Elucidating the Effect of Perceived Power on Destructive Responses during Romantic Conflicts

Abstract Prior research has indicated that the people one loves the most, such as their romantic partners, ironically, are also the people toward whom they often direct destructive behaviors in times of conflict, and such destructive responses become one of the most challenging relationship problems. Identifying the conditions that promote destructive (vs. constructive) conflict-resolution strategies is a crucial gap requiring study to help individuals build healthier and happier relationships. Across three studies (total N = 728), we examined whether (a) power is related to direct destructive (vs. constructive) responses during romantic conflicts; (b) this effect was moderated by the seriousness of the conflict and the relationship’s inclusiveness. In Study 1, participants involved in romantic relationships completed scales assessing interpersonal power, the conflict’s seriousness, their relationship’s inclusiveness, and conflict-resolution responses. In Studies 2-3, the participants were randomly assigned to complete an essay in which the conflict’s seriousness and power were experimentally manipulated. Findings from hierarchical regression analyses consistently showed that power led to destructive (and lower constructive) responses. However, this only occurred when the participants faced severe conflicts and their partner was not central to their self-concept. An internal meta-analysis of the studies confirmed the reliability and significance of these relationships; |r’s| =.13-37. Together, these results support the proposition that power asymmetries can threaten relationships by driving destructive responses during romantic conflicts, and untangle the conditions under which this happens. The conflict’s seriousness and the inclusiveness of the relationship may be considered to provide skills that help individuals navigate their relationships’ life challenges.

Romantic couples often struggle in situations in which the partners argue about financial issues, how to spend time together, family responsibilities, and bad habits. Indeed, relationship conflict is inevitable with no risks of damaging if they are adequately addressed (Overall, 2020). However, ironically, people often respond destructively to their loved ones in times of conflict, and such destructive responses become one of the most challenging relationship problems (Overall & McNulty, 2017). While the potential (inter) personal consequences of conflict resolution are well-known (e.g., Overall, 2020), identifying, clearly, the conditions that promote direct destructive (vs. constructive) conflict-resolution strategies is crucial to help individuals build healthier and happier relationships.
To shed light on this issue, recent research has indicated that power dynamics are essential when negotiating conflict (Pietromonaco et al., 2020). Yet, surprisingly, relatively little theoretical or empirical work has focused on the influence that power differences have on relationship partners and their conflictresolution responses. Because people often feel the need to influence others to achieve a desirable solution when goals and needs conflict in relationships (Overall et al., 2016;Rusbult & van Lange, 2003), the present work investigates whether the ability or capacity to influence the partner's ideas-namely power within the relationship-plays an important role in determining the type of response used during romantic conflicts. Building on the power and close relationship literature, the current research sought to (a) uncover if powerful people (rather powerless) are more willing to use a direct destructive (vs. constructive) response to their partner during romantic conflict discussions; and if so, (b) untangle the conditions under what powerful people may use a destructive response during romantic conflicts.

Power in Relationships as Determining of Conflict Resolution
Although in many romantic relationships partners strive for equality, asymmetries in levels of dependence commonly result in one partner perceiving to have more relative power than the other across different types of interactions (e.g., Simpson et al., 2015). According to the Principle of Least Interest (Kelley & Thibaut, 1978), the partner in a relationship who is less invested or interested in continuing the relationship has relatively more power to influence their partner in their favor because is less affected by their partner's actions (i.e., has less to lose if the relationship ended) and thus, is more independent of him/her. Therefore, in line with previous literature, holding power in relationships includes both being more independent of the partner as well as being more able to influence the partner to reach desired outcomes (Overall et al., 2016;Simpson et al., 2015). These power dynamics may have important consequences when negotiating conflicts. In particular, because powerful people are relatively more independent of others, they have the option to decide between displaying constructive or destructive responses during a romantic discussion (Foulk et al., 2020;Rusbult & Zembrodt, 1986). Constructive responses refer to an attempt to maintain the relationship via cooperative or prosocial behaviors such as seeking help, explaining one's point of view, and suggesting solutions to problems. Destructive responses allow a relationship to atrophy or end it via oppositional behaviors such as being selfish, coercive, or ending the relationship (Rusbult et al., 1986).
In line with Kipnis's (1972) early proposition that power corrupts, we propose that powerful, as compared to powerless individuals, may be more likely to respond to their partner via opposition (rather than cooperative) behaviors-that is, using direct destructive conflict resolution. We draw our assumptions from diverse theoretical perspectives on power and close relationships-such as the approach-inhibition theory (Keltner et al., 2003), and the interdependency theory (Kelley & Thibaut, 1978). Despite their differences, these theoretical models share the assumption that people high in power are less dependent from others, basically because they have the potential to influence and retaliate against others by delivering punishments and withholding rewards. This makes them less vulnerable to potential threats and thus, individuals may directly engage in courses of action that advance their aims rather than broader partner or relationship's interest. Stated differently, elevated power seems to evoke a focus on personal benefits or desires and hinders power-holders from behaving in favor of relationships welfare (Case et al., 2015;Galinsky et al., 2003). Supporting this view, traditional research has demonstrated that holding power reduces perspective-taking (Galinsky et al., 2006), compassion (van Kleef et al., 2008, social affiliative motives (Case et al., 2015), willingness to help strangers (Lammers et al., 2012) and, sacrifice in romantic relationships (Riguetti et al., 2015). Likewise, importantly, a higher power relative to others increases dehumanization (Gwinn et al., 2013), abusive behaviors (Cislak et al., 2018), and intimate partner violence (Martín-Lanas et al., 2019). Together, this theoretical and empirical reasoning guided us to suggest power as a factor that determines individuals' destructive rather than constructive responses in times of conflict. Yet, to our knowledge, no prior research has specifically analyzed whether greater power guides individuals' conflict-resolution responses within relationship interactions in a close relationship context. This link is particularly important in romantic relationships because they are a highly interdependent context where people are also communally oriented and should care or feel responsible about each other's needs and well-being (Clark & Mills, 2012). In fact, romantic relationships cannot survive long terms if people always act in a dysfunctional manner (Karremans & Smith, 2010). Therefore, albeit the antisocial view of power may predominate the power literature, powerful individuals might not invariably behave destructively in any conflictual situation (Foulk et al., 2020). Additionally exploring how relationship power may interlock with the other situational and relational factors provides refined insights (Tost & Johnson, 2019). In this line, the current investigation expands prior research by testing whether higher power is associated with destructive rather constructive responses in a conflictual situation towards their romantic partner, testing the moderating roles of conflict seriousness and the inclusion of the partner in their self.

Untangling when Power Does Promote Destructive Responses in Romantic Relationships
Considering power heightens goal pursuit during relevant situation contexts (Guinote, 2007), we suggest that variables that constrain those interests during romantic conflicts may prompt individuals' destructive rather constructive responses to face such conflicts. The seriousness of the conflict-a contextual factor that may affect people's responses and behaviors by highlighting their own needs (Deutsch et al., 2011)-could, firstly, explain under what circumstances powerful individuals may use destructive responses to face conflicts. Evidence has shown that how severely someone interprets the conflict influences how they will react to it . Specifically, the perception of conflict as severe increases the demand for solutions, the need to vent frustration, and the use of direct oppositional responses during the conflict (Overall & McNulty, 2017;Rusbult et al., 1986). For example, McNulty and Rusell (2010) explained that the discussion of severe relationship problems increases the use of direct oppositional behaviors to effectively resolve problems; conversely, when problems are minor, the use of this response is likely to be perceived as unnecessary. Why would the seriousness of the conflict specifically affect the responses of powerful people? Given its fundamental role in structuring social interactions, conflict might exacerbate the effects of some dimensions of interdependence, for example, relationship power (Rusbult & van Lange, 2003). When the conflict of interest is strong and relevant situations arise, powerful individuals experience important demands that may outweigh their resources, resulting in a greater threat to their position and therefore their goal fulfillment . In other words, power-holders might suspect others' withdrawal of cooperation and fear of losing power (Mooijman et al., 2019). As not every conflict of interest implies a power imbalance (and vice versa), this suggests that the confluence of both factors may render social interactions particularly troubling from a relationship perspective. Indeed, previous work has shown that when threats to power are high, restoring power has priority, leading to a cascade of defensive responses that aim at protecting individuals from harm (Bukowski et al., 2017;Deng et al., 2018;Weiss et al., 2020). Thus, we suggest that the higher the perceived degree of conflict, the more the power-holder sees reasons to suspect the possibility of their own goals could be blocked and thus, they might respond to perceived constraints upon their power by increasing negative or antisocial behaviors to get their way during interpersonal conflicts (Carpenter, 2017;Deng et al., 2018;Foulk et al., 2020). As a result of these dynamics, we predicted that the perception of having power would be associated with an increase in destructive responses during romantic conflicts (and a decrease in constructive ones), to the extent that people identify romantic conflict as severe.
Secondly, going beyond the previous postulation, different levels of relationship motivations may also differentially affect conflict-resolution responses. Drawing from an interdependence theory framework, individuals goals could be composed by others needs besides the personal ones (Rusbult & van Lange, 2003). Indeed, prior research has shown that romantic relationship partners may have mostly a communal relationship in which individuals' main goal might be to make happy their partner or take care of their relationship welfare and maintenance, beyond their selfinterests (Clark & Mills, 2012). In this line, the conflict resolution of powerful partners may also depend on the degree to which individuals perceived their partner as a central part of their self-concept and thus, include their partner into their conception of the self (IOS; Aron et al., 1992). Evidence in interpersonal contexts has shown that when individuals have a high IOS, they are motivated to act constructively, in favor of this person or relationship's needs, moving their own personal need to the background . Conversely, having low feelings of connectedness does not promote each partner's awareness of the other's needs (Rusbult et al., 1991). Why would IOS affect the conflict responses of powerful people? As outlined above, powerful individuals should better respond in an important situation, especially to what is accessible during these situations (Guinote, 2007). When the other person is close to the powerful one and thus, likely to be included in their self (Aron et al., 1992), power should heighten pursuit of relationship-maintenance goals and the powerful person could behave and communicate constructively according to the relationship's needs (Chen et al., 2001;. Conversely, when the identification is low, the personal self should be salient and individuals probably behave destructively for the relationship, focusing on their personal goals. Accordingly, previous researches have shown that power induces prosocial behaviors when people feel-or are induced to feel--closely connected to the other person (DeMarree et al., 2014;Gordon & Chen, 2013;Karremans & Smith, 2010). However, following a traditional view of power, other studies suggest that powerful people tend to be less committed and connected to their relationships, and thus engage in destructive responses for the relationship (Chen et al., 2001;Lammers et al., 2012;Riguetti et al., 2015). Because not all romantic relationships are characterized by the same level of feelings and motivation of connectedness which drives force for engaging in pro-relationship behavior (Rusbult et al., 1991), we suggest that the higher perceived power relative to the romantic partner would be associated with an increase in destructive (not constructive) conflict resolution to the extent that the partner is not considered a central part of their self-concept.

The Current Research
In the present research, three studies were conducted to test the general hypothesis that power would be positively associated with destructive responses and negatively associated with constructive responses to face Power and Conflicts in Romantic Relationships 3 conflicts (Hypothesis 1). We further examined under what circumstances the perception of power would be associated with conflict resolution, that is, whether the effects of power were moderated by the seriousness of the conflict and IOS. We expected that power would result in greater destructive and lower constructive responses when (a) the conflict was severe rather than mild (Hypothesis 2), (b) one had a low rather than high IOS (Hypothesis 3).

Sample
Participants were 347 Spanish individuals (207 women) who were involved at least 6 months in a romantic relationship. Originally, 400 participated in the study but 53 were excluded because they did not follow the instructions properly. Participants' mean age was 22.91 years (SD = 8.79). On average, they reported being involved in a romantic relationship for 4.47 years (SD = 6.97), and 19.6% lived together. A sensitive power analysis was conducted using the linear multiple regression R 2 increase tested in G*Power (Faul et al., 2009) to determine our ability to detect the contribution of interactions between power, seriousness, and IOS. Sensitivity analysis suggests that with this sample (N = 347, α = .05), we have the ability to detect small effect sizes (f 2 ≥ 0.034) with a power level of .80.

Measures and Procedure
Participants were recruited through an incidental sampling procedure in different public locations (e.g., local transportation stations) in a southeast Spanish city. Specifically, a trained evaluator requested the participants' collaboration and informed them of the estimated study's duration (approximately 15 min) and the agreement with the principles of confidentiality and anonymity regarding their responses. After signing an informed consent form, participants who affirmed being involved in a romantic relationship were asked to complete the questionnaire booklet individually and voluntarily. First, they were asked to indicate to what extent they have power in their romantic relationship using the Spanish adaptation of the Sense of Power Scale (e.g., "I can make my partner listen to me"; Anderson et al., 2012;Willis et al., 2016), composed by 8 items rated on a 7-point Likert scale (1 = totally disagree, 7 = totally agree); Cronbach's alpha =. 73. Then, individuals were asked to conceptualize their own experiences of relationship closeness using an adapted Spanish version of the Inclusion of the other in the self (IOS) scale (Aron et al., 1992;Gómez et al., 2011). This pictorial scale includes five sets of two circles in which one of the circles represents the "self" of the participant and the other circle represents their "partner". Each set depicts the circles with different degrees of overlap (1 = totally independent, 5 = almost completely overlapping). Therefore, participants had to indicate which model most closely represented their relationship. Next, individuals were asked to vividly describe, in writing, a conflict situation that they had experienced with their intimate partner. After writing about this incident, they were asked to complete another short questionnaire about this conflict situation. Specifically, participants rated the severity of the conflict described using 4 items (e.g. "At the time of the incident you described above, how trivial or severe was the problem?"; Rusbult et al., 1986;Valor-Segura et al., 2014) on a 6-point Likert Scale (1 = not all severe, 6 = extremely severe); Cronbach's alpha = .89. Then, participants were asked to report how they responded to the described conflict situation with their partner using the Adaptation of the Accommodation in Romantic Couples Scale (ARCS) to the Spanish Population. This measure was composed by 27 items rated on a 9-point Likert scale (1 = I did not do that, 9 = I showed that type of behavior; Valor-Segura et al., 2020) that included a direct "destructive" conflict resolution subscale (e.g. "When I am very upset with my partner I think about terminating our relationship"; α = .91) and the direct "constructive" conflict resolution subscale (e.g. "When things aren't going well between us, I suggest changing things in the relationship to solve the problem"; α = .75). Once participants filled the questionnaire, they were fully debriefed and thanked. 1

Statistical Analysis Strategy
Descriptive statistics and correlations for all variables are displayed in Table 1. Hierarchical regression analyses were performed to test the predictive contribution of power, the seriousness of the conflict, IOS, and their interactions terms to destructive and constructive responses, using SPSS Version 21. Specifically, sociodemographic 2 factors were entered in Step 1 (method: Enter); power (standardized), seriousness (standardized), and IOS (standardized) were included in Step 1 The study was approved by the corresponding college's Institutional Review Board. All participants gave informed written consent in accordance with the Declaration of Helsinki.
2 (method: Enter); and the two-way interactions of Power Â Seriousness, Pwer Â IOS, and Seriousness Â IOS were incorporated in Step 3 (method: Enter) of the regression model. 3 When the expected interactions emerged, we performed simple slopes analyses to interpret the interactions for high (+1SD) and low (À1SD) conflict seriousness and IOS.

Key Analyses
As Table 2 displays, power was related to destructive responses; that is participants were more likely to respond destructively as they reported perceiving increasing amounts of power in the relationship. Moreover, as expected, the effect of power on destructive responses was moderated by the seriousness of the conflict, as well as, by IOS. Simple slope analyses (Figure 1) revealed that power was not associated with destructive responses in the case of mild conflict situations (À1 SD), β = .13, t = 1.78, p = .077, 95% CI [À0.01, 0.26]. Conversely, this association emerges in the case of severe conflict situations (+1 SD), β = .30, t = 5.16, p < .001, 95% CI [0.17, 0.51]. Likewise, as Figure 2 shows, power was not associated with destructive responses for people high (+1 SD) in IOS, β = À.01, t = À0.01, p = .993, 95% CI [À0.15, 0.15], but this association emerges for people low (À1SD) in IOS, β = .27, t =3.44, p = .001, 95% CI [0.11, 0.42]. Therefore, the more power a person perceives in their relationship, the more destructively respond in a conflict situation when the conflict was severe, or they were lowly identified with their partner (low IOS).
For constructive responses, there was no significant main effect of power and no significant second order interaction of power and the seriousness of the conflict, and power and IOS.
In sum, the findings provide preliminary evidence that powerful individuals are more likely to respond destructively (not constructively) if the conflict was severe and to the extend participant had a low IOS; this does not occur during mild conflicts if they have a high level of IOS. Despite our results confirm that severe incidents would facilitate the power-holders' destructive responses as an attempt to defend their own interests (Bukowski et al., 2017;Deng et al., 2018), previous research has shown that, in general, people with high power perceive less conflict given their relative independence to others (Weiss et al., 2020). Therefore, Study 2 aimed to replicate the previous findings and extend them by experimentally inducing the seriousness of the conflict to make most salient an important conflict situation in which aims and desires may be strongly threatened.

Sample
Participants were 221 Spanish individuals (115 women) who were involved at least 6 months in a romantic relationship. Originally, 240 participated in the study but 19 were excluded because they did not follow the instructions properly. Participants' mean age was 34.72 years (SD = 11.09). On average, they reported being involved in a romantic relationship for 10.91 years (SD = 10.57), and 58.8% lived together. As in the previous study, the sensitive power (Faul et al., 2009) suggests that with this sample (N = 221, α = .05), we have the ability to detect small effect sizes (f 2 ≥ 0.055) with a power level of .80.

Measures and Procedure
Participants were recruited following the same procedure as in Study 1. They provided informed written consent and then, individually, completed the questionnaire in which they, first, rated their perceived power (α = .86) and IOS as in Study 1. Next, individuals were asked to vividly describe, in writing, a conflict situation that they had experienced with their intimate partner. However, in this case, the conflict experienced with the partner was manipulated depending on it seriousness 3 Additional analyses were also conducted in each study including the three-way interaction of Power Â Seriousness Â IOS in Step 4. Results reveals that the effect of power on destructive and constructive responses, respectively, was fully moderated altogether by both, the seriousness of the conflict and IOS in Study 1 (β = À.14, p = . , in the case of severe conflict situations for participants with low IOS (À1 SD). Despite the consistency of the previous results and the addition of the three-way interaction in Step 4 accounted for incremental criterion variance (1-3%), these results were not included in the main test because larger sample sizes are required to test a three-way interaction with enough statistical power. Note. IOS = Inclusion of the Other in the Self; Higher scores on continuous variables indicate greater standing on the variable (e.g., greater Power). a In Study 3, a dichotomy measure of power was used (0 = powerless; 1 = powerful). b In Studies 2 and 3, a dichotomy measure of seriousness was used (0 = mild; 1= severe).
(mild or severe). Specifically, participants were randomly assigned to complete a priming essay-adapted from Valor-Segura et al. (2014)-intended to elicit conflicts of different seriousness (mild or severe). Their task instructions were: "Please recall a particular incident in your relationship in which you experienced a [mild/severe] conflict or discussion. Please describe this situationwhat happened, how you felt, and so forth". After writing about the assigned incident, they were asked to complete another short questionnaire about this conflict situation. Specifically, to check the effectiveness of the manipulation, participants were asked to rate the severity of the conflict described using 1 item (e.g., "How severe do you consider the previous discussion that you described?") rated on a 6-point Likert Scale (1 = not severe, 6 = extremely severe). Then, as in Study 1, participants were asked to report their conflict-resolution responses: destructive (α = .91) and Note. N = 347; IOS = Inclusion of the Other in the Self; SE = standard error; CI = confidence interval. Higher scores on continuous variables indicate greater standing on the variable (e.g., greater Power). CI that does not include 0 indicates statistically meaningful associations.

Statistical Analysis Strategy
Descriptive statistics and correlations for all variables are displayed in Table 1. To verify the effectiveness of the experimental manipulation a Student's t-test with the experimental condition (mild vs. severe conflict) as the independent variable and "perceived seriousness" as the outcome variable was conducted. Then, as in Study 1, we performed hierarchical regression analyses to test the predictive contribution of power (standardized), the seriousness of the conflict (0 = mild, 1 = severe), IOS (standardized), and their interactions terms to destructive and constructive responses.

Manipulation Check
The results yielded statistically significant differences depending on condition t = À13.43, p < .001, d = À1.82. As expected, participants in the severe condition (M = 4.55, SD = 1.27) reported having described a conflict more severe that did participants in the mild condition (M = 2.40, SD = 1.09).

Key Analyses
As Table 3 displays, there was no significant main effect of power on destructive responses. However, as expected, the effect of power on destructive responses was moderated by the seriousness of the conflict, as well as, by IOS. Simple slope analyses (Figure 3, Panel A) revealed that power was not associated with destructive responses in the case of mild conflict situations (À1 SD), β = À.04, t = À0.41, p = .680, 95% CI [À0.13, 0.08].
Conversely, this association emerges in the case of severe conflict situations (+1 SD), β = .38, t = 4.35, p < .001, 95% CI [0.26, 0.69]. Likewise, as Figure 4 shows, although power was not associated with destructive responses for people high (+1 SD) in IOS, β = À.01, t = À0.01, p = .993, 95% CI [À0.15, 0.15], this association emerges for people low (À1 SD) in IOS, β = .16, t = 1.96, p = .052, 95% CI [0.00, 0.35]. Therefore, the more power a person perceive in their relationship, the more destructively respond in a conflict situation when the conflict was severe, or they were lowly identified with their partner (low IOS). For constructive responses, there was a significant main effect of power; that is participants were less likely to communicate constructively if they were the power-holder in the relationship. Moreover, as expected, the effect of power on constructive responses was moderated by the seriousness of the conflict. Simple slope analyses (Figure 3, Panel B) revealed that power did not led to constructive responses in the case of mild conflict situations, β = .06, t = 0.59, p = .556, 95% CI [À0.12, 0.21]. Conversely, this association emerges in case of severe conflict situations, β = À.44, t = À5.02, p < .001, 95% CI [À0.72, À0.31]. Lastly, the effect of power on constructive responses was not moderated by IOS.
In sum, Study 2 corroborated the findings of Study 1. Powerful individuals are more likely to respond destructively (and not constructively) during a severe conflict and to the extend participant had a low IOS; this does not occur during a mild conflict or if they have a high level of IOS. Despite these findings provide strong external validity because were found regarding self-reported past conflicts, the data precludes any conclusions about causality. To resolve this issue, in Study 3, besides the seriousness of the conflict, power was manipulated using a well-established experiential priming procedure (Galinsky et al., 2003).

Sample
Participants were 160 Spanish individuals (82 women) who were involved at least 6 months in a romantic relationship. Participants' mean age was 27.69 years (SD = 9.93). On average, they reported being involved in a romantic relationship for 6.19 years (SD = 8.14), and 40.9% lived together. As previous studies, the sensitivity analysis suggests that with this sample (N = 160, α = .05), we have the ability to detect small effect sizes (f 2 ≥ 0.077) with a power level of .80. Note. N = 221; IOS = Inclusion of the Other in the Self; SE = standard error; CI = confidence interval. Higher scores on continuous variables indicate greater standing on the variable (e.g., greater Power). CI that does not include 0 indicates statistically meaningful associations.

Measures and Procedure
Participants were recruited following the same procedure as in Studies 1 and 2. They provided informed written consent, and then, individually completed the questionnaire in which they first rated their IOS. Next, individuals were asked to vividly describe, in writing, a conflict situation that they had experienced with their intimate partner in which the seriousness of the conflict (mild or severe) and the participants' power (powerless or powerful) were manipulated. Specifically, participants were randomly assigned to complete an essay in which the first priming intended to elicit conflicts of different seriousness (situations that could be described as mild or severe) as in Study 2, and the second priming intended to elicit either feeling of low or high power. For the second aim, we used an experiential priming procedure-adapted from a manipulation of power by Galinsky et al. (2003), and previously used in the literature by Laurin et al., (2016)-which manipulates people's relative perception of power in a specific situation. Specifically, participants were divided into two conditions. In the powerless condition, the task instructions were: "Please recall a [mild/severe] incident in your relationship [in which your partner had power over you]. By power, we mean a situation in which your partner had control of your ability to get something you wanted, or was in a position to evaluate you. Please describe this situation in which you did not have power-what happened, how you felt and so forth". In the powerful condition, the task instructions were: "Please recall a [mild/severe] incident in your relationship [in which you had power over your partner]. By power, we mean a situation in which you controlled the ability of your partner to get something they wanted or were in a position to evaluate him/her. Please describe this situation in which you had power-what happened, how you felt, and so forth". After writing about this incident, they were asked to complete another short questionnaire about this conflict situation. Specifically, to check the effectiveness of the conflict' seriousness and power manipulations, participants were asked to rate the severity of the conflict described using 1 item (e.g., "How severe do you consider the previous discussion that you described?") rated on a 6-point Likert Scale (1 = not severe, 6 = extremely severe); and to what extend they had power during the conflict described using 1 item (e.g., "Do you think that you had control and influence on the discussion previously described with your partner?") rated on a 6-point Likert Scale (1 = no control and influence, 6 = much control and influence). Then, as in Study 1 and 2, participants were asked to report their conflict-resolution responses: destructive (α = .92) and constructive (α = .75). Once participants filled the questionnaire, they were provided with information about the study. 1

Statistical Analysis Strategy
Descriptive statistics and correlations for all variables are displayed in Table 1. To verify the effectiveness of the experimental manipulations two Student's t-test were conducted with the seriousness (mild vs. severe conflict) and power (powerless vs. powerful) experimental condition as the independent variables and perceived seriousness and power, respectively, as the outcomes variable. Then, as in Studies 1 and 2, we performed hierarchical regression analyses to test the predictive contribution of power (0 = powerless, 1 = powerful), the seriousness of the conflict (0 = mild, 1 = severe), IOS (standardized), and their interactions terms to destructive and constructive responses.

Manipulation Checks
The results yielded statistically significant differences depending on seriousness condition (t = À5.52, p < .001, d = À0.88), and power condition (t = À7.53, p < .001, d = À1.23). Participants in the severe condition (M = 4.34, SD = 1.59) reported having described a conflict more severe that did participants in the mild condition (M = 3.01, SD = 1.43). Moreover, participants in the powerful condition (M = 4.59, SD = 1.22) reported having described a conflict in which they feel more power that did participants in the powerless condition (M = 3.01, SD = 1.36).

Key Analyses
As Table 4 displays, power was related to destructive responses; that is participants were more likely to respond destructively as they reported perceiving increasing amounts of power in the relationship. Moreover, as expected, the effect of power on destructive responses was moderated by the seriousness of the conflict and IOS. Simple slope analyses ( Figure 5, Panel A) revealed that power was not associated with destructive responses in the case of mild conflict situations (À1 SD), β = .20, t = 1.79, p = .077, 95% CI [À0.04, 0.67]. Conversely, this association emerges in the case of severe conflict situations (+1 SD), β = .49, t = 4.89, p < .001, 95% CI [0.63, 1.50]. Likewise, as Figure 6 shows, although power was not associated with destructive responses for people high (+1 SD) in IOS, β = .14, t = 1.15, p = .256, 95% CI [À0.18, 0.65], this association emerges for people low (À1SD) in IOS, β = .43, t = 4.36, p < .001, 95% CI [0.52, 1.40]. Therefore, replicating previous results, the more power a person perceives in their relationship, the more destructively respond in a conflict situation when the conflict was severe, or they were lowly identified with their partner (low IOS). For constructive responses, there was a significant main effect of power; that is participants were less likely to respond constructively if they were the power-holder in the relationship. Moreover, as expected, the effect of power on constructive responses was moderated by the seriousness of the conflict. Simple slope analyses ( Figure 5, Panel B) revealed that power did not led to constructive responses in the case of mild conflict situations and, β = À.04, t = À0.37, p = .711, 95% CI [À0.51, 0.35]. Conversely, this association emerges in case of severe conflict situations, β = À.30, t = À2.79, p = .007, 95% CI [À1.07, À0.18]. Lastly, the effect of power on constructive responses was not moderated by IOS. In sum, Study 3 corroborated that powerful individuals are more likely to respond destructively (not constructively) Note. N = 160; IOS = Inclusion of the Other in the Self; SE = standard error; CI = confidence interval. Higher scores on continuous variables indicate greater standing on the variable (e.g., greater IOS). CI that does not include 0 indicates statistically meaningful associations.
during a severe conflict, and to the extend participant had a low IOS; this does not occur during a mild conflict or if they have a high level of IOS.

Internal Mini Meta-Analysis
To determine the strength, reliability, and robustness of our conclusions, we conducted an internal mini metaanalysis to test our hypotheses across studies per the recommendations of Goh et al. (2016).

Statistical Analysis Strategy
First, we carried on a mini meta-analysis of the main effect of perceived power on destructive and constructive responses across the three studies. Second, we tested if the associations between perceived power and destructive or constructive responses are moderated by seriousness and IOS. Next, we conducted the mini meta-analysis using the effect sizes for the relation between power and destructive or constructive responses for participants high and low in moderation variables (e.g., 1 SD above and below the mean of conflict seriousness and IOS). We used fixed effects in which the correlations (a measure of effect size) were calculated from the sample size along to the t-value of the multiple regressions. Positive correlations indicate that a higher level of power is associated with a higher level of destructive/ constructive responses, whereas negative correlation associated a higher level of power with a lower level of destructive/constructive responses. Analyses were conducted using the most recent version of the mini metaanalysis Excel template created by Goh et al. (2016).

Moderation of Power by the Seriousness of the Conflict (Hypothesis 2)
Results revealed that the interaction effect of power with seriousness was positively related to destructive responses (r = .14, p < .

General Discussion
Existing studies suggest that having (or lacking) power has many effects on people and their actions (Cho & Keltner, 2020;Deng et al., 2018;Willis & Rodriguez-Bailon, 2010). However, until now, the role of power as the main factor that promotes individuals' conflict resolution responses in romantic contexts and the conditions that stimulate power-holders' responses have seldom been explored. Across three studies, the primary findings of this research showed that power was positively associated with destructive responses and negatively associated with constructive responses. Bolstering work on the approach-inhibition theory of power (Keltner et al., 2003) and interdependency theory (Kelley & Thibaut, 1978), results suggest that the perception of power is associated with increased freedom and independence from others, which in turn can promote individuals to prioritize self-interest over the relationship's earnings, as reveals their destructive responses (Case et al., 2015;Galinsky et al., 2003;Cho & Keltner, 2020). Therefore, these findings extend prior research that shows the negative consequences of social power for interpersonal relations (e.g., Case et al., 2015;Cislak et al., 2018;Galinsky et al., 2003;Keltner et al., 2003;Lammers et al., 2012;Riguetti et al., 2015;Willis & Rodriguez-Bailon, 2010), and provide evidence that the psychological perception of power could also emerge as an important factor that guides people's destructive responses towards their loved ones in times of conflict. The second, perhaps most noteworthy, finding from this research is that the perception of having power is associated with greater destructive and lower constructive responses to the extent that the conflict was severe and participants had a low IOS. We provide evidence that powerful individuals do not invariably behave destructively in a conflictual situation instead, these findings advance the close relationship literature by identifying the particular conditions under destructive responses can be triggered in powerful individuals (Foulk et al., 2020). Specifically, as we have shown, these oppositional responses may be particularly relevant in situations of severe conflict, which presumably endanger their important personal goals (Carpenter, 2017). Following Guinote's (2007) situated focus theory of power, these results suggest that power-holders, who are better at distinguishing potential threats to their personal goals, might be particularly oriented to the fulfillment of those goals using destructive responses probably in a direct attempt to control their partners, thereby maintaining and even imposing their own powerful position, needs, and desires (Bukowski et al., 2017;Deng et al., 2018;. Therefore, our findings confirm that given its fundamental role in structuring social interactions, the perception of conflict can exacerbate the effects of power (Weiss et al., 2020).
Beyond the seriousness of the conflict, because romantic relationships are characterized by higher levels of communal strength (Clark & Mills, 2012), we additionally considered that the effect observed thus far may be due to individuals' goals and to what extend their partner is included in their prior needs and goals. Interestingly, our results indicated that when powerful individuals who, according to the self-concept expansion model (Aron et al., 1992) do not include other into the self, seem less concerned with others' needs-in short, they communicate destructively with others. We thus propose that the low IOS or the perception of disconnection between a person and his or her partner may make powerful people move their personal interests rather than relational goals to the foreground. When this happens and individuals' personal goals are highlighted, either because their IOS with the 4 We additionally carried on a mini meta-analysis of the main effect of gender (0 = men; 1 = women) on destructive and constructive responses across the three studies. Results showed a significant relation between gender and constructive responses (r = .28, p < .001, CI [0.22, 0.35]). Gender was not related to destructive responses (r = À.05, p = .446, CI [À0.12, 0.02]). Our findings suggest that women had a higher use of constructive responses to face conflicts than men. partner is low and they respond more to personal goals or because such goals are prioritized or threatened by a severe conflict, power-holder may increase their tendency to think, feel, and act in typically powerful ways to restore their control position, according to the classic power literature (Galinsky et al., 2006;Kipnis, 1972). In the short-run, this could be functional because conflicts may vanish quickly and powerful individuals may get their goals (Overall, 2020). However, in the long run, serious problems could remain unaddressed (Overall & McNulty, 2017). For example, on a stable basis, powerful partners may continue problematic behavior that may adversely impact the relationship quality (Overall, 2020). Thus, scientists and therapists should understand which mechanisms could change this destructive pattern. Our findings provide critical evidence that power dynamics in relationships are contingent on the seriousness of the conflict and the levels of partner's inclusiveness. Marital therapy interventions could use this information to provide skills regarding behavioral patterns that will be more helpful versus detrimental in resolving the challenges of social life. For instance, individuals could analyze targeting issues related to the perception of threat and disconnection between a person and his or her partner when severe relationship problems appear, and whether their response may be more or less effective at navigating interpersonal interactions. Future studies could complement these findings and analyze whether individual differences (e.g., attachment styles), adherence to certain social norms (e.g., traditional gender beliefs), and partner's responsiveness could also shape how individuals respond to relationship conflicts.
Finally, some limitations of the current research should be acknowledged when interpreting results. First, the sample of our studies was withdrawn from the young general population. Because previous studies have revealed that there is a developmental shift in the way that conflict is handled in later life (e.g., older couples perceive conflict as less stressful and cope with them more proactively; Neubauer et al., 2019), future studies using a larger sample composed by a representative participants' selection of each cohort from the general population should corroborate our finding in more detail. Second, albeit the findings regarding past conflict situations have strong external validity, the use of retrospective bias could also trigger all sorts of memory biases. Hence, power and the seriousness of the conflict were manipulated, thereby allowing us to exercise more control over our findings and minimize other causal pathways. However, to gain convergent validity, utilizing different methodologies may be useful in future studies. Third, in the present research, the sizes of some effects presented are small. This might be due we conceptualized power based on a general person's perceived ability to control decisions and outcomes in a relationship. Future studies may benefit from testing our predictions at a couple level to calculate the power each partner has in a specific domain and their relative influence across domains . Lastly, previous research has shown that direct destructive conflict communication have negative consequences in relationships when it was stable, but is less harmful when partners' negative-direct behavior varied across time (Overall, 2020). Thus, the identification of variability versus stability as a consequential feature of social behavior could offer fruitful directions for examining how power and personal or contextual factors shape interpersonal patterns such as romantic conflicts.
Despite these limitations, we should highlight that the utilization of samples of adults involved in established romantic relationships and the replication of results across studies are notable strengths of this research.
It is an irony of life that the people we love the most, such as our romantic partners, are also the people towards which we display destructive behaviors in times of conflict. A key contribution of this research stems from introducing relative power as a factor to determine such conflict-resolution strategies and the underlying circumstances that explain when this happens. The seriousness of the conflict and IOS thus may be necessary to understand individuals' conflict-resolution responses and may help marital intervention researchers develop specialized intervention strategies for couples based on their relationships characteristics and dynamics. We also expect that this research will open up interesting avenues for future work on the psychosocial mechanisms that emerge from close relationship interactions and the ways in which they can help people to constructively cope with conflict in the context of romantic relationships.