prosocial) analyses of variance (ANOVAs). To be able to compare the results corresponding with directional hypotheses with exploratory research findings, I used two-tailed tests for all analyses. Significant interaction effects were decomposed using simple-effects analysis (Winer, 1981). I also ran simple effects analyses to examine the pattern of means in the various conditions when ANOVA revealed unexpected results. Finally, I tested for conditional indirect effects (or mediated moderation effects; see Baron & Kenny, 1986) using conditional process analyses (PROCESS) developed by Hayes (2013). The term conditional indirect effect reflects the fact that the causal effect of an independent variable (X) on a dependent variable (Y) is transmitted by one or more mediators (M), and that this effect depends on the value of one or more moderators (W) (see Preacher, Rucker, & Hayes, 2007). In my study it refers to the hypothesized effects of gossip on group members’ contributions to the group that is transmitted by several mediators: group opinion concerns, social responsibility, reciprocity, sucker role and orientation of work motivation. These effects potentially depend on group members’ social value orientation and the identifiability of their contributions. For example, I expect that group opinion concerns mediate the effect of blame gossip on contributions to the group for proself group members, especially when their contributions are identifiable, but not for prosocial group members. Testing for these conditional indirect effects “involves first showing an interaction effect of X and W on Y, then introducing a mediator of that effect […] to clarify the nature of key
Manipulation Checks
ANOVA on the manipulation check for gossip valence revealed that participants in the blame gossip conditions indicated their received gossip comment about another group member to be more negative (M = 4.64, SD .72) than participants in the praise gossip conditions (M = 1.43, SD = .78), F (1, 168) = 728.11, p = .00, η² = .81. No other effects were significant. ANOVA on the manipulation check for identifiability revealed that participants in the identifiability conditions indicated their decisions to be more identifiable (M = 3.99, SD 1.23) than
participants in the no identifiability conditions (M = 1.42, SD = .90), F (1, 168) = 239.15, p = .00, η² = .59. Thus, both manipulations were successful.However, the ANOVA also revealed that participants in the blame gossip conditions indicated their decisions to be more
identifiable (M = 2.91, SD 1.65) than participants in the praise gossip conditions (M = 2.50, SD = 1.69), F (1, 168) = 4.39, p = .04, η² = .03. Therefore, I also ran simple-effects analyses to examine the pattern of means in the various conditions. Interestingly, these analyses revealed that whereas in prosocial group, gossip valence did not significantly affect the manipulation check of identifiability, F (1, 169) = 1.24, ns; Mblame gossip = 2.81, SD = 1.70, Mpraise gossip = M = 2.28, SD = 1.72, in the proself group, proself participants indicated their decisions to be more identifiable in response to blame gossip (M = 2.96, SD = 1.64), than in response to praise gossip (M = 2.66, SD = 1.66), F (1, 169) = 3.84, p < .05. This supports the idea that proselfs are more susceptible for the controlling nature of blame gossip than
prosocials.
Contributions to the Group
ANOVA revealed a significant Gossip valence X Social value orientation interaction on contributions to the group, F (1, 168) = 8.04, p = .00, η² = .05. This interaction effect was decomposed using simple effects analyses. These tests demonstrated that in the praise gossip conditions, prosocial participants contributed more to the group (M = 58.44, SD = 31.74),
than proself participants (M = 44.90, SD = 35.53), F (1, 169) = 6.25, p = .01. This supports Hypothesis 1d, in which I predicted that praise gossip increases contributions of prosocials, relative to proselfs. Prosocial participants in the praise gossip conditions also seemed to contribute more to the group (M = 58.44, SD = 31.74), than in the blame gossip conditions (M = 53.27, SD = 27.26). However, this effect was not significant and therefore does not support Hypothesis 1e, F (1, 169) < 1, ns. In contrast, proself participants contributed more to the group in response to blame gossip (M = 65.49, SD = 25.72), relative to prosocial participants (M = 53.27, SD = 27.26), F (1, 169) = 4.75, p = .03, and relative to praise gossip (M = 44.90, SD = 35.53), F (1, 169) = 13.66, p = .00. Thus, praise gossip increases prosocials’
contributions to the group, relative to proselfs, and blame gossip increases proselfs’
contributions to the group, relative to prosocials and relative to praise gossip (see figure 2 and 3).
Although the expected three-way interaction between Gossip valence, Identifiability and Social value orientation was not significant, F (1, 168) = 1.63, ns, I ran simple effects analyses to examine the pattern of means in the various conditions. Interestingly, these analyses showed that in the identifiability conditions, the interaction between Gossip valence and Social value orientation significantly affected contributions to the group, F (1, 169) = 3.88, p = .05. This effect was decomposed using simple simple effects analyses. These tests demonstrated that blame gossip increased contributions to the group of proself participants when their contributions were identifiable for their group members (M = 73.06, SD = 25.49), relative to prosocial participants (M = 54.64, SD = 29.51), F (1, 171) = 7.43, p = .01, relative to praise gossip (M = 52.07, SD = 36.14), F (1, 171) = 7.47, p = .01, and relative to
unidentifiable contributions (M = 57.11, SD = 23.68), F (1, 171) = 4.40, p = .04. This
supports Hypotheses 1a - c, in which I predicted that blame gossip increases the contributions to the group of proselfs when their contributions are identifiable for their group members,
relative to prosocials (1a), relative to praise gossip (1b) and relative to unidentifiable contributions (1c).
Furthermore, the additional analyses showed that in the no identifiability conditions, Gossip valence X Social value orientation had a marginally significant affect on contributions to the group, F (1, 169) = 3.35, p = .07. Simple simple effects tests demonstrated that praise gossip lowered contributions to the group of proself participants when their contributions were unidentifiable for their group members (M = 35.45, SD = 33.16), relative to prosocial participants (M = 57.00, SD = 32.55), F (1, 171) = 5.55, p = .02; relative to blame gossip (M = 57.11, SD = 23.68), F (1, 171) = 5.09, p = .03; but not relative to identifiable contributions (M = 52.07, SD = 36.14), F (1, 171) = 2.36, ns. This supports Hypothesis 1f and g, in which I predicted that praise gossip lowers the contributions to the group of proselfs when their contributions are unidentifiable for their group members, relative to prosocials (1f), and relative to blame gossip (1g).
ANOVA also revealed a significant effect of identifiability on contributions to the group, F (1, 168) = 4.21, p = .04, η² = .02, indicating that participants contributed more to the group when their contributions were identifiable (M = 61.25, SD = 31.74), than when their contributions were unidentifiable (M = 50.75, SD = 29.907). In search of a possible
explanation, I ran additional simple effects analyses to examine the pattern of means in the various conditions. These simple effects tests demonstrated that whereas in the prosocial group, identifiability did not significantly affect contributions to the group, F (1, 169) < 1, ns; M identifiability = 57.68, SD = 30.11, M no identifiability = 54.92, SD = 29.70, in the proself group, proself participants contributed more to the group when their contributions were identifiable for the other group members (M = 62.92, SD = 32.59) than when their
contributions were unidentifiable (M = 47.58, SD = 29.97), F (1, 169) = 7.93, p = .01. The significant effects of Gossip X Identifiability and Gossip X Social value orientation confirm the presumption that proselfs are more sensitive for identifiability of their contributions and
provide some support for Hypothesis 1h, in which I predicted that, relative to identifiable contributions, proselfs contribute less to the group in response to praise gossip, when contributions are unidentifiable (see figure 2).
Group Opinion Concerns
ANOVA revealed a significant effect of identifiability on group opinion concerns, F (1, 168) = 8.38, p = .00, η² = .05, indicating that participants experienced more group opinion concerns when their contributions were identifiable (M = 2.84, SD = .97), than when their contributions were unidentifiable (M = 2.36, SD = .95). This is in line with Hypotheses 2a and b, in which I predicted that participants experience more group opinion concerns when their contributions are identifiable for their group members, than when their contributions are unidentifiable. However, I specifically expected that proselfs would experience more group opinion concern when their contributions were identifiable, especially in response to blame gossip. Therefore, I ran simple-effects analyses to examine the pattern of means in the various conditions. Interestingly, simple-effects tests demonstrated that whereas in the praise gossip conditions, identifiability did not significantly affect group opinion concerns, F (1, 169) = 1.47, ns, M
identifiability = 2.68, SD = .89, M no identifiability= 2.41, SD = .90, in the blame gossip conditions, participants experienced more group opinion concerns (M = 2.99, SD = 1.02), than when their contributions were unidentifiable (M = 2.31, SD = 1.00), F (1, 169) = 8.40, p = .00. Furthermore, whereas in the prosocial group, identifiability did not significantly affect group opinion concerns, F (1, 169) = 1.88, ns; M identifiability = 2.65, SD = .96, M no
identifiability = 2.37, SD = .86, in the proself group, proself participants experienced more group opinion concerns when their contributions were identifiable (M = 2.93, SD = .97), than when their contributions were unidentifiable for their group members (M = 2.35, SD = 1.01), F (1, 169) = 10.03, p = .00. Within the identifiability conditions, proself participants also experienced more group opinion concerns (M = 2.93, SD = .97), than prosocials (M = 2.65,
