Rival or comrade? A systematic review and conceptual framework of when and why the powerful act prosocially or antisocially towards each other

We present a systematic review of the literature on power and its interpersonal consequences. Our review, comprising 339 studies published in 145 research articles, shows that this line of research has primarily examined how powerholders attend to and act towards powerless individuals, or others in general. We therefore know surprisingly little about how powerholders attend to and act towards other powerhold-ers. To address this issue, we present a conceptual framework that outlines how an actor’s power interacts with a target’s power to influence prosocial and antisocial beliefs, attitudes, and behaviors. We identify two routes in the literature detailing how powerholders respond to one another. First, building on rivalry literature, we present a competitive route suggesting that powerholders rival each other and engage in conflict. Second, building on social identity and social dominance literature, we present a harmonious route suggesting that powerful peers will show compassion and care for each other. Finally, we bring forth suggestions for how future research could test these two perspectives, by presenting moderators that determine when each of these two routes is activated. In doing so, we offer important implications for the power literature and open a new line of inquiry for future research.

The results of our review demonstrate that research so far has largely neglected the broader hierarchical context, where social perceptions and behaviors can also occur in lateral relationships (i.e., among powerful peers; for exceptions see van Dijke & Poppe, 2003;Hildreth & Anderson, 2016).Consequently, we know little about how high-power individuals attend to, and act towards, other people with power.
The present article therefore develops a conceptual framework detailing how an actor's power interacts with a target's power to influence prosocial (i.e., being compassionate, caring, and cooperative; George, 1990) and antisocial beliefs, attitudes, and behaviors (i.e., being unfriendly, rude, and uncooperative; Robinson & O'Leary-Kelly, 1998).We identify two routes in the literature.First, adopting an individual perspective (Kilduff et al., 2010), we present a competitive route, suggesting that powerholders will rival each other and thus act antisocially toward each other.Second, adopting a group perspective (Hornsey, 2008;Pratto & Sidanius., 1999;Tajfel & Turner, 1979, 1986), we present a harmonious route, suggesting that powerful peers will show compassion and care for each other and thus act prosocially towards one another.Finally, we present suggestions for how future research could test these two perspectives, by identifying moderators that determine when each of these two routes is more salient, and hence, when powerful actors will behave antisocially versus prosocially towards other high-power targets (See Figure 1).
By doing so, the present research offers important contributions to the power literature.First, our framework shows the importance of examining the interpersonal consequences of power in a dyadic context.To date, many studies have examined interpersonal consequences of power without specifying the characteristics of the target of the attitudes and/or behaviors.Hence, it is hard to draw decisive conclusions on how power influences interpersonal outcomes based on this work.
Second, our research contributes to a growing body of literature that paints a more nuanced picture of power and its social consequences.Although power was initially often portrayed as a corrupting force, leaving its beholders inattentive to the needs of the people around them (e.g., Fiske, 1993;Galinsky et al., 2006;Stamkou et al., 2016), a growing body of research shows that the powerful can also be attentive to other people's needs (e.g., Galinsky et al., 2003;Schmid Mast et al., 2009).Our proposed framework supports this notion, as we argue that in certain situations powerholders express empathy and compassion for individuals that occupy similar high-power positions.
Finally, our research has important societal implications.While there is great value in understanding how high-power individuals attend to low-power individuals, it is also important to understand when and why powerholders do and do not get along with each other.Such peer assessments among the powerful may have important implications in organizational, institutional, and political contexts where powerholders interact and work together to make decisions that affect millions of lives.

| Method and results
To examine whether past studies have investigated the high-power actor/high-power target dyad, we conducted an extensive systematic review of the literature on the interpersonal consequences of power.We followed the guidelines of the Preferred Reporting Items for Systematic Reviews (PRISMA, see Appendix 1).The present review was not registered.

| Search strategy
We used three search approaches to identify relevant empirical studies.First, we searched major academic databases (PsycINFO, PsycArticles and Business Source Premier) for empirical studies published in a list of 20 journals in organizational behavior, psychology and marketing (see Appendix 2).We focused on these 20 journals to ensure that we included all mainstream work on the interpersonal consequences of power.We used the search terms "power" and "hierarchy" to search for relevant literature.These items were combined with Boolean F I G U R E 1 Conceptual framework operators according to the rules of each database.Second, to complement the database searches, we reviewed the reference sections of a number of theoretical articles on power (namely Galinsky et al., 2015;Guinote, 2017;Schaerer et al., 2018;Sturm & Antonakis, 2015).Third, we used Google Scholar to obtain studies that cited articles with the main operationalizations of power in field research (namely Anderson et al., 2012;Fast & Chen, 2009;Lammers et al., 2010).

| Selection criteria
We applied two criteria to select the papers for our review.First, the paper required one or more experimental or field studies that used power as an independent variable.Second, the study should focus on one or more interpersonal outcomes (i.e., behaviors or attitudes directed at others) as a dependent variable.For the papers with field studies, we included studies that complemented experimental studies in the selected articles, or that used one of the main operationalizations of power in the literature: the sense of power (Anderson et al., 2012), or the hierarchical power measures developed by Fast and Chen (2009) or Lammers and colleagues (2010).We excluded work that focused on power differences at the group-level of analysis (e.g., Ronay et al, 2012), because in these studies it is unclear which specific group members are targeted in the interactions.The literature review included all published work prior to January 15, 2021.Our keyword search in the databases returned 1387 results.
We screened the titles, abstracts and method sections for inclusion.
Two papers were included based on the search of reference sections and two papers were included based on the Google scholar citations.The final review included 339 studies collected from 145 papers (See Table 1).

| Coding procedure
For all included studies, we collected the publication year, journal, power variable (manipulated or measured), and the type of manipulation (power prime, role assignment or other) or measure (hierarchical power, sense of power or other).We then coded whether the study manipulated or measured the power level of the actor, the power level of the target, and the direction of the power dyad (high-low, low-high, low-low, high-high).Two co-authors independently coded 10 randomly selected studies.The coding was then discussed, and any discrepancies and problems were resolved.
One of the co-authors then proceeded to code all the remaining studies.The review protocol can be accessed by contacting the author.

| Results
We categorized all studies based on whether they experimentally manipulated power or measured actual power in the field.We found that 279 studies manipulated power (82.3%; e.g., with a power prime or role assignment), 56 studies measured actual power (16.5%; e.g., with hierarchical power or sense of power measure), and 4 studies included both a manipulation and a measure of power beyond the manipulation check (1.2%).We then coded whether the power level of the actor, target, or both were specified in these studies.
Of the 279 experimental studies, 163 studies (58.4%)only manipulated the power level of the actor (and not that of the target) and 13 studies (4.7%)only manipulated the power level of the target (but not of the actor).Furthermore, 103 studies (36.9%) specified both the power level of the actor and the target.The vast majority of these studies included high-power actor/low-power target relationships (n = 98) and/or low-power actor/high-power target relationships (n = 76).Three studies included low-power actor/ low-power target relationships (Hildreth & Anderson, 2016;Studies 1a and 3;Van Dijke & Poppe, 2003; Study 1) and three studies included high-power actor/high-power target relationships (Hildreth & Anderson, 2016;Studies 1a, and 3;Van Dijke & Poppe, 2003;

Study 1).
Of the 56 studies that measured actual power of the actor, only 17 studies (30.4%) also specified the power level of the target to which the interpersonal outcome was directed.These included mostly high-power actor/low-power target relationships (n = 14), and/or low-power actor/high-power target relationships (n = 9).
Finally, one study included low-power individuals (i.e., subordinates) who rated a peer at the same hierarchical level (Van Prooijen et al., 2014; Study 2) and one study compared interactions between TA B L E 1 Frequencies and percentages of studies (columns 1 and 2) and frequencies of type of dyads within studies that specified both the power level of the actor and target (column 3) included in the systematic review Note: Several studies include multiple dyads.Therefore, the sum of the dyads exceeds the number of studies that specified both the power level of the actor and target.
Lastly, the four studies that included both a manipulation and measure of power (beyond the manipulation check), only specified the power level of the actor.

| Conclusion
More than half of the studies that empirically examined the interpersonal consequences of power only specified (i.e., measured or manipulated) the power level of the actor, and not of the target.The research participants in these studies were asked, for instance, to write an E on their forehead (i.e., measure of perspective taking; Galinsky et al., 2006; Study 1), identify the degree to which specific emotions were expressed by a random person in a photograph (Côté et al., 2011; Study 1) or rate the degree of conflict they experienced with their colleagues in general (Anicich et al., 2016;Study 4).Of the studies that did specify the power levels of both the actor and the target in their dyadic power relationship, the vast majority exclusively focused on unequal power relations, that is, high-power actor/low-power target and/or low-power actor/high-power target relationships.
Moreover, only one of the four studies that examined high-power actor/high-power target relationships included a full crossover design that allowed for a direct comparison between the high-power actor/high-power target cell on the one hand, and the high-power Studies 1a, 2, and 3).
Overall, it seems fair to conclude that although power research has generated important insights into how powerful individuals attend to and act towards others in general, and low-power individuals in particular, we know little to nothing about how highpower individuals attend to and act towards powerful peers.In order to address this gap in the research, we present a conceptual framework that outlines how an actor's power interacts with a target's power to influence prosocial and antisocial beliefs, attitudes, and behaviors.

| CON CEP TUAL FR AME WORK
In the remainder of this paper, we conceptualize how powerholders may attend to and behave towards other powerholders.First, we draw on research at the individual level (Kilduff et al., 2010) to present a competitive route, which suggests that powerful peers will rival each other, and hence, behave antisocially towards each other.Second, we draw on research at the group level (Hornsey, 2008;Pratto & Sidanius, 1999;Tajfel & Turner, 1979, 1986), to present a harmonious route, which suggests that powerholders care for each other, and will act prosocially towards one another.Finally, we present avenues for future research to test these two perspectives, by presenting moderators that regulate the salience of these routes and thus determine when powerful peers are likely to get along or not.

| Competitive route
At the individual level, research on rivalry and competition portrays a pessimistic picture of how individuals in general are likely to see each other.Rivalry is defined as a "subjective competitive relationship that an actor has with another actor that entails increased psychological involvement and perceived stakes of competition for the focal actor, independent of the objective characteristics of the situation" (Kilduff et al., 2010, p. 945).Rivalry leads to a wide range of anti-social beliefs, attitudes, and behaviors, such as judging rivals as less warm (Russell & Fiske, 2008), sabotaging their success (Huang et al., 2019) and acting aggressively in order to win (Tjosvold et al., 2003; for a review of the antisocial consequences of rivalry and competition; see To et al., 2020).
Theory on rivalry and competition proposes that similarity is an important predictor of rivalry, such that individuals are more likely to rival other individuals who are similar to them (Kilduff et al., 2010).Rivals who are similar, for instance in the amount of power they possess, are more likely to strive for the same resources, and are therefore more likely to challenge one another.
This potential of the probable rival is an important predictor of rivalry, such that individuals are particularly likely to compete with others if they feel that they are evenly matched (Chen et al., 2007;Kilduff et al., 2010).
Rivalry research thus suggests that powerholders are more likely to compete with other powerholders as opposed to powerless individuals, because other powerholders strive for the same resources and are stronger competitors.Indeed, it is the politician with similar access to power who is most likely to steal another powerful politician's votes and it is the CEO of a company who is awarded the title "CEO of the year" at the expense of other CEOs.Politicians and employees lower in rank are no "match" and are therefore not considered competitors or rivals to these highpower individuals.
Empirical work supports this theorizing.This research shows that high-powered individuals in particular are inclined to end up in conflict and power struggles when working together (Greer et al., 2011(Greer et al., , 2018;;Greer & Chu, 2020).Hildreth and Anderson (2016), for instance, showed that high-power individuals (compared to individuals with medium or low levels of power) fought more over status and shared less information with each other when working together.Overall, this research suggests that powerholders are more likely to compete with other powerholders (compared to the powerless), and therefore tend to act antisocially towards high-power individuals.

| Harmonious route
At the group-level, on the other hand, the social identity approach (i.e., comprising both social identity and self-categorization theory; Hornsey, 2008;Tajfel & Turner, 1979, 1986) suggests that powerholders get along and cooperate with other powerholders.
According to this approach, people tend to categorize themselves and others into different social groups, such as old/young, male/ female, and also powerful/powerless.This categorization of people has important implications for how individuals see and treat the people around them.Specifically, in order to protect and enhance selfesteem, people tend to favor their "own groups" and the individuals that belong to these groups over "other groups" and the individuals that belong to them (Abrams & Hogg, 1988).For example, people evaluate members of their own group more positively and allocate more resources to members of their own group (Brewer, 1979;LeVine & Campbell, 1972;Mullen et al., 1992;Perdue et al., 1990;Tajfel & Turner, 1979).
A social identity perspective further suggests that these identification processes are stronger for the powerful than for the powerless (Sachdev & Bourhis, 1985).Comparison between the powerful and powerless results in favorable outcomes for the powerful, but not the powerless.It are the powerful who have access to important resources and who are granted more status (Magee & Galinsky, 2008).Therefore, being in a position of power is an important part of how people see themselves (Ashforth & Mael, 1989; DeRue & Ashford, 2010).High-power individuals thus more strongly identity with their powerful peers than the low-power individuals identify with the powerless (Joshi & Fast, 2013).Considering these identification processes among the powerful, it seems particularly likely that powerholders perceive their group and its members (i.e., other powerholders) in a positive manner.
Social dominance theory (Pratto et al., 2006;Pratto & Sidanius, 1999) further argues that powerholders are motivated to legitimize power differences between the powerful and powerless.This motivation stems from the desire to see the world as a just place where people get what they deserve and deserve what they get (Jost & Banaji, 1994;Lerner, 1980).Moreover, social dominance theory argues that the powerful are especially motivated to justify existing social hierarchies because such justifications protect their privileged positions (Jost & Burgess, 2000;Pratto & Sidanius, 1999).
The powerful justify social hierarchies through the endorsement of beliefs that enhance and legitimize existing hierarchies and that consist of the notion that the relatively powerful and powerless deserve their positions in the social hierarchy and the (lack of) resources and privileges that accompany these positions.Accordingly, high-power individuals tend to "enhance" themselves, and feel more entitled to have access to valued resources (De Cremer, 2003;De Cremer & Van Dijk, 2005) and to deviate from ethical and social norms (Galinsky et al., 2003;Keltner et al., 2010;Lammers et al., 2010).A logical extension of these ideas is that powerholders "enhance" other powerholders as well, as they too represent the existing hierarchical structure (Magee & Galinsky, 2008).Justifying other powerholders' positions will therefore help maintain and protect the hierarchical system in a similar manner.
Although social identity and social dominance theory theorize about group-level phenomena, they are informative for generating predictions about interactions between powerful individuals as well (Georgesen & Harris, 2006).Indeed, these theories suggest that since powerful peers belong to the same social group or category, they will identify with each other and will be motivated to legitimize and justify each other's standing in the hierarchical system.
Therefore, these theories suggest that powerful individuals will act antisocially towards the powerless, but prosocially towards their powerful peers.
Overall, the two perspectives described above lead to different predictions with respect to powerholders' antisocial and prosocial responses to each other.The competitive route describes how individuals with power are likely to rival and compete with each other.
The harmonious route, on the other hand, suggests that powerholders get along well because they belong to the same group of "the powerful".The two routes are grounded in different theoretical approaches, and have a different focus with regards to the level of comparison-individual or group.The degree to which those in power perceive themselves as individual powerholders versus part of the powerful group will thus determine the route that powerholders follow (Simon & Oakes, 2006;Turner, 2005).In the remainder of this paper we will explore the conditions under which powerholders are likely to consider themselves "individual powerholders" versus part of "the powerful group", and thus will or will not get along with other powerholders.

| IDENTIF YING MODER ATOR S
Whether people construe themselves as individuals or as part of a certain group is determined by their personal readiness to belong to the group, their fit with the group, and contextual factors (Oakes, 1987;Oakes et al., 1991).Below we address all three of these to predict when powerholders are likely to follow either the competitive or the harmonious route, and consequently act prosocially or antisocially towards each other.

| Personal readiness
First, research suggests that an individual's personal readiness is an important factor in determining whether he or she sees him or herself as an individual or as part of a social group or category.Personal readiness refers to how easily accessible a certain category or social identity is to someone (Turner et al., 1987(Turner et al., , 1994)).In this regard, people who have often adopted a certain category have this category readily accessible, and hence, are more likely to adopt it again (Hogg & Terry, 2000).This implies that the longer an individual has occupied a position of power, the more likely he or she will be to consider him or herself part of the powerful group or category.In support of this reasoning, research has shown that power is addictive and that individuals tend to value positions of power more over time (Fehr et al., 2012;Kets de Vries, 1991).Overall, this suggests that new powerholders are likely to act antisocially towards other powerholders and that the longer individuals occupy a position of power, the more prosocially they will behave towards other powerholders.
The accessibility of a certain identity is, however, not only related to how often a person has adopted a social category before, but also applies to how important someone finds a certain social identity.
Individuals are more likely to adopt a social identity that they value (Hogg & Terry, 2000;Turner et al., 1994).An individual's social dominance orientation (SDO) is likely to play an important role in this regard.SDO refers to the extent to which an individual supports group-based hierarchies (Pratto et al., 1994;Pratto & Sidanius, 1999).
Moreover, individuals higher in SDO strongly value their own power and are motivated to protect it (Altemeyer, 1998;Duckitt, 2006;Feenstra et al., 2017).This research thus suggests that powerholders higher in SDO are especially likely to identify as powerholders and hence act prosocially towards other powerholders (Altemeyer, 1998;Joshi & Fast, 2013).On the other hand, powerholders lower in SDO, who find being part of the category "the powerful" not such an important part of how they see themselves, are more likely to identify with other groups, such as their work teams, which might also include individuals in low-power positions.Hence, these individuals are more likely to behave antisocially towards other powerholders.
Finally, a social category can also be made accessible by making it salient.This implies that powerholders will especially favor other powerholders when their high power is made accessible or salient in a particular context or situation.For example, when a high-power individual is at work and stands at the executive desk overlooking his or her workers, he or she will be more likely to feel powerful than at home when there are few situational cues signaling his or her high-power position (Carney et al., 2010;Garrison et al., 2016).In a high-power-salient context, powerful individuals are reminded of their power, and hence, we propose that powerholders are especially likely to identify with other powerholders and act prosocially towards them in such situations.In contexts that lack cues to their high power, on the other hand, powerholders will be more likely to identify with other groups, or perceive themselves as individual powerholders, and hence, will be more likely to compete with and behave antisocially towards other powerholders.

| Fit with the group
Second, research suggests that the degree to which individuals feel that they fit the prototype of the powerful determines whether powerholders see themselves as members of the powerful group or see themselves as powerful individuals (Hogg & Terry, 2000;Hogg & Turner, 1987).Prototypes refer to sets of attributes that define certain groups or categories.An individual matches the prototype of a group or category to the degree that they are representative of exemplary or ideal members of that group.
Although what is considered a prototypical powerholder may vary between contexts, research has shown that the prototype of a powerful or "good" leader is generally male/masculine (Koenig et al., 2011), white (Gündemir et al., 2014), heterosexual (Morton, 2017), and dominant (Powell et al., 2002).Individuals who have these attributes or see themselves as someone who possesses these qualities are thus more likely to perceive themselves as prototypes of the powerful and identify as powerholders.This suggests that a white heterosexual man, for instance, likely identifies with the powerful and acts prosocially towards other powerholders.Members of minority groups, such as women or gay men, however, fit the prototype of a powerholder less and are therefore less likely to identify as part of the powerful, which stimulates competition with antisocial behavior towards other (prototypical) powerholders.
In addition to looking like a prototypical member of a certain group, the degree to which people behave in line with a certain prototype determines their feelings of fitting to this prototype, and hence, the likelihood of identifying with the social category (Turner et al., 1994).Individuals who behave like a "typical powerholder" by, for instance, showing dominance, ambition, and/or planning and organizing activities, are more likely to identify as powerholders.Overall, this research suggests that powerholders who fit the powerful prototype more in terms of attributes, qualities, and/or behaviors, are more likely to identify with the powerful, and are therefore more likely to get along better with other powerholders compared to individuals who fit the powerholder prototype less.

| Contextual factors
Finally, social categorization theory proposes that contextual factors determine the salience of a certain identity.With regard to power identities, the degree to which power is under threat seems to be particularly important.Indeed, research suggest that powerholders' motivation to justify and protect existing power arrangements is amplified when the powerless threaten to usurp their power (Maner & Mead, 2010).In such threatening situations, powerholders feel a strong need to protect their power and, therefore, derogate (Georgesen & Harris, 2006) and demean the powerless (Fast et al., 2012) while prioritizing their own feelings and interests (Sidanius et al., 2013).We propose that powerholders who are threatened by the powerless may similarly prioritize and favor other powerholders who reflect the social hierarchical system, in order to protect that system.Therefore, while a threat coming from the powerless triggers antisocial behavior towards the powerless, we expect that powerholders will act particularly empathic and protective towards their powerful peers in such threatening situations.
In addition to the powerless threatening the position of the powerful, powerful peers can also threaten a powerholder's position (Greer, 2014;Greer & Chu, 2020).This may happen, for example, when one powerholder (e.g., director of finance) interferes with an important project of another powerholder (e.g., the implementation of important HR strategies that take up company resources).
Another situation in which this may occur is when different powerholders strive for similar rewards, such as awards (e.g., leader of the year).We argue that such threats from the powerful cause powerholders to see themselves as individuals with power or identify with, for instance, their work teams, who have the potential to boost a powerholder's position relative to other powerholders (Turner, 2005).Hence, we argue that threats from other powerholders will cause power struggles, and lead to rivalry and competition among powerful peers.

| DISCUSS ION
We presented a systematic review of the literature on the interpersonal consequences of power.Our review showed that this line of research primarily examined how powerholders perceive and act towards the powerless or others in general, while very few studies examined how powerholders interact with their powerful peers.
We therefore presented a conceptual framework in which we outlined two possible routes that determine how powerholders might or might not get along with other powerholders.Adopting an individual perspective, we contended that powerholders compete with each other for similar resources and therefore act antisocially towards each other.In addition, adopting a group perspective, we argued that powerholders identify with their powerful peers and are therefore likely to act prosocially towards each other.Finally, we put forward several moderating variables which determine whether powerholders see themselves as "individuals with power" or as part of the "powerful group", and hence whether they will compete or cooperate with other powerholders.

| Future research suggestions
Our framework provides an important new line of inquiry for future research: instead of focusing primarily on how powerholders attend to and act towards the powerless, we urge future research to examine how powerholders interact with each other.In this regard, the challenge for future research will be to empirically test the competing predictions that we put forward and test the different mediating and moderating processes of our framework.It will be important that these studies test the interpersonal consequences of power in the dyadic power context, and thus specify and/or systematically vary the power level of the actor as well as the power level of the target.
In experimental research designs, researchers could, for instance, follow the procedure of Hildreth and Anderson (2016) and assign individuals to roles with high or low power (or a control condition) and then, in a subsequent task, let individuals with similar roles interact and work together in order to see how they get along.Moreover, such a research design could further be extended by simultaneously including both high-power actor/high-power target dyads (or groups) and high-power actor/low-power target dyads (or groups), in order to directly assesses the differences in how powerholders look at and treat the powerless (and participants in control conditions) versus their powerful peers.Similarly, in survey research, it will be important for researchers to measure the (perceived) power of the target when examining the interpersonal consequences of power.
Finally, researchers could combine measures and manipulations of power and, for instance, measure an actor's power while systematically varying the power level of the target.

| Limitations
Our research has a few noteworthy limitations.First, our framework focused primarily on lateral relations that can be found higher in the power hierarchy.Importantly, one question that could be raised is whether powerholders in a similar way experience their power in interactions with their powerful peers (instead of with powerless targets).Drawing from the social identity approach, we argue, however, that identification processes are especially strong for individuals with higher power (as compared to lower power individuals; Joshi & Fast, 2013).As a consequence, powerholders may especially "feel their power" when interacting with their high-power peers (Tost, 2015).Imagine, for example, a board of directors making a strategic business decision that will affect their respective subordinates.In such a situation, their powerful position is highly salient and therefore likely affects their interpersonal interactions-also among each other.Accordingly, we argue that there are numerous situations in which power matters to interactions, even when these interactions happen among powerful peers.
We further emphasize the importance of studying these highpower dyads because the interactions and decisions made at this level have a tremendous impact on organizations, institutions, and societies as a whole.
Another potential limitation of this research is that we focused primarily on how powerholders tend to see themselves as either individual powerholders or as part of the powerful group, and how this focus in turn determines the route that powerholders will follow.
The social identity approach to leadership (Ellemers, 2018;Steffens et al., 2014) suggests, however, that under certain conditions the powerful and powerless can also identify with the same group.A leader and subordinate can, for instance, both strongly identify with the team that they belong to.Future research could benefit from exploring the interpersonal consequences of this route as well.
A final limitation of this research is our choice to focus the systematic review on articles published within 20 journals in organizational behavior, psychology and marketing.Although we believe that for our purposes (identifying a gap in the literature) the scope of the review was sufficiently broad, we note that we may have potentially missed studies in other fields and trends that stem from publication bias.

| Theoretical implications
Our framework makes several important contributions to the power literature.First, our framework shows the importance of studying the interpersonal consequences of power in a dyadic context.While we know a considerable amount about how powerholders attend to and act towards others, and the powerless in particular, our framework is the first to develop hypotheses about when and why powerholders do and do not get along with other powerholders.
By doing so, our framework provides a more nuanced portrayal of the interpersonal consequences of power.While abundant research has identified detrimental interpersonal consequences of power, our framework identifies conditions under which powerholders are likely to compete with or care for other powerholders.
Specifically, our framework suggests that powerholders can be empathic and compassionate towards other powerholders when they identify with them, and/or are motivated to legitimize existing power arrangements.

| Practical implications
Our framework also has important implications for practice.In this regard, it is imperative to understand when and why powerholders do and do not get along, because powerholders often make important decisions together.Boards of directors, for example, decide on the futures of companies, and heads of state decide on pressing world issues such as how to address crises related to health and climate change.Moreover, powerholders often operate as gatekeepers and have the ability to protect and reinforce others that are in positions of power.Our rivalry perspective suggests that powerholders might not work together because powerholders will perceive power as a zero-sum entity.Fights over power and status might thus cause powerholders to severely harm each other, which can explain why powerholders may sabotage and hinder each other in certain contexts.
From the harmonious perspective, our framework suggests that powerholders will work together in relative harmony and will defend each other in times of need because they identify with each other and are motivated to legitimize existing power arrangements.
Despite having positive consequences (as illustrated by the example of Steve Jobs supporting Heidi Roizen during a difficult time), this might also help explain more toxic processes by which powerholders defend and conceal each other's wrongdoings.An important phenomenon in this regard is the tendency of powerful men in particular, to sympathize with other powerholders who are accused of misconduct-and to feel more empathy towards them instead of towards the victims of their wrongdoing (i.e., also referred to as himpathy; Manne, 2017).

| Conclusion
Overall, the present research developed a new conceptual framework that outlines when and why powerholders act prosocially or antisocially towards their powerful peers.We hope that our framework inspires future research to empirically examine these lateral relationships higher up the power ladder, and identify when and why powerholders do and do not get along.By doing so, we are confident that we will gain a more complete and balanced understanding of the interpersonal consequences of power.

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to declare.The prisma checklist is designed for both systematic review and meta-analysis.Our paper includes a systematic review, but not a meta-analysis of study results.Therefore some points to not apply here.We indicate this in the table by specifying NA.

R E FE R E N C E S
actor/low-power target cell on the other(Van Dijke & Poppe, 2003;    Study 1).Due to the relatively small sample size (N = 117) in relation to the number of conditions (18), this study could not identify meaningful differences between such cell means.The final three studies that examined interactions between powerful individuals could only compare interactions between high-power individuals with interactions between low-power individuals(Hildreth & Anderson, 2016;

Section and Topic Item # Checklist item Location where item is reported
://doi.org/10.1016/S0022-1031(02)00524-3 van Kleef, G. A., Oveis, C., van der Löwe, I., LuoKogan, A., Goetz, J., & Keltner, D. (2008).Power, distress, and compassion: Turning a blind eye to the suffering of others.Psychological Science, 19(12), 1315-1322.https://doi.org/10.1111/j.1467-9280.2008.02241.xvanProoijen, J.-W., Coffeng, J., & Vermeer, M. (2014).Power and retributive justice: How trait information influences the fairness of punishment among power holders.Journal of Experimental Social Psychology, 50, 190-201.https://doi.org/10.1016/j.jesp.2013.10.004Effectmeasures 12 Specify for each outcome the effect measure(s) (e.g., risk ratio, mean difference) used in the synthesis or presentation of results.Describe the processes used to decide which studies were eligible for each synthesis (e.g., tabulating the study intervention characteristics and comparing against the planned groups for each synthesis [item #5]) all statistical syntheses conducted.If metaanalysis was done, present for each the summary estimate and its precision (e.g., confidence/credible interval) and measures of statistical heterogeneity.If comparing groups, describe the direction of the effect 11Specify the methods used to assess risk of bias in the included studies, including details of the tool(s) used, how many reviewers assessed each study and whether they worked independently, and if applicable, details of automation tools used in the process p. 6 relations.Journal of Experimental Social Psychology, 39(2), 149-160.httpsHow to cite this article: Feenstra, S., Oedzes, J., van der Vegt, G. S., & Stoker, J. I. (2022).Rival or comrade?A systematic review and conceptual framework of when and why the powerful act prosocially or antisocially towards each other.Journal of Applied Social Psychology, 52, 158-183.https://doi.org/10.1111/jasp.12849NA A PPE N D I X 1

(Continued) Section and Topic Item # Checklist item Location where item is reported Discussion
Discussion23a Provide a general interpretation of the results in the context of other evidence p. 16,1723b Discuss any limitations of the evidence included in the review p. 18,19 Describe sources of financial or non-financial support for the review, and the role of the funders or sponsors in the review Report which of the following are publicly available and where they can be found: template data collection forms; data extracted from included studies; data used for all analyses; analytic code; any other materials used in the review p. 37-49From: Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C.,Mulrow, C. D., et al. (2021).The PRISMA 2020 statement: An updated guideline for reporting systematic reviews.BMJ, 372(71).10.1136/bmj.n71.