Interpersonal Dominance in Relational Conflict: A View from Dyadic Power Theory

This investigation uses dyadic power theory (Dunbar, 2004; Dunbar & Burgoon, 2005a; Rollins & Bahr, 1976) to offer competing hypotheses examining the relationship between power and dominance in close relationships. Forty-seven couples engaged in a conversation while being videotaped; the tapes were coded by third-party observers for dominance. Participants rated themselves to be the most dominant when they were equal to their partners in power, followed by those who perceived they were more powerful relative to their partners. Men and women had different perceptions of power and dominance in their relationships. Men’s perceptions of power were not related to their behavioral dominance whereas when women saw themselves as more powerful, they viewed their partners as more dominant.


Distinguishing Power and Dominance
Many scholars in communication and related fields have identified power and dominance as fundamental constructs in the study of human relationships (Burgoon & Hale, 1984;Jacobson, 1986;Olson & Cromwell, 1975). These constructs have been conceptualized in a variety of ways, ranging from ethological approaches that emphasize hierarchy, to linguistic or psycho-social approaches that emphasize particular behaviors (Burgoon, Johnson & Koch, 1998). From a communication perspective, power is generally defined as the capacity to produce intended effects and, in particular, the ability to influence the behavior of another person (Berger, 1994;Burgoon & Dunbar, 2006;Dunbar, 2004;Foa & Foa, 1974;French & Raven, 1959;Gray-Little & Burks, 1983;Huston, 1983;Neff & Harter, 2002;Olson & Cromwell, 1975).
Dominance, on the other hand, refers to context-and relationship-dependent interactional patterns in which one actor's assertion of control is met by acquiescence from another (Burgoon et al., 1998;Mack, 1974;Rogers-Millar & Millar, 1979). Although dominance is sometimes viewed as a personality trait, from a communication perspective it is based upon a combination of individual temperament and situational features that encourage dominant behavior (Aries, Gold, & Weigel, 1983;Burgoon & Dunbar, 2000;Dunbar & Burgoon, 2005b). Thus, power is the ability to influence another person, whereas interpersonal dominance is a relationally-based communication strategy dependent on the context and motives of the individuals involved (Burgoon & Dunbar, 2000).

Interpersonal Approaches to Power and Dominance in Close Relationships
In a social setting, dominance and power are relative to one's social partner and are not absolute (Dunbar, 2004). As Emerson (1962) and other theorists have noted, power is rarely in the hands of one person, but is shared as people become dependent upon one another. Power and dominance are especially important in close relationships, because people often depend upon one another to attain their goals, and this dependence creates power (Molm, 1994).
Interpersonal approaches, in contrast with psychological or sociological approaches, view power as the product of interaction between people (Burgoon & Dunbar, 2000;Dunbar, 2004). Perhaps most prominent in the theoretical perspectives with implications for power are the social exchange theories. In general, social exchange theorists assume that individuals will act to maximize their interpersonal rewards and minimize their interpersonal costs (Blau, 1964;Emerson, 1972;Homans, 1958;Molm & Cook, 1995;Thibaut & Kelley, 1959). Individuals' dependence on a particular exchange partner is based on both value and alternatives, such that people are more dependent on those whose exchange relationships they value highly, especially when alternatives are few (Molm & Cook, 1995). According to the social exchange perspectives, power is achieved dyadically when a person is valued as an exchange partner and there are few alternatives (see, e.g., Emerson, 1962, Thibaut & Kelley, 1959. However, the social exchange perspective does not explicitly postulate what communication behaviors be enacted when interactants are confronted with an imbalance of power. Although they specify general outcomes, such as increases or decreases in commitment to the relationship (Floyd & Wasner, 1994;Rusbult & Buunk, 1993;Sprecher, 2001), attitudes toward divorce (Amato & Rogers, 1999), or the willingness to confront one's partner (Molm, 1994), such studies do not discuss what particular dominance strategies will be used or avoided, or how partners will react to compliance-gaining attempts or challenges.
One communication approach that does address the behavioral responses to power is the chilling effect, whereby individuals who feel powerless or who fear aggression from their partners will avoid conflict (Cloven & Roloff, 1993;Roloff & Cloven, 1990, Solomon & Samp, 1998. In their program of research, Cloven (now Solomon) and Roloff have elucidated and tested a perspective of power in interpersonal relationships that asserts that a partner's power has the tendency to quell the expression of interpersonal complaints. That is, lower power persons will withhold grievances and avoid conflict because they fear the response of their relational partner (i.e., that the partner will leave them or become aggressive). Cloven and Roloff have found that individuals' perceptions of their partner's potential actions creates the chilling effect, rather than their partner's actual behavior.
Interactive approaches treat power as relational and dominance as behaviorally manifest in interpersonal interaction. They offer us important perspectives from which to study dominance and power behaviorally. Although power and dominance are distinct conceptually (Burgoon & Dunbar, 2006;Dunbar, 2004;Dunbar & Burgoon, 2005b), they are clearly related insofar as dominance is one behavioral manifestation of the relational construct of power.

A Theoretical Approach to the Relationship between Power and Dominance
One theory that illustrates this connection between dominance and power is dyadic power theory (DPT; Dunbar, 2004). This theory builds on the interactive dominance tradition in four ways. First, it extends the social exchange perspective to include communication strategies that become manifest during interaction. Consistent with the social exchange perspective, DPT illustrates that resources and dependence are key sources of power in close relationships. Second, the theory also recognizes from the literature on relational control that the authority to utilize those resources in interactions is often granted to individuals by societal (and often patriarchal) norms as well as the interactants' own relational history (Ayres & Miura, 1981;Rogers, Castleton, & Lloyd, 1996;Rogers & Farace, 1975). Third, as the chilling effect demonstrates, DPT emphasizes that power should be viewed relatively, or in relation to others. Thus, it acknowledges that power is a dynamic, multidimensional construct that incorporates the perspectives of both individuals in the interaction. Fourth, consistent with a communication rather than a sociological or psychological perspective, it puts the interaction itself on center stage. The attempts to take control of any given interaction through dominance, although determined by power, are the focus of the theory because they determine the outcome of the process --the final decision that has ramifications for the future of the relationship.
Originally proposed by Rollins and Bahr (1976) and later revised by Dunbar (2004;Dunbar & Burgoon, 2005a), DPT emphasizes the dyadic nature of power. It asserts that power is an integral part of any relationship, especially close romantic relationships, because it determines how the partners relate to each other and how decisions are made. It assumes that perceptions of legitimate authority to make decisions and access to a variety of resources increase individuals' perceptions of their own power compared to their partner.
Perceptions of power, in turn, increase the likelihood of using dominance as a way to control the interaction through what Rollins and Bahr refer to as control attempts. The field of marital conflict has advanced a great deal in the thirty years since Rollins and Bahr originally proposed their theory and although their theory has been cited numerous times by the literature, no one has conducted a direct test of their propositions regarding the theoretical link between perceptions of power and the use of control attempts. The purpose of this research is to test the competing predictions between what Rollins and Bahr originally proposed in 1976 and what we now know about the relationship between dominance and power. Rollins and Bahr (1976) originally argued for a linear relationship between dominance and power stating "an increase in a spouse's perception of her (his) own power in a marriage role relative to that of her (his) partner will produce an increase in her (his) attempts to control her (his) spouse in the role" (p.623). Their rationale for making this prediction is that individuals will use control attempts only when they believe they have a probability of success with their attempts. Rollins and Bahr suggest that dissonance or strain might be created if one has power that is not exercised, suggests that control attempts will be increased as perceived power increases (Festinger, 1957). Thus, according to their theory, the more powerful one feels, the more dominant one acts: H1: The more powerful individuals perceive they are relative to their partner, the more they will display dominance in interaction with their partner.

Competing Predictions
In contrast to this proposition by Rollins and Bahr (1976), Dunbar (2004;Dunbar & Burgoon, 2005a) has argued that it is more likely that the relationship between perceived relative power and manifest dominance is curvilinear, such that partners who perceive their relative power as extremely high or low will use fewer control attempts than partners who perceive their relative power differences as small or moderate. Extremely powerful individuals do not need to make a large number of control attempts; by virtue of their latent power, they can maintain control without even appearing dominant. A prominent example of this is the demand-withdraw interaction pattern. Heavey (1990, 1993) found that the common wife-demand/husband-withdraw interaction pattern was more pronounced when the topic was of greater importance to the wife than the husband. Women typically want more changes in the areas of closeness, housework, and childcare. Husbands, who are typically presumed to be the overbenefitted and more powerful partners in a marriage, may resist cooperating in those areas through withdrawal from communication.
Because they have nothing to gain from discussing problems with their partner, they can preserve the status quo and their position of power by avoiding conflict rather than making their power manifest through confrontation (Christensen & Heavey, 1990).
On the other hand, in line with the chilling effect (e.g., Cloven & Roloff, 1993;Roloff & Cloven, 1990), powerless individuals are unlikely to express grievances if they fear that retaliation, violence, or termination of the relationship will result from their control attempt. Powerless individuals weigh the potential gain or loss of engaging in conflict and find that tolerating or accommodating a conflict at a minor cost is more beneficial than running the risk of pursuing the conflict and disrupting the relationship (Leung, 1988). Thus, extremely powerful or powerless individuals (compared to their partner) should be disinclined to use control attempts and more likely to exhibit control avoidance. Dunbar and Burgoon (2005a) found support for a curvilinear relationship between power and dominance. They examined whether perceptions of power would be related to several verbal and nonverbal indicators of dominance in couples' problem-solving discussions. Although their analyses revealed a curvilinear trend that was statistically significant, a closer inspection of the data revealed that individuals were more likely to engage in dominance behaviors such as interrupting and gesturing when their partner was more powerful. However, the restricted range in that dataset (few participants reported any power differences at all) limited the generalizability of the findings. They argue that at the lower levels of reported power, the relationship between dominance and power curves upward but there was not enough data to see the entire curve.
Although it is difficult to recruit couples with large power discrepancies, especially with volunteer samples, our study extends Dunbar and Burgoon's (2005a) test of DPT in two important ways. First, this is the first study to examine the original predictions made by Rollins and Bahr (1976). Dunbar and Burgoon examined their own hypothesis arguing for a curvilinear relationship between power and dominance but did not also test the linear prediction.
Second, in Dunbar & Burgoon's (2005a) study, couples were observed having a problem-solving discussion about money. The participants were given the task of deciding together how to spend a fictional gift of $1000. While the interactions were probably indicative of "everyday" non-conflict conversations (VanderVoort & Duck, 2000), the present study examines interpersonal conflict which presents unique challenges for couples. Argyle and Furnham (1983) demonstrated that conflict is highest in our closest relationships, such as those with our spouse, but that those relationships also represented the highest level of support and satisfaction. This suggests that conflict can be both a negative and a positive part of close relationships depending on how it is carried out. The decisions couples make about how to manage or avoid a conflict can influence their relational outcomes. Our goal is to test whether the curvilinear relationship proposed by Dunbar (2004) will be observed not only when couples are engaged in routine conversation but also discussing problematic areas of their relationship. Therefore, we propose the following hypothesis to test the contention that the relationship between power and dominant behavior is a nonlinear one: H2: Individuals will display more dominance when they perceive they are relatively equal in power to their partners than when they perceive they have more or less power than their partner. Rollins and Bahr (1976) argue that the "effectiveness of the control attempts is influenced by the relative power of the initiator as perceived by the recipient such that the greater the relative power of the initiator, the more likely the control attempts will be complied with" (p. 623). From this perspective, the more powerful someone is, the more dominance he or she will exert through control attempts, and the more his or her partner will comply with his or her requests. However, in interpersonal conflict, disagreements are often on-going and are not resolved in one particular interaction. In other words, "compliance" is often judged by the outcome of the conflict, such as whether or not the conflict was resolved or was escalated during the interaction as well as whether or not individuals perceive they "won" the argument (Dunbar & Allen, 2005).

Power, Dominance, and Conflict Outcomes
In contrast with Rollins and Bahr's (1976) proposition that greater power leads to greater control over the outcome, previous research suggests that having power does not necessarily result in others submitting to your will. Some researchers have found that persistence can lead to successful compliance gaining, especially for those without the power to use persuasive strategies based on legitimate power. For example, Boster, Kazoleas, Levine, Rogan, and Kang (1995) found that participants whose power was experimentally manipulated used more diverse compliance-gaining messages and were more persistent when their power was unequal than equal. Boster et al. suggest that this result, which was contrary to their predictions, occurred because the high power participants used their greater resources as a threat against their weaker partners resulting in aggression from the weaker partners and a "negative threat spiral" (Boster et al., 1995, p. 142). It is possible that because the participants were strangers in a role play with a low likelihood of future interactions, the low power partner may have felt they had "nothing to lose" by bargaining aggressively with their higher power counterpart. There was no chilling effect for the low power partner because they did not fear retaliation from their partner.
However, persistence from low-power partners has been found in the family context as well, where there is a continuing relationship among the participants. A large, nationally representative study of children ages 12 to 17 in the U.S. found that 59% use the strategy of nagging their parents an average of nine times to get something they want and 55% said the strategy was effective (Earthcare International, 2002). In sum, research supports the proposition that those who make repeated control attempts are more likely to eventually gain control than those who make few control attempts. Given that there is evidence to suggest that it could be either perceptions of power or actual behavioral dominance that influences the outcome of a conflict, we ask the following research question: RQ1a: Does individuals' perception of power relative to their partner influence their perception that the outcome of the conflict is more favorable to them?
RQ1b: Does the amount of dominance displayed in an interaction, influence the individuals' perception that the outcome of the conflict is more favorable to them?

Differing Perspectives of Dominance and Power
Finally, Rollins and Bahr (1976) argue that in marital interactions, husbands' perceptions must be distinguished from wives' perceptions. Differences between male and female socialization and unbalanced division of labor in families (e.g. Beckwith;1992;Cunningham, 2005;Zipp, Prohaska, & Bemiller, 2004) suggests that perceptions of authority and power by men and women frequently differ. However, even when women are "objectively overburdened" with household work, they often report that it is a fair arrangement because equity in relationships is subjectively based on their sex role ideology (Van Willigen & Drentea, 2001). Furthermore, Canary and Hause (1993) reviewed 15 metaanalyses of sex differences of communication-related variables and found that sex differences tend to be small, are moderated by a number of factors, and account for only 1% of the variance in communication behavior. Specifically, in the area of conflict management, some researchers have reported support for the stereotype of the aggressive male, but other researchers have found women to be as assertive as or more assertive than men (Canary & Hause, 1993). There are also contradictory findings with regard to dominance, with some researchers finding sex differences (Felmlee, 1994;Halloran, 1998;Kimble & Musgrove, 1988;Snodgrass & Rosenthal, 1984) but other researchers finding none (Walker, 1991;Dunbar & Burgoon, 2000). Thus, we asked the following research question to address the ambiguity regarding sex differences in power and dominance:

Procedures
The research was conducted at our communication research laboratory (CRL), which consists of a series of rooms for videotaping and viewing interaction. The main room in the CRL is a living room environment with comfortable chairs facing each other, and is decorated with bookshelves, art on the walls, and an area rug. This is to make the space appear home-like and encourage natural conversation. The room is equipped with two small video cameras positioned on tripods and two small omni-directional microphones hung from the ceiling. The cameras and microphones were visible but out of the line of site of the participants so as to be as unobtrusive as possible.
Upon arriving at the CRL at their appointed time, both members of the couple were placed in separate rooms. After signing the informed consent form, they individually listed five or more potentially contentious topics within their relationship to discuss (either current conflicts or not yet resolved past conflicts). The researchers compared the lists of both partners and selected a topic (and a back-up topic) that overlapped between the lists. In rare cases where no topics overlapped, the researchers asked participants to re-think their lists and repeated the procedure until overlapping topics were found. The participants also completed pre-interaction questionnaires.
Next, the participants were seated face-to-face in the living room to discuss their designated topic selected while being videotaped. Couples were instructed that they needed to speak for the full time of seven minutes; they moved to the "back-up" topic if they exhausted the first topic. Participants were instructed that they should speak about the topic naturally, as if they were at home, but should not feel as though they must resolve the issue in the brief time allotted. They were left alone in the room and the researchers observed from a remote location. After the interaction, participants were separated again to complete postinteraction questionnaires.

Measurement
Power. Because we have defined power or potential influence as a perceptual variable, it was necessary to use a self-report measure to capture the individuals' perceptions of relative power (but see Safilios-Rothschild, 1970, andOlson &Cromwell, 1975, who demur on this approach). Many researchers (e.g. Glidden, 1986;Felmlee, 1994, Neff & Harter, 2002 use single items to measure the balance of power in relationships. In order gauge reliability, we created a scale for the measurement of power that included the following items: "In general, who has more power in your relationship?"; "Who do you think makes most of the decisions in this relationship?"; "Who influences the decisions of the other person more?"; and "Who can more easily persuade the other to do things he/she does not wish to do?" All items were rated on a seven-point Likert scale anchored with Me and My partner on either end of the scale and Both Equally on the midpoint of the scale. Cronbach's coefficient alpha for the 4-item scale was .82. In general, the individuals in our study reported they were relatively equal to their partner in power (M =4.07, SD = 1.14). Scores on the power scale ranged between 1.00 and 7.00 but 70% of the sample was clustered around the mean between 3.25 and 5.25.
Dominance. Measurement of dominance behaviors in a discrete episode warranted a triangulated approach. Some research has found low levels of agreement between self-report and objective measures of the same construct (Sypher & Sypher, 1984), whereas others (Burgoon & Dunbar, 2000;Dunbar, Ramirez, & Burgoon, 2003) found a fair degree of consensus among senders, receivers, and trained coders. Burgoon and Newton (1991) put it this way: "Participants, due to their proximity to one another, are awash in a stream of subtle and visceral nonverbal cues that the observer, standing on the banks, as it were, cannot detect" (p. 109). However, third-party observers bring a level of detachment to their observations that participants are not capable of, as participants are more involved in the interaction and are therefore more cognitively busy (Burgoon & Newton, 1991;Burgoon & Dunbar 2000). Whereas participants are more attuned to their partners' communication style, they may be unable to distinguish their partners' current behavior from their normal behavior and therefore are not necessarily reporting on the same behaviors that the trained observers have seen. Thus, in this study, perceptions of interactional dominance were measured by both of the participants themselves (and about each other) and by two thirdparty observers.
Dominance was assessed by both participants and observers using seven-point semantic differential scale items taken from previous studies of dominance (Burgoon et al., 1998;Burgoon & Dunbar, 2000;Dunbar & Burgoon, 2005b). These scale items, which included "dominant/submissive," "confident/unconfident," and "outgoing/withdrawn" were intended to capture more general impressions of an individual's dominance over the course of an entire encounter. Participants completed their ratings of their own and their partner's behavior after their interaction; observers completed their ratings while observing the videotaped interaction. The inter-item reliabilities of the dominance measure were .89 when participants were rating their own dominance, .88 when rating their partners, and .89 and .90 for the two third-party observers. Inter-rater reliability was .89 for the two observers.
The average ratings of the two coders were used in analyses. A paired-samples t-test was used to compare the observer ratings to the ratings of participants. The results revealed that observers (M = 4.39, SD = .87) saw the participants as significantly more dominant than they saw themselves t (93) = 4.29, p < .01, (M = 3.97, SD = .32) or their partners t (93) = 3.80, p < .01, (M = 4.02, SD = .39).
Conflict Outcomes. The measure of conflict outcomes was developed by Bippus (2003).
This measure contains 16 seven-point items tapping perceptions that the conflict escalated ( = .78), whether or not the couple had made progress on the issue ( = .82), and who "won" the conflict ( = .78). The escalation sub-scale included nine items such as "Communication between my partner and me was difficult," "I feel the fight was worse than before we had talked," and "I feel worse." The progress sub-scale included five items such as "I feel a sense of accomplishment about the conflict," "I feel we made good progress in resolving the conflict" and "I feel we have solved our problem." To determine whether or not individuals felt their partner had "won" the argument, the following two items were used: "I feel like my partner's perspective won out in our argument" and "My partner won the argument." In general, the individuals in our study reported their conflicts had not escalated a great deal (M = 1.70, SD = .73) and they had made some progress on their issue (M = 4.02, SD = 1.40) during their interaction. The participants also generally disagreed that their partner had "won" the argument (M = 3.03, SD = 1.49), although it is not clear if they felt they had won the argument themselves or if it they were unable to declare a "winner".

Overview
As the data used in this study were gathered from couples in dyadic relationships, there is interdependence between the individuals. This violates the independence-ofobservations assumption required by many statistical tests. To address this, we used the statistical method of dealing with dyadic data proposed by the Actor-Partner Interdependence Model (APIM; Kenny, Kashy, & Cook, 2006). Kenny et al. recommend first establishing the interdependence of the data, which we did by conducting bivariate correlations between each actor and partner dyad. The actor and partner perceptions of power r (93) = -.66, p < .01 and self-reports of dominance r (93) = -.23, p < .05 were significantly correlated with one another, suggesting that the APIM is the appropriate statistical analysis. Though a more stringent alpha level may have been set to adjust for the number of tests, we retained an alpha of .05 for all tests so as not to increase our chance of Type II error due to low power.

Hypotheses
Based on the original propositions made by Rollins and Bahr (1976), we predicted in H1 that the more powerful individuals perceive they are relative to their partner, the more they will display dominance in interaction with their partner. H1 was tested with a separate mixed model analysis for self-reported dominance, ratings of partner dominance, and coder-rated dominance with both the actor's self-reported perception of relative power and the partner's self-reported perception of relative power as predictors (following the recommendations by Kenny et al., 2006, p. 160-161). The results revealed that neither selfreported dominance nor coder-rated dominance was influenced by perceptions of power; however, the actor's perception of his/her power was a significant predictor of his/her perception of his/her partner's dominance = .12, t(64.37) = 2.72, p < .01. Contrary to H1 which predicted perceptions of power would increase one's own dominance, the results reveal that more power increased perceptions of the partner's dominance. In other words, the more powerful an actor felt going into an interaction, the more dominant he or she perceived the partner was during the interaction. Thus, H1 was not supported.
H2 predicted a curvilinear relationship between power and dominance such that those who perceive themselves to be equal to their partner in power will exhibit the most dominance. A separate multiple regression analysis was conducted for each of the dominance measures (self-rated, ratings of partner, and coder-rated). Following the method of Dunbar and Burgoon (2005a), the predictor variables included in each model were the actor's perceived power, the partner's perceived power, the actor's power squared, the partner's power squared, and the person's role (either actor or partner). The polynomials were included to test for the curvilinear effect. For self-reported dominance, the analysis revealed that both actor power = .34, t(93) = 2.23, p < .05, R 2 = .09 and actor power squared were significant predictors = -.05, t(93) = -2.40, p < .05, R 2 = .09 indicating significant effects for both the linear and curvilinear trends (see Figure 1). For coder-rated dominance, only actor power squared was a near-significant predictor = .08, t(93) = 1.71, p = .09, R 2 = .09.
There were no significant predictors of partner dominance. These results provide partial support for H2. The presence of both the linear (for self-reported dominance) and the curvilinear trend (for self-reported dominance and coder-rated dominance) suggests that although those in equal power positions are reporting the most dominance, both the high power and the low power may not be experiencing the same level of submissiveness.

Research Question
Research question 1 tested whether it was an individuals' perception of power or dominance that most influenced their perception of the outcome of the interaction. RQ1a was tested with a mixed model analyses, similar to that of H1, for the three interaction outcomes (perceptions of conflict escalation, progress, and "who won" the interaction) with both the actor's self-reported perception of relative power and the partner's self-reported perception of relative power as predictors. None of the analyses were significant, finding that perceptions of power did not have a direct effect on the perceptions of the conflict's outcome.
RQ1b was tested with a similar mixed model analysis for the three interaction outcomes (perceptions of conflict escalation, progress, and "who won" the interaction) with the actor's self-reported dominance, the ratings of partner dominance, coder-rated dominance and the person as predictors with "person A or B" included as a repeated factor (see Kashy, Jellison & Kenny, 2004). The results revealed that neither self-reported dominance nor partner dominance influenced whether or not the participants saw that they had made progress in resolving the conflict or whether or not they perceived they had "won" the conflict. However, the actor's perception of his/her own dominance was a significant predictor of his/her perception that the conflict had escalated = .38, t(63.98) = 1.87, p = .05.
Taken together, these results suggest it is more likely that it is the behavioral dominance displayed in the interaction, rather than the perceptions of power prior to the interaction that influence conflict outcomes.
The second research question asked if the relationship of power to dominance would differ between men and women. A series of bivariate correlations, conducted separately for men and women between the power variables (actor and partner), the dominance variables (self-rated, ratings of partner, and coder-rated), and the conflict outcome variables (perceptions of conflict escalation, progress, and "who won" the interaction) revealed some different results for men and women. For men, the only significant correlation was between the actor's perception of power and his partner's perception of power r(46) = -.66, p < .01. For women, there were significant correlations (see Table 1) between their perceptions of power and their partner's perception of power as well as and how dominant they thought their partner acted in the interaction and how much progress they thought had been made on the issue.
To test for the curvilinear relationship between power and dominance, a curve-fit regression analysis was conducted for each of the dominance measures (self-rated, ratings of partner, and coder-rated) separately for men and women. As men and women are now treated as separate samples, the independence of observations assumption has not been violated. For men, neither the linear nor the quadratic trends for coder-rated dominance and self-rated dominance were significant. The linear trend for partner dominance approached significance = .08, t(46) = 1.75, p = .09, R 2 = .06 but the quadratic trend was not significant.
The quadratic trend for coder-rated dominance approached significance = .12, t(46) = 1.95, p = .06, R 2 = .10. In examining the plots provided in figures 2 and 3 there is evidence of a curvilinear trend, for women only. As predicted by H2, when women rated their own dominance after interacting with their partners, they indicated they were most dominant when they were equal in power to their partners. When the women rated their partner's dominance, however, they were more likely to see their partner as more dominant when the women self-reported they felt more powerful.
As these results revealed different findings for men and women, a series of independent samples t-tests were conducted to compare for direct differences. There were no significant differences between men and women on self-reported power t(92) = .86, ns, self-rated dominance t(92) = .43, ns, partner dominance t(92) = .18, ns, or coder-rated dominance t(92) = .83, ns.

Discussion
This study offers competing hypotheses for the relationships between perceptions of power and displays of dominance in interactions. We argue that in contrast with the predictions made by Rollins and Bahr (1976), power and dominance are curvilinearly related and dominance displays have greater effects on the outcome of the conflict than pre-existing impressions of power. We also explored the consequences for differing perspectives between men and women.
Based on the extant literature since the publication of Rollins and Bahr's (1976) theory, Dunbar (2004) theorized that the relationship between perceived power and the dominance displayed in interactions would more likely be curvilinear, in that those who perceive themselves to be more or less powerful than their partner would engage in less dominance than those who perceive they are equal to their partners. The study described here is more supportive of this view than the original linear view although the results should be viewed with some caution due to the limitations in the sample discussed below.
Hypothesis one and the direct relationship between power and conflict outcomes tested in RQ1a , which were consistent with Rollins and Bahr's proposal of a linear relationship, were not supported. The curvilinear pattern, on the other hand, is evident both in the results from H2 and the results from the sex-segregated analyses for RQ2. In addition, the data here from the tests of RQ1b supports the contention that power does not directly result in favorable outcomes but the interaction itself plays a more powerful role in predicting the outcome of the conflict.

The Relationship between Power and Dominance
For self-reported dominance, there is both a linear and curvilinear relationship between dominance and power for both the dyadic model and the model for female participants only. In both cases, there is a positive linear equation and a negative quadratic equation, which generally means that the linear and quadratic term compete with one another because the quadratic term is exerting a downward force on the equation. It is presumed that because of the presence of the curvilinear trend, the trend will eventually level off and head downward but this is beyond the maximum of the data (Simon, 2006). One possibility for this effect is that the curve is somewhat skewed; that is, the two tails of high and low power are not identical in how power relates to dominance. It is possible that, as H2 predicts, those who perceive they are equal in power to their partner exert the most dominance, but that those who are relatively high in power exert more dominance than those who are relatively low in power. Perhaps not surprisingly, even though high power partners do not feel the need to exhibit much dominance during a conflict episode, they do so more than partners who have low power (who might be exhibiting the chilling effect).
Thus, how much dominance one displays in a power-imbalanced relationship depends on whether or not one is in the position of high or low power.
In the case of coder-rated or ratings of partner dominance, the results are less clear than that of self-rated dominance. Previous investigations have found that participants and observers do not view dominance behavior as having the same meaning (Burgoon & Dunbar, 2000;Dunbar & Burgoon, 2005a;Dunbar, et al., 2003). Both the dyadic analyses and the female-only analyses revealed near-significant positive quadratic equations for coder-rated dominance, suggesting that there is a difference between the coders' perceptions of those who self-reported they were equal in power and that of the participants themselves. Selfreported power did not seem to influence partners' perceptions of dominance during the interaction. However, these findings must be viewed with caution given that many of them were not significant at the .05 level. Future studies with greater statistical power should explore the possibility that partners and outside observers have differing perceptions of dominance enacted in conversation from individuals rating themselves. Perceptions of both participants should be included in future studies as well. Rollins and Bahr (1976) make the argument that husbands and wives have different perceptions of power and control in their relationships based on their different roles and culturally-prescribed norms, but they do not specify how those differing perceptions will result in different behavior for men and women. In fact, their model of husband-wife interaction depicts the relationship between authority, resources, power, and control attempts to be identical for men and women. They acknowledge the weakness of this approach: "Although there is good reason to expect husband and wife differences in perceptions, the theory does not adequately account for these differences….these perceptual differences result from factors not explicitly identified in the theory" (Rollins & Bahr, 1976, p. 625). In addition, the literature on sex differences is rife with inconsistencies (Canary & Hause, 1993), spawning our research question about differences in the perceptions of men and women.

Sex Differences in Perceived Power and Dominance
The results revealed that men and women do see both their own behavior and that of their partner differently. For men, the only significant correlation between power, dominance, and the interaction outcomes was between the actor's and his partner's perception of power in the relationship. In other words, men's perceptions of power were related to their partners' perceptions, in that men who saw themselves as powerful tended to have partners who shared their view. However, as none of the linear or quadratic trends hypothesized by H2 were significant for men, it appears that men's perceptions of power do not appear to be related to their behavioral dominance as reported by themselves, their partners, or the third-parties.
For women, the story is quite different. In examining the correlations presented in Table 1, it appears that women saw their partners as more dominant when they saw themselves as more powerful. They also saw a connection between their own displays of dominance and their partners' displays of dominance. In addition, according to the curve-fit regression analysis, women were most likely to report they engaged in dominance behaviors when they were equal in power but more likely to report their partner engaged in dominance when the female was the most powerful. Combined, these results indicate that when women perceive they occupy the most powerful position in a relationship, they believe that they receive the most resistance from their male partners. This is consistent with the research that suggests our culture does not support powerful women, especially in relational contexts; or, at least our results support that women have internalized this notion. In addition, marital adjustment and relational satisfaction are often found to be lowest in wife-dominant marriages (Coleman & Straus, 1986;Gray-Little, 1982;Harrell, 1990), and female-dominated courtships are the most likely to dissolve (Felmlee, 1994). Perhaps women highest in power see their husbands as the most dominant, challenging their authority in conflict, despite the fact that this view is not shared by their partners or third-party observers. Future research should investigate further not just the perceptions of power related to satisfaction, but if resistance to control mediates its effect to satisfaction with the outcomes of conflict interactions.

Limitations and Future Directions
To our knowledge, this study is one of the first to examine power in close relationships using DPT's theoretical framework. In fact, though many studies have cited the original Rollins and Bahr (1976) theory, this is the first to test the theory empirically.
One limitation of this study was the reliance on a volunteer sample solicited by students, which may have influenced the type of couples that were included. Individuals with extreme differences in power were not present in this study and limited the range of the findings. While other researchers have been able to hand-pick dissatisfied couples from larger samples for observation in laboratory studies, they routinely pay these couples $150-200 for their participation (Gottman, 1993;Gottman & Levenson, 1999;Jacobson, Gottman, Waltz, Rushe, Babcock, & Holtzworth-Munroe, 1994;Levenson, Carsten & Gottman, 1993;Roberts, 2000;Roberts & Krokoff, 1990). Future research would benefit from a more diverse range of power configurations in the relationships sampled. Perhaps future researchers should consider conducting experimental tests of DPT in which power could be manipulated and thus guaranteeing there would be equal sized groups with high, low, and equal power.
In addition, in order to increase the number of couples that were included, both dating and married couples were combined for the analyses of this study. Rollins and Bahr (1976) only discussed married couples in their original iteration of the theory. Ideally, larger samples should be collected to allow testing of the model within married and dating groups.
Perhaps somewhat different associations would emerge between the correlations within these separate samples.
Future researchers should also consider other ways to test the relationship between dominance and power. Although we still expect that extremely powerful or powerless individuals will not engage in dominance behavior, these individuals are probably rare and so this proposition made by DPT will prove difficult to test in naturally occurring relationships. A more likely scenario is that those highly imbalanced relationships will not survive to marriage or even cohabitation. If they do happen to survive, finding those couples and including them in a voluntary study will prove very difficult. The results of this study suggest that most couples see themselves as relatively equal in power, all things considered, and the deviations from this trend toward equality are typically very small. As equity theory suggests (Sprecher, 2001), equally balanced relationships are the most comfortable and so people may either convince themselves they are equal (even when they are not) or if they recognize inequity, may use their verbal and nonverbal control attempts to change this power discrepancy. Thus, those who are relatively equal in power will use more control attempts but those who are extremely powerless will use fewer control attempts than those who are equal or more powerful. This is a complex relationship but is borne out by the findings reported here. Further research needs to continue to investigate this relationship between power and communication behavior. The findings of the current investigation provide a springboard for future research by illustrating the significance and complexity of power dynamics in close relationships.